TALLINNA ÜLIKOOL HUMANITAARTEADUSTE DISSERTATSIOONID TALLINN UNIVERSITY DISSERTATIONS ON HUMANITIES

Väljavõte

1 TALLINNA ÜLIKOOL HUMANITAARTEADUSTE DISSERTATSIOONID TALLINN UNIVERSITY DISSERTATIONS ON HUMANITIES

2 ANASTASSIA ZABRODSKAJA RUSSIAN-ESTONIAN LANGUAGE CONTACTS: GRAMMATICAL ASPECTS OF LANGUAGE USE AND CHANGE Abstract Tallinn

3 TALLINNA ÜLIKOOL HUMANITAARTEADUSTE DISSERTATSIOONID TALLINN UNIVERSITY DISSERTATIONS ON HUMANITIES ANASTASSIA ZABRODSKAJA Russian-Estonian Language Contacts: Grammatical Aspects of Language Use and Change Abstract Institute of Estonian Language and Culture, Tallinn University, Estonia The thesis is accepted for the commencement of the degree Doctor of Philosophy (Linguistics) on November 6, 2009 by the Doctoral Committee of Humanities of Tallinn University. Supervisor: Anna Verschik, PhD (Tallinn University, Professor) Opponents: Albert Marie Backus, PhD (Tilburg University, Associate Professor) Martin Ehala, PhD (Tartu University, Professor) The academic disputation on the thesis will be held at Tallinn University, Lecture Hall T-409, Narva mnt 25, Tallinn, on December 16, 2009 at Copyright: Anastassia Zabrodskaja, 2009 Copyright: Tallinn University, 2009 ISSN (doctoral thesis, online, PDF) ISBN (doctoral thesis, online, PDF) ISSN (analüütiline ülevaade, online, PDF) ISBN (analüütiline ülevaade, online, PDF) 3

4 4

5 Minu isa mälestuseks и для моей любимой мамочки ( TO THE MEMORY OF MY FATHER and for my beloved mother ) 5

6 RUSSIAN-ESTONIAN LANGUAGE CONTACTS: GRAMMATICAL ASPECTS OF LANGUAGE USE AND CHANGE Abstract The thesis presents a study on Russian-Estonian code-switching (CS) from the perspective of bilingual language use and change. The main goal is to give an overview of the major findings on Russian-Estonian contact phenomena. The analytic overview opens with introductive Chapter 1 where research questions and theoretical background of the thesis as well as its outline is described. This is followed by Chapter 2 that explores various aspects of methodology employed for data collection. Chapter 3 provides an informative overview of current grammatical approaches to CS analysis. I also summarize and discuss wider implications for the understanding of the term convergence. Chapter 4 of the overview comprises the analytic digest of the Papers I III. The findings are discussed in the context of the Matrix Language Frame model principles and point out at its insufficient basis for distinction between content, late / early system and outsider morphemes. Moreover, the MLF model relies only on morphosyntax but it is also important to take into consideration the phonetic realization of Estonian items. Both Estonian and Russian have a developed inflectional morphology. For example, morphological integration (gender assignment, case / case + PL) of Estonian nouns into the Russian matrix depends on individual preferences, types of discourse, oral or written CS, etc and not structural factors exclusively. Full integration of nouns means gender assignment and adding of case/number markers. Empirical observations show that a full morphological integration (and phonological adaptation) of an Estonian single noun into the Russian matrix is not always the case. As far as changes in Estonia s Russian spoken by Russian-speaking students are concerned, a number of phenomena are found. Patterns of borrowing of Estonian (compound) nouns that belong to university domain involve more than just lexical borrowing. Whole morphosyntactic patterns that often contradict monolingual Russian norms have been borrowed, i.e. the word order in NP in genitive. The findings also prove that proposed constraints on CS can be hold as well as violated. In Chapter 5, I discuss the similarities and differences between the three approaches proposed by Myers-Scotton (2002), Muysken (2000) and Johanson (2002) for identifying and analyzing the amalgamated constructions. Finally, together with general conclusions a number of directions for future work in this area are suggested in Chapter 6. The thesis takes one step more towards understanding of (potential) contact-induced morphosyntactic change in Estonia s Russian. Keywords: sociolinguistics, language contact, bilingual language use, language change, codeswitching, Matrix Language Frame model, matrix language, embedded language, convergence, congruent lexicalization, amalgamated constructions, Estonian, Russian, Estonia s Russian contact variety 6

7 VENE-EESTI KEELEKONTAKTID. KEELETARVITUSE JA -MUUTUSE GRAMMATILISED ASPEKTID Resümee Artikliväitekiri kujutab endast vene-eesti koodivahetuse uurimust kakskeelse keelekasutuse ja -muutuse perspektiivist. Selle peamine eesmärk on anda ülevaade vene-eesti keelekontaktidest tulenevatest ilmingutest. Analüütiline ülevaade algab sissejuhatava peatükiga, kus püstitatakse uurimiseesmärgid, kirjeldatakse töö teoreetilist tausta ning ülesehitust. Sellele järgneb 2. peatükk, mis valgustab andmete kogumise metodoloogia eri aspekte. 3. peatükk annab informatiivse ülevaate nüüdsetest grammatilistest lähenemistest koodivahetuse analüüsile. Võetakse kokku termini konvergents eri tähendused ja arutletakse selle erinevate kasutuste laiemate implikatsioonide üle. Ülevaate 4. peatükk koosneb kolme artikli analüütilisest kokkuvõttest. Tulemusi valgustatakse maatrikskeele raamistiku mudeli põhimõtete kontekstis ning juhitakse tähelepanu mudeli ebapiisavale alusele leksikaalsete ja grammatiliste morfeemide eristamiseks. Maatrikskeele raamistiku mudel põhineb ainult morfosüntaksil, kuid niisama oluline oleks arvestada eesti keeleüksuste foneetilist realiseerumist. Et nii eesti kui ka vene keeles on rohkesti grammatilisi tunnuseid, siis näiteks eesti nimisõnade morfoloogiline integratsioon vene maatriksis (soo määramine, kääne / kääne + mitmus) sõltub paljuski kõneleja individuaalsetest eelistustest, diskursuse tüübist, suulisest ja kirjalikust koodivahetusest, mitte üksnes strukturaalsetest faktoritest. Nimisõnade täielik integratsioon eeldaks soo määramist ja käände/arvu tunnuse lisamist. Empiiriline vaatlus näitab, et eesti nimisõna täielikku morfoloogilist integratsiooni (ja fonoloogilist kohanemist) vene maatriksis alati ei toimu. Mis puutub muutustesse venekeelsete üliõpilaste eestivene keelevariandis, siis on leitud terve rida kontaktist tulenevaid nähtusi. Ülikooli valdkonda kuuluvate eesti (liit)nimisõnade laenamise mallid hõlmavad rohkemat kui ainult leksikaalset laenamist. Laenatud on terve morfosüntaktiline raam, mis räägib vene ükskeelsetele normidele vastu, eeskätt genitiivse NP-sõnajärje puhul. Uurimistulemused kinnitavad sedagi, et pakutud piirangud koodivahetuse kohta ei ole absoluutsed nad võivad küll paika pidada, kuid võivad olla ka rikutud. 5. peatükis võetakse vaatluse alla sarnast ja erinevat kolme lähenemise vahel, mida pakuvad Myers-Scotton (2002), Muysken (2000) ja Johanson (2002) sulamkonstruktsioonide (amalgamated constructions) tuvastamiseks ja analüüsimiseks. Viimases, 6. peatükis, esitatakse üldkokkuvõte ning arutatakse suundi edaspidiseks uurimistööks antud valdkonnas. Artikliväitekiri katsub selgemini kaardistada (potentsiaalseid) kontaktidest johtuvaid morfosüntaktilisi muutusi eestivene keelevariandis. Võtmesõnad: sotsiolingvistika, keelekontaktid, kakskeelne kõne, keelemuutus, koodivahetus, maatrikskeele raamistiku mudel, maatrikskeel, sisestuskeel, konvergents, ühildav leksikaliseerumine, sulamkonstruktsioonid, eesti keel, vene keel, eestivene keelevariant 7

8 CONTENTS ACKNOWLEDGEMENTS... 8 LIST OF PUBLICATIONS... 9 LIST OF ABBREVIATIONS TRANSLITERATION OF RUSSIAN ITEMS PREFACE INTRODUCTION Rationale for the thesis Research questions in the current study Organization of the thesis METHODOLOGICAL FRAMEWORK A brief look at the sociolinguistic situation in Estonia Methodology of data collection Synchronic and diachronic dimensions in collected data THEORETICAL FRAMEWORK The grammatical approach to CS General assessment of the MLF model The Equivalence Constraint Government Constraint The triangle model Monolingual bias and constraints Convergence Convergence as the result Convergence as a type of change Convergence as the mechanism Relationship with CS Directionality of convergence Relevance of the grammatical approach to the study of language change CONTACT PHENOMENA IN ESTONIA S RUSSIAN Types of CS in the collected data Implications of the data for the MLF model Morphological integration and phonological adaptation Double marking Choice of infinitive forms Amalgamated constructions Internationalisms and lexical facilitation Word order DISCUSSION Amalgamated constructions: a composite ML, congruent lexicalization or selective copying/ mixed copies Introducing contact draught-board metaphor CONCLUSIONS AND DIRECTIONS FOR FUTURE RESEARCH KOKKUVÕTE (Summary in Estonian) REFERENCES

9 ACKNOWLEDGEMENTS This thesis is a part of research conducted on a Russian-Estonian code-switching corpus. The creation of the corpus was supported by Estonian Ministry of Education and Research, Project SF s03 The analysis, modelling and control of the development of the Estonian linguistic environment ( ), Estonian Science Foundation grant no 6151 Child language, interlanguage and code-switching corpora creation and description ( ) and by Project EKKM09-85 under grant agreement Vene-eesti ja inglise-eesti koodivahetuse ja koodikopeerimise korpuse koostamine ja haldamine [Russian-Estonian and English-Estonian code-switching and code-copying corpora creation and management] ( ). The bilingual data I use was also collected within the Project New System of Estonian and Bilingual Teaching in Tallinn University ( ). I would like to acknowledge all scholarships that supported financially my participation in international conferences and other research activities in (in alphabetic order): Alfred Kordelini Sihtasutuse Eesti Fondi stipendium, Eestikeelse terminoloogia sihtstipendium, Eesti Üliõpilaste Toetusfond USA-s, European Social Fund s Doctoral Studies and Internationalisation Programme DoRa, NordLing Mobility grant, NordLing Travel grant, ÜEKN (Ülemaailmse Eesti Kesknõukogu) Margot M. ja Herbert R. Linna stipendium, Tallinna linna stipendium. I am very grateful to all senior colleagues who generously have given me letters of support and recommendation (in alphabetic order): Prof. Martin Ehala (University of Tartu), Prof. Em. Anu-Reet Hausenberg (Tallinn University), Dr. Leelo Keevallik (Uppsala University), Prof. Helle Metslang (University of Tartu), Prof. Aneta Pavlenko (Temple University), Prof. Mart Rannut (Tallinn University), Prof. Anna Verschik (Tallinn University), and Prof. Jüri Viikberg (Tallinn University). I would like to thank the following researchers for discussing various aspects of my work with me (in alphabetic order): Prof. Michael Clyne (The University of Melbourne), Dr. Kapitolina Fedorova (European University at St. Petersburg), Prof. Em. Carol Myers-Scotton (Michigan State University), Prof. Aneta Pavlenko (Temple University), and Prof. Sarah Thomason (University of Michigan). For the insightful comments and helpful suggestions I would also thank the reviewers of my dissertation: Ass. Prof. A. M. Backus (Tilburg University) and Prof. Martin Ehala (University of Tartu). I would like to express my gratitude to my supervisor Prof. Anna Verschik whom I got to know as a teacher from I think that the PhD defence is a most suitable gift I could make for her on the occasion of celebrating the 10 th anniversary of our collaboration. Last but not least: Спасибо моей любимой мамочке. Without my mother s great love and support I would never complete my dissertation. 9

10 LIST OF PUBLICATIONS This thesis is primarely based on the publications: I. Anastassia Zabrodskaja Towards establishing the matrix language in Russian-Estonian code-switching: A corpus-based approach. Edited by Tsiplakou, Stavroula, Marilena Karyolemou and Pavlos Pavlou. Language Variation European Perspectives II. Selected papers from the Fourth International Conference on Language Variation in Europe (ICLaVE 4), Nicosia, Cyprus, June, Studies in Language Variation 5. Amsterdam: John Benjamins Publishing Company, II. Anastassia Zabrodskaja. Evaluating the Matrix Language Frame Model on the basis of a Russian-Estonian code-switching corpus. International Journal of Bilingualism. Special issue Language Contacts in the Post- Soviet Space. Guest editor Anna Verschik, 2009, 13 (3), to appear. III. Anastassia Zabrodskaja. Morphosyntactic contact-induced language change in young Estonia s Russian speakers variety. The interplay of variation and change in contact settings Morphosyntactic studies. Edited by Isabelle Léglise & Claudine Chamoreau. Studies in Language Variation. Amsterdam: John Benjamins Publishing Company (to appear in 2010). The data and other materials leading to the results presented in this thesis were also published in the following issues: Monograph I. Anastassia Zabrodskaja Vene-eesti koodivahetus Kohtla-Järve vene emakeelega algkoolilastel [Russian- Estonian code-switching among Russian-speaking schoolchildren in Kohtla-Järve]. Monograph. Tallinna Ülikooli eesti filoloogia osakonna toimetised 6. Tallinn: TLÜ kirjastus. Peer-reviewed publications I. Anastassia Zabrodskaja Expansion of a marked genitive construction as an example of contact-induced transfer. Special issue on Transfer. Guest editors Jeanette Sakel and Jeanine Treffers-Daller. International Journal of Bilingualism (to appear in 2010). II. Anastassia Zabrodskaja Language testing in the context of citizenship and asylum: The case of Estonia. Special issue Language Assessment for Immigration, Citizenship, and Asylum. Guest editors Elana Shohamy and Tim McNamara. Language Assessment Quarterly 6 (1), III. Anastassia Zabrodskaja Code-switching and L2 students in the university: bilingualism as an enriching resource. Journal of Linguistic and Intercultural Education (JoLIE), Volume 1, IV. Anastassia Zabrodskaja Code-switching and L2 students in the university: Bilingualism as an enriching resource. Ed. by Cornelia Ilie, Teodora Popescu, Rodica Pioariu, Rodica Chira. Proceedings of the 1st International Conference on Linguistic and Intercultural Education CLIE Centre for Research and Innovation in Linguistic Education CIEL. Alba Iulia, Romania: Aeternitas, V. Anastassia Zabrodskaja Kakskeelse klassiruumi uurimisvõimalusi [Research opportunities of bilingual classroom]. Toimittaneet Pirkko Muikku-Werner, Ossi Kokko, Hannu Remes. VIRSU III Suomalais-ugrilaisia kohdekieliä ja kontakteja. Studies in Languages 42. Joensuu: Joensuun yliopisto, VI. Anastassia Zabrodskaja Russian-Estonian Code-Switching in the University. The Arizona Working Papers in Second Language Acquisition & Teaching, Volume 14. Managing Editors Lance Askildson and Robert Coté, VII. Anastassia Zabrodskaja Emergent bilingual speech: from monolingualism to code-switching. A case of young Estonian Russian-speakers. Ed. Andrey G. Kirillov. Samara AltLinguo, Volume 1; VIII. Anastassia Zabrodskaja Vene-eesti koodivahetuse korpus: kodeerimispõhimõtete väljatöötamine [Russian-Estonian code-switching corpus: elaboration of encoding principles]. Toim. Helle Metslang, Margit Langemets, Maria-Maren Sepper. Eesti Rakenduslingvistika Ühingu aastaraamat 3. Tallinn: Eesti Keele Sihtasutus, IX. Anastassia Zabrodskaja Russian-Estonian Conversational Code-Switching in Student Interviews. Ed. by Ineta Savickienè. Language, Diversity and Integration in the Enlarged EU: Challenges and Opportunities. Regioninès Studijos (2). Kaunas: Vytautas Magnus University,

11 X. Anastassia Zabrodskaja Russian-Estonian code-switching among young Estonian Russians: developing a mixed linguistic identity. Ed. by Pedro Mateo, Sabri Al-shboul, Wang Jing. Kansas Working Papers in Linguistics (KWPL), Volume 28. Lawrence: The University of Kansas, ; XI. Anastassia Zabrodskaja Eestivene keel(evariant): kas samm segakoodi poole? [Russian-Estonian Language (variety): a Step Towards a Fused Lect?] Keel ja Kirjandus, 9, ; XII. Anastassia Zabrodskaja Keelekontakt ja kakskeelsus nüüdisaegse kontaktlingvistika poolt vaadatuna [Language contact and bilingualism in modern contact linguistics]. Peatoimetaja Mati Erelt, tegevtoimetaja Maria-Maren Sepper. Emakeele Seltsi aastaraamat 51. Tallinn: Eesti Teaduste Akadeemia Emakeele Selts, XIII. Anastassia Zabrodskaja Vene-eesti koodivahetuse funktsioonid Kohtla-Järve venekeelsete laste vestluses [The conversational functions of Russian-Estonian code-switching among Russian-speaking children in Kohtla-Järve]. Toimetaja Helle Metslang, Margit Langemets. Eesti Rakenduslingvistika Ühingu aastaraamat 2. Tallinn: Eesti Keele Sihtasutus, XIV. Anastassia Zabrodskaja Eesti keelt õppiv Kohtla-Järve vene koolide muukeelne laps [The Non-Native Child Learning Estonian in a Kohtla-Järve Russian School]. Keel ja Kirjandus, 11, Other scientific publications I. Anastassia Zabrodskaja Tugiseminar kui kakskeelse hariduse õppevorm [Support Seminar as a Form of Bilingual Education]. Tegevtoimetaja Natalja Zorina. Acta et commentationes collegii Narovensis. Tartu Ülikooli Narva Kolledži toimetised. Multikultuurilise hariduse aktuaalsed küsimused kaasaegses ühiskonnas: pedagoogiline teooria ja praktika. Rahvusvahelise teaduskonverentsi ettekannete põhine artiklite kogumik. Актуальные вопросы мультикультурного образования в современном обществе: педагогическая теория и практика. Сборник статей по итогам международной научной конференции. IX/1. Narva, , II. Anastassia Zabrodskaja Rühmatööst sisupõhises keeleõppes [Group work in Content and Language Integrated Learning]. Toimetaja Pille Eslon. Õppijakeele analüüs: võimalused, probleemid, vajadused. Eesti filoloogia osakonna toimetised 10. Tallinn: TLÜ Kirjastus, III. Anastassia Zabrodskaja Keel meie ümber ehk mida võib Eesti keelemaastik pakkuda sotsiolingvistile? [Language around us or what can offer the Estonian linguistic landscape to a sociolinguist?] Oma Keel, 15, IV. Anastassia Zabrodskaja Code-switching and contact-induced language change in Estonia s Russian. Toimetaja Pille Eslon. Tallinna Ülikooli keelekorpuste optimaalsus, töötlemine ja kasutamine. [Tallinna Ülikooli eesti filoloogia osakonna toimetised 9 / Publications of the Department of the Estonian Philology of Tallinn University 9]. Tallinn: TLÜ kirjastus, 13 64; rodskaja.pdf V. Anastassia Zabrodskaja О прагматическом и психолингвистическом аспектах переключения кода [Pragmatic and psycholinguistic aspects of code-switching]. Tegevtoimetaja Irina Tabakova. ACTA et commentationes collegii Narovensis VIII. Tartu Ülikooli Narva Kolledži toimetised ja a rahvusvaheliste üliõpilasteaduskonverentside materjalid. Материалы международных студенческих научных конференций 2006 и 2007 гг. The Proceedings of Students International Scientific Conferences 2006 and Narva: Tartu Ülikooli Narva Kolledž, VI. VII. Anastassia Zabrodskaja Muukeelsete algkoolilaste vene-eesti/eesti-vene koodivahetus Kohtla-Järvel [Russian-Estonian/Estonian-Russian Code-Switching Among Russian-Speaking Schoolchildren in Kohtla- Järve]. Finest Linguistics. Proceedings of the Annual Finnish and Estonian Conference of Linguistics. Tallinn, May 6 7, Eds. Krista Kerge, Maria-Maren Sepper. [Tallinna Ülikooli eesti filoloogia osakonna toimetised 8 / Publications of the Department of the Estonian Philology of Tallinn University 8.]. Tallinn University: Tallinn University Press, Anastassia Zabrodskaja Koodivahetus kokkuleppimise strateegiana Kohtla-Järve vene-eesti kakskeelses suhtluses [Russian-Estonian code-switching as negotiation strategy in Kohtla-Järve bilingual language practices]. Toim. Helen Koks ja Jan Rahman. Konverentsikogumik Mitmekeelsus ja keelevahetus läänemeresoome piirkonnas. Võro Instituudi toimõndusõq 18. Võro: Võro Instituut, Book reviews for peer-reviewed research journals I. Anastassia Zabrodskaja. Multidisciplinary approaches to code switching (Studies in Bilingualism 41), L. Isurin, D. Winford and K. de Bot (eds.). Amsterdam/Philadelphia: John Benjamins Publishing Company, International Journal of Bilingual Education and Bilingualism (to appear in 2010). 11

12 II. III. Anastassia Zabrodskaja. Crosslinguistic influence in language and cognition. New York and London: Routledge. S. Jarvis, A. Pavlenko Sociolinguistic Studies (to appear in 2010). Anastassia Zabrodskaja. Anna Verschik, Emerging Bilingual Speech: From Monolingualism to Code-Copying. London, New York: Continuum, International Journal of Bilingual Education and Bilingualism (to appear in 2009). IV. Anastassia Zabrodskaja Douglas W. Blum, National Identity and Globalization: Youth, State, and Society in Post-Soviet Eurasia. Cambridge & New York: Cambridge University Press, 2007, viii + 226pp., h/b. Europe-Asia Studies, November, Vol. 61, No. 9, V. Anastassia Zabrodskaja Heidi Byrnes (ed.). Advanced Language Learning. The Contribution of Halliday and Vygotsky. Language in Society, Volume 38, Issue 2, VI. Anastassia Zabrodskaja Book review on The Sociolinguistics of Identity edited by Tope Omoniyi and Goodith White. Journal of Sociolinguistics, Volume 13, Issue 1, VII. Anastassia Zabrodskaja Alexander Bergs, Social networks and historical sociolinguistics: Studies in morphosyntactic variation in the Paston letters ( ). Language in Society, Volume 37, Issue 02, VIII. Anastassia Zabrodskaja Nicole Müller (ed.). Multilayered Transcription. San Diego Oxford Brisbane: Plural Publishing Inc, Journal of the International Phonetic Association 37 (3), IX. Anastassia Zabrodskaja Jessner, Ulrike. Linguistic Awareness in Multilinguals. English as a Third Language. Edinburgh: Edinburgh University Press, Laura Buszard-Welcher (ed.). LINGUIST List Eastern Michigan University; X. Anastassia Zabrodskaja Asesõnadest soome ja eesti vestluses [Pronouns in the Estonian and Finnish conversation. Ritva Laury (ed.). Minimal reference: The use of pronouns in Finnish and Estonian discourse. Helsinki: Finnish Literature Society, lk]. Keel ja Kirjandus, 12, XI. XII. Anastassia Zabrodskaja National Corpus of the Russian Language: A Good Example for Minor Languages. International Journal of Bilingual Education and Bilingualism. Volume 10, Issue 5, Anastassia Zabrodskaja Koodivahetusest ja keelevahetusest [On code-switching and language change. Helena Bani-Shoraka, Language Choice and Code-Switching in the Azerbaijani Community in Tehran: A conversation analytic approach to bilingual practices. Studia Iranica Upsaliensia 9. Uppsala: Uppsala Universitet, 2005.]. Keel ja Kirjandus, 8, Conference overviews for peer-reviewed research journals I. Anastassia Zabrodskaja. Seitsmes rahvusvaheline kakskeelsuse sümpoosion [International Symposium on Bilingualism 7]. Keel ja Kirjandus (to appear in 2009). II. Anastassia Zabrodskaja Anna Verschik. Vanad ja uued tuuled sotsiolingvistikas [Old and New Winds in Sociolinguistics]. Keel ja Kirjandus, 10, III. Anastassia Zabrodskaja Üleilmselt pragmaatikast [Pragmatics worldwide]. Keel ja Kirjandus, 11, IV. Anastassia Zabrodskaja Mitmekeelsete uuringute mosaiigist [Mosaic of multilingual research]. Keel ja Kirjandus, 10, V. Anastassia Zabrodskaja Põhjamaade keeled teise keelena [Nordic languages as second language]. Keel ja Kirjandus, 7, VI. Anastassia Zabrodskaja Sotsiolingvistika sümpoosionist doktorandi pilguga [Sociolinguistics symposium from the doctoral student s point of view]. Keel ja Kirjandus, 11, In addition, some unpublished data and materials are used. 12

13 LIST OF ABBREVIATIONS ACC ADES(S) ADJ CS DAT EL EST FEM FUT GEN ILL IMP INESS INF INSTR IPS L1 L2 LOC MASC ML MLF N NOM NP PL PART PARTC PREF P(A)ST Q RUS SG SLA SUF accusative adessive adjective code-switching dative embedded language Estonian feminine future genitive illative imperative inessive infinitive instrumental impersonal first language second language locative masculine matrix language Matrix Language Frame noun nominative noun phrase plural partitive participle prefix past question particle/marker Russian singular second language acquisition suffix 1PL, 2PL, 3PL 1SG, 2SG, 3SG 1 st person, 2 nd person, 3 rd person plural 1 st person, 2 nd person, 3 rd person singular 13

14 14 TRANSLITERATION OF RUSSIAN ITEMS Cyrillic Transliteration а a б b в v г g д d е e/je ё jo ж ž з z и i й j к k л l м m н n о o п p р r с s т t у u ф f х x ц c ч č ш š щ šč ъ ы y ь э é ю ju я ja In the examples, the Russian part is in italics, Estonian in bold. In the glosses and translations, Estonian items are in UPPER-CASE. Phonologically in-between items are underlined.

15 PREFACE The metaphors Life with Two Languages (Grosjean 1982), One Speaker, Two Languages (Milroy and Muysken 1995a) and One Mind, Two Languages (Nicol 2001) became a part of my life in 1999 when I began my studies at the university. Using Russian and Estonian became suddenly my everyday reality. Choosing different languages for different purposes and code-switching (henceforth CS) have all been part of my daily language use for the last ten years. Therefore, the issues discussed in the thesis have always been very close to my heart. As a multilingual speaker, a second language educator and a researcher, I have always been fascinated by the notion of language contact and change. Presently, my research interests comprise language contact, dynamics and change in Estonia, linguistic landscape and ethnolinguistic vitality. The main aspects of my research in were: 1. Language choice and CS patterns in the speech of Tallinn and Narva Russian-speaking students; 2. Regional dimension of language contacts in Estonia. At the beginning, I supposed that bilingual Tallinn and predominantly Russian-speaking Narva would probably show entirely different patterns of CS functions and language choice. The questions of my interest here were: Is the Russian language resistant to the Estonian influence or converging towards Estonian? And what kinds of change occur in the Russian language? 3. Self-reports on language use. Self-reported data are not always reliable, which is a disadvantage in sociolinguistic research. This type of data has been accused of being subjective and impossible to check. As Romaine (2000: 27) describes it self-reports are subject to variance in relation to factors such as prestige, ethnicity, and political affiliation, etc. Doing my research, I tried to answer the following questions and make the best use of the answers got: Can use of questionnaire data be valuable in understanding the processes of language choice, CS and language change? What is the best way of using questionnaires in sociolinguistic studies? 4. My role as a researcher in the fieldwork perspective. Among sociolinguistics, Labov (1972) was one of the first to systematically highlight the problems associated with participant observation. Another problem, which is relevant, is the effect of the observer s presence on the linguistic behaviour of the persons under study (Milroy 1987a: 60). As Milroy (1987b: 43) argues, the fieldworker as a participant and the fact that tape recording is taking place are important components of the total communicative situation. She herself claimed that she acquired a status which was neither that of insider nor that of outsider but something of both a friend of a friend, or more technically, a second order network contact (Milroy 1987b: 43). How successfully could I be an insider and outsider at once? My first micro-sociolinguistic research dealt with Russian-Estonian CS functions among Russian-speaking schoolchildren of Kohtla-Järve (Zabrodskaja 2005) where I belonged to the speech community. In , dealing with bilingual language behaviour of Russian-speaking students in Tallinn University and the Narva College of the University of Tartu, I had a number of advantages in my fieldwork as a teacher, because I had an access to casual speech behaviour of the students. Still, the question remained of in what way I could influence participant language behaviour and tape recordings of informal speech events during interviews. All above-mentioned issues I discuss in my thesis that is primarily based on the following three publications (in the text of an analytic overview referred as Paper I, Paper II and Paper III): 15

16 I. Anastassia Zabrodskaja Towards establishing the matrix language in Russian- Estonian code-switching: A corpus-based approach. Edited by Tsiplakou, Stavroula, Marilena Karyolemou and Pavlos Pavlou. Language Variation European Perspectives II. Selected papers from the Fourth International Conference on Language Variation in Europe (ICLaVE 4), Nicosia, Cyprus, June, Studies in Language Variation 5. Amsterdam: John Benjamins Publishing Company, II. Anastassia Zabrodskaja. Evaluating the Matrix Language Frame Model on the basis of a Russian-Estonian code-switching corpus. International Journal of Bilingualism. Special issue Language Contacts in the Post-Soviet Space. Guest editor Anna Verschik, 2009, 13 (3), to appear. III. Anastassia Zabrodskaja. Morphosyntactic contact-induced language change in young Estonia s Russian speakers variety. The interplay of variation and change in contact settings Morphosyntactic studies. Edited by Isabelle Léglise & Claudine Chamoreau. Studies in Language Variation. Amsterdam: John Benjamins Publishing Company (to appear in 2010). Please notice that the bibliography of the whole thesis is listed at the end of the analytic overview. It means that the part References includes every single cited source from Preface of the analytic overview to the Papers I III. 16

17 1. INTRODUCTION The last decade has witnessed a rise in scholarly interest towards the post-soviet language situation. The agenda is dominated by research in language policy and macro-sociolinguistics (Korth 2005) as well as overall descriptions of the status change of Russian (Pavlenko 2008a). The post-soviet language situation can be viewed as a language laboratory since one can explore contacts between the same languages and Russian in a changed sociolinguistic situation (Verschik 2009: 2, 6). 1 To reword it in a more poetic way, I would quote Jacobs (2005: 271): The list of languages present in the [sociolinguistic] soup remained the same, while the recipe changed. While Russians settled in Estonia, they mostly remained monolingual, whereas the titulars became bilingual (M. Rannut 2008). After the collapse of the Soviet Union, Russian-speakers are facing the necessity of becoming bilingual. Case-studies on contact-induced change in Russian and the languages of post-soviet space are scarce (Muhamedowa 2006) and seldom discuss what happens when two languages rich in inflectional morphology come together in a language contact situation (Verschik 2008). The thesis examines this issue in the context of Russian-Estonian language contact. Estonian is a Finno-Ugric language spoken by approximately 1 million people, mostly in the Republic of Estonia. Estonian is an agglutinating inflectional SVO language, which is known for its rich morphology (for a comprehensive overview see Erelt 2003). Russian, a fusional language, belongs to the family of Indo-European languages (for a comprehensive overview see Comrie, Stone and Polinsky 1996). Within the Slavic branch, Russian is one of three living members of the East Slavic group, the other two being Belarusian and Ukrainian. Russian has a developed inflectional morphology but to a lesser extent than Estonian. The contact situation between languages with highly developed inflectional morphology provides new empirical data relevant to test different approaches introduced in language contact theory Rationale for the thesis The two primary aims of this Section are to provide a brief historical overview of the Estonian sociolinguistic studies that took place prior to the preparation of my doctoral thesis and to define the scope and terminological apparatus of the current research. The novelty of the Estonian context quite naturally includes some challenges. On the theoretical level, I can note several issues. Most of the contact linguistic theories have been developed in the USA and Western Europe, often containing English or other European language. These theories specifically address a Western sociolinguistic context. It is a challenge to analyze bilingual extracts when at once both languages in the language pair are most probably unknown to Western academic readers. What is more, both languages in contact are rich in inflectional morphology whereas Estonian has even more developed inflectional morphology than Russian. Also the current sociolinguistic context is new for Western contact linguists. For example, the situation of Russian-speakers in post-soviet Estonia is different from that of Turkish immigrants in Holland (see Backus 1996) or of the Frisian autochthonous minority in Holland (see Spolsky 2004: 199) etc. 1 Two years earlier Blum (2007: 5) stated that post-soviet Russia, Kazakhstan, and Azerbaijan have been a wonderful laboratory for studying globalization and national identity since

18 The fact that the sociolinguistic situation in Estonia has changed a lot attracted the attention of Estonian sociolinguists; among them I would mention works by (in alphabetic order): Ehala (1994, 2009) who mainly works on Russian influences on Estonian, M. Rannut (1994, 2004, 2008) who does his research on language policy in Estonia, Ü. Rannut (2005) who has described in details different language environments in Estonia, and Verschik (2002, 2004a, 2004b, 2005, 2006, 2007, 2008) who has contributed numerous papers on Russian-Estonian CS and Russian-Estonian language contacts. Last but not least, my master thesis, later published as a monograph, was the first attempt to systematize and analyze the linguistic and pragmatic consequences of multilingual communication among Russian-speaking schoolchildren with a special focus on Russian-Estonian CS in the town of Kohtla-Järve (Zabrodskaja 2005). Thus, sociolinguistics is prominently represented in Estonian linguistics and there are previous studies and data on bilingual behaviour I can rely on. Before proceeding to the terminology used in my thesis, I wish to mention the most recent developments in the field of Russian-Estonian language contact research that are given attention throughout my thesis and that have led to the recognition of the self-evident valuable fact that Estonian Russians can interact in the Estonian sociolinguistic context nowadays. To the best of my knowledge, Verschik (2008) is the first English language book specifically dedicated to language contacts in Estonia. Drawing on Johanson s (1993) code-copying model, Verschik presents an analysis of code-copying in Estonian compound nouns of the type N NOM-N NOM and N GEN-N NOM, Estonian analytic verbs, and discourse-pragmatic words. I also explore the genitive constructions of the type noun + noun, but analyze them as typical congruent lexicalization patterns. As argued by Jarvis and Pavlenko (2008: 28), the methodology one adopts for the linguistic analysis depends on the scope of the investigation. The scope of my thesis, on the one hand, is the exploration of Russian-Estonian contact phenomena in terms of convergence whose result can be seen in the emergence of a new language variety that I call Estonia s Russian. On the other hand, I also look at sociolinguistic and grammatical factors that can both trigger and constrain this process, and claim that the emerging patterns are not equivalent to the two monolingual ones. In my terminology, I refer to convergence as the point at which, as a result of contact-induced morphosyntactic changes, amalgamated constructions (or congruent lexicalization) appear. The primary impetus for this approach is my belief that during language contact not only phenomena emerge that can be linked to a particular monolingual variety but also new items that are not Russian anymore and cannot be distinguished as Estonian either (see Papers I III). To some extent my approach has similarities to Johanson s (2002) code-copying framework, but there are also important differences that will be dealt with in Chapter 5. This is also the reason why I do not use his terminology and theoretical approach. Another challenging terminological question is raised by the different theoretical accounts of convergence used in the fields of contact linguistics and bilingualism. One of the main problematic issues I raise in Section 3.2 is to what extent the linguistically-based analysis of the mechanisms of language change is fragmentized because of the different terminology involved. Of course, much further research needs to be done on this question but I do the first step in my thesis and employ only the term convergence when talking about contact-induced morphosyntactic similarities in two languages. 18

19 Although I refer to Verschik s (2002, 2004a, 2004b, 2005, 2006, 2007, 2008) research where appropriate, more data collection on Russian-Estonian contact phenomena is needed and the findings cannot be said to be representative for the entire Russian-speaking community in Estonia. The possible development of future research is addressed in Chapter Research questions in the current study On the basis of Russian-Estonian bilingual data, my thesis presents a study of CS between Russian and Estonian and also introduces some morphosyntactic features of Estonia s Russian that have been emerging as a result. Following Grosjean (1982: 145), the term CS is employed here as a general term that designates the alternate use of two or more languages in the same utterance or conversation. The major goals of this thesis are to answer the following research questions: 1. What kind of morphosyntactic changes occur in Estonia s Russian spoken by Russianspeaking students? 2. What Russian-Estonian contact data can and cannot be explained by the Matrix Language Frame model (henceforth MLF model) and its modifications? 3. Can other models provide a better explanation to Russian-Estonian contact data that the MLF model fails to explain? 4. What Estonian structural features appear to be attractive to bilingual speakers? Contributing to the research on contact-induced language change, the present study takes a closer look on grammatical features of Russian-Estonian CS and on the convergence of local Russian towards Estonian Organization of the thesis This thesis contains an analytic overview and three papers. The analytic overview is organized as follows. Chapter 2 functions as the methodological framework of the research. Chapter 3, Theoretical framework, introduces the reader to the grammatical approaches to CS. This chapter is a basis for the analysis conducted in the empirical studies (see Papers I III) presented in Chapter 4, where contact phenomena in Estonia s Russian is described. In Chapter 5, the findings are described and viewed in the light of three theoretical approaches: congruent lexicalization (Muysken 2000), composite ML (Myers-Scotton 2002), and selective copying/mixed copies (Johanson 2002). General conclusions from the results and directions for future research are given in Chapter 6. One of the major themes running through the three papers is the explicit concern with methodological refinement of the MLF model. The matrix language (henceforth ML) is determined by which language s grammar dominates in a bilingual clause. The ML sets the grammatical frame in mixed constituents into which items from the embedded language (henceforth EL) are inserted. ML + EL constituents have morphemes from both languages. Alongside methodological issues, and especially with regard to the establishment of the ML, the papers explicitly address theoretical issues, such as the morpho-phonological, morphosyntactic and sociolinguistic aspects in the exploration of language use and change. Introducing the MLF model (Myers-Scotton 1993a, 1997, 2002) through the example of Russian-Estonian bilingual data, Paper I, Towards establishing the matrix language in Russian- Estonian code-switching: A corpus-based approach, lists some cases where the principles for 19

20 the determination of the ML and EL do not work. The paper also covers variation between the various types of Russians in Estonia. The current MLF model deals only with morphosyntax and says nothing about phonological adaptation of code-switched items. Paper II, Evaluating the Matrix Language Frame Model on the basis of a Russian-Estonian code-switching corpus, continues the discussion raised in Paper I and addresses the issue of identifying the ML in a more detailed way, pointing out that in many instances the ML cannot be determined unambiguously and emphasizing that the theory of the MLF model should take into consideration degrees of morphological integration as well as the phonological adaptation of EL items. As integration and adaptation are gradient phenomena that include many aspects, some challenges for the MLF model are presented. Here are some examples: Estonian code-switched items exhibit different degrees of morphological integration and/or phonological adaptation. Furthermore, morphosyntactic integration of Estonian code-switched nouns into the Russian matrix does not happen on an all or nothing basis. With regard to identifying the ML, these findings should be used as a complementary tool. A number of interesting findings focusing on the connections between classic CS, congruent lexicalization (also a type of CS, to be considered below), and convergence are discussed in Paper III, Morphosyntactic contact-induced language change in young Estonia s Russian speakers variety. The paper starts with an overview of the overt use of Estonian (compound) nouns in terms of congruent lexicalization. Here, the research shows the link between the use of internationalisms (or common stems/words) and lexical facilitation of CS leading to structural changes in bilingual genitive constructions. Congruent lexicalization also includes cases where a code-switched clause consists of an auxiliary and an infinitive form, in which not only Estonian grammatical influence but also double marking is present. The paper aims to contribute to the identification of crucial factors in contact-induced morphosyntactic change. All three contributions also address issues of methodology in various aspects of data collection and analysis. The discussion is based on Russian-Estonian data taken from the Russian- Estonian CS corpus that is described in Papers II and III in more details. By shifting the focus to data collection procedures, I now turn to the methodological discussion. 20

21 2. METHODOLOGICAL FRAMEWORK In the following Chapter, methodological issues regarding the data are presented and discussed. As I have already mentioned in Section 1.1, so far, most of the Western literature on contact linguistics addresses European contact situations where a sociolinguistically dominant language is one of the Indo-European languages (English, German, French or Dutch) and this gets into contact with an immigrant language (e.g. Turkish or Arabic). In this situation, my study on post-soviet Russian-Estonian language situation can certainly add new empirical evidence from a novel sociolinguistic context. Russian-Estonian language contact can be characterized with an adjective multiple because, on the one hand, Estonians have some competence of Russian as L2 and, on the other hand, Russians have started extensively using the Estonian language as L2 (see also Verschik 2007). The language situation in Estonia (see also Pavlenko 2008b on Russian in the post-soviet space in general) presents a sociolinguistic constellation previously unknown in Western sociolinguistics. While Russian minorities in the classical sense existed and do exist here (Old Believers who live on the western coast of Lake Peipus), the rest of Estonia s Russianspeakers cannot be equalled with minorities and immigrants in the conventional sense (see Pavlenko 2008c on terminology problems, also see Ozolins 2003 on the Baltic countries). In Section 2.1, I will describe the current sociolinguistic situation in Estonia. This overview will be relevant for a general understanding of linguistic processes in Estonia. In Section 2.2, the methodology of data collection for the current study will be addressed. To introduce the mechanisms and nature of contact-induced change, the grammar of CS, and the emergence of Estonia s Russian contact variety, Section 2.3 contains some discussion on the balance between the synchronic and diachronic dimensions in the current study A brief look at the sociolinguistic situation in Estonia Despite its small territory and population size (1.3 million), Estonia offers a lot of challenges for a scholar who does research on Russian-Estonian language contacts. These include internal diversity within the Russian-language community (indigenous group vs. Soviet era newcomers and their descendants, see Verschik 2005: 288, 307), regional variation in language environments (bilingual Tallinn, predominantly Russian-speaking North-East and predominantly Estonian-speaking rest of the country, see Ü. Rannut 2005), inter-generational differences in language proficiency (see Vihalemm 2008), etc. What makes the Estonian context especially interesting is its turbulent transition from Soviet to post-soviet. When Estonia became independent in 1991, Estonian became the single official language. Most non-estonians residing in Estonia are second generation settlers and their families, who came to live here after World War II during the period of Soviet occupation as late as 1945 Estonians made up more than 97% of the population (see M. Rannut 1994, 2008). During the Soviet occupation, the percentage of Russians was constantly on the increase and reached 30.3 % according to the last Soviet census of Due to the fact that the migration of Russian-speakers was part and parcel of demographic and language policy of the Soviet authorities, the Soviet-time newcomers and their descendents are better described as colonizers (Ozolins 2002). At the same time the developing situation between Estonians and Russian-speakers cannot be simply described as a classical case of majority-minority, indigenous-immigrants or indigenous-colonizers because this is a 21

22 particularly new sociolinguistic context (Pavlenko 2008c). As Western approaches to linguistic minority rights cannot be applied wholesale to post-soviet Estonia, special adaptations are needed to current theories of minority rights (see Ehala 2008). In present day Estonia, Russians constitute 25.7% of the total population. Estonians comprise the bulk of the population 68.6% (Verschik 2005). Knowledge of the Estonian language among L1 speakers of Russian has increased from 14% in 1989 to 44.5% in 2000, according to 2000 Population and Housing Census data. Russian is a language widely spoken in Estonia nowadays and has recently become influenced by Estonian, the official language. Thus, observations of this contemporary and very dynamic language situation are important because of its uniqueness. To reflect this diversity, the bilingual data I analyze come from different sources collected for the Russian-Estonian CS corpus that is currently under construction Methodology of data collection The major research question in the sociolinguistic study of language change is the explanation [my emphasis A.Z.] of innovation. This is the starting point of change and of the diffusion or spread of changes in society. To answer this question, corpus data is required because it reflects spontaneous bilingual speech (Backus 2009: 308). The data analyzed in my thesis give evidence of contact-induced language change occurring in Estonia s Russian spoken by Russian-speaking students in two different language environments. Furthermore it is assumed that a grammatical analysis of Russian-Estonian CS can show how a new non-monolingual variety emerges, i.e. what types of morphosyntactic changes can be registered and what role is played by both languages in the emerging grammar of Estonia s Russian. Milroy (1987a: 27) argues that the objectives of a piece of research to a very large extent dictate methods of speakers selection. In my case, it means that methods of bilingual informants selection necessarily need to be related to the research goals. Most of the data for this particular study were collected by me among young Russian-speakers in Tallinn University and at the Narva College of the University of Tartu, that is, in bilingual Tallinn and the predominantly Russian-speaking North East respectively. The research was carried out between 2005 and The data come from recorded interviews, informants self-reported language use, and an analysis of written homework in Estonian and/or Russian. Held both in Russian and Estonian, interviews were individual as well as in-group. Having graduated from schools with Russian as the primary language of instruction, the informants have acquired Estonian in school and their proficiency in Estonian varies considerably. As all of the students have the Russian language as their mother tongue and home language, it can be argued that the informants have almost equal knowledge of their L1. They are all receiving higher education in the Estonian language and use Estonian as their L2 in both official and unofficial situations, in oral and in written communication. I subscribe to the view proposed by two experts in ethnomethodology, micro-ethnographers Varenne and McDermott (1998: 177), who notice that it is not easy to capture people in the real time of their practice. This means that it is important to use multiple sources of data because researchers should not rely on any single source of data, interview, observation, or instrument (Mills 2003: 52). In research terms, such approach is called triangulation. Jarvis 22

23 and Pavlenko (2008: 34 and references therein) also point to the fact that the most useful studies are often those that investigate patterns of language use across different types of data. To show the most exact and sufficient picture of students language behaviour as possible, my data also include: 1. Various pieces of student work and curriculum material; 2. A field diary ( ); 3. Audio-recorded lectures, seminars, and students spontaneous everyday language practices; 4. Interviews with Russian-speaking students on the following topics: university, local life, language use etc. There are various ways of data collection, processing and analysis. To store the data, I use the Russian-Estonian CS corpus structured on the basis of LIDES (LIPPS Group 2000). In addition to my data, the corpus consists of a wide range of speech and written text samples collected by different researchers. Several sub-corpora have been constructed within the corpus: (a) bilingual TV talk shows; (b) data from bilingual Tallinn; (c) data from the predominantly Russian-speaking North East (Narva and Kohtla-Järve); and (d) data from the predominantly Estonian-speaking Ämari (Estonian Armed Forces). The former three are still worked on all the time. The corpus contains CS samples from a variety of speakers with differing levels of proficiency in Estonian, and in different settings, and together gives a comprehensive picture of sociolinguistic, stylistic and intergenerational variation. For a detailed description of the corpus, including an explanation of the transcription and encoding processes, see Papers I III and Zabrodskaja (2007a, 2007b). On the one hand, such a wide range of data allows analyzing common language patterns produced by bilingual speakers in different settings. On the other hand, a critical contact linguist might ask if such a broad variety of data allows drawing reliable conclusions about a model that could subsume all existing phenomena in bilingual speech. I would argue that this new linguistic data undoubtedly enriches the understanding of CS and contact-induced language change mechanisms. My data are diverse and allow the analysis of grammatical aspects of CS; however, my purpose here is not the formulation of a new model but rather to test the MLF model on the example of languages that are rich in inflectional morphology. I am not convinced that constraint-based approaches to CS are the only way to deal with it. The history of the grammatical approach [as well as my own study] has shown that no constraints are absolute (see Section 3.1 together with its subsections). Johanson s (2002) code-copying model is an example of a productive approach that is not constraintbased. Another statement relevant to this discussion is that a sociolinguistic situation can change very quickly, especially in such a relatively new and dynamic society as Estonia and new patterns of bilingual speech can supplant the previous ones. Although these new patterns and phenomena are relevant, their interpretation would gain a lot from our knowledge of the previous context. Thus, collecting as much data as possible is necessary. In addition, there is an important empirical question that has not been asked before, namely, whether any regional differences in Russian-Estonian CS grammar and Russian language use exist. This question can be answered only if a researcher uses all possible data sets. Therefore, the data should be diverse. Among other things, it is necessary to answer the what-question, i.e. what contact-induced language changes occur if at all, because they are not universal. It should be pointed out that this question remains unanswered so far. Only then it is possible to ask how and why these changes occur and why in this way. 23

24 Finally, as for diachronic research, it would be too premature to speak about any reliable time-proved model the validity of which was tested, as the history of intensive Russian- Estonian language contacts is too short to find correlations between linguistic and social phenomenon. Now I look specifically at what phenomena are of a synchronic and which ones more of a diachronic nature in my data Synchronic and diachronic dimensions in collected data Developing the metaphor of language laboratory, I would emphasize that multilingual speech spread out very quickly after the collapse of the Soviet Union in 1991 and became everyday reality in the linguistic repertoire of Estonia s Russian-speakers. Thus, a researcher has a unique possibility to observe contact phenomena developing from zero (see also Verschik 2008: 42 43). The most difficult matter of my data analysis is seeking an answer to this challenging question: when are we dealing with innovations or change in progress, and when with change that has already occurred? This question can not be answered very easily. Intensive Russian- Estonian language contacts are a very recent phenomenon and it is difficult to separate the synchronic and diachronic dimensions. This is noteworthy because a good deal of research in the field of contact linguistics has dealt either with grammar of CS (incl. constraints) or with convergence phenomena (or contact-induced change) between two languages in contact. Looking specifically at CS grammar, my study has detected some innovations or, maybe, some changes in progress. I would claim that if some contact-induced phenomena have been registered (like word order change in seemingly monolingual genitive constructions); there is a good chance that it may be a change in progress or, maybe, even completed change. My research has also shown that to the question whether the variety Estonia s Russian exists now, a clear yes-no answer can not be given. A sociolinguist can talk about a continuum, or a cluster of varieties that can be labelled under the umbrella of Estonia s Russian. Bilingual speech of Tallinn and Narva Russian-speaking students comprises the bulk of my data. Unfortunately, no comparable data are available from the 1990s or early 2000s that would allow making any firm conclusions about the nature of change diachronically. At present, I label some phenomena with the terms innovation or change in progress, because the border between synchronic and diachronic dimensions is really fuzzy in my study. Other contact data serve as a background for making some generalizations for the synchronic study of the grammar of CS. With the methodological discussions in mind, I now turn to the presentation of the theoretical paradigm for the current research. 24

25 3. THEORETICAL FRAMEWORK The following discussion is intended to provide a state-of-the-art description of research on the grammar of CS including a look at the sometimes confusing use of the term convergence in the literature on language contact and change. The theoretical background of my study is discussed in order to inform the reader about the current situation in the field of contact linguistics and to facilitate understanding of some challenges for further research in this area that will be discussed in Chapter 4, in addition to presenting the results on contact phenomena found in Estonia s Russian The grammatical approach to CS There have been many attempts to describe the structure of code-switched utterances and to identify the linguistic principles and constraints that govern their production (e.g. Pfaff 1979 on Spanish-English CS; Berk-Seligson 1986 on Spanish-Hebrew CS; Leung 1988 and Chan 1998 on Cantonese-English CS). To date, researchers analyzing bilingual speech have asked whether there are cases where CS cannot occur. Because of this focus, the grammatical approach dealt mostly with intrasentential switches (see an excellent overview by Halmari 1997: 67 69). Some researchers have also worked on formulating universally valid linguistic constraints on CS patterns (Sankoff and Poplack 1981; Berk-Seligson 1986; Di Sciullo, Muysken and Singh 1986). Others investigate a number of purely synchronic constraints that allegedly restrict CS at the syntactic, phonological or morphological levels (e.g., Poplack 1980; MacSwan 1999). Research in this field has largely concentrated on finding universally applicable grammatical constraints on CS, but it has not had much success so far (Gardner-Chloros and Edwards 2004: 104). Below I provide a brief overview of several structural constraints that have been claimed to hold for intrasentential switches General assessment of the MLF model The most important reason for choosing the formal grammatical approach introduced by Myers-Scotton in her numerous contributions (frequently together with Jake) is that her theoretical framework is the most developed and influential one and has attracted a lot of critical attention over the years (e.g. Kamwangamalu 1997, 2000; Nivens 2002; Boussofara-Omar 2003a, 2003b; Yu Liu 2008). As is argued by Muysken (2000: 18), Myers-Scotton was the first among CS researchers who brought psycholinguistic, sociolinguistic and structural perspectives together, providing a deeper explanatory level. From 1993, Myers-Scotton and Jake have been working on the MLF model, improving it through notions such as the System Morpheme Principle, the Morpheme Order Principle, the 4-M model, the Abstract Level model, the Uniform Structure Principle (henceforth USP), and Asymmetry Principle (see Myers-Scotton 1993a, 1997, 2002, 2006, 2008; Jake 1998; Myers- Scotton and Jake 2000, 2001, 2009; Jake and Myers-Scotton 2009 to name some major contributions). Myers-Scotton (1997: 82) states that the ML is the base language which sets the grammatical frame in mixed constituents, that is, the frame into which items from the other language, the 25

26 EL, are inserted. Myers-Scotton has undoubtedly done much to clarify her model, trying to differentiate morpheme types to allow better understanding of the ML and the EL. One of the major problems with her work is that she tends to introduce revised principles for ML determination in quick succession. For example, when introducing the System Morpheme Principle, she claims that the language that is identified as the ML is the only one which can supply a certain type of system morphemes in bilingual constituents (Myers-Scotton 1997). Three years later, Myers-Scotton and Jake (2000) label such morphemes late outsider system morphemes and predict them to come from the ML. This idea is served under the umbrella of the 4-M model morpheme classification. For the first time, Myers-Scotton and Jake (2000) introduced the 4-M model where four types of morphemes are described. These are content morphemes and three types of system morphemes. The basic criterion for the division is whether morphemes are conceptually-activated or structurally-assigned. The former are content morphemes (such as nouns, verbs, adjectives) and the latter are early system morphemes (plural affixes, many determiners, derivational affixes, and verbal prepositions that can modify thematic structure) or late system morphemes that, in turn, divide into bridges and outsiders. While bridges join phrases together to create a larger constituent, outsiders depend on information outside of the constituent in which they occur. Jake and Myers-Scotton (2009) talk about the USP when trying to identify the ML in mixed constituents. According to them, the USP recognizes the universality of what constitutes a language (Myers-Scotton 2002; Myers-Scotton and Jake 2009). They claim that any linguistic constituent has a uniform structure that is maintained whenever it appears. For bilingual constituents, the USP preferences the structure of the language that frames the bilingual clause and determines the ML. To conclude, their approach distinguishes between the ML and the EL. Although both may supply all types of morphemes, the role of EL is potentially more limited. And the ML supplies the morphosyntactic frame of a bilingual utterance. I will return to the MLF model in connection with Russian-Estonian bilingual data analysis in Section The Equivalence Constraint The Equivalence Constraint introduced by Poplack (1980) and Poplack, Wheeler and Westwood (1990) aims to predict where switches are likely to occur. If two languages have similar syntactic structures, a switch can take place because the switch does not violate the structure of either language. Various studies have shown this constraint to hold (see Poplack 1980; Poplack and Meechan 1995; Deuchar 2005) or not (Bentahila and Davies 1983; Berk-Seligson 1986; MacSwan 2000; Toribio 2001) in bilingual data from all over the world. To conclude, although finding constraints is certainly legitimate, the Equivalence Constraint doesn t provide an explanation. It rather proposes a generalization which has then been proven empirically problematic Government Constraint Similarly, another constraint on intra-clausal CS appeals to the notion of government referring to the relation between the head of the construction and its complement. The Government Constraint has been applied to CS by Di Sciullo, Muysken and Singh (1986). According to 26

27 these authors, CS (they call it mixing ) can occur at such points of a sentence where there is no government relation between the elements, but is prohibited where a government relationship prevails (Di Sciullo, Muysken and Singh 1986: 4). Elements related to one other by government must be drawn from the same lexicon, or, in their terms, must have the same language index The triangle model In this section, I discuss the notions of insertion, alternation and congruent lexicalization, stressing their role in contact-induced language change. Muysken introduces these as the three main types of code-mixing. To avoid the terminological mess, in my thesis the same phenomena are described as three types of CS or the triangle model. In Muysken s (2000: 60) view, insertion is used to refer to such types of switching where a base language structure is the matrix frame into which lexical items or entire constituents are inserted from another language. In case of alternational CS, L1 and L2 remain relatively separate in the clause (see Muysken 2000: 96). Congruent lexicalization takes into account both morphemes and combinatorial principles (morphosyntax, word order etc). Here, both languages jointly provide the grammatical structure of the clause and the vocabulary elements (see Muysken 2000: 122). Muysken (2000: 9) tries to connect CS types with the duration of the contact situation (e.g. insertions exemplify the speech of first generation immigrants and as the extent of bilingual contact grows, insertional code-mixing may give way to either alternation or congruent lexicalization). In my data, the situation is quite complex because it is difficult to distinguish between generations, and Russian-speakers are not classic immigrants or members of an ethnic/linguistic minority (see Ehala 2008; M. Rannut 2008; Verschik 2008: 25 39; Zabrodskaja 2009: ). In the Estonian context, young people who technically belong to the second or even third generation of Russian settlers in Estonia are the first generation of Estonian L2 speakers. Peeling the CS onion, findings of the empirical studies conducted in Papers I III show that Russian-Estonian CS can be intrasentential, i.e., take place within a sentence or within a clause, or intersentential, i.e., occur across two sentences or clauses as well as constituent- or word-internal. Such mixed production is probably caused by the fact that the social and demographic conditions favour different degrees of proficiency in the Estonian language, different types of interaction (one talks only about everyday topics while another gives lectures or works in the Estonian language environment) and that is why the production of differing types of CS by Russian-speakers varies Monolingual bias and constraints In Section 3.1, I reviewed four distinct but interrelated theoretical approaches to the grammatical analysis of CS. One feature that they all have in common is that they unquestionably deal with morphosyntactic features of CS. While considering the constraints and principles, we have seen that each direction represents a different position with respect to CS. If constraints state what is possible, then general principles of the MLF model suggest what is probable. If we compare Myers-Scotton s (1993) principles to Poplack s constraints, then we see that both approaches focus on the morphosyntactic level. While the latter lists restrictions 27

28 on possible switches, the former suggests principles that explain morphosyntactic structures on the basis of monolingual equivalents. Poplack and Meechan (1995: 199) posed the following research question: Do speakers operate with a single base grammar which is on occasion overlaid with lexical items from other languages, or are different grammars activated at different times? They further ask: If the latter is the case, what structural principles govern the juxtaposition? And they start comparing code-switched structures with their monolingual equivalents. In all studies, constraints are based on the rules of two monolingual grammars, and the two languages are hypothesized to be code-switched in accordance with these rules. Comparison of bilingual utterances to monolingual grammars brings up the notion of monolingual bias, i.e., a tacit assumption that a monolingual speaker can serve as a yardstick (as often happens in mainstream SLA research). Monolingual bias have governed the field of contact linguistics since times when its acknowledged pioneer Weinreich ([1953] 1966: 7) stated that based on the assumption that every speech event belongs to a definite language (Lotz 1950: 712) it is possible to determine in an utterance some elements which belong to another language. He calls the non-belonging elements borrowed or transferred. I argue against the claim that every speech event belongs to a definite language. One should bear in mind that sometimes it can be hard to determine what element belongs to what language and why, for example in Finnish-Estonian data where contact is between two closely related languages. Verschik (forthc.) shows very convincingly that there are compromise forms and new creations that are not unambiguously determinable. Thus, an analysis of CS as something that combines two monolingual varieties is wrong per se. It is not always possible to establish firmly what item belongs to what language in Russian-Estonian bilingual utterances, even though the languages belong to different language families. In the middle of the 1990s, Milroy and Muysken (1995b: 2 3) pointed out that in linguistics a monolingual speaker in a homogenous speech community is regarded as the norm. In contact linguistics, researchers are not concerned with deviation from the respective monolingual norms. However, a monolingual bias exists in the sense that bilingual speech is frequently analyzed in terms of two monolingual varieties, e.g. when linear CS models postulate constraints on switches that create morphosyntactic conflict ; constituents in CS that are wellformed in the terms of respective monolingual grammars etc (Backus 1999a; Auer 2007). The results of bilingualism for an individual speaker cannot be considered a simple overlay of one language over another (Auer 2007) because the resulting amalgamated constructions and items cannot always be labelled as belonging to a particular variety (see Paper III, Sections 5, 6.1 and 6.2). To criticize the monolingual bias more fundamentally, I would argue that change presupposes a dynamic system, and then a monolingual bias simply makes no sense (see also Cook 1997; Backus 1999a; Ellis 2006; Auer 2007; Jarvis and Pavlenko 2008: 17; Jessner 2008). Researchers such as Myer-Scotton assume oftentimes (unconsciously, I presume) as if the ML and the EL are indeed monolingual varieties. Accordingly, two monolingual grammars of the monolingual standard varieties that are in contact allegedly operate in the code-switched clause. But the phenomenon is more complex: grammar can mean different things. As Gardner-Chloros and Edwards (2004: 105) demonstrate, the term grammar can have a variety of meanings (dictated by particular linguistic rules or theories of principles and constraints that underlie the syntax and morphology etc) and it is not clear whether and how common assumptions about (monolingual) grammars are applicable to non-monolingual speech. 28

29 Alvarez-Cáccamo (1998: 36) argues against the monolingual bias of constraints: In order to argue convincingly for or against the existence of CS constraints and CS grammars based on the two monolingual ones (Sankoff and Poplack 1981: 10), research should first convincingly prove that (a) speakers who code-switch possess two (or more) identifiable linguistic systems or languages, each with its identifiable grammatical rules and lexicon; and (b) code-switched speech results from the predictable interaction between lexical elements and grammatical rules from these languages. None of these assumptions, I believe, is proven yet. In this study, the continuously changing sociolinguistic situation and growing knowledge of Estonian among Russian-speakers has to be taken into account. For a microsociolinguist, patterns of Russian-to-Estonian communication are more subtle and diverse than just the opposition between two monolingual varieties. This is especially true of Tallinn, where people with differential degrees of command in Estonian and in Russian interact on an everyday basis in institutions, over the counter, in universities, etc (Verschik 2002, 2004b; Zabrodskaja 2006a, 2006b, 2007). Another question that a microsociolinguist might ask is exactly what kind of Russian (or Estonian) the speakers use. The utterances can be produced from a theoretically postulated system called Estonia s Russian (or Estonian Russian), a new variety (see also Verschik 2006, 2008). I would also emphasize that Estonia s Russian is not one particular language variety; it is a quite broad label because this new phenomenon has not got crystallized yet and definitely varies across speakers. How internally consistent it is, and how different from standard Russian, is an empirical matter, and will decide on whether we can indeed legitimately talk about Estonia s Russian as a new variety, or as a cluster of varieties. To conclude, my data show that code-switched constituents may (see Paper III, Section 6.2) but do not have to be, well-formed grammatically (see Paper III, Section 6.1). Amalgamated constructions violate both the Equivalence Constraint and the MLF model principles because they lead to a typical congruent lexicalization pattern. Russian-Estonian bilingual data show that CS triggers convergence, which, in its turn, can trigger a switch (see also Backus 2005). If CS causes convergence, then it is possible to argue against the claim that some synchronic constraints, restricting CS at whatever level, can potentially restrict convergence and, as a result, make impossible the potential structural change itself. However, the greatest challenge for a reader here is the question of what is meant when different researchers talk about, at first glance, the same topic convergence because, in the field of contact linguistics, the term can be used in different, sometimes strongly opposed ways, as I demonstrate below. Convergence is defined step by step in the next Section Convergence In this section, different notions of convergence and different approaches underlying use of the term are briefly discussed, pointing out where possible their mutual similarities. First, I want to start with a small overview of the terminology used in the field of contact linguistics when describing contact-induced grammatical changes. A look at the history of studies on contact-induced language change so far makes clear that to date researchers use notions such as interference (Weinreich [1953] 1966), code-copying (Johanson 2002), cross-linguistic influence (Jarvis 2002; Jarvis and Pavlenko 2008), transfer (Heine and Kuteva 2005), convergence (Auer, Hinskens and Kerswill 2005), but still 29

30 talk about more or less the same topic how morphosyntactic similarities between the two language systems increase. I use the term convergence to refer to such phenomena. Of course, I recognize that the compared terms have different terminological capacity but I speak here only about morphosyntactic interference, morphosyntactic transfer etc. Of all the current concepts, I think that the term convergence is the most relevant to contactinduced language change description, but I also agree with Heine and Kuteva s (2005: ) conclusion that convergence is not the only result of contact-induced language change, and that sometimes, as a result, new morphosyntactic structures can emerge [in the ML] that do not have an analogue in the EL (or model language in their terminology). Although I do not have such examples in my data at present, I would not argue that this is in principle impossible and I would not exclude the possibility of such phenomena occurring in Estonia s Russian in later years (or decades). Heine and Kuteva (2005: 4) speak about contact-induced change as transfer of linguistic material from one language to another that is based on the morphosyntactic equivalence between the model and the replica languages. 2 Transfer itself is defined by Odlin (1989: 27) as the influence resulting from the similarities and differences between the target language and any other language that has been previously (and perhaps imperfectly) acquired. Bullock and Toribio (2004: 91) argue for a broader application of the term convergence (see also Verschik 2008: 81 on discussion of its use in SLA). There is no reason to create a terminological jungle in the thesis, so I use the term convergence because the mechanisms that underlie transfer/interference (emergence of structural similarities between a bilingual s two grammars) are probably the same as in convergence. Another argument for implementing the term convergence is that it does not have the negative connotations of transfer / interference ; in Weinreich s ([1953] 1966: 1) classic book it was used meaning: those instances of deviation from the norms of either language which occur in the speech of bilinguals as a result of their familiarity with more than one language [my emphasis A.Z.]. (See also my discussion in Section ) Before talking about purely linguistic convergence, I would like to note that there is an entirely different phenomenon that also happens to be covered by the term convergence. 3 I would like to bring this up not only with the aim to avoid terminological confusion but also to show that there is a link between this kind of convergence and the purely linguistic kind. Giles and Smith (1979) describe convergence as a strategy by which interlocutors accommodate to each other s speech in a variety of ways: by adjusting pronunciation and other linguistic features, or even paralinguistic features such as speech rate, pauses, and length of utterance. This so-called Communication Accommodation Theory is similar to Thomason s (1997) notion of negotiation. Thomason (1997, 2001c) justly lists negotiation among the mechanisms of contact-induced language change. Negotiation as used by Thomason is not a question of which language should be used in the interaction but rather the modifications based on the speakers hypothesis concerning what can and cannot be understood by the interlocutor. 2 3 Here I would make a small remark that Heine and Kuteva s (2005: 4) notions model language and replica language are relative, i.e. a given language can be associated with both roles. See also Hinskens et al (2005: 5 7) who talk about accommodation as a short-term manifestation of convergence and divergence. 30

31 Giles and Smith (1979) talk about speaking the same way that is, using the same words, etc. It is the phenomenon that people who often talk to one another will start resembling each other in the way they talk: their speech styles converge. However, notwithstanding the different domains of application, this conversational principle is probably what is behind the process of entire varieties becoming more similar to each other because of bilingualism: since certain Russian speakers speak to Estonians all the time, their Russian starts to resemble Estonian, i.e. converges towards it. There are also compromise forms: they result when a Russian-speaker uses neither monolingual Russian nor anything approximating monolingual Estonian (see Zabrodskaja 2006b). People do this when they do not know enough Estonian but still want to be polite. Such language behaviour can also be called accommodation. Ehala and Üprus (2008) show that accommodation also happens in communication between Estonians and Russians: Estonianspeakers take over accent traits of Russian-speakers who speak Estonian as L2. To conclude, the behaviour in communication of speakers who wish to accommodate their speech (= converge) employs mechanisms that can be described with the help of convergence as a process and this leads to linguistic convergence as a result. As Auer and Hinskens (2005: 343) state accommodation is the first step to linguistic changes on the community level. In the following Sections , I briefly look at three central notions of convergence starting from its highest level Convergence as the result Toribio (2004) concludes that the simultaneous presence of languages in CS favours the search for parallels between them and enhances convergence. It is often debated whether unidirectional change that results in the increase of structural similarities, may be called convergence. The problem is addressed in a systematic way by Heine and Kuteva (2005: 11), Bullock and Gerfen (2004) and Bullock and Toribio (2004), who point out that the term is understood in the literature in two different ways. For some scholars, convergence means a reciprocal increase in similarities (Silva-Corvalán 1994/2000; Thomason 2001c), while others claim that convergence may also be understood as change affecting only one of the languages, as often happens in situations of asymmetrical bilingualism, where one language has a higher status (Myers-Scotton 2002; Clyne 2003; Bakker 2006: 154). Bullock and Toribio (2004: 91) conclude that the directionality of convergence is not so important. What is more important, in their view, is how convergence differs from other types of contact-induced language change Convergence as a type of change The role of structural convergence in language contact, and what its definition should be, has been discussed by numerous scholars (Thomason and Kaufman 1988, Thomason 2001c, Myers-Scotton 2002, Clyne 2003, Sebba 2009 to name just some major contributions). Structural convergence often occurs where languages are spoken in close geographical proximity and the degree of multilingualism is high, like in the Balkan Sprachbund. While for Clyne (2003: 79) convergence describes, in general, languages becoming more similar to each other, Myers-Scotton and Jake (2000) propose a much more specific notion of convergence, reserving it for the use of lexical items and grammatical morphemes from one 31

32 language and combinability rules and abstract meaning from the other. This definition overlaps with semantico-syntactic transference suggested by Clyne (2003) Convergence as the mechanism Backus (2004) employs the term convergence to define the mechanism associated with language change, emphasizing that it does not cover all types of change, as some of them can result from, for example, grammaticalization and reanalysis. Discussing the question of how convergence relates to the causes and mechanisms of change, and to the various types of change, Backus (2004: 179) characterizes it as a processual mechanism, which is a diachronic phenomenon that links what speakers do in conversation with what happens to the language as a result of it. However, he also acknowledges that few empirical studies have addressed these matters. The role of interpersonal accommodation ( convergence as a mechanism ) in a theory of language change ( convergence as a result ) is also addressed by Auer and Hinskens (2005). The authors present a model that shows the links between verbal communication in face-toface situations and the implementation of structural language change. Figure 1 is a simplified illustration of the relationships among the three levels of convergence (or accommodation in their terminology). Highest level Speech community: Language change [convergence as the result] Middle level The individual: Long-term accommodation [convergence as one type of changes] Lowest level Interactional episode: Short-term accommodation [convergence as one of the mechanisms] Figure 1. Hierarchy of convergence [modified from Auer and Hinskens (2005: 336)] More thorough and detailed explanations of the types can be found in Auer and Hinskens (2005: ). In Section 5.2, I will say more about convergence as the mechanism and about the connections between the three types of explanations of convergence presented in Sections Also I will introduce my contact draught-board metaphor, which aims to aid the understanding of the different shapes of convergence. 32

33 Relationship with CS The definition of CS as the alternate use of two or more languages in the same utterance or conversation (Grosjean 1982: 145) is broad enough to encompass any kind of language alternation. Myers-Scotton (2006: 234) defines intraclausal CS as any clause that includes elements from two or more languages. According to her, the elements that make a clause bilingual may be actual surface-level words from two languages as well as abstract rules. According to Myers-Scotton (2006), the use of lexical items from another language is called CS, and the use of rules and structures from that language is labelled convergence. In Chapter 4, I return to her definitions in relation to my findings. It should be clear that CS and convergence are related in the sense that both are linked to the use of another language. I state that CS helps to account for what morphosyntactic structures are more likely to be converged and why. To differentiate between CS and convergence, I propose Toribio s (2004: 172) classification of these factors, which I feel reflects best the findings of empirical studies in the area of contact linguistics to date: The simultaneous presence of languages in CS further favours the searching for parallels between them, and hence promulgates the striving towards convergence. In my study, I take a similar approach to the investigation of CS effects on convergence in the area of morphosyntax (see Paper III) Directionality of convergence In Sections , I demonstrated that convergence has been defined in several ways, and that researchers make different claims about it. To conclude, a term convergence means that two languages become more and more alike, and that can be defined in a unidirectional or a bidirectional way. To make matters even more complex, the notion is used for both the process and the result of change. This complicates the discussion among researchers. Additional evidence that supports my statement comes from Jarvis and Pavlenko (2008: 164) who talk about: convergence between two systems, whereby a unitary conceptual is created that incorporates both L1 and L2 features. In fact, this process may be better described as divergence [my emphasis A.Z.], since the new category does not fully resemble either the L1- or the L2-mediated category, but is rather a hybrid or amalgam of the two. 4 I subscribe to their point of view only partly because, in my thesis, I try to reduce the terminological mess to a minimum by adopting the term amalgamated constructions (used also by Sarhimaa 1999, see more in Paper III) and by emphasizing that divergence means something else according to Hinskens, Auer and Kerswill (2005: 1 2). It is amounts to linguistic diversification, growing diffuseness and heterogenisation making the varieties more distinct from each other. In the same vein, Bullock and Toribio (2009: 202) use the term divergence to describe maintained or enhanced contrast between the two languages (L1 L2). 4 The question of differentiation between divergence and convergence remains out of the scope of the present study. I would only make a short remark that the relevant use of the notions depends on the approach. If a researcher looks only at the Russian language use, then its regional varieties diverge from standard Russian but if he or she deals with contacts of Russian, then Russian regional varieties converge towards local languages. 33

34 As I have already mentioned, with regard to directionality, convergence can be unidirectional and bidirectional. In my thesis, unidirectional convergence is documented (convergence of Russian towards Estonian as a result). It should be clear that during the analysis I have to distinguish between convergence as a process and convergence as a result and the use of two different notions congruent lexicalization and amalgamated constructions can help here. In my view, the process of convergence produces joint structure. These patterns are congruent lexicalization (Muysken 2000: 122). This is a very useful notion for the description of cases where the vocabulary comes from two languages and both languages jointly provide the grammatical structure of the clause, for example, Paper I (excerpt 3a), Paper II (excerpt 2), Paper III (Sections 5 and 6.2). Muysken s (2000) category congruent lexicalization has a synchronic nature but my study can be characterized as both a synchronic and a diachronic one. This suggests that a separate term for the diachronic context is also needed. There are also a number of instances of non-monolingual constructions that I call amalgamated constructions because they involve overt use of Russian lexical items and Estonian combinatorial principles. Thus, for the description of genitive constructions in seemingly Russian phrases I adopt the term amalgamated constructions because Russian lexical items behave according to Estonian word order rules (see Paper III, example 9a). Can this phenomenon also be labelled unidirectional structural change? It is a challenging question whether it is change or change in progress, and it cannot be answered very easily. Little time has passed and it is difficult to say now where the synchronic dimension finishes and where the diachronic dimension starts. I have introduced these issues in Section 2.3, so I will not discuss them further here Relevance of the grammatical approach to the study of language change Within the last twenty years, the search for universally valid grammatical constraints on CS has remained popular among some researchers (Kamwangamalu 1997, 2000; Muysken 2000; Myers-Scotton 1993, 1997, 2006), but no agreement has been reached. On the one hand, Sarhimaa (1999), Muysken (2000) and Clyne (2003) show how numerous approaches, models, and formulations of yet other grammatical and typological constraints on CS have failed to produce a universally valid account of CS phenomena. On the other hand, adherents of formal linguistic approaches continue to develop models of morphosyntactic constraints on CS, stating that the principles governing CS are the same everywhere (Myers-Scotton and Jake 2000; Myers-Scotton, Jake and Gross 2002 to name but a few). Some CS studies have endeavoured to take into account other factors in explaining the grammar of CS. Still, very few scholars combined these factors in one approach, paying attention to sociolinguistic (see Myers-Scotton 1997), psycholinguistic (Green 1986, 1998; Levelt 1989; Grosjean 1995, 1997, 2001), and conversational factors (Auer 1998; Li Wei 2005). A glance at the history of research so far makes clear though that interest in the grammar of CS cannot be separated from the sociolinguistic context because there is the connection between CS, structural changes caused by it, and sociolinguistic factors that make contactinduced language change possible at all (see Backus 2005). For the qualitative empirical study of Russian-Estonian language contact phenomena, a morphosyntactic approach has been chosen to analyze the grammar. Following the approach in Romaine (2000), Thomason (2001b), and Gardner-Chloros and Edwards (2004), it can be claimed that in addition to grammatical factors, sociolinguistic and conversational factors may 34

35 play a role in CS and can override the rules of the two monolingual grammars. Thus, another central issue in the analysis is the Estonian sociolinguistic situation. With the definitions and discussions presented above in mind, let us now turn to the findings of empirical studies on Russian-Estonian contact phenomena. 35

36 4. CONTACT PHENOMENA IN ESTONIA S RUSSIAN First, this Chapter introduces different kinds of CS found in the data. Second, to find out more about contact phenomena that present challenges for the MLF model, I present major findings on the basis of Russian-Estonian bilingual data in three subsections of Section 4.2. Some insight into congruent lexicalization and convergence is provided by the analysis described in Sections and Types of CS in the collected data As my approach is qualitative, I primarily discuss what contact phenomena are registered at all and whether they follow the MLF model principles. The aim of my thesis is not to detect every type of CS produced anywhere in Estonia by different Russian-speaking people. Theoretically, differences could exist in the extent of CS and in the pace of contact-induced language change in different regions (i.e. bilingual Tallinn, the predominantly Russian-speaking North-East and the predominantly Estonian-speaking rest of Estonia). My data are basically from Tallinn and Narva, two linguistically different language environments, but from similar settings (university). The same phenomena are found in both settings and there is no regional variation in the speech of students in the two localities. My educated guess would be that this is because Russian-speaking students study in universities with Estonian as a language of instruction. In other settings, this could be different but, as there is no research, this remains just a hypothesis at this point. In the bilingual conversations and interviews with Russian-speaking students, Russian-Estonian switches are mostly intrasentential, i.e., within a sentence or within a clause, or, more rarely, intersentential, i.e., across two sentences or clauses. In Muysken s (2000: 60 61) terminology, therefore, most switches are of the insertion type, because Russian is a base language into which lexical items or NP, VP and other types of constituents are inserted from Estonian. Constituents could also be labelled as alternation because boundaries are blurry between these two types of CS, for instance, genitive constructions, compound nouns and other multiple word items. As I have already mentioned in Chapter 2, I used data collected by other researchers as well. In the corpus, longer stretches of Estonian within Russian occur quite rarely. In the data from bilingual TV programs, CS is mostly alternational but this phenomenon is affected by the genre. For instance, in the bilingual program Unetus/Bessonica Insomnia there were two hosts, one Russian-speaking and one Estonian-speaking. They predominantly used their mother tongue and switched codes for pragmatic reasons; citation and repetition were the most frequent reasons for CS. Both hosts had a particular role. They were bilingual and the situation (the program) was bilingual, so it was difficult to stick to monolingual mode. Still, at the same time they had the roles of Russian-speaking and Estonian-speaking host. That is why CS was really frequent and alternational. Insertional CS was also detected, especially in the case of semantically specific things. To summarize, Russian-Estonian intra-sentential and constituent-internal or word-internal CS (or insertion) prevails in the corpora. Instances of alternational CS, where Estonian and Russian remain relatively separate in the clause (see Muysken 2000: 96) and of congruent lexicalization, where Estonian and Russian jointly provide the grammatical structure of the clause, and the vocabulary also comes from both languages (see Muysken 2000: 122) can also 36

37 be found. I do not separate switches on the basis of grammatical criteria for a quantitative statistical analysis because clauses with a clear ML are registered alongside clauses with a mixed ML and clauses with a questionable (or undeterminable) ML. Between them, one cannot draw a border very easily. In my analysis, attention is paid to intra-sentential switches because this is where Russian and Estonian grammars come into contact. It will be seen that morphosyntactic combinatorial principles present some challenges for the MLF model, which I will be dealing with in the next sections Implications of the data for the MLF model Focusing on the grammar of Russian-Estonian CS, it is important to determine the ML in bilingual data. Before accounting for the role of the ML in language contact, we should be aware that there are differences between, on the one hand, the mutual influences of languages in contact, and on the other hand, their functioning as the ML and the EL. The general assessment of the MLF model and its place in the CS literature was presented in Section I show in Paper I how the MLF model and its modifications can be applied to Russian-Estonian data. As I argue there, the principles of the MLF model are not sufficient because certain phenomena remain unexplained. In what follows, I criticize some of these principles on which the MLF model is based because my findings show that taking only morphological and syntactic presuppositions into account is not enough for comprehensive description of the existing phenomena. My statements are: 1) besides morphosyntax, phonology plays a role, too and 2) there are instances where the ML cannot be determined at all. Below I analyze cases where an ML is not clearly determinable and show how phonology is relevant for the ML determination. There is no need to search for a new unit of analysis that would replace the ML nor invent a new basis for the determination of the ML. However, the definition of the ML could gain from the inclusion of phonological factors. It would allow for a more detailed analysis. What is more, there are instances of Russian-Estonian CS where the ML cannot be determined unambiguously (see Paper I). For example, if we deal with the phonological compromise form like se minar seminar (Est se minar, Rus semina r) in a Russian monolingual clause, then how should this word be defined? Should we analyze it as an item in the ML or in the EL? Although morphologically it is Russian, there are phonologic changes under the influence of Estonian, the EL. Thus, se minar can not be defined as a clear example of either the ML or the EL. It is something else. I would propose that this is a converged element: it belongs to this new emerging variety that is neither monolingual Russian nor Estonian any more. In the following, I propose some challenges for the ML determination. I would also suggest that it is important to take the ML as an empirical notion that sometimes may be, but is not always clearly determinable, rather than as an abstract conception that always exists in theory; cf. Myers-Scotton and Jake (2009: 337) who construct the principles for the MLF model claiming crucially a dominant role in the bilingual clause for only one of the participating languages, the ML. 37

38 Morphological integration and phonological adaptation My point of departure here is that the MLF model looks only at morphemes, categorizing them as content or system (and, later, bridge or outsider). Jake and Myers-Scotton (2009: 214) state that the 4-M model classifies morphemes and not CS. My question is how a researcher can deal only with morpheme status, categorizing them sometimes as an outsider morpheme, another time as a late system morpheme, depending on the language pair involved (see my example on v kapi-s in CLOSET-INESS to be discussed in Section 4.2.2), and not look at the CS phrase and the sentence as a whole, and not take into consideration on what level a particular EL item is morphologically integrated or phonologically adapted. I would emphasize that even if an EL item does not seem to have ML early system morphemes, it can be phonologically adapted to the ML. Thus, such a code-switched item cannot be equal to the EL item that is used in an EL monolingual grammar. Here, I would refer to my example päev DAY (see Paper II). Pjaév day is not the EL item because it is not pronounced in the way it should be pronounced in a monolingual variety (päev). Bullock and Toribio (2009: 190) state that: The component languages of a bilingual engaged in CS should not be expected to be invariable, monolingual-like systems but, instead, these languages used independently or in conjunction, as in CS can provide us with evidence of a rich repertoire of bilingual forms The variable outcomes produced by bilinguals are probably not reducible just to differences in proficiency or language dominance... CS data reflect various sociophonetics strategies that bilinguals have at their disposal to maintain contrast between their languages while, at the same time, articulating them together. It should be clear that a code-switched form pjaév cannot be labelled as an item belonging to the EL. This is a new creation and the MLF model cannot potentially deal with such cases. There are a lot of examples with internationalisms in the data that, without taking phonological features into account, the item could be classified as an ML item according to the MLF criteria. Let us analyze the clause ty polučil di plom? Have you received the diploma? The item for diploma is an internationalism that is used in both languages but with different pronunciation. In standard Russian, it is pronounced with the stress on the second syllable diplo m diploma, while, in Estonian, the stress is on the first syllable di plom DIPLOMA. In the code-switched clause, the pronunciation was close to Estonian (not only stress on the first syllable but also the consonant -d was not as voiced as it would be expected in Russian). Thus, while there seems to be just Russian content and system morphemes, in reality, we have an EL item inserted into the Russian matrix. Another explanation is also possible: a researcher can propose that di plom, se minar etc with the Estonian stress pattern are typical EL elements. But the investigator can not prove this. It is a general tendency that stress in internationalisms in Estonia s Russian (in so-called monolingual speech ) shifts to the first syllable or, in any case, follows the Estonian model, e.g. kasi no (< Est kasiino), not kasino casino like in standard Russian. This phenomenon is more or less conventionalized. Such converged items as lo to lottery, di plom diploma, se minar seminar are typical examples. If a researcher does not know about the idiolect of the speaker, various explanations can be proposed. Based on my experience with Russianspeaking students whose speech I had a chance to observe, I think that in a seemingly monolingual utterance (such as Ty poluchil di plom?), the item di plom is more likely to be an EL item inserted into the Russian matrix than an ML item pronounced according to the Estonian rules. This is a common internationalism that has become almost identical across the two languages. And this fact, of course, makes the search for a possible explanation more difficult. 38

39 Estonian items may be integrated phonologically into Russian in additional ways: a short vowel instead of a long one, and one stress in compounds, unlike in Estonian. To conclude, phonologically different forms of the same item should not be analyzed as pure EL items. Things become more complicated when we turn to the morphology of EL items. Both Estonian and Russian have complex inflectional morphology. Full integration of an Estonian item into Russian means gender and case assignment, but in fact this does not occur in all codeswitched instances. Russian-Estonian data demonstrate a wider range of possibilities than just fully integrated and completely unintegrated items. Thus, there is a need for detailed description of integration type and degree. Let me specify six gradations of morphological integration (see examples in Paper II where the principles are formulated): 1) Full morphological integration of an Estonian single noun into the Russian matrix. This means that the Estonian noun receives Russian gender and case assignment; 2) Non-integration of the Estonian noun, even though it fits structurally into the Russian declension classes. Such non-integrated Estonian nouns appear in the Estonian nominative, although the syntactic position does not call for the nominative; 3) Instances of CS where there is no morphological integration, but where the syntactic position indeed does not require the addition of case markers. This applies to the nominative singular, and, in certain noun classes, to the accusative where that form is identical to that of the nominative; 4) Morphological integration of the Estonian noun only for gender, indicated by an agreeing adjective, demonstrative pronoun, verb in singular past tense, etc, but not for case; 5) In some situations the assignment of gender (and sometimes the addition of a case marker too) is impossible if the Estonian noun is in plural. 6) Two types of ambivalent cases: 6.1 The gender cannot be determined on the basis of the ending of an unintegrated Estonian noun (for example because agreeing adjectives or pronouns are absent). 6.2 The final vowel of an Estonian stem can theoretically be interpreted as a case marker. However, indicators of its role, such as agreeing adjectives or past tense singular verbs are absent. Such constructions occur with the locative case, which is -e for the 1 st and 2 nd declension classes. These types show that the EL item can behave differently in different situations and the MLF model does not have an adequate explanation for why sometimes the EL noun is integrated (early system morphemes are added) and sometimes not. An especially tough situation for the MLF model is type 4 where no early system morphemes from the ML are used but the EL still behaves according to the ML rules. My findings show that morphological integration and phonological adaptation do not depend on each other: an Estonian item may show high phonological adaptation and no morphological integration into Russian, and vice versa. This evidence should not be interpreted in favour of the MLF model because the sociophonetics of CS may influence asymmetric or bidirectional convergence, as has been shown by a recent study (Bullock and Toribio 2009), or trigger CS (Broersma et al 2009). 39

40 The importance of phonology is underrated and that fact is nicely pointed out by Bullock and Toribio (2009: 190): The lack of attention to the phonological repercussions of CS may ensue from the traditional division between borrowing and CS. The former is generally assumed to involve the phonological adaptation of words from the source language into the lexicon of the recipient language, whereas CS, which occurs spontaneously, does not. Returning to the notions of ML and EL, I would propose that the longitudinal analysis of the phonetic properties of CS can contribute to a better understanding of contact-induced change. In future research, the issue of phonological adaptation of Estonian items should be examined more closely and not only in relation to the CS ~ borrowing continuum, for example along the lines of Sarhimaa (1999: 179), who analyzes examples of Russian-Karelian CS where only stress can differentiate a Russian form integrated into Karelian from an unintegrated one. Research should also look at phonologically (non)-adapted EL items from the perspective of psycholinguistics. It is also observed that during the integration of Estonian items into a Russian matrix, double marking of some morphosyntactic categories occurs, that is: a grammatical function is marked by two functionally equivalent but structurally divergent strategies from both languages (see Auer 1999: 328). In the next Section, I discuss this phenomenon and its application to furthering our understanding of the content and system morpheme division Double marking During an discussion about the example v kapi-s in CLOSET-INESS, Myers-Scotton (2009, p.c.) stated that there is no double marking here from her point of view. I am going to present Myers-Scotton s explanation of this. She claims that Estonian (like Hungarian in this sense) has two kinds of case affixes: the fist kind indicates grammatical relations, while the second kind indicates direction and other spatial relations. For Myers-Scotton, the Estonian locative cases are so-called semantic cases. Semantic case markers are not morphemes that she calls outsiders. They are more like early system morphemes that add content to their heads (like the inessive case marker -s in Estonian). That is, this second kind can come from the EL in CS clauses (kapi-s). So, if we get double marking with a Russian preposition (v), which is an outsider and then an Estonian locative affix (-s), it is really not double marking as far as the MLF model is concerned. Here, I have to stress that Myers-Scotton and Jake (2000) assume that only early system morphemes can double in CS (see also Myers-Scotton 2002). According to Viitso (2003: 32-33), nominative, genitive and partitive are grammatical cases in Estonian. The other 11 cases are adverbial and fulfill the same tasks as prefixes or suffixes in many other languages. The interior local cases illative (directional), inessive (static), and elative (separative) express mostly relations in a closed space and are formed by adding a certain case ending (-sse, -s, -st) to the genitive singular or genitive plural forms. First, if we agree that Estonian has different groups of cases grammatical and semantic, then in Myers-Scotton s terminology endings of the semantic cases should always be early system morphemes. But here one encounters the next tough question for the MLF model. In comparison to Hungarian and Finnish, the Estonian language has changed towards a more fusional type and that is why the borderline between stem and affix is not always clear (see Sutrop 1997). 40

41 The illative case formation includes two parallel but fundamentally different morphological patterns (Viitso 2003: 40), i.e. it is possible to say pane kappi PUT CLOSET:ILL and pane kapi-sse PUT CLOSET:GEN-ILL. I would note here that a short form of the illative case coincides with the partitive (kappi) which, in turn, is a content morpheme (this form can exist on its own). This shows that the principle for the division of morphemes into content and system can be ambiguous. Second, one may ask why Myers-Scotton categorizes the Russian preposition v as an outsider but the Estonian case-ending -s as an early system morpheme. In Jake and Myers-Scotton (2009: 215), it is clearly stated: Examples of content morphemes are nouns, verbs, adjectives, as well as some prepositions and some COMP-like elements. Examples of early system morphemes are plural affixes, many determiners, derivational affixes, and verbal prepositions Early system morphemes depend on content morphemes for their form and cannot appear on their own. And they continue (Jake and Myers-Scotton 2009: 215): Outsider system morphemes are called outsiders because their form depends on information outside of the phrase in which the outsider morphemes occur Similarly, case marking on nominals depends on elements in other parts of the clause in which they occur [my emphasis A.Z.]. In Russian, the preposition v requires the locative case ending: v škaf-u in the closet. Why does v depend on information outside the phrase v škafu (as Myers-Scotton argues), if it contains the ending -u in this particular instance of the locative case? When answering this question, I can agree with Myers-Scotton by adopting the term early system morpheme to describe the ending -u that adds meaning to the content morpheme because there are two types of locative endings in some nouns of the 2 nd declension class. If it is an example of a clear locative (like v les-u in the forest-loc, v škaf-u in the closet-loc, na pol-u on the floor-loc ), then the noun obviously has the ending -u. But if the preposition + locative has a different function (i.e., not spatial), then the ending is -e (o les-e about the forest-loc, o škaf-e about the closet-loc, o pol-e about the floor-loc ). Third, in Russian, there is no distinction between grammatical and semantic cases. Compared with Estonian, Russian has fewer cases and their functions and semantics are quite vague. Prepositions play an important role in Russian because they add meaning to a content morpheme (v škaf-u in the closet, na škaf-u on the closet ). In Estonian, their role is more marginal in comparison to Russian. It is not clear why v is claimed to be an outsider morpheme and -s an early system morpheme. A Russian preposition v is an early system morpheme exactly in the same way as the Estonian case ending -s because it adds content in a particular situation and cannot depend on the information outside of the clause. Forth, and more important, if in the Russian example, there would be only the preposition, and the case ending would be incorrect or absent, the semantics would still be clear (locative or lative). What is more, if a speaker does not use the preposition v but uses only the ending - u (škaf-u pro v škaf-u), it is still easy to understand what (s)he is talking about when answering the question Gde on byl? Where was it [the ball A.Z.]?, thanks to the context, because an ending -u can be used only with the prepositions v in or na on here. However, I would propose that a preposition has a more important role in semantic marking. In Estonian, it does not happen this way: the whole semantics is in the case ending because prepositions, while possible in Estonian, are much less frequent than Russian prepositions. Another issue I would like to point out is that, in Estonian, it is also possible to say kapi sees, 41

42 literally CLOSET:GEN IN. In this case, sees IN is an early system morpheme and kapi CLOSET:GEN is an amalgamated form of system and content morphemes because it is morphologically unmarked and represents a unity of a stem allomorph and its case function (see Viitso 2003: 33 for similar examples). In Estonian, there are many examples like these because the nominative singular and genitive singular are always unmarked. This means that the division into system and content morphemes is hard to apply, and therefore easily falsifiable. Similar results have been shown by Muysken who has also noticed that the definition of system morphemes is problematic (Muysken 2000: 18). He shows that they differ across languages and it is problematic to classify them universally (Muysken 2000: ). Following Myers-Scotton s logic, I can claim that, according to Russian grammar rules, a preposition v is an early system morpheme because it adds content to the stem. This evidence gives me the opportunity to claim that Myers-Scotton s explanation is not entirely correct because it is in conflict with her own morpheme definition rules. In the case of the codeswitched prepositional phrase v kapi-s in the CLOSET-INESS we deal with double marking of the early system morphemes because one semantic feature (that something is located in something) is expressed by early system morphemes belonging to two different languages: while in Estonian it is an ending (-s IN ), in Russian it is a preposition (v in ). To conclude, the same semantics that, in Estonian, is expressed with the help of a locative case and in Russian with a preposition is double marked. This phenomenon has notable similarities with Russian-Finnish CS where double marking of locative cases also occurs. Leinonen (1994: 230) describes instances like na hylly-llä stoit on shelf-adess stands. Here, the same strategy is applied by a bilingual speaker, as the spatial relation is expressed by a Russian preposition and a Finnish adessive case-ending. To summarize, the MLF and 4-M models do not differentiate instances where the same semantic relation is marked twice. For Myers-Scotton (2002, 2009 p.c.), double marking means that the same type of morpheme comes from both languages. It is relevant to point out that Russian and Estonian use different strategies to express spatial relations. Thus, the distribution of morphemes is not the same in these languages, but the semantics is (Russian prepositions and Estonian case endings both express spatial meaning). I argue that this is double marking even if the two languages use different kinds of morphemes, in this case, according to Myers-Scotton s (2009, p.c.) classification, an outsider (Russian preposition) and an early system morpheme (Estonian inessive marker). Crucially, these morpheme types fulfill (at least partly) the same functions; consequently, I would argue they are both early system morphemes. Let us go back to the System Morpheme Principle that allows double marking of early system morphemes, but not of late system morphemes. Myers-Scotton (1997: 110) calls this phenomenon simultaneous system morphemes (double morphology) : The formulator accesses not only ML system morphemes but also those EL system morphemes which are at the same lemma address as an EL noun or verb stem (or somehow automatically accessed with the stem). Here, however, the resemblance between the morphological doublets ends: these EL system morphemes have no relationships external to their heads; i.e., they show no interrelations with other items in the sentence, such as agreement. 42

43 If one follows this approach, then looking at Russian-Estonian data, one immediately sees that double marking is also possible when a Russian infinitive and an Estonian infinitive (or supine) meet in one verb (see Paper III, Section 6.2). It is double marking because infinitive endings fit fine into the definition (Jake and Myers-Scotton 2009: 215): Early system morphemes add meaning to content morphemes by adding various forms of specificity. In Section 4.2.3, a brief summary of amalgamated constructions of the type Russian auxiliary + Estonian infinitive is given Choice of infinitive forms The issue of the choice of a supine form (-ma) or an infinitive (-da) in CS phrases was studied and analyzed. Russian has only one infinitive form (-t ) which corresponds to two Estonian forms: a supine and an infinitive, whose distribution is governed by strict rules. The question is which infinitive form is chosen when CS occurs between a verb and an infinitive: cf. monolingual Est lähen maga-ma [supine], monolingual Rus idu spat and CS idu maga-ma I go to sleep ; Est tahan maga-da [infinitive], Rus xoču spat, CS xoču maga-da I want to sleep. In Estonian, the distribution of two different infinitive forms is determined by the head. Russian-Estonian CS data demonstrate that bilingual speakers typically tend to follow Estonian grammar rules when choosing between the illative supine and infinitive (see Viitso 2003: 53, on the Estonian verb supine and infinitive forms). Some central questions which arise from the discussion concern the choice of Estonian infinitive form and its analysis. On what grounds should the form of a verb used by a speaker in the CS phrase be analyzed? Should it be done on the basis of monolingual Estonian grammar or on the basis of free variation because the monolingual Russian infinitive covers the functions of both Estonian infinitives? My data show that the question has an easy answer because speakers follow the Estonian rules, so the inserted verbs are ordinary ML + EL constituents so far. I expand upon infinitives in Paper III, Section 6.2. Still, I argue against a monolingual bias: even though these particular examples are well-formed, this is not an indicator that every single emerged element should be formed only in a right way (= according to a particular monolingual grammar). It is up to future research to find out how frequently such well-formedness occurs in bilingual s non-monolingual speech and what consequences they have on a bilingual s monolingual varieties Amalgamated constructions This section is dedicated to non-monolingual constructions. My study is in line with Sarhimaa (1999) and Muysken (2000); I do not use the notion of a composite ML because although the idea itself is useful, the definitions suggested by Myers-Scotton are vague (see also Papers I and II). To remind the reader, a composite ML can be applied to cases where one level or parts of one level may come from one variety and other levels or their parts from another (Myers-Scotton 2001: 52). This means that it cannot help in the analysis of non-monolingual 43

44 constructions where the morphosyntax of one part can be provided jointly by Russian and Estonian (see Paper III, Section 6.1). 5 To me, Myers-Scotton s (1998b) classification of contact phenomena into intra-sentential CS and convergence appears problematic. According to her, while classic intra-sentential CS means the use of morphemes from two or more linguistic varieties in the same projection of complementizer, then convergence is the use of morphemes from a single linguistic variety, but with parts of their lexical structure coming from another source (Myers-Scotton 1998b: 290). My empirical studies suggest that such division is imprecise because the overt lexical material can come from Russian but the syntactic relationship between the constituents can be governed by Estonian rules. Next, I will look at cases that involve CS and congruent lexicalization and can result in convergence of Estonia s Russian towards Estonian Internationalisms and lexical facilitation Section 5 in Paper III shows that loan words (mainly NPs or established expressions that belong to the university domain) in the speech of Russian-speaking students are oftentimes internationalisms and/or bilingual homophones that facilitate CS. It means that some borrowed words or bilingual homonyms are materially similar to each other in L1 and L2, the similarity may trigger CS at almost any point in the sentence. Clyne (2003: 168) calls this phenomenon lexical facilitation or triggering (see Clyne 1967, 1987). As we will see in Section 4.3.2, internationalisms not only facilitate CS; they can also play a role in structural changes. 5 At the same time, the concept of a composite ML cannot be applied either in relation to instances where no morphemes from the L2 occur but where the morphosyntactic pattern is clearly foreign. As for the definitions of a composite ML given by Myers-Scotton, they vary in different contributions, complicating its use considerably: cf. in a composite ML, the abstract lexical structure projecting surface structures comes from more than one linguistic system (Jake and Myers-Scotton 1997: 322); a composite ML is composed of lexical structure from different sources (Myers-Scotton 1998a: 213); In classic CS, the ML happens to be synonymous with the frame of one of the participating varieties. However, because the ML is an abstract construct, it is possible for it to be composed of abstract structure from more than one source variety. Such an ML is called a composite ML to distinguish it from the ML of classic CS (Myers-Scotton and Jake 2000: 2); One can define a composite ML as an abstract frame composed of grammatical projections from more than one variety (Myers-Scotton 2002: 22); composite CS is defined as bilingual speech containing surface level morphemes from two (or more) languages, and with both languages supplying abstract input to morphosyntactic frame of the bilingual CP [projection of complementizer A.Z.] (Myers-Scotton 2004: ); In composite CS, both participating languages contribute abstract structure, as in convergence, and also surface forms. This means that the MLF model alone does not account completely for composite CS because it is based on the premise that only one language supplies the abstract morphosyntactic frame in composite CS, there is abstract structure from the EL in bilingual constituents, not just overt EL morphemes, and this abstract structure can contribute to the morphosyntactic frame (Jake and Myers-Scotton 2009: ). As these quotations show, in her latest contributions Myers-Scotton is inclined to use the term composite CS instead of composite ML and admits that the MLF model cannot deal with composite CS (Jake and Myers-Scotton 2009: 234). But she does not provide an explanation about how composite CS differs from a composite ML. Her definitions for both terms seem very similar. I will return to the issue of vague and not exactly user-friendly definitions of composite ML in Section

45 Word order This section presents the findings on ongoing word order change in Russian genitive constructions that have emerged under the influence of Estonian via Russian-Estonian congruent lexicalization patterns. The process can be nicely described using Heine and Kuteva s (2005: 44 62) formula from minor to major use pattern. Drawing on the empirical material, Heine and Kuteva (2005: 44) stress that although language contact sometimes leads to a new emerging pattern that did not exist in either monolingual variety, it is more common when there is already some (rarely used) collocation (in their terminology minor use pattern ) that develops into a major use pattern under the influence of L2 (in their terminology model language ). Heine and Kuteva (2005: 50) conclude that A widely observable process triggered by language contact concerns infrequently occurring, minor use patterns that are activated because there is a model provided by another language. Such patterns tend to be pragmatically and functionally marked, being restricted to specific contexts and associated with some specific function; but under the influence of the other language they come to be used more frequently and their functions tends to be desemanticized with the effect that they may turn into more widely used major use patterns. This is how new word-order structures can arise, in that speakers of the replica language [the ML A.Z.] activates one of their minor use patterns that matches the word order of the model language [the EL A.Z.] and generalize it to a major, unmarked, pattern. This approach is related to a process identified by Matras and Sakel (2007: 830) as pivotmatching, that is the identification of a structure that plays a pivotal (changing) role in the model construction [the EL A.Z.], and matching it with a structure in the replica language [the ML A.Z.], to which this new role is assigned. Following Heine and Kuteva s (2005) discussion, Matras and Sakel (2007) also suggest that the process itself leads to grammaticalization. Next, Matras and Sakel (2007: 852) add that the most powerful factor leading speakers toward identifying potential pivots in the EL is the semantic potential of a structure in the ML to cover the (lexical or grammatical) semantics represented by the EL. Let us now turn back to Russian-Estonian contact phenomena. My findings support very clearly the views presented above, i.e. a marked construction becomes unmarked and this is the contact-induced change. Colloquial Russian allows genitive inversion (i.e., GEN precedes NOM) in certain contexts. The contexts of use and the semantics involved expand on the basis of the Estonian model. The word order in genitive constructions (NOUN GEN + NOUN NOM) is a marginally occurring construction that has become more frequent under Estonian influence and has lost its pragmatic force in Estonia s Russian. Another point to keep in mind when looking at word order in code-switched genitive constructions is that it does not tend to follow the main principle of the MLF model, i.e. the ML determines word order inside the ML structure. Here, I would like to quote Myers-Scotton and Jake (2009: ) who firmly state no chaos allowed. I object to their ambitious and categorical statement that the structure of the ML is always preferred in Paper III. To conclude, this chapter has identified different types of Russian-Estonian language contact phenomena. Factors related to the learning environment, everyday language use and structural features of the languages play a role in understanding the nature of contact-induced morphosyntactic change. This last issue has to do with the fact that my approach reflects to some extent Johanson s (2002) ideas, which focus on the ways in which non-linguistic and structural factors foster and inhibit contact-induced change and what structural features are perceived to be attractive. 45

46 In Chapter 5, I discuss similarities and differences found in the following conceptual categories: composite ML (Myers-Scotton 2002), congruent lexicalization (Muysken 2000) and selective copying / mixed copies (Johanson 2002). All of these aim to analyze the morphosyntactic properties of an amalgamated construction in relation to the corresponding monolingual lexical and grammatical categories. 46

47 5. DISCUSSION In Chapter 4, I presented the different types of contact-induced morphosyntactic change that have been investigated in Russian-Estonian bilingual data. The results show that different contact-induced phenomena occur simultaneously: CS, lexical borrowings (loan words), different degrees of morphological integration and phonological adaptation of amalgamated (or clearly Estonian) items in Russian grammar, amalgamated constructions, double marking of certain features, and congruent lexicalization. The analysis of my data shows that various constraints are often violated. The result of contact-induced change is convergence of the Russian grammar towards Estonian, which, in turn, results in the emergence of a new variety. At this point the findings are similar to those of Bakker (2006), who also shows that convergence can take place in a relatively or extremely brief period. The field of contact linguistics still has a long way to go in the direction of better understanding of convergence. In Section 5.2, I propose a contact draught-board metaphor that clarifies the complexity of the phenomena that are included under the term convergence. A good example of convergence of L1 towards L2 is provided by the amalgamated constructions, in which the ML cannot be unambiguously established (if the researcher follows the rules of the MLF model and its modifications). In the following section, I will try to analyze what it means that there is no clear ML in the code-switched clause. I pose the following questions for my discussion: If for a particular utterance or clause none of the two languages sets the matrix, then what rules are followed? Should it be labelled with notions such as amalgamated matrix, emergent matrix? Does a scholar really have to search for a particular matrix at all? And, in this case, what notions allow describing such instances in the most comprehensive way: is it a composite ML (Myers-Scotton 2002), congruent lexicalization (Muysken 2000) or rather selective copying / mixed copies (Johanson 2002)? 5.1. Amalgamated constructions: a composite ML, congruent lexicalization or selective copying/mixed copies The purpose of Section 5.1 is not to give an exhaustive overview of all of the models that exist in the field of contact linguistics nowadays and that could be applied to moderate contact-induced change in Estonia s Russian, but instead to present in a structured manner the discussion over possible analytical tools for amalgamated constructions. In cases, where the MLF and 4-M models are insufficient for the analysis, I find Muysken s notion of congruent lexicalization more helpful. The MLF model, which looks only at morphosyntactical level of classic CS, clearly fails to account for all Russian-Estonian contact phenomena, due to the reasons described above in Chapter 4. For example, although in some cases the distinction between the two kind of morphemes may be clear, it is also true that sometimes it is not possible to differentiate morphemes because, in fusional languages, and Estonian has been developing into one, the borders between the morphemes are blurred (many stem variants, sound alternation, etc). As a prelude to my discussion, I wish to point also out that, to determine the ML, there is no need to count morphemes (see the example leg-k-o easy in Paper II, see footnote 4). Although this example involves one content and two early system morphemes, it is obvious that, for a bilingual speaker, this word is an undividable entity that is taken from the lexicon like a stem, and not formed with the separate morphemes. 47

48 It remains beyond the scope of the present study to investigate with the help of psycholinguistic tools how Estonian compound nouns are interpreted by a L2 speaker. Maybe they are perceived as one content morpheme (or even a content word) by a learner (e.g. keele-teadus LINGUISTICS, õppe-kava CURRICULUM ). It is possible that an Estonian as L2 speaker does not form these words from two components but memorizes and uses such items as a whole, without analysing them into components. There can be a debate over whether this principle is at work all the time, because in the cases where one of the components is a common internationalism (like ainepunkt CREDIT, seminaritöö WORK PAPER ) it does not necessarily work. It is obvious that sometimes a Russian-speaking student can differentiate components of a compound noun. Otherwise it would be difficult to explain such amalgamated constructions as seminar-sk-oj peatük-оv seminar-adj-gen CHAPTER-GEN PL or seminar-sk-oj rabot-y osa-d seminar-adj-gen work-gen PART-NOM PL. In these instances, a speaker makes a difference between Russian language elements (an adjective and a noun) and its Estonian equivalents compound nouns. A common stem, an internationalism seminar seminar facilitates CS. This can be interpreted as potential evidence that L2 user of Estonian can recognize the components of a compound noun. Still, I would be cautious in making radical far-reaching claims because it is not known what exactly happens in the head of a learner. To quote Jarvis and Pavlenko (2008: 74): For bilingualism and second language research, a fundamental question is whether the words we know in different languages are mentally interconnected with each other, both directly through inter-lingual word-word associations and indirectly through links to extralinguistic representations. They further state (Jarvis and Pavlenko 2008: 159 and references therein): Education may also require the internalization of concepts that involve new academic realities and relationships, such as academic credits, syllabi, and tutoring. Thus, if the investigator takes into account the productivity of constructions and lexicalization factors, then the way in which the ML is determined will be different. I would go even further, and state that if there are cases where the ML cannot be established, then it is better not to speak anymore about asymmetric relations between the two languages that act as ML and EL, but rather to establish which Estonian structural features are so attractive for L2 speakers that they overrule the basis of the model claimed by Myers-Scotton (e.g. 1997, 2001, 2005) to be universal. Before proceeding to this issue, I will review the main criteria used by the MLF and 4-M models. As can be gathered from Chapter 4 of the analytic overview and Papers I III, the determination of the ML has been primarily concerned with the establishment of a monolingual variety, to which content, early and late system and outsider morphemes belong, as well as what particular language sets the morphosyntactic frame for the clause. Myers-Scotton (2005: 327) argues that a bilingual speaker unconsciously selects a ML to provide morphosyntactic structure for the bilingual clause. She also believes that only one language supplies the morphosyntactic structure (i.e. can be the ML). The frame is the set of abstract well-formedness constraints on how the clause is structured. The second asymmetry refers to the source of the surface-level morphemes that indicate syntactic relations across mixed constituents in the frame. The prediction is that, although the EL contributes content morphemes to mixed constituents, only ML morphemes indicate grammatical relations within such constituents. (Myers- Scotton 2005: 334) 48

49 According to Myers-Scotton (2005: 334): In composite CS, some of the abstract structure underlying the frame comes from more than one of participating languages even though one language still dominates in supplying the frame. Four years earlier she argued that, in the case of a composite ML, one level of structure (that can be split) may come from one language and other level(s) from the other (Myers-Scotton 2001: 52). I consider her notion of a composite ML quite imprecise because often she adds that it can lead to language shift (ML turnover hypothesis e.g. Myers-Scotton 2001, 2005). According to her, a bilingual speaker may shift towards primary use of the EL and this fact causes composite ML use. In short, continuous re-wording of the definition of a composite ML complicates its use and application. Non-monolingual constructions require unambiguous explanations and that is why the terminology proposed by Muysken (2000) and Johanson (1993, 2002, 2006) is adopted in the following discussion. My analysis of amalgamated constructions in genitive (N GEN + N NOM) shows that the notion of composite ML cannot be applied in its present form because of its unclear definition. One cannot speak about asymmetrical influence that would give evidence of a clear ML and EL. What is more, it has not yet been described by Myers-Scotton how this asymmetry should be measured in examples where, for instance, Russian lexical items are combined according to Estonian morphosyntax rules. What is a more weighty argument in favour of the ML Russian content and system morphemes or an Estonian covert morphosyntactic pattern and semantics? How exactly does the asymmetry manifest itself here? I will leave these questions unanswered now because on the non-lexical level the MLF model does not provide an explanation. Muysken (2000) contributes to the understanding of amalgamated constructions presenting the concept of congruent lexicalization that allows analyzing amalgamated constructions, i.e. phonologically-lexically Russianized constructions that follow Estonian morphosyntactic patterns. Until now, I have used his terminology (see Paper III) but I acknowledge that the notion can also be misleading or incomplete. The reason is that it refers simultaneously to two interrelated phenomena: a type of CS (Paper III, Section 6.1 on genitive construction) and an amalgamated item (Paper III, Section 6.2, kajf-i-ma-t [groove-suf-inf]-inf ). Two complex questions emerge, i.e., what is CS and what is a lexical borrowing (a conventionalized amalgamated item). Further, one could ask according to what criteria the CS vs. lexical borrowing continuum should be measured. At this point it would be instructive to look at contact phenomena in a broader sense, as Johanson (2002) does, and claim that amalgamated constructions can be creatively re-shaped. One important theoretical development that pertains more specifically to research on amalgamated constructions has been the code-copying model that outlines different degrees of copying (Johanson 1993, 2002, 2006). A few studies furnish arguments for the need for such a model (Backus 1999b; Verschik 2008). The theory is quite close to the combined approach I use in my dissertation, because it suggests that two monolingual varieties should not be used as the yardstick for the grammar of amalgamated constructions. My main concern has been the applicability of the MLF model and its modifications on the example of Russian-Estonian bilingual data because little research has previously been done on this topic (only Verschik 2002; Zabrodskaja 2005: 51 54). Johanson s approach is unquestionably more flexible than the MLF model. It does not use constraints, because it does 49

50 not rely on the asymmetric division of labour between the two languages in contact, and he also points out the creative nature of new codes. In addition, Johanson does not deal with CS only but with morphosyntax as well. His model covers all possible contact-induced changes. He looks at morphosyntactic changes that have already occurred in a language A under influence of a language B, in which A designates the socially dominated language acting as the basic code; and B the socially dominant language from whose code elements are copied (Johanson 2002: 9). Notably, Johanson (2002: 9) disregards cases of alternation or CS, i.e., the alternating use of several codes (the same phenomenon is described as alternation by Muysken 2000). According to his interpretation, copying is neither a switch from the basic code to the foreign code nor a fusion of the two (Johanson 2002: 9). Johanson (2006: 167) goes on to say that: Copying means the insertion of copies of elements of a model code into a basic code. Global copying concerns morphemes or morpheme sequences as a whole with their material, semantic, combinational and frequential properties. Selective copying concerns individual material, semantic, combinational and frequential properties. Copying never just means the transfer of elements from one code to another code, but always implies creative adaptation. Another point worth mentioning is that Johanson s model addresses an area that is insufficiently explored amalgamated constructions that contain elements and patterns of both codes. According to him, these are mixed copies that to some extent differ from the originals and, therefore, structural equivalence is not an issue here. His study suggests that mixed copies are combinational copies containing at least one global copy. In the sentence syntax, a copied combinational pattern is often linked to a globally copied grammatical function marker (Johanson 2002: 18 19). Thus, see že on legko, literally THIS but IS easy is an excellent example of mixed copies because morphosyntactic principles are drawn jointly from both Russian and Estonian (see Paper I). Verschik (2008: 111, ) discusses other examples of mixed copies available from the Russian-Estonian bilingual data. According to her, mixed copies can include: fixed expressions: Pae gimnazija PAE gymnasium < Pae gümnaasium PAE GYMNASIUM ; compound nouns: lóto-magazin LOTTERY shop < loto-pood LOTTERY SHOP ; and analytic constructions: pähe učila learned by HEART < õppisin pähe LEARNED BY HEART. The proposed degrees of code-copying allow a contact linguist to capture and examine amalgamated constructions. Diachronically, the continuum can be presented as the following: complex global copies > mixed copies > purely combinational copies (Johanson 2002: 19). I concur with Johanson that we are dealing with creative production and creativity cannot be formulated as a set of strict principles or constraints. Thus, my response to the questions postulated earlier is that, in certain situations, the ML loses its explanatory power since it can not be distinguished in amalgamated constructions. What makes Johanson s (2002) research particularly interesting is the matter of attractiveness of structural features that is manifested through morphosyntactic features of the languages. I take into account his argument that as a rule the more morphological classes and inflectional patterns language A contains, the more extensive are its modification strategies (Johanson 2002: 13) and go a step further arguing that, in the case of two languages 50

51 with highly developed morphology the amount of modification strategies increases even more (up to double-marking of certain features). Having said that, I want to emphasize that the Russian-Estonian data analyzed in Chapter 4 and Papers I III show that, for a bilingual speaker, different degrees of morphological integration and / or phonological adaptation may be attractive, because these levels have been seen to vary to a great extent. In addition, Estonia s Russian-speakers seem to be attracted by the possibilities afforded by the distinction between different infinitive forms in Estonian and by its local cases that allow expressing spatial relations only with the help of a case ending. For Russian-speaking students, Estonian compound nouns belonging to the university domain are attractive because of their frequency, economy and high productivity. These morphosyntactic features are apparently so attractive that they appear regardless of the MLF model predictions. To sum up, several morphosyntactic and phonological features have been noted to be attractive and to increase the likelihood of convergence of Russian towards Estonian, resulting in Estonia s Russian variety. In Figure 2, the basic interrelations of Russian and Estonian in relation to Estonia s Russian are represented, based on the process of relexification proposed by Lefebvre (1998: 16, cited from Winford 2008: 137). Though I do not discuss my data as cases of relexification, this scheme is helpful for the characterization of contact-induced change processes in Estonia s Russian. Russian lexical entry phonological representation R syntactic features R semantic features R morphological features R Estonian lexical entry phonological representation E syntactic features E semantic features E morphological features E Estonia s Russian lexical entry phonological representation R E syntactic features R E semantic features R E morphological features R E Figure 2. Representation of lexical entry on the example of Estonia s Russian [modified from Lefebvre 1998: 16] The reader will notice there are bidirectional and unidirectional arrows. These are used in order to clarify relations between the two languages. Unidirectional arrows show that features of Russian tend to converge towards Estonian, resulting in Estonia s Russian production. Bidirectional arrows signal bidirectionality of convergence and/or compromise forms: phonology of Estonia s Russian is not Russian nor Estonian any more and double marking of grammatical features occurs. In Estonia s Russian, under the influence of Estonian, the marked genitive construction looses its pragmatic markedness: N NOM + N GEN > N GEN + N NOM (see Paper III, Section 6.1). As a result, a once marginal word order becomes customary. 51

52 The model of code-copying has been applied to assess the linguistic effects of contact situations between Turkic and non-turkic languages (Johanson 2002, 2006). To my knowledge, this model has not been tested in the broader perspective, with examples of language contacts involving Russian and Finno-Ugric languages (excl. Verschik 2007, 2008 with the research on Russian-Estonian bilingual data), so this may be a fruitful avenue for future research. In Chapter 6, I will single out other general directions for future research in detail. Now I turn to the contact draught-board metaphor, to further model bilingual phenomena Introducing contact draught-board metaphor This section provides some general observations and suggests some explanations of contactinduced change, based on the Russian-Estonian contact phenomena described in the analytic overview and Paper III (Section 6.1). To begin with, I would like to quote Federman (1993: 84) My French and my English play with one another as two children do in a playground only two languages that would exist, or rather co-exist outside of their origin, in the space of their own playfulness. He sees languages as living organisms that can exist by themselves, playing according to the rules that are known only to them. I think that this position is justified and I try to show language contact as a game between two languages, using the metaphor of draughts. Let us imagine that, on a draught-board, we do not have two groups of draughts but two languages in contact. Black draughts are elements from Russian and white draughts are elements from Estonian. Languages start interacting in different phrases, clauses, sentences produced by a bilingual speaker. Thus, draughts start moving. As the Russian-Estonian data show (see Papers I III for the examples), a white draught can represent any Estonian item or rule a morpheme (a stem, an ending), word order, government relation, noun phrase (where not only CS occurs but also Estonian word order) etc (see Picture 1). Picture 1. Russian and Estonian in contact 52

53 The point of departure for my metaphor is that we have two languages in contact Russian (L1) and Estonian (L2) that both participate in CS and can equally construct an emerging variety. Crucially, there is no need to search for a concrete ML and EL in any particular example of CS, as it tells us nothing for the whole picture of language contact. Thus, the languages can not and should not be determined as the ML or the EL because if one of the languages provides more morphosyntactic elements in one particular instance of CS, then it does not mean that the other language cannot do the same in another excerpt of CS. Let us consider one example: Skol ko seminarskoj peatük-оv ty sdelal? How many CHAP- TERS of your work paper have you done? (see Paper III, Section 6.1). Here, we have a situation where a Russian word seminarskaja meets with Estonian seminaritöö and the latter not only triggers the following Estonian word (the switch is facilitated by the common stem, an internationalism) in the sentence because of the sociolinguistic context (the Estonian equivalent is used more frequently than the Russian one in everyday spontaneous speech among students in this community of practice), but also brings along the Estonian word order (see more in Paper III), cf. the following monolingual equivalents: RUS glav seminar-sk-oj [rabot-y] chapter:gen PL seminar-adj-gen [paper-gen SG] EST seminaritöö peatükk-e WORK PAPER:GEN CHAPTERS:PART PL Thus, the morphological item and the syntactic rule are inserted into the Russian part a word and a word order. Please note that convergence is treated as a mechanism here, not as a result (see Picture 2). Picture 2. Mechanism of occurrence of phonologically-lexically Russianized construction that follows an Estonian morphosyntactic pattern Continuing my metaphor, this new unit should become a so-called king or a crowned draught (see Picture 3). 53

54 Picture 3. What language can a new form be linked to? At this point one runs into the monolingual bias of bilingualism research, because it is necessary to determine what language a king belongs to. It seems logical now to start searching for the ML because a certain element entering the language comes from the other. Recall again that convergence has been treated as a mechanism so far. Next I try to present an analysis of convergence as the result relying on my metaphor of draughts as much as possible. We can see on Picture 3 that an Estonian draught disappears from the Estonian (white) side and appears in the Russian (black) part, whereas two Russian black draughts are eaten. Here, the metaphor seems inaccurate. However, another important matter is the fact that a white draught does not actually disappear from the white side and nothing is eaten on the black part. As a result, no language really looses anything: as a mechanism convergence functions exactly as a draught that starts moving from one language, gets in touch with elements or rules from the other language, and this contact process itself results into a crowned draught, a new element that emerged as a result of convergence. A critical reader might ask why the crowned draught is white and not black. I would say that, in reality, the new draught should have neither white nor black colour (see Picture 4). Picture 4. Emergence of a new form on one particular example 54

55 This blue draught is not yet an element of the Russian language, but it is not an element of the Estonian language anymore either, because it is surrounded by the Russian matrix (and can be morphologically integrated and/or phonologically adapted). It is premature to assume what kind of changes this king can give rise to or, speaking metaphorically, what Russian draughts will be jumped over or even eaten. As Heine and Kuteva (2006) show, such developments often take hundreds of years, for example for a demonstrative attribute to become a definite article, or for the numeral one to become an indefinite article. Thus, I would suggest that, differently from Picture 4, this Estonian draught (a phonologically-lexically Russianized construction that follows Estonian morphosyntactic patterns) will not stop here but will continue its way causing further contact-induced language changes. Language contact is obviously not a board with the type of artificial boundaries imposed by an 8 8 grid. What change will happen next because of this particular construction only time can tell. To conclude, contact-induced morphosyntactic change should not be analyzed with reference to two monolingual grammars, because it is impossible to move the blue draught back to its starting position. It simply has no start position: it has emerged as the result of moving draughts. When analyzing some code-switched units (e.g. an amalgamated construction or a congruent lexicalization pattern), it is impossible to segregate EL elements from the ML because the EL is not only manifested through the overt use of the lexis but also, for instance, through a syntactic government pattern that can broaden and be used in constructions with overt ML elements. This is another aspect represented well by my metaphor. The two languages cannot be distinguished on a draught-board as black or white draughts anymore, because of the actual results of the language contact situation. In my opinion, the metaphor shows nicely how the mechanism of convergence works, but it is impossible to show it in all its details with the help of just black and white draughts, because a new item emerging as the result should have neither black nor white colour. Here, I refer to the numerous works by Grosjean (1982, 1989, 1992, 1998) who has repeated over and over again the important argument that a bilingual speaker is not a sum of two monolinguals. Thus, a bilingual item is not a sum of two monolingual ones. So the draught that emerges should not be white or black, or fifty-fifty black and white. It should have a new colour: the complex combination of white, black and grey and new features that cannot be differentiated that easily. Therefore, I have used blue here. I want to show that one deals with a new emerging item that, as a whole, belongs to a new emerging variety. At this point, I wish to argue again that there is no such things as an ML per se, because, looking at this contact draught-board, how can we be sure that some draughts are not kings left in the ML from the previous game of languages that now seem to be purely ML items. Language contact is not something that happens during a short period one researcher could study so that every change would be detected. Any language is in contacts with other languages. Any language interacts, changes, develops for hundreds of years. Hinskens et al (2005: 1) claim that dialect convergence and dialect divergence have probably existed for as long as dialects have been in use. An element that now seems to be the ML element could in reality easily be an EL element that simply came into use some hundreds years ago and became an ML element due to the history of contact (see Picture 5). Example 6 in Paper II is a case in point: Raamat BOOK is a loan word that could have come to Estonian from ancient Russian (< gramota, literally written document, a piece of writing ) or any other old Slavic language variety approximately before 13 th century. I analyze it as an Estonian item incorporated and morphologically integrated into the Russian matrix. But how Estonian is this noun? This question remains 55

56 unanswered here because different researchers would give different answers depending whether they are language planners, etymologists, sociolinguists etc. I state that there is no monolingual grammar because standard languages are idealized constructs, and no variety in the world could remain unaffected by contacts during its entire history. Thus, the comparison of code-switched items with monolingual equivalents that aims at pointing out its well-formedness is a priori wrong. Picture 5. The same languages in contact in X hundreds years from now I further wish to claim that using my metaphor of a contact draught-board is convenient for understanding the different approaches to convergence I discussed in Section 3.2. In general, purely linguistic convergence is seen as a type of change, whether as the final result or as a mechanism. When talking about a type of change, a researcher compares corresponding features in both languages (Picture 6a). Analyzing the result only, a scholar pays attention to this new element that has emerged under influence of L2 and has been used in L1 grammar (Picture 6b). When a contact linguist analyzes the mechanism of change, he or she tries to follow the path the converged element has undergone (Picture 6c). Looking at the three pictures, one notices that, in general, all deal with the same phenomenon, except that they focus on different aspects. 6a 6b 6c Picture series 6a c. What is meant by convergence? 56

57 Convergence is also considered as a type of accommodation: this is not purely linguistic convergence but also interactional. It is also a way to describe the speech of a bilingual who, I would suggest, sets up all draughts on the contact draught-board closer to the opponent s side when accommodating to the interlocutor, with the aim to reduce differences (see Picture 7). In this perspective, the researcher looks at the draught-board as a whole trying to answer the question how a bilingual interaction is connected to contact-induced morphosyntactic changes occurring in one or both of the languages. My position here is based on Auer and Hinsken s conclusion (2005: 356) that the driving force of language contact is interpersonal accommodation (convergence). Picture 7. Process of accommodation When a researcher analyzes two languages in contact, he or she cannot put all the contact phenomena on one draught-board because it is impossible to detect all types of switches and to decide what concrete type will get conventionalized. A reader should bear in mind that various blue draughts constantly appear on the white side of a draught-board (at least in the case of the individual speaker). And the new variety emerges if these blue draughts disperse at the speech community level. I concur with Thomason (2000b) that one never knows what types of switches will get conventionalized. However, we could argue that what will get conventionalized is likely to be something that occurs frequently, and what occurs frequently is likely to be something that the constraints allow (although this is not the case with contact-induced change in word order in Estonia s Russian genitive constructions). Foreign words used in CS become part of the repertoire of speakers of the borrowing language (that is, as established borrowings or formulaic phrases). Which words those are is another issue altogether, and to the best of my knowledge no one has succeeded to explain which loanwords will conventionalize and why, although many scholars have made efforts to do that (e.g. Clyne 1987; Croft 2000; Myers-Scotton 1993b, 2000; Backus 2003). It is currently too early to predict anything about conventionalization regarding Russian- Estonian CS because the history of the current Russian-Estonian language contacts is too short for this (the period started after the restoration of independence in 1991, when the Russianspeaking part of the population began to use Estonian intensively, as it is the single official language). It can also in principle be impossible to make any predictions because of the kinds 57