Aspects of Languages in Contact in Algeria: the Case of the Mzabi and Kabyle Minorities of Oran

Democratic and Popular Republic of Algeria

Ministry of Higher Education and Scientific research University of Oran

Faculty of Letters, Languages and Arts

Department of Anglo-Saxon Languages

Section of English

DOCTORAL DISSERTATION SUBMITTED IN PARTIAL FULFILMENT FOR THE DEGREE OF DOCTOR IN LINGUISTICS

ASPECTS OF LANGUAGES IN CONTACT IN ALGERIA: THE CASE OF THE MZABI AND KABYLE MINORITIES OF ORAN

SUBMITTED BY : SUPERVISED BY:

Benhattab AbdelkaderLotfi Professor Bouhadiba F.A.N

Members of the Jury

Bouamrane Ali Professor University of Oran President Bouhadiba Farouk Professor University of Oran Directeur of thesis Riche Boutheldja Professor University of Tizi Ouzou Examiner

Benmoussat Smail Professor University of Tlemcen Examiner

Benali Med Rachid Maitre de Conference ‘A’ University of Oran Examiner

Dendane Zoubir Maitre de conference ‘A’ University of Tlemcen Examiner

Academic Year 2010/2011 1

Dedications

„Incomplete‟ though it is, this work is dedicated to my parents because it is owing to their faith that I was equal to the task and that it is now as good as I can make it. I thank them for always being there to provide support and encouragement through my many years of graduate and post graduate studies. They didn‟t always understand what I was doing, but they understood my need to devote time and energy to my dissertation.

To My beloved Sihem, the kindest, the most loving, and the most tolerant wife in the world. I thank her for helping me raise our fourth child (The present dissertation), for taking care of me, my children, and everything when I was in the other world. She has been here as always at times of extreme stress and demand, always selflessly and without complaint. She is the one who resisted to my hard times and had to put up with my mood in times when things became difficult for me. To my shining, bright and boundless children: Saleha, Mohamed Amine and Ismahen. To Soraya my sister, to my aunts and uncles. To Larbi, and Abla, my parents in law who took care of my garden when I was drafting my dissertation. To my brothers in law Abdou, Kader, Amine, Sofiane, and their wives who prove to be true brothers in hard times. To Issam, the financial analyst who still believes in virtue. To Karima, my sister in law

To my brothers Houari and Mokhtar who have always been here for me in all times.

To Samir a true Kabyle and a faithful friend.

To my friends, because without their friendship I am empty and incomplete

To the Berber students who helped me in the data gathering process (Siham, Kahina, and others).

Acknowledgements

„ Whoever is thankful (to God) is in fact thankful for his own self. But if anyone is ungrateful, god is self-sufficient and glorious‟ (Coran 31 :12) … Thank God

Anything of substance is rarely the product of a lone individual – this dissertation is no exception. It is not possible to thank everyone who contributed to this physical product or my intellectual growth, but I sincerely hope that you all know what your support, encouragement, and wisdom have meant to me.

Nevertheless, I would like to single out a few people for special recognition.

I would first of all like to acknowledge the work and insights of my thesis to my supervisor, Pr Bouhadiba Farouk, His rigorous attention to the details over the course of the study from beginning to end have greatly improved my skills as a researcher. His demonstrated confidence in me and his words of encouragement were more valuable than he knew. His continual and unwavering support and guidance has helped to give this study a breadth it would otherwise lack. I thank him for his never lacking enthusiasm for research and for his professionalism which never ceased to push me forward.

Besides, I also owe a debt of gratitude to Pr Bouamrane Ali whose expertise and academic excellence added considerably to my post-graduate experience. I want to thank him for his advice, support and encouragement from the start of my course work to the day of my defence.

Pr Bouamrane has helped shape my reasoning as a researcher. His undergraduate and post graduate courses helped direct my attention to contact linguistics as a field of study. He opened my eyes to the realm of code switching research which is now my field of interest. Pr Bouamrane I consider as one of the pioneers of contact linguistics in

Algeria. Not only this work but also my whole career as a researcher has been shaped by the knowledge he transmitted to me.

I would like to extend my gratitude and appreciation to Dr Benali for his continuous assistance in preparing this thesis. His kind guidance, support and his helpful discussions and valuable suggestions kept me on the right track. Dr Benali is to my eyes a true model of academic excellence. He initiated me to research in Berber and I thank him indeed for having helped in my making as a researcher. Dr Benali has always believed in my capacity as a researcher, and his words mean much more to me than he will never imagine.

My appreciation and thanks are extended to the members of my thesis jury who are participating in the making of the future of the country at the expense of their own lives. I thank you indeed as your local product for being what you are for us university will be teachers and researchers. The knowledge and life insights you are transmitting us are shaping our lives as human beings.

I am also grateful to the Department of Anglo-Saxon Languages at the University of Oran for their financial support in the form of fully- financed trainings offered to me every year to undertake bibliographical research.

Last but not least, I am deeply indebted to Pr Ben Rampton, Pr Ad Backus, Pr Louis Boumans, Dr Gardner Chloros, Dr Malcolm Edwards, Dr Ziamari Karima, for opening my eyes to new perspectives of research in the domain of contact linguistics, and for providing me with considerable amount of literature relevant to my research.

Thanks are also due to my colleagues at the section of English - Department of Anglo-Saxon Languages. My years as a doctoral student would not have been as much fun without you: Thank you for your moral help, encouragements, and for having so strongly believed in me.

As I thanked the people who helped me so much in this endeavour, please remember that wherever I have failed in any regard, it is my failure alone.

List of Abbreviations

AA: Algerian Arabic accompl :accomplished

AdjP Proj: Adjective Phrase projection

AdjP: Adverb Phrase

AdvP Proj: Adverb Phrase Projection

AdvP: Adverb Phrase aff :affix

Ber: Berber

BR: Borrowing

Clit: Clitic Affix

CM: Code mixing

Comp: complementizer

CompP Proj: Complement Phrase projection

CP: Complement Projection

CS: Code switching

Def: Definite Article

EL Isl: Embedded Language Island

EL: Embedded Language fem: feminine

Fr: French

Gend: Gender Affix

Inaccompl: unaccomplished 6 mas:Masculine

Mixed Isl: Mixed Matrix+Embedded Island

ML Isl: Matrix Language Island

ML: Matrix Language neg: Negative affix, Negative marker

Neut: Neutral affix

Nom aff: Nominal Affix

NP Proj: Noun Phrase Projection

NP: Noun Phrase

Num: Number Affix

Obj: Object plu: Plural pref: Prefix

PrepP Proj: Prepositional Phrase projection

SA: Standard Arabic sing: singular sub: Subject

Suf: Suffix

VP Proj: Verb Phrase Projection

VP: Verb Phrase

TABLE OF CONTENTS

Dedications

Acknowledgements List of Abbreviations……………………………………………………………… I

Table of contents…………………………………………………………………... III

Abstract……………………………………………………………………………. X

Résumé de la thèse...... XI

XIII ...... ملخص البحث

Presentation of the Thesis and Research hypotheses...... XV

General Introduction...... 01

CHAPTER ONE:

A PRELIMINARY TO THE STUDY

1.1 Introduction……………………………………………………………………. 12 1.2 Berber in Algeria………………………………………………………………. 13 1.1.1Berber Varieties or Berber languages………………………………………... 16 1.3 The Present Study……………………………………………………………... 19 1.4 A Brief Overview of the Literature...... 23 1.4.1 Early Prescriptive Approaches to Code Switching...... 23 1.4.2 Macro social studies on Code Switching during the 50ths and the 60ths...... 27 1.4.2.1 Diglossia…………………………………………………………………… 29 1.4.3 Micro Sociolinguistic Approaches to Code Switching...... 34 1.4.3.1 The Use of Code Switching as a Conversational Strategy...... 40 1.4.3.2 Myers Scotton (1993) Markedness model………………………………… 44 1.4.3.3 The conversation Analysis Approach to code switching...... 53 1.4.3.3.1 Discourse related switching…………………………………………… 54

1.4.3.3.2 Extra conversational knowledge………………………………………… 55 1.4.3.3.3 Preference related Switching…………………………………………… 57 1.4.3.3.4 Participants‟ Related Switching…………………………………………. 58 1.4.3.4 The Social Interpretative Approach to Code Switching...... 58 1.4.3.5 Integrating the Social and Linguistic Approaches for a Better Interpretation of Code Switching...... 61 1.4.3.5.1 Integrating the Social and the Linguistic Approaches in the interpretation of the Present data...... 63 1.4.3.5.2 Bourdieu‟s (1991) application of the „Theory of Language Practice‟ in the Interpretation of Code Switching...... 66 1.4.3.5.2.1 Presenting key concepts in Bourdieu‟s (1991) model with a special reference to the data of the present study...... 66 1.4.3.5.2.2 Criticisms to Bourdieu‟s (1991) „Theory of language Practice‟ in the literature...... 71 1.4.3.5.3 Milroy‟s (1987) and Milroy & Li Wei‟s (1995) Investigation of Language Choice in Relation to Social Networks...... 74 1.4.3.6 Bell‟s (1984, 2001) Speech Accommodation Approach to bilingual discourse...... 76 1.4.3.7 Gardner Chloros‟ (1991, 1995) Continuum Approach to Code Switching and other Contact Phenomena...... 81 1.5 Conclusion…………………………………………………………………….. 88

CHAPTER TWO:

GRAMMATICAL APPROACHES TO CODE SWITCHING (1): EQUIVALENCE BASED STUDIES EQUIVALENCE STUDIES

2.1 Introduction……………………………………………………………………. 91 2.2 Methodological Considerations of the Study...... 92 2.2.1 The Informants of the Study and data collection...... 92 2.2.2 Transcription of the Data in the Study...... 94 2.2.3 The Approach Advocated in the Study...... 95

2.3 Grammatical Approaches to Code Switching…………………………………. 97 2.3.1 Early Interests in the Grammar of Code Switching...... 97 2.3.2 Modern Approaches to the Grammar of Code Switching...... 98 2.3.2.1 Distinguishing Code Switching and Borrowing………………………… 100 2.3.2.2 The Equivalence Based Approaches…………………………………….. 102 2.3.2.2.1 The Equivalence Constraint……………………………………………... 103 2.3.2.2.1.1 Criticisms and counter Evidences to the Equivalence Constraint Drawn from Other Bilingual Corpora...... 106 2.3.2.2.1.2 Counter Evidences to the Equivalence Constraint from the Corpus of the Study...... 112 2.3.2.2.1.2.1 Mismatch in the positions of adjectives in relation to their head nouns(Ordinary adjectives, possessive adjectives, demonstrative adjectives)…… 112 2.3.2.2.1.2.2 Mismatch in patterns of Berber-Algerian Arabic-French switching involving prepositional constructions...... 116 2.3.2.2.1.2.3 Code switching patterns involving gender assignation mismatches.... 117 2.3.2.1.1.2.4 Code switching patterns involving the sub-categorization of object accusative and dative clitic affixes...... 119 2.3.2.2.1.3 Poplack‟s Response to the Criticisms Put Forward in the Literature in Relation to the Equivalence constraint...... 122 2.3.2.3 The Free morpheme Constraint……………………………………………. 129 2.2.2.3.1 Criticisms to the Free Morpheme Constraint in the Code Switching Literature...... 130 2.3.2.3.2 Some Violations of the Free Morpheme Constraint in the Data of this Study...... 132 2.4 Conclusion…………………………………………………………………… 137

CHAPTER THREE:

GRAMMATICAL APPROACHES TO CODE SWITCHING (2)

GOVERNMENT AND MINIMALIST BASED STUDIES

3.1 Introduction…………………………………………………………………… 140 3.2 The Government and Binding Approaches to code switching...... 141 3.2.1 Woolford‟s (1983) Government and Binding Model of Code Switching...... 141 3.2.1.1 Violations to Woolford‟s (1983) constraints in Instances of code switching from the present data...... 143 3.2.2 Di Sciullo et al(1986)Government and Binding Approach to Code Switching………………………………………………………………………….. 145 3.2.2.1 Counter evidence to Muysken‟s et al Government and Binding Approach to Code switching in the Literature...... 146 3.2.2.2 Visiting Muysken et al‟s Government and Binding model of code switching using the present data...... 148 3.2.2.2.1 Subject verb switching………………………...... 149 3.2.2.2.2 Verb/Object code switching……………………………………………. 150 3.2.2.2.3 Subject/Predicate (copula-less predicate) Switching...... 151 3.2.2.2.4 Verb-Direct object-Indirect Object Switching...... 152 3.2.2.2.5 Verb-Direct object-Object Complement switching……………………... 154 3.2.2.2.6 Switching patterns involving Prepositions and their objects...... 155 3.2.3 Minimalist Approaches to Code Switching…………………………………. 159 3.2.3.1 The Single Model (Belazi, Rubin, and Toribio 1994)...... 159 3.2.3.1.1 Testing the Predictions of the Single Model (Belazi et al, 1994) on the Present Data...... 162 3.2.3.1.1.1 Switching between a complementizer (a relative pronoun or a subordinator) and its subordinate clause……………………………………….... 163 3.2.3.1.1.2 Switching between numerals and/or quantifiers and their complement Noun Phrases...... 164 3.2.3.1.1.3 Switching between a negative marker and its Complement verb...... 167 3.2.3.1.1.4 Switching between Modals and their Lexical verbs...... 169 3.2.3.1.2 Further Remarks on the Single Model (Belazi et al, 1994)...... 171 3.2.3.2 Mahootian and Santorini‟s (1993, 1996) Null Approach to Code switching...... 173 3.2.3.2.1 Some Criticisms to Mahootian and Santorini‟s (1993, 1996) Null Approach to Code Switching in the Literature...... 176

3.2.3.3 Mac Swan‟s (1999, 2000) minimalist Approach to code switching...... 178 3.2.3.3.1 Mac Swan‟s (2005) applications of minimalism to the counter evidence of the Equivalence Model (Poplack, 1982)...... 180 3.2.3.3.1.1 Facilitating Strategies in Code Switching and Borrowing (with a Special Reference to the Present Data)...... 181 3.2.3.3.2 Criticisms to Mac Swan‟s Minimalist Approach to Code switching in the Literature...... 183 3.2.3.3.3 Testing Mac Swan‟s (1999, 2000, 2005 Predictions on some instances from the Present Data...... 186 3.3 Conclusion…………………………………………………………………… 190

CHAPTER FOUR:

GRAMMATICAL APPROACHES TO CODE SWITCHING (3): INSERTIONAL STUDIES INSERTIONAL STUDIES

4.1 Introduction…………………………………………………………………… 193 4.2 Insertional approaches to code switching…………………………………… 193 4.2.1 Joshi‟s Insertional Model of Code Switching...... 194 4.2.1.1 Criticisms to Joshi‟s Constraints in the Literature...... 195 4.2.1.2 Possible Violations to Joshi‟s Constraints from the Present data...... 196 4.2.2 Myers Scotton Insertional Approach to code switching (the Matrix Language Frame Model)...... 197 4.2.2.1 Identifying the Matrix Language………………………………………… 199 4.2.2.1.1 The Language Affiliation of the Verb (Klavans (1986) and Treffer Daller (1990))...... 199 4.2.2.1.2 The Discourse Criterion (Berk Selingson, 1986)...... 202 4.2.2.1.3 The Statistical Criterion (Myers Scotton, 1993; 1997)...... 202 4.2.2.1.4 The Verb Inflections Criterion (Treffer Daller, 1994; Boumans, 1998).... 205 4.2.2.1.5 The Language Activation Psycholinguistic Criterion…………………… 207 4.2.2.2 The Different Hierarchies Operating within the MLF Model and its Subsequent Models...... 208

4.2.2.2.1 The Matrix/ Embedded language hierarchy…………………………….. 209 4.2.2.1.2 The Uniform Structure Principle……………………………………….. 212 4.2.2.1.2.1 Applying the Uniform Structure principle to the present data...... 212 4.2.2.1.3 The system/content morpheme Hierarchy………………………………. 215 4.2.2.1.3.1 Content Morphemes…………………………………………………… 215 4.2.2.1.3.2 System morphemes……………………………………………………. 217 4.2.2.1.3.2.1 Early System Morphemes…………………………………………… 220 4.2.2.1.3.2.1.1 Testing the Early System Morphemes principle against the present data...... 221 4.2.2.1.3.2.2 Late System Morphemes…………………………………………….. 223 4.2.2.1.3.2.2.1 Bridge Late System Morphemes…………………………………... 224 4.2.2.1.3.2.2.1.1 Testing the Prediction of the Bridge System Morpheme Principle against the Present Data...... 225 4.2.2.1.3.2.2.2 Outsider Late System Morphemes……………………………… 228 4.2.2.1.3.2.2.2.1 Testing the Outsider Late System Morpheme Principle on the Present Data...... 229 4.2.2.3 Problematic Cases to the MLF and the 4M Models with a Special 235 Reference to the Data……………………………………………………………… 4.2.2.3.1 Possible Problematic Instances in Relation to Negative Marking in the data...... 236 4.2.2.3.2 Switching of Algerian Arabic and French Discourse Markers with a Special Reference to Complementizers, and Coordinators...... 241 4.2.2.3.3 Insertions of Berber singly occurring elements and Islands into Algerian Arabic and French Matrices...... 249 4.2.2.4 Alternational Patterns of Switching from the Data...... 252 4.2.2.5 Insertions of Bare Forms as a possible problematic issue in Myers Scotton‟s (1993, 1997, 2000, and 2002) MLF and 4M models...... 253 4.2.2.5.1 Trial Interpretations of Algerian Arabic, Standard Arabic, and French Bare inserted Forms in the Data...... 255 4.2.2.5.1.1 Algerian Arabic, Standard Arabic, and French Bare Nouns inserted into Berber matrices...... 255 4.3Conclusion…………………………………………………………………...... 262

General conclusion...... 263 Bibliography...... 276

A B S T R A C T

The present study investigates the patterns of speech displayed by Berber speakers in Oran. These patterns are essentially bilingual as they involve Berber, Algerian Arabic, Standard Arabic, and French. Our study focuses on only two groups of the Berber minority of Oran: The Mzabi, and the Kabyle groups. The rationale behind our interest in these groups is that they outnumber the other Berber speaking groups in Oran. The patterns of speech of the above cited groups are indicative of the contact phenomena under light in our investigation.

The Kabyle and Mzabi Berber communities of Oran may be considered as quite interesting minority groups in the sense that they have as their mother tongue Berber. These minorities are the only ones whose mother tongue is totally unintelligible with the varieties that are in use in the verbal repertoire of the Algerian Arabic majority speakers. Mzabi and Kabyle speakers use Berber as the sole means of communication in their inner social networks such as the one of the family, and the one of friends. They also use the other varieties that make up their verbal repertoire i.e. Algerian Arabic, standard Arabic, and French. This situation creates contacts between the Members of these two groups and the Algerian Arabic majority speakers. Such type of contact triggers bilingual contact phenomena such as Code switching, code mixing and borrowing which are the main focus of the present investigation. The behaviour of the members of the Mzabi and Kabyle Berber minorities of Oran is at the centre of the present study.

Our investigation of the bilingual language behaviour of Kabyle and Mzabi speakers will be conducted from a micro sociolinguistic perspective. This perspective is couched within the general field of contact linguistics. We will be observing contact phenomena such as code switching and borrowing from two different angles i.e. the social angle, and the grammatical one. The investigation of the grammatical patterns of 14 code switching, code mixing and borrowing phenomena will be conducted in the light of four different approaches to the grammar of contact phenomena. These approaches have investigated code switching and related phenomena from equivalence, government, insertional, and alternational perspectives. All along our investigation we shall be applying the findings of studies couched within these four perspectives to the Kabyle and Mzabi Berber bilingual data.

R E S U M E

Cette étude sur des locuteurs Berbères habitants à Oran vise à découvrir les différentes attitudes et comportements chez un groupe minoritaire par rapport au groupe majoritaire représenté par des locuteurs de parlé arabe algérien qui est la prétendue langue ou variété dominante.

L‟approche que nous avons suivie est une approche sociolinguistique. Nous avons essayé d‟utiliser les implications des résultats des recherches faites sur les minorités linguistique afin de voir si celles-ci pouvaient être appliqué dans notre contexte.

Pour réaliser ceci, nous avons basé notre recherche sur la linguistique de contact. Ceci afin d'explorer des phénomènes tels que l‟alternance de codes (Code switching), le mélange de codes (Code mixing), et l‟emprunt (Borrowing). L'observation de tels phénomènes nous a menés à l'étude puis à l‟application de modèles linguistiques tels que celui de Myers Scotton avec ces différents réajustements, celui de Muysken ainsi que celui d'Auer qui nous ont aidés à interpréter un certain nombre de processus observables utilisés par les locuteurs Berbères qui sont bilingues et qui vivent à Oran. Ces processus se sont montrés très dynamiques et leurs interprétations nous ont obligé à être très critique envers les models établis qui ne marchent pas forcement dans le contexte des minorités que nous avons étudié. Nous avons essayé d‟utiliser une approche éclectique afin de contrecarrer les différents problèmes survenus a travers notre analyse du corpus.

Les objectifs initiaux dans ce travail sont d‟observer, d‟analyser et de discuter des aspects de certain phénomènes de contact entre les locuteurs minoritaires Berbères et les locuteurs majoritaires d‟arabe algérien vivant à Oran. Ces objectifs ont mené à une meilleure compréhension des stratégies et des ressources que les minorités Berbères

16 utilisent afin de communiquer dans leurs réseaux sociaux, et comment ceux-ci fonctionnent pour maintenir la langue de Berbère.

Un certain nombre d'hypothèses sont exposées dans ce travail. Elles permettent d'expliquer et d'interpréter non seulement ces stratégies bilingues (alternance de codes, mélange de codes, emprunt) en tant qu’éléments principaux qui ont un rôle déterminant à jouer dans la maintenance de la langue Berbère mais elles permettent également de s’interroger si elles favorisent le déplacement du Berbère à Oran par l’Arabe algérien.

Les principaux résultats obtenus dans ce travail révèlent que l’alternance de codes est employée pour signaler l’identité et l’appartenance à une communauté donnée (la communauté Berbère à Oran). Elle est également utilisée pour signaler l'adhésion aux différentes communautés linguistiques à Oran. Ceci s’explique par la tendance générale des informants à utiliser des mélanges de codes où le Berbère est à chaque fois présent. Les informants de cette étude utilisent aussi l’alternance et le mélange de code comme stratégie conversationnelle et discursive dans leurs réseaux sociaux.

الملخــص:

Contact Linguistics Code Code switching Borrowing mixing

Auer Myers scotton

Presentation of the thesis and research hypotheses

The present study represents an exploration in the field of sociolinguistics with particular reference to bilingual phenomena such as Code Switching, code mixing, and borrowing. It is essentially based on observation, study and analysis of social phenomena that involve not only the languages in contact in a given speech community, but also speakers’ behaviour towards one language or another. This has led us to investigate the Kabyle and Mzabi speakers in the city of Oran. This city is an urban area where for various socio-economic and even security reasons populations from different parts of the country have settled and sticked to their own verbal heritage (be it a dialect of Algerian Arabic as for the case of rural exodus, or a variety of Berber). As to the case which interests us in this particular study the communities being investigated have come to Oran for more economic and trading reasons than those of the new dwellers of the city of Oran.

The contact between these new dwellers and the population that is native to Oran triggers contacts that are not only of a social nature, but also of a linguistic one. These contact phenomena include code switching, code mixing, and borrowing. The use of these latter in the communities in question has given rise to social attitudes that led to perhaps new parameters for the sociolinguist in terms of minority majority speakers. Parameters that we shall be presenting in this work in the light of theoretical models such as the Equivalence based model as expounded by Poplack and her associates(1981, 1988, 1995, 2000), the government based model as it has been expounded by Muysken et al(1986, 1990, 1995, 2000), and insertional models as expounded by Myers Scotton and her associates(1993, 1997, 2000, 2001, 2002) in that we consider to be adequate to some extent to explain these merging phenomena in oran as a welcoming city for new comers as the case may be for all big cities in Algeria. We may suppose that these phenomena are also observed in other major urban centres.

The present thesis is fleshed out into four chapters. Chapter one presents an overall vision of the sociolinguistic situation in Algeria and the one in Oran; besides, this chapter briefly presents the sociolinguistic situation of Berber in the context that we shall investigate. In this chapter, we also presented a brief overview of studies conducted from micro social, macro social and integrative perspectives on code switching and related bilingual phenomena.

Chapter two starts with a presentation of the methodological considerations of the work. It also investigates the grammatical approaches to code switching with a particular reference to equivalence based studies. The equivalence based model advocated by Poplack (1981, 1988, 1995, and 2000) is considered with a critical view using the data of the present study.

Chapter three, which represents a continuation of the issues at stake in this line of thought, describes code switching and code mixing in the light of government and minimalist studies on the interpretation of bilingual phenomena. The theoretical constructs selected for this section of the investigation are the ones put forward by Muysken et al(1986, 1990), Belazi and Toribio(1994), Mahootian and Santorini(1993, 1996, and Mac Swan(1999, 2000, 2004).

Chapter four raises issues related to insertional patterns of code switching. A number of insertional models are tried out on the basis of the data at hand before we opt for the most appropriate model (or models) that helped us explain the sociolinguistic phenomena that we investigate in this work

Research hypotheses

The present work is based on the following hypotheses:

 Berber speakers living in Oran are minority speakers in the sense that they display contact linguistic phenomena similar to minority speakers observed elsewhere.

 Minority is not taken here to mean emigrants as observed for Moroccans in Holland. In fact they share the same religion, traditions, and culture. We shall thus use this key concept differently from the cover or blanket term of minority which is generally associated in the general English as well as in sociolinguistics with emigration,

emigrants, new settlers in a community that shares not only different languages but also different religion, traditions and culture.

 We hypothesize that the Kabyle, as minority speakers in Oran, are more prone to integration into the Oran speech community than the Mzabi speakers. This may be related to their different patterns of code switching. This leads us to conclude that Kabyle Berber is more influenced by Algerian Arabic than Mzabi Berber in Oran.

 We put forward the hypothesis that the theoretical models that we used in this work; namely equivalence, government, minimalist, and insertional ones are not totally operative to the data that we have at hand. The reason being probably that they based on sociolinguistic settings different from the one that we investigate in this work.

 We also put forward the hypothesis that language typologies have an influence on the patterns of language contact. This relates to the typology of Berber which is an agglutinative language.

General Introduction

The present research work is an attempt to shed light on a number of sociolinguistic phenomena that take place in an interactional situation where two or more varieties of Berber and Arabic ( and French)are involved and which trigger off insertional and alternational patterns of bilingual speech. We observe how these patterns combine and operate in a conversational context and how they are influenced in one way or another by the nature, time, and space where the conversational interaction takes place. A case in point would be a situation where a Mzabi speaker as a seller interacts with an Algerian Arabic speaker as a buyer. We shall see that this interactional context is characteristic of Algerian Arabic as being the sole language of interaction while Berber will be absent or perhaps slightly present in the interaction as the seller is a native speaker of Mzabi and the presence of Mzabi words or phrases will attributed in this particular case to slips of the tongue. Another illustrative case would be that of a Mzabi speaker interacting with a Kabyle speaker. The interaction triggers of a mixture of Mzabi and Kabyle as the Matrix Languages with perhaps a slight presence of Algerian Arabic. A last but not least example would be that of a Kabyle speaker interacting with an Algerian Arabic speaker where Algerian Arabic will be the language of Interaction.

Such interactional settings led the present research work to interpret instances of code switching, code mixing, and borrowing as they are displayed by the Berber minority speakers in Oran. We thus tried to see how syntactic models that have been used in the interpretation of contact phenomena in other bilingual settings may be used in the analysis of the data we gathered for this study. From this point we started to question the practicality of these models to the patterns of contact phenomena in the minorities under light here. Let us contemplate the following example: /--   “   - / très bien // ma fille /./ that we translate literally as the teacher the school she told me ‘well you are wit and intelligent very good about you my girl, carry on’. This is a characteristic utterance of multi insertional codes in which the speaker has inserted // (teacher) and // (school) that are elements from Algerian Arabic into Berber respecting the syntactic well formedness requirements of this latter. Producing such a sentence may have been

23 a demanding task which requires competence in both the inserted language and the receiving language. She has signalled a quotation using a change in Language. This is again a quite demanding task for it requires conversational competence. To add to the complexity of the utterance, the speaker has used French in alternation with Algerian and Standard Arabic. This speaker seems to be intelligent to some extent as she has got the ability to insert elements from a language to another language without violating the syntactic system of the receiving language. She also seems to have a good conversational ability as to attract the attention of the interlocutor to a piece of discourse that is not her through the use of language alternation.

This utterance is, however, an utterance which has been produced by a child aged four; besides this is not an idiosyncratic behaviour but a recurrent one in the community where the utterance has been observed. Similar structures are attested in social interactions and verbal behaviours by minority communities where more than one language or language varieties are used. They constitute data worth analysing in bilingual settings or multilingual settings where minority speakers are found. Phenomena such as these, whereby languages are juxtaposed, inserted, and take from each other, have been noticed in the majority of human societies. Speakers use different languages or language varieties in their verbal repertoire. The human mind being a highly efficient and economic system, human beings could be contend with using a single language or variety. Cross linguistic research has shown that this has proven to be an exception rather than the rule. Most speakers of the world use as least two varieties of a language in their speech.

Human beings have been continuously questioning the reasons behind this state of affairs. The interest in utterances such as these started as early as human kind; although the puzzle is still unresolved. The phenomena listed above remain one of the most challenging characteristics of human language. Their interpretation has been at the core of many studies which have been carried by many scholars from diverse disciplines such as psychology, sociology, anthropology, neuroscience and linguistics. This indicates that the interpretation of the alternate use of language, the mixture of languages, the insertion of elements from a language to another, and the integration of elements from a variety or a language to another is not a straight forward one. Neither is it a purely linguistic one. In fact, the interest in such phenomena did not start in linguistics per se, but it started in sociology, psychology and anthropology. Linguistics took it up quite recently. The field of linguistics that is involved in the investigation of the contact between language and language varieties is 24 contact linguistics. The reasons behind this late interest are related to the nature of contact itself. Linguists thought that the interpretation of contact falls out of the reach of linguistics.

The fact that speakers use different languages, different varieties or elements from languages different from their own in their verbal repertoire is related to the nature of human societies. These societies are in perpetual contact with one another. The nature and motivations of contact may be different. One of the reasons behind contact is populations’ movements. Another reason of contact may be economic exchange.

There have been movements of populations all along human history for reasons as diverse as the search for security, food, water, warm weather, economic prosperity, and the escape from armed conflicts. When populations move, they may leave their belongings, lands and assets behind them, but they do not leave their culture, customs, social organization, beliefs, and most importantly language or language varieties. These populations’ movements are particularly noticeable in bordering areas and countries. The geopolitical situation of these areas favours contact in all its forms. One of these forms of contact is a linguistic one. Contact is also attested in big urban centres where dwellers from smaller towns settle for the reasons listed above.

The linguistic picture of Algeria is a one of a multilingual country that is highly heterogeneous. It is characterized by the co-existence of different language varieties. These include namely standard Arabic, Algerian Arabic, Berber, and French. The relationships between these varieties are complex ones. This is not a case in point as all the contact situations are highly complex. These relationships include diglossia, Bilingualism and other contact phenomena. Diglossia is not the scope of our study as we will be concentrating mainly on bilingualism and its different manifestations in one the Algerian speech communities i.e. the one of Oran. This speech community is a small picture of what is taking place in the cities of Algeria. We have chosen Oran because of its being a large economic centre. It is in fact the second largest capital of the country in terms of economy and also in terms of demographic density. Many people from different regions moved and are still moving in into this city for many reasons (economic, security, housing opportunities). This movement of population to Oran has always existed and it creates linguistic contact which sometimes leads to the making up of linguistic minorities. 25

One of the minorities in question is the Berber minority. This minority has the particularity of having as a mother tongue a variety which is unintelligible with the one of the other linguistic groups of Oran. The contact of this minority with the other speech communities gives rise to bilingual linguistic phenomena involving Berber, Algerian Arabic, French and standard Arabic. These phenomena include code switching, Code mixing, and borrowing. They are studied within the field of contact linguistics.

Contact linguistics is a branch of sociolinguistics. Its scope of interest is contact phenomena. Human contact being multi dimensional, this branch of linguistics draws its methods of investigation from sociology, psychology, and anthropology. There has been a growing interest in contact phenomena in general and in the above listed phenomena in particular. In the past, linguists depicted them as deviant and abnormal; consequently, they were not studied for their own right. However; there has been a growing awareness of the importance of understanding these phenomena to the understanding of human languages, social organization, and the processing of language as a mental construct.

Contact linguistics is one of the most interesting and challenging branches of linguistics. It highlights the dynamicity of languages in contact, the ability of the human brain in processing different linguistic systems, and the nature of human contacts that are rarely peaceful. As it has been stated above contact linguistics is a hybrid field. We believe that conducting our study within a contact linguistics perspective will permit us interpret the contact phenomenon that are displayed by bilingual speakers in the city of Oran as an urban centre. It will allow us understand more clearly the behaviour of Berber minority speakers in the main stream of Oran.

The present study concentrates on two Berber minority groups of Oran (The Kabyle, and the Mzabi groups). These two minority groups have been chosen because they outnumber the other Berber minorities of this city. We shall be investigating the sociolinguistic situation of Oran with a special reference to bilingual contact phenomena as they are displayed by the members of the communities under light here i.e. The Kabyle and the Mzabi communities. The contact of these

26 minorities with the Algerian Arabic majority speakers triggers contacts between Berber, Algerian Arabic, Standard Arabic and French. The outcomes of these contacts are contact phenomena such as code switching, code mixing and borrowing.

We shall be observing code switching, code mixing and borrowing as they are displayed by Mzabi and Kabyle Berber minority speakers in Oran. Our work not only investigates these phenomena from the linguistic point of view, but it also looks at Berber speakers’ behaviour and attitudes towards the languages that are in use in their verbal repertoire and the ones of Algerian Arabic majority speakers. These attitudes and Behaviour are important to our understanding of the terms majority/minority speakers.

The work is partly based on the assumption that the way we will use the concept ‘minority’ is different from its common use in the sociolinguistic literature. Unlike minorities investigated elsewhere, the minorities being investigated throughout this work share the same religious beliefs, culture, and social organization as the Algerian Arabic majority speakers. we shall be trying to test the underlying hypothesis from which this work generated i.e. Berber speakers in Oran may be considered as a minority group in the sense that they display sociolinguistic phenomena similar to the one used by minority speakers in other parts of the world. The example may be given of the Turkish minority of the Netherlands. This minority displays code switching involving Turkish and Dutch. In addition to that Turks have social, cultural and, religious beliefs that are different from the Dutch majority speakers. The same holds for the Indian minority of England for example. In spite of that, Berber speakers manifest language patterns, attitudes that may be comparable to the ones of minority groups in other sociolinguistic contexts.

We shall also try to understand the strategies used by the informants of this study i.e. Mzabi and Kabyle informants to maintain their mother tongue in a sociolinguistic environment that is far from being favourable for that. The work also investigates the grammatical interpretation of these phenomena using theoretical framework that have been commonly used in the interpretation of the above cited phenomena in bilingual contexts in general and in minority group environments in particular.

These theoretical models include the equivalence based model as it has been expounded by Poplack and her associates (1981, 1988, 1995, 200), the government based models as they have been advocated by Muysken et al(1986, 1990, 1995, 2000), minimalist models as expounded by Mac Swan (1999, 2000, 2004), and insertional models as expounded by various scholars such as Myers Scotton (1993, 1997, 1995, 1998, 2000, 2001, 2002), Boumans(1998), Backus (1996). We will consider these different perspectives to code switching in the light of the bilingual phenomena as they are displayed by members of the Kabyle and Mzabi minorities of Oran. Our morphosyntactic analysis of code switching and related phenomena centres around the possible applicability of the above cited models on the bilingual context of Oran. As it has been stated above, the sociolinguistic context in which the Berber minorities under light here evolve is different from the ones in which the theoretical models we shall be using have been devised and tested. Throughout our work we shall consider the universality of the different models presented above in the light of possible violations to the predictions of the models as this have been claimed by the scholars who devised them.

The equivalence perspective to code switching code switching and borrowing as it has been expounded by Poplack and her associates (1981, 1988, 1995, 1998, and 2000) advocates that code switching involves symmetry between the linguistic systems participating in bilingual processes. Symmetry is used by Poplack and her associates to mean that the languages participating in code switching have to respect the syntactic requirements of one another. Symmetry is also related to the degree of bilingualism of the speakers using code switching and related phenomena. Poplack and her advocate that, in order to code switch, bilingual speakers no to have enough competence in the two languages as to produce well formed stretches of talk from these languages. The equivalence model also stipulates that the elements that are worth analysing and that are therefore considered as code switchings are not singly occurring code switchings. Poplack and her associates have been making a strong point for the distinction of code switching and borrowing.

The government binding, and minimalist perspectives to code switching as they have been advocated by Muysken et al (1986, 1990, 1995, 2000), Mac Swan(1999, 2000, 2005), and Mahootian(1993, 1996) stipulate that bilingual patterns of language production in general, and the ones of code switching in particular are generated by the same mental processes as monolingual ones. They, thus, obey the same generative rules as their monolingual counterparts. Scholars working within the government perspective argue that elements in a sentence be it a monolingual or 29 a bilingual one are bound with one another. These bounds operate between heads and their governed elements at different levels within the sentence. The minimalists go even farther and argue that code switching is not governed by any bilingual constraints apart from the requirements of the language from which the elements are drawn.

The notion of asymmetry is at the core of the insertional approaches (Myers Scotton, 1993, 1997, 2000, 2002) to code switching, borrowing and code mixing. This asymmetry operates at different levels during the production of bilingual utterances. The first level on which it operates is the one of the language systems. One of the systems participating in code switching is the base, or the dominant language. This language establishes the morpho-syntactic frame of elements within the bilingual strings. The second language is the dominated, embedded or inserted language. The second hierarchy operating during the process of insertion is the one between functional and content elements. Functional elements are supposed to be from the base or matrix language. While content elements may be drawn from both codes.

Our work also explores the possible interaction between the typologies of the languages involved in the above cited bilingual phenomena and the bilingual patterns of code switching, code mixing, and borrowing in the minorities under light for this study. The influence of language typologies on code switching and related phenomena is being considered as a promising avenue of research. Scholars haven’t started until quite recently to explore this area. The preliminary findings presented by some researchers who pointed to this direction of research are amazing. Muysken (2000: 158) points to the relationship between language typology and borrowing types in agglutinative languages. We shall consider this avenue of research in the light of the typological characteristics of Berber, Algerian Arabic, Standard Arabic, and French. We hypothesized that as Berber is an agglutinative language we would normally obtain patterns of borrowing, and code switching that are similar to the ones observed in language pairs which involve agglutinative languages.

The models presented above are not the only models which treat bilingual sociolinguistic phenomena. Our choice to deal with these particular models and not with other ones does not mean that these are the right models that may be applicable to all bilingual contexts. Code switching, code 30 mixing and borrowing, as they are displayed by Kabyle and Mzabi speakers, are not always rule governed. At least they do not always follow the predictions of the models listed above.

Some instances of code switching do not have any type of marking from either languages, or language varieties. Patterns of code switching such as these clearly stand in the face of most models listed above. This relates to the fact that in any language, items need to have some kind of marking. Investigation of bilingual data indicates that this is not valid in all sociolinguistic environments. Our data contains many elements that lack the necessary inflections or functional elements. We hypothesized that this may be due to the morphologies of the languages involved in bilingual linguistic processes. Another possible interpretation may be related to a lack of congruence or mismatch between the languages being in contact.

In cases where the grammatical models do not apply, we shall try to look at alternative avenues of research to interpret the violations or the problematic bilingual instances. One of these alternatives is the alternational one. It is a quite interesting perspective in the analysis of code switching and related phenomena. Scholars working within the alternational perspective argue that all types of bilingual phenomena are worth analysing. The alternational approach to code switching also argues that the juxtaposition of languages together or the use of elements from a language in the boundary between sentences produced in another language is interesting. The relevance of the alternational perspectives to bilingual phenomena resides in their ability to interpret patterns of code switching that are considered as unworthy of analysis by scholars dwelling within the insertional paved paths.

The present work also investigates patterns of bilingual phenomena in elements of language that are salient at the supra-segmental level i.e. the ones that are beyond the sentence level i.e. discourse markers, complementizers, coordinators, and tags. These elements seem to have a functioning that is different from the elements that are within the boundary of the sentence. We have noticed that bilingual phenomena involving such elements are a recurrent feature in the speech of the Berber minorities that we have investigated. Their investigation will be conducted using theoretical frameworks that are different from the ones listed above (the equivalence, the

government, and the insertional models) in the sense that their interpretation requires a different perspective in research.

Our work is subsumed under a micro sociolinguistic perspective to contact phenomena. The approach that we shall be using in the analysis of our data is to its majority a qualitative one. The interpretation of the bilingual phenomena listed above will be made using individual utterances that have been produced by informants in the data. The qualitative perspective in the analysis of code switching and related phenomena has shown its efficiency in the interpretation of bilingual phenomena. These phenomena are often called strategies by scholars working within this field. This makes generalisation in this field of research quite difficult. The qualitative perspective in research relies rather on the intelligence and the perspicacity of the researcher to induce the generalisations from individual language production. It also grants him more freedom in the approach that he/she applies to reach the results wished at the beginning of his research.

Our investigation is however not entirely qualitative, as it uses statistical quantitative procedures when possible. The challenge to the use of a quantitative procedure in the analysis of code switching and related phenomena relates to the practical problem of generalisations from individual utterances. We shall be using a quantitative perspective to interpret the bilingual patterns that are recurrent in the data. Bilingual research has shown that the only part of a research in sociolinguistics is the one in which there are recurrent observable features that may be quantified.

The data gathering follows the general procedures applied in studies carried in bilingual environments. Members of the two minority groups in light here were recorded in natural situation of daily life. The recordings were carried by one of the informants or by the investigator without the knowledge of the speakers. Though this procedure of data gathering has been questioned for ethical reasons, we thought that this seems to be the only way to obtain spontaneous natural speech. Such types of recordings are considered by the bulk of researchers working in the investigation of bilingual phenomena as a reliable procedure in data collection. Another elicitation technique could have been interviews. We thought that using interviews with the Berber minority would have been a bit delicate. One of the minorities under study in the present work i.e. the Mzabi minority is said to be a close minority. 32

So far, we haven’t discussed the social insights of the work. We consider that we ought to give it justice in this part of the paper. The investigation of the social functions of code switching, code mixing, and borrowing in the Kabyle and Mzabi minorities have been studied in the light of a micro perspective. The approach advocated is based on a brief review of the literature in relation to the social, the discursive and the conversational functions of bilingual phenomena. The perspective that we take in our overview of the literature is a diachronic one. We visited the social studies on bilingual data. However, we only chose the studies that we consider pertinent to our work. The implied hypothesis of this part of the work is that Berber minorities in Oran may be compared to other minority groups all over the world in relation to their social conduct as a minority group in a majority of Algerian Arabic speakers.

The present investigation of bilingual phenomena does not consider the phonological and phonetic side of the Berber varieties. We are rather interested in the morpho-syntactic functioning of the two varieties under light here i.e. Kabyle and Mzabi Berber. These varieties alongside the other Berber varieties do share many morphosyntactic features which are the main interest of the present work. There are however differences between them. These differences are not the focus of the present work. Consequently, we will be using a simplified phonological transcription for the instances of code switching in our investigation.

Throughout the last three chapters and even in parts of the first chapter, we will be using Berber as a cover term to refer to both Kabyle and Mzabi. In cases where the differences are morphosyntactic which represent exceptions, the Berber variety used will be signalled. In the same vein we will be using the term Algerian Arabic as a cover term to the varieties of Arabic used by Berber speakers in the study. The variations between the varieties of Arabic used by the Kabyle and Mzabi minorities of Oran fall out of the scope of this study.

Chapter One:

A preliminary to the study

1.1 Introduction

The present Chapter is a preliminary to our investigation of the different perspectives in the analysis of code switching in the light of a Berber-Algerian Arabic-Standard Arabic-French data. As it has been stated in the general introduction, our study is couched in the sociolinguitic investigation of code switching and other contact phenomena. We will investigate these phenomena as they are displayed by the Berber minorities of Oran.

We believe that a non-negligible part of our study will be devoted to the grammatical investigation of code switching (the next three chapters). Our choice has been guided by the fact that doing both a grammatical and a social investigation of a phenomenon as complex as code switching falls out of the scope of a doctoral dissertation. However, we will try to give the social investigations of code switching and other contact phenomena some justice in our study.

We truly believe that the social aspects of language in general and contact linguistics in particular are of a paramount importance. These social aspects are quite important for our understanding of the social groups’ dynamics operating in the minority groups that we will be investigating.

The present chapter is devoted to the scholarly interests in the social factors that intervene in language contact situations in general. Throughout our presentation, we shall keep in mind the main goal of the present study. We aim at interpreting the contact phenomena as they are displayed by the Berber speakers in Oran. The studies that we will be presenting are the ones that we consider as relevant to our understanding of the social conduct of the Berber minorities under light in this work. Consequently, this chapter will, be devoted to a brief overview of the literature in relation to the contact phenomena that we will investigate i.e. Code Switching and code mixing. Nevertheless, before we engage in this enterprise let us present some preliminary geographical and historical informantion on Berber

35 in Algeria with a particular interest on the minorities we will be working on along this study i.e. the Kabyle and Mzabi minorities.

1.1 Berber in Algeria

Berber is one of the official languages of Algeria. It is the mother tongue of approximately 30%of the population1. The different Berber varieties used in Algeria are distributed as follows:

 Takbaylit: It is the variety used in the central north of the country (approximately 100kms east of Algiers). This variety is quite important in the sense that two quarters of the total number of Berber speakers in Algeria are Kabyle2 . The Kabyle community is also important in the sense that its members have been the leaders in the continuous demands for the official recognition of Berber as an official language in Algeria. This community is made up of speakers living in the cities of Tizi Ouzou, Bejaia Boumerdes, and Bouira

 Tashawit: This variety is used by a group of speakers who Dwell In the Aures mountains area in the eastern south of Algeria( Souk Ahras, Batna, and khenchla) this is again another important group in terms of the number of speakers3

 Tamzabit: It is the variety of Berber used In the Mzab valley which is situated in the centre of Algeria. This variety is also of an importance because of its high Ethno linguistic vitality and because of its speakers who may be regarded as a leading group in the trade life of the country. Members of This group settled in many big urban centres of the country. They seem to have managed to establish minorities and to maintain their mother tongue as a minority variety.

1 The percentages are taken from Haddadou (2000 :22) 2 Cf : Chaker Salem in Jochen Pleines( 1991 :238) 3 Unfortunately there are presently no exact findings on the number of tashawit speakers in the Aures 36

 Tamahaq: This variety is used By the Touaregs in the Ahaggar and the Tassili areas in the south of Algeria. Tamahaq is also an important variety of Berber in the sense that it is considered by specialists of Berber as the variety which has been best preserved from the influence of contact with other languages such as Arabic4. Tamahaq is also important in the sense that it is also the mother tongue of large groups of speakers in Mali and Niger.

 Tachenwit: The variety of the Chenoua in the cities of Tipaza and Cherchell to the west of Algiers. This speech group is of lesser importance because of its small number of speakers compared to the important groups such as the Kabyle, the Chaouia, and the Mzab groups.

 There are other small Berber speaking populations of a lesser importance because of their small numbers. These include the Berber populations living in Ouargla, (the Touat’ Ksour), Bechar, Ain Sefra, The Gourrara’ Ksour in Timimoun, Jbel Bissa (in the Area of tenes).

 In addition to the above cited Berber speaking areas we may include the Berber minorities living in different cities of the country (mainly the biggest urban and economic centres which include Algiers, Oran, Constantine, Annaba, etc…). These minorities are as important as the speech communities living in Berber areas in the sense that they play a paramount role in the overall maintenance of Berber in the Maghreb in general and in Algeria in particular.

The Berber speaking areas do not make up a linguistic or a geographical continuum as they are often separated by large Arabic speaking areas. They are also separated by natural barriers such as mountains and deserts. Some scholars(Chaker: 1997, Haddadou: 2000) discuss the idea that one may talk about Berber languages in the sense that the Berber varieties share a low degree of mutual intelligibility. In the next section we shall be tackling this issue of uniformity in the different varieties of Berber.

4 Cf : Chaker (1996 :225) 37

From a historical point of view Berber belongs to the Chamito Semitic family of languages5 which also includes Arabic, Hebrew, ancient Egyptian, and other African languages from Tchad and Nigeria. The relationship between these languages has been put forward by many scholars such as Cohen (1924: 10, 1947:42) and Cohen (1975:52). Present Berber is said to be one of the daughters of Libyan which was attested in North Africa during the antique era (Haddadou 2000:224). There are few records or descriptions of Libyan for it was considered by the Phoenicians and by the Roman colonisers as the indigenous language of the North African population or the Barbarous as the Romans used to call them (Haddadou 2000:20). There is no consensus among specialists on the uniformity of Libyan. Haddadou (2000: 25) and Chaker (1995: 223) argue that it has many varieties due to the fact that the speakers using that language scattered in large areas stretching from morocco to Libya. Their proponents put forward the hypothesis that there was a Proto Form of Berber which used to be uniform and which used to be spoken in North Africa (militarev1991:125). This debate hasn’t been settled up to now as the controversy has moved in the present time to Berber varieties which are considered as the descendants of Libyan. Boukous (1991:26) talks about a process of dialectisation whereby Libyan has undertaken diachronic changes on the phonological level due to linguistic inherent causes which are not a particularity of Berber. These are processes all languages of the world undertake in their history. Boukous (ibid: 26) puts forward the example of the consonantal movement from stops to affricates to fricatives to glides.

E.g.: the changes that happen to Stops which become affricates such as the change of the stop // as in / /, to the affricate / / as in / / (butter).

Other scholars such as Chaker (1996: 217), and Haddadou (2000: 67) partly explain the change with causes that are extra linguistic i.e. sociolinguistic. They relate variations to the fact that Berber speakers may have few opportunities of inter regional contact because of the long distances which separate them. Another sociolinguistic feature is that the Berber communities are separated by large communities of Arabic speaker which make up the majority in addition to the fact that the political regimes sometimes stigmatize Berber as a minority language.

5 Cf Chaker(1996:201) 38

These regimes are different in the way they tackle the issue of The Berber minorities and do not have a unified vision about this given issue. The Algerian, the Moroccan, and the Libyan regimes are strongly turned to the Arab nation and officially consider their nations as Arab whereas the Mali and Niger governments are rather turned to France and to western Africa (Chaker1991:140)6. These are strong factors which may result in different evolutions of Berber in These countries.

In Addition to the linguistic and the sociolinguistic causes of variation in Berber, we may add the factor of long lasting contact that Berber has always had with different languages throughout its history (Punic, Phoenician, Latin, Arabic, Spanish, Turkish, and French). The balance of power during the different contact has hardly been in favour of Berber as the other languages have always had more power and prestige as they were the languages of the conquerors and the colonisers. Chaker7 (1991:205) asserts that Berber has always been downgraded because it always lacked a written form and a written body of literature and because it relied on the Oral transmission of customs and cultural heritage. This did not prevent Berbers to write literature in the conquerors languages (Phoenician, Latin, and Arabic)8. Even during the period of Berber dynasties such as the ones of Massinissa, Jughurta, and Al Kahina the ruling affairs were carried out in varieties other than Berber namely in Latin during the reign of Massinissa and Arabic during the reign of Al Kahina( Haddadou 2000:87).

1.1.1Berber Varieties or Berber languages

The issue of having the different varieties of Berber as regional varieties or as languages per se has attracted many researchers in the field of language classification, the one of Semitic studies, and more lately the one of Berber studies. Vicichl (1991:77) takes a strong position by considering Berber varieties as languages per se. He goes even further by comparing them to the roman, to the Germanic and the Slavic languages in Europe. The argument put forward by

6 In Pleines ( 1991 :204) 7 In Pleines( 1991:205) 8 Haddadou(2000: 45) mentioned many famous Berber authors such as Apulée(around 125BC), Saint Augustin Who was a famous Christian priest ( 391), and abu al kacem Who was a specialist of Islam and a fine theologian ( 1417) 39 this scholar is that of the lack of mutual intelligibility between the different Berber languages as he puts it:

“…Les berbères qui habitent loin les uns des autres ne se comprennent pas et parlent arabe ou français….”9

Following the same line of thought, Miltarev (1991:91) does not consider the Berber varieties as a dialect continuum but as a big language family with a history of more than 3000years. His argument is based on recent scientific methods of glottochronology.

On the other hand Basset (1969:42) an eminent figure in Berber studies takes an opposite position and states that

“…. Berber is only one language and each dialect is just a regional variety of it….” 10

Chaker (1996:6) also clearly asserts that recent research is corroborating this view as he puts it:

“….Tout les travaux récents confirment les constats et enseignements classiques de la berberologie française: L’enchevêtrement trans-dialectal infini des isoglosses; la variabilité interdialectal très grande… Traits caractéristiques qui interdisent de considérer, sur des bases strictement linguistiques, le berbère de telle ou telle région comme langue particulière11. ”

9 „The Berber speakers who live far from each other do not understand each other and use French or Arabic” translation is mine

10 The translation is mine. The original text reads as follows „ la langue berbère est une et chaque dialecte n‟en est qu‟une variante régionale‟ 11 “all the recent studies confirm the classical assumptions made by the French Berber studies: The infinite trans dialectal interrelations of the isoglosses; The high interdialectal variability, even in the central points of linguistic structures… characteristic features which prevent us to put forward that the Berber variety of a given region is a language” translation is mine. 40

Chaker (ibid: 8) talks about a hierarchy in Berber. The different varieties of this language may be divided into two groups being in a hierarchical relationship. This may be based on the degree of mutual intelligibility and also on the distance between the areas in which these varieties are used.

 Regional dialects: these have some degree of mutual intelligibility. They are commonly known by the Berber speakers as being the most important varieties (Kabyle, Chleuh and Tamazight12). They are also separated regarding the geographical point of view  Local dialects: These are characterized by phonetic, lexical and even grammatical variations which are less salient than the ones between regional varieties. These variations signal the origin of the speaker E.g.: The Kabyle used in Bejaia, The one used in Bouira and the one used in Tizi Ouzou are not regional varieties of Berber but are rather local varieties of Kabyle.

Chaker (1996:10) also asserts that in spite of the variability in the Berber varieties the core phonological inventory is the same one in all these. This core phonological inventory has been established by basset in 1945 and it has been accepted by the specialists since then. We may mention the opposition voiceless/ voiced which exist in almost all the varieties e.g. // and // and the vowel system which is a triangular one /a/, /i/, /u/. Chaker also gives syntactic features such as the cyclical morphology which combines a verbal or a nominal stem with cyclical morphemes showing aspect, number and gender on both sides of the stem

E.g. // (a cat) and // (a pussy cat) here the stem is // and to obtain its feminine a cyclical feminine gender morpheme i.e./t……..t/ is added.

We may also give the example of prepositions and verb particles which are very similar in spite of slight vowels or consonants variations

12 Chleuh is the variety of Berber used in the south west of Morocco. Tamazight is the variety of Berber used in the centre of Morocco 41

E.g., the future particle /ad/ has regional variants such as /ad/ in Kabyle /a/ in Mzabi. Another good example would also be the prepositions which are very similar in the different varieties of Berber

E.g. between is produced // in Kabyle, // in Mzabi, and // in Tamahaq

From the lexical point of view we may also say that there is variation in the Berber varieties Chaker(1996:15) asserts that it is the level in which the different varieties are most divergent but this is again relative as the divergence is sometimes due to slight differences in the semantic functions given to such or such words. This means that the words are used with different meanings but if one looks closer to these differences he will notice that the uses are in reality semantically part of a similar field.

E.g. // means house in Moroccan Berber varieties. It comparatively means family or domestic unity in Kabyle. These two lexemes are related to each other as they belong to the same semantic field which is the one of the home.

From the sociolinguistic point of view the question of whether to consider Berber varieties as different languages or as entities belonging to the same language becomes more salient as the debate becomes more complex and the question more debatable because of its subjectivity. The complexity of the question lays in the fact that the different Berber speaking communities are not united in one geopolitical sphere i.e. they are scattered in morocco, Algeria, Mali, Niger, Libya, and Tunisia. The other source of complexity is the question of whether a Berber common self and consciousness exists or not. Some specialists of Berber (Chaker, 1991: 120; Haddadou, 1990: 54) put forward the premise of a pan Berber unity but, this is still a debatable question which need to be verified more objectively.

1.2 The present Study

As it has been stated above in the introduction of this chapter, the present study will be conducted on Berber from a contact, and a sociolinguistic perspective. Our investigation will be devoted to the study of contact between the Berber minorities of Oran and the Arabic majority speakers in this city. We have chosen to deal with this minority because it has as the peculiarity of being different from the other Arabic speaking minorities of Oran. In addition to French, Standard Arabic, and dialectal Arabic, It has at its disposal Berber. This permits it to display contact phenomena and strategies that no other minority could display.

The present study also aims at understanding and interpreting the different bilingual resources and strategies used by this minority or minorities to socialize, to communicate, and to maintain their inner social networks operative in an environment which may be hostile to the maintenance of their language. The questions we will be trying to answer are these ones: Are members of this minority going to survive as a minority or are they going to assimilate to the mainstream of Arabic speaking communities? Will bilingual strategies such as code switching, Code mixing, and Borrowing have a role to play in the maintenance of the variety of this minority, or will it favour its displacement by Arabic the language of the majority in such a context? And it would be also interesting to see whether the contact phenomenon cited above obey rules. We will also be trying to correlate these contact phenomena to the social context of the community in question. But before we go in the details of our study we would like to have a brief overview of the literature in language contact research in addition to some brief clarifications about the concepts we will be using. We will start by the term code switching that we shall be using all along our study.

The issue of terminology in language contact research is an important one in the sense that there is no consensus among specialists on the operating terminology to be used. Not all researchers’ use the same terms in the same way, nor do they agree on the territory covered by terms such as code-switching, code-mixing, borrowing, or code-alternation. In particular, at issue here is the difference made between code-switching and code-mixing. Several criteria have been proposed to distinguish between these two technical terms. But before looking at

43 them more closely in the coming chapters, we shall consider the definitions of these concepts as they have been put forward by specialists. We also shall try to make our way through the different definitions given by the community of researchers in contact linguistics and propose a working definition of these and other concepts that we shall be dealing with throughout our study.

The term code is a relatively neutral conceptualization of a linguistic variety be it a language or a dialect, However; not many researchers really explicate this concept in their definitions. Milroy and Muysken (1995: 7), two important figures in contact linguistics research, state that CS is:

“The alternative use by bilinguals of two or more languages in the same conversation”

They (ibid: 17) use code switching as a blanket term under which different types of bilingual behaviour are implied. The term intra-sentential code switching is used to refer to the switching within sentence, in contrast with the term inter-sentential code switching, which is on its turn, the switching at sentence boundaries or between sentences. These two concepts are used by Milroy and Muysken as the units of analysis. Myers-Scotton (1993b:1) also uses code-switching as a cover term and defines it as:

“The alternations of linguistic varieties within the same conversation”

We shall come back to the distinction between intrasentential and intrasentential code switching in due time when we will be dealing with the structural analysis of code switching in the coming chapter. Other researchers working on dialect switching such as Gardner-Chloros (1991:69; 1995: 8113) and Romaine (1995:121) also emphasize that switching can occur not only between languages, but also between dialects of the same language. Romaine (ibid: 122)

13 In L.Milroy and L. Wei ed (1995:81) 44 talks about “monolingual code-switching” extending the meaning of code-switching to include what others (Labov, 1966:301; Trudgill, 1974:45) would call style-shifting in monolingual speech as she puts it:

“I will use the term ‘code’ here in a general sense to refer not only to different languages, but also to varieties of the same language as well as styles within a language” (Romaine, 1995: 122)

Gumperz (1982:59) refers to the concept of code switching with more neutrality as:

“The juxtaposition within the same speech exchange of passages of speech belonging to two different grammatical systems or subsystems”

We will be using the term code-switching as a cover term unless otherwise stated throughout our study to cover the phenomena of alternating between two languages or dialects of the same language within the same conversation. At times we will also refer to this phenomenon as code-alternation with a similar meaning to the one that has been adopted by conversational analysts such as Auer (1995:119), Gafaranga (2000:310), and Wei (2000:1). These researchers use the term code-alternation as a hyponym to replace CS, but it is marginally used in that sense by the community of contact linguistics researchers.

The term alternation is, in fact, used in the literature to refer to instances of one language being replaced by the other halfway through the sentence, and it is mostly, but not always, associated with longer stretches of CS. This is what is commonly known as the alternational approach to code switching whereby large elements such as phrases, clauses, and even sentences from language ‘A’ are used alternatively with other large elements from language ‘B’. This state of affairs is opposed to other situations where relatively small singly occurring elements of language ‘A’ (nouns, verbs) are inserted in otherwise larger structures from other languages (sentences, clauses, and phrases). The term insertion is then used in such situations. 45

Insertion and alternation represent two distinct but generally accepted processes at work in CS utterances (Muysken, 1995, 2000).

Others (Sridhar & Sridhar, 1980; Kachru, 1983; Singh, 1985), however, reserve the term code-switching for inter-sentential switches only, and instead prefer to use code-mixing for intra-sentential switches. The reason is that only code-mixing (i.e., intra-sentential CS) requires the integration of the rules of the two languages involved in the discourse. But as far as the structural constraints are concerned, the intra- vs. inter-sentential distinction can equally well distinguish the two types of switches. This terminological issue remains a matter of individual preference, but at the same time it creates unnecessary confusion.

Other researchers such as Muysken (2000:320) avoid using the term code-switching as a cover term because they believe that switching suggests alternation only, as in the case of switching between turns or utterances, but not necessarily insertion. Instead, they prefer to use code-mixing to refer to intrasentential code-switching and to use code switching to cover instances of intersentential code switching. We shall come back to this thorny issue when we will be dealing with the different approaches to code switching in chapter three.

The other debatable question is the one which deals with the distinction between code switching and borrowing. Pfaff (1979), along with Poplack, raises the question of the need to distinguish between code-switching and borrowing. This distinction attracted since then most researchers dealing with the two contact phenomena (Gumperz, 1982; Gardner Chloros, 1995; Muysken, 1995; Backus, 1998:74; Myers Scotton, 2002). This is a much more complicated issue than the perceived distinction between code switching and code mixing. We shall come back to the differentiation between the two phenomena in chapter two.

1.4 A Brief Overview of the Literature

Contact linguistics is a branch of linguistics that specialises in contact phenomena. Its main interest is bilingual or multilingual contact situations. In the past this area of research was the concern of sociology, Anthropology, and psychology. More recently it has been taken up by linguists and sociolinguists. This has been due to a growing interest and awareness of the important role language plays in the shaping of the personalities of individuals and the social constructs in not only bilingual but also monolingual communities. This field of sociolinguistics is challenging for it deals with actual language use. It is a hybrid field that takes its strengths and methods of analysis from linguistics, Psychology, sociology, and anthropology.

1.4.1 Early prescriptive approaches to code switching

Contact linguistics has been integrated as a branch of sociolinguistics relatively recently because in the past contact induced phenomena were regarded by the bulk of researchers as deviant in nature and as irregular. This argument was partly based on the principle of the ideal monolingual speech community advocated at that time. Many prominent researchers overtly claim that code switching is aberrant and even not worth investigating. Bloomfield (1927: 395) in a description of the linguistic profile of a Native American speaker makes this observation:

“White Thunder, a man around 40, speaks less English than Menomini, and that is a strongIndictment, for his Menomini is atrocious. His vocabulary is small, his inflections are often barbarous, and he constructs sentences of a few threadbare models. He may be said toSpeak no language tolerably.”

It appears from the observation that Bloomfield does not see his informant as a fully competent in either English or Indian. He defined this speaker as a semilingual14

Bloomfield was not the only linguist who had such a dramatic view towards code switching. There was a tendency during his era to consider this phenomenon as deviant, random and not worth studying. Espinosa (1917: 408) was one of the pioneers in Spanish- English code switching studies in the United States. He (Ibid, 408) overtly dismisses this phenomenon as random and states that:

“Code Switching was just a random mixture of the languages available to a bilingual” (Espinosa, 1917: 408)

This negative prescriptive15 view about code switching was a long established view among specialists, and it was reiterated by many other important scholars. Similarly Haugen (1950: 211) defines code-switching stating that:

“The speakers may switch rapidly from one language to the other but at any given moment they are speaking only one, even if they resort to the other for assistance”

14 Semilingualism may be considered as the popular belief that bilingual speakers who code-switch do so because of their lack of linguistic competence in their repertoire (Edelsky, Hudelson, Flores, Barkin, Altwerger, & Jilbert, 1983).

15 Prescriptivism is known as the view that one variety of language is given an inherently higher value than other varieties and that this ought to be imposed on the whole of the speech community, especially through educational means (Crystal, 1997).

Haugen (Ibid: 38) also argues that “the speaker is to be speaking only one language”. He (Ibid: 38) explicitly states that items in code switching must be assigned to only one of the languages involved. Accordingly,

“Any item that occurs in speech must be a part of some language if it is to convey any meaning to the hearer” (Ibid: 38)

This view of code-switching is in accordance with the tendency of the time in the sense that it is similarly monolingual in perspective. In this view, it was crucial to define which language a bilingual is using at a given moment, i.e., the bilingual is supposed to use only one language at a time even if he inserts elements from another language. Haugen (1956:40) defines code-switching as:

“A linguistic process which occurs when a bilingual introduces completely unassimilated words from another language into his speech”

He integrates the term into a continuum of diffusion which also includes interference and integration

“Switching being the first stage in the continuum and meaning the alternate use of two languages, Interference regarded as the overlapping of two languages and integration being the regular use of materials from one language in another” (Haugen, 1950: 41)

During this period there were other challenging approaches to code switching. These were mainly psychological, phonological, and social approaches. We may mention the studies of Jackobson (1952), Fante (1952), Halle (1952), Fano (1948), and the one of Hoijer (1948).

Jackobson, Fante and Halle (1952)16 refer to the phenomenon of switching code based on Pike and Fries’ (1949) work on co-existent phonemic systems. They tried to demonstrate that two or more phonemic systems may exist in the speech of a monolingual. The evidence that was presented was based on the existence of phonemes which were alien to the speaker’s inner phonemic system. One may cite as an example, the case of Arabic phonemes in French loans words such as // (a board) or // (a table).

Hoijer (1948)17 established the pair of concepts “phonemic alternation” and “phonemic alteration” which roughly parallel those of code-switching (where phonological systems alternate), and borrowing (where an aspect of the target language grammar is altered). These studies were all directed towards the investigation of code switching and Borrowing but no explicit mention was made to these words in any of them. One of the first explicit mentions to code switching is found in Vogt (1954:36) who considers that language contact phenomena are natural and quite common. His approach is considered as a starting point in the variationist- based approaches to code switching and language change (Muysken, 2000: 12).Vogt (ibid: 368) assumes that:

“Bilingualism is a universal phenomenon, since no languages we know have been spoken over long periods of time in complete isolation. It is even possible that bilingualism is one of the major factors in linguistic change- a point of view which could be defended by good arguments.”

Vogt (1954) interest was not really directed towards code switching proper, but it was rather directed to the effect of the alternate use of languages on linguistic change (Backus, 1998: 23). Vogt (ibid: 46) also considers this process from a psychological perspective. He defines it stating that:

16 .In C. Caccamo (1998 : 41) 17 .In C. Hoffman (1991:90) 50

“Code switching in itself is perhaps not a linguistic phenomenon but rather a psychological one and its causes are obviously extra-linguistic”

Fano (1948)18 also argues that switching code is a strictly psychological phenomenon. He states that it consists of altering one’s internal speech mechanisms for the identification of phonemic symbols. He defines the term as

"A mechanism for the unambiguous transduction of signals between systems"19

Following the same line of thought, Jackobson (1961: 142) accounts for language alternation between French and Russian among 19th century Russian aristocracy. He integrates the notion of switching code into a broader information theory. Codes are meant according to Jackobson (ibid: 144) to decode and decipher the speech of others. They are not considered as the speech material itself but as psychological mechanisms at the disposal of the bilingual. All the above cited investigations of code switching are directed towards psychological and phonological interests. Now let us turn to the socially centred approach to code switching during this period.

1.4.2 Macro social studies on Code Switching during the 50ths and the 60ths

One of the first trials to relate code switching to social issues has been made by Barker 1947) in his work on the Mexican American community of Tucson, Arizona in the United States. Barker tried to correlate the distinct varieties used by this Mexican American community (Southern Arizona, Standard Mexican, Pachuco, and Yaqui, which are Spanish varieties and standard alongside sub-standard English) with identity. Barker study (1947: 12) focuses on the symbolic values attributed to these varieties by the members of this minority

18 .In C. Caccamo (1998 : 42) 19 in C. Caccamo (1998: 42) 51 group and the other linguistic groups in Tucson. He (ibid: 147) put forward the hypothesis that English, and Mexican Spanish are used as identity markers by Tucson residents stating that

“For individuals both inside and outside the ethnic group, the ethnic language comes to symbolize the group and its cultural background, or in terms of its social function to identify the group as a group… the individual’s skill in using the language of a second culture comes to symbolize his status in the new community.” (Barker, 1947: 187)

Barker makes the observation that intimate family discussions tend to be carried I Spanish, whereas the discussions about formal issues tend to be carried in English. In addition to this differing functional distribution, Barker (ibid: 188) talks about the youngsters tendency to mix elements from Spanish and English into their discourse as a sign of multiple identity construction. He argues that:

“In a field of informal relations among bilinguals, a field in which most of the social life of the younger Mexican- Americans takes place, we find that rapid shifting from one language to another is common. Often two languages may be used in the same sentence or phrase…It may be said that the mixing of the two languages is indicative of he speaker’s participation in the urban social life of the younger native-born group in the Tucson-Mexican community” (Barker 1947: 195-196).

Weinreich (1953: 120) strongly criticizes Barker’s (1953) vision of code switching and states that the latter’s division is not explicitly articulated; besides, Weinreich (ibid: 120) argues that this taxonomy of speech i.e. intimate language, informal, formal, and inter-group discourse, does not explain the motives of the younger members of the Tucson Mexican- American community when they alternate Spanish and English in the same conversation. Weinreich (1953:45) strongly put into question the possibility for bilingual speakers to switch languages, and put forward the hypothesis that bilingual speakers use in an ideal fashion and in a clearly separate manner the varieties at their disposal. He (1953:73) even goes further and

52 denies that the “ideal bilingual” would even engage in code switching. He states that this phenomenon is a lack of competence in the languages involved in code switching as he puts it:

“The ideal bilingual switches from one language to the other according to Appropriate changes in the speech situation (interlocutors, topics, etc.), but not in an unchanged speech situation, and certainly not within a single sentence. If he does include expressions from another language, he may mark them off explicitly as’ quotations’ by quotation marks in writing and by special voice modification (slight pauses, changes in tempo, and the like) in speech” ( Weinreich 1953: 73)

Weinreich (ibid: 75) clearly defines switching languages as a social behaviour in which different varieties or languages are attributed distinct social functions. He (ibid: 75) goes on and defines the insertion of single elements from a host language into a guest language as an abnormal behaviour displayed by bilingual speakers because of their familiarity with the languages at their disposal as he states that:

“Deviations from the norm of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language” (Weinreich, 1953: 75)

The two studies of Barker (1947) and Weinreich (1953) launched the tradition of socially based research in the field of code switching as an important field in contact linguistics research alongside the psychological and the phonological ones. The social scholarly studies of code switching may be divided into two branches: micro sociolinguistically based research represented by Hymes, Bloom and Gumperz and macro sociolinguistic research represented by Fishman, Fergusson. We will start by discussing the macro sociolinguistic research for it is an important trend in bilingual and sociolinguistic research. Many sociolinguist such as Fergusson (1959; 1966; 1972), and Fishman (1967; 1971; 1980) devoted their research to the study of the composition of speech communities and to the language these groups actually use. They developed concepts such as Diglossia, Domains of language use, language maintenance and shift, and reversing language shift. For the needs of our study we will discuss some of these 53 concepts. We will start by the one of diglossia. This sociolinguistic concept has been defined differently by scholars such as Marçais (1930), Fergusson (1959), Fishman and others.

1.4.2.2 Diglossia

Diglossia may be defined as a language situation whereby different language varieties be it genetically related or unrelated are attributed different statutes, and functions. These statuses and functions are attributed by the speech communities because of official language policy or because of social agreement in the community. The first introduction of this term has been made by the French linguist William Marçais (1930) who studied Arabic. He defines la ‘diglossie’ (diglossia) as:

“The Arabic language appears under two perceptibly different aspects: 1) A literary language so called written Arabic, or regular or literal, or classical; the only one that has always and everywhere been written in the past; the only one in which today are written literary or scientific works, newspapers articles, judiciary, acts, private letters, in a word everything that is written, but which, exactly as it is has perhaps never been spoken anywhere and which in any case is not spoken now anywhere; 2) Spoken idioms, patois…none of which has ever been written …but which everywhere and perhaps for a many time are the only language of conversation in all popular and cultured circles” (Marçais, William.,1930:401)20

In his definition of diglossia Marçais made a clear cut distinction between what he calls a literary, purely written language and spoken language or patois. The functions of the two varieties are also clearly stated. The first variety is used for sciences, newspaper and literature writing, and also for judicial matters. The second one is used for daily conversation by all the speakers and is never written anywhere. This definition seems a bit outdated. It also seems not

20 The English translation of this quotation is taken from Bouamrane (1986:2). The original text reads as follows: “ La diglossie arabe se présente à nous sur deux aspects sensiblement différents: 1) une langue littéraire, dit arabe écrit ou régulier ou littéral, ou classique, qui a été partout et toujours écrite dans le passé dans laquelle seul aujourd‟ hui sont rédigés les ouvrages littéraires ou scientifiques, les articles de presse, lez actes judiciaires, les lettres privées, bref, tout ce qui est écrit, mais pas exactement telle qu‟elle se présente à nous n‟a peut –être jamais été parlée nulle part. 2) les idiomes parlés, des patois … dont aucun n‟a jamais été écrit mais qui, partout, et peut- être depuis longtemps, ( sont) la seule langue de la conversation dans les milieux populaires et cultivées” 54 to be valid for certain Arabic speech communities especially in the Maghreb. Dialectal Arabic is starting to be used as a written medium in some popular newspapers (sports newspapers such as Al Haddaf in Algeria21 - people’s magazines such as Panorama Arabiya, and Oyoun). It is also used in courts alongside with Standard Arabic as the variety in which the courts cases are conducted. Dialectal Arabic is also used as one of the favourite languages to write Short mobile messages and to chat on the internet.

Diglossia has been introduced in the English language and linguistics by Charles Fergusson. He uses this concept to describe the sociolinguistic situation of countries like Greece, Switzerland, Haiti, and countries of the Arab world. He defines diglossia as:

“A relatively stable language situation in which, in addition to the primary dialects of the language (which may include a standard or regional standards), there is a very divergent, highly codified (often grammatically more complex) superposed variety, the vehicle of a large and respected body of written literature, either of an earlier period or in another speech community, which is learned largely by formal education and is used for most written and formal spoken purposes but is not used by any sector of the community for ordinary conversation.” (Fergusson, 1959b: 336)22

The important concepts in Fergusson’s(1959b:336)1 definition are the ones of genetic relationship between the varieties being used in the speech communities in question, the complementary distribution of these varieties across different situations, the specialization of functions attributed by the community to the (H) and (L) varieties, and the stability of the situation in question. Fergusson(1959b:336) Talks about the high (H) variety which is used in official domains such as government, education, literature, and judiciary in opposition to the low (L) variety or varieties which is/are used in more private domains such as the family, friendship, and neighbourhood domains. To illustrate his concept, he mentioned the alternation of vernacular and standard Arabic in the Arab world, the alternation of Swiss, and

21 This example is taken from Benali Mohamed(1993: 19) 22 In D. Winford (2003 :112) 55 standard Deutch in German speaking Switzerland, Dhimhotiki and khatarevousa in Greece, and Haitian Creole, and Standard French in Haiti.

Some criticisms have been made to Fergusson’s definition of diglossia. The first one is the fact that diglossic situations are not always stable. These situations may change in the sense that we may have a turnover through time by which the H variety becomes L and the L variety becomes H. The L variety may also gain prestige and become H alongside the standard. We may give the example of Cairo Egyptian Arabic which has become the standard variety of Egypt. We may also give the example of Australian English, and Black vernacular English which were considered in the past as corrupted varieties of English and which have become standard varieties. Many other researchers (dejean1993:68; Schiffman 1993:186; Myers Scotton 2005:83,) point out to these H varieties turnovers in the four classical cases of diglossia described by Fergusson i.e. Switzerland, Greece, Haiti, and Arabic countries. The concept for example no longer exists in Greece where Demothik(the former L variety) is now in general use and a 1976 language reform established it as the language of publication administration, law, and education( Myers Scotton, 2007 :83)

Another criticism is related to the issue of middle varieties. The clear cut distinction that Fergusson makes between the different levels is not that evident. This issue has attracted many scholars from different perspectives (Blanc1960, Kaye 1972, Al Hassan 1977, Meisles 1980, Bouamrane1986, and Benali Mohamed 1993). They see these middle varieties as being rather part of a continuum than clear cut distinguished varieties.

Downes (1998:80) put forward another criticism. He puts into question the idea that H symbolizes the community’s historical identity and values. He( ibid:83) goes on stating that this is only true in cases in which the H variety enjoys prestige and positive attitudes among the speech community, but as long as these positive attitudes change, the status of H will be challenged by a local variety or by local varieties as Downes(ibid:84) puts it:

“Social changes could destabilise the system. These include Changes such as the competing prestige of a local nationalism, loss of belief in the values which the H symbolises, or loss of clearly defined domains which must be conducted in the H”

This is for example what is currently taking place with Caribbean and Asian Stylized English (Rampton 2005:64) in England. This variety is gaining prestige especially among the young generations (ibid 2005:67). It has even started to challenge the prestige and the functional allocation of British English in certain domains such as TV Broadcasting, literature and newspaper writing.

Fishman (1968:50) extended the definition of diglossia to cover not only genetically related varieties but also genetically unrelated ones. He then extended the occurrence of diglossic situations to not only monolingual situations but also to bilingual or multilingual ones as he states it:

“Diglossia exists not only in multilingual societies which recognize several languages and not only in societies that utilize vernacular and classical varieties, but also in societies which employ several dialects, registers, or functionally differentiated varieties of whatever kind”( Fishman, 1972:92)

Fishman’s definition of diglossia is partly based on a study of Rubin (1968) on the sociolinguistic situation of Paraguay. This country is characterized by a diglossic situation between Spanish the official language of the country, and Guarani (an Indian indigenous language which is genetically unrelated to Spanish). We can also make the same observation between Berber and Standard Arabic in Algeria with Berber being used as an L variety and standard Arabic being used as the H variety.

In an improvement of the definition of diglossia Fishman (1970: 35, 1980a: 04) recognizes four sociolinguistic situations in which diglossia and bilingualism are interwoven: 57

 H as a classical, L as vernacular where the two are genetically related, e.g. Classical and vernacular Arabic, Sanskrit and Hindi;

 H as classical, L as vernacular, where the two are genetically unrelated, e.g. textual Hebrew and Yiddish, Standard Arabic and Berber;

 H as written/ formal spoken and L as vernacular where the two are not genetically related to one another, e.g. Spanish and Guarani in Paraguay;

 H as written/ formal spoken and L as vernacular where the two are genetically related to one another, e.g. Urdu and Punjabi

There are other cases in which the speech community has got two H varieties and one or more L varieties such as the case of Algeria whereby Standard Arabic is the official language alongside French, and Algerian Arabic and Berber are the L varieties.

What is important in Fishman’s (1972:92) definition of diglossia is that it includes all societies in which “two or more varieties are used in given circumstances”23. The idea of circumstances is of a paramount importance in his definition of diglossia. In two influential papers Fishman (1964, 1965) introduced the concept of domain specialisation relevant to language use. He focuses on the relationship between code or language choice and kinds of social activity. In his approach, Fishman (1964:35) is first concerned with relatively stable socially agreed norms of choice and daily use of language. He clearly asserts that:

“There is an almost direct relationship between linguistic codes and social activities in the speech community. This would mean that appropriate language usage imposes on the speakers the obligation that only one of languages available or varieties will be chosen by

23 In R.A. Hudson (1998 :49) 58 particular types of interlocutors on particular situations to discuss particular kinds of topics” (Fishman, 1972, p. 437)

This would mean that, social information are not interpreted within code switching itself but are rather perceived in the association between language use, and the norms of language choice. This means that the norms of use give meaning to speakers’ individual choice of language. This is made manifest in Fishman’s (1965, 1972) key concept of ‘domain’ and ‘allocation paradigm’24

If we relate this to our discussion on code switching social interpretation, it would be worth mentioning Fishman’s (1971) typical example of English/Spanish CS between a boss and his secretary, both Puerto Ricans. The boss uses English as he dictates a letter to his secretary, but he then switches to Spanish for an informal conversation with her about the person who is supposed to receive the letter. We may Note that social meaning does not lie within the act of switching to Spanish, but in the correlation between the type of activity and language or code choice (e.g., Spanish for informal conversation vs. English for business). As it has been mentioned above the investigation on language choice and alternation made by Fergusson (1959), and Fishman (1965, 1972) were oriented towards the macro sociolinguistic interpretation of Code switching. This means that the social situation, which is defined prior to the conversation by social well established macro norms, dictates language choice and language alternation. We now move to the micro oriented sociolinguistic perspective in the study of code switching.

1.4.3 Micro Sociolinguistic Approaches to Code Switching

The micro study of Code switching within the frame of a single conversation was initiated by Blom (1957, 1961, and 1964) and later by Blom and Gumperz (1972). It has become since then one of the major topics in code switching and bilingual research. Blom and Gumperz put

24 Cf : C. Myers scotton(1993c : 49)

59 more emphasis on the individual rather than on society. They placed much of the responsibility within the individual in the choice between the languages. Their approach allows the individual speaker some kind of flexibility as they state that:

“The same individual need not be absolutely consistent in all his actions. He may wish to appear as a member of the local team on some occasions, while identifying with middle-class values on others” (Blom & Gumperz: 1972, 421)

Blom and Gumperz (1972) described several types of sociolinguistic bilingual situations including Code switching situations where there isn’t any change in the social situation. Gumperz (1957, 1958, 1961, and 1964) introduced the notion of interactional code switching25 in a study of this phenomenon in India. Code switching according to Gumperz (1964:58) is considered as a conversational phenomenon. He also emphasized on the need to study the social its functions when he studied Hindi-Punjabi code-switching in Delhi (India).Gumperz tried to relate code switching to social situation from an individual scope i.e. by concentrating on the communicative functions of the codes being involved in bilingual situations. He asserts that each code has a given communicative function within society. Codes are maintained separate by members of the community even if they are sometimes structurally similar. The reason for that is because they encode social functions. An illustration of this would be code- switching between varieties of the same language. Blom and Gumperz (1972) gave the example the two dialects of Norwegian in the village of Hemnesberget in Norway, i.e. Bokmal and Ranamal. They found that the speakers in this village consider that they were code switching while, in reality, they were moving in a continuum of intelligible dialects of the same language.

Blom and Gumperz (1972) made a study on the speech habits of the speakers of a fishing village called Hemnesberget in Norway. They found that these speakers code switch between Bokmal (the vernacular local dialect) and Ranamal (the standard) for different social purposes.

25 We may think that this concept was later redefined by Bloom and Gumperz (1982:59) as conversational code switching. 60

Blom and Gumperz (1972:78) related code switching to the social values attributed to these two varieties. They argued that:

“Code switching was ultimately explicable in terms of the values which Ranamal and Bokmal encoded” (Ibid: 79)

There where two sets of values encoded by the two varieties:

 Local values including Solidarity, and local Identity, Which were represented by the local variety (Ranamal)  Nation wide values, Which were associated with economic, political, and pan Norwegian cultural activities, and which were represented by the standard variety( Bokmal)

Following this social paradigm Blom and Gumperz could explicate individual switching. They further made a distinction between three types of code switching: Situational, metaphorical, and conversational code switching26.

Situational Code Switching: The varieties present in the speech community i.e. the standard and the local dialect are associated with social situations which are themselves conditioned by social norms. Blom and Gumperz (1972: 417) argue that:

“The linguistic separateness between dialect and standard is conditioned by social factors"

The premise behind this assumption is that the social allocations attributed to the two varieties trigger patterns of situational code switching. In such type of code switching, the

26 Besides situational and metaphorical Code switching, Conversational Code switching was added by Blom and Gumperz in 1982 61 speakers switches language to signal a change in one or more components of the social situation such as the interlocutor, the topic, the context, the setting (time, place) etc… Blom and Gumperz (1972:86 )give the example of a formal conversation between a local resident and an employee in a government office which is conducted in the Standard (Bokmal), but if the two engage in an informal chat they switch to Ranamal, the L variety. In general Bokmal is used in Church services, School matters, Interaction with strangers, and other formal contexts, While Ranamal is used by the members of the speech community at home, and in public places. Situational code switching has been equated by many scholars (Downes 1998:84, Winford 2003:116) to Diglossia in the sense that language choice is guided by social norms of appropriate language use in society. The only difference between the two is that language choice is individual in situational code switching while it is nation wide in Diglossia.

A good example of situational code switching in the Berber minority of Oran would be a Kabyle university English teacher using English in his lectures, French with his colleagues to talk about formal work matters, Algerian Arabic to have an informal Chat with them at work, a mixture of Kabyle, and Algerian Arabic to talk with a kabyle colleague at university, and kabyle to talk to members of his family at home. Here is an example taken from one of the tape recordings. The situation takes place in a teachers’ meeting room at the University of Oran.

A: Comment Vous avez fais pour corriger votre examen de magistère en Littérature Mabrouk?

(How did you proceed to correct your literature master’s exam Mokthar.)

B : je prends des critères bien définis comme la Grammaire, le style, l’inspiration entre autre

(Lit: I take well-defined criteria such as …)

C: je vous fais confiance pour cela corrig-/  /

(Lit: I have got confidence in you for that correct as usual)

B: Allo /    / Rayan /  -    /

(Lit: hello. What? Let me talk to him stay Rayan! Do not go! I will come)

( hello( answering on the phone) What(do you want) ? keep calm Rayan( talking to his son) Do not go( talking to his son again). If you still want to go, I will come (to punish you)

C: Mabrouk / /

( Lit: Mabrouk there is not something?)( Mabrouk is everything alright home)

A: / -  /

(Everything is fine)

Here is another example of situational code switching between Algerian Arabic and Mzabi in a Mzabi owned hardware shop of Oran.

1- A: // les projecteurs // cinq cents

(Lit. Do you have floodlight bulbs of five hundred?)

2- B: / / mille

(Lit. there are ones of one (1) thousand)

3- B: /m  Salah/

(Lit. How much do you count them Salah)

4- C: huits cents

(Lit. Eight hundred)

5- B: /d'origine i/

(Lit. Authentic are they?). 63

The code switch to Mzabi in line 3 signals a change in the interlocutor. Speaker A is an Arab, while speakers B and C are Mzabi. Speaker B switches to Mzabi as it is the language of solidarity among the Mzabi minority of Oran. He does so to talk about the price of the floodlight and more importantly to talk in line 5 about the quality of the product which is a matter he does not want speaker A (an Oranian) to know.

Metaphorical Code switching: In such cases of code switching is used to signal a change in topical emphasis (Blom and Gumperz 1972:409). This means that bilingual speakers for example use language varieties as a rhetorical device i.e. for irony, or to redefine the social relations within the conversation, or to show their attitudes towards a topic or towards an interlocutor. We may illustrate metaphorical code switching with this example of an informal situation in a newspapers shop in a Kabyle owned newspapers shop in Oran:

1 A: / - / Flexi

(Lit:Salam load me a hundred dinars Flexi(top up)

(Salam I would like a hundred dinars top up on my sim card)

2 B: /  -  -/

(Lit: How are you our dear fine about you)

(How are you dear, is every thing alright)

3 A: /    Houari / .    /

(Hamdullah oh! You are here Houari, how are you my friend, you look tired)

4 C: /  -  --  - - -  /-comptoir

(I am exhausted and I want to sleep I will ask him to arrange a bed(for me) behind the till)

5 B : /  -  -  - - /

(Lit: stand! Stand better than you sleep and this(is) not a place you sleep)

(stand up! Stand up this is not a place for sleeping)

The code switch to Kabyle made by the shopkeeper (a Kabyle) in line 5 has been made to show his mocking attitude towards speaker C, and to show his opinion about the idea of sleeping in a shop to a Kabyle friend of his. Speaker B was supposed to carry on the conversation in Algerian Arabic in the presence of Speaker A who is an Arab.

Compared to situational code switching which is guided by speakers external factors (social factors), conversational Code Switching is motivated by speakers’ conversational or interactional purposes (Downes 1998:80). Blom and Gumperz (1982:59) state that conversational code switching is:

“The juxtaposition within the same speech exchange of passages of speech belonging to two different grammatical systems or subsystems”

They go (ibid: 60) on stating that:

“Conversational code switching can be between participants turn so that one speaker speaks in one code, and the response comes in another, or alternatively, any one speaker can code switch between sentences, intersententially, or within a sentence intrasententially”(Blom and Gumperz, 1982, 59-60)

Here is an example of a conversational switch taken from the data:

/m  -/ les étagères (Mzabi-French conversational code switching)

(Lit. Bring a container and arrange on the shelves)

This process is also known by some specialists as code mixing (Bouamrane, 1986:6, Hudson, 1996:53; Muysken, 1995: 183, 2000:48).

1.4.3.1 The Use of Code Switching as a Conversational Strategy

As it has been said above, conversational code switching may be regarded as a conversational strategy. Gumperz (1982:412) proposes six major functions for this process. These are mainly:

 Quotation i.e. switching language to signal a quotation or to make a reported speech. Gal (1979:109) argues that:

“all one needs to know to predict the language in which most quotes will be spoken is the language in which the original utterance was spoken”

E.g.:

1. Faiza: Sarah (. 1.5) / -- “- -  ”/

(Lit: Sarah tell to- aunt-your “Come inacomp- stay with us)

( Sarah tell your aunt “come to stay with us”)

(.30 silence)

2. Sarah:     “- - - .”

(Lit: Mum! Mum! She did not want. She told you “tired- 1st sing inacomp- go- 1st sing inacomp- sleep-1st sing)

(Mum! Mum! She did not want (to join you) she told you” I am tired I am going to sleep”)

L’histoire Geo  sujet    “deux sujets au choix”      sujet

(Lit: The history and geography; they gave them one subject only. At the beginning, they told them “two subjects to choice” in addition they gave them only one of subject)

(In the history and geography subject, they gave them (the pupils) one subject only. At the beginning, they told them (that) they would have a choice between two subjects; however, they gave them one subject only

In the first interaction, the code switch operated by Sarah in line two is a code switch to signal a quotation or a direct speech. Sarah seems to be afraid to upset her mother, and she uses the exact terms used by her aunt when she (Sarah) asked her to join them. In fact Sarah was using Algerian Arabic in her previous turns in the conversation. We did not want to transcribe the whole interaction because it is a bit long. We preferred to concentrate on the part in the interaction where there was a switch for quotation.

In the second example, the French code switching “deux sujets au choix” is used to signal a quotation and to report what has been said by the administration to the students. This is shown by the coming back to Berber right after the quotation.

 Addressee specification, addressee change i.e. switching language either to signal a change in the interlocutor, or to specify the interlocutor one is addressing As Auer(1995) puts it:

“The switch serves to direct the message to one particular person among several addressees present in the immediate environment or to signal a change in the interlocutor”

 Interjection: Switching language to use an interjection in one’s mother tongue, or to use discourse markers, or to use sentence fillers. This simply serves to mark sentence fillers as in the insertion of the English filler you know in an otherwise completely Spanish utterance27. We may illustrate this type of switching with the expression //(By the way) which functions a sentence filler is Algerian Arabic, but which is often used in otherwise Kabyle or Mzabi utterances. Below is an example

/ - -- - - -   --/

(Lit:by the way and you and brother-your and you the one like this that you said the one who learned)

(By the way what about your brother(the one who learned

27 This example is commonly cited by specialists of Spanish English code switching such as torribio( 2000, 12), Mc Swan(1997 :11), 68

 Reiteration: Bilingual speakers use code may use code switching as a strategy to reiterate their message. They may either resort to a quasi translation of the message in the other language for purposes of emphasis, clarification, attracting attention28. E.g. /.  -- ---  --  --/

(Its over I told you I say I repeat to you I say I repeat You drive me crazy)(Berber-Algerian Arabic code switching)

The Speaker repeats the same message /-- ---/(lit: I say, I repeat) and then / --/(Lit: I say, I repeat)in Berber and Algerian Arabic to show to her interlocutor that she is fed up and to emphasize her message.

 Message qualification: Switching code to qualify one’s message. Bilingual speakers may code switching to operate a change in the communicative function that they are using e.g. to operate a change from an informative function to a directive function (a command for example).

 Personification versus objectification: Bilingual speakers may code switching to signal a change in the degree of their involvement in a message as in the case of, for example, giving one’s statement more authority in a dispute through CS. We may illustrate this function using this example in which the speaker switches code i.e. from Algerian Arabic to Kabyle Berber to show that he takes his message for very serious.

E.g. / -   -.  /

(come today or I get upset with you be careful look!)

(Lit: Come today or I will be upset becareul watch it!)

28 Cf. J.Mac Swan (1997 :51) 69

These functions have been extended later on by other researchers such as Hill and Hill (1986:35) who added the functions of reducing and enforcing the illocutionary29 effects of the message, and Auer (1995:127) who added these functions:

 side comments( switching language to signal a personal comment on a statement)

 Topic change (switching language to signal a change in the topic of conversation)

All these functions signalled by code switching in bilingual interactions may be compared to monolingual interactions functions. These functions are are fulfilled in monolingual language by the Prosodic features such as: The paralinguistic features including face expressions, body movements, the lexical features such as synonymy and paraphrasing and the syntactic features such as the use of reported speech and the use of passive voice.

The findings of Blom and Gumperz (1982) in the analysis of code switching at the micro level of conversation have been taken up by two groups of researchers with somehow different perspectives and approaches. The Approach of the first group represented by Auer (1985, 1995) is based on the conversational analysis of code switchings by relating them first to speakers interactional intentions, and then to conversational norms, and then to sociocultural norms that are shared by speakers(Auer 1995:324). This approach is definitely a conversational one. Winford (2005: 117) describes Auer’s approach stating that:

“The meaning of a code switch depends both on its sequential position in the discourse context itself, and on the broader situational and sociocultural context which make up the background knowledge of the participants….the approach attempts to link the micro level of conversational interaction to the macro- level societal setting in which it occurs”

29 Cf. Mac Swan (1997 :53) 70

Auer (1995) is not the only scholar who advocated a conversational analytical approach to code switching. Myers Scotton applied an approach somehow similar. She considers that speakers do not code switch to display social identities, or to obey sociocultural norms, but do so to convey intentional meaning. This meaning is of a socio pragmatic nature (Myers Scotton1993a:89). This approach emphasizes the role speaker’s creativity and freedom to comply with the norms (sociocultural and conversational norms) or to violate them. Myers Scotton (1993a: 90) argues in this vein that:

“The Speaker may wish to negotiate and to change the nature of the interactional situation and the social relationships among participants”

Following these principles, Myers scotton (ibid) developed a whole model in the social interpretation of code switching. She calls this model The Markedness Model. It is based on speakers’ socio-psychological motivations for code switching.

1.4.3.2 Myers Scotton (1993) Markedness model

This model is partly based on Grice’s (1975) co-operative principle30. It may be necessary to clarify the different premises of this principle as it is paramount to our understanding of the Markedness model. The co-operative principle is one of the principles of the Communication Accommodation Model(Giles 1982:213) which states that language is not related only to the membership of the speakers who use it, but also to what they wish to accomplish with it( Myers Scotton 2006:155). The model also tries to explain why people modify their linguistic behaviour in different interactional situations (blanc2000:260). The modifications are used by the interlocutors as strategies to obtain two opposite goals: High degree of communication or

30 This is one of the basic principles of interaction, whereby speakers need to accept to have conversations before this latter starts. This principle is also part of a broader theory of Speech accommodation (Giles 1982:38) in which all types of interaction are either linguistically convergent (there is an agreement between speakers) or linguistically divergent( there is divergence between speakers). 71 low degree of communication i.e. they either converge to each other using language, or on the contrary diverge from each other. These strategies are themselves governed by four principles:

 The linguistic competence principle: the language of the interaction as a whole is the one in which the interlocutors share the highest degree of competence i.e. the one in which the speakers share the highest degree of mutual intelligibility. We may take the example of two Berber speakers, who are bilinguals in Arabic, but who share the highest linguistic competence in Berber, consequently; these will use Berber instead of Algerian Arabic in their conversation to attain the highest degree of communication. This may be described as a convergent language choice.

 Ethnolinguistic Affirmation principle: The speaker weighs the costs and disadvantages of using a given language in a conversation. He may, for example not use a language that represents threats to his ethnic identity, even if this language is the appropriate one following the linguistic competence principle. This is characteristic of divergent language choices.

 Interlocutors perceived intention principle: If the speaker perceives signals of a hostile intention or attitudes on the part of his interlocutor, he may decide to display a divergent language choice using a communicatively inappropriate code.

 Personal, situational, and social factors principle: The prominent factors influencing divergent and convergent language choices between interlocutors ,and which have also been termed contextualisation cues by Gumperz(1982:132) are:

1. Interlocutors Roles within the conversation i.e. is any one of them in control of the conversation 2. The topic of conversation. 3. The setting (time and place). 4. The degree of formality of the conversation.

5. The social norms guiding the conversation. 6. The status of the languages involved (in the broader society). 7. The ethnic Origins of the interlocutors in addition to, their age, sex, instruction level, and socio-economic level.

These personal, situational, and social factors are highly important to any interaction, accordingly; bilingual communicative strategies vary in relation to any change in one of them. This would mean that language choice is partly dependent on these contextualisation cues. They are used by speakers as cues to select the appropriate language to be used in the interaction, but language choice is not as easy as that. This is particularly true in situations in which no prior information is available to the interlocutors. In such case, we would have instances of language negotiation (Auer, 1998:09). Not only bilingual, but also monolingual interactional situations are subject to these instances of language negotiation.

Following the same line of thought, Myers Scotton (1993a:113) asserts that any piece of conversation is based first of all on some kind of negotiation between participants. This negotiation principle has been defined by Myers Scotton (1993a: 113) as such:

“ Choose the form of your conversation contribution such that it indexes the set of rights and obligations which you wish to be in force between speaker, and

Addressee for the current exchange”

The Markedness model as it has been expounded by Myers Scotton (1993a) seeks to relate the Communication Accommodation Model to her bilingual data. Myers Scotton’s model is based on the principle that there are norms of communication i.e. what she calls ‘Rights and Obligations sets’ (Myers Scotton 1993a:45). She also asserts (Ibid: 47) that types of interactions are highly conventionalized. The social conventions in question are drawn from the contextualisation cues listed above. Interactional situations are partly predetermined by these cues.

In bilingual situations, some languages tend to prevail more than others in certain conversational situations. As a result each language comes to be associated with Rights and obligations sets of a given interactional type. Each variety acquires a socio psychological value in the linguistic market of the speech community as the result of their daily encounters. This principle is similar to the one that has been put forward by Blom and Gumperz (19725:59) allusion to the idea that varieties encode social meanings. The only difference between the two principles is that of a perspective. The one of Myers Scotton (1993a) is micro social in the sense that it is directed to the analysis of conversation, whereas Blom and Gumperz’ perspective is macro social as it considers the varieties as encoding macro community wide values. Myers Scotton (1998:225) states in this light that:

“The knowledge of the unmarked RO (rights and obligations sets) and the knowledge of socio-psychological values of the varieties present in the speech communities are exploited by the bilingual speakers to fulfil specific communication goals”

The speaker has got the freedom to either comply with the Rights and obligation sets (Communication norms) or to disagree with these norms. In the first situation, the speaker would be using the appropriate language or code i.e. the unmarked language, whereas he would be making a marked language choice if he refuses to comply with the norms. Unmarked language choices have been defined by Myers Scotton (2007:159) as:

“ Unmarked choices are those that are more or less expected given the ingredients of the interaction( participants, topic, setting, etc)…Myers Scotton refers to a rights and obligation set as part of the normative expectations. These expectations refer to an unmarked way to behave…When makes the unmarked choice, he or she is causing no social ripples because participants expect such a choice” (Myers Scotton, 2007: 159)

To relate Language unmarked choices in the Markedness Model to The Communication Accommodation Theory (hereafter CAT), we may say that the speakers would be making a convergent language choice when they are showing an unmarked language choice. In bilingual situations, speakers switch language whenever there is a new right and Obligation set (hereafter RO set). This means that the speaker changes languages or codes according to the new RO set. This is what Myers Scotton (1993a: 50) calls sequential code switching. This type of code switching has been compared by many scholars (Downes 1998:92; Winford2003:118) to situational code switching. We may illustrate this with the case of a Berber speaker in Oran who uses Arabic with an Arab, but when this speaker is having an interaction with a Berber he switches to Berber.

Another type of unmarked code switching is the case of bilingual communities where Code switching is not sequential but is itself the Unmarked or appropriate choice. Speakers in these communities use mixed codes to signal multiple group membership as (Myers Scotton 2007:268) puts it:

“Such code switching conveys the message of dual identities or memberships in both of the cultures that the languages index”

This kind of unmarked code switching may be equated with Blom and Gumperz’ (1982:78) conversational code switching. Myers Scotton (2007; 167) gives the example of Nairobi teenagers from different ethnic groups who use as an in-group language a mixture of Swahili and English (Swahili being one of the standards in addition to being one of the linguafrancas in Kenya) which shows features of their ethnic varieties. Myers Scotton (ibid: 162) argues that the young men in Nairobi are not satisfied with either the identity associated with speaking English alone or the one the one associated with speaking Swahili alone when they talk to each other. They rather see the rewards in indexing both identities (Myers Scotton, 1993b:122 (in Meewis and blommaert, 1998:78)). They solve this problem by using code switching between the three languages. This is not a case in point in Kenya, but is a common feature of many bilingual situations involving minority groups living in otherwise different majority communities (Texmex in the Puerto Rican and Mexican minorities in The USA, Arabic 75

French mixed codes used by the Maghrebi Minority in France). This kind of unmarked code switching is also a recurrent feature of bilingual communities with a stable and long lasting bilingualism such as Algeria with its French- Algerian Arabic Algerian mixed codes. To come back to our Berber Minority speakers we may give the example of highly educated kabyles speakers in Oran who use a mixture of Kabyle and French as their in group language

Now let us turn to marked language choices, which are more important to the Markedness Model than are unmarked choices as those former highlight the negotiation process. Myers Scotton (2007:160) states that:

“One of the main features of the Markedness Model is not what it has to say about unmarked choices, but what it says about marked choices. Marked choices are those that are not predicted … The model keystone is a negotiation principle”

This means that choosing a marked language choice signals a speaker’s negotiation for a right and obligation set different from the existing one at the time of the interaction (Myers Scotton 2007:160). This means that the speaker switches from the unmarked (appropriate code) to a code which is not relevant for the situation. If we relate this to the CAT this language choice would be a divergent choice.

Myers Scotton (1993a:132) argues that marked choices represent a departure from the normal state of affairs. It is used by speakers to signal a change in the social and psychological distance towards their interlocutors. In bilingual situations making a marked language choice may be signalled using what Myers Scotton (Ibid: 133) calls marked code switching. This type of code switching is used as a strategy to signal anger, Authority, to exclude outsiders from in- group interaction, to emphasize messages via repetition, and to show more familiarity and solidarity with an interlocutor in a situation which needs a different language choice( e.g.: using a minority language in a situation which calls for a majority one).

Myers Scotton (2007:160) gives the example of a young man who meets his girlfriend’s father for the first time. In most societies the unmarked choice would be relatively formal such as “It’s a pleasure meeting you sir”. What if instead the young man’s greeting is just a “Hi there” or some equivalent and no more? This is a marked choice. By saying this, the young man is not complying with the RO set (conversational, social, and situational norms) of the situation. One interpretation of his choice may be that he may signal he does not care for his girlfriend’s father as a person of higher status.

She (ibid: 160) gives another example from a bilingual situation in Nairobi (the national capital of Kenya). The situation is the one of a person coming from a rural to visit his brother, who has an important position at a national company. The unmarked choice would be for the brothers to greet each other in their shared ethnic group language. Instead the city brother greets the rural brother in English (one of the official languages in Kenya). By doing so the city brother is making a marked language choice. He is kind of misconsidering their shared Ethnic identity to signal that rural ethnicity does not hold in such a situation. The city brother is making a marked code switching. If we relate marked code switching to CAT terminology, we would say that it is a clear case of a divergent language choice. Let us consider few illustrative examples from the data

1)A: serrure étoile petit modèle // Salah.

(Lit. star door lock small model there is)

(2)B: // (yes)

(3)A: / / (Lit. how much do you count it?)

(4)B: / / (Lit. I count it fifty thousands)

(5)A: /  / (Lit. lower me the price)

(6) B: / / (no the price is fixed)

(7)A: /     / (Lit. We are brothers make me a lower price)

(8)B: /   / (it is the whole sale price dear Brother)

(9)A: /  / (give me five locks)

This conversation took place between two Mzabi speakers in a hardware shop. The interaction is carried first in Mzabi, which is the in group language of this minority in Oran; there; however, is a switch to Arabic made by the shop owner in line 6. This code switch is a marked one for Algerian Arabic is not the appropriate code in such a situation. The shop keeper switches to Algerian to redefine the RO set of the situation (by redefining his relationship with his interlocutor). He signals to the customer that no bargaining is possible even if they both belong to the same minority group i.e. the Mzabi one. He also signals to him that he and no one else is in control of the rules of conduct in his shop.

The premises of the Markedness model finally assert that we may have instances of exploratory code switching whereby the speaker is not sure about the language which best fits the RO set and the context of the interaction. This is another situation in which a negotiation process takes place (Myers Scotton, 1993c: 142, 1998:267, 2007:167). In a later development of the Markedness model, Myers Scotton (2004: 1045, 2007:161) and Myers Scotton and Bolonyai (2001:42) introduced the idea of rationale choices in code switching. She (2007:161) states that

“An important premise of the Markedness model is that speakers not only make choices of which choices are more unmarked or more marked, but also make choices with a sense about choices will bring them the best outcome…in regard to language choice they weigh the relative costs and rewards of speaking one language rather than another…This brings the Markedness model into the tradition of Rational Choice models.”

Rationale is not employed here to mean objective Choices, but is rather related to the speaker’s subjective perceptions of the possible outcomes of his choices in addition to his social and personal values and beliefs (Myers Scotton, 2007: 162). We will come back to this

78 important issue when we will deal with the conversational analysis of Code switching in chapter four.

The Markedness model has been strongly criticized by many conversational analysts (Auer, 1998:46, Gafaranga, 2004:654, Jorgensen, 2004:670, Gardner Chloros, 2005:48). They argue that that Myers Scotton (1993a: 135) interprets conversational events using factors which are external to the inner dynamics of conversation. They go on stating that Myers Scotton only considers interactions in terms of RO sets which are a social normative device. Conversational analysts clearly put in question the idea that Markedness can account for all cases of code switching. Auer (1998:3) talks about the need to avoid using pre established external categories in the social interpretation of code switching. He goes on stating that the conversation is to some extent sufficiently autonomous from larger societal factors. In the same vein, Auer (1995:123) states that:

“The situated meaning of code alternation cannot be stated unless a sequential analysis is carried out. The same cue may receive a different interpretation on different occasions”

Myers Scotton (2007: 172) herself acknowledges that the difference between the two approaches is rather a matter of degree in the emphasis than a matter of absolute opposition as she asserts that:

“The difference between a CA Analysis of bilingual conversations and that of other approaches such as the Markedness model seems to be a matter of degree, not absolutes” Myers Scotton(2007:172)

The Markedness model does not have drawbacks only, but should be used alongside a conversational analysis approach to interpret bilingual conversations. Joseph Gafaranga

(2000:294)31, a leading figure in the Conversation analysis approach, and an opponent to the Markedness model, states that:

“It is important to recognize this contribution by Myers Scotton to the rehabilitation of bilingual language alternation among bilingual speakers”

We now turn to the conversational approach to code switching as it has been expounded by conversational analysts such as Auer(1984, 1988, 1995, 1998, 2000), Gafaranga(1988,2000), and Jorgensen(1998).

1.4.3.3 The conversation Analysis Approach to code switching

Peter Auer (1984) established a model in the analysis of code switching (what he (ibid: 45) calls code alternation. His model is based on the application of conversational analysis methods and procedures in the analysis and the interpretation of bilingual speech within the frame of conversation as he states that:

“A promising alternative approach to code alternation might therefore consist of analysing the signalling value of the juxtaposition of languages and deriving the conversational meaning of code alternation from it” (Peter Auer 1995:119)32

Bilingual and monolingual conversations are, according to Auer (1998:1), considered as sequential conversational events. Another important principle behind Auer’s approach is that these conversational events involving different varieties of the same language or different languages may be interpreted from a conversation internal perspective i.e. without resorting to social and cultural community wide factors, which is considered as the norm in most studies

31 In P. Auer and L.Wei ed ( 2000 :294) 32 In L. Milroy and P. Muysken ed( 1995 :118) 80 on Bilingualism33. To depart himself from the conventional procedure in sociolinguistic research, Auer (1995: 119)34 clearly asserts that his approach is:

“Quite different from Fishman’s association since, no association between languages and speech activities needs to be presupposed. Obviously, it requires a sequential account of language choice, in which the language chosen for one speech activity must be seen against the background of language choice in the preceding utterance”

Auer (1998:17) further argues that Social meaning is not inexistent in bilingual conversation, but is rather locally produced at the micro conversational level. The analysis of bilingual exchange, according to him (ibid: 19), needs to integrate both sociocultural parameters and conversational ones. This would mean that background knowledge of the conversation is to be derived from the social and cultural context in which the interaction takes place; however, it has to be linked to information derivable from the internal development and dynamic of conversation. He clearly puts a lot of emphasis on the sequential (conversational analysis of code alternation) as he puts it:

“Any theory of conversational code alternation is bound to fail if it does not take into account that the meaning of code alternation depends in many ways on its ‘sequential environment’. This is given, in the first place by the conversational turn immediately preceding it, to which code alternation may respond in various ways”( Auer 1995:116)

Auer (1998:14) gives four types of code switching which integrate extra-conversational knowledge with an internal analysis of the conversation. He states that these are organized according to two axes. The first axis is the use of code switching to construct conversational

33 This is related to Fishman‟s languages domains allocation whereby specific languages or language varieties are associated with specific social and speech activities 34 In L. Milroy and P. Muysken ed( 1995 :118) 81 meaning, whereas the second one is used to construct interpersonal meaning as he puts it in a previous paper (1984:12)35:

“Faced with language alternation, participants ask themselves: Is the language alternation in question providing cues for the organisation of the ongoing interaction or about attributes of the participants?”

1.4.3.3.1 Discourse related switching

This type of code alternation may be defined as the use of code switching as a strategy to organize the conversation. This means that code switching is used by bilinguals as an additional device to the other monolingual conversational strategies such as pauses, hesitations, voice reduction, and stress change to construct interactional meaning. The conversational functions of code switching that have been described by Auer (1995:124) include: marking quotation (code switching to signal that a quotation has started), reiteration for emphasis (the repetition of the same word or the same group of words in a different language to emphasize one’s message) and others which will be dealt with in more details in chapter four which is devoted to the conversational analysis of the corpus., Auer (1995:123)36, in a discussion about the conversational function of code switching states that:

“Code switching is one of an array of devices such as intonation, Rhythm, gesture or postures which are used in the situated production and interpretation of language…Code switching works in many ways just like these other cues, a fact that calls for uniform analysis”

Here is an example of a discourse related code switching taken from the data

35 In J. Gafaranga(2000:300) 36 In L.Milroy and P.Muysken( 1995:123) 82

E.g.: /      / (Lit. the Muslim the Muslim we implore God to be Muslims)

This example highlights the use of code switching as a conversational device. This utterance in Mzabi contains a Standard Arabic code switch/ /( the Muslim). The alternation from Standard Arabic to Mzabi may be interpreted as an emphatic device in the sense that the speaker( a Mzabi starts his utterance in Standard Arabic and then moves to Mzabi repeating the same word to emphasize his message.

1.4.3.3.2 Extra conversational knowledge

Auer (1998:14) argues that the analysis of bilingual conversation may as well be derived from wider context factors. This is particularly true in situations where the direction of switching is constant i.e. when it goes from a variety to another without any change in the direction of switching. We may talk, in such a situation, of code switching as indexical of the larger socio-cultural context. In a study conducted on young Italian emigrants in Germany, Auer and Jorgensen (1998:86) found that these speakers use German in their interaction while their parents use Italian as an in-group language.Jorgensen (1998: 89)37 talks about “in group”, and “out-group” language in such a case.

The in-group language, according to Jorgensen (ibid: 89), is the language of the conversation among peers. This language may function as the dominant language among these peers even if their parents do not have it as their in-group code in the sense that they are members of an Italian minority which uses Italian as the in-group minority language. This may be considered as an intergenerational language shift whereby the in- group language of a first generation emigrants cease to be so in the other generations. The same has been noticed in

37 In. P. Auer (1998 :89) 83 the community we are working on, but this is just an observation which needs to be studied in more details. Here is an example taken from our corpus to illustrate this situation:

(1) X: /.  / Samira (Lit. will you come with me to Samira)

(2) Y: (5 seconds silence)

(3) X: /    / (talk talk don't you come with me to (see) her?)

(4) Y: / / T.D (no, I have a seminar)

This is a conversation between two young female speakers. The code switch in line (3) / / (talk talk) is used as a discursive device by speaker X to emphasise her message. We may also suggest an extra conversational interpretation. Berber may be regarded as the out group language and Algerian Arabic as the in group language of these peers. Speaker X uses Berber in line (1) but receives no answer. The non response of speaker Y in line (2) may signal a wrong language choice. The switch in line (3) to Algerian Arabic may be used as a repair initiating device to repair the wrong language choice in line (1). The switch to Algerian Arabic here has a conversational meaning (emphasising one's message) still it gives some pieces of information about in group and out group language among these speakers.

1.4.3.3.3 Preference related Switching

This type of code alternation is observed when there is a divergence between bilingual speakers about which language to use in the conversation. This is a common feature in bilingual situations. Preference related switching is different from discourse related switching in the sense that in the former there are divergent language choices whereas in the latter it is a matter of convergent language choice. On the one hand, discourse related switching takes place when speakers share and follow the same norms about the language to be used; even if, there are instances of divergent language choice from time to time, the speakers try to repair them by converging to their interlocutors’ language. On the other hand, preference related switching is characterized by instances of persistent divergent language choice; accordingly, the speakers do not try to repair these divergent choices for different reasons. Here is an example in point from our recordings:

X: vous avez des lampes? (Do you have bulbs?)

Y: /./ à vis // à baïonnette. (You need screw bulbs or bayonet fitted bulbs)

X: je voudrais des lampes à vis? (I would like screw bulbs)

Y: /    / (of one hundred (watts) or sixty (watts)?)

X: les plus fortes. (The most powerful)

Y: /  / (take the ones of one hundred)

X: il comprend ce que je lui dis mais il me répond en arabe. (He understands what I tell him but answers in Arabic)

Y: je comprends mais je parle comme je veux /  ./ (I understand but I talk the way that pleases me, is that all you need?)

This conversation, which took place in a popular market in Oran, calls for Algerian Arabic to be used as the appropriate code. Speaker X uses French which is a wrong language choice.

Speaker Y uses Algerian Arabic which is the language to be used in such a context; whereas, speaker X persists in using French making a divergent language choice. Speaker Y code switches to French in line 8 but soon comes back to Algerian Arabic even if he knows that it is not the language by means of which he is addressed. This signals a preference related code switching used by speakers X and Y.

1.4.3.3.4 Participants’ Related Switching

It has been defined by Auer (1995:116) and by Gafaranga (2000:302) as the “use of code switching by participants to show their attitudes towards their interlocutors”. Gafaranga (ibid:117) argues that this notion is quite interesting in the sense that it paves the way for language negotiation sequences in the sense that the speakers negotiate about the language they will be using in the ongoing conversation Gafaranga(2000:302). This type of code switching may be equated with Myers Scotton’s (1993b) exploratory code switching.

Our discussions of the conversational analysis of the Markedness models have been a bit lengthy for two reasons. The first one being that our study builds heavily on the findings of these two approaches to interpret the social and conversational motivations of code switching in our data. The second one is that these models are the ones which are the most hotly debated, and the most widely used by the community of researchers working on bilingual data.

Other micro conversational approaches include the social interpretive approach as it has been put forward by Jacobson (1982; 1990), and the social network approach as it has been expounded by Milroy (1987:64) and by Milroy and Wei (1995:136).

1.4.3.4 The Social Interpretative Approach to Code Switching

Jacobson (1982: 190) studied code switching in the Spanish community of the United States. He tried to interpret code switching in this community using Gumperz’ (1982: 45) micro sociolinguistic interactional paradigm i.e. the interpretation of code switching according to contextual cues such as change in interlocutor, topic, setting, domain. Jacobson (1982: 192) further elaborated the interpretation of code switching by relating it to an additional feature i.e. the one of emotion. He (ibid: 195) gives instances of code switching in which the speakers switch to their dominant language (Spanish) to signal a strong, or a surprising emotion. Below is one of the examples given by Jacobson in his study of Spanish-English code switching in the Spanish speaking minority of the United States. Spanish in Italics (emphasis is Jacobson’s):

I lose my temper porque a mi me da mucho coraje

(I lose my temper because it makes me so mad) (Spanish- English code switching) (1982: 195)

The switch to Spanish (the speaker’s dominant language) in this example has been interpreted by Jacobson (1982: 195) as a sign of a big emotional load that the speaker wants to convey in her dominant language to be sure that her interlocutor is going to understand what she(the speaker) really feels. Jacobson (ibid: 196) clearly asserts that emotion may be regarded as a valid contextualisation cue which may trigger off switching from the weaker language to the dominant one38. Here is an example in point from the corpus of this study:

1: Mother: /  .  / Leila

(I saw this photo before! Leila!)(Addressing her daughter)

2: Baby: /  / (Mum, Mum) (The baby crying repeatedly)

3: Mother: /  /

(Please leave me alone! You!) (Addressing the baby)

38 Dominant language is not used here to mean the language of the majority speakers, but it rather means the language in which the minority speakers are most competent( Spanish in Jacobson study); similarly, the weaker language is meant to be the language of the host community, and the one in which the minority speakers are less competent( English in Jacobson study) 87

4: Mother: / ! -- /

(For god sake! Take it( the baby) away from me there) (addressing Leila)

The mother code switches from Algerian Arabic to Kabyle in line 3. This code switch may be interpreted as being triggered by the fact that the baby is crying repeatedly. This seems to strongly disturb the mother who switches to Kabyle to signal to her baby that this is really making her nervous, and she also does the same thing with her daughter in line 4 when she asks her to take the baby out of her sight.

Jacobson (1982: 196) further argues that some occasional switches are only due to a desire to use the other language instead of being always related to some definite contextual cues or to other sociolinguistic or even to psychological motives. Below is an instance of code switching from the data to illustrate this point.

1: Manal: /  / la crèche

(Lit: Me I went to the nursery school)

2: Mother: /    ?/

(Lit: the nursery teacher what she did tell you?)

3: Manal: //? //?    /

(Lit: (what? What? Teacher told me you (are) witty). (Kabyle- Algerian Arabic code switching)

In this conversation between a mother and her daughter the switch operated by the daughter in line 3 does not have any emotional value, nor does it have any social meaning. It is also not due to a lack of lexical availability in the child repertoire in the sense that the child uses the Kabyle word // (nursery school teacher) few lines later in the conversation. We may interpret this code switch along the lines of Jacobson (1982: 195) as a desire to use the other language .i.e. Algerian Arabic in this case.

Many scholars(Mac Swan, 2000: 21; Myers Scotton, 2002: 135) criticized Jacobson for over generalizing the use of this category as a ‘catch all category’( James Smith, 2002: 44) in which all instances of code switching which are impossible to interpret may be put. The assumption put forward by Jacobson may be regarded as a good resting place where code switching researchers may rest when they fail to interpret code switching instances, but the danger lays in the fact of resorting to this category too often to the expense of loosing one’s explanatory validity.

In a further investigation of Spanish-English bilingualism in the United States, Jacobson (1990: 124) argues (ibid: 125) that the socio-economic position of speakers is not enough when one wants to interpret code switching patterns; accordingly, he (ibid: 127) asserts that age and gender, in addition to socio-economic standing, have a clear correlation with the different types of code switching displayed by Spanish speakers in the United States. Jacobson (ibid: 127) still concludes that a reliance on these three variables only is not sufficient in the sense that it does not give the researcher an exhaustive interpretation of code switching.

On the other extreme side of the code switching research community, Myers scotton (1993b: 65) proposes that the social interpretation of most instances of code switching is possible under her social Markedness model. The Markedness model tries to see code switching as being constrained by socio-psychological constraints which are based on the Marked versus Unmarked language distinction as it has been discussed above; however, Myers Scotton fails to correlate the different types of code switching with social features. Her Markedness model has been strongly criticized by many scholars (Gardner Chloros and Edwards, 2002: 730; James Smith, 2002: 27; Wei, 2005: 1050). The criticisms relate to the fact 89 that the Markedness Model does separate the social interpretation of code switching (Myers Scotton, 1993b: 27) from the linguistic interpretation of the different patterns of code switching. The interpretation of these patterns is part of her Matrix language frame model39 (Myers Scotton, 1993a: 12).

1.4.3.5 Integrating the Social and Linguistic Approaches for a Better Interpretation of Code Switching

James Smith (2002: 28)40 tried to investigate the relationship between Myers Scotton’s (1993b) Markedness model, which deals with the micro-social characteristics of code switching, and her (1993c) Matrix language Frame model which deals with the structural linguistic aspect of code switching. The premise behind his approach is that there should be a relationship between the social attributes of code switching and the linguistic ones. James Smith (2002: 28) believes that there is a gap between the investigations related to the structural aspect of code switching and the ones related to the social aspect of this phenomenon. Myers scotton put forward two models for the understanding of code switching, but as James Smith rightly points, they are not related. In this vein, James Smith (ibid: 29) States that:

“Myers-Scotton’s MLF model for the “structural consequences” of bilingual speech production includes at its base a “conceptual level” in which are determined the “speaker’s intentions regarding referential information and pragmatic and socio-pragmatic messages” (1995: 236-237). She also says that EL insertions (even single morpheme insertions) into an ML structure “must match ‘in terms of their discourse or pragmatic functions’” (1993b: 121; 1997b:

39 The Matrix Language Frame model as it has been expounded by Myers Scotton in 1993 is a morphosyntactic model of code switching which has as its most important principle that the languages contributing to code switching do not contribute on an even scale. The other principle underlying this model is that there are two different types of morphemes in code switching data (system morphemes and content morphemes). We shall come back to this model of when we will be dealing with the structural analysis of our data in chapter two. 40 James Smith carried a study in 2002 on the Spanish community of Georgetown in the United States where he related patterns of code switching in this community to social variables. The approach that James Smith takes is radically different from the one advocated by Myers Scotton when she devised her two models. 90

6).Yet she treats her MLF model separately from her Markedness model, so much so that a separate volume of her work on CS is devoted to each model (1993b for the MLF model and 1993c for the Markedness model).

In a similar micro- Macro integrative approach, Dabène and Moore (1995:25)41 conducted several studies on migrants Iberian (Spanish and Portuguese), and Algerian communities in France. They concentrated on the possible correlation between the generation to which the speakers belong and the code switching patterns that they display. Dabène and Moore (1995: 189) noticed that the two generations of Iberian and Algerian speakers display Different types of Code switching on the basis of switching points in the sentence. A correlation has been made between switching types and the generation to which speakers belong (first or second generation emigrants). Dabène and Moore (ibid: 193) further relate these differences to several factors. These factors include:

 The differing levels of proficiency in the host language, i.e. French, displayed by first and second generation Iberian and Algerian emigrants.

 The necessity of using both languages for communication that may be displayed by members of the second generation

 The symbolizing of different social identities. i.e. the use of a mixed Spanish-French; Algerian Arabic-French code as a social marker of identity on its own in addition to the social functions attributed to the use of Spanish or Arabic alone and the ones attributed to the use of French alone.

 The different patterns of identification that speakers entertain with the host and home language. The young or second generation seem to identify more with the host language and they also seem to consider themselves more fluent in that language i.e.

41 In L. Milroy& P. Muysken ed( 1995:25) 91

French. The patterns of identification of the first generation speakers seem to be quite different.

1.4.3.5.1 Integrating the Social and the Linguistic Approaches in the interpretation of the Present data

Similar tendencies have been observed in our investigation with second and first generation Mzabi speakers in Oran. The first generation Mzabi speakers seem to use an insertional type of code switching in the sense that they mostly insert singly occurring nouns or verbs into otherwise Algerian Arabic stretches of talk. The second generation members of this community tend to display another type of code switching which is characterized by the alternate use of long stretches of talk from Algerian Arabic and Mzabi in addition to the insertion of singly occurring elements (Nouns, verbs and Adjectives); besides, These second generation speakers seem to display a more mixed language (Mzabi-Algerian Arabic) in which they, for example, make use of Algerian Arabic discourse markers in Mzabi interactions. We have chosen to mention the example of discourse markers42 in mixed codes because they are considered by many researchers in contact linguistics (Auer: 1998; Blanc: 2000; Gafaranga: 2002; Li Wei: 2005) as good signs of the emergence of bilingual mixed codes. They also signal a shift from the alternate use of languages to the mixing of these.

We made a calculation of the number of Algerian Arabic and French discourse markers inserted in Mzabi utterances. We compared the number of these discourse markers (Algerian Arabic and French discourse markers) used by first and second generation Mzabi speakers in Oran43. We observed that out of an overall number of 353 Discourse markers used by first generation Mzabi speakers, 33% are Algerian Arabic while 07% are French. The tendency

42 Discourse markers are connectors such as conjunctions and extra sentential phrases such as „well, you know, and come on‟ e.g. well I am moving out now are you coming with me or not. Come on be courageous. That‟s your problem, you are a contend person. You should react you know!

43 The data that we used for this calculation is based on two of the several tape recordings of the corpus. The first tape recording has been made in a Mzabi owned hardware shop. The two Mzabi speakers who took part in the interaction are first generation Mzabi speakers aged 43 and 36 who came to Oran in 1986 for the first one, and in 1989 for the second one. The second tape recording contains interactions between 4 Mzabi youngsters who were born and brought up in Oran. At the time of the recording these youngsters were aged 27 and 25.

92 seems to be higher with second generation Mzabi speakers. Out of a total number of 465 discourse markers, 47% are Algerian Arabic, and 12% are French.

The tendencies in the two groups may be partly related to the differing degrees of proficiency in Algerian Arabic, and French in the first and second generation Mzabi speakers of Oran. We have noticed, for example, that the second generation group seems to be more proficient in Oran Arabic than is the first generation group. Their Algerian Arabic seems to be close to Oranian. The first generation group still keeps a variety which is rather close to the varieties of the centre of the country.

The analysis of the Mzabi corpus also revealed that the second generation speakers seem to display a mixed code which contains a lot of Algerian Arabic. This mixed code is even considered by the first generation speakers as bad and corrupted Mzabi Berber. We may cautiously hypothesize that this code is a marker of a multiple social identity i.e. a Mzabi and an Oranian social identity. We noticed a similar mixed code in the speech of second and third generation Kabyle speakers in Oran. Here is an example of this mixed code in an interaction between two Mzabi speakers in a Wholesale hardware shop:

1. Azziz: /   ./?

(Lit: What do you … ? What do you need?

2. Mounir: / / la commande /    /

(I am looking for the order which (is) there. By the way give me

/  ..    /

(The invoice but Inacomp-check- 2nd sing aff the price of the list (the list in the invoice) 93

3. Mounir: /   /

(By the way are you coming with us )

4. Azziz: /         /

(I do not come with you I mean you do excuse me today I am forgiven (I have some stuff to do) (Mzabi Berber- Algerian Arabic code switching)

What is of a particular interest for us in the above interaction is the use of the Algerian Arabic discourse markers // (by the way), /.. / (but), and /  / (I mean). We may interpret them as a sign of a change in the patterns of code switching in the speech of second generation Mzabi speakers of Oran. These speakers may be in the process of developing a mixed code (Mzabi Berber-Algerian Arabic). This mixed code is quite different from the one that they used to have in the past few years44. In our previous investigation (Benhattab, 2004: 52) of this particular aspect of code switching in the Mzabi community, we noticed that they (Mzabi speakers) rarely used discourse markers from French or Algerian Arabic in their speech (Benhattab, 2004: 51). There has been a change since then even if seven years is said to be a quite short period from the sociolinguistic point of view. The point that we are making here is only a preliminary observation that needs not to be generalized. We intend to explore this area of research in a future investigation on the differences between first and second generation Mzabi speakers in terms of code switching patterns.

Other approaches integrating Micro and macro sociolinguistic perspectives include Bourdieu (1977; 1991), Gardner Chloros (1991; 2009) and Milroy& Li Wei (1995). The

44 This observation is based on the study that we carried on the Kabyle and the Mzabi communities in 2004. One of the hypotheses that we made is that the second and third generation Kabyle speakers were in an advanced phase in the making up of a Kabyle- Algerian Arabic mixed code. The second generation Mzabi speakers were in the beginning of this process( Benhattab, 2004: 51)

94 investigations conducted by these scholars are different in their methodological procedures, but they share a common integrative vision.

1.4.3.5.2 Bourdieu’s (1991) application of the Theory of Language Practice in the Interpretation of Code Switching

Bourdieu (1991) developed a whole theory in the interpretation of monolingual and bilingual linguistic behaviour. This theory is the one of ‘language and symbolic power45’ (Bourdieu, 1977; 1991). It is itself embedded into a broader theory of social practice (Bourdieu, 1977; 1991; 1993, 1998). His approach is interesting in the sense that besides integrating micro and macro social variables in the understanding of code switching, he puts symbolic power of speakers at the centre of his research procedures. Bourdieu (1977: 657) argues that speakers’ actions are closely linked to the possible outcomes of these actions in linguistic markets46, but before we go any further in the discussion of Bourdieu’s approach let us define some key concepts in his theory. These include the notions of ‘Field’, ‘habitus’ and ‘symbolic capital’.

1.4.3.5.2.1 Presenting key concepts in Bourdieu’s (1991) model with a special reference to the data of the present study

‘Field’ has been defined by Bourdieu (1991: 232) as a “social space or network of objective relations between positions”. This means that field is a social context or a social network i.e. a set of social ties that are established and maintained through the use of language by a given social group. Bourdieu (ibid: 232) does not use the concept of position as a stable social position in society but he uses this term in a dynamic manner. Positions are dynamic in the sense that they are related to symbolic social capital that can be cashed in the

45 Bourdieu (1991: 220) uses the concept of „Power‟ as the possibility to gain access to powerful social positions thanks to the mastery of officially acknowledged varieties and thus to the successful participation in society at large. This consequently grants speakers some kind of symbolic capital of various kind( blommaert, 2005: 70)

46 The concept of „market‟ is used by Bourdieu (1977:655) with its sociological meaning i.e. the social context, the social setting, or the social situation. 95 social group. They are also dynamic for the concept of field is considered by Bourdieu (1991: 230) as:

“Highly unstable and subject to continuous symbolic struggles involving forces of transformation and conservation of the relative value of different kinds of capitals”

This is relatively true for the linguistic situation of the Berber minorities under study here. The positions of Kabyle and Mzabi speakers are presently quite different from the positions they used to have previously in Oran Mainstream Algerian Arabic speech groups. In a previous language preference investigation that we carried out in 2004 we found that members of these minority groups stigmatized Berber as being the language which is outdated and the one they use the least right after Standard Arabic. Their attitudes have been changing since then. The elicitation procedure that we carried out this in 2009 on these groups revealed that Kabyle and Mzabi informant alike display quite positive attitudes towards their Berber varieties (Mzabi and Kabyle). We will come back to this issue in the coming chapter which deals with Berber speakers attitudes towards the languages which make up their verbal repertoire.

‘Symbolic Capital47’ may be equated with the outcome that one obtains thanks to his linguistic competence. It is related to the social prestige that the speaker accumulates through his linguistic practices. As it has been stated above this capital may be cashed in different forms. It is not only economical (money or property) but also cultural (Knowledge, skills, and education) and most importantly social (accepted membership to a social network, newly established acquaintances, and memberships to particular social groups).

If we relate the notion of ‘symbolic capital’ to the study that we carried on the Berber minorities of Oran, many observations have been made in relation to the outcomes of using

47 Capital, in Bourdieu (1991: 656) terms, is not necessarily linguistic, but also economic like cash or assets, and even cultural (knowledge acquired previously, skills, and formal education).

Kabyle and Mzabi in the Berber inner networks of this city . We noticed, for example, that speaking Mzabi may be a very good help in Mzabi social networks. This may bring immediate economic capital48. Being a member of the Mzabi Social network also grants cultural capital in the form of formal education. Children of Mzabi Parents are granted free private courses in most subjects in the Mzabi cultural associations. They are at the same time given courses in Mzabi to maintain this variety in their verbal repertoire; they are besides given courses in religious education to promote and to maintain the Ibhadite49 Islamic trend in these children.

‘Habitus50’ refers to a set of predefined social linguistic behaviour. These linguistic pre- dispositions are responsible for the speaker’s attitudes towards the languages or the varieties at his disposal. These pre-dispositions are also responsible for the way he perceives his linguistic behaviour, and the way he uses language. The origin of these dispositions can be traced back as far as in the language experiences of early childhood. The nature of these pre- dispositions is a well structured and a long lasting one. They are structured in the sense that they are strongly tied with the social context in which they have been originally acquired (Childhood); accordingly, they are long lasting in the senses that are deeply rooted in One’s life history51. Madsen (2008: 50) argues that:

“Habitus is based on the speaker’s experience as a participant in various fields of social life and develops in accordance with one’s life trajectories. It is to be understood as practical sense rather than an intellectual consciousness Bourdieu develops the notion of

48 In wholesale trade in general and in wholesale hardware trade in particular, Mzabi traders may have access to a range of trade support from their fellows. They are for example granted a 0% interest loan to start their businesses. They are also given facilities in terms of payment of orders i.e. they are allowed to pay for the goods they purchase after they sell them. They are granted a 3% discount on all goods, and they are granted priority for important goods. Members of the Mzabi community have an informal system of finance that is based on the Bait al Maal (the treasury house). This informal financial institution is present in all the cities of Algeria. It is compulsory for all members to donate yearly from their revenues. The money collected may be used by the community to finance marriages, businesses, houses and flats purchasing, and to help the poor Mzabi members( Bakir Mohamed 1993:87)

49 The majority of Mzabi speakers follow the Ibhadite religious trend(Bakir Mohamed 1993: 12)

50 „Habitus‟ is an old sociological term, but Bourdieu uses this term to mean a set of dispositions. These dispositions generate practices and attitudes that are very much like norms( Myers Scotton, 2007: 114) 51 Cf : E. Botzepe( 2002: 15) 97

‘habitus’ to account for an observed general tendency of conservation with respect to societal social differences” (Madsen 2008: 50)

This is related in Bourdieu’s terms (1991:235) to the notion of acceptance of one’s social position in the group. This acceptance is tacit in nature. If we correlate acceptance to the notion of symbolic power, speakers may either accept the power relation and respect their social position or reject these positions and challenge the power hierarchy in the social group. In this vein Bourdieu (1991: 163) states that:

“ It is as structured and structuring instruments of communication and knowledge that ‘ symbolic systems’ fulfil their political function, as instruments which help to ensure that one class dominates another( symbolic violence) by bringing their own distinctive power to bear on the relations of power which underlie them and thus by contributing in Weber’s terms to the domestication of the dominated” Bourdieu (1991: 167).

Language is implied as a one of the symbolic systems available to members of social groups. It is also a strong factor in the shaping of the power relations within these social groups. Mastery of languages may function as a linguistic capital which grants “profit of social distinction to speakers” (Bourdieu, 1991: 77)52. Symbolic power relations are rarely peaceful; accordingly members of social groups are continually fighting to gain representation in their groups. They also try to impose their own rules of conduct or the rules of conduct of the groups to which they belong. Bourdieu (1991:234) states in this vein that:

“The most resolutely objectivist theory must take account of agents’ representation of the social world, and more precisely, of the contribution they make to the construction of the vision of this world, and thereby to the very construction of this world via the labour of representation (in all senses of the term) that they continually perform in order to impose their own vision of the world or the vision of their own position in this world, that is their social

52 Cf Madsen( 2008: 51) 98 identity. The perception of the social world is the product of a double social structuring.” (Bourdieu, 1991:234)

This statement clearly highlights a vision of society which functions on the basis of control patterns and challenges to these control patterns. These challenges to the power relations have been noticed in the Mzabi community of Oran. There are Different groups of Mzabi Speakers in Oran. These constitute first, second and third generation groups. The first generation group constitutes the most conservative group in terms of social conduct, language use, and assimilation in the main stream of Oran society while there are two different tendencies in the second generation group. Some members of this group accept the social rules of conduct and conform to them while others simply put into question these norms and strongly challenge them53.

Bourdieu’s approach is not only relevant to our investigation of Berber minorities in relation to power in minority groups, but it is also relevant to our study in the sense that it integrates micro sociolinguistic instances of language use. Bourdieu (1977b: 657)mentioned, for example, patterns of language use involving varieties of French in code switching, and style shifting. Here is the example he gives of an old woman who lives in a small village in Béarn (a south western province in France)

“The old lady+at one moment used “provincialised French” to address a shopkeeper’s wife, a young woman originating from another large market town in Béarn; *…+ the next moment, she spoke in Béarnais (the local dialect) to a woman who lived in the town but who was originally from (the villages) and more or less of her own age; then she used a French that if not “correct” was at least strongly “corrected” to address a minor official in the

53 The Mzabi first generation group is conservative in the sense that they do not assimilate to the Oran majority group in terms of social, cultural, and most importantly in terms of linguistic behaviour. They have a philosophy of life that is based on an avoidance of spending one‟s money easily, an avoidance of mainstream culture (music, sports…), and an utmost attachment to the original customs and traditions. This group is considered as the norms protector. This group establishes the rules of conduct in the whole Mzabi minority. Some members of the second generation Group strongly challenge this supremacy.They display their disapproval in their daily routines and social conducts. They, for example, overtly show their economic status and assimilate to the mainstream of Oran culture. They even show that in their linguistic behaviour by displaying a kind of Oran Algerian Arabic-Mzabi mixed code. 99 town; and finally she spoke in Béarnais to a road worker in the town, aged about fifty”. (Bourdieu, 1977b: 657)

The situation described by Bourdieu (ibid: 657) in this example resembles the one which has been described by Gumperz (1972: 82) when he illustrates situational code switching in Norway. Bourdieu (ibid: 658) argues that, the speaker evaluates the different contextual cues inherent to the situation before making language choices. He i.e. the speaker also tries to anticipate his interlocutors’ reception of the message. These two criteria guide his language choice, and they also function as constraints to the type of code switching that he is going to display.

Bourdieu’s (1991) theory of language, power and identity has been taken up many researchers in contact linguistics. Myers Scotton (2007: 30) states in this vein that:

“Bourdieu (1991) popularized the notion of thinking of the linguistic varieties we command as symbolic capita54l. This idea applies not just to dialects, but also to the different languages in a bilingual repertoire.”(Myers Scotton 2007:30)

Myers Scotton (ibid: 30) goes even further and acknowledges the relevance of Bourdieu’s (1991) model to her socio-psychological ‘Rationale Model’55 (2001), and states that:

“Such a model can be applied to linguistic choices. The basic idea is that speakers are selecting their choices rationally if they make choices based on weighing costs and rewards of competing choices” ( Myers Scotton, ibid: 114)

54 Emphasis is mine 55 The rationale model as it has been discussed earlier in this chapter has been devised by Myers Scotton and Bolonay in 2001. 100

1.4.3.5.2.2 Criticisms to Bourdieu’s (1991) ‘Theory of Social Practice’ in the literature

Bourdieu (1991) theory has been challenged by some scholars (Wodak, 1997; Chouliaraki& Fairclough, 1999; Agha, 2007: 229) regards one of his central principles which is the one of ‘Habitus’. The criticism put forward by these scholars is that Bourdieu’s definition of ‘Habitus’ does not consider the important role of the individual in the process of Power enforcement. The second criticism is related to the idea that habitus is considered by Bourdieu as a static construct which entails potential changeability (Madsen, 2008: 51). A third criticism points to the idea that Bourdieu accounts rarely for meta-linguistic processes in the making up of ‘Habitus’. These meta-linguistic processes include ‘cognitive and discursive processes’ (Agha, 2007: 229). This counter argument is particularly evident in contemporary discourse and conversational studies. Madsen (2007: 229) argues in this vein that:

“Within contemporary discourse studies, discourse is generally understood as involving more meaningful human semiotic activities than merely linguistic” (Madsen, 2007: 51)

Bourdieu theory (1991) has also been criticized, for he gives a paramount importance to the concept of symbolic power at the expense of downplaying the equal importance of solidarity in vernacular varieties (Mesthrie, Swann, Deumert, & Leap, 2000). Recent conversational and discursive investigations of code switching in minority groups contexts demonstrated that code switching functions primarily as a marker of solidarity and as a symbol of group identity (Botzepe, 2002: 16). In the same vein Gumperz (1982: 97) talked about the dichotomy we/they codes in code switching. He (ibid: 98) made a clear categorization of the two codes in terms of the functions that they fulfil in the social group. Gumperz (ibid: 98) states that the primary criterion that may be used to differentiate the two codes is the solidarity one. The two codes are distributed as follows:

 The ‘we’ code tends to be associated with in group relations and informal activities. It also tends to be aesthetically undervalued in the social group

101

 The ‘they’ code is associated with the majority group; accordingly, it is used as the means of out group communication with the main stream community.

Gumperz (1982: 100) also relates the ‘we code’ ‘they code’ distinction to the attitudes of Minority speakers towards the languages which make up their linguistic repertoires, and towards code switching of these languages or varieties. He (ibid, 100) talks, for example, of match guise procedure in which speakers are asked to comment on recorded stretches of talk in the varieties in use in their community. They display attitudes ‘personal’, ‘subjective’, involvement’ with the ‘we code’, while they display deference attitudes such as ‘objective’, ‘detached’, and ‘distance attitudes’ towards the ‘they code’. This match guise procedure has been put in question by many code switching researchers (Woolford; 2003: 78; Myers Scotton, 2007: 138) on the basis that it may fail to show objectivity by the informants. This may be related, for example, to the gender of the person being in the recording. The issue of the scientific validity of this procedure is still a controversial one. The classical elicitation procedure based on the interviewing of informants via questionnaires may be an alternative to the match guise procedure.

In a similar vein, Gal (1978) investigated the patterns of language choice in the Hungarian-German bilingual community of a border town called Oberwartin in eastern Austria56. The study which has been conducted by this scholar (ibid) was directed to the observation of language choice patterns in relation to different social contexts. Gal’s study shows that Hungarian symbolizes the local traditional agricultural culture, whereas German seems to symbolize modernity and non local values. Gal (1978) could also demonstrate the strong correlation between the individual’s language choices and his/her age. Language choice patterns in Gal’s study (ibid) showed for example that old speakers preferred Hungarian, and that young speakers chose German even if they were addressed in Hungarian. These results

56 Oberwart is situated near the borders between Austria and Hungary. People in this town and surrounding farms found themselves part of Austria after 2nd world war; hence, German became the official language while Hungarian became a minority language in this town. Oberwart is regarded as a town which has an agricultural tradition and vocation; this town has however undergone some deep social changes due to a rapid economic growth. This growth offered the natives of this community the chance to work in jobs other than agriculture. As a result an opposition is created between the values signalled by the waged work and the ones signalled by peasants. These values are represented in the two languages of the community i.e. Hungarian and German.

102 pushed her (ibid: 231) to talk about a process of language shift which was in progress. The argument put forward by Gal (ibid: 234) is that activities which were in the past typically associated with Hungarian are now associated with German. This process of language shift is, presently, at an advanced stage. In an on the sociolinguistic situation of Oberwart, Myers Scotton (2007: 91) states that:

“The number of Hungarian speakers is declining. The small size of the community makes the difference here, as well as the norm to speak German in the town when German monolinguals are present. All the Hungarians are bilingual in German, the only official language in Austria. Further, German is the language associated with socio-economic mobility and jobs in the town.” (Myers Scotton, 2007: 92)

The correlation between age and patterns of language use seem to have an important role in language shift in Oberwart community. This correlation has been confirmed in many other bilingual communities (Young & Tran, 1999 on the Vietnamese minority in the united states; Al Khatib, 2001 on the Armenian minority of Jordan; Langman, 2002 on the Hungarian minority of Slovakia; Raschka, li, and lee, 2002 on the Chinese minority of Tyneside England). The arguments made in these studies are to their majority based on a questionnaire elicitation procedure.

In a previous investigation (Benhattab, 2004: 110) of language preferences in the Kabyle and Mzabi Minorities of Oran, we observed similar tendencies in our Berber informants. Age seems to be correlated with language preferences in both the Kabyle, and the Mzabi communities of Oran. We made the hypothesis that this may signal a pattern of intergenerational language shift.

Gal (1978) also demonstrated in her investigation that the interlocutor is an important factor to the individual’s language choice. Gal (ibid: 229) could demonstrate the correlation between the speaker’s network and his patterns of language choice. This idea has been taken

103 up by Milroy (1987) in her study of the relationship between speaker’s social networks nature and their patterns of language choice.

1.4.3.5.3 Milroy’s (1987) and Milroy & Li Wei’s (1995) Investigation of Language Choice in Relation to Social Network

Milroy (1987) investigated the relationship between social network57 and monolingual language choice in the working class language variety in Belfast. In her study of the working class vernacular speech of three inner city communities in Belfast, she found that the dense and multiplex nature of a working class individual’s social network58 gives rise to its imposing the vernacular form, which symbolizes in-group solidarity, on his or her code choice as Milroy (1987:139) puts it:

“In Belfast, ‘a network strength scale’ was devised to examine the relationship between strength of tie and variation in monolingual language behaviour in Belfast”

Milroy and Li Wei (1995:136)59 tried to compare the findings reached in Milroy’s (1987) monolingual study in Belfast with a bilingual context i.e. the one of the Tyneside Chinese minority (in Newcastle). A similar relationship between social network and language choice has also been found among second generation Chinese/English bilinguals in Newcastle upon Tyne

57 “A social network may be defined as a boundless web of ties which reaches out trough a whole society, linking people, to one another for, however remotetely, however for practical reasons, social networks are anchored to individuals, and analysis is limited to between twenty and fifty individuals ”( L.Milroy & L.Wei 1995:138)

58 A high-density network, as opposed to a low-density network, is the one where one’s contacts also know each other. In a multiplex network, on the other hand, individuals interact with each other in more than one capacity (e.g. neighbour and co-worker at the same time).

59 In L. Milroy & L. Wei ed ( 1995 :138) 104 in England (Li Wei, 1995; Milroy & Li Wei, 1995). They found that this social variable is the one that best explains language choice in code switching. Milroy and Li Wei (1995:155)60 state that:

“While network interacts with a number of other social variables such as generation, gender, and occupation, it is capable of accounting more generally than any other single variable for patterns of code switching language choice.”

Similarly to Gal (1978: 234), Milroy and Wei (1995:157) put the speaker’s social networks at the centre of their approach. This is again putting emphasis on the interlocutor. This tendency has not only been taken up by scholars interested in bilingual data, but it has also been used by scholars interested in so called monolingual contexts. What is common in the two approaches is that code switching accomplishes for the bilingual speaker what style shifting accomplishes for the so called monolingual speaker. Romaine (1995: 170) argues in this vein that:

“Although Gumperz has concentrated on multilingual and bidialectal situations where these functions are mapped onto different languages and varieties, the value of his approach lies in the fact that we can see style shifting and code switching as equivalent at the level of discourse. Style shifting accomplishes for the monolingual what code switching does for the bilingual” Romaine (1995: 170)

1.4.3.6 Bell’s (1984, 2001) Speech Accomodation Approach to bilingual discourse

Bell (1984: 150) also puts the interlocutor or the audience at the centre of his speech accommodation approach to monolingual style shifting and variation. The basic principle in the approach put forward by Bell (1984:152) is that:

60 In L. Milroy & L. Wei ed ( 1995 :155) 105

“Speakers design their style of speaking primarily for and in response to their audience i.e. they vary their way of speaking depending on who they are talking to, or who their audience is( actually present or via radio or TV61)” (Bell, 1984: 152)

Bell (2001: 144) goes even further by extending his definition of audience design to bilingual situations62. He states that code switching may fulfil the same conversational functions as monolingual style shifting. Code switching is seen in the ‘audience design’ fashion as one of the possible alternatives available to the speaker to accommodate to his interlocutors’ linguistic preferences (Coupland, 1985; Sachdev and Bourhis, 1990; Wei Zhang, 2005)63. The concept of audience design has also been used by conversational analysts as an alternative to interpret the functions of code switching in bilingual conversations (Gardner Chloros, 2009: 64; Wei Zhang, 2005: 364). Gardener Chloros (2009: 64) argues that:

“The concept of audience design developed by Bell (1984, 2001) and Coupland (1984) can help explain many cases of code switching”

Gardner Chloros( ibid: 78) mentioned instances of code switching where the speaker has to accommodate his language because of a mismatch between his level of competence in language and the one of his interlocutors. This type of accommodation has been considered by many researchers (Ait Siselmi, 200364; Burt 1994 Coupland, 2001; Gardner Chloros and Charles, 2007; Woolard, 1988) as a compromise function of code switching to which speakers resort in bilingual and bidialectal settings.

61 The premise behind Bell‟s (2001) approach is that the interlocutor may be physically present or not as in cases of TV or radio broadcastings. A good illustration of audience design may be the study conducted by Wei Zhang (2005) on radio interactions in a bidialectal area of china whereby the Radio broadcaster switches, according to the variety he is addressed in by the phone caller, between Cantonese, one of the standards in China, and Putonghua, a provincial variety. 62 Most of Bell‟s (1984; 2001) work is embedded within a speech accommodation approach to monolingual style shifting and variation (Myers Scotton 2007: 156). 63 Cf. Gardner Chloros( 2009: 78) 64 Cf. Gardner Chloros(ibid: 79) 106

Gardner Chloros and Charles (2007: 99) described processes of compromise code switching between English and Punjabi on Zee TV65. Ait Siselmi 66(2003) mentioned the example of Beur FM, which is a French radio channel directed to the Maghrebi minority of France. Ait Siselmi67 observed an extensive use of code switching between Dialectal Arabic (Algerian Arabic, Tunisian Arabic, and Moroccan Arabic) and French in the programs broadcasted by this channel. Gardner Chloros (2009: 79) also mentioned the example of a London Greek Chypriot radio.

We may also mention the example of ‘Alger Chaine Deux’( Algiers Channel two)which broadcasts in Berber, but in which we may observe many instances of code switching between Berber (with its varieties), Arabic, and French. Broadcasters in this radio often use Code switching as a compromise to accommodate their language to the one of the audience. Guests of this channel also resort to code switching as a compromise to maximize the understanding of their audience when it comes to technicalities of fields such as medicine, electronic, literature and the like.

The idea that the interlocutor is an important factor in code switching has also been put forward by Gumperz (1982: 45). He (ibid: 47) argues that Code switching may have the conversational function of addressee specification68 in interactions as he puts it:

“Using the appropriate language to address different interlocutors allows the participants to continue the conversation smoothly, without undo flagging69 of who they mean to address” (Gumperz, 1982: 45)

65 Zee TV is a TV channel whose audience is essentially first and second generation minority Indian speakers. The first language of the parents and grant parents of this community may be one of the Indian varieties. Their language of literacy is Hindi. The second generation members of this group have English as their first language. To attract an optimum audience, the TV programs in this channel contain much Code switching between 66 In Gardner Chloros (ibid: 78) 67 Cf. Gardner Chloros( ibid: 79) 68 This function of code switching has already been discussed on page 28 where we have been introducing the different conversational functions of code switching, but we decided to come back to addressee specification in this section because it is an important premise in the „Audience design‟ as it has been advocated by Bell(1984; 2001)

107

Code switching is used in such cases to signal a change in addressee in the interaction or to specify the addressee being addressed by using, for example, a language with one of the addressees, and another language with the other addressees without interrupting the flow of conversation. Conversational analysts(Li Wei, 2005: 382; Gardner Chloros 2009: 80; Gardner Chloros, Charles, and Cheshire, 2000: 1334) have demonstrated that bilingual speakers have got the capacity to conduct more than one conversation in a single interaction by resorting to code switching as a strategy to show changes in the addressee in conversation. This bilingual strategy may be compared to changes operated in tone of voice or in pitch in monolingual conversation to signal an addressee specification. Here is an example taken from our corpus. The extract is a conversation between three students from the department of English.

01. Linda: /  -/ (Lit: and you who 2nd object prefix- called (plural)) (And who did call you)

02. Amira: // Rafik( noun affix- Rafik) 03. Linda: / -  / (Lit: What which want-3rd sing accomp like that?)(What did he want from you?)

04. Amira: -- -  l’examen (Lit: He called to her Inf affi-ask- 3rd sing pre about the exam)

(He called her to ask about the exam)

05. // Malika ça va les notes/     / (What’s up Malika is it alright with marks or not have you learnt or not)(Addressing Nacera)

06.   DS, Amira,   (You attend to the Make up exam or not)( addressing Amira

69 Flagging is used here by Gardner Chloros to mean the insertion of conversational markers such as you know, well or comments at the point where the switch is made. It is used in the same way as Poplack‟s definition of Flagged code switching( see page 81)

108

07.  les notes   ( did you count your marks or not yet)( addressing Malika) 08. Malika: ça va il me reste juste quelque module /  / 09. ( Things are alright I only have three modules( to make up). We are we are learning perfectly well) 10. Amira:   Linda!  -     --   - (Interj (come on) you are making us crazy Linda go! Change place isolate yourself (to learn) god will help you)

What is interesting in this conversation is the code switch to Algerian Arabic displayed by Linda in line 5 when she wants to address Malika (She was addressing Amira in Kabyle in the previous line). Linda carries on two conversations, one with Amira in Kabyle, and the other one with Malika in Algerian Arabic and French). She resumes using Kabyle with Amira in line 06, and then comes back to Algerian Arabic with Malika in line 07. The type of code switching displayed in this conversation looks like a language accommodation pattern (a convergence one). Gardner Chloros( 2009: 80) argues that studies which try to correlate code switching with patterns of accommodation in an ‘Audience design trend’ are few in the sense that this latter has been applied mainly to so called monolingual contexts.

Lawson Sako and Sachdev (2000: 1358) tried to investigate the degree of convergence of Tunisians to requests70 made in Tunisian Arabic, French, and in a Tunisian Arabic French mixed code. They (ibid: 1356) found that convergence to the language used by the questioner seems to be the overall tendency. This is especially true when the questioner looked as if he belongs to the in-group or used the in-group code which is either Tunisian Arabic or a mixture of Tunisian Arabic and French. Tunisian Arabic-French mixed variety seems to be used as one

70 Lawson Sako and Sachdev elicited their respondents by sending researchers who approached in a random manner 711 people in the street and asked them in Tunisian Arabic or in French directions to the local post office. The language of the reply was recorded by researchers as either convergent (a reply in the language of the question), divergent (a reply in another language from the one of the question), or code switching (using both Tunisian Arabic and French in the reply). 109 of the in-group varieties in the Tunisian street. This mixed variety is also positively valued by Tunisian speakers (Lawson Sako and Sachdev, 2000: 1351), and it is even reported by respondents that they use code switching more than Tunisian Arabic in their daily life. Lawson Sako and Sachdev (ibid: 1351) argue that:

“Initial patterns of reported language use revealed that the varieties most spoken in Tunisia were Code switching (42%), Tunisian Arabic (38%), while the use of French (5%), and Modern standard Arabic (1%) was relatively low…Tunisian Arabic and Code switching were reported to be the varieties most used in discussing family matters, and in-group ones. The differences obtained as a function of ethnicity suggest that code switching may be a distinct variety used in intra group interaction” (Lawson Sako and Sachdev, 2000: 1356)

These results may suggest that Tunisian speakers report a high degree of code switching use in their interaction. This tendency is contradictory with their attitudes towards the varieties which make up their verbal repertoire. The respondents in Lawson Sako and Sachdev study stigmatize Tunisian Arabic-French code switching in their reported attitudes. Similar tendencies have been observed in the Berber Minorities under investigation in this study. We shall discuss this issue of reported language use in the Berber minorities of Oran, and the one related to the attitudes of these speakers towards the varieties which make up their verbal repertoire in the coming chapter.

The relationship between accommodation and code switching is not as simple as it seems for accommodation is not related only to language choice but also to other sociolinguistic factors such as Gender ethnicity and social class. These factors may have as much importance in code switching patterns as attitudes to varieties being present in the verbal repertoire of the community under study. This view of code switching is similar to the ones that we have seen above in the sense that it also integrates micro and macro sociolinguistic factors relevant to the interpretation of code switching patterns. What is interesting with Bell’s (1984; 2001) approach and its related approaches is that they put the interlocutor at the centre of their procedure.

110

1.4.3.7 Gardner Chloros (1991, 1995) Continuum Approach to Code Swtiching and other Contact Phenomena

Gardner Chloros (1995)71 is to be singled out for her original approach to code switching. She (ibid: 70) could demonstrate, among other things, that code switching and Borrowing are not distinct phenomena72, but are rather part of a contact continuum which also includes interference, pidginization, and creolization as she puts it:

“Even within what are generally accepted as code switches, we are dealing with a number of overlapping phenomena” (Gardner Chloros, 1995: 71)

Gardner Chloros (1995: 75) clearly departs herself from the general tendency in the contact linguistics research community when it comes to the distinction between contact phenomena. This tendency is represented by Poplack and her associates who advocate that code switching should be clearly distinguished from the other contact phenomena as she (1987: 72)73 puts it:

“What is important is that this phenomenon (i.e. code switching) be clearly distinguished, first conceptually, and then operationally as much as possible, from all other consequences of bilingualism which involve not alternate use, but the truly simultaneous use of elements from both codes …Not least important, all of these phenomena should be

71 Gardner Chloros (1991) conducted an investigation on the speech community of Strasbourg (in Alsace). Alsace is a bordering area between France and Germany. The members of Strasbourg‟s speech community display different patterns of code switching which are related to both micro and micro sociolinguistic variables. The micro sociolinguistic factors include, for example, conversational functions attributed to code switching. Macro sociolinguistic factors include the overt status enjoyed by the varieties in the speech community.

72 For the time being, we shall dismiss the discussion on the distinction between code switching and Borrowing. This discussion will be resumed in chapter two where we shall be discussing borrowing in ample details.

73 In Gardner Chloros(1995: 75) 111 distinguished from speech errors which involve elements of both languages, and which may be properly considered ‘interference’.”(Poplack 1987: 72)

This clear cut distinction as it has been expounded by Poplack (1987) may be true for her data but it needs not be generalized to other bilingual contexts. It has been shown in many studies that code switching and interference74 go, for example, hand in hand in a speech community

Gardner Chloros (ibid: 70) also points out the need to draw group macro generalisations from micro individual sociolinguistic behaviour. She argues that regarding Code switching as a purely micro sociolinguistic phenomenon implying a binary choice between two codes or language system is too simplistic; furthermore, She(ibid: 70) observes that the differing patterns of code switching displayed by the speakers are not only related to Micro sociolinguistic variables, but are also related to macro ones:

“The type of language mixing which occurs in given setting depends on a number of factors including the relative prestige of the varieties in question and the extent to which they are considered separate, identifiable languages; moreover, it is very unlikely that in any language contact situation only one kind of interlingual phenomenon is to be found, and that this kind can be parcelled off from the rest”( Gardner Chloros, 1995: 70)

Gardner Chloros (ibid: 75) further argues that the very notion of languages being used by speakers as separate and distinguishable entities is problematic and controversial even if many renown scholars such as Poplack(1987: 72)75 are strong proponents of this argument. The evidence put forward by Gardner Chloros (1995, 76) in this vein is the one of lexemes

74 Interference and code switching are used here following Haugen( 1956)who differentiates code switching and interference on the ground that the former takes place when the properties of the codes being used are preserved, whereas the latter i.e. interference and integration take place when there are instances in which there is a levelling of the differences between the two codes.

75 In Gardner Chloros(1995: 75) 112 which may belong to language A or B and which are used with morphological markings from both languages. Here are some examples from a Punjabi-English code switched corpus in Leeds (Agnihotri, 1987)76

Aunt-// aunt-// (English lexeme Aunt marked by the plural morpheme markers // from English and// from Punjabi

Week-/ week-(English week marked with both morpho-phonemes standing for plural i.e. the // plural marker in English and the // plural marker in Punjabi.

- (Punjabi prayer marked with //the bound morpho-phoneme marking plural in English), and marked with the Punjabi Morpheme // standing for plural in Punjabi. (Punjabi-English code switching) (Agnihotri: 1987)

Similar examples have been found in our data. At first glance we thought that they represent idiosyncratic interference mistakes made by individual speakers, but their recurrence in different social settings involving speakers from both the Kabyle and Mzabi speech groups obliged us to interpret them. A possible interpretation may be that they could be classified as parts of a continuum between code switching and interference. A further investigation needs to be conducted on the continuum between code switching and interference in the Berber community of Oran. Here are some examples:

// (sweet) has been attested in three instances in the corpus. This adjective is made up of a nominal stem from Kabyle and an Algerian Arabic morphology marking i.e. // at the end of the adjective standing for singular feminine marking in Algerian Arabic. // (sweet) the equivalent in Kabyle is also in current use in the Berber minority of Oran. The interesting here is that // does not exist in Algerian Arabic.

76 In Gardner Chloros(1995: 76) 113

/-/ (She made small circles with them i.e. with the minced meat). This verb exists in Kabyle, but not in Algerian Arabic. What is striking in this example is that it is used with an Algerian Arabic morphosyntax; even if the verb does not exist as a lexical item in Algerian Arabic. The Kabyle normal production of this verb is /- / (She turned them).

These examples highlight contact situations where the languages are so mixed that code switching becomes the unmarked choice (Myers Scotton, 1993: 76). Such situations not only, favour the overlapping of contact phenomena, but they also make it difficult to separate the languages or the varieties in use in the community. These situations may also signal an ongoing language shift operating in different generations in the community. The above cited examples hace been observed in second and in third generation Kabyle and Mzabi speakers in Oran. They haven’t been observed in first generation speakers. This may lead us to put forward the hypothesis that the unmarked language among the members of the Berber community of Oran is different.

The language contact continuum described by Gardner Chloros (1995: 77; 2009: 33) also includes the pidginization77, creolization78 and decreolization processes. Many scholars (Gumperz, 1964: 1121; Crowley, 1990: 384; Myers Scotton, 2006: 281; Gardner Chloros, 2009: 33;) have investigated the relationship between code switching and creolization. Gardner Chloros argues in this vein that:

77 “Pidgins were and are developed by speakers for short lived and for very specific interaction types. Most researchers accept that the word Pidgin comes from the English word business. Records indicate that a pidgin developed on the southern China coast in the early eighteenth century between traders and local businesspersons…Most pidgins developed as medium of communication between people speaking different languages, but wanting to do business together. Other pidgins developed in master servant relationships, often involving colonial civil personnel and their indigenous subordinates or house servants.”( Myers Scotton, 2006: 279)

78 “Before linguists started studying the term “creole”, it was already in use as the term for people who lived in European colonies, especially people of mixed descent. Then the languages that developed in these colonies came to be called creoles. These languages were created by speakers of non-European languages who worked on European plantations, whether as slaves, indentured workers, or indigenous workers. Other types of creoles developed among speakers who all were non-Europeans. There may also be new creoles developing today( Atlantic Caribbean creoles, South Pacific creoles such as Tok Pisin on Papua new Guinea, and Hawai‟i creole English”(Myers Scotton, ibid: 280) 114

“Gumperz( 1964) and Gumperz and Wilson(1971) identified the close relationship between code switching and creolization in the early days of code switching research…these processes clearly co-occur, derive from similar social factors and may sometimes lead to similar outcomes”( Gardner Chloros, 2009: 33)

Romaine (1992: 45) investigation of Tok Pisin (the creole used in Papua New Guinea) shows that there is a clear correlation between code switching and creolization. She (ibid: 323) describes the situation in Papua New Guinea as follows:

“In town Standard English, English spoken as a second language with varying degrees of fluency, Highly anglicized Tok Pisin, more rural Tok Pisin of migrants, and the creolized Tok Pisin of the urban-born co-exist and loosely reflect the emerging social stratification”( Romaine, 1992: 323)

This situation may be regarded as a one in which there are patterns of situational code switching in which the defining social variables would be the social backgrounds and statutes of speakers. The patterns of code switching would involve, according to some creolists (Siegel, 1994: 147), Tok Pisin and English. Romaine (ibid: 322), However, argues that it is in fact very difficult to decide to which language some utterances belong to. This brings us back to Gardner Chloros (2009: 33) important premise i.e. that code switching and other contact phenomena may be interwoven and that they may make up parts of a single contact continuum.

Le Page and Tabouret Keller (1985: 125) also illustrate the overlapping that may exist between code switching and creolization in their study of the Caribbean based English creoles in London. They (ibid: 126) 79 argue that code switching in such creole situation does not always involve switching between clearly distinguished varieties, but it rather involves the shifting at particular linguistic levels. Le Page and Tabouret Keller (1985: 179) even argue that,

79 In Gardner Chloros(2009: 33) 115 sometimes as is the case of London Jamaican, code switching is more symbolic than it is real as they put it( ibid: 180):

“London Jamaican is more a set of norms to be aimed at than an internally coherent and consistent system. Speakers behave as if there were a language called “Jamaican”, but often all they( Perhaps all they know how to do) is to make gestures in the direction of certain tokens associated with Jamaican Creole which have a stereotypical value”( Le Page and Tabouret Keller, 1985: 180) 80

Gardner Chloros (2009: 34) talks about mixed compound verbs as another outcome of language contact. Mixed compound verbs show the haziness of the line separating the different contact phenomena in general and between code switching and creolization in particular. Different scholars (Romaine, 1995: 133; Muysken, 2000: 78; Myers Scotton, 2002a; Gardner Chloros, 1995: 78, 2009: 34; Edwards and Gardner Chloros, 2007: 75) have investigated this phenomenon.

Romaine (1995: 131) investigated the functioning of Punjabi-English mixed compound verbs in the Indian minority of Birmingham. The Punjabi verbal operator ‘kerna’ which stands for the verbal operators do and make in English is commonly combined with the major grammatical categories( verb, noun, adjective)taken from English to make new verbal compounds. These compound verbs are not recognized by bilingual speakers as representing two different items being code switched, but are rather considered as single syntactic and lexical units. Romaine mentions the following examples81:

Ple- kerna (Ple comes from the English verbal stem play while kerna functions as a verbal operator meaning do or make in Punjabi). The mixed compound means to play. Its equivalent in Punjabi monolingual language would be khelna (to play) (Romaine, 1995: 131).

80 In Gardner Chloros(2009: 34) 81 the Punjabi code switched part is in Bold following the original emphasis from Romaine 116

Phonam-Kerna (Phonam comes from the English verbal stem to phone while the mixed compound means to phone)

Here are other examples given by Romaine (ibid: 133): Mix-Kerna (to mix), Pick-up-Kerna (to pick-up), Depend-Kerna (to depend), use-Kerna (to use).

The phenomenon of mixed compound verbs has been observed in other bilingual contexts. The similarity in most contexts in which these mixed formation have been attested is the fact that they involve languages which are typologically distant (Gardner Chloros, 2009: 34). Appel and Muysken (1987: 127) report mixed compound verbs involving English and Surinam Hindustani. Gardner Chloros (1995: 78; 2009: 34) investigated the occurrence of Greek Chypriot- English mixed compound verbs in the Greek Chypriot Minority of London. She (2009: 34) could demonstrate that English major categories i.e. verbs, nouns and adjectives are combined with a Greek Chypriot verb ‘kamno’ (meaning do, make). Here are some examples given by Gardner Chloros (2009: 34)82:

Kamno use (to use)

Kamno respect (to respect)

Kamno developed (to develop)

What is interesting with these examples is that neither English nor Chypriot Greek has any constructions resembling these compound formations; consequently, we may say that they highlight some kind of creolization and that they also show that the borderline between the languages being in use in the speech community is sometimes very hazy from the pint of view of speakers perceptions as Gardner Chloros (2009: 34) puts it:

82 the Greek Chypriot part are spelled in bold following original emphasis by Gardner Chloros. 117

“This may also be viewed as a product of the fuzziness of the borders, in linguistic terms, between different types of contact languages. But whereas prototypical pidgins and creoles arise as linguafrancas in situations involving contact between several languages, other types of mixed languages arise where there is widespread bilingualism…” Gardner Chloros (ibid: 34)

118

1.5 Conclusion

The micro and macro sociolinguistic approaches to code switching we have dealt with in this overview are the ones that we found most relevant to our study. Mentioning all the socially based studies dealing with this phenomena falls out of the reach of a doctorate dissertation. The different studies that we have presented may be divided into two groups. The first group is represented by studies whose perspective is to investigate the macro social functioning of languages in contact. The second group’s interest is directed towards the microanalysis of the social outcomes of language contact. We consider the perspective of scholars in this group as the most relevant to the study that we have conducted. The third group takes integrative approaches to the analysis of code switching. These approaches are as follows: linguistic and sociological, Macro and micro sociolinguitic, or directed towards power and solidarity as features influencing language contact.

We have observed that many of the social approaches we discussed do not make a relation between the structural or grammatical side of code switching and the social one. The few trials made to integrate the social and the grammatical approaches include, for example, James Smith (2002) who has made a trial to correlate the Matrix Language Frame and the Markedness models put forward by Myers Scotton in 1993. Another interesting approach is the one advanced by Dabène and Moore (1995) in their study of the Iberian and the Algerian minorities in France. Still specialists in code switching research tend to go either for grammatical investigations or for social investigations. Many contact linguistics studies also separate the two aspects but do not prevent themselves from tackling both in separate sections.

We will be trying to proceed in a similar vein. In the coming chapters, we shall turn to another important aspect of code switching studies, which is the one of the grammar of code switching. We will be dealing with the most influential, and the most popular approaches which have attempted to explain and interpret code switching from a grammatical point of

119 view. We shall do this using a critical view testing the different approaches on the Kabyle and Mzabi bilingual data before hand. We shall also try to go into the problematic aspects of each model when our data permits that.

The Grammatical approaches to code switching have taken three perspectives in the analysis of code switching and other related contact phenomena. The first persepective is the equivalence-based perspective. The second perspective is the government-based perspective, and the third perspective is the language asymmetry based perspectives. Each one of the coming chapters will address one of these perspectives with a special reference to the Berber data we have investigated. The next Chapter tackles the equivalence-based perspective in the analysis of code switching.

120

Chapter Two

Grammatical Approaches to Code Switching (1)

Equivalence based Studies

121

2.1 Introduction

Investigations of the grammatical characteristics of code switching and other contact phenomena such as borrowing, interference have been important fields in contact linguistics research. These studies have also been very prolific, and they have helped enormously in the understanding of language systems, psychological processing of languages, and the social functioning of languages in human groups. The studies carried on the grammar of code switching may be divided into different groups in terms of their implying principles. These are mainly:

 Scholars who have a monolingual perspective to code switching. Those scholars have as their implying principle the idea that code switching is a deviant behaviour from the established monolingual norm (the one which states that languages are separate entities.

 Scholars with a bilingual perspective to code switching. Those scholars have as their implying principle the idea that code switching is the product of a language bilingual competence; and thus consider this process as a normal outcome of contact situations.

The first group portrays code switching from a monolingual perspective. This perspective considers that code switching represents a danger, an anomaly, or a malady. Muysken (2000: 1) argues that:

“In most models portraying the functioning of the speaker/listener, pictures we carry in our minds or see portrayed in a textbook, a single grammar and a single lexicon are embedded in the network of relations that constitutes the model. This is so commonplace that the essential enrichment of having several grammars and lexicons participate in it at the same time is often seen as a threat, a disruption, and a malady.”(Muysken, 2000: 1)

122

This first tendency definitely considers code switching from a monolingual perspective. The second tendency in code switching research clearly departs itself from the first one in the sense that it sees code switching from a bilingual or a multilingual perspective. A big number of investigations of code switching could for example demonstrate that bilingual speakers produce code switched utterances much more often than one may think. Code switched varieties may even be the unmarked choice in some speech communities within some social occasions (Myers Scotton, 1993b: 45) Before We engage in the grammatical approaches to code switching we felt the need to discuss some issues related to the present study

2.2 methodological considerations of the study

Any research works needs to have a methodological procedure. This procedure needs to be clarified to the reader ahead, for this warrantees a good visibility of the work, and a better understanding of the insights of the investigation.

2.2.1 The Informants of the Study and data collection

The data of this study is based on a 12 hours recorded corpus. This corpus has been recorded in five sociolinguistic environments:

1. a university class(with female Kabyle students) 2. an apartment(with a Mzabi family) 3. an insurance company(with Kabyle workers) 4. a wholesale hardware shop in Oran(with Mzabi workers and customers) 5. a detached house in the neighbourhood of Oran(with a Kabyle family

The informants of the study are divided as follows

123

1. 10 female students studying at the English department of Es Senia university. Both of these female informants are from a Kabyle origin. Three of them are from Tizi Ouzou; the remaining seven girls are from Bejaia. 2. A family of four from Bejaia. A father aged 37, his wife aged 29, and their two daughters aged 5.5 and 9months. 3. Seven workers in a Mzabi owned whole sale hardware shop. These informants are aged between 22 and 45. 4. A Mzabi family living in Oran. The family is composed of 05 members. The father aged 41, the mother aged 30, a son aged 16, and two daughters aged 14 and 7 5. Three workers in an insurance company. Both are aged between 25 and 42.

The recordings were made by one of the informants from each group. The other members were not aware that they were being recorded. Though this procedure of data gathering is ethically controversial, we resorted to it to obtain spontaneous and natural speech. The transcription of the data was done with the help of the informants that made the recordings. This was done to be as faithful as possible to the original contexts in which the recordings were done. Another reason why we asked for the help of the informants is that the investigator is a non native speaker of Berber. Consequently, our competence in the varieties83 of Berber that we have investigated is far from being good. The informants that made the recordings have got a native control of the above cited varieties.

The second phase after transcription was the morphosyntactic analysis phase. During this phase we first arranged the data into chunks of discourse. During the second phase in the analysis, we segmented the conversations into turn takings and then we took each turn and segmented it into sentences, clauses, and phrases, the language index of the different parts of the data was systematically signalled to detect the instances of code switching84. This phase

83 We have investigated two varieties of Berber in this study. These include Kabyle Berber with its two varieties Tizi Ouzou Kabyle and Bejaia Kabyle and Mzabi

84Code switching is the main focus of this study. The data of this includes other contact linguistic phenomena such as interference and accommodation, but these phenomena do not fall within the scope of the present study. We believe that covering all the contact phenomena in a single research investigation is not possible as doing this would not give justice to these processes. We thus decided to concentrate on code switching alone to try be as exhaustive as possible with its investigation.

124 has been the most problematic in the study. The reason behind that relates to the fact that we are doing an investigation of spoken discourse. The challenge with such type of discourse is that it does not always contain fully fledged elements such as sentences, clauses, or phrases. As a consequence to that, morphosyntactic analysis of such type of language becomes problematic. The fact that the data we have investigated is a bilingual85 one adds to the complexity of the task.

Some instances of the data are repeated throughout the investigation. This was done in order to maximize our interpretation of the patterns of code switching displayed by the informants. Some examples may be interpreted differently from different perspectives and from different angles. The practice of exploiting data several times is quite common in contact linguistics research in general and in code switching research in particular. We haven’t numbered the examples for the above cited reasons.

2.2.2 Transcription of the Data in the Study

The present work being couched in a contact linguistics perspective, we felt the need to use investigations drawn from bilingual linguistic environments different from the one that we have studied. Many recent contact linguistics studies in general and code switching studies (Boumans, 1998; Muysken, 2000; Myers Scotton, 2002; Mac Swan, 2005; Gardner Chloros, 2009) in particular make use of such procedure to illustrate their arguments and to strengthen their positions. We will follow the same procedure in the present work as we think that cross linguistic comparisons is at the essence of contact linguistic research.

The instances of code switching drawn from other studies will be transcribed in their original form. The languages involved in these instances of code switching will be systematically signalled. The general tendency in code switching research that we have

85 In fact the data that has been analysed is not only bilingual but trilingual as for the Berber-Algerian Arabic-French code switching data. In some instances the data is even quadrilingual (Berber-Algerian Arabic-Standard Arabic-French. WE thus decided to use the term bilingual as a blanket term that covers both types of code switching. 125 observed throughout our readings is to separate code switched elements from the other elements. This is generally done through three different procedures:

1. the code switched parts may be transcribed in bold 2. the code switched parts be transcribed in italics 3. the code switched elements may be underlined

As we have been working mainly on spoken discourse, the instances of code switching that we investigated are mainly spoken. The present data involves Berber, Algerian Arabic, Standard Arabic, and French. We, thus, decided to use the SIL Manuscript phonetic alphabet to transcribe parts of the data that are from Berber, Algerian Arabic, and standard Arabic. The elements of the data that are French are not transcribed. They are spelled instead. The transcription procedures that we have adopted along the investigation are the following ones:

1. Berber elements are transcribed in normal fonts. 2. Algerian Arabic elements are transcribed in Underlined fonts. 3. Standard Arabic Elements are transcribed in bold underlined fonts. 4. French elements are not transcribed, but are spelled in bold letters.

2.2.3 The approach advocated in the Study

Contact linguistics research resembles research in the other domains of sociolinguistics in terms of methodological considerations. The problematic issues applying in sociolinguistic research in general apply to contact linguistics as well. One of these challenging issues is the qualitative/quantitative debate.

Some scholars working within the quantitative research tradition investigate contact linguistics phenomena such as code switching from a quantitative perspective. They thus describe patterns of language contact with statistical methods. Their perspective relies on 126 quantification of data. Findings in the qualitative approach are quantified and objectivity is taken for granted, thus leaving very little freedom to the investigator to interpret the data. The only valid interpretations are drawn from statistics.

There have been groups of researchers (Poplack, 1981, 1990, Polack and Meechan, 1998, Owens, 2004) who advocate that this approach is the best to the interpretation of contact phenomena in general and code switching data in particular. They use rigorous statistical methods in the interpretation of bilingual data. These methods draw from statistical procedures such as the one used by computer software e.g. the CLAN software and the Combo statistical software86. The challenge of the qualitative approach to code switching relates to the very definition of code switching. Some researchers consider this phenomenon as a strategy used by individual speakers. Drawing statistical assumptions from the speech of individual speakers seems a complicated task.

Scholars working within the second approach i.e. the qualitative one do not rely on statistical and quantitative methods. They rather rely on individual speakers, or on small scale micro sociolinguistic environments such as interactions conversations. This research perspective gives them more freedom in the analysis and the interpretation of sociolinguistic phenomena. The qualitative research perspective has been employed in contact linguistic research. It has proven its efficiency in inducing generalisations from studies focussing on environments that are as small as interactions between individuals. Modern studies on code switching focus on the speech patterns of individual speakers. Others focus on conversations between two to three speakers maximum. Many studies having such qualitative procedures have been quite successful. The findings of these qualitative studies have helped tremendously in the advances made in code switching and contact linguistics research.

The use of an integrative research methodology which integrates both perspectives strengthens the insights and the final product of sociolinguistic investigations. Many studies in

86 These softwares have been devised to help researchers working with spoken monolingual and bilingual discourse to transcribe and code data. They are also used to do statistical investigations on these data. 127 code switching research do in fact integrate both perspectives in a single research work. This is, however, not an easy task.

As far the present study is concerned, we shall be using a qualitative approach. We will rely on the speech of individual speakers for our analysis of the grammatical patterns of code switching under light in the study. In addition to our use of the qualitative perspective, we shall be using a quantitative approach when the data requires that. This approach has been advocated by renowned scholars in the area of sociolinguistics research such as Milroy and Li Wei (1995), Muysken (2000), Myers Scotton (2002), and Clyne (2003). We shall be using the quantitative perspective in the data when we will observe patterns of code switching that are recurrent in most recordings of the data. By doing so, we shall be trying to draw interpretation from the recurrence of these features.

2.3 Grammatical Approaches to Code Switching

Scholars’ interest in the grammar of code switching is not recent. This interest has witnessed different periods and tendencies. These tendencies reflect the general tendencies in language study in general and the general context in which the scholarly studies have been done. There are certain exceptions in which scholars prove to be ahead of their context and ahead of the general tendency. An example of scholars having a vision different from the one of his contemporaries may be Clyne (1967, 1973, and 1992) who has investigated bilingualism in Australia. Clyne investigated the German, Dutch and Swedish minorities of Australia. The approach that he advocated is quite different from the ones of scholars in his era. Bu for the present time let us briefly visit the different approaches to the grammar of code switching with a critical eye. The procedure that we will follow in our discussion of these approaches is a chronological one.

128

2.3.1 Early interests in the Grammar of Code Switching

Early studies on contact linguistics tried to depict Code switching as a deviant behaviour and as an ungrammatical use of the rules of the languages in contact in a bilingual speaker. Labov (1971:457), one of the most important figures in sociolinguistic studies states that:

“No one has been able to show that such rapid alternation is governed by any systematic rules or constraints and we must therefore describe it as the irregular mixture of two distinct systems”(Labov, 1971: 457)

This view is currently still advocated by many renowned scholars such as Chomsky who overtly excludes bilingual code switching from his I language87.He states that:

“The language of the hypothesized speech community, apart from being uniform, is taken to be a pure instance of UG (Universal Grammar) in a sense that must be made precise, and to which we will return. We exclude for example, a speech community of uniform speakers, each of whom speaks a mixture of Russian and French (say an idealized of the nineteenth century Russian aristocracy). The language of such a speech community would not be pure in the relevant sense, because it would not represent a single set of choices among the options permitted by UG but rather would include contradictory choices for certain of these options”( Chomsky, 1986b:17)88

Many counter arguments have been put forward by Muysken (2000: 43) to question the very notion of E languages as clearly distinct constructs (fortresses) in the sense that these

87 In his I versus E language approach ,Chomsky defines the E language as the „totality of utterances that can be made in a speech community‟ whereas he defines the I language as „ some elements of the mind of the person who knows the language acquired by the speaker and used by the speaker- hearer‟( Chomsky 1986b:22)

88 Emphasis is mine 129 are social constructs rather than being linguistic ones; besides, the perfect match between the I language and the E language is sometimes risky and problematic. Muysken (ibid: 45) argues in this vein that “the fortress may be built on quicksand”. He (ibid: 43) mentioned the example of the Balkan languages( Serbian, Bosnian, and Croatian) which are considered as separate languages by their speakers but which in reality share most of their grammar, morphology, and vocabulary.

2.3.2 Modern approaches to the Grammar of Code Switching

Code switching research has made a lot of progress since then. Many researchers showed that Code switching is possible at specific points within the sentence89 and is not at other points. Contact linguistics researchers in general, and code switching researchers in particular tried to devise constraints to this phenomenon. Some others tried to describe the grammatical properties of code switching. The first proposals started to appear in the 70ths, but the problem with findings is that they are not uniform. This means that the different proposals are controversial in the sense, that as long as a constraint appears, counter examples abound to infirm it. Three schools emerged in the study of the grammar of code switching. The three of them tried to predict possible sites for code switching (Backus, 1996:61).

The first school is led by Shana Poplack, and David Sankoff (1980, 1988, 1998), two leading figure in code switching research. The most important premise of this trend in code switching research is that there should be some kind of equivalence between the languages involved in code switching. This equivalence does not have to be violated at code switching points. Equivalence has to be present at the surface morpho-syntactic level to allow for code switching. Counter examples to this assumption abound from many studies especially those involving typologically remote languages. We shall discuss the two main constraints put forward by Poplack and her associates in the 80ths i.e. the free morpheme and the equivalence constraint in subsequent sections.

89 Only Intrasentential code switching is concerned with the relevance of grammatical (morphosyntactic) constraints. In intersentential code switching, the grammar of only one of the two languages is operative. 130

The abundance of counter examples to these two constraints led a second group of scholars to formulate alternative constraints. Their main principle is that code switching as monolingual speech is guided by universal generative rules such as government and binding rules as these have been advocated by Chomsky in (1983). This trend of thought is represented by Di Sciullo, Muysken, and Singh (1986). The weaknesses of this approach are its lack of consideration for the social context in which code switching takes place, in addition to the many counter examples put forward by other researchers.

This led a third group to formulate constraints that are less universal. These constraints predict that code switching does have such and such characteristics in such and such social contexts. This approach truly departs itself from the two preceding trends of thought in the sense that it correlates linguistic constraints with the social environment in which code switching takes place. Myers scotton (1993, 1998, 2000, and 2002) is considered as the spokeswoman of this third trend of thought. The three schools of thought are said to be complementary in their approaches and in their findings. Each one of them concentrates on a given aspect of code switching. Code switching is especially recurrent in certain points in utterances (the first trend represented by Poplack and her associates). It does follow universal rules at other points in the utterances (the view of the second trend represented by Muysken and his associates). The structure of code switching does finally depend on the specific social environment (Backus 1996: 64).

2.3.2.1 Distinguishing Code Switching from Borrowing

Before we engage in the analysis of the data in the light of three perspectives presented above, we felt the need to discuss briefly the distinction between code switching and borrowing. Distinguishing these related concepts is considered as a central issue in most investigations of contact linguistic phenomena. Myers Scotton (2002: 153) states in this vein that: 131

“A discussion that will not go away is the issue of distinguishing borrowing from code switching” (Myers Scotton: 2002: 153)

This task seems to be a fairly easy one, but it has proven to be a highly controversial issue; besides, Even if contact linguistics has evolved a lot since its beginnings, the issue of distinguishing code switching and borrowing hasn’t moved a lot forward. As recently as a year ago, Gardner Chloros (2009: 10) wrote that:

“Eastman90(1992: 1) wrote in the introduction to a volume on code switching “Efforts to distinguish code switching, code mixing are doomed” little has occurred since then to lighten this pessimistic view: terminology has been endlessly discussed in the code switching literature without any real commonality of practice being achieved”(Gardner Chloros, 2009: 10)

The distinction between borrowing and code switching has divided the community of contact linguists into two distinct groups. The first group considers that the demarcation line between the two phenomena is a clear cut one whereas the second group considers that the line is very hazy. Some researchers within the second tendency go even further by considering that there is almost no difference between the two processes at the synchronic level (Myers Scotton, 1997). A third group considers that the difference between the two processes exists and that this difference may be exemplified in terms of a continuum. This continuum has as its extremes code switching and borrowing. There are other contact phenomena ranging between the two according to this perspective. This position is advocated by scholars such as Heath (1989), Haust (1995), and Gardner Chloros (2009: 31). The argument given by these scholars relates to processes whereby. Code switchings are the beginning of loans life. Intermediate stages include different integration processes that occur during different periods in the history of the loan. During the latest phases of integration the loan or borrowed item may look quite different from the original code switched form.

90 Cf Gardner Chloros(2009) 132

Some scholars (Poplack, 1988; Poplack and Meechan, 1995; Sankoff, 2001) still claim that there is a clear distinction between code switching and borrowing in relation to the length of the element being inserted in the host language. These scholars argue that singly occurring inserted items are clear cases of borrowing. This argument is, according to Poplack and her associates, even valid for items that haven’t yet been adapted to the host language. The argument that they put forward is that the conduct of singly occurring code switchings and the one of longer stretches is different in bilingual corpora.

The aim of our investigation being the study of code switching patterns, we will interrupt our discussion of the differences between the phenomena of code switching and borrowing at this stage. Presently we may assume that code switching and borrowing are structurally akin in the sense that they are treated similarly at the morpho-syntactic level91. Backus (1996:65) states in this vein that:

“We can give up the idea that there are constraints specific to code switching and try to constrain its structure in some other way. Universal linguistic principles would then apply to borrowing and to code switching alike. This path has been taken by most researchers, for example Myers Scotton (1993b), Johanson (1993), Di Sciullo, and Singh (1986), Pandit and Stenson (1990)” (Backus, 1996: 65)

We shall now deal with equivalence-based approach with a little critical eye and illustrative examples from the original data of this approach, and with examples from our own corpus. The equivalence based approach is essentially an alternational trend. The basic premise of the equivalence based approaches to code switching stipulates that the languages contributing to code switching are in a symmetrical relationship i.e. they contribute on an even scale, and there is no hierarchy between them. Proponents of this approach include, Poplack

91 We here follow the stand taken by many researchers such as Myers scotton(1993b), Backus(1996), Boumans(1998) who consider that borrowing and single word code switchings are structurally very similar 133 and her associates (1981, 1988, 1995, and 2000), Auer (1988, 1995, and 1998) and researchers working within this tradition.

2.3.2.2 The Equivalence Based Approaches

The equivalence based approaches to code switching are said to have as their pioneers the researchers who studied Spanish English code switching in the United States during the 70ths and the 80ths, but it is possible to go farther back in the history of code mixing or code switching research. In fact, the first mention to surface equivalence as a pre requisite for any type of code switching has been made by Meri Lehtinen in the 60ths in his study of Finnish- English code switching. Meri Lehtinen (1966:153) argues that:

“In order for any intra-sentence code switching to be possible at all, there must exist in the two languages some constructions which are in some sense similar, so that certain syntactic items from each language are equivalent to each other in specific ways. Further reflection supported by an examination of the corpus, shows that the similarities must exist in what is known as surface grammar of sentences” (Lehtinen, 1966; 153)

Lehtinen seem to be ahead of her era as she also investigated the importance of the close and open class items distinction and hierarchy in code switching, which is at the core of the insertional approach. Lehtinen (ibid: 177) goes on stating that:

“It would seem that switching at words which belong to a close class is not allowed by code switching rules except in cases where such a switch is forced by structural considerations” ( Lehtinen, 1966:177).

Code switching research has tremendously progressed since then, but many things written in the code switching grammar literature seems to be a comment on the two assumptions 134 made by Lehtinen. Since the 60ths up to now a considerable part of the research has been concentrating on the syntactic properties of code switching (Muysken, 2000:12). Areas that have been investigated include the possible places of code switching elements in a sentence, and the rules governing this phenomenon. This trend of research has undergone three phases:

 An early phase in which grammatical rules specific to given language pairs and to given grammatical elements and constructions in these language pairs have been investigated. The results reached during this phase were not claimed to have any universal validity.

 An intermediate phase in which universal constraints have been investigated. This phase started in the 80ths onward.

 The present phase which is characterized by the search for new perspectives. Research during present phase has been trying to find alternatives to the so called universal constraints which to their majority are debatable, and even controversial in some cases.

The quest for structural constraints to code switching was initiated with different studies on Spanish English code switching in the US. Gumperz and Hernandez-Chavez (1971), Redlinger(1976), Valdes- Fallis(1976), Timm(1978), Pfaff(1979), and MClure( 1981) all devoted much of their research to finding points where switching is possible. These studies culminated in Poplack (1980) two constraints on Spanish English code switching after the studies that she carried on the Puerto Rican community of the US. These two constraints are the free morpheme constraint and the equivalence constraint. Poplack and her associates claimed some kind of universal validity to these constraints. Let’s start first by the equivalence constraint.

2.3.2.2.1 The Equivalence Constraint

135

This constraint is directed towards the issue of word order surface equivalence. The main premises behind Poplack constraint are linear equivalence and grammatical sub- categorization equivalence. The former principle means that switching is only possible in points where there is no overlap between the surface structures of the languages involved in code switching, and the latter is related to the grammatical sub-categorization properties of these languages. Poplack and her associates (Poplack, 1980; 1981; Sankoff and Poplack, 1981) put forward the equivalence constraint stating that:

“Code switching will tend to occur at points in discourse where juxtaposition of L1 and L2 elements does not violate a syntactic rule of either language i.e. at points around which the surface structures of the two languages map onto each other. According to this simple constraint, a switch is inhibited from occurring within a constituent generated by a rule from one language which is not shared by another.” Poplack (1980:586)

This constraint seems to underlie an important principle which is the one of balance or symmetry between the languages involved in code switching. This principle is one of Poplack’s main lines of thought. She seems to stick to it even if she has been overtly criticized for not taking into consideration the fact that code switching is not always an alternation between languages but also an insertion of elements of a language into longer stretches of another language as she states that:

“Code Switching is the juxtaposition92 of sentences or sentence fragments, each of which is internally consistent with the morphological and syntactic (and optionally, phonological) rules of the language of its provenance” (Poplack, 1993:85):

92 The Emphasis is from Shana Poplack. By putting emphasis on juxtaposition, Poplack wants to state that code switching is most of the times alternational and not insertional. This is true for contact situations which are characterized by balanced bilingualism such as the Spanish communities that have been the main focus of Poplack and her associates, but the word juxtaposition seems not to be valid in situations which are characterized by an asymmetry between the varieties involved in code switching. These include for example situations such as the one of ancient colonies where colonisers‟ languages (such as French in Algeria) are still in use in addition to local varieties.

136

Here is an example which has been put forward by Poplack to illustrate this juxtaposition:

(A) Se me hace que (B) I have to respect her (C) porque ‘ta …Older’ (It appears to me that I have to respect her because she is older) (Spanish-English code switching)(Poplack, 1993:85)

Fragment (B in Spanish) is a complement to (A) and (C) modifies (B). Data such as these can not be dealt with using an insertional approach in the sense that they are made up of long fragments from different languages or language varieties. Below is another example taken from our corpus which may illustrate this alternational state of affairs.

(A)(B)     ---(C) ..   (D) - 

(Lit: May be I love one Arab he is good better than the Kabyle but God is the strongest, the parents do not want)

(May be I fall in love with an Arab who may be better than a Kabyle but unfortunately my parents do not accept.)

The first part (A) is a discourse marker from Arabic. It is followed by (B) which is an independent clause from Berber. This independent clause is followed by a coordinating conjunction from Arabic (but) and(C) an independent clause from Arabic which is itself followed by (D) a verbal phrase from Berber. This example is a clear case of alternation as it involves the juxtaposition of elements as large as independent clauses in code switching.

Another characteristic of the “Equivalence Constraint” is that it prevents or at least greatly inhibits grammatically incorrect utterances because of language differences in word order (e.g., adjective position with respect to their head nouns). The code switching may also

137 not be possible in points where there is an overlap in lexical sub categorization between the languages involved. The position Poplack and Sankoff (1981:6) take is that in addition to the grammars of the two varieties involved there is an independent code switching grammar (code switching style) which draws from the two monolingual grammars. They (ibid: 11) state that:

“Ability to code switches results from Knowledge of the rules of grammars, their similarities, and their differences.”(Poplack and Sankoff, 1981, 6)

The evidence Poplack and Sankoff (1981:8) put forward is that switching involves no hesitations, pauses, or corrections, or other interruptions or disruptions in the rhythm of speech. Their positions opposes to the idea that in code switching only one language is on during the process. This language acts as the base or matrix language of code switching utterances (Winford, 2003: 128). They(ibid: 11) give the example of Spanish- English multiple switched sentences in which it would be very difficult to state which language is the base language. Here is the example they give (Spanish in bold):

There was a guy, you know, que he se monto he started playing with congas, you know, and se monto, y emprezo a brincar

(Lit: There was a guy, you know, that he got up, he started playing with congas, you know, and got up and started to jump). (Sankoff and Poplack, 1981: 11)

Muysken (2000: 45) calls this type of frequent code switching “congruent lexicalisation”. He defines (ibid: 45) it as:

“A situation where the two languages share a grammatical structure which can be filled lexically with elements from either language…The rules used to construct code switching

138 utterances may be drawn at times from one language and at times from the other.” (Muysken 2000:45)

2.3.2.2.1.1 Criticisms and counter Evidences to the Equivalence Constraint Drawn from Other Bilingual Corpora

The equivalence constraint received much attention and criticism from the community of code switching researchers. It has first been criticized on the ground of its lack in methodological rigour. This is related to the fact that the claims put forward by Poplack (1981) and her associates are not always based on large quantitative investigations. These claims seem to be based on judgement of acceptability by native speakers rather than being built on naturally occurring data.

The second criticism made to the methodological assumptions behind the equivalence constraint is related to the very notion of equivalence which seems to be problematic in the sense that the syntactic categories are sometimes different from a language to another (gardener Chloros, 2009:96). These categories may for example exist in a language and not exist in another one. In the same vein, Muysken (1995:193) argues that:

“There is assumed to be a match between both the terminal nodes in the syntactic tree of the languages involved in the switch. This idealisation is unwarranted; in fact there is no exact match between categories in different languages. Well documented problem areas in categorical equivalence include clitic versus non clitic pronouns, types of determiners and demonstratives, and types of auxiliary elements” (Muysken, 1995:193)93

We may cite the example of the category of definite article which exists in Arabic and French and which does not exist in Berber. This creates a mismatch in code switching

93 In L. Milroy, & P. Muysken ed (1995:193) 139 instances. Here is an example of a French noun, which lacks a definite article, and which is inserted in a Mzabi sentence:

// stilo /  /

(Lit: give me pen writes glass)(Give me the pen which writes on glass)(French – Mzabi Berber- Algerian Arabic code switching)

In this example the French code switch “Stilo” (pen) and the Algerian Arabic one //(glass) have got no definite articles which represents a clear violation of the French and Algerian Arabic syntactic systems; however, this is not a violation of the Mzabi syntactic system in which nouns do normally not subcategorize for any definite article. We will deal in more details with the issue of

A third criticism has been made to the equivalence constraint on psycholinguistic, and / or sociolinguistic grounds in the sense that in some cases different codes are in fact varieties of the same language and they therefore are part of a continuum rather than being discrete and separate entities. Blom and Gumperz (1972) observed that the inhabitants of Hemnesberget thought that the difference between the standard Bokmal and the local dialect Ranamal is inherent to the linguistic structure of the two varieties while in reality the distinction is only a social one .i.e. it is related to the social attitudes and meanings attributed to each variety by the community. Blom and Gumperz (1972:78) accordingly found that while the speakers of this community thought that they were alternating codes, they were moving up and down in a continuum of the same language (Downes 1998:91). Sebba94 (1998:272) goes even further in the same line of thought. He states that codes in code switching are not made up of inherent differences between languages, but are rather “constructed”95 by individuals on the basis of their value in conversation.

94 In P.Auer Ed (1998: 268). 95 The emphasis is Sebba‟s. 140

From the structural point of view, a number of counter examples from diverse bilingual situations have been put forward. Nartier (1982) provides counter examples from a data of Adangme96/English switched code where the word order equivalence is violated. Adangme word order is (SOV) with Head first NP i.e. Noun+det constructions, Whereas English word order is (SVO) with head last NP. Here is one of the counter examples given by Nartier:

/    help-/

3 Pl cop me help pres prog

(Lit: They are me helping) (Adangme/ English; Nartier, 1982: 185)

In this example the non equivalence is in the position of the object and the verb i.e. in English the verb precedes its object and in Adangme the verb follows its object. The English code switch help is preceded by an Adangme object pronoun //.

Bentahila and Davies (1983: 119) provide many counter examples to the morpho syntactic sub-categorization equivalence constraint from their Moroccan- French code switching data. They talk about the mismatch in the position of adjectives in noun phrases. In Arabic, nouns generally precede their modifying adjectives whereas in French they may follow it, or precede it. If we follow the equivalence constraint, we would say that the only possible occurrence of a code switch is when there is equivalence between the two languages i.e. cases in which the noun precedes the adjective and not ones in which the French adjective precedes its head noun. In spite of this prediction Bentahila and Davies cite examples of the occurrence of a French adjective preceding an Arabic noun. Here is the example they give:

J’ai vu un ancien / -/.

(I have seen an old student of mine).

96 C.f. Myers Scotton (1993b:28): Adangm is a Kwa language spoken in southeastern Ghana. 141

The counter examples put forward by Bentahila and Davies have been seriously put into question by Bouamrane (1988: 148) on the basis that these seem not to hold for Algerian Arabic- French code switching97. The counter examples given by Bentahila and Davies have also been seriously challenged by Benali Mohamed (2007: 86) on the basis of his data of Berber- French code switching. Benali Mohamed (ibid: 86), however, claims that the equivalence constraint seems to hold water for his Berber-French code switched data98. We shall come back to this issue after looking at the other counter examples to the equivalence constraint given from other contact linguistics studies.

Berk Selingson (1986: 132) presents a number of violations of the equivalence constraint in her Spanish- Hebrew bilingual corpus in Israel99. She argues that when languages are typologically remote, it is difficult to keep the category of equivalence. Berk Selingson (1986: 332) gives the example of the omission of the indefinite article in Spanish switches. She argues that one possible interpretation of this violation of the Spanish grammatical system is the inexistence of this category in Hebrew.

The same has been noticed in our data where there have been violations to the French Grammatical system concerning definiteness (the use of a definite article) in many Berber- French code switches in addition to violations to the Berber syntax in situations where French and Arabic code switchings in Berber are used with their articles. We made the hypothesis that, in the case of the violations to the French grammar, this may be due to the fact that Berber has no category of definite articles. This inhibits its use with French and Arabic code switchings100. The second hypothesis we have made is that the use of definite articles with French and Arabic code switched nouns may be due to the long lasting contact that there has

97Algerian Arabic, being very similar to Moroccan Arabic, should function similarly to this variety in the contact of French in code switching situations.

98 We will try to put into question this assumption as we have found that there are counter evidence to the equivalence constraint in our Berber-French and Berber-Algerian Arabic code switching data.

99 Berk Selingson presented a corpus of both Spanish emigrants to Israel and Local inhabitants born in Israel. 100 Definite articles do normally not exist in Berber while they are attested in both dialectal and standard varieties of Arabic in addition to being used in French. 142 been between Berber and these two varieties; thus creating long term profound influences in the Berber syntactic system. We found examples of French code switchings in Berber where the code switch keeps its Definite article, besides French and Arabic code switchings in Berber where the definite Articles are omitted. These two situations are clear violations to the equivalence constraint. Here are some examples taken from our corpus:

// riveteuse? Inter 1st plu verb (Mzabi) noun (French) (Don’t we have blind rivet pliers?)

This sentence contains a Mzabi- French intrasentential code switch (riveteuse) which has no definite article. This is a clear overlapping between French and Mzabi in the sense that Mzabi does not have any definite article while French nouns subcategorize for Definite and indefinite articles. Some scholars ( Boumans, 2004: 45; Myers scotton, 1993b:190; Owens, 2006:26) argue that these may be considered as bare forms in code switching whereby code switched items are used without grammatical markers from any of the languages involved in code switching. These grammatical markers may be signalling aspect, tense, gender, number, definiteness, and case. In a study of Nigerian Arabic-English code switching in Nigeria, Owens (2006:781) cites examples of English bare forms in otherwise Nigerian Arabic strings. Here is an example of a bare inserted English noun phrase in a Nigerian Arabic sentence. English inserted noun phrase is in Bold:

/ / Ø-kampeen masta // (Lit: He just campaign manager here).

Romaine (1989: 124) provides other counter examples to the equivalence constraint from her Panjabi- English code switching data. She gives the example of the mismatch in the position of preposition/postposition in English and Panjabi i.e. Prepositions are generally put before their noun complements in English (what she calls right branching position), whereas they are used as postpositions (after their noun complements) in Panjabi (what Romaine calls left branching position). Romaine (2005:42) argues that in spite of this overlapping switching 143 can occur in a postpositional phrase (between a postposition and its object noun). She gives examples such as (Panjabi in Bold):

Parents  depend 

Parents post depend be aux

(It depends on the parents)(Punjabi-English code switching)(Romaine 2005: 45)

Similarly Pandit (1990:45) provides examples, from his English- Hindi code switching data, of entire Hindi postpositional phrase switched for an English prepositional phrase (Hindi in Bold):

John gave a book  

a girl to

(John gave a book to a girl)

Myers Scotton (1993b: 35) provides counter examples from her Swahili-English code switching data, among which the mismatch in the order of noun adjective within NPs. Below is an example she gives in which the English code switched noun and adjective follow an otherwise Swahili word order:

Ni-ka-i taste- ni-ka-ona i-na taste lousy sana

1 sing-consec-obj-cl.9 1 sing-consec-percieve it-with taste lousy very

(And I thought it had or was with a very lousy taste)(Swahili- English code switching) (Myers Scotton 1993b:29)

144

2.3.2.2.1.2 Counter Evidences to the Equivalence Constraint from the Corpus of the Study

In the data under investigation in this study, we found a number of counter examples to the equivalence constraint. Some of them concern the issue of syntactic sub-categorization of certain syntactic categories such as nouns, adjectives, and prepositions. Another mismatch may reside in the gender assignation attributed to the verbs, nouns and adjectives in the languages involved in Code switching. A third mismatch may be in the neutralization of gender and number in the categories listed above. We will be dealing with the cases of mismatch that seem worth-analysing to us, as dealing with all the cases of mismatch between the languages involved in code switching in our study falls out of a the humble research work that we are conducting.

2.3.2.2.1.2.1 Mismatch in the positions of adjectives in relation to their head nouns (Ordinary adjectives, possessive adjectives, demonstrative adjectives)

We may for example mention the case of the overlapping between adjectives and their head nouns in NPs. Berber, being a member of the Chamito Semitic language family, is a language in which adjectives generally follow their head nouns. In Kabyle and Mzabi, the two Berber varieties under study here, the normal order is NPN+ADJ whereas this order in French would be NPADJ+N. The equivalence constraint would predict that there would be no switch between these categories; in spite of this prediction we found many counter examples such as the ones listed below:

/-/ les chanteurs /--  Kusayla/ (Me like-1st sing accompl aff the singers nom aff-new-Plu aff like Kusayla) 145

(Lit: Me I like the singers new like Kusayla) (I like new singers like Kusayla)(Berber-Algerian Arabic-French code switching)

/.. Différence  --/ (But Difference not fem aff-big-fem aff) (Lit: But(the difference not big)(but not a big difference)(Berber-Algerian Arabic- French code switching)

The first example contains ‘les chanteurs’/-- / Kusayla)(the singers new like Kusayla), an object NP, the Object NP is made up of a French code switched noun (Les chanteurs)(singers) followed by its Berber modifying adjective(/101/)( new). The mismatch is in the position of the adjective which does not follow French syntactical paradigm but a Berber one.

The second example displays a similar pattern of violation to the predictions of the equivalence constraint. (Différence’  --/)(Lit: difference not big) is also an NP made up of ‘Différence’(difference), a French code switched noun that is modified by /--/(new) a Berber adjective.

The mismatch between French nouns and their possessive pronouns or adjectives may be used to further illustrate Noun/adjective mismatch in Berber-French code switching. French nouns normally subcategorize for pre-positional possessive pronouns. E.g. mes etudes(my studies), tes amies(your friends) and not *amies tes(*friends your).

Berber is a language in which Nouns subcategorize for Post-positional possessive pronouns. The equivalence constraint would predict that there should be no switching between Berber and French for Nouns and their possessive pronouns. Poplack et al’s(1981)

101 This adjective is in fact made up of // (new) an adjectival stem from Algerian Arabic, but the adjectival inflections are from Berber /…/. We considered this adjective as a Berber one, for the aim of this subsection is not related to the analysis of the internal make up of words. Recall that the aim of this subsection is to investigate cases of mismatch between French and Berber in relation to nouns and their modifying adjectives. 146 predictions seem to be unpractical to the data that we have investigated. Below are some instances in which a French noun is combined with a Berber possessive pronoun.

/-- - --/ les vacances //

(Inaccompl aff-go- 1st sing aff to Oran inaccompl aff-finish-1st sing aff the holidays-1st sing posse pron)

( Lit: I go to Oran I finish the holidays my)

(I will go to Oran to finish my holidays (there))(Berber-Algerian Arabic-French code switching)

/  --/ la famille / - ./

(So You 2nd sing accompl aff-have-2nd sing aff the family 2nd sing posse pron in Setif)

(Lit: So you you have(members of)the family your in Setif)

(So you have members of your family in Setif)(Berber-Algerian Arabic-French code switching)

/- -- -les proches /

(Like-1st sing Inaccompl aff-go-1st sing aff to-the near our/

(Lit: I like I go to the family our)

(I like to go(visit members of) our family)(Berber-Algerian Arabic-French code switching)

The mismatch in these three instances of code switching is related to the position of the possessive pronouns // (my), // (your), and// (our) in relation to their

147

French code switched head nouns that are respectively ‘les vacances’ (the holidays), ‘la famille’(the family), and ‘les proches’(the family).

The position of the demonstratives (this –that) and their head nouns may also be mentioned as a case of mismatch between French and Berber. Berber is a left parsing language (the modifiers follow their head nouns in noun Phrases) Whereas French is a right parsing language in which the modifiers precede their head nouns in noun phrases. The equivalence constraint would therefore prevent any switching between a head noun from French and a Berber demonstrative (in a right position). Instead of that, instances of French head nouns followed by a Berber demonstrative abound. Here are some examples:

1 A: / /

(Who is this?)

2 B: Bébé -Amira -

(Lit: baby Nom aff-Amira fem aff-this)

(This is Amira’s baby.)

3 Mais l’année /  ..- / (French- Berber Code switching).

( Lit: but the year this neg picture 1st sing aff- neg)

( But this year I did not take any picture)(Berber-Algerian Arabic-French code switching)

In this example the French code switch ‘bébé’( baby) in line 2 is followed by the Berber demonstrative Pronoun // (this). This represents a clear violation to the equivalence constraint in the sense that it (the constraint) would normally inhibit such instances of switching from occurring. The French code switchings in line 3 are the coordinator ‘Mais’(

148 but) and the noun ‘l’année’( the year ). This noun is followed by its demonstrative Pronoun // from Kabyle. This is again a clear case of mismatch between French and Berber for this noun only subcategorizes for demonstratives in a left branching position e.g.

Cette Année, nous avons eu un nombre élevé d’étudiants. (This year, we had a big number of students)

 Année cette, nous avons eu un nombre élevé d’étudiants. (Lit : Year this, we had a big number of students).

Below is another instance of French noun-Berber demonstrative code switching

/ --/ tricot //

(Where buy-2nd sing accompl aff top demonst pron)

(Lit: Where did you buy top this)(Mzabi Berber-French code switching)

Another case of structural non-equivalence relates to the use of a French code switched prepositional phrase or clause with a Berber preposition.

2.3.2.2.1.2.2 Mismatch in patterns of Berber-Algerian Arabic-French switching involving prepositional constructions

// la différence // tu me dis le protocole TCP et IP

 (Lit: Tell me the differences between you tell me…)  Tell me what is the difference between the two protocols TCP and IP( Kabyle Berber- French Code switching)

149

The two French code switchings are indicated in bold. The complement phrase (// tu me dis le protocole TCP et IP) is made up of a kabyle complementizer //( between) which is followed by a French complement phrase tu me dis le protocole TCP et IP( you tell me the protocol TCP and IP). The structural non equivalence is in the object of the preposition which should be a noun phrase or a noun clause starting with a subordinating conjunction in French, but which functions differently in Berber. Here are two illustrative examples from French

 e.g. quelle est la difference entre une femme et un homme, (what is the difference between a woman and a man)  e.g. quelle est la difference entre ce que tu me dis et ce que je viens de te dire (what is the difference between what you are telling me and what I 've just told you).

When functioning as a complement of the object, prepositional phrases, in Berber, are made up of a preposition which is followed by a verbal phrase, without a subordinating conjunction (the case in French grammatical structure, cf. Naït Zerrad, 1997: 119). The equivalence constraint would inhibit switching in such contexts because of the overlap between the two systems. Instead the example given above has been attested in our corpus and proves not to be an exception.

We may also mention the example of the use in code switched sentences of a Berber preposition “Dal” (of) before a numeral determiner and a noun construction which is prohibited under the equivalence constraint. The reason for its prohibition is the fact that French nouns phrases do not subcategorize for any preposition in a premodification position after a premodifying numeral. Here is an example from French:

Est-ce que tu pourrais m’acheter trois DVD ?

(Would you be able to buy me three DVDs?)

* Est ce que tu pourrais m’acheter trois de DVD ?

(Would you buy me five of DVDs?)

150

Berber and Arabic on the other hand may subcategorize for such a preposition (// (of) in Berber and // (of) in Arabic) before a noun phrase when it functions as the object of the verb. Here is an example that we found in our corpus

/  / les croissants.

(Lit: give me five of the French croissants)102

(Kabyle Berber- French code switching)

In this example the French code switch “Les croissants” (The French croissants) has got two pre modifiers: (five) which is an Arabic numeral and the preposition (of) from Berber. This is a clear case of structural non equivalence. And it seems not to be an exception but a recurrent feature of French- Berber code switched noun phrases where the head noun is French and where the phrase functions as the object of the sentence.

2.3.2.2.1.2.3 Code switching patterns involving gender assignation mismatches

Gender assignation attributed to certain concepts in Arabic, French and Berber may also be mentioned as a further case of mismatch in Berber-Arabic, Berber- French code switching. The equivalence constraint would predict for example that a noun which has a feminine gender in one language and a masculine gender in another one would not be switched. It would also predict that when there is a mismatch in the noun-verb, noun-adjective agreement there would normally be switching. This proves not to be valid for our data, for we found many instances where there are nouns which sub categorize for a masculine adjective in Arabic, but which are used with a feminine adjective in Berber. An illustrative example would be:

102(Give me five French croissants) 151

/  .      /

(Lit: Perhaps I wanted I picture like this. This is during the feast fem affi- big- fem affi)

(May be, I wanted to take a picture like this. This was during the slaughter feast)

The mismatch in this example is in the prepositional phrase /  / (during the slaughter feast) which contains an Algerian Arabic masculine code switched noun // (feast) which subcategorizes for a masculine modifying adjective103. Instead of that the Berber adjective is marked by a feminine gender circular affix i.e. /…./. This is a clear case of mismatch in the sense that there is a structural non equivalence between Algerian Arabic and Berber in the concept of feast (feast being Feminine in Berber and Masculine in Algerian Arabic). In the example below the direct object of the verb is supposed to

There are other cases of mismatch in gender assignation for nouns where Arabic feminine nouns lose their feminine gender and become neutral. This neutralization strategy seems to be operative for both noun and verb categories this is again a clear violation of the equivalence constraint. Below are some instances of code switching showing such type of mismatch:

//( the school- 3rd sing null aff)

/-/( full- 2nd sing neutral aff)( you are full)

//( step out)

//?

103 Nouns and their modifying adjectives in Both Arabic and Berber agree on number and gender i.e. masculine nouns in these languages subcategorize for masculine adjectives( Adjectives having masculine markings), and feminine nouns subcategorize for feminine adjectives( Adjectives having affixes having feminine markings ) 152

(2nd sing accompl aff- upset- 2nd accompl aff)

( were you upset)

The same form of the verb is used whether it is the second person masculine or feminine that is not the case in Arabic where this verb subcategorizes for feminine and masculine affixes in agreement with the subject. The opposite process has also been observed whereby. A noun subcategorizes for a neutral modifying adjective or possessive pronoun in Arabic. It is used with a feminine and masculine modifier in Berber. Here is an example:

Mother: /- -/ Manal?

( take- 2nd sing accompl verb aff medecine- 2ND fem poss clitic affi)

( did you take your medecine manal)

Manal: / /( Lit: over(finished) I did mum)

The arabic code switched noun (medecine) subcategorizes for a neutral gender possessive clitic postpositional pronoun in Algerian Arabic()(Lit: medecine-your( your medecine), but in this example it is used with a femnine gender possessive clitic postpositional pronoun. This is clear case of violation of the equivalence constraint.

2.3.2.1.1.2.4 Code switching patterns involving the sub-categorization of object accusative and dative clitic affixes

The position of the object dative clitic affix attached to the verb represents another case of mismatch between Berber and Algerian Arabic morpho-syntactic systems. Algerian Arabic Verbs subcategorize for clitic suffixes which are attached to it and which function as direct(accusative) or indirect object(dative) clitics; on the other hand verbs in Berber may

153 sub-categorize for either prefixes or suffixes functioning as object accusative, or dative clitic pronouns. Here are some examples from both Arabic and Berber.

--- -

( tell- 2nd sing fem accompl suf- 1st sing clitic object dative case suf love-1st sing possessive aff )

--

(give-(to)-1st sing accusative object suf)(give me!)

 ---

( in fact neg aff- 3rd sing clitic object accusative aff- know-2nd sing inaccompl aff- neg marker)

( In fact you do not know it)

The two examples from Berber highlight two positions of the clitic accusative Object prefix. The clitic takes a suffix position in the first example and it takes a prefix position in the second example.

The equivalence constraint would predict that switching of Algerian Arabic verbs would only be possible in cases where the object dative clitic pronoun follows its head verb. This is not always true in the sense that in some instances we found Algerian Arabic verbs which subcategorize for a prefix object clitic dative pronoun. Here are some examples:

 --- 

(like 1st sing aff- obj clitic accusative pref-work- 1st sing accompl Aff in the past)

154

( like( the same way as) I did her in the past)

/--- --  - - -./

( 2nd sing obj clitic accusative pref- 2nd sing aff- take-2nd sing you will pay( You will be beaten) now your sister is sleeping)

( If you take the radio I will beat you. Your sister is now sleeping)

/   -----/

( look look how 3rd plu aff(my legs)- 1st sing obj clitic accusative pref- 3rd plu fem inaccomp aff- hurt- 3rd plu fem inaccomp aff)

( look look how my legs are hurting me) (Kabyle Berber-Algerian Arabic switching)

/  ---     -/

( interj- here obj clitic accusative pref- 2nd sing fem inaccomp aff- do- 2nd sing fem inaccomp aff my hands come on mum press on them)

( interj here here put your hands here come on mum press on them)

/ . --  --/

( Lit: Yes ok obj clitic accusative aff- warm- 1st sing clitic object clitic accusative suff because obj clitic accusative aff- hurt- clitic object accusative aff)

( yes(mum) warm me( my legs) because they are hurting me)(Berber-Algerian Arabic code switching)

155

/ --- - /?

(What that-2nd sing obj clitic dative aff- hurry(verb)-3rd plu inaccompl aff like-that)

(Lit: What(who) is hurrying you like that?)( Kabyle Berber-Algerian Arabic code switching)

/-  ---/?

(And-you, who 2nd sing clitic object accusative aff- ask- 3rd plu accompl aff)

(And, who asked for you again?)(Kabyle Berber-Algerian Arabic code switching)

/-- ---/

(2nd sing fem accompl aff-like-2nd sing fem accompl aff 1st sing object accusative aff-talk)

(You want to make me talk again)(Kabyle Algerian Arabic Code switching)

/   ----/

(Look look how they-1st sing object accusative aff-3rd fem plu accompl aff-hurt-3rd plu accompl aff)

(Lit: look! Look! How they (my legs) me hurted) (Kabyle Berber-Algerian Arabic code switching)

In these examples Algerian Arabic code switched verbs are preceded by Berber clitic object accusative prefixes. The examples that we have listed are a sample of a recurrent process i.e. the affixation of a clitic accusative object prefix to a code switched verb from Algerian Arabic. This represents a clear violation to the equivalence constraint principle. The Berber clitic accusative object pronoun is not supposed to precede an Algerian Arabic verb since this represents a violation to the Algerian Arabic morpho-Syntactic system. These counter examples may be partly interpreted using the dominance configuration principle as

156

Myers Scotton (1993, 1997: 65) has put it forward. We shall come back to this very notion of dominance when we will be dealing with Myers Scotton (1993, 1997) model in next chapter.

2.3.2.2.1.3 Poplack’s Response to the Criticisms Put Forward in the Literature in Relation to the Equivalence constraint

Poplack and her associates (1990:74) try to overcome most of these criticisms by stating that most counter examples cited in the literature are in fact single word switchings. They call these single word switchings ‘nonce borrowings”104 which function differently from multiword longer switchings. They argue that the two types of switchings are governed by otherwise different constraints as Poplack (2006:158) puts it:

“Most single word switches are the result of a process which is fundamentally different from the process which produces multiword switches. In the former, borrowing, only the base language is spoken, while in the latter, Code switching, there is a real switch to the other language… unambiguous Code switching forms are multiword sequences which remain lexically, syntactically, and morphologically unadapted to the recipient language”

Poplack (1988: 99) argues that there are criteria to distinguish nonce borrowings, and code switchings. These criteria include roughly the degree of integration (phonological- morphological, and syntactic integration), the frequency of use in the receiving speech community and the acceptability by native speakers. We are not going to get into the details of the distinction between established loans, nonce borrowings, and code switchings as this is a

104 The concept of nonce borrowings has been described first by Haugen (1950:94), and it has been used recently by Poplack and Sankoff (1988b:97) to mean the bilinguals use of borrowed items which are still yet not recognized as borrowings by the receiving speech community. Nonce borrowings have been opposed by Poplack(ibid:99) and her associates to established loans in the sense that nonce borrowings are used only by bilingual speakers while established loans are used by monolingual speakers and bilingual ones alike. Poplack and her associates also define these nonce borrowings as a once occurrence borrowings(once life borrowings) 157 quite thorny issue which falls out of the scope of the present chapter. Poplack and her associates are not the only researchers who have taken such a stand. This stand seems to facilitate the work as it dismisses the most difficult cases of code switching i.e. the singly occurring lexemes that are often assimilated and integrated into the receiving language. Myers Scotton (1997:23) states that:

“In the 1970s and far into the 1980s, most researchers accepted the idea that only full constituents( or clauses or sentences) qualify as true Code switchings obviously, such structures represent the easiest ‘cases’ since they( largely avoid the issue of assimilation/integration.”(Myers Scotton, 1997: 23)

Myers Scotton (ibid: 23) goes on stating that the tendency in code switching research has changed since then:

“Researchers increasingly reject ‘nonce Borrowings’ as a resting place out of harm’s way for single lexemes which are not clearly established borrowings (Bokamba, 1988; Elliasson, 1989; 1991; Nortier, 1990; Scotton, 1988a; 1988b.”(Myers Scotton, 1993: 23)

In spite of all these criticisms Poplack (2000: 243) and her associates still make a case for Nonce borrowing). Other researchers (Swigart, 1992:42) tried to use a compromise designation which i.e. ‘Code mixing’ to cases of intrasentential code switching be they single morphemes or phrases or larger constituents within the sentence boundary.

Poplack’s (ibid: 243) claim that singly occurring code switchings have to be ruled out from code switching research has been strongly challenged by many researchers. Pat, Troike, and Mun (1989: 7) made the observation that 60% of the entire switches in their Korean- English code switching data are composed of single nouns. Berk Selingson (1986: 81) made a similar observation on her Hebrew Spanish code switching data. Nortier made the following comment on her corpus of Moroccan Arabic- Dutch code switching: 158

“By far the largest group of switches concern insertion single words in one language (usually Dutch) in sentences of the other language.”(Nortier, 1990:140)

We found similar tendencies in our code switching data involving Kabyle and Mzabi with more than 51.66% of singly occurring Algerian Arabic code switched Nouns, and 48.13% of singly occurring Algerian Arabic verbs. The tendencies highlighted for the two classes of Adjectives and Adverbs are lower than the ones of Nouns and verbs. We have decided to take the stand of researchers who believe in the principle that singly occurring code switched constituents are worth-investigating.

These items represent by far the largest percentages of code switched items in our data. Many researchers (Joshi, 1985: 190-191; Myers Scotton, 1998: 26; Muysken, 2000: 120) strongly believe that these elements highlight a certain type of asymmetry in code switching data in the sense that these single nouns, verbs, adjectives, or adverbs find themselves vulnerable in a sea of the other language items (Myers Scotton, 1998: 181).

Poplack and her associates have also been strongly criticized for not taking into consideration this notion of asymmetry between the languages participating in code switching. The notion of asymmetry in code switching stands like a competing vision to the equivalence vision of code switching and we will come back to it when we will deal with the insertional models of code switching as they have been expounded by Backus(2000), Boumans( 1998; 2004), Joshi(1985), Myers scotton(1993; 1998; 2000; 2002), and many other researchers.

In another trial to bypass the criticisms against their equivalence constraint Poplack and Sankoff (1988: 1175) developed methods to differentiate cases of code switching which obey the equivalence constraint and others which do not by stating that these two types of switching operate differently. These are divided into

159

1. Smooth code switching: This type of code switching obeys the equivalence constraint. The switching is made only in places where word order, or/and sub categorization prosperities of the languages involved are followed. It is characterized by a smooth flow of language without any pauses hesitations or interruption. Here is an example:

1) Mother: //?

(What is it(feminine)?)

2) Leila: /      / portable

(Lit: This fill- 1st sing aff with the speech of our not a Mobile phone)

(I fill it with our speech. It is not a mobile phone.)( kabyle- Algerian Arabic- French code switching)

The Algerian Arabic Code switch /  / (the speech of ours not) in line 2 takes place without any interruption, pauses or hesitations. It is embedded in an otherwise Kabyle string. The transition between the two languages seems to be seamless and smooth (Muysken, 2000:101).

2. Flagged Code Switching: Flagging has been described by Poplack as the insertion of a discourse marker, an interjection, or a tag to separate the two languages in switching. Compared to smooth code switching which is by definition smooth, flagged switching is not smooth in the sense that it is characterized by a brief interruption in speech, a pause, a hesitation, a tag, an interjection, a discourse marker, a coordinator, or a complementizer, to signal things such as a change in the language being used. Here is an example given by Treffer- Daller (1994: 204) from her French-Dutch code switching data in Brussels.

Tu sais l’affûteur de scies hein daan gink bij die beenouwers, de zager (French- Dutch Code switching) (Treffer- Daller, 1994 :204)

160

In this example the discourse marker ‘hein’ is used to separate the two languages( French and Dutch). Here is a an example showing a similar pattern from our corpus:

// enregistri/  ( 0.3) ..  ?/

(Lit: record- 2nd sing aff mother- Poss 1st sing but what fill 2nd sing aff?)

(Lit: You record mother my, but what do you fill (record about her)?)

(Are you going to record my mother? But what are you going to record?) (Kabyle- Algerian Arabic code switching)

In this example the Algerian Arabic Code Switch /../ (but) is used as a flagged code switch. The flow of speech is interrupted before this code switch and it signals a change in the language of the sentence. Below is another example:

// Déjà je ne suis pas contre, déjà /     déjà /    / (Kabyle- French- Algerian Arabic code switching)

( Lit: Me I like them, right, I am not against , right, me I study with them, right, me I look only the religion)

(I like them (the Arabs), right, I am not against (them), right, me I study with them, right, I only consider religion as a criteria)

In this example, the first code switch to French, the discourse marker ‘déjà’ (right) is used after a Kabyle string to signal a change in language in the second sentence. The second discourse marker ‘déjà’ (right) is again used to signal a change in language (Kabyle), and so is the third discourse marker which is used to signal a third change in language (Algerian Arabic).

161

Poplack (1980:198) also calls this type of code switching emblematic code switching or tag switching She argues that this type of switching is characteristic of alternation.

3. Constituent Insertion: As it is indicated by its name, constituent insertion is the insertion of an entire constituent such as a noun phrase, a verb phrase or a Prepositional phrase from the host language into the receiving language even if there already is a determiner from the receiving language right before the switch point). Here are some examples:

/-/ la Sixieme Meriem /-- -/

(Lit: she got the sixth Meriem inacomp-go- 1st plu to-her)

(She (Meriem) got the primary school certificate. We will go to visit her) (Berber- French- Algerian Arabic code switching)

/.  .  / la carte mémoire

(Lit: got-1st sing one (fem) got-1st sing them in the memory stick)

(I have got one (picture). I have got them in the (Photographer) memory stick)

In these examples the French code switches ‘la Sixieme’ and ‘la carte mémoire’( the memory stick) are constituent insertions which are made of a Noun and its definite article for the first code switch and a compound noun with its definite article for the second one. If we consider the equivalence constraint these code switches are violations to the syntactic system of Berber which normally has got no category of definiteness as it has been indicated above. Poplack considers that most of the counter examples to the equivalence constraint given in the literature are in fact constituent insertions (for cases of multiword insertions) or nonce borrowings (for cases of singly occurring insertions). We now turn to the free morpheme constraint or the intra-word constraint.

162

The big number of counter examples to the equivalence constraint led Poplack and Sankoff (1988: 90) to acknowledge that equivalence based code switching as exemplified by Spanish-English code switching in New York is an extreme case for it highlights a contact situation which is characterized by stable bilingualism, as well as an equal proficiency of the members of the community in the two languages. In a sharp criticism against Poplack, Backus (1996: 64) argues that:

“Chicanos and Puerto Ricans in New York soon turned out to represent a type of bilingualism which may not be the most typical around the world. Their bilingualism is fairly symmetrical, or balanced. In many emigration contexts, the situation is different, especially for the first generation of emigrants. Speakers are dominant in one of the languages, and their code switching is characterized by strong asymmetry” (Backus, 1996: 64)

Winford (2003: 132) also argues that the symmetrical bilingual situations prove to be an exception rather than a general tendency out of which one can make any generalisations. Most common contact situations are not as stable as the one in New York and are mainly characterized by instability as well as by a mismatch in the linguistic structure of the languages involved in code switching.

The general tendency in the code switching literature is that the more typologically remote the languages are the more problematic it becomes to keep the equivalence constraint operative. This tendency seems to hold true with our corpus. The equivalence constraint seems to operate somehow well with Algerian Arabic-Berber code switching as these two languages are to some extent typologically similar. Comparatively the constraint seems to display weaknesses in relation to Berber-French code switching. Similar tendencies have been found in Bouamrane (1988:140) data of Algerian Arabic- French code switching and in Benali (2008:88) data of Algerian Arabic-Kabyle code switching in Algeria. We will come back to these tendencies when we will be dealing with the analysis of some possible code switching occurrences from our personal corpus.

163

We shall now turn to the other constraint put forward by Poplack and her associates in the 1980ths. This constraint is the free morpheme constraint or the little sister of the equivalence constraint as some scholars (Romaine, 1995: 126; Myers Scotton, 1998: 27; Muysken, 2000: 46) like to call it.

2.3.2.3 The Free morpheme Constraint

Poplack (1980: 585) added an additional constraint, the free morpheme constraint, to the equivalence constraint. Compared to the equivalence constraint which operates at the sentence segmental level, this constraint is an intra-word constraint. It operates at the internal morphological level of words. The free morpheme constraint states that:

“Codes may be switched after any constituent in discourse provided that constituent is not a bound morpheme” (Poplack, 1980: 585).

Poplack (1981: 5) further elaborates her constraint stating that:

“Switching is prohibited between a lexical item and a bound form unless the latter has been phonologically integrated into the language of the bound form” (Poplack, 1981: 587)

The free morpheme constraint thus predicts that there will be for example no switching between a free morpheme and a bound morpheme (Backus, 1996: 62). In addition to that the constraint states that there will be no switching before and after a bound morpheme. Here are some examples given by Poplack (ibid: 590) to support her argument:

*eat-iendo (eating), iendo

164

*watch-ar (to watch, look)

*quit-ear (to quit)( Spanish-English code switchings Poplack, 1981)

Poplack excludes intra-word switchings containing a free morpheme from a language and a bound morpheme from another. She also excludes bundles of bound morphemes coming from different languages. eat-iendo( eating) has got the asterix standing for a mistake in syntactic terms because the bound morpheme iendo( stading for –ing marker in Spanish) may not be attached to the verbal root eat from English( ibid: 587). She(ibid: 587) argues, in this vein, that the instances which violate the constraint are either free morphemes which have undergone phonological adaptation in the receiving language105, or borrowings which may have not been integrated phonologically.

In a criticism that she put forward to the free morpheme constraint, Myers-Scotton (1993b: 30-31) argues that Poplack manages thus to exclude any form such as ‘*eat-iendo’ by automatically putting it into the category of Borrowings, not a form of Code switching. Myers Scotton(1993: 30) gives many examples from her data in which English free morphemes are combined with Swahili Bound inflectional affixes. The examples include cases where the Swahili Inflectional morphemes are in both pre and post positions. Other specialists point to the the inadequacy of the free morpheme constraint to an exhaustive understanding of code switching as a phenomena. Muysken (2000: 14) argues in this vein that one of the main issues is the typological status of the languages involved in code switching.

“ soon a number of criticism were raised with respect to both the Equivalence and the Free morpheme constraint often involving typologically more different pairs, where code mixing appears to violate both constraints at the same time”( Muysken, 2000: 14)

2.2.2.3.1 Criticisms to the Free Morpheme Constraint in the Code Switching Literature

105 Poplack gives this example to illustrate phonological integration(una buena excuse[eh kjuws]. She states that excuse has been adapted to the Spanish phonological system by being aspirated

165

Language pairs involving agglutinative languages are the ones in which the free morpheme constraint encounters most troubles( Bantu languages-English (Myers Scotton 1993: 34), Maori-English( Elliason, 1989: 24), Turkish-Dutch( Backus, 1998: 61), Hebrew- English, Hebrew-Spanish( Berk Selingson, 1986: 320), Arabic-French( Bouamrane, 1986: 134), and Berber-French(Benali, 2007: 145). All these pairs involve agglutinative languages.

In Addition to code switching situations involving agglutinative languages that represent a clear threat to the Free Morpheme Constraint, contact situations involving non agglutinative languages also highlight counter evidence to this constraint. Clyne(1987: 743) gives many counter examples from his data involving German-English and Dutch-English code switching in Australia. He seriously puts into question the very notion of equivalence and symmetry in code switching. He argues that code switching situations are rarely symmetrical. Gardner Chloros (2009: 96) argues in this vein that:

“The Free Morpheme constraint stated that a switch is prohibited from occurring between a bound morpheme and a lexical form unless the latter has been phonologically integrated into the language of the former…Such coinages do in fact occur, in both agglutinative and non agglutinative language. In some communities, this is one of the commonest forms of switching” (Gardner Chloros, 2009: 96)

The cases that are the commonest and the most threatening to this constraint, seem however to be the ones which involve agglutinative languages. It seems that keeping the idea of equivalence operative in these languages is a difficult enterprise. Many researchers (Levelt, 1989: 68, Myers Scotton, 1993: 31; Muysken, 2000: 54) support the hypothesis that this may be due to the fact that agglutinative languages generate new words using combinations between existing free and bound morphemes. The basic premise of this hypothesis is that roots and free morphemes are stored in a different location from the bound morphemes and inflectional affixes in the mental lexicon. Levelt (1989: 78) hypothesized that:

166

“Different languages may have different types of lexical entry in a mental lexicon. In Turkish, for example, Affixes have a separate lexical entry from stem. The stored form will probably consists of all stems, all possible affixes, and a certain number of frequently used multimorphemic words, English at the other extreme. English speakers use words they have probably used before and these are stored in their mental lexicon as full words (not as stems and affixes)” (Levelt, ibid: 78)

Levelt(ibid: 83) argues that if the mental lexicon of the speaker consists of stems and affixes, this predisposes him to combine a stem from one language with an affix from another. This state of affairs has been observed in many agglutinative languages where stems from the guest language are combined with affixes from the host language.

2.3.2.3.2 Some Violations of the Free Morpheme Constraint in the Data of this Study

The present data shows many code switching instances in which verbal, nominal, and adjectival stems from French and Arabic are combined with inflectional and derivational affixes from Berber. Other instances include the affixation of Berber clitic pronominal affixes to Algerian Arabic and French code switched verbs and nouns. Here are some examples:

/ -/( feel- 1st sing inaccomp aff)

( lit: fill it( I tape record on it)

Enregistr-( record-1st sing inaccomp verb aff) (I tape recorded )

/-/( 3rd sing fem inaccomp verb aff-own) ( She owns)

/-/ ( remember- 1st sing inaccomp aff) ( I remember)

167

/---/ ( 3rd sing fem accompl verb aff- spoil- 1st sing accusative cli aff- 3rd sing fem accompl verb aff) ( she spoiled me( she durtied me)

/-.---/ ( 2nd sing fem accompl verb aff- picture(took a picture)-3rd sing accusative affix- 2nd sing fem accompl verb affix)(lit: you pictured her( you took a picture of her)

/-/ (mix- 2nd sing infix- 3rd sing accusative affix)

/-/ (poor- 3rd plu fem adj aff)

/-/(medicine-your)

---(fem sing aff-teacher-Algerian Arabic fem sing aff-fem sing aff)(the teacher)

These combinations represent counter examples to the free morpheme constraint, even if Poplack and her associates would consider some of them as borrowings. We could find many instances which may not be considered as borrowings, for they are used idiosyncratically by speakers. Here is an example:

/   / Melissa //!

( Look(at) your sister how (she) nice and sweet and dirty(cute she is))

168

What is interesting in this example is that code switching does not take place from Berber to French or to Algerian Arabic, but the process is done the other way round. The sentence is from Algerian Arabic and the code switch // is from Berber. This Berber Adjective has got an Algerian Arabic morphological construction since it is normally produced as /--/

Berber-French code switching instances highlight a contact situation which is quite common in the sense that it involves an inflectional language with an agglutinative one (many studies cover such situations); in comparison to that, Berber-Arabic code switching instances are considered differently in the sense that they highlight a contact situation between two languages which are somehow typologically similar (both of them being agglutinative languages in some sense). The free morpheme constraint would predict that there should be fewer problematic cases in such type of code switching instances, for there are few clashes in the morpho-syntactic construction of the two languages. This does not seem to be case with our data which show many critical cases where the equivalence and the free morpheme constraints are violated at the same time. Here are some instances:

/ / ( lit: religion race- 3rd sing nul aff( god sake)

This code switching violates both the Equivalence and the Free Morpheme constraints at the same time. The equivalence constraint is violated because in Algerian Arabic the word race is feminine; consequently, it takes a feminine inflectional suffix at the end. But in Berber it is neutral when it is in 3rd person singular, so it does not need any feminine suffix. It comparatively takes a null neutral suffix. The free morpheme violation lays in the fact that / / (the race) takes a null bound morpheme which neutralizes the 3rd sing feminine Arabic affix. This phenomenon seems to be a recurrent feature in the speech the informants of this study. We have observed that 77, 34% of Arabic code switched 3rd person feminine nouns lose their feminine gender to the benefit of a neutral null gender. This represents a clear violation to the equivalence constraint as it has been stated above in the discussion about the violations to the equivalence constraint.

169

-/-/ - (lit: you go the school (school)

What is of particular interest in this utterance is the code switched noun/ /( lit: the school) which is supposed to have a singular feminine suffix at the end. It is however used by the speaker with a neutral gender null suffix.

The equivalence and the free morpheme constraints seem to have undergone much trouble with different language pairs and the attempts made by Poplack and her associates as late as in 2000 seem to be of a relative value and of a little strength. The arguments put forward by Poplack (2000: 242) that borrowing could be the explanation to most counter examples does not seem to hold water. The fact that she makes a strong claim for nonce borrowings that she describes as a once off process (Gardner Chloros, 2009: 97) does not solve the problem either. Gardner Chloros (ibid:97) talks about cases of Greek Chypriot-English mixed verbs containing a verb from a language and its particle from another one. She also gives instances of verbal constructions in which the verb is English and the infinitive marker is French. Here is one of the examples that she gives:

Tu peux me pick-up-er

You can me pick-up- inf suff

You can pick me up( French-English code switching Gardner Chloros)

This example may not be considered as a borrowing since it is not used by the whole of the community. The same would be true for the example //106( the dirty which normally exists neither in Kabyle Berber( the equivalent in kabyle being /-- /(the dirty( the cute), nor in Algerian Arabic(the equivalent in Algerian Arabic being //( the dirty( the cute) and which has been used idiosyncratically by one of the

106 Examples such as the ones that we are discussing in this part have been already discussed as part of the discussion on the social attributes of code switching above. They have been considered by many scholars (Myers Scotton 1993b: 65; Backus, 1998: 45; Gardner Chloros, 2OO9: 125) as evidence to the existence of a continuum between code switching and interference. These instances of code switching also show that the border line between code switching and other contact phenomena is not that evident and clear cut. 170 informants. Other instances of Berber verbal and nominal stems used with an Algerian Arabic Morphology have been attested in our

/     /

(come with me I am going, come we take a shower)

The verb // which is originally from Berber does not exist in Algerian Arabic, as the equivalent in Arabic is supposed to be //( we take a shower), but the verb in Berber is supposed to be produced as//( I take a shower). Another similar construction has been attested in the data These constructions have been attested in various contact situations. They represent counter evidence to the principle of discreteness of linguistic system involved in code switching as this has been argued by Poplack and her associates.

Investigations carried by researchers working in the same linguistic environment as Poplack i.e. on Spanish-English code switching found constructions which have been initially rejected by Poplack and which were attested in the speech of Spanish-English bilinguals in the US. James Smith (2002: 39) gives the example of verbs such as quit-ear ( quit), watch-ar( to watch), send-er (to send), Check-ear( to check). All these verbs have got an English stem which has been attached to a Spanish affix. All of them have been attested in natural conversations of Mexican emigrants in the US. James Smith (2002: 36) states that:

“The pattern of creating a verb from an English stem with Spanish verb ending is a productive one, at least incipiently, in the Hispanic community of this study... ‘-ar’ and ‘-ear’ are actively affixed to English verb stems to create new verbs in Spanish/English contact situations”( James Smith, 2002: 37)

171

2.4 Conclusion

The analysis of the data of this study indicates that the predictions of the Equivalence Model (Poplack et al, 1981, 1990, 1998) seem to be unpractical for Berber Algerian Arabic, Berber-French and Berber-Standard Arabic code switching. At the beginning of this chapter we made two implied hypotheses. We first hypothesized that the Equivalence Based Model should be operative on Berber-Algerian Arabic code switching. This hypothesis is related to the typological similarities of Berber and Algerian Arabic. It is also related to the pattern of bilingualism between the two varieties which has existed for a long time and which is stable107 to some extent. The second hypothesis that we made is states that applying this model to Berber-French code switching may be problematic. This is related to the typological differences between Berber and French; besides relates to the type of bilingualism between Berber and French. Berber-French bilingualism hasn’t got the same history as its Berber- Algerian Arabic counterpart. The two types of bilingualism are, thus, different. The latter hypothesis seems to be confirmed with the data that we investigated, whereas the former seem to be infirmed. Our test of the Equivalence model in the light of Berber-Algerian Arabic data seem to be inconclusive.

The results of analysis that we have conducted seem to be in accordance with investigations of bilingual data from other environments. As it has been stated above, many scholars working on various contact situations showed the inadequacy of the notion of equivalence to the analysis of code switching. As a consequence of the growing number of criticisms to the equivalence approaches, research started to be directed towards asymmetry and power as factors which have an influence on language contact and the processes that it entails. The power configuration is displayed on two levels. It operates between languages, or between elements in these languages.

107 This is just a hypothesis and not an assumption. It is based on our observation of Berber bilingualism in Algeria in General and of the patterns of bilingualism in the Berber minorities of Oran. Recall that these minorities are the Kabyle and Mzabi minorities of Oran. 172

Two different trends of research were launched. Each trend concentrated on one side of the process. Some scholars worked on the notion of asymmetry in code switching (Azuma, 1983; Nishimura, 1986; Joshi, 1985; Myers Scotton, 1993, 1997, 2002). Others (Wolford, 1983; Disculio, Muysken and Singh, 1986; Pandit, 199O; Belazi, Rubin and Toribio: 1994) worked on the idea of dependency between elements in the sentence as an important factor determining the making up of code switching instances and the constraints on this process.

The coming two chapters will address the findings of each of these two perspectives on its own. We shall start in the next chapter by the dependency approach to the grammar of code switching.

173

Chapter Three

Grammatical Approaches to Code Switching (2) Government and Minimalist Studies

174

3.1 Introduction

The dependency perspective to bilingual phenomena may be considered as a counter current to the equivalence based approaches. Scholarly interest in the application of dependency relations to bilingual data came in a favourable context. This perspective started to appear in the mid eighties. This period witnessed strong and bitter criticisms to the equivalence based models to bilingual phenomena. The equivalence model as it has been expounded by Poplack and her associates in the 80ths started to show its weaknesses as more studies showed its inadequacy to bilingual data from typologically different languages. As a consequence of these circumstances, some scholars tried to change perspective, and to look at Government as a serious alternative to equivalence. The basic premise of the dependency approach is that dependency relations being universal, applying them on bilingual data should be a promising avenue of research. Muysken (2000: 19) argues in this vein that:

“This approach stresses rather on dependency than equivalence, assuming that code mixing108 obeys a general constraint of lexical dependency”( Muysken, 2000: 19)

Dependency approaches to code switching were put forward by these scholars on the basis of the relation of code switching to formal grammar. They tried to apply Chomsky’s (1981) formalisms to code switching data, and they consequently came up with some constraints to code switching. The main principle behind these constraints is that bilingual language is not that different from monolingual language, and that utterances are generated by similar processes whether it is monolingual or bilingual speech.

Scholars working within the dependency perspective to code switching data may be divided into two groups. The first group which is represented by Muysken et al(1986, 1990, 1995, 2000), claims that there are constraints operating on code switching and related bilingual phenomena. These constraints may be drawn from the dependency relations holding

108 Muysken(1995: 180, 2000: 5) uses the term code mixing as an alternative to code switching to refer to intrasentential code switching in which elements of a language or a variety are inserted into stings of another language or a language variety. Muysken (1995: 180) does not use code switching because he considers that this process takes place when two language system alternate. This takes place when there is Exta or inter-sentential code switching; he argues (1995: 180). 175 between the elements of the bilingual sentence. The principles advocated by this group of scholars are again subsumed under Chomsky’s formalisms related to his Government and Binding theory of syntax (Chomsky, 1981). The second group is represented by Mahootian and Santorini (1993, 1997) and Mac Swan (1999, 2000, and 2005). The perspective of this group is not a government and binding one, but rather a minimalist one. The Minimalist theory of syntax at it has been advocated by Chomsky in the 90ths is an amendment of the Government and binding theory. We shall now start with the perspective of the first group i.e. the Government and Binding perspective.

3.2 The Government and Binding Approaches to Code Switching

Wolford(1983: 523),Di Sciullo, Muysken, and Singh(1986) made attempts to interpret Code switching patterns using Chomsky’s(1981) syntactic theory of Government and Binding. This theory is based on the principle that within the sentence there are elements which govern other elements on the basis of lexical relations. It is also based on the principle that this government relation binds elements to each other. These two principles haven’t been devised by Chomsky (1981) to be applied to bilingual data but to account for universal Grammar.

3.2.1 Woolford’s (1983) Government and Binding Model of Code Switching

Woolford (1983) may be considered as one of the first scholars who applied Chomsky’s (1981) Government and Binding formalisms to bilingual data. She investigated Spanish-English Code switching and came along with her own model which seems to support two of Chomsky’s principles of government and Binding. The first one is the constituent and structure within the NP as it has been formulated by Chomsky within the x-bar theory109 and lexical projection that

109 X-bar theory is based on the principle that most phrases (Noun , verb, adverb, prepositional phrases etc..) are governed by a lexical head. These heads are represented by a X (standing for a head noun, head verb, head preposition...etc) which govern a other elements represented by a Xs in a hierarchical manner until we reach the smallest elements. The topmost node is the one which has got the most bars. These hierarchical relations represent maximal projections. A verb has, for example as its maximal projection a noun Phrase or an Adverb phrase or a prepositional phrase.

176 apply to parts of the VP. Woolford (1983: 523) proposes that when there is a similarity between the syntactic systems of the languages involved in code switching speakers may use lexical elements from both systems. She(ibid: 525) also states that when there is a phrase structure that is peculiar to only one system, only lexical elements from this system can fill the terminal node110.

Woolford (ibid: 530) argues that in her data i.e. Spanish-English code switching there is, for example, no switching between a noun and a following modifying adjective. She(ibid: 531) goes on arguing that This is related to the fact that the noun and its modifying adjective are part of the same node(N’ maximal projection). She gives the example of the Spanish N’ which contains N+ adj and English N’ which contains Adj+N. So when N is Spanish Adj in this case may not be filled by an English Adj. We will discuss counter examples to this assumption using Berber-French Code switching instances from our data. The second constraint that she has proposed is related to clitic object pronouns positions in relation to their head verb. Wolford(ibid: 530) proposes that it would be impossible to use a Spanish clitic pronoun with an English verb in the sense that the English syntactic system does not have any clitic pronoun; However, many researchers could find counter examples to Wolford’s(1983: 530) strong claims using similar data i.e. Spanish-English code switching in addition to other contact situations.

Myers Scotton (1993: 43) seriously questions the validity of Woolford claim in relation to maximal projections. She (1997: 43) rightly states in this vein that:

“The claim that an expansion rule which is unique to one of the languages must be filled from that language’s lexicon simply does not hold.”(Myers Scotton, 1997: 42)

110 The terminal node is used here to refer to the dependencies of the X element( the ones that are attached to it with a government bind e.g. the terminal node of a verb includes among other elements: a noun, an adverb, a noun phrase, an adverb phrase, a noun clause, an adverb clause, a prepositional phrase, while the terminal node of a preposition includes its a noun, a noun phrase or a noun Clause 177

Myers Scotton (ibid: 42) goes even further by comparing Woolford’s (1983) constraints to the equivalence model constraints, and states (ibid: 43) that these constraints are simple reformulations of Poplack’s equivalence principles. Myers Scotton (1997: 43) points to the counter examples that have been put forward to the equivalence constraint and state that they also represent counter evidence to Woolford (1983) constraints.

3.2.1.1 Violations to Woolford’s constraints in Instances of code switching from the present data

We could find many instances of Berber-French, Berber-Algerian Arabic, and Berber- Standard Arabic code switching from the data to which the predictions made by Woolford (1983) do not seem to operate. We may take the position of clitic pronouns as a counter example to Woolford (1983) constraint stipulating that there should be no switching between a verb which cliticises for a post positional clitic accusative object and a pre positional object clitic pronoun. Below are some instances from the data displaying such pattern of switching.

/  --/

( now she is sleeping(your sister) 3rd sing obj accusative clitic aff- switch( 2nd sing imperative)- Neg marker)

( Now your sister is sleeping don’t switch it on( the radio)) (Berber-Algerian Arabic code switching)

/  -- - ---  --- -/

(I swear now Inaccompl aff-3rd sing aff-bring the-water 3rd sing fem obj accusative clit aff- inaccompl aff-3rd sing mas aff- throw with-fem nom aff-bottle-fem aff of-water)

(Lit: I swear now he brings the water it he throw with the bottle of water)

178

(I swear(that) now he brings(the bottle of) water, he throws it i.e. the bottle of water(on me))(Berber-Algerian Arabic code switching)

/-- ----/

(2nd sing aff- want-2nd sing aff 1st sing obj accusative clit aff- causative aff111-articulate-2nd sing aff)

(Lit: You want me you make me talk)

(You want to make me talk)(Berber-Algerian Arabic code switching)

In these instances of code switching, the clitic object accusative affixes precede the verbs to which they are attached. They are distributed as follows:

1. The second person object accusative clitic affix // is attached to the Algerian Arabic code switched verb //(switch on) in the negative verb phrase /-- /(do not switch it on).

2. the 3rd person object accusative clitic affix /--/ precedes the Algerian Arabic code switched verbal stem //(throw on) in the Verbal phrase /--- -/(he throws it on).

3. The First person singular object accusative clitic affix is attached to the verbal stem // ( cause to articulate) in the VP /----/(you are pushing me to talk)

These instances of code switching represent clear violations to Woolford (1983) principle which says that there ought to be no switching in that case, for Algerian Arabic does not sub-categorize for a pre positional clitic accusative affix. The only possible position for such

111 Berber uses a clitic derivational affix that precede the verb to signal a causative, or a factitive thematic relation that is assigned by the verb on the object e.g. // enter (introduce or make enter, (go out), //(dismiss, make go out)(Naït Zerrad 1995: 98). 179 an affix with an Algerian Arabic code switched verb should be in a post positional location which is also a possibility in the Berber syntactic system as it has been stated above.

3.2.2 Di Sciullo et al (1986) Government and Binding Approach to code switching

In a trial to sort out the drawbacks of Woolford (1983) constraints to code switching, Di Siciullo, Muysken, and Singh (1986) put forward a Government approach which seems more elaborate than Woolford’s(1983)model . This approach to code switching is said to be more principled in the presentation of its argument( Myers Scotton, 1998: 43).

Muysken et al (1986; 3) argue that there could be no possible switching between a governor and the elements that it governs within a maximal projection. The maximal projection is here used to mean the highest projection of a category such as N or V with the full projection being either NP( Noun Phrase) or VP(Verb Phrase) for example. If we relate these formalisms to code switching, Switching would be impossible between elements related with a government relation. Switching could therefore be impossible between the syntactic or lexical head and its governed elements.

The Government and Binding approach to code switching, as it has been put forward by Di Siciullo, Muysken, and Singh (1986), predicts that the relationship between a lexical element and its syntactic environment determines the language from which the governed element comes. Di Siciullo et al(1986) formulated this constraint as follows:

“The traditional assumption behind X-bar theory is that syntactic elements are endocentric, i.e. that their properties derive from those of their head. Thus a noun phrase is characterized by the same features as its head noun; the internal constituency of a verb phrase in terms of numbers of objects derives from the properties of the verb. Another way of saying this is that the head noun or head verb project their features within the phrase, but not beyond it. The central notions involved here are exploited in the code mixing literature under the 180 government constraint: Not only the categorial and semantic features of a lexical head are projected in the constituent, but also its language index112. The language index was assumed to be something specified in the lexicon, since the lexicon is a language-specific collection of elements.”(Muysken, 2000: 21).

This constraint predicts that there should be for example no switching in cases such as these:

 A head verb and its object  A head preposition and its object (NP complement)  Verb and its clausal complement  Noun and its modifying adjective

Possible cases of switching include elements that are not bound with a government relation. These are mainly:

 Switching between a subject and a verb  Switching of tags  Switching of exclamations, interjections  Switching of adverbs

Other scholars (Bentahila and Davies, 1983; Klavans, 1985: 34, Pandit, 1990: 54) put forward similar proposals using Government and binding assumptions. Klavans (1985: 34), argues, in the same vein, that it is the language of the inflected main verb or the auxiliary of the clause that dictates the language index of the whole clause, for these two elements function as syntactic and lexical heads to the clause as a whole.

3.2.2.1 Counter evidence to Muysken’s et al Government and Binding Approach to Code switching in the Literature

112 Emphasis is mine 181

The arguments put forward by Muysken et al(1986) and by Klavans(1985) were soon criticized by many specialists(Romaine, 1989: 125; Nortier, 1990: 132; Pandit, 1990: 52; Myers Scotton, 1993: 43: 131) working on language situations different from the ones investigated by these scholars. Muysken (2000: 21) himself confesses that:

“The government constraint fails badly for a large number of cases...A definition of government in which the noun governs the whole noun phrase its maximal projection would rule out switching between nouns and their modifying adjectives which is extremely common. Similarly with adverbs which often are within the verb phrase and which switched very often” (Muysken, 2000: 21)

Muysken (1986b: 5) tried to limit the very notion of head to lexical heads (Major lexical heads or categories include N, V, A, P) alone excluding the functional heads such as the determiners and the auxiliaries. In such case code switching or mixing as Muysken calls it would be possible when it involves functional heads, while it would be impossible when it involves lexical heads, but even this revised version of the constraint was soon put into question by many specialists as Muysken(2000: 24) himself acknowledges that:

“As it turns out, however, even this restricted versions runs into grave difficulties, due to abundant more recent counter-evidence... given all these counter examples, the government constraint, even in the revised form of Muysken(1989b) is clearly inadequate”

The data presented by Nortier in 1990 involving Dutch-Moroccan Arabic code switching in the Netherlands proves to be the most damaging to the government constraint model. This data seem to be in complete contradiction with the constraint put forward by Muysken et al (1986). Nortier (1990) Presents counter evidence to every single case of code switching that would have been ruled out by the Government constraint. She gives many instances of switching inside the verb phrase (these involve roughly switching between a verb and its direct

182 object, an indirect object and a direct object, and a copula and its predicate. Here are some examples that she (ibid: 131) gives:

 - een glass water of zo. (verb/indirect object/direct object)

( get for-me a glass of water or so)

L-islam - is echt love. (subject/dutch copula predicate)

(The islam all of it is truly love)

- strak hé stroef.(Subject/ non verbal copula predicate)

(The-moroccans tight,huh, rough)

- -zelfde tijd (prep- noun phrase code switching)

(And- we spent in-same time)

  maatschappelijk werkster(prep-noun phrase code switching)

(Appointed as social workers)(Moroccan-Dutch code switching Nortier; 1990: 139)

Romaine (1989: 130) also presents counter evidence to Muysken et al’(1986) constraints from her Punjabi-English data. She presents instances of code switching between a head verb and its noun phrase. She also discusses instances of switching in mixed compound Punjabi- English verbs with Punjabi operator verbs such as ‘kerna’( to do) and ‘honda’(to be or to become) mixed with lexical verbs from English. Here are some examples that she gives:

183

Exam pass kerna(lit:exam pass do)(do pass exam)

Time waste kerna(lit: time waste do)(do waste time)(Punjabi-English code switching; Romaine: 130)

3.2.2.2 Visiting Muysken et al’s Government and Binding model of code switching using the present data

We made the hypothesis that, since Berber, Algerian Arabic and Moroccan Arabic are languages which are typologically somehow similar113, we may find counter evidence and examples to the government and binding model which may be similar to the ones that have been put forward by Nortier (1990). We also intend to investigate the conduct of Berber and Algerian Arabic code switched items in relation to the government model of code switching. We decided to investigate this aspect of code switching in our data as we believe that it seems to be a perspective in the study of code switching which has been covered by few other studies. Most studies, in the literature, that tested the practicality of the Constraints advocated by the Government and Binding model have been conducted on language pairs that are typologically remote. This is not the case for Berber and Algerian Arabic.

3.2.2.2.1 Subject verb switching

/-/ drag/--   / (Subject/verb)

(Neg - you were chat up-neg marker one of them)

(you weren’t chat up by one of them)(Berber-French switching)

113 The view that Algerian Arabic and Moroccan Arabic can be grouped as varieties of Maghrebi Arabic starts to gain recognition by the community of specialists of these two varieties. This idea has been introduced by Elimam (2000) in a lenghty discussion of the origins of these varieties which are said to be from punic; besides, these two varieties descend from the same ancestry as Berber( the Chamito-Semetic family of languages). This gives the three varieties many typological similarities. 184

// la famille / -  -/ Meriem (Subject/verb)

(nom aff-the family our inaccomp future aff- come clitic accusative aff- bring meriem)

( a member of our family will come and bring Meriem)(Berber-French code switching)

/  --/

(My brother the poor neg-win-neg)

(Lit: my brother the poor he did not win)

(My poor brother did not succeed)

/- -  ---/

(The-sauce fem accompl aff-be a bit nom aff-def-high-fem nom aff)

(Lit: The sauce it was a bit high)

(The sauce was a bit hot)

/ -- -/

(God 3rd sing aff-send-1st plu accusative nom aff- saver)

(Lit: God sent us a saver)(Berber-Algerian Arabic code switching)

In the first two examples, the subjects of the two sentences are French and their verbs are Berber. The subjects of the third, the fourth, and the fifth examples are Algerian Arabic. These are clear cases of subject verb code switching. In the above cited sentences, the subjects precede their verbs. Below is an example in which the code switched subjects follows the Berber verb114. Below are some instances showing this pattern of switching.

114 Both Berber and Algerian Arabic may have subject first sentences or verb first sentences. 185

A: / --/ les resultats /-/ ça va /   /

( The day they be known the results 1st sing inaccomp aff-see its alright of her how is it like)

( The day the results will be known, we will see how ‘its alright’ sounds like)

B: / -  -/

(with god will he brings god the blessing)(Berber-Algerian Arabic code switching)

This is a conversation between two friends. In turn A ‘les resultats’ (the results) functions as the subject of the sentence whose verb is from Berber. In the second turn the subject of the sentence is ‘’ (god). This subject follows a verb which is from Berber

3.2.2.2.2 Verb/Object code switching

/- -/ / bac

(Sister-2nd sing possessive aff fem inaccomp aff-attend the bac(examination)(Verb/ Object)

( My sister will attend the bac examination)(Berber-French code switching)

Nadia /--/ l’africanism

(Nadia 2nd sing fem accomp aff- learn the africanism)(verb/Object

(Nadia did you learn the africanism course)(Berber-French code switching)

/ -- - -- /

186

(I must inaccompl marker- go-1st sing aff to-house inaccomp marker- do- 1st sing aff- business)(Verb/object)

(I must go home as I have some business to do there)(Berber-Algerian Arabic code switching)

/-  gateau  - /

(make-1st sing accompl aff the cake then help-1st sing accompl aff my mother)(Verb/ Object)

(I made the cake and the helped my mother(in the cleaning))(Berber-Algerian Arabic code switching)

/---- - /

(inaccompl aff- Causative aff- boil-1st sing aff The-tea sweet)

(Lit: I am making (it) boil the tea sweetened)(Verb-Object)

(I am boiling sweetened tea)(Berber-Algerian Arabic code switching)

Instances of verb direct object code switching are very frequent in the data under study. The most numerous cases are the ones involving Algerian Arabic objects combined with Berber verbs. The instances of French objects combined with Berber verbs are far less numerous. We may hypothesize that this seems to be related to the prominence of Algerian Arabic in comparison to French in the verbal repertoire of our informants.

3.2.2.2.3 Subject/Predicate (copula-less predicate) Switching

L’histoire-Geo/ - /

(Lit: The history-Geography one-affix only)(Subject/copula-less predicate)

187

(History and geography alone only) Berber-French-Algerian Arabic code switching)

// anglais / / deux sujets

(No English easy they gave them two subjects) (Subject/ copula-less predicate)

(No, the English subject( was) easy. They wer given(choice between) two subjects)(Berber- French-Algerian Arabic code switching)

/  ./

(Thank god! The house care alot)(Subject/ copula-less predicate)

(Thank god! But the house care(was) too much)(Berber-Algerian Arabic code switching)

These subject/copula-less instances of code switching represent clear cases of violation to the Government and binding constraint in the sense that subjects and their copula-less predicates are bound with a government relation. There should be, according to this constraint, no switching between these two elements in the sentence. The examples above highlight cases of French subjects code switched with Berber copula-less predicates for the first two examples, and a Berber subject code switched with an Algerian Arabic copula-less predicate. The Berber-French code switching instances are highly problematic to the Government and Binding constraint, for French subject sub-categorize for a compulsory linking verb (copula).

3.2.2.2.4 Verb-Direct object-Indirect Object Switching

/-/ l’enfance

(Lit:They gave them the childhood)(Verb/Indirect object/Direct Object switching)

(They were examined in the subject of childhood)(Berber-French code switching) 188

/-  - -  /bac /.-   - /

(told- 3rd sing aff-2nd sing accusative ind obj aff the minister of education: get-3rd plu inaccompl aff all def- bac told- 3rd sing aff- 2nd sing accusative 70% get-3rd plu inaccompl aff)(Verb/indirect object/direct object)

(The Education’s minister said to you (That) they (the pupils) will all get the BAC exam: He said to you70% will get (it))(Berber-Standard Arabic-French code switching)

/-  /gateau / - -- /

(make-1st sing accompl aff the cake then help-1st sing accompl aff mother make-1st sing accompl aff-3rd sing clitic accusative aff the lunch)(Verb:/ indirect object/ direct object switching)

( I made cakes and then I helped my mother preparing lunch)(Berber-French-Algerian Arabic code switching)

The three utterances contain three ditransitive verbs. These verbs with their indirect objects affixed to them /-/(he told you) and /-/( they Gave them) and /-/(I helped her). The first two verbs take Algerian Arabic code switched direct objects in the form of noun phrases consisting of nouns and their definite articles. The third verb takes a noun clause which functions as a direct object to the verb of the main clause. The instances of verb-Indirect object-Direct object switching are not negligible in the data that we have investigated.

This pattern of code switching represents counter evidence to the Government and Binding constraint to code switching, for Verbs are supposed to govern Both Indirect and Direct Objects in the sentence (these latter elements are supposed to have the same language

189 index as their head verb. This does not seem the case with our data especially with Berber- Algerian Arabic code switched instances.

The most numerous cases of switching for the pattern Verb/Indirect object/direct object involve Berber and Algerian Arabic. Berber-Arabic code switching instances seems to outnumber Berber-French code switching in our data. We may hypothesize that this may be due to the similar conduct of Berber and Arabic with regards to indirect Objects. Verbs in these two languages sub-categorize for affixes which are attached to the verbal root and which function as clitic indirect Objects. French, on the other hand is a language in which verbs do not sub-categorize for clitic indirect object affixes, but for free morpheme indirect objects115. Code switching in such a case seems to be easier between Berber and Algerian Arabic. This hypothesis is to be verified in some future research using a quantitative study of the occurrences of Verb-Indirect Object-Direct Object switching in our data. The most frequent patterns of Verb-Indirect object-Direct object switching displayed by the informants are as follows:

 Verb(Algerian Arabic)-Indirect object clitic suf(Berber)/Direct Object(Algerian Arabic)  Indirect Objet clitic pref-Verb(Berber)/ Direct object(Algerian Arabic)  Indirect object clitic pref-Verb(Algerian Arabic)/Direct Object(Berber)  Verb(Berber)-Indirect object clitic suf/Direct Object(French)

3.2.2.2.5 Verb-Direct object-Object Complement switching

A: -  

(Wash-1st sing accompl aff the dishes (of) lunch)

(I wash the dishes of lunch)

115 The only exception, in French, is the 1st singular object clitic pronoun which may be attached to the verb e.g. Tu m‟as donné tout ton amour(You gave me all your love) 190

B:  --   --  

(we 1st plu accompl aff-call-3rd plu clitic accusative object suf the dishes you 2nd sing accompl aff-call-3rd plu clitic accusative object suf the dishes) (Verb/Object/Object Complement) (Berber-Algerian Arabic switching)

Objects and their object complements are bound with government, and this is supposed to prevent any type of switching between them if we apply Muysken’s et al(1986: 3) constraint to code switching. The conversation above proves that this constraint does not hold to our data. In a single turn taking two instances of Object-Object complement switching have been displayed by the informants. Instances such as these do not represent isolated cases, but seem to be a recurrent feature in our data. Here are other instances from the data.

/ -- - -/

(give-1st sing accusative indirect object affix the-bottle-fem sing aff of –the- lemonade)(indirect Object/Direct Object/Object complement

(give me the bottle of lemonade)( Berber-Algerian Arabic code switching)

/ - -  . - /

(Yeah hurry-1st sing inaccompl aff have-1st sing inaccompl aff business alot at- home)(Verb/Object/Object complement)

(Yeah I am in a hurry I have a lot of things(to do) at home )(Berber-Algerian Arabic code switching)

191

3.2.2.2.6 Switching patterns involving Prepositions and their objects

Patterns of code switching involving prepositions and their objects seem to be, with patterns of Verb/Object switching, the most numerous in our data. These patterns of switching are the most threatening and problematic counter evidence to the Government and Binding constraint.

Government between prepositions and their objects (be it a single noun, a noun, phrase or a noun Clause) represents one of the strongest bounds that may operate between lexical and syntactic elements. There should be then no permissible switching between prepositions and their objects. This argument has been put in question by Nortier (1990: 139) in her data of Moroccan Arabic-Dutch code switching in the Netherlands. She argues that code switching between prepositions and full noun phrases or single nouns is very frequent in Moroccan Arabic-Dutch code switching in the Moroccan minority of the Netherlands. Here is one of the examples that she(ibid: 139) gives:

/ /- Zelfde tijd

(and we spent in same time)

/ / maatschappelijk werkster

(appointed as social workers)

The data that we have investigated show that the patterns of code switching involving prepositions and their objects are quite frequent. The analysis of the corpus reveals that the different informants code switched 324 times with the pattern Preposition/Object of preposition switching. The distribution of switching patterns is as follows:

 Berber Preposition/Algerian Arabic code switched single noun 95 instances

192

 Berber preposition/Algerian Arabic noun phrase: 78 instances  Berber preposition/Algerian Arabic noun clause: 12  Algerian Arabic Preposition/Berber single noun object: 65  Algerian Arabic preposition/Berber noun Phrase 47  Algerian Arabic/Berber noun clause 5  Berber preposition/French single noun 17  Berber preposition/ French noun Phrase 4  Berber preposition/French noun Clause 1

Here are some instances from the data:

/ - -- -   / Meriem

(God speed inaccompl aff- 3rd sing fem aff- happy(verbal root) in-all-and with- success who-succeed Meriem)(Berber preposition/Algerian Arabic single)

(with god help, she(meriem)will be happy in everything with all the success, she is the one who will succeed)(Berber-Algerian Arabic code switching)

Sarah,   --- -.

(Sarah, this now neg aff-3rd sing accompl aff-come-neg aff from-morning)(Berber preposition/Algerian Arabic object of preposition)

( Sarah has not come back yet from the morning)(Berber-Algerian Arabic code switching)

 -   -  -- AM LIT  

-   Tableau  mercredi

193

(Me think-1st sing accompl aff day first think-1st sing accompl aff tomorrow 1st plu inaccompl aff-have American literature, but when 1st plu accompl aff- see on-the board wednesday)( BerberPreposition/ French Object of preposition)

(I thought the first day(at first glance), I thought(that)tomorrow We would have American literature, but when we saw the board(I discovered that we would have this exam) on Wednesday)(Berber-Algerian Arabic-French code switching)

/ --/

(what(which of-anxiousness)(Berber preposition/Algerian Arabic object of preposition)

(What anxiousness)(Berber-Algerian Arabic code switching)

/ -  ---/

(Lit:Before at-school it was 1st sing fem object acusative aff- 3rd sing accompl aff- hurt-1st sing object accusative aff)(preposition/noun phrase)

(When I was at school it(my leg) was hurting me)(Berber-Algerian Arabic code switching)

/ - /-les toilettes

(Lit: Now, she went out from-the toilettes)(Berber preposition:French noun+article)(Berber- French code switching)

/ --- ()/- toilette

(This when-2nd sing fem inaccompl aff-enter-2nd sing fem inaccompl aff to- (the)toilet)

194

(this(a prayer that you recite) when you enter the toilet)(Berber preposition-French Bare noun)(Berber-French code switching)

/  - /

(She said he increased(his weight with-three kilos/( Berber preposition-Algerian Arabic noun phrase(Noun+modifier)(Berber-Algerian Arabic code switching)

The most frequent pattern of switching involving prepositions seem to be the ones in which a Berber preposition is associated with an Algerian Arabic object of preposition. This tendency seems to be followed directly by the cases of Algerian Arabic prepositions which are combined with a Berber object of preposition or an Algerian object of preposition. What is striking in these tendencies is the fact that Berber prepositions are used interchangeably with Algerian Arabic ones by some informants. This tendency seems to be making its way, as we have already discussed that in our previous investigation of the Berber community in (2004: 65). At that time we put forward the hypothesis that this may be the signal of an ongoing mixed language formation. This hypothesis seems to become stronger, but we still could not verify it.

In order to have a clearer picture of the distribution of Algerian Arabic and Berber prepositions in the speech of the informants that we have been recording for this study, we intend to look at their distribution in a subsequent section where we will be dealing with code switching from a different perspective. This perspective is the alternational one, and as it has been stated above, it does not consider that code switching is the insertion of elements of one language or variety into strings from another language or variety116.

On the light of the numerous counter examples that we have listed, we may assume that the Government and Binding constraints to code switching seem to be inoperative for our data. This is similarly true for both types of code switching (Berber-Algerian Arabic, Berber-

116 The alternational approach to code switching has as its main principle the idea that when there is bilingual speech, elements from languages are not inserted, but they rather alternate. This approach has been advocated by renown scholars such as Muysken(2000: 83), Auer(1998:4), and Sebba(2000: 34) 195

French code switching). We consequently felt the need to look at other perspectives which would fit our data.

3.2.3 Minimalist Approaches to Code Switching

Minimalism is an important evolution in Chomsky’s syntactic theory. Many scholars (Halmari, 1993, 1997; Mahootian, 1993 Belazi, Rubin and Toribio, 1994; Chan, 1999, Mac Swan, 1999, 2000) work in this perspective. These scholars have been trying to overcome the inadequacy of the government model by devising minimalist models to the analysis of code switching and other bilingual phenomena. The similarity between the minimalist and the government approaches to code switching models is that they are both subsumed under Chomsky’s (1981, 1986, 1993) principles of sentence generation. We shall see if these models may apply to the analysis of our data or not.

3.2.3.1 The Single Model (Belazi, Rubin, and Toribio 1994)

This model is embedded within the minimalist approach to syntax as it has been expounded by Chomsky in 1991. Belazi, Rubin and Toribio (1994) investigated Tunisian Arabic- French bilinguals in Tunisia and Spanish English bilinguals in the US. They (1994: 229) put forward the constraint that switching is not permitted between functional heads and the elements that they govern. Belazi et al (ibid: 228) argue that:

“The language feature of the complement f-selected by a functional head117, like all other relevant features, must match the corresponding feature of that functional head...As predicted, code switching between a lexical head and its complements remains unimpeded... The Functional Head Constraint; thus, restricts switching between a functional head and its

117 Emphasis is mine. I have put stress on the syntactic head because Belazi et al(1994) purposefully eliminated the lexical heads from the constraint as they thought that including them would be repeating the same error as the one made by Muysken et al(1986) when they put forward the government and Binding constraint

196 complement by invoking the strong relation that exists between them.” (Belazi, Rubin and Toribio, 1994: 228)

Functional heads include elements such as Conjunctions, Complementizers, determiners, and numerals. This constraint is partly based on the assumptions made by Abney(1987) and Chomsky(1993) with regard f-selection which predicts that lexical heads contain the same features as their specifiers and complements(these features are projected on the specifiers and the complements)118. The same applies to functional heads119 which carry features which are projected on the elements that they govern. Belazi, Rubin and Toribio (1994: 231) concentrated their focus on the functional heads excluding the lexical heads from their constraint to code switching. On the light of the analysis of their data and bilingual speakers elicitation about imaginary examples120, Belazi, Rubin and Toribio(Ibid, 231) ruled out switching such as these:

 Switching between a Quantifier and its complement Noun Phrase(the quantifier being the syntactic head

 Switching between a complementizer(a relative pronoun or a subordinator) and its subordinate clause

 Switching between a determiner and its noun or noun phrase

 Switching between a negative particle and its verb

118 These features include morpho-syntactic information such as gender number, case and agreement)

119 The functional heads that have been identified by Belazi Rubin and Toribio (1994) are the following ones: complementizers which govern subordinate clauses, gender, number and case inflections which govern noun phrases, determiners that govern Nouns or noun phrases, quantifiers which govern nouns or noun phrases, and negative markers which govern verb phrases.

120 Researchers working in Chomsky‟s government tradition rely on both naturally occurring data analysis, and artificial examples that they create and that they present to native speakers to see the degree of their acceptability. The latter approach is highly controversial in the sense that the instances of code switching created do not occur in natural environments. 197

Belazi Rubin and Toribio (1994: 234) put forward an additional constraint to the functional head constraint. This constraint is called the Word Grammar Integrity Corollary. As the Functional Head constraint this constraint draws its principles from Chomsky’s minimalism(1993). It is based on the assumption that lexical items have to obey the rules of the language from which they are drawn. Belazi et al(1994: 234) put forward the following constraint:

“A word of language X, with grammar Gx, must obey grammar Gx “

(Belazi, Rubin, and Toribio, 1994: 234)

This constraint predicts, for example, that switching between a Noun and its modifying adjective is possible if the modifying adjective obeys the grammatical rules of its lexical head (its noun). Belazi et al give the following examples:

j’ai une voiture // (lit: I have a car nice)(Tunisian Arabic-French code switching Belazi, Rubin, and Toribio, 1994: 232)

This example is said to be possible according to Belazi et al(1994: 234), for the Tunisian Arabic code switched adjective// follows the Tunisian Arabic grammatical rule which says that adjectives take a post nominal position. Being the syntactic head, the adjective controls its lexical head ‘voiture’(car) which is from French and which is not supposed to have a post-nominal adjective. The following example is ill formed according to them because the adjective ‘Belle’(nice) which heads the noun controls the noun from the syntactic point of view may not be put in a post-nominal position:

*/ / belle(Lit:I have car nice)(Tunisian Arabic-French code switching: Belazi, Rubin, and Toribio, 1994: 234)

198

This constraint shows some similarities with the constraints that have been put forward using the equivalence tradition (Poplack, 1981; Woolford, 1983). What is similar is that both imply that code switching has to respect the grammars of the languages involved in the process. But what is different between the two is that the constraint put forward by Belazi, Rubin, and Toribio (Ibid: 231) do not consider the languages on an even scale. The most important language being the language contributing with the syntactic heads which themselves are going to control both the distribution and the language features of the elements that they govern. The equivalence constraint implies that both grammatical systems contribute evenly in the process of code switching.

Though they hold for Tunisian Arabic-French code switching in Tunisia and Spanish English bilinguals in the US, The constraints put forward by Belazi et al(ibid) have been seriously put in question by other researchers(Mahootian and Santorini, 1996: 468; Halamari, 1997: 84; Boumans, 1998: 76). Halamari (1997: 84) presented counter evidence to the Functional Head Constraint using her Finnish-English code switching data in the US. In the example below, she (ibid: 85) presents a counter example to the Functional Head constraint:

Luen yhde+n magazine + n

(I read one magazine)

Muysken also put into question both constraints as they have been put forward by Belazi et al (1994). He (2000: 26) argues in this vein that t:

“As shown by Mahootian and Santorini (1996), the predictions made by the Belazi, Rubin, and Toribio (1994) proposal can easily be shown to be incorrect on the basis of published data from the code switching literature. The main reason is again that categorial equivalence undoes the effect of the government (here f-selection by functional categories) restrictions”(Muysken, 2000: 26)

199

3.2.3.1.1 Testing the Predictions of the Single Model (Belazi et al, 1994) on the Present Data

The data that we have dealt with seems to violate the two constraints put forward by Belazi et al (1994). Especially with instances of Berber-Algerian Arabic code switching. We may hypothesize that this could be related to the typological features of Berber and Algerian Arabic which are somehow similar languages in the sense that they both belong to the Chamito Semitic family of languages. This remains a preliminary observation that needs to be verified with a deeper investigation of the relationship between the typologies of Berber and Algerian Arabic and the conduct of their functional heads. We shall now expose the different violations to the Functional head constraint that we have noticed in the analysis the data.

3.2.3.1.1.1 Switching between a complementizer (a relative pronoun or a subordinator) and its subordinate clause

We could find many instances of code switching which violate the functional head constraint restriction on switching between Complementizers (Functional heads) and their complement phrase or clause. The restriction stipulates that:

“Switching is disallowed between a C(complementizer and its IP(complement phrase)”(Belazi, Rubin and Toribio, ibid: 228)

This prediction proves to be particularly inoperative with our informants. They seem to use complementizers from Berber, Algerian Arabic, and to a lesser extent French interchangeably regardless the complement phrase or subordinate clause that follows the complementizer or the head verb in the main clause. We will come back to this tendency in the informants for such behaviour in a subsequent section where we will be trying to demonstrate using a statistical procedure that complementizers operate at a suprasegmental level which goes beyond the segmental level of the sentence. And that they are not controlled

200 by any segmental restrictions, at least not with the corpus that we are investigating. Belowe nstances.

Melissa / /, Melissa - / Melissa /- . -  ---/ parceque /-      -- -/ à la creche  --/

(Melissa better than me, Melissa 3rd sing accompl aff- stay because Melissa in-morning 3rd sing accompl aff- sleep me in-morning neg aff-3rd sing Accompl aff-sleep- neg aff because wake(up)-1st sing accompl aff early, she not early, she neg aff-3rd sing fem accompl aff-go- to the nursery school me fem aff- go-1st sing accompl aff)(Complementizer/Subordinate clause)

(Melissa is better than me, Melissa stayed(until late in night) because Melissa slept in the morning and I didn’t sleep(in the morning) because I woke up early, she (did not wake up) early, she did not go to nursery school, me I go)(Berber-Algerian Arabic-French code switching)

/--  --- - -/

(inaccompl marker- want to know-1st sing aff only if- 3rd sing accompl aff-hurt- 2nd sing accusative aff fem nominal aff- cat-fem nominal aff)(If complementizer/ subordinate clause)

(I want to know only if it hurted you(my littl) cat(taliking to her daughter))

/- / tu n’es pas gentille/   -- - - --/

(And-you, you are not nice, now if Inaccompl aff- buy(take,marry) one- fem aff 3rd sing fem aff-speak fem aff-arabic-fem aff)

( And you are not nice, what if you now marry a woman(who speaks Arabic)

201

/-- -     -- /

( inaccompl aff-jump-2nd sing inaccompl aff to-there then you stay you jump like this neg aff- 2nd sing inaccompl aff- rehearse-2nd sing fem inaccompl aff nothing)

( you are jumping from here to there then you keep jumping like that you will not learn anything)

3.2.3.1.1.2 Switching between numerals and/or quantifiers and their complement Noun Phrases

Belazi et al(ibid: 229) following Ritter(1991)121 argue that switching is disallowed between the noun and its determiner in the sense that they are linked by the DP requirements(D stand for determiner and P stand for Phrase) which is a government relation whereby the quantifier represents the syntactic head. Belazi et al(1994: 230) argue that:

“We will predict that switching is not possible between Nom(the quantifiers and the numerals) and their complement Noun phrase”(belazi, Rubin, and Toribio, 1994: 230)

This prediction does not seem to be valid for the data that we have studied. The instances of Algerian Arabic and to lesser extent French nouns which are combined with Berber quantifiers are very frequent. We have also observed that Algerian Arabic numerals are combined with Berber nouns quite frequently. We may assume that this may be due to the fact that the Berber numeral system is to its majority borrowed from Arabic. Apart from using the Berber numerals// ‘one’ and // ‘two’, our informants seem to switch directly to the Arabic system from three. The examples that we are going to list below are the

121 Ritter (1991) worked within the X-Bar tradition. He has identified the Nom category in which we may include numerals and quantifiers 202 ones in which the informants do not use borrowed Algerian Arabic numbers but code switched ones 122.

/ -.- --/

(be careful 2nd sing fem inaccompl aff- disturb-2nd sing fem inaccompl aff the-brain-my)

(be careful you are disturbing brain my)(Algerian Arabic noun /Berber possessive pronoun switching)

/ -- /

(One Nominal aff-the-thing-only)(lit: one thing only)(only one thing)(Berber numeral/Algerian Arabic noun/Berber quantifier)(Berber-Algerian Arabic code switching)

/ --/

(He attended two-of-courses)(He attended two courses)(Berber numeral/ Standard Arabic switching)

/ / les ampoules economique -250w

(lit:Give me two the bulbs energy saving of-250w)(Berber numeral/French noun phrase switching)

(Give me two 250w energy saving bulbs)

/--   -.- -/

122 For this part of the study we will assume that a borrowed form is a form which has been phonologically adapted to the host language, and that a code switched form is a one which has not been adapted to the host language phonologically. We believe that this criterion is not valid for the distinction between the two forms, but this is not the aim of this chapter. 203

(nominal aff-the-thing only that 2nd sing aff-repeat- 2nd sing aff)(Berber numeral/Algerian Arabic noun+article/ Berber quantifier switching)

(Lit: the thing only that you repeated)( The only thing that you repeated)

/  sujet 

(They gave them one subject only) )(Berber numeral/Algerian Arabic noun+article/ Berber quantifier switching)

/   /

(Twenty five hundred centimes only)(Algerian Arabic numeral+Noun/Berber quantifier)

/ - ---,   --- /

(Many girls who-faint-3rd plu accompl aff, many girls who-3rd plu accompl aff-scream- 3rd plu accompl aff)(Berber quantifier/Algerian Arabic noun switching

((there were)many girls who fainted, (there were) many girls who screamed)

// les examens -DS / --- --  -revise-/

(I swear (that during) the exams of-synthesis neg aff-1st plu aff-neg aff all-the-time 1st plu aff- revise)(Berber quantifier/Algerian Arabic noun switching)

(I swear that during the make up exam we do not have time enough to revise)(Berber Algerian Arabic-French code switching

204

3.2.3.1.1.3 Switching between a negative marker and its Complement verb

Belazi et al(ibid: 231) argue that negative markers have been identified as functional elements by Pollock(1991: 390). He (ibid: 390) named this functional category Neg. Following Pollock’s claim, Belazi et al, state that code switching is not permitted between NEG (the negative markers) and their verbs, for they are attached with a syntactic government relation or bound. This assumption proves to be inoperative with our data especially with Berber- Algerian Arabic code switching. Here are some instances:

/ - -  ---/

(not all(the verse), and-not like that-this-2nd sing inaccompl aff- rehearse-2nd sing inaccompl aff)(Algerian Arabic Neg/Berber verb switching)

(it is not all the verse, it is not like this that you rehearse)

/  -- ---/

(No, no inaccompl aff-stand-1st sing aff on(by)-myself)(Algerian Arabic Neg/Berber Neg/ Berber verb switching)

(No I will stand by myself(an old woman talking to her daughter)

/ Khadidja /-- /-l’examen / ---  -- /

(What’s up khadidja 2nd sing fem accompl aff-work-2nd sing fem accompl aff in-the exam of- fem nom aff- arabic-fem nom aff or neg marker- sing fem accompl aff-work-2nd sing fem accompl aff nothing)(Algerian Arabic Neg/Berber verb switching)

(What’s up Khadidja did you do well in the Arabic exam or not)(Berber-Algerian Arabic code switching)

205

/ -- - - - --?/ (Neg marker 1st plu accompl-aff-agree we go to the Fem nom aff-shoe-fem nom aff- plu aff) (Didn-t we agree to go to buy shoes)(Algerian Arabic Neg/Berber verb switching)

/  -- / (No not Neg aff-Like-1st sing accompl aff-Neg aff)(Algerian Arabic Neg/Berber Neg/Berber verb switching)(Berber-Algerian Arabic code switching)

/--   -- /-sujet (Neg-remember-1st sing accompl aff- I swear Neg- remember-1st sing accompl aff about-subject)(Berber Neg/Berber verb/Algerian Arabic Neg/Berber verb switching) (I do not remember, I swear I do not remember about the subject)

/ --/ (I swear Neg-know-1st sing aff)(I swear I did not know)(Algerian Arabic Neg/Berber verb)

The Neg/verb restriction to code switching seems to be particularly inoperative with our data especially with Berber-Algerian Arabic code switching. Berber-French code switching seems to be more restrictive in terms of the relations holding between Neg and verbs. This differential conduct of Algerian Arabic and French in relation to this feature may be related to the way Berber and Algerian Arabic construct their negative forms. This is done in a similar manner with the two languages which attach affixes to the left and right of the verb. French does not function like that in relation to such construction.

We may hypothesize that Berber speakers sometimes draw negative markers from both Berber and Algerian Arabic to construct negative statements. This hypothesis goes

206 in the same direction as the one that we put forward above where we have been discussing the idea that there is a mixed code which is in the making. The tendency with the category of negative markers is less accentuated than the one with complementizers, but still the percentages of Algerian Arabic negatives markers is not negligible in the speech of the informants that we have been working with. We shall come back again to the use of Negative markers in a subsequent section where we will be discussing code switching repetition phenomena in the speech of our informants.

3.2.3.1.1.4 Switching between Modals and their Lexical verbs

Belazi et al (ibid: 229) argue that switchings between modals and auxiliaries and their verb phrases (lexical verbs) are not permitted. Many researchers have pointed to the inadequacy of this prediction. The most threatening studies are the one which have investigated bilingual compound verbs formation. Romaine (1994: 133) investigated Punjabi-English compound verbs, and Gardner Chloros (2009: 115) investigated Greek Chypriot-English mixed compound verbs in London. These two scholars presented bilingual data in which there are plenty of mixed compound verbs made up of a native language modal and a lexical verb from the guest language. The corpus that we have investigated shows few instances of Berber-Algerian Arabic or Berber-French mixed compound verbs. Below are some instances.

Ça va on revise /- - / Am Lit //-Brit Lit (It‟s OK we are revising be-1st plu inaccompl aff 1st plu inaccompl aff-learn Am Lit and Brit-Lit)(Berber modal/Algerian Arabic verb switching)

/ - -- - -/ (Before at-the school 3rd sing fem accompl aff- be 3rd sing fem accompl aff- hurt- 1st sing obj accusative clit aff) (Lit: Before(when)at school it was it hurt me) (When I was At school before it was hurting me)

207

The mixed Berber-Algerian Arabic verbal constructions /-- - -/ (it was hurting me) and /- - / (we are we revise) are made up of Berber auxiliaries (the verb be) and Algerian Arabic lexical verbs (the verbs hurt and revise or learn

/ --  - / (You must inaccompl aff-go-2nd sing aff sit(stay) 1st plu aff-talk a bit)(Algerian Arabic modal/Berber verb switching) ((Are you sure)you must go stay(so that) we have a little talk)

// Sakina /, ---  -/ ((she)Must Sakina come, I took it from her)(Algerian Arabic modal/ Berber verb switching)

/ --   --/ ((we) Must 1st plu-think me I am in-heart-my) (I must think about what is inside my heart)(Algerian Arabic modal/Berber verb code switching)

/ -- -/ ((It)must fem aff-competence-fem aff inaccompl aff-exist)(Algerian Arabic Modal/Berber verb switching) (The competence must exist)(Algerian Arabic modal/Berber verb code switching)

The instances of Modal/verb code switching that we have observed seem to involve the Algerian Arabic modal /(must) this may be due to the fact that this verb is used interchangeably with its Berber counterpart ..... by the informants. Auxiliaries are, according to many scholars (Belazi et al, 1994: 231; Mahootian, 1996: 465; Myers Scotton, 2002: 64), said be functional elements by excellence. Switching within this category is also a strong sign of the development of a mixed language. Even if their frequency is negligible compared to the ones of complementizers and 208 prepositions, the instances of modals‟ code switching corroborate more the hypothesis put forward above.

Recall that the above cited hypothesis says that our informants are in the process of developing a mixed code. This mixed code is characterized by the fact that functional elements are drawn from Berber, Algerian Arabic, and French to a lesser extent. This mixed code seems to be at its early age, but it is still in a constant development. In a previous study (Benhattab: 2004: 51), we detected signs of this code, but the tendencies seem to be more accentuated with the present study. The percentages of Algerian Arabic code switched functional elements seem to be higher than the ones that we found in the previous study. We shall come back to this issue when we will be dealing with suprasegmental code switching instances123 as part of the alternational approach to code switching.

3.2.3.1.2 Further Remarks on the Single Model (Belazi et al, 1994)

On the light of the instances of code switching that we have discussed above, the assumptions made by Belazi et al (1994) seem, to their majority, to be inoperative in our study. We may hypothesize that this may be due to the fact that the typological differences between Berber and Algerian Arabic are less salient than the ones between Tunisian Arabic and French. We have noticed that Berber-Algerian Arabic code switching involves Algerian Arabic code switched functional elements in most positions in which Belazi et al predicted severe constraints. Instances of French code switched functional elements seem to be less salient than the ones of their Algerian Arabic counterparts. We may hypothesize that the typological differences between French and Berber do not act in favour for Functional elements switching. This is again a preliminary hypothesis made on the light of our observations and analysis of the corpus. This hypothesis needs to be verified with a more thorough study of the relationship between code switching and typological differences between the languages involved in

123 These instances include complementizers (relative pronouns, subordinators and subordinating conjunctions), coordinators and coordinating conjunctions, and discourse markers and fillers to cite only these.

209 code switching. Muysken (2000: 53) points to this relationship as a possible explanation to the patterns of code switching as he argues that:

“It is clear that some mixing patterns have to do with the linguistic typology124 of the languages concerned. The degree of linguistic kinship and the lexical morphosyntactic similarity between the languages are quite important. Exploiting typological differences can lead to new, more directed research strategies in code mixing research”(Muysken, 2000: 53)

The universality of the Functional head constraint has been challenged by many scholars working both within a similar Government X-Bar tradition and in other insertional traditions. As it has been stated above, the assumptions made by Belazi et al face strong and bitter criticisms. These criticisms even emanated from scholars working in the same government tradition (Mahootian, 1996: 464; Chan, 1999: 164; Eppler, 1999: 283;125 Mac Swan 1999: 39, 2004: 290). Counter evidence also come from scholars working with other perspectives such as the insertional one (Boumans, 1998: 95; Myers Scotton, 2002: 154).

One of the alternatives to Belazi et al’s model is Mahootian’s (1993) minimalist model of code switching. The approach advocated by Mahootian (1993) is drawn from Joshi’s (1985) Tree Adjoining Grammar126. The most important premise in Mahootian’s approach is that code switching does not require any specific bilingual constraints apart from those of the languages involved in code switching. She (1997: 470) argues in this vein that:

“The analysis of code switching relies on general principles of phrase structure rather than on constraints that are specific to code switching such as the Functional Head

124 Emphasis is Muysken‟s(2000) 125 In Gardner Chloros(2009: 99)

126 Tree adjoining grammar (TAG) is a grammar formalism defined by Joshi (1985). Tree-adjoining grammars are somewhat similar to context free grammar, but the elementary unit of rewriting is the tree rather than the symbol. Whereas context-free grammars have rules for rewriting symbols as strings of other symbols, tree-adjoining grammars have rules for rewriting the nodes of trees as other trees(Wikipedia)

210

Constraint and the Grammar Integrity Corollary... Heads determine the syntactic properties of their complements in code switching and monolingual contexts alike” (Mahootian and Santorini, 1996: 470)

This assumption clearly put into serious doubt127 the two constraints put forward by Belazi et al (1994). It also sets the pace for a different approach to the analysis of code switching in which the rules governing code switching are drawn from monolingual rules of the languages participating in code switching. This approach is known as the Null Approach.

3.2.3.2 Mahootian and Santorini’s (1993, 1996) Null Approach to Code switching

As it has been stated above, Mahootian’s (1993, 1996) approach can be subsumed under Tree Adjoining formalisms128. Mahootian and Santorini (1996: 471) focus on the idea that constraints on code switching may be drawn from derivational rules which are operative at the monolingual level. In their approach, Mahootian (1993) and Mahootian el al (1996: 469) clearly reject the idea of equivalence as it has been implied by Belazi et al (1994) approach to the notion of equivalence between the languages involved in code switching. Mahootian and Santorini (1996: 469) states in this vein that:

“Since the Word GrammarIntegrity Corollary requires the placement of each single word of a language to be consistent with the languages grammar, it reduces in effect to the well-known Equivalence Constraint. As a result, the Word Grammar Corollary, like the Equivalence Constraint is incompatible with the large body of evidence that has been accumulated against the latter”(Mahootian and Santorini, 1996: 470)

127 Mahootian and Santorini criticized Belazi et al (1994) constraints using counter evidence from Moroccan Arabic, Farsi-English and Spanish-English code switching. They could demonstrate that none of the constraints put forward by Belazi et al(1994) was operative to these types of code switching. They(ibid: 470) further criticized the conceptual considerations made in Belazi et al(1994)approach 128 Mahootian is not the only scholar who has worked within a Tree Ajoining grammar approach to analyze code switching. Chan (1999) has developed a similar model in the analysis of code switching. The only difference between the two models is that Mahootian‟s model is subsumed under tree adjoining grammar while Chan‟s one is rather directed towards a more minimalist approach.

211

Mahootian (1993: 45) also assumes that there is no difference between syntactic functional elements and Lexical functional elements129. Mahootian and Santorini (1996: 471- 473) were able to present counter evidence and examples to all the assumptions made in Belazi et al (1994). The cases of switching that they were able to present are the following ones:

 Switching between Verbs and their modifying adjectives coming from languages with different parsing systems

posta le cailin Brazilian married with girl

'married to a Brazilian girl' (Irish-English; Stenson 1990:171, (7b))

We observed similar patterns of code switching involving nouns and their modifying adjectives within the Berber-French code switching instances. French being a left parsing language in terms of adjective placement in relation to their head nouns, and Arabic being a right parsing language in relation to these two elements, switching involving Nouns and their modifying adjectives would then be difficult. This seems, however, to be a common practice in bilingual environments. Below is an example:

Ma cousine/ - - /

(My cousin poor-fem aff 3rd sing aff- be tired somehow)

(My poor cousine got somehow tired)

129 Mahootian and Santorini(1996: 471) argue that: “In contrast to what happens under Belazi, Rubin and Toribio analysis, under our analysis lexical heads behave no differently in code switching contexts than functional heads.” 212

Mahootian rightly points to the fact that there is no consensus among scholars about the concept of head when it comes to nouns and their modifying adjectives. Bentahila and Davies (1983: 84) assume that in code switching data the head is the adjective. They backed up their position using evidence from their data. Other scholars (Di Sicilio, Muysken, and Singh, 1986: 9; Pandit, 1990: 42) assume that the head is the noun in the sense that it represents the syntactic head which is accompanied by its complement. Mahootian and Santorini (1996: 473) simply reject the application of the principle of head and government in code switching data involving nouns and their modifying adjectives. They argue (ibid: 473) that:

“Modification cannot be forced into the mould of the head-complement relation, but is better represented by adjunction structure130… Since there is no head in this case to unilaterally determine the order of the modifying and the modified phrases, we conclude that the direction of adjunction can be determined by either of the grammars of the languages involved in code switching” (Mahootian and Santorini, 1996: 473)

The other possible instances of switching that have been discussed by Mahootian and Santorini (1996) are the following ones:

 Switching between a complementizer and its subordinate clause

 Switching between a preposition and its object be it a single noun, a noun phrase, or a complement clause

 Switching between a verb and its object

130 This argument is restricted by Mahootian et al (1996: 476) to ordinary modification (excluding certain adjectives which exhibit the power of heads and which do not function as adjuncts). Adjuncts which are either Adverbs or adjectival adjuncts may not postulate as heads.

213

We may observe that Mahootian and Santorini (1993, 1996) give a considerable freedom to elements in code switching environments131. This is partly why the approach is labelled as a ‘Null Model’

This does, however, not give a complete freedom to elements in bilingual environments. On the basis of their data analysis Mahootian and Santorini put forward another principle in which they restrict code switching occurrence and in which they try to emphasize the importance of heads in code switching. They (1996: 472) state that:

“Their model does not allow code switching at every phrase structure node. Rather, Heads play a pivotal role, imposing their syntactic requirements and thereby determining the phrase structure position, syntactic category, and feature content of their complements”(Mahootian and Santorini, 1996: 472)

Mahootian and Santorini (ibid: 471) applied this principle to the head vebs/ Complement object switching. They ruled out, for example switching between Head verbs from a VX132 language which are preceded by an Object from an XV language. Here is an example of an impossible code switching that they give:

*You'll xune- ye jaedid buy.

House Poss new

(You'll buy a new house)

131 Freedom is not used here with an absolute meaning, but it means that code switching is not governed by bilingual constraints, but by predictions drawn from the two grammars separately. Freedom is used under the Tree Adjoining Grammar in relation to the fact that parsing directions of nouns are for example on both left and right directions.

132 A VO language is a language in which objects follow their head verbs (English). An OV language is on the contrary a language in which the object precedes its head verb( Japenese,Hindi,Farsi, Tamil) 214

This means that a head (a verb, for instance) determines the phrase structure position and the content feature of its complement (the complement’s sub-categorization features).

3.2.3.2.1 Some Criticisms to Mahootian and Santorini’s (1993, 1996) Null Approach to Code Switching in the Literature

The constraints put forward by Mahootian and Santorini (1996: 65) have been criticized on a number of grounds by many scholars. Muysken (2000: 26) state that, even if the approach advocated by these two scholars seems to be new, the principle underlying their constraints adopts some kind of equivalence between the languages involved in code switching. This idea is not that new. It has proved its inefficiency through time.

Gardner Chloros and Edwards (2004: 116) and Gardner Chloros (2009: 99) also criticize Mahootian’s et al approach for it reliance on the government rules which apply between elements at the lexical level of phrases. Gardner Chloros and Edwards (bid: 116) relate the shortcomings of an over-reliance on these surface syntactic categorizations to the counter evidence that have been put forward by many sholars (Boumans, 1998: 89; Eppler, 1999: 287; Mac Swan, 2005: 13). These counter evidences are mainly drawn from instances of code switching involving mixed forms, bare forms and avoidance strategies. They directly challenge the idea that in code switching there are elements which govern other elements, or the one that in such a process languages or language varieties govern other languages or language varieties as Gardner Chloros and Edwards (2004: 116) put it:

“The existence of mixed forms, bare forms, and avoidance strategies suggests, on the contrary, that there is more going on in code switching than can be accounted for by models which assume a base language. At an intuitive level, Many bilinguals find the idea that one language is always dominant in code switching speech does not correspond with their experience” (Gardner Chloros and Edwards: 2004: 116)

215

The instances of bare code switched forms seem to challenge the morphosyntactic and the syntactic constraints to code switching put forward by most scholars. We will come back to their challenging nature when we will discuss the drawbacks of the next model i.e. Myers Scotton’s insertional model. Below are some instances of bare forms from Berber-French and Berber-Algerian Arabic code switching:

/--  -/

(Talk-2nd sing fem imperative-1st sing object accusative affix a bit about- boyfriend)

(Lit: Talk to me a bit about (your) boyfriend)(Berber-Algerian Arabic code switching)

/  --  -/

(Lit: I only who-do-infinitival affix house work at-house)(Berber-Algerian Arabic code switching)

The bare forms in these examples are // (boyfriend) and // (housework) which are nouns that lack determiners or possessive pronouns from either Berber or Algerian Arabic)

Even if it has been put into serious question by some scholars(Boeschoten and huybregts, 1997; Boumans, 1998; Chan, 1999; Eppler, 1999), Mac Swan (1999, 2000) proposed an approach to code switching which shows some similarities with the one advocated by Mahootian et al(1996). He adopts a similar minimalist approach to the analysis of code switching.

3.2.3.3 Mac Swan’s (1999, 2000) minimalist Approach to code switching

Mac Swan (1999, 2000) investigated Nahuatl133-Spanish code switching in the city of Mexico. He applied the principles of Unified Grammar134 as they have been expounded by

133 Nahuatl is an Indian language used in central America 216

Chomsky in the 90ths to the interpretation of code switching. Mac Swan (1999: 42) argues that the assumption of languages as fortresses has no reason d’étre in the investigation of code switching as he puts it:

“As has been frequently noted, the question of where one language ends and another begins is more a matter politics than of linguistics…languages, as such, are theoretically spurious entitles in the sense that “French” or “German” are political rather than grammatical constructs”(Mac Swan, 2005: 4)

The second premise behind Mac Swan’s (1999, 2000, and 2005) approach to the analysis of code switching is that code switching is no more than “the mixing of grammars which is itself the mixing or union of two lexicons” (Mac Swan, 2005: 5). As a result, the lexical requirements have the priority on the grammatical ones. These lexical requirements are checked before the production of code switching utterances135.

This features checking is not done differently from monolingual language production. This means that lexical items may be drawn from either of the languages involved in code switching. The only syntactic requirements are the ones of the mixed grammar and not at any point in the process of switching the requirements of the monolingual grammars of the languages involved in code switching. Mac Swan’s (1999: 44) position is that:

134 Universal Grammar (Chomsky: 1995) is a development of generative grammar and its latest development is the minimalist program. It is based on the underlying principle that differences between languages are neutralized at the lexical conceptual level rather than being based on purely syntactic parameters. When applied to code switching this principle is supposed to help scholars interpret the phenomenon of code switching at a level which is more abstract and refined than the morpho-syntactic level. The applications of minimalism to code switching are not intended to tackle surface aspects i.e. performance, processing or production aspects of code switching. These applications are rather directed towards the mechanisms that operate at the brain level to permit bilingual speakers use grammatical or well formed utterances (Mac Swan, 2005: 4).

135 As it has been stated above Mac Swan‟s model is not a production model, but rather a competence model which addresses the mental mechanisms of checking at lexical interface. These checking processes are mainly devised to block lexically ill formed code switched utterances ahead before they are produced. 217

“Nothing constrains code switching apart from the requirements of the mixed grammars. Put differently, this assumption posits that all of the facts of code switching may be explained just in terms of the principles and requirements of the specific grammars –and embedded principles of Universal Grammar –involved in each specific utterance under consideration…A sound account of code switching will appeal to no code switching specific constraints or mechanisms. A theory of code switching consistent with leading assumptions in linguistic theory would include no code switching-specific rules” (Mac Swan, 1999: 146)

This may be regarded as a clear departure from the other grammatical approaches to code switching which argue to their majority that code switching has to comply to some extent with the requirements of the monolingual grammars of the languages participating in code switching and or with some constraints specific to code switching.

Mac Swan approach is also couched in one of the most important assumptions made by Chomsky (1995) in his minimalist program. This assumption relates to the influence of lexical items on their environments. These lexical items when operating at a cross-linguistic (bilingual) level may trigger conflicts. These conflicts are sometimes due to the requirements of the lexical items themselves.

Recall that these requirements are mainly lexical. However, there are other features checking. These include phonological or morphological checking. In his approach to minimalism in general and to code switching in particular, Mac Swan(1999, 2000, 2005) assumes that phonological and morphological features are part and parcel of the lexical constructs as he (2005: 6) puts it:

“The lexical items in a bilingual’s repertoire are mentally compartmentalized in some sense, with a specific set of phonological and morphological rules associated with each “lexical compartment”. Lexical items are collections of formally substantiated morphological and phonological rules which applies to a specific set of lexical items, contained in a LEXICON” (Mac Swan, 2005: 6) 218

3.2.3.3.1 Mac Swan’s (2005) applications of minimalism to the counter evidence of the Equivalence Model (Poplack, 1982)

Mac Swan (2005:7-14) tried to see whether his approach to code switching can help bypass the drawbacks of the Free Morpheme constraint as it has been expounded by Poplack in 1981. He seems to make a strong point for Poplack (1981, 1990) word internal constraint. He states that that many of the counter examples put forward in the literature (Nartey, 1982; Bokamba, 1989; Myers-Scotton 1993; Halmari, 1997; Chan, 1999; Hlavac, 2003) do not take into consideration the interface between the phonological, the morphological and the lexical features of the languages involved in code switching. Looking at the cross influence between these features may explicate many of these counter examples, he (ibid: 7) argues. Muysken (2000: 54) has a similar position as he states that:

“Exploiting typological differences can lead to new more directed research strategies in code mixing research. We need to look at the lexical and morphological typology. Morphological typology plays a role in code mixing in so far as we consider the type of word internal mixing involved in morphologically integrated borrowing to be a type of code mixing”(Muysken, 2000: 54)

Here are some examples given by Mac Swan (2005) in which he contradicts the opponents of the free morpheme constraint:

a. Juan esta parqu-e-ando su coche.

Juan be/3SS park-DUR his car

(Juan is parking his car.)

b. Juan parqu-e-o su coche.

219

Juan park-PAST/3SS his car

(Juan parked his car.)

c. Juan parqu-e-ara su coche.

Juan park-FUT/3SS his car

(Juan will park his car.)(Spanish-English code switching Mac Swan 2005: 8)

This example has been judged as acceptable by the informants of Mac Swan (2005) even if it clearly violates the free morpheme constraint. Mac Swan explicates its acceptance by the argument that the English borrowed item park has been adapted to the Spanish phonology by adding a Spanish phoneme /e/.

3.2.3.3.1.1 Facilitating Strategies in Code Switching and Borrowing (with a Special Reference to the Present Data)

Many Algerian Arabic and French borrowings and code switchings in our corpus have been adapted through the affixation of Berber Morphemes. Here are some instances:

/-/ (place), /-/ (boyfriend), /- -/ (ten past ten), /-./ (difficult), /-/ (easy), /-/-gateau (the cake), //- micro (computer), -la Chaine (the (TV) channel), //-enregisteur (tape recorder)

In all these instances of Algerian Arabic and French noun switchings, the code switched nouns are preceded by the nominal affixes //, // and //. These affixes are used in Berber before nouns and nominal adjectives. Some scholars (Treffer Daller, 1994: 91; Muysken, 2000: 57; Gardner Chloros and Edwards, 2004: 107) put forward the hypothesis that there are facilitating processes that enable bilingual speakers to code switch in a smooth

220 manner without violating the phonological, morphological and/or grammatical requirements of the host language.

We may hypothesize that in the case of our corpus this is done with the category of nouns by encapsulating them into the morphological system of Berber by the affixation of a nominal affix from the host language to the nominal stem 136 from the guest language. Muysken (2000:31) calls this strategy ‘Morphological agglutination’. He (2000: 32) argues that it represents a tiny part of the manifold competencies in the hand of bilinguals to by-pass the system’s constraints:

“If we take a strong system oriented view and conceive of the juxtaposition of material from different languages…Then we can imagine then to be various strategies to neutralize mixing and make it less offensive; thus, neutralizing the system conflicts by bending it rules a bit.”(Muysken, 2000: 31-2)

Clyne (2003: 65) observed similar facilitating strategies used by bilingual speakers in his investigation of Chinese and Vietnamese-English bilinguals and in his study of Dutch and German-English bilinguals in Australia. He (ibid: 75) discusses the use of ‘bridge words and morphemes’ from the host language that are used by bilinguals to facilitate code switching. We have observed similar bridge strategies used by the informants of the present study. Below are some instances:

/ -/ (two of- lessons)

/ -/ (what of-(house) work)

/  -sujet/ (one only of-subject)

 -Rotatif(bread of-Rotative(oven)

136 Nouns and adjectives are said to originate from similar nominal stems in the sense that Adjectives are not supposed to exist as a category per se in Berber(Nait Zerrad, 1995: 45) 221

In these instances, the bridge morpheme is the preposition // (of) which is not supposed to be used with numerals. It is used to facilitate the switching to the Standard Arabic noun (lessons), the Algerian Arabic code switched noun //(house work), and the French code switched noun ‘sujet’(subject)

Other compromise strategies as Clyne (2003: 65) calls them include bare forms i.e. code switched forms in which there are morpho-syntactic features from neither of the languages participating in code switching. This strategy has been observed in many bilingual contexts.

Many scholars (Backus, 1996: 87; Boumans, 1998: 45; Clyne, 2003: 75; Owens, 2005:27) have put forward the hypothesis that they highlight one of the most interesting strategies used by bilinguals to neutralize conflicts and/or overlapping between the systems of the languages participating in code switching. We will come back to these bare forms in our corpus as we will be tackling them in a subsequent section which deals with the problematic cases to Myers Scotton’s insertional model of code switching.

The above assumptions made by Clyne(2003: 65), Mac Swan(2005: 7), and Muysken(2000: 54) seem to be very interesting to our investigation in the sense that we have noticed that switching and borrowing of single verb and noun categories seem to be the most recurrent in our corpus.

We may hypothesize that this may be due to the morphological typology of Berber which is said to be an agglutinative language to some extent. This seems to be an interesting avenue for future research. The insights made by Nortier137 (1990), and Backus (1998) seem to be worthy exploiting to verify our hypothesis, for they investigated code mixing and borrowing

137 Nortier investigated Moroccan Arabic-Dutch code mixing and Borrowing in the Moroccan minority of the Netherlands. Backus (1996) investigated Turkish-Dutch code mixing and borrowing in the Turkish minority of the Netherlands. 222 involving two languages which are somehow similar to Berber with regard to their morphological typologies.

3.2.3.3.2 Criticisms to Mac Swan’s Minimalist Approach to Code switching in the Literature

As we have observed, Mac Swan (1999, 2000, and 2005) and other researchers working within the minimalist and the government traditions (Toribio, Rubin, and Belazi, 1993; 1996; Mahootian and Santorini, 1996; Boeschoten and huybregts, 1999) have been arguing that minimalism is sufficient alone to the analysis of code switching. Their assumptions have been criticized by scholars working within the insertional approach138who claim that this is too strong an argument especially in a field which involves highly competent bilingual processes. Jake, Myers Scotton, and Gross (2002: 69) state in this vein that:

“Some researchers studying code switching suggest that the Minimalist Program may offer a parsimonious explanation for the structures occurring in this bilingual phenomenon. However, their claim is, in fact, stronger. That is, they claim nothing more than a minimalist approach to monolingual data is needed to explain code switching structures…Only a modified minimalist approach can account for the substantial body of code switching data. Such an approach may even only succeed if it incorporates a basic asymmetry between the languages participating in code switching” (Jake, Myers Scotton and Gross, 2002: 69)

This criticism clearly gives a paramount importance to the idea of asymmetry between the languages involved in code switching. It also sets direction to a more eclectic approach in the analysis of code switching, but the problem seems to be still valid for the insertional

138 This approach is mainly based on the assumption that code switching may involve languages which co-exist in an asymmetrical environment i.e. an environment in which one language is the dominant language and the other one is the embedded language. This type of code switching is mainly characterized by the insertion of singly occurring, small stretches of elements from the embedded language into longer stretches of the matrix language.

223 approaches show some weaknesses in code switching contexts where there is some kind of symmetry between the languages involved in the process139.

Scholars working within minimalism have been further criticised on questions related to methodological issues. These scholars (Toribio and Rubin, 1993; 1996; Mahootian and Santorini, 1996; Chan, 1999; Mac Swan, 1999; 2000; 2005) partly rely in their investigations of code switching on invented utterances that they present to native speakers to check their acceptability. Gardner Chloros (2009: 98) rightly points out to the problem of which bilingual speakers’ competence to appeal to.

Another important criticism to the Minimalist approach to code switching is related to the dismissal of singly occurring code switchings to the status of borrowings. In the same vein, intraword switchings are disallowed within Minimalism (Mac Swan, 1999; 2000). The argument given by Mac Swan is related to the checking of Phonological features. This checking is different from a linguistic system to another and if there is a mismatch at this level. This may result in the inhibition of code switching at an early stage140. According to Mac Swan (1999: 45; 2005: 7), most code switched elements that do not obey this constraint are defined as ‘Nonce Borrowings141’. He (2005142: 8) states that one of the prerequisites for singly occurring code switchings to be well formed is their phonological integration to the host language. This

139 Spanish-English code switching in the US may be considered as a good case of symmetrical code switching where the two languages participate somehow equally to the process of code switching.

140 This is argument is subsumed under the PF disjunction theorem which states that there is a phase of features checking before any surface production and that phonological checking is an important part of this phase. Mac Swan (2005: 3) correlates checking and bilingual code switching in the sense that phonological checking is different from a language to another. This implies that switching within the word is blocked as it implies a mismatch between different phonological systems. The only possible alternative would be the phonological integration ore adaptation of the other language item to the host system according to Mac Swan(ibid: 5)

141 This concept is used by Mac Swan with the same meaning as Poplack and her associates (1988b: 97, 99) i.e. borrowings which haven‟t yet reached a full status in the receiving community. These are mainly singly occurring code switched forms that are inserted into larger constituents framed by the receipt language( Poplack and Meechan, 1998: 127)

142 Jake et al (2002: 70) criticized earlier proposals from Mac Swan (1999, 2000) of a minimalist approach to code switching, but our discussion of his approach is based on a more recent paper (Mac Swan, 2005) in which he keeps the same assumptions as the earlier proposals and in which he only adds criticisms to the Matrix Language Frame model and a tentative response to the criticisms put forward by Jake, Myers Scotton and Gross in 2002. 224 assumption has been seriously put into question by Jake, Myers Scotton and Gross (2002: 70)143. They (ibid: 71) point to the idea that singly occurring code switchings may be analyzed in the same way as their larger counterparts (Larger code switched elements). They (ibid: 71) further state that:

“Mac Swan and others are missing a generalization. Mac Swan avoids features mismatch that would arise when singly occurring forms appear in mixed constituents by classifying all such forms as borrowings. Under his PF Disjunction Theorem, morphemes from one language inflected with morphemes from another language are blocked” (Jake, Myers Scotton and Gross, 2002: 71)

Mac Swan responds to this very criticism by arguing that the strength of his argument is related to the assumption that monolinguals and bilinguals store their lexical items in a similar manner i.e. as lexical entries which have morphological and phonological features. This assumption seems to be controversial in the sense that there are few objective arguments supporting it in bilingual research; besides it is based on results reached in error analysis and neurolinguistic research144.

A competing version asserts that the lexical entries and their morphological and phonological features are stored separately in the mental lexicon of bilinguals. The debate between the two versions is still valid presently. Research in Neuro science advances very slowly; besides, there is still no real and valid arguments to support either versions

144 Researchers working on error analysis, and on bilingual speakers suffering from pathologies such as aphasia and accidental brain damage put forward the hypothesis that lexical items and their morphological and/or morphological features are not stored in the same areas in the brain. The techniques that they have used to investigate this are the magnetic resonance imagery and neuro-surgery experiments(Grosjean, 1997: 236; Fabro, 1999: 37; Myers Scotton and Jake, 2000: 1067) 225

3.2.3.3.3 Testing Mac Swan’s (1999, 2000, 2005 Predictions on some instances from the Present Data

Our analysis of the data pushes us to put the unverified hypothesis that Lexical items in Berber bilingual language are stored in a compartment different from the one of the inflectional and derivational affixes. The informants of the present investigation display the capacity to produce a non negligible number of singly occurring code switched morphemes from Algerian Arabic, Standard Arabic and French. These morphemes are inflected with bound morphemes from the host language (Berber seems to be the host language for the present investigation in the sense that it is considered by many or our informants as their mother tongue). Here are some instances of this pattern of code switching from the corpus of this investigation.

/- /(go- 1st sing accompl aff) (I went)

/- /(full- 2nd sing accompl aff) (You are full)

/-/(3rd sing accompl aff-like)(she liked)

/--/(3rd sing inaccompl aff-leave-2 sing accusative obj aff) (It leaves you)

/-- -/( Berber verb fem accompl aff-Algerian Arabic verb fem aff-hurt- 1st sing accusative obj aff)(It hurt me)

/--/ (2nd sing fem accompl aff-upset-2nd fem accompl aff) (you were upset)

/-/ (3rd sing fem accompl aff-dirty)

/--/ (Future aff-repeat-1st sing inaccompl aff)

/-/ (the brain-1st sing poss clitic pronoun)

/--/ (fem nom aff-teacher-fem nom aff) (teacher)

226

/- / (medecine-2nd poss clitic pronoun) (your medicine)

These instances of code switching represent an evidence that the Berber bilingual speakers that we are investigating have the ability to combine a stem(be it a verbal or a nominal stem from either French, Algerian Arabic, or standard Arabic) with inflectional, derivational, or clitic morphemes from Berber(Mzabi or Kabyle Berber). The abundance of such code switching instances obliges us to consider them as being worth analyzing145.

We, thus, decided to take Jake et al(2002: 75) stand which states that single code switchings are code switchings items which may function differently from longer code switchings. This does, however, not prevent them from being code switched items which need to be interpreted on their own right. The argument put forward by Jake et al (ibid: 75) is that such singly occurring code switchings are the most numerous in code switching research corpora and that they need to be interpreted in the same way as their longer counterparts.

Some of the singly occurring code switchings that we found in our corpus may be considered as borrowings, for they have been phonologically integrated to the host language. Others may not be considered as they have not followed the same process of integration146. Another criterion that may be interesting is the use in the monolingual repertoire of the host community. Some of the instances of singly occurring code switchings are not use in monolingual discourse even if they have been integrated to Berber in code switching instances. Below is an example of a code switched form which is integrated but which has been used only once in several recordings.

145 The most significant tendencies of code switching that we encountered in our corpus are the ones of singly occurring code switchings. The highest percentages seem to be the ones of Algerian Arabic code switched nouns. This tendency seems to be somehow similar for both Kabyle and Mzabi Berber. The second highest percentages are the ones of Algerian Arabic verbal stems. The switching of singly occurring adjectival stems follows directly to some extent.

146 Phonological integration is however not always a valid criterion in distinguishing borrowing forms from code switching ones, for there are in the literature cases of long established borrowings that have been integrated neither morphologically nor phonologically. 227

/ - / specializ-/ - /-plantes médicinales

(Me start-1st sing aff inaccompl aff- specialize-1st sing aff in-Bread of-medicinal plants)

(Lit: Me I start I specialize in bread of medicinal plants)

(I started to specialize in Medicinal Plants’ breads)(Berber-Algerian Arabic-French code switching)

What attracted our attention in this example is the French code switched verb ‘specializ’. This noun is not a borrowed form in the sense that it is not used by Berber speakers in their monolingual language147. In spite of not being a borrowed form, this singly occurring code switched verb

Even if Mac Swan’s (1999, 2000) model of code switching is a minimalist one, it underlies a clear influence of Poplack’s equivalence vision of the rules and the processes governing code switching148. The approach advocated by Jake et al (ibid) may be subsumed under another tradition which is the insertional one. As it has been stated above the basic premise of most insertional models is that there is an asymmetry between the languages participating in code switching. One of the languages thus functions as the base language, and sets the syntactic frame of utterances while the other language only participates with items that are inserted in the frame set by this base language.

147 The strategy that we have used to check if a form is borrowed or code switched is by presenting it to different informants from the ones that we have recorded. Adaptation to the host language seems to fail badly in this test. There are in fact some instances that have been in use for a long time in the Berber community without being adapted to the Berber system phonological and morphosyntactic systems. There are also some instances in the data that are perfectly integrated to Berber, but which are not in use in the monolingual repertoire.

148 Mac Swan(2005: 7) overtly states that his Disjunction Theorem “ is subsumed under the Free Morpheme constraint, and it significantly overlaps with the Free morpheme constraint in terms of range of empirical predictions”

228

Although the fact that the two approaches are clearly different, Jake et al (2002) tried to integrate minimalism as it has been expounded by Mac Swan in their insertional approach to code switching. They argue that their approach is also couched in the

Minimalist Program as it has been expounded by Chomsky (1995, 2001), but they most importantly state that one of the main principles underlying their approach is the idea of asymmetry between the languages involved in code switching.

229

3.3 Conclusion

The government and minimalist models of code switched that have been presenting in this chapter have proven to be unpractical on Berber bilingual data. The very essence of the Government models i.e. the notion of dependency seems to be problematic to code switching data in general and to the present data in particular. Recent research in syntax has gone even further by questioning a notion which is considered as being central to the understanding of modern syntactic theory i.e. the one of dependency. Contact linguistics and code switching research on cross linguistic grammar have been following the same path. Leaving the challenge of the grammatical interpretation of code switching unresolved.

Throughout our analysis of the Berber bilingual data in the light of the government and minimalist models, we have felt uncomfortable with the concept of head. What to consider as a head syntactic or lexical heads? The answer is that after our analysis of the data we feel short of an answer. The ideas that we discussed in relation to this very notion are still debateable even if most of the references that we have been using in this chapter are at least three years old. This period of time is considered by syntax theorists as a long period of time. Many steps have been made in the theory of syntax; however the hazy lines remain hazy and some concepts haven’t yet been settled.

Remaining with our thirst for the right interpretation for our data, we shall consider another approach in the investigation of code switching and related phenomena. The next Chapter will address the contributions of the insertional models and their applications to cross linguistic environments. As we have been arguing above the applications of equivalence based, and Government based models149 proved to be partly inoperative to our data. We will consequently direct our attention to the insertional models as they represent an alternative to the equivalence and to the government based models of code switching. We will also try to

149 Recent research on minimalism abandoned the notion of Government because of the controversies about the class of governing categories and because of the difficulties in defining the operative concepts which are highly complex and subject to little consensus among scholars working in minimalism.

230 address the contributions of alternational approaches150 to code switching. The rationale behind using these alternational approaches is that they interpret patterns of code switching that are considered as not worth investigating. Another reason behind using the alternational approaches is that they look at the drawbacks of the insertional approaches with a trial to interpret their problematic cases i.e. the ones of the insertional approaches.

These approaches are considered by many contact linguists who have worked on the grammar of code switching as the most suitable (Backus, 1996; Boumans, 1998; Muysken, 2000; Gardner Chloros, 2009), and as the strongest in the interpretation of code switching in many contact situations involving asymmetrical bilingualism151. The bilingual situation that we have been investigating is an asymmetrical one in the sense that the patterns of code switching that we have focussed on are asymmetrical as they involve the following patterns of

 Berber-French code switching  Berber-Algerian Arabic code switching  Berber-Standard Arabic code switching

150 The alternational approaches(Muysken, 2000: 120; Auer, 1998: 10; Li Wei, 2002) to code switching investigate the enter and extra sentential patterns of code switching that are considered as not worth studying by scholars working within the insertional tradition.

151Asymmetry is used here with its different perspectives i.e. asymmetry in speakers‟ bilingual competence and in the statuses of the languages being in use in the speech communities under study. 231

Chapter Four

Grammatical Approaches to Code Switching (3)

Insertional Studies

232

4.1 Introduction

The present chapter tackles the interpretation of code switching as the Kabyle and Mzabi Berber communities of Oran display it. It is a continuation of the previous chapters that were also devoted to the grammatical investigation of code switching and related phenomena. This chapter is devoted to the investigation of code switching patterns in the data from an insertional perspective. The approaches that we will be using in this part are couched within the insertional morpho-syntactic tradition of code switching analysis. This tradition was launched by Joshi (1986) and it has further elaborated by Myers scotton and her associates (1993, 1995, 1997, 2000, and 2002) through their well-known Matrix Language Frame. In addition to Myers Scotton’s approach we may discuss other insertional approaches such as the one advocated by Boumans (1998). We will present other studies that are of some relevance to the insertional perspective in code switching152.

4.2 Insertional approaches to code switching

These approaches are considered by many contact linguistics researchers who have worked on the grammar of code switching as the most suitable (Backus, 1996; Boumans, 1998; Muysken, 2000; Gardner Chloros, 2009), and as the strongest in the interpretation of code switching in many contact situations that involve asymmetrical bilingualism153. The bilingual situations that we have been investigating seem to be asymmetrical ones in the sense that the patterns of code switching that they highlight are asymmetrical as they involve the following types of code switching:

152 IN this part of the chapter, we will tackle the insertional approaches that we consider as relevant to the data under study. We will also try to present alternative approaches to Myers Scotton‟s insertional model (1993, 1997, 1995, 2000, and 2002) such as the Monolingual structure approach advocated by Boumans (1998). The relevance of these approaches to our data is related to the fact that there are some counter evidences to Myers Scotton‟s model in the data that we have investigated that are interpreted in these alternative approaches.

153Asymmetry is used here with its different perspectives i.e. asymmetry in speakers‟ bilingual competence and in the statuses of the languages being in use in the speech communities under study.

233

 Berber-French code switching(the asymmetry here may not only be related to the bilingual competence of the speakers, but it may also be related to the status attributed by the Berber community to the languages in Oran Speech community)

 Berber-Algerian Arabic code switching(Asymmetry may be here related to the status of the languages involved and to a lesser degree to the bilingual competence of the speaker in the sense that many informants of this study are balanced bilinguals in Berber and Algerian Arabic, some of them have even got a native like control of Algerian Arabic)

 Berber-Standard Arabic code switching (asymmetry may be related to the differing competence of the Berber informants in Standard Arabic. It may also be related to the status of Standard Arabic in the Berber community as part of the Algerian Arabic speech community)

4.2.1 Joshi’s Insertional Model of Code Switching

The study carried by Joshi (1985) may be considered as one of the first applications of the notion of asymmetry to bilingual code switching data. The underlying principles subsumed under Joshi’s model of code switching are drawn from error analysis research in the differential conduct of functional and lexical elements in Speech errors. Joshi (ibid) applied the finding of Garrett154 (1975) to his Marathi-English code switching data.

He hypothesized that the differential conduct of elements in monolingual discourse may be applied to bilingual discourse as well. Joshi (ibid) first applied Garrett’s (distinction between open and closed class items to his bilingual data. He put forward the hypothesis that there are two types of asymmetries operating when code switching takes place:

154Garrett (1975, 1988, and 1990) investigated monolingual language processing using a psycho-linguistic error analysis procedure. Garrett (1990) found that there are two classes of morphemes which are retrieved differently and during different steps in language production. These morphemes are closed class items and open class items. Garrett‟s argument is based on the assumption that open class items and close class items do not pattern together when errors occur. This means that when there is an error in the production of an open class item, the following errors may be predicted as there is a high probability that they will be from the same type (open class). 234

 An asymmetry between the Matrix language (the base language or the dominant language and the embedded language. Joshi states in this vein that:

“Despite extensive intrasentential switching, speakers and Hearers generally agree on which language the mixed sentence is coming from. We can call this language the Matrix Language and other language the embedded language” (Joshi, 1985:190)

 An asymmetry between the major categories or open class items i.e. nouns, verbs, adjectives and adverbs and the minor categories or close class items such as determiners, quantifiers, prepositions, possessives, etc. Joshi (ibid: 194) claims that switching open class items is permitted whereas switching closed class items is not.

4.2.1.1 Criticisms to Joshi’s Constraints in the Literature

One of the constraints put forward by Joshi (ibid: 194) states that there is no possible switching between a complementizer and its subordinate clause. The argument that he presents is that complementizers are close class items; consequently, they cannot be switched. This constraint has been has been tested by Bokamba (1988) on his Kinshasa-Lingala code switching data. Bokamba (ibid) found that the French complementizer ‘que’ (that) is very often used in complex sentences where the main clause or subordinate clauses are either from French or from Lingala. Pandit (1990: 51) also tested Joshi’s constraints on close class items, and he found that they are inoperative on Hindi-English code switching data. Mahootian (1993c, 1996: 476) similarly puts into question the constraint prohibiting switching of close class items using her Farsi-English code switched data. She states in this vein that:

“A number of authors have observed that the close class constraint is indeed empirically inadequate” (Mahootian, 1996: 476)

235

Here are two examples given by Mahootian (ibid: 476) as a criticism to the constraint on close class items switching:

Anyway, I figured /ke/ if I worked hard enough, I'd finish in the summer.

that

(Farsi complementizer switching in English framed complex sentence)(English-Farsi; Mahootian 1996: 476)

Maen through taemam- e ina raeftaem.

I all Poss these I went

(I've gone through all of this.) (English preposition switching in a Farsi framed sentence)(Farsi- English; Mahootian, 1996: 476)

4.2.1.2 Possible Violations to Joshi’s Constraints from the Present data

The data that we have investigated seems to violate Joshi’s (1985) constraints on a number of grounds. The constraint stating that complementizers switching is not permitted proves to be highly ineffective for our corpus. As it has been stated above, our informants seem to use Berber and Algerian Arabic complementizers interchangeably. They also seem to use French complementizers to a lesser extent. Here are some instances:

/-   -- - -  /

(And you now if fut aff-buy (marry)-2nd sing aff one-fem aff 3rd sing fem –speak Arabic)

(And you if you marry a woman (who speaks Arabic))(Berber Algerian Arabic-code switching)

236

-  - 

(inaccompl aff- kaka(verb) then Inaccomlp-drink milk)

(she does Kaka then she drinks milk)

/ -  ---/

(before at-school it was hurting-1st sing obj accusative aff)

(before at school it was hurting me)

We have noticed in the corpus of this study that Algerian Arabic complementizers are sometimes used interchangeably with their Berber counterparts. Some scholars working within the insertional approach argue that complementizers are not worth analysing, others argue that, since complementizers are situated outside the level of the sentence, they are to be analyzed in a different manner from intra-sentential code switchings. This does not prevent them from having an importance from the grammatical point of view.

Myers Scotton (1993) took over the notion of asymmetry between languages in code switching as it has been expounded by Joshi (1985: 190). She also took the terms ‘Matrix language155’ (hereafter ML) and ‘Embedded language’ (hereafter EL) from Joshi156. Her approach to the grammatical analysis of code switching is similarly insertional in the sense that the contact situations she (1988, 1993b, 1997, 2002) has been investigating are characterized by an asymmetry between participating languages.

155 Wentz (1977) is considered by the bulk of the community of contact linguists as one of the first scholars to talk about the matrix language. She (1977: 85) argues that the matrix language is the language of the sentence.

156 Joshi (1985:191) himself credited Sridhar and Sridhar (1980) with the terms host language (the equivalent of matrix language in Joshi (1985) terms, and guest language (which is the equivalent of embedded language in Joshi‟s terms). Sridhar and Sridhar (ibid) are worth mentioning when dealing with insertional approaches to code switching. They are among the first scholars who investigated the differential roles of the languages participating in code switching, and the power hierarchy between them in such process.

237

4.2.2 Myers Scotton Insertional Approach to code switching (the Matrix Language Frame Model)

At it has been stated above Myers Scotton (1993, 1997, 2002) elaborated a model of Code switching that is based on the asymmetry between the languages participating in the code switching and other bilingual phenomena. The perspective advocated by Myers Scotton and her associates is an insertional one. She (1997b: 4) defines code switching as:

“Code switching is the selection by bilinguals or multilinguals of forms from an embedded language (or languages) in utterances of a matrix language in a single conversation” (Myers Scotton, 1997: 4)

This view of code switching is clearly different from the one advocated by both the linear and the government approaches. Some scholars such as Muysken (2000: 15) argue that the asymmetry between languages as it has been advocated by Myers Scotton(1993, 1997) and other scholars working within the insertional approaches to code switching may be related to the social contexts in which they have undertaken their investigations. Some others such as Backus (1996: 75) argue that it may be related to the degree of bilingualism displayed by the speakers, as he (ibid: 75) puts it:

“Balanced bilinguals tend to code switch in a different way than L1 dominant bilinguals do, the former often prefer alternation, Which means by definition that both ML and El units whereas the latter prefer insertion”(Backus, 1996: 75)

Some other scholars such as James Smith (2002: 45) relate code switching types to the typologies of the languages involved in code switching. James Smith (ibid: 45) argues that insertional code switching tend to occur in situations involving typologically different languages; besides, he states that alternation in code switching involves situations where the

238 participating languages are typologically similar157. Using the data that we have investigated, we will argue that, even if Berber and Algerian Arabic are typologically somehow similar, our informants make an extensive use of Algerian Arabic insertions into otherwise Berber utterances.

As it has been stated above, one of the most important premises behind the approach advocated by Myers Scotton (1993b, 1997; 2002) is the notion of asymmetry. This asymmetry does not apply only between the languages participating in code switching, but also between the elements and morphemes being code switched. Before looking at the hierarchy between the elements being code switched, it is important to look at the way scholars working within insertional approaches identify the Matrix language.

4.2.2.1 Identifying the Matrix Language

Identifying the Matrix language is crucial to most insertional models in general and to the Matrix Language Frame model in particular. This seems to be an easy and simple enterprise at first glance; however, the concept itself is highly problematic and controversial (Muysken, 2000: 65, Gardner Chloros, 2009: 101). The source of controversy is the criteria used in establishing this language. Myers scotton (1993, 1997, 2002) herself revised her criteria in defining the Matrix language throughout the developments of her models because of the criticisms that were put forward by other scholars.

157 According to James Smith (2002: 45) Spanish-English code switching in the US exemplifies the alternational type of code switching whereas Swahili-English code switching in Kenya exemplifies insertional code switching. 239

4.2.2.1.1 The Language Affiliation of the Verb (Klavans (1986) and Treffer Daller (1990))

Early insertional approaches (Klavans, 1985; Treffer Daller, 1990) define the Matrix language (hereafter ML) as the language of the verbal stem in the sentence. This claim proved to be inoperative in bilingual contexts where speakers incorporate verbal stems from the embedded language into host languages. This is particularly true for some languages such as the agglutinative158 ones in which such process of incorporating host language verbal and nominal stems is very productive. Some scholars (Backus 1996, 2002; Muysken, 2000) relate this tendency to the morphological typology of these languages in which generating new lexemes is obtained using combinations of existing lexemes with inflectional and derivational affixes. Counter examples to Klavans‟ (1985) and Treffer Daller‟ (1990) claims are not drawn from bilingual situations involving agglutinative languages only, but also from bilingual situations involving other language typologies. Clyne (1992: 31) argues that:

“Klavans‟ (1985) claim that all sentences can be assigned to a matrix language according to the language affiliation of the verb is not practical in Dutch- English code switching spoken by Dutch emigrants in Australia” (Clyne, 1992: 31) Berber may be considered as an agglutinative language; thus, finding counter evidence to the claim put forward by Klavans (1985) and Treffer Daller (1990) in relation to the identification of the Matrix language may be quite an easy task. The data that we have investigated displays a big number of Verbal and nominal stems from either Algerian Arabic or French with inflectional and derivational morphemes from Berber. The number of instances involving such patterns of code switching is not negligible. We, thus, may consider them as counter evidence to Klavans’ (ibid) and Treffer Daller’s (ibid) criterion. Here are some instances from the data:

/-  --/

158 This process of incorporating host language verbal and nominal stems seems to be productive in agglutinative languages. Some scholars (Backus 1996, 2002; Muysken, 2000) relate this tendency to the morphological typology of these languages in which generating new lexemes is obtained using combinations of existing lexemes with inflectional and derivational affixes. 240

(Talk- 1st sing accusative object aff a bit about-boyfriend)

(talk to me a bit about your boyfriend)(Berber-Algerian Arabic code switching)

/ -  -/

(I do- 1st sing accompl aff only the sauce)

(I prepared the sauce only)(Berber-Algerian Arabic code switching)

/ --- - - -./

(We 1st plu inacomp aff-play-3rd obj accusative aff 1st plu inaccompl aff- study for- nom aff- truth)

(We were play it, (as if) we study for true)(Berber-Algerian Arabic code switching)

/ /-derang-//

(Now 1st plu obj accusative aff- disturb- 3rd plu inaccompl aff)

The verbal stems in these code-switching instances are either Algerian Arabic, or French (as for the last example of the above list in which the verb ‘derang’(disturb)is French) and their inflectional, derivational and clitic morphemes are Berber. We would then argue that the Matrix language is supposed to be Algerian Arabic for the first three examples and French for the fourth example. This does not seem to be the case in the sense that the verbal stems / / (talk), // (do), // (play), and ‘derang’ (disturb) are respectively Algerian Arabic and French code switchings159 which have been inserted into a Berber Matrix

159 These verbal stems are in fact Algerian Arabic borrowings, but we have already stated above that borrowing and code switching forms are treated in the same manner from the morpho-syntactic point of view. By arguing so, we are in fact taking the stand of many researchers in code switching and borrowing who avoid getting into the thorny and confusing issue of the differentiation between code switching and borrowing. 241 frame. This means that it would be more logical to hypothesize that Berber is the Matrix language in these examples.

The second process described by Muysken(2000: 64) whereby mixed verbal constructions are made up using verbal stems from the guest language combined with auxiliary verbs from the host language does not seem to be as productive as the above-cited process. We still could find some instances of such switching in the data. Still, it would be problematic to argue that the Matrix language is the one of the verbal stem in Berber-Algerian Arabic and Berber-French mixed constructions. Here two instances from the data:

/  --/

(By the way, you must go.)

The reverse process has occurred in this instance of code switching. The mixed compound verb is made up of an Algerian Arabic modal verb //(must) combined with a Berber verb constructions in which the verbal stem is Algerian Arabic and the aspect marker and the derivational and inflectional affixes are Berber. We have noticed the regular occurrence of the Algerian Arabic modal// (must) in the present study. We have encountered 56 tokens of this modal in a total number of 231 modals. The Berber speakers that we have investigated seem to use a somehow big number of modals among which the above cited modal, //(may be)

Ça va /- - /

(We are fine be- 1st sing inaccompl aff 1st plu inaccompl aff-learn)

(Everything is fine we are revising fine.)

242

4.2.2.1.2 The Discourse Criterion (Berk Selingson, 1986)

Other scholars (Berk Selingson, 1986: 323) use a discourse approach to identify the Matrix language. She (ibid: 323) claims that the Matrix language is the language of the whole discourse or conversation. Conversational analysts (Auer, 1998: 78; Gafaranga and Torres, 2001: 201, Gafaranga, 2005: 281; Li Wei, 2005: 379) rejected this claim. Gafaranga(2005: 282) argues that conversations are dynamic constructs in which interlocutors often negotiate on the language to be used in terms of macro social and conversational micro outcomes160i.e. that it is often difficult to define the ML in some bilingual conversations where speakers code switch languages quickly.

Another criticism to the claim put forward by Berk Selingson (1986) relates to the concept of code switching as a marked choice whereby bilingual speakers use a code switched variety as the unmarked maxim161 of conversation. Auer (1998a: 13) also points to the fact that arguing that the Matrix language is the language of the conversation or the one of discourse implies or presupposes that conversations are constrained by stable social norms operating in uniform communities. This seems not to be true in most bilingual or bidialectal162 situations where contacts between the varieties being in use are rarely peaceful, and where there are social pressures.

160 As we have stated it in chapter one a whole model of conversational code switching has been devised by Myers Scotton(2001) on the social outcomes of using such or such variety in bilingual contexts (the Rationale model of code switching). Gafaranga(2005: 285) developed an approach somehow similar to Myers Scotton‟s rational model in the analysis of code switching by arguing that bilingual speakers construct conversational meaning(the only difference being that Gafaranga sees conversation as the sole at the micro level of the conversation.

161 We have already discussed briefly the three maxims of the Markedness social model set by Myers Scotton (1993b) among which we may cite code switching as an unmarked choice.

162Auer(1998: 12)discussed the investigation of bi-dialectal switching in Italy(the two varieties involved being Frisian and 243

4.2.2.1.3The Statistical Criterion (Myers Scotton, 1993; 1997)

Myers Scotton (1993b: 68) proposed a statistical approach to identify the Matrix language. She (ibid: 68) argues that the Matrix language is the one participating with more morphemes in the sample being analysed, but this sample need not be only one sentence in the sense that taking a single sentence as the sample is often misleading. Myers Scotton (1993b, 1997: 68) argues in this vein that:

“The ML is the language of more morphemes in interaction types including intrasentential code switching… the ML can only be identified in sentences containing code switching materials if sentences are considered as part of a larger sample…How large is large enough remains an unresolved issue”(Myers Scotton, 1997: 68)

Bentahila and Davies (1998: 31) seriously put into question this criterion in relation to conversations where the speakers move from a language to another several times. Identifying the ML in such situations is simply impossible on statistical grounds they (ibid: 31) argue. Below is an example where it would be difficult to identify the Matrix Language on statistical grounds:

/    -      --   - - -/ il a tellement grandi maintenant que je me rappelle même pas c’est fou

(There(in the picture) they were(the months) maybe eight months like this, or nine months, or seven months I don’t remember how much he was at that time he has so much grown up that I do not even remember it is crazy.)

244

Auer (1998: 14) criticized Myers Scotton’s criteria in defining the ML because of the ambiguity of the concept of the CP163. He (ibid: 14) states that working with spoken discourse make it difficult to postulate that the CP is the unit of analysis in the investigation of code switching. Auer (ibid: 15) argues that it would be better to take the utterance as the unit of analysis. The argument that he gives relates to the fact that spoken discourse contains many unfinished sentences.

Myers Scotton amended her definition of the ML several times because of the bitter criticism put forward by different scholars about an issue that is central to her model of code switching. As early as in 1995, she (1995, 1997, and 2002) has been changing her definition of the ML construct continuously throughout the developments of her Matrix Language Frame model (Hereafter MLF). Myers scotton (1997: 246) rejected her statistical criteria in her after word. She (2002: 62) later acknowledges that taking a discourse sample as representative of the language configuration within the community as a whole is often problematic. The issue here is that how large the sample is large enough remains an unresolved question up to now (Gardner Chloros, 2009: 101). Myers Scotton (2002: 64) also revised her definition of the Matrix language as a fully-fledged language. She (ibid: 66) claims that the ML is not a concrete entity but rather an abstract, theoretical construct as she puts it:

“The matrix language becomes the label for the frame providing the morphosyntactic structure for the bilingual CP. The Matrix language is not to be equated with an existing language; rather one should view the Matrix language as an abstract construct for the morpho-syntax of the bilingual CP” (Myers Scotton, 2002: 66)

163 Myers Scotton (2002: 55) defines the CP (complement Phrase) as the maximal projection of a lexical unit e.g. a verb, a noun, a preposition, or an adverb. The maximal projection of a CP may be represented as an S- maximal projection where the projection may be filled with the category Spec (specifier). Taking the CP as the unit of analysis instead of the clause or the sentence helps avoid the confusing problems of code switching within a single sentence. Myers Scotton (2002: 56) illustrates her argument with code switchings within a compound sentence where the first part would be from a language and the other one from another language, or switching within a complex sentence where the main clause is from a language and where the subordinate clause is from another language). The other advantage of using the CP as the unit of analysis is related to the status of extra sentential code switchings such as „What!‟ or „never!‟ that may be defined as CPs, which contain null elements. 245

This definition of the Matrix language presupposes that there should be a frame at all times in constituents of bilingual CPs. This is, however, not always the case. There are for example bare forms that are inserted without any syntactic frame. Here is an example:

/-/ stilo /- /

(Give me pen inaccompl aff- write glass/

(Give me (a) pen (that) writes (on) (the) glass) (Berber-Algerian Arabic code switching)

The code switched bare forms are stilo (pen) and // (glass in Algerian Arabic) which are inflected neither in Berber, nor in Algerian Arabic. They do not receive any inflections for definiteness, number. In addition to that the second bare form // lacks a preposition (on). We shall come back to these bare forms in a subsequent section where we will be tackling the problematic cases of the MLF and other insertional approaches to code switching.

4.2.2.1.4 The Verb Inflections Criterion (Treffer Daller, 1994; Boumans, 1998)

Other insertional approaches (Treffer Daller, 1994; Boumans, 1998) try to identify the Matrix language in relation to the inflections of the verb within the clause. Boumans establishes a different insertional model of code switching which is more dynamic in terms of the ML construct. Boumans (1998) basic principle is that the ML may be identified at the individual phrase or clause level and that this ML is not necessarily the one of the whole discourse. He (ibid: 130) further argues, that the inflections of the finite verb within the finite clause are indicative of the ML as he puts it:

“In code switching with languages marking Tense and/or Aspect of the verb, verbal inflection is a reliable indicator of the ML on the finite clause level.” (Boumans, 1998: 100)

246

This definition of the ML represents a clear departure from Myers Scotton’s (1997, 2002). Myers scotton (2002: 64) does not allow for any change of the matrix language within the CP; besides, she argues that:

“A change within the sentence would be ‘an extreme case’ not the usual state of affairs. In fact, there are very few corpora in the code switching literature in which the matrix language changes at all. Certainly many other researchers have found that the Matrix Language is a robust and stable construct in classic code switching.” (Myers Scotton, 2002: 64)

Boumans (1998) definition of the ML seems to be more suitable in the sense that it espouses better the dynamicity of bilingual language in which speakers may change ML several times within a single conversation. Nortier (1990:188), similarly, argues that the ML has to be differentiated from the base language. The ML operates, according to him (ibid: 188) at the sentence level while the base language operates at the level of the whole discourse or conversation.

The criteria postulated by Nortier (ibid: 188), Treffer Daller (1994), and Boumans (1998) share the fact that the ML may change in a single conversation. The difference between the criteria postulated by these scholars is related to the data that they have investigated which in fact reflect bilingual situations that are quite different; besides, there seems to be a clear correlation between language typologies, and the criteria that these scholars have postulated to identify the ML.

Another point worth noting in the approach advocated by Boumans (1998) is the concept of layered insertion164. The ML of a main clause may become embedded within the

164 One of the first allusions to layered insertions has been made by Nishimura (1986: 130) who points to the idea that within some code switching corpora, it is possible for the Matrix language to change, for example, several times within a single complex sentence.

247 subordinate clause in such type of insertions. This highlights the dynamicity of Boumans’ (ibid) approach in the sense that he demonstrates using his Dutch-Moroccan Arabic data that it is possible for the ML to change several times within a complex sentence. Boumans (ibid) goes even further and argues that this dynamicity may even be observed within a simple sentence. Boumans(ibid)is not the only scholar who made such a claim, Nishimura (1986: 130) illustrates this pattern of code switching using English-Japanese bilingual sentences in which a Japanese postpositional phrase (made up of a Japanese postposition and an English object of postposition) is inserted within an otherwise English sentence. Here is an example165 given by Nishimura (1986: 130):

I slept with her basement 

Locative

(Lit: I slept with her basement in) (I slept with her in the basement.)(English-Japanese code switching Nishimura)

Here is an instance of layered code switching from the data that we have investigated:

/--/une idée générale //le protocole

(Give us a general idea about the protocol)(Berber-French code switching)

In this instance of code switching the ML of the verb phrase /--/ (give us) is Berber. The direct object of the verb phrase is the French Embedded Noun phrase ‘une idée générale’ (a general idea). The ML within this embedded noun phrase is French, but the noun phrase is modified by the prepositional phrase // ‘le protocole’ (about the protocol) in which there is a function element// (about) and a French embedded noun phrase ‘le protocole’ (the protocol). The Matrix language in the prepositional phrase is Berber. This instance of code switching highlights a change of the ML within a simple sentence in the

165 In this example, the Japanese part is in Bold following Nishimura‟s words. 248 sense that in the beginning of the sentence, The ML is Berber, then within the embedded noun Phrase, the ML is French, and then there is a matrix turnover at the end of the sentence where the ML is Berber again. This instance of code switching may be considered as a case of layered insertion because the Prepositional phrase //le protocole (about the protocol) functions as an adjective phrase which modifies the noun phrase ‘une idée générale’ (a general idea).

We shall come back to the issue of layered insertions in the section where we will be trying to analyse instances of code switching that represent challenges to the insertional approaches to code switching. We will do this with a special reference to the data that we have analysed.

4.2.2.1.5 The Language Activation Psycholinguistic Criterion

Some scholars postulated that psycholinguistic criteria might be used to identify the ML. The argument that they give is that the ML is the language which is most activated by the speaker in his bilingual speech. This is related to the fact that it is also the language in which the bilingual speaker is most proficient. Muysken (2000: 66) points to the unreliability of proficiency as an indicator of the ML. Muysken (ibid: 66) rightly argues that the speaker may switch to the language in which he is less proficient at some point in the conversation. Gardner Chloros(2009: 103) also points to the fact that establishing a direct relationship between the language which is activated in the brain of the bilingual and the frame of the sentences he or she is using is very difficult166.

Identifying the ML167 seems to be a thorny issue on which scholars have not yet reached any consensus. Muysken (2000: 68) argues in this vein that:

166 Making the relationship between monolingual language processing in the brain and language performance is still at an experimental stage. Research in Neuro-science has not yet started to tackle this issue in bilingual speakers.

167 The issue of identifying the ML is not the only difficult source of difficulty in the face of researchers. Defining the ML as a construct is another issue in itself in the sense that up to now there is not any 249

“A generally valid criterion for defining a single matrix language for a sentence or a conversation is hard to find”

Presenting all the criteria postulated in the code switching literature falls out of the scope of the present study. We shall consequently consider that Berber seems to be the ML in the inner networks of the Berber speakers that we have investigated. There are some exceptions in the data, but these instances will be treated differently. We will automatically signal any change of the ML from Berber to the other languages being in use in the data i.e. Algerian Arabic, French, or Standard Arabic. This pattern again represents an exception rather than a common pattern of code switching in the speech of the Berber bilinguals that we have investigated168.

Identifying the matrix language is at the heart of the Matrix Language Frame model and other insertional approaches to code switching; besides, the asymmetry and opposition between this matrix language and the embedded language is another important component of the insertional model advocated by Myers Scotton (1993, 1997, and 2002). Asymmetry operates at different levels or with different hierarchies.

conventional way of defining this concept which is central to most insertional approaches to code switching.

168 The analysis of one of the recordings that we have made highlights a sociolinguistic situation whereby French is the Matrix language of a discourse given by a theatre actor who presented a monologue translated from Algerian Arabic to Berber that he played before a Berber audience. The audience of the conference was made up of Berber specialists of theatre plays. A debate followed the presentation. Berber regained progressively the status of ML along the debate.

250

4.2.2.2 The Different Hierarchies Operating within the MLF Model and its Subsequent Models

The insertional approach advocated by Myers Scotton and her associates in its original form and in its revised versions has continuously based on asymmetry. This asymmetry operates on different levels. This triggers different hierarchies between the languages or codes involved in code switching. Asymmetry also triggers hierarchies between the elements being code switched. The first hierarchy postulated by Myers Scotton and her associates is the one between the Matrix and the Embedded Languages.

4.2.2.2.1 The Matrix/ Embedded language hierarchy

The first hierarchy is the one between the Matrix and the embedded languages in bilingual language production. As we have been discussing this hierarchy above, prominence is given to the ML in most insertional models. Either the embedded language either participates with small elements being inserted in otherwise larger chunks from the ML, or it participates with combinations or sequences that consist of at least two elements that follow the syntactic requirements of this embedded language169. There are, consequently, three types of elements in most code switching corpora characterized by an asymmetry between the languages being in use. Two of them are entirely in one language or in the other (Myers Scotton, 2002: 58). These elements must be composed of at least two morphemes being in a dependency relationship. Below are the three different types of elements or islands

 Matrix language islands: These islands consist of morphemes from the matrix language only. These elements are well formed according to the requirements of the matrix language. Myers Scotton (1997: 78) argues that these morphemes must show internal dependency

169 these combinations of elements are called embedded language islands in the MLF model 251 relations (they must at least make up a maximal projection or a combination of maximal projections e.g. a noun with its modifier, a verb, and its object noun phrase. Matrix Language islands do not represent any problem in the analysis of code switching, for they are monolingual in the Matrix Language. Below is an instance of code switching from the data where there is a matrix language island:

/    / on dirait pas / /

(Look just how he is here we would not say at all (that’s) he)

(Just look how he is (in this picture) here we would not at all imagine (that’s) him) (Berber- French code switching)

The Matrix Language island in this instance of code switching is the complex sentence /    / (just look how he is here). This complex sentence is entirely in Berber. We may also say that this island is in fact composed of two Matrix islands: A verb phrase /  /(Lit: look just) and the subordinate noun clause /  / which contains a complementizer //(how) and a verb phrase / /(he is here)which is composed of a verb and a place adjunct.

 Embedded language islands: they are made up of elements that are exclusively from the embedded language. These islands function the same way as the matrix language islands in the sense that they are well formed according to the requirements of the embedded language. They must also consist of at least two morphemes in a hierarchical relation e.g. a noun with its modifier, a noun with its maximal projection (NP), a verb with its maximal projection (a verb phrase), an adjective with its maximal projection (adjective phrase), an adverb with its complement(an adverb phrase) a preposition with its object( a prepositional phrase). Even if the Embedded Language islands (hereafter EL islands) are well formed according to the requirements of the EL, they generally occur in ML environments170, and therefore remain

170 Some EL islands are inserted in a maximal projection set by the ML and are therefore in a dependency relationship with the elements governing the maximal projection e.g. a noun phrase from the EL being inserted in a VP from the ML will consequently under the control of the verb of this VP 252 under its control. These islands must show some kind of congruence with the requirements of the ML. Here is an instance of code switching showing an EL island from the data:

/-  / la carte mémoire que j’ai acheté cent cinquante mille /  - --/

(Lit: I have these (pictures) in the memory stick that I bought year this in-the-feast-small) (I have these pictures on the memory stick that I bought this year during the end of fast month feast) (Berber-French code switching)

The French EL island in this instance of code switching is the Noun Phrase ‘la carte mémoire que j’ai acheté’ (the memory stick that I bought) which is composed of a compound noun and its maximal projection which is made of the adjective clause que j’ai acheté cent cinquante mille.

 Mixed islands (Embedded+Matrix language islands): these islands are at the core of the MLF model in the sense that they best highlight the asymmetry as it is displayed in bilingual code switching. The order of elements in these islands is dictated by requirements of the ML. Functionally active morphemes (which are called system morphemes in the MLF and subsequent models) also come from the ML in these islands. Content morphemes may come from either language. As it has been stated above, these islands may be composed of EL islands that are inserted following a dependency relationship into larger ML islands. Below is an example of a ML+EL(a mixed) island from the data: / -- - --/

(How inaccompl aff- change-2nd sing aff nom aff-place inaccompl aff- go-2nd sing aff)

(How you change place (where)you go)(Berber-Algerian Arabic code switching)

One of the mixed constituent in the above instance of code switching is the VP /- - - / which is made up of an Algerian Arabic verbal stem ‘’

253

(change) inflected with /-/ a Berber derivational affix standing for unaccomplished aspect and // an inflectional affixes standing for 2nd person singular.

/--  -/

(Talk-to me a bit about-(your) boy friend)(Berber-Algerian Arabic code switching)

This instances of code switching is made up of a mixed (ML+EL) island composed of a Berber Verb Phrase -( talk-1st sing object accusative aff). This verb phrase is composed of an Algerian Arabic verbal stem //, a Berber null inflectional affix standing for 2nd sing imperative, and a Berber derivational accusative object clitic affix standing for 1st sing indirect object.

4.2.2.1.2The Uniform Structure Principle

Myers Scotton (2002: 8-9) claims that the occurrence of most components in bilingual corpora in general and the occurrence of the three islands listed above in particular is guided by the Uniform Structure Principle which stipulates that:

“A given constituent type in any language has a uniform abstract structure and the requirements of well formedness for this constituent type must be observed whenever the constituent appears. In bilingual speech, the structures of the Matrix Language are always preferred171, but some Embedded Language structures are allowed if certain conditions are met” (Myers Scotton, 2002: 8-9)

171 Emphasis is our. We have emphasized this very notion of predominance of the ML even in situations where EL islands are in use because this is what we have noticed in the data that we have investigated. We shall come back to this point in a subsequent section where we will be arguing that it is true that the structures and the requirements of the ML are most generally preferred in cases of mismatch between the ML and the EL in some instances of code switching. 254

4.2.2.1.2.1 Applying the Uniform Structure principle against the present data

We have noticed in the data that we have analysed that this principle is followed to some extent. We have noticed for example that the requirements of the ML and its patterns are generally followed in mixed constituents. This seems to be particularly true in cases where there is a mismatch between the requirements of the ML and the ones of the EL. Below is an instance of code switching where there is a mismatch (lack of congruence) between Berber (the ML of the sentence) and French (the EL in the same sentence).

// la différence // tu me dis le protocole TCP et IP

(Lit: Tell me what the difference between you tell me the protocol TCP and IP)

(Tell me what is the difference between the two protocols TCP and IP)(Berber-French code switching)

The lack of congruence or mismatch in this instance of code switching is in the mixed island (// tu me dis le protocole TCP-IP (between you tell me the protocol TCP-IP). At first glance, this mixed island seems to be ill formed according to the syntactic requirements of French that is the embedded language here. This mixed island is composed of a preposition and its maximal projection i.e. its object that is a noun clause in this case). The Prepositional phrase172 functions as a complement to the object ‘la différence’(the difference). The lack of congruence lies in the object of the preposition. In French, it is a noun, a noun phrase, or a subordinate clause functioning as a noun and starting with a subordinating conjunction:

E.g. quelle est la différence entre une belle petite fille et un adorable petit garçon?

(What is the difference between a nice little girl and a lovely little boy?)

172 The concept prepositional phrase is used here to mean a combination of a preposition and its maximal projection(a noun clause in this instance) 255

E.g. quelle est la différence entre ce que tu enseignes à tes enfants la maison et ce que tu leurs enseignes quand vous êtes dehors.

(What is the difference between what you teach your children at home and what you teach them when you are outside?)

In Berber, when the prepositional phrases functions as a complement of the object. It is made up of a preposition that is followed by a verb phrase, which is not introduced by any subordinating conjunction (Naït Zerrad, 1997: 119).

The syntactic requirements of the two languages differ in this mixed island. This is why the maximal projection of the preposition seems ill formed in French. It is however well formed according to the requirements of Berber. The Uniform Structure Principle seems to apply well to this instance of code switching in the sense that Berber syntactic requirements have prevailed on the ones of French. This highlight one the underpinnings of the approach advocated by Myers Scotton, which states that even if EL islands173 occur they still have to be congruent with the requirements of the ML(Berber in this case).

Instances of code switching showing both patterns of code switching (Berber-French and Berber-Algerian Arabic code switching) highlight to some extent that the data analysed in this study follow the requirements of the Uniform Structure principle. Algerian Arabic imperative verbs switching seems to be another evidence for the validity of the above-cited principle. We have observed, for example, that the informants that we have investigated seem to follow the requirements of Berber in relation to the gender inflectional morphemes attached to the verb in the imperative form174. Here is an instance from the data:

173 We may consider the Noun Clause „tu me dis le protocol TCP IP‟( you tell me the protocol TCP and IP) as a French EL island which is itself part of a larger mixed island, for it is linked to the preposition with a dependency relation.

174 Imperative verbs in Berber are attributed a neutral null gender. This is not the case in some varieties of Algerian Arabic. This is particularly true for Oran Arabic (the variety being in use alongside Berber, French, and Standard Arabic in the majority of the data that we have investigated). The syntactic 256

Mother: /  / -Manal!

(How like that talk oh-Manal! (A mother addressing her daughter)) (Berber-Algerian Arabic code switching)

Daughter: // mama!

(Step back! mum!)

The lack of congruence in this conversation is in the Algerian inserted verbs / / (talk!) and //(step back!) which are imperative verbs but which should be assigned gender affixes(plural in this case), but which in spite receive a Berber null neutral gender affix. Almost all imperative Algerian Arabic inserted verbs in the data receive an invariable null neutral gender affix175. We may put the hypothesis that this pattern of code switching may highlight a dominance configuration176 that we will be trying to verify all along our test of the MLF model and other insertional models on the data. Instances of code switching such as these are also in favour of the Uniform Structure Principle in the sense that Berber syntactic requirements have been preferred instead of Algerian Arabic requirements. The second hierarchy on which we may test the uniform structure principle is related to the system and content morphemes statuses and conduct. The next sub-section will focus on these different types of morphemes.

requirements of Oran Arabic require that there should be a masculine and feminine gender signalled using the two suffixes// (standing for singular feminine) and standing for singular masculine.

175 In the whole data that we have analysed, we could find only one instance of code switching in which there is an imperative Algerian Arabic inserted verb which kept its gender affix: /--  /(talk-fem aff-1ST sing object accusative aff a bit)(talk to me a bit!)

176 This dominance configuration seems to be for Berber (the so-called Matrix language in most of the data that we have investigated). This state of affairs may be considered as predictable, for Berber seems to be the unmarked language used by the informants of this study in their inner networks(The elicitation of the patterns of language use of the informants have shown that Berber is the language used in the social inner networks of the family and friends. Cf chapter 4)

257

4.2.2.1.3 The system/content morpheme Hierarchy

Myers scotton (2002: 9) argues that morphemes are not equal whether they are used in monolingual or in bilingual speech177. She (1997: 58) also argues that there are two types of morphemes (System and content morphemes). These morphemes operate on an asymmetrical basis in bilingual corpora. This asymmetry is related to the occurrence of system and content morphemes in mixed islands (mixed ML+EL constituents) which are the most important in the interpretation of code switching. Myers Scotton (2002: 15) argues in this vein that:

“All the participating languages may contribute content morphemes to bilingual CPs, but not all can contribute critical system morphemes”(Myers Scotton, 2002: 15)

The distinction between content and system morphemes is one of the most important principles of the MLF model and its subsequent developments (Myers scotton, 1993, 1997, 2000, and 2002). We thus find it important to discuss briefly these two types of morphemes and their oppositions in the light of the data that we have investigated.

4.2.2.1.3.1 Content Morphemes

Content morphemes usually include nouns, verbs, adjectives, and most prepositions. They are activated at the early conceptual level of the formulation of the message. These morphemes are activated as heads of maximal projection i.e. verbs are activated as heads of VPs; nouns are activated as heads of NPs; adjectives are activated as heads of AdjP; prepositions are activated as heads of the maximal projection PreP. Content morphemes may

177This argument is based on Garrett‟s (1990: 165) findings in relation to the psycholinguistic investigation of speech errors. Garrett (1990) could demonstrate that there are two types of morphemes (content and system morphemes). These morphemes have different statuses, and consequently behave differently in language production. Systems morphemes are part of the frame and content morphemes are part of the semantic and pragmatic load of the lexical message. 258 be from either ML or El in mixed bilingual constituents i.e. in ML+EL constituents or islands. Below are some instances from the data that we have investigated:

/ ---  -- / Zaina

(What inaccompl aff-2nd sing fem aff- tell-2nd sing aff you Oh- aunt-1st sing pos clit aff Zaina)

(What can you tell us aunt Zaina)(Berber-Algerian Arabic code switching)

/---  -  --  /

(Neg aff- 2nd sing-have-2nd sing aff what that-3rd sing mas-have a bit of which-3rd sing aff-bet)

(Lit:you do not have(any idea about) what that he has some(money), he bets some(some of his money)

(You never know the amount of money that he has, for he spends some of his money in betting and some of it for the house expenditures) (Berber-Algerian Arabic code switching)

/ -  ---/ betise

(What nom aff- appraisal when-2nd sing fem aff-do-2nd sing fem aff fault)

(What praise (you ask when you make a mistake)(Berber-Standard Arabic-Algerian Arabic- French code switching)

/ -.- -- /

(watch(it)-null imperative aff 2nd sing fem aff-annoy-2nd sing aff def-brain-1st sing possessive clit aff) 259

(Be careful you started to annoy my brain)( Berber-Algerian Arabic code switching)

The content morphemes in these instances of code switching are:

 The Algerian Arabic verbal stems // (tell) ,// (earn), //(do), .(annoy), (watch(it)), and (bet)  The Nouns  ( aunt), -(the brain), (appraisal), and betise(fault)

We have observed that switching of nominal and verbal stems represents a non- negligible percentage of the bilingual instances in the data that we have analysed. Verb stems (apart from the copula verbs) are prototypical thematic (semantic) role assigners while Nouns are prototypical thematic role receivers178. These two categories are the most reliable and unproblematic content morphemes across languages. We may hypothesize that this may be a reason why our informants switch verbal and nominal stems from Algerian Arabic and French so often. This does However, not infirm the hypothesis we put forward above that says that the big number of nouns, verbs, and to a lesser extent, adjectives switching are related to the morphological typology of Berber179.

4.2.2.1.3.2 System morphemes

While content morphemes assign or receive thematic roles, system morphemes180 do not fulfil any semantic or pragmatic function. These latter are also labelled functional elements (Muysken: 2000: 158) in the sense that they have a syntactic function within the

178 Thematic roles assignments have been used by Radford (1988) as theta roles that are the roles of people and entities in the clause (James Smith 2002: 34).

179 We hypothesized that the typology of Berber facilitates the penetration of verbal and nominal stems from Algerian Arabic, French and to a lesser extent Standard Arabic that are adapted to Berber morphology through the affixation of derivational and inflectional Berber prefixes, suffixes, and circular affixes.

180Bolinger (1968) was the first scholar to use the concept System Morpheme. This concept is, according to Bolinger, a cover term used for both inflections and function words. 260 clause e.g. definite and indefinite articles, quantifiers, plural affixes, gender affixes. Myers Scotton and Jake (2000: 1054) argue in this vein that:

“The characteristic properties of content morphemes should be largely self- evident; they convey the core semantic/pragmatic content of language. The three types of system morpheme carry the relational aspects of language” (Myers Scotton and Jake, 2000: 1054)

The hierarchy between the ML and the EL operates during the distribution of system morphemes in bilingual language production. This is an important reason why they are at the core of Myers Scotton’s (1993, 1997, and 2002) insertional approach to bilingual corpora. This claim is formalized by Myers Scotton (1997: 83) in her System morpheme principle, which stipulates that:

“In ML+EL constituents, all system morphemes which have grammatical relations external to their constituent (i.e. which participates in the sentence’s thematic role grid) will come from the ML.” (Myers Scotton, 1997: 98)

Myers Scotton and her associates refined the definition of system morphemes several times (1997, 2000, and 2002) because of the prominence of these morphemes in the interpretation of code switching and because of the strong criticisms and the counter evidence put forward in the code switching literature(Boumans, 1998; Muysken, 2000, Li Wei, 2000). Myers Scotton and Jake (2000) put forward the 4M model in order to elaborate the concepts of content and system morphemes. They (ibid: 1062) divided these into four different types of morphemes according to two criteria:

261

1. +/-conceptual activation181: Content morphemes are definitely conceptually activated, for they are accessed at an early stage. So are early system morphemes, which are differentiated from the other types of system morphemes because they are conceptually activated. We will come back to the early system morphemes below

2. +/-looks outside its maximal projection: This criteria is used to see if the system morpheme looks outside its immediate maximal projection or not. It is used to separate system morphemes that operate at the suprasegmental level of the CP and others that operate within the maximal projection.

Having presented the content morphemes above, we shall concentrate on how Myers Scotton and Jake (2000: 1062) present system morphemes182. Using the criteria listed above, Myers scotton and Jake (ibid: 1062) argue that there are three different types of system morphemes. The relevance of this new division is that it helps account for most bilingual corpora involving code switching and other contact phenomena. It also helps by pass the counter evidence183 put forward to the system morpheme principle as it has been expounded by Myers Scotton (1997: 98). The three different types of system morphemes are the following ones: Early system morphemes, bridge system morphemes, and outsider system morphemes. We shall now try to present the three types of morphemes in the light of the data that we have analysed.

181 This criterion is related to the way morphemes are accessed (Myers Scotton and Jake, 2000: 9) some morphemes are accessed at an early stage in language production i.e. simultaneously to the formulation of the semantic or pragmatic message or at the lemma level. This stage is called the conceptual level. Examples of these include definiteness. Others are on the contrary accessed at a later stage which is called the formulator stage or level

182 System morphemes have been given a paramount importance in insertional approaches to code switching. This may be related to the fact that they are more salient than the content morphemes at the syntactic level. This may also be due to the fact that defining system morphemes has been one of the biggest challenges to syntactic theory in general and to cross linguistic research in particular. Content morphemes seem not to be problematic as there is a relative consensus among contact linguists as to what constitutes a content morpheme and in relation to the conduct of content morphemes in bilingual corpora.

183 These counter evidence include for example cases of Embedded language definite articles that are inserted with their nouns, 262

Before we proceed in our discussion of the different types of system morphemes, it would be interesting to start by the general picture (in the form of a figure) of the 4M. This model as it has been expounded by Myers Scotton and Jake (2000: 1062) divides the different types of morphemes into content morphemes, early system morphemes, bridge system morphemes, and outsider late system morphemes.

Feature-based classification of morphemes in the 4M Model (Myers Scotton and Jake, 2000: 1062)

4.2.2.1.3.2.1 Early System Morphemes

These morphemes are the only system morphemes that are activated at the early phase of conceptual level i.e. at the early stage of the formulation of the semantic message; therefore, they behave differently from the two remaining system morphemes. Early system morphemes satisfy the criterion of conceptual activation together with content morphemes. 263

These two types of morphemes are associated to transform the speaker’s intentions into a semantic and pragmatic message. Early system morphemes do not however assign or receive any thematic roles. They are elected indirectly by the content morphemes that point to them (Myers Scotton and Jake, 2001: 98); besides, they are realized within the maximal projection of the head which elects them(the Noun in NPs , the Verb in VPs, the Adjective in AdjPs, the Preposition in PREP-Ps, and the Adverb in ADVPs). Early system morphemes include, for example, indefinite and definite articles, plural affixes, gender affixes, possessive adjectives, and prepositions related to verbs(preposition functioning as particles in verbal phrases. These morphemes map onto the form of the content morphemes, which elected them within the maximal projection. Myers Scotton and Jake (2000: 18) argue that under her 4M Model early system morphemes may come from both ML and EL as they put it:

“In the original statement of the model, the system-morpheme principle states that ‘‘all system morphemes which have grammatical relations external to their head constituent ... will come from the ML’’ (Myers-Scotton 1993, 1997: 83). Under the new 4-M model, this class of system morpheme is more explicitly identified as the late outsider system morpheme. While other types of system morpheme may come from the embedded language (EL184), in fact, almost all — not just those required by the System Morpheme Principle”(Myers Scotton and Jake: 2000: 1071)

4.2.2.1.3.2.1.1 Testing the Early System Morphemes principle against the present data

Below are some instances of code switching from the data where there are early System morphemes. These early system morphemes are not only from the Matrix Language (Berber in many instances) but also from the embedded one (Algerian Arabic, French, and Standard Arabic)

/ ----   - /

184 Emphasis is our. This part of the quotation shows that the 4M model permits the occurrence of EL early system morphemes and other system morphemes that are elected inside the maximal projection. 264

(And-fem nom aff-teacher-fem nom aff (AA)-fem nom aff (Ber) what she told day-this)

(And(your)teacher what did she tell you today)(Berber-Algerian Arabic code switching)

In this instance of code switching, the mixed constituent /----/ (and (your teacher)) contains a content morpheme // (teacher) and two early system morphemes the Berber circular affix /…/ that is a feminine nominal affix in Berber(The ML in the mixed constituent), and the suffix // which is a feminine nominal affix from Algerian Arabic(the EL in this mixed constituent).

/  - - - /-la pommade

(Not over take-1st sing accompl aff nom aff-medicine take-1st sing accompl aff nom aff (BR)-def (Fr)- pomade)

(Isn’t it over I took the medicine; I took the pomade)(Berber-Algerian Arabic-French code switching)

In this code switching instance, the mixed constituent /- /-‘la pomade’ (I took the pomade) is a verb phrase and its maximal projection (the object NP /-‘la pommade’ (the pomade). The NP mixed island is composed of two early system morphemes: // which is a Berber nominal affix and ‘la’ (the French definite Article)

/-- . ---- -- - /

(And hurt-3rd sing accompl infix-1st sing accusative obj clit aff a lot causative aff-show-1st sing accompl aff-it (accusative obj affix) to-def(AA)-teacher-nom fem aff(AA) )

(And it hurted me a lot I showed it to my teacher)(Berber-Algerian Arabic code switching)

265

The Matrix language of this example is Berber. This instance of code switching contains two mixed constituents: /-- ./and-it was hurting me a lot) and /-- ---/. We will concentrate on the second mixed constituent or island /-- ---/. This island is composed of a VP maximal projection. This projection is made up of two content morphemes the Berber verb // (show) and the Algerian Arabic noun  (teacher). There are two early system morphemes in this projection. These are namely the Algerian Arabic definite article // and the Algerian Arabic feminine nominal affix //. These early system morphemes are both from the EL (Algerian Arabic)

The three instances of code switching presented above highlight the assumption that early system morphemes may be from both the ML (Berber in this case) and the EL (Algerian Arabic and French).

4.2.2.1.3.2.2 Late System Morphemes

Early system morphemes got their metaphorical name from the fact they are accessed or elected early in the process of language production. Late system morphemes are comparatively accessed at a later stage during speech production process. These morphemes are not accessed during the conceptual phase of the speaker’s intention, but during the formulator phase. This phase is the second phase in language production.

Late system morphemes obey different requirements from their early counterparts. They are not activated conceptually as the early system morphemes. They do not convey any thematic pieces of information, but they rather convey syntactic relations. Myers scotton and Jake (2000: 1063) argue in this vein that:

“The information contained in late system morphemes is grammatical as opposed to conceptual. The two types of late system morpheme are not elected to complete a semantic 266 and pragmatic feature bundle with their heads; rather, they are structurally assigned to indicate relations between Elements when a larger constituent is constructed”(Myers Scotton and Jake, 2000: 1063)

Late system morphemes are accessed (elected) at the phase of the formulator in the sense that the pieces of lexical information needed to build them are sometimes not available until the formulator level. Myers Scotton and Jake (2000: 1064) further divided these system morphemes into two types. This division relates to the criteria ‘+/- looks at its maximal projection’ for its syntactic requirements. This means that first type of late system morphemes (outsider system morphemes) have to look outside of their maximal projection to have the form that they are going to take and the lexical information that they need to be produced. The second late system morphemes (the bridge system morphemes) look inside their maximal projection in order to have their form and the lexical pieces of information needed for their production.

4.2.2.1.3.2.2.1 Bridge Late System Morphemes

Bridge system morphemes’ forms and syntactic requirements depend on the lexical and syntactic information within their maximal projection. This feature is shared with early system morphemes. The difference between these two types of morphemes is related to their syntactic requirements. Early system morphemes such as articles are directly elected by their heads; consequently, their syntactic requirements are the ones of their heads. The syntactic requirements of bridge system morphemes are decided later and they are the ones of the maximal projection not the ones of the heads of the maximal projection.

The function of bridge system morphemes is to join content morphemes to each other within the maximal projection (Myers scotton and Jake, 2000: 1064). Examples of bridge system morphemes discussed by Myers Scotton and Jake (ibid: 1065) and by Myers Scotton (2002: 75) include the possessive ‘of’ in English, and the genitive‘s, the French ‘de’. The preposition ‘of’ is used to link two adjacent nouns and it is in no way under the control of any 267 of them. One of the nouns being the head and the other one being the subordinate, this subordination needs to be signalled syntactically using either ‘of’ or the genitive case

E.g. the students of the class, my father’s friend (the friend of my father)

Bridge system morphemes may be from either the ML or the EL as long as they are within the maximal projection. Myers Scotton and Jake(2000: 1070) argue that the only system morphemes that are necessarily from the ML are the outsider late system morphemes185.Below are some instances of code switching to illustrate this discussion of bridge system morphemes

4.2.2.1.3.2.2.1.1 Testing the Prediction of the Bridge System Morpheme Principle against the Present Data

/---  ---/186

(def-news-plu infi of-fem nom aff-Kabyle-fem nom aff)

(Lit: the news of Berber)(Berber news)(Berber-Algerian Arabic code switching)

/ - -/

(daughter of-aunt-1st sing possessive aff)

(Lit: Daughter of aunt my)( my cousin)

185Myers Scotton and Jake (2000: 1070) claim in this vein that: “Under the new 4-M model, this class of system morpheme is more explicitly identified as the late outsider system morpheme. While other types of system morpheme may come from the embedded language (EL)”

186 This instance of code switching and the one following it are in fact borrowings, as it has been stated above the morpho-syntactic treatment of borrowings and code switching is similar. 268

//-la sauce /--/

(Nom affi- the sauce of- me/

(Lit: the sauce of me)(my sauce)(Berber-French code switching)

Les toilettes /- /

(Lit: The toilet of- house this)

(This house’s toilet)

The matrix language in these examples is Berber. The content morphemes in these examples are the nouns // (Algerian Arabic piece of news), // (Kabyle), // (daughter), // (aunt), and ‘sauce’ (sauce). The early system morphemes are the Algerian Arabic definite article // (the) in /--- (the news), the feminine nominal circular affix /…/ in /---/ (Kabyle), // the 1st person singular possessive clitic affix in -(Aunt my), the nominal affix// in //-la sauce(the sauce), and the French plural definite article ‘les’(the) in ‘Les toilettes’(the toilet’. The bridge system morpheme in these examples is the Berber preposition // and // (of). These prepositions are used to link the above-cited content morphemes (nouns). These examples highlight a code switching pattern whereby the Berber bridge system morphemes //(of) and //(of) are used to link an EL noun with an ML one. This possibility is not the only one in ML+EL constituents. Myers Scotton (2002: 80) discusses another possibility in which an ML late bridge system morpheme is used to link two ML content morphemes (two nouns for example). Here are some examples from the data of this study whereby the Berber preposition // (of) is used to link two nouns ( a head noun and its subordinate or complement noun)

Les vacances //-. /

(Lit: The holidays of the summer)(Summer holidays)(French-Berber-Algerian Arabic code switching)

269

/  -/

(Lit: The mentality of person)(the person’s mentality)(Berber-Algerian Arabic code switching)

L’examen --/

(Lit: The exam of-Arabic)(The Arabic exam

The investigation of the data shows that there are also some other late bridge system morphemes that are used to link verbs, to link modals with verbs and to link verbs with their objects within VPs maximal projections. These late system morphemes are for example used to link a modal with its lexical verb. Below is an example which highlights this state of affairs.

 -  - Melissa

(Look! 3rd sing fem aff-like(want) to 3rd sing fem aff- talk Melissa)

(Look! She wants to talk (to you) Melissa)

In this example // may be considered as a late bridge system morpheme in the sense that it is used within the maximal projection of the VP to link the verbs // (like or want) and // (talk) which are both content morphemes187 within this mixed VP maximal projection where Berber seems to be the Matrix language.

/ --  -- /

(if 3rd sing accusative object clit aff-tell-1st sing affi now to-def-teacher/

187 We may also assume that in some contexts such as the one of this example, // is used as a modal with the meaning „want‟. It may consequently be considered as a system morpheme that is used with a lexical verb. 270

(Lit: if to her I tell now to the teacher)(If now I tell to the teacher)

One of the components of this mixed VP maximal projection where Berber is the ML is /-/which functions as the object to the verb // (tell). //(to), which is a bridge late system morpheme is used to link the Berber verb// (tell) with its Algerian Arabic object/-/ (the teacher). A similar constructions exists in Algerian Arabic whereby the Algerian Arabic preposition // (to) is used to link the verb // with its object when this object is a person. Here is an example from Algerian Arabic:

- - -

(Tell to-that-mad (person)

Some instances of code switching are problematic in the sense that they lack bridge system morphemes. In the examples below the bridge system morpheme // (of) is missing

- -

(Religion -the race) (By the religion (of) race)

//-Bon / /

(The invoice (of) the-route)

We may consider these examples as bare forms188, for the omission of the bridge system morpheme neither follows the syntactic requirements of Berber (the Matrix language of the two instances), nor does it obey the requirements of Algerian Arabic or French (the EL languages in the two instances respectively). There is no mismatch in the syntactic

188 Bare forms are code switched elements that lack system morphemes such as inflections derivations. These bare forms are generally accessed as content morphemes. WE shall come back to the interpretation of these forms below.

271 requirements of the three languages. The hypothesis, which states that bare forms are accessed when there is a lack of congruence between the languages involved in code switching, does not hold to these two instances.

4.2.2.1.3.2.2.2 Outsider Late System Morphemes

The 4M model (Myers Scotton and Jake, 2000) stipulates that late system morphemes are divided into two types. The criterion used to differentiate them is +/- looks at its maximal projection for its form and grammatical requirements. Outsider system morphemes are elected during the last phase preceding language production. Their function is to integrate maximal projections into larger units with the highest units being the CP or a sequence of CPs. Myers Scotton (2002: 79) argues in this vein that:

“Outsider System morphemes also integrate content morphemes and X- projections into larger constituents but contrast with bridges in regard to the source of information about their form. For outsider system morphemes, this information (usually) is not available until the highest-level projection, the CP, is assembled. Outsider late system morphemes perform the arguably more important function of showing co-indexical relationships across maximal projection189s.” (Myers Scotton, 2002: 79)

The syntactic information needed to produce late outsider system morpheme operate at a higher level than the one of the maximal projection in which they occur. These pieces information are used to unify the maximal projections together. An example would be the grammatical and lexical information used to unify two clauses together. One of these pieces of information would be the logical relationship holding between the maximal projections of these two clauses. This information is needed to access different types of complementizers. This represents a higher order assembling procedures. It is also possible to move to lower order procedures. Myers Scotton and Jake (2000: 1064) discuss the example of outsider

189Emphasis is our 272 system morphemes used to assemble subject NPs with their VPs; examples of outsider late system morphemes would be verbal inflections attached to the verb to co-index number, Gender and other features.

4.2.2.1.3.2.2.2.1 Testing the Outsider Late System Morpheme Principle on the Present Data

Late system morphemes in Berber include affixes that are attached to the verb to signal number, gender, aspect. Other types of outsider late system Clitic affixes attached to the verb to signal indirect objects, direct objects and the like as these affixes are used to link VPs maximal projections with the object NPs. Myers Scotton and Jake(ibid: 1071) argue that under their refined system morpheme principle, late system morphemes are obligatorily from the ML.

The data that we have investigated indicate that this prediction applies relatively well to Berber-French, Berber-Algerian Arabic, and to Berber-standard Arabic code switching. Below are some instances from the data where Berber is the ML language and where Algerian Arabic, French, and Standard Arabic are the EL languages. In some instances, however, there is a Matrix language turnover where, for example, EL languages of the data become the ML languages and where Berber becomes the EL. below are some instances from the data where displaying outsider late system morphemes.

/ -- /

(Like third sing obj accusative aff- do-1st sing accompl aff before)

(Like I did to her yesterday) (Berber-Algerian Arabic code switching)

273

In this mixed CP190 the late outsider morphemes are the Berber complementizer // (like) and// the Berber object accusative clitic affix both of them have to look (for their form and grammatical requirements) outside the maximal projection in which they occur.

/- -- -/

(3rd sing fem aff-say when 3rd sing accompl aff-come to-me)

(Lit: She said when she came to me)(Berber-Algerian Arabic code switching)

The outsider late system morpheme in the mixed CP /-- -/ is the Berber complementizer //. This complementizer is used to link the mixed CP, which functions as a noun clause, to its main clause /-/ (she said)

/--  -  --    - -- /

(inaccompl aff-jump-2nd sing aff to here then inaccompl-jump-2nd sing aff to there like-that Neg aff-learn-2nd sing aff nothing)

(Lit: You jump to here then you jump to there like that you will not learn nothing)

((If) you jump from here to there like that you do not learn anything)

This sequence of CPs (mixed complex sentence) contains three CPs. The first one is/--  -/ (you jump to here). The second CP is / --  / (then you jump to there). The outsider late system morpheme in this Algerian Arabic mixed CP is the Berber complementizer // (then).

190 The CP means the complementizer projection. This roughly corresponds to a clause with all its maximal projections. Myers Scotton (2002: 54) uses the CP as the unit of analysis to avoid using the sentence as the unit of analysis. She (ibid: 54) argues that using the sentence as the unit of analysis is confusing especially in complex sentences which by definition contain more than one clause. Myers Scotton (2002: 54) also uses the Cp instead of sentence to avoid getting in the trap of intersentential code switching. 274

The third CP is /- -- / (like that you do not learn nothing) the outsider system morphemes in this CP are the Berber negative affix // and the Algerian Arabic complementizer/-/ (like that). The outsider system morpheme (the complementizer) introducing the second CP does not look inside the maximal projection in which it occurs i.e. within the VP maximal projection /--  / (you jump to there) to have its form and lexical requirements. It rather looks at the logical relationship holding between its maximal projection and the one preceding it i.e. a logical relationship of chronological order. The second outsider late system morpheme (the complementizer /-/ unifies the whole argument stated in the first and the second CPs and relates the third CP to them with a relationship of consequence.

/--  -- -- /- les toilettes

(And-this when-2nd sing fem aff-like(want) inaccompl aff-enter-2nd sing aff nom aff-toilet)

(And this (prayer you say it) when you want to enter the toilet)(Berber-Algerian Arabic-French code switching)

The Matrix language in this mixed CP is Berber. The three content morphemes are The Algerian Arabic verbal stem // (like or want in this case, the Berber verbal stem //(enter), the French noun toilettes (the toilet). The early system morphemes are the Berber demonstrative pronoun // (this), the Berber nominal affix // attached to the French code switched noun ‘toilettes’, the French definite article ‘les’191(the) and the French plural affix /s/. The outsider late system morphemes are:

1. The Berber coordinating conjunction // which is not a free morpheme but a clitic in Berber.

191 Unlike English articles that are neutral, French definite articles sub-categorize for Gender and number 275

2. The Berber complementizer //(when) which is used to introduce the Berber-Algerian Arabic-French CP / -- -- /- les toilettes/(when you want to enter the toilet). 3. The Berber unaccomplished marker // attached to the verb (enter). This affix signals unaccomplished case in Berber192. 4. The circular inflectional affix /… /attached to the content morphemes //(Algerian Arabic want to) and //(enter). This circular affix stands for 2nd person singular feminine conjugation.

/ ---- --/

(of def-alive def-mighty 1st plu inaccompl aff- read-3rd sing )

(Lit: (The Koran Verse) of the alive and the almighty we read it)

(Shall we read the Koran verse of the all alive and the almighty)(Berber-Algerian Arabic- Standard Arabic code switching)

This mixed CP contains a VP maximal projection /-/ (We read) and a NP maximal projection / ----/ (of the all alive the almighty). The VP /--/ is composed of:

1. The Algerian Arabic verbal stem //(read), a content morpheme 2. The Berber 1st person plural unaccomplished inflectional affix // 3. The Berber 3rd person singular feminine accusative object clitic affix, an outsider late system morpheme. This morpheme is an outsider late system morpheme as it looks outside of the maximal projection in which it occurs (The VP maximal projection) to have its form and grammatical requirements. /  ----/

(Look! Look how 3rd plu aff-1st sing obj accusative aff-3rd plu aff-hurt-3rd plu inaccompl aff)

192 In this context // is used to introduce an aspect which resembles infinitive in Berber, but in which the verb receives inflections for gender and number. Verbs in Berber have got a functioning which is somehow similar to Algerian Arabic in which verbs when following the modal want to (like to) are assigned gender and number e.g. /  --/ (Lit: He wants he jokes). 276

(Lit: look! Look how they me hurt)

(Look! Look how my legs are hurting me!)

This complex sentence is made up of a main CP //(look) and/ -- --/(how my legs are hurting me) which is a subordinate CP which functions as a noun clause. This subordinate CP is composed of

1. Two content morphemes: the Berber verb //(look) and the Algerian Arabic verbal stem //(hurt) 2. The outsider system morphemes // (how) the Berber complementizer, the 3rd plural clitic pronoun //, the 1st person object accusative clitic affix193 //, and the verbal circular affix /…/ which stands for 3rd person feminine plural in Berber.

This type of code switching represents a clear violation to the constraints of the equivalence model of code switching as it has been advocated by Poplack and her associates (1982, 1988)194. It however is normally interpreted under the predictions of the MLF approach. This may be considered as an evidence of the practicality of using an insertional approach based on the idea of asymmetry between the languages in the data under study here. The content morphemes in this instance, for example, are from both Berber and Algerian Arabic, while the outsider system morphemes are from Berber only.

/-- -  - --- /

193 Myers Scotton (2002: 80) discusses similar cases of accusative case affixes in Hungarian. She similarly argues that these affixes are outsider system morphemes par excellence.

194 Examples such as this one involve a pattern of code switching whereby a Berber object clitic prefix is attached to an Algerian Arabic code switched verb. We have claimed above that this represents a violation to both the equivalence constraint and the free morpheme constraint as it has been expounded by Poplack and her associates (1982, 1988) in the sense that Algerian Arabic verbs do not subcategorize for a Berber clitic object accusative affix in a pre verbal position. See the discussion above on the violations to the equivalence constraints.

277

(And-fem aff -this want-1st sing accompl aff I fem aff-this Neg aff-learn-1st sing accompl aff- Neg aff)

(Lit: and this want I this did not learn)

(And I want (to learn) this verse, I did not learn (it)) (Berber-Algerian Arabic code switching)

The mixed CP /- ---/ (this I did not learn it) is composed of a two content morphemes i.e. the Algerian Arabic code switched verbal stem // (learn) and the demonstrative adjective //(this).

There is only one early system morpheme in this mixed CP, the feminine affix // which is attached to the demonstrative adjective195. The outsider system morphemes are the Berber negative discontinuous affix /…/ and the Berber verbal 1st person singular accomplished affix //. The outsider system morphemes of this Berber-Algerian Arabic mixed CP are all Berber.

There seems to be a violation in this CP as there is a complementizer missing. The predictions of the MLF and 4M models stipulate that the occurrence of ML outsider system morphemes is obligatory in bilingual corpora. The complementizer, which should be used in this context, is the Berber complementizer // (that). We have observed the dropping of the complementizer ‘that’ in the data that we have been analysing196. There seems not to be any lack of congruence between Berber and Algerian Arabic in relation to this complementizer (the equivalent in Algerian Arabic being // (that)). We may hypothesize that this may be a feature of spoken discourse.

195 The demonstrative adjective in Berber subcategorize for two different affixes standing for Gender//(standing for feminine) and //(standing for masculine)

196 Bouamrane (1986, 1988) and Benali Med (2007) observed the same process of complementizer dropping in general and the one of „that‟ dropping in particular during their analysis of bilingual data in Algeria.

278

Both Monolingual and bilingual speakers sometimes drop features of their languages or use weak or contracted forms by means of economy. The dropping of complementizers is one of these features. We may hypothesize that this may be considered as a discursive strategy used by the bilingual speakers under interest in this study. This remains a hypothesis that needs to be verified using a quantitative investigation of the use of complementizers in Berber- Algerian Arabic code switching corpora in Oran.

The instances of code switching analysed above seem to obey the principles advocated by Myers Scotton and Jake (2000, 2001) and the ones advocated by Myers Scotton (1997, 2000, 2002) under her MLF model ands its subsequent 4M model in relation to the distribution and behaviour of content and system morphemes197. Some instances of the data that we have analysed, however, seem to violate the predictions198 of the 4M in relation to the form, grammatical and lexical requirements of the Outsider system morphemes. This will be the focus of the next sub-section.

4.2.2.3 Problematic Cases to the MLF and the 4M Models with a Special Reference to the Data

AS it has been stated above, outsider late system morphemes are supposed to be from the ML only (Berber in most of the CPs of the data). This is one of the most important premises of Myers Scotton’s insertional approach in addition to being one of its latest developments. In

197The different refinements and changes brought by Myers Scotton (1997, 2000) and by Myers Scotton and Jake (2001) to the definition and classification of different system morphemes may be considered as a response to the strong criticism and to the counter evidence (Boumans 1998; Muysken 2000; Gardner Chloros 2009) to the MLF model in general. These refinements are also a response to the criticisms in relation to the way this model i.e. interprets functional elements. Functional elements being at the centre of most competing insertional models, Myers scotton herself tried to be as exhaustive as possible with this aspect of code switching data. This, however, does not prevent even the latest developments to Myers Scotton model to show weaknesses in the interpretation of some bilingual corpora.

198 These predictions state that the outsider late system morphemes are elected from the ML language only. The argument put forward by Myers and Jake (2002: 1063) is that these morphemes are elected at the latest phase in language production (the formulator phase). The requirements of these morphemes are, consequently, the ones of the ML in the sense that they intervene during the phase of uniting Maximal projections, CPs, or sequences of CPs together.

279 the present sub-section, we will test this prediction on more instances of code switching from the data.

/--  - !/ (Talk-2nd sing imperative aff-1st sing obj accusative clit aff a bit about-boyfriend!) (talk to me a bit about(your boy friend)(Berber-Algerian Arabic code switching)

Berber is the Matrix language of the mixed CP /-- /(talk to me a bit). The first outsider system morpheme in this CP is the 2nd person singular imperative Algerian Arabic affix //. This affix is part of the VP maximal projection/-/(talk!). The second outsider late system morpheme is the Berber object accusative clitic affix// which also occurs within the VP maximal projection. The violation to the System Morpheme principle in its refined version199 is the occurrence of //(an outsider late system morpheme) from Algerian Arabic. Berber subcategorizes for a null neutral gender affix for the imperative case e.g. (look!). This is not the case for this instance of code switching, for the verb receives an Algerian Arabic feminine affix for imperative. The second violation is the bare noun // (boyfriend). This noun receives gender, number assignments neither from the ML (Berber in this context), nor from the EL i.e. Algerian Arabic from which the bare noun has been inserted.

4.2.2.3.1 Possible Problematic Instances in Relation to Negative Marking

Switching of negative markers within the CP has always been considered as one of the most challenging problems to the insertional approaches to code switching. Negative markers, negative affixes, and negative words are system morphemes par excellence. These

199 The 4M model (Myers Scotton and Jake, 2000; Myers Scotton, 2002) is the refined version of the system morpheme principle (Myers Scotton, 1993, 1997) in the sense that this model (the 4M model) subdivides the system morphemes in a more precise manner into early system morphemes, late bridge system morphemes, and outsider late system morphemes. This model also predicts that only late outsider system morphemes must be from the ML. 280 morphemes are elected at a late stage during language production. Below are some instances showing different patterns of code switching involving negative system morphemes.

/ - -- /

(I swear Neg (AA)-learn-1st sing accompl aff-Neg aff a bit)

(Lit: I swear I did not learn somehow) (Algerian Arabic code switching)

IN this instance of code switching, at first sight, Berber seems to be the Matrix language of the mixed CP /- --  / (did not learn somehow) which contains a VP maximal projection /- -- / (did not learn somehow).

The content morpheme of the maximal projection is the Algerian Arabic code switched verbal stem //. The early system in the VP maximal projection is the adverb// (a bit). The outsider system morphemes are the Algerian Arabic negative affix //, the Berber negative affix /-/, and the Berber 1st singular accomplished inflectional affix //. This VP maximal projection seems to violate Myers Scotton (2002: 79) and Myers Scotton and Jake (2000: 1064)) prediction stipulating that outsider system morphemes bilingual data are only from the ML (Berber in this context). The negative affix preceding the verb is Algerian Arabic.

This pattern of switching whereby Algerian Arabic negative particles are attached to verbs having Berber inflections standing for number, gender, and aspect is not exceptional. We made a calculation of the negative constructions in VP maximal projection in the data, and we found that out of 364VPs constructions 122are negative ones. 19.7% of these highlight the pattern that we have just discussed (AA negative affixes that are embedded into an otherwise Berber pattern in which the remaining system morphemes in the Verbal construction are Berber).

281

Some scholars (Backus, 1996: 145; Boumans, 1998: 78; Muysken, 2000: 158,) call this type of bilingual data ‘double morphology’ that occurs when the morphologies of the systems participating in code switching overlap or resemble each other. We may hypothesize that this may be an interpretation for this violation to the 4M model. This hypothesis seems to be strengthened by the fact that we found no overlapping between Berber and French in the data in relation to Negative markers attached to verbs200.

/--   -- /-sujet (Neg-remember-1st sing accompl aff-Neg I swear Neg(AA)- remember-1st sing accompl aff about-subject) (I do not remember, I swear I do not remember about the subject.)(Berber-Algerian Arabic code switching)

In this instance of code switching where Berber seems to be the ML, there are two mixed CPs. The first mixed CP is /-- / (I do not remember). The second mixed CP is / -- /-sujet (I swear that I did not remember about the subject). The content morphemes in these two CPs are The Algerian Arabic code switched verb //(remember) and the French code switched noun „sujet‟(subject). The first mixed CP contains two outsider late system morphemes i.e. the Berber 1st person singular accomplished affix // and the Berber discontinuous negative affix /…/. This CP is in agreement with Myers Scotton‟s prediction on the distribution of outsider system morphemes in mixed CPs; however, this does not seem to be true for the second CP.

The outsider late system morphemes in the CP/-- /-sujet(I swear that(I)did not remember about the(exam‟s) subject) are: the Algerian Arabic negative affix //, and the Berber 1st person singular accomplished affix //. The Matrix language of this CP and the one preceding it is Berber. Outsider system morphemes should be Berber according to the predictions of the 4M model. The

200 These are just hypotheses that need to be tested against more Berber bilingual data. Another hypothesis could be that bilingual speakers use this type of double morphology as a conversational device. 282 negative affix //(a late outsider system morpheme) used in this CP is form Algerian Arabic even if it is supposed to be from Berber in the sense that Berber is the Matrix language of the verb maximal projection /-- /-sujet(I did not remember about the subject) within which the negative marker occurs. Berber is also the ML of the sequence of mixed CPs /--   --  /-sujet(I did not remember I swear that I did not remember about the (exam‟s) subject)); as a logical consequence to that Late outsider morphemes in this sequence of CPs are all supposed to be from Berber.

/   --  / (No No Neg-accept-3rd plu aff-Neg in-house) (Lit: no no they do not accept in house) (No no my parents do not accept)

The ML in this mixed CP is Berber. The Algerian Arabic negative marker // (no) that is an outsider late system morpheme has been used in addition to its Berber counterpart // (no) which is used alongside another Berber outsider late system morpheme i.e. the Berber negative discontinuous affix /…/. This doubling has been interpreted differently by scholars working within the conversational analysis of code switching (Auer, 1998; Li Wei, 2003; Gardner Chloros, 2009) tradition. They interpret doubling as a conversational strategy used by bilingual to emphasize their message in the same way as monolinguals use reiteration to emphasize their messages

/ ----/ (no Neg-3rd sing fem obj accusative clit aff-understand-1st sing aff accompl aff-Neg) (Lit: no not it I understand not) (No I did not understand it)

This mixed CP shows double morphology in terms of outsider system morphemes. There is a doubling of the negative markers. The first negative marker is //(no) from Algerian Arabic. This outsider system morpheme is accompanied 283 by the Berber discontinuous negative verbal affix /…/ which is attached to the Algerian Arabic verbal stem //(understand). The other violation to the MLF and the 4M models is the combination of the Algerian Arabic negative marker with the Berber verbal affix //(a late outsider morpheme) which stands for 1st person singular accomplished case in Berber.

Bousfoura (2003: 39) discusses similar violations to the system morpheme principle using a data of Tunisian and Standard Arabic diglossic switching in Tunisia. She(ibid: 42) presents instances of code switching where Tunisian outsider system morphemes such as tense and negative markers are associated with standard Arabic negative markers within mixed CPs. Here is one of the examples she gives:

/  ‟ - / (Because Neg-3rd sing masculine inaccompl aff-improve the-literature) (Because literature will not improve)(Tunisian Arabic-Standard Arabic code switching, Bousfoura, 2003: 39)

Many examples discussed as part of this subsection show the occurrence of outsider system morphemes from Berber (the ML) and Algerian Arabic (the EL). These system morphemes are even combined within a single CP in some instances. Negative morphemes are, for instance, indicative of the ML language according to Myers scotton and Jake‟s (2000: 1064) predictions. They, therefore, should be drawn from Berber in many of the instances analysed. They however occur in Algerian Arabic. This strongly challenges one of the underpinnings of the MLF and the 4M models in relation to system morphemes distribution and conduct in code switching corpora. In addition to that, negative markers are even doubled in both Algerian Arabic and Berber in some instance. Instances of negative markers switching stand in the face of Myers Scotton (2002:92) claims in relation to the doubling of morphemes in bilingual data. She (ibid: 92) overtly states that:

“A hypothesis under the 4M model is that early system morphemes are salient at the same time as their content morphemes heads. Thus, they are available if 284 any mistiming201 is going to occur. In contrast, late system morphemes are not available yet, but only become salient at the level of the formulator. There are no examples in the literature to show they are doubled in code switching. All examples in the code switching literature showing double morphology are early system morphemes202” (Myers Scotton, 2002: 92)

The next subsection tackles another troublesome issue in the insertional interpretation of code switching i.e. the one of discourse markers such as coordinators, complementizers, adverbs, and discourse fillers.

4.2.2.3.2 Switching of Algerian Arabic and French Discourse Markers with a Special Reference to Complementizers, and Coordinators

The category of discourse markers is quite difficult to define in terms of what represents a discourse marker and what does not. Crystal (2008: 59) defines discourse markers as:

“A word or phrase that marks a boundary in discourse typically as part of a dialogue. Discourse markers do not belong to the syntactic or semantic structure of the sentence. Discourse markers are often poly-functional elements. Discourse markers can be understood in two ways. Firstly as elements which serve to the union of utterances (in this sense they are equivalent to the term connective). Secondly, as elements which serve to a variety of conversational purposes” (Crystal, 2008: 59)

The definition and the classification of discourse markers are problematic in monolingual and bilingual language alike. Myers Scotton (1997: 150) classifies subordinators, complementizers, and other types of discourse markers as system morphemes in her earlier MLF model (1993, 1997). Myers Scotton and Jake (2001a: 93) then redefine them as content morphemes. The argument that they (ibid: 93) put forward is that discourse markers are thematic role assigners in the sense that they

201 Mistiming has been used by Myers Scotton(1997 : 100)to explain cases of double morphology in her data. Double morphology being the doubling of elements from both languages in code switching 202 Emphasis is our. 285 assign thematic roles to CPs and unite them at a suprasegmental level. Myers Scotton and Jake argue in this vein that:

“Under a revision of the MLF model, the notion of thematic role has been extended. At the discourse level, discourse particles and some subordinators are content morphemes because they have thematic structure in the sense that they determine thematic discourse thematic roles” (Myers Scotton and Jake, 2001a: 93)

Myers Scotton and Jake (2001b: 1093) elaborated their argument in a second paper where they treated the division of morphemes in more details. They (ibid: 1093) argue in this paper that:

“While discourse elements are closed-class elements, we argue that they are content morphemes at the discourse level. We are under no illusions that they participate in the thematic grid of the IP; however, it can be argued that they assign discourse-relevant thematic roles at the discourse level. Discourse elements clearly restrict the interpretation of CP or other phrasal categories they head. Some subordinators are obviously discourse markers since they signal contrast or reason, although they may be realized as part of a larger lexeme combined with system morphemes, as in the Arabic subordinators that agree with their complement subject (e.g. li?anni ‘because/1SG’). However, we argue that such subordinators including relative pronouns also structure the discourse. Consequently, such subordinators are content morphemes”(Myers Scotton, 2001b, 1093)

This classification of discourse markers is problematic on a number of grounds. The use of „some subordinators‟ is not defined and arguing that discourse markers are content morphemes is problematic in the sense that these morphemes operate outside the level of the CP in most cases203. Their syntactic requirements are

203 The discourse markers are considered by Myers Scotton (1997) as extra-sentential or intersentential code switchings. She, thus, dismisses them as instances of code switching that are unworthy studying. Some others(Boumans, 1998: 136; Ziamari , 2003: 163) consider that this is an elegant way to dismiss instances of code such as these because of the difficulty in interpreting them properly as Boumans puts it: “while the „CP analysis‟ thus offers an elegant way out for some intricate problems concerning the status of adverbs, discourse markers and conjunctions in a matrix language model”

286 found outside the CP in which they occur. Other researchers (Boumans, 1998: 157; Muysken, 2000: 162; Boumans and Caubet, 2000: 125) acknowledge the difficulty of classifying these discourse markers as either system, or content morphemes. Boumans and Caubet (2000: 117) point to this difficulty (ibid: 117) stating that:

“The concept of Matrix language applies less well to various types of elements that function at the discourse level. Various complications emerge with respect to „discourse markers‟, a heterogeneous group of particles, adverbs and expressions that either order the text into sequences (e.g. through, so, and), or express the speaker attitude toward what is being said or toward his/her interlocutor (already, really, you see).” (Boumans and Caubet, 2000: 117)

Boumans (1998: 157) classifies discourse markers as system morphemes within his Monolingual Structure Approach. Muysken (2000: 162), similarly, classifies them as functional elements. He (ibid: 163), further, classifies complementizers as clear cases of system morphemes; although, he gives no classification for coordinating conjunctions and connectives. He remarks that not only discourse markers, but also most function words do not behave in the same way across languages as he(ibid: 156) puts it:

“The functional effect derives from the lack of equivalence of functional elements across different languages. Although the evidence is not conclusive, I think a fair case can be made for this hypothesis” (Muysken, 2000: 156)

For our analysis of this data, we will consider discourse markers as content morphemes following Myers scotton and Jake‟s (2001b: 1093) argument. This stand seems to be safer for us for the time being. Considering these elements as functional, which is what we truly believe204, would be problematic for the treatment of most

204 The general tendency in code switching research is that discourse markers are functional elements, but there is still no consensus on the real status of these elements. However, from Myers Scotton and Jake‟s classification of morphemes in the 4M model we understood that discourse markers are functional elements that may be classified as outsider system morphemes. This relates to the fact that they are elected at the formulator level, which is the latest phase in language production. Content morphemes are elected at the conceptual phase, which occurs much earlier. The second reason we think that discourse 287 instances of discourse markers switching. We have observed above that the switching of these elements in general and complementizers in particular is a recurrent feature in the data. This would represent a clear violation in the face of the MLF and the 4M models.

/ - --   ----/ (Neg(AA) one-fem aff girl not then(later) inacomp-2nd sing obj accusative cli aff-tell-1st sing aff) (Lit: (It‟s) not one girl, not then, I tell you) (Its not the girl (we were talking about), then I (do) not tell you)

Berber is the ML in the mixed CP /   --- -/ (lit: not then, I tell you); accordingly, all the late outsider system morphemes are supposed to be from Berber. This is, however not the case. The negative marker // (not) is from Algerian Arabic.. The insertion of // (not) represents a clear violation to one of the most important predictions of the MLF and the 4M (its subsequent model). The complementizer// is a content morpheme from Algerian Arabic. We have noticed that instances of Algerian Arabic and French complementizers inserted into Berber CPs abound. These patterns of switching are not idiosyncratic205 but seem to be recurrent in the data that we have investigated. The following instances of code are illustrative of Algerian Arabic and French complementizers and coordinators‟ insertions into Berber CPs from the data.

/.. -  --/ (But with-how much how much 3rd sing obj accusative clit aff-buy-2nd plu fem accompl aff)

markers are outsider system morphemes is that they operate at the suprasegmental level(they join CPs together)

205 The recordings that we made have been done on speakers from different sociolinguistic profiles to obtain a good representativity of the data. The analysis of the data has been done on a qualitative basis; however, the recurrent features in the data have been investigated on a quantitative basis. This has been done for example for switching in the categories of nouns, verbs, complementizers, coordinators, and other lexical items

288

(Lit: But how much it did you buy)(But how much did you buy it)(Berber-Algerian Arabic code switching)

Berber is the ML of the CP. The coordinator/../(but) and the complementizer // (how much) are content morphemes from Algerian Arabic. Berber being the ML of the whole piece of discourse in which the CP occurs, the outsider late system morphemes are supposed to be from this language. This CP displays two interesting features that seem to be problematic to the predictions of the MLF model. The Berber complex complementizer /-/ (with how much) is repeated in Algerian Arabic// (how much). Myers Scotton (2002: 97) discusses double morphology in bilingual speech whereby content morphemes and early system morphemes (number, gender inflectional affixes) are doubled. Backus (1996: 90) also discusses doubling of plural affixes (early system morphemes) in his Dutch-Turkish data in the Netherlands. The doubling of complementizers has, however, rarely been signalled in code switching corpora.

The above instance of double morphology is clearly not indicative of alternation, but a case of insertion, for the Algerian Arabic complementizer // (how much) is inserted into a Berber CP. Doubling of complementizers is not the only case of doubling that we have observed in the data. It however seems to be the most frequent one. We have also noticed the doubling of negative markers, which are outsider system morphemes. Below is an instance of such pattern of switching from the data

/ Chabott   ---    - (he Chabott not necessarily inaccomp aff- 3rd sing obj accusative clit aff do- 2nd sing-know-2nd sing aff do-2nd sing imperative null aff like that 3rd sing accompl aff- have) (Lit: He (is) Chabott not necessarily it you know, do like (as if) that he had) ((his name is) Chabott, it is not necessary (that) you know it (his name) just do(as if) that(is) the name he has).(Berber-Algerian Arabic code switching)

289

The matrix language in the mixed CP /   - /(do like that he had) is Berber. The whole piece of discourse in which the CP occurs is from Berber. Most outsider system morphemes are from Berber. The outsider system morphemes are

1. The neutral gender imperative null affix from Berber. This null affix is used with the Algerian Arabic inserted verbal stem // (do). 2. The Berber verbal inflection // which is used with the Algerian Arabic inserted verbal stem //(has).

The complex mixed complementizer / / (like that) is a combination of two content morphemes. What is interesting in this complementizer is that it is composed of the Berber preposition // (a content morpheme as it is the head of the PREP maximal projection) and // (a content morpheme for it is a discourse marker). This Combination seems not to exist in Algerian Arabic *//

/ -- -/ (Then inaccompl aff-ask-1st sing aff girls) (then I will ask girls)(Berber-Algerian Arabic code switching)

The outsider late system morphemes of this Berber-Algerian Arabic mixed CP in which Berber is the Matrix language are: The Berber unaccomplished affix //, and the 1st singular verbal inflection //. The Algerian Arabic complementizer// (then) is a content morpheme.

/ - - --/ (So the-writing Fem aff- wrong 2nd sing aff know-2nd sing aff) (so the writing(was) wrong you know)(Berber-Algerian Arabic code switching)

The Matrix language in this mixed CP is Berber, The Algerian Arabic coordinator (so) is a content morpheme. This content morpheme operates at the suprasegmental level as it is not part of the CP/- - - 290

-/(the writing(the spelling is)wrong you know). The morphemes of the CP are:

1. The Algerian Arabic noun //(writing), a content morpheme. This noun is adapted to the Berber phonological system through the replacement of // by //206. The Algerian Arabic definite pronoun //( an early system morpheme) accompanies the noun //. This is in accordance with the predictions of the 4M model which stipulates that early system morphemes may be from both the ML(Berber in this context) and EL(Algerian Arabic in this instance) languages.

2. The Algerian Arabic inserted verbs //(mistaken), and //(know). These inserted verbal stems are content morphemes.

3. The affix //, which stands for third person singular feminine accomplished Berber conjugation and which is attached to the verb//(mistaken), and the circular affix /…/ which stands for 2nd person feminine singular accomplished in Berber, and that is attached to the verb//(know). These two affixes are outsider late system morphemes.

The recurrence of Algerian Arabic complementizers in the data that we have analysed pushes us to reconsider the way we should interpret them. Using the MLF model in their analysis proves to be unpractical on many grounds. Their classification within the MLF and the 4M models proves to be controversial and unsatisfactory for the data that we have analysed. Myers Scotton and her associates (1997, 2000, and 2001) seem to have found a way to get rid of the annoying problem of discourse markers by compartmentalising all of them as content morphemes. The idea that they operate at the discourse level also seems a bit deceiving to our eyes as doing that dismisses the importance of discourse markers at the syntactic level.

206 This noun may be considered as a borrowing, which has been adapted to the phonological system of Berber, but the purpose of the analysis in this chapter is not to differentiate between borrowing and code switching. These phenomena are treated in the same manner from the morphological and the syntactic point of view. This is why differentiating them is neither productive nor relevant in this context. 291

An interesting alternative has been advocated by Muysken (2000: 112). He argues that discourse markers operate at a level of language that is different from the other functional elements. Muysken(ibid: 96) compares the research findings of studies carried on quite diverse language pairs in relation to discourse markers and other problematic elements such as prepositions. He (ibid: 96) claims that the processes whereby these elements are inserted are alternational rather than being insertional as he puts it:

“These elements can be incorporated into a language via alternation; thus, there is a second path in addition to the insertional path. Content words such as nouns and adjectives are likely to be insertions, while discourse particles and adverbs may be alternations” (Muysken, 2000: 97)

Muysken (ibid: 97) goes even further and rightly points to the difference between sentence grammar and discourse grammar. He points to the autonomy of the two systems. Treffer Daller (1994:91) takes a similar stand in her analysis of French-Dutch in Brussels. Treffer Daller analysis of her data shows that the direction of switching in insertions is unidirectional i.e. from the embedded to the Matrix language while it is bidirectional in alternation.

Treffer Daller (ibid: 91) analysed patterns of switching from French into Dutch matrices and patterns of Dutch switching into French matrices. Using a rigorous quantitative investigation of switching tokens in both types of switching, she could show that the types of tokens code switched are indicative of the type of switching. The most numerous are the category of nouns. Switching in this category is indicative of insertion in the sense that it highlights an asymmetry between Dutch and French. French being the dominant language, the number of French nouns inserted in Brussels Dutch is much bigger than the one of Dutch nouns inserted into Brussels French. This is indicative of an asymmetry between French and Dutch in relation to insertions. The direction of switching in insertion is thus unidirectional i.e. from French, the language with more influence to the one with less influence (Dutch). The other patterns of switching involve conjunctions, complementizers, and interjections. Asymmetry is less salient for switching within this pattern. In addition to that, the direction of switching is 292 not unidirectional i.e. from the Matrix language to the embedded language, but is bidirectional

Similarly, the analysis of the data of this study indicates that there is a clear correlation between the patterns of switching and the direction of switching. The insertional patterns of switching i.e. the ones that involve the insertion of nouns, verbs, adjectives and other elements within the segmental level are unidirectional. They, generally, are from Algerian Arabic (the variety with most influence) to Berber (the variety with less influence.

4.2.2.3.3 Insertions of Berber singly occurring elements and Islands into Algerian Arabic and French Matrices

The patterns of code switching whereby Berber elements are inserted into otherwise Algerian Arabic or French matrices have been observed in the data. We may assume that such patterns are less frequent than the pattern whereby Algerian Arabic and to a lesser extent French and Standard Arabic morphemes are inserted into otherwise Berber matrices We may consider them as challenging to the predictions of the MLF in relation to the ML principle. This principle stipulates that the ML does not normally change within discourse. Myers Scotton (2002: 101) argues that the stability of the ML within bilingual corpora has been attested by diverse investigation of bilingual corpora207. . Below are some instances of Berber morphemes inserted into Algerian Arabic and French matrices.

/  -  -- --     - / Melissa /-/!

( Interj stand(up) like-that now I take you I arrange(your bed) stand(up) look, look (at) your sister how (she) nice and sweet and dirty(cute she is))

207 Myers Scotton illustrates her point with a statistical investigation of the ML in the speech of indigenous African bilingual speakers in Pretoria (Finlayson, Calteaux, and Myers Scotton, 1998). The statistics indicate that the ML is highly stable in a data of numerous bilingual CPs

293

(Lit: Interj Stand(up) like that now I take you I Arrange(you to sleep) stand(up) look look your sister how nice and sweet Melissa Dirty)

(Stand up and let me take you to bed, be quiet and look at your sister, look how nice, sweet she is)(Berber-Algerian Arabic code switching)

This example has already been discussed in Chapter two as part of the discussion on the Equivalence and the Free Morpheme constraints. There is a change in the Matrix Language from Berber to Algerian Arabic. The first CP is/  -  -- --  /(stand up like that now I take to bed stand up look). Its matrix is Berber which is the matrix of the whole piece of discourse.

The second CP is   - / Melissa /-/!(look at your sister how nice and sweet Melissa Dirty). The Matrix Language of this utterance seems to be Algerian Arabic. Berber functions as an embedded language in the utterance. It participates with the inserted adjectival stem // (dirty). This adjective may be considered as a content morpheme. // (dirty) is accompanied by the Algerian Arabic inflectional affix // which is an outsider late system morpheme. This system morpheme indicates singular feminine on Adjectives in Algerian Arabic. This form does not exist in Berber. The normal realisation of this adjective in Berber is /-- /(the dirty). The equivalent in Algerian Arabic is roughly /-/ (the dirty)

Linda: /, ---  /

(isn’t it, 2nd plu fem-do-2nd plu aff- Fem aff Tiquranin true)

(Lit: isn’t it you made them, Tiquranin true)

(isn’t it true you made these Tiquranin)

Amira:   -- -- - -

294

No mother circle-3rd sing fem accompl aff-3rd plu obj accusative aff nom aff- mother 3rd plu obj accusative clit aff-turn)

(Lit: No mother circled them (gave them circular forms), mum turned them(gave them circular shapes)

(No my mother made these (Tiquranin), mum gave them their circular forms)(Berber-Algerian Arabic code switching)

Amira’s turn taking starts with Arabic as the Matrix language. What is striking in the Algerian Arabic part of the utterance is the Berber inserted verbal stem //(turn). This verbal stem is inflected with Algerian Arabic morphology. It receives an inflectional affix for 3rd person singular feminine accomplished in Algerian Arabic /-/. It also receives //, an Algerian Arabic 3rd person plural object accusative clitic affix. This verbal stem does not exist in Algerian Arabic. It is therefore not used by the Algerian Arabic speech community of Oran. The second part is a reiteration of the first in Berber.

The phenomenon, whereby the same message is repeated in both varieties, has been often observed in bilingual corpora. Some scholars (Auer, 1998; Li Wei, 2003; Gafaranga, 2000; Gardner Chloros, 2009) claim that bilinguals sometimes need to reiterate their messages. Instead of doing this in the same language, they do it using the other varieties at their disposal. Code switching becomes in such case a conversational device.

 -- ---- -

(I swear Neg-know-1st sing accompl aff how-inaccompl aff-3rd plu inaccompl aff-say-3rd plu aff with-the presidence)

(I swear I do not know how they say ‘under the presidency’)(Algerian Arabic-Berber-Standard Arabic Code Switching)

295

The Matrix language of this complex sentence is Algerian Arabic, for the piece of conversation in which it occurs is in its majority in Algerian Arabic. What interests us in this instance of code switching is the mixed VP maximal projection /--/( I do not know). This mixed maximal projection contains:

1. The Algerian Arabic negative marker //(an outsider system morpheme). 2. The Algerian Arabic verbal stem //(know) that is a content morpheme. 3. The Berber affix // (an outsider system morpheme that stands for first person singular accomplished conjugation in Berber.

This mixed VP maximal projection violates the predictions of the MLF and the 4M models which stipulates that outsider late system morphemes come from the ML(Algerian Arabic in the case of the above example).

4.2.2.4 Alternational Patterns of Switching from the Data

The alternational patterns of switching i.e. the ones involving complementizers, conjunctions, and other discourse markers are not unidirectional. They are not from Algerian Arabic, Standard Arabic, or French to Berber. We have, for example noticed that Berber, Algerian Arabic, and French complementizers are used interchangeably in the data. Below are instances from the data involving discourse markers from Algerian Arabic and French in Berber clause boundary positions.

.. -        --  ( but, I am in a hurry, I have business me alone me only who-do(inf) business

 --- aprés - -gateau  -- 296

I washed the dishes of-lunch then I made the cake then I helped mother- my après -  -   -  - then I cleaned up(the floor) then I washed up the clothes then I watched a bit feuilleton   the series of the lost years)

These instances of code switching exemplify the process whereby Berber bilingual speakers alternate between Berber Algerian Arabic, and French discourse markers within their speech. These discourse markers seem to be used interchangeably by Berber speakers in a Random manner. However, different studies on the distribution of Discourse markers in bilingual data indicate that they are not used randomly. Many studies investigating the conversational characteristics of bilingual conversation could demonstrate that discourse markers fulfil conversational functions in bilingual discourse.

1. Insertions of Bare Forms as a possible problematic issue in Myers Scotton’s (1993, 1997, 2000, and 2002) MLF and 4M models

As it has been stated above bilingual speakers use bare forms as a compromise strategy to by pass the restrictions of the languages that they are alternating, and “to reduce the distance between the languages involved in code switching” as Clyne(2003: 140) puts it. Bare forms are used by bilinguals to sort out mismatches between the languages being code switched. Bare forms may be any content morpheme (A Verb, a noun, an adjective, an adverb etc…). From the morphosyntactic point of view, these forms are said to be anomalous. Accordingly, they challenge the predictions made by most studies of code switching. Myers Scotton (2002:115) defines them as

“Embedded Language content morphemes appearing in mixed constituents framed by the Matrix Language, but they are missing the Matrix Language system 297 morphemes that would make them well formed elements in such frames” (Myers Scotton, 2002: 115)

Myers Scotton‟s (2002, 116-126) interpretation of bare forms is qualitative as she tries to generalize over individual instances of content morphemes being uninflected with ML system morphemes. Myers scotton (2002, 118) relates the occurrence of bare forms to lexical and grammatical lack of congruence and mismatch between the languages participating in bilingual speech. One of the examples she (ibid: 126) discusses is:

/ Ø-probleem

(a Problem)(Moroccan Arabic-Dutch code switching, Myers Scotton, 2002: 126 after Boumans, 1998)

The bare form in this example is the Dutch EL inserted noun ‘probleem’(problem). The Matrix language of the whole piece of discourse is Moroccan Arabic. This noun receives no determiners from neither Moroccan Arabic (the ML), nor from Dutch(the EL). The interpretation Boumans (1998a: 95) and Myers Scotton (2002: 117) give is that there is mismatch between the morphosyntactic features of Dutch and Moroccan Arabic. Moroccan Arabic has a Masculine/Feminine gender system while Dutch has a Masculine/neuter gender system. The two systems are thus not fully congruent. Myers Scotton argues that this lack of congruence impedes (blocks) the procedures of Moroccan Arabic (the ML).

Budzhak Jones and Poplack (1997, 223) took a different perspective in the interpretation of bare forms for their analysis of bare forms in English-Ukranian bilingual data in Canada. Unlike Myers-Scotton, they do not consider the occurrence of bare forms as a problematic issue, but rather consider them as a type of ‘non-standard forms208’. Besides, Budzhak and Poplack (1997: 230) do not interpret bare forms from a qualitative perspective, but rather

208 Budzhak Jones and Poplack (1997, 227) define non-standard forms as forms that do not follow the general requirements of the languages in which they occur. „Non standard forms‟ include instances of double morphology, „bare forms‟, and other unconventional forms that occur in bilingual settings. 298 from a purely statistical quantitative angle. The approach advocated by Poplack in the treatment of bare forms is also couched under the general premises of her model of code switching. Recall that Poplack and her associates dismiss singly occurring code switchings to the status of Nonce borrowings. Most bare forms being singly occurring content morphemes, their statuses are, according to Budzhak and Poplack (1997: 235) assimilated to borrowing forms.

Both approaches being very interesting, we decided to apply the approach advocated by Myers Scotton (2002, 115) in the interpretation of bare forms in the data of this study.

4.2.2.5.1 Trial Interpretations of Algerian Arabic, Standard Arabic, and French Bare inserted Forms in the Data

As we have just stated we shall interpret bare forms in the data using a qualitative approach. The bare forms that shall be interpreted are the ones that we noticed as being the most numerous. These include bare content nouns, verbs, adjectives, and adverbs to a lesser degree. We shall concentrate in the present study on bare nouns as a sample of bare forms from Algerian Arabic, French, and Standard Arabic. WE have chosen to deal with bare nouns alone, for we believe that treating bare forms from all the categories listed above falls out of the reach and scope of the present chapter. Our choice relates to the differing frequencies of each category. Bare nouns seem to be, by far, the most frequent in the data.

4.2.2.5.1.1 Algerian Arabic, Standard Arabic, and French Bare Nouns inserted into Berber matrices

Before embarking into the analysis of bare nouns in the data, we hypothesized that the possibility of finding Algerian Arabic, Standard Arabic and French bare nouns in the data is big. This relates to the fact that Berber morphology in relation to nouns is quite different from Both Algerian Arabic, Standard Arabic and French morphologies.

299

The morphological system of Berber normally does not have any definite article. Arabic and French have a dual gender definite marking system. There are prefixes in Berber which signal nouns (they signal both feminine and masculine nouns). Algerian Arabic, Standard Arabic, and French do not normally have such prefixes. We, thus, hypothesized that these differences might have an influence on the conduct of inserted code switched nouns from these languages in Berber.

Bare nouns represent by far the highest percentage of bare forms in the corpus. This seems somehow logical in the sense that singly occurring code switched nouns are the most numerous in the data that we have investigated. Below are some instances from the corpus.

/ Ø-   -/

(Ø-teacher what she told you day-this)

((the teacher)what did she told you today)(Berber-Standard Arabic Code switching)

The Bare noun in this instance of code switching is the noun //(teacher), a content morpheme that lacks its definite article from Arabic(the EL of the utterance). It also lacks the Berber early system morpheme //, // or // that are morphemes which generally precede nominal stems in the Berber system. This is clearly not a borrowing item. There are many cases in which//(teacher) is inserted with its Algerian Arabic definite article into Berber matrices here is an example:

/ --/

(I showed it to-the teacher)

We found Bare French nouns in the data. These bare nouns are generally inserted without any system morphemes from either French or Berber. The system morphemes that are generally lacking are the early system morphemes such as articles, determiners. Below are some instances. 300

/ -/ Ø patisserie

(Yesterday you ate Ø pastry)(Berber-French code switching)

- --  Ø Science

(and-that 1st mas sing aff-pass-3rd sing accompl aff def-sciences science)

(Lit: and that one he passed the natural sciences, science)

( and he attended the(natural)sciences(exam))

// ça va /--/ Ø-DS /  --/ les notes / /?

(What is new is (everything) all right inaccompl aff-enter-2nd sing aff Ø-synthesis exam or no, 2nd sing fem accompl aff- count-2nd sing fem accompl aff the marks or no)

(Lit: what is new, is everything alright, do you enter Ø-synthesis exam or no?)

(What is new, is everything alright, will you attend (the synthesis exam or no)(Berber-Algerian Arabic-French code switching)

/--- - -- /- Ø-sujet

(Neg-remember-1st sing accompl-Neg and God Neg-remember-1st sing accompl aff about-Ø- subject)

(Lit: I do not remember, god(is my witness I do not remember about(the exam’s) subject)

301

/--- -- - - .. -/ Ø- racisme

(Not-like-3rd plu accompl aff-Neg Inaccompl aff-change-3rd plu accompl aff def-mentality fem aff-this, but have-3rd plu accompl aff Ø racism)

(Lit: They do not want they change the mentality this, but they have racism)

(they do not want to change this mentality, but they have racism)(Berber-Algerian Arabic- French code switching)

The bare nouns in these examples are ‘patisserie’ (pastry), DS (synthesis exam), sujet (subject), racisme (racism) and ‘Science’ (science). These are French inserted code switched nouns which receive system morphemes from neither French (the EL), nor Berber (the ML). These nouns are supposed to receive a feminine singular definite article in French. They, however, do not receive system morphemes from the ML 209(Berber) either. We may interpret them as cases of mismatch between Berber and French in the sense that Berber does not have any indigenous definite article system. French on the other side has no prefixes, or marking nouns for number or gender. This mismatch may have inhibited system morphemes from both languages to appear.

Ø -- --  

(Ø sinner-plu 1st plu inaccompl aff- read-1st sing obj accusative clit aff 1st plu inaccompl aff-read also)

(Lit: (the) sinners(Surat) we read it, we read it also)(Berber-Standard Arabic code switching)

209 Nouns in Berber subcategorize obligatorily for a prefix in the form of a vowel (//, //, //) for masculine nouns in general, a consonant and a vowel e.g. //, // (for feminine nouns in general). Without these morphemes, nouns may not be recognized as such in the Berber morphosyntactic system. This rule is common to most Berber varieties ( Naït Zerrad, 1995: 43-45). 302

The Bare noun in this instance is the standard Arabic noun /-/(sinners) which receives // a suffix standing for masculine plural in Standard Arabic but which lacks system morphemes from Berber. It also lacks the Arabic definite article //

/ - Ø -  - -/

(Yesterday-morning I told you rince-1st sing accompl aff def-cloth)

(Lit: yesterday morning, I told you (that) I rinsed the clothes)(Berber-Algerian Arabic code switching)

/Sarah   -- - Ø -/

(Sarah this now Neg-come-Neg from- Ø-morning)

(Lit: Sarah this now she hasn’t(yet) come from(the morning)

(Now, Sarah hasn’t yet come back)(Berber-Algerian Arabic code switching)

/ - - Ø- /

(Lit: I am hurried have-1st sing aff business)

(I am in a hurry I have business)(Berber-Algerian Arabic code switching)

/-- . /

(see-1st sing accompl aff Ø-picture this )

(Lit: I saw picture this)(I saw this picture)(Berber-Algerian Arabic code switching)

303

/-/ Østylo /- Ø- /

(give-1st sing obj accusative clit aff Ø-pen Inaccompl aff-write Ø-glass)

(Lit: give me Ø-pen writes Ø-glass)

(Give me a pen(that)writes(on the) glass)(Mzabi Berber-Algerian Arabic-French code switching)

In the above instances of code switching, the Algerian Arabic nouns Ø- //(morning), Ø-//(business), Ø- //(glass), and Ø- /./(picture) occur as bare nouns. All of them are Algerian Arabic content morphemes that have been inserted as bare forms in Berber matrices. They lack ML Berber affixes for masculine. They also lack the Algerian Arabic // with its regressive form // that should be used in this context.

We may draw some general assumptions from the analysis of bare nouns from the data. The most frequent bare nouns are the ones that are accessed without their definite articles from the ELs (Algerian Arabic, standard Arabic, and French). These bare nouns generally lack early system morphemes from Berber. They are also undefined when they are embedded in Berber matrices i.e. they receive no definite marking. Recall that these system morphemes are prefixes that are affixed to nouns in pronominal positions to permit nouns to be differentiated from adjectives. The hypothesis that we put forward in the beginning of this section seems to hold water well. It seems that this hypothesis is the most plausible to interpret the instances of bare nouns presented above.

The lack of congruence between the Berber, the French and Arabic systems in relation to definiteness seems to be an interesting alternative at least for the examples discussed above. This hypothesis needs however to be tested on more instances of bare nouns. What makes our hypothesis plausible is the fact that the bare nouns discussed above are not borrowings. Many of them occur in free variation with inflected forms. An example could be the bare noun Ø-// (business) which occurs in the data in free variation with its inflected form -// (the business).

304

Another hypothesis that we could draw from our observation is the influence of the phonological environment on the occurrence and the omission of Berber prefixes that mark nouns. We have noticed that, for example, there seems to be signs of regressive phonological assimilation. This assimilation seems to be particularly operative with masculine nouns that start with a voiceless consonant. An example would be the bare noun Ø -//(morning) which seems to be bare in environments where the phonological environment is not favourable for either a Berber vowel prefix, or an Algerian Arabic definite article in its weak form //. Its Algerian Arabic definite article may have been dismissed in such environments. */--/. Seems to be improbable in the sense that the sound preceding the supposed definite article// (a back vowel) is // (a central vowel). The movement of the tongue from the back to the centre seems to be difficult especially when the sound coming just after is // a fricative pronounced at the centre of the mouth. Owens (2005: 29) put forward similar interpretations in his investigation of English bare forms in Nigerian Arabic in Nigeria. He (ibid: 29) hypothesized that:

“It is worthwhile to look in greater detail at linguistic factors which may, or may not, Influence the occurrence of inflections on the English nouns. The occurrence of the definite article and of the plural morpheme will be examined here in terms of possible phonological influence…Beginning with the definite article, it may be asked whether the lack of the definite article isn’t due to phonological factors” (Owens, 2005: 29)

The importance of phonological features and the cross influence of these features on language contact seems to be an interesting avenue of future research for the interpretation of the challenging issues in code switching. Muysken (2000: 156) argues in this vein that:

“Three hypotheses appear with respect to special behaviour of functional elements in code mixing: The first effect relates to the special status of functional elements within the mental lexicon and speech production as argued by Myers Scotton (1997) and Azuma (1993) on the basis of psycholinguistics findings. The second effect

305 of functional elements derives from the lack of equivalence of functional elements across languages. A third hypothesis worth exploring is that the effect is related to the phonological strength of the elements involved. I think a fair case can be made for the third hypothesis” (Muysken, 2000, 156)

The occurrence of bare nouns in the present data seems to corroborate the hypothesis that we made above. Recall that this hypothesis states that our informants may be in the process of developing a Berber-Algerian Arabic mixed code.

306

Conclusion

The application of the different insertional models presented in this chapter indicates that the earlier models such as the one advocated by Joshi (1985) are inoperative to the Berber data before hand in this study. The developments and the amendments brought by Myers Scotton (1993, 1997, 2000, 2001, and 2001) to the insertional perspective in code switching seem to have strengthened the insights of the investigations of code switching. The elaboration of the distinction between content and system morphemes made by Myers Scotton and Jake (2000, 2001) has also been very profitable to the MLF and its subsequent models.

Myers Scotton refinement of the category of system morphemes proves all the very insightful to our analysis of Berber bilingual data. The possible problematic cases to the system morpheme principle in the present data are few. They include Embedded Language negative morphemes, instances of double morphology and bare forms to cite only these. The analysis of the present data indicates that our findings corroborate the general tendency in code switching research, for these problematic cases have already been well illustrated in other bilingual corpora.

Cross bilingual research points to the inefficiency of the insertional models of code switching in the interpretation of switching within the category of discourse markers. The analysis of the present data highlights the same tendency as we have felt unsatisfied with the insights of Myers Scotton’s(1993, 1997, 2000, 2001, 2002) MLF model and its continuous developments in relation to the switching of discourse markers. Discourse markers switching may be considered alongside Noun, verb and adjective switching as one of the most recurrent features in the speech of the Berber speakers that we have investigated. The insights given by Muysken (2000: 150) in relation to code switching in this category seem to be quite interesting for a future research. Muysken points to the idea that discourse markers operate at a level of grammar which is beyond the segmental level of the CP. This avenue of research seems to be of a higher relevance especially with the data that we have investigated.

307

General Conclusion

The study that we have carried is concerned with contact phenomena as they are witnessed in the Berber minorities in Oran. The work has a special focus on the Mzabi and Kabyle minority groups of Oran. At the beginning of our study, we thought that choosing these two groups would be interesting. They are easily noticeable in the sociolinguistic environment of Oran. The Mzabi minority is well represented in the economic net of Oran. So is the case for the Kabyle minority. Some dwellings of Oran known for dense trade activities (such the ones where wholesale shops are situated) are shared by Mzabi and Kabyle sellers. The Kabyle minority is also present in the city in service activities such as accommodation, and catering activities. These groups have also been chosen for demographical reasons. They outnumber the other Berber minorities of Oran. The third reason behind our choice for these minority groups relates to their Ethnolinguistic vitality which is said to be higher than the one of the other Berber minorities in Oran.

Oran has been chosen as a sample for it is an example of Algeria’s urban centres that welcome new dwellers from both smaller towns and other urban centres. This relates to its geographical and economic situation. Oran is a coastal city par excellence. It is also the second biggest city of Algeria in terms of economic activities.

The contact phenomena that we investigated are code switching, code mixing and borrowing. All along the study we have been using the term code switching as a blanket term which covers both concepts. The rationale behind our use of code switching as a cover term is that the main focus of our work is the investigation of the morphosyntactic characteristics of these contact phenomena. We followed the general tendency in contact linguistics research which states that the treatment of these phenomena from the morphosyntactic point of view is the same. The other reason we resorted to this blanket term instead of operating with a clear cut distinction of the three phenomena throughout the work is that we wanted to avoid getting into the thorny, controversial, but unproductive distinction of the three phenomena

308 listed above. We consider that as we have been involved in the morpho-syntax of contact phenomena in general, questioning the differences between code switching, code mixing and borrowing falls out of the scope of the study that we have conducted.

When we started the study, we were cautious when using the word minority. In a prior study of these minorities (Benhattab: 2004) we used the term community which is said to be more neutral. Our apprehensions were related to the fact that the minorities that we investigated are as Algerian as all the other communities living in Oran. The second reason is that members of these minorities share the same religion, systems of belief, customs and traditions and social organisation with the other groups.

We started the study with the implied hypothesis that these are minority groups which share many characteristics with other minorities of the world. We built our hypothesis on the readings that we did on minorities as diverse and different as the Turkish minority of the Netherlands (Backus, 1996), the Moroccan minority of the Netherlands (Boumans, 1998), the Chinese, the Greek Chypriot, and the Indian minorities of England (Romaine, 1994; Li Wei, 2003; Gardner Chloros, 2009), the Dutch minority of Australia (Clyne, 2003). The game seemed a bit risk taking as all the minorities just listed above are composed of emigrants who left their countries. They have different cultures, social organizations, perhaps different religions, and most importantly different languages from the majority groups with which they live. The Berber minorities that we have investigated are different from these. The only shared characteristic between the Berber minorities and these emigrants’ minorities is the one of language that is different from the language used by the majority speakers.

The hypothesis we built our reasoning on throughout the work is that the Berber communities under study here may be considered as minorities for they display bilingual patterns of language use similar to the ones displayed by minorities in other sociolinguistic contexts in which the term sounds more appropriate. Our apprehensions started to dissipate as we got into the analysis. The social investigation revealed that Berber speakers use code switching and related phenomena to signal social features to one another such as identity to one another. 309

The application of the Markedness model as expounded by Myers(1993c) on Berber minorities of Oran indicate that these minorities use Berber as the unmarked code in their inner social networks. The tendencies are not similar for the Mzabi and Kabyle minorities. The Mzabi minority seem to use Berber more often in its social networks than the Kabyle Berber minority does. Throughout the social analysis of code switching in the data, we detected signs of the development of a Berber-Algerian Arabic mixed code that is used as an unmarked language in the Berber minorities. The use of Algerian Arabic, and French discourse markers are characteristic of this unmarked mixed code. We observed that the tendencies are again different for the two communities. The Kabyle minority seems to be at a more advanced stage than the Mzabi one with regards to this mixed code.

Similarly, the use of code switching as an unmarked choice is characteristic of minorities all over the world. The only difference is that the tendencies are different from a minority to another. Backus (1996: 87) noticed that the Turkish minority of the Netherlands uses a Turkish- Dutch mixed code at the unmarked language in its inner networks. So are the German and Swedish minorities of Australia (Clyne, 2003: 45).

The analysis of the data revealed that code switching is used by Berber speakers as a conversational device. We have noticed, for example that code switching is often used by the informants of this study as an emphasizing device. The application of Auer’s (1995, 1998) pragmatic perspective to code switching and related phenomena proves to be useful to the interpretation of many instances of code switching in the data. The use of code switching as a conversational device adds to the strength of our hypothesis that the Berber communities in question are minorities. Scholars working within the pragmatic tradition in code switching advocate that such type of code switching is characteristic of language use in minority groups. Many applications of the pragmatic functions of code switching have in fact been made on Minority groups (Alfonzetti, 1998; LiWei, 1998, 2003; Sebba, 1998; Blommaert, 1998).

310

Throughout our social interpretation of code switching in the Berber minorities studied here, we considered the insights of integrating a social perspective with a linguistic one for a better understanding of the data at hand. Few scholars(Milroy and Li Wei, 1995; James Smith, 2002; and Gardner Chloros, 2009) point out to the relevance of this integrative approach in the analysis of code switching. This perspective permitted us to have interesting insights on the code switching patterns as they are displayed by the Mzabi and Berber minorities of Oran. An example of such insights would be the relationship between code switching types and the generation to which the speakers belong. The analysis of the present data indicates that First and second generation Mzabi speakers display patterns of code switching that are different from the ones displayed by their second generation counterparts. Second generation Mzabi speakers use a mixed variety of Mzabi. This variety is characterized by an extensive use of code switchings from Algerian Arabic. To come back to our initial hypothesis, this process, whereby second generation speakers display code switching patterns different from their first generation fellows, is also indicative of minority groups in bilingual settings. This change in language patterns is considered by many scholars as a sign of an ongoing language displacement process.

The second part of the present study focuses on the grammatical analysis of code switching as members of the Berber minority display it. We kept the same hypotheses put forward at the beginning of the investigation. These are mainly that the Berber minorities display similar bilingual processes as minorities in other bilingual contexts. We added some more hypotheses for our grammatical analysis of code switching and related phenomena in the Berber minorities under light here. We hypothesized that the theoretical models that have been accepted by the bulk of researchers in code switching studies may not be necessarily applicable to the patterns of language use in the present study. We also hypothesized that there may be an influence of typological factors on code switching, code mixing, and borrowing in the Mzabi and Kabyle groups investigated here.

The application of the first theoretical model i.e. the Equivalence based model as expounded by Poplack and her associates (1981, 1988, 1998, indicates that the predictions made in the Equivalence based model in its different versions are unpractical to Berber-French code switching in the data. Many instances of code switching in the data involving these two 311 languages seem to stand in the face of this model. We hypothesized that this mismatch may be related to the type of bilingualism in the communities that we have investigated. Even if Poplack and her associates claim universality for their model, Poplack (2007) confessed that cases of bilingualism such as the one of the Puerto Rican community of the US whereby the speakers have an equal competence in both languages are extreme cases. She also confesses that most bilingual communities rather display symmetrical bilingualism whereby speakers do have an unequal competence in the languages involved in code switching and other bilingual phenomena.

Another hypothesis that made its way in our data is the influence of typological differences between the languages involved in code switching. Before we started to look for possible problematic instances, we looked at the morpho-syntactic features that are different across the languages involved in code switching i.e. Berber, Algerian Arabic, Standard Arabic, and French. The predictions of the Equivalence constraint as expounded by Poplack and her associates stipulate that code switching is impossible in cases of mismatch between these languages. We then searched for cases of code switching involving patterns where there are mismatches having in mind that these would be clear violations to the equivalence constraint. We also started our analysis with the hypothesis that Algerian Arabic code switching would be unproblematic to the equivalence constraint. This hypothesis is motivated by the typological similarities of Algerian Arabic and Berber that weakens the possibilities of having mismatch between the two languages.

The analysis indicates that there are plenty of violations to the predictions of the equivalence constraint. Some examples include switching between a French noun and a modifying Berber adjective(the mismatch being in the parsing system of French and Berber in relation to nouns and their modifiers), Code switching involving a mismatch in gender assignments as for cases of code switching of Algerian Arabic, French or Standard Arabic nouns that are feminine with Berber adjectives that are masculine.

The analysis of Berber-Algerian Arabic code switching shows that the hypothesis put forward when we started to consider the equivalence constraint holds water to some extent. 312

Recall that this hypothesis states that Berber-Algerian Arabic code switching should not be problematic to the equivalence constraint because of typological similarities between the two languages. The analysis of the data reveals that there are few violations to the equivalence constraint especially in instances of code switching involving object clitic pronouns, and others involving gender assignments in nouns and verbs.

Most violations to the equivalence constraint in the present data seem to relate to asymmetry between the languages involved in code switching. Asymmetry seems to be the right interpretation to these cases of mismatch between the languages participating in code switching in the present study. There seems to be a control of the Berber morpho-syntactic system on code switching instances that are inserted in Berber stretches of discourse.

Many code switching instances from the present data that represent violations to the predictions of the equivalence model as expounded by Poplack and her associates (1980, 1981, 1988, 1995, 1998, 2000, and 2007) are singly occurring nouns, verbs, adjectives and adverbs. When these elements are inserted in Berber they tend to follow the morphosyntactic requirements of this language. This observation is in complete contradiction with the basic premise in Poplack’s perspective in the interpretation of code switching.

The findings of the present study are in accordance with studies carried on other bilingual contexts which involve minority speakers. These studies highlight for, example, the idea that the biggest percentage of code switching instances involve singly occurring code switched nouns, verbs and adjectives. These patterns of code switching are characteristic of asymmetrical code switching as it is displayed by minority speakers.

We may interpret these tendencies differently with the present data. We hypothesized that there is a possible relationship between the typology of Berber (the receiving language) and code switching types in the data. Berber is an agglutinative language that generates new words using combinations between existing morphemes. This highlights the way Berber speakers store function morphemes and lexical items in their brain. This manner of 313 compartmentalizing lexical and functional elements in the brain facilitates the insertion of code switched nouns, verbs, adjectives into Berber. This is again a preliminary hypothesis that needs to be verified against more Berber bilingual data. This hypothesis will help us explore an avenue of research which seems to be a promising one.

The free morpheme constraint as expounded By Poplack (1981) hasn’t been in a safer compartment than its sister the equivalence morpheme. We hypothesized, at the beginning of our analysis of the data in the light of the free morpheme constraint, that it would be easy to find counter evidence to this constraint. The code switching literature indicates that the instances of code switching that are the most problematic to the free morpheme constraint come from agglutinative languages. These languages integrate code switched elements using native affixes. Similarly, Berber being an agglutinative language shows patterns of code switching such as the ones highlighted in the literature whereby Algerian Arabic, Standard Arabic and to a lesser extent French lexical items(free morphemes) are inserted to the Berber morphology using Berber Affixes.

Some scholars (Levelt, 1989; Backus, 1996, 2003; Myers Scotton, 1998; and Muysken, 2000: 54) relate these code switching patterns to the way bilingual speakers who have agglutinative languages as their native languages store morphemes in their brains. They hypothesized that free and bound morphemes are stored separately in the bilingual brain. This allows bilingual speakers to use bilingual combinations of lexical items and free morphemes from either the host or the guest language with bound morphemes from the host language.

We put forward a similar hypothesis in relation to the possible violations of the Equivalence constraint above. The analysis of the present bilingual data in the light of the Free Morpheme Constraint strongly supports the hypothesis made by the above cited scholars. The analysis also represents counter evidence to this constraint. The point made by Poplack and her associates as recently as in 2007 that the violations to the Free Morpheme Constraint are nonce borrowings does not seem to hold for the present data. Many instances that have been presented as counter evidences to this constraint are neither adapted to the Berber

314 morphological nor are they adapted to its phonological system. We, accordingly, consider them as clear and unquestionable instances of code switching.

Our analysis of the data in the light of the equivalence model further supports the hypothesis that Mzabi and Kabyle speakers are in the process of developing a mixed code. The signs of this mixed code that we detected in the equivalence perspective are the use of Algerian Arabic prepositions, and the Algerian Arabic constructions in which Berber elements are inserted.

The consideration of the Government models of code switching in the light of the present data reveals that the Government and Binding Model as it has been expounded by Muysken et al(1986, 1990, 1995, 2010) is somehow inoperative. We tried both versions of government i.e. the Lexical head, and the syntactic head versions, but both of them fail badly against Bilingual code switching as it is displayed by the Mzabi and Kabyle minorities in Oran.

The bilingual studies that prove to be the most damaging to the government perspective to code switching is the one carried by Nortier in 1990 on the Moroccan minority of the Netherlands. We hypothesized for this part of the work that it is easy to find counter evidence to the predictions of the Government Model for Berber, Algerian Arabic and Berber share some typological features. In fact, our hypothesis has been confirmed through the analysis of the data.

It was not difficult to find counter examples to the predictions made in the Government and Binding perspective to code switching. We proceeded the same way as for our consideration of the Equivalence model. Before starting the analysis, we detected possible syntactic heads and looked at their governed elements in Berber and in the other languages involved in this study. We then looked for instances of code switching involving these elements in the data. Using this strategy, we could find many counter instances of code switching involving for example head verbs with their governed objects, head prepositions with their governed objects. 315

We also tested the second version of government i.e. the one of syntactic or functional heads as it has been expounded by Belazi el al (1994). It also fails the test of the present bilingual data. The most threatening instances of code switching to constraints advocated by Belazi et al (1994) in the present data are the ones that involve Algerian Arabic, and to a lesser extent French code switched complementizers combined with Berber complementizers’ phrases. These violations to the Functional head Constraint support the hypothesis that we have been testing throughout the work. The Kabyle and to a lesser extent Mzabi Berber minorities of Oran are in the process of developing a Berber-Algerian Arabic-French mixed code.

The data that we investigated in the light of the Government model as expounded by Muysken et al (1986, 1990, 1995, and 2000) is different from the data that have been investigated by scholars who have been working on other language pairs. Most of the studies that considered the practicality of Government and binding to bilingual code switched data have been carried on language pairs that involve typologically remote languages. The present study presents code switching involving Berber and French which resembles other data, for Berber and French are remote languages. In addition to this language pair, it presents two other pairs that are more interesting in terms of their insights to the relevance of government as a universal construct. These language pairs are Berber-Algerian Arabic and Berber-Standard Arabic with all the other possible combinations. These patterns of code switching are more interesting for they involve languages that are typologically close to one another to some extent, as they belong to a single ancestry i.e. the Chamito Semitic family of languages.

The study that we carried explores an alternative to the government perspective to code switching which is also couched in Chomsky’s formalisms. This perspective is the minimalist one. It has been advocated by Mahootian & Santorini (1993, 1996) and by Mac Swan (1999, 2000, and 2005). In his minimalist approach to code switching Mac Swan (1999, 2000, and 2005) advocates that code switching is governed by no rules apart from the ones of the monolingual grammars. Mac Swan discusses facilitating strategies that are used by bilingual speakers to make code switching smoother.

316

The analysis of the present data indicates that the informants resort to facilitating strategies such as double morphology and morphological integration through the affixation of a native affix to the code switched item. These facilitating strategies have been observed in many investigations of minority groups in the world. They signal the development of mixed codes in bilingual communities. An extreme case of these facilitating strategies would be the grammatical patterns of creoles that have been simplified by the speakers in bilingual environments.

The bilingual strategies support the hypotheses which state that the Mzabi and Kabyle communities display bilingual patterns of language similar to the ones displayed by the other minorities of the world.

The investigation of patterns of code switching as they are displayed by Kabyle and Mzabi minorities of Oran has been conducted in the light of a third theoretical perspective i.e. the insertional perspective. The analysis of the data in the light of this perspective revealed that it seems to be the most appropriate to the investigation of Berber bilingual phenomena in Oran. Similar observations have been made by many scholars ((Backus, 1996; Boumans, 1998; Muysken, 2000; Gardner Chloros, 2009) working bilingual settings which involve minority groups. These scholars argue that minority settings highlight an asymmetrical pattern of bilingualism. They also argue that such type of bilingualism triggers insertional patterns of code switching.

The MLF model and its subsequent models as they have been expounded by Myers Scotton and her associates(Myers Scotton, 1993, 1997, 2002; Myers Scotton and Jake, 2000, 2001) proves to be relevant to the interpretation of code switching and related phenomena in the Berber minority of Oran. The relevance of this model may be related to the sociolinguistic situation of Oran. This situation seems to be somehow similar to the one where Myers Scotton conducted her research to devise her insertional model in the interpretation of code switching. Both sociolinguistic situations involve ancient colonial languages. They both involve a

317 superimposed standard and finally they both involve colloquial languages and varieties. The other similarity between the two situations is related to the patterns of code switching that are much insertional.

The study that we carried on the Berber minorities visits the most important hierarchies on which the MLF and subsequent models are built. The first hierarchy is the ML/EL hierarchy. Identifying the ML for our data hasn’t been an easy enterprise. We thus decided to test the relevance of different criteria postulated by scholars working within the minimalist tradition to the present data. Some of these criteria fail the test of the present bilingual data (the Main verb criteria). Some other criteria prove to be highly controversial (the discourse criteria, the statistical criteria). We thus decided to consider Berber as the Matrix language for convenience reasons.

Throughout the analysis of the data we noticed that the different hierarchies of the MLF and subsequent models hold water for code switching as they are displayed by the Mzabi and Kabyle minorities of Oran. The uniform Structure Principle holds for most instances of code switching from the present data. It even rightly interprets the possible violations to the preceding models. As usual we put forward a hypothesis at the beginning of our consideration of the MLF in the light of the present data. This hypothesis stipulates that with regards to the present data, there is a dominance configuration in many instances of code switching. Berber seems to be in control in many patterns of code switching. This control is highlighted at different levels in the bilingual discourse displayed by the informants.

The investigation of the present data indicates that the predictions made by the MLF and its subsequent models in relation to the distribution of morphemes are to some extent respected. However there are some violations to these predictions. These include instances of code switching from the data where there are outsider late system morphemes from both languages participating in the code switching. Cases such as these include for example negative markers from both Algerian Arabic and Berber, Bare forms that are inserted without any markings from either of the languages.

318

Other code switching instances problematic to the MLF are the ones of discourse markers switching. Many scholars working within the insertional perspective point to the difficulty of interpreting these elements from an insertional perspective. We personally haven’t been convinced with Myers Scotton (2002) classification of discourse markers as content elements that assign thematic roles to Cps or to clauses. In spite of our disapproval we decided to consider discourse markers from Algerian Arabic and French as content morphemes. An alternative consideration is the one given by Boumans (1998) who considers them as functional elements. Muysken (2000) argues that there two levels of grammar sentence grammar and discourse grammar. Muysken also claims that discourse markers operate at the level of discourse grammar. They therefore fall within a category of code switchings which are the ones of alternational code switchings. We tried to interpret these discourse markers in the light of the alternational perspective and the application seems to be more satisfactory than the one of the insertional perspective.

The application of the different perspectives in the interpretation of Code switching as it is displayed by the Mzabi and Kabyle Berber minorities of Oran indicate that the perspective that seems to be the most relevant to the present data is the insertional one. This does however not prevent to show some shortcomings. The short comings of the insertional approach that we have noticed in the data are somehow similar to the ones that have been observed in other bilingual contexts involving minority speakers.

In the light of our investigation of contact phenomena in the Berber communities of Oran we could support the hypothesis that we made in the beginning of the work This hypothesis says that Berber speakers are minority speakers that are different from other minorities of the world, but they display similar patterns of bilingual use.

The present study also points to the relevance of typological factors in the interpretation of bilingual phenomena as they are displayed by the Berber minorities of Oran.

319

This correlation needs to be investigated in future research with a particular interest in the typological differences between language typologies and code switching patterns.

In the light of the present work we could confirm that the Kabyle and to a lesser extent Mzabi minorities of Oran are in the process of developing a mixed code that is characteristic of minority groups in the world. This mixed code may also be considered as a sign of an ongoing language shift process. This hypothesis needs to be deepened using an investigation of the patterns of language use and language attitudes in the Berber minorities of Oran.

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339