Convergence and Divergence in Language Contact Situations [1 ed.] 9789027288820, 9789027219282

This book deals with the consequences of converging and diverging processes and their development in language contact si

195 70 2MB

English Pages 252 Year 2009

Report DMCA / Copyright

DOWNLOAD PDF FILE

Recommend Papers

Convergence and Divergence in Language Contact Situations [1 ed.]
 9789027288820, 9789027219282

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Convergence and Divergence in Language Contact Situations

Hamburg Studies on Multilingualism (HSM) Hamburg Studies on Multilingualism (HSM) publishes research from colloquia on linguistic aspects of multilingualism organized by the Research Center on Multilingualism at the University of Hamburg.

Editors Peter Siemund Christoph Gabriel Barbara Hänel-Faulhaber Research Center on Multilingualism, University of Hamburg

Volume 8 Convergence and Divergence in Language Contact Situations Edited by Kurt Braunmüller and Juliane House

Convergence and Divergence in Language Contact Situations Edited by

Kurt Braunmüller University of Hamburg

Juliane House University of Hamburg

John Benjamins Publishing Company Amsterdam / Philadelphia

8

TM

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

Library of Congress Cataloging-in-Publication Data Convergence and divergence in language contact situations / edited by Kurt Braunmüller, Juliane House. p. cm. (Hamburg Studies on Multilingualism, issn 1571-4934 ; v. 8) Mostly based on the contributions to a symposium held in October 2007 at the Hamburg Research Centre on Multilingualism. Includes bibliographical references and index. 1. Languages in contact--Congresses. 2. Bilingualism--Congresses. I. Braunmüller, Kurt, 1948- II. House, Juliane. III. Universität Hamburg. Sonderforschungsbereich Mehrsprachigkeit. P130.5.C685    2009 417’.2--dc22 2009034587 isbn 978 90 272 1928 2 (Hb ; alk. paper) isbn 978 90 272 8882 0 (Eb)

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

The production of this series has been made possible through financial support to the Research Center on Multilingualism (Sonderforschungsbereich 538 “Mehrsprachigkeit”) by the Deutsche Forschungsgemeinschaft (DFG).

Table of contents Introduction   Kurt Braunmüller & Juliane House

1

part i. Challenges to accepted views of convergence and divergence in language contact situations Divergence, convergence, contact: Challenges for the genealogical classification of languages   Georg Bossong

13

Increases in complexity as a result of language contact   Östen Dahl

41

Converging genetically related languages: Endstation code mixing?   Kurt Braunmüller

53

part ii. Convergence and divergence in different varieties in oral and written discourse Converging languages, diverging varieties: Innovative relativisation patterns in Old Swedish   Steffen Höder

73

Converging verbal phrases in related languages: A case study from Faro-Danish and Danish-German language contact situations   Karoline H. Kühl & Hjalmar P. Petersen

101

Convergence and divergence of communicative norms through language contact in translation   Viktor Becher, Juliane House & Svenja Kranich

125

On the importance of spontaneous speech innovations in language contact situations   Robert E. Vann

153

 Convergence and Divergence in Language Contact Situations

part iii. Phonological processes of variation and change in bilingual individuals Gradient merging of vowels in Barcelona Catalan under the influence of Spanish   Susana Cortés, Conxita Lleó & Ariadna Benet

185

Comparing the representation of iambs by monolingual German, monolingual Spanish and bilingual German-Spanish children   Javier Arias & Conxita Lleó

205

Author index

235

Subject index

239

Introduction Kurt Braunmüller & Juliane House

Hamburg University and Research Centre on Multilingualism

The origins, conditions, trajectories and consequences of language contact are a major topic in linguistics. Language contact and contact-induced variation and change which lead to convergence and divergence are ubiquitous phenomena of languages in use. They are as old as the use of more than one language by human beings. Since all languages have been influenced – more or less – by contact with other languages, they all are mixtures. No existing language can be said to be a “pure” monolithic and homogeneous entity, and “as far as the strictly linguistic possibilities go, any linguistic feature can be transferred from any language to any other language” (Thomason & Kaufman 1988: 14). More recent surveys on this topic can for instance be found in Thomason (2001) and in Aikhenvald and Dixon (2007), especially in Aikhenvald’s comprehensive introductory overview on what happens when languages, as represented by their grammars, come into contact with one another. One major reason for language contact has always been migration, conquest, intermarriage, travel, language learning and translation as well as bilingualism and multilingualism of individuals and societies, with convergence – the increase in similarity and divergence, the decrease of similarity between languages and varieties, being key concepts in this regard (cf. Hinskens et al. 2005). Language contact may involve face-to-face interaction among groups of speakers using different languages, but it may also involve non-personal contact of persons with texts available in the written medium. So for mutual influencing, immediate face-to-face social contact is actually not a necessary condition: lexical borrowing for instance can occur through book reading and learning by writers, teachers or students, who then pass on the new items to other people through other written texts. The result of language contact can be variation and change, mixing and of course convergence and divergence of the languages involved. Processes of convergence and divergence which occur whenever languages come into contact with one another involve both individuals and societies and also embrace language acquisition in bilingual and multilingual individuals spanning both social contact and linguistic transfer or interference. They may involve genetically related languages affecting frequently realised structures such that similar or



Convergence and Divergence in Language Contact Situations

parallel structures eventually come to be preferred. With genetically non-related languages, linguistic replication can also occur such that transfer from a “model language” to the “replica language” takes place by code copying. Alongside such processes of convergence there are also cases where the languages involved have, for instance for social reasons, come to be used in different domains (as in situations of diglossia) and are kept separate thus paving the way for processes of divergence setting in. An important and abiding issue relates to the role of individual speakers in processes of divergence and convergence in language contact. In this connection questions such as the following have frequently been asked: –– To what extent speakers (can) make deliberate choices and consciously manipulate their language(s)? –– How powerful are speakers’ linguistic actions when it comes to changing the form and content of their languages? –– What is the nature of the relationship between individual deliberate changes and their (long-term) effect? The field of language contact covers a wide range of linguistic, psychological and social phenomena. It has grown out of historical linguistic research dating back at least to 19th century research with its focus on dialects in contact, language evolution and its quest for a common ancestor, and later to studies of the formation of pidgins and creoles, borrowing, code switching, language maintenance and attrition, loss and death as well as the multilingual individual’s social and cognitive competences. In historical comparative linguistics in the 19th century a controversy arose about how to explain contact-induced change, and in particular convergence processes, i.e., whether the “family tree” model of genetic relationships among languages can be maintained in the face of the growing evidence that all languages are in fact mixtures of elements from different model languages. Traditional comparative linguistics considered phenomena of mutual influence between languages through language contact as hard to explain, exceptions and complications unnecessarily messing up the clear internal laws of the historical development development of genetic language trees – influenced in no small measure by the (supposed) analogies of trees and branches or Darwinian biology and Linnéan botany. Gradually, the viewpoint widely held today and erstwhile pioneered by Schuchardt, and continued by Hesseling, Broch and others gained ground supporting the lexical and grammatical mixture view challenging the single parent and family tree model’s implication that all language change is the result of internal causes. Many scholars today even go as far as suggesting that the processes of language contact, variation and change found in highly mixed languages such as



Introduction

creoles are in fact – to varying degrees – found in all languages, i.e., we are here dealing with a continuum, a more or less condition, not a qualitative difference (cf. e.g., Mufwene 1997). Language contact was also an important topic in 20th century linguistics for E. Sapir and L. Bloomfield and other early structuralists such as N.S. Trubetzkoy in the Prague school circle, and linguistically-oriented research on language contact was soon complemented by work on the social context of language contact, its social motivation and the psycho-social factors involved. Leading figures such as J. Fishman and D. Hymes working in what they called “The Sociology of Language” showed great interest in language maintenance and shift. In a similar vein, and denying the isolation and autonomy of the linguistic system, W. Labov emphasized that internal structural linguistic pressures and external, socio-cultural and sociopsychological pressures systematically alternate in the mechanisms of language contact, variation and change. Labov argued for “the stronger claim that it is not possible to complete an analysis of structural relations within a linguistic system, and then turn to external relations.” (Labov 1972: 537), pleading for the recognition of a “re-cycling process” as the best way to find answers to the questions of the causes, the mechanisms and the adaptive functions of language change. In Britain the theory of communication adaptation theory (CAT) formulated by H. Giles and his associates (1991) further contributed important insights into the motivation of variation and change following language contact situations and the concomitant processes of convergence and divergence. In a sense, the goal of the communication adaptation theory is to go beyond descriptions of the “what” to the “why”, i.e., to concentrate on explaining the reasons for phenomena such as code switching and other linguistic convergence and divergence phenomena in the speech of individuals when they seek to increase or decrease the social differences between themselves and their interlocutors. The hypothesis put forward in CAT is that speakers in trying to seek approval from their interlocutors in social interactive situations, would make their speech converge with their interlocutors’ speech, for instance in terms of accent, choice of words and phrases, morpho-syntax, features of regional or social dialect etc. Conversely, speakers may opt to diverge from the speech of their interlocutors. In diverging, speakers will emphasize the social distance between themselves and their interlocutors by (consciously or unconsciously) increasing the use of linguistic elements that would identify them as members of their own in-group and, by implication, sets them apart from the out-group. Further milestones in research on language contact are of course the important studies by Weinreich and Haugen in the 1950’s, both emphasizing the need to combine intra-linguistic considerations with socio-cultural perspectives, giving new impetus to “contact linguistics” and laying the ground for integrating linguistic





Convergence and Divergence in Language Contact Situations

analysis and contextual, socio-psychological, socio-cultural explanations to account for contact induced processes of language change. More recently, the work by Thomason and Kaufman (1988) attempted to build bridges between the ‘comparative method’ favoured by historical linguists with their focus on internal motives, and the role of external influences on language, urging for a more comprehensive interdisciplinary approach. In her more recent work, however, Thomason (2007) maintained that social factors override linguistic factors of contact-induced change, claiming that social factors are ultimately more important than linguistic ones. Since the publication of Weinreich’s (1953) seminal work, studies of contactinduced processes of language variation and change have multiplied. Weinreich’s classic distinction between borrowing and interference as the two fundamental types of cross-linguistic influence was picked up and further differentiated by Thomason and Kaufman (1988) and integrated into a wide variety of contact situations. Another, more theoretical approach was launched by Van Coetsem (2000), who distinguishes between recipient language agentivity processes (pull transfer) such as borrowing, adaptation and integration on the one hand and source language agentivity processes (push transfer)on the other hand, the latter referring to cases where foreign linguistic items get imposed on a language that has a weaker status than the dominant contact language (cf. Van Coetsem 2000: 104). As to definitions offered of the concepts of ‘convergence’ and ‘divergence’, Myers-Scotton (2002) refers to the kinds of language mixtures found in multilinguals’ mixed languages as ‘convergences’ defining them as processes underlying multilingual persons’ mixed languages. In general, convergence is a term referring to various ways in which languages or language varieties become more similar to each other in all grammatical subsystems. Convergence can be both a process and a product, and both conscious or non-conscious. Conscious, deliberate convergence changes with different social motivations including for instance factors such as state actions of language standardization and widespread bilingualism. While convergence involves linguistic unification and homogenisation of the linguistic repertoire, divergence leads to diversification and heterogenisation, such that languages or varieties grow more distinct from one another (Hinskens et al. 2005: 2). Processes and products of divergence are often said to go together with an attitude associated with “language trees” and a belief that languages are systems “où tout se tient” (A. Meillet), closed in themselves and different. So we can say that the type of 19th century linguistic research described above, was generally more divergenceoriented than convergence-oriented, ignoring external factors as facilitators of change – mainly because they did not easily offer themselves for classification and systematisation. And as we implied above, a focus on convergence of languages and language varieties is probably more common today (cf. also Romaine 2005).



Introduction

The dimension of time is of course important in any consideration of the relative susceptibility of grammatical structures to processes of convergence. It is therefore necessary to appreciate the diachronic dimension of language contact and trace possible gradual typological alignments. In language contact involving both linguistic systems and the multilingual speaker, it is the latter who is the proper locus of contact as Weinreich famously pointed out. However, various linguistic phenomena associated with linguistic contact situations such as transfer, interference, borrowing, code switching, convergence and divergence and other signs of cross-linguistic influencing which occur first in the speech of bilingual speakers may later become so conventionalised that over time their use can be described as independent of bilingualism and bilingual speakers. So there is a place for both explanations: the internal systemic one stressing the structural characteristics of language, i.e., how the various parts of the grammatical system make up a coherent and functional whole, and the external social explanations of contact and change, emphasising the speakers, their varieties, contexts, social networks, and the relative prestige of the languages involved as motors of variation and change and/or resistence to them. The present volume is largely based on the contributions to a symposium held in October 2007 at the Hamburg Research Centre on Multilingualism on the topic which we tried to unfold above, and it covers a range of the issues we have referred to. We have divided the volume into three parts: the first one dealing with fundamental general issues related to, and essentially challenging, established views of convergence and divergence processes; the second part tracing phenomena of convergence and divergence occurring, synchronically and diachronically, in situations of contact in different oral and written varieties leading to contact-related variation and change, and the third part dealing with phonological processes of language change where two languages impact on each other in different generational groups and areas of use under conditions of bilingualism. The first chapter is by Georg Bossong who sets out to challenge the paradigm of traditional genealogical linguistics with its focus on a genealogical tree and tidy classifications. Bossong presents a stimulating counter-proposal suggesting a ‘multiple descent approach’ which he describes as more appropriate for classifying a huge number of languages and varieties than a dogmatic ‘unique descent approach’. He gives a number of examples of complex patterns of divergence and convergence processes, and he suggests that convergence as a universal principle is definitely as important as divergence, and furthermore that language mixing is far more common than is generally conceded by genealogically oriented linguists. Another challenge to accepted views is presented by Östen Dahl in his chapter on “Changes in complexity as a result of language contact”. Dahl makes the interesting point that language contact can actually lead not to a reduction of complexity





Convergence and Divergence in Language Contact Situations

but rather to an increase in the complexity of the language system that is found in neither of the languages involved. Such an increase in complexity, he suggests, is found whenever a borrowed item or construction does not simply replace something already present in the language but is rather added to the repertoire already in existence. He exemplifies this process, which he claims is similar to what happens through internal restructuring without external influence, with examples from the development of definite articles in Scandinavian languages. The third paper in the first section of this volume is by Kurt Braunmüller. He mainly looks at convergence phenomena in the genetically related languages of Scandinavia but also at some varieties of German which, he claims, display some very special types of convergence and language mixing. Braunmüller challenges the views presented by P. Muysken and C. Myers-Scotton who restricted their descriptions to cases of ‘congruent lexicalisation’ and ‘composite matrix languages’, respectively. He suggests that there are many more and much more complex forms of code mixing pointing to the crucial role of linguistic societal norms and the latent systemic possibilities inherent in all linguistic subsystems involved in contact situations and leading to convergence varieties. Braunmüller gives examples of Faroese and Danish, German and Danish as well as other local (Germanic) dialects. He finishes his paper outlining a new model of language mixing between genetically closely related languages and dialects which might prove useful in better understanding the principles underlying processes of convergence, standardisation and mixing. His challenging claim is that code mixing is inevitable when genetically closely related language come into contact. The second part starts with another paper dealing with the Scandinavian linguistic area, this time from a historical perspective: Steffen Höder examines the emergence of Written Old Swedish focussing on the contact between Old Swedish and Latin in the Late Middle Ages, which led to morpho-syntactic changes in Old Swedish that Höder describes as grammatical replication in the sense of Heine and Kuteva (2006) and linguistic convergence. He argues that these convergence processes have only affected one social variety of Old Swedish in only one medium, the written language, leading to the divergence of the written language from other varieties of Old Swedish, to the development of a diglossic situation between the standard written language and vernacular forms, and to cases of areal convergence between functionally similar varieties of different languages. The next chapter in part two is by Karoline Kühl and Hjalmar P. Petersen. Focussing on the Verb Phrase, the authors present and compare two case studies on similar situations of language contact and change involving the Germanic languages Faroese-Danish on the one hand and German-Danish on the other in the bilingual speech both produced in Danish but heavily influenced by Faroese and German, respectively. In both these language contact situations there has been extensive synchronic and diachronic contact between the two languages



Introduction

involved. The authors compare their two data sets on the basis of an application of Myers-Scotton’s (2002) Abstract Level Model and Fredsted’s (2008) hypothesis with regard to inter-language transfer in the Verb Phrase. Their findings point to a basic comparability of the two contexts examined: both are characterised by widespread individual bilingualism that has led to language contact in practically all domains of everyday life. The authors conclude that the Verb Phrase is indeed highly vulnerable to convergence in both cases examined, which they take as proof that the grammatical system is more susceptible to convergence processes than has hitherto been assumed. The third paper in part two is by Viktor Becher, Juliane House and Svenja Kranich. They examine a situation of language contact often ignored in the literature: translation. The authors deal with both divergence and convergence processes in written texts and in the mind of the bilingual individual, the translator. The authors first describe the larger context of the case studies they present in the body of the paper, outlining the hypothesis of a larger project “Covert Translation”, namely that due to the growing socio-cultural dominance of the English language as today’s global lingua franca, German communicative norms in both translated and non-translated texts gradually converge with Anglophone ones, such that the traditional use of a “cultural filter” is becoming obsolete. The authors base their investigation of this hypothesis on a corpus of original English popular science and economic texts, their German translations and comparable German texts. The case studies they present refer to the areas of modality and connectivity – two areas that diverge in terms of the availability of interlingual identification of linguistic forms. While the expressions of modality are very different in the two languages involved, there is substantial formal and functional equivalence in the expressions of connectivity. These differential conditions lead to divergence and convergence processes respectively. The final chapter in part two is by Robert E. Vann. He looks at spontaneous linguistic innovations which occur in oral language contact situations. The author argues for a recognition of the importance of spontaneous speech innovation in language contact situations, and gives a number of examples from Spanish in contact with other languages both in the Americas and in Europe. The author recommends better use of the data from oral contact situations, and makes a strong plea for their documentation in controlled, publicly accessible corpora. Drawing on the findings of his analysis of the numerous examples of spoken language contact situations presented in this chapter, the author points to the possibility that spontaneous speech innovation may lead, over time, to systemic language change, and that spontaneous speech innovation plays an important role as a pragmatic resource in discourse strategies which perform extralinguistic functions. He stresses the sociolinguistic potential of such innovations and their contribution to creating ethnographic richness in culture-specific inter-actional styles.





Convergence and Divergence in Language Contact Situations

The third part of this volume starts with a chapter by Susana Cortés, Conxita Lleó and Ariadna Benet. In their study, the authors suggest that the phonology of Catalan in Barcelona is developing towards the loss of some of its phonemes, i.e., there is a gradual divergence from standard Catalan. They show that this change affects Catalan segments which do not exist in Spanish, whenever Catalan is in more or less intense contact with Spanish. They examine contact situations faced by two age groups in three different districts of Barcelona. The results of their auditory and acoustic analysis of recorded and transcribed oral data shows that under conditions of greater contact with Spanish, processes of convergence are more noticeable than is the case in areas of less contact with Spanish. In comparing the production of different vowel pairs by the selected generations in different districts of Barcelona, the authors are able to show that the Spanish and the Catalan vowel systems are converging in the speech of children in a district with potentially high Spanish-Catalan contacts. The final paper in this volume is by Javier Arias and Conxita Lleó. Leaning on Metrical Phonology, they examine the divergent typological structures in the phonology of Spanish and German, especially their manifestations in bilingual SpanishGerman children with respect to iambic-shaped words. As both languages are based on trochees, iambic word-structures represent a special challenge for speakers of those languages, particularly for young children acquiring them. The sample comprises two very young monolingual speakers of each language and two young bilingual children. Arias and Lleó succeed in showing that the acquisition of iambic-shaped words shows differences in the development in the two languages, German and Spanish, at the earliest stages. The results of the bilingual children diverge, however, with respect to truncation of word structures, a fact which might be due to divergent underlying representations. Their investigation shows that iambs seem to be acquired at an early age, beginning at 2;2. Moreover, it seems as if iambs are acquired before trochees. Iambic structures pose, however, difficulties in the acquisition of Spanish and German, the development of the bilingual children being divergent. The question remains as to which factors might lead to such divergence, markedness and/or language dominance being the most suitable candidates.

References Aihkenvald, A.Y & Dixon, R.M.W. 2007. Grammars in Contact. A Cross-Linguistic Typology. Oxford: OUP. Fredsted, E. 2008. Convergence in verb phrases. Linguistics 46: 949–982.



Introduction

Giles, H., Coupland, J. & Coupland, N. (eds). 1991 Contexts of Accommodation. Cambridge: CUP. Heine, B. & Kuteva, T. 2006. The Changing Languages of Europe. Oxford: OUP. Hinskens, F., Auer, P. & Kerswill, P. 2005. The study of dialect convergence and divergence. Conceptual and methodological considerations. In Dialect Change. Convergence and Divergence in European Languages, P. Auer, F. Hinskens & P. Kerswill (eds), 1–48. Cambridge: CUP. Labov, W. 1972. Principles of language change. In Directions in Sociolinguistics. J. Gumperz & D. Hymes (eds), 533ff. New York NY: Holt, Rinehart and Winston. Mufwene, S. 1997. What research on creole genesis can contribute to historical linguistics In Historical Linguistics, M. Schmid, J. Austin & D. Stein (eds.), 315–338. Amsterdam: John Benjamins. Myers-Scotton, C. 2002. Contact Linguistics. Bilingual Encounters and Grammatical Outcomes. Oxford: OUP. Romaine, S. 2005. Language contact studies. In Socioliguistics – Soziolinguistik. An International Handbook of the Science of Language and Society, U. Ammon, N. Dittmar, K. Mattheier & P. Trudgill (eds.), 49–58. Berlin: de Gruyter. Thomason, S.G. 2001. Language Contact. An Introduction. Edinburgh: EUP. Thomason, S.G. 2007. Language contact and deliberate change. In Journal of Language Contact 1: 41–62. Thomason, S.G. & Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley CA: University of California Press. Van Coetsem, F. 2000. A General and Unified Theory of the Transmission Process in Language Contact. Heidelberg: Winter. Weinreich, U. 1953. Languages in Contact. New York NY: Publications of the Linguistic Circle of New York.



part i

Challenges to accepted views of convergence and divergence in language contact situations

Divergence, convergence, contact Challenges for the genealogical classification of languages Georg Bossong

University of Zurich The genealogical paradigm, classically formulated by Meillet, still dominates language classification. Despite of its weaknesses, the tree model is at the base of all global or particular language lists. Distinguishing common heritage from subsequent borrowing is a matter of hot debates in all parts of the world. This contribution challenges the universal validity of this paradigm. Genealogical trees are based on divergent evolutions, leading to the formation of new species. But this biological metaphor has its limits when applied to language. The formation of a new language is only a “pseudo-speciation” (Erikson), communication remains possible even after the split, and convergence may counterbalance the effects of divergence at any moment. The evolutionary model of Indo-European does work in some cases, but not in all.

1.  Introduction Up to the present time, the genealogical model has dominated both historical linguistics and language classification, the latter almost without restriction. All current language catalogues are arranged according to genealogical criteria, from Ruhlen’s “Guide to the world’s languages” (1987) to the latest internet version of the “Ethnologue”, which is updated every month. The belief of the great scholar in Indo-European linguistics, Antoine Meillet, that the genealogical classification is the only one with scientific value, still seems to be valid. In the foreword of his “Les langues du monde” published in 1924 he formulated the principles of the genealogical classification with unsurpassable clarity: La trop fameuse classification en langues isolantes, agglutinantes et flexionnelles ne se laisse pas poursuivre exactement, et, pour autant qu’elle se laisse formuler, elle n’a ni portée scientifique ni utilité pratique. La seule classification linguistique qui ait une valeur et une utilité est la classification généalogique, fondée sur l’histoire des langues.… Le principe en est connu: lorsqu’une langue est parlée sur un domaine étendu et que les individus qui l’emploient viennent à cesser d’avoir les relations

 Georg Bossong

régulières et continues qui maintenaient l’unité de la langue, les changements qui ont lieu dans les diverses parties du domaine ne sont pas identiques; et, au bout d’un temps, variable suivant les cas, les différences entre les parlers locaux qui continuent la langue d’abord commune deviennent telles que les occupants des diverses régions cessent de se comprendre aisément entre eux. Dans la mesure où les habitants de provinces différentes cessent de se comprendre, on peut dire que la langue commune est remplacée par des langues nouvelles. (Meillet & Cohen 1924: I, 1)

However, Meillet did not see the principles of historical-comparative linguistics in such a one-dimensional and abbreviated way as many of his successors did. But this matter cannot be followed up here in any detail. I simply quote him here as the chief witness for the thesis that genealogical classification is the only scientifically respectable one. What is so attractive about the genealogical model? First of all, its lack of ambiguity: each language has its own position, a direct 1:1 assignment. A genealogical tree is the prototype of order. Different types of parental relationships can be determined with almost mathematical precision. In biology clarity prevails: each human being, each member of a sexually reproducing species has exactly one father and one mother; the ascendence, the positioning on the genealogical tree, is unquestionably clear (even if, in some cases, the individual does not personally know the identity of his parents, they still have to exist). Can this biological metaphor be easily transferred to language? Are languages biological creatures, can they be described as species? Metaphors are useful; they can shed light on phenomena. But they can also be misleading when their relativity is not taken into consideration. A metaphor makes certain traits of reality clear; this is helpful as long as the metaphor is not taken for the reality itself. The metaphor of the genealogical tree had an incomparable impact on linguistics. It has paved the way for recognizing certain connections which had not been detected in earlier stages of linguistics. But it has also been questioned from its very beginning, when Johannes Schmidt developed his wave theory, and Hugo Schuchardt presented his more sophisticated views about the somewhat simplistic equations of the Neogrammarians. Given the metaphor of the genealogical tree, the results of half a century of research were represented around 1850 in one unified, clearly systematized model. August Schleicher, not by chance a contemporary of Charles Darwin, paved the way for the future; he formulated a paradigm, which we still follow today. But we should consider the limits of such a model in a more systematic and basic way than we have done until now. Maybe the clarity of the tree model is simply too good to be true – at least too good to reflect the truth in its full complexity. In order to begin such a critical examination I would like to point out a number of examples which can be regarded as challenges for the genealogical language



Divergence, convergence, contact 

classification: creolisation; relexification by influential languages of a dominating culture; structural convergence without (a close) genetical relationship. This will be followed by general considerations of the consequences of the biological tree metaphor. Moreover, departing from Dixon’s (1997) “punctuated equilibrium model”, I will present the outline of a uniform genealogical language model that overcomes the previous unilateral emphasis on divergence by giving more weight to convergence. Finally, I will briefly discuss the possible consequences of such a model in the current discussions about hypothetical far-reaching genetic relations (Greenberg’s 1987; 2000–2002 Amerindian and Euro-Asiatic) and prehistoric language contact (Vennemann 2003).

2.  Case studies 2.1  The genealogical classification of Creole languages According to the – implicitly or explicitly presumed – basic dogma of the genealogical model, there is no double or multiple ascendency of a language. Each language has exactly one and only one position in the uniform genealogical tree which comprises all languages of mankind without exception. If we are not (yet) sure about the genealogical classification of one or the other language, then this is due to our lack of knowledge. In principle it has to be possible to classify each language in an unambiguous way. In spite of this dogma, it is generally recognized that in at least one field things are not so easy: in Creole languages. The fact that this group of languages constitutes a challenge for genealogical classification has often been discussed in the past, but it has never been taken very seriously, nor has it been discussed conclusively. In the major language catalogues, Creole languages are listed as their own group at the end. What does this mean? Do they form a genetic unit, just as the IndoEuropean, the Austronesian or the Arawacan languages? Certainly not, since we find in this group such different languages as Haitian (← French), Papia Kristang (← Portuguese), Tok Pisin (← English) and KiNubi (← Arabic). This cannot be a valid genetic group. The criteria of the unambiguous assignment are not applied consistently in this case. Instead others are being employed: it is not the descent that counts, but the special conditions which have led to the genesis of the languages in question, and therefore possibly also to special structural features. With this last criterion – if it is accepted to be valid for the classification of Creole languages – we definitely leave genealogy and enter typology; so we move from the area of historical comparative linguistics to the equally legitimate, but fundamentally different, area of general comparative linguistics. What would be the alternative? Should we, as

 Georg Bossong

some linguists have proposed, assign the Creole languages to their corresponding languages of origin? In this way we would do justice to the genealogical model and its reflection of a certain historical reality, for somehow or other it is not wrong to say that Creole languages “descend” from their languages of origin. But then we ignore the special nature of this kind of descent, and we deliberately ignore typological parallels. This kind of genealogical classification has been brought up theoretically, but has never been thought completely through as to all its consequences. If we did so, the genealogical tree of the Germanic and Romance languages would look quite different from what is presented in current handbooks. Thus, not only French and Portuguese would belong to the Romance language family, but also Seychellois and Principense. Seychellois would be a Gallo-Romance language, and Principense an Ibero-Romance one. Do we want such consequences? Would this be intuitively convincing? In any case, the structural parallels of the verbal system – not only between Seychellois and Principense, but, for example, also between these languages and the Arabic-based KiNuba – are completely neglected. Whatever alternative, whatever genealogical tree we choose, we can never do full justice to all aspects of the complex reality. 2.2  Language mixing as double descent In these brief considerations about Creole languages I have not alluded to their genesis (monogenesis, polygenesis, universalistic hypothesis), nor to the process of “relexification”, a term which was coined in the context of the genesis of Creole languages. This process will now be focused upon from a more general perspective, for relexification can also be found in other languages; it is so common that it seems to be totally inappropriate to treat it as a marginal phenomenon. Relexification is an extreme case of language mixing. To illustrate this point I will discuss two languages that can only be classified as mixed languages, although, according to the strict genealogical dogma, mixed languages do not exist. The two languages in question have been described in the literature on contact linguistics, so that it will suffice here to present a few illustrative examples. 2.2.1  Michif between French and Cree Michif (from Canadian-French métif “métisse”) came into being at the beginning of the 19th century as the language of the children of French trappers and their Indian wives. It still exists today as the native language of small groups of people in the Canadian provinces of Manitoba and Saskatchewan, as well as in the US American state of North Dakota. Michif is a mixture of Cree, an Algonquian language, and French. Remarkably enough, present-day speakers know neither Cree nor French;



Divergence, convergence, contact 

they are bilingual in Michif and English. Within Michif, the limit between the two components follows the borderline between nouns and verbs: basically all verbs are Cree, including all the special features of this morphologically complex Algonquian language, and all nouns stem from Canadian French. Here are a few examples (taken from Bakker 1997: 317, 328, 329):

Michif

(1) εn s�čyr d-ušt-am-ā-šu-n une ceinture I-make-BEN-TA-refl-NON3 “I am making myself a belt.” (2) nu či-kī-tūtam-an st-afεr non COMP-past-do.TI-TI.2→4sg cette affaire “You shouldn’t do such a thing.” (3) kat v�ĵis gī-ajāw-ak lı mũd la promji žurni dı lã quatre-vingt-dix 1.past-have-3pl le monde la 1ère journée de l’an “I had 90 people over on New Year’s day.”

What should we call this language? French relexified Cree? We might argue that this is the case if the verb is the structural center of the phrase, building the fundamental matrix filled with elements of “foreign” origin. Undoubtedly such an argumentation has certain advantages. But if we consider the following examples, we can clearly see that the nominal, i.e. French, part of the lexicon was developed with great liberty and creativity (in French orthography, adapted from Bakker 1997: 333f): (4)

volage “theft” sucrage “candy” trostable “trustworthy” beggueur/-euse “beggar” avariceux “avaricious” agencerie “agency” contenterie “contentment” prouvasse “proof ” une fais-la-bonne “a prude” un fais-le-saint “a hypocrite” un fais-le-gros “a boaster” l’argent pas vrai “false money”

Such a series of examples creates the impression of a native, completely French vernacular language. It has nothing “foreign” in it. Should Michif therefore be considered a French dialect? What is Michif? Is this question difficult to answer? Or is it principally misleading?

 Georg Bossong

2.2.2  Media Lengua between Spanish and Quechua One additional example will be given. During his field work on Ecuadorian Quechua, the Dutch linguist Muysken discovered by chance a language that is positioned between Spanish and Quechua: Media Lengua. It is spoken in Salcedo, about 100 km south of Quito. Present day speakers use it as their primary language, not knowing Quechua and learning Spanish as their second language. This language emerged in the 1920es, when young Quechua speakers came to Quito to work on the construction of railway lines. With the help of this language they tried to express their own identity, a language which neither corresponded exactly to Spanish nor to Quechua. In Media Lengua, the linguistic border lies between lexemes and grammemes. The entire grammar stems from Quechua, the whole lexicon (nouns and verbs) from Spanish (examples adapted from Muysken 1997):

Media Lengua

(5) Q:

unuˉ fabur-ta pidi-nga-bu bini-xu-ni shukˉ fabur-da maña-nga-bu shamu-xu-ni a favour-acc ask-NOMIN-BENEF come-PROGR-1sg “I come to ask a favour (vengo para pedir un favor)”

(6) kuyi-buk yirba nuwabi-shka Q: kuyi-buk k’iwa illa-shka cavia-BENEF grass there is not-“sudden discovery” “There turns out to be no grass for the cavias (parece que no hay hierba para los cuyes)” (7) Q:

yo-ga awa-bi kay-mu-ni ñuka-ga yaku-bi urma-mu-ni I-THEM water-loc fall-“come after some event”-1sg “I come after having fallen into the water (vengo después de caer en el agua)”

What is Media Lengua? Is it profoundly Quechua, as the complex and highly specific grammatical categories are purely Quechua? Is the “inner form of the language”, the underlying mental structure the decisive factor for its genetic classification? But can Media Lengua really be classified as Quechua, since practically the whole lexicon, including the free personal pronouns, is Spanish? How far do we get with an essentialist question such as “What is X?”? 2.3  Relexification by dominating languages (Kultursprachen) It could be argued that languages like Michif and Media Lengua are marginal and rare. Indeed, cases with such a clear double descent are not frequent. However, they are probably not as rare as is generally assumed. Cases have been reported where whole areas of the lexicon are completely replaced within one generation,



Divergence, convergence, contact 

under the pressure of tabooization, for example in Papua New Guinea. However, I do not want to focus on this now, but rather ask the question of whether language mixing and re-lexification are everyday phenomena. The degree of language mixing in Michif and Cree may seem extreme, but it is not fundamentally different from the kind of language mixing that is normal and forms the basis of countless languages all over the world. Mixing of the same basic nature is ubiquitous. To illustrate this I would like to point out two languages which are neither exotic nor artificial, but great national languages (or Kultursprachen) with long and complex histories: Japanese and Persian. 2.3.1  Chinese lexicon in Japanese Through the contact with Chinese, the Japanese lexicon has been modified to such a degree that its original character has profoundly changed. Borrowings from Chinese left their mark on all domains of the lexicon. The majority of the Japanese vocabulary, counted in dictionary entries (types), consists of Sino-Japanese elements. Here is an example from everyday speech: Japanese (8) kyûkaC negai-wa kanarazu bunshoC-de isshûkanC -mae holiday application-THEM specific writing-in one+week-before made-ni shozokuC buchôC made teishutsuC suru koto until-in responsible head+of+department until presentation make NOMIN “applications for holidays must be presented one week before in written form to the responsible head of department” (Shimamori 2001: II, 249)

Even in a simple text taken from daily colloquial speech, there is such a large number of Sino-Japanese words that the formulation of a trivial fact would be almost impossible without this supposedly “foreign”, but nonetheless fully integrated, element. In our example, only the grammatical endings and a few fundamental verbs and adverbs are genuinely Japanese, whereas the whole semantically relevant vocabulary is of Chinese origin. The image we obtain is similar to that of Media Lengua. The cited phrase is neither sophisticated nor stylistically marked; long passages in natural Japanese can be formulated where the ratio between native and borrowed elements would be similar. 2.3.2  Arabic vocabulary in Persian In contrast to Japanese, Chinese does not have inflectional morphology. Therefore the grammatical elements in Japanese are not of Chinese, but rather exclusively of Japanese origin. The relationship between Persian and its primary contact language Arabic is somewhat different. In this case, not only vocabulary but also grammatical features are borrowed. Anyone learning Persian has to master a fair bit of the

 Georg Bossong

Arabic grammatical system. In the following stylistically unmarked text, we find a situation similar to that observed in the cases of Japanese and Media Lengua: Persian (9) dar tab‛A-e avvalA vaqâye‛A-e târixiA-ye Irân-râ in print-EZF* first events-EZF historical-EZF Iran-acc be-entehâA-ye saltatA-e šâh-e šahidA NâserA ad-DinA with-end-EZF reign-EZF king-EZF martyr proper name xatmA karde bud-am seal made was-1sg

* EZF means ezâfet, an Arabism meaning “addition”; it stands for the head-dependent, suffixed determination marker of Persian, used in genitive, adjective, and relative clause constructions. “In the first edition I had finished the historical events in Persia with the reign of the martyr king Nasr ad-Din.” (adapted from ‛Ali Xân 1908: 1, see Bossong 1985: 146)

Practically the whole semantically relevant vocabulary is of Arabic origin (except for the names Irân and šâh “king”). Most grammatical endings are Persian; also of Persian origin are those verbs which express the predicative function in noun-verb combinations, i.e. verbs like “to do”, “to be”, “to become” and so on (e.g. xatm kardan “seal make → finish”). But the formation of nominal plural frequently follows Arabic rules: instead of the regular agglutinated Persian plural ending (-hâ and -ân), Arabic nouns are borrowed together with their very irregular “inner” plurals. Thus we find the “broken” plural vaqâye‛ “events” from the Arabic waqâyi‛, plural of wâqi‛a. The influence of Arabic not only concerns the lexicon, it penetrates deeply into the grammatical system. 2.3.3  Further remarks on cultural relexification Both Japanese and Persian have noun-verb combinations in the form of a nominal semanteme plus a functional verb, with the nominal semanteme stemming from the corresponding donor language. In our two examples we have teishutsu suru “to make handing in → to hand in” for Japanese and xatm kardan “to make seal → end” for Persian. Thousands of verbal terms are created according to this model in both languages. The borrowings themselves are nominal, but they are used in verbal constructions. In this way, the rule according to which the borrowed vocabulary of mixed languages is mostly or exclusively nominal can be avoided; the use of functional verbs makes the verbalisation of borrowed nouns possible on a large scale. Closer to us than Japanese and Arabic we find languages such as English: if we extracted the non-Germanic, especially Greco-Latin and Romance elements,



Divergence, convergence, contact 

this would lead to a complete collapse of the English language; with the Germanic vocabulary alone it is almost impossible to communicate in a reasonable way. Language mixing on a large scale is thus an everyday phenomenon. Another example is modern Judeo-Spanish; this language has borrowed so many French elements that it is sometimes referred to as “fragnol”. The classification of such mixed languages is problematic. The historical circumstances under which language mixing took place are of course well-known for all of the cases mentioned above. From the 7th century onwards, Japanese has become inundated with Chinese vocabulary to such a degree that we can actually speak of a kind of (partial) re-lexification. In the same way, we know exactly how the Persian language came under Arabic influence, namely in the context of the islamisation of the country, which failed to lead to the extinction of Persian. Modern Persian has survived; it was not – like Aramaic, Coptic and other languages – simply abandoned in favour of Arabic; but it was profoundly altered under Arabic influence. Around the turn of the millennium, it reappeared in history as a completely modified language compared to Middle Persian. Likewise, the significance of the battle of Hastings (1066) for the history of English is universally known. Finally, Judeo-Spanish has changed from a Hebrew-marked to a French-marked language after the opening of the schools of the Alliance Israélite Universelle, where a new elite bilingual in French and Judeo-Spanish was educated. In other words, because of well-known historical facts we are aware of when and how Japanese came under Chinese, Persian under Arabic, and English and JudeoSpanish under French influence. We can distinguish the original inherited lexicon from the later borrowed vocabulary. We can exactly separate the different lexical components since we know the influencing languages of culture. In this way, it is not difficult to separate a Japanese basic layer from the Chinese cultural layer, an Indo-­European core vocabulary from the Islamic superstrate, a Germanic or IberoRomance basic vocabulary from its Gallo-Romance overlay. Thus, when asked to classify these languages genetically, no one would hesitate: Persian, for instance, “is” fundamentally Indo-European. Two questions must be asked, however. First: Is the affirmation that Persian “is” Indo-European so much more important than that that it is Islamic? Some core elements of its basic grammar and lexicon assign it a place among the Indo-European languages, together with Singhalese, Armenian, Russian, or Icelandic; but in its modern use it belongs no less to the Islamic world together with languages such as Turkish, Urdu, or Malay. Second: How do we treat the countless languages whose historical conditions of development are not as well-known as in the above mentioned languages of culture? What do we do if there is no written tradition that allows us to distinguish between native basic vocabulary and cultural borrowing? Countless efforts have been dedicated to the distinction between inherited and borrowed parts of the

 Georg Bossong

lexicon. This question is fundamental for genealogical classification. But are we able to answer it in all cases? And moreover, does the answer always make sense? Let us summarize: There are cases of language mixing where the matrix structure, the grammatical frame, is maintained, whereas the vocabulary is profoundly altered. These are cases of convergence in the lexicon with simultaneous divergence in grammar. We can postulate a scale of convergence from total relexification, as in some Creole languages or in Media Lengua, passing by cases of profound alteration, such as in Japanese or in Persian, up to a relatively superficial influence in special semantic fields. It is essential to recognize that total re-lexification and partial lexical convergence are not fundamentally different, they differ only in degree. 2.4  Structural convergence of genetically distant languages The opposite case is also found. Because of long lasting language contact it is possible that two languages with originally separate genetic affiliations can converge structurally, although their vocabularies continue to be different. I would like to discuss two examples of languages with a possible genetic relationship. The hypothesis of such a relationship is highly controversial in both cases. If a relationship in fact exists, it is a very distant one, much older than the genetic relationship between, say, the Indo-European languages. 2.4.1  Japanese and Korean The first pair of languages is Japanese and Korean. As an example, the structure of the postpositional paradigm will be analyzed. Postpositions mark the fundamental grammatical and adverbial functions such as the nominative/accusative/ dative/locative/directional/terminative, etc. In addition to this system, we find a grammaticalized marker of thematicity. This thematic marker replaces the nominative marker compulsorily (12/13) and the accusative marker optionally (14), being added freely to all other postpositions (15). In both languages a structure combining the thematic and the nominative marker in the same sentence (16) is not permitted. Schematically:

Japanese – Korean

(10)

Function – THEM + THEM nom   nom   THEM acc acc THEM or acc +THEM loc […] loc loc and THEM

Here are a few illustrative examples which show the perfect structural resemblance between Japanese and Korean that makes a morpheme-to-morpheme translation possible, and at the same time the almost total difference between



Divergence, convergence, contact 

these two languages as far as the phonetic substance is concerned (Korean examples­ from Lewin and Kim (1976: 116); translation into Japanese GB.): (11)

watashi-ga uchi-ni sake-o yomi-masu nae-ga chib-esŏ sur-ŭl masi-mnida I-nom house-in rice wine-acc drink-POLITE “I drink rice wine in the house.”

(12) watashi-wa uchi-ni sake-o yomi-masu nae-nŭn chib-esŏ sur-ŭl masi-mnida-TOP “As for me, I drink rice wine in the house.” (13) sake-(o)-wa watashi-ga uchi-ni yomi-masu sur-(ŭl)-ŭn nae-ga chib-esŏ masi-mnida “As for rice wine, I drink it in the house.” (14) uchi-ni-wa watashi-ga sake-o yomi-masu chib-esŏ-nŭn nae-ga sur-ŭl masi-mnida “As for (the inside of the) house, I drink rice wine in it.”

This is a highly specific structure, i.e. there are no observable parallels in any other language pairs. The so-called “double subject construction” is another case in point. The structural scheme is as follows: (15) S1 NPthem + S2 S2 → NPnom + V in linear form: [NPthem [NPNOM V]]

A frequently quoted example follows: (16)

zō-wa hana-ga nagai k‛okkili-nŭn k‛o-ga gilda elephant-TOP nose-nom be long “As for the elephant, the nose is long. →The elephant has a long nose.”

The double-subject construction can also be found in other languages, strikingly in Chinese for example, but without morphological marking: the functions of THEM and nom are expressed exclusively through their position in the phrase: Chinese (17) xiàng bízi cháng elephant nose be long (Li and Thompson 1981: 92)

The structural convergence between Japanese and Korean is complete. At the same time we observe an absolute divergence in the phonetic substance; we do not find

 Georg Bossong

correspondence either in the lexematic vocabulary or in the form of the structurally corresponding grammemes (the correspondence with -ga “NOM” seems to be accidental- in Japanese this particle originally had a genitive function; in Korean it alternates with its allomorph -i). Such a typological structural correspondence cannot be taken as proof of a genetic relationship. If such a relationship existed (the “Altaic hypothesis”), it lies so far back in history that it cannot be proven with the classical methods of historical comparative linguistics. Structural convergence has nothing to do with such a speculative genetic relationship, being rather based on relatively recent language contact, even though the exact course of events is unknown. It is precisely the example of Indo-European which shows that all traces of structural typological similarity have long been lost in members of a language family separated for a relatively short time (ca. 6000–8000 years): what do Irish, Swedish and Hindi share structurally and typologically? If Japanese and Korean are supposed to have had a genetic relationship in the remote past, their common proto-language has to be at least twice as old as Proto-Indo-European. 2.4.2  Quechua and Aymara The second language pair to be dealt with here is Quechua and Aymara. The controversial debate on whether or not these two languages are genetically related has become a classic case of historical comparative linguistics (see e.g. Campbell 1997b; Cerrón Palomino 2000, 2008). Scholars have fought for decades using a variety of arguments. In this context, however, the question of the historical depth of such a genetic relationship and its link to typological convergence has scarcely been raised. In my opinion this hypothetical genetic relationship, if accepted, would be situated on the level of Greenberg’s hypothetical Amerindian, i.e. – if it ever existed – at the level of a 12,000 year old proto-language. Such a hypothetical remote ancestor cannot explain the striking similarities between these two languages. The result is similar to the language pair Korean and Japanese: there is a striking structural convergence including highly specific details. This congruence in grammatical categories allows not only a word-by-word but frequently a morpheme-by-morpheme translation from one language into the other. On the other hand, there is no resemblance in terms of the material substance. The cases of lexical convergence are easily recognizable as recent borrowings on semantic and phonetic grounds: they refer to material goods and cultural artifacts. Phonetically they are so similar that they cannot stem from a proto-language dating back millennia. As for the basic vocabulary and grammemes, no trace of genetic relatedness can be detected. As a first example I will look at the system of thematic and rhematic markers. Cross-linguistically, grammemes explicitly expressing the function of the theme are not very frequent; grammemes expressing the function of the rheme are a highly marked typological rarity (see (19)). Still more idiosyncratic is the combination of



Divergence, convergence, contact 

the thematic marker with a grammeme combining the two functions of “question” and “negation”. The basic value of this grammeme seems to be assertion blocking. Used alone it signalizes interrogativity (20); in combination with the rhematizing negator of the phrase it forms a bipartite negation (21). Schematically: Quechua – Aymara (18) [S] ^ –ASSERT → INTERROG [S] ^ neg+RHEM ^ –ASSERT → NEGAT

Concrete examples include the following (adapted from various sources, see especially Cerrón-Palomino 1987: 296f; thanks to my anonymous consultants for translations of Quechua examples into Aymara): (19)

musoq wasi-qa sumaq-mi mačaq uta-h xiwaki-wa new house-THEM nice-RHEM “The new house is nice.”

(20)

čay-qa kesu-ču? uka-h kesu-ti? this-THEM cheese-INTERROG “Is this cheese?”

(21)

čay-qa mana-m kesu-ču uka-h xani-w kesu-(ki)-ti this-THEM not-RHEM cheese-NEGAT “This is not cheese.”

A second set of examples concerns different nominalisations and their functions: simple nominalisation expresses obligation (22) and finality in the case of different subjects (23); agentivised nominalisation expresses finality in the case of identical subjects (24): (22)

ruwa-na-n-mi lura-ñ-pa-wa do-NOMIN-3sg-RHEM “He has to do it.”

(23)

hamu-ni riku-na-nki-paq xut-ta uñxa-ñ-ma-taki come-1sg see-NOMIN-2sg-FINAL “I have come in order for you to see it.”

(24)

llaqta-man ri-sunchis misa-ta uyari-q marka-ru sara-ñani misa ist’-iri village-ALL go-1PLINCL mass(-acc) hear-AGENT “We go to the village in order to hear the mass.”

 Georg Bossong

2.4.3  General reflections on structural convergence What we can observe in both language pairs is structural convergence to the point of nearly complete isomorphism, and at the same time completely different phonetic substance. Apparently the question of linguistic relationship cannot be answered as uniformly and unambiguously as the study of classical and well-known language families suggests. Let us draw a provisional conclusion based on the examples presented so far. The semantically relevant vocabulary, i.e. the lexemic part, and the morphosyntactic apparatus, i.e. the grammemic part, are independent of each other. The cases presented above exhibit a separation of both areas, in different mixtures and combinations. We observe convergence in the lexicon with simultaneous divergence in the grammemes, as well as divergence in the lexicon combined with convergence in the grammar. Grammemic convergence can mean that the elements of morphosyntax are materially similar, but it can also mean that they converge categorically, while remaining different in their material expression. Using the words of Wilhelm von Humboldt, the “inner form of language”, i.e. the formation of grammatical categories, can converge up to complete congruence, whereas the external form, i.e. its material realization, shows no similarity at all. How does the genealogical language classification – with its claim of presenting an exhaustive and unambiguous positioning of all languages of the world – deal with these disturbingly incongruous facts? Let us continue with the example of Japanese. The genealogical classification of Japanese is controversial. Three possibilities have been discussed: genetically isolated, Altaic (and therefore ultimately Eurasian, if this is taken as a valid genetic unit), and Austronesian, having split from the proto-language at a very early date. I will not argue for or against any of these alternative hypotheses, but will instead ask the question as to the benefits of such a classification. Admittedly it is not uninteresting to find out from which distant, scarcely reconstructible proto-language Japanese stems, but such knowledge yields only limited insights. No less fundamental is the fact that the grammatical system of modern Japanese developed in close contact with Korean, and that its vocabulary has predominantly been borrowed from Chinese. The details of the genesis of the Sino-Japanese vocabulary are well known, but we know very little about the prehistorical scenario that led to structural convergence between Korean and Japanese. In any case it would be misleading to reduce Japanese in an essentialist way by saying that it is “isolated” or it is “Altaic”. In other words, genealogical linguistics should consider not only divergence but also convergence, and recognize both as equal. Only in this way can we hope to obtain a comprehensive picture of the languages of mankind and overcome the inherent contradictions and shortcomings of the genealogical tree model.



Divergence, convergence, contact 

3.  Limits of the tree metaphor 3.1  Communication barriers in biology and language The fundamental question can be formulated this way: how far does the biological metaphor advance our knowledge? The motor of biological evolution is the molecular clock (Kimura 1983; 1991; 1994). As a result of unpredictable mutations at the molecular level, changes occur in the genotype which accumulate over the course of time. When populations split into separate groups, particularly through emigration to different habitats, the accumulation of mutations over time renders genetic communication impossible: individuals belonging to a formerly unified species have grown so much apart that they are no longer able to conceive fertile descendants. The result is the splitting of the species; speciation is based on the genesis of an interspecific communication barrier, due on the one hand to the “blind” random action of the molecular clock and on the other to population splits. The steadily unfolding division of new species constitutes the genealogical tree of life. This is the evolutionary process that was first described by August Schleicher’s contemporary Charles Darwin. Nowadays we understand the details of this process more deeply and precisely than was the case in the middle of the 19th century, but with respect to its fundamental elements Darwin’s model is still valid today. Undoubtedly the biological evolutionary process is in certain respects similar to the development of language. In language as well as in biology the main feature of both linguistic and biological evolution is the genesis of communication barriers; blindly acting forces of change which cannot be stopped exist in both areas. In both areas the combined effect of accidental mutation and the ecology of habitats leads to the separation of species. But it is necessary to have a closer look at these analogies. The forces of language change present an exact analogy to the molecular clock. This is most obvious at the phonetic level. Let us briefly comment on one example. In Romance, and perhaps universally, final -s has a kind of half-life period: in the long run this sound is doomed to disappear because of the general laws of erosion of phonetically weak elements. We are not able to predict exactly the moment when it will disappear, but we can specify the likelihood of its disappearance. In the Romance language family, Italian -s vanished in early stages; in French it survived until the 13th/14th century; in Andalusian and Argentinean Spanish we can hear it disappearing right now; in Castilian and Mexican Spanish it enjoys great vitality; and in Sardinian it has been backed by a paragogical final vowel in such a way that its disappearance seems quite unlikely in the foreseeable future. If the average of these different values is measured, something like a half-life period appears, a decay time as a probabilistic value. But nobody is able to predict if and when final -s vanishes effectively in a given language.

 Georg Bossong

To the best of my knowledge, no serious attempts have been made to investigate and empirically check such analogies between phonetic and molecular decay, although this would be a promising field of investigation. On the other hand, intensive work has been done in another field with the half-life period, analogous to the molecular clock, namely in that of the lexicon, a field which is actually less suitable for such comparisons, since the imponderability of historical contingent circumstances play a major role. Glottochronology operates with a constant rate of decay in basic vocabulary, which can be captured by the two versions of the so-called Swadesh list. Scholars hoped, and still hope today – even more so than a few years ago – to reconstruct the chronology of language branching not only in relative, but also in absolute terms, even for prehistoric times. The inherent problems of this method have been sufficiently discussed, so that it is not necessary to go into details here. Be that as it may, there is undoubtedly something like a molecular clock in language, or more exactly, various molecular clocks, even though they do not tick with the same regularity and uniformity as the molecular clock of unfolding life. Diachronic change in language can never be stopped under any circumstances. No language stops its evolution at a given point. The speed of development is different depending on the language as an entire system, depending on the subsystems within one language, and depending on the epoch. But change always takes place. This is the motor of the genesis of language in the sense of language splitting, just as in speciation in biology. New languages arise from accumulating changes when populations separate from each other. Just as in biology, the combined effect of accumulating mutation and ecological separation leads to communication barriers. This is how far the analogy goes. The genesis of established languages families, Indo-European, Austronesian, Tupí-Guaraní and many others, can be described very well with a model from biology: from an original homeland, different groups disperse in all directions, lose contact and therefore branch more and more into language communities unable to communicate with each other. This is the classic picture that was drawn when Indoeuropean studies first began to flourish, roughly at the same time when evolution theory emerged. Many arguments have been brought forward to locate the original homeland; speculations range from the shores of the Baltic Sea over the Ukrainian steppe to Eastern Anatolia. But the basic scenario was always the same: starting from an original centre, there is expansion, resulting in the split of the separated groups. Such a split, in combination with accumulative language change, inevitably leads to the genesis of new languages and language families, and this can be compared with the genesis of new species. But at this point the following question arises: are biological and linguistic communication barriers of the same nature? This question is to be answered in the negative.



Divergence, convergence, contact 

There is a fundamental difference between biology and linguistics: biological barriers cannot be crossed, but linguistic barriers can; the biological split of species is irreversible, whereas linguistic splits are reversible. In biological speciation an interspecific barrier emerges from an intra-specific one. Once the barrier has become inter-specific, there is no way back: members of different species are forever unable to conceive descendants. Once biological lineages have separated they will never meet again. Cats and dogs cannot conceive descendants for all eternity. In contrast to biology, the development of human language remains intra-specific. Communication barriers are never insurmountable, since every human being is able to learn every human language. This is the point where the biological metaphor has its definite limits: language communication with the aim of mutual understanding is not the same as sexual communication with the aim of biological reproduction. Erik Erikson (1966) coined the term “pseudo speciation”. This term was adopted and refined by Konrad Lorenz in his epistemological treatise “Die Rückseite des Spiegels” (1977: 242). I think that this concept helps clarify the interrelations discussed in this paper. As a consequence of cultural developments, real communication barriers emerge between human groups. These barriers are in certain respects analogous to barriers between species, but they are not identical in nature. It is not a true case of speciation, but rather a pseudo-genesis of a species that biologically remains intra-specific. All members of the species homo sapiens have the universal capacity of human language in common, i.e. what Saussure called langage. Due to this universal capacity, each human is in principle theoretically able to learn any historical human language (langue). Communication barriers do in fact exist, but they are never definite. This is valid at the individual level: no human being is inescapably enclosed within the language of the community into which he or she was born. The limits of one’s language can always be transcended, simply by learning another language. But this principle is also valid at the social level: languages do not develop in isolation; rather, they are integrated in a network of contacts which have a more or less significant impact on their development. Populations split, lose contact and develop new languages; but these same populations come into contact with different populations, or they may re-establish lost contacts, thus leading to convergence on all levels of the linguistic system. In order to achieve a comprehensive understanding of language development, convergence is no less fundamental than divergence; only the combination of both tendencies, divergence and convergence, results in a complete picture of linguistic evolution. Genealogical comparative linguistics has unilaterally privileged divergence. The family tree model is based on divergence; convergence phenomena are treated as secondary, they are considered irrelevant for genealogical classification. Of course,

 Georg Bossong

they are taken into consideration when the history of the individual languages is described in detail, but they do not play any role for the classification itself, which is based on an essentialist approach: a language fundamentally assumes its assigned place within the universal family tree of human languages; all other properties are considered supplementary and marginal. Such a point of view is questionable in its very essence. 3.2  Limits to the generalisation of the Indo-European model The family tree model is based on the example of the Indo-European languages: expansion from an original centre and subsequent emergence of more and more subfamilies. From the Indo-European prototype the model was applied to other language families with resounding success in some cases: Finno-Ugric, Semitic, Bantu, Sino-Tibetan, Austronesian, Algonquian, Uto-Aztecan, Maya, Arawak, Tupí-Guaraní, to name just a few universally recognized language families from different parts of the world. In these cases, as well as in numerous others, the transfer of the Indoeuropean model worked perfectly well. But in fact there are many regions, many language groups which tenaciously resist the application of the Indo-European model, despite patient and longstanding efforts of researchers. Substantial parts of native America, Papua New Guinea, indigenous Australia, parts of South East Asia and native Siberia are cases in point. In his standard work “Catálogo de las lenguas de América del Sur” the great Spanish comparatist Antonio Tovar postulated not more than eight clearly definable language families and divided all the numerous remaining languages into a number of geographically defined regions, thus renouncing a more detailed genetical classification – intentionally, and without expecting to further elucidate the picture through additional investigation. If we apply the strict criteria of IndoEuropean linguistics – according to which there are strict sound laws, and a protolanguage can be reconstructed with nearly mathematical precision – we do not get very far in many parts of the world. Of course, as has been repeatedly stressed, one reason for this situation is our limited knowledge of many languages and language families, especially those which cannot be traced back in history because there is a lack of written documents and because they are known only in their contemporary spoken form. But it is legitimate to ask whether it might be more than ignorance that impedes the reconstruction of genealogical relationships in some cases. It may be the case that the unilateral generalisation of the Indo-European model itself is systematically misleading, as this generalisation presupposes that the only motor of language development is expansion with subsequent rupture of contact. Admittedly, such a historical scenario is quite frequent; nevertheless, it cannot be assumed to be universal for all times and all locations. In many parts of the



Divergence, convergence, contact 

world, language change seems to take place in a different way – less dramatically, through mutual penetration and influence between languages and language families. Language genesis not only takes place through expansion and isolation but also through diffusion and contact; not only through divergence, but also through convergence. The Indo-European model cannot claim to be universal. 3.2.1  Australia and Dixon’s “punctuated equilibrium” Let us have a look at the world from an Australian point of view. The Australian linguist Robert Dixon has formulated in his noteworthy book “The rise and fall of languages” (1997) a theory which he calls “punctuated equilibrium model”, a metaphor also taken from biology. According to this model there are long periods of language contact and mutual penetration of languages and language families without expansion of one dominant people at the expense of others. Such periods are interrupted by “punctuations”, i.e. by dramatic events like goal-oriented migration and land seizure, military conquests, establishment of supranational empires and religious expansions. During such punctuations language families emerge in the classical, i.e. Indo-European (or Semitic, Austronesian, Mayan etc.) way: within a relatively short period of time, the descendants of a uniform language spread over vast territories. Because of the rupture of contact, new languages arise, which are not mutually understandable but still exhibit some characteristics revealing their common origin. The crucial point in Dixon’s model is the fact that he does not consider such punctuations to be the normal case, but rather the historical exceptions. Dixon had the situation in Australia in mind while constructing his model, probably the first region which mankind reached on its way out of Africa. Australia was settled about 50,000 years ago over a land bridge from Indonesia and New Guinea. After that it developed in isolation from the rest of the world. Empires were never established. Bigger and smaller groups lived together and side by side, without one dominating group, with the geography of the mostly flat continent making migration and establishment of contacts easier. Multilingualism was natural in such an environment, even essential for survival, until today. Nick Evans, another wellknown Australian linguist, reports that his Ilgar teacher, Charlie Wardaga, speaks Marrgu, Garig, Managkari, Gunywinygu and Iwaidja, and of course English in addition to his mother tongue Ilgar (Evans 1998: 160). It is inevitable that manifold convergences in all linguistic subsystems occur whenever such a high level of individual and social multilingualism is at stake. The result is a diffuse picture of a zone of language contact, several tens of thousands of years old and of continental dimensions; the original genetic relatedness and later borrowings have long become indistinguishable.

 Georg Bossong

3.2.2  New Guinea New Guinea, the linguistically densest region on earth, is another example. Here, about 7.7 million people speak more than 1,000 languages, one sixth of the languages of mankind, 750 of which are not of Austronesian origin. William Foley (1986: 22f) cites the great anthropologist Margaret Mead who wrote (1938: 151): Each local community, sometimes only a hamlet, sometimes several hamlets, occasionally three or four villages, presents an aggregation of widely diffused traits peculiar to it. From this narrow vantage ground each individual sees the behaviour of the members of neighbouring communities as becoming steadily more diversified from his own as the distance increases between the communities involved. Each community is a centre of many lines of diffusion, which cross and re-cross in arbitrary ways, variously determined by the topography of the country, the natural resources, the immediate state of feuds and alliances, all only partly interdependent factors.

In such an ecolinguistic context the criteria developed according to the model of the Indo-European expansion are of limited value. I am not saying that they are not valid at all but that they are by no means sufficient for coping with the real complexity of relationships. 3.2.3  The Romania After these excursions to far away continents I would like to stress that such reflections are also relevant for our immediate environment, too. The Romania is a prototypical example of language genesis according to the family tree model: Latin, the local dialect of an insignificant village in the lower reaches of the river Tiber, became a world language due to the political and military success of its speakers. At the height of the Roman Empire it was a relatively uniform language, but as a consequence of the collapse of traffic routes and commercial exchange in Late Antiquity a series of independent languages emerged. We thus observe expansion with subsequent isolation, which means divergence as the primum movens of language genesis – exactly the same picture that is also valid, on a higher level, for the Indo-European language phylum as a whole. But by reducing genealogical sub-classification in such a way we fail to account for its real complexity. After a time of divergence there was a long-lasting, variegated period of convergence, first in Western Europe, much later also in the East. The Romance languages of the West (i.e. all languages except Romanian) were more or less intensely exposed to the influence of Latin as an inalterable classical model, fixed by a written norm. The lasting influence of Latin brought the Romance languages closer together than expected as the result of separate development for fifteen centuries. In the domain of what Amado Alonso has called the “Romania continua”, such secondary convergence phenomena began to appear at the beginning of the second millennium, and they have become increasingly prominent since the Renaissance. The eastern



Divergence, convergence, contact 

“outlier” of the Romania, Rumanian, developed in its own way, deeply marked by Church Slavonic. This Romance language came under the influence of the West only in the 19th century; it was not influenced directly by Latin, but indirectly by the profoundly re-latinised French. We are well aware of all of these developments because they are documented in the full light of history. Thus we are able to distinguish with great precision between primary divergence and secondary convergence. But what is the purpose of attempting such a differentiation in regions where languages do not have any recorded history – languages whose unwritten prehistory lasts until the present day? And how successful will be our attempts to establish such a distinctions in the prehistory of our own languages?

4.  New research impulses 4.1  Joseph Greenberg: Macrofamilies in Eurasia and America The genealogical classification of languages has evoked much interest in recent years, even in the general public, since it has reached people outside the narrow circle of specialists. This becomes evident in newspaper articles in the papers such as The New York Times as well as in numerous Internet panels. This new interest arose in the wake of two spectacular advances made by a linguist who distinguished himself several times through innovative ideas: Joseph Greenberg. As a result of lifelong research he published his theses about the linguistic prehistory of the Americas in 1987; analyses of the grammar and lexicon of his postulated “Eurasian” proto-language followed in 2000 and 2002. The last mentioned work, finished just before Greenberg’s death, carries the provocative title “Indo-European and its closest relatives”. Does the Indo-European family have relatives? And if yes, are they just “close” relatives? Are there more distant relatives if we go back in time? How far into prehistory does this relatedness reach, what are its ultimate roots? Before Greenberg, such questions had been raised by those Russian, Israeli and later also American circles (Bomhard and Kerns) which became generally known as Nostratics – and which are regarded as outsiders by most mainstream linguists. Greenberg put forth new ideas and new empirical evidence in favour of this old hypothesis. He postulates a genetic relationship with not only, and not even primarily, highly estimated language families such as Kartvelian, Dravidian and Semito-Hamitic, but also with exotic languages scattered over north-eastern Eurasia, i.e. Uralic, Altaic and Paleosiberian languages up to Eskimo-Aleut. From this perspective, languages such as Chukchee and Kamchadal appear as comparatively close relatives, as brothers or cousins of the Indo-European languages.

 Georg Bossong

Greenberg’s publication of “Language in the Americas” resulted in a scientific earthquake. Rarely has a book of a highly esteemed colleague been met with such vehement, overtly hostile reactions (see Campbell 1988 as an example). Greenberg dared to reduce the overwhelming genetic abundance of a cautiously estimated 70 or a more extreme 140 language families to only three: Eskimo, Na Dene, and a family he calls Amerindian, consisting of all remaining languages. One has to look more closely at Greenberg’s publication in order to understand the general outcry which followed. We poor “pedestrians” carefully weigh a lot of arguments for and against the hypothesis that Quechua and Aymara are genetically related; the more cautious among us prefer to consider the undeniable similarities in the lexicon as instances of cultural contact and borrowing. And all of a sudden Greenberg shows up declaring these debates as sham battles within an all-encompassing genetic unit called Amerindian! The methodological, anthropological and historical consequences resulting from this scientific controversy were incalculable. 4.2 Outlines for a divergence-convergence-model: Inheritance, borrowing, or simply contact? Both older and contemporary problems of genealogical classification are reflected in the Amerindian and Eurasian debates like rays concentrated in a concave mirror. These fundamental questions are an object of passionate discussions; for reasons of space they cannot be expounded here in depth. Instead I would like to reflect on certain basic principles of methodology and sketch an outline of a model which generalises Dixon’s “punctuated equilibrium model”, and which could be called “divergence-convergence-model”. The crux of the discussions concerning hypothetical deep genetical relations like Amerindian and Eurasian (or North-Caucasian, Niger-Kordofanian, Austric, Trans-Guinea etc.) is the distinction between inherited and borrowed linguistic features, i.e. what is based on a common origin and what was adopted as a secondary borrowing at a later historical moment. A great deal of sophisticated argumentation has been dedicated to these basic questions. In these debates, two criteria are used as arguments: sound development leading to sound reconstruction, and the semantic field. The methodological problem with the first criterion is the fact that there is no constant speed of sound change. Rather, phases of rapid change alternate with long periods of stagnancy. Moreover, individual languages may show very different speeds of sound change. The accuracy of sound reconstruction, which made the Indo-European of the Neo-grammarians so attractive, decreases proportionally with the time spans of the compared elements. It is utopian to expect the same accuracy in reconstruction of elements reaching 12,000 years back as with those from 6,000 years ago.



Divergence, convergence, contact 

The methodological problem with the second criterion is that there is no uncontroversial semantic field. Practically all criteria of proven value in IndoEuropean studies have been empirically falsified. Nothing is irrefutably constant, not the numerals, not the concepts for elementary relatives, not even the personal pronouns. In Southeast Asia, for instance, pronouns are very volatile and especially susceptible to borrowing. The most reliable criteria are grammatical irregularities. The decipherment of Hittite offers a good illustrative example: once the equation aszi : sanzi = ist : sind had been discovered, there were almost no further proof needed to show that this language was of Indo-European ascent. But even such phenomena do not offer absolute protection from mistakes. In languages of the agglutinative type with its regular morphology this criterion is of limited value. Isolating languages lack morphology altogether, so this criterion does not work at all. I would even go one step further and claim that the distinction between inherited and borrowed elements ceases to make any sense beyond a certain threshold of time depth. The genesis of a clearly defined language family of the Indo-European type, if viewed in a Dixonian framework, can be described as a punctual event within a greater framework of events characterized by diffusion and contact. Such punctual events occur again and again, but it would be wrong to consider them the only motor of language development. Long periods of convergence have inevitably left their marks. Congruence between languages detected today may be a sign of a common origin as well as of historical convergence, but who can delimit these two factors once a certain time horizon has been passed? After all, we are unable to make truly reliable statements about a postulated common protolanguage. What do we really know about reconstructed Proto-Indo-European? Today nobody would be as venturesome as August Schleicher, who wrote a fable in this proto-language. There must have been profound dialectal varieties from the very beginning. All languages come from somewhere, they all have different dialects, and there is no language family which can be traced back to an absolutely uniform, absolutely homogeneous origin. No such language exists at present, and it is safe to assume that it has never existed in the past. It does not make sense to assume something like a “singularity” at the origin of language families, which then split asunder in a kind of “big bang”. At all times, alternative varieties were in contact and competed with each other, even within a hypothetically reconstructed proto-language. Inevitably we get into an endless regression: is not the proto-language itself as mixed and differentiated as its offspring? Do not its components come from multiple sources? There is no such thing as a single, ultimate source. New currents and new sources show up for every origin we reconstruct. The re-constructed ProtoIndo-European has its position within an endless chain, just as all other real human

 Georg Bossong

languages. From this perspective it is absolutely legitimate to search for relatives of Indo-European. But it would be misleading to reconstruct a hypothetical Eurasian as a homogeneous self-contained entity. What we can detect are partial similarities which result from some kind of contact. Even parenthood is nothing else but a specific kind of contact; descent is contact not in space but in time. Such contacts are of course highly interesting from a historical perspective. Let us look at personal pronouns. Greenberg (1987: 49ff) not only claimed but also proved in detail that his two macro-groupings show the following distribution regarding the form of the pronouns: (25) Pronominal forms Eurasian Amerindian

1st person m n

2nd person t/ s m

This is, put in a simplified way, a fundamental and insightful statement that permits us to distinguish between Eurasian and Amerindian as two different (hyper-) macro-phyla. According to a vivid controversy between Johanna Nichols and David Peterson (1996; 1998) and Lyle Campbell (1997a), mere coincidence can be excluded on statistical grounds as an explanation for such a distribution of features. Although nasal consonants cross-linguistically show an above-average frequency in personal pronouns, such a specific distribution can only be interpreted – with the help of elementary statistical tests – as the result of a historically contingent event. Of course we can ask the usual question: common origin or borrowing? But this question makes little sense. The similarities in the pronominal systems are a clear indicator of prehistorical contacts, this stands to reason and is universally agreed upon. But the question of whether these contacts have the form of a hypothetical proto-language or of a convergence between formerly separated language branches is not just unanswerable, it is simply irrelevant. What is historically interesting is the contact as such. Therefore, Greenberg’s findings are of fundamental historical importance, even if the reconstruction of a Eurasian or Amerindian proto-language remains impossible. Prehistorical contacts have left their mark; that is the only fact that counts. What I have tried to formulate was pointed out inimitably by Hugo Schuchardt, the great opponent of the Neo-grammarians. In a paper published in 1925 he presents a summary of his life-long reflections on the genealogical classification of Basque: Es scheint die Vorstellung zu herrschen, dass wenn wir das Baskische von all dem seit zwei Jahrtausenden Eingedrungenen reinigten, es vor uns als eine gleichartige Masse läge, als alt- oder echtbaskisch. Denkbar wäre es ja; aber weitaus wahrscheinlicher dass die Sprache als deren Fortsetzung wir das Baskische betrachten, einem damaligen Sprachforscher nicht minder große Rätsel aufgegeben hätte als uns das



Divergence, convergence, contact 

heutige Baskisch. Und Lehnwörtern würden wir immer begegnen, bis zu welchem erdenklichen Anfang wir auch emporstiegen; ja jedes Wort ist einmal ein Lehnwort gewesen… Sprachverwandtschaft kann nicht als ein streng wissenschaftlicher Begriff gelten; aber wir dürfen ihn auch nicht verpönen, wir mögen uns seiner vielleicht eher mit einer gewissen Lässigkeit als mit übertriebener Vorsicht bedienen. Diese Mahnung kann man auch verallgemeinern: unsere pädagogischen Triebe lassen uns zu oft übersehen welch feiner, ewig wechselnder Stoff die Sprache ist und dass sie eine anschmiegende Behandlung verlangt. (Schuchardt 1925: 29f)

4.3  Theo Vennemann: Language contacts in Europe At this point it seems appropriate to say a few words about the work of Theo Vennemann. He has developed the traditional methodology of contact linguistics in a highly sophisticated and innovative way and made it available for the elucidation of European prehistory. Substrate and superstrate, the old terms that are often used without the necessary precision, have been sharply redefined by Theo Vennemann. He applied them to areas which until then had remained in darkness – not because this darkness was impenetrable, but because up until then no one had dared to conceive of the ideas required for penetrating it. With his two great discoveries, the Vasconic substrate and the Semitic superstrate, he was able to shed new light on the linguistic prehistory of our continent. Joseph Greenberg made important discoveries, but his postulate of a common genetic origin of macro-families may sell be illusionary. This does not diminish the importance of his discoveries, but it does limit the perspective in which he situates them himself. Theo Vennemann is not hunting chimerical proto-languages; he is not in fact looking for anything similar. His research is targeted at what is verifiable as a historical fact, namely language contact. Nothing in the life of languages is as constant and omnipresent as contact. The discovery of Vasconic and Semitic elements in the languages of Western Europe did not lead to theories about their genealogical essence. These elements are important ingredients in the basic vocabulary of our Indo-European languages, and they can be explained by a realistic scenario of prehistoric language contact. This is not only an important discovery but the perspective in which the discoverer himself situates it, is reasonable and convincing. One other point must be stressed: English is the prototype of a mixed language. The relativity of the common essentialist classification becomes particularly evident in this case. What does it mean if we classify English as essentially Germanic? Vennemann has shown in an impressive way how a Semitic superstrate first profoundly altered the “inner form” of the westernmost subgroup of Indo-European, namely Insular Celtic, and how such a semiticized language form passed indirectly from Celtic to English. English is Indo-European in some, Germanic in

 Georg Bossong

other aspects, but it also has Semitic and Celtic (and of course Latin and Romance) features. All of these factors contribute to an essence which cannot be unambiguously classified because of its fundamental multidimensionality. If contact and convergence are consistently taken into account, such phenomena find their place quite naturally.

5.  Final remarks Divergence and convergence together constitute the life of languages. This makes languages fundamentally different from biology, where genetic convergence cannot exist per definitionem. The neglect of convergence, the genuinely human element in language development understood as pseudo-speciation, has lead to misleading questions and misinterpretations. Classical historical comparative methodology aiming at an exhaustive and unambiguous genetic classification of the languages of mankind has without any doubt led to immense advances in knowledge and insight. But by favouring divergence over convergence, the comparative method has remained unilateral in this decisive point. Its claim to be exhaustive and unambiguous is utopian. Languages cannot be reduced to one single identity; in this respect they resemble human beings. From a biological point of view, every human being has his irrevocable position in the genealogical tree; but at the same time the same human being has multiple identities, being classifiable, for instance, as male, German, European, a member of a political party, professor, dean, husband, cook, poet, musician, and so on. Something similar can be said of languages. Languages are not biological species but highly complex entities shaped out of social norms of different origins. They are products of human culture. Multiple ascendence is the rule, not the exception. Assigning a language a place in a genealogical tree does not determine its essence. Languages are modular entities consisting of many components. Their individual components can very well have different origins and develop independently. And languages are always in contact, even on the most remote islands of the Pacific. Divergence is relativised by convergence.

References ‛Ali-Xân, M.M. [1908]. Dowre-ye muxtasar-e târix-e Irân (Short compendium of the history of Iran), Tehrân 1326 H. Alonso, A. 1945. Partición de las lenguas románicas de Occidente. In Estudios lingüísticos. Temas españoles, 84–105. Madrid: Gredos 1974.



Divergence, convergence, contact 

Bakker, P. & Papaen, R. 1997. Michif: A mixed language based on Cree and French. In Thomason (ed.) 1997: 295–363. Bomhard, A.R. & Kerns, J.C. 1994. The Nostratic Macrofamily. A Study in Distant Linguistic Relationship. Berlin: Mouton de Gruyter. Bossong, G. 1985. Empirische Universalienforschung. Differentielle Objektmarkierung in den neuiranischen Sprachen. Tübingen: Narr. Campbell, L. 1988. Review article on Greenberg 1987. Language 64: 591–615. Campbell, L. 1997a. Amerind personal pronouns: A second opinion. Language 73: 339–351. Campbell, L. 1997b. The Quechumaran proposal. In American Indian Languages. The Historical Linguistics of Native America, L. Campbell (ed.), 273–283. Oxford: OUP. Cerrón-Palomino, R. 1987. Lingüística quechua. Cuzco: Centro de Estudios Rurales Andinos ‘Bartolomé de las Casas’. Cerrón-Palomino, R. 2000. Lingüística aimara. Lima: CBC. Cerrón-Palomino, R. 2008. Quechumara. Estructuras paralelas del quechua y del aimara. La Paz: Plural. Dixon, R.M.W. 1997. The Rise and Fall of Languages. Cambridge: CUP. Erikson, E. 1966. Ontogeny of ritualisation in man. In Transactions of the Royal Society London 251 B: 337–349. Evans, N. 1998. Myth 19: Aborigins speak a primitive language. In Language Myths, L. Bauer & P. Trudgill (eds), 159–166. London: Penguin. Foley, W.A. 1986. The Papuan Languages of New Guinea. Cambridge: CUP. Greenberg, J. 1987. Language in the Americas. Stanford CA: Stanford University Press. Greenberg, J. 2000–2002. Indo-European and Its Closest Relatives. The Eurasiatic Language Family, Vol. 1: Grammar, Vol. 2: Lexicon. Stanford CA: Stanford University Press. Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge: CUP. Kimura, M. 1991. New Aspects of the Genetics of Molecular Evolution. Tokyo: Japanese Scientific Societies Press. Kimura, M. 1994. Population Genetics, Molecular Evolution, and the Neutral Theory. Chicago IL: University of Chicago Press. Lewin, B. & Kim, T.D. 1976. Einführung in die koreanische Sprache, 2nd edn. Heilbronn: Gustav Scheerer. Li, C.N. & Thompson, S.A. 1981. Mandarin Chinese. A Functional Reference Grammar. Berkeley CA: University of California Press. Lorenz, K. 1997. Die Rückseite des Spiegels. Versuch einer Naturgeschichte menschlichen Erkennens. München: dtv. Mead, M. 1938. The Mountain Arapesh: An importing culture. American Museum of Natural History, Anthropological Papers 36: 139–349. Meillet, A. 1924. Introduction. In Les langues du monde, A. Meillet & M. Cohen (eds). Paris: Champion. Muysken, P. 1997. Media lengua. In Thomason (ed.), 365–426. Nichols, J. & Peterson, D.A. 1996. The Amerind personal prounouns. Language 72: 336–371. Nichols, J. & Peterson, D.A. 1998. Amerind personal prounouns: A reply to Campbell. Language 74: 605–614. Ruhlen, M. 1987. A Guide to the World’s Languages. London: Edward Arnold. Schleicher, A. 1861–1862 Compendium der vergleichenden Grammatik der indogermanischen Sprachen. Weimar: Böhlau.

 Georg Bossong Schleicher, A. 1873. Die Darwinsche Theorie und die Sprachwissenschaft. Weimar: Böhlau. Schuchardt, H. 1925. Das Baskische und die Sprachwissenschaft. In Sitzungsberichte der Wiener Akademie 202: 1–34. Shimamori, R. 2000–2001. Grammaire japonaise systématique. 2nd edn. Paris: J. Maisonneuve. Tovar, A. 1984. Catálogo de las lenguas de América del Sur. Buenos Aires 1961; 2nd edn. Madrid: Gredos 1984. Thomason, S.G. ed. 1997. Contact Languages. A Wider Perspective. Amsterdam: John Benjamins. Vennemann, T. 2003. Europa Vasconia – Europa Semitica. Berlin: Mouton de Gruyter.

Increases in complexity as a result of language contact Östen Dahl

Stockholm University The paper discusses the possibility of partially or entirely new linguistic patterns developing in a contact situation and how this influences the complexity of the language system, in particular the grammar. In earlier work, I have argued that such linguistic complexity often arises out of the competition between different patterns or by the incomplete expansion of one pattern into the territory of another. As an example, I discuss the principles governing the choice of “long” and “short” Russian adjectives. Language contact plays an obvious role in that the competition may be between an inherited pattern and a borrowed one. Thus, since a borrowed pattern does not always replace existing patterns completely, distinctions are created that did not exist in either of the contributing languages. There is also a geographical aspect to this in that the effects of language contact are not likely to be equally strong everywhere, with “truces” between old and new patterns being most likely in intermediate areas (“buffer zones”). I exemplify this by two examples from Scandinavian. One is the rise of a suppletive verb paradigm as the result of the competition between two verbs for “become” in Swedish: varda and bli(va), where the latter is a loan from Low German. In parts of Sweden, forms of varda are retained in parts of the paradigm, notably in the past and the supine. The second example concerns the two definite articles in Scandinavian: the preposed free articles (den, det, de), similar to the ones of West Germanic, and thee suffixed articles (-en, -et etc.). In definite noun phrases with preposed modifiers, there are two major patterns: (i) only a preposed article is used (Danish and marginally Norwegian); (ii) both a preposed article and a suffixed one are used (Swedish and Norwegian). However, the actual picture shows considerable diversity both within and between varieties. There is a general tendency for the two articles to have different degrees of strength in the Scandinavian languages, with the preposed article being strongest in Danish and the postposed being strongest in Swedish. The different geographical distribution of the two articles suggests that two separate grammaticalization have been operating in partially overlapping geographical areas. In the “buffer zone” between these, both preposed and suffixed articles are in use, but a division of labour has been found, the details of which vary depending on the location.

 Östen Dahl

1.  Introduction Language contact can have consequences of different kinds. Not all of them can be described in terms of convergence or divergence between the languages. In this paper, I want to discuss the possibility of partially or entirely new linguistic patterns developing in a contact situation, and how this influences the complexity of the grammar. The relevant notion of complexity here is the one referred to as “system complexity” in Dahl (2004). In this context, it refers to the complexity of the language system (the grammar, lexicon, and phonology) and the basic criterion is that a more complex system is one that needs a longer description. Complexity in this sense is not necessarily reflected in the difficulty of learning the language. Interest in the application of the notion of complexity to linguistic issues has increased markedly in recent years. In particular, the questions of whether crosslinguistic comparisons between languages makes sense and whether differences in complexity between languages exist have been in the focus of the discussion (see for instance McWhorter (2001) and the responses in the same issue of Linguistic Typology; further the papers in Miestamo et al. (2008) and Sampson et al. (2009)). One recurrent idea has been that language contact tends to favour the reduction of complexity, which predicts that high-contact languages would be characterized by low complexity and vice versa (see for instance Trudgill (1983)). Loss of grammatical structure and reduced grammatical complexity is likely in particular in language shift or other situations where there are large numbers of second-language speakers. Less attention has been paid to the possibility that language contact may lead to the rise of new grammatical structures and increased complexity. This is what will be discussed in the present brief paper, which essentially summarizes data published elsewhere.1 2.  Competition between patterns leads to increased complexity In Dahl (2004), I argued that linguistic complexity in the sense of “system complexity” often arises out of the competition between different patterns or, which

1.  About at the same time as submitting the first version of this paper, I received Kuteva (2008), which covers much the same ground, looking at “linguistic accretion” in contactinduced grammaticalization. Kuteva’s focus is narrower in being restricted to grammaticalization rather than treating language change in general, but is also wider in one sense, by including not only cases of complexity that is not found in any of the input languages, but also cases where complexity is borrowed from one language to another. However, both papers use Scandinavian definite articles, with data from Dahl (2003) as a central example, and both employ the notion of a “buffer zone”.



Increases in complexity as a result of language contact 

may amount to the same thing, by the incomplete expansion of one pattern into the territory of another. One example is that of the Russian adjectives. Russian adjectives have two sets of forms, “long” and “short”, e.g. pustoj:pust. Originally, long forms had a definite meaning but then they completely took over the attributive function. At this point, then, the difference between the forms was entirely syntactic; the old semantic distinction had disappeared. However, in a further development, the long forms acquire also predicative uses. At this stage, new semantic distinctions between the forms arise, such as that between ‘disposition’ and ‘temporary state’: (1) On bol’noj he ill.m.sg.nom.LONG ‘he is sickly’ (2) On bolen he ill.m.sg.SHORT ‘he is ill (now)’

The point here is that what makes this distinction possible is the co-existence of the two sets of forms in predicative function, or in other words, the fact that the long forms have encroached on the territory of the short ones but have not yet conquered it totally. In many cases, the result is a whole set of more or less strong factors that favour one or the other mode of expression. Thus, the Russian normative grammar Rozental’ (1968) devotes six pages to a discussion of the semantic, syntactic, and stylistic differences between the long and short forms of the adjective. The two sets of forms of Russian adjectives are examples of a situation where two patterns are in competition with each other. That some of the differences are stylistic implies that speakers often have a choice between different ways of saying the same thing; it also suggests that the situation is unstable or at least somewhat fluid. It does look as if the long forms have been gaining ground for a long time, and that this process is still on-going – we cannot exclude the possibility that the long forms will eventually win a total victory. With respect to grammatical complexity, this actually means a return to the initial simple situation, where only one set of forms existed. Another possible outcome of the competition, however, is a “truce”, that is, a stable division of labour between the competing patterns is reached – and this may mean the establishment of a new grammatical distinction. For instance, there is a general tendency in languages for demonstrative pronouns to take over the role of non-demonstrative third person pronouns. This tendency is stronger with inanimate referents, which may lead to the rise of an animate-inanimate opposition in pronouns, as in standard Danish and Swedish, where the erstwhile demonstrative pronoun den has replaced han ‘he’ and hun/ hon ‘she’, but only when referring to inanimate referents. Similarly, when new

 Östen Dahl

possessive constructions arise, old constructions may survive but be restricted to certain types of noun phrases. Thus, the older possessive construction may come to be used only with kin terms and/or body part nouns, which gives rise of a distinction between alienable and inalienable possession (Koptjevskaja-Tamm 1996), as in Maltese, where the inherited Semitic “construct state” construction is used with kin terms and body part nouns and a newer periphrastic construction is used for other head nouns: (3) bin is-sultân son def-king ‘the king’s son’ (4) id ir-raġel hand def-man ‘the man’s hand’ (5) is-siġġu ta’ Pietru def-chair of Peter ‘Peter’s chair’

3.  Competition often arises in language contact situations So far, I have been discussing competition between forms and constructions within one language. But contact between different languages plays an obvious role in that the competition may be between an inherited pattern and a new one that originates in another language. When words or constructions are borrowed in one language from another language, they do not always replace existing patterns completely. For instance, it is common for words that are borrowed to receive a narrower interpretation than they had in the source language. Thus, famously, English nouns denoting domestic animals are largely of Germanic origin, whereas the denominations of the corresponding kinds of meat come from French, although in French, these words were also used for the animals. Neither Old English nor Old French seems to have had more than one word corresponding, for example, to Modern English sheep and mutton; when Old French moton ‘sheep/meat of sheep’ was borrowed into Middle English with the more restricted meaning ‘meat of sheep’ a distinction was created that did not exist in either of the contributing languages. 4.  The geographical dimension But there is also a geographical aspect to this: the effects of language contact are not likely to be equally strong everywhere. A contact-induced change is likely



Increases in complexity as a result of language contact 

to be implemented more consistently in the part of the target language area that is closest to the source language. Where the influence of the source language is weaker, we may obtain the kind of compromise or “truce” that was spoken of above, leading to the rise of new grammatical or lexical distinctions through a division of labour between old and new patterns. One example of such a development from the Scandinavian languages concerns the position of adnominal possessive pronouns. The original dominant Germanic word order is Noun + PossPron, as in Gothic waurd mein ‘my word’. The PossPron + Noun order expanded into Scandinavia from the West Germanic languages during the early Middle Ages. However, it met resistance from the old pattern and halted in mid-Scandinavia. In many varieties of Peninsular Scandinavian, postposed possessive pronouns are still used quite generally, as in Norwegian bilen min ‘my car’ or Southern Westrobothnian katta minä ‘my cat’, but in a large area in southern and central Sweden, they are possible only with kin terms, e.g. far min or min far ‘my father’ but only min bil ‘my car’ (see Map 1, which is based on data presented in Delsing 2003). We thus obtain a grammatical distinction that is in essence parallel to the split between alienable and inalienable possession found in many other languages. The result is an increase in grammatical complexity in the intermediate area in which the change in word order is not wholly implemented.

Predominantly postposed possessive pronouns

Postposition of possessive pronouns only with kin terms

Only preposed possessive pronouns

Map 1.  Possessive pronouns in Sweden

Some postposed non-kin possessives

 Östen Dahl

5.  Contact-induced suppletive patterns A second example of contact-induced complexity which I want to discuss and which has a somewhat similar geographical distribution concerns the rise of a suppletive verb paradigm as the result of the competition between two verbs for ‘become’ in Swedish: varda and bli(va). The former goes back to a common Germanic verb for ‘become’ which is derived from the Indo-European root *u̯ert- ‘turn’. In Swedish (like in the other Continental Scandinavian standard languages), varda has been largely replaced by bli(va/ve), derived from Low German bliven ‘remain’. It is unclear when and where the shift of meaning from ‘remain’ to ‘become’ took place; it is not well attested in written Middle Low German (Markey 1969), but it is striking that a development ‘remain/ stay’→‘become’ is found not only in Scandinavian but in a number of languages spoken in the area around the Baltic Sea where the Hanseatic League was active: Polish zostać, Lithuanian palikti, Latvian palikt, Estonian jääma. It can therefore be speculated that the change originated in some spoken variety that was current in the Hanseatic area. What is of interest here though is the suppletive paradigm that has arisen as a result of this development. If we ignore the existence of varda in very archaic forms of written Swedish and instead look at spoken standard and non-standard varieties, looking at the Swedish dialect area, we find that the replacement process was complete only in the southern part (Götaland) and on Gotland, i.e. the parts that had the most direct contacts with the Hanseatic area. Some northern vernaculars have preserved the old varda paradigm fully, but in a large area in central and northern Sweden bli has taken over only in the present and the infinitive, whereas forms of varda survive in the preterite (vart) and the supine (most common form: vurti). In urban spoken varieties of central Sweden (i.e. the spoken varieties closest to the standard), the only form that has survived is the preterite vart. The facts are summarized in the following table: Table 1.  Variation in bliva/varda paradigms in Sweden

Southern/Western and written standard Central urban colloquial Central/Northern vernaculars Conservative Northern vernaculars

infinitive

present

supine

preterite

bli bli b(l)i varda

blir blir b(l)ir varder/vaL

blivit/blitt blivit/blitt vurti vurti

ble(v) blev/vart vart vart

6.  Buffer zones Again, we can see that an incomplete process of change gives rise to a more complex situation in an intermediate zone between two areas with consistent patterns.



Increases in complexity as a result of language contact 

The term “buffer zone” was used in Stilo (2004) to refer to the more general phenomenon of a typologically “inconsistent” zone between two areas with consistent typological patterns. As an example of such a zone, consider the Uralic languages in Europe. It is well-known that there is a high correlation between postpositions and verb-final word order; languages such as Finnish, Estonian, North Saami and Erza Mordvin are exceptions in having postpositions although they are SVO languages. This buffer zone is squeezed in between the western VO/preposition area and the eastern OV/postposition area. The reasonable hypothesis here is that the SVO word order has arisen under influence from the west. It is perhaps less clear that we are here dealing with differences in grammatical complexity.

Map 2.  Map text: Dark grey area – preposed articles only; light grey area – suffixed articles. Solid line enclosed area – preposed article only weakly represented, dotted enclosed area – complementary distribution of articles

7.  Definite articles in Scandinavian: Two grammaticalization waves meet What I want to discuss now is an example of a “buffer zone” where we are not dealing with one spreading innovation but rather with two interacting grammaticalization processes that operate in partly overlapping areas. (A more detailed account is found in Dahl 2003.)

 Östen Dahl

In the Scandinavian languages, two kinds of definite articles appear: preposed free articles (den, det, de), similar to the ones of West Germanic, and suffixed articles (-en, -et etc.). In addition, definiteness is also reflected in the choice of “weak” rather than “strong” endings of adjectives. The story usually told about definiteness in Scandinavian is that there are two main patterns. Thus, in Danish (and marginally in Norwegian), suffixed articles are used in definite noun phrases which do not contain a preposed modifier, and prefixed articles when the noun phrase contains an adjective or a quantifier. In Danish, then, the two types of articles obey a rather strict complementarity principle. By contrast, in Swedish and most forms of Norwegian, while single nouns take an S-article, like in Danish, definite NPs with preposed attributes typically contain both a P-article and an S-article. This is commonly referred to as “double determination”, “over-determination” and will be illustrated here by Swedish examples: Table 2. noun only adjective + noun

indefinite

definite

ett hus ‘a house’ ett stor-t hus ‘a big house’

hus-et ‘the house’ det stor-a hus-et ‘the big house’

The actual picture is a bit more complicated than the standard account suggests. There are more than two patterns, and there are some special contexts where the patterns are in competition. Thus, in Swedish, we find variation above all between the double determination pattern and one where there is only a suffixed article. One context where the latter pattern is the normal one in Standard Swedish is what we can call name-like uses of definite noun phrases, as exemplified by Vita huset ‘the White House’. In Danish, a single preposed article is the most common option. Nynorsk Norwegian uses double determination, whereas Bokmål Norwegian exceptionally sides with Danish. This is summarized in the following table: Table 3.

Danish Norw. bokmål Norw. nynorsk Swedish

Definite description

Proper-name use

det hvide hus det hvite huset det kvite huset det vita huset

Det Hvide Hus Det Hvite Hus Det Kvite Huset Vita Huset

Another context where we find similar variation is in noun phrases which contain pre-modifiers that are usually treated as belonging to quite different



Increases in complexity as a result of language contact 

categories, but for which I have coined the term “selectors”. This includes (i) words like samma ‘same’, sist(a) ‘last’, nästa ‘next’, förra ‘previous’, höger/högra ‘right (hand)’, vänster/vänstra ‘left’ (hand) etc.; (ii) ordinal numerals; (iii) superlatives. In addition to the standard patterns described above, we find here also the suffixed article used alone (but only in Norwegian and Swedish) and a further pattern with no article at all, which shows up in all three languages. The possibilities (based on an Internet search) are shown (in somewhat simplified form) in the following table, where an asterisk means that the pattern is unattested. The preferred options are in boldface. Table 4.  Variation in definiteness marking with “selectors” in the Continental Scandinavian standard languages

no article preposed article only suffixed article only

Danish

Norwegian

Swedish

højre hånd første gang *ældste sønn den højre hånd

høyre hånd første gang *eldste sønn (den høyre hånd)

höger hand första gång *äldste son *den högra hand

‘the right hand’ ‘the first time’

den første gang den ældste søn *højre hånden

den første gang den eldste sønn høyre hånden

*den första gång *den äldste son högra handen

‘the first time’ ‘the eldest son’ ‘the right hand’

første gangen eldste sønnen den høyre hånden den første gangen den eldste sønnen

första gången äldsta/äldste sonen den högra handen den första gången den äldsta/e sonen

‘the first time’ ‘the eldest son’ ‘the right hand’ ‘the first time’ ‘the eldest son’

*første gangen *ældste sønnen double articles *den højre hånden *den første gangen *den ældste sønnen

‘the right hand’

What we can see in name-like uses and NPs containing selectors is that there is a general tendency towards more restrictive use of articles than in other contexts. However, this tendency works differently in the Scandinavian languages: in Swedish, it is primarily the preposed article that is omitted, while in Danish, the tendency is generally weaker but when it applies, it results in articlelessness. Norwegian is a more mixed case. Arguably, this is not an isolated fact but part of a general tendency for the two articles to have different degrees of strength in the Scandinavian languages, with the preposed article being strongest in Danish and the postposed being strongest in Swedish. This becomes clearer if we widen the perspective to other constructions and to non-standard varieties of Scandinavian. Thus, going from the north to the south in Sweden, we can see what looks like a cline in the use of preposed articles. In the north (Norrland and Dalarna),

 Östen Dahl

many vernaculars have only a very limited use of the preposed article, which may be explainable as relatively late influence from the standard language. In this area, adjectives are frequently incorporated (nybiln ‘the new car’). The dialects of Central Sweden do not differ from the standard language here, but it does appear that southern vernaculars are less prone to omit preposed articles with selectors than Standard Swedish. And finally, the vernaculars of South Jutland (sønderjysk) use preposed articles across the board, in the same way as the West Germanic languages. Conversely, suffixed articles, which are not found at all in the southwestern corner of Scandinavia, generally become stronger as we move northwards. Thus, demonstratives are usually followed by definite forms of nouns (i.e. nouns with suffixed articles) in Swedish and Norwegian, while in Danish the head noun is indefinite,2 e.g. Swedish den här boken ‘this book’ and Norwegian denne boka vs. Danish denne bog ‘this book’. Whereas Standard Swedish behaves like Danish in having indefinite forms after an s-possessive (Pelles hus ‘Pelle’s house’), some vernaculars (traditional dialects) spoken in the eastern and northern parts of the Swedish dialect area (Gotland, Estonia, Finland and northern Sweden) use definite forms of the head nouns of such NPs (Pelles huset). In an area overlapping with the one just mentioned, there are also many other extended uses of definite forms, most notably for what I have called “non-delimited” uses of mass nouns and plurals, coinciding roughly with the cases where French would have a partitive article, as in nôppä bära ‘to pick berries’ (Selånger, Medelpad, Sweden), where bära is a definite plural form. The different geographical distribution of the two articles suggests that we are dealing with two separate grammaticalization processes which have been operating in partially overlapping geographical areas. While the preposed article is part of a general West European tendency, the suffixed article appears to be a more local Scandinavian affair. In the “buffer zone” between the two grammaticalization areas both preposed and suffixed articles are in use, but a division of labour has been found between them, the details of which vary depending on the location. Most saliently, Danish opts for a solution in which the articles have a strict complementary distribution, while the peninsular varieties of Scandinavian display an extensive use of “double determination”. As we have seen, however, there are many cases of deviation from the major patterns. The Scandinavian definite articles thus illustrate how language

2.  In Standard Swedish, denne/denna takes an indefinite form, but this is most probably a feature of the written language under Danish influence. In the Swedish dialects where denne occurs naturally, it is usually combined with a definite form of the noun, like in Norwegian.



Increases in complexity as a result of language contact 

contact gives rise to a complex grammatical system through the competition between two grammaticalization processes.3 8.  Conclusion Summing up, competition between linguistic patterns is a frequent source of linguistic complexity, in that new distinctions emerge when a division of labour is established between the patterns. New patterns are often introduced through language contact and get into competition with already existing ones. This leads to the rise of complexity that was not previously found in either of the languages involved. Moreover, since patterns tend to spread in a wave-like fashion, there is also a geographical dimension, in that the strength of a pattern will decrease as the distance to the source language increases. More intricate situations can arise for instance when two grammaticalization waves meet and a “buffer zone” is created in which either a division of labour between the grammaticalizing elements is established (Danish) or they combine in one pattern (Peninsular Scandinavian). While the examples in this paper were taken from North Germanic, I expect more cases to turn up in other languages, once linguists become more aware of the phenomenon.

References Dahl, Ö. 2003. Competing definite articles in Scandinavian. In Dialectology Meets Typology, B. Kortmann (ed.), 147–80. Berlin: Mouton de Gruyter. Dahl, Ö. 2004. The Growth and Maintenance of Linguistic Complexity [Studies in Language Companion Series 71]. Amsterdam: John Benjamins. Delsing, L.-O. 2003. Syntaktisk variation i nordiska nominalfraser. In Dialektsyntaktiska atudier av den nordiska nominalfrasen, Ø.A. Vangsnes, A. Holmberg & L.-O. Delsing (eds), 11–64. Oslo: Novus. Koptjevskaja-Tamm, M. 1996. Possessive NPs in Maltese: Alienability, iconicity and grammaticalization. Rivista di Linguistica 8(1): 245–74. Kuteva, T. 2008. On the frills of grammaticalization. In Rethinking Grammaticalization: New Perspectives, M.J. López-Couso & E. Seoane (eds), 189–217. Amsterdam: John Benjamins.

3.  This account glosses over a number of further complications. In particular, preposed articles seem to have arisen from demonstratives not just once but several times in the Scandinavian area. For example, the distal demonstratives den där and han där are frequently found in the function of definite articles before modifiers in several different locations in the Swedish dialect area.

 Östen Dahl Markey, T.L. 1969. The Verbs varda and bliva in Scandinavian [Studia Philologiae Scandinavicae Upsaliensia Uppsala 7]. Stockholm: Almqvist & Wiksell. McWhorter, J.H. 2001. The world’s simplest grammars are creole grammars. Linguistic Typology 5(2/3): 125–66. Miestamo, M., Sinnemäki, K. & F. Karlsson. 2008. Language Complexity: Typology, Contact, Change [Studies in Language Companion Series 94]. Amsterdam: John Benjamins. Rozental, D.E. 1968. Praktičeskaja stilistika russkogo jazyka. Izd. 2, ispr. i dop. ed. Moskva: Vysšaja škola. Sampson, G., Gil, D. & Trudgill, P. 2009. Language Complexity as an Evolving Variable. Oxford: OUP. Stilo, D. 2004. Iranian as buffer zone between the universal typologies of Turkic and Semitic. In Linguistic Convergence and Areal Diffusion: Case Studies from Iranian, Semitic and Turkic, É.Á. Csató, B. Isaksson & C. Jahani (eds), 35–63. London: Routledge/Curzon. Trudgill, P. 1983. On Dialect: Social and Geographical Perspectives. Oxford: Blackwell.

Converging genetically related languages Endstation code mixing? Kurt Braunmüller

Hamburg, University and Collaborative Research Centre on Multilingualism The main goal of this paper is to investigate and model the impact of genetically close relationships between languages in contact situations and to describe the consequences for bilingual speakers living in bilingual communities. Both Muysken’s concept of “congruent lexicalization” and Myers-Scotton’s definitions of “composite matrix languages” will be evaluated and an outline of a code mixing hierarchy designed to achieve a better understanding of convergence processes will be proposed. The main source of the data is an ongoing research project on “Variation in Multilingualism on the Faroe Islands”, supplemented by other parallel data from the German-Danish border region and German vernaculars. The results of the data analysis strongly suggest that, in the long run, code mixing is inevitable, preconditions being (balanced) bilingualism and the frequent alternating use of the two contact languages/varieties.

1.  Introduction1 1.1  The investigation of contact phenomena between genetically closely related languages has been a stepchild both within contact linguistics and research on bilingualism in general, although some very special types of convergence and mixing occur which cannot be found in other bilingual settings. Therefore it is important to gain more insight into bilingual speech constellations involving structurally related languages. Peter Muysken (2000: Ch. 5) has thoroughly analysed such contact phenomena and classified them as “congruent lexicalization”, a term which primarily focuses on structural aspects (syntax, morphology). Such a focus is however clearly not detailed enough to give a comprehensive view of all mechanism that actually occur in such bilingual settings, especially when the two (or more) languages in question

1.  I would like to thank two anonymous reviewers for their comments on an earlier version of this paper. All resting errors and deficiencies are mine, of course.

 Kurt Braunmüller

are used in frequent alternation in everyday life, often accompanied by various forms of code-switching. Muysken defines his point of view as follows: The languages share the grammatical structure of the sentence, fully or in part. The vocabulary comes from two or more different languages. When only parts of the grammar are shared by the two languages, it is often the alignment of the major constituents, but not all the internal structure of these constituents. Thus grammatical convergence leads to congruent lexicalization. (Muysken 2000: 122) If congruent lexicalization is frequent in a bilingual setting, it could be due to two things:



a. There is an overabundance of homophonous words, diamorphs, that serve as bridges or triggers for the code-mix; b. There is a general structural equivalence, both categorial and linear, making code-mixing possible, without there necessarily being any lexical correspondence. (p. 123, his emphasis)

Carol Myers-Scotton (2002: esp. Ch. 5) defines the simultaneous use of two languages which share basic grammatical structures as “composite matrix language” in the following way: From a structural standpoint, convergence and attrition are very similar as outcomes, perhaps, even identical. As outcomes, they have two distinctive features: (i) all surface morphemes come from one language; (ii) the abstract lexical structure projecting these morphemes no longer comes from one language, but includes some abstract structure from another language. The result is a composite morphosyntactic frame or Matrix Language. (p. 164) Both convergence and attrition necessarily involve a composite Matrix Language. That is, the source of the grammatical frame for the relevant CP [complementizer phrase/clause; K.B.] is not a single language, but more than one. Thus we speak of the frame as a composite. […] the bilingual CP […] includes codeswitching (meaning morphemes from two or more varieties) and convergence (abstract grammatical structure from two or more varieties). (p. 165)

1.2  There are, however, far more forms of code mixing which have not been described sufficiently or have even not been detected before, although the convergence phenomena involved are immediately obvious when one compares the languages as they are used by bilingual speakers with the non-contact varieties of the languages used by monolinguals (cf. Fredsted (2008) for the shared structural use of related verb phrases by bilingual German-Danish school children). For instance, covert code mixing may occur where the impact from the (non-dominant local) linguistic variety in contact remains invisible if only the



Converging genetically related languages 

grammatical surface structures and the phonology of these sentences are taken into account. Since both the typological and systemic structures between genetically related languages are quite similar due to their, in part, common linguistic history, the first step towards code mixing between closely related languages would affect not only the lexicon via borrowing or grammatical transfer (as this has been stated by Thomason and Kaufman 1988: 50, 74ff. and many others for language contact situations in general). It clearly also affects the societal linguistic norm. In most cases, bilingual speakers of genetically related languages use the societal norm of the dominant local language also for the other language they use alternatively in other contexts. This may, however, lead to various kinds of norm conflicts (see e.g. Braunmüller 1990). In some cases, native speakers of the non-dominant language coming from outside may have problems fully understanding what has been said. Without a certain command of the model language, it can be difficult to grasp what the bilingual speakers actually want to say. Grammatical details for various kinds of such comprehension difficulties will be discussed in the next sections. But even performance phenomena of the linguistic behaviour of bilingual speakers can trigger serious irritations on the part of addressees who have no experience with for example, the default or ‘normal’ use of discourse particles in informal face-to-face communication. I would like to give an example of a Faroese native speaker, who frequently switches into Danish (and sometimes even into English) when speaking to nonnatives – a situation which is typical for most of the inhabitants of the Faroe Islands, where everybody also has an excellent command of the Danish language and is used to frequently switching into Danish in every-day life: Ha?2 (‘I beg your pardon; sorry, I did not understand you’) uttered by a native speaker of Faroese when using his/her second language Danish in a dialogue with a (monolingual) Dane causes some embarrassment or even confusion, because ha? will not be interpreted as an equivalent to Danish hva’ beha’er? or hvad? but rather as a harsh and impolite discourse particle with an unknown semantic function.

The explanation for this communicative behaviour is quite simple: The Faroese speaker has replicated (or transferred) the discourse particle ha? from his/ her mother tongue Faroese using it in exactly the same way as in his colloquial Danish speech. He or she is clearly not aware of the fact that an emphatic ha? uttered in a monolingual Danish conversation is regarded as a very impolite and crude interruption of his or her interlocutor in a dialogue. Its use is clearly face-threatening.

2.  Letters in bold face indicate emphasis and/or a very distinct pronunciation.

 Kurt Braunmüller

The Faroese speaker does not realize this because s/he thinks that the same pragmatic norms apply for either language. But a more general explanation offers itself: a common pragmatic (here: conversational) norm is obviously operative between native bilinguals on the Faroe Islands without attracting attention – but not beyond, especially not in Denmark. Moreover, an extensive use of all, i.e. also currently non-activated, systemic possibilities inherent in the linguistic systems involved, can be observed.3 This may encourage spontaneous linguistic creativity by bilinguals in the alternating use of these languages and may in the long run, lead to the unimpeded transfer of grammatical structures from the model language to the replica language (cf. Heine & Kuteva 2005: esp. Ch. 2): Grammatical replication has the effect that the replica language (R) acquires some new structure (Rx) on the model of another language (M). […] the new structure Rx is in most cases not entirely new: rather, it is built on some structure (Ry) that already existed in the replica language, and what replication then achieves is that it transforms Ry into Rx (Heine & Kuteva 2005: 40).

The main variables of grammatical replication [are:] a. the sociolinguistics setting b. the structure of the model language c. the structure of the replica language d. communicative needs and intentions e. cultural values characterizing the communities associated with the model and the replica languages, and f. creative behavior activated to the use of linguistic resources available in novel ways. (p. 35) In which way such processes may have been active over a longer period in history has been summarised in Diercks and Braunmüller (1993) and Braunmüller (2004), where we investigated the transfer and contact induced creativity (esp. in word formation; see e.g. Diercks 1993) from the model languages Middle Low and, later, Early High German on the replica languages in Mainland Scandinavia, Old Danish and Old Swedish. The parallel use of Latin and German – both functioning as models for (elaborate) written languages at that time – and the vernacular Swedish by trilinguals in late mediaeval Sweden and its consequences for the emergence of many structural features of early modern Swedish texts has been investigated in detail by Höder (2008). 3.  For the psycholinguistic aspects of these processes see for example Grosjean (1999) and de Bot (2004).



Converging genetically related languages 

1.3  The social acceptance of convergence varieties is often very high throughout bilingual communities, for example between speakers of (local) dialects and the standard language, or in societies where both languages have coexisted and were used side by side for centuries. Convergence phenomena, though often stigmatised by non-local or monolingual speakers of the standard language, represent an important part of the speakers’ local (linguistic) identity: Some bilingual speakers may be proud of being different from the other (often dominant) part of the population. Any attempt to criticise language convergence phenomena, for whatever reasons, will therefore be met with reluctance and considered an assault on speakers’ personal local integrity. Other bilinguals are, however, very well aware of the fact that their command of the standard language is insufficient and feel therefore handicapped that they are not able to live up to the norm of the monolingual speakers of that language. Examples for areas where converged varieties of genetically closely related languages have been in intense contact for many centuries are, as far as Germanic languages are concerned, for example (a) the Faroe Islands with the vernacular Faroese, a West Scandinavian language similar to modern Icelandic, and Danish as an elaborate standardised language,4 (b) Norway where the local (rural) dialects and the H-variety Danish and, later, Norwegian Riksmål, respectively, were used or (c), the German-Danish border region with the standard varieties of German and Danish on the one hand and South Jutish, the local (transnational) Danish dialect, on the other. Regionally restricted contact varieties have, however, also emerged, for example South Schleswig Danish, Sydslesvigdansk, or North Schleswig German, Nordschleswigerdeutsch, or Norwegian Riksmål/Bokmål, in English often explicitly called “Dano-Norwegian” (Haugen 1966; 1976), a Danish based creoloid,5 and today the prevailing variety of the two written standard forms of the Norwegian language. I would like to give some simple illustrations of such locally deviant linguistic varieties. The morphological divergences of modern Norwegian compared to the original standard Danish language used in Norway for centuries are small, Norway having been an integral part of the Danish kingdom (between

4.  The Faroe Islands have been part of the Danish kingdom since 1380, when Norway and its colonies in the north Atlantic Sea lost their independence. Since 1948 the Faroe Islands got more autonomy (home rule) but are still a part of Denmark (but they are not a member of the European Union). 5.  A creoloid is a result of language contact that “experienced simplification and admixture [i.e. language contact], but has not undergone the reduction associated with full pidginization (nor, therefore, the expansion associated with creolization)” Trudgill (1992: 22; his emphasis).

 Kurt Braunmüller

1380 and 1814).6 In Norwegian, you have to say ¼ på 7 ‘6:45 [a.m./p.m.]‘, whereas in Danish only ¼ i 7 is possible. På is the most general, less restricted Danish preposition and has therefore been easier to acquire for the (formerly) second language learners of Danish in Norway. Bilingual speakers of the local variety of Danish (sydslesvigdansk) in the northernmost parts of Germany (Schleswig) often use prepositions in the same way as they would use them in German, the majority’s language: Jeg har pakket mine bukser ind i kufferten. ‘I have put my trousers [lit.] in (between) my suitcase’, Germ. Ich habe meine Hose in den Koffer gepackt. (‘… into my suitcase’). In (monolingual) standard Danish, one only can say: Jeg har pakket mine busker ned i kufferten ([lit. … down into …]; cf. Christoffersen 1991: 28–31). In both cases the model was already available in the replica language; its scope has, however, been extended, as described by Heine and Kuteva (2005: 40) mentioned above. 1.4  In the following, I will first discuss characteristic examples from the Germanic languages in contact mentioned above. I will then outline a model of language mixing between genetically closely related varieties, their conditions and restrictions. Furthermore, I will demonstrate that language standardisation follows the same principles in terms of converging divergent dialects into one unified roofing language: The aim of all these processes is (a) to reduce variation and (b) to create a new linguistic system with a new societal norm in its own right. Or, to put it more generally, code mixing may, in principle, occur in any language contact situation, but it remains often unrevealed – especially when genetically closely related varieties are involved. 2.  From overt language contact phenomena to covert code mixing 2.1  Lexical code switching Lexical code switching is one of the most basic and most evident forms of any language contact. For instance, a widespread case is when a speaker is, for a moment, not able to remember the equivalent or an otherwise appropriate term in the other language (cf. (1a/b/c) below). If there is a majority language, it may happen that the speaker borrows the term from this language in order to make sure s/he will be understood immediately and properly (cf. (1d/e) below), because the locations or things referred to are only known by everybody in the majority’s language.

6.  The ‘union’ of the common written language Rigsmaal, later written Riksmål, in Norway did not end before 1907, when the Norwegian way of pronouncing Danish replaced the original Danish spelling of certain consonants in the interlude and the coda; they became then devoiced obstruents (b, d, g > p, t, k) in spelling as well.



Converging genetically related languages 

Furthermore, there may be no directly corresponding term in the other language at all, because the institution referred to does not exist in the country, which prohibits a direct or even a spontaneous loan translation for obvious reasons. The Faroese-Danish examples are, for the most part, taken from our SFB-project K8: “Variation in multilingualism on the Faroe Islands”, and occasionally also from Faroese websites and books on Faroese; the German-Danish examples are taken from Braunmüller (1996, 2001) and Fredsted (2008).

(1) a.



Far.-Dan. – og tá hevði eg eisini eina romantiska fyristilling um at … ‘and then I also had a romantic view about …’ (Dan. [en] romantisk∅ forestilling [om]) (www.kvinna.fo; Sept. 10, 2008)

b. Far.-Dan. – Tað er forundarligt, at tú ikki skalt forstanda ein dunklan hentydning. ‘It is strange that you are unable to [lit.: that you not shall] understand a vague hint.’ (Dan.: Det er forunderligt, at du ikke kan forstå en dunkel hentydning.) (Petersen & Adams 2008: 17 [7e]) Dan.-Far. – Og hun var ikke tíðarhóskandi, hun var stærk … men hun var for smøl. ‘And she was not suitable for the time, she was strong but she was too thin.’ (Dan. [in ital.]: 1. no equivalent; 2. smal; without umlaut in the feminine, an Embedded Island) c.

d. Dan.-Far. – Det er ved siden af Bartalstovan (Far.: við síðuna av Bartalstovuni [dative]) ‘That’s close to the B.’ [with an enclitic definite article in both versions] e. Dan.-Germ. – o.k, vi mødes så ved ZOB kl. 4; ‘o.k., let’s meet at thecentral bus station [in Flensburg] at 4 o’clock; – min bil skal til TÜV i næste uge; ‘my car has to be checked by theTechnical Control Board next week’ – Dansk Alleinunterhalter Svend P., Wallsbüll, tfl. … [FlA7 1994/12/1] ‘Danish solo-entertainer Sv. P., W., phone …’ 7.  “FlA”: Flensborg Avis, the local bilingual daily newspaper in Flensburg. Most of the content is, however, written in Danish; at that time (1994/95) with many structural and lexical loans from German, especially in word order and in phraseology – not to speak of the many obvious loan translations, which in most cases were incomprehensible for any native speaker of Danish in Denmark (for a more detailed analysis see Braunmüller 1996).

 Kurt Braunmüller

Lexical code switching as demonstrated in the above examples is quite common as long as the addressee is familiar with the other language as well and needs no further explanation. Things are more complicated when loan translations are used which do not exist in the other language (see (2 a/b) below). The main reasons for using loan translations are either ignorance, insufficient familiarity with the political, societal etc. structures (in this case) in Denmark or laziness, because you can take for granted that you will be understood anyway due to widespread bilingualism. The examples in (2b) would, however, not (or not properly) be understood in Denmark itself, esp. not the first and the third example of (2b). (2) a.

Dan. hvad hedder det, trolari, trawlbåd … (switch to Far.; cf. Dan. trawler) ‘how do you call this, trolari, trawlboat’ [sc. trawler] (K8-corpus)

b. Dan. – Forfatningsbeskyttelsen skal have halveret sit buget ‘The secret service’s budget shall be halved.’ (Dan. efterretningstjeneste; Germ. Verfassungsschutz: loan translation) [FlA 1994/12/9]

– I et lokale i Lolfod kom to gæster op at skændes. ‘In a restaurant/inn in Lolfod two guests started quarrelling.’ (Dan. værtshus; Germ. Lokal, Dan. lokale ‘room’) [FlA 1995/1/3]

– Men telefonen i kunstnerhuset vil ikke stå stille. ‘But the phone in the artist’s house does not stop moving around [sc. ringing]’ (Dan. kimede uafbrudt, Germ. will nicht still stehen).[FlA 1995/1/10]

At first glance, one might regard the Danish examples in (2b) as evidence of an insufficient command of Danish or as clear cases of (careless) linguistic transfer from German. But the situation is actually more complicated. As German had great influence on the Danish language since the days of the Hanseatic League, i.e. from the 13th century onwards up to the middle of the 19th century (and beyond), many loan words and idiomatic phrases entered the Danish language. They were not looked upon as foreign, primarily due to their close genetic relationship, and became thus integrated without any problems. Since Danish as an indigenous minority language south of the German-Danish border is still in lively contact with German due to obvious political reasons, this symbiotic development between these two languages has never come to an end in this area. The situation can therefore be considered equivalent to the linguistic situation of the Faroe Islands (cf. Kühl and Petersen in this volume), where Danish has also been a spoken and a written language since the late Middle Ages. We are here dealing with a question of following a national linguistic norm whether these loans are acceptable or not. As far as Schleswig is concerned, one can check the local usage of Danish as a minority language against the (monolingual) language use in Denmark. It is therefore difficult for persons who have not spent



Converging genetically related languages 

a longer time in Denmark to know precisely which ones are old and therefore accepted loans, and which ones are ad hoc loan translations or cases of code copying (cf. Johanson 2008) not acceptable to native speakers in Denmark. 2.2  Covert lexical code switching Covert lexical code switching is the next step towards language convergence. It may become difficult or sometimes even impossible to recognise words as loans because their phonological form is indistinguishable from inherited forms, such as in (3a) vs. (3b): (3) a. 

Abends brachte[!] S. und B. ein Bukett Blumen ‘in the evening, S. and B. came along with a bouquet’ (Braunmüller 2001: 122)

b. Nachm[ittags]8 süsselte ich im Garten ‘in the afternoon I did some work in the garden‘ (Braunmüller 2001: 123)

(3a) seems to be stylistically marked. Normally one would expect the native German noun Strauß and not the more sophisticated loan word from French Bukett ‘bouquet’, but anyway (3a) can be considered a perfect (colloquial)9 German sentence. In (3b), one gets the impression that the verb süsseln, a typical regional word (in Schleswig), may have been taken from the local vernacular Low German. Without knowing the background of the speaker – a maritime painter from Flensburg, raised by a Norwegian mother in Flensburg, who scribbled down more or less noteworthy things in his diary during the Second World War (here in 1940) – will never guess that this person is actually bilingual with a very strong active competence in Danish, and certainly a good passive command of the regional vernacular and the local dialect as well. (3a) might be translated directly word by word into Standard Danish: (3a') Om aftenen kom S. og B. med en buket blomster by making use of congruent lexicalisation according to Muysken’s definition. (3b) contains a Low German verb süsseln, which has an equivalent in the local south Jutish Danish dialect (å sysl). Other examples of this kind, which can be found both in (Low) German and (dialectal, here: Jutish) Danish, are püseln ‘to do some minor work‘ or Nükken haben ‘to be capricious/fickle’ and many others (cf. also Lindow 1984). My argument here is: “you can never tell where a word comes from” – as LePage (1992) has stated after having analysed a Caribbean scenario where creole languages were involved.

8.  Abbreviation, often found in this diary. 9.  The grammatically correct verb form would normally be “brachten” (‘brought’ [3pers.pl.]). Singular verb forms (such as here brachte) referring back to coordinated NPs are, however, quite frequent in colloquial German.

 Kurt Braunmüller

2.3  Covert replication of the verbal frame The next step in converging genetically closely related languages is the replication of the verbal frame (or a part of it) from either language, which is filled lexically by elements of the other language, such as in (4a/b) and in (4c/d) on the one hand and (4e/f) on the other. The Faroese examples in (4a/b) show that the other co-activated language Danish supplies the verbal frame, and the articulated language, Faroese, the wording. Strictly speaking, we observe in these cases some sort of abstract code switching, which normally will not be detected due to widespread bilingualism. The same can, pari passu, be said about the Danish counterparts, as documented in (4c/d): the verbal frame has been transferred from Faroese [4c] into Danish (cf. Far. halda > Dan. [1] synes ‘to think/to feel’ and [2] tro ‘to believe/ suppose’) or from English [4d] into Danish (cf. Engl. to go > Dan. (at) gå ‘to walk’, equivalent to Far. ganga; to go means fara in Far.). The Examples (4e/f) document similar phenomena: due to phonological identity between South Jutish and German and German and Danish, respectively, the frame from the other co-activated language may be used without causing any form of misunderstanding. (4) a. Vit skuldu svara uppá okkurt skrivtstað… ‘We should reply on some Bible verse … (Dan: Vi skulle svare på [prepositional construction] et eller andet skriftsted, cf. Far. Vit skuldu svara onkrum skrivtstaði [dative]) (K8-corpus, Far. part) b. eg havi aldrin tonkt um hatta spælið fyrr … ‘I have never thought about that game before …’ (Dan. jeg har aldrig tænkt på denne leg før…; cf. Far. eg havi aldrin hugsað um hatta spælið fyrr…) (K8-corpus, Far. part; same person) c.

Der er omkring tredive, fyrre mennesker, synes jeg. ‘[lit.] there are about thirty, forty people, I think.’ ‘There are about thirty or forty people, I believe.’ (K8-corpus, Dan. part)

d. Jeg ønsker at gå til Kalifornien. ‘[lit.] I wish to walk to California.’ [not possible from the Faroe Islands!] ‘I want to go to California.’ (K8-corpus, Dan. part) e. oh ha hvo de nerve[r] ‘oh, how they get on your nerves’ (Germ. nerven; cf. Dan. enervere, at gå en på nerverne: ‘to get s.o. on his nerves’) (Fredsted 2008: 995 [2]) f. dann sollen wir heute aufräumen ‘then we are going to clear up today’ (Dan. skal = Germ. 1. sollen, 2. werden [future]; with negation 3. (nicht) dürfen; cf. also 4. in (4e)) (Fredsted 2008: 975 [29])



Converging genetically related languages 

The modal sollen with another, divergent reading than the one mentioned in (4f) has been replicated from Low German turning into an integral part of the regional modal system. This is illustrated in (4g), which sounds quite odd for most speakers of German who are not familiar with that vernacular. They would immediately ask for the person who told you to get these rolls, especially if the speaker is not a small child.10 (4) g. Ich soll sechs Brötchen haben. ‘I would like to have six rolls [lit. I have been told to get …]’ (Germ. wollen; gerne [etwas] haben) (Braunmüller 2000: 327)

If you asked that question – who has told you to do so? –, the addressee will not necessarily understand your point because everybody speaks like that in this area, and s/he cannot imagine that something might be odd. 2.4  Overt replication of the verbal frame The verbal frame can also be overtly taken from the contact language, as demonstrated in (5 a/b).11 Similar types of data can also be found in our Faroese-Danish corpus, which is documented in the examples (5c/d). The grammatical procedures can, in both cases, be considered absolutely parallel to and typical of convergence processes between genetically closely related languages. (5) a.

nej, de skal da forkleide jer ‘no, they must dress you up, you know’ (Germ. verkleiden)

b. så ville jeg probere ‘then I would try‘ (Germ. probieren)

(Fredsted 2008: 956 [4]) (Fredsted 2008: 957 [6])

10.  Further examples of that kind are discussed in Kühl (2008: 175 [112] and 188 [120]):

[112] “sie sollten sich treffen” (DH6) ‘they are going to meet each other’ and [120] “ich und Ole sollen mit nach Dänemark” (HfA1) ‘I and Ole will join our trip to Denmark’.

11.  Further examples are presented in Kühl (2008: 131 [41, 42] and 133 [45, 46] where the matrix language is Danish and the embedded language German (in both cases, the infinitival marker -e is Danish):

[42] “han kan angebe med hende” (HH1) ‘he can have a show-off with her’ [45] “og så vil jeg beobachte den” ‘and then I will observe_ this’.

 Kurt Braunmüller

c. Nej, mange, nej det var ikke, det er forskelligt, det hehe, forskelligt hvor vel det lykkes og så meget sne, så dør burt- doyr burturav. Nej, det er forskelligt. ‘No, many, no, it was not that, it is different, it {laughs} [is] different how good the outcome may be, and then so much snow, so [it] dies, dies’  (K8-corpus, Dan. part) d. Og, men jeg kan, hvis jeg ikke har god belysning, så kan jeg også løse ­krydsogtværs, lægge kabale ogspasera og strikke. ‘Well, but I can. If I do not have any good light, then I do also solve crosswords, lay patience and walk about [cf. Dan. spadsere] and knit’.  (K8-corpus, Dan. part)

In some cases, problems occur with linearisation, as demonstrated in ((6a); cf. also Petersen 2006 and (6b)): (6) a.

… so hevur mann givið ringin til onkran [… så har man givet ringen til en eller anden] (Dan. wording)

(K8-corpus, Far. part)

a.' … so hevur mann givið onkrum ringin (Far. without convergence) ‘then somebody has given the ring to anybody’ b. [D1]: ich äh schreib über [sic.] mittelstufenfest also ich fehl nur noch eine linje oder so dann wär ich fertig; (cf. Dan. … jeg mangler kun en linje …, [my emphases, diverging from the original markings]) ‘I am writing about [the/our?] middle grade party, so I just need one line or so, then I would be ready’) (Westergaard 2008: 226; [6.11])

According to normative grammar which would favour (6a'), the Faroese example in (6a) shows a marked order in the verb phrase, which has obviously been triggered by the co-activated L2 language Danish, but which has become more and more usual in colloquial Faroese. The wrong linearisation in (6b) … ich fehle … is, however, grammatically worse but can be explained by an underlying switch to Danish, which becomes lexically overt due to the use of linje (‘line’) as part of a covert Dan. verbal frame: [jeg mangler en] linje. Here Danish has clearly been activated and becomes, in part, lexically overt because the German verb fehlen needs, contrary to Danish, a dative and not a nominative subject (mir fehlt). Thus, the default subject rule for both languages and the congruent lexicalisation principle override, in this case, a German morphosyntactic rule which would have held for Danish throughout: ‘to miss’ demands, like all other Dan. verbs, a nominative subject: jeg mangler (cf. also Far. eg mangli pening with nominative subject and strictly parallel to the Danish example vs. the same sentence in more ‘orthodox’ Far. mær vantar pening ‘I lack



Converging genetically related languages 

money’ [lit. me lacks money] with a dative subject. A non-nominative subject is, however, obligatory in Icelandic: mig [acc.] vantar pening). 2.5  On the rise of regional vernaculars and the process of standardisation The emergence of regional varieties, i.e. mixture between elements of the roofing (standard) language and a local dialect, shows the same principles of code mixing: one just inserts, overtly or covertly, lexical or morphosyntactic structures into the matrix language, which in most cases is the standard variety. Consider (7a) and (7b): (7) a. 

–Jeden Samstag um ¾8 hole ich frische Weckle/Semmeln vom Bäcker (Swabian/Bavarian vernacular) –Jeden Sonnabend um ¼ vor 8 hole ich frische Rundstücke/Schrippen vom Bäcker  (Hamburg/Berlin vernacular of German) ‘Every Saturday at 7:45 a.m. I fetch fresh rolls from the baker.’

b. Ich komme dann am Mittag zu dir und bring dir dann das Schwarzbrot vorbei. (Meaning in a southern German speech situation: ‘I will be there around 3 or 4 p.m. with a brown bread’; in northern German context, however: ‘I will be there around noon or 1 p.m. with a kind of pumpernickel’) → ‘I will come and see you at [??] bringing along a [??] bread’

It is well known that diasystems may have divergent lexical items for referring to the same objects, as demonstrated in (7a). In many cases, neutral terms exist in the roofing language; in example (7a) it would be 7 Uhr 45 ‘7:45’ and Brötchen ‘rolls’, respectively. But in some cases, one has no choice, as in the case of Saturday. In cases like (7a), a vernacular with overt code mixing between the standard and a regional variety, any native speaker of German will always know where the speaker comes from accommodating accordingly to his/her way of speaking, if this is an option which can be fulfilled. An utterance like (7b) is more complicated because the covert code mixing is difficult to detect if you are not familiar with the speaker’s background and his/her regional use of lexical items. In (7b), the correct referential semantic interpretation depends either on the region where the utterance occurred or on the speaker’s regional background. The ultimate aim of standardisation is thus (a) to reduce variation and (b) to establish uniformity in surface structure. One may therefore either select a certain dialect for this purpose or create it by making compromises with respect to grammar and lexicon. In the latter case, one may consider this kind of a standard language a compromise or a convergence language, i.e. a selection of lexical but also syntactic elements of divergent dialects where most other regional variants are no longer accepted. Standard languages are normally the result of language planning and

 Kurt Braunmüller

language regulation, if based on unified dialects, the motivation of which is to achieve a larger range in communication. Regional varieties/languages are, however, the result of individual and/or spontaneous convergence, a result of overt and/or covert code mixing due to accommodation to the addressee (cf. also Munske 1983).

3.  Conclusion: Outlining a code mixing hierarchy All these observations show that convergence between genetically closely related languages is the default case – no matter whether the converging items are overtly or covertly marked. A good starting point for linguistic convergence is a situation where two languages – for whatever reasons – come into close contact and are used in frequent alternation: A. The first step in a code mixing hierarchy is to make use of (even marked) patterns and lexical entries which enable congruent lexicalisation and a wide use of cognates (cf. e.g. the detailed graphical representations of utterances in Ridell 2008: 62–82). B. The next step would be to increase the frequency of such parallel entities in order to re-define the linguistic norms of the dominant language. One can also make use of more or less universal strategies when simplifying word order or over-generalising other grammatical patterns. This is often the case when a socially influential group consists of adult second language learners, as was the case in the era of the Hanseatic League (cf. Braunmüller 2006 for a discussion of some syntactic developments, especially the loss of V1-declarative sentences and the (almost) loss of the hinn/hin/hit-demonstrative pronouns in Mainland Scandinavia, accompanied by a clear preference for the other demonstrative pronouns and the preposed definite article þæn/þat [ðεn/ ðad], due to the phonetic similarity with Low Germ. de [dε; +def. article]). C. Especially in long(er)-term accommodation inflectional endings from the L1 (e.g. Swedish) are frequently preserved in attempts to accommodate lexically to the genetically very closely related neighbouring target language L2 (e.g. Danish). This has also been documented in Ridell (2008: 10, 11, 41 and passim). Similar code mixing phenomena have been reported by Nordenstam (1979) for long-term contacts of Swedes who have been living in the city of Bergen in Norway for years and who subconsciously accommodate to the local Norwegian variety without giving up their L1 Swedish. Both cases represent examples of step three in a code mixing hierarchy.



Converging genetically related languages 

D. Overt, but especially covert, code mixing is the most effective way to achieve social acceptance within a short period of time – if most speakers are bilingual. Bilinguals will understand and, in most cases, also tolerate any overgeneralised or incomplete acquisition of the grammar of the target language, because they always know what the speaker intended to say. Such a development can be observed when studying the decline of Low German and the learning of Standard Germany by native speakers of Low German. Some of the (formerly) intermediate dialects and results of code mixing had been called Missingsch. Its Hamburg variety is known as Paulunsch (for details see Paulun 1957), a kind of L2-German, showing many structural features of the former local Low German dialect. Petuh, once widely spoken in Flensburg, can also be regarded as a product of converging languages, i.e. Danish, the vernaculars South Jutish and/or Low German, and Standard German, where the German standard language was the target which was, however, never achieved (cf. Fredsted 2003 for the situation in Flensburg in the 19th century). Code mixing may result in new languages. Some of them may be classified as creoloids, double- (or triple)-source creoloids or jargons (cf. Trudgill 2000). DanoNorwegian and especially Afrikaans12 may be considered examples of creoloids, source language agentivity and even pidginisation (see Van Coetsem 2000: 73ff., 167ff.) within the Germanic language family. In any case, linguistic convergence between genetically closely related languages/varieties inevitably results, sooner or later, in overt or covert code mixing. Regional varieties can tell us how convergence works in individuals, whereas standardised and mixed languages in stable bilingual communities inform us about such a development on a societal scale. In any case, the endstation will always be code mixing (cf. Ross 1972) – and never a clear separation of genetically related varieties.

12.  So e.g. the classification in Trudgill (1992: 22). McWhorter (2007: 101) categorises Afrikaans as “semicreole”. Van Coetsem (2000: 79) gives more details: “Afrikaans is one of those multiple-source languages, but it unambiguously reveals the effects of S[ource] L[anguage] agentivity in its verbal morphology. We find here a striking example of reduction interacting with internally induced change, and, as we have noted, reduction is a (diachronic) diagnostic of SL agentivity. Reduction in Africaans is exemplified by virtually complete and comparatively abrupt abandonment of verbal ablaut; internally induced change, compensating for the loss of verbal ablaut, occurs with the generalization of the ‘weak preterite formation’. […] This strong reduction and the abrupt character of the development are revealing factors of pidginization. [my emphasis, K.B.]”

 Kurt Braunmüller

References Bot, K. de 2004. The multilingual lexicon: Modelling selection and control. International Journal of Multilingualism 1: 17–42. Braunmüller, K. 1990. Sprachkonflikte als Sprachnormenkonflikte (am Beispiel der interskandinavischen Semikommunikation). In Language Attitudes and Language Conflict/Spracheinstellungen und Sprachkonflikte, P.H. Nelde (ed.), 29–39. Bonn: Dümmler. Braunmüller, K. 1996. Sydslesvigdansk – et regionalsprog?. In Studier i talesprogsvariation og sprogkontakt. Til Inger Ejskjær på halvfjerdsårsdagen den 20. maj 1996, Institut for dansk Dialektforskning (ed.), 33–44. Copenhagen: Reitzel. Braunmüller, K. 2000. Direkter Transfer und strukturelle Adaption. Überlegungen zum Sprachwandel und zur Mehrsprachigkeit. In Wortschatz und Orthographie in Geschichte und Gegenwart. Festschrift für Horst Haider Munske zum 65. Geburtstag, M. Habermann et al. (eds), 315–332. Tübingen: Niemeyer. Braunmüller, K. 2001. Verdeckte Mehrsprachigkeit. In Vulpis Adolatio. Festschrift für Hubertus Menke zum 60. Geburtstag, R. Peters et al. (eds), 117–128. Heidelberg: Winter. Braunmüller, K. 2004. Niederdeutsch und Hochdeutsch im Kontakt mit den skandinavischen Sprachen. Eine Übersicht. In Deutsch im Kontakt mit germanischen Sprachen, H.H. Munske (ed.), 1–30. Tübingen: Niemeyer. Braunmüller, K. 2006. Wortstellung und Sprachkontakt: Untersuchungen zum Vorfeld und Nebensatz im älteren Dänischen und Schwedischen. In Amsterdamer Beiträge zur älteren Germanistik 62: 207–241. Christophersen, H. 1991. Sprog i Sydslesvig. Det danske sprog i Sydslesvig. Die gesprochene Umgangssprache in Norddeutschland. Det plattyske sprog, 2nd edn. Sorø: Rostras Forlag. Diercks, W. 1993. Zur Verwendung prä- und postmodifizierender Morpheme im Mittelniederdeutschen und in den skandinavischen Sprachen. In Niederdeutsch und die skandinavischen Sprachen I, K. Braunmüller & W. Diercks (eds), 161–194. Heidelberg: Winter. Diercks, W. & Braunmüller, K. 1993. Entwicklung des niederdeutsch-skandinavischen Sprachkontakts. Untersuchungen zur Transferenz anhand von volkssprachlichen Texten des 15., 16. und 17. Jahrhunderts – eine Projektübersicht. In Niederdeutsch und die skandinavischen Sprachen I, K. Braunmüller & W. Diercks (eds), 9–40. Heidelberg: Winter. Fredsted, E. 2003. Language contact and bilingualism in Flensburg in the middle of the 19th century. In Aspects of Multilingualism in European Language History, K. Braunmüller & G. Ferraresi (eds), 35–59. Amsterdam: John Benjamins. Fredsted, E. 2008. Converging verb phrases. Linguistics 46: 949–982. Grosjean, F. 1999. The bilingual’s language modes. In One Mind, Two Languages. Bilingual Language Processing, J. Nicol (ed.), 1–22. Oxford: Blackwell. Haugen, E. 1966. Language Contact and Language Planning: The Case of Modern Norwegian. Cambridge MA: Harvard University Press. Haugen, E. 1976. The Scandinavian Languages. An Introduction to Their History. London: Faber and Faber (revised German edition: Hamburg: Buske 1984). Heine, B. & Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: CUP. Höder, S. 2008. Sprachausbau im Sprachkontakt. Eine theoretische und empirische Analyse zum Syntaxwandel im Altschwedischen. Dr. phil. dissertation, University of Hamburg. Johanson, L. 2008. Remodeling grammar: Copying, conventionalization, grammaticalization.” In Perspectives on Language Contact, P. Siemund & N. Kintana (eds), 61–79. Amsterdam: John Benjamins.



Converging genetically related languages 

Kühl, K.H. 2008. Bilingualer Sprachgebrauch bei Jugendlichen im deutsch-dänischen Grenzland. Hamburg: Kovač. LePage, R. 1992. ‘You can never tell where a word comes from’: Language contact in a diffuse setting. In Language Contact. Theoretical and Empirical Studies, E.H. Jahr (ed.), 73–101. Berlin: de Gruyter. Lindow, W. 1984. Plattdeutsches Wörterbuch. Leer: Schuster. McWhorter, J. 2007. Language Interrupted. Signs of Non-Native Acquisition in Standard Language grammars. Oxford: OUP. Munske, H.H. 1983. Umgangssprache als Sprachkontakterscheinung. In Dialektologie. Ein Handbuch zur deutschen und allgemeinen Dialektologie, Vol. 2, W. Besch et al. (eds), 1002–1018. Berlin: de Gruyter. Muysken, P. 2000. Bilingual Speech. A Typology of Code-Mixing. Cambridge: CUP. Myers-Scotton, C. 2002. Contact Linguistics. Bilingual Encounters and Grammatical Outcomes. Oxford: OUP. Nordenstam, K. 1979. Svenskan i Norge. Språklig variation hos svenska invandrare i Bergen. [Nordistica Gothoburgensia 11]. Göteborg: University. Paulun, D. 1957. Missingsch. Studien in Hamburger Hochdeutsch. Hamburg. Petersen, H.P. 2006. Føroyskt-danskt málsamband: athall, tillaging, eingangstillaging og málbygging. In Fróðskaparrit 54: 8–20. Petersen, H.P. & Adams, J. 2008. Faroese. A Language Course for Beginners. Grammar. Tórshavn: Stiðin. Ridell, K. 2008. Dansk-svenska samtal i praktiken. Språklig interaktion och ackommodation mellan äldre och vårdpersonal i Öresundsregionen [Skrifter utgivna av Institutionen för nordiska språk 76]. Ph.D. dissertation, University of Uppsala. Ross, J.R. 1972. The category squish: Endstation Hauptwort. In Papers from the 8th Regional Meeting of the Chicago Linguistic Society, P.M. Peranteau, J.N. Leviand & G.C. Phares (eds), 316–328. Chicago IL: University of Chicago Press. Thomason, S.G. & Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley CA: University of California Press. Trudgill, P. 1992. Introducing Language and Society. London: Penguin. Trudgill, P. 2000. On locating the boundary between language contact and dialect contact – Low German and continental Scandinavian. In Språkkontakt – Innverknaden frå nedertysk på andre nordeuropeiske språk, E.H. Jahr (ed.), 71–85. Copenhagen: Nordic Council of Ministers. Van Coetsem, Frans. 2000. A General and Unified Theory of the Transmission Process in Language Contact. Heidelberg: Winter. Westergaard, A. 2008. Strukturelle und pragmatische Verwendungsmuster im bilingualen Sprachgebrauch bei Jugendlichen. Hamburg: Kovač.

part ii

Convergence and divergence in different varieties in oral and written discourse

Converging languages, diverging varieties Innovative relativisation patterns in Old Swedish* Steffen Höder

Hamburg, Collaborative Research Centre on Multilingualism In the late Middle Ages, when Old Swedish develops into a written language it acquires simultaneously several innovative syntactic features, such as new relativisation patterns. On the basis of an annotated digital corpus of Late Old Swedish texts, appositive relative clauses and the pronominal relativisation strategy are singled out as the typologically most salient innovations. In this contribution the author argues that the emergence of these features has to be explained as a grammatical replication of Latin features in a process of language Ausbau. Furthermore, it is argued that these changes affect only the emerging written variety of Old Swedish and mark the beginning of a medial split, with the written language converging towards Latin and diverging from the spoken varieties.

1.  Introduction This paper looks at how the development of Old Swedish1 into a literary language in the Late Middle Ages led to both typological convergence between languages and typological divergence between varieties of Old Swedish.2 The goal is to show

* This paper is based on a talk at the international colloquium on Convergence and Divergence in Language Contact Situations in Hamburg (Germany), 18–20 October 2007, and subsequent studies. It is part of the project Scandinavian Syntax in a Multilingual Setting at the Collaborative Research Centre on Multilingualism hosted at the University of Hamburg and funded by the German Research Foundation (DFG). I wish to thank the editors of this volume as well as two anonymous reviewers for their helpful comments. All remaining errors are mine. .  The onset of the Old Swedish period is usually defined by the earliest texts written in Latin script (c. 1225). Within Old Swedish, it is common to differentiate between Early Old Swedish (c. 1225–1375) and Late Old Swedish (c. 1375–1526). The year 1526, when the first translation of the New Testament into Swedish was printed and published, is meant to mark the beginning of the Protestant Reformation in Sweden. .  Convergence is usually understood as a type of contact-induced language change whereby certain structural (rather than simply lexical) features are transferred from one language into

 Steffen Höder

that the emergence of certain relative clause types and constructions should be explained as a result of language contact in a process of language Ausbau. I will argue that these syntactic changes affected only the emerging distinct written variety of Old Swedish and hence reflect the beginning of a medial split. The empirical basis of this study is a morpho-syntactically annotated digital corpus of Old ­Swedish texts which contains about 113,000 words (corresponding to about 14,500 clauses) and consists of eight Late Old Swedish texts and one Early Old Swedish text covering different text types (legal texts as well as religious and secular prose).3 This study is by no means the first to suggest that language contact might have influenced some of the changes in the domain of Old Swedish relativisation. One relative pronoun in particular, hviliken, has often been proposed as a syntactic loan from Latin (cf. Lindblad 1943: 134ff.; Wollin 1983: 134ff.). One goal of this paper is therefore to corroborate this hypothesis. Beyond that, it aims (a) to extend this claim beyond single relativisers to a specific set of relative clause types and relativisation strategies, and (b) to relate these changes to a wider range of syntactic changes induced by the establishment of Swedish as a medium of written communication under Latin influence. The structure of the paper is as follows: In Section 2, I discuss some general ideas about the importance of intralingual variation for diachronic studies. Section 3 provides background information about the development of Old Swedish into a written language and the sociolinguistic context in which this process takes place. Section 4 deals with innovative syntactic features and presents an explanation for their emergence. In Section 5 the results are summarised. 2.  Intralingual variation and the Uniformitarian Principle In the past, research on convergence and divergence phenomena has concentrated in particular on the role of such phenomena in the emergence of linguistic areas,

another, eventually resulting in these languages sharing an increasing number of similar grammatical structures. Divergence on the other hand is defined as a change in the opposite direction, leading to the structural similarities between two or more languages being reduced (cf. Weinreich 1954: 395). .  All texts have been digitalised and annotated according to the standards set by the ­Medieval Nordic Text Archive (www.menota.org), which in turn are based on the more general scheme defined by the Text Encoding Initiative (www.tei-c.org) widely used in corpus linguistics. The corpus used for this study is part of a larger corpus of Old East Nordic texts assembled within the project Scandinavian Syntax in a Multilingual Setting at the Collaborative Research Centre on Multilingualism in Hamburg.



Converging languages, diverging varieties 

which are mostly thought of as geographical entities consisting of entire languages.4 Consequently, typological studies have scarcely discussed intralingual – let alone medial – variation systematically, although cross-linguistically, written and spoken varieties are known to differ considerably with respect to their lexicon, morphology, and syntax (cf. Biber 1995; Koch & Oesterreicher 1994, 2007) and might play different roles in convergence and divergence processes as well as in the formation of linguistic areas.5 Recently, however, there has been an increasing interest in dialect studies and not least dialectal data in typological research (cf. the contributions in Kortmann 2004), which aims at a wider, more representative empirical basis that allows for generalisations over a sample of actually spoken varieties instead of a (convenience) sample of easily accessible grammars of written standard varieties. Additionally, some recent studies focus on convergence and divergence between varieties of the same language (cf. Hinskens, Auer & Kerswill 2005), and particular features found in standardised languages are discussed from a typological point of view by, for example, McWhorter (2007). Intralingual variation is also of great importance (although often neglected or underestimated) for historical linguistics in general. The Uniformitarian Principle (cf. Labov 1994: 21ff.) states that nothing that is now impossible can have been the case in the past and vice versa, or, as Lass (1997: 28f.) puts it, the ‘likelihood of any linguistic state of affairs (structure, inventory, process, etc.) has always been roughly the same as it is now.’ While this view is generally accepted, it is still necessary to emphasise that this principle, instead of being merely a truism, has some important implications for historical (or diachronic) linguistics. Language change is still often treated differently depending on whether it concerns languages in the remote (i.e. pre-modern) past or languages of today: Firstly, language change in the past is still often described and explained primarily in terms of intralinguistic developments affecting whole languages, whereas it is obvious that language change in the present is usually connected with intralingual, rather than only geographical, variation and is motivated by social factors. Secondly, while contemporary studies clearly show that language contact is an important motivating factor for language change, its role for the past is often considered marginal. This is mainly due to the .  In addition to the classic linguistic area in Europe, the Balkan Sprachbund, several other areas of different types and sizes have been proposed and discussed extensively in recent years, including the Mediterranean languages (Ramat & Stolz 2002), the Circum-Baltic languages (Dahl & Koptjevskaja-Tamm 2001) and Standard Average European (Haspelmath 2001). .  In Koch and Oesterreicher’s terms, the difference between spoken and written varieties is not primarily one of medium, but rather one of communicative distance: Written communication tends to be used in formal contexts, whereas spoken language is prototypically used in familiar contexts. Communicative distance constitutes a continuum where different communicative settings trigger the use of different varieties or, in polyglossic situations, different languages.

 Steffen Höder

fact that the lack of evidence for the influence of language contact is misinterpreted as evidence against its importance (cf. Milroy 2003). A more realistic approach, then, should be in accordance with the sociolinguistic and contact linguistic knowledge about modern languages. It should acknowledge, at least as a working hypothesis, that historical languages – like their modern offsprings – were characterised by intralingual variation and structured heterogeneity. Likewise, if language contact does matter in present-day situations, it should be expected to have done so in the past as well.6 Turning to Old Swedish, we can assume that this language was not a homogeneous monolithic entity, but rather a complex system consisting of different social, dialectal, situational and medial varieties which were not isolated from, but interacted with, each other and influenced one another quite substantially.7 Yet, historical corpora do not represent this complexity statistically, i.e. they do not contain texts in different varieties in proportion to their quantitative distribution in actual language use. In any corpus of Old Swedish, for instance, a. urban dialects are highly overrepresented in comparison to rural ones; b. there are more texts written by members of social elite groups than by speakers with a low socio-economic status; c. on the whole, spoken varieties are not extant (for obvious reasons). Basically, any corpus of Old Swedish texts can therefore be expected to be highly variety-specific, while the majority of Old Swedish varieties are not documented at all and therefore not directly accessible. This implies, then, that while any lexical or grammatical element found in the Old Swedish sources may be shared by all varieties of that language (or at least a majority), such elements may as well be particular features of the varieties that the sources represent. Consequently, we 6.  Moreover, language contact should be expected to exert its influence through similar mechanisms both in the past and in the present. As Braunmüller (2006: 18f.) phrases it, ‘[w]hen reading historical descriptions of certain forms of very intense language contacts, you may be confronted with the impression that some kind of dark cloud has moved across a linguistic area […] which drops a new, perhaps prestigious language like rain over the respective country, or that a huge wave has covered large parts of a territory the result of which is finally a more or less intensive change.’ In fact it is of course multilingual speakers’ creativity in language use that results in interlingual transfer and eventually can bring about contact-induced language change. It is therefore essential to take into account the social, cultural, and communicative context in which texts are written, copied, translated, distributed, and read, as well as the biographical background of the individual bi- or multilingual speakers behind the text production in the relevant period, in order to determine what kind of variety is actually represented in the written sources. 7.  Of course, the complexity of intralingual variation may be negligible in minor languages spoken by very small speaker groups.



Converging languages, diverging varieties 

should expect that the Old Swedish texts exhibit at least some specific features of an elaborated written variety that were absent in the everyday spoken language. 3.  Background: The emergence of Written Old Swedish 3.1  Old Swedish as a contact language Traditionally, the North Germanic languages are classified into two genetically determined groups, West Nordic (Norwegian, Icelandic, Faroese, Norn) and East Nordic (Swedish, Danish, Gutnish). This family tree model, however, suggests a clear-cut division within Old Nordic, which is greatly exaggerated, since North Germanic constituted a single dialect continuum up to the Middle Ages, and was not as isolated from West Germanic as it might seem from the genetic classification (cf. Harbert 2007: 6ff.).8 In contrast, Modern North Germanic can be synchronically divided into two groups, Insular Nordic (Icelandic, Faroese) and Continental Scandinavian (Swedish, Danish, Norwegian). This classification is essentially based on a number of salient typological differences such as the complex inflectional paradigms in Insular Nordic as opposed to the largely agglutinative morphology in Continental Scandinavian (cf. Fig. 1): North Germanic West Nordic Icelandic

Faroese

Insular Nordic

East Nordic Norwegian

Danish

Swedish

Continental Scandinavian

Figure 1.  Genetic and typological classification of the North Germanic languages

Such differences and divergences in the lexicon are usually explained by language change as a result of the intensive contact between Continental Scandinavian and other languages in the Late Middle Ages and in Early Modern Times (cf. Norde 1997). The language discussed here, Old Swedish, was in continuous contact with two major allochthonous languages from the European continent, which resulted from two important cultural developments, namely Christianisation and the

8.  This view is also supported by the fact that there are no documented glottonyms referring to any single Nordic language until the 14th century (Kjartan Ottósson 2002: 789f.).

 Steffen Höder

expansion of the Hanseatic League in the Baltic region. In the Late Middle Ages (13th–15th century), the Swedish-speaking area itself can be considered trilingual, with Old Swedish, Middle Low ­German and MedievalLatin in a triglossic distribution (cf. Braunmüller 2005; 2007):9 a. Swedish was used as a typical Low variety, primarily spoken in domains of everyday life; b. Latin was spoken as a lingua franca, and more importantly served as the written High variety in religious domains, administrative contexts, and secular literature; c. Low German served as an additional lingua franca within the Hanseatic area, but was also used for administrative purposes in the towns and spoken as the L1 of the influential German minority in Sweden. However, this trilingualism at the societal level does not imply that a large proportion of the population was in fact productively trilingual. On the contrary, productive and receptive competence varied depending on the geographical location as well as the social, economic and educational background of the speakers. 3.2 Language Ausbau Within the Old Swedish period, and within the trilingual situation described above, Swedish began to develop into a literary language, with text production written in the Latin alphabet starting in the early 13th century and replacing the older runic script used mainly for short inscriptions. Besides the application of the Latin script for writing in the vernacular, this development also includes a complex process of language Ausbau in the sense of Kloss (1978 [1952]). Kloss’s notion of Ausbau as a diachronic process is primarily based on his more widely known synchronic distinction between ‘Ausbau language’ and ‘Abstand language’ (language by development vs. language by distance). Kloss (1967: 29) defines Ausbau languages as languages that ‘are recognized as such because of having been shaped or reshaped […] in order to become a standardized tool of literary expression.’ As a general type of diachronic process, Ausbau can be defined as a continuous development resulting in a language gradually taking over domains of the written language in which it has not been used before. More

9.  I use the terms ‘diglossia’ and ‘triglossia’, in a wider sense than Ferguson (1959) and others, to describe a functionally complementary distribution of different languages or varieties regarding their communicative domains (cf. Kremnitz 2004). For a discussion of the history of the (socially determined) intralingual diglossia in older Swedish see Wollin (2003).



Converging languages, diverging varieties 

precisely, Ausbau is a process in which a group of literate and socially influential speakers increasingly produce texts of certain text types (relevant to them) in their L1, whereas they have previously used other languages or none at all for this purpose (Höder 2008: 88ff.). The result of such a process is the emergence of a distinct written variety of the language in question. In the Old Swedish case, the most influential group of authors, translators, scribes, and of course readers of Old Swedish literature in the Late Middle Ages comprised only a small minority of the entire population, namely primarily the clergy, many of whom were highly educated, literate and bilingual in Swedish and Latin (cf. Öberg 1994). The most important centre of Old Swedish literacy was undoubtedly Vadstena Abbey in central Sweden, established in 1384, where the majority of all Late Old Swedish texts were written. Since Written Old Swedish was decisively shaped by a rather small group of bilinguals, the Ausbau process itself as well as the emerging written variety were heavily influenced by Latin, as it constituted the established written variety and the prestigious model language. As a consequence, text production in Old Swedish concentrated on liturgical, monastic and other religious text types adapted from Latin-based models and was increasingly dominated by paraphrases and translations of Latin source texts (cf. Wollin 1991; Lönnroth 2005). Apart from translation interference, interlingual transfer from Latin is clearly recognisable in the language system of the new variety itself. Contact-induced phenomena range from lexical loans to morphological and syntactic innovations. Later on, the newly emerged Written Old Swedish – as represented in the religious texts from the 14th and 15th century – served as a model for the Reformation bible translations, which in turn had a normative impact on language standardisation from the 18th century onwards (Teleman 2002; 2003). Old Swedish Ausbau thus marks the beginning of a medial split between the spoken and the written varieties that still exists today, with Written Old Swedish gradually replacing Latin as the High variety and the grammatical differences between the spoken and the written language slowly increasing. The result was an intralingual diglossia between Spoken and Written Swedish, respectively, instead of the previous interlingual diglossia between Swedish and Latin (cf. Fig. 2). Although generally diverging from Spoken Swedish, Written Swedish was of course influenced by grammatical developments originating in the spoken language, but since literacy was restricted to a rather small proportion of the population, specific traits of the written variety generally did not spread into Spoken Swedish. Unlike morphology, the syntax of Written Swedish remained particularly conservative until the beginning of the 20th century, when standardisation efforts were extended to the syntactic level, and syntactic and textual norms were deliberately chosen to be based on the spoken language (Teleman 2003: 418ff.).

Spoken

Written

 Steffen Höder

interlingual diglossia

intralingual diglossia

Latin

Written Old Swedish

pre-Ausbau Old Swedish

Spoken Old Swedish

Written Modern Swedish

Spoken Modern Swedish

Figure 2.  Medial split in Old Swedish

On the basis of this scenario, an Old Swedish construction can be said to qualify as a contact-induced Ausbau phenomenon (Höder 2008: 136ff.) if it a. is an established feature of Written Old Swedish, i.e. it occurs regularly and frequently in the Old Swedish texts; b. is an innovation, i.e. did not exist in older stages of the language; c. is variety-specific, i.e. a feature only of the written variety (this can be determined indirectly by comparing the Old Swedish sources with written and spoken Modern Swedish, since, as noted above, specific features of the written variety generally did not affect the spoken variety). d. has structural and translational equivalents in Latin; e. is typologically rare (so that a coincidental parallel between Old Swedish and Latin can be ruled out). 4.  Relativisation patterns and syntactic Ausbau One fundamental cross-linguistic difference between spoken and written varieties concerns the way information is presented and organised in texts or utterances.­ Spoken language is predominantly used in contexts where all participants are familiar with each other and share a great amount of common knowledge that they can refer to implicitly when communicating. There is no need to verbalise all information that the speakers want to convey, and speakers need not mark different relations between utterances or sentences explicitly or unambiguously.



Converging languages, diverging varieties 

In written communication, however, the writer and the reader tend to not have as much knowledge in common (or at least cannot rely on having it). As a consequence, they have to make sure that all information they want to communicate is verbalised and that the relations between different sentences are made explicit. One result is an increased informational density and syntactic complexity in written­ texts.10 Some of the syntactic changes that are observable in the Old Swedish Ausbau are, as I have shown elsewhere (Höder 2008: 281ff.), at least partly traceable to an increasing need in written communication to distinguish explicitly different kinds of semantic relations between distinct bits of information, e.g. certain developments in the inventory of subjunctions, the syntactic integration of subordinate clauses in their matrix clauses, or the development of clause type-specific word order patterns. Relativisation constitutes a highly complex clause linking strategy, in which identical referents in two or more clauses are marked by a particular construction that concurrently subordinates a relative clause, in which the relativiser occurs, to a matrix clause containing the antecedent, if overtly expressed. Since relative clauses provide an additional structural possibility to increase informational density and syntactic complexity within the text, they can be regarded as likely candidates for Ausbau-induced changes in the language system. In the following sections I will argue that one particular relativisation strategy, viz. pronominal relativisation, in combination with one specific type of relative clauses, viz. appositive ones, has to be explained as a contact-induced innovation within the process of language Ausbau in Late Old Swedish. The transfer of this specific pattern from Latin to Old Swedish will be explained as an instance of grammatical replication within the framework developed by Heine and Kuteva (2003; 2005). 4.1  Pronominal relative clauses From a typological perspective, different structural types of relative clauses or constructional relativisation strategies can be distinguished, the most common of which (at least in the languages of Europe) include a. a subjunctional, b. a zero-marking and c. a pronominal strategy (Lehmann 1984: 43ff.; 1995; Zifonun 2001). All three strategies appear in Old Swedish as well.

10.  Of course, I am only referring here to one – albeit an important one – of many communicative differences between written and spoken language. For an extensive discussion of other differences see Koch and Oesterreicher (2007).

 Steffen Höder

In subjunctional relativisation, the relative clause is introduced by a subjunction similar to Modern Swedish som, English that or German (dialectal/regional) wo (in the following examples, coreferential elements are marked by a subscript i, relativisers are in italics):

(1) The woman speaks about [the man]i [thati we saw yesterday]

Such relative subjunctions generally do not specify which constituent of the matrix clause they refer to; the antecedent is only identifiable in a given semantic context.11 Similarly, the syntactic function of the relativiser within the relative clause can only be disambiguated by the semantic or grammatical context: Since the transitive verb saw in (1) lacks a direct object, the relativiser that has to be interpreted as its object. Relative subjunctions in Old Swedish include sum (> Modern Swedish­som) and þär, both of which can refer to any nominal or adverbial constituent in the matrix clause and can function as any type of constituent in the relative clause:

(2) ST, 2 (p. 78):

Pynaranei [somi them pynto] sagho … torturers-def  rel them tortured saw ‘the torturers who [refers to noun/functions as subject] tortured them saw …’ (3) ST, 2 (p. 88): … iærnkambai framlæggias [somi thera køth skulde mz slitas]   iron-combs forth-lay-pass  rel their meat should with tear-pass ‘… bring iron combs with which [noun/adverbial (prepositional phrase)] their meat should be torn’

(4) ST, 2 (p. 87):

… [then høxte oc waloghaste konunger]i [theri nokon    the highest and mightiest king  rel any tidh war fødder] time was born ‘… the highest and most powerful king that [noun phrase/subject] was ever born’

In zero-marked relative clauses, relativisation is expressed by the absence of any lexical subordinator:

(5) The woman talks about [the man]i [∅ we saw yesterday]

11.  It could be argued that since relative subjunctions do not indicate which constituent in the matrix clause they refer to, it is the relative clause as a whole – rather than the subjunction itself – that refers to the antecedent (cf. the alternative analysis for (1): The woman speaks about [the man]i [that we saw yesterday]i). The difference between these two analyses is, however, peripheral to the line argument proposed here.



Converging languages, diverging varieties 

Zero-marked relative clauses do not allow for an unambiguous identification of their antecedents in the matrix clause or the syntactic function of the antecedent within the subordinate clause, unless they are restricted to certain grammatical contexts, as is the case in, for example, Modern English or Modern Swedish, where zero-marking cannot be employed if the antecedent functions as subject in the relative clause (cf. *she talks about the man ∅ saw us yesterday). In Old Swedish, however, no such restriction exists:

(6) MU, 1.5 (p. 19):

… jnflyther j themi [∅ wan härra tagha]   in-flow in those   our lord take ‘… flows into those who [refers to pronoun/functions as subject] accept our Lord’

(7) MU, 1.8 (p. 23f.):

… [alth thz godha]i [∅ jak giordhe j minom mandom]   all that good    I did in my humanity ‘… all the good that [adjectival phrase/object] I did in my humanity’

(8) JP (p. 348):

[Varth härberge]i [∅ wi soffwom i ] …  our dwelling   we sleep in ‘our residence in which [noun phrase/adverbial (prepositional phrase)] we sleep…’

Pronominal relative clauses are marked by a clause-initial pro-form (either a relative pronoun comparable to Modern Swedish vilken, English who[m] or German der or a deictic adverb12 similar to Modern Swedish där, English where, German wo) referring to the antecedent in the matrix clause or an implicit antecedent:

(9) The woman talks about [the man]i [whomi we saw yesterday]

(10) … [at the station]i wherei we got off yesterday

Relative pronouns agree in some morphological categories (typically gender and number) with their antecedents, while other categories (often the case) are determined by a verb or preposition governing the relative pronoun in the relative clause. Inflected relative pronouns thus allow for an unambiguous identification of both the antecedent and the syntactic function of the relativiser in the relative clause. Relative adverbs belong to a specific semantic category and therefore also specify the possible semantic context of the antecedent, meaning that, for example, local relatives can only refer to local noun phrases or adverbials in the 12.  The Old Swedish corpus contains only relatively few adverbial relative clauses. These are counted among the pronominal relative clauses with which they share a relative pro-form as a common feature.

 Steffen Höder

matrix clause. Hence, pronominal relativisation constitutes the most explicit relativisation strategy, encoding more specific information about the relation between antecedent and relative clause than the others. Relative pronouns in the Old Swedish corpus include hviliken (> Modern Swedish vilken), þän, and hvar (only in free relative clauses). The most frequent relative adverbs are hvar ‘where’ and hvart ‘whither’. In (11), the relative pronoun hwlkin (a spelling variant of hviliken) agrees in gender and number with its antecedent kærlekin ‘the love’ (masculine singular), while the case (nominative) is determined by the syntactic function within the relative clause (subject). Likewise, hwlkith in (12) agrees in gender and number with klædith ‘the clothing’ (neuter singular), and the case (accusative) is governed by the preposition om ‘about’. The relative hwar in (13) corresponds to the local adverb ther ‘there’ in the matrix clause: (11) BU1, 7.5 (p. 149): … kærlekini / [hwlkini høxth ær j allom dygdom]   love-def  rprn-nom.sg.m highest is in all virtues ‘… love, which [refers to noun/functions as subject] is the highest among all the virtues’ (12) BU1, 7.13 (p. 177): … [thet klædith]i [om hwlkithi hans korsfæstare dobbladho]   that clothing-def  about rprn-acc.sg.n his crucifiers diced ‘… the garments for which [noun phrase/adverbial (prepositional phrase)] his crucifiers cast lots’ (13) BU1, 7.4 (p. 143): [hwari tith ligghiandhe fææ ær]/ theri ær tith hyærta  where your lying possession is there is your heart ‘where [adverb/adverbial] your treasure lies, there your heart is’

The additional hybrid relativisation strategy represents a combination of the pronominal and the subjunctional strategies in that both types of relativisers, inflected pronouns (or relative adverbs) and uninflected subjunctions, are used. In (14) and (15), the relative pronouns hwlke and the agree in number with their respective antecedents (mangha pawa ‘many popes’, alla ‘all’; gender is undistinguishable in plural forms), while the case is determined by the syntactic function in the relative clause (subject). In both cases, the relative subjunction som is redundant: (14) BU1, 7.7 (p. 160):

[mangha pawa]i waro før iohannez pawa  many popes were before John pope



[hwlke somi føro till hælwitis] rprn-nom.pl rel went to hell

‘… there were many popes before Pope John, who went to Hell’



Converging languages, diverging varieties 

(15) BU1, 7.8 (p. 162): … allai [the somi sægya pawan ey vara sannan pawa]   all  rprn-nom.pl rel say pope-def not be true pope ‘… all those who say that the pope is not the true pope’

A quantitative analysis of the corpus reveals that all relativisation strategies are established in the Late Old Swedish texts (cf. Tab. 1). Note that the numbers for the pronominal and hybrid relativisation strategies include all relative clauses that are potentially interpretable as pronominal or hybrid; hence, they also include critical contexts (revisited in 4.3). Table 1.  Relativisation strategies in Late Old Swedish texts

PP BA BU1 BU2 MU ST HA+JP total

subjunctional

zero-marked

pronominal

hybrid

n

74.5% 38.9% 39.1% 43.2% 36.5% 43.9% 55.6% 50.1%

  5.1% 16.7%   3.8%   1.6%   4.1% 15.9% 15.6%   6.1%

12.0% 38.9% 40.3% 28.0% 40.0% 32.1% 13.3% 27.6%

  8.4%   5.6% 16.8% 27.3% 19.4%   8.1% 15.6% 16.1%

  510    36   340   579   170   358    45 2038

Although subjunctional relativisation can be identified as the most and zeromarking as the least frequent option in an overall rank order (zero-marked < hybrid < pronominal < subjunctional), this general tendency does not necessarily apply to the individual texts. In BA, BU1 and MU, the percentage of pronominal relative clauses is about 40%, i.e. pronominal and subjunctional relative clauses are approximately equally frequent. In BU2 on the other hand, the proportion of hybrid relative clauses is remarkably high (27.3%, i.e. they are almost as frequent as the pronominal ones with 28.0%). Strikingly, the earliest text in the corpus (PP, dated between 1300 and 1350),13 has the highest proportion of subjunctional relative clauses (74.5%) and the lowest

13.  The digital version of the Pentateukparafrasen (PP) used for this study is based on a manuscript from the first third of the 15th century (Cod. Thott 4), according to the edition by Klemming (1848–1855). Another manuscript (Cod. Holm. A 1, edited by Thorell 1959), which dates from 1526 but is nevertheless considered to resemble the older original text more closely and would thus be a more appropriate representative of Early Old Swedish religious prose, is currently being annotated and added to the corpus at the Collaborative Research Centre on Multilingualism in Hamburg. A quantitative evaluation of a sample (ca. 10% of the whole text) yielded no significant difference between the two manuscripts with respect to the

 Steffen Höder

proportion of pronominal relative clauses (12.0%). This finding indicates that pronominal relativisation may be – or at least may relate to – an innovative pattern in (Late) Old Swedish. An analysis of the Early Old Swedish text (UL) however, provides no further evidence to support this claim, since subjunctional, zero-marked and pronominal relative clauses are approximately equally frequent here (34.4% vs. 27.3% vs. 26.6%; hybrid relative clauses make up 11.4%; n=297). Therefore, pronominal relativisation in itself cannot be shown to be an innovative pattern in Old Swedish. However, the picture becomes clearer if we also take different relative clause types into account. 4.2  Appositive relative clauses According to the semantic and informational relation to their antecedents, relative clauses can be classified into three functional types (following Lehmann 1984): a. restrictive relative clauses, which restrict the possible referents of the antecedent (‘she speaks about exactly that man that we saw yesterday’): (16) The woman talks about [the man]i [thati we saw yesterday]

b. free relative clauses, which do not have an explicit antecedent (‘we can sell the things that we have’): (17) We can sell [whati we have]

c. appositive relative clauses, which have an explicit antecedent but add new information about a given referent rather than restricting a set of possible referents: (18) The boy talks about [his father]i [whomi we met at the station]

In appositive relative clauses, relativisation can be regarded as an alternative strategy of expressing (roughly) the same meaning that can also be expressed in paratactic sequences like the ones in (19) or (20), where coreference is marked by repeating the noun phrase or using an anaphoric pronoun: (19) The boy talks about [his father]i. We met [his father]i at the station. (20) The boy talks about [his father]i. We met himi at the station.

The main difference between those competing strategies is the fact that the second clause in (18) is subordinated to, and syntactically integrated in, the first one. Within this sequence, relativisation is therefore used exclusively to establish and relativisation strategies and clause types investigated in this study. The numbers given for PP can therefore be considered to be representative for both versions.



Converging languages, diverging varieties 

formally mark a specific clause linkage type. Consequently, appositive relative clauses are predominantly used cross-linguistically in written language, while in spoken language the alternative strategies are more common. If their emergence is due to the Ausbau process, this should be observable in a corpus-based study of texts from the relevant period. A quantitative analysis of the corpus reveals that all three relative clause types, as illustrated in (21)–(23), are established in the Late Old Swedish texts (cf. Tab. 2): (21) BU1, 7.13 (p. 177): … [the thymana]i [i hwlkai han syndadhe]    those hours-def  in rprn he sinned ‘… the hours in which he sinned [restrictive relative clause]’ (22) BU2, 4.47 (p. 89): [huilkin somi hona faa] han ma vara sua oräddir suasom …  rprn rel her gets he may be so unafraid so-as ‘he who obtains my mercy [free relative clause] may feel as secure as …’ (23) ST, 2 (p. 88): Jak ær christus [thin konunger]i [somi thu thiænar] I am Christ  your king  rel you serve ‘I am Christ, your king, whom you serve [appositive relative clause]’ Table 2.  Relative clause types in Late Old Swedish texts

PP BA BU1 BU2 MU ST HA+JP total

restrictive

free

appositive

n

59.8% 41.7% 45.3% 48.4% 48.8% 45.5% 46.7% 50.1%

11.6% 11.1%   7.9% 12.1% 15.9% 15.6% 17.8% 12.3%

28.6% 47.2% 46.8% 39.6% 35.3% 38.8% 35.6% 37.6%

  510    36   340   579   170   358    45 2038

The overall rank order of the three relative clause types (free < appositive < restrictive) does not apply to each text (appositive relative clauses are more frequent than restrictive ones in two of the texts). Strikingly, the percentage of appositive relative clauses is again considerably lower in the earliest text (PP) than in the other sources, indicating that the use of appositive relative clauses has increased in the (Late) Old Swedish period and reflects an innovative pattern

 Steffen Höder

in Old Swedish. An analysis of the Early Old Swedish text (UL) provides further evidence for this claim (72.4% restrictive, 3.7% appositive, 23.9% free relative clauses; n=297). A combined analysis of relative clause types and relativisation strategies in the Late Old Swedish corpus yields illuminating results (Tab. 3): Table 3.  Combined analysis: Relative clause types and relativisation strategies

restrictive free appositive total

subjunctional

zero-marked

pronominal

hybrid

n

67.2%   1.6% 43.2% 50.1%

11.0%   0.0%   1.7%   6.1%

14.0% 40.6% 41.5% 27.6%

  7.8% 57.8% 13.6% 16.1%

1021   251   766 2038

This evidence indicates that different relativisation strategies are preferred depending on different relative clause types: Subjunctional relativisation prevails in restrictive relative clauses, whereas pronominal and hybrid relativisation are the only options used frequently in free relative clauses. In appositive relative clauses, subjunctional and pronominal as well as hybrid relativisation occurs. Zero-marking is possible almost exclusively in restrictive contexts. These findings suggest a functional difference between the different relativisation strategies. This difference becomes even clearer if we apply a simpler classification. Both pronominal and hybrid relativisation employ a pronominal element and can thus be counted as [+pronominal], as opposed to subjunctional relativisation and zero-marking which are [–pronominal] strategies. Applying this classification gives the distribution in Tab. 4, which can be summarised as follows: a. Restrictive relative clauses are predominantly constructed using [–pronominal] strategies; b. free relative clauses are always [+pronominal]; c. appositive relative clauses can be either [–pronominal] or [+pronominal]. Table 4.  Combined analysis: Relative clause types and [±pronominal] strategies

restrictive free appositive total

–pronominal

+pronominal

n

78.2%   1.6% 44.9% 56.2%

21.9% 98.4% 55.1% 43.8%

1021   251   766 2038



Converging languages, diverging varieties 

The figures are quite similar for the Early Old Swedish text in the corpus (UL): Restrictive relative clauses are 81.4% [–pronominal] vs. 18.6% [+pronominal], free relative clauses are 4.2% [–pronominal] vs. 95.8% [+pronominal], and appositive relative clauses are 54.5% [–pronominal] vs. 45.5% [+pronominal]. In Tab. 3 and 4 the focus was on the percentage of different relativisation strategies in relation to the number of relative clauses of the respective type. If we now look at the token frequency of the different clause type/strategy combinations and compare Early and Late Old Swedish, we find that these are distributed as follows (Tab. 5): Table 5.  Token frequency of clause type/strategy combinations Late Old Swedish

restrictive free appositive total

Early Old Swedish

–pronominal

+pronominal

–pronominal

+pronominal

39.2%   0.2% 16.9% 56.2%

10.9% 12.1% 20.7% 43.8%

58.9%   1.0%   2.0% 61.9%

13.5% 22.9%   1.7% 38.1%

Two prototypical combinations are identifiable in the Early Old Swedish text, viz. [–pronominal] restrictive relative clauses and [+pronominal] free relative clauses, which in sum make up 81.8% of all relative clauses. In Late Old Swedish, in contrast, these prototypical combinations only add up to 51.3%. This difference is largely due to the high proportion of appositive relative clauses in Late Old Swedish, which do not show a clear preference for a specific relativisation strategy. Hence, the use of appositive relative clauses does not only seem to constitute an innovative feature in Late Old Swedish, it also leads to the diminishing prominence of prototypical clause type/strategy combinations. 4.3  Ausbau phenomenon The claim that Old Swedish pronominal relativisation in non-prototypical contexts (i.e. primarily appositive relative clauses) is an Ausbau-related innovation is supported by a qualitative analysis of the development and history of the pronominal relativisers. All pronominal or adverbial relativisers occurring in the Old Swedish texts are homophonous with other pro-forms. Firstly, the relative pronoun þän is homophonous with the demonstrative þän ‘that’ (> Modern Swedish den); the local relative adverb þär is homophonous with the demonstrative þär ‘there’ (> Modern Swedish där). This homophony indicates that the pronominal (and adverbial) relativisers

 Steffen Höder

have emerged through the grammaticalisation of demonstratives via ambiguous or critical contexts (cf. Diewald 2002: 109ff.) such as the ones in (24) and (25): (24) BU1, 7.13 (p. 179):

Thei ærw mangh otalik thwsandh / [the somi alle ærw they are many countless thousand rprn rel all are  

j myne twngo scriffwadha] in my tongue written ‘there are many thousand [sins], all of which are written on my tongue’ (25) BU1, 7.13 (p. 183): iach mynnes nw inthei [theti han æpter myn vilia giordhe] I remember now nothing  rprn he after my will did ‘now I do not recall anything that he did after my wishes’

In (24), the sequence the som can either be analysed as a hybrid relativiser (relative pronoun + relative subjunction) or as a right-dislocated demonstrative followed by a relative subjunction (‘there are many thousand sins, those that are all written on my tongue’). Similarly, sequences like the one in (25) can either be interpreted as a relative pronoun or as a demonstrative followed by a zero-marked relative clause (‘I recall nothing, [no] things ∅ he did after my wishes’). In such cases, the ambiguity depends on the case of the pronominal element agreeing with the function of its referent both in the matrix clause and in the relative clause. Secondly, the relative pronouns hviliken and hvar are homophonous with the interrogative forms hviliken ‘which’ (> Modern Swedish vilken) and hvar ‘who’ (and suppletive forms, cf. Modern Swedish vem); the local relative adverbs hvar (static) and hvart (directional) are homophonous with the interrogative adverbs hvar ‘where’ and hvart ‘whither’ (> Modern Swedish var, vart). However, these deinterrogative relativisers cannot have emerged in similarly ambiguous contexts as illustrated above for the de-demonstrative ones. Interrogatives may instead have grammaticalised to relativisers in ambiguous contexts as in (26): (26) PP, 27 (p. 212): komber honum wäl i hugh [hwathi gudh hawer iättath] comes him well in mind  rprn god has promised ‘he recalls well what God has promised’

In this sequence the pronoun hwath (after a verbum dicendi) is followed by a clause that can be analysed both as an interrogative and a free relative clause (‘he recalls the answer to the question as to what God has promised’ vs. ‘he recalls the things that God has promised’; cf. also Lehmann 1984: 325ff.; Heine & Kuteva 2002: 251). In the Old Swedish corpus, however, ambiguous contexts similar to the one in (26) are quite rare and only involve free relative clauses, whereas restrictive or appositive relative clauses with de-interrogative relativisers cannot be interpreted as interrogative clauses.



Converging languages, diverging varieties 

The transparent etymology of the pronominal relativisers, their homophony with other pro-forms, and the occurrence of ambiguous or critical contexts seem to be indicative of a recent grammaticalisation process in Old Swedish. This view is supported by the fact that the earliest Swedish manuscripts as well as the Swedish runic inscriptions (in the so-called Younger Fuþark) do not use [+pronominal] relativisation strategies; the inventory of relativisers consists only of relative subjunctions such as sum, þär, es, är and än.14 The older runic inscriptions in Ancient Nordic (before 800 AD) contain only two relative clauses that can neither be classified as [+pronominal] nor as [–pronominal]. These two free relative clauses are found in two near-identical and obviously formulaic inscriptions (from Stentoften and Björketorp, cf. Düwel 2001: 42ff.); they read sa[r] þat bar[i]ut[i]þ ‘who breaks this’. The clause-initial sa[r] can be analysed either as a demonstrative pronoun (sa) followed by a relative clitic (-r) or as a relative pronoun. Even in Old Norse, relative clauses are normally non-pronominal (cf. Nygaard 1905: 256ff. [§ 254ff.]; Faarlund 2004: 259ff.). The most frequent relative subjunctions are sem and er, as illustrated in (27): (27) Njáls saga 29.3 (Faarlund 2004: 259): er [sú kona]i illa gift, [eri þú átt] is  this woman bad married  rel you own ‘this woman that you have made a bad marriage’

Interestingly however, relative clauses that resemble hybrid relativisation strategies with demonstrative or interrogative pronouns do occur in Old Norse, but represent an infrequent, stylistically marked alternative construction and are restricted to certain text types (cf. Nygaard 1905: 263f. [§ 264]). Such patterns are shown in (28) and (29): (28) Gamal norsk homiliebok 5.4 (Faarlund 2004: 264): … með illgjǫrnumi … [þeir eri frið hafa …]   with evil-minded …  those rel peace have ‘…with the evil-minded, who have peace …’ (29) Diplomatarium Norvegicum (Faarlund 2004: 265): … [tvau bref]i … [í hverjum eri konungrin býðr …   two letters   in which rel king-def orders ‘… two letters in which the king orders …’

Even though such constructions could qualify as (predecessors of) [+pronominal] relative clauses, they are not fully developed in Old Norse and most certainly not established in the language as a whole, but only as a specific feature of some text types. 14.  For an overview of relative clauses and relativisers in Old Swedish cf. the grammars by Wessén (1956: 227ff.), Noreen (1904: 409ff.) and de Boor (1922: 122ff.) as well as the classic studies by Lindblad (1943) and Wenning (1941).

 Steffen Höder

As for Modern Swedish, both [+pronominal] and [–pronominal] relativisation strategies do occur, but have different functions. On the whole, [+pronominal] constructions can be used in free relative clauses in all varieties: (30) … den rike borde sälja [vadi han har]   the rich should sell  rprn he has ‘… the rich man should sell what he has [free relative clause]’

In Spoken Modern Swedish, however, restrictive and appositive relative clauses are generally constructed non-pronominally, i.e. with a relative subjunction (som) or without a relativiser (exclusively in restrictive clauses, cf. Stroh-Wollin 2002: 53ff.). (31) Matcheni [∅ vi kollade på] gick bra. match-def    we watched on went good ‘The match that we watched ended well [restrictive relative clause]’ (32) Han lyckades skjuta bolleni [somi var strax bakom honom] he managed shoot ball-def  rel was right behind him ‘He managed to shoot the ball that was right behind him [appositive relative clause]’

In Written Modern Swedish, on the other hand, [+pronominal] constructions are used in such contexts as well, particularly in older literary texts and in conservative text types such as administrative and religious texts (cf. Pettersson 1976). In (33), the relative pronoun vilka agrees in number (plural) with its antecedent (två resolutioner ‘two resolutions’); in (34), vilken agrees in gender and number (common gender, singular) with its antecedent (den riktning ‘the direction’): (33) Administrative text:15

Parlamentet uppmanar i [två resolutioner …]i [vilkai parliament-def demands in  two resolutions  rprn



antogs den 15 maj 1997…] adopted-pass the 15 May 1997

‘In two resolutions …, which were adopted on May 15, 1997, the [European] Parliament demands … [appositive relative clause]’ (34) Religious text:16 … [den riktning]i [i vilkeni vi skall tänka oss Gud …]   the direction  in rprn we shall think us God ‘… the way in which we shall conceive of God … [restrictive relative clause]’

15.  Source: www.europarl.europa.eu/factsheets/4_1_7_sv.htm, 20 October 2008. 16.  Source: www.katolik.nu/html/artcl_treen.htm, 20 October 2008.



Converging languages, diverging varieties 

In addition, Modern Swedish has two special relativisers that are used to refer to a whole clause or verb phrase (vilket) and to a possessor expressed in the matrix clause (vars), respectively. Despite the fact that both historically derive from inflected pronominal forms, they are uninflected in Modern Swedish and hence do not qualify as pronouns. Vilket and vars are used in written and spoken language. 4.4  Contact-induced innovation If we assume that [+pronominal] relativisation strategies in non-prototypical contexts are an innovation that came about during the Old Swedish Ausbau process and that they affected only the emerging written variety, the question remains whether or not this innovation was induced by language contact with Latin. If it was, then Latin should possess an equivalent construction that was transferred into Swedish by bilinguals, and the way in which the transfer proceeded should be explained. Indeed, although (Medieval) Latin possessed a large inventory of relativisers, all of them represent the same relativisation strategy, viz. pronominal relativisation. The most frequent and semantically least specific relativiser was the pronoun qui ‘who, which’. Besides qui, there were various morphologically complex pronouns like quicumque and quisquis ‘whoever, whichever’ which occur in generalising contexts, i.e. free relative clauses, and such pronouns as quantus ‘as much as’ or qualis ‘of the kind that,’ which functioned as relativisers as well. The most frequent adverbial relativisers were the local adverbs ubi ‘where’, quo ‘whither’ and unde ‘whence’. Moreover, appositive relative clauses were a firmly established and frequent clause type in (Medieval) Latin. A comparative analysis of Latin originals and their Old Swedish translations shows that appositive relative clauses in Latin tended to be rendered as relative constructions in Old Swedish as well, and that pronominal relativisation in Latin was frequently translated with a [+pronominal] relative clause in Old Swedish.17 This is illustrated in (35) as compared to (36): (35) St. Birgitta’s revelations, 7.30 (Bergh 1967: 210): Et [quedam virgo]i stabat ante sedentem in throno [quei erat fulgencior sole]. [Quami omnes illi assistentes celestis milicie honorabant reuerenter vt reginam celorum.

17.  In BU1 (13,793 words, 1,659 clauses), for instance, I found that 99 appositive relative clauses in the Old Swedish text were formally equivalent to the Latin source with its Latin equivalent (cf. Bergh 1967); 11 of those were constructed non-pronominally.

 Steffen Höder

(36) BU1, 7.30 (p. 222f.): ok [en iomffrw]i stodh før hanom som saath aa stolenom [hwilkin somi war meer skynandhe æn solen] [hwlkai all the hymmilska hærskapin ther till standandhe hedradhe mædh wørdningh swa som hymmeriikis drøthningh] ‘And a virgin stood before him who sat on the throne, who [= the virgin] was more radiant than the sun, whom [= the virgin] all those of the heavenly host who stood nearby reverently honoured as the Queen of Heaven’18

Latin thus qualifies as a possible source for the innovative relativisation pattern in Old Swedish, because a. both the relative clause type and the relativisation strategy in question are established in the language, and b. Old Swedish translators obviously viewed and treated Latin and Old Swedish [+pronominal] appositive relative clauses as translational equivalents. If it is true that the innovative relativisation patterns are the result of interlingual transfer from Latin, then this transfer represents a specific type of what Heine and Kuteva (2003; 2005) label as grammatical replication, viz. contact-induced grammaticalisation. Grammatical replication is basically defined as the transfer of grammatical meanings when they are not parts of form-meaning units (Heine & Kuteva 2005: 2). Contact-induced grammaticalisation involves a three-step process (Heine & Kuteva 2005: 80ff.): a. Bilingual speakers notice that in language M (the model language) there is a grammatical category Mx that language R (the replica language) is lacking; b. they create an equivalent category Rx in R on the basis of an available category, Ry, drawing on universal grammaticalisation strategies (or paths); c. Ry grammaticalises to Rx. A process of this type apparently lies behind the emergence of [+pronominal] strategies in Old Swedish, and can be described as follows: Latin-Swedish bilinguals realised that Latin (the model language) had a range of [+pronominal] relative­constructions that could be used in all relative clause types (qui, ubi etc.). They then created an equivalent category in Old Swedish (the replica language) by conventionalising existing use patterns which involved existing pronouns or adverbs (þän, þär etc.). These pro-forms grammaticalised to [+pronominal] relativisers.­ Intermediate

18.  Note that (36) contains also a (prototypical) non-pronominal restrictive relative clause (hanomi [somi saath aa stolenom] ‘him who sat on the throne’) which translates a Latin participial clause (sedentem in throno ‘[the one] sitting on the throne’).



Converging languages, diverging varieties 

steps are still reflected in critical contexts such as the ones in (24) and (25), where pronominal elements can either be analysed as relatives or as demonstratives. While such a scenario can adequately account for the emergence of de-­ demonstrative relativisers, the grammaticalisation of interrogatives as relativisers – particularly the most frequent relative pronoun hviliken – seems to involve the mechanism called replica grammaticalisation by Heine and Kuteva (2005: 92ff.). Replica grammaticalisation is a special type of contact-induced grammaticalisation­ in which speakers not only replicate a particular grammatical category in the replica language, but even the grammaticalisation process itself that they assume to have taken place in the model language. In this case, a. bilingual speakers noticed that Latin had a relative pronoun, qui, which was homophonous and etymologically identical to an interrogative determiner qui (‘which’);19 b. they identified (or established) interrogative qui as the interlingual equivalent of Old Swedish hviliken; c. they then replicated the (assumed) grammaticalisation process (interrogative > relative) in Old Swedish; d. they started using hviliken as a translational equivalent of Latin qui in relative clauses; e. hviliken eventually grammaticalised as a relative pronoun in Written Old Swedish. Compared to ordinary contact-induced grammaticalisation, replica grammaticalisation represents a more creative linguistic behaviour that requires a higher degree of receptive and productive bilingualism (cf. Heine & Kuteva 2005: 34ff.). This difference explains why relative hviliken, unlike the de-demonstrative relativisers, is totally lacking in the earliest texts analysed in this study (UL and PP) and first appears in the Old Swedish texts as late as the end of the 14th century: Relative hviliken entered the grammar of Written Old Swedish at a time when the text production was dominated by the monasteries and, hence, by writers who were educated and highly proficient in Latin, and who were, furthermore, experienced translators routinely using both languages interchangeably. The grammaticalisation of demonstratives and interrogatives to relativisers might in principle – as suggested by e.g. Pettersson (1976: 348) – be entirely contact-­independent and language-internal (or variety-internal, as it were). Indeed, this grammaticalisation path is recognised in other languages as a potential way in

19.  Note that the Latin interrogative determiner qui ‘which’ is functionally and phonetically distinct from the interrogative pronouns quis ‘who’ and quid ‘what’, although it shares some inflectional forms (e.g. the genitive cuius) with these elements.

 Steffen Höder

which pronominal relatives evolve (cf. Heine & Kuteva 2002: 113ff., 251). However, though possible, such a coincidence is highly unlikely, since [+pronominal] relativisation strategies (except in free relative clauses) are typologically rare and cluster mainly in European languages, where their occurrence does not derive from Proto-Indo-European,20 but appears to be due to later developments (Comrie­ 1998; 2002; Comrie & Kuteva 2008a,b; Haspelmath 2001: 1494f.): Although the earliest attested Indo-European languages in Europe had relative pronouns (such as Ancient Greek ὅς), most of them lost them later on in favour of uninflected relativisers (cf. Modern Greek που) and subsequently, additional redeveloped pronouns (cf. Modern Greek ο οποιος).

5.  Conclusion This study suggests that the development of [+pronominal] relativisation patterns in non-prototypical (i.e. restrictive and especially appositive) contexts represents a contact-induced typological change as a part of the language Ausbau in Old Swedish. The grammaticalisation processes behind this innovation can be analysed as instances of grammatical replication. Combined with other contact-induced syntactic innovations related to the dominance of Latin in the Ausbau process, the innovative relativisation patterns lead to an increasing informational density and syntactic complexity of written Swedish. They also contributed to an increasing structural difference between written and spoken Swedish, resulting in a medial split between spoken and written varieties of Swedish, whereby the structural properties of Latin relativisation strategies and relativisers were transferred into the emerging written variety. Thus, the intralingual divergence within Old Swedish was associated with a variety-specific interlingual convergence towards structural similarity – i.e. isomorphic and translationally equivalent constructions – between Written Old Swedish and Latin. Finally, the long-term result was that Written (Old) Swedish shared a certain typological feature, [+pronominal] relativisation, with the Standard Average European languages (cf. Haspelmath 2001), whereas Spoken (Old) Swedish did not. This observation makes it plain that the literal view of linguistic areas as flat, geographical areas is somewhat simplistic. Rather, linguistic areas such as Standard Average European have to be regarded, or redefined, as coherent regions within a multilingual and multidimensional communicative space, which may

20.  For a detailed discussion of relative clauses in the older Indo-European languages cf. Kurzová (1981).



Converging languages, diverging varieties 

also cross language boundaries and consist of non-geographical – i.e. social or medial – varieties of different languages.

Glosses acc def m n nom pass pl rel rprn sg

accusative definite masculine neuter nominative passive plural relative subjunction relative pronoun singular

Sources BA = Birgitta-autografen A. 1951. In Heliga Birgittas originaltexter [Samlingar utgivna av Svenska fornskriftsällskapet 1.58], B. Högman (ed.), 73–78. Uppsala: Almqvist & Wiksell. BU1 = Birgittas uppenbarelser, 1:a redaktionen. 2000. In Heliga Birgittas uppenbarelser bok 7 efter Cod. Ups. C 61. Diplomatarisk utgåva med kommenterande inledning [Samlingar utgivna av Svenska fornskriftsällskapet 1.84], I. Lindell (ed.). Uppsala: Svenska fornskriftsällskapet [book 7]. BU2 = Birgittas uppenbarelser, 2:a redaktionen. 1860. In Heliga Birgittas uppenbarelser. Efter gamla handskrifter, Vol. 2 [Samlingar utgivna av Svenska fornskriftsällskapet 1.14.2], G.E. Klemming (ed.). Stockholm: Norstedt [book 4, chapters 1–20 and 40–60]. HA = Herr Abboten. 1887–1889. In Prosadikter från Sveriges medeltid [Samlingar utgivna av Svenska fornskriftsällskapet 1.28], G.E. Klemming (ed.), 351–355. Stockholm: Norstedt. JP = Aff Joan prest aff India land. In Prosadikter från Sveriges medeltid [Samlingar utgivna av Svenska fornskriftsällskapet 1.28], G.E. Klemming (ed.), 343–350. Stockholm: Norstedt. MU = Mechtilds uppenbarelser. In Hel. Mechtilds uppenbarelser (Liber spiritualis gratiæ) öfversatta från latinet af Jöns Budde. Efter gamla handskrifter [Samlingar utgivna av Svenska fornskriftsällskapet 1.32], R. Geete (ed.). Stockholm: Norstedt [chapters 1–20]. PP = Pentateukparafrasen. 1848–1855. In Svenska medeltidens bibel-arbeten, Vol. 1 [Samlingar utgivna av Svenska fornskriftsällskapet 1.9.1], G.E. Klemming (ed.). Stockholm: Norstedt [Book of Genesis]. ST = Själens tröst. 1954. In Siælinna Thrøst. Første delin aff the bokinne / som kallas siælinna thrøst. Efter cod. Holm. A 108 (f.d. cod. Ängsö). Kritisk upplaga [Samlingar utgivna av Svenska fornskriftsällskapet 1.59], S. Henning (ed.). Uppsala: Almqvist & Wiksell [Introduction and Second Commandment]. UL = Upplandslagen. 1834. In Uplands-Lagen [Samling af Sweriges gamla lagar 3], C.J. Schlyter (ed.). Stockholm: Norstedt [Ärvdabalken and Manhelgdsbalken].

 Steffen Höder

References Bergh, B. (ed.). 1967. Den Heliga Birgittas Revelaciones. Bok VII [Samlingar utgivna av Svenska fornskriftsällskapet 2.7.7]. Uppsala: Almqvist & Wiksell. Biber, D. 1995. Dimensions of Register Variation. A Cross-linguistic Comparison. Cambridge: CUP. de Boor, H. 1922. Studien zur altschwedischen Syntax in den ältesten Gesetztexten und Urkunden [Germanistische Abhandlungen 55]. Breslau: Marcus. Braunmüller, K. 2005. Language contacts in the Late Middle Ages and in Early Modern Times. In The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol. 2 [Handbooks of Linguistics and Communication Science 22.2], O. Bandle et al. (eds), 1222–1233. Berlin: de Gruyter. Braunmüller, K. 2006. Dialects, varieties and historical linguistics. In Nordisk dialektologi og sociolingvistik. Foredrag på 8. Nordiske Dialektologkonference, Århus 15.–18. august 2006, T. Arboe (ed.), 11–32. Århus: Peter Skautrup Centret for Jysk Dialektforskning. Braunmüller, K. 2007. Receptive multilingualism in Northern Europe in the Middle Ages. A description of a scenario. In Receptive Multilingualism. Linguistic Analyses, Language Policies and Didactic Concepts [Hamburg Studies on Multilingualism 6], J.D. ten Thije & L. Zeevaert (eds), 25–47. Amsterdam: John Benjamins. Comrie, B. 1998. Rethinking the typology of relative clauses. Language design 1: 59–86. Comrie, B. 2002. Rethinking relative clause types: The Mediterranean area. In Mediterranean Languages. Papers from the MEDTYP Workshop, Tirrenia, June 2000 [Diversitas linguarum 1], P. Ramat & T. Stolz (eds), 87–98. Bochum: Brockmeyer. Comrie, B. & Kuteva, T. 2008a. Relativization on objects. In The World Atlas of Language Structures Online, Ch. 123, M. Haspelmath et al. (eds). Munich . Comrie, B. & Kuteva, T. 2008b. Relativization on subjects. In The World Atlas of Language Structures Online, Ch. 122, M. Haspelmath et al. (eds). Munich . Dahl, Ö. & Koptjevskaja-Tamm, M. (eds). 2001. The Circum-Baltic Languages [Studies in ­Language Companion Series 54/55]. Amsterdam: John Benjamins. Diewald, G. 2002. A model for relevant types of contexts in grammaticalization. In New Reflections on Grammaticalization [Typological Studies in Language 49], I. Wischer & G. Diewald (eds), 103–120. Amsterdam: John Benjamins. Düwel, K. 2001. Runenkunde. 3rd edn. Stuttgart: Metzler. Faarlund, J.T. 2004. The Syntax of Old Norse. Oxford: OUP. Ferguson, C.A. 1959. Diglossia. Word 15: 325–340. Harbert, W. 2007. The Germanic Languages. Cambridge: CUP. Haspelmath, M. 2001. The European linguistic area: Standard Average European. In Language Typology and Language Universals. An International Handbook [Handbooks of Linguistics and Communication Science 20.2], M. Haspelmath et al. (eds), 1492–1510. Berlin: de Gruyter. Heine, B. & Kuteva, T. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Heine, B. & Kuteva, T. 2003. On contact-induced grammaticalization. Studies in language 27: 529–572. Heine, B. & Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: CUP. Hinskens, F., Auer, P. & Kerswill, P. 2005. The study of dialect convergence and divergence. Conceptual and methodological considerations. In Dialect Change. Convergence and Divergence in European Languages, P. Auer, F. Hinskens & P. Kerswill (eds), 1–48. Cambridge: CUP.



Converging languages, diverging varieties 

Höder, S. 2008. Sprachausbau im Sprachkontakt. Eine theoretische und empirische Analyse zum Syntaxwandel im Altschwedischen. Ph.D. dissertation, University of Hamburg. Kjartan Ottósson. 2002. Old Nordic. A definition and delimitation of the period. In The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol. 1 [Handbooks of Linguistics and Communication Science 22.1], O. Bandle et al. (eds), 787–793. Berlin: de Gruyter. Kloss, H. 1967. ‘Abstand languages’ and ‘Ausbau languages’. Anthropological linguistics 9(7): 29–41. Kloss, H. 1978 [1952]. Die Entwicklung neuer germanischer Kultursprachen seit 1800, 2nd edn [Sprache der Gegenwart 37]. Düsseldorf: Schwann. Koch, P. & Oesterreicher, W. 1994. Schriftlichkeit und Sprache. In Writing and its Use. An Interdisciplinary Handbook of International Research, Vol. 1 [Handbooks of Linguistics and Communication Science 10.1], H. Günther & O. Ludwig (eds), 587–604. Berlin: de Gruyter. Koch, P. & Oesterreicher, W. 2007. Schriftlichkeit und kommunikative Distanz. Zeitschrift für germanistische Linguistik 35: 346–375. Kortmann, B. (ed.). 2004. Dialectology Meets Typology. Dialect Grammar From a Cross-Linguistic Perspective [Trends in Linguistics. Studies and Monographs 153]. Berlin: de Gruyter. Kremnitz, G. 2004. Diglossia – polyglossia. In Sociolinguistics. An International Handbook of the Science of Language and Society, Vol. 1 [Handbooks of Linguistics and Communication Science 3.1], 2nd edn, U. Ammon et al. (eds), 158–165. Berlin: de Gruyter. Kurzová, H. 1981. Der Relativsatz in den indoeuropäischen Sprachen. Hamburg: Buske. Labov, W. 1994. Principles of Linguistic Change, Vol. 1: Internal Factors [Language in Society 20]. Oxford: Blackwell. Lass, R. 1997. Historical Linguistics and Language Change [Cambridge Studies in Linguistics 81]. Cambridge: CUP. Lehmann, C. 1984. Der Relativsatz. Typologie seiner Strukturen, Theorie seiner Funktionen, Kompendium seiner Grammatik. Tübingen: Narr. Lehmann, C. 1995. Relativsätze. In Syntax. An International Handbook of Contemporary Research [Handbooks of Linguistics and Communication Science 9.2], J. Jacobs et al. (eds), 1199–1216. Berlin: de Gruyter. Lindblad, G. 1943. Relativ satsfogning i de nordiska fornspråken [Lundastudier i nordisk språkvetenskap 1]. Lund: Gleerup. Lönnroth, L. 2005. From Old Nordic to Early Modern Nordic: The development of types of text. In The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol. 2 [Handbooks of Linguistics and Communication Science 22.2], O. Bandle et al. (eds), 1189–1195. Berlin: de Gruyter. McWhorter, J. 2007. Language Interrupted. Signs of Non-Native Acquisition in Standard Language Grammars. Oxford: OUP. Milroy, J. 2003. On the role of the speaker in language change. In Motives for Language Change, R. Hickey (ed.), 143–157. Cambridge: CUP. Norde, M. 1997. Middle Low German-Middle Scandinavian language contact and morphological simplification. Multilingua 16: 389–409. Noreen, A. 1904. Altnordische grammatik, Vol. 2: Altschwedische grammatik mit einschluss des altgutnischen [Sammlung kurzer Grammatiken germanischer Dialekte A 8]. Halle: Niemeyer. Nygaard, M. 1905. Norrøn Syntax. Kristiania: Aschehoug. Öberg, J. 1994. Vem kunde latin i medeltidens Sverige? In Medeltida skrift- och språkkultur. Nordisk medeltidsliteracy i ett diglossiskt och digrafiskt perspektiv II. Nio föreläsningar från

 Steffen Höder ett symposium i Stockholm våren 1992 [Opuscula 2], I. Lindell (ed.), 213–224. Stockholm: Runica et mediævalia. Pettersson, G. 1976. Om vilken som relativt pronomen. In Nordiska studier i filologi och lingvistik. Festskrift tillägnad Gösta Holm på 60-årsdagen den 8 juli 1976, 342–352. Lund: Studentlitteratur. Ramat, P. & Stolz, T. (eds). 2002. Mediterranean Languages. Papers From the MEDTYP Workshop, Tirrenia, June 2000 [Diversitas linguarum 1]. Bochum: Brockmeyer. Stroh-Wollin, U. 2002. Som-satser med och utan som [Skrifter utgivna av Institutionen för nordiska språk vid Uppsala universitet 58]. Uppsala: Institutionen för nordiska språk vid Uppsala universitet. Teleman, U. 2002. Ära, rikedom och reda. Svensk språkvård och språkpolitik under äldre nyare tid [Skrifter utgivna av Svenska språknämnden 85]. Stockholm: Norstedts Ordbok. Teleman, U. 2003. Swedish. In Germanic Standardizations. Past to Present [IMPACT: Studies in Language and Society 18], A. Deumert & W. Vandenbussche (eds), 405–429. Amsterdam: John Benjamins. Thorell, Olof (ed.) 1959. Fem Moseböcker på fornsvenska enligt Cod. Holm. A 1 [Samlingar utgivna av Svenska fornskriftsällskapet 1.60]. Uppsala: Almqvist & Wiksell. Weinreich, U. 1954. Is a structural dialectology possible? Word 10: 388–400. Wenning, A. 1941. Relativ satsfogning i fornsvenskan. Lund: Berling. Wessén, E. 1956. Svensk språkhistoria, Vol. 3: Grundlinjer till en historisk syntax [Stockholm Studies in Scandinavian Philology 19]. Stockholm: Almqvist & Wiksell. Wollin, L. 1983. Svensk latinöversättning, Vol. 2: Förlagan och produkten [Samlingar utgivna av Svenska fornskriftsällskapet 1.74.2]. Lund: Blom. Wollin, L. 1991. Ska vi forska om fornsvenska? In Studier i svensk språkhistoria 2 [Nordistica Gothoburgensia 14], S.-G. Malmgren & B. Ralph (eds), 239–257. Gothenburg: Acta Universitatis Gothoburgensis. Wollin, L. 2003. Swedish and Swedish. On the origin of diglossia and social variation in the Swedish language. In Aspects of Multilingualism in European Language History [Hamburg Studies on Multilingualism 2], K. Braunmüller & G. Ferraresi (eds), 145–171. Amsterdam: John Benjamins. Zifonun, G. 2001. Grammatik des Deutschen im europäischen Vergleich. Der Relativsatz [Arbeitspapiere und Materialien zur deutschen Sprache 2001.3]. Mannheim: Institut für Deutsche Sprache.

Converging verbal phrases in related languages A case study from Faro-Danish and Danish-German language contact situations* Karoline H. Kühl & Hjalmar P. Petersen Hamburg, Research Centre on Multilingualism

The paper deals with Danish-Faroese and Danish-German language contact, in particular Faro-Danish and Danish-German verb phrases that are constructed under heavy influence of Faroese and German, respectively. The theoretical framework is the Abstract Level Model (2002) as well as further developments by Fredsted (2008) and K. Kühl (2008). The focus of the work described is on the process and the results of convergence. It will be shown how semantic information, syntactic information (predicate-argument-structure) and the morphological realization patterns are transferred extensively from one language to another, without this being evidenced by an insertional kind of code-switching.

1.  Introduction In this article we will concentrate on verb phrases (VPs) in bilingual speech, produced in Danish but heavily influenced by Faroese and German, respectively. The language contact situations referred to are Danish as L2 on the Faroe Islands spoken by Faroe Islanders with Faroese as their L1, and Danish as L2 in the Danish minority in Northern Germany spoken by speakers from the Danish minority with L1 German. Both situations involve intense language contact conditions, with highly proficient speakers of Danish who use the language daily and extensively. In both language contact situations there has been extensive historical language contact between Danish and Faroese and between German and Danish. Why are we concentrating on VPs? Verbs are pivotal links of inter-linkage processing. By distributing theta-roles, main verbs are crucial for the syntax of the sentence. They assign case and semantic roles, subcategorize for prepositional

*We would like to thank E. Fredsted and K. Braunmüller for comments on earlier versions of this paper and A. MacDougall for proof-reading.

 Karoline H. Kühl & Hjalmar P. Petersen

phrases and are, in many languages, inflectional elements marked for tense, voice, mode and possibly aspect and person as well. The crucial role played by the main verb as a category in appointing grammatical and semantic roles implies a close link between verbal syntax and meaning (Fredsted 2008: 954). This article will concentrate on how VPs are influenced by the extensive (synchronic and diachronic) contact of related languages. We will compare our data from the two different language contact areas mentioned above and apply them to the Abstract Level Model (Myers-Scotton 2002), as well as to Fredsted’s hypotheses concerning the possibilities of inter-language transfer within verb phrases (Fredsted 2008; K. Kühl 2008). The paper is organized as follows: First, we provide an introduction section on bilingualism and language contact on the Faroe Islands and in Northern Germany. We then proceed to a section on methodology, which is followed by a section with some theoretical terms. Finally, we present the data from Faro-Danish and Danish-­German and finish with a discussion and a conclusion.

2.  Danish in Northern Germany and on the Faroe Islands The Danish minority in Northern Germany can be described as a historical, regional community that became a national minority in its present host country when the Duchy of Schleswig was divided in 1920 into two parts by a national border line between Denmark and Germany. Despite being a historically substantiated minority, it is not (and never has been) a language minority in the traditional sense of the word, as the connection between nationality and preferred language (usage) never played a decisive role in the multilingual and multinational region of Schleswig. Today, the Danish minority constitutes approximately 50,000 members.1 It has its own political representation, a complete educational system (from Danish Kindergarten to Danish secondary schools) and various institutions such as libraries, sports clubs and other organizations for recreational activities. Danish is recognized as a minority language by the government of SchleswigHolstein (the State Government), the Bundesregierung (the Federal Government) and the Council of Europe’s Framework Convention from 1995.

1.  These numbers represent a rough calculation, as the government is not allowed to survey members of minorities. The numbers mentioned have resulted from counting the members of the minorities’ institutions (cf. J. Kühl 2004).



Converging verbal phrases in related languages 

Almost all members of the minority are bilingual in German and Danish, although the degree of competence in both languages varies considerably. The spoken variety of Danish is not necessarily Standard Danish, but rather the so-called ‘Sydslesvigdansk’, a contact variety of Danish and German. ‘Sydslesvigdansk’ differs considerably from Standard Danish with regard to phonological, semantic and syntactic features as well as the use of code-switching. It does not have its own written norms, but can nevertheless be found in written form, both in informal texts (e.g. blogs, chat or text messages produced by members of the Danish minority) and produced presumably non-intentionally (e.g. in ‘Flensborg Avis’, the daily newspaper of the Danish minority). ‘Sydslesvigdansk’ is geographically restricted and does not have any influence on the Danish language spoken in Denmark. Although the features of ‘Sydslesvigdansk’ have been established (Pedersen 2000; Braunmüller 1996; Fredsted 1986, 2001), the case of VPs has not yet been accounted for (cf. however Fredsted 2008; K. Kühl 2008). As it is difficult to decide or anticipate whether a certain feature is an (established) feature of ‘Sydslesvigdansk’ or instead produced ad hoc, we will label the examples in this article produced in Danish and influenced of German not with ‘Sydslesvigdansk’, but rather with the neutral term German-Danish. The label “Sydslesvigdansk” implicates a condescending notion, implying that it is not as good as Standard Danish. Therefore it shall be avoided. According to the Home Rule Act (Heimastýrislógin) from 1948, both Danish and Faroese are official languages on the Faroe Islands. Typically, Faroese is spoken as L1 and Danish as L2. Again, this is actually a variety of Danish called ‘Gøtudanskt’ (street-Danish, or Danish from the village Gøta). For the same reasons mentioned above relating to ‘Sydslesvigdansk’, we will not use this label, but rather the term Faro-Danish. According to a report from a language committee on the Faroe Islands, 80% of all teaching materials in upper and secondary schools are in a language other than Faroese, primarily in Danish. The same report concluded that 20% of all ­lessons in upper and secondary schools were given in Danish (Málmørk 2007: 45). A survey from 1990 shows that 45% of the teachers in upper secondary schools were Danes, 49% were Faroese, and 5% foreigners who taught in Faroese. One cannot get along on the Faroe Islands without Danish (Hansen et al. 2001). Illustrative of the situation is a woman from Eastern Europe. In an interview in the newspaper Dimmalætting (2007) she says that it is necessary to know Danish in order to manage in the course of daily life on the islands. These few figures give a rough idea of what the Faroese-Danish language situation looks like. Both language situations are cases of stable individual bilingualism and language contact. The general attitude toward Danish in both the

 Karoline H. Kühl & Hjalmar P. Petersen

Faroese-Danish situation as well as in the German-Danish situation is a positive one; see, for example Søndergaard (1988) and Jacobsen (2008). Both Faroese and Danish are North Germanic languages. The former is of WestScandinavian, the latter of East-Scandinavian origin. They are syntactically quite similar. Faroese has preserved more of the Old Norse lexicon than Danish, but has also borrowed a large number of words from Danish. This makes the languages similar with regard to vocabulary and even syntax, though they remain dissimilar in terms of morphology and phonology. One example is that Faroese has three genders (masc., fem. and neutr.), while Danish only has two, common gender and neuter. This gives rise to very different morphological systems with different realizations of the controller genders on different targets (Petersen 2009). German is a West-Germanic language. Danish has borrowed many words from Low German, words that are still in use in today’s Danish (cf. Winge 2000). Nevertheless, there are noticeable differences with regard to syntax, morphology and phonology. German has three genders (masc., fem. and neutr.), as does Faroese, and four cases. In Danish there is no case marking of nouns except for the genitive; only personal pronouns are marked for case (oblique case and genitive). In Danish, the possessive is often expressed by a fronted genitive (‘Lisa’s house’), while in Modern Colloquial German it is constructed using a prepositional phrase (‘the house of Lisa’). As far as verbs are concerned, Danish verbs are not marked for person; other differences in the verb phrase will be shown in the course of the following data analysis. To sum up: We have two situations of extensive individual bilingualism and extensive (historical) language contact, both concerning Germanic languages. In these situations there are few limitations concerning the language competence of the speakers or the intensity and scope of language contact. Thus, we are here dealing with conditions which enable us to observe the far-reaching possibilities of language contact. 3.  Methodology The German-Danish data corpus consists of 28 hours of oral language recorded in different situations (interviews, free talk, group conversation, team work for school-related tasks) produced by 45 speakers. The speakers were pupils in the 7th grade attending two different Danish minority schools. The recordings were made by an interviewer who is a bilingual speaker of Standard Danish and Standard German. These data were collected as part of the project Bilingualer Sprachgebrauch von Jugendlichen, 2004–2006, at the University of Flensburg, funded by the German Science Foundation (Deutsche Forschungsgemeinschaft).



Converging verbal phrases in related languages 

The majority of the speakers had German as their dominant L1. This might be surprising, but as a general rule the children attending the Danish minority schools do not have Danish as their dominant language. They typically start acquiring the minority language in Kindergarten. There are some exceptions (3 speakers were Danish-German balanced bilinguals and 3 speakers displayed Danish as their dominant L1). The pupils were proficient bilinguals by the time they reached the 7th grade. The Faro-Danish data are part of a larger study on Faroese-Danish language contact and bilingualism on the Faroe Islands. This study is currently being conducted at the Research Centre 538: Multilingualism at the University of Hamburg (Sonderforschungsbereich Mehrsprachigkeit) and is also funded by the German Science Foundation (Deutsche Forschungsgemeinschaft). The first part of the Danish data is based on 24 informal interviews conducted by a Danish research assistant on the Faroe Islands during the summer of 2006. She is a native speaker of Danish and she does not understand or speak any Faroese, so the informants had to speak Danish with her. The length of each interview was approximately 30 to 50 minutes, of which 30 minutes were transcribed in Praat. The informants spoke about various topics such as children’s games, school and so on. They came from two generations: 70+ and between 16–20. 7 women were from the older generation, 6 from the younger and 5 men from the older and 6 from the younger generation.

4.  Theoretical approach Convergence is defined by Myers-Scotton (2002) and (2006) as [bold face in original]: “[…] speech by bilinguals that has all the surface-level forms from one language, but with part of the abstract lexical structure that underlies the surface-level patterns coming from another language (or languages).”  (Myers-Scotton 2006: 271)

This definition is not unproblematic, as it fails to mention the difference between processes and results on the one hand and the convergence between languages (level of language contact) and in individual bilingualism (speaker’s level) on the other. In this paper, we only use the term ‘convergence’ to refer to results of individual bilingualism. The Abstract Level model was outlined by Myers-Scotton and Jake in 1995 and was expanded upon by the same authors in 2000. It refers explicitly to Levelt’s (1989) language production model. The Abstract Level model makes it possible to

 Karoline H. Kühl & Hjalmar P. Petersen

discuss certain convergence phenomena and the nature of an abstract morphosyntactic frame “that structures bilingual clauses in types of contact phenomena other than classic code-switching” (Myers-Scotton 2002: 19). According to Levelt (1989), lexical items consist of two parts: on the one hand the lemma as a concept and on the other hand the morpho-phonological form or lexeme. The lexical entry’s meaning and syntax are represented in the lemma, while morphological and phonological properties are represented in the lexeme. The Abstract Level model by Myers-Scotton postulates that all lemmas in the mental lexicon consist of three levels of abstract lexical structure: “[…] this model is based on premises about the nature of the mental lexicon. The major premise underlying the Abstract Level model is that all lemmas in the mental lexicon include three levels of abstract lexical structure. The three levels contain all the grammatical information necessary for the surface realization of a lexical entry. The levels refer to (i) lexical-conceptual structure, (ii) predicate-argument structure and (iii) morphological realization patterns.” (Myers-Scotton 2002: 19)

Myers-Scotton suggests that each of the mentioned structures can be transferred to another language, implying an inter-linguistic recombination: “The types of abstract structures found in bilingual data showing convergence indicate that any of the three levels posited in the Abstract Level model for one participating language can be copied to the other language in a bilingual clause.”  (Myers-Scotton 2002: 21)

Fredsted (2008: 961) pushes this hypothesis further and shows how the three levels can be split in the lemma of one language and recombined with levels from the verbal lemma of another language so that the three levels become a composite of two languages. The “splitting and recombining” (Fredsted 2008: 968) of the abstract but still language-specific lemma structures leads to convergence in verbal clauses. Based on a combination of the 4-M-model2 and the Abstract Level Model, Myers-Scotton postulates the following hypotheses describing the roles different levels play in language contact: “Of the three levels of abstract lexical structure, the level of lexical conceptual structure in content morphemes is most susceptible to change through attrition/ convergence” (Myers-Scotton 2002: 196).

2.  The so-called 4-M-model was established by Myers-Scotton and Jake (1995 ff.): The aim of the model is to provide universal predictions for possibilities and constraints in code-switching. The model categorizes morphemes on the basis of their formal morphosyntactic properties (whether or not they are conceptually activated, whether they are thematic role assigners or thematic role receivers and whether they refer to grammatical information outside the maximal projection of head) and by their point of activation in the language production process.



Converging verbal phrases in related languages 

“The level of morphological realization patterns is more likely to show modification in attrition than the level of predicate-argument-structure”  (Myers-Scotton 2002: 200).

Following these hypotheses, the lexical-conceptual structures (semantics and pragmatics) are easily borrowed. According to Myers-Scotton’s predictions, morphological realization (surface morphological elements and their order) and predicate-argument structures (how the thematic roles are realized in the phrase structure) are least likely to come from the Embedded Language (EL)3, while the lexical-conceptual information is transferred rather easily across languages. She summarises this in the following hierarchy: Predicate-argument structure < morphological realization patterns < lexical-conceptual structure (Myers-Scotton 2002: 231).

5.  Converging VPs Speaking in terms of Levelt’s production model and Myers-Scotton’s Abstract Level model, we see a clear tendency in our data towards the recombination of verb lexemes from one language with the semantic and/or syntactic features of the correspondent verb lemma from the other language. From a processing model’s point of view, interlingual transfer could be explained as follows: The lemma of the Embedded Language is activated along with the corresponding syntactic and/or semantic and/or morphological information, but the Matrix Languagelexeme is selected in the surface structure. Other research has mentioned this possibility of interlingual transfer, cf. de Bot and Schreuder (1993), Green (1993) and Myers-Scotton (2002), and see Fredsted (2008) and K. Kühl (2008). The convergences that appear in the German-Danish- and Faro-Danish-VPs can be divided into three different types: a. Semantic convergences, where the semantic field of the main verb is extended to the VP in the other language. b. Syntactic VP-convergences, where the predicate-argument-structure of a verb-lemma in language A is transferred to the corresponding verb-lexeme in language B.

3.  In the following we use the terms Matrix Language (ML) and Embedded Language (EL) in the sense of Myers-Scotton, according to whom the matrix language is the variety which structurally dominates the morphosyntactic frame of the linguistic unit (the clause). The structurally less dominant language, the embedded language, contributes largely by supplying lexical elements that are integrated or inserted into the ML-frame (Myers-Scotton 2002: 15f.).

 Karoline H. Kühl & Hjalmar P. Petersen

c. Convergences concerning the morphological realization pattern of VPs or verbs.4 These three types of VP-convergences will be outlined in the following Sections (5.1–5.3). 5.1  Semantic convergence in VPs Consider Example (1a), where a speaker of the older generation on the Faroe Islands tells how she changed her school. There were only two schools in Tórshavn, and she went from one to the other, asking if ‘I may come to school there’. Note that she uses kunne ‘may’, and not Standard Danish måtte ‘may’, with this use of kunne in Faro-Danish corresponding to, as we will mention, the deontic use of kunna ‘may’ in Faroese.

(1) a.

Faro-Danish:



Så gik jeg bare ned til den anden skole, then walked I(N) just down to the other school(OBL),



og spurgte, om jeg kunne komme i skole dér. and asked, if I(N) am able to come in school(OBL) there

‘Then I just walked down to the other school, and asked if I was able to attend school there.’ [intended meaning: ‘…and asked if I might attend school there’]. b. Standard Danish: Så gik jeg bare ned til den anden skole og spurgte om jeg måtte komme i skole der. then walked I(N) just down to the other school(OBL), and asked, if I(N) might attend school(OBL) there c.

Standard Faroese:



So fór eg bara oman í hin skúlan, og spurdi, um eg kundi sleppa í skúla har.

Then went I(N) just down in other school-the(N) and asked if I(N) might get in school(OBL) there.

4.



German-Danish corpus

Faro-Danish corpus

Semantic VP-convergences

123 tokens in 28 hours of recording

56 tokens in 54 hours of recordings

Syntactic VP-convergences

210 tokens in 28 hours of recording

81 tokens in 54 hours of recordings

Convergence in the morphological realization pattern

6 tokens in 28 hours of recording

32 tokens in 54 hours of recordings.



Converging verbal phrases in related languages 

Epistemic possibility5 is expressed in Danish with kunne ‘can’ and with kunna ‘can’ in Faroese (Thráinsson & Vikner 1995: 55; Thráinsson et al. 2004: 306ff):

(2) a.



Faroese: Tað kann vera satt.

b. Danish:



Det kan være sandt. ‘It may be true.’

Deontic modality6 is expressed in Danish with måtte ‘may’ and in Faroese with kunna ‘may’:

(3) a.



Danish: Hun må godt tage min bil. she may well take my car ‘It’s fine that she takes my car’.

b. Faroese:

Hon kann gott fara við mínum bili. she may well take with my car ‘It’s fine that she takes my car.’

That is, the differences and partial overlap between the languages can be described as:

Danish: Faroese:

Epistemic, possibility

Deontic, permission

kunne ‘can’ kunna ‘can’

måtte ‘may’ kunna ‘may’

The partial overlap is the reason why deontic permission is expressed with kunne instead of måtte in (1a) from the Danish-Faroese data.7 This example shows how a partial overlap in meaning leads to a semantic convergence based on a verb lemma. According to Myers-Scotton (2002: 196), such transfers represent convergence on the lexical-conceptual level (cf. Abstract Level model, above). In language

5.  ‘With epistemic modality speakers express their judgments about the factual status of the proposition’ (Palmer 2001: 8). 6.  ‘Deontic and dynamic modality refer to events that are not actualized, events that have not taken place, but are merely potential, and may, therefore, be described as ‘event modality’ (Palmer 2001: 70). 7.  We also observed a similar example in Faroese. In (1a), which is taken from a kind of a underground Faroese homepage, the writer, after seeing the picture of a ‘babe’, says (1a) Handa mátti gjarna kopiera fyri meg = ‘That one may make a copy for me’. He uses the Danish verb måtte in a Faroese form mátti, but in a deontic sense, meaning that the speaker refers to an event that has not yet taken place, as the ‘babe’ has not copied anything at all for the writer.

 Karoline H. Kühl & Hjalmar P. Petersen

combinations where the languages share a common ancestry and have been in contact with one another for centuries, such as Danish with Faroese or German, a semantic double assignment would not have any consequence in the case of synonyms. Such cases of semantic convergence can only be described in terms of a covert convergence, when items with inter-linguistic congruent semantic fields are preferred in usage (cf. Thomason 2001: 289). However, when the verb lemma under consideration is not entirely co-extensive, the transfer of the semantic field from the language A verb lemma to the language B verb lexeme will result in an obvious semantic convergence.8 We will illustrate such semantic convergence in VPs with two more examples, one from the Faro-Danish data and another from the Danish-German data. The Faroese verb at fara ‘to move; to go; to travel…’ denotes broadly a movement or traveling towards a place. The Danish language uses more specific verbs to specify the kind of movement (at gå, at rejse, at tage til, at køre, at cykle …). The following examples show that although the speaker uses a specific Danish verb at gå ‘to walk’, he does this with the same broad semantic field of Faroese at fara:

(4) a.



Faro-Danish: Jeg ønsker at gå til Kalifornien. I(N) wish to walk(inf) to California(obl) Lit.: I wish to walk to California. [intended meaning: ‘I intend to go to California’].

b. Standard Faroese:

Eg ynski at fara til Kalifornia. I(N) wish to go(inf) to California(A) ‘I intend to go to California.’

c.

Standard Danish:



Jeg ønsker at tage til/rejse til Kalifornien. I(N) want to take to(inf)/travel(inf) California(obl) ‘IˉintendˉtoˉgoˉtoˉCalifornia.’

Although the semantic fields of Faroese at fara and Danish at gå overlap in terms of the movement, the phenomenon observed here is clearly a case of the broadening of the semantic field of a Danish verb lexeme according to the Faroese verb lemma.

8.  The difference between a semantic convergence and a loan translation lies in the overlap of the semantic fields of the lemmas (not only verb lemmas). If there is an overlap, however small, the semantic field can be enlarged according the other language; this is considered a semantic convergence. If overlap does not exist between the source word and the newly created word (either displaying a totally different meaning or being a non-existing word in the target language), this is considered a loan translation.



Converging verbal phrases in related languages 

The same thing happens in the Danish-German data with the German verb fahren. Fahren is a verb with a very broad semantic extension that is used for almost all kinds of locomotion and movement/traveling towards a place (not unlike the English ‘to go to’). It can additionally be constructed together with a substantive specifying the vehicle used (Fahrrad fahren = ‘to go by bike’, Schiff fahren = ‘to go by ship …’). As mentioned above, Danish disposes of locomotion verbs that are more specific than fahren, as they include information about the vehicle used (at cykle = ‘to go by bike’, at sejle = ‘to sail, …’). However, the Danish verb at køre (at køre = ‘to drive’; a partial translation equivalent of German fahren) is employed in the data with the broad semantic field and all of the semantic extensions of fahren:

(5) a.



German-Danish: Jeg skal først køre med cykel til busstoppestedet. I(N) must first drive(inf) by bike(obl) to bus-stop-the(obl) ‘Before [that], I have to drive/go by bike to the bus stop.’ [intended meaning: ‘Before [that], I have to bike to the bus stop’].

b. Standard Danish:

Jeg skal først cykle til busstoppestedet. I(N) must first bike(inf) to bus-stop-the(obl) ‘Before that, I have to bike to the bus stop.’

c.

Standard German:



Ich muss erst mit dem Fahrrad zur Bushaltestelle fahren. I(N) must first with the bike(D) to bus stop-the(A) go(inf) ‘Before that, I have to bike to the bus stop.’

The three examples (1a), (4a) and (5a) show how the semantic fields of VPs merge, resulting in semantic convergence. 5.2  Syntactic convergence in VPs Syntactic convergence implies a Danish VP, but with the heavy influence of an underlying language – Faroese or German – on the construction of the VP, as in (6)

(6) German-Danish:

Jeg tror, vi har hele tiden slet ikke I(N) think we(N) have whole time-the at all not taget det op? lifted(PPP) it up(PRT) ‘Iˉthinkˉduringˉtheˉwholeˉtimeˉweˉdidn’tˉliftˉitˉup.’ [intended meaning: ‘I don’t think we recorded it the whole time.’]

 Karoline H. Kühl & Hjalmar P. Petersen

This VP is formed through the activation of the predicate-argument-structure (verbal syntax) of language A (German), while the speaker speaks language B (Danish), implying that the lexical-conceptual information and the morphological realization pattern are realized in this language, language B. This is a “splitting and recombining” (Fredsted 2008: 968) of the predicate-argument-structure of a language A-lemma (German or Faroese) with a language B-lexeme (Danish). The example in (6) shows how a syntactic VP convergence can be caused by the transfer of rules for the separation of verbs and particles (satellites): In Danish, inseparable compound verbs developed to a great extent as transfers from German (and Latin) as a result of language contact of varying intensity beginning around 1100 (Diderichsen 1971; Winge 2000); inseparable compound verbs rarely occur in Old Norse and Old Danish. In many cases, the VPs that were originally adopted from German or Latin display a more abstract meaning (Diderichsen 1971: 236f.). As a consequence, the same verb can be found in today’s Danish as an inseparable compound verb with a more abstract meaning on the one hand and as a separable compound verb with a tangible meaning on the other (for example at opstå = “to rise from the dead” vs. at stå op = “to rise, to get up”). We discuss (6) in more detail below. The example in (7a) shows how the rules for the separation of verb and particle can be transferred, creating a syntactic VP convergence. Note that Danish has an inseparable compound verb optage ‘to record’ (7b), while German has a separable compound verb (7c).

(7) a.

German-Danish:



Jeg tror, vi har hele tiden slet ikke I(N) think we(N) have whole time-the(obl) atˉall not



taget det op? lifted(PPP) it up(PRT)



‘I think during the whole time we didn’t lift it up.’ [intended meaning: ‘I don’t think we recorded it the whole time’].

b. Standard Danish:

Jeg tror vi hele tiden slet ikke I(N) think we(N) whole time(obl) at all not



har optaget det. have recorded(SUP) it(obl)



‘I think we didn’t record it.’

c.

Standard German:



Ich glaube, wir nehmen es die ganze Zeit nicht auf. I(N) think we(N) record it(A) the whole time(A) not on(PRT) ‘I think during the whole time we didn’t record it.’



Converging verbal phrases in related languages 

It becomes clear that a compound in Danish is required instead of a verb + particle construction, with verb + particle resulting from the influence of German (aufnehmen = ‘to record’, ich nehme etwas auf = ‘I record something’). The separation of particle and verb (a change in the predicate-argument-structure) results in a change in the semantic-syntactic information: The verb at tage ‘to take’ requires an Agent and a Patient. Direction is added to the thematic grid when combined with the particle op ‘up’, e.g. tage noget op fra gulvet ‘to lift something up from the floor’. The compound verb optage ‘record’ requires an Agent, a Patient and an (implicite) Instrument; optage med en båndoptager ‘record with a cassette-recorder’, for example. The change from compound to verb + particle is not merely a change in the morphological realization pattern, but also a change in the predicate-argument-structure, as it affects the syntactic order. In addition, in (7a) it entails a complete change of meaning. Many compound verbs in Faroese are borrowed from Danish. As this is the case, compound and simple verbs co-occur in colloquial speech e.g.: afturhalda ‘hold back’ (lit.: after + hold) and halda aftur ‘hold back’ (lit.: hold after) (Petersen & Adams 2009; Gullbein 2006). (8a) offers an illustrative example in which the Danish VP give ud ‘to distribute’, e.g. give penge ud ‘to spend money’, is modeled on the Faroese verb + particle construction geva út ‘to publish’. The Danish target, however, is the compound udgive ‘to publish’:

(8) a.



Faro-Danish: Han har givet så meget ud. He(N) has given(SUP) so much out(PRT) ‘He has spent a lot (of money)’. [intended meaning: ‘He has published a lot’].

b. Standard Faroese:

Hann hevur givið so nógv út. He(N) has given(SUP) so much out(PRT) ‘He has published a lot.’

c.

Standard Danish:



Han har udgivet så meget. He(N) has outgiven(SUP) so much ‘He has published a lot.’

Another kind of syntactic convergence based on VPs is the transfer of reflexivity. The following example shows how the reflexivity of the German reflexive verb-lemma sich drehen ‘to spin around’ is transferred to its Danish non-reflexive translation equivalent at dreje rundt ‘to rotate’:9 9.  In fact, dreje sig also exists in Danish, meaning someone turning him/herself around actively and autonomously, implying that Agens and Patiens are the same person. Dreje rundt (rotate) does not differentiate between the roles of Agens and Patient.

 Karoline H. Kühl & Hjalmar P. Petersen



(9) a.



German-Danish: …og så drejede de sig i luften. …and then turned they(N) themselves(OBL) in air-the(OBL) ‘…and then they turned themselves in the air’. [intended meaning: ‘…and then they rotated in the air’].

b. Standard Danish:

…og så drejede de rundt i luften. …and then rotated they(N) round in air-the(OBL) ‘…and then they rotated in the air’.

c.

Standard German:



…und dann drehten sie sich in der Luft. …and then turned they(N) themselves(A) in the air-the(D). ‘…and then they rotated in the air’.

We find similar examples in Faro-Danish in the following complex Example (10), where two Faroese constructions are mixed with a Danish compound verb: (10) a.

Faro-Danish:



Det kommer for sig. it(N) comes-3.p.sg. fore SELF ‘It happens’.

b. Standard Danish:

Det forekommer. it(N) fore-comes(3.p.sg) ‘It happens’.

c.

Standard Faroese:



Tað kemur fyri. it(N) comes-3.p.sg. fore ‘It happens’.

d. Standard Faroese:

Tað hendir seg. it(N) happens self(A) ‘It happens’.

(10a) builds on the Faroese verb + particle construction kemur fyri (lit. comes fore) ‘happens, occurs’ as shown in (10c) instead of the Danish compound verb (10b). The reflexive sig in (10a) is adopted from another Faroese verb with a similar semantic field, henda seg [lit. happens self] ‘happen, occur’ (10d).10

10.  This is a kind of transfer that goes both ways, as the following Faroese example shows: The VP is constructed according to the Danish compound verb at foregå ‘to occur’ and not ac-



Converging verbal phrases in related languages 

Convergence in word order can be analysed as another kind of syntactic ­convergence: The rules for placing the words in a syntactic string in main clauses in Danish and German are determined by the main verb. The verbal syntax in ­Danish and German is characterized by two different mirror-inverted principles. The complements of the V are placed left to right in Danish according to their semantic and syntactic closeness to the main verb. In German, this positional principle is mirror-inverted: the complements under the head V are placed right to left according to their semantic and syntactic closeness to the main verb. These positional principles are part of the subcategorizations of the category main verb (Fredsted 1986). This means that the verbal building process and syntactic information are not connected with a specific verb lemma, but rather with the language specific syntactic information of the category main verb as such. If the speaker activates a certain lemma, the specific syntactic information encoded in this lemma will activate the relevant syntactic structure. This justifies the categorization of the change in word order as a convergence based on a verb (Fredsted 1986; 2008). Analyzed as such, word order change can – in terms of the 4-M-model (Myers-Scotton 2002: 245) – be seen as an early system morpheme, when a content morpheme exists as the head of the structure. According to the 4-M-model, a content morpheme is conceptually activated and acts as a thematic role assigner (e.g. a verb, a noun or an adjective) or receiver (e.g. a noun or a preposition). An early system morpheme is conceptually activated, but not a thematic role assigner or receiver. It is shaped by the underlying content morpheme and reinforces the realisation of the semantic-pragmatic features of the content morpheme. ‘Early’ refers to the point of activation in the mental lexicon. According to predictions made by the 4-M-Model (Myers-Scotton & Jake 1995, 2000), the activation as a

cording to the corresponding Faroese verb, which would be the reflexive verb + particle ganga fyri seg [lit.: go fore SELF] = ‘to occur’. Nevertheless, the reflexive is maintained:

Faroese:



Alt fyrigekk seg á donskum. all(N) happened SELF(A) on Danish(D) ‘Everything was in Danish’.



Standard Danish:



Alt foregik på dansk. all(N) happened on Danish(OBL) ‘Everything was in Danish.’



Standard Faroese:



Alt gekk fyri seg á donskum. All(N) went for SELF(A) on Danish(D) ‘Everything was in Danish.’

 Karoline H. Kühl & Hjalmar P. Petersen

concept of both content and early system morpheme entails that they can be transferred easily from one language to another, in the sense that they may come from the Matrix Language or the Embedded Language. The following example from the Danish-German data shows how the main verb lide ‘like’ and the auxiliary kan ‘can’ are placed according to German word order rules (11a, cf. 11c), and not according to Standard Danish word order (cf. 11b). (11) a.

German-Danish:



Jeg kan alle lide. I(N) can all(obl) like(1sg) ‘I like [them] all .’

b. Standard Danish:

Jeg kan lide alle. I(N) can like(1sg) all(A) ‘I like [them] all.’

c.

Standard German:



Ich kann alle gut leiden. I(N) can all(A) like(1sg) ‘I like [them] all.’

This is just one example of many in which the word order in Danish follows the German word order. We find changes in the word order in Faro-Danish as well: In subordinate clauses, Standard Danish uses the construction adverb + verb. In Icelandic (Insular Scandinavian), however, the string seen after bridge- and non-bridge verbs is verb + adverb. Faroese allows both the adverb + verb and verb + adverb strings after bridge verbs, as well as after non-bridge verbs (see Petersen (2000); Thráinsson (2004) and Thráinsson et al. (2004: 438 ff)). Given the distribution of verb + adverb in Faroese subordinate clauses, the sentence below is not unexpected. The Faro-Danish sentence in (12a) with the string vil ikke ‘will not’ is based on Faroese word order (12b), not on Standard Danish word order (12c). The informant speaks about an American preacher and remarks that he is: (12) a.

Faro-Danish:



En meget interessant mand, som mange mennesker vil ikke a very interesting man(N) who many people(N) will not



høre om. hear about



‘A very interesting man, who many people do not want to hear about.’



Converging verbal phrases in related languages 

b. Standard Faroese:

Ein ógviliga áhugaverdur maður, sum nógv fólk vilja ikki a very interesting man(N) who many people(N) will not



hoyra um. hear about



‘A very interesting man, who many people do not want to hear about’.

c.

Standard Danaish:



En meget interessant mand, som mange mennesker ikke vil a very interesting man(N) who many people(N) will not



høre om. hear about



‘A very interesting man, who many people do not want to hear about’.

Another cause for the adoption of the syntactic property ‘main verb’ from a language A-lemma to a language B-lexeme is another cause of syntactic convergences in VPs. This happens quite often in the Danish-German data: a Danish verb lexeme takes over the role of a main verb, as the corresponding German verb would have done in the same construction. The following example shows a case in which the Danish modal verb at ville ‘to want’ takes over the features of the corresponding German main verb wollen ‘to want’ in the correspondening German verbal phrase. The necessary Danish main verb have ‘to have’ is thereby deleted, leaving only the model verb: (13) a.

German-Danish:



…det ville Aase sådan. …it(obl) want Aase(N) such ‘Aase wanted it so.’ [intended meaning: ‘…Aase wanted it like that’].

b. Standard Danish:

…sådan ville Aase have det. …such wanted Aase(N) it(obl) ‘…Aase wanted it like that.’

c.

Standard German:



…das wollte Aase so. …it(A) wanted Aase(N) such ‘Aase wanted it like that.’

This kind of syntactic convergence becomes visible through the lack of the infinite verb. Another kind of syntactic convergence arises from the fact that German VPs defining the theta-roles by case often correspond with Danish VPs selecting a prepositional phrase. If this syntactic feature, i.e. case vs. a PP, is in fact transferred interlingually, this will cause crucial changes in the VP-construction.

 Karoline H. Kühl & Hjalmar P. Petersen

In the following example (14), the predicate-argument-structure of the ­German verböffnen ‘to open’ (selecting case, cf. (14b) is transferred to the corresponding Danish verb åbne ‘to open’, which requires a preposition, see (14a). This syntactic transfer leads to an important change in meaning: (14) a.

German-Danish:



Ja, Pia åbner ham. yes, Pia(N) opens him(obl) ‘Pia opens him up’ [intended meaning: ‘Yes, Pia opens [the door] for him].’

b. Standard Danish:

Ja, Pia åbner døren for ham. yes, Pia(N) opens door-the(obl) for him(obl) ‘Yes, Pia opens the door for him.’

c.

Standard German:



Pia öffnet ihm [die Tür]. Pia(N) öffnet him(D) (the door(A)) ‘Pia opens (the door) for him’.

The speaker’s target meaning in (14) was ‘Pia opens up the door for him’ (cf. (14b)). This can be realized in German with the verb öffnen ‘to open’, which selects two objects: an NP-dative/benefactive (ihm ‘him’) and a NP-accusative/affected (die Tür ‘the door’) (14c). Although Danish personal pronouns are marked for oblique and non-oblique case, this distinction only marks non-subject and subject. This implies that this case distinction as such cannot define the benefactive and affected object. This distinction must be realized in a prepositional phrase (as shown in (14b). It is quite obvious that the speaker in (14a) tries to maintain the German construction while speaking Danish, making the most of the Danish case possibilities. The syntactic transfer results in an involuntary but crucial change in meaning. In Danish, an adverb of direction (hen ‘to’, op ‘up’, ned ‘down’, langs ‘alongside’…) adds a telic meaning to a verb as seen in de går hen til huset ‘they walk over to the house’. A telic reading means that the action has a beginning and an end, as opposed to an atelic reading as in de går ‘they walk’. In German, for example, telicity is expressed by preposition + case, cf. the following sentences: (15) German preposition + accusative: Sie liefen auf das Fußballfeld. they(N) ran on footballfield-the(A) ‘They ran to the football field.’



Converging verbal phrases in related languages 

(16) German preposition + dative: Sie liefen auf dem Fußballfeld. they(N) ran on the football-field(D) ‘They were running on the football field.’

The difference in the realization of telicity in Danish (preposition + adverb of direction) and German (preposition + case) results in syntactic convergences that are meant to express a telic reading. There are many examples in the Danish-­German data in which the Danish VP is constructed without the directional adverb, following the German construction. Such an example is shown in (17a): (17) a.

German-Danish:



…så kom politiet, og de løb på taget. …then came police-the(N), and they(N) run on roof-the(obl). ‘…then the police came, and they were running on the roof.’ [intended meaning: ‘…then the police came, and they ran up onto the roof ’].

b. Standard Danish:

…så kom politiet og de løb op på taget. …then came police-the(N), and they(N) ran up on roof-the(obl) ‘…then the police came and they ran up onto the roof ’.

c.

Standard German:



…dann kam die Polizei und sie liefen auf dem Dach. …then came the police(N) and they(N) ran on the roof(D) ‘…then the police came, and they ran on the roof ’.

Note again the involuntary change of meaning that is caused by syntactic convergence. Another kind of verbal convergence stems from the differences between supine attraction in Faroese and Danish. Faroese, Norwegian and Swedish dialects in Jämtland exhibit supine attraction or supine spreading (Sandøy 2001: 140) in which an infinitive changes to a supine as the result of the supine in the preceding main verb, as seen in (18a). Supine attraction is a phenomenon that is unknown in Standard Danish. The way in which this causes changes in VPs is shown in the following example: The target in (18a) is Standard Danish at arbejde and not at arbejdet ‘to work’, which is impossible in Standard Danish: (18) a.

Faro-Danish example:



Jeg har prøvet at arbejdet på en café. I(N) have tried(SUP) to work(SUP) on a café(D) ‘I have tried to work at a café.’

 Karoline H. Kühl & Hjalmar P. Petersen

b. Standard Faroese:

Eg havi prøvað at arbeitt á einari kafé. I(N) have tried(SUP) to work(SUP) on a café(D) ‘I have tried to work on a café.’

c.

Standard Danish:



Jeg har prøvet at arbejde på en café. I(N) have tried(SUP) to work(inf) on a café(obl) ‘‘I have tried to work on a café.’

To sum up: The examples above show the numerous ways in which different syntactic features in VPs can be transferred inter-linguistically, resulting in a combination of the predicate-argument-structure from language A (Faroese or German, respectively) and the lexical-conceptual structure and the morphological realization pattern from language B, Danish. 5.3  Convergence in the morphological realization patterns of VPs In the Faro-Danish data we find different cases of convergence in the morphological realization pattern of verb phrases: One such case concerns the past participle. The past participle in Standard Danish is expressed with -t, as in kommet ‘arrived’ (Christensen & Widell 2001: 85). Faroese exhibits different past participle endings depending on the verb class (Petersen & Adams 2008; Thráinsson et al. 2004; Petersen 2001). Ignoring the details, we shall merely mention that the past participle of strong verbs is -in, as in komin ‘arrived’, and that it is this suffix that is activated in (19). An example with a convergence in the morphological realization pattern of the past participle is presented below. In the Faro-Danish example (19), the speaker uses the suffix -en and transfers it to Danish. (19) a.

Faro-Danish example:



…som jeg var kommen at kende. which I(N) was come (PPP) to know ‘which I had learned to know.’

b. Standard Faroese:

…sum eg var komin at kenna. which I(N) was come(PPP) to know(INF) ‘which I had learned to know’.

c.

Standard Danish:



...som jeg var kommet til at kende. which I(N) was come (PPP) to(PRT) to know(INF) ‘which I had learned to know’.



Converging verbal phrases in related languages 

The Faro-Danish verbs that are presented in the table below all have a “wrong” past tense. Three of the verbs are strong verbs in Standard Danish: forstod ‘understood’, past tense of at forstå ‘to understand’; gad ‘liked (to)’, past tense of at gide ‘to like (to)’; and frøs ‘froze’, past tense of at fryse ‘to freeze’. It should be noted that these forms show that the past tense suffix -ede is more productive in the interlanguage (FD) than in Danish and Faroese. Far.-Dan.

bestemmede

gadede

forståede

frysede

kaldede

visede

Dan. Far.

bestemte bestemmaðu ‘decided’

gad

forstod forstóðu ‘understood’

frøs frystu ‘froze’

kaldte kallaðu ‘called’

viste vístu ‘showed’

‘liked to’

The explanation of the phenomenon observed in the example above is relatively straight-forward, and it is one based on analogical leveling: Two suffixes exist in the past tense in the weak declination in Faroese verb inflection, sg./pl. -aði/-aðu, and -ti/tu (allomorphs: -di/-du; -ddi/ddu; Petersen 2001). Of these, the -aði/-aðu suffixes of the first inflection of the weak verbs are productive, as exhibited by loanwords and historical changes; loanwords automatically enter this class and a strong verb such as bjóða ‘to invite’ and past tense beyð ‘invited’ changes the past tense to bjóðaði ‘invited’ instead of beyð ‘invited’. The ending corresponds synchronically (and mostly also historically) with -ede in Danish and the Old Norse ō-verbs. As the suffix is productive, the past tense of the verb-inflections above is not unexpected. Note that the endings have the Danish form -ede, not the Faroese [-aji] or [-avu]. According to the 4-M model, one should not expect to find an underlying cross-linguistic influence on the morphological realization pattern of verbs. Tense (and agreement) are considered outsider system morphemes and should not be among the prime candidates to be transferred in a language contact situation.

6.  Discussion and conclusion We have compared bilingual VPs from two different but nevertheless comparable language contact situations, where the common features of the two situations and consequently the data pose an interesting juxtaposition: Both the situation on the Faroese Islands and in Northern Germany build upon widespread and extensive individual bilingualism as well as extensive language contact in all domains of daily life, the same holding true for the historical language contact. The languages exposed to this synchronic as well as diachronic language contact – Danish with Faroese and German – are in both cases Germanic languages. The situations described show only few limitations concerning the language competence of the

 Karoline H. Kühl & Hjalmar P. Petersen

speakers or the intensity and scope of language contact. These are conditions which enable us to observe the far-reaching possibilities of language contact through a comparative analysis of data from two different language contact situations. Through the application of a common analytical approach to both data sets we were able to show that VPs are highly susceptible to convergence in both cases. While Myers-Scotton (2002: 231) predicts that the predicate-argument-structure and the morphological realization pattern should be influenced at least by ­language contact, this is not what we observe in our data, where both the morphological realization patterns as well as the predicate-argument-structure are transferred inter-linguistically. In contrast, the comparison of the two different language contact situations confirms the results of Fredsted (2008) and K. Kühl (2008) showing how syntactic features of verb phrases can be activated in another language without being obviously – in the sense of a language shift – connected to a verbal lexeme from the first language. This suggests the possibility of a splitting and recombining of the abstract features of VPs (according to the Abstract Level model). Obviously, VP-features such as predicate-argument-structure and lexical-conceptual information can be transferred extensively across languages and recombined with verb lexemes from another language: The transfer of morphological realization patterns in Faro-Danish shows that it is indeed possible for tense (and agreement) to be transferred, despite other predicions. Whether the systematic takeover which we saw in the Faro-Danish data – compared to the rather random takeover in the German-Danish data – is due to factors on the speakers’ level (e.g. a certain tradition developed by the Faroese speakers) or to factors on the level of language contact (structural equivalence between Faroese and Danish) remains to be seen. These results show how the grammatical system is definitely more susceptible to change than previously assumed. Further, our results do not confirm the general assumption that while changes in the semantic field are quite frequent, changes in the grammar of a language are strongly resisted. It must be mentioned, however, that the convergences in the grammatical system do not imply mixed VP-features (for instance, a mixed predicate-argument-structure). The convergences shown here are always composites of two languages, e.g. predicateargument-structures of language A combined with the lexical-conceptual information of language B. The results also fail to comply with the traditional concept of Matrix Language and Embedded Language (terms referring to Myers-Scottons MLF-model, cf. footnote 3), indicating a clear distribution of roles set up by a dominant and a subordinating variety. In the examples shown, the interlingual transfer of the VP features does not seem to entail an actual obvious language shifts. This has certain implications for understanding the roles of the Matrix Language and the Embedded



Converging verbal phrases in related languages 

Language. Regarding the syntactic VP-convergences shown above, it seems that the language contributing to the phrase with the predicate-argument-structure should be considered the dominant language, as it determines the number and form of complements as well as the word order (Myers-Scotton, p.c., accepts this view). However, this means that the Matrix Language in these phrases is completely invisible on the surface. This point of view represents a considerable departure from traditional attempts to determine the dominant language. It also points towards a new dimension in determining the roles of the languages participating in language contact. Whether these kinds of language contact phenomena are due to the special features of the two situations described – relatedness of languages, extensive language contact, early sequential and highly competent bilingual speakers – or whether they can also be found under different circumstances remains to be seen.

References Braunmüller, K. 1996. Sydslesvigdansk – et regionalsprog? In Talesprogsvariation og sprogkontakt. Til Inger Ejskjær på halvfjerdsårsdagen den 20. maj 1996. Institut for dansk dialektforskning (ed.), 33–44. Copenhagen: Reitzel. Christensen, C.B. & Widell, P. 2001. Nudansk Grammatik. Copenhagen: Politikens Forlag. De Bot, K. & Schreuder, R. 1993. Word production and the bilingual lexicon. In The Bilingual Lexicon, R. Schreuder & B. Weltens (eds), 191–214. Amsterdam: John Benjamins. Diderichsen, P. 1971. Elementær dansk grammatik. Copenhagen: Gyldendal. Fredsted, E. 1986. Syntax, Deutsch-Dänisch kontrastiv. In Arbeiten zur Skandinavistik. U. Groenke, (ed.), 19–36. Frankfurt: Peter Lang. Fredsted, E. 2001. Parallelføring af nært beslægtede sprog. In Møde om Udforskningen af Dansk Sprog 8. P. Widell & M. Kunøe (eds), 83–93. Århus: Nordisk Institut. Fredsted, E. 2008. Convergence in verb phrases. Linguistics 46: 949–982. Green, D. 1993. Towards a model of L2 comprehension and production. In The Bilingual Lexicon. R. Schreuder & B. Weltens (eds), 249–277. Amsterdam: John Benjamins. Gullbein, S. 2006. Soleiðis segði mamma. MA thesis at Føroyamálsdeildin, Fróðskaparsetur Føroya. Hansen, Z.S. et al. 2001. Udredning om sprogpolitiske initiativer og domænetab i færøsk. Inspireret af den sprogpolitiske referencegruppe under Nordisk Ministerråd. Ms. Jacobsen, J.í Lon. 2008. Álvaratos, who cares? Ein samfelagsmálvísindalig kanning av hugburði og nýtslu av tøkuorðum og nýggjum orðum í føroyskum. Dr. phil. dissertation, University of Bergen. Kühl, J. 2004. Nationale Identität und kulturelles Gedächtnis. Die dänische Minderheit in Schleswig-Holstein. In Nordlichter. Geschichtsbewusstsein und Geschichtsmythen nördlich der Elbe [Beiträge zur Geschichtskultur 27], B. Lundt (ed.), 321–340., Köln: Böhlau. Kühl, K. 2008. Bilingualer Sprachgebrauch von Jugendlichen im deutsch-dänischen Grenzgebiet. Hamburg: Kovač. Levelt, W. 1989. Speaking: From Intention to Articulation. Cambridge MA: The MIT Press. Málmørk. 2007. Málstevnunevndin. Tórshavn. Mentamálaráðið.

 Karoline H. Kühl & Hjalmar P. Petersen Myers-Scotton, C. 2002. Contact Linguistics. Bilingual Encounters and Grammatical Outcomes. Oxford: OUP. Myers-Scotton, C. 2006. Multiple Voices. An Introduction to Bilingualism. Oxford: Blackwell. Myers-Scotton, C. & Jake, J. 1995. Matching lemmas in a bilingual language production model: Evidence from intrasentential code-switching. Linguistics 33: 981–1024. Myers-Scotton, C. & Jake, J. 2000. Four types of morpheme: Evidence from aphasia, codeswitching and second language acquisition. Linguistics 38: 1053–1100. Palmer, F.R. 2001. Mood and Modality. Oxford: OUP. Pedersen, K.M. 2000. Dansk sprog i Sydslesvig. Aabenraa: Institut for Grænseregionsforskning. Petersen, H.P. 2000. IP or TP in Modern Faroese. Working Papers in Scandinavian Syntax 66: 75–83. Petersen, H.P. 2001. Hovedverber i færøsk: En studie af ordstadiet. Norsk lingvistisk tidsskrift 19: 3–28. Petersen, H.P. & Adams, J. 2009. Faroese Grammar for Foreigners. Tórshavn: Stiðin. Petersen, H.P. 2009. Gender Assignment in Modern Faroese. Hamburg: Kovač. Sandøy, H. 2001. Færøsk i vestnordisk språkhistorie. In Moderne lingvistiske teorier of færøsk, K. Braunmüller & J.í Lon Jacobsen (eds), 125–154. Oslo: Novus. Søndergaard, B. 1988. Motivationsprofilen ved indlæring af L2 med henblik på dansk på Færøerne. In A. Holmen et al. (eds), 291–297. Clevedon: Multilingual Matters. Thomason, S.G. 2001. Language Contact: An Introduction. Edinburgh: EUP. Thráinsson, H. & Vikner, S. 1995. ‘Modals and double modals’. Working Papers in Scandinavian Syntax 55: 51–88. Thráinsson, H., Petersen, H.P., Jacobsen í Lon, J. & Hansen, Z.S. 2004. Faroese. An Overview and Reference Grammar. Tórshavn: Føroya Fróðskaparfelag. Thráinsson, H. 2004. Um sagnbeygingu, sagnfærslu og setningagerð í færeysku og fleiri málum. Íslenskt mál 23: 7–70. Winge, V. 2000. Pebersvend og poltergejst. Tysk indflydelse på dansk. Copenhagen: Gyldendal.

Convergence and divergence of communicative norms through language contact in translation Viktor Becher, Juliane House & Svenja Kranich

Hamburg, University and Research Center on Multilingualism This contribution addresses the question of whether and how translation as a classic case of language contact can act as a trigger for convergence and divergence phenomena between two languages. We present two studies which indicate that translation-induced convergence does not occur unconditionally: while we found no signs of English-German convergence in the use of modal verbs (study 1), the use of sentence-initial concessive conjunctions in translated and comparable German texts shows convergence with Anglophone usage patterns (study 2). Explaining these disparate results, we hypothesize that divergence occurs when bilinguals perceive profound differences between source and target language (as is the case in English and German lexicogrammatical means for expressing modality), while convergence takes place when bilinguals perceive items as equivalent in form and function (as is the case in English and German concessive conjunctions).

1.  Introduction This paper presents some recent results from the project Covert Translation.1 The general assumption underlying this project is that the dominant status of the English language in global communication influences and changes communicative norms in other languages through language contact in covert translation and comparable text production. Covert translations (House 1977; 1997) are translations in which the communicative purpose of the original text is upheld via the .  The project is directed by Juliane House, current research fellows are Viktor Becher and Svenja Kranich. Claudia Böttger, Julia Probst, Nicole Baumgarten and Demet Özçetin were members of the research team in previous project phases. The project is part of the Research Center on Multilingualism (Sonderforschungsbereich “Mehrsprachigkeit”) at the University of Hamburg. It has been funded by the German Science Foundation (Deutsche Forschungsgemeinschaft) since 1999. We gratefully acknowledge this ongoing generous support.

 Viktor Becher, Juliane House & Svenja Kranich

use of “cultural filtering”, i.e. adapting the original text in accordance with conventionalized expectation norms of the target audience. If, however, the influence of global hegemonic English is pervasive enough, it will effectively prevent cultural filtering and instead impose Anglophone norms on texts in other languages. To test this hypothesis we relied on the results of House’s (1996; 2006) and Clyne’s (1987; 1994) empirical contrastive pragmatic work and its postulation of certain dimensions of communicative preferences along which English and German text production tend to differ. Particularly relevant to the case studies presented in this article are the two dimensions ‘directness vs. indirectness’ and ‘content-orientation vs. addressee-orientation’. In general, English discourse in many genres tends to be more indirect and oriented towards the addressee, while German discourse may be described as more direct and oriented towards content (House 1996; 2006).2 Given the absence of comparable contrastive pragmatic research using different language pairs, we concentrated in this project on the language pair English and German. To test the project hypothesis of global English influence on textual norms we constructed a multilingual diachronic corpus of texts in two genres which we assumed were particularly prone to global Anglophone influence: popular science and economic texts. The bulk of the corpus consists of English originals, their German translations as well as comparable, i.e. non-translated German texts.3 For an analysis of the translation relation, source and target texts were manually aligned sentence by sentence, so that an electronic search of the originals automatically retrieves the relevant segments in the translations, and vice versa. The popular science texts which we will examine in the present study were mostly taken from the journal Scientific American and its German daughter publication, Spektrum der Wissenschaft. The total number of words in the popular science part of our corpus is about 500,000. The texts were sampled from two distinct time frames: 1978–1982 and 1999–2002. Given this corpus design, we are in a position to conduct comparative research into language-specific and genre-specific conventions and into the nature of the translation relations. However, in order to make valid statements about whether it is in fact Anglophone texts that have led to variations and changes of German textual norms over the past 25 years and not any independent development of norms in the two languages involved, we also have to draw

2.  Of course, the characteristics of English and German discourse summarized here should not be viewed as absolute dichotomies but rather as points on a cline. 3.  Additionally, the corpus contains translations in the opposite direction (German → English) as well as translations from English into French and Spanish for validation purposes.



Convergence and divergence through language contact in translation 

on larger German and English reference corpora like for instance the DWDSKerncorpus4 and the British National Corpus. The methodology used in this project is a combination of qualitative and quantitative methods. The qualitative part of the research consists of detailed contrastive analyses of some 70 texts, the quantitative part combines frequency counts with statistical operations, and the re-contextualized qualitative analysis includes an isolation of occurrences of vulnerable items and manual annotations to locate collocations and co-occurrences with several pragma-linguistic categories. We can thus divide our project work into three phases: in a first qualitative phase, texts were analyzed and compared in their entirety on the basis of House’s systemic-functional (register-theory based) translation evaluation model (1997), such that textual profiles were established reflecting whether and how over time cultural filtering has changed or has indeed been completely abandoned. Results of the analyses in this phase show that changes did indeed occur in the realization of subjectivity and addressee orientation. The concept of subjectivity is related in systemic-functional theory to Stance and captured along the register dimension of Tenor (Halliday & Matthiessen 2004). Similar concepts discussed in the literature are point of view and perspective (Smith 2003), metadiscourse (Crismore & Farnsworth 1990), evaluation (Hunston & Thompson 2000) and emotive prosody (Bublitz 2003). Subjectivity is also clearly related to the function certain linguistic elements have in influencing hearers/writers (intersubjectivity or addressee-orientation). For our quantitative work, the complex and rather fuzzy concepts subjectivity and addressee orientation were operationalized to include phenomena such as speaker-hearer deixis, mental processes (in the sense of Halliday), commenting parentheses, and expressions of modality and connectivity. Results of the second, quantitative project phase essentially confirmed the changes over time in regard to the frequency with which functional categories such as speaker-hearer deixis or coordinate conjunctions as means of expressing subjectivity and addressee orientation were realized. In the current third re-contextualized qualitative project phase, where we focus on the translation relation and isolate complex collocation and co-occurrence patterns, we are examining whether equivalent elements occur in the same linguistic environments, and whether they are used for the same communicative functions. Our findings concerning adaptations of German communicative norms in the areas of subjectivity and addressee orientation suggest that in the space of 25

4.  The DWDS-Kerncorpus is a diachronically balanced corpus of written German covering the entire 20th century, each decade being represented by approximately 10 million words of text (Geyken 2007). The corpus can be queried free of charge at http://www.dwds.de.

 Viktor Becher, Juliane House & Svenja Kranich

years there has indeed been an adaptation of popular science and economic texts to Anglophone norms with regard to the use of linguistic means of expressing speaker-hearer deixis (cf. Baumgarten & Probst 2004; Baumgarten & Özçetin 2008; Baumgarten 2008), as well as coherence (cf. Baumgarten’s 2007 detailed study of sentence-initial conjunctions And/Und ). What these phenomena have in common is the fact that they constitute formal and functional equivalents in the two languages involved such that a bilingual speaker (or a translator) is capable of “interlingual identification” (Weinreich 1953). Interlingual identification, in turn, facilitates transfer (cf. also Heine & Kuteva 2005: 4, 219–234). Conversely, in cases where there is no clearly discernible formal and functional equivalence in markers of subjectivity and addressee-orientation, Anglophone influence could not be established in our analyses. Cases in point are the use of mental processes (Özçetin 2008), the realization of explicitness in the description of events and states of affairs (House 2008; Baumgarten et al. 2008), and the employment of linking constructions (Bührig & House 2004; 2007). These findings have given rise to a more specific project hypothesis, namely that the impact of English communicative norms on German communicative norms is facilitated in regard to those linguistic expressions where form-functionequivalence can easily be established by the bilingual individual. In the following, two recent project case studies which further examine this hypothesis will be presented. The first study investigates the use of modal markers – as a case of non-equivalence of form and function in German and English discourse – while the second study investigates the use of sentence-initial concessive conjunctions as a case of formal and functional equivalence. Both case studies were carried out on the popular science part of our corpus described above.

2.  The use of epistemic modal markers 2.1  The textual function of epistemic modal markers Broadly speaking, epistemic modal markers function as “a speaker’s comment on the status of information in a proposition. They can mark certainty (or doubt), actuality, precision, or limitation […]” (Biber et al. 1999: 972). Speakers can use these linguistic elements to “express[…] their lack of confidence in the propositions expressed in these utterances” (Coates 1995: 59). This can be seen in the following examples from our popular science corpus (English original and German translation plus English gloss):

(1) a.

Thus some individuals may have a defective natural interferon system.

b. Somit haben wahrscheinlich einige Patienten ein defektes Interferonsystem. ‘Thus some patients probably have a defective interferon system.’



Convergence and divergence through language contact in translation 

In the above examples, the speaker (i.e. the author of the text), makes it clear that he is not one hundred percent certain of the truth of the proposition. In the English Example (1a), may indicates a relatively low degree of certainty, while wahrscheinlich in the German translation (1b) marks a rather high degree of conviction. That is, in both the English original and the German translation, the speaker modifies the statement, but the modal strength of the epistemic modal element used is different. Clearly, epistemic elements with high and low modal strength (i.e. marking both high and low certainty as to the truth of the proposition) are found in both the English and the German system. However, in accordance with the typical communicative preferences in English and German texts referred to in Section 1 above, we may assume that epistemic modal markers in general and those of low modal strength in particular will be more commonly used in the English texts, because they contribute to making texts more indirect as well as more addressee-oriented. The use of epistemic modal markers is thus not always motivated by a real lack of conviction in the truth of the proposition, but may also be used as a particular conventionalized communicative strategy. The following example illustrates this:

(2) a. Planetary rings are not only striking, exquisite structures; they may be the Rosetta stones to deciphering how planets are born.

One might suppose that the author has more conviction in the correctness of the proposition than the use of may betrays, and that he strives to conceal this in order to make the statement less open to criticism. Such a use of an epistemic modal expression as a hedging device may be motivated by a wish to state matters less directly and leave more room for non-face-threatening intervention (such as disagreement) on the part of the addressee. Hyland (1998: 351) notes that “deference, humility, and respect for colleagues’ views” are conveyed through the hedging use of epistemic modal markers (cf. also White 2003; Graefen 2007). White and Sano (2006) stress that the use of such devices produces a more ‘dialogic’ text which allows for alternative positions and voices. They explain that hedging devices such as epistemic modal markers … signal a recognition that, in the current communicative context, these are contentious matters and thereby […] signal recognition that those being addressed may query, reject or at least find such propositions novel or otherwise problematic. By this mechanism, the interpersonal cost to any who might advance alternative views is lowered as their position is recognised as a valid one in the current ongoing colloquy. (White & Sano 2006: 194)

Thus we come back to the communicative preferences summarized above: the hedging use of an epistemic modal makes the text more interpersonal, more addressee-oriented, more indirect, and therefore more typical of Anglophone than

 Viktor Becher, Juliane House & Svenja Kranich

of German texts. If we look at the translation of (2b), we actually find that the German text again makes a somewhat stronger statement: (2) b. Planetenringe sind nicht nur schöne und spektakuläre Gebilde: Sie könnten geradezu der Schlüssel für die Erkenntnis sein, wie Planeten entstehen. ‘Planetary rings are not only beautiful and spectacular formations: they could in fact be the key to the understanding of how planets are born.’

This can be understood as an adaptation to German textual preferences, i.e. the use of a cultural filter: the translator chooses to strengthen the proposition by adding the modal particle geradezu, thus making the passage better conform to a preference for a direct and content-oriented communicative style rather than a more indirect and addressee-oriented style typical of English texts. 2.2  German translation strategies of English epistemic modals In order to investigate how common such adaptations in the translations are, four English modals – may, might, can and must – were searched for in the corpus and their epistemic uses were isolated. All instances in which the proposition containing the modal was translated were analyzed.5 The first and crucial criterion concerns the modal strength of the translation choice,6 i.e. the degree of certainty expressed concerning the truth of the proposition. Five different translation types can be distinguished in this respect. The first one is a translation using a modally unmarked proposition. Here complete conviction in the truth of what is asserted in the translation is expressed (as in 3b):

5.  Since we are looking at covert translations (cf. House 1997), not all passages of the English original are present in the translations. Sometimes passages are simply left out, possibly because they are not deemed interesting for the German readership. 6.  It is important to keep in mind that epistemic modal markers are not always chosen in order to express a lack of conviction in the truth of the proposition on the part of the speaker, but may just as well serve as a hedge. Using an expression of low modal strength can thus either be motivated by a low conviction in the truth of the proposition, or it can function as a communicative strategy which leaves a great deal of room for non-face-threatening disagreement. Markers of high modal strength, on the other hand, either mark a high degree of conviction in the truth of the proposition, or, as communicative strategy, only leave a certain amount of room for a different opinion. They ‘hedge less’, as they give clear preference to the opinion expressed. Such markers would thus seem to be more in line with typical German than with English communicative preferences.





Convergence and divergence through language contact in translation 

(3) a. But the vast majority of the matter in the universe may not emit photons of any wavelength.

b. Doch der größte Teil der Materie im Kosmos strahlt gar keine Photonen ab. ‘But the greatest part of the matter in the universe does not emit any photons.’

In the majority of cases, however, the English modal element is translated by a German modal element. These instances were classified according to the relation between the modal strength in the English original and in the German translation. Three types were distinguished: higher modal strength, same modal strength, and lower modal strength in the translation. First a more fine-grained classification was attempted. But then a basic classification was found to be more intersubjectively reliable. For the two-way distinction between high modal strength and low modal strength, we applied an easy test: expressions conveying low modal strength can be combined with the negated expression using the same modal element without creating a contradiction. One can thus say: This may or may not be the case. Maybe it is true, maybe it is not true. Expressions of high modal strength or probability, where the speaker clearly favors one possibility over all alternative possibilities, on the other hand, disallow such a combination, e.g. *This should or should not be the case. *Probably it is true, probably it is not true. This test was used in the present study to decide how to classify the modal strength of the translations. The present classification is thus based on a two-way distinction between markers of high and of low modal strength. Additionally, cases where the German translation adds an element that upgrades the strength of the assertion, e.g. when English might is translated by könnte tatsächlich ~ durchaus ~ geradezu (all roughly translatable as ‘indeed, in fact’) were always classified as having ‘higher modal strength in the translation’. Similarly, when another modal element was added which downgraded the overall modal strength, we included it in the category ‘lower modal strength in translation’ (an example of this is found in (6) below): The translation type ‘higher modal strength’ has already been evidenced in Examples (1) and (2). A further example can be seen in (4):

(4) a. The cause of these losses remains unknown, but it may reflect the complexity of the genetic reprogramming needed if a healthy offspring is to be born.

b. Der Grund für diese hohen Verluste ist unbekannt, liegt aber wohl darin, daß die Eizelle das Erbgut des untergeschobenen “älteren” Zellkerns reprogrammieren muß […]. ‘The reason for these high losses is unknown, but probably lies in the fact that the egg needs to reprogram the introduced “older” cell nucleus […]’

 Viktor Becher, Juliane House & Svenja Kranich

Here again the test allows us to perceive that the English modal may is of low modal strength (This may or may not be true is possible), while the German modal element has high modal strength (it is not possible to say *Das ist wohl wahr, oder es ist wohl nicht wahr). The categories ‘same modal strength in translation’ and ‘lower modal strength in translation’ are evidenced by Examples (5) and (6) respectively:

(5) a. Imaging studies over the past decade have indicated which brain regions might malfunction in patients with ADHD.

b. Welche Gehirnregionen beteiligt sein könnten, lassen in den letzten zehn Jahren angefertigte Gehirnaufnahmen vermuten. ‘Which brain regions might be involved is indicated by brain images produced in the last ten years.’

(6) a. Such models might then be used to understand in greater detail the coevolution of memes.

b. Gegebenenfalls könnten die Modelle helfen, die Koevolution von Memen und Genen gründlicher zu verstehen. ‘Possibly the models might help to understand the coevolution of memes and genes more thoroughly.’

Finally we find translations where the modally marked proposition is rendered by a particular lexical choice rather than through recourse to a grammatical marker of modality, as in the following example:

(7) a.

… because mammary cells grow vigorously at this stage of pregnancy, indicating that they might do well in culture.

b. … zu diesem Zeitpunkt wachsen die Zellen des Euters lebhaft, was gute Kultivierbarkeit versprach. ‘… at this point the cells of the udder grow vigorously, which promised good culturability.’

In instances of the last type, the expression of less than full conviction regarding the truth of the proposition via a modal marker in the English original is rendered lexically in German, often in quite creative ways. These cases were very difficult to classify with regard to the modal strength they express, so that they have been put into a category called ‘other’. Concerning the four types where the modal strength of the translation choice could be determined, we may consider the types evidenced in (3) and in (4) as adaptations to German textual preferences, marking full or higher conviction and thus being more direct and more content-oriented, while translations of the types evidenced in (5) and in (6) remain close to the (more indirect, more addressee-oriented) English original.



Convergence and divergence through language contact in translation 

The results for the two different time-windows included in the corpus are shown in Figures 1 and 2: 100%

unmarked

80%

higher

60%

same

40%

lower

20%

other

0% might

may

can

must

Figure 1.  Modal strength in German translations of English modals 1978–1982* * The absolute numbers for this time-span are as follows: might n = 14, may n = 57, can n = 105, must n = 20. Total n = 196. 100%

unmarked

80%

higher

60%

same

40%

lower

20%

other

0% might

may

can

must

Figure 2.  Modal strength in German translations of English modals 1999–2002* * The absolute numbers for this time-span are as follows: might n = 148, may n =184, can n = 327, must n = 29. Total n = 688. The reason for the elevated numbers compared to the earlier time-span mainly lie in the corpus make-up, which contains 42,500 words of English text for the 1978–1982 time-span, but 122,866 words for the 1999–2002 time-span. This has to do with the fact that for the earlier time-span popular scientific English texts which have covert translations into German are hard to come by, as the genre was only getting established in German.

Already from these two figures it becomes apparent that no drastic diachronic change can be observed between the two time-spans. We may note that, overall, translators seem to choose a ‘freer’, more creative translation (as exemplified in (7) above) more often in the latter time-span. This is, however, not immediately relevant for the present investigation. What seems more interesting in the context of the present study is that apparently, translators use modally unmarked propositions as translations of must more often in the first time-span than in the second one. However, one should note that epistemic uses of must are, overall, not all that frequent in our corpus, having an average frequency of only 5.3 (1978–1982) and

 Viktor Becher, Juliane House & Svenja Kranich

2.6 (1999–2002) occurrences per 10,000 words. Taken individually, the findings do not allow strong generalizations. However, considered together, we can certainly see a general trend, as Figure 3 shows: 100% unmarked

80%

higher

60%

same

40%

lower

20%

other

0% 1978–1982

1999–2002

Figure 3.  Modal strength in German translation of English modals: 1978–1982 vs. 1999–2002

Thus one can say that when translators introduce changes regarding the strength of an assertion, there is a clear tendency to increase the modal force compared to the English expressions containing might, may, can or must, i.e. to use a modal marker with a higher modal strength or translate the proposition using a modally unmarked indicative. Translations with a lower modal strength, on the other hand, are much less frequent. We can regard this as an adaptation of the text to German communicative preferences, i.e. as the use of a cultural filter. Instead of a decreasing use of cultural filtering in the English-German translations due to the prestige associated with the English language, which the project hypothesis would have led us to expect, we find, in fact, an increasing divergence from the English model. We shall now see how this could be explained. 2.3  Possible reasons for increasing divergence Looking at Figure 3 we could see that quite a remarkable proportion of the translations do not choose an expression of the same modal strength in German. This is true of almost 30% of the translation choices of the four epistemic modals analyzed in the earlier translations, and of a little over 40% of the translation choices in the later translations. We may thus ask why the German texts diverge so commonly from the model present in the English originals. The reasons may lie in a lack of a form-and-function equivalence between English and German in the field of modality. Other studies conducted in our project which have shown significant influence of English communicative styles on German styles in the German translations have dealt with linguistic elements showing a high degree of formal and functional similarity between the two languages (see for instance Baumgarten 2007



Convergence and divergence through language contact in translation 

on sentence-initial And and Und, and Baumgarten & Özçetin 2008 on the use of we and wir in business communication, as well as the following case study on sentence-initial But and Aber/Doch). These linguistic elements are easily identifiable as equivalents by any bilingual speaker and translator, and they thus provide a gateway for an influence on the frequency and/or functions of these elements. In the field of modality, on the other hand, such obvious equivalences are difficult to find in the language pair English-German. One of the major differences lies in the kind of categories predominantly recruited for the expression of modal meanings in the two languages. While in present-day English the modals are a clear-cut class of their own (cf. e.g. Denison 1993), having grammaticalized to a high degree, the German modal verbs still share many more characteristics with ordinary verbs, as is evident for instance in the existence of non-finite forms (Example 8), the ability to take a direct (Example 9) or a prepositional object (Example 10): (8) a. Maria hat das Gedicht am besten gekonnt. b. *Mary has could [past participle] the poem best. (9) a. Peter darf noch ein paar Kekse. b. *Peter may some more cookies. (10) a. Julia muss nach London. b. *Julia must to London.

A further difference, which can be seen as a consequence of a higher degree of grammaticalization, is that in English the modals are the predominant means of expressing modal meaning. In German, however, a variety of linguistic elements are used to this end. Among the grammatical elements expressing epistemic modal meanings, German commonly uses modal adjectives and adverbs (e.g. möglich, wahrscheinlich), as well as modal particles (e.g. wohl), a category not present in the English inventory (cf. Nehls 1989; König 2001; Teich 2003: 111). From the present state of further investigations, in which we compare the use of epistemic modal elements in the English originals, their German translations and the original texts in our corpus, these other categories seem to be clearly preferred over modal verbs in the untranslated, comparable German texts.7

7.  The current work in progress compares expressions of epistemic modality in English originals, German translations and German originals, taking the functional side of these expressions as our departure point. In a manual analysis of a subpart of the corpus, all expressions that function as a modification of the speaker’s conviction in the truth of the proposition will therefore be isolated. These expressions are then categorized according to modal strength and linguistic

 Viktor Becher, Juliane House & Svenja Kranich

Furthermore, German texts show a certain predilection for lexical expressions rather than grammatical elements. Thus, phrases such as unseres Erachtens (‘in our opinion’) and so weit bisher bekannt (‘as far as it is known so far’) fulfil the same function as epistemic modal markers in that they modify the certainty with which the truth of the proposition is asserted, but they add further information (e.g. the grounds on which it is asserted), thus satisfying the preference for explicitness characteristic of German texts (cf. House 2004a, 2004b). In other words, the linguistic expressions predominantly chosen for the epistemic modification of the proposition are clearly different in the two languages. The translations of the English modal verbs also reflect these systemic differences. The realization of English might, may, can and must in the translations can be categorized into the following types: .

a. b. c. d. e.

zero translation modal verb modal adjective/adverb/particle a combination of several modal elements (e.g. modal verb + modal particle) a creative translation which renders the modal meaning via other means (typically lexical choice)

For reasons of space, we shall just look at the translations of all four modal verbs together. This gives us a good impression of the general tendencies: 100% 80%

a) zero

60%

b) modal verb c) modal particle/adv. /adj.

40%

d) multiple modal marking

20%

e) other

0% 1978–1982

1999–2002

Figure 4.  Formal categories of German translations of English modals 1978–1982 vs. 1999–2002

category of the element used. This allows us to gain both a more detailed view of the contrasts between English and German and the potential differences between German translations and German originals in the genre of popular scientific texts. Additionally, further studies of translation relations, comparable to the present one, will be conducted in the field of modal expressions.



Convergence and divergence through language contact in translation 

Figure 4 shows the following diachronic development: there is a certain rise in the use of modal particles, adverbs or adjectives (c) as well as in the use of combinations of several modal elements (d). Also, the creative choices grouped together here under (e) become more frequent. The rise in categories (c) through (e) happens at the expense of the more straightforward translation of an English modal verb by a German modal verb. As far as the formal properties of the modal expressions are concerned, there is also an increasing divergence from the English original expression. Summarizing, we can conclude that as far as both the functional and the formal sides are concerned, German translations of popular science texts do not seem to be converging with the English model – on the contrary, they show increasing divergence from it (cf. also Kranich 2009).

3.  The use of sentence-initial concessive conjunctions Sentence-initial concessive conjunctions (SICCs) were chosen for analysis because their different use in English and German is indicative of two interesting contrasts between the two languages. These contrasts account for the fact that SICCs (Engl. But, German Aber and Doch,8 both meaning ‘But’) are much rarer in German than in English (cf. Table 1 below). Accordingly, in English-German translations it is often the case that a German equivalent of sentence-initial But such as Aber or Doch is avoided. Specifically, English-German differences in the use of SICCs are due to contrasts in terms of (1) word order and (2) textual norms. In the following, the two contrasts will be discussed in turn. I. Word order. In contrast to the largely grammatically determined word order of the English sentence (SVO), German word order is pragmatically controlled. That is, German observes a stricter organization of the sentence in terms of given and new information than English (Hawkins 1986; Doherty 2001; 2005). And connectives like but “compete with the propositional parts for the initial position in a sentence” (Doherty 2001: 223). This is why German prefers to use a sentenceinternal connective in many situations where English would use a sentence-initial one. Cf. the following two examples from our corpus.

8.  But, aber and doch are often called “contrastive” conjunctions. This is somewhat misleading, as only in a minority of cases these conjunctions are actually used to convey a purely contrastive relation. Their default use seems to be the signalling of a concessive relation (Becher 2007; cf. also the following examples, of which only (13) can be analyzed as contrastive).

 Viktor Becher, Juliane House & Svenja Kranich

(11) a.

But this process is not automatic.

b. Eine solche Entwicklung vollzieht sich allerdings nicht automatisch. ‘Such a development, however, does not take place automatically.’ (12) a.

But that scenario is now thought to contradict observations […].

b. Dieses Szenario scheint jedoch den Beobachtungen zu widersprechen. ‘This scenario, however, seems to contradict the observations.’

In both examples, the respective translator most probably preferred to place aller­ dings and jedoch (both are roughly equivalent to ‘however’) sentence-internally in order to leave the first position in the sentence to the subject, which presents given information. In some cases, an English-German translator may even achieve an adequate translation of sentence-initial But without using a connective at all: (13) a. But the newfound organism not only appeared to poison fish – it ate them as well! b. Nicht so unser Neuling: Er vergiftet die Fische nicht nur – er frisst sie auch! ‘Not so our newcomer: it not only poisons fish – it eats them as well!’

In (13), the contrast expressed by But has been translated into German solely by means of word order. The topicalization of the negated adverb so has placed new information at the beginning of the clause. This departure from the default given-new ordering of the German sentence gives rise to a strong implicature of contrast (cf. Büring 1999, 2006). The example shows that in the case of sentenceinitial But, omission does not necessarily equal non-translation: the informationstructural expectations of German readers are so strong that a departure from the canonical given-new ordering of the German sentence will suffice to reproduce the effect of sentence-initial But. The use of a direct equivalent like Aber or Doch is often unnecessary. II. Textual norms. Another contrast which accounts for the finding that SICCs are more common in English than in German is located in the domain of culturally determined discourse norms: while English (spoken and written) discourse may be characterized as interactional,9 dialogous and addressee-oriented, German discourse has been described as ‘transactional’, monologous and content-oriented (cf. the literature quoted in Section 1 above and Byrnes 1986; Kotthoff 1989;

9.  We follow Thompson and Thetela (1995) in distinguishing interactive from interactional properties of text. While every text is interactive, e.g. in the sense that an author needs to anticipate reader expectations to communicate successfully, a text is only interactional if its interactive quality is explicitly verbalized; for example, possible reader expectations may be ‘hard-wired’ into the text as questions. The term interactional as we use it can therefore be paraphrased as ‘overtly interactive’.



Convergence and divergence through language contact in translation 

Baumgarten & Probst 2004). Among the most important devices that authors can use for “arguing with the reader” (Thompson 2001) are concessive conjunctions (cf. Grote et al. 1997). In our corpus data, we find that especially in sentenceinitial position But is used for the purpose of simulating an interaction between author and reader, an expository strategy which is often not reproduced in German texts by English-German translators, as it is not in line with German textual norms. Specifically, sentence-initial But regularly occurs in two highly interactional discourse patterns (Thompson 2001: 65f), Question–Answer and Claim–Response (cf. Hoey 2001).10 The following two examples illustrate the use of But in the Question part of a Question–Answer pattern: (14) a.

But what caused these calamities in the first place […]?

b. Wie kam es zu der erstaunlichen Klima-Instabilität […]? ‘How did the astonishing climate instability come about ?’ (15) a.

But why do these inhibitory events occur during waking in narcoleptics?

b. Warum die Bewegungshemmung bei Narkoleptikern zur falschen Zeit auftritt, ist noch nicht klar. ‘Why the movement inhibition occurs in narcoleptics at the wrong time is not clear yet.’

While it is not made explicit whether the question in (14) and (15) is attributable to the author, the reader or even to both parties, it is clear that this way of introducing a new subtopic evokes the impression of an interaction between author and reader. If we take the view that rhetorical questions are attributable to the ‘reader-in-thetext’ (Thompson & Thetela 1995; Thompson 2001) – which need not be identical with any real reader – the use of sentence-initial But makes sense here, as it makes the question seem like an objection to what was said before and thus even heightens the interactional appearance of the text. Another effect of But is that it signals the beginning of the Question–Answer pattern. (The Answer part of the pattern is omitted in (14) and (15).) Finally, the use of sentence-initial (or rather: utteranceinitial) But in this context is strongly reminiscent of face-to-face conversation, where it occurs more frequently than in written texts (Biber et al. 1999: 84).

10.  Note that the use of SICCs is not invariably interactional. Rather, it appears that SICCs are put to use very differently in different registers. For example, Bell (2007) finds sentenceinitial But in academic writing predominantly used as “a tool which allows speakers to subtract unintended meanings away from an existing meaning” (Bell 2007: 197). On the other hand, he quotes an example from his corpus where But appears as part of a Question–Answer pattern (no. 28, Bell 2007: 196). A comparative study of the use of sentence-initial But in academic and popular scientific texts seems worthwhile.

 Viktor Becher, Juliane House & Svenja Kranich

The translator of (14) has decided to retain the question, but has weakened its ‘interactionality’ by refraining from translating the objection marker But. In this way, the translator has shown his or her concern for conforming to German textual norms. The translator of (15) has gone even further; not only has s/he dropped the SICC, but also dispensed with the question of the English original by explicitly verbalizing the lack of knowledge it implies (… is not clear yet). We can say that, in this case, the translator has applied a cultural filter that has completely converted the interactional orientation of the source text into a transactional orientation of the target text. The following two examples illustrate the use of But as a signal of the discourse pattern Claim–Response. While in the Question–Answer pattern But introduces the Question, i.e. the first part of the pattern, in the Claim–Response pattern it usually signals the transition to the second part of the pattern: (16) a. The public perception is that we have looked extensively for signs of life elsewhere. But in reality, we have hardly begun to search. b. Für die Öffentlichkeit scheint es, als hätten wir unsere Fühler bereits weit ins All ausgestreckt. Tatsächlich hat die Suche nach außerirdischem Leben aber kaum erst begonnen. ‘To the public, it often seems as if we had already extended our feelers far into the universe. In reality the search for extraterrestrial life has however just begun.’ (17) a. Still, for some it may seem disturbing that life, certainly in its physical incarnation, must come to an end. But to us, it is remarkable that even with our limited knowledge, we can draw conclusions about such grand issues. b. Manchen mag die Vorstellung befremden, daß jedes Leben irgendwann einmal zu Ende geht. Für uns ist es jedoch bemerkenswert, wenn wir überhaupt solche Schlußfolgerungen ziehen können. ‘To some it may be disturbing that every life has its end. To us, however, it is remarkable that we can draw such conclusions at all.’

As in the Question–Answer pattern, the use of But in the Claim–Response pattern of (16) and (17) is reminiscent of spoken conversation, where the discourse marker is commonly used by speakers for claiming the floor even in the absence of a transition-relevance place (Gumperz et al. 1984: 11; Schiffrin 1987). This, in combination with the plural speaker deictic we/us, gives an impression of a high degree of author involvement: the authors in the two examples present the opinion of a more (example. 16) or less (example 17) closely specified group of people as a Claim, with which they contrast their own opinion as a Response. Again, we can observe the tendency of English-German translators to mitigate the interpersonal/ interactional quality of the source text: instead of using sentence-initial Aber,



Convergence and divergence through language contact in translation 

which shares the discourse function of But as a “turn-getter” (Schlobinski 1992), the translators of (16) and (17) have resorted to sentence-internal connectives. Independently of the discourse pattern in which sentence-initial But occurs, it regularly appears in short and pointed sentences. Such sentences typically occur in spoken rather than written discourse and are thus particularly suitable for creating the illusion of spontaneous and interactional speech: (18) a. If some singular, unavoidable flaw caused every cell to fail eventually, no animal would escape aging. But some do. b. Gäbe es ihn, würde kein Tier dem Altern entkommen. Solche Lebewesen existieren aber. ‘If [such a flaw] existed, no animal would escape aging. Such creatures exist, however.’ (19) a. In 1990 a new Peruvian “environmental code” finally made the eradication practice illegal. But pesticides remain pervasive. b. Erst 1990 verbot ein neuer peruanischer “Umweltkodex” die Ausrottungs­ praxis, doch Pestizide beherrschen nach wie vor die Landwirtschaft. ‘Not until 1990 did a new Peruvian “environmental code” forbid the eradication practice, but pesticides are still pervasive in agriculture.’ (20) a. Scientists’ search for life beyond Earth has been less thorough than commonly thought. But that is about to change. b. Die wissenschaftliche Suche danach ist bislang allerdings auch weniger gründlich gewesen als gemeinhin angenommen. Doch das wird sich bald ändern. ‘The scientific search for it has so far however been less thorough than commonly assumed. But that will soon change.’

Of the three Examples, (19) is especially interesting. The translator has merged the two short sentences of the English original into a single sentence, in this way abandoning the colloquial and incremental style of the English original in favor of a more formal and content-oriented style of exposition. (Cf. Fabricius-Hansen 1998, who suggests that English texts tend to ‘build up’ the discourse in an incremental manner while German generally favors a hierarchical and syntactically more compressed mode of writing.) Summing up, the discussion of the above examples has shown that the use of SICCs in German texts is generally dispreferred for (1) information-structural and (2) stylistic reasons.11 We do not want to overstate the case, though: as

11.  It is interesting to note that German textual norms regarding the use of SICCs as they are presented here seem to be a fairly recent development. A query in the DWDS-Kerncorpus reveals that the frequency of Aber has decreased by more than 50 percent around the end of the 1950s,

 Viktor Becher, Juliane House & Svenja Kranich

examples  3 and 20 indicate, German aber and doch do occur sentence-initially. However, this happens less frequently than in English texts, and usually entails a strong rhetorical effect. The examples discussed above show that an adequate translation of sentence-initial But into German often means its omission (But → 〈null〉) or its replacement by a sentence-internal connective (But → aber, doch, jedoch, or the like). In line with our project hypothesis, it is hypothesized that 1. sentence-initial Aber and Doch occur more frequently in the English-German translations than in the non-translated comparable German texts of our popular science corpus (cf. Section 1), 2. this over-use of SICCs in the English-German translations is due to an interference of the English source texts and 3. the comparable, i.e. non-translated German texts ‘inherit’ over time the more frequent use of SICCs from the translations. 4. It is further hypothesized that the driving force behind these changes is the adoption of the interaction-oriented textual norms of English, i.e. EnglishGerman translators and German authors start using SICCs more frequently in order to imitate the author-reader interaction characteristic of (prestigious) English popular science texts. The remainder of this section will present some evidence in support of these hypotheses. 3.1  Quantitative findings Table 1 shows the frequency of sentence-initial But and Aber/Doch respectively in the English source texts, their German translations and the non-translated comparable German texts of the popular science corpus. Since the number of sentences – and hence the number of opportunities to use a SICC – differs between the individual subcorpora, the frequency counts given in the table are normalized on the basis of 1000 sentences.

and has subsequently stayed on this low level until the end of the 20th century. (On the other hand, Doch shows no comparable development.) It would be an interesting topic for further study to assess whether this rather abrupt decline in the use of Aber is part of a more extensive change in German communicative norms around the middle of the 20th century or whether it is an isolated phenomenon (e.g. induced by an emergence of certain prescriptive norms).



Convergence and divergence through language contact in translation 

Table 1.  Frequency of sentence-inital But in English texts and Aber/Doch in their German translations and comparable German texts (per 1000 sentences; n = 621)

English source texts (But) English-German translations (Aber/Doch) Non-translated German texts (Aber/Doch)

1978–1982

1999–2002

development

32.6 22.7   9.0

32.6 30.1 19.8

– +32.5% +119.8%

Let us first consider only the figures for the earlier time-frame (1978–1982). Here, we see that the frequency of But in the English texts is indeed considerably higher than the combined frequency of Aber and Doch in the non-translated German texts (32.6 vs. 9.0 occurrences per 1,000 sentences). The English-German translations are almost exactly in the middle, which could mean that they are (1) affected by a “shining-through” (Teich 2003) of the English source texts and (2) only partly follow the textual conventions of German (at least as far as the use of SICCs is concerned). In any case, hypothesis 1 is supported by the frequency counts. Taking the later time-frame (1999–2002) into account, we can observe two remarkable diachronic developments. First, the number of SICCs in the EnglishGerman translations has risen by 32.5 percent, which indicates that the degree of source language shining-through has increased. Second, the non-translated texts seem to have followed suit: here, the frequency of SICCs has more than doubled (+119.8%), almost reaching the frequency level of the translations in the earlier time-frame. Summing up the results obtained from the frequency counts, it appears to be the case that the English-German translations have – at least in terms of frequency – aligned their use of Aber and Doch to the use of But in their English source texts. And the non-translated German texts, in turn, appear to have followed the lead of the English-German translations, which indeed seem to function as role models for German popular science texts (cf. hypothesis 4 in the previous section, and cf. also Baumgarten 2008). Thus, the frequency counts presented in Table 1 make an English influence on German textual norms via translation seem very likely. However, two questions must be answered before we can think about accepting this explanation: 1. Is the frequency rise of Aber and Doch in the English-German translations actually due to an increased shining-through of the English source texts and an accompanying adoption of their textual norms? 2. Do the non-translated German texts in their use of Aber and Doch actually follow the example of the English-German translations or have they developed idiosyncratic uses of the two conjunctions? The questions will be answered in the following two subsections.

 Viktor Becher, Juliane House & Svenja Kranich

3.2  Translation analysis Table 2 compares the translational equivalents of sentence-initial But chosen by translators in the earlier time-frame to those chosen in the later time-frame.12 Table 2.  Translations of sentence-initial But into German (n = 208) 1978–1982

1999–2002

development

But → Aber, Doch

26.8%

47.4%

+20.6

But → aber, (je)doch, allerdings

46.4%

36.2%

–10.2

But → 〈null〉

17.9%

7.2%

–10.7

But → 〈other〉

8.9%

9.2%

+ 0.3

total

100%

100%

A definite trend is evident: while the number of ‘literal’ translations (But → Aber, Doch) has almost doubled (+20.6 percentage points), the number of ‘free’ translations (But → aber, (je)doch, allerdings and But → 〈null〉) has decreased by an equivalent amount (–10.2 and –10.7 percentage points respectively). The results indicate that the first question posed at the end of the last subsection is to be answered affirmatively: the frequency rise of Aber and Doch in the EnglishGerman translations is due to a change in translation behavior. The table shows that over time, translators hesitate less and less to adopt sentence-initial But as Aber or Doch. And as the following examples illustrate, the changed translation behavior seems indeed to result from – quite successful – attempts to imitate the interactional style of the English source texts: (21) a. Something must have drained away its [sc. the sun’s] angular momentum. But what? b. Irgendwann muß sie den Drehimpuls verloren haben. Aber wie? ‘At some point it must have lost the angular momentum. But how?’ (22) a. … the entire endeavor of robotics has failed rather completely to live up to the predictions of the 1950s […] It is true that industrial robots have transformed the manufacture of automobiles, among other products. But that kind of automation is a far cry from the versatile, mobile, autonomous creations that so many scientists and engineers have hoped for.

12.  Occurrences of But in citations and in sentences that were left out in the German translation were excluded from the analysis (cf. Note 5).



Convergence and divergence through language contact in translation 

b. Gewiß, Industrieroboter spielen in der Produktion von Autos und anderen Gütern inzwischen eine wesentliche Rolle. Aber die sind weit entfernt von dem, was so viele Forscher und Ingenieure sich erträumt hatten. ‘Certainly, industrial robots play an essential role by now in the production of cars and other goods. But they are far away removed what so many researchers and engineers have dreamed of.’

In (21) the short, elliptical question prefaced by the objection marker But gives rise to a high degree of interactionality and author-reader involvement. The author creates the illusion of following a train of thought together with the reader. It is interesting to note that the translator did not mechanically transfer the two sentences into German: he has changed the subject of the first sentence (something → sie [the sun]) and found an equivalent metaphor for expressing the event described (drained away → verloren). This shows that the retention of But as Aber is the result of a conscious decision, namely not to apply a cultural filter. (22) exemplifies what Antaki and Wetherell have aptly described as “making a show of conceding” (1999: 7). The author pretends to concede a possible objection (Claim) by the reader, even emphasizing its truth (It is true that …). In the immediately following Response (introduced by But), however, the author completely marginalizes the ‘conceded’ objection by showing that it is not at all relevant to her argument that “the entire endeavor of robotics has failed rather completely to live up to the predictions of the 1950s”. What we observe here is thus a hardly concealed attempt to persuade the reader, who is pictured as (possibly) objecting to the argument of the author. As in (21), the translator apparently does not feel the need to mitigate the overtly expressed author-reader-interaction and almost literally renders both It is true that … (as gewiß ‘certainly’) and the objection marker But (as Aber) in German. 3.3  Qualitative analysis of non-translated German texts The question that remains to be answered is whether the frequency increase of Aber and Doch in the non-translated German popular science texts in fact results from an adoption of textual norms introduced by the English-German translations. A qualitative analysis of all occurrences of sentence-initial Aber and Doch in the non-translated texts indicates that this is indeed the case. The following examples are representative of the findings. (23) Gewiss sind unter den Lockstoffen für bestäubende Insekten auch Monoterpene. Aber das kann nur ein Nebeneffekt sein… ‘Certainly, among the attractants of pollinating insects are also monoterpens. But that can only be a side-effect…’

 Viktor Becher, Juliane House & Svenja Kranich

(24) Keine Region auf unserem Planeten ist so entlegen und fern jeder menschlichen Zivilisation wie die Polargebiete. Nirgendwo sonst sollte also die Luft so rein und der Himmel so klar sein. Doch weit gefehlt! ‘No region on our planet is as remote and far from every human civilization as the pole areas. Nowhere else should the air be so clean and the sky so clear. But far from it!’ (25) … eine Folge von sehr vielen Einzelanweisungen […] Erst deren Ausführung in ihrer Gesamtheit erweckt den geschilderten Eindruck. Aber wer ist es, der diese Einzelanweisungen ausführt? ‘… a sequence of very many single orders. Only their execution in their entirety gives rise to the described impression. But who is it that carries out these single orders?’

Like (22) above, (23) and (24) may be described as ‘show concessions’ following a Claim–Response pattern. Note that the use of gewiß (‘certainly’) is exactly parallel to the use of the adverb in the translation of (22). This is remarkable because gewiß in this function is extremely rare in contemporary German, as a search for this adverb in the 1990–2000 time-frame of the DWDS-Kerncorpus confirms. (24) is a particularly interesting case of a show concession. The author makes use of the discourse pattern Hypothetical–Real (Winter 1994), a variant of the Claim– Response pattern (Hoey 2001: 179, 188): he deliberately leads the reader up the garden path by incrementally drawing the false picture of unpolluted pole areas. The short, elliptical Doch-sentence then destroys this illusion. Finally, (25) features Aber as an objection marker prefacing a rhetorical question (cf. Examples 14, 15 and 21 above). The interactional (and incremental) mode of presenting information achieved by means of sentence-initial Aber and Doch evidenced in Examples 23 through 25 is (or used to be?) highly untypical of German texts. The findings of the qualitative analysis point to an influence of English-German translations on original German popular science texts. We can thus argue that the increase in frequency of Aber and Doch in these texts is the result of an adoption of Anglophone communicative norms, which have been introduced to the genre of popular science via translations from English. 4.  Conclusion Both case studies presented in this paper confirm the findings of previous contrastive pragmatic studies regarding different communicative preferences in English and German texts by House (cf. e.g. 1996; 1997; 2006). The use of epistemic modal



Convergence and divergence through language contact in translation 

markers, particularly those of low modal strength, which can be associated with less direct and more addressee-oriented rather than content-oriented discourse styles, was shown to be more typical of English than of German popular scientific texts: German translators tend to omit these markers or replace expressions of low modal strength with markers of high modal strength. And the use of the SICCs But, Aber and Doch, which serve to make a text more interactional, was also shown to be originally more associated with English textual conventions. However, as regards our hypothesis that English text conventions have an increasing influence on German communicative preferences, the results of the two case studies presented here differ greatly. In the case of modal markers, the German covert translations show a continuous adaptation to German communicative preferences: epistemic modal markers are left out so that unmodified propositions can take the place of modified ones, and markers of low modal strength (which make the text most indirect and least content-oriented) are frequently replaced by modal markers of higher strength. And both these tendencies are apparent in both time-spans, 1978–1982 and 1999–2002, without any clear indication of changing preferences. In regard to the SICCs investigated in the present study, on the other hand, we found a clear trend towards a decreasing application of a cultural filter. In the second time-span, the German translations of sentence-initial But quite commonly also employ a SICC (Aber and Doch), while translations in which adaptations to German communicative preferences, such as the rendering of the contrast through word order inversions, were undertaken, are used much less frequently compared to the earlier time-span. Thus, we were able to establish an increasing convergence of German covert translations towards the model represented by the (prestigious) English source texts. Moreover, we found that the Anglophone text conventions introduced by the English-German translations are also adopted by authors of originally German texts. Anglophone text norms were thus shown to find their way into the German genre of popular scientific writing. Other studies carried out within the project ‘Covert Translation’ have pointed to similar convergence phenomena, as pointed out in the introduction. For instance, in the use of the personal pronouns we-wir a certain impact of the English preferences on the German covert translations was established (Baumgarten & Özçetin 2008). And similar to what we reported here on the use of SICCs, in the use of sentenceinitial And-Und, convergence towards the English model also extended to German comparable texts produced without an English model (Baumgarten 2007). The main difference between elements such as we-wir, And-Und and ButAber/Doch and the epistemic modal verbs, where we found no signs of increasing convergence, lies in the different degrees of perceivable form-function equivalence.

 Viktor Becher, Juliane House & Svenja Kranich

While in regard to the former linguistic elements, bilingual speakers, and thus also translators, can easily establish the elements as equivalents of each other, the fundamental differences in the linguistic inventory of expressions of epistemic modality in the two languages effectively block such a process in this domain. We could see that English mainly uses a set of highly grammaticalized modals, while German tends to mark epistemic modality with a variety of different elements, some of which (the modal particles) do not exist in English at all. One can therefore assume that no clear one-to-one equivalence relations are apparent to the bilingual speaker/ translator. Such a lack of a transparent equivalence relation can be assumed to prevent convergence towards source language conventions. This would confirm Teich’s (2003: 211–218) findings that typological characteristics of the languages involved have a clear impact on the kind and degree of source-language induced interference phenomena in translations. Of course, this general hypothesis will profit from further investigations into the question of convergence and divergence of textual conventions in translation.

Acknowledgements This paper has profited from many stimulating discussions we had with Volker Gast while he was visiting scholar at our research centre. Furthermore, we would like to thank the two anonymous reviewers for their helpful comments.

References Antaki, C. & Wetherell, M. 1999. Show concessions. Discourse Studies 1(1): 7–27. Baumgarten, N. 2007. Converging conventions? Macrosyntactic conjunction with English and and German und. Text & Talk 27: 139–170. Baumgarten, N. 2008. Writer construction in English and German popularized academic discourse: The uses of we and wir. Multilingua 27(4): 409–438. Baumgarten, N. & Probst, J. 2004. The interaction of spokenness and writtenness in audience design. In Multilingual Communication [Hamburg Studies on Multilingualism 3], J. House & J. Rehbein (eds), 63–86. Amsterdam: John Benjamins. Baumgarten, N. & Özçetin, D. 2008. Linguistic variation through language contact in translation. In Language Contact and Contact Languages [Hamburg Studies on Multilingualism 7], P. Siemund & N. Kintana (eds), 293–316. Amsterdam: John Benjamins. Baumgarten, N., Meyer, B. & Özçetin, D. 2008. Explicitness in translation and interpreting: A critical review and some empirical evidence (of an elusive concept). Across Languages and Cultures 9(2): 177–203. Becher, V. 2007. Die pragmatischen Funktionen von BUT und ABER im Vergleich. Eine Untersuchung anhand von englischen populärwissenschaftlichen Texten, ihrer deutschen Übersetzung und deutschen Vergleichstexten. MA thesis, University of Hamburg. Bell, D. 2007. Sentence-initial And and But in Academic writing. Pragmatics 17(2): 183–201.



Convergence and divergence through language contact in translation 

Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. 1999. Longman Grammar of Spoken and Written English. London: Pearson Education. Bublitz, W. 2003. Emotive Prosody: How attitudinal frames help construct context. In Anglistentag 2002. Proceedings, E. Mengel, H.-J. Schmidt & M. Steppat (eds), 381–391. Trier: WVT. Bührig, K. & House, J. 2004. Connectivity in translation: Transitions from orality to literacy. In Multilingual Communication, J. House & J. Rehbein (eds), 87–114. Amsterdam: John Benjamins. Bührig, K. & House, J. 2007. Linking constructions in discourse across languages. In Connectivity in Grammar and Discourse, J. Rehbein, C. Hohenstein & L. Pietsch (eds), 345–366. Amsterdam: John Benjamins. Büring, D. 1999. Topic. In Focus – Linguistic, Cognitive, and Computational Perspectives, P. Bosch & R. van der Sandt (eds), 142–165. Cambridge: CUP. Büring, D. 2006. Intonation und Informationsstruktur. In Text – Verstehen. Grammatik und darüber hinaus, H. Blühdorn, E. Breindl & U.H. Waßner (eds). Berlin: de Gruyter. Byrnes, H. 1986. Interactional style in German and English conversations. Text 6: 89–106. Clyne, M. 1987. Cultural differences in the organization of academic texts. Journal of Pragmatics 11: 211–47. Clyne, M. 1994. Inter-Cultural Communication at Work: Discourse Structures across Cultures. Cambridge: CUP. Coates, J. 1995. The expression of root and epistemic possibility in English. In Modality in Grammar and Discourse, J. Bybee & S. Fleischmann (eds), 55–66. Amsterdam: John Benjamins. Crismore, A. & Farnsworth, R. 1990. Metadiscourse in popular and professional science discourse. In The Writing Scholar: Studies in Academic Discourse, W. Nash (ed.), 118–136. Newbury Park CA: Sage. Denison, D. 1993/2004. English Historical Syntax: Verbal Constructions. London: Longman. Diewald, G. 1999. Die Modalverben im Deutschen. Grammatikalisierung und Polyfunktionalität. Tübingen: Niemeyer. Doherty, M. 2001. Discourse relators and the beginnings of sentences in English and German. Languages in Contrast 3 (2): 223–251. Doherty, M. 2005. Topic-worthiness in German and English. Linguistics 43: 181–206. Fabricius-Hansen, C. 1998. Informational density and translation, with special reference to German – Norwegian – English. In Corpora and Cross-Linguistic Research: Theory, Method, and Case Studies, S. Johansson & S. Oksefjell (eds), 197–234. Amsterdam: Rodopi. Geyken, A. 2007. The DWDS corpus: A reference corpus for the German language of the 20th century. In Idioms and Collocations: Corpus-Based Linguistic and Lexicographic Studies, C. Fellbaum (ed.), 23–40. Birmingham: Continuum. Graefen, G. 2007. ‘Hedging’ als neue Kategorie? Ein Beitrag zur Diskussion. . (accessed March 2008) Grote, B., Lenke, N. & Stede, M. 1997. Ma(r)king concessions in English and German. Discourse Processes 24: 87–118. Gumperz, J.J., Kaltman, H. & O’Connor, M.C. 1984. Cohesion in spoken and written discourse: ethnic style and the transition to literacy. In Coherence in Spoken and Written Discourse, D. Tannen (ed.), 3–20. Norwood NJ: Ablex. Halliday, M.A.K. & Matthiessen, C.M.I.M. 2004. An Introduction to Functional Grammar, 3rd edn. London: Arnold. Hawkins, J. 1986. A Comparative Typology of English and German: Unifying the Contrasts. London: Croom Helm. Heine, B. & Kuteva, T. 2005. Language Contact and Grammatical Change. Cambridge: CUP.

 Viktor Becher, Juliane House & Svenja Kranich Hoey, M. 2001. Textual Interaction: An Introduction to Written Discourse Analysis. London: Routledge. House, J. 1977. Translation Quality Assessment. A Model for Analyis. Tübingen: Narr. House, J. 1996. Contrastive discourse analysis and misunderstanding: The case of German and English. In Contrastive Sociolinguistics, M. Hellinger & U. Ammon (eds), 345–361. Berlin: Mouton de Gruyter. House, J. 1997. Translation Quality Assessment. A Model Re-visited. Tübingen: Narr. House, J. 2004a. Explicitness in discourse across languages. In Neue Perspektiven in der Übersetzungs- und Dolmetschwissenschaft, W. Koller, J. House & K. Schubert (eds), 185–208. Bochum: AKS. House, J. 2004b. Linguistic aspects of the translation of children’s books. In Übersetzung –Translation – Traduction: An International Encyclopedia of Translation Studies, H. Kittel, A.P. Frank & N. Greiner (eds), 683–697. Berlin: de Gruyter. House, J. 2005. Politeness in Germany – Politeness in Germany?. In Politeness in Europe, L. Hickey & M. Stewart (eds), 13–29. Clevedon: Multilingual Matters. House, J. 2006. Communicative styles in English and German. European Journal of English Studies 10(3): 249–267. House, J. 2008. Beyond intervention: Universals in translation? trans-kom. Zeitschrift für Translationswissenschaft und Fachkommunikation 1(1): 6–19. Hunston, S. & Thompson, G. (eds). 2000. Evaluation in Text: Authorial Stance and the Construction of Discourse. Oxford: OUP. Hyland, K. 1998. Boosting, hedging, and the negotiation of academic knowledge. Text 18: 349–382. König, E. 2001 Kontrastive Analysen Deutsch-Englisch: Eine Übersicht. In Deutsch als Fremdsprache: Ein internationales Handbuch, Vol. I, G. Helbig et al. (eds), 324–330. Berlin: de Gruyter. Kranich, S. 2009. Epistemic modality in English popular scientific texts and their German translations. trans-kom. Zeitschrift für Translationswissenschaft und Fachkommunikation 2(1): 26–41. Kotthoff, H. 1989. Pro und Kontra in der Fremdsprache. Frankfurt: Peter Lang. Lyons, J. 1977. Semantics. Cambridge: CUP. Nehls, D. 1989. German modal particles rendered by English auxiliary verbs. In Sprechen mit Partikeln, H. Weydt (ed.), 282–292. Berlin: de Gruyter. Özçetin, D. 2008. Die Versprachlichung mentaler Prozesse in englischen und deutschen Wirtschaftstexten. Arbeiten zur Mehrsprachigkeit/Working Papers in Multilingualism 88. (Hamburg: Sonderforschungsbereich Mehrsprachigkeit). Quirk, R., Greenbaum, S., Leech, G. & Svartvik, J. 1985. A Comprehensive Grammar of the English Language. London: Longman. Schiffrin, D. 1987. Discourse Markers. Cambridge: CUP. Schlobinski, P. 1992. Funktionale Grammatik und Sprachbeschreibung: Eine Untersuchung zum gesprochenen Deutsch sowie zum Chinesischen. Opladen: Westdeutscher Verlag. Smith, C. 2003. Modes of Discourse. Cambridge: CUP. Teich, E. 2003. Cross-linguistic Variation in System and Text. A Methodology for the Investigation of Translations and Comparable texts. Berlin: de Gruyter. Thompson, G. 2001. Interaction in academic writing: Learning to argue with the reader. Applied Linguistics 22(1): 58–78. Thompson, G. & Thetela, P. 1995. The sound of one hand clapping: The management of interaction in written discourse. Text 15(1): 103–127. Weinreich, U. 1953. Languages in Contact. The Hague: Mouton.



Convergence and divergence through language contact in translation 

White, P.R.R. 2003. Beyond modality and hedging: A dialogic view of the language of intersubjective stance. Text 23: 259–284. White, P.R.R. & Sano, M. 2006. Dialogistic positions and anticipated audiences: A framework for stylistic comparisons. In Pragmatic Markers in Contrast, K. Aijmer & A.-M. Simon-Vandenbergen (eds), 189–214. Amsterdam: Elsevier. Winter, E. 1994. Clause relations as information structure: Two basic text structures in English. In Advances in Written Text Analysis, M. Coulthard (ed.), 46–68. London: Routledge. Zifonun, G., Hoffmann, L. & Strecker, B. 1997. Grammatik der deutschen Sprache, Vol. 3. Berlin: de Gruyter.

On the importance of spontaneous speech innovations in language contact situations Robert E. Vann

Western Michigan University This paper builds on previous research by offering new perspectives on the importance of spontaneous speech innovation in linguistic, social, and cultural processes of language contact and change in Spanish-speaking communities of Spain and America. The author reconsiders existing data concerning such innovation and synthesizes previous independent findings in a novel, unified approach that relates previously unrelated but functionally similar linguistic phenomena. With the purpose of moving toward a more comprehensive and verifiable model of language contact and change, the paper underlines some complex linguistic and extralinguistic functions associated with spontaneous speech innovations in contact varieties and outlines a coordinated program of interdisciplinary sociolinguistic research to establish appropriate archival tools for documenting corpus-based spontaneous speech innovations for posterior analysis.

1.  Introduction For over half a century now, at least formally since Weinreich (1953), sociolinguists have concerned themselves with understanding and explaining the fundamental issues involved in language contact and language change. One of these issues concerns linguistic innovations that occur in language contact situations. The metalinguistic nomenclature in vogue to classify such innovations has come and gone throughout the years (cf., e.g., interference, transfer, transcodic markers, etc.), yet the basic facts of the matter remain mostly the same. Two such facts are that contact-related linguistic innovations occur both spontaneously and systemically, and usage of such innovations may be restricted or widespread. While the importance of systemic linguistic innovations seems both obvious and inherently sexy in research on language contact and change situations, the relative importance of spontaneous speech innovations may at first glance appear less so. Nevertheless, despite their lower profile, spontaneous speech innovations can be very important

 Robert E. Vann

in linguistic, social, and cultural processes of language contact and change as well, as discussed below. This chapter offers new perspectives on the linguistic, sociolinguistic, and ethnographic importance of spontaneous speech innovations in language contact situations, citing evidence from Spanish. The main purpose of this investigation is not to analyze new linguistic data but rather to reconsider existing data and synthesize previous independent findings in a more unified approach that relates previously unrelated but functionally similar linguistic phenomena.1 Following a status quaestionis in Section 2 that underlines the crucial roles that spontaneous linguistic innovation can play in language use and language change, the chapter continues in Section 3 with a discussion of the need for stronger emphases on the documentation and subsequent analysis of spontaneous speech innovation in research on language contact and change. A summary and conclusion are given in Section 4, along with an explicit call for the establishment of a freely accessible digital archive of spoken language corpora dedicated to documenting contact dialects in various languages for the purposes of linguistic and social analysis. Such an archive is a critical step in the process of moving toward a more comprehensive and verifiable model of language contact and change.

2. The importance of spontaneous speech innovation in language contact situations This section explains how spontaneous speech innovation in language contact situations can and does have linguistic, sociolinguistic, and ethnographic importance. Spontaneous speech innovation in language contact situations is important at the linguistic level if and when such innovation leads to a widespread, systemic language change. Spontaneous speech innovation in language contact situations is of sociolinguistic import when, for example, such innovation serves as a pragmatic resource in discourse strategies that perform extralinguistic functions. Spontaneous speech innovation in language contact situations has ethnographic import when, for example, ways of speaking that exploit such innovation become constitutive of culture-specific interactional styles. These scenarios are respectively

1.  An earlier version of this paper was given as a keynote address in October, 2007, at the Colloquium on Convergence and Divergence in Language Contact Situations, sponsored by The German Science Foundation’s Research Center on Multilingualism, University of Hamburg, Germany. I would like to thank an anonymous reviewer for the idea to state the nature of this paper clearly and transparently in its introduction.



Spontaneous speech innovations in contact situations 

discussed and illustrated below with examples from published analyses of distinct varieties of Spanish in contact with other languages. 2.1 Spontaneous speech innovation as the source of widespread systemic language change in Spanish Throughout history, the Spanish language has on many occasions been in frequent contact with other languages, some Romance (e.g., Catalan, Galician-Portuguese, etc.), others distinctly not Romance (e.g., Basque, Arabic, various Amerindian languages, English, etc.). During many periods, language contact situations involving Spanish in Europe and America have led to language changes. Such changes have occurred and will continue to occur in contact situations when innovations propagate and become accepted by a large enough portion of society. The question is, then, has widespread systemic language change in Spanish ever been attributed to spontaneous speech innovation in a contact situation? Though normally not documented as such, the answer is yes. Weinreich (1953:11) told us that interference in speech amounts to spontaneous speech innovation among bilinguals, given that it “occurs anew in the utterances of the bilingual speaker as a result of his personal knowledge of the other tongue” whereas interference in language, “having frequently occurred in the speech of bilinguals”, becomes habitualized and established and thus “is no longer dependent on bilingualism”. How frequent must a spontaneous speech innovation be in order for it to qualify as a systemic language change? As Blas Arroyo and Tricker have noted (2000:107), in the analysis and characterization of language contact phenomena, identifying interference phenomena based on frequency criteria can be problematic. Due to the lack of lexical predictability in discourse, the low frequency of a particular innovation in conversation does not necessarily mean that it is “spontaneous speech interference” (Wesch 2002:1862). The innovation could easily be established, integrated, recognized, accepted, and even used by many people though its appearance may simply be limited. Weinreich himself even stipulated that interference in language could be determined from a single usage, nuancing his earlier definition: When a speaker of language X uses a form of foreign origin not as an on-thespot borrowing from language Y, but because he has heard it used by others in X-utterances, then this borrowed element can be considered, from the descriptive standpoint, to have become a part of language X. (1953:11, italics added)

At this point it becomes relevant and necessary to expand on the notion of “used by others” in the above quote from Weinreich. On the one hand, spontaneous speech innovations among bilinguals can lead to changes in the linguistic norms of monolinguals who participate in the same speech community. As

 Robert E. Vann

Romaine has reminded us (1995:51), over time “the effects of interference in a bilingual speech community can be cumulative, and lead to new norms, which are different from those observed by monolinguals who use the languages elsewhere”. On the other hand, these new norms can become more widespread, reaching even monolinguals outside the speech community in which they began. This second form of language change is more rare, yet still productive. As Thomason (2003) has stated, “any feature that can appear in a single person’s speech at any time— e.g. in speech errors caused by fatigue or drunkenness or mere carelessness—can turn into a permanent change in the entire language”. Let us consider now a few examples of spontaneous speech innovations in Spanish that, having started in the speech of bilinguals, have successfully “crossed over” to the speech of monolinguals outside the bilingual areas, thus constituting widespread systemic language change in Spanish. Such innovations have started spontaneously in the speech of bilinguals in Spanish-speaking communities in both Spain and America. 2.1.1  Contact-induced language change in Spain Analyses of contact-induced language change permeate descriptions of peninsular Spanish, both historical (Lapesa 1991) and dialectological (Zamora Vicente 1967). Let us consider how such changes have started in the speech of bilinguals and have successfully crossed over to the speech of monolinguals outside the bilingual areas, thus constituting widespread systemic language changes in Spanish. 2.1.1.1 Changes due to adstratum influence during the history of the Spanish ­language Historical linguistics has attributed many widespread systemic language changes in Spanish to substratum, superstratum, and adstratum influences. Indeed, as Thomason (2003: 687) has stated, “in a sense, most of what historical linguists study under the designation ‘language change’ is due to contact”. One of the most well known such arguments in Spanish involves the historical change of word initial /f/ in Latin to /h/ in Castilian Spanish following its intensive contact with Basque. The traditional explanation for the change relies on the influence of the Basque adstratum (Menendez Pidal 1962), given that Basque is thought to have lacked the voiceless labiodental fricative [f]. There was likely close linguistic contact between Basque and protoCastilian for centuries, so much so that among the first writings in the Spanish language (the so-called Glosas Emilianenses) we find Basque translations as well. Lloyd (1987:219) has asserted that Basque speakers, when speaking the emerging IberoRomance language that would become Castilian Spanish, would have either substituted a voiceless, aspirated bilabial stop [ph] or favored the generalization of a voiceless bilabial fricative [φ] that already existed in protoCastilian as a variant of /f/. In either case, this supposed case of phonological interference would lead to the



Spontaneous speech innovations in contact situations 

reinterpretation of /f/ as /h/ in certain prevocalic contexts. Once established by the 10th century in a bilingual Castile, the new pronunciation [h] was spread throughout Spain during the Reconquest, reaching many speakers outside the original Basque/Castilian speech community. Other systemtic language changes have also been attributed to Basque influence, such as the trinary deictic system that Castilian Spanish developed, etc. The inherent problem, of course, with all historical analyses that have attributed widespread systemic language changes in Spanish to substratum, superstratum, and adstratum influences is that, given the diachronic nature of such investigations, it becomes impossible to observe the role, if any, that spontaneous speech innovation may have played in the genesis and social diffusion of the changes under investigation. Fortunately, recording and archiving technologies developed in the 20th century and refined in the 21st century have enabled synchronic research on Spanish in contact the world over. We can now observe, record, and preserve spoken language data. Thanks to these advances in linguistic methods, spontaneous speech innovations have been documented in spoken language corpora of Spanish in contact with other languages, and linguists can more easily track the social diffusion of language changes over time and space. 2.1.1.2 Spontaneous speech innovations in modern spoken language corpora of Catalan Spanish In the Spanish spoken by bilinguals in Catalonia, for example, many spontaneous innovations have been documented in spoken language corpora over the last two decades (Casanovas Catalá 1996a; 1996b; 2002; 2005; Hernández García 1997; 1998a; 1998b; Sinner 2002a; 2002b; 2004; Vann 1995; 1998a; 1999; 2000; 2001; 2002; 2005; 2009; Vila Pujol & Etxebarria 2005; Wesch 1997; 2002, among others). Many such innovations concern lexical borrowing. Wesch (1997), among others, has documented crossovers in the area of Spanish vocabulary. Examples include lexical borrowings from Catalan that began spontaneously in bilingual speech in the Països Catalans and are now used regularly outside the Catalanspeaking areas of Spain. One of the most well known of such borrowings is chafardero ‘gossiper’. First documented as a spontaneous speech innovation in a spoken language ­corpus of Spanish in Barcelona gathered by Wesch (1997), this borrowing is now used by monolinguals throughout Spain and appears in the 22nd edition of the dictionary of the Spanish Royal Academy (Real Academia Española 2001). At the level of systemic language change, several innovations begun in the spontaneous spoken language of speakers who are bilingual in Catalan and Spanish have successfully crossed over to spoken (and written) monolingual Spanish, both within and outside the Països Catalans. The author’s own work has documented

 Robert E. Vann

systemic language change in several pragmatic systems of Spanish in Catalonia (Vann 1998a; 1998b). These investigations demonstrated that the Spanish spoken in Barcelona has developed innovative norms governing the systemic usage of deictics (e.g., motion verbs venir ‘to come’ and traer ‘to bring’, the demonstrative este ‘this’, and the locative aquí ‘here’) as well as the word order representing information structure (e.g., rhematization). Furthermore, this research also found that the innovative pragmatic norms are not restricted to bilinguals in the language contact zone. In addition, many authors have identified other systemic (grammatical) language changes in the Spanish of Catalonia. Corpus-based studies of Catalan Spanish have observed the following innovative grammatical characteristics, among others: animated direct objects without the preposition a, comparisons with que no ‘than not’, subcategorization of de ‘of/from’ by permitir ‘to permit’, absolute interrogative sentences headed by the particle que ‘that’, omission of definite articles in NPs following todos ‘all’, partitive constructions with de ‘of ’ before AdjP, past perfect tense in place of simple past (preterit), postverbal negative indefinite adjectives with positive values, preverbal double negation, switches in nominal gender, verb paradigm alterations and innovative verb conjugation. Example (1) documents the use of an animated direct object without the preposition a. In contrast, this preposition generally accompanies animated direct objects in monolingual varieties of Peninsular Spanish outside the contact zone of Catalan and Spanish. (1) Abandonó su familia. abandon-pret3s his family ‘He abandoned _ his family.’

(Atienza et al. 1997)

Example (2) documents a comparison with que no ‘than not’. Such usage is the norm in Catalan but less frequent in monolingual varieties of Peninsular Spanish outside the contact zone. (2) Prefiero éste que no el otro.ˉ prefer-pres1s this.one than not the other.one ‘I prefer this one over the other one.’

(CasanovasˉCataláˉ1998)

The innovative spoken language documented in Example (3) illustrates de ‘of / from’ as a transition between two verbs. Such usage is lexically patterned in monolingual varieties of Spanish with verbs such as acabar ‘to finish’ and preocuparse ‘to worry’. In monolingual varieties outside the contact zone, however, permitir ‘to permit’ is not one of the Spanish verbs observed to commonly subcategorize this preposition.



Spontaneous speech innovations in contact situations 

(3)

Noˉ me permitieron no dat1s permit-pret3P

de largarme te lo of leave-inf1S dat2sˉ acc3S



juro. swear-pres1s



‘Theyˉdidn’tˉletˉmeˉleave;ˉIˉswear.’

(Wesch 1997)

Research on spoken language corpora in the Spanish of Catalonia has also documented absolute interrogative sentences headed by the particle que ‘that’ as illustrated in Example (4). Though this particle commonly introduces such sentences in Catalan, such usage does not generally characterize monolingual varieties of Peninsular Spanish. (4) ¿Que me entiendes? that acc1S understand-pres2s ‘Do you understand me?’

(Wesch 1997)

Another innovation that has been documented in the Spanish of Barcelona is the omission of definite articles in NPs following todos ‘all’, as illustrated in Example (5). In contrast, linguistic practice in monolingual varieties of Peninsular Spanish outside the contact zone generally includes the definite article in such situations. (5) Hombres, mujeres, niños, de todas edades. men women children of all ages ‘Men, women, children of all _ ages”.

(Sinner 2004)

Unlike Catalan and French, Peninsular Spanish outside of the Catalan contact zone has a weakly developed partitive system. Nevertheless, in the Spanish of Catalonia, there is documentation of partitive constructions that employ de ‘of ’ before adjective phrases. Partitive reference in Catalan Spanish is illustrated in Example (6) (Atienza et al. 1997). (6) En multitud de personajes extraños incluso en alguno de extranjero.ˉ in multitude of characters strange even in some of foreign ‘Inˉaˉnumberˉofˉstrangeˉcelebrities,ˉevenˉinˉsomeˉofˉforeignˉones.’

An innovative characteristic of Catalan Spanish that the author has personally documented (Vann 2001) concerns the use of past perfect tense in reference to simple past with perfective aspect. Such use, innovative when compared to the preterit tense found in monolingual Peninsular Spanish varieties outside the contact zone, is illustrated in Example (7). (7) Los toros, de pequeño había ido a los toros. the bulls of small have-aux-imp1s go-part to the bulls ‘Bullfights, as a youngster I went to the bullfights.’

 Robert E. Vann

Catalan Spanish also displays norms of negation that are innovative compared to those of monolingual varieties of Peninsular Spanish outside the contact zone. Postverbal negative indefinite adjectives with positive values have been documented in the Spanish of Barcelona, as have cases of preverbal double negation (Atienza et al. 1997). Examples (8) and (9) illustrate such usage. (8) ¿Hay ningún libro que trate de to have.in.existence-pres3s none book that treat-subj-pres3s of

este tema? this theme

‘Is there no book that treats this topic?’ (9) Nadie no lo diría. nobody not acc3S say-cond3s ‘Nobody would say it.’

Switches in nominal gender further attest to the innovations that have led to new linguistic norms in Catalan Spanish. Example (10) illustrates the assignment in ­Catalan Spanish of feminine gender to a noun that, outside the contact zone, remains masculine. (10) Todas las alicates que tenía...ˉ (HernándezˉGarcíaˉ1998b) all-fem-p the-fem-p pliers-fem-p that have-imp1s ‘All the pliers that I had...’

Finally, verb paradigm alteration and innovative verb conjugation have been documented in Catalan Spanish as well (Vann 2001). Some examples of such innovations are given in (11) through (16). (11) Depiende. depend-pres3s ‘It depends.’ (12) Entuviera. understand-subj-past1s ‘I were to understand.’ (13) Lleguese. arrive-subj-past1s ‘I were to arrive.’ (14) Ponimos. put-pret1s ‘We put.’ (15) Ser respuesto. be-inf respond-part ‘To be responded to.’



Spontaneous speech innovations in contact situations 

(16) ¡Ves! go-imperative2s ‘Go!’

Of course there are various ways of using Spanish in Catalonia, in accordance with various sociolinguistic factors (Hernández García 1995; Montolío & Vila 1993; Sinner 2002b; Vann 2009). Though none of the innovations documented in Examples (1) through (16) is found in the spoken language of all Spanish-speaking residents of Barcelona, and though all of the grammatical innovations may be found with much lower frequency in monolingual varieties elsewhere, in Barcelona, these innovations constitute new norms that may appear variably in the spoken language of monolingual Spanish speakers who have grown up in Barcelona or otherwise acquired their Spanish there (including foreign nationals). Given their elevated frequencies, following Klein (1980), innovations such as those presented in (1) through (16) may be considered systemic language changes.2 Spanish speakers in Catalonia do not necessarily need to know Catalan nor be aware of its structures in order to appropriate one of the linguistic traits discussed above. In fact, many monolingual Spanish speakers in Catalonia use these and other, similar forms without realizing that they may be contact-induced (López del Castillo 1984; Montolío & Vila 1993; Vann 1996; 1998a). As far as an example of an innovation that has successfully crossed over to monolingual speech both within and outside the Països Catalans, Seco (1989) has documented a syntactic change in progress, originated in the Spanish of bilinguals in the Països Catalans about fifty years ago, that has been successively penetrating monolingual Spanish throughout the Iberian Peninsula for the last thirty years: the omission of the definite article after partitive Spanish expressions such as ma­yoría de ‘most’, parte de ‘part of ’, por ciento de ‘percent of ’, etc.

2.  Of course, as an anonymous reviewer pointed out, the degree to which changes indicated by innovative frequencies of occurrence may be contact-induced will depend on both linguistic and extralinguistic conditioning factors, and researchers should not conclude contactinduced change based exclusively on relative frequencies of use in one variety vs. another. Indeed, an extreme conservative view might only admit language change as contact-induced when a given phenomenon observable in a contact variety of language A does not occur at all in historically earlier varieties of language A or in contemporary non-contact varieties of language A. The perspective offered in the current investigation is that, when compared to monolingual usage outside the contact zone, elevated frequencies of occurrence of a particular linguistic form or structure among speakers of a contact variety of language A can at times be taken as an artifact of language contact and, indeed, may sometimes eventually lead to systemic change.

 Robert E. Vann

2.1.2  Contact-induced language change in modern Spanish in the Americas In the Spanish spoken in the Americas, historical, dialectological and sociolinguistic accounts have all cited language contact as a cause of systemic change (Munné & Guitart 1982). Historical analyses have held that the most important and definite area in which the Amerindian languages have contributed to Spanish is in the lexicon (Lapesa 1991:556). Indeed, many Amerindian words that entered the Spanish language in America through direct language contact not only were incorporated into Peninsular Spanish but, through Peninsular Spanish, were then passed on to other European languages as well. Nevertheless, Lope Blanch (1968) recognized as early as 1968 that the influence of Amerindian languages was also present in the phonetics, the phonology, the morphology, and the syntax of American Spanish and that the sociolinguistic study of this influence had yet to begin. It was no longer enough to simply conclude that, for example, the raising of mid vowels in Peruvian Spanish was likely due to the influence of Quechua, or that the expression of agents via by-phrases even in active voice in Yucatan Spanish was likely due to the influence of Mayan languages. As late as 1989, Amastae (1989:811) felt it necessary to point out that our understanding of how contact-induced language change occurs was still quite limited beyond lexical borrowing, with little emphasis on underlying grammar. With regard to spontaneous speech innovation in contact situations leading to systemic language change in the Spanish language, Silva-Corvalán has argued, repeatedly and convincingly, that language contact tends to accelerate internallymotivated change but does not introduce radical changes in linguistic structure (Silva-Corvalán 1986; 1990; 1993; 1994; 2001). Her now classic study of estar ‘to be’ in Los Angeles Spanish (Silva-Corvalán 1986) was groundbreaking in terms of illuminating the role that spontaneous speech innovation among bilinguals can play in the process of systemic language change. This study examined the progressive extension of estar ‘to be’ in Los Angeles Spanish into what, in monolingual varieties outside the contact zone, would be the semantic/pragmatic territory of ser ‘to be’, the other copular verb in Spanish. The diffusion of estar was found to be gradual and syntactically context-selective, led by speakers born in the USA and those who immigrated from Mexico before the age of six, supporting the hypothesis (Labov 1972a) that change-related variation observed in apparent time actually reflects stages of diffusion in real time. Upon further analysis of Los Angeles Spanish, Silva-Corvalán documented innovations in clitic placement involving a tendency to favor preverbal position, an aspectual simplification reducing the distinction between preterit and imperfect morphology, omission of the complementizer que ‘that’, SV word order, as well as innovations in subject and indirect object pronoun expression (Silva-Corvalán 1990; 1994; 1996). The only direct transfer observed by Silva-Corvalán (1994) in



Spontaneous speech innovations in contact situations 

her corpus of Los Angeles Spanish, however, involved lexico-syntactic calques, multiple word calques that alter the grammatical features of Spanish with syntactic implications, which she divided into six categories. The first type of lexico-syntactic calque identified by Silva-Corvalán (1994) involves syntactic units that reproduce in Spanish meanings from English, as in Example (17). (17) Para atrás. for behind ‘Back.’

The second type of lexico-syntactic calque identified by Silva-Corvalán (1994) involves multiple-word units that bring about changes in combinational restrictions, as illustrated in Example (18). In contrast to (18), in monolingual varieties of American Spanish outside the contact zone of English, the expression to know how to generally is not observed to include como ‘how’. (18)

Sabía como hablar. know-imp3s how speak-inf ‘He knew how to speak.’

The third type of lexico-syntactic calque identified by Silva-Corvalán (1994) involves innovative use of prepositions in Spanish matching usage in English, as in Example (19). The usage in (19) can be considered innovative because, in American Spanish outside the contact zone, usage of the preposition en is not generally observed in this context. (19) Voy en el jueves. go-pres1s on the Thursday ‘I go on Thursday.’

The fourth type of lexico-syntactic calque identified by Silva-Corvalán (1994) involves subcategorization modifications, as illustrated in Example (20), in which the verb gustar ‘to be pleasing’, normally a psych verb, is used as a transitive verb in active voice. (20) Gusta golf mucho. like-pres3s golf a.lot ‘He likes golf a lot.’

The fifth type of lexico-syntactic calque identified by Silva-Corvalán (1994) involves pragmatically neutral word orders from English reproduced with Spanish lexicon. This situation is illustrated in Example (21), in which the prenominal adjective is presented as ostensibly unmarked. Such syntax in Spanish is innovative and uncommon in American Spanish outside contact zones of English.

 Robert E. Vann

(21) Hablaba machucado español. speak-imp3s chop-part Spanish ‘She spoke chopped up Spanish.’

The final type of lexico-syntactic calque identified by Silva-Corvalán involves relexification. An example of relexification is given in (22). (22) Eso es porque DEMS be-pres3s because ‘That is why.’

Silva-Corvalán (1994) considered the lexico-syntactic calques illustrated in Examples (17) through (22) to be forms of modeling. Her analysis demonstrated a modeling distribution across the bilingual continuum, such that modeling increased across generations as Spanish proficiency decreased. Taken together, her findings led Silva-Corvalán (1994) to propose that contact-induced language change will be noticeable first and foremost in the loss of semantic-pragmatic restraints, the extension of discourse-pragmatic functions, and the increased frequency of parallel structures at the expense of nonparallel structures, especially when exposure to and use of monolingual norms are limited. Thus, Silva-Corvalán’s investigations demonstrated that spontaneous speech innovation in Spanish among bilinguals of Los Angeles has linguistic import insofar as it has favored widespread, systemic language change. The influence of English on Spanish has been mainly indirect, however, and the “change” has involved acceleration of internal processes as almost no new structures have been found to have entered the Spanish language. A more striking case of contact-induced language change can be observed in Andean Spanish. Escobar (1978) has claimed, among others, that prolonged contact with Quechua has led to systemic morphosyntactic changes in Andean Spanish. In more recent investigations Klee (1990; 1996) has revisited this question in light of Silva-Corvalán’s proposals. Klee has documented that the use of the archmorpheme lo neutralizes number and gender in marking the third person direct object even in the speech of monolingual Spanish speakers from the Andean region, due to a widespread, systemic language change induced by contact with Quechua. Furthermore, she has demonstrated quantitatively that OV structures occur with more frequency in Andean Spanish than in non-contact dialects of Spanish. In a clear example of how spontaneous speech innovation in language contact situations is important at the linguistic level, García Tesoro (2005) was able to trace the social diffusion of several systemic language changes in Spanish from spontaneous speech innovations among Mayan/Spanish bilinguals to widespread acceptance and usage of new norms among monolingual Spanish speakers in Guatemala.



Spontaneous speech innovations in contact situations 

Many of the innovations studied concern features common to other varieties of Spanish in contact with Amerindian languages in Peru, Ecuador, Bolivia, Paraguay, and Argentina (Escobar 2000; Godenzzi 1986; 1991; Granda 2002; Palacios 1996; 2000). In an attempt to trace the effects of language contact on language change, García’s investigation specifically examined the Spanish spoken by four groups: (1) monolinguals who do not have contact with bilinguals; (2) monolinguals who live in bilingual environments; (3) symmetrical bilinguals; and (4) instrumental bilinguals. Her data confirmed systematic lack of gender and number agreement in groups (2) through (4), with higher frequencies among bilinguals. Her interpretation was that higher frequencies among bilinguals indicated the likely origin of the change and that systematic usage among monolinguals in group (2) indicated a characteristic of what has now become a regional variety. In addition, García’s data confirmed null object instantiation in all four groups, with the highest frequencies among bilinguals, from which she concluded that this feature began among bilinguals and has crossed over to monolingual Spanish even outside the contact zone. The data García obtained regarding the collapse of gender and number in the accusative clitic system demonstrated that all four groups used the single reduced form lo, with decreasing frequencies from group (4) to group (1), from which we may posit a systemic change in progress, led by bilinguals, that is already reaching monolinguals outside the contact zone. García’s study is significant in several ways. First and foremost, the study demonstrates that morphosyntactic variation and change in Guatemalan Spanish are socially determined by degree of bilingualism. Not surprisingly, García also found social class to be a determining factor as well. In a related fashion, the recent process of industrialization was cited as a major social change that has led to increased contact between speakers of Mayan varieties and speakers of Spanish. In this regard, the social mechanisms of systemic language change in Spanish at work in bilingual Guatemala are not unlike mechanisms that have been previously identified in monolingual Spanish-speaking communities (cf., e.g., Holmquist 1985). The aforementioned studies have detailed systemic, contact-induced language changes reaching monolinguals in major subsystems of the Spanish language, and they have traced these changes to spontaneous speech innovation among bilinguals. Spontaneous speech innovation in these language contact situations is therefore important at the linguistic level. Consequently, theories of language variation and change involving languages in contact should include spontaneous speech innovation in contact dialect situations as a potential source for widespread systemic language changes. Significantly, research on languages in contact is not equivalent to research on contact dialects.

 Robert E. Vann

2.1.3 The social mechanisms of widespread systemic language change resulting from spontaneous speech innovation in Spanish During the middle third of the 20th century, much work in bilingualism was focused on the linguistic mechanisms of contact-induced language change. In addition to Weinreich (1953), others tried to determine the extent of linguistic transfer that is possible while along the way classifying different types of linguistic influence (Clyne 1967; Gumperz & Wilson 1971; Haugen 1956; Mackey 1968). Thomason and Kaufman (1988) was one of the first works to focus on the social mechanisms of contact-induced language change. Thomason and Kaufman’s theoretically fundamental, discrete distinction between language shift and maintenance has been justly criticized for its lack of practical applicability in the field (cf., e.g., Romaine 1995). Furthermore, their idea that substratum interference results from imperfect group learning during language shift disregards the potential social significance attached to innovative linguistic norms that may arise during language contact and the role that such significance might play in motivating linguistic transfer (Vann 1996:179). Nevertheless, Thomason and Kaufman (1988) catalyzed further research into the social mechanisms of linguistic change in areas of language contact. After 20 years, Thomason and Kaufman’s claim that, “it is the sociolinguistic history of the speakers, and not the structure of their language, that is the primary determinant of the linguistic outcome of language contact” (1988:35) is not without its problems, yet the examples discussed above illustrate the importance of social factors in diffusing contact-induced language changes that began among bilinguals and have now crossed over to monolingual speech. While it is now clear that the motivations for contact-induced language change may, at times, be language internal, it appears to be even clearer that the mechanisms by which spontaneous speech innovations may become widespread, systemic, contact-induced language changes are at least linguistic, psychological, and social in nature. Thomason (1997:181) has described six different social mechanisms through which features may be transferred from one language to another: (1) intrasentential code-switching, (2) learner error transfer in SLA, (3) code alternation, (4) passive linguistic knowledge, (5) accommodation, and (6) change brought about by conscious and deliberate decisions on the part of speakers. In regard to “any change in a language whose entire or partial source lies in linguistic properties of another language”, clearly Thomason views interference in speech and interference in language as opposing ends of the same continuum. Mechanisms and social processes that mediate this continuum must allow for linguistic creativity, language processing, and individual behaviors. Furthermore, they must account for speech errors and nonce borrowings alongside permanent features of a language. As Thomason (2003:687) has noted, the process



Spontaneous speech innovations in contact situations 

of spread is a function of contact between speakers who employ a series of negotiation mechanisms in simply trying to make themselves understood. It is worth noting, in relation to larger issues of contact-induced language change, that this process would at times seem to parallel, at least in part, the process of spread involved in language shift. Indeed, in her monograph on Spanish and Guarani in Paraguay, Solé (2001:85) has commented that, at a basic level, language choice in bilingual areas can also be a matter of linguistic convergence and divergence. She notes that linguistic tensions can lead to mutual linguistic adaptations as bilingual speakers try to express their needs and desires without incurring the feelings of insecurity, ineptitude, and shame that may be associated with inappropriate code selection. The research discussed in Sections 2.1.1 and 2.1.2 has demonstrated that, among others, the following social variables may be involved in the spread of spontaneous speech innovations to and among monolinguals in language contact situations: (1) length of time living in a language contact zone, (2) relative linguistic exposure, and (3) socioeconomic status. Furthermore, migration patterns, language identities, language ideologies, and the symbolic capital associated with certain ways of speaking may also play a role in the spread of such innovations to and among monolinguals. In this regard research on Spanish in contact supports Weinreich, Labov, and Herzog’s (1968) theory of sociolinguistic change. To better understand the ways in which spontaneous speech innovations in language contact situations can be socially distributed and diffused, the next section examines some ways in which these innovations may have sociolinguistic and ethnographic import, citing further evidence from Spanish. 2.2 The sociolinguistic and ethnographic import of spontaneous speech innovation in Peninsular and American dialects of Spanish in contact with other languages Spontaneous speech innovation in language contact situations is of sociolinguistic import when, for example, such innovation serves as a pragmatic resource in discourse strategies that perform extralinguistic functions. Spontaneous speech innovation in language contact situations has ethnographic import when, for example, ways of speaking that exploit such innovation become constitutive of culture-specific interactional styles. Sociolinguistic and ethnographic import have been shown to favor the diffusion of spontaneous speech innovations in language contact situations resulting in linguistic change from above as well as linguistic change from below. The research discussed in this section presents cases involving the Spanish language in contact situations in both Spain and America.

 Robert E. Vann

2.2.1 Sociolinguistic/ethnographic import of spontaneous speech innovation in Catalan Spanish The sociolinguistic and ethnographic import of spontaneous speech innovations in the Spanish of Catalonia have been documented on multiple occasions. Sinner (2004:599) claims that many in Catalonia perceive the construction of ethnic groups through discourse strategies that correspond to the Spanish spoken by Catalans (the dominant class),3 the Spanish spoken by immigrants in Catalonia (the dominated class), and Castilian Spanish. He claims that Spanish speakers in Catalonia manifest their affiliations, or lack thereof, with one or more of these groups through their ways of speaking. Of course, such ways of speaking can and do include spontaneous speech innovation, even to the degree that some innovations may come to characterize particular ways of speaking. The author’s own research (Vann 1995; 1996; 1998a; 2000) has demonstrated that, in particular, spontaneous innovative deictic usage is a linguistic stereotype that can be reinterpreted as a functional linguistic resource to (1) practice the Catalan habitus; (2) communicate the symbolic capital of Catalan identity; and/ or (3) perform Catalanism or Catalanist discourse. This usage may be socially significant in terms of a speaker’s intentional discourse strategy or in terms of an interlocutor’s interpretation of meaning with or without speaker awareness that the usage may be in any way innovative or the result of transfer. Furthermore, the author has argued that such usage is regulated by communicative competence such that understanding of extralinguistic messages that may be associated with innovative usage of Spanish deictics is culturally and pragmatically bound. Given its sociolinguistic and ethnographic correlates, the innovative use of the deictics may be spread by certain individuals who mark themselves as members of a certain speech community through shared linguistic practices. Indeed, Thomason (2003) has claimed that a sizeable body of evidence points to the possibility of large-scale changes that result from speech communities’ deliberate manipulation of linguistic resources.

3.  I would like to thank an anonymous reviewer for pointing out the fact that the definition of the term Catalans as it is used here by Sinner (2004) is not unproblematic. For example, it is not clear whether Sinner’s usage refers to Catalans whose first language is Catalan or Spanish. Of course, these are two different sociolinguistic groups with relevant ethnographic differences. Presumably, from the context, Sinner’s use of the term in contraposition to immigrants in Catalonia refers more to birthplace and descent than to L1. These and other building blocks of Catalan identity are discussed further in Vann (1995). In any case, the reason for citing Sinner’s research here is that his claim (regarding Spanish discourse strategies as markers of membership in distinct social and ethnic groups) supports the idea that spontaneous speech innovations in the Spanish of Catalonia have both sociolinguistic and ethnographic import.



Spontaneous speech innovations in contact situations 

In other publications (Koike et al. 2001; Vann 2007), the author has argued that ways of speaking Spanish in Catalonia include additional spontaneous speech innovations as pragmatic resources in discourse strategies that perform extralinguistic functions. These ways of “doing Catalan Spanish”, available to bilinguals as well as monolinguals, make use of bilingual simultaneity (Woolard 1998) and conventionalized indirectness (Blum-Kulka et al. 1989) in discourse strategies to construct cultural messages of self and Other, expressing Catalan culture in Spanish language discourse. Two or more speakers may do Catalan Spanish together to co-construct common ground and, given audience design, to construct crosscultural messages. Turell (2001:20) has documented ethnographically the use of contact phenomena as pragmatic resources with which communicative strategies are formed. Indeed, one important thing that Spanish speakers in Barcelona do through ways of speaking that exploit bilingual simultaneity is respect and reflect the different components of the speech event: (1) the setting, (2) the participants and their identities, (3) the topics, (4) the goals and agendas of each participant, etc. More specifically, through spontaneous innovations in spoken language that make use of bivalent discourse markers such as entofuido (Vann 2007; 2008), bilinguals begin the co-construction of certain acts of identity (Le Page & Tabouret-Keller 1985) by literally searching for common ground between two languages and two cultures. For a moment they are speaking both languages simultaneously, subtly inviting their interlocutor to respond to a translingual stimulus in anticipation of potential phatic benefits. The perlocutionary effects of such innovative speech acts and the negotiation of their linguistic and cultural meaning depend on pragmatic competence and correspond to norms that, though available to bilinguals and monolinguals alike, are likely particular to the speakers in the contact zone. Within the contact zone, the potential for sociocultural benefits may have led to a high incidence of bilingual simultaneity even among monolinguals. Through a spontaneous innovation involving conventionalized indirectness, ways of speaking Spanish in Catalan speaking territories of Spain now include a query preparatory indirectness strategy that is used to soften requests, lessening the face threatening acts (FTAs) inherent in them through the sociopragmatic device ¿o qué? ‘or what?’ (Vann 2007). This device ostensively communicates the speaker’s lack of a presupposition concerning a preferred positive response. Playing on the duality of absolute interrogatives, the query preparatory inquiry makes apparent the possibility that a request may be merely a question and not a directive. In opening the door to alternative interpretations of the interrogative this sociopragmatic device, conventionalized in the Spanish of Catalonia, serves to mitigate the FTA inherent in the request. Thus, for interlocutors in on the

 Robert E. Vann

convention, use of the query preparatory expression ¿o qué? is a conventionally indirect redressive act meant to counter a simultaneous FTA. Given that politeness strategies are generally culturally bound, outsiders to this cultural convention such as native speakers of other dialects of Spanish will likely only discern (literal) utterance meanings of the question containing the query preparatory expression ¿o qué? as opposed to the actual speaker’s meaning of a polite request. The extent of the diffusion of this strategy in nonCatalan-speaking territories of Spain is not currently known. Within the contact zone the innovation clearly has both sociolinguistic and ethnographic import and its usage is available to bilinguals and monolinguals alike. 2.2.2 Sociolinguistic/ethnographic import of spontaneous speech innovation in the Americas Spontaneous speech innovations in the contact dialects of Spanish spoken in the Americas have been analyzed as sociolinguistically and ethnographically important on multiple occasions. Indeed, the sociolinguistic and ethnographic import of such innovations has been cited in the process of diffusion of spontaneous speech innovations that has resulted in linguistic change in these varieties of Spanish. In regards to spontaneous speech innovation amongst Spanish speakers in the US, Zentella (1997:83) has commented that speakers not only anticipate benefits and select among appropriate ways to achieve their desired goals; they also monitor the reactions of their interlocutors in relation to the anticipated benefits. Regarding the ethnographic importance of contact-induced language change in America, Silva-Corvalán (2001:277) has commented that linguistic contact is in truth one more expression of cultural contact and transfer reflects a process of cultural diffusion or acculturation. Silva-Corvalán’s research on grammatical changes in progress in the Spanish of Los Angeles (1994) underlines the sociolinguistic and ethnographic import of spontaneous speech innovation in language contact situations. Silva-Corvalán concluded in this regard that spontaneous spoken language innovations, particularly involving lexico-syntactic calques, serve as pragmatic resources in clear discourse strategies (p. 185). Furthermore, she found that innovations increased across generations along the bilingual continuum as positive attitudes towards the Spanish language and commitment to maintaining ancestors’ language and culture decreased (pp. 210–212). These findings suggest that ways of speaking Spanish in Los Angeles that exploit spontaneous speech innovation of the kind observed by Silva-Corvalán (1994) may be in the process of becoming constitutive of a culturespecific interactional style, specifically, a style that emphasizes the differentness of Spanish speakers in Los Angeles vis a vis their mostly Mexican ancestors.



Spontaneous speech innovations in contact situations 

Klee has found (1996) that college-educated professionals in Calca, Peru, who are bilingual in Spanish and Quechua have developed systemic norms for the use of third person object pronouns lo ‘it (masc)’, la ‘it (fem)’, los ‘them (masc)’, and las ‘them (fem)’ that differ from those found in monolingual zones of Peru. Usage of these bilingual norms is variable and can be considered a spontaneous speech innovation with ethnographic import that may be leading a change from above, given Klee’s contention that such uses “reflect the sociolect of those who set the cultural and linguistic standards for the city” (p. 79). The innovation likely has sociolinguistic import favoring social diffusion as well, given Klee’s statement that “the stages of second language acquisition are reflected in the clitic pronoun system of the Spanish of the region, even in the speech of those who learned Spanish as a first language” (p. 79). Going back to García’s research in Guatemala (2005), we recall that systemic linguistic change in Spanish originated in bilingual groups and spread from there to monolinguals; however, variable usage decreased with increased social and cultural distance from Mayan varieties and peoples. These findings suggest that ways of speaking that are characterized by the spontaneous speech innovations studied by García and described above may have become constitutive of a culturespecific interactional style among the indigenous population that enjoys some covert prestige among monolinguals who live in bilingual environments. Furthermore, given that Mayan varieties are low varieties in this bilingually diglossic community, this ethnographic import may now be playing a role in leading this change from below. In describing and analyzing the patterns of use of languages and dialects within specific cultures, we uncover the ways in which speakers draw upon the resources of their language to perform certain functions (Labov 1972b). As Thomason (2003) has stated, speakers’ creativity in the domain of contact-induced change is the main reason for the claim that anything goes. When spontaneous speech innovations serve as pragmatic resources of the Spanish language to perform the function of constructing and expressing culture, they essentially serve as ethnolinguistic tropes in discourse strategies whose role may be sociolinguistically significant in understanding language change.

3. The need to document and analyze spontaneous speech innovation in research on language contact and change Given the linguistic, sociolinguistic, and ethnographic importance of spontaneous speech innovation in language contact situations involving Spanish, there

 Robert E. Vann

needs to be a stronger emphasis on the documentation and subsequent analysis of spontaneous speech innovation in research on language contact and change in Spanish and, more generally, in all languages. Thomason (2003:687) has noted that, currently, the likelihood that any historical linguist will become aware of an individual’s spontaneous speech innovations is slim. Crucially, however, she also notes that it is only after other speakers have adopted an individual speaker’s spontaneous innovation that it will typically become part of the database of historical linguistics. In other words, in the past, by the time historical documents began to reflect language changes, typically the origins of such changes in spontaneous spoken language were nowhere to be found. It is in this regard that we have the technology today to change the historical linguistics of tomorrow. 3.1  Digital spoken language corpora Digital spoken language corpora represent our best hope of finding spontaneous speech innovation in contact dialects, as well as our best way to reference it. They can provide a linguistic record of spontaneous spoken language that can be archived indefinitely in nonproprietary, open source formats for preservation and future consultation (Bird & Simons 2002). Corpora can take many forms, including audio recordings, video recordings, transcriptions, field notes, grammars, lexicons, grammatical analyses, ethnographic information, etc. The OLAC Working Group on Linguistic Data Types (Aristar Dry & Johnson 2006) divides all linguistic resources into the following three types of language data: primary texts, lexicons, and language descriptions. For the purposes of investigating the role that spontaneous speech innovations in language contact situations may play in catalyzing widespread, systemic language change, it is essential that researchers have open access to spontaneous spoken language corpora in the form of primary texts and/or language descriptions that provide extensive representations of such primary texts in the form of orthographic transcripts. Digital spoken language corpora facilitate both access to and transcription of primary texts. With access to spoken language corpora in the form of digital primary texts and/or the published transcriptions of such texts, researchers can investigate not only linguistic forms, but pragmatic, sociolinguistic, and ethnographic functions as well, based on linguistic context and extralinguistic metadata. In the fields of Spanish sociolinguistics and dialectology, such thinking has led to more and more corpus-based analyses every day (Briz 2005:7). Regrettably, however, accessible spontaneous spoken language corpora from contact dialects of Spanish remain largely unavailable (Pusch 2002:245) despite



Spontaneous speech innovations in contact situations 

lofty projects such as CREA (“Corpus de Referencia del Español Actual”) and ­PRESEEA (“PRoyecto para el Estudio Sociolingüístico del Español de España y América”). Existing corpora documenting spontaneous speech innovation in contact dialects of Spanish could of course be made more accessible if distributed via the World Wide Web in the form of digital audio and/or video files representing primary data. In practice, however, Web distribution of spontaneous spoken language corpora may often be limited by technological or financial resources or by the need to respect the privacy and anonymity of participants. In such cases, digital corpora still facilitate transcription; moreover, print edition language descriptions that include extensive orthographic transcripts representing spontaneous spoken language corpora offer a highly desirable way to render primary texts both discoverable and accessible (for a recent example of such a publication concerning Spanish, cf. Vann 2009). Without greater access to spontaneous spoken language corpora in the form of digital primary texts and/or the published transcriptions of such texts, researchers in sociolinguistics and dialectology will be forced to continue to work mostly on proprietary corpora to document spontaneous speech innovation in contact dialects. This is true for Spanish as well as for other contact languages. Unfortunately, studies based on proprietary corpora are difficult to verify, expand upon, or refute. Even worse, without expanded access of the sort described above, our scientific knowledge with respect to spontaneous speech innovation in contact dialects will accumulate largely based on studies that cite corpus-based linguistic examples, i.e., secondary sources, rather than (representations of) primary source data, i.e., the corpora themselves. For this reason, published corpora of spontaneous spoken language must be at the center of future research on the documentation and subsequent analysis of spontaneous speech innovation in language contact and change. 3.2  Preserving and accessing digital language data Sadly, spoken language corpora with primary texts that have been appropriately designed for linguistic research of the sort described above are scarce and scattered, not easily discovered or accessed. In today’s day and age, many organizations and resources store and provide digital language data for various purposes, but none does so expressly for the specific purpose of supporting research on spontaneous speech innovation in language contact situations. Some of the more established organizations known to store and provide digital language data are given in Table 1.

 Robert E. Vann

Table 1.  Established organizations that store digital language data and serve these data online Abbreviation

Internet Address and full organization name

AILLA

http://www.ailla.utexas.org/site/welcome.html Archive of the Indigenous Languages of Latin America http://www.uaf.edu/anlc/ Alaska Native Language Center http://www1.aiatsis.gov.au/aseda/specialproj/aseda/index.html Aboriginal Studies Electronic Data Archive http://www.indiana.edu/˜libarchm/index.html Archives of Traditional Music http://www.mip.berkeley.edu/blc/la/ Audio Archive of Linguistic Fieldwork http://www.ime.usp.br/˜tycho/ Comparative Corpus of Spoken Portuguese http://childes.psy.cmu.edu/ CHIld Language Data Exchange System http://childes.psy.cmu.edu Child Language ANalysis http://www.delaman.org/ Digital Endangered Languages And Musics Archive Network http://www.mpi.nl/dobes/ DOkumentation BEdrohter Sprachen http://www.elra.info/ European Language Resources Association http://emeld.org/ Electronic Metastructure for Endangered Languages Data http://lacito.vjf.cnrs.fr/archivage/index.html LAngues & CIvilisations à Tradition Orale http://www.ldc.upenn.edu/ Linguistic Data Consortium http://www.linguistlist.org/ The Linguist list http://www2.fmg.uva.nl/lpca/ Language and Popular Culture in Africa http://www.mpi.nl/world/ Max Planck Institute http://www.language-archives.org Open Language Archive Community http://ota.ahds.ac.uk/ota/ Oxford Text Archive http://paradisec.org.au/ Pacific And Regional Archive for DIgital Sources in Endangered Cultures http://www.rae.es/ Real Academia Española http://www.rosettaproject.org/ The Rosetta project http://accent.gmu.edu/ Speech Accent Archive

ANLC ASEDA ATM AALF CCSP CHILDES CLAN DELAMAN DOBES ELRA E-MELD LACITO LDC Linguist LPCA MPI OLAC OTA PARADISEC RAE Rosetta SAA

(Continued)



Spontaneous speech innovations in contact situations 

Table 1.  Continued Sinica THDL UHLCS

http://www.sinica.edu.tw/ftms-bin/kiwi1/mkiwi.sh?language=1 Academia Sinica Balanced Corpus of Modern Chinese http://www.thdl.org/ Tibetan and Himalayan Digital Library http://www.ling.helsinki.fi/uhlcs/ University of Helsinki Language Corpus Server

The organizations listed in Table 1 store linguistic data online for different purposes, but the kind of data varies, and so does organizational commitment to data preservation (Vann 2006). For example, E-MELD and Rosetta provide exemplary linguistic data, but they do not generally archive materials, while ELRA and LDC may be more dedicated to accumulating and selling electronic linguistic resources than they are to preserving them for posterity. Overall, as discussed in Vann (2006), organizations that store digital language data may vary based on at least the following six factors: (1) the actual languages archived (open, restricted to language families, restricted to indigenous or endangered languages, restricted to a particular language, etc.); (2) the type of language data archived (audio/video recordings, transcriptions, field notes, lexicons, literature, etc.); (3) the type of speech data available, if any (spontaneous conversational speech, elicited speech, interview speech, readings of word lists, etc.); (4) the levels of access available; (5) the fee for service; and (6) audience design and user-friendliness of linguistic archives. The first three criteria fall under archival content, whereas the last three fall under archival access. Many archives now emphasize primary texts of digital audio data where possible (cf., e.g., ATM). Both within and across archives, however, linguistic corpora may be deposited for different purposes, even when the types of language data archived are similar. For example, content in the form of digital audio files is generally deposited with ATM for purposes of academic research in ethnomusicology, anthropology, and linguistics. ELRA, like ATM, also primarily maintains audio corpora; however, much of ELRA’s content is deposited for commercial purposes in language engineering, telecommunications, etc. In this regard, content will affect discoverability, insofar as linguistic resources created for the purposes of academic research might be out of place and overlooked in an archive whose other resources were predominantly deposited for purposes of commercial exploitation. Complicating matters, most linguistic archives do not seem to be organized by the type of spoken language data that they archive, nor do they necessarily present the same spoken language data types. 3.3  Archiving spoken language data from contact dialects What the field needs is the creation of a freely accessible, online digital archive of spoken language corpora openly dedicated to documenting contact dialects in

 Robert E. Vann

various languages. This archive could preserve and serve audiovisual recordings of spontaneous, colloquial spoken language data produced in language contact situations for the purposes of interdisciplinary linguistic research. Ideally, the spoken language data would be produced in peer group conversations, with multiple participants gathered from naturally occurring social groups. The inclusion of corpus metadata with orthographic transcriptions would create a highly valuable documentary resource for subsequently verifiable linguistic and sociolinguistic investigation of contact-related linguistic innovation. Cooperation between academic resource providers and consumers would be key to this venture given that existing digital archives of spoken language corpora often do not contain data and metadata that are appropriate for sociolinguistic research on spontaneous speech innovation in language contact situations. Presumably, diverse and varied approaches to the analysis of spontaneous speech innovation are crucial to a comprehensive understanding of this phenomenon. This archive should provide for various levels of web access free of charge. The search engine should allow linguists to select resources by language and, within each language, by contact language. In other words, one should be able to choose Spanish in a drop down menu of languages, and then one should be able to choose Catalan, Quechua, Mayan languages, etc. in a drop down menu of languages with which Spanish is or has been in contact. Spoken language resources of various types could be organized by genre, style, register, participants, spontaneity, etc. Once resources have been located, users should be able to download both the recordings and the corresponding metadata, including resource citation information, in open formats. To date, no such archive currently exists and, therefore, its creation remains among the desiderata that will most enable and facilitate interdisciplinary research on language contact, spontaneous speech innovation, and language change. 4.  Conclusion This chapter has reflected on the multiple complex functions of spontaneous speech innovation in language contact situations based on published analyses of distinct varieties of Spanish in contact with other languages. The chapter has revealed the linguistic potential of such innovation in terms of language contact and change. Indeed, as demonstrated with examples in Spanish from both Spain and America, spontaneous speech innovation in such language contact situations can and does lead to widespread systemic language change. Furthermore, the chapter has underscored the sociolinguistic and ethnographic potentials of such innovation in terms of pragmatic resources, discourse strategies, ways of speaking, and culture-specific



Spontaneous speech innovations in contact situations 

interactional styles, all of which may well be relevant to the social mechanisms that help to spread linguistic change. Clearly, both language internal and language external analyses are crucial to a comprehensive phenomenological understanding of spontaneous speech innovation and its role in systemic language change. This chapter has also outlined a coordinated program of sociolinguistic research to more appropriately document spontaneous speech innovation in contact dialects for later analysis. One of the best ways to investigate such innovation, synchronically and diachronically, is through the digital preservation of spontaneous spoken language in contact varieties, creating linguistic resources that can be analyzed from multiple complementary perspectives, now and into the future. To house these resources, a call is hereby issued for the establishment of a freely accessible digital archive of spoken language corpora dedicated to documenting contact dialects in various languages for the purposes of linguistic and social analysis. It is essential that researchers of contact dialects publish and archive their corpora, no matter how small, if the international academic community is to make the most of our limited resources. To better identify and understand the many roles and meanings of spontaneous speech innovation in language contact situations we need a closer integration and cross-fertilization of research in contact linguistics, sociolinguistics, linguistic anthropology, and crucially, documentary linguistics, with a dedicated agenda of identifying and understanding the many roles and meanings of spontaneous speech innovation in language contact situations. Only through consideration of both structural and functional aspects of language contact and change can we understand sociolinguistic divergence as a matter of linguistic convergence, as well as matters of linguistic convergence in terms of sociolinguistic divergence.

References Amastae, J. 1989. Contact and bilingualism in the Hispanic world: Reflections on the state of the art. Hispania 72(4): 810–820. Aristar Dry, H. & Johnson, H. 2006. OLAC linguistic data type vocabulary. (October 1, 2007). Atienza, E., Battaner, P., Bel, A., Borrás, L., Díaz, L., Hernández, C., Hurtado, E., López, C. & Martínez, R. 1997. Una tipología de interferencias catalán-castellano a partir de las producciones escritas de los estudiantes universitarios. In Didáctica de la Lengua y la Literatura para una Sociedad Plurilingüe del Siglo XXI, F.J. Cantero, A. Mendoza & C. Romea Castro (eds.), 577–582. Barcelona: Universitat de Barcelona. Bird, S. & Simons, G. 2002. Seven dimensions of portability for language documentation and description. In Proceedings of the Workshop on Portability Issues in Human Language Technologies, Third International Conference on Language Resources and Evaluation, 23–30. Paris: European Language Resources Association.

 Robert E. Vann Blas Arroyo, J.L. & Tricker, D. 2000. Principles of variationism for disambiguating language contact phenomena: The case of lone Spanish nouns in Catalan discourse. Language Variation and Change 12: 103–140. Blum-Kulka, S., House, J. & Kasper, G. 1989. Cross-cultural Pragmatics: Requests and Apologies. Norwood NJ: Alblex. Briz, A. 2005. Los corpus de español hablado: Presentación. Oralia 8: 7–12. Casanovas Catalá, M. 1996a. Algunos rasgos propios del español en las comunidades de habla catalana: Fonética, morfosintaxis, y léxico. Analecta Malacitana 36(1): 149–160. Casanovas Catalá, M. 1996b. El contacto lingüístico en Lleida: Algunas consecuencias en el léxico español de los catalanohablantes. Sintagma 8: 57–63. Casanovas Catalá, M. 1998. Aproximación al estudio de la interferencia lingüística en la lengua escrita. Revista de Filología de la Universidad de La Laguna 16: 9–17. Casanovas Catalá, M. 2002. Modelos de incorporación léxica en un caso de contacto de lenguas: Cuando el español es segunda lengua. Verba 29: 261–289. Casanovas Catalá, M. 2005. Español y Catalán en Contacto: La Expresión Deíctica en el Castellano Hablado en Lleida. Aachen: Shaker Verlag. Clyne, M. 1967. Transference and Triggering. The Hague: Martinus Nijhoff. Corpus de Referencia del Español Actual. (October 12, 2007). Escobar, A. 1978. Variaciones Sociolingüísticas del Castellano en el Perú. Lima: Instituto de Estudios Peruanos. Escobar, A.M. 2000. Contacto Social y Lingüístico: El Español en Contacto con el Quechua en el Perú. Lima: Pontificia Universidad Católica del Perú. García Tesoro, A.I. 2005. Español en contacto con lenguas mayas en Guatemala. In Variedades Lingüísticas y Lenguas en Contacto en el Mundo de Habla Hispana, C. Ferrero & N. Lasso-von Lang (eds.). Bloomington IN: AuthorHouse. Godenzzi, J.C. 1986. Pronombres de objeto directo e indirecto del castellano en Puno. Lexis 10(2): 187–201. Godenzzi, J.C. 1991. Discordancias gramaticales del castellano andino en Perú. Lexis 15(1): 107–118. Granda, G.d. 2002. El noreste argentino, área lingüística andina. In El Indigenismo Americano III, A. Palacios & A.I. García (eds.), 61–81. Valencia: Universitat de València. Gumperz, J.J. & Wilson, R. 1971. Convergence and creolization: A case from the Indo-Aryan/ Dravidian border in India. In Pidginization and Creolization of Languages: Proceedings of a Conference Held at the University of the West Indies, Mona, Jamaica, April, 1968, D. Hymes (ed.), 51–167. Cambridge: CUP. Haugen, E. 1956. Bilingualism in the Americas: A Bibliography and Research Guide. Tuscaloosa AL: University of Alabama. Hernández García, C. 1995. Algunas reflexiones sobre el español en las zonas bilingües: ­Propuesta de variables sociolingüísticas para la obtención de los infomantes del corpus de la variedad de español de Barcelona y su area metropolitana. Anuari de Filologia 18(F6): 87–105. Hernández García, C. 1997. ‘Va, mujer, arréglate y vámonos a cenar’: El va en español, un caso de interferencia del catalán. Espéculo 6. (January 3, 2007). Hernández García, C. 1998a. Anàlisi comparativa del nivell d'interferència lingüística catalàcastellà d’un grup d’universitaris de Barcelona (UPF). In Actes de la Cinquena Trobada de Sociolingüistes Catalans: Barcelona, 24 i 25 d’abril de 1997. Barcelona: Generalitat de Catalunya, Departament de Cultura.



Spontaneous speech innovations in contact situations 

Hernández García, C. 1998b. Una propuesta de clasificación de la interferencia lingüística a partir de dos lenguas en contacto: El catalán y el español. Hesperia: Anuario de Filología Hispánica 1: 61–79. Holmquist, J.C. 1985. Social correlates of a linguistic variable: A study in a Spanish village. Language and Society 14: 191–203. Klee, C. 1990. Spanish-Quechua language contact: The clitic pronoun system in Andean Spanish. Word 41: 35–46. Klee, C. 1996. The Spanish of the Peruvian Andes: The influence of Quechua on Spanish language structure. In Spanish in Contact: Studies in Bilingualism, J.B. Jensen & A. Roca (eds.), 73–91. Somerville MA: Cascadilla. Klein, F. 1980. A quantitative study of syntactic and pragmatic indicators of change in the Spanish of bilinguals in the United States. In Locating Language in Time and Space, W. Labov (ed.), 69–82. New York NY: Academic Press. Koike, D., Vann, R. & Busquets, J. 2001. Spanish no, sí: Reactive moves to perceived face threatening acts, Part II. Journal of Pragmatics 33(6): 879–899. Labov, W. 1972a. Language in the Inner City. Philadelphia PA: University of Pennsylvania. Labov, W. 1972b. Sociolinguistic Patterns. Philadelphia PA: University of Pennsylvania. Lapesa, R. 1991. Historia de la Lengua Española, 9th corrected and augmented edn. Madrid: Gredos. Le Page, R.B. & Tabouret-Keller, A. 1985. Acts of Identity. Cambridge: CUP. Lloyd, P. 1987. From Latin to Spanish. Philadelphia PA: American Philosophical Society. Lope Blanch, J.M. 1968. El Español de América. Madrid: Alcalá. López del Castillo, L. 1984. Llengua Standard i Nivells de Llenguatge, 2nd edn. Barcelona: Laia. Mackey, W.F. 1968. The description of bilingualism. In Readings in the Sociology of Language, J. Fishman (ed.), 554–584. The Hague: Mouton. Menendez Pidal, R. 1962. En Torno a la Lengua Vasca. Buenos Aires: Espasa-Calpe Argentina. Montolío, E. & Vila, M.R. 1993. La enseñanza del español lengua extranjera (E/LE) en una ciudad bilingüe: Barcelona. Anuari de Filologia 16(F4): 89–105. Munné, J.Z. & Guitart, J. 1982. Dialectología Hispanoamericana, 2nd edn. Salamanca: Almar. Palacios, A. 1996. Discordancias pronominales en el español de América. In Actas del XI Congreso de la ALFAL II, 1187–1196. Las Palmas de Gran Canaria: Universidad de Gran Canaria. Palacios, A. 2000. El sistema pronominal del español paraguayo: Un caso de contacto de lenguas. In Teoría y Práctica del Contacto: El Español de América en el Candelero, J. Calvo Pérez (ed.), 123–143. Madrid: Vervuert-Iberoamericana. PRoyecto para el Estudio Sociolingüístico del Español de España y América. October 12, 2007. Pusch, C.D. 2002. A survey of spoken languge corpora in Romance. In Romanistische Korpuslinguistik – Korpora und Gesprochene Sprache / Romance Corpus Linguistics – Corpora and Spoken Language, C.D. Pusch & W. Raible (eds), 245–264. Tübingen: Narr. Real Academia Española. 2001. Diccionario de la Lengua Española, 22nd edn, augmented and updated online to March 23, 2003. (November 10, 2008). Romaine, S. 1995. Bilingualism, 2nd edn. Oxford: Blackwell. Seco, M. 1989. Un catalanismo sintáctico en el español de hoy. In La Corona de Aragón y las Lenguas Románicas, G. Holtus, G. Lüdi & M. Metzeltin (eds), 309–317. Tübingen: Narr. Silva-Corvalán, C. 1986. Bilingualism and language change: The extension of estar in Los Angeles Spanish. Language 62(3): 587–608.

 Robert E. Vann Silva-Corvalán, C. 1990. Current issues in studies of language contact. Hispania 73: 162–176. Silva-Corvalán, C. 1993. On the permeability of grammars: Evidence from Spanish and English in contact. In Linguistic Perspectives on the Romance Languages, W. Ashby, M. Mithun, G. Perissinotto & E. Raposo (eds.), 19–43. Amsterdam: John Benjamins. Silva-Corvalán, C. 1994. Language Contact and Change: Spanish in Los Angeles. Oxford: OUP. Silva-Corvalán, C. 1996. Cambios sintácticos en situaciones de contacto lingüístico. In Actas del X Congreso Internacional de la A.L.F.A.L., M. Arjona Iglesias (ed.), 302–308. Mexico: Universidad Nacional Autónoma de México. Silva-Corvalán, C. 2001. Sociolingüística y Pragmática del Español. Washington DC: Georgetown University. Sinner, C. 2002a. Análisis constrastivo de un corpus oral de diferentes variedades del caste­ llano: Aspectos cuantitativos y cualitativos. In Romanistische Korpuslinguistik – Korpora und Gesprochene Sprache/Romance Corpus Linguistics – Corpora and Spoken Language, C.D. Pusch & W. Raible (eds), 279–292. Tübingen: Narr. Sinner, C. 2002b. The construction of identity and group boundaries in Catalan Spanish. In Us and Others: Social Identities across Languages, Discourses and Cultures, A. Duszak (ed.), 159–185. Amsterdam: John Benjamins. Sinner, C. 2004. El Castellano de Cataluña. Tübingen: Niemeyer. Solé, Y.R. 2001. Valores Comunicativos y Emblemáticos del Español y del Guaraní. Bogotá: Instituto Caro y Cuervo. Thomason, S.G. 1997. On mechanisms of interference. In Language and its Ecology: Essays in Memory of Einar Haugen, S. Eliasson & E. Hakon Jahr (eds.), 181–207. Berlin: de Gruyter. Thomason, S.G. 2003. Contact as a source of language change. In A Handbook of Historical Linguistics, R. Janda & B.D. Joseph (eds.), 687–712. Malden MA: Blackwell. Thomason, S.G. & Kaufman, T. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley CA: University of California. Turell, T. (ed.). 2001. Multilingualism in Spain: Sociolinguistic and Psycholoinguistic Aspects of Linguistic Minority Groups. Clevedon: Multilingual Matters. Vann, R.E. 1995. Constructing Catalanism: Motion verbs, demonstratives, and locatives in the Spanish of Barcelona. Catalan Review 9(2): 253–274. Vann, R.E. 1996. Pragmatic and Cultural Aspects of an Emergent Language Variety: The Construction of Catalan Spanish Deictic Expressions. Ph.D. dissertation. The University of Texas at Austin (University Microfilms No. 9633318). Vann, R.E. 1998a. Aspects of Spanish deictic expressions in Barcelona: A quantitative examination. Language Variation and Change 10: 263–288. Vann, R.E. 1998b. Pragmatic transfer from less developed to more developed systems: Spanish deictic terms in Barcelona. In Romance Linguistics: Theoretical Perspectives, A. Schwegler, B. Tranel & M. Uribe-Etxebarría (eds), 307–317. Amsterdam: John Benjamins. Vann, R.E. 1999. Language exposure in Catalonia: An example of indoctrinating linguistic ­ideology.Word 50(2): 191–209. Vann, R.E. 2000. Constructing reality in bicultural communication: Catalan ways of speaking Spanish. Intercultural Communication Studies 10(1): 113–124. Vann, R.E. 2001. El castellà catalanitzat a Barcelona: Perspectives lingüístiques i culturals. Catalan Review 15(1): 117–131. Vann, R.E. 2002. Linguistic ideology in Spain’s ivory tower: (Not) Analyzing Catalan Spanish. Multilingua 21(2/3): 227–246.



Spontaneous speech innovations in contact situations 

Vann, R.E. 2005. El español de los Países Catalanes. In Variedades Lingüísticas y Lenguas en Contacto en el Mundo de Habla Hispana, C. Ferrero & N. Lasso-Von Lang (eds.), 183–193. Bloomington IN: AuthorHouse. Vann, R.E. 2006. Frustrations of the documentary linguist: The state of the art in digital language archiving and the archive that wasn't. Proceedings of the 2006 E-MELD Workshop on Digital Language Documentation (Tools and Standards: The State of the Art), June 20–22, 2006, Michigan State University, East Lansing, MI. (January 3, 2007). Vann, R.E. 2007. Doing Catalan Spanish: Pragmatic resources and discourse strategies in ways of speaking Spanish in Barcelona. In Selected Proceedings of the Third Workshop on Spanish Sociolinguistics, J. Holmquist, A. Lorenzino & L. Sayahi (eds.), 183–192. Somerville MA: Cascadilla. Vann, R.E. 2008. Una derivación transcódica en el español de los Països Catalans y el papel del conocimiento lingüístico en su génesis. In El Castellano en las Tierras de Habla Catalana, C. Sinner & A. Wesch (eds.), 269–286. Madrid/Frankfurt: Iberoamericana/Vervuert. Vann, R.E. 2009. Materials for the Sociolinguistic Description and Corpus-Based Study of Spanish in Barcelona: Toward a Documentation of Colloquial Spanish in Naturally Occurring Groups. Lewiston NY: Mellen. Vila Pujol, M.R. & Etxebarria, M. 2005. Corpus para el estudio de las interferencias lingüísticas: Los corpus de Barcelona, Lérida y Bilbao. Oralia 8: 213–242. Weinreich, U. 1953. Languages in Contact. The Hague: Mouton. Weinreich, U., Labov, W. & Herzog, M. 1968. Empirical foundations for a theory of languge change. In Directions for Historical Linguistics, W. Lehmann & Y. Malkiel (eds.), 95–195. Austin TX: University of Texas. Wesch, A. 1997. El español hablado de Barcelona y el influjo del catalán. Esbozo de un programa de investigación. Verba 24: 287–312. Wesch, A. 2002. La investigación sobre variedades del español hablado en contacto con el catalán (particularmente en Cataluña y Baleares): Estado de la cuestión y perspectivas para el futuro. In Actas del V Congreso Internacional de Historia de la Lengua Española, M.T. Echenique Elizondo & J. Sánchez Méndez (eds), 1857–1872. Madrid: Gredos. Woolard, K. 1998. Simultaneity and bivalency as strategies in bilingualism. Journal of Linguistic Anthropology 8(1): 3–29. Zamora Vicente, A. 1967. Dialectología Española, 2nd edn. Madrid: Gredos. Zentella, A.C. 1997. The bilingual/multidialectal repertoire of el bloque. In Growing Up Bilingual, 41–55. Malden MA: Blackwell.

part iii

Phonological processes of variation and change in bilingual individuals

Gradient merging of vowels in Barcelona Catalan under the influence of Spanish* Susana Cortés, Conxita Lleó & Ariadna Benet

Hamburg, University and Research Centre on Multilingualism The present article examines the influence of Spanish phonology on the production of Catalan vowel contrasts. Vowels produced by Catalan bilinguals in three different districts in Barcelona are both auditorily and acoustically analyzed in order to find out whether differences in exposure to Catalan and Spanish could trigger differences in vowel pronunciation. Differences in production are found between the groups of children in the different districts. In particular, the case of the Catalan vowel contrast /ε/ vs. /e/ seems to be at an advanced stage of merging. The children in the highly Spanish-influenced district hardly produce this vowel contrast in comparison with children in districts where Catalan is more present in the environment.

1.  Introduction 1.1  Catalan-Spanish contact in Barcelona In Barcelona, the capital city of Catalonia, Catalan coexists intensively with Spanish. The population in this region has been historically bilingual, with varying degrees of dominance by one or the other language from the Middle Ages on. One of the last influential periods in this respect was Franco’s dictatorship (1939–1975), when Catalan was prohibited in public domains. The presence of Spanish in Catalonia, and especially in Barcelona, kept on gaining strength in the second half of the

*This article is part of the research carried out within the project “Phono-prosodic development of Catalan in its current bilingual context”, directed by Conxita Lleó, in the Research Centre on Multilingualism. We would like to express our gratitude to the Deutsche Forschungsgemeinschaft (German Research Foundation) and the University of Hamburg for their support. We would also like to thank all of our subjects who kindly agreed to take part in the study, without whom this research would not have been possible.

 Susana Cortés, Conxita Lleó & Ariadna Benet

20th century through two immigration waves. The first one took place from the 1950s to the 1970s and brought a huge number of non-Catalan speakers from the rest of Spain. One third of the population (33.3%) living in Barcelona in 1981 belonged to this wave (Ajuntament de Barcelona 2007: 73). Catalan was reintroduced at school and in the mass media after Franco’s time, which implied an increase in the linguistic competence and in the number of speakers, including some of the newcomers or their descendents. As an illustration, more than one third of the second generation of this immigration wave coming to Catalonia from the rest of Spain uses Catalan within their newly created families (Pons & Vila 2005: 71). In 2007 the percentage of inhabitants in Barcelona who were born in regions of Spain other than Catalonia had considerably diminished compared to 1981: 21.3% (Ajuntament de Barcelona 2007: 73). This is not only due to the natural evolution of the population, but also to the arrival of the second immigration wave, in this case coming from abroad, which has taken place from the end of the nineties until today. At the beginning of 2008 the population coming from abroad constituted 17.3% of the total population in Barcelona. From this group, 46% came from Latin America (Ajuntament de Barcelona 2007), i.e., they were mostly Spanish speakers. All these facts from the last decades, which affect most strongly a big metropolis like Barcelona, explain why this city and its surroundings are the area in Catalonia with the lowest percentage of population with Catalan as their first language1 (31.91% in 2003; IDESCAT 2006) and, therefore, where Spanish, the other main language of this bilingual society, has an outstanding presence. When two languages are in such a contact situation like Catalan and Spanish in Barcelona, they are likely to influence one another. Catalan seems to be currently changing, which has been attributed to the strong influence of Spanish on the Catalan population. Especially vulnerable areas are those where the languages differ, one of them being more complex than the other. One such area are the Catalan vowels. Catalan2 has a total of eight vowels: /i/, /e/, /ε/, /a/, /f/, /o/ and /u/ in stressed position and /i/, /u/ and schwa [ә] in unstressed position. Spanish has five vowels: /i/, /e/, /a/, /o/ and /u/, all of which make up both the stressed and the unstressed paradigms. The two opening degrees in the Catalan

1.  The data from the Institut d’Estadística de Catalunya (2006) show three categories: people who have Catalan as their “first language (the first language spoken at home)”, as their “own language (the language which the interviewed feels identified with)”, and as their “usual language (the language mostly used by the interviewed)”. People who answered with both languages, Catalan and Spanish, indistinctly were not included in the counting. 2.  We report here only on Central Catalan, the variety spoken in Barcelona. Other dialects present some variation as regards the vocalic system.



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

mid vowels, /ε e/ and /f o/, are the distinctive elements with respect to the Spanish system, as far as the stressed vocalic system is concerned. Thus, Catalan minimal pairs such as ['pε.nә] ‘pear’ and ['pe.nә] ‘Peter’, or ['sfw] ‘salary’ and ['sow] ‘you are, pl.’ are not part of the Spanish phonological domain. Another prototypical feature of the Catalan vocalic system is the existence of schwa in the unstressed inventory. Stressed /e/, /ε/ or /a/ are realized as [ә] in words where these underlying vowels surface in unstressed position3, for instance in [pә.'lεt] ‘little hair’, or ‘little stick’, from ['pεl] ‘hair’ or ['pal] ‘stick’. Throughout the years, it has been observed that some traditional Catalan features have been lost (Recasens 1991, Pla Fulquet 1995 and Herrick 2006). Lleó et al. (2008) have already confirmed those statements, even though they have not found an equal loss of features across districts of Barcelona. They found significant differences in rate of target-like Catalan vowels in Gràcia and Nou Barris. These differences suggest that some traditional Catalan vowels are hardly produced as such in Nou Barris, a district where the influence of Spanish is very strong. Lleó et al. (2008: 201) even observed that the production by children in each district did not parallel the language spoken at home. Instead, the children’s productions were more influenced by the dominant language in the district (i.e., Catalan in Gràcia and Spanish in Nou Barris). These two districts studied in Lleó et al. (2008) were at both ends of the spectrum in the report about level of linguistic competence (IDESCAT 2006). Therefore, here we address the question whether a district reported to be in the middle of the ranking such as Eixample will show intermediate values between those of Gràcia and Nou Barris. That is, the production of vowels by children and adults in Eixample will be presented in Experiment I, additionally to that by children and adults in Gràcia and Nou Barris described in Lleó et al. (2008).4 1.2  Qualitative differences in vowel production A further question we want to address in the present paper is what exactly the acoustic realization of vowels perceived as target-like by native Catalan listeners is like. When native speakers identify members of the Catalan vowel pairs (i.e., 3.  The realization of schwa exclusively in unstressed position, as allophone of the corresponding stressed vowels /a/, /ε/ and /e/ justifies the transcription between brackets that we have adopted here. However, there exists a long debate in the phonology of Catalan about the status of this segment, which in some analyses is considered to be a phoneme (see e.g., Badia Margarit 1965 and Viaplana & DeCesaris 1984). 4.  The production percentages have minimally changed with respect to those in Lleó et al. (2008) due to updates, extensions and improvements in the corpus. The significant differences in the statistical tests remain unchanged anyway.

 Susana Cortés, Conxita Lleó & Ariadna Benet

/ε/ vs. /e/, /f/ vs. /o/, and [ә] vs. /a/), how distinct are the two members in each pair in the different districts? The rationale behind this question is that there might be qualitative differences across districts apart from the quantitative differences noted in Lleó et al. (2008). We can observe how convergent or divergent the vowels in each vowel pair are by analyzing the acoustic quality of the vowels perceived as target-like by native Catalan speakers. In their acoustic and perceptive study of the open and close mid-vowels /ε/, /e/, /f/ and /o/ in different Catalan dialects, Carrera-Sabaté and Fernández-Planas (2005) have concluded that, in all cases, differences in F1 seem to play a more important role than differences in F2 between members of each pair to define vowel quality. We will analyze whether our data, uttered by speakers from different districts in Barcelona, show this trend too. It is also worth mentioning that, out of the four dialects studied in Carrera-Sabaté and Fernández-Planas (2005), they found that Barcelona’s dialect has the mid-open and mid-close vowels with the smallest degree of opening, i.e., the lowest F1 values, when compared with other dialects. However, this does not mean that the F1s of those vowel pairs merge, since both, the mid-open and the mid-close vowel in each pair, have a low value. Unfortunately, Carrera-Sabaté and Fernández-Planas only describe their subjects according to the broad areas speakers belong to (e.g., Barcelonès, Segrià, La Safor and Mallorca occidental) and the cities they represent (i.e., Barcelona, Lleida, Gandia and Palma de Mallorca). They do not provide any further details about the origin or the linguistic background of their speakers, apart from indicating that they are university students in those cities. Therefore, we do not know whether the speakers they describe from Barcelona belong to the same or different districts and whether the districts they live in coincide with the districts analyzed in the present paper.5 Experiment II is devoted to the acoustic analysis of the data in our corpus.

2.  Method The present section describes the subjects analyzed, materials devised and equipment used for the data recordings for both Experiment I and Experiment II. The

5.  The subjects they analyzed are described as twelve men (i.e., 3 per dialectal variety) who are between 22 and 32 years old, whose pronunciation displays no Spanish influence (Carreras-Sabaté & Fernández-Planas 2005: 17).



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

elicited data are auditorily analyzed in Experiment I and acoustically analyzed in Experiment II. 2.1  Subjects The data was elicited from 60 subjects in total, comprised of three generations. G1 consisted of 10 children from every district under analysis; they were interviewed at school. The children in Gràcia attended the pre-school section in the schools CEIP Patronat Domènech, and CEIP Josep Maria Jujol, those in Eixample the pre-school section in CEIP Ramon Llull, CEIP Fort Pienc and IPSI, and the children in Nou Barris the pre-school section in the schools CEIP Gaudí and Sant Lluís. Subsequently, a session with either the father or the mother of each child was also recorded (G3). The interviews with the parents were carried out either in a room at the child’s school or at home. Table 1 gives a detailed description of the children and adults in our corpus, whose speech production is analyzed in the present paper. Table 1.  Subject groups’ description by generation and district (gender and mean age provided) Generation

G1

G3

District

Gràcia

Eixample

N. Barris

Gràcia

Eixample

N. Barris

Gender

8 female 2 male 3;10

4 female 6 male 4;7

5 female 5 male 4;6

9 female 1 male 36;4

7 female 3 male 39;10

6 female 4 male 37;1

μ age

2.2  Materials and equipment A series of materials were devised in order to elicit data from subjects. At least ten word types containing each target segment were elicited by picture and object naming. The pictures were found in children books, and some note cards were devised to fit the rest of the words not available in the books. Some questions were also asked to elicit abstract concepts not easy to represent in a picture. A total of 33 items were elicited from each subject for the analysis of vowels, 11 items for /ε/, 12 for /f/ and 10 for [ә]. As an illustration, we list a few of the target words here: for /ε/, cel ['sεl] ‘sky’, setze ['sεd.zә] ‘sixteen’, pedreta [pә.'ðnε.tә] ‘little stone’; for /f/, escola [әs.'kf.lә] ‘school’, flor ['flf] ‘flower’, nou ['nfw] ‘nine’; for [ә], pedreta [pә.'ðnε.tә] ‘little stone’, papallona [pә.pә.'ˆo.nә] ‘butterfly’, petit [pә.'tit] ‘small’.

 Susana Cortés, Conxita Lleó & Ariadna Benet

The data were recorded with a Sony ECM-CS1- unidirectional lapel microphone, connected to a portable Hi-MD Walkman MZ-RH10 Sony Mini-Disc recorder.

3.  Experiment I 3.1  Procedure Given the auditory results reported in Lleó et al. (2008), the data from speakers in Eixample have been auditorily analyzed, and they are reported here together with the results from the other two districts. Each token was transcribed by one of the authors and a native Catalan speaker.6 3.2  Results of vowel production across districts 3.2.1  Production of /ε/ The production of /ε/ across districts by the different generations shows high variation. The adults (G3) show very similar percentages of target-like production and no significant differences are found between all possible comparisons between districts (i.e., Gràcia vs. Eixample, Gràcia vs. Nou Barris, and Eixample vs. Nou Barris). However, percentages of target-like production by the children in Gràcia and Nou Barris and between those in Eixample and Nou Barris are significantly different (Gràcia vs. Nou Barris: [χ2(1, N = 20) = 230.75; p ≤ .001]; Eixample vs. Nou Barris: [χ2(1, N = 20) = 174.92; p ≤ .001]). Even the difference between the production by G1 in Gràcia and that by G1 in Eixample reaches significance [χ2(1, N = 20) = 7.23; p = .007]. The G1 /ε/ production going from more target-like in Gràcia to less target-like in Nou Barris, and with Eixample somewhere in between, is what we expected due to the description of language proficiency by the linguistic census of the districts of Barcelona (IDESCAT 2006). Nevertheless, the adults in these districts do not display differences. The difference in production by generations in the same district does indeed reach significance in Gràcia and Nou Barris (see Lleó et al. 2008) but not in Eixample. Children and their parents produce target-like /ε/ at similar high rates in Eixample.

6.  We are grateful to Laura Batalla, Laia Puig and Anna Schreibweis for their invaluable work doing data segmentation and transcription.

Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

% of target-like production



100 90 80 70 60 50 40 30 20 10 0

Gràcia Eixample Nou Barris

G1

G3

Figure 1.  Percentages of /ε/ production

% of target-like production

3.2.2  Production of /f/ As far as /f/ is concerned, Eixample displays the highest production rates in both generations. When comparing the children across districts, those in Eixample produce /f/ in a target-like manner significantly more often than those in Gràcia [χ2(1, N = 20)= 9.20; p = .002] and drastically more often than those in Nou Barris [χ2(1, N = 20)= 267.85; p ≤ .001]. The difference between production in Gràcia and Nou Barris is also statistically significant. Interestingly, here the ranking Gràcia > Eixample > Nou Barris in the group of children is not followed. In the group of adults, those in Eixample also produce /f/ in a target-like manner more often than those in Gràcia [χ2(1, N = 20)= 23.62; p ≤ .001] and than those in Nou Barris [χ2 (1, N = 20)= 17.50; p ≤ .001]. However, the difference in production between Gràcia and Nou Barris does not reach significance (see Lleó et al. 2008). As in the case of /ε/, the production by children and by adults in Eixample does not differ significantly. 100 90 80 70 60 50 40 30 20 10 0

Gràcia Eixample Nou Barris

G1

Figure 2.  Percentages of /f/ production

G3

 Susana Cortés, Conxita Lleó & Ariadna Benet

3.2.3  Production of [ә] The overall target-like production of [ә] is lower than that of the other vowels under study, especially by the children in Gràcia and in Eixample. The children in Gràcia pronounce [ә] target-like more often than those in Eixample [χ2(1, N = 20)= 11.027; p ≤ .001] or those in Nou Barris [χ2(1, N = 20)= 83.48; p ≤ .001]. The difference in [ә] production between the children in Eixample and those in Nou Barris also reaches significance [χ2(1, N = 20)= 32.93; p ≤ .001]. Children’s production of [ә] shows the ranking Gràcia > Eixample > Nou Barris observed in /ε/ production already. As for adults, the group displaying the highest production of [ә] is that of Eixample. The difference in production between the adults in Eixample and Gràcia is not quite statistically significant [χ2(1, N = 20)= 3.43; p = .06], whereas the difference in production by adults in Eixample and Nou Barris is clearly significant. [χ2(1, N = 20)= 17.80; p ≤ .001]. The difference in production by adults in Gràcia and Nou Barris is also significant [χ2(1, N = 20)= 5.19; p = .025].

% of target-like production

100 90 80 70 60

Gràcia

50

Eixample

40

Nou Barris

30 20 10 0 G1

G3

Figure 3.  Percentages of [ә] production

3.3  Discussion Lleó et al. (2008) found crucial differences between G1 percentages of target-like production of Catalan vowels in Gràcia and in Nou Barris, the latter being much lower. Significant differences between G3 and G1 in Nou Barris were also found. Both facts mean that the phonology of the language is very different across the two districts, correlating with the force of the Spanish language in them, much stronger in Nou Barris than in Gràcia; and they also show that the impact of the Spanish language is stronger than the language of the parents, as G1 percentages of target-like



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

production of Catalan vowels in Nou Barris significantly differ from those of their parents. The present study has added one more district to this picture, Eixample, which was selected because of being in between the other two districts as far as Catalan competence is concerned, according to sociolinguistic surveys. The results related to percentages of target-like production of the open-mid vowels /ε/ and /f/, as well as schwa, do not always match the predictions based on the position of this district in between the other two. In the case of G3, Eixample unexpectedly tends to have higher percentages of /f/ and [ә] than Gràcia, the difference being even statistically significant in the case of /f/; the production of /ε/ shows equivalent percentages in Eixample and Gràcia. For G1, both the production of /ε/ and [ә] lead to the expected results, Eixample being between Gràcia and Nou Barris, with statistically significant differences in both sides, i.e., both in relation to Gràcia and in relation to Nou Barris; however, in the production of /f/ Eixample takes the lead, with significantly higher percentages than both Gràcia and Nou Barris. These results point to a constant, namely that in the district with a very strong presence of Spanish, the young generation is losing the Catalan vowels, and this is thus heading to language change at least in the realm of phonology. As for Eixample, its G1 never approaches the low values of G1 in Nou Barris, rather it alternates with G1 in Gràcia in producing the highest percentages of Catalan vowels. That is, putting all vowels together, Eixample is comparable to Gràcia regarding the percentages of target-like production of Catalan vowels in G1, which points to the maintenance of the language with its phonological characteristics in these two districts. 4.  Experiment II When transcribers thought that the segment produced by the speaker was not target-like or when there was no agreement between them, those tokens were not acoustically analyzed. Only the tokens that were judged to correspond to the target vowel under study were acoustically analyzed. This way, the vowels that were judged by native speakers to be instances of the vowels in each pair could be compared in a qualitative manner. 4.1  Procedure Formant values of the selected vowels were obtained by using the formant track function on Praat 5.0.05 (Boersma & Weenik 2008) and running a script.7 The 7.  We are grateful to Francisco Torreira for providing us with the Praat script.

 Susana Cortés, Conxita Lleó & Ariadna Benet

script measures the formant structure of each vowel at the middle of the vowel on the time axis. Hence, measurements avoid the beginning and end of the vowel, where co-articulation effects are likely to affect the formant structure. With this procedure, the formant values of the steady state of the vowel were provided. Given that the automatic formant tracker does not provide 100% reliable outcomes, the values provided by the automatic procedure were double-checked by a phonetician, who changed inadequate values for her own measurements on the spectrograms.8 The mean F1 values for the clear instances of vowel production by each speaker were calculated. The same was done for the F2 values. The F1 mean values of all speakers in each district and generation were converted from Hertz into Bark scale, and then averaged. The mean F2 values were also averaged to obtain a group value. Those mean values give us group formant structures that are represented in the graphs included in the following section. The average values obtained are compared to the standard values provided by Recasens (1991).9 4.2  Results of the acoustic analysis 4.2.1  Production of /ε/ vs. /e/

16

15.5

15

14.5

F2 Bark 14 13.5

13

12.5

12

3.0 3.5 4.0 4.5 5.0 5.5 6.0

F1 Bark

/ε/ Std /e/ Std /ε/ G3 /e/ G3 /ε/ G1 /e/ G1

6.5 7.0 7.5 8.0

Figure 4.  Formant structure of vowels /ε/ and /e/ in Gràcia

8.  Thanks are due to Annekatrin Sell for doing this laborious job. 9.  The standard values for Catalan target vowels represented in Figures 4–12 are the Bark values corresponding to the Hz formant values in Recasens (1991). They are provided for reference only.

Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

16.0

15.5

15.0

14.5

F2 Bark 14.0 13.5

13.0

12.5

/ε/ Std /e/ Std /ε/ G3 /e/ G3 /ε/ G1 /e/ G1

12.0 3.0 3.5 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0

F1 Bark



Figure 5.  Formant structure of vowels /ε/ and /e/ in Eixample

16.0

15.5

15.0

14.5

F2 Bark 14.0 13.5

13.0

12.5

12.0 3.0 3.5 4.0 4.5 5.0 5.5 6.0

F1 Bark

/ε/ Std /e/ Std /ε/ G3 /e/ G3 /ε/ G1 /e/ G1

6.5 7.0 7.5 8.0 Figure 6.  Formant structure of vowels /ε/ and /e/ in Nou Barris

In Figures 4, 5 and 6, we can observe that the standard formant structure pattern is followed both by the group of children and adults in Gràcia and Eixample to a greater or lesser extent. However, the patterns noticeably deviate from the norm in the case of Nou Barris. Several comparisons were carried out by running t-tests on the F1 mean values for each speaker for the open and closed vowels. These tests were run separately per age group and per district. The same was done with the F2 mean values. The

 Susana Cortés, Conxita Lleó & Ariadna Benet

results of those tests are presented in Table 2, with shaded cells marking differences that reach significance. Table 2.  T-test results for /ε/ and /e/ per age group and district G1

F1 F2

G3

Gràcia

Eixample

N. Barris

Gràcia

Eixample

N. Barris

t(9) = 9.36 p ≤ .001 t(9) = – 4.40 p = .002

t(9) = 8.06 p ≤ .001 t(9) = – 5.47 p ≤ .001

t(2) = 1.84 n.s. t(2) = – 1.39 n.s.

t(8) = 5.72 p ≤ .001 t(8) = – 5.98 p ≤ .001

t(9) = 10.08 p ≤ .001 t(9) = – 3.05 p = .014

t(9) = 7.16 p ≤ .001 t(9) = .091 n.s.

There is a noticeable difference between the group conformed by Gràcia and Eixample when compared with Nou Barris. In Gràcia and Eixample both G1 and G3 show a significant difference between /ε/ and /e/ in F1 and in F2. In Nou Barris, however, only G3 shows a difference in F1. Given that dubious data were filtered out to carry out the acoustic analysis, we were left with a small sample of data for G1 in Nou Barris. We could only analyze data from 3 out of the 10 speakers examined in Experiment I. That is why it is difficult to generalize from the results of this group, even though they can be regarded as a tendency. 4.2.2  Production of /f/ vs. /o/

12

11.5

11

10.5

F2 Bark 10 9.5

9

8.5

8

4 4.5 5 5.5 6 6.5 7 7.5 8

Figure 7.  Formant structure of the vowels /f/ and /o/ in Gràcia

F1 Bark

/ɔ/ Std /o/ Std /ɔ/ G3 /o/ G3 /ɔ/ G1 /o/ G1

Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

12

11.5

11

10.5

F2 Bark 10 9.5

9

8.5

8

4 4.5

/ɔ/ Std /o/ Std /ɔ/ G3 /o/ G3 /ɔ/ G1 /o/ G1

5 5.5 6 6.5

F1 Bark



7 7.5 8 Figure 8.  Formant structure of the vowels /f/ and /o/ in Eixample

11.5

11

10.5

F2 Bark 10 9.5

9

8.5

8

4 4.5

/ɔ/ Std /o/ Std /ɔ/ G3 /o/ G3 /ɔ/ G1 /o/ G1

5 5.5 6 6.5

F1 Bark

12

7 7.5 8 Figure 9.  Formant structure of the vowels /f/ and /o/ in Nou Barris

Figures 7, 8 and 9 show that all groups follow the standard formant structure to some degree. Interestingly, all G3 groups show a very similar pattern with very small differences in the F2 axis and somewhat larger differences on the F1 axis. All G1 groups show large differences in F1 values between the two vowels compared. In Table 3, the differences observed in the graphs are ratified by the t-tests. All groups produce /f/ and /o/ with significantly different F1 values, but only G1 in Gràcia and Eixample also keep the F2 values of these vowels apart. Table 3.  T-test results for /f/ and /o/ per age group and district G1

F1 F2

G3

Gràcia

Eixample

N. Barris

Gràcia

Eixample

N. Barris

t(9) =10.39 p ≤ .001 t(9) = 5.76 p ≤ .001

t(9) = 8.36 p ≤ .001 t(9) = 3.19 p = .011

t(8) = 4.18 p = .003 t(8) = 1.92 n.s.

t(8) = 7.19 p ≤ .001 t(8) = .59 n.s.

t(9) = 9.62 p ≤ .001 t(9) = .181 n.s.

t(9) = 5.35 p ≤ .001 t(9) = .830 n.s.

 Susana Cortés, Conxita Lleó & Ariadna Benet

13.00

12.50

12.00

F2 Bark 11.50 11.00

10.50

/ә/ Std /a/ Std /ә/ G3 /a/ G3 /ә/ G1 /a/ G1

10.00 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0

F1 Bark

4.2.3  Production of [ә] vs. /a/

13.0

12.5

12.0

F2 Bark 11.5 11.0

10.5

[ә] Std /a/ Std [ә] G3 /a/ G3 [ә] G1 /a/ G1

10.0 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5

F1 Bark

Figure 10.  Formant structure of the vowels [ә] and /a/ in Gràcia

Figure 11.  Formant structure of the vowels [ә] and /a/ in Eixample

13.0

12.5

12.0

F2 Bark 11.5

11.0

10.5

10.0 4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5

Figure 12.  Formant structure of the vowels [ә] and /a/ in Nou Barris

F1 Bark

[ә] Std /a/ Std [ә] G3 /a/ G3 [ә] G1 /a/ G1



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

At first sight, the production of the pair of vowels [ә] and /a/ seems to pose more difficulties than any of the two other vowel contrasts. The F2 values for the two vowels produced by the children in Gràcia and Eixample, albeit displaying the right pattern, did not differ significantly. In many cases, however, the pattern regarding F2 seems to have been reversed. That is, a mirror image of the standard formant structure pattern on the F2 axis is observed. The standard pattern prototypically defines /a/ with lower F2 values than [ә] but in our data for all G3 groups across districts and for G1 Nou Barris, the opposite is the case. As observed in Table 4, even the swapped F2 values reached significance when produced by G3 from Eixample and Nou Barris. However, the difference between the swapped F2 values for [ә] and /a/ in G3 Gràcia and G1 Nou Barris did not reach significance. In Pla Fulquet (1995), some speakers also display swapped patterns. Such findings might be explained by the large dispersion of [ә] when compared to that of other vowels. When we refer to dispersion here, we mean the great variation of [ә] across speakers. Recasens and Espinosa (2006: 645) state that Majorcan Catalan stressed /ә/ “appears to be targetless or specified for a widely defined mid central target”. If we assume that to be the case for unstressed Catalan [ә] too, then the broad acoustic space targeted for [ә] production could lead to these mirror-image patterns. Flemming (2007) also found high F2 variability, especially in word-internal English schwa. He accounts for such variability in terms of coarticulation and duration, stating that “one reason why schwa is expected to be subject to stronger coarticulatory effects than other vowels is that it is shorter, and thus more subject to undershoot – i.e., assimilation to its context” (Flemming 2007: 13). If the area targeted for schwa is large and not clearly defined for a certain formant, then the values of /a/ for that same formant could easily be pushed away from the area belonging to [ә], if the instances of [ә] happened to be located on the very area originally targeted by /a/. As an illustration, Figure 13 shows the formant structure for [ә] and /a/ by a speaker exhibiting the pattern described for the standard, i.e., with higher F2 values for [ә] and lower for /a/, whereas Figure 14 shows the formant structure for those two vowels with swapped F2 values produced by one speaker in Nou Barris.10 It should also be noted that, given that only clear examples of [ә] were included in the acoustic analysis, we were often left with three or fewer tokens per speaker, like in Figure 13, and, therefore, we cannot see a big dispersion of schwa with respect to that of /a/. Nevertheless, the big dispersion of [ә] can even be observed in the speaker’s production plotted in Figure 14. Tokens for [ә] by this speaker have F2 values ranging between 1000 and 1500 Hz.

10.  Large symbols for [ә] and /a/ in Figures 13 and 14 represent the location of the average value of all tokens for each vowel.

 Susana Cortés, Conxita Lleó & Ariadna Benet

Table 4.  T-test results for [ә] and /a/ per age group and district G1

F1 F2

G3

Gràcia

Eixample

N. Barris

Gràcia

Eixample

N. Barris

t(9) = – 4.37 p = .002 t(9) = 1.31 n.s.

t(9) = – 5.39 p ≤ .001 t(9) = .43 n.s.

t(6) = – 5.06 p = .002 t(6) = .193 n.s.

t(9) = – 6.82 p ≤ .001 t(9) = -.164 n.s.

t(9) = – 6 p ≤ .001 t(9) = - 4.47 p ≤ .001

t(8) = – 9.56 p ≤ .001 t(8) = - 3.09 p = .015

2300 2100 1900 1700 1500 1300 1100 900 F2

700

500 200 300 400 500 600

әә ә

ә

a

a

a

aa a a aaa aa aa

700 a

a

800

a

900 1000 F1

Figure 13.  Formant structure of the vowels [ә] and /a/, showing the standard pattern for F2 2300 2100 1900 1700 1500 1300 1100 900 F2

700

500 200 300

ә

ә

a әә a a ә ә a ә ә a әa aa a ә ә ә a a a a a ә

400 500 600 700 800 900 1000 F1

Figure 14.  Formant structure of the vowels [ә] and /a/, showing the swapped pattern for F2



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

4.3  Discussion Overall, we can state that F1 is differentiated for all vowel contrasts, except in the case of the /ε/ vs. /e/ contrast by G1 Nou Barris. Our results provide further evidence supporting Carrera-Sabaté and Fernández-Planas (2005)’s claims on the decisive role of F1 in order to maintain the contrasts in mid vowels. We can extend such a claim to the contrast between [ә] and /a/. It is interesting to note that the children in the districts with a stronger presence of Catalan (i.e., Gràcia and Eixample) keep the F1 and F2 values for mid vowel contrasts: /ε/ vs. /e/ and /f/ vs. /o/. Children in the districts with less influence from Spanish produce the mid vowels with clearly differentiated formants.

5.  General discussion The general question whether pairs of Catalan vowels, namely /ε/ and /e/, /f/ and /o/ as well as [ә] and /a/, are merging cannot be categorically answered, as we have seen both in Experiment I and II that it is especially in Nou Barris, the district with a stronger presence of Spanish, where the merging is taking place, as manifested by the results of G1. Also the question whether a district like Eixample, with a linguistic competence of Catalan that lies between Gràcia and Nou Barris (from more to less linguistic competence) has more in common with one or the other of the two extremes, must be relativized: Experiment I has shown that G1 in Eixample, like G1 in Gràcia, always shows a statistically significant higher percentage of target-like production for the three pairs of vowels when compared to Nou Barris. However, in the case of /ε/ vs. /e/ and [ә] vs. /a/, Gràcia exhibits a significantly higher percentage of target-like production than Eixample, whereas in the case of /f/ vs. /o/ it is G1 in Eixample that exhibits a significantly higher percentage of target-like production than Gràcia. Thus, according to Experiment I, only if a certain threshold of presence of Catalan is reached, as in Gràcia or Eixample, are language features maintained. In contrast, Nou Barris does not reach the necessary threshold values to maintain the vowel contrasts. The results of Experiment II support the findings from Lleó et al. (2008) and those of Experiment I, as it is mainly in Nou Barris that certain formant values of the two vowels of each pair are merged. Such merger takes place mainly in the production by G1, but in some cases in that by G3, too. However, merged formant values for G3 are mainly exhibited in the case of F2, which is in a way a less reliable formant, as it often shows the effects of co-articulation with the preceding and/or following consonant (see Herrick 2007). Consequently, we take F1 to be the decisive formant indicating the height status of the respective vowel. In this sense, only the opposition between /ε/ vs. /e/ seems to be merged by G1 of Nou Barris: that is, F1

 Susana Cortés, Conxita Lleó & Ariadna Benet

shows no difference between /ε/ and /e/ in the younger generation living in that district. The other vowels are also getting closer to one another when uttered by the young generation in Nou Barris, although they still exhibit significant differences in relation to F1. G3 maintains the oppositions more clearly in all districts, with differences mainly regarding F2. Putting all results together, we can say that, on the one hand, G1 in Nou Barris is effecting some language change, in that many vowels that belong to Catalan but not to Spanish, namely /ε/, /f/ and [ә] are being merged with the corresponding /e/, /o/ and /a/, respectively, in the younger generation. On the other hand, the older generation (G3) shows a more homogeneous language state across districts. Differences between G1 and their parents (G3) show that language transmission is not predictable generationally because school, language of the district, peer group, etc. play a more important role than the information transmitted by the parents. The low results of G1 found in Nou Barris suggest that these children, when speaking either in one or in the other language, may be using only one phonological system, the Spanish one, with only five vowels instead of the Catalan system with eight vowels. Besides, note that the process of sound change that is taking place in Nou Barris, through which the Spanish system of mid vowels with only one degree of opening is substituting for the Catalan system of mid vowels with two degrees of opening, involves markedness, in the sense that systems with two degrees of opening in the mid vowels are more marked than those with only one degree of opening (see Maddieson 1984). Thus markedness may be favoring the vocalic merge of the mid vowels (for a more detailed account on the question of markedness in this context, see Lleó et al. 2008). Although the percentages of target-like production are very low, it is not always the case that the speakers of G1 in Nou Barris systematically pronounced the target-like Catalan vowels /ε/, /f/ and [ә] with the vocalic values of their Spanish counterpart. From production alone it is difficult to determine whether the few occurrences of some Catalan vowels, namely the open mid vowels and the schwa, are just produced by phonetic imitation. That is, the child may not be using the Catalan phonological system, and may have memorized a few items with the corresponding Catalan vowel. Running perception tests on the subjects in this study could provide us with further insights into the question of the phonological system of the children in Nou Barris by testing whether they actually perceive the difference of each vowel pair. Research on perception of Catalan vowel contrasts reports that Catalan speakers with Spanish dominance have difficulties in discriminating between the vowels in such contrasts, whereas those with Catalan dominance do not (Bosch & Sebastián-Gallés 2003).



Gradient merging of vowels in Barcelona Catalan under the influence of Spanish 

6.  Conclusion This paper has compared the production of three vowel pairs in three districts of Barcelona by two different generations. Quantitatively, the Spanish and Catalan vowel systems are converging in the speech of the children in Nou Barris, whereas the children in Gràcia and Eixample keep producing the different vowel pairs in a target-like manner at high levels. The qualitative study has also shown that the children in Nou Barris do not produce /ε/ and /e/ with such a noticeable difference in vowel quality as the children in Gràcia and Eixample. Therefore, the contact between Spanish and Catalan seems to be causing some convergence in the Catalan phonology of this highly Spanish-influenced district of Barcelona.

References Ajuntament de Barcelona. 2007. Característiques de la població de Barcelona segons el Padró municipal. Juny 2007. Ajuntament de Barcelona, Departament d’Estadística. , November 28, 2008. Ajuntament de Barcelona. 2008. La població estrangera a Barcelona. Gener 2008. Ajuntament de Barcelona, Departament d’Estadística. , November 28, 2008. Badia Margarit, A.M. 1965. Función significativa y diferencial de la vocal neutra en el catalán de Barcelona. Revista de Filología Española 48: 79–93. Boersma, P. & Weenink, D. 2008. Praat: Doing Phonetics by Computer. Version 5.0.05. , January 22, 2008. Bosch, L. & Sebastián-Gallés, N. 2003. Simultaneous bilingualism and the perception of a languagespecific vowel contrast in the first year of life. Language and Speech 46(2/3): 217–243. Carrera-Sabaté, J. & Fernández-Planas, A.M. 2005. Vocals mitjanes tòniques del català. Estudi contrastiu interdialectal. Barcelona: Horsori. Flemming, E. 2007. The Phonetics of Schwa Vowels. Ms, MIT. , June 3, 2009. IDESCAT (Institut d’Estadística de Catalunya). 2006. ‘Societat: llengua’. , December 13, 2006. Herrick, D. 2006. Mid vowels and schwa in Eastern Catalan: Five non-Barcelona dialects. In New Perspectives on Romance Linguistics: Vol. II: Phonetics, Phonology and Dialectology, I.-P.Y. Montreuil (ed.), 113–126. Amsterdam: John Benjamins. Herrick, D. 2007. An acoustic description of Central Catalan vowels based on real and nonsense word data. Catalan Review 21: 231–256. Lleó, C., Cortés, S. & Benet, A. 2008. Contact-induced phonological changes in the Catalan spoken in Barcelona. In Language Contact and Contact Languages, P. Siemund & N. Kintana (eds), 185–212. Amsterdam: John Benjamins. Maddieson, I. 1984. Patterns of Sounds. Cambridge: CUP.

 Susana Cortés, Conxita Lleó & Ariadna Benet Pla Fulquet, J. 1995. L’ obertura de [ә] a Barcelona: El xava i altres varietats. In La Sociolingüística de la variació, M.T. Turell (ed.), 139–162. Barcelona: PPU. Pons, E. & Vila, F. X. 2005. Informe sobre la situació de la llengua catalana (2003–2004). Observatori de la Llengua Catalana. , July 10, 2006. Recasens, D. 1991. Fonètica descriptiva del català. Barcelona: Institut d’Estudis Catalans. Recasens, D. & Espinosa, A. 2006. Dispersion and variability of Catalan vowels. Speech Communication 48: 654–666. Viaplana, J. & DeCesaris, J.A. 1984. La ‘vocal neutra’ del català central: Fonema o al.lòfon? Estudis Gramaticals 1. Working Papers in Linguistics, 341–380. Bellaterra: Universitat Autònoma de Barcelona.

Comparing the representation of iambs by monolingual German, monolingual Spanish and bilingual German-Spanish children Javier Arias & Conxita Lleó

Hamburg, Research Centre on Multilingualism This article examines the production of iambic-shaped words by two monolingual German, two monolingual Spanish and two German-Spanish bilingual children, aiming to contribute to the understanding of stress acquisition in early childhood. Target iambic words produced at ages 1;0 to 2;6, have been auditorily and acoustically analyzed, focusing on rhyme duration. Results show that whereas German monolinguals at first often truncate the unstressed syllable, Spanish monolinguals hardly show any truncation, but at about 1;8 convert iambs to trochees. These diverging behaviors respond to different analyses, based on target language differences: Whereas German monolinguals analyze iambic words as comprising a moraic trochee preceded by an unfooted syllable, Spanish monolinguals analyze them as quantity-insensitive iambs. The bilinguals show some interaction between both systems.

1.  Introduction Iambic-shaped words have traditionally posed a challenge for theories of stress within the generative framework, most significantly, standard Metrical Stress Theory, as advocated by Hayes (1995). His Iambic-trochaic Law, by which trochees may be quantity-sensitive or quantity-insensitive, whereas iambs are always quantity-sensitive, has inspired much work on stress in Metrical Phonology. However, the Iambic-trochaic Law has recently been called into question, as researchers have encountered languages that base their stress patterns on iambic feet, while lacking a distinction between short and long vowels, which precludes such iambs from involving quantity differences (see Altshuler 2006; Everett 2003). In Spanish, feet tend to be left-headed and binary, i.e. the language is based on trochees (Harris 1983), although alternative accounts have been proposed in terms of iambic feet (Roca 1988). The Spanish language has a majority of words stressed on the penultimate syllable, although words ending in a stressed

 Javier Arias & Conxita Lleó

syllable are common, too. Adverbs like aquí ‘here’ and allí ‘there’, or infinitives like a comer ‘food is ready’, a dormir ‘time to sleep’, a jugar ‘let’s play’, used as imperatives, are relatively frequent, especially in child language. An analysis that allows iambs along with trochees in the language poses at least two difficulties. First, such a duality of feet in one single language is not quite acceptable in Metrical Theory. Second, iambs with a light last syllable (as e.g. Spanish aquí or allí) cannot be handled by means of the Iambic-trochaic Law, because, as mentioned above, the presence of iambs in a language presupposes quantity-sensitivity, and Spanish is not a quantity-­sensitive language. Another analysis of such iambic forms in Standard Metrical Theory is carried out by means of moraic trochees (as opposed to syllabic trochees). However, such an analysis poses some problems as well, since moraic trochees are only allowed in quantity-sensitive languages (e.g. English, Latin), which makes them difficult to analyze in the case of languages like Spanish. That is, these words must be marked lexically, which leads to a problem for the Standard Theory. German metrics is also essentially based on trochees, with only a few iambs (e.g. kaputt ‘broken’). Iambs are not very common in child German, although they are not as challenging for the theory as is the case with Spanish, because German is a quantity-sensitive language with vowel-length distinctions. Thus, although they would not challenge the Iambic-trochaic Law, their existence is doubtful in German, as this language is massively trochaic, exhibiting both syllabic and moraic trochees. Given that the main foot is the trochee in both languages, the existence of some iambs notwithstanding, quite often in child Spanish but relatively seldom in child German, the question must be asked of how young children acquire the corresponding stress system of each language. In the literature on stress acquisition it is widely assumed that the development takes place by means of an algorithm, rather than on a lexical basis (Fikkert 1994; Klein 1984). The claim has also been made that the occurrence of stress errors in the language of the child should provide critical evidence in support of this view (Hochberg 1988). From a theoretical standpoint, stress errors can be interpreted as the outcome of overgeneralization of the prosodic pattern that lies at the core of the stress system being considered. Besides, children may also produce errors due to their imperfect command of the phonetic parameters they have to put together when realizing a prominent prosodic unit. As we will see, it can be assumed that both languages are based on an algorithm that builds trochaic feet, and that the most plausible analyses have iambic feet construed as iambs in Spanish and as moraic trochees in German. In both cases, iambic-shaped words are exceptions to the trochaic algorithm, and we expect children to overgeneralize the trochaic pattern and/or to resolve the exceptions, i.e. the iambic forms, in some way, which will be characterized as errors.



Comparing the representation of iambs 

In this paper, we will present and compare the production of target iambicshaped words by two monolingual Spanish children, two monolingual German children, and two bilingual German-Spanish children. The time span comprises ages 1;0–2;6. Our data show that monolingual Spanish and monolingual German children analyze iambic-shaped words differently in their respective languages. Whereas German iambic-shaped forms are truncated for a relatively long period of time, the Spanish data hardly manifest truncation. We interpret the lack of truncation in Spanish by means of some recent proposals (Altshuler 2006; Everett 2003), which count on quantity-insensitive iambic feet for the analysis of some languages. That is, Spanish children begin producing trochees and a few iambs; these are converted into trochees during a few months. But given their relative frequency in the target language, iambs are soon produced target-like. For German, the argument can be made that it contains syllabic as well as moraic trochees, and that iambic-shaped words are analyzed as moraic trochees preceded by an unfooted syllable. This representation accounts for the frequent truncation of the metrically weak syllable. The paper is organized as follows. In Section 2, after a brief description of the metrical systems of Spanish and German, we present the hypotheses related to the acquisition of stress in the two languages within a monolingual and a bilingual context (Section 2.3). The empirical study is presented in Section 3, along with the results of acquisition in Section 4, followed by the discussion in Section 5 and some conclusions in Section 6.

2.  Brief description of metrical feet and stress in the target systems Metrical Theory has stimulated an enormous amount of work on stress in German and Spanish. We will only mention some crucial and well-known studies, and will introduce a general description of both stress systems, without tackling several polemic issues that have been discussed in the field and that are far from being settled. In Metrical Theory, the prosodic units that receive stress are metrical feet, which can be stressed on the left or on the right. Generally, languages have only one type of foot: the one with prominence on the left (left-headedness) is a trochee and the one with prominence on the right (right-headedness) is a iamb. Prominence is generally considered to be the grouping of several phonetic parameters, duration, intensity and pitch, which depending on the language are more or less salient. 2.1  German Word stress in German is based on trochees, built from left-to-right (Giegerich 1985; Vennemann 1991; Hayes 1995; Wiese 1996; Zonneveld, Trommelen, Jessen,

 Javier Arias & Conxita Lleó

Bruce & Árnason 1998; Janßen 2003; Domahs, Wiese, Bornkessel-Schlesewsky & Schlesewsky 2008; Knaus & Domahs 2009). German being quantitive-sensitive (QS), it has both, syllabic and moraic trochees, as shown in the following example:

(1) a.—Syllabic trochee

b.—Moraic trochee

Ft

Ft

σ [vε

σ lə]

µ [ma

µ n]

The word Welle ‘wave’ comprises two syllables, stress being on the syllable to the left, i.e. it is a syllabic trochee (1a).1 Mann ‘man’ comprises only one syllable, which also constitutes a foot, namely a moraic trochee (1b). The theory requires that feet be binary, binarity being satisfied by the presence of two syllables, as in (1a), or by the presence of two morae, as in (1b). Feet are the building blocks of prosodic words; the latter generally comprise one foot and something else. This something else usually consists of one or even two further feet. Morphologically simple words are stressed on the last foot to the right, whereas morphological compounds, as e.g. aufmachen ‘open’, made up of two feet, auf and machen, bear the main prominence mark on the left-hand foot, in this case on auf. Besides, there are words with iambic shape, as e.g. kaputt ‘broken’, with stress on the right-hand side. The question that this type of word raises is related to the type of foot that constitutes it. There are at least two possibilities: either the word comprises an iambic foot, with stress on the right-hand syllable, or it comprises an unfooted syllable followed by a moraic trochee, i.e. a monosyllabic trochee (Féry 1996; Van de Vijver 1998; Revithiadou 2004; Kager 2006). The two possibilities are shown in (2a) and (2b), respectively.2

(2) a.—Iamb

b.—Moraic trochee

PW

PW Ft

Ft σw [ka

σ

σs pʊt]

σ [ka

µ pʊ

µ t]

1.  It is rather common to analyze such words with an ambisyllabic [l], to render the stressed syllable bimoraic (Hall 2000, hinging on the analyses of English ambisyllacitity by Kahn 1976, and Rubach 1996). This analysis, which is absolutely plausible for German, does not alter the thrust of our argument that Welle should be parsed as a syllabic trochee.] 2.  Instead of an unfooted syllable, a degenerate foot could be posited. But such feet are bounded to strong restrictions in the literature (Hayes 1995: 87), which are not fulfilled here, so that we prefer to ignore this further alternative.



Comparing the representation of iambs 

Apart from formulating the phonological and metrical description of the system, it is important to determine the phonetic parameters that contribute to prominence in each language. If we draw on the literature in order to determine the phonetic properties of stressed and unstressed syllables in comparison, we find the following pieces of information: Duration of a stressed syllable is twice that of an unstressed syllable in German (Mengel 1997; Dogil 1999). Intensity of the stressed syllable is somewhat 2.2 dB greater than that of an unstressed syllable (Delattre 1966). And the stressed syllable has a higher pitch than the unstressed syllable. Final Lengthening, by which the last syllable of a Phonological Phrase is lengthened, must also be considered (Berkovits 1994; Schneider 2007). Although we will be analyzing words produced by the children, not necessarily phrases, in many cases the words analyzed are produced in isolation, thus constituting their own phrases. 2.2  Spanish The basic foot of the Spanish stress system is the trochee, as well. The most prominent foot in a word is generally the one to the right-hand side. Alternative analyses, as e.g. Roca (1988) propose to base word stress on the stem (excluding final elements, like final vowels, according to Harris 1991a), instead of on the complex word including final elements; this proposal involves iambs rather than trochees, as the majority of stems bear stress on the last vowel. We prefer the analyses based on trochees, as they better correspond to young children’s experience. That is, young children have not yet undertaken a morphological analysis of words, which would allow them to isolate stems. Such an analysis is too abstract for the purpose of trying to identify the underlying representation of certain words in young children’s vocabularies, which should be based on what children hear. We assume that Spanish is a quantity-insensitive language (QI), although this is a controversial issue in the literature on Spanish stress (e.g. Harris 1983, 1991b, 1995). Words in Spanish can have prominence on one of the last three syllables to the right, which corresponds to the well-known three-syllable window. In our view, certain phenomena that have been interpreted by some phonologists in favor of quantity-sensitivity are just vestiges of the ancestor language, Latin, which was indeed quantity-sensitive. Under such vestiges, we count the tendency to not having preparoxitonic stress if the penultimate is a closed syllable (i.e. *cáramba), which goes back to the Latin stress rule. Moreover, the fact that oxitonic stress is more common in words containing a final coda is not a strong generalization, as there are many paroxitonic words that end in a consonant, too (i.e. balcón coexists with móvil). As far as stress is concerned, putting aside the massive number of paroxitonic or trochaic words, Spanish also contains a respectable number of oxitonic or iambic-shaped words, which, as mentioned in the introduction, pose a problem to any metrical analysis of Spanish. In principle, the alternatives for analysis are the same as those stated under (2a) and (2b) for German. That is, iambic-shaped words

 Javier Arias & Conxita Lleó

can in principle be analyzed in two ways: as a disyllabic foot with prominence to the right (that is, an iamb) or as a moraic trochee preceded by an unfooted syllable. Both of these solutions are questionable in a quantity-insensitive system, as both iambs and moraic trochees are based on counting morae rather than syllables. As in the case of German, it is also important to determine the phonetic properties of stressed syllables, that is, to establish which phonetic parameters contribute to prominence and in what relation. Duration of a stressed syllable is greater than that of an unstressed syllable (Navarro Tomás 1916). The difference amounts to some +50 ms for trochaic-shaped words, and to about +150 ms in iambic patterns. Intensity of the stressed syllable is about 1.3 dB greater than that of an unstressed syllable (Delattre 1966). The stressed syllable has a higher and wider pitch contour than the unstressed syllable. However, Spanish tends to show a L*H pitch accent in pre-nuclear position (Sosa 1999; Face 2001a, 2001b; Hualde 2002; Prieto 2004). This late peak, which is typical of many Romance languages, often renders the stressed syllable in a lower tone than the posttonic; in those cases, though, it is common to still find a wider pitch contour in the stressed syllable. Final Lengthening, by which the last syllable of a Phonological Phrase is lengthened (Medina Murillo 2005; Snow 1994 for child language), has also to be considered when the words analyzed appear in final position. 2.3  Hypotheses for early acquisition: Monolingual and bilingual In this section the differences in the metrical make-up of the two languages, as briefly discussed above, will be taken into consideration, as they may have far-reaching consequences for the representation of iambs. We will assume the proposal made in (2) above, namely that languages may differ in relation to the prosodic structure that they attribute to iambic-shaped words. German is a quantity-sensitive language, based on trochaic feet, and the best representation for iambic-shaped words thus seems to be that of an unstressed unfooted syllable followed by a moraic trochee, as in (2b), so that basically one foot type is available in this language, namely, the trochee (see, among others, Giegerich 1985; Vennemann 1991; Hayes 1995; Wiese 1996; Hall 2000). In Spanish, moraic trochees are not accessible, as this language does not count morae, but syllables. Iambic feet have traditionally been restricted to quantity-sensitive languages in Metrical Theory. However, several authors have recently shown that the universal inventory of metrical feet should also include syllabic iambs, i.e. two-syllable feet with prominence on the right in quantity-insensitive languages (see Altshuler 2006; Everett 2003). In accordance with this proposal, we will consider representation (2a) as the most plausible one for iambs in Spanish. This main difference between the representations (2b) in German and (2a) in Spanish for iambs may in its turn lead to important differences in acquisition. We will thus search there for an explanation for the different behavior that German and Spanish young children exhibit in relation to the production of iambic words.



Comparing the representation of iambs 

As for the phonetic parameters that contribute to prominence in the two languages, mean differences in duration between stressed and unstressed syllables are greater in German than in Spanish. The difference in amplitude is also greater in German (2.2 dB) than in Spanish (1.3 dB) (Delattre 1966). Finally, the pitch contrast marking syllable prominence should show greater values in German than in Spanish, as German generally places the stressed syllable in a peak (i.e. H*). In Spanish, the stressed syllable is aligned with a H tone only in final phrases, whereas in pre-final phrases, there is a tendency to align the stressed syllable with a L tone (i.e. L*H) (Sosa 1999; Face 2001a, 2001b; Hualde 2002; Prieto 2004). However, this difference, which is clear for adults, may not have much of a correspondence in very young Spanish children’s productions. In Lleó, Rakow and Kehoe (2004) it was shown that at age 3;0 Miguel, the child analyzed in this study, had a relatively high number of L*H pitch accents in pre-final position. However, at 2;0 years of age, only one child out of three, María, showed the relevant difference between pre-final and final pitch accent (Lleó & Rakow 2006). Moreover, the L*H pitch accent is generally aligned with trochaic words; iambic words have the stressed syllable at the end of the word, so the pitch accent L*H can only be realized if a further syllable follows; since the words considered have generally been produced phrase-finally, we may expect the stressed syllable of a iamb to be the highest one in Spanish, as well. Final Lengthening does not seem to exhibit a difference between the two languages. On the basis of such similarities and differences between the two language systems, we can formulate several hypotheses (H1–H5) related to the acquisition of iambs by young monolingual and bilingual children, which will be tested in the empirical part of the study: H1. Assuming that stress is acquired by means of an algorithm, and taking into consideration that there is a majority of trochaic-shaped words in both languages, iambic-shaped words should present difficulties both in German and in Spanish, as they differ from the most frequent prosodic pattern of the corresponding language. H2. The difference between the two languages regarding the representation of iambic-shaped words, as discussed at the beginning of this section, will result in differences in the acquisition of those word-patterns in the two languages, as the initial syllable is unfooted—and thus loose from foot—only in German. In Spanish, both syllables have footed status. Thus, a tendency to omitting the unfooted syllable is to be expected in German, whereas in Spanish both syllables should be produced. H3. As far as the production of stressed syllables is concerned, they will have more duration, more intensity and be higher in relation to the unstressed syllable in German than in Spanish, as all these prosodic cues have higher values in German than in Spanish.

 Javier Arias & Conxita Lleó

H4. Bilingual children will acquire word-stress in the corresponding languages the same way as monolinguals do, in the sense that some difficulties will show up in relation to iambs. The rationale for this hypothesis is based on the relative similarities between the stress systems of the two languages, namely, the overwhelmingly greater frequency of trochees over iambs. H5. Concomitantly, given the different representation of iambs assumed here for the two languages, and given the interaction that has been often found in the bilingual acquisition of prosody, either iambs will be facilitated in German or they will be made more difficult in Spanish, depending on whether German or Spanish is the most influential language in a particular bilingual child’s development. 3.  Empirical study In what follows, we will describe the present study on the acquisition of iambicshaped words by German and Spanish monolingual and bilingual children, aged 1;0 to 2;6. Trochaic-shaped words have been described elsewhere (see Lleó & Arias 2006; Lleó & Arias 2009). 3.1  Subjects For the purposes of this study, iambic-shaped words produced by two German monolingual (Bernd and Marion), two Spanish monolingual (José and Miguel) and two German-Spanish bilingual children (Jens and Simon) were selected from the beginning of word production until 2;6. The monolingual data stem from the project PAIDUS, conducted at the University of Hamburg by Conxita Lleó, and the bilingual data stem from project TP E3, also conducted by the same author, at the Research Center on Multilingualism of the University of Hamburg.3 In both projects, children were visited at home by two investigators, and were stimulated to communicate in spontaneous play with the interviewers. The bilinguals and the German-speaking monolinguals were recorded in Hamburg. The Spanish-speaking monolinguals were interviewed under the direction of Antonio Maldonado in Madrid, where they lived. The bilingual children were visited by two different teams, a German-speaking team and a Spanish-speaking team. As to the method of data collection, it should be pointed out that the children were not tested with regard to some specific target words, but were free to produce any utterance, which partly explains the lack of data in some areas, as e.g. lack of iambs in German, where children tended to restrict their iamb production to various realizations of kaputt ‘broken’ and of a few other words (see Sec. 4.1).

3.  We want to express our gratitude to the German research Foundation (DFG) and to the University of Hambug for their support.



Comparing the representation of iambs 

3.2  Data and analysis One-word utterances made up of iambic-shaped target words were selected for analysis. We first inspected the transcriptions of these words, in order to identify which syllable had been transcribed as being the stressed syllable according to the children’s productions. Not all words had been transcribed as iambs, as we will see below. We then proceeded to the acoustic analysis of all target iambic-shaped words for the values of duration, pitch and amplitude. The analyses were done by means of Praat for the MacIntosh. Both duration of vowels and duration of rhymes was measured, whereby the latter seemed the most relevant measurement, as many of the syllables produced by the children, especially in German, were closed syllables. Finally, the data were submitted to statistical analyses: two-tailed t-tests were conducted regarding the phonetic parameters responsible for stress prominence. Thus, the present article comprises three different types of analyses, namely, an auditory analysis, as rendered in transcriptions, acoustic measurements and a statistical account of the results. Although most of the data analyzed comprised iambs, trochees were considered as well, in cases a comparison seemed necessary, as e.g. in Sec. 5.1. Table 1 shows all iambs and trochees considered, divided in types and tokens. An example illustrative of an iamb is the word Musik ‘music’, produced [dᴧ˜dᴧ˜tһ] by Bernd at (2;3), and an example illustrative of a trochee is the word Schaukel ‘swing’, produced ['thaʊ̯phɛl] by Bernd at (2;6). Notice that whereas the transcription of the latter word, the trochee, indicates the position of stress, that of the former word, an iamb, does not; the lack of stress in the transcription indicates that the transcribers could not decide where stress was placed in the child’s production. Table 1.  Number of iambic- and trochaic-shaped words analyzed IAMBIC-SHAPED WORDS

TROCHAIC-SHAPED WORDS

Spanish monolinguals

Types

Tokens

Types

Tokens

José Miguel German monolinguals Bernd Marion Bilinguals (Spanish) Jens Simon Bilinguals (German) Jens Simon

5 6

20 18

16 9

20 20

2 4

2 4

28 18

41 33

4 8

8 14

13 12

20 20

4 8

6 17

33 18

46 44

 Javier Arias & Conxita Lleó

4.  Results of the child data on metrical feet and stress In this section, we present the percentages of truncation of iambic-shaped words, the percentages of stress misplacement, according to the auditory analysis, and the values for the phonetic parameters in the stressed syllable in contrast with the unstressed one. 4.1  Truncation The data were first inspected for truncation. For this purpose, besides the productions of the monolingual and bilingual children introduced above (in Sec. 3.1), we also considered one further monolingual child in each language group, Thomas for German and María for Spanish. Because of lack of data, the bilinguals have only been inspected in Spanish.4 The child data were separated into two stages, from the beginning of child production until 1;10 (indicated in the diagrams by the name of the child followed by “1”) and from 1;11 to 2;1 (name of child followed by “2”). The following three diagrams show the percentages of truncation by German monolinguals (Fig. 1), Spanish monolinguals (Fig. 2) and German-Spanish bilinguals in Spanish (Fig. 3). 100 Bernd1 80

Bernd2 Marion1

60

Marion2 Thomas1

40

Thomas2

20 0

Monolingual German

Figure 1.  Percentages of truncation in German monolinguals

4.  As already mentioned in Sec. 3.1, recordings were done in a naturalistic context, and iambic words were not abundant in the German productions.



Comparing the representation of iambs  100 José1

80

José2 Maria1

60

Maria2 Miguel1

40

Miguel2 20 0 Monolingual Spanish Figure 2.  Percentages of truncation in Spanish monolinguals

100 80

Jens1

60

Simon1

Jens2 Simon2

40 20 0

Bilingual Spanish

Figure 3.  Percentages of truncation in bilinguals

Note that two of the German monolingual children hardly attempt to produce any iambs at the first stage; only one German monolingual child (Thomas) attempts to produce them, with about 40% of truncation at stage 1. At stage 2, truncation reaches 40% in the case of Marion, but almost 100% in the case of the other two children, Bernd and Thomas. The three monolingual Spanish children have a very low percentage of truncation at all times, never reaching the mark of 20%. Bilinguals in Spanish at both stages have low values of truncation, relatively similar to those of monolinguals, although they tend to be a bit higher, especially in the case of Simon, who reaches about 40% of truncation at the first stage. As pointed out above, in German, the data contain very few iambic words.

 Javier Arias & Conxita Lleó

4.2  Stress misplacement The auditive analysis carried out on the basis of transcriptions shows that iambs are generally stressed targetlike by the monolinguals, i.e. stress generally falls on the syllable that is also stressed in the target language. However, from 1;9 until 2;6, both Spanish monolingual children, José and Miguel, convert many iambs to trochees. In German, once truncation is overcome, there are hardly any stress errors in this sense, i.e. converting iambs to trochees. In the bilinguals we find some productions of trochees for iambs, but the attested errors first come up at 2;0, which is a point in time when the child lexicon contains many trochees, and errors are not numerous. Stress errors for iambs are quantified in Table 2, with regard to the total number of errors per child. On the basis of transcriptions the table illustrates the percentages of iambs produced as trochees out of all stress errors noted; raw numbers are given in parentheses. The percentages have been calculated out of all stress errors and not on the basis of iambs targeted. There are not many iambs showing stress misplacement, and they are more numerous later on, when the child reaches the age of 1;9 in the case of Spanish monolinguals and 2;0 in the case of the bilinguals. The table also shows that Spanish monolinguals produce more stress errors than the bilinguals. In German, both monolinguals and bilinguals exhibit few errors. Table 2.  Percentages of iambs transcribed as trochees out of all stress errors found children

Iambs as trochees Percentages (tokens)

Bernd (from 1;11 on) Marion (from 1;11 on) GERMAN (MONOLINGUALS) José 1 (until 1;8) José 2 (from 1;9) Miguel 1 (until 1:8) Miguel 2 (from 1:9) SPANISH (MONOLINGUALS) Jens (from 2;0 on) Simon (from 2;0 on) GERMAN (BILINGUALS) Jens (from 2;0 on) Simon (from 2;0 on) SPANISH (BILINGUALS)

4 % (2) 3 % (4) TOTAL: 6 9 % (2) 88 % (16) 15 % (3) 82 % (14) TOTAL: 35 90 % (9) 100 % (14) TOTAL: 23 86 % (6) 94 % (17) TOTAL: 23

4.3  Phonetic parameters of prominence in contrast In this section, we first consider the values for duration of the two syllables of iambic-shaped words produced by German monolinguals (Table 3); the first syllable



Comparing the representation of iambs 

is the unstressed syllable of the iamb and the second syllable is the stressed one. In order to establish prominence, only rhymes have been taken into consideration. Because very young children tend to omit target codas in their words, especially in Spanish, we first proceeded to analyse the two vowels, the unstressed and the stressed one. The following figures show the duration values of vowels. Fig. 4 shows the mean duration of the two vowels in ms, as produced by the two Spanish children, José and Miguel, Fig. 5 shows duration for the two German children, Bernd and Marion, and Fig. 6 shows the duration values for the two bilinguals, Jens and Simon, in Spanish and in German. Finally, we present the values for the other two parameters of prominence, amplitude and pitch. 350

DURATION (GERMAN)

300 250

ms

200

Simon Jens

150 100 50 0

FIRST VOWEL

SECOND VOWEL

Figure 4.  Values for vowel duration in German produced by monolingual children 450

DURATION (SPANISH MONOLINGUALS)

400 350

ms

300 250

José Miguel

200 150 100 50 0

FIRST VOWEL

SECOND VOWEL

Figure 5.  Values for vowel duration in Spanish produced by monolingual children

 Javier Arias & Conxita Lleó

350

DURATION (SPANISH)

300 250 200 ms

Simon Jens

150 100 50 0

350

FIRST VOWEL

SECOND VOWEL

DURATION (GERMAN)

300 250 200 ms

Simon Jens

150 100 50 0

FIRST VOWEL

SECOND VOWEL

Figure 6.  Values for vowel duration produced in Spanish and German by bilinguals

A comparison of the group values for vowels (2nd column of Table 3), the unstressed and the stressed one, does not lead to any statistically significant difference, in the case of the two German monolingual children. In the case of the monolingual Spanish children, we can see a statistically significant difference in duration for Miguel. As far as the bilinguals are concerned, they do not exhibit any significant difference of amplitude or pitch. However, Simon has a significant



Comparing the representation of iambs 

difference of duration in Spanish, and Jens in German. Consideration of whole rhymes (3rd column of Table 3) shows quite a different picture in the case of the German monolingual children: both children, Bernd and Marion, have significantly different values for duration of the stressed and the unstressed rhyme. Both children clearly make use of codas in the stressed syllable, in order to make such syllables longer than the unstressed ones. The same is true in the case of the German of the bilinguals, as both Jens and Simon have statistically significant differences of stressed and unstressed rhymes, once codas are taken into consideration. Both bilinguals exhibit an analogous behavior in German to that of Bernd and Marion. Figs. 7 and 9 show duration of rhymes in German both for monolinguals (Fig. 7) and for bilinguals (Fig. 9). As for Spanish, we already found a difference between vowels in one of the monolinguals; this difference is certainly boosted when considering whole rhymes, as shown in Fig. 8. The difference in duration of the stressed rhyme vs. the unstressed one results in statistically significant differences for both monolingual children. As for the bilinguals, both children show a statistically significant difference in Spanish, too, when whole rhymes are taken into consideration, as shown in Fig. 9 and Table 3. Table 3.  Comparison of duration of stressed vs. unstressed vowels and rhymes of iambs: results of the t-tests Children

Duration of vowels

Duration of rhymes

Bernd Marion GERMAN (MONOLINGUALS)

p > 0.19649 p > 0.06266 non-significant

José Miguel SPANISH (MONOLINGUALS) Jens Simon GERMAN (BILINGUALS)

--**p < 0.00441 high significant (Miguel) *p < 0.03818 p > 0.11028 significant (Jens)

Jens Simon SPANISH (BILINGUALS)

p > 0.24588 ***p < 0.00038 high significant (Simon)

*p < 0.01979 ***p < 0.00022 significant (Bernd) and extremely high significant (Marion) ***p < 0.000483 ***p < 0.00038 extremely high significant **p < 0.00481 ***p < 0.00000065 high and extremely high significant *p < 0.04771 ***p < 0.000646 significant (Jens) and extremely high significant (Simon)

A comparison of amplitude and pitch of the stressed vs. the unstressed vowels does not lead to any statistically significant differences. The results of the t-tests for these phonetic parameters of the two vowels are shown in Table 4. They lack significance in all cases, monolinguals and bilinguals.

 Javier Arias & Conxita Lleó

Table 4.  Comparison of amplitude and pitch of stressed vs. unstressed vowels of iambs: results of the t-tests Children

Amplitude

Pitch

Bernd

p > 0.88695

p > 0.92275

Marion

p > 0.36474

p > 0.66684

GERMAN (MONOLINGUALS)

non-significant

non-significant

José

p < 0.07297

p > 0.35438

Miguel

p < 0.08337

p > 0.12448

SPANISH (MONOLINGUALS)

non-significant

non-significant

Jens

p < 0.38373

p > 0.26891

Simon

p > 0.29496

p > 0.66016

GERMAN (BILINGUALS)

non-significant

non-significant

Jens

p > 0.50282

p > 0.84340

Simon SPANISH (BILINGUALS)

p < 0.09857 non-significant

p > 0.09454 non-significant

450

GERMAN MONOLINGUALS

400 350

ms

300 250

Bernd Marion

200 150 100 50 0

FIRST RHYME

SECOND RHYME

Figure 7.  Duration of rhymes in German iambs produced by monolingual children

5.  Discussion The results regarding the acquisition of iambic-shaped words show that such words have a different development in the two languages, German and Spanish, at the earliest stages. In German, iambic-shaped words tend to be truncated during several months. At this early age, monolingual German children tend to limit

Comparing the representation of iambs 

ms



RHYME DURATION IN MONOLINGUALS (SPANISH) 450 400 350 300 250 José Miguel 200 150 100 50 0 FIRST RHYME SECOND RHYME

Figure 8.  Duration of rhymes in Spanish iambs produced by monolingual children

600

RHYME DURATION IN BILINGUALS (GERMAN)

500

ms

400 Simon Jens

300 200 100 0

350

FIRST RHYME

SECOND RHYME

RHYME DURATION IN BILINGUALS (SPANISH)

300

ms

250 200

Simon Jens

150 100 50 0

FIRST RHYME

SECOND RHYME

Figure 9.  Duration of rhymes in German and in Spanish produced by bilinguals

 Javier Arias & Conxita Lleó

their production to trochees, either syllabic or moraic trochees, i.e. disyllables or monosyllables (Lleó & Arias 2006; Lleó & Arias 2009). This behavior can be accounted for by means of the representation of iambs that was introduced under (2b), with a moraic trochee preceded by an unfooted syllable. This unfooted syllable is interpreted as extrametrical, and for some time it tends to be deleted. Once the unfooted syllable is adjoined to the PW, it is produced and iambic-shaped words cease to be a problem for German monolingual children. Moreover, a few productions of trochees for iambs emerge, but they are not numerous. This can also be explained by the representation of iambs as in (2b) in the lexicon of German children: an unfooted syllable is not eligible for stress, which prevents such iambic-shaped words from being stressed on the left-hand side. In Spanish, after being correctly produced at the earliest stages, a caesura is found within the monolingual data at about age 1;7. After this point in time, a stage appears where iambic-shaped words are realized as trochees, due to overgeneralization of the basic (syllabic) trochaic pattern of the Spanish system (Lleó & Arias 2006; Lleó & Arias 2009). According to our hypothesis, Spanish children represent iambs as in (2a), and this representation does not prevent stressing the syllable on the lefthand side, as this syllable is integrated within the foot. This stage is short, as iambs make up an important part of the Spanish children’s lexicon (in tokens, if not in types). An account of the Spanish data in terms of the quantity-insensitive iamb (Altshuler 2006) allows for a unitary approach to all iambic shapes in the language, disregarding syllable weight. While the German bilingual data do not differ from the German monolingual data, the Spanish bilingual data crucially differ from the monolingual data in two respects: in the bilingual data there is some more truncation and the truncation stage is longer, especially in the case of one of the children, Simon. Interestingly, this child hardly goes through a stage in which iambs are produced as trochees. This is again consistent with the representation (2b), assumed in H2. The other child, Jens, presents less truncation, but also incurs few cases of trochaic adaptation or stress misplacement. Thus, the two bilinguals behave differently in Spanish at the beginning in relation to truncation, but later on show a similar behavior regarding stress misplacement, as both Simon and Jens hardly incur trochaic adaptation of iambs. These apparently similar behaviors may be conditioned by different underlying representations. In the case of Simon, he seems to base his production of iambic words on representation (2b). However, the case of Jens can be explained in two different ways. On the one hand, his representation may be influenced by German and he may be analyzing iambic-shaped words by means of moraic trochees preceded by an unfooted syllable. This unfooted syllable takes longer to be acquired, and is not prone to receiving stress. On the other hand, Jens may be representing iambic words in Spanish by means of iambic feet, as Spanish



Comparing the representation of iambs 

monolinguals do. The fact that he hardly incurs any stress misplacements may be a positive result of his bilingualism, as German has a phonemic vowel length distinction, which may make bilingual children sensitive to vowel and rhyme length distinctions. In German, stressed syllables tend to have one additional mora (consonantal coda). Duration of syllables with coda is greater than duration of syllables without coda in all children. Iambic-shaped words come up at a later stage, and after being truncated, are realized target-like as regards duration from 2;2 on. In iambicshaped words, vowel duration proves to be statistically significant for Jens. Once codas are considered, duration becomes significant for Bernd and highly significant for Marion and for both bilingual children, Simon and Jens. In Spanish, the iambic words of the bilingual child Simon group together with those of the monolinguals, with duration alone playing the relevant role for prominence, whereas in the case of Jens no phonetic parameter shows statistical significance. If the various phonetic parameters are considered, certain differences between the children emerge. In the German monolinguals, no phonetic parameter of vowels seems to be statistically significant. But when duration of the whole rhyme containing a coda is measured, the stressed syllable of iambs appears to be significantly longer than the unstressed syllable. In the Spanish monolinguals, duration (but no other parameter) of vowels plays a significant role. In fact, both stressed vowels considered by themselves and whole rhymes are significantly longer than unstressed vowels or rhymes, respectively. Regarding the German of the bilinguals, vowel duration proves significant in Jens, and rhymes including the coda in Simon. With respect to the Spanish of the bilinguals, only vowel duration in the child Simon shows significant values. Summarizing the results, it can be claimed that iambs seem to be acquired at an early age, beginning at 2;2. In fact, it seems as if iambs are acquired before trochees, as a high percentage of trochees, especially in Spanish, show the effect of Final Lengthening (Lleó & Arias 2006). This poses the question whether the apparent acquisition of iambs could be due to an effect of Final Lengthening. Given that the last syllable of the phrase is lengthened, this might lead to a confounding of the results. Is it iambs that children produce, or are they rather producing trochees with a lengthened last syllable? In order to answer this question, it is necessary to compare the final stressed vowel of iambs with the non-final stressed vowel of trochees, as the former should be longer than the latter, due to Final Lengthening. Moreover, a comparison of the non-final unstressed rhyme of iambs with the final unstressed rhyme of trochees should also result in duration differences, as the former should be shorter than the latter, given that the unstressed rhyme of trochees might show Final Lengthening. Figs. 10 and 11 clarify such contrasts for monolinguals and bilinguals. Fig. 12 also shows that the final vowel of iambs

 Javier Arias & Conxita Lleó

(independently of the presence of a coda or not) is much longer than the final vowel of trochees. The values for the duration of trochees have been taken from data analyzed in Lleó and Arias (2006). All trochees analyzed corresponded to isolated utterances produced by the children. 350

SPANISH MONOLINGUALS

300 250

ms

200

José Miguel

150 100 50 0

TROCHEES

IAMBS

Figure 10a.  Comparison between the stressed rhymes of iambs and trochees: Spanish monolinguals

350

GERMAN MONOLINGUALS

300 250 ms

200

Bernd Marion

150 100 50 0

TROCHEES

IAMBS

Figure 10b.  Comparison between the stressed rhymes of iambs and trochees: German monolinguals



Comparing the representation of iambs 

250

BILINGUALS (SPANISH)

200

150 ms

Simon Jens 100

50

0 TROCHEES

250

IAMBS

BILINGUALS (GERMAN)

200

150

ms

Simon Jens 100

50

0 TROCHEES

IAMBS

Figure 10c.  Comparison between the stressed rhymes of iambs and trochees in Spanish and German: bilinguals

 Javier Arias & Conxita Lleó

350

SPANISH MONOLINGUALS

300 250

ms

200

José Miguel

150 100 50 0 TROCHEES

IAMBS

Figure 11a.  Comparison between the unstressed rhymes of iambs and trochees: Spanish monolinguals

300

GERMAN MONOLINGUALS

250

ms

200 Bernd Marion

150 100 50 0 TROCHEES

IAMBS

Figure 11b.  Comparison between the unstressed rhymes of iambs and trochees: German monolinguals



Comparing the representation of iambs 

300

BILINGUALS (GERMAN)

250

ms

200 Simon Jens

150 100 50 0

250

TROCHEES

IAMBS

BILINGUALS (SPANISH)

200 150 ms

Simon Jens

100 50 0

TROCHEES

IAMBS

Figure 11c.  Comparison between the unstressed rhymes of iambs and trochees in Spanish and in German: bilinguals 450

SPANISH MONOLINGUALS

400 350

ms

300 250

José Miguel

200 150 100 50 0

TROCHEES

IAMBS

Figure 12a.  Comparison between final vowels: Spanish monolinguals

 Javier Arias & Conxita Lleó GERMAN MONOLINGUALS

450 400 350

ms

300 250

Bernd Marion

200 150 100 50 0

TROCHEES

IAMBS

Figure 12b.  Comparison between final vowels: German monolinguals 350

BILINGUALS (SPANISH)

300

ms

250 200

Simon Jens

150 100 50 0 350

TROCHEES

IAMBS

BILINGUALS (GERMAN)

300

ms

250 200

Simon Jens

150 100 50 0

TROCHEES

IAMBS

Figure 12c.  Comparison between final vowels: bilinguals



Comparing the representation of iambs 

Table 5.  Comparison of the mean duration of the stressed rhymes of iambs to the stressed rhymes of trochees: results of t-tests Children

Duration

Bernd Marion GERMAN (MONOLINGUALS) José Miguel SPANISH (MONOLINGUALS) Jens Simon GERMAN (BILINGUALS) Jens Simon SPANISH (BILINGUALS)

*p > 0.05162 p > 0.09244 marginally significant (Bernd) ***p < 0.000429 **p < 0.00938 high (Miguel) and extremely high significant (José) *p < 0.02497 *p < 0.04582 significant p > 0.53896 p > 0.61103 non-significant

Table 6.  Comparison of the mean duration of the unstressed rhymes of iambs to the unstressed rhymes of trochees: results of the t-tests Children

Duration

Bernd Marion GERMAN (MONOLINGUALS) José Miguel SPANISH (MONOLINGUALS) Jens Simon GERMAN (BILINGUALS) Jens Simon SPANISH (BILINGUALS)

p > 0.35427 p > 0.06125 non-significant *P < 0.01484 *p < 0.02580 significant p > 0.33803 p > 0.11433 non-significant p > 0.21274 p > 0.70659 non-significant

5.1  Comparison between trochees and iambs Figures 10, 11 and 12 seem to confirm the predictions in a global manner, as unstressed rhymes of trochees are longer than unstressed rhymes of iambs, which must be attributed to Final Lengthening in the case of trochees. For all children, monolinguals as well as bilinguals, the last stressed rhyme of iambs has greater duration than the non-final stressed rhyme of trochees, which certainly results from the joint effect of stress prominence and Final Lengthening in the case of iambs. The effect of Final Lengthening is even clearer in Fig. 12, which compares the duration of final vowels. In order to show these contrasts more conclusively,

 Javier Arias & Conxita Lleó

we have conducted t-tests. First, we have applied t-tests to the comparison of the mean duration of the two rhymes of iambs, the stressed and the unstressed ones. The mean values are shown in Fig. 7 for German monolinguals, in Fig. 8 for Spanish monolinguals, and in Fig. 9 for the German of bilinguals (9a) and for the Spanish of bilinguals (9b). Table 3 shows the results of the t-tests comparing stressed and unstressed rhymes of iambs. The table shows that the difference in duration is significant in all cases, which confirms that iambs have been acquired, that is, the stressed rhyme of iambs is more prominent than the unstressed rhyme as regards duration. However, since this difference could result as an effect of Final Lengthening, we have also compared the stressed rhyme of iambs with the stressed rhyme of trochees, as the former should be more prominent than the latter in terms of duration, due to Final Lengthening. Results (see Table 5) confirm that in most cases, except for Marion, monolingual Spanish and monolingual German children, as well as bilinguals in Spanish are producing a statistically significantly longer stressed rhyme in iambs than in trochees, which must be attributed to Final Lengthening. The bilinguals do not show this difference in German. We have also compared the unstressed rhymes of iambs to the unstressed rhymes of trochees. Figs. 11 show that the latter are longer than the former, as the unstressed rhyme of trochees is submitted to the effect of Final Lengthening. However, here the results of the t-tests are not as clear-cut (see Table 6). Out of the monolinguals, only the Spanish children produce a significant difference in the unstressed syllable. With the bilinguals we did not find a statistically significant difference either in German or in Spanish. The fact that the difference between the unstressed rhymes is only significant in Spanish seems to show that Spanish children clearly distinguish final syllables. This is attributable to Final Lengthening, which is more effective in Spanish than in German. In a way, it is not surprising that unstressed rhymes show a weaker effect of Final Lengthening than stressed rhymes. All these results taken together point to target-like acquisition of stress in iambic-shaped words. 5.2  The hypotheses put to the test Going back to the hypotheses that were introduced in Section 2.3, we note the following: H1 predicted that iambs would be difficult to acquire in both languages, as the great bulk of words children first learn are trochaic. The hypothesis has been confirmed, as iambic-shaped words in general posed more difficulties than trochaic words. H2 predicted differences in the acquisition of iambs in German and Spanish, due to the different representations that such feet receive in the two languages. This was also confirmed, as in German, children tended towards truncation of the



Comparing the representation of iambs 

initial syllable of iambs, whereas Spanish children did not have much truncation. Moreover, Spanish children went through a stage in which they tended to produce iambic words with trochaic stress. This strategy was almost absent in the German children’s data. According to H3, greater values for duration, amplitude and pitch of stressed over unstressed syllables were expected in German than in Spanish. This hypothesis was not confirmed, as only duration seemed to systematically contribute to the prominence of stressed syllables, but the results for duration did not have higher values in German than in Spanish. H4 predicted no differences between bilingual and monolingual acquisition of iambic-shaped words. This was indeed the case for German, whereas in Spanish one bilingual (Jens) might have treated iambic words like the Spanish monolinguals did (representing them by means of (2a)), but the other bilingual treated them as German children did, namely with more truncation at first and no stress misplacement later on. Finally, H5 fleshed out the previous hypothesis, by proposing the plausibility that in the bilingual context, acquisition of iambs in German might be facilitated by the influence of Spanish, and/or made more difficult in Spanish, due to the influence of German. That is, this hypothesis predicts interaction between the two prosodic systems of the bilingual child, and we did find interaction, as shown with regard to H4. However, interaction only appeared in one direction, namely German seems to be influencing Spanish in the case of Simon, as this child shows more truncation and less stress misplacement than the Spanish monolinguals. Certainly, this finding is only preliminary, as the number of subjects and the number of items observed is very small. We do not know whether Simon might be influenced by German, because German is the ambient language, or whether this child might be applying the less marked or easiest solution. Certainly, the influence of German on Simon’s Spanish is very limited, and points to a result that often appears in the studies of bilingual prosody, namely that bilingual children tend to apply compromise values to their two languages. 6.  Conclusions Iambs pose difficulties both in German and in Spanish. This is reflected in the stress errors in which iambic-shaped words are levelled as trochees to adapt to the most general stress pattern of the language. Such errors are found in Spanish more frequently than in German, and more often in the monolinguals than in the bilinguals. Differences in the acquisition of iambic-shaped words in the two languages have been attested by the different degrees of truncation in the acquisition of the two

 Javier Arias & Conxita Lleó

languages. Bilingual children acquire iambic word-stress in German the same way as monolinguals. However, deviations from the monolingual pattern could be found in Spanish both in the case of Jens and Simon, and especially at the earliest stage in the case of Simon.

References Altshuler, D. 2006. Osage fills the gap: The quantity insensitive iamb & the typology of feet. Rutgers Optimality Archive Nr. 870. Berkovits, R. 1994. Durational effects in final lengthening, gapping, and contrastive stress. Language and Speech 37: 237–250. Delattre, P. 1966. A comparison of syllable length conditioning among languages. Applied Linguistics 4: 183–198. Dogil, G. 1999. The phonetic manifestation of word stress. In Word Prosodic Systems in the Languages of Europe H. van der Hulst (ed.), 273–334. Berlin: de Gruyter. Domahs, U., Wiese, R, Bornkessel-Schlesewsky, I. & Schlesewsky, M. 2008. The processing of German word stress: Evidence from the prosodic hierarchy. Phonology 25: 1–36. Everett, D.L. 2003. Iambic feet in Paumari and the theory of foot structure. Linguistic Discovery 2(1): 22–44. Face, T.L. 2001a. Intonational Marking of Contrastive Focus in Madrid Spanish. Ph.D. dissertation, The Ohio State University. Face, T.L. 2001b. Focus and early peak alignment in Spanish intonation. Probus 13: 223–246. Féry, C. 1996. German foot and word stress in OT. In Papers from the 2nd Workshop on Comparative Germanic Phonology , [Tromsø University Working Papers on Language and Linguistics 24], P. Bye, O. Lorentz & C. Rice (eds), 63–96. Tromsø: Tromsø University. Fikkert, P. 1994. The Acquisition of Prosodic Structure. The Hague: HAG. Giegerich, H.J. 1985. Metrical Phonology and Phonological Structure: German and English [Cambridge Studies in Linguistics 43]. Cambridge: CUP. Hall, T.A. 2000. Phonologie. Eine Einführung. Berlin: de Gruyter. Harris, J.W. 1983. Syllable Structure and Stress in Spanish. A Nonlinear Analysis. Cambridge MA: The MIT Press. Harris, J.W. 1991a. The exponence of gender in Spanish. Linguistic Inquiry 22: 27–62. Harris, J.W. 1991b. With respect to accentual constituents in Spanish. In Current Studies in Spanish Linguistics, H. Campos & F. Martínez-Gil (eds), 447–473. Bloomington IN: Indiana University Linguistics Club. Harris, J.W. 1995. Projection and edge marking in the computation of stress in Spanish. In The Handbook of Phonological Theory, J. Goldsmith (ed.), 867–887. Cambridge MA: Blackwell. Hayes, B. 1995. Metrical Stress Theory: Principles and Case Studies. Chicago IL: University of Chicago Press. Hochberg, J.G. 1988. Learning Spanish stress: Developmental and theoretical perspectives. Language 64: 683–706. Hualde, J.I. 2002. Intonation in Spanish and the other Ibero-Romance languages: Overview and status quaestionis. In Romance Phonology and Variation. Selected Papers From the 30th Linguistic Symposium on Romance Languages. C. Wiltshire & J. Camps (eds), 101–116. Amsterdam: John Benjamins.



Comparing the representation of iambs 

Janßen, U. 2003. Untersuchungen zum Wortakzent im Deutschen und Niederländischen. Ph.D. dissertation, Heinrich-Heine-Universität, Düsseldorf. Kager, R. 2006. Feet and metrical stress. In The Cambridge Handbook of Phonology. P. de Lacy (ed.), 195–228, Cambridge: CUP. Klein, H.B. 1984. Learning to stress: A case study. Journal of Child Language 11: 375–390. Knaus, J. & Domahs, U. 2009. Experimental evidence for optimal and minimal metrical structure of German word prosody. Lingua 119, 1380–1395. Lleó, C. & Arias, J. 2006. Foot, word and phrase constraints in first language acquisition of Spanish stress. In Optimality-Theoretical Studies in Spanish Phonology, S. Colina & F. Martínez-Gil (eds.), 470–496. Amsterdam: John Benjamins. Lleó, C. & Arias, J. 2009. The role of Weight-by-Position in the prosodic development of Spanish and German. In Phonological Domains: Universals and Deviations. J. Grijzenhout & B. Kabak (eds.), Berlin: Mouton de Gruyter. Lleó, C. & Rakow, M. 2006. The prosody of early two-word utterances by German and Spanish monolingual and bilingual children. In Interfaces in Multilingualism: Acquisition and Representation, C. Lleó (ed.), 1–26. Amsterdam: John Benjamins. Lleó, C., Rakow, M. & Kehoe, M. 2004. Acquisition of language-specific pitch accent by Spanish and German monolingual and bilingual children. In Laboratory Approaches to Spanish Phonology, T. Face (ed.), 3–27. Berlin: Mouton de Gruyter. Medina Murillo, A. 2005. Alargamiento final en español. Signos Lingüísticos 1: 45–59. Mengel, A. 1997. Das akustische Korrelat des deutschen Wortakzents. Conference at the ESSV meeting, Cottbus. Navarro Tomás, T. 1916. Cantidad de las vocales acentuadas. Revista de Filología Española 3: 387–408. Prieto, P. 2004. The search for phonological targets in the tonal space: H1 scaling and alignment in five sentence-types in Peninsular Spanish. In Laboratory Approaches to Spanish Phonology. T. Face (ed.), 29–60, Berlin: Mouton de Gruyter. Revithiadou, A. 2004. The iambic/trochaic law revisited. In Leiden Papers in Linguistics 1(1), B. Arsenijevic, N. Elouazizi, M. Salzmann & M. de Vos (eds.), 37–62. Roca, I. 1988. Theorical implications of Spanish word stress. Linguistic Inquiry 19: 393–423. Rubach, J. 1996. Shortening and ambisyllabicity in English. Phonology 13: 197–237. Schneider, K. 2007. Acquisition of word stress in German: Vowel duration and incompleteness of closure. Proceedings of the 16th International Conference of Phonetic Sciences (Saarbrücken), 1565–1568. Snow, D. 1994. Phrase-final syllable lengthening and intonation in early child speech. Journal of Speech and Hearing Research 37: 831–840. Sosa, J.M. 1999. La entonación del español: Su estructura fónica, variabilidad y dialectología. Madrid: Cátedra. Van de Vijver, R. 1998. The Iambic Issue: Iambs as a Result of Constraint Interaction. Den Haag: HAG (HIL dissertations). Vennemann, T. 1991. Skizze der deutschen Wortprosodie. Zeitschrift für Sprachwissenschaft 10: 86–111. Wiese, R. 1996. The Phonology of German. Oxford: Clarendon Press. Zonneveld, W., Trommelen, M., Jesen M., Bruce, G. & Árnason Kristján. 1998. Wordstress in West-Germanic and North-Germanic languages. In Word Prosodic Systems in the Languages of Europe, H. van der Hulst (ed.), 477–604. Berlin: de Gruyter.

Author index

A Adams, J.  59, 113, 120 Aikhenvald, A.Y.  1 Alonso, A.  32 Altshuler, D.  205, 207, 210, 222 Amastae, J.  162 Antaki, C.  145 Arias, J.  8, 205, 212, 222–224 Aristar Dry, H.  172 Atienza, E.  158–160 Auer, P.  75 B Badia Margarit, A.M.  187 Bakker, P.  17 Baumgarten, N.  125, 128, 134f., 139, 143, 147 Becher, V.  7, 125, 137 Bell, D.  139 Benet, A.  8 Berkovits, R.  209 Biber, D.  75, 128, 139 Bird, S.  172 Blas Arroyo, J.L.  155 Bloomfield, L.  3 Blum-Kulka, S.  169 Boersma, P.  193 Bomhard, A.R.  33 Bornkessel-Schlesewsky, I.  208 Bosch, L.  202 Bossong, G.  5, 20 Braunmüller, K.  1, 6, 55f., 59, 61, 63, 66, 76, 78, 101, 103 Briz, A.  172 Bruce, G.  208 Bublitz, W.  127 Bührig, K.  128 Büring, D.  138 Byrnes, H.  138 C Campbell, L.  24, 34, 36 Carrera-Sabaté, J.  188, 201

Casanovas Catalá, M.  157f. Cerrón-Palomino, R.  24f. Christensen, C.B.  120 Clyne, M.  126, 166 Coates J.  128 Cortés, S.  8 Crismore, A.  127 D Dahl, Ö.  5, 42, 47, 75 Darwin, Ch.  14, 27 De Bot, K.  56, 107 DeCesaris, J.A.  187 Delattre, P.  209–211 Delsing, L.-O.  45 Denison, D.  135 Diderichsen, P.  112 Diercks, W.  56 Diewald, G.  90 Dixon, R.M.W.  1, 15, 31, 34 Dogil, G.  209 Doherty, M.  137 Domahs, U.  208 E Erikson, E.  13, 29 Escobar, A.  164f. Espinosa, A.  199 Etxebarria, M.  157 Evans, N.  31 Everett, D.L.  205, 207, 210 F Fabricius-Hansen, C.  141 Face, T.L.  210f. Farnsworth, R.  127 Fernández-Planas, A.M.  188, 201 Féry, C.  208 Fikkert, P.  206 Fishman, J.  3 Flemming, E.  199 Foley, W.A.  32

Fredsted, E.  7, 54, 59, 62f., 67, 101–103, 106f., 112, 115, 122 G García Tesoro, A.I.  164f., 171 Geyken, A.  127 Giegerich, H.J.  207, 210 Giles, H.  3 Godenzzi, J.C.  165 Graefen, G.  129 Granda, G.d.  165 Greenberg, J.  15, 24, 33f., 36f. Green, D.  107 Grosjean, F.  56 Grote, B.  139 Guitart, J.  162 Gullbein, S.  113 Gumperz, J.J.  140, 166 H Halliday, M.A.K.  127 Hall, T.A.  208, 210 Hansen, Z.S.  103 Harris, J.W.  205, 209 Haugen, E.  3, 57, 166 Hawkins, J.  137 Hayes, B.  205, 207f., 210 Heine, B.  6, 56, 58, 81, 90, 94–96, 128 Hernández García, C.  157, 160f. Herrick, D.  187, 201 Herzog, M.  167 Hinskens, F.  1, 4, 75 Hochberg, J.G.  206 Höder, S.  6, 56, 79–81 Hoey, M.  139, 146 Holmquist, J.C.  165 House, J.  7, 125–128, 130, 136, 146 Hualde, J.I.  210f. Humboldt, W. von  26

  Author index Hunston, S.  127 Hyland, K.  129 Hymes, D.  3 J Jacobsen, J. í Lón  104 Jake, J.  105f., 115 Janßen, U.  208 Jessen, M.  207 Johanson, L.  61 Johnson, H.  172 K Kager, R.  208 Kahn  208 Kaufman, T.  1, 4, 55, 166 Kehoe, M.  211 Kerns, J.C.  33 Kerswill, P.  75 Kim, T.D.  23 Kimura, M.  27 Klee, C.  164, 171 Klein, F.  161 Klein, H.B.  206 Knaus, J.  208 Koike, D.  169 König, E.  135 Koptjevskaja-Tamm, M.  44, 75 Kotthoff, H.  138 Kranich, S.  7, 125, 137 Kristján Árnason  208 Kühl, J.  102 Kühl, K.H.  6, 60, 63, 101–103, 107, 122 Kuteva, T.  6, 42, 56, 58, 81, 90, 94–96, 128 L Labov, W.  3, 75, 162, 167, 171 Lapesa, R.  156, 162 Le Page, R.B.  61, 169 Levelt, W.  105–107 Lewin, B.  23 Li, C.N.  17, 23 Lindow, W.  61 Lleó, C.  8, 185, 187f., 190–192, 201f., 211f., 222–224 Lloyd, P.  156 Lope Blanch, J.M.  162 López del Castillo, L.  161 Lorenz, K.  29

M Mackey, W.F.  166 Maddieson, I.  202 Markey, T.L.  46 Matthiessen, C.M.I.M.  127 McWhorter, J.H.  42, 67, 75 Mead, M.  32 Medina Murillo, A.  210 Meillet, A.  4, 13f. Menendez Pidal, R.  156 Mengel, A.  209 Miestamo, M.  42 Montolío, E.  161 Mufwene, S.  3 Munné, J.Z.  162 Munske, H.H.  66 Muysken, P.  6, 18, 53f., 61 Myers-Scotton, C.  4, 6–7, 53f., 102, 105–107, 109, 115, 122f. N Navarro Tomás, T.  210 Nehls, D.  135 Nichols, J.  36 Nordenstam, K.  66 O Özçetin, D.  125, 128, 135, 147

Rozental, D.E.  43 Rubach, J.  208 Ruhlen, M.  13 S Sampson, G.  42 Sandøy, H.  119 Sano, M.  129 Sapir, E.  3 Schiffrin, D.  140 Schleicher, A.  14, 27, 35 Schlesewsky, M.  208 Schlobinski, P.  141 Schmidt, Johannes  14 Schneider, K.  209 Schreuder, R.  107 Schuchardt, H.  2, 14, 36f. Sebastián-Gallés, N.  202 Seco, M.  161 Shimamori, R.  19 Silva-Corvalán, C.  162–164, 170 Simons, G.  172 Sinner, C.  157, 159, 161, 168 Smith, C.  127 Snow, D.  64, 210 Solé, Y.R.  167 Søndergaard, B.  104 Sosa, J.M.  210f. Stilo, D.  47

P Palacios, A.  165 Palmer, F.R.  109 Paulun, D.  67 Pedersen, K.M.  103 Petersen, H.P.  6, 59f., 64, 104, 113, 116, 120f. Peterson, D.A.  36 Pla Fulquet, J.  187, 199 Pons, E.  186 Prieto, P.  210f. Probst, J.  125, 128, 139 Pusch, C.D.  172

T Tabouret-Keller, A.  169 Teich, E.  135, 143, 148 Thetela, P.  138, 139 Thomason, S. G.  1, 4, 55, 110, 156, 166, 168, 171f. Thompson, G.  127, 138f. Thompson, S.A.  23 Thráinsson, H.  109, 116, 120 Tovar, A.  30 Tricker, D.  155 Trommelen, M.  207 Trudgill, P.  42, 57, 67 Turell, T.  169

R Rakow, M.  211 Recasens, D.  187, 194, 199 Revithiadou, A.  208 Ridell, K.  66 Roca, I.  205, 209 Romaine, S.  4, 156, 166 Ross, J.R.  67

V Van Coetsem, F.  4, 67 Van de Vijver, R.  208 Vann, R.E.  7, 153, 157–161, 166, 168f., 173, 175 Vennemann, Th.  15, 37, 207, 210 Viaplana, J.  187

Author index    Vikner, S.  109 Vila, F.X.  186 Vila, M.R.  161 Vila Pujol, M.R.  157 W Weinreich, U.  3–5, 74, 128, 153, 155, 166f.

Wesch, A.  155, 157, 159 Westergaard, A.  64 Wetherell, M.  145 White, P.R.R.  48, 129 Widell, P.  120 Wiese, R.  207f., 210 Wilson, R.  166 Winge, V.  104, 112

Winter, E.  146 Woolard, K.  169 Z Zamora Vicente, A.  156 Zentella, A.C.  170 Zifonun, G.  81 Zonneveld, W.  207

Subject index

A adjectives  41, 43, 48, 50, 135, 137, 158, 160 adstratum  156f. American Spanish  162f. Amerindian languages  155, 162, 165 amplitude  211, 213, 217–220, 231 Ancient Nordic  91 Andean Spanish  164 appositive relative clauses  73, 86–90, 92–94 Arabic   15, 19–21, 155 attrition  2, 54, 106f. Ausbau  73f., 78–81, 87, 89, 93, 96 Australia  30f. Aymara  24f., 34 B Basque  36, 155–157 bilingual continuum  164, 170 bilingual speaker  104, 128, 135, 148, 155 bilingualism  1, 4f., 7, 53, 60, 62, 95, 102–105, 121, 155, 165f., 223 bimoraic  208 borrowing  1f., 4, 5, 13, 21, 34–36, 55, 155, 157, 162 buffer zones  41, 46 C Castilian Spanish  156f., 168 Catalan  8, 155, 157–161, 168f.–169, 176, 185–188, 190, 192–194, 199, 201–203 Catalan Spanish  157–160, 168f. Catalan vowels  186f., 192f., 201f. Catalonia  157–159, 161, 168f., 185f.

Chinese  19, 21, 23, 26, 175 code copying  2, 61 code mixing  6, 53–55, 58, 65–67 code mixing hierarchy  53, 66 code switching  2f., 5, 58, 60–62 co-existence  43 cognates  66 common pragmatic norm  56 Communication Adaptation Theory (CAT)  3 communicative norms  125, 127f., 142, 146 communicative preferences  126, 129f., 134, 146f. competition  41–44, 46, 48, 51 complementizer  54, 162 complexity  5f., 14, 32, 41–43, 45–47, 51, 76, 81, 96, 131 composite matrix language  54 concessive conjunctions  125, 128, 137, 139 congruent lexicalization  53f. contact dialects  154, 164f., 170, 172f., 175, 177 contact-induced grammaticalisation  94f. contact-induced language change  73, 76, 156, 162, 164, 166f., 170 contact-induced variation  1 convergence  1–8, 11, 13, 15, 22–24, 26, 29, 31–36, 38, 42, 53f., 57, 61, 63–67, 71, 73–75, 96, 101, 105f., 108–113, 115, 117, 119, 120, 122, 125, 147f., 154, 167, 177, 203 convergence varieties  6, 57 conversation  55, 104, 139f., 155

copular verbs  162 covert lexical code switching  61 covert replication  62 covert translation  7, 125, 147 Cree  16f., 19 creole languages  2f., 15f., 22, 61 creolisation  15 creoloid  57 critical contexts  85, 90f., 95 cultural filter  7, 130, 134, 140, 145, 147 D Danish  6, 41, 43, 48, 49–51, 53–64, 66–67, 77, 101–105, 107–122 Dano-Norwegian  57, 67 definite articles  6, 41f., 47f., 50f., 158f. demonstrative pronouns  43, 66 dialects  2, 6, 35, 50, 57–58, 65–67, 76, 119, 154, 164f., 167, 170–173, 175, 177, 186, 188 diglossia  2, 78f. dispersion of schwa  199 districts in Barcelona  185, 188 divergence  1–8, 11, 13, 15, 22–23, 26, 29, 31–34, 38, 42, 71, 73–75, 96, 125, 134, 137, 148, 154, 167, 177 duration  199, 205, 207, 209–211, 213, 216–220, 223f., 229–231 E Eixample  187, 189–193, 195–201, 203

  Subject index English  7, 15, 17, 20f., 31, 37, 44, 55, 57, 62, 82–83, 111, 125–148, 155, 163f., 199, 206, 208 epistemic modal verbs  147 F F1 values  188, 194, 197 F2 values  194, 197, 199, 201 Faro-Danish  101–103, 105, 107f., 110, 113f., 116, 119, 120–122 Faroe Islands  53, 55–57, 59f., 62, 101–103, 105, 108 Faroese  6, 55–57, 59, 62–64, 77, 101, 103–105, 108–117, 119, 120–122 free relative clauses  84, 86, 88–93, 96 French  15–17, 21, 27, 33, 44, 50, 61, 126, 159 G Galician-Portuguese  155 genealogical classification  13–16, 22, 26, 29, 33, 34, 36 genealogical model  13–16 geographical distribution  41, 46, 50 German  6–8, 38, 41, 46, 53f., 56–65, 67, 73, 78, 82f., 101–105, 107, 110–122, 125–148, 154, 185, 205–220, 222–232 German vernaculars  53 German-Danish border region  53, 57 Global English  126 Gràcia  187, 189–201, 203 grammatical replication  6, 56, 73, 81, 94, 96 grammaticalisation  90f., 94–96 Guatemalan Spanish  165 H Hanseatic area  46, 78 Hanseatic League  46, 60, 66, 78 hedging  129 historical-comparative linguistics  14

hybrid relativisation strategy  84 I iambic-shaped words  8, 205–207, 209–214, 216, 220, 222–223, 230f. iambic-trochaic law  205f. Iambs  8, 205–207, 209–213, 215f., 219f., 222–227, 229–231 indefinite adjectives  158, 160 individual bilingualism  7, 103–105, 121 innovations  7, 73, 79, 96, 153–157, 160–162, 164–172 instrumental bilinguals  165 interactional styles  154, 167, 177 interference  1, 4f.–5, 79, 142, 148, 153, 155f., 166 interrogative sentences  158f. intralingual variation  74–76 J Japanese  19, 20–24, 26 jargons  67 K Korean  22–24, 26 Kultursprachen  18f. L language contact  1–3, 5–7, 11, 15, 22, 24, 31, 37, 41f., 44, 50f., 55, 57f., 73–76, 93, 101–106, 112, 121–123, 125, 153–155, 158, 161f., 164–167, 170–173, 176f. language mixing  5f., 16, 19, 21f., 58 language mixture  4 language standardisation  58, 79 Latin  6, 20, 32f., 38, 56, 73f., 78–81, 93–96, 112, 156, 174, 186, 206, 209 lexical code switching  58, 60f. lexico-syntactic calques  163f., 170 linearisation  64 Lingua franca  7, 78

linguistic areas  74f., 96 linguistic creativity  56, 166 linguistic norms  66, 155, 160, 166 literacy  79 local (linguistic) identity  57 long-term accommodation  66 Los Angeles Spanish  162f. Low German  41, 46, 61, 63, 67, 78, 104 M macrofamilies  33 Mayan languages  162, 176 Media Lengua  18–20, 22 medial split  73–74, 79–80, 96 medial varieties  76 Medieval Latin  78 metaphor  13–15, 27, 29, 31, 145 metrical phonology  8, 205 Michif  16–19 minority language  60, 102, 105 mixed languages  2, 4, 16, 20f., 67 modal strength  129–135, 147 modality  7, 109, 125, 127, 132, 134f., 148 model language  2, 55f., 79, 94f. Modern Swedish  56, 80, 82–84, 89f., 92f. monolingual varieties  158–163 monolinguals  54, 155–157, 165, 167, 169–171, 205, 212–216, 219, 220, 223f., 226–232 multilingualism  1, 5, 31, 53, 59, 73f., 85, 101, 105, 125, 154, 185, 205, 212 multiple-word units  163 mutual influencing  1 N national linguistic norm  60 new norms  156, 161, 164 norm  32, 55–58, 60, 158, 195 Northern Germany  101f., 121 Norwegian  41, 45, 48–50, 57, 58, 61, 66f., 77, 119 Nou Barris  187, 189–193, 195–199, 201–203 NP  118

Subject index    O Old Norse  91, 104, 112, 121 Old Swedish  6, 56, 73f., 76–91, 93–96 overt replication  63 P Paulunsch  67 penultimate  205, 209 performance phenomena  55 Persian  19, 20–22 Peruvian Spanish  162 Petuh  67 phonological identity  62 phonology  8, 42, 55, 104, 162, 185, 187, 192f., 203, 205 pidginisation  67 pitch  207, 209–211, 213, 217–220, 231 popular science texts   143 popular scientific writing  147 pronominal relative clauses  81, 83, 85f. Q quantity differences  205 Quechua  18, 24f., 34, 162, 164, 171, 176 R regional varieties  65–67 regional vernaculars  65 relative adverbs  83f., 90 relative clauses  73, 81–96 relative pronouns  83f., 90, 96 relative subjunctions  82, 91 relativisation  73f., 80–82, 84–86, 88f., 91–94, 96 relexification  15f., 18, 20, 22, 164 replica grammaticalisation  95 replica language  2, 56, 58, 94f. restrictive relative clauses  86, 88f. rhyme duration  205 Romania  32f. runic script  78 S Scandinavian  6, 41f., 45–51, 57, 73f., 77, 104, 116 semantic convergence  108–111 sentence-initial concessive conjunctions  125, 128, 137

similarity  1, 24, 26, 66, 96, 134 societal linguistic norm  55 South Jutish  57, 61f., 67 Spain  153, 156f., 167, 169f., 176, 186 Spanish  7f., 18, 21, 27, 30, 126, 153–173, 176, 179, 185–188, 192f., 201–203, 205–207, 209–220, 222–227, 229–232 spoken  7, 18, 30, 45f., 50, 60, 67, 73, 75–81, 87, 92f., 96, 101, 103, 138, 140f., 154, 157–159, 161f., 165, 168–170, 172–177, 186f. spontaneous speech innovations  153–157, 164, 166–172 standardisation  6, 58, 65, 79 stress acquisition  205f. stress misplacement  214, 216, 222, 231 stylistic  43, 141 subjunctional relativisation  82, 85, 88 substratum  156f., 166 superstratum  156f. suppletive patterns  46 suppletive verbs  41, 46 Swedish  6, 24, 41, 43, 46, 48–51, 56, 66, 73f., 76–96, 119 Sydslesvigdansk  57f., 103 syllable  205–211, 213f., 216f., 219, 222f., 230f. symmetrical bilinguals  165 syntactic convergence  111, 113, 115, 117, 119 system complexity  42 systemic language change  7, 154–158, 162, 164–166, 172, 176f. T textual norms  79, 126, 137–143, 145 transfer  1f., 4f., 7, 30, 55f., 60, 76, 79, 81, 93f., 102, 107, 110, 112–114, 118, 122, 128, 145, 153, 162, 166, 168, 170 translation  1, 7, 22–24, 59f., 73, 79, 110f., 113, 125–134, 136–138, 142–144, 146–148

translation evaluation model   127 translational equivalents  80, 94, 144 translingual stimulus  169 tree model  2, 13f., 26, 29, 30, 32, 77 triglossic distribution  78 trochaic feet  206, 210 trochees  8, 205–210, 212f., 216, 222–227, 229–231 truce  43, 45 truncation  8, 205, 207, 214–216, 222, 230f. U uniformitarian principle  74f. unstressed syllables  209, 211, 231 V variation  1–5, 46, 48f., 53, 58f., 65, 74–76, 162, 165, 183, 186, 190, 199 varieties  1, 4–6, 35, 41, 45f., 49f., 53f., 57f., 65–67, 71, 73, 75f., 78–80, 92, 96f., 153, 155, 158–163, 165, 170f., 176f. variety-specific  76, 80, 96 verbal frame  62–64 verbs  17–20, 41, 46, 64, 101, 104, 108, 110–113, 116, 120f., 125, 135f., 147, 158 vowel contrast  185 vowel production  187, 190, 194 VP  107f., 111–114, 117, 119, 122f. W waves  47, 51, 186 West Germanic  41, 45, 48, 50, 77 written  1, 5–7, 19, 21, 30, 32, 46, 50, 56–60, 71, 73–81, 87, 90, 92f., 95, 96, 103, 127, 138f., 141, 157 Y Yucatan Spanish  162 Z zero-marked relative clauses   82f., 85f., 88, 90

In the series Hamburg Studies on Multilingualism the following titles have been published thus far or are scheduled for publication: 11 Rinke, E. and Tanja Kupisch (eds.): The Development of Grammar. Language acquisition and diachronic change. Expected Forthcoming 10 Gabriel, Christoph and Conxita Lleó (eds.): Intonational Phrasing at the Interfaces. Cross-linguistics and bilibgual studies in Romance and Germanic. Expected Forthcoming 9 Meyer, Bernd and Birgit Apfelbaum (eds.): Multilingual Communication at the Workplace. Expected Forthcoming 8 Braunmüller, Kurt and Juliane House (eds.): Convergence and Divergence in Language Contact Situations. 2009. viii, 241 pp. 7 Siemund, Peter and Noemi Kintana (eds.): Language Contact and Contact Languages. 2008. x, 358 pp. 6 Thije, Jan D. ten and Ludger Zeevaert (eds.): Receptive Multilingualism. Linguistic analyses, language policies and didactic concepts. 2007. x, 328 pp. 5 Rehbein, Jochen, Christiane Hohenstein and Lukas Pietsch (eds.): Connectivity in Grammar and Discourse. 2007. viii, 465 pp. 4 Lleó, Conxita (ed.): Interfaces in Multilingualism. Acquisition and representation. 2006. xiv, 284 pp. 3 House, Juliane and Jochen Rehbein (eds.): Multilingual Communication. 2004. viii, 359 pp. 2 Braunmüller, Kurt and Gisella Ferraresi (eds.): Aspects of Multilingualism in European Language History. 2003. viii, 291 pp. 1 Müller, Natascha (ed.): (In)vulnerable Domains in Multilingualism. 2003. xiv, 374 pp.