Linguistic Universals and Language Variation 9783110238068, 9783110238051

The volume explores the relationship between linguistic universals and language variation. Its contributions identify th

233 56 4MB

English Pages 480 Year 2011

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Table of contents
List of contributors
Universals and variation: an introduction
Part 1 Varieties and cross-linguistic variation
Causal clauses: a cross-linguistic investigation of their structure, meaning, and use
Two euroversals in a global perspective: auxiliation and alignment
The diachronic development of article-possessor complementarity in the history of Italian and Portuguese
Variation as a window on universals
Part 2 Contact-induced variation
Gender and contact – a Natural Morphology perspective on Scandinavian examples
Universals of structural borrowing
Part 3 Methodological issues of variation research
Variation and reproducibility in linguistics
Parameters of morphosyntactic variation in World Englishes: prospects and limitations of searching for universals
Comparing varieties of English: problems and perspectives
Part 4 Variation and linguistic theory
Norwegian main clause declaratives: variation within and across grammars
Syntactic variation and change in Romance: a Minimalist approach
A localistic approach to universals and variation
Lexicon, phonology and phonetics. Or: rule-based and usage-based approaches to phonological variation
Index
Recommend Papers

Linguistic Universals and Language Variation
 9783110238068, 9783110238051

  • 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

Linguistic Universals and Language Variation

Trends in Linguistics Studies and Monographs 231

Editor

Volker Gast Founding Editor

Werner Winter Editorial Board

Walter Bisang Hans Henrich Hock Matthias Schlesewsky Niina Ning Zhang Editor responsible for this volume

Hans Henrich Hock

De Gruyter Mouton

Linguistic Universals and Language Variation

Edited by

Peter Siemund

De Gruyter Mouton

ISBN 978-3-11-023805-1 e-ISBN 978-3-11-023806-8 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data Linguistic universals and language variation / edited by Peter Siemund. p. cm. ⫺ (Trends in linguistics: studies and monographs ; 231) Includes bibliographical references and index. ISBN 978-3-11-023805-1 (hardcover : alk. paper) 1. Language and languages ⫺ Variation. 2. Linguistic universals. I. Siemund, Peter, 1965⫺ P120.V37L5194 2011 417⫺dc22 2011009502

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, Berlin/New York Printing: Hubert & Co. GmbH & Co. KG, Göttingen ⬁ Printed on acid-free paper Printed in Germany. www.degruyter.com

Table of contents Universals and variation: an introduction Peter Siemund 

1

Part 1. Varieties and cross-linguistic variation Causal clauses: a cross-linguistic investigation of their structure, meaning, and use Holger Diessel and Katja Hetterle 

23

Two euroversals in a global perspective: auxiliation and alignment Michele Loporcaro 

55

The diachronic development of article-possessor complementarity in the history of Italian and Portuguese Tanja Kupisch and Esther Rinke 

92

Variation as a window on universals Sali A. Tagliamonte 

128

Part 2. Contact-induced variation Gender and contact – a Natural Morphology perspective on Scandinavian examples Hans-Olav Enger 

171

Universals of structural borrowing Yaron Matras 

204

Part 3. Methodological issues of variation research Variation and reproducibility in linguistics Walter Bisang 

237

Parameters of morphosyntactic variation in World Englishes: prospects and limitations of searching for universals Bernd Kortmann and Benedikt Szmrecsanyi 

264

vi

Table of contents

Comparing varieties of English: problems and perspectives Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund 

291

Part 4. Variation and linguistic theory Norwegian main clause declaratives: variation within and across grammars Kristin Melum Eide and Hilde Sollid 

327

Syntactic variation and change in Romance: a Minimalist approach Guido Mensching and Eva-Maria Remberger 

361

A localistic approach to universals and variation Brian D. Joseph 

404

Lexicon, phonology and phonetics. Or: rule-based and usage-based approaches to phonological variation Frans Hinskens 

425

Index 

467

List of contributors Walter Bisang University of Mainz, Department of English and Linguistics [email protected] Julia Davydova University of Hamburg, Linguistic Diversity Management in Urban Areas (LiMA) [email protected] Holger Diessel University of Jena, Institute of British and American Studies [email protected] Kristin Eide NTNU Trondheim – Norwegian University of Science and Technology, Department of Scandinavian Studies and Comparative Literature [email protected] Hans-Olav Enger University of Oslo, Department of Linguistics and Scandinavian Studies [email protected] Katja Hetterle University of Jena, Institute of British and American Studies [email protected] Michaela Hilbert University of Bamberg, Institute of British and American Studies [email protected] Frans Hinskens Meertens Institute (KNAW) [email protected] Brian Joseph Ohio State University, Department of Slavic Languages and Literatures [email protected] Bernd Kortmann University of Freiburg, Department of English [email protected]

viii

List of contributors

Tanja Kupisch University of Hamburg, Institute of Romance Studies [email protected] Michele Loporcaro University of Zurich, Institute of Romance Studies [email protected] Yaron Matras University of Manchester, School of Languages, Linguistics and Cultures [email protected] Guido Mensching Free University of Berlin, Institute of Romance Languages and Literatures [email protected] Lukas Pietsch University of Hamburg, Department of English and American Studies [email protected] Eva-Maria Remberger University of Konstanz, Department of Romance Studies [email protected] Esther Rinke University of Hamburg, Institute of Romance Studies [email protected] Peter Siemund University of Hamburg, Department of English and American Studies [email protected] Hilde Sollid University of Tromsoe, Department of Linguistics [email protected] Benedikt Szmrecsanyi University of Freiburg, Department of English [email protected] Sali Tagliamonte University of Toronto, Department of Linguistics [email protected]

Universals and variation: an introduction Peter Siemund 1.

Universals and variation

This volume focuses on two notions that appear incompatible at first sight – if not contradictory: universals and variation. Linguistic universals ideally are generalizations that capture properties of language or languages that are essential to and stable across all possible languages and language types. In principle, they even extend to languages that are not attested, and hence also apply to languages that are extinct or have not yet come into existence. Linguistic universals also define the scope within which languages can vary and make predictions concerning the co-occurrence of structural properties. Language variation, in contrast, is about the variable parts of language and languages and students of variation are interested in charting the range of variation and in correlating variation, i. e. a choice out of several competing options, with external or internal factors that can be used as predictors for the observable variation. Linguistic universals have predominantly been discussed within Language Typology and Universal Grammar, albeit based on different conceptions of what constitutes a linguistic universal. Language variation is a domain traditionally associated with dialectology and socio-linguistics, and more specifically variationist linguistics. The present volume continues a highly successful line of investigation into the relationship between universals and variation in different domains of language and from diverse theoretical perspectives and methodological approaches. Among its most important predecessors are Cornips and Corrigan (2005), Dufter et al. (2009), Filppula et al. (2009), Good (2008), Hinskens et al. (1997), Kortmann (2004), Nevalainen et al. (2006), Scalise et al. (2009), Siemund and Kintana (2008). Of course, this list is not complete and could easily be extended. Linguistic universals are highly theory dependent and can hardly be discussed outside a specific model or framework. The two major strands of linguistic research in which universals of language are currently discussed can broadly be characterized as either functionalist-inductive or formalist-deductive, instantiated by Language Typology and Universal Grammar respectively (cf. Newmeyer 1998; Siemund 2009). Each of these frameworks has developed its own conception of universals and modes of

2

Peter Siemund

explanation. A fundamental assumption of Universal Grammar is that linguistic universals are innate, i. e. a more or less direct consequence of the genetic endowment of the human species. Universal properties of linguistic systems in this sense are, for example, constituent structure and recursion (cf. Roeper 2007). In Language Typology and in Functionalism more generally, universals, by contrast, are viewed as a result of constraints on language use (e.g. speaker goals, processing restrictions, memory limitations, time constraints, etc.). The two approaches also differ in what they admit as an explanation of linguistic observations, as Universal Grammar prefers theory-internal explanations while Functionalism holds that explanation must be external to the linguistic system – even though recent approaches to formal grammar do offer explanations in terms of ‘interface conditions’ that resemble external explanations to a considerable extent. As pointed out above, language variation is a domain traditionally associated with sociolinguistic and dialect studies. This conventional association has somehow clouded the fact that various other fields of investigation are also concerned with variation in rather essential ways (cf. Bisang 2004). First and foremost we need to mention language typology with its focus not only on linguistic universals but cross-linguistic variation. Besides their interest in universals, typologists take a keen interest in charting the scope of cross-linguistic variation (cf. Haspelmath et al. 2005). There are at least four further domains of investigation that crucially involve language variation. These are language contact, language change, language acquisition and the investigation of grammatical (i. e. system-internal) variation (cf. Siemund and Kintana 2008). Language contact studies investigate modifications and innovations (i. e. pieces of variation) that are the result of language contact processes. Moreover, they try to make predictions with respect to the scope of variation that language contact can induce (Heine and Kuteva 2005; Thomason 2001; Winford 2003). Studies on language change analyze how synchronic variation can lead to diachronic change. Language acquisition processes – both first and second acquisition – introduce variation. These are eventually filtered out in successful first language acquisition, but may remain as so-called ‘fossilizations’ in second language acquisition. In other words, so-called ‘interlanguages’ can be seen as specific instances of language variation. Grammatical variation in the sense used here also refers to options built into grammatical subsystems whose distribution is determined by grammar-internal and also functional constraints (cf. Rohdenburg and Mondorf 2003). Linguists seriously interested in language variation try to bring order into the observable spectrum of variation that often looks chaotic at first

Universals and variation: an introduction

3

sight. They are interested in identifying generalizations that determine the limits of variation. It is probably fair to say that all approaches to language variation with theoretical aspirations per se try to detect the systemic aspects behind the observable variation. This is reasonable (and indispensable) as theories want to make predictions, and this in turn, necessitates systemic aspects. These do not necessarily have to be deterministic, but may just as well be of a probabilistic nature trying to assess the statistical impact of a factor or a factor group on an instance of variation. These approaches develop statistical models of variation. At the same time, variation research is interested in the genesis of variation, i. e. the emergence of new grammatical categories and constructions. For example, language contact can generate variation while language change proceeds as speakers identify new patterns in the pool of available variation (via reanalysis and analogy). Over the past ten or so years, the boundaries between the research paradigms introduced above have become increasingly blurred in so far as there has been – and still is – tremendous cross-fertilization in terms of the methodologies employed and also the modes of explanation. For example, much of linguistic research has become radically empirical, and this shift towards empiricism is by no means restricted to work in the functionalistic paradigm, but has been successfully taken up by researchers working in formalist paradigms, too. A good example is the recent volume by Cornips and Corrigan (2005), though empirically based work in the formalist paradigm has a much longer history (cf. the work by Anthony Kroch, Susan Pintzuk and Anthony Warner, to name just a few important scholars). In the present volume it is perhaps the contribution by Kupisch and Rinke that illustrates this approach best. As a consequence of this turn towards empiricism, the methods of data analysis have become substantially more sophisticated. Tools for statistical analysis make sense where extensive data sets need to be tackled, and such data sets have a long tradition in socio-linguistics and language acquisition studies. In these research fields the use of statistical analyses has more or less become a benchmark of good scientific practice. Extensive data sets have increasingly come to be used in language typology, language contact studies and in research devoted to grammatical variation, too, and in these research fields we also find statistical tools used regularly today. In the present volume, no fewer than half of the contributions present statistical analyses of the data, including significance testing, logistic regression and principal component analysis. These statistical tools are tremendously useful for filtering out the noise and identifying the systemic aspects concealed in seemingly chaotic data sets.

4

Peter Siemund

The research fields put together in this volume show remarkable overlap in the modes of explanation they offer for their empirical findings and generalizations. One explanatory tool that has gained widespread acceptance and use is implicational hierarchies (or scales). Originally perhaps best known from language typology, implicational hierarchies now can be regularly found in socio-linguistic studies, in dialectology, in language contact studies and in studies on grammatical variation. To be sure, one needs to be careful in making statements who used them when and where first. For instance, even though implicational hierarchies have enjoyed a widespread use in typological studies for many years, they are discussed as early as 1980 in the dialectological monograph by Chambers and Trudgill. Implicational hierarchies also figure prominently in the contributions to the present volume and are explicitly referred to in the papers by Michele Loporcaro, Sali Tagliamonte, Yaron Matras, Walter Bisang, Bernd Kortmann and Benedikt Szmrecsanyi, Julia Davydova et al., Frans Hinskens. Of course, these implicational hierarchies cover different domains and are formulated on the basis of restricted language samples. Nevertheless, they explicitly or implicitly claim universality. Loporcaro proposes an implicational hierarchy that explains the use of the perfective auxiliaries ‘have’ and ‘be’ in relation to the immediate syntactic context. Tagliamonte discusses a hierarchy of noun types and its predictive power concerning subject-agreement phenomena. Yaron Matras proposes a wide range of new implicational hierarchies restricting borrowing processes (borrowing hierarchies). Davydova et al. test the validity of established typological hierarchies, such as the animacy hierarchy and the person hierarchy, on varieties of English.

2.

Structure of the volume

While it is not possible within the confines of the current volumes to address all aspects of linguistic universals and language variation, the articles assembled here cover many of the issues introduced above. As far as the fields of inquiry are concerned that are interested in language variation, there are papers on cross-linguistic variation, syntactic variation across different varieties of one language, and variation across genetically related languages. Two papers explicitly address variation introduced through language contact and nearly all papers concern themselves with grammar-internal variation. The issue of variation resulting from language acquisition is also addressed in a few places.

Universals and variation: an introduction

5

On the theoretical side, the articles span the whole range from formalist to functionalist approaches and I explicitly introduce this volume as an attempt to help to bridge this often rather artificial theoretical divide. I find it revealing to see similar grammatical domains treated from different theoretical perspectives and in some cases the division between these two broad theoretical approaches becomes rather blurred. Studying variation phenomena with the overall aim of contributing to linguistic universals research requires sound methodologies and reliable empirical data. There can be no doubt that all papers included here fulfill these requirements, but some of them explicitly address methodological issues and point to limitations on what the empirical data can tell us about linguistic universals and what our methodologies can possibly reveal (cf. section 2.3). These papers are a warning against jumping to fitting conclusions where several alternative interpretations are possible or where conflicting data points appear had one analyzed the data in slightly different ways or simply considered related data sets. In view of the empirical, methodological and theoretical scope of the papers, I have organized the volume into four broad sections. These areas are (i) varieties and cross-linguistic variation, (ii) contact-induced variation, (iii) methodological issues of variation research and (iv) variation and linguistic theory. I introduce and discuss these areas in what follows. It goes without saying that all papers are relevant beyond the section that they have been assigned to and that some papers can be assigned to more than one section. I also try to explicate these broader connections.

2.1.

Varieties and cross-linguistic variation

Over the past ten or so years it has become increasingly obvious that researchers interested in language variation (i. e. dialectal variation, but also socially conditioned variation) share important scientific goals and methodologies with scholars working in the field of language typology (cf. Bisang 2004). Since typologists usually do comparative work on samples of (genetically unrelated) languages, the similarities are most obvious in those cases when dialectologists and sociolinguists do comparative work across dialects and other varieties of some language. As far as I can see the approaches taken by these scholars have the following in common: (i) they look for patterns in the data based on formal and functional similarities and equivalences, (ii) they offer hypotheses on determinants constraining the observable variation and (iii) they interpret these constraints within a theoretical framework thereby advancing the model of language

6

Peter Siemund

they are working on. The last point especially is subject to individual preferences and belief systems. These commonalities are nicely illustrated through the papers by Holger Diessel and Katja Hetterle, Michele Loporcaro, Tanja Kupisch and Esther Rinke, and Sali Tagliamonte. While Diessel and Hetterle work on the basis of a cross-linguistic sample of genetically unrelated languages (a ‘probability sample’, to be precise), the other papers in this section are concerned with variation within one genetic phylum or even one language. Loporcaro compares dialects and languages of the Romance family, where the distinction between language and dialects is notoriously difficult to draw. Kupisch and Rinke compare Italian and Portuguese, both along the synchronic and diachronic dimension. Tagliamonte analyzes variation across several English dialects on the British Isles and in North America. The papers collected in this section look for patterns in different grammatical domains. Diessel and Hetterle offer a cross-linguistic survey of causal adverbial clauses focusing on their position relative to the associated main clause as well as their integration into the main clause. Loporcaro’s paper deals with the distribution of the auxiliaries ‘have’ and ‘be’ in perfect constructions – a well-known parameter of variation across dialects and languages of Romance. Kupisch and Rinke study the combinatorial properties of determiners and possessive pronouns in Present and Old Italian and Portuguese. Sali Tagliamonte in her paper provides an analysis of the distribution of English was and were focusing on contexts of default agreement, i. e. contexts where a pronominal subject in the plural occurs together with a verb form in the singular (was instead of were). All papers provide highly detailed empirical data and are based on statistical analyses of large data sets. The authors of the contributions in this section develop very clear and convincing ideas concerning the factors constraining the observable variation. Diessel and Hetterle find several universal tendencies in the grammar of adverbial causal clauses: they are usually postposed to the associated main clause; there is an intonational break between main clause and causal clause; the causal clause often contains verb forms and an argument structure found in main clauses. On the whole, adverbial causal clauses are comparatively loosely integrated into complex sentence structures. Loporcaro shows that ‘have’/‘be’ auxiliation across Romance languages and dialects is highly sensitive to the alignment properties of the respective linguistic systems (nominative/accusative alignment, ergative alignment), and less so to the semantic properties of the subject and the predicate as well as the overall conceptualization of the proposition expressed. Kupisch and Rinke provide detailed accounts of the distribu-

Universals and variation: an introduction

7

tion of determiner/possessive combinations from Old Italian to Modern Italian and Old Portuguese to Modern Portuguese. Unlike in the development from Old to Modern Italian, they find a strong increase in the occurrence of determiner/possessive combinations in the historical documents of Portuguese reflecting a systematic expansion of the definite article to new contexts. Tagliamonte studies a whole battery of constraints on default agreement including the grammatical person of the subject, negation, conjoined pronominal subjects, demonstrative pronouns, relative and indefinite pronouns. It is clear from her study that default agreement occurs in all varieties of English though the constraints identified show highly variable strengths and may not even be relevant for some varieties. Diessel and Hetterle explain the special formal properties of causal adverbial clauses in terms of the discourse structure in which they are used. The observable cross-linguistic similarities result from a common discourse structure that give rise to the use of such clauses. Causal clauses are frequently used to support a disputed proposition and thus function as assertions. This, in turn, explains their loose integration into a complex clause structure. The mode of explanation is clearly functional, as it is proposed that speakers can find similar linguistic solutions to similar communicative challenges. Loporcaro discusses the distribution of ‘have’ and ‘be’ as auxiliaries in perfective contexts against the background of socalled ‘euroversals’, i. e. common traits of European languages, addressing two euroversals taken from van der Auwera (1998), namely “the use of ‘have’ and ‘be’ as auxiliaries” and “accusativity” (accusative alignment). On the basis of the data drawn from Romance languages and dialects, Loporcaro argues that the two euroversals are not independent of one another, as previously assumed, and that European languages are not consistently accusatively aligned as he also finds actively/inactively aligned features/properties. Finally, Loporcaro stresses that these results are highly theory-dependent and that the relationship between ‘have’/‘be’auxiliation and alignment requires a specific theoretical perspective to become visible. Kupisch and Rinke approach the synchronic and diachronic variation observable in the Italian and Portuguese noun phrase from a theoretical angle that presupposes a parametric distinction between ‘adjectival-genitive languages’ and ‘determinative-genitive languages’. They argue that the variation they find in their data does not threaten a binary division into these two language types as it can be attributed to processes that do not interfere with the parametric distinction. Tagliamonte argues that default agreement occurs in all varieties of English and that in this sense it is a universal property of English. Variation of default agreement, i. e. different patterns of it, turns out to be a function of the context

8

Peter Siemund

in which it occurs. For example, default agreement with plural noun phrases is widespread and in some locales categorical, giving rise to the well known Northern Subject Rule.

2.2.

Contact-induced variation

The articles in this section concern themselves with variation that arises as the result of language contact. It is probably fair to say that today the main focus of language contact studies lies on finding universally valid constraints on processes of borrowing and transfer, i. e. processes introducing linguistic material (i. e. forms, functions, frequencies, combinatorial properties) from one language to another under conditions of language contact (“code copying” in the sense of Johanson 2008). Universals of language contact thus constrain the amount of variation as well as the type of variation that language contact processes can introduce. A theory of language contact ideally makes predictions regarding what can happen in such situations and what is ruled out, or at least what is likely or less likely to happen. It can be considered an established fact that there are no simple universal constraints on language contact processes that would allow us to formulate watertight predictions concerning the outcome of any situation of language contact. Language contact situations are extremely heterogeneous and complex, and one needs to consider a substantial number of factors to offer convincing explanations for language contact phenomena or to make predictions on the results of on-going language contact processes. I briefly review here the most important factors and then discuss how the papers placed in this section work towards a refinement of these factors. The relevance of these papers for the volume as a whole lies in the overlap of these factors with those constraining the other types of variation discussed here (dialectal variation, cross-linguistic variation, grammar-internal variation, etc.). A long-standing dispute in language contact studies centers around the notions of ‘simplification’ and ‘complexification’ (cf. Dahl 2004; McWhorter 2001, 2005; Miestamo et al. 2008; Sampson et al. 2009) where, according to a widely held belief, language contact leads to a simplification of the languages involved. To be sure, much depends on our understanding of simplification and in those studies that see language contact as a trigger of simplification processes what is usually meant is the reduction or loss of morphological categories. In other words, language contact triggers or leads to the reduction of inflectional morphology. Moreover, the reduction processes positively correlate with the intensity of the contact

Universals and variation: an introduction

9

situation. In his contribution to the present volume, Hans-Olav Enger illustrates this general finding on the basis of Scandinavian dialectal gender systems where various cases of contact-induced gender reduction are reported (e.g. the cities of Bergen and Stockholm, which experienced considerable influence from Low German during the Hansa period). However, the literature on language contact also reports numerous cases of complexification where the inventory of morphological categories – or grammatical categories in more general terms – has increased as a result of language contact. Especially in the typological literature we find numerous cases of language contact reported that resulted in an increased morphological complexity of at least one of the languages involved (Aikhenvald 2002, 2006; Nichols 1992). In view of these results Enger argues that the traditional focus on simplification processes in the study of Scandinavian gender systems falls short of empirical reality and needs to be replaced by a viewpoint that also takes pronominal gender and more generally the syntagmatic aspects of gender into consideration. Adopting this perspective, one will notice that the systems of pronominal gender found in the contact varieties became more elaborate and that in the modern contact varieties significantly more gender exponents (e.g. pronouns) are used than, for example, in Old Norse. Enger concludes that contactinduced processes of simplification and complexification may operate side by side albeit in different grammatical subsystems (paradigmatic versus syntagmatic). Another important research area in language contact studies concerns systematic constraints on processes of borrowing and transfer. According to a well-known differentiation (cf. Thomason 2001), instances of contactinduced language change can be the result of two fundamental processes (i. e. ‘borrowing’ and ‘transfer’) that mainly arise as a consequence of the social parameters determining the contact situation as well as the ways along which code from one language enters another. Moreover, the linguistic results of these processes are quite different. Processes of borrowing can arise when two or more social groups speaking different languages interact and the social structure of the groups involved is left intact. In such situations speakers may exchange lexical material, though they are less likely to influence the grammatical systems of the contact languages. Transfer processes, in contrast, typically arise in highly unbalanced social situations when, say, one social group becomes socially dominated by another group and the dominated group is forced to learn the language of the dominating group. Such constellations can trigger widespread L2 learning as a consequence of which grammatical features of the substrate language may be transferred into the superstrate language.

10

Peter Siemund

According to Thomason (2001) the best predictor for the outcome of language contact situations is “intensity of contact”, i. e. the degree of interaction between the linguistic groups involved. A low degree of contact intensity can be expected to lead to the borrowing of lexical items, while a high degree of contact may have substantial effects on grammatical subsystems of the contact languages. For instance, it is widely accepted that inflectional morphology normally resists borrowing or transfer, but in a highly intense language contact situation pieces or even subsystem of inflectional morphology may find their way from one language into another. Several researchers have tried to identify restrictions on the borrowing or transfer of grammatical subsystems typically resulting in the formulation of implicational hierarchies. A well known example is the following implicational hierarchy formulated in Ross (2001: 149–50) capturing the sequence of changes in cases of contact-induced change that he refers to as ‘metatypy’: lexical semantic patterns > discourse structure > clause linkage > clauseinternal structure > phrase structure > word-internal structure. In his contribution to the present volume, Yaron Matras – based on an extensive sample of cross-linguistic contact situations – offers several proposals for fine-tuning existing borrowing hierarchies leading to important new constraints on the expectable variation in situations of language contact. In addition to a number of more general constraints such as “nouns > non-nouns, function words”, “free morphemes > bound morphemes”, “derivational morphology > inflectional morphology” and “agglutinating affix > fusional affix”, Matras also identifies several specific constraints on the borrowing of lexical material, including inter alia “remote kin > close kin”, “numerals in formal contexts > numerals in informal contexts”, “higher numerals > lower numerals” and “peripheral local relations > core local relations”. Matras embeds and discusses these findings within a new and unique conception of bilingualism that does not view the bilingual mind as a harbor of two separate linguistic systems, but rather as a complex repository of linguistic structures from two (or more languages) complemented by an extensive set of conditions on their usage (selection principles). Borrowing, on this view, comes to be interpreted as a violation of these selection principles such that linguistic forms originally tied to specific usage contexts appear in new contexts. Matras identifies and discusses several motives why speakers might relocate the “demarcation lines” between coexisting and alternative linguistic structures and forms, including accidental errors as a result of cognitive overload, which disturbs the relevant selection principles. The research findings of language contact studies are highly relevant for our understanding of linguistic universals as universally valid con-

Universals and variation: an introduction

11

straints on the architecture and the processing of languages should manifest themselves very clearly in language contact situations. As I see it, we can plausibly consider language contact as a generator of language variation and linguistic universals as filters, as it were, that separate the instances of permissible variation from those that violate these universals, should these be generated at all.

2.3.

Methodological issues of variation research

Even though linguistic variation is readily observable, it is difficult to turn it into an object of scientific investigation. An important problem, as Walter Bisang points out in his contribution, is reproducibility. Any scientific investigation must be based on data that are in principle reproducible, meaning that matching pieces of empirical evidence can be created at different points in time as long as the parameters conditioning the empirical evidence are kept constant. To take a simple example, the boiling or freezing of water can be replicated provided a certain number of external parameters is maintained. The problem is that there is no similar guarantee for the reproducibility of linguistic data since these data by necessity are subject to social and functional variation as well as the psychological state of the subjects. In equivalent experimental settings (elicitation) different subjects will produce heterogeneous data because all subjects are different at the individual level. Moreover, the same subjects may produce different data at different points in time even if the experimental setting is kept constant. Observational evidence (text corpora, audio material, etc.) is per se unique and can never be recreated in exactly the same way. More often than not, this type of data lacks metainformation (as e.g. the sex, age, education, job type of the speaker). To be sure, variation research – especially in the socio-linguistic tradition – is well aware of these problems and has laid great emphasis on controlling external parameters as well as the importance of statistical generalizations. Put in a nutshell, the main idea of statistical generalizations is that even if one and the same individual will never produce the same data point twice and even if a set of individuals will never produce exactly corresponding data points, the statistical means derived from such production and elicitation tasks can be expected to converge very strongly. This line of argumentation also carries over to observational evidence where balanced samples of text converge around some statistical mean. Bisang discusses this point drawing on Featherston (2007) who states that:

12

Peter Siemund There is no such thing as an ideal speaker-listener in one person, but the mean of a reasonably sized sample has exactly the characteristics one would expect an ideal speaker-listener to have. (Featherston 2007: 29).

Even though it is not obvious that the statistical mean yields the “ideal speaker-listener”, the data points thus obtained possess one important property: they are reproducible. However, as is also well known the statistical mean says nothing about the individual; ironically, there even may be no individual corresponding to the statistical mean. Bisang summarizes this problem as follows: Even if the social perspective is disregarded, it is highly questionable whether Featherston’s (2007) statistical “mean” can be immediately associated with processes that take place in an individual’s brain with its innate language capacity. The statistical results produced by Featherston’s (2007) sampling method may be due to such a large number of different processes in the brain that no reliable conclusions can be drawn. In fact, strict grammaticality judgments that reflect the manifestation of UG in an individual’s cognition are no longer accessible in such a statistical approach. (Bisang, this volume)

There are two papers in the present volume that deal with the methodological problems addressed in Walter Bisang’s contribution. To be sure, these problems also play a certain role in some of the other papers of the volume, but two papers clearly stand out insofar as they try to give at least partial answers to the issues of reproducibility and also representativity identified by Bisang. Bernd Kortmann and Benedikt Szmrecsanyi, in their contribution, discuss the status of ‘vernacular universals’ in the sense of Chambers (2004), i. e. recurrent phonological and morpho-syntactic features that can be observed across many, if not all, varieties of English and that should in principle also be observable in the vernacular forms of other languages. The notion of ‘universals’ was introduced into the study of English varieties primarily as a consequence of the fact that varieties of English in different parts of the world manifest similar or even identical non-standard grammatical forms although these non-standard forms cannot be traced back to a common ancestor. According to Chambers (2004: 129), nonstandard phenomena such as conjugation regularization (Mary heared the good news), default singulars (They was the last ones), multiple negation (I don’t/ain’t know nothing) and copula absence (She smart) qualify as vernacular universals. Chambers’ vernacular universals have been criticized because they cannot be found in all vernacular forms of English (they are primarily restricted to vernaculars of North America) and because some of the fea-

Universals and variation: an introduction

13

tures are rather common typologically (multiple negation, copula absence) and hence have little to do with vernacular forms. In other words, these features cannot be reproduced for all speakers who use vernacular forms and occur with speakers who do not use vernacular forms. It should also be noted that their reproducibility in the individual speaker can never be guaranteed since speakers typically control more than one lect. In view of these (and other problems), Kortmann and Szmrecsanyi propose to analyze bundles of non-standard features rather than consider such features in isolation – something they call “strategy-based approach”. This approach is based on a statistical analysis of a substantial set of nonstandard features across a large sample of English varieties. Using Principal Component Analysis, Kortmann and Szmrecsanyi are able to show that non-standard features form bundles correlating with the variety type (L1 variety, L2 variety, high contact variety, etc.), i. e. the variety type is a good predictor for the occurrence of a non-standard feature. This approach at least partially solves the problem of reproducibility as it effectively turns it into a statistical problem. Of course, this is far away from reproducibility at the individual level, which is the main concern of Bisang’s paper. Kortmann and Szmrecsanyi offer a second alternative to the search for individual non-standard features qualifying as vernacular universals, and in doing so again address the problem of reproducibility. Based on a comparison of corpus material sampling varieties of English from the different areas of the world, they performed a morphological analysis of 1000 randomly chosen word tokens from each variety in terms of bound/free morphemes and closed class/open class membership. The results show that varieties of English systematically differ along these parameters and, moreover, that these differences again pattern with the variety type. Since the composition of most of the corpus material that went into this study is controlled for many parameters (spoken/written, male/female, etc.), a relatively high degree of reproducibility of the results is guaranteed. To be sure, this concerns the reproducibility of some statistical mean, which, again, is different from reproducibility at the level of the individual. Even though a statistical mean is usually reproducible in corpus-based approaches, linguistic analyses based on such statistical information are not without problems. The main problem addressed by Walter Bisang, as discussed above, is that statistical generalizations say little about the grammatical knowledge at the individual level. In the contribution by Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund, another important problem is identified and analyzed in great detail. In the same way as Kortmann and Szmrecsanyi, these authors problematize the notion of ‘vernacular universals’ in varieties of English, pointing to the additional

14

Peter Siemund

complication that recurrent surface phenomena presented as candidates of vernacular universals may in fact hide significant distributional and functional differences. Statistical generalizations across varieties of English that are based on surface similarities run the risk of comparing incompatible phenomena: “what looks the same may not in fact be the same”. Davydova et al. therefore propose that varieties should be studied and compared across different dimensions, which can include typological hierarchies, the proximity of a variety to some reference variety and the source of origin of a non-standard grammatical phenomenon. One of the phenomena these authors discuss is subject/auxiliary inversion in embedded interrogative clauses. They demonstrate that this phenomenon – even though it is attested in a significant number of varieties – eschews comparison, as there are highly significant differences with regard to the verbs that can be found in the relevant contexts. Moreover, these distributional differences point to independent historical developments of subject/auxiliary inversion in embedded interrogative clauses, which apparently is a consequence of L2 learning strategies in Indian English but due to substrate influence in Irish English. Such considerations show that quantitative analyses must be complemented by careful qualitative interpretations. To sum up, it is clear that reproducibility remains a highly important touchstone of scientific investigations. It is also clear that linguistic studies sampling individual human subjects will never be fully reproducible, as the experimental conditions usually cannot be replicated. What can be reproduced with a good chance of success, however, are statistical means, but these carry the danger of not being representative of individual grammars. Moreover, they may conceal distinctions that more fine-grained qualitative studies of these individual grammars would reveal.

2.4.

Variation and linguistic theory

The traditional concern of most (mainstream) linguistic theorizing focuses on the systemic aspects of language. As pointed out above, this is understandable as theories aim for predictive power and hence are difficult – if not outright impossible – to design for chaotic data sets. Nevertheless, recent linguistic theorizing has begun to tackle variation data and has produced a number of highly insightful proposals. With Dufter et al. (2009), for example, there is a whole volume dedicated to the exploration of the theoretical aspects of variation research. In their introduction to Dufter et al. (2009), the editors argue that linguistic theories have been developed for essentially two types of variation

Universals and variation: an introduction

15

data and that the resulting theories are not necessarily compatible. On the one hand, linguistic theories have been developed to capture variation (i. e. differences) between linguistic systems where “linguistic system” can mean languages or regional (and possibly other) varieties of the same language. On the other hand, we find theories that have their focus on grammar-internal variation. As grammar-internal variation is less systematic and hence less predictable, the relevant theories are frequently based on probabilistic models. Even though it cannot be the task of the present volume to illuminate every theoretical corner of variation research, it contains four contributions explicitly addressing the theoretical constructs found behind variation data. What is more, they discuss theoretical constructs of rather different persuasions. Kristin Melum Eide and Hilde Sollid, in their contribution, promote the concept of parallel grammars while Guido Mensching and Eva-Maria Remberger work in the framework of Minimalism, where variation can be built in via different feature specifications. These are two radically different approaches in spite of the fact that both come from the deductive (i. e. top-down) paradigm. Brian Joseph’s article nicely complements these two lines of attack in arguing for a bottom-up approach to the analysis of variation. According to his view, variation is an emergent phenomenon that is not built into a grammar right away. Frans Hinskens, finally, argues that the assessment of the relative merit of competing theoretical approaches to the analysis of variation data is first and foremost an empirical problem, meaning that it should be decided for each empirical problem separately if it is more adequately captured in, say, a rule-based or usagebased model. The empirical problem taken up for discussion by Kristin Melum Eide and Hilde Sollid concerns exceptions to the verb-second rule in Norwegian. This otherwise surprisingly stable rule turns out to have exceptions (verb-third) in several regional varieties of Norwegian – especially contact varieties – and specific registers such as hymns, nursery rhymes and what these authors generally refer to as “festive syntax”. Kristin Melum Eide and Hilde Sollid argue that grammars are context-sensitive in the sense that speakers choose from a set of grammars the one that is contextually adequate. This means that speakers have several parallel grammars at their disposal where each grammar is assumed to form a coherent computational system. According to Kristin Melum Eide and Hilde Sollid the main distinction in Norwegian concerns “verb raising” and “non-verb raising” grammars. The main thrust of the paper by Guido Mensching and Eva-Maria Remberger lies in the development of a basic comparative syntax of

16

Peter Siemund

Romance languages, specifically addressing word order patterns in basic clause types. The resulting syntactic model should be able to yield the common and divergent properties of Romance languages in this domain and explain the observable differences in terms of a few underlying parameters. The authors propose a model framed within the Minimalist Program where the feature specifications of lexical items yield the observable differences in surface structure. The authors argue that this approach is in principle also applicable to individuals who control more than one lect where diverging feature specifications can be the result of coexisting (i. e. parallel) lexicons or lexical elements containing alternative feature specifications. The former approach is the preferred choice for capturing regional variation while the latter approach seems more adequate for social and stylistic variation. Brian Joseph contributes a highly topical and central paper to the volume in which linguistic universals are discussed in relation to language variation. The major question addressed is the origin of linguistic universals – especially in view of the fact that languages are full of variation. Brian Joseph argues that speakers first and foremost make local generalizations and that this is the place where universals must originate. On the basis of a wide array of examples taken from many languages, it is shown how small-scale local generalizations can turn into major generalizations. Universals arise as the result of problem solving strategies that speakers employ in order to cope with communicative challenges. Linguistic universals, on this view, clearly have a functional basis and must be conceived of as emergent phenomena. To put it in Brian Joseph’s own words: … speakers in the process of using – and thus of changing – their language often act as if they are in a fog, by which is meant not that they are befuddled but that they see clearly only immediately around them, so to speak, and only in a clouded manner farther afield … they thus generalize only ‘locally’ … and not globally over vast expanses of data, and they exercise their linguistic insights only through a small ‘window of opportunity’ over a necessarily small range of data. (Joseph, this volume)

As mentioned above, Frans Hinskens argues that theoretical debates should be permanently fuelled by empirical data and empirical analyses. The theoretical divide that he intends to inform in his contribution concerns that of ‘rule-based’ models of grammar as opposed to ‘usage-based’ approaches. In investigating the reduction and deletion of unstressed vowels in modern spoken Dutch, Frans Hinskens shows that usage-based parameters (various kinds of type frequency and token frequency) are only weak predictors for the occurrence of vowel reduction and deletion

Universals and variation: an introduction

17

whereas grammatical factors such as the morphological embedding of unstressed vowels (in stem or affix, in function word versus content word) or phonological factors (metrical, segmental, prosodic) turn out to be robust predictors for reduction. These results are based on an extensive quantitative case study and show that rule-based models have substantially more explanatory power than usage-based models in this specific empirical domain. To be sure, Frans Hinskens does not claim that these results necessarily generalize to other empirical domains, nor does he claim that rule-based models per se can be assumed to have more explanatory power as such far-reaching and general conclusions can only be taken against the yardstick of comprehensive empirical analyses. In conclusion, Hinskens hypothesizes that, with respect to phonological variation, usage-related parameters may play a specific role in what is often referred to as the ‘life-cycle’ of a sound change or rule typology.

3.

Concluding remarks

The overarching aim of the present volume is to explore the relationship between linguistic universals and language variation. Most scholars will probably agree that these notions are valuable research objects even though their extensions are heavily debated and it is far from trivial to draw a demarcation line between the universal and variable aspects of language. It is to this general debate that the present volume intends to make a fresh contribution. In view of the scope of the research area the contribution can certainly only be a modest one. Several articles in the present volume originate from a workshop on Linguistic Universals and Language Variation (UniVar) held at the University of Hamburg’s Research Centre on Multilingualism in July 2007. The financial aid made available by the German Research Foundation is herewith gratefully acknowledged. Each of the papers in the present volume was read and commented on by one external reviewer in addition to the editor. The external reports especially greatly increased the coherence of the papers as well as the coherence of the volume as a whole. I am extremely grateful to the external reviewers for their time and commitment. Last but not least, I would like to express my thanks to Martin Schweinberger, Isabel Peters, Maike Berger, Hanne Brandt and Leonie Fölsing for their help and commitment during the editorial process.

18

Peter Siemund

References Aikhenvald, Alexandra and Robert Malcom Ward Dixon (eds.) 2001 Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics. Oxford: Oxford University Press. Aikhenvald, Alexandra 2002 Language Contact in Amazonia. Oxford: Oxford University Press. Aikhenvald, Alexandra 2006 Grammars in contact: A cross-linguistic perspective. In: Alexandra Aikhenvald and Robert Malcom Ward Dixon (eds.), Grammars in Contact: A Cross-Linguistic Typology, 1–26. Oxford: Oxford University Press. Bisang, Walter 2004 Dialectology and typology – an integrative perspective. In: Bernd Kortmann (ed.), Dialectology Meets Typology. Dialect Grammar from a Cross-Linguistic Perspective, 11–45. Berlin: Mouton de Gruyter. Chambers, Jack K. 2004 Dynamic typology and vernacular universals. In: Bernd Kortmann (ed.), Dialectology meets Typology, 127–145. Berlin: Mouton de Gruyter. Chambers, Jack K. and Peter Trudgill 1980 Dialectology. Cambridge: Cambridge University Press Cornips, Leonie and Corrigan, Karen (eds.) 2005 Syntax and Variation. Reconciling the Biological and the Social. Amsterdam: Benjamins. Dahl, Östen 2004 The Growth and Maintenance of Linguistic Complexity. Amsterdam: Benjamins. Dufter, Andreas, Jürg Fleischer and Guido Seiler (eds.) 2009 Describing and Modeling Variation in Grammar. Berlin: Mouton de Gruyter. Featherston, Sam 2007 Data in generative grammar: The stick and the carrot. Theoretical Linguistics 33: 269–318. Filppula, Markku, Juhani Klemola and Heli Paulasto (eds.) 2009 Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond. London: Routledge. Good, Jeff (ed.) 2008 Linguistic Universals and Language Change. Oxford: Oxford University Press. Haspelmath, Martin, Matthew s. Dryer, David Gil and Bernard Comrie (eds.) 2005 The World Atlas of Language Structures. Oxford: Oxford University Press.

Universals and variation: an introduction

19

Heine, Bernd and Tania Kuteva 2005 Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Hinskens, Frans, Roeland van Hout and Leo Wetzels (eds.) 1997 Variation, Change and Phonological Theory. Amsterdam: Benjamins. Johanson, Lars 2008 Remodeling grammar. In: Peter Siemund and Noemi Kintana (eds.), Language Contact and Contact Languages, 61–79. Amsterdam: Benjamins. Kortmann, Bernd (ed.) 2004 Dialectology Meets Typology. Dialect Grammar from a Cross-Linguistic Perspective. Berlin: Mouton de Gruyter. McWhorter, John H. 2001 The world’s simplest grammars are Creole grammars. Linguistic Typology 5 (2/3): 125–166. McWhorter, John H. 2005 Defining Creole. New York: Oxford University Press. Miestamo, Matti, Kaius Sinnemäki and Fred Karlsson (eds.) 2008 Language Complexity: Typology, Contact, Change. Amsterdam: Benjamins. Nevalainen, Terttu, Juhani Klemola and Mikko Laitinen (eds.) 2006 Types of Variation: Diachronic, Dialectal and Typological Interfaces. Amsterdam: Benjamins. Newmeyer, Frederick J. 1998 Language Form and Language Function. Cambridge, Mass.: MIT Press. Nichols, Johanna 1992 Linguistic Diversity in Space and Time. Chicago: Chicago University Press. Rohdenburg, Günter and Britta Mondorf (eds.) 2003 Determinants of Grammatical Variation in English. (Topics in English Linguistics 43.) Berlin: Mouton de Gruyter. Roeper, Tom 2007 The Prism of Grammar: How Child Language Illuminates Humanism. Cambridge, Mass: MIT Press. Ross, Malcolm 2001 Contact-induced change in oceanic languages in north-west Melanesia. In: Malcom Ward Dixon and Alexandra Aikhenvald (eds.), Areal Diffusion and Genetic Inheritance: Case Studies in Language Change, 134–166. Oxford: Oxford University Press. Sampson, Geoffrey, David Gil and Peter Trudgill (eds.) 2009 Language Complexity as an Evolving Variable. Oxford: Oxford University Press.

20

Peter Siemund

Scalise, Sergio, Elisabetta Magni and Antonietta Bisetto (eds.) 2009 Universals of Language Today. Berlin: Springer. Siemund, Peter 2009 Linguistic universals and vernacular data. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, 321–346. London: Routledge. Siemund, Peter and Noemi Kintana (eds.) 2008 Language Contact and Contact Languages. Amsterdam: Benjamins. Thomason, Sarah 2001 Language Contact: An Introduction. Edinburgh: Edinburgh University Press. van der Auwera, Johan 1998 Conclusion. In: Johan van der Auwera (ed.), Adverbial Constructions in the Languages of Europe, 20–23. (Empirical Approaches to Language Typology, EUROTYP.) Berlin: Mouton de Gruyter. Winford, Donald 2003 An Introduction to Contact Linguistics. Oxford: Blackwell.

Part 1 Varieties and cross-linguistic variation

Causal clauses: a cross-linguistic investigation of their structure, meaning, and use Holger Diessel and Katja Hetterle Abstract In this paper we investigate the form and function of causal adverbial clauses, which have never been systematically studied from a cross-linguistic point of view. Using data from 60 languages, it is shown that causal clauses tend to be more independent of the associated main clause than other semantic types of adverbial clauses. In contrast to temporal and conditional clauses, causal clauses predominantly follow the (main) clause, include the same non-reduced verb forms as independent sentences, and are often intonationally separated from the semantically related clause. We argue that the particular structural properties of causal clauses are motivated by their communicative function in speaker-hearer interactions. Drawing on conversational data from English, German, Mandarin Chinese and Japanese, we show that causal clauses are commonly used to support a previous statement that has been challenged by the hearer. In this use, causal adverbial clauses function as independent assertions that are only loosely combined with the associated main clause. Keywords: adverbial clause, causal clause, coordination-subordination continuum, clause order, grammar and language use, grammaticalization, language acquisition

1.

Introduction

The cause-effect relationship is one of the most fundamental concepts of the human mind that has been studied extensively in various subfields of cognitive science (cf. Sowa 2000; Meyer 2000). Linguistically, causal relationships are commonly expressed by complex sentences consisting of a main and a subordinate clause, but they can also be expressed by two coordinate sentences. In this paper, we examine the structure, meaning, and use of causal clauses from a cross-linguistic point of view. Our analysis concentrates on causal adverbial clauses, but since the distinction between adverbial subordination and sentential coordination is fluid, we will also

24

Holger Diessel and Katja Hetterle

look at causal clauses at the borderline between subordination and coordination. In fact, the subordination-coordination continuum plays an important role in our study. It is the central claim of this paper that causal adverbial clauses are structurally more independent of the associated main clause than other semantic types of adverbial clauses and often are reminiscent of coordinate sentences. In the typological literature, causal adverbial clauses have been studied together with other semantic types of adverbial clauses, such as conditional and temporal clauses (cf. Thompson and Longacre 1985; CouperKuhlen and Kortmann 2000), but in contrast to these other types, causal clauses have never been systematically investigated from a cross-linguistic point of view. The current study seeks to fill this gap. It is the first crosslinguistic investigation that systematically examines the form and function of causal clauses from a typological point of view (cf. Hetterle 2007). Drawing on data from a wide range of languages, the paper argues that adverbial clauses constitute a family of related constructions that vary as to the degree to which they are integrated into a complex sentence. Some adverbial clauses are only loosely adjoined to a neighboring clause, resembling a coordinate sentence, whereas other adverbial clauses are tightly integrated into the main clause (cf. Diessel 2001, 2004: chap. 3). The degree of formal integration is determined by several features: the morphosyntactic properties of the verb and its arguments in the adverbial clause (cf. Cristofaro 2003), the positioning of the adverbial clause relative to the main clause (cf. Diessel 2001, 2005), and the intonational link between main and adverbial clauses (cf. Chafe 1984; Ford 1993). The paper shows that causal clauses tend to be less tightly integrated into complex sentences than other semantic types of adverbial clauses. In contrast to temporal and conditional clauses, causal clauses typically include the same verb forms and arguments as ordinary main clauses, are usually placed after the semantically associated clause, and are commonly expressed by a separated intonation unit. Taken together these features suggest that causal clauses are only loosely combined with the associated main clause; they are commonly realized by constructions that exhibit the same morphosyntactic properties as main clauses and thus may be analyzed as coordinate sentences rather than adverbial clauses. The paper argues that the particular structural properties of causal clauses are motivated by their communicative function in speaker-hearer interactions. Drawing on conversational data from several unrelated languages, it is shown that causal clauses are commonly used to support a problematic statement. More precisely, the paper argues that causal clauses are often embedded in a particular discourse pattern involving

Causal clauses: a cross-linguistic investigation of their structure, meaning, and use

25

three verbal actions: (1) a statement that the hearer does not accept or understand, (2) the hearer’s response to this statement indicating disagreement or lack of understanding, and (3) a causal clause providing a reason for the problematic statement (cf. Ford 1993). Causal clauses that are embedded in this discourse pattern are structurally independent sentences that lack the usual properties of subordinate clauses, which they may exhibit in other contexts, notably in written genres. However, in conversational discourse causal clauses are commonly used as independent assertions that are structurally independent of the semantically associated clause.

2.

Database

Cross-linguistic investigations are often based on a variety sample including several hundred languages, but we decided to use a probability sample, which tends to be smaller than a variety sample because it is more rigorously designed. A probability sample includes a group of languages that are independent of each other and therefore can be submitted to statistical analysis (cf. Rijkhoff et al. 1993; Rijkhoff and Bakker 1998; Jansen et al. 2006). A pure probability sample would include languages that are selected at random from all languages across the world, but since languages are genetically and geographically related, the selection procedure cannot be entirely at random; instead, the researcher has to control for genetic and areal factors in order to compile a sample in which the languages are independent of each other (cf. Rijkhoff et al. 1993; Bakker 2010). In our study, we used a stratified probability sample controlled for genetic and areal dispersion. The sample includes 60 languages selected from 60 different genera, i. e. linguistic groups that roughly correspond to the subfamilies of Indo-European (Dryer 1992), distributed across six large areas: Eurasia, Africa, South East Asia and Oceania, Australia and New Guinea, North America, and South America (Dryer 1989, 1992). Thus, the languages in our sample are genetically and geographically unrelated (or only distantly related). A complete list of the languages we examined is given in the appendix. In what follows, we investigate the cross-linguistic properties of causal clauses in comparison to other semantic types of adverbial clauses. Specifically, we compare causal clauses to temporal and conditional clauses, which are often expressed by similar types of constructions that are historically related (cf. Matthiessen and Thompson 1985; Givón 1990, 2006).

26

Holger Diessel and Katja Hetterle

However, despite this relationship there are some important differences between them, which are the focus of the current study.

3.

The form and function of causal clauses in cross-linguistic perspective

3.1.

The morphosyntactic properties of causal clauses

To begin with, adverbial clauses can be expressed by a wide range of constructions that differ in terms of the verb forms and arguments they include. In a recent study, Cristofaro (2003) argued that subordinate clauses can be divided into two basic types: balanced and deranked clauses (cf. Stassen 1985). Balanced subordinate clauses include the same verb forms and arguments as main clauses, but deranked subordinate clauses are different. They include a reduced verb form and do not always require an overt subject. For instance, in English adverbial clauses are commonly realized by participial constructions that do not carry tense and person markers and lack an overt subject.1 (1)

And he encountered the problem of conducting with one hand while holding the reins with the other eighteen months ago ….

Both the missing subject and the uninflected verb forms are characteristic of deranked subordinate clauses. Since the interpretation of a deranked subordinate clause relies on semantic features of the main clause, it is reasonable to assume that a deranked subordinate clause is more tightly integrated into a complex sentence than a balanced subordinate clause, whose semantic interpretation is more independent of the associated main clause. Note that in English the occurrence of deranked adverbial clauses is most typical of temporal clauses, notably temporal clauses of simultaneity, whereas causal clauses are primarily expressed by balanced constructions. However, in other languages, causal adverbial clauses are also commonly deranked. For instance, in Evenki (Tungusic, Russia) all adverbial clauses are expressed by converbs (cf. Haspelmath 1995). A converb is a particular non-finite verb form that indicates the semantic relationship to the associate main clause by an affix, as in example (2), in which two con-

1 The English examples in this paper are taken from the International Corpus of English, British Component.

Causal clauses: a cross-linguistic investigation of their structure, meaning, and use

27

verbal clauses are marked by the suffix -kA(n)im indicating a temporal link of anteriority between the converbs and the main clause. Example (3) shows that the same type of construction is used to realize a causal adverbial clause. (2)

Evenki (Nedjalkov 1997: 45) Asi-va ga-kaim oron-mo rege-keim tar beje suru-re-n. wife-ACC take-CONV reindeer-ACC sit-CONV that man go.awayNONFUT-3SG ‘Having taken a wife and having sat on a reindeer that man left.’

(3)

Evenki (Nedjalkov 1997: 53) Engesi bi-mi nungan homo:ty-va strong be-CONV he bear-ACC.DEF ‘He overcame the bear because he was strong.’

davdy-ra-n. win-NONFUT-3SG

Interestingly, Cristofaro (2003) showed that causal clauses are less often expressed by deranked subordinate clauses than others semantic types of adverbial clauses. Examining a variety sample of 80 languages, she found that causal clauses are commonly expressed by balanced constructions whereas other semantic types of adverbial clauses, notably temporal and purpose clauses, are typically deranked (see also Cristofaro 2005). Our data are consistent with these findings. Following Cristofaro, we distinguished between deranked and balanced adverbial clauses based on the criteria she proposed. Deranked adverbial clauses are expressed by reduced or special verb forms such as infinitives, participles, converbs, and nominalizations, which lack at least some of the inflectional distinctions of verbs in main clauses (such as tense, aspect, mood, or agreement distinctions) and often do not include an overt subject. Balanced adverbial clauses, in contrast, are expressed by finite constructions including the same verb forms and arguments as independent sentences. Note that we disregarded cases of simple clause juxtaposition (i. e. asyndetic parataxis), which Cristofaro included in her sample if juxtaposition served as a general means for the expression of an adverbial relationship. Figure 1 shows that in our sample causal clauses are less often deranked than temporal and conditional clauses, which we lumped together in one class (cf. data in the Appendix).

28

Holger Diessel and Katja Hetterle

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

38,33 %

31,7 % 11,7 %

33,33 %

deranked mixed balanced

56,7 % 28,3 %

temporal/conditional

causal

Figure 1: Deranking in temporal, conditional, and causal adverbial clauses.

As can be seen, in more than half of the languages in our sample (56.7 %), causal clauses are exclusively expressed by balanced constructions; only a minority of the languages employs deranked causal clauses. If we look at the latter more closely we find that their occurrence is restricted to languages in which temporal and/or conditional clauses are also deranked; there is not a single language in the entire sample in which deranking is an exclusive feature of causal clauses. A randomized 2×3 Ȥ-analysis revealed a significant difference between causal clauses and temporal/conditional clauses (Ȥ=12.30, p you > we > they (Chambers

138

Sali A. Tagliamonte

2004: 141). Existentials, for example, are widely known to promote default agreement more than any other category: “was is most frequent after there” (Britain and Sudbury 2002: 19–20; Chambers 2004: 132). Britain and Sudbury (2002: 19–20) refer to this correlation as “the existential constraint”. As far as the other grammatical subjects are concerned, 2nd person singular ranks highest, then 1st person plural and 3rd person plural pronouns have the least proportion of default agreement of all.4 The consistency of this ranking (in addition to the world-wide diffusion of the phenomenon) has been used to bolster the argument for interpreting default agreement as a universal (e.g. Walker 2007) Figure 2 plots the frequency of default agreement in existential context, i. e. there was, by geographic scale across the X-axis.

85,7 64,3

95,7

90,7 84

80 68

63,5 65,5 56,8 48

30,2

nt Pr o es uy g to sb ht En n or ou c gh lav tV e Cu illa lly ge ba Po ckey rta vo gi e Bu ck Cu ie m no M ck W ary p he at ort le y H ill Y o Ti rk ve W rto n in ca nt on

18

ou

G

G

uy

sb

or

N

or

th

To

ro

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

Figure 2: Distribution of default agreement ‘was’ in existential contexts. 4 The frequency of grammatical person tokens across corpora is remarkably similar: 3rd singular > existential > 1st singular > 3rd plural > 1st person plural > 2nd person (singular and plural). The lower frequency persons, namely 3rd person plural, 2nd person and 1st plural have a high proportion of default agreement. However, a higher proportion of default agreement occurs with existentials, which are the second highest construction in frequency across all corpora.

Variation as a window on universals

139

Figure 2 reveals that the use of was as there was is a pervasive pattern. It occurs in all the communities, notably even the two urban contexts York and Toronto. The difference in frequency of default agreement between the two cities is great: 56.8 % vs. 30.2 % respectively. Could this be due to the contrast between Britain and North America? The fact that the other Canadian varieties in Nova Scotia have similar rates to York suggests that an over-arching contrast between these two major varieties of English is not the explanation. At the same time, although Toronto has one of the lowest rates of default agreement in existentials, it still has a healthy proportion and this despite the fact that it is well known to be a standard middle-class variety.5 However, there is another variety – Cullybackey – in Northern Ireland where use of default agreement is even lower, 18 %. Yet the low rate cannot be due to the imposition of prestige norms as might be expected for middle-class Toronto; this is a small fishing village inhabited by elderly fishermen, their wives and families. Thus, although default agreement in existential subjects is widespread and generally robust, the results in Figure 2 reveal that in some places it is very frequent – such Wheatley Hill and Guysborough Village – while in other places it is very infrequent, – such as Cullybackey and Toronto. So far these discrepancies are not easily explained. The two communities where default agreement is most frequent are in two different countries in two different types of communities. The same is true of the places where default agreement is not so frequent. Indeed, these two contexts are entirely antithetic: a small Irish fishing village (Cullybackey) vs. Metropolitan Toronto. Despite the fact that the people are the same age these are widely divergent situations. This is a strong indication that there is something additional in operation beyond a simple tendency either towards regularization or to vernacular primitives. I will return to this observation later.

4.2.

Grammatical person

The next most frequently-attested context for default agreement in the grammatical paradigm is 2nd person singular you. This subject type has often been singled out as having a high degree of default agreement. Cham-

5 In Toronto default agreement is virtually restricted to existential constructions. The number of tokens in other contexts is extremely rare and isolated to a handful of elderly people.

140

Sali A. Tagliamonte

bers (2004) has noted that this may be due to the syncreticism between singular and plural. However, in more recent research 2nd person you has demonstrated what Chambers refers to as “unruly behavior” (Chambers 2004: 136), suggesting that this is not the regular pattern it was once thought to be. The remaining personal pronoun subjects – we and they are often reported to be ordered regularly such that you has the most default agreement, then we, and the 3rd person plural pronoun they has the least. However, there is really “no semantic or grammatical feature that distinguishes” among these personal pronouns (Chambers 2004: 141). Thus, there is no reason for them to behave systematically in this way either. Figure 3 displays the frequency of was by personal pronouns you, we and they and plots their frequency by geographic scale across the X-axis. you

we

they

G

uy

sb

N

or t or h P re o G uy ugh sto sb tE n or ou ncla gh v tV e Cu illa lly ge ba c Po key rta vo gi e Bu ck Cu ie m no M ck W aryp he or at le t y H ill Y or Ti k ve W rton in ca nt on

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

Figure 3: Distribution of default agreement ‘was’ by grammatical person.

Figure 3 shows that some localities do indeed have frequent use of was 2nd person singular, including North Preston, Guysborough Enclave, Buckie and Tiverton. Yet others have almost none. Comparing you with we, the Figure reveals higher rates of default agreement in North Preston, possible Guysborough Enclave, Buckie. Maryport and Tiverton. Since you is an infrequent grammatical subject in each data set and the frequencies

Variation as a window on universals

141

of default agreement are generally low, the ranking for the other locales is not very revealing. Note that default agreement with 3rd person personal pronouns they is rare, here too confirming the prevailing view that it inhibits default agreement. Yet there are three locales where it does occur – in the African Nova Scotian villages in Canada and in Wincanton in Southwest England. Once again there is no simple explanation for the intercommunity patterns. Although it seems that only rural locations have default agreement with you and we, only certain ones do and among them there is no obvious commonality. Thus, as with 2nd person, here too there is no striking regularity to the use of default agreement for these grammatical subjects.

4.3.

The Northern Subject Rule

Another famous constraint involving default agreement is the so-called Northern Subject Rule, a feature described by Ihalainen (1994), Klemola (1996), Pietsch (2005), McCafferty (2003) and others based on earlier dialectological observations by Murray (1873). This rule refers – in large part – to the contrast between 3rd person plural noun phrases, such as these boys or the receivers, as opposed to the 3rd person personal pronoun, they, as in (4).6 (4)

a. And these boys is earning a mint. Okay they’re off in the summer… (PVG/1) b. They’re bust, the receivers is in. (Scotland/CMK/6)

The original Northern Subject Rule dictated that default agreement would never occur with the 3rd personal plural pronoun they when immediately preceding the verb, i. e. they was. However, as Figure 2 shows, spoken data typically displays variability. Nevertheless, a constraint ranking of more default agreement after a noun phrase than after a personal pronoun would still indicate the application of the rule, albeit an extension of it into personal pronouns context. Figure 4 displays the frequency of was according to this grammatical contrast and once again plots the frequency by geographic scale across the X-axis. 6 Another facet to the Northern Subject Rule is the proximity of the subject to the verb. Personal pronouns that are separated from their subjects with an adverbial or a conjunction may exhibit default agreement (Montgomery and Chapman 1992; Montgomery, Fuller and DeMarse 1993).

142

Sali A. Tagliamonte

3rd p. pl. NOUN PHRASE

3rd p. pl. PRONOUN ‘they’

uy G

G

uy

sb

or

N

or th

ou Pres g t sb ht E on or n c ou gh lave tV Cu illa lly ge ba c Po key rta vo gi e Bu ck Cu ie m no M ck W aryp he or at le t y H ill Y or Ti k ve W rton in ca nt on

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

Figure 4: Distribution of default agreement ‘was’ by NP vs. plural pronoun ‘they’.

Figure 4 shows that full noun phrases occur with default agreement in all locations. Some varieties however exhibit default agreement with plural pronoun they as well. Notice that in each of these locales the predicted constraint hierarchy is maintained – the pronoun contexts have less default agreement and full noun phrases more. Thus, it appears that in some locales the Northern Subject Rule has undergone modification by extension. The essential nature of the pattern (differentiating the 3rd plural personal pronoun from full noun phrases) is maintained; however, instead of categorical contrast, it is variable.

4.4.

The Existential Contrast

Figures 2–4 have demonstrated that the existential constraint marks a split in the nature of default agreement. Due to the over-whelming use of default agreement in this context and the lower usages in most locales in the other contexts (Figures 3 and 4), Figure 5 plots the use of was in existential constructions versus all other contexts combined.

Variation as a window on universals

existential ‘there’

other

uy G

G

uy

sb

or

N

or th

ou Pres g t sb ht E on or ou ncla gh v tV e Cu illa lly ge ba c Po key rta vo gi e Bu ck Cu ie m no M ck W aryp he or at le t y H ill Y or Ti k ve W rton in ca nt on

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

143

Figure 5: Distribution of default agreement with ‘was’ in existential contexts contrasted with all non-existential contexts.

Figure 5 reveals a scale independent constraint. First, each locale shows this contrast. Second, every locale but one shows the same direction of effect. In Cullybackey in Northern Ireland existentials tend to have standard agreement whereas plural noun phrases and the other personal pronouns tend toward default agreement.

4.5.

Summary

The over-arching generalization that emerges from the results for the grammatical person constraint is that the specific hierarchy of grammatical subjects, i. e. Existential > you > was > they, that has been reported in the literature does not hold. Nevertheless, there are fairly consistent scale independent contrasts within the constraint ranking. The clearest one contrasts existentials with everything else. Yet there is also a regular contrast between plural noun phrases and pronominal they.

144

Sali A. Tagliamonte

5.

Analysis of contrast

I now turn to an analysis of contrast following the Horvath and Horvath method. A contrast analysis examines the strength of the constraint ranking on the use of was, according to the following procedure (Horvath and Horvath 2003: 153–154): ± Plot the degree of contrast between promotor and inhibitor. ± Measure by using multiple regression. ± Calculate the difference between highest and lowest probability weights. ± The degree of contrast is strong if values are high; weak if values are low. ± Is there variation in the consistency of the contrast? For these contrast analyses the degree of contrast between the promoting category (such as noun phrase) and the inhibiting category (such as pronoun) are measured. The values come from factor weights derived from a logistic regression analysis using Goldvarb X (Sankoff, Tagliamonte and Smith 2005). For each factor, the difference in the factor weights has been calculated and then plotted on the graph (Horvath and Horvath 2003: 152). For example, if the factor weight for the probability of default agreement for full noun phrases is 1 and 0 for pronouns, as in BCK in Figure 8, then the value on the graph is the difference between those two numbers, i. e. 1. Such a comparison across the geographic scale reveals variation in the consistency of the strength of the contrast. ± If the values are large, then the effect is strong. If it is low, the effect is weak. ± If there is consistency in the contrast across varieties, this adds another measure of comparability to the analysis.

Variation as a window on universals

145

Geographic Scale

uy G

G

uy

sb

or

ou

gh

To ro nt sb t E o or ou ncl gh av e t N Vil or l th age Pr Cu est lly on ba Po ckey rta vo gi e Bu ck Cu ie m no M ck W ary p he at ort le y H ill Y o Ti rk ve W rto n in ca nt on

Range of probability weights

Contrast: Promoter-Inhibiter 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Figure 6: Model of contrast analysis following Horvath and Horvath (1993).

Figure 6 illustrates an idealized contrast analysis. The way to interpret the Figure is basically this: values close to the X-axis are weak. The higher the values are, the stronger they are. The line in Figure 6 measures what an effect would look like if it was neither strong nor weak and was the same across varieties. The next series of Figures provide a contrast analysis of the influence of different grammatical persons on the use of default agreement.

146

Sali A. Tagliamonte

5.1.

‘you’ versus ‘we’

Figure 7 plots the contrast between you and we.

W rton in ca nt on

k

ve

or

Ti

H y

Y

ill

t or

le at

he W

M

ar

yp

no

ck

ie m

ck

Cu

Bu

vo

gi

e

ey rta Po

Cu

lly

ba

ck

ag

e

ve

ill

la

tV

sb

or

ou

gh

nc tE

gh ou uy G

G

uy

sb

or

N

or

th

Pr

es

to

n

factor weight

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Figure 7: Contrast analysis of ‘you’ vs. ‘we’.

Figure 7 reveals that overall the strength is moderate. It never rises above 0.4. It is also erratic; some points are as high as 0.4, in locales where the contrast between the two pronouns is great and as low as 0 where there is no difference.

Variation as a window on universals

5.2.

147

Noun phrase versus pronoun

Figure 8 plots the contrast between noun phrases and pronouns (the Northern Subject Rule).

W rton in ca nt on

k

ve

or

Ti

Y

ill

t y

H

or

le at

he W

M

ar

yp

no

ck

ie m

ck

Cu

Bu

vo

gi

e

ey rta Po

Cu

lly

ba

ck

ag

e

ve

ill

la

tV

uy

sb

or

ou

gh

nc tE

gh ou G

G

uy

sb

or

N

or

th

Pr

es

to

n

factor weight

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Figure 8: Contrast Analysis: noun phrases vs. personal pronoun ‘they’.

Figure 8 reveals quite a different result. The degree of contrast is extreme in some locales and weak in others. This indicates considerable scale dependence. Some varieties simply do not have any default agreement with 3rd person plural pronouns – such as Buckie. In this case the effect is categorical. Many others are positioned at 0.6 or above, indicating a very strong effect. Two hover around 0.3 – such as Maryport and Wheatley Hill indicating a moderately strong effect. Finally there are four where the effect is minimal – the two African Nova Scotian villages along with York and Tiverton.

148

Sali A. Tagliamonte

5.3.

Existentials

Figure 9 plots the contrast between existentials and everything else.

W rton in ca nt on

ve

or

Ti

Y

H y

k

ill

t or

le at

he W

M

ar

yp

no

ck

ie m

ck

Cu

Bu

vo

gi

e

ey rta Po

Cu

lly

ba

ck

ag

e

ve

ill

la

tV

sb

or

ou

gh

nc tE

gh ou uy G

G

uy

sb

or

N

or

th

Pr

es

to

n

factor weight

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Figure 9: Contrast Analysis: existential vs. non-existential.

Figure 9 reveals yet another pattern. Notice that there are no extremes here. Instead there is a modulating scale independent effect. In most locales this contrast is strong or very strong with only a couple of cases where the effect is weak. In sum, the analysis of contrast adds an important dimension to the overall findings. While the constraint ranking analyses showed us the order of the constraints, the contrast analysis enables us to view the strength of the effect and assess its scale independence. The result is that only the existential constraint is consistent on the measures of constraint and contrast.

Variation as a window on universals

5.4.

149

Negation

After grammatical person, the most important constraint reported for default agreement is the effect of negation (e.g. Britain and Sudbury 2002; Tagliamonte and Smith 2000; Anderwald 2002). The examples in (5) illustrate the extent of variability in the communities under investigation. (5)

a. Well no we wasn’t hardly big enough then. We were only very tiny what, very small. (SMT/g) b. There weren’t even a lock on mi Granny’s door, it was just a bar. (CLB/b) c. And of course, that time, there was-nae pensions. (BCK/g) d. And you wasn’t allowed to follow. (GYE/I) e. You wasn’t allowed to use their toilets. (NPR/p) f. ’Cos they wasn’t safe, ’cos they were magnesium. (SMT/g)

The constraint ranking for the effect of negation has a number of different patterns. The first is what has been referred to as “Vernacular Pattern I”. This is where the was variant occurs regardless of type of sentence. This pattern is said to be the simpler and more basic pattern. A second pattern, labeled “Vernacular Pattern II” is the case where weren’t occurs in negatives but the default agreement was occurs in affirmatives (Chambers 2004: 131). This pattern is attested in North American dialects (North Carolina), in southwest England (Reading) (Cheshire, Edwards, and Whittle 1989), in Britain’s research on the Fens in southeast England (Britain and Sudbury 2002) and elsewhere in Britain (Anderwald 2002). Indeed, Anderwald (2002) argues that the single form weren’t for all past be negatives in non-standard varieties reveals a greater conformity to markedness than the standard language. However, a third pattern has also been reported.This is wasn’t for negatives and were for affirmatives (Tagliamonte and Smith 2000: 160–161). Here too Anderwald’s (2002) markedness analysis applies since there is only one form, albeit a different one, for negatives. Vernacular Pattern I ± Use of was/wasn’t for affirmative and negative, no constraint ranking. Vernacular Pattern II ± More weren’t for negative contexts and more was for affirmative. Vernacular Pattern III ± More wasn’t for negative contexts; and more were for affirmative.

150

Sali A. Tagliamonte

Figure 10 plots the contrast between negative and affirmative contexts to see how these patterns distribute by geographic scale. Negative

o

es

nt

Pr

ro

or

G

uy

or

sb

ou

gh

th

To

or N sb uy G

Affirmative

t E ton ou ncl gh av tV e Cu illa lly ge ba Po ckey rta vo gi e Bu c Cu kie m no M ck W ary p he at ort le y H ill Y o Ti rk ve W rto n in ca nt on

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

Figure 10: Overall distribution of default agreement by negation.

Vernacular Pattern I is identifiable by an equal proportion of default agreement for negative and affirmative contexts. Notice that this exists in four locales. Two in Canada: Guysborough Village, one of the small towns in Nova Scotia, Toronto in Ontario and two in the United Kingdom, Wheatley Hill in Durham (marginally), and Wincanton in Somerset. It is difficult to imagine what these varieties have in common. Although only existential contexts are represented in Toronto, it is notable that Vernacular Pattern I prevails. Recall that Vernacular Pattern I is thought to be the most primitive, basic, pattern, yet this is a complex urban conservative/ standard variety. Vernacular Pattern II is identifiable by a contrast between a higher proportion of default agreement for affirmatives than negatives. Notice that this ranking of the constraint turns up in four communities – Tiverton in Devon in the southwest, York in the northeast and Cullybackey in County Antrim in Northern Ireland. Again, the common linguistic constraint does not offer insights into common characteristics of the communities. Vernacular Pattern III is identifiable by a contrast between a higher proportion of default agreement for negatives than

Variation as a window on universals

151

affirmatives. This constraint hierarchy is found in North Preston, Guysborough Enclave, Portavogie, Cumnock and Maryport. What do all these localities have in common? They too are rural, non-urban, peripheral but the fact that at least the United Kingdom locales are all in the north or in Northern Ireland suggests a northern British pattern. However, the contrast is also present in both of the African Nova Scotian in Canada. In sum, there are actually three patterns for the negation contrast. All possible configurations exist and there are no easy parallels across the varieties that share one pattern or the other. As before, a contrast analysis is performed on the same data in order to test the relative strength of this effect by geographic scale, as in Figure 11.

W rton in ca nt on

k

ve

or

Ti

Y

ill

t y

H

or

le

he at W

M

ar

yp

no

ck

ie m

ck

Cu

Bu

vo

gi

e

ey rta Po

Cu

lly

ba

ck

ag

e

ve

ill

la

tV

uy

sb

or

ou

gh

nc tE

gh ou G

G

uy

sb

or

N

or

th

Pr

es

to

n

factor weight

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

Figure 11: Contrast analysis: negative vs. affirmative.

Figure 11 shows that there are two places where this effect is strong: the African Nova Scotian communities, and York, in England. Recall however (Figure 10) that in North Preston and Guysborough Enclave the effect goes in the opposite direction to York. In every other case the negation effect is weak, regardless of direction, hovering below or well below 0.3. On the one hand the negative effect is scale dependent, that is local, by constraint ranking, but on the other hand it exhibits characteristics of a

152

Sali A. Tagliamonte

universal by the measure of contrast since it constrains the use of default agreement, one way or the other, in nearly every locale. In sum, the findings thus far support a building picture in which some effects which have appeared to be universal are not (the grammatical person hierarchy) and other effects which have appeared to be local (negation) may be underlyingly the same. These findings strongly support the suggestion that “great care must be exercised in attempting generalizations across all of varieties of English” (Siemund 2009).

6.

Narrowing the focus

In order to explore inter-variety differences more closely, I now conduct a comparison among four of the varieties that are most intimately linked historically yet can be distinguished on one key characteristic – region. These are the four dialects spoken in Southwest Scotland, Northwest England and Northern Ireland: Cumnock, Maryport, Cullybackey and Portavogie. These varieties offer unique corpora for the exploration of universals and variation due to their close geographic and historical connections. All the varieties have evolved from a single cultural region and from a common ancestral root and are often referred to under an umbrella term as Northern English (e.g. Wales 2006). However, because of their separate geographic and social developments, involving in part the current division into different countries (Scotland, England and Northern Ireland), but also varying histories, settlement patterns, population mixes, etc. they might well be differentiated in nuanced ways as well. This provides an opportunity to view the universal/local dichotomy by controlling as much as possible for similarity and then assessing where the differences may lie. Research in dialect syntax by Henry (1995) and more recently by Adger and Smith (2005) has shown that regional differentiation in vernacular data from northern Ireland and northern Scotland offer decisive insights into how variation can be embedded into a formal theory of grammar.7 To date, however, this research has focused on deriving an account based on a single dialect’s vernacular patterns without generalizing an 7 Henry found that Belfast English had default agreement in various contexts that differed from Standard English. She explained this difference by suggesting that there was a difference in parameter setting between the two varieties. Henry argued that the presence of default agreement in Belfast English was due to a parametric choice between: 1) move either to Tense or AGRs (the Belfast case) as opposed to 2) Standard English where verbs are required to

Variation as a window on universals

153

explanation to account for variable patterns across dialects. The dialects in the current archive provide a particularly good and timely opportunity to tease apart universal as opposed to local differences and eventually to more fully understand the mechanism that permits variation. To this point, the analyses have revealed that personal pronouns – you, we and particularly they – while variable, tend to require agreement. Yet Henry (1995: 18) argued that “it is not the case that all pronouns require agreement”, but that only a subset of them do. Indeed, personal pronouns are not the only type of pronoun. There are also conjoined subjects with pronouns, demonstrative pronouns, relative pronouns, and indefinite pronouns, as in (6).8 (6)

Conjoined subjects with pronouns a. Nelly and I was going to the pictures. (MPT/W) Demonstrative pronouns b. Them was all throwed away. (CLB/³) c. These was very small boats, like open boats. (PVG/3e) Relative pronoun d. They sat in chairs [that] was made out of barrels. (NI/CLB/å) e. Shirley looked at the three woman [that] was on it. (NI/PVG/8) f. There were some of them [ø] was nae dead at all. (NI/CLB/©) Indefinite pronoun g. Some of them was working overtime. (MPT/U)

In Belfast English, Henry discovered that these types of pronouns (i. e. conjoined pronouns and demonstratives) did not pattern with personal pronouns, but instead behaved like plural noun phrases – taking default agreement. Thus, distinguishing different types of pronominal subjects may be a critical diagnostic for understanding agreement phenomena in the regional dialects in this part of the world. Figure 12 shows the distribution of default agreement distinguishing all these subject types across the dialects in Scotland, England and Northmove to AGRs. The application of such an analysis to the corpora studied here is in progress. 8 Conjoined subjects were not differentiated in the previous constraint analysis due to their sparseness in the data. Indeed the narrowing focus of this investigation necessitates a more qualitative approach since the number of tokens of these rarer subject types is significantly reduced in comparison to the gross contrast between full noun phrase subjects vs. person pronouns.

154

Sali A. Tagliamonte

ern Ireland. For comparison, default agreement with existential subjects has been superimposed on the distributional patterns as shown in the solid black line. FULL NP

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

Conjoined NP Relative Prounoun Demonstrative Pronoun Indefinite Pronoun Existential pronoun ‘there’ Northern Northern Ireland, Ireland, Cullybackey Portavogie

Scotland, Cumnock

England, Maryport

Figure 12: Distribution of default agreement distinguishing subject types.

Figure 12 reveals that in Cullybackey in Northern Ireland existential pronouns have a low frequency of default agreement (see also Figure 2), whereas all the other pronoun types have a high frequency. This is precisely the pattern Henry reported for Belfast and so appears to confirm her analysis for another Northern Ireland dialect. However, notice that the other Northern Ireland dialect in this analysis – Portavogie – behaves quite differently. In Portavogie existentials pattern with relative pronouns – while the other subject types are lower and demonstratives lowest of all. Cumnock, in Scotland, shares this propensity for default agreement with existentials and relative pronouns, but here default agreement is much less likely with the other subject types. Finally, in Maryport another pattern emerges, existentials have highly frequent use of default agreement, but here full noun phrases as well as all the other pronoun types pattern together at much lower rates. Abstracting slightly away from this complex picture, Figure 13 now groups “other” subject types together as opposed to existentials and full noun phrases in Cullybackey. Note that full noun phrases and other categories pattern together.

% default agreement

Variation as a window on universals

100 90 80 70 60 50 40 30 20 10 0

FULL NP

Other

155

Existential pronoun ‘there’

Northern Ireland, Cullybackey

Figure 13: Pattern I, subject type.

% default agreement

,Q3RUWDYRJLHDQG&XPQRFNLQ)LJXUHRWKHUSURQRXQVDQGH[LV WHQWLDOVSDWWHUQWRJHWKHU 100 90 80 70 60 50 40 30 20 10 0

FULL NP

Northern Ireland, Portavogie

Other

Existential pronoun ‘there’

Scotland, Cumnock

Figure 14: Pattern II, subject type.

In Maryport, in Figure 15 existentials pattern uniquely, noun phrases are lower and other subject types are lower still. This exercise has revealed that the pronouns across varieties exhibit three distinct patterns of agreement.

% default agreement

156

Sali A. Tagliamonte

100 90 80 70 60 50 40 30 20 10 0

FULL NP

Other

Existential pronoun ‘there’

England, Maryport

Figure 15: Pattern III, subject type.

Examples of these patterns may be viewed in the examples in (7–9). (7)

Northern Ireland, Cullybackey Existential agreement; Full NPs default agreement; other subjects default agreement a. There were some of them [ø] was nae dead at all. (NI/CLB/©) b. You dropped them and [ the drills] was closed. (NI/CLB/å) c. But [Lily and er Maggie and Sid and them] was all in the choir. (NI/ CLB/å)

(8)

Portavogie (NI) and Cumnock (Scotland) Existential default; other subjects default agreement; personal pronouns agreement a. [Them] was all throwed away the minute the electric come in [they] were all throwed out. (CMK/³) b. Well ald banks, [they] ’re rascals; [Them] ’s crooks, aye. (PVG/2)9

(9)

Maryport (Northern England) Existential default; Full NP default agreement; other subject types, agreement. a. [There] was about seventy ships that were sent over. (MPT/R) b. [ Lads ] was in ya bit and [ lassies ] was in t’ other. (MPT/u)

9 Although this example shows the verb to be is in the present tense; it suggests that the same contrast is found across present and past tense.

Variation as a window on universals

157

factor weight

c. And a fore they knew where [they] were at [there] was cobbles and all sorts flying at windows. (MPT/R) d. But [Jean and my twin brother and I] were too young. (MPT/c) e. [Them] were great hobbies for me. (MPT/d)

1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Cullybackey

Maryport

Cumnock

Portavogie

Figure 16: Contrast analysis subject type.

Figure 16 reveals that the constraint on subject types operates in all communities; however, its effect is strong in Cullybackey and Maryport – albeit operating in the opposite direction – and weak in Portavogie and Cumnock. Not only does the constraint hierarchy differ across dialects, so does its contrast (or strength). Thus, while all the dialects have default agreement in every context, they differ qualitatively in how the constraint manifests, particularly with regard to the minor category – the other pronominal subject types. To sum up the findings for these four dialects, Table 5 shows the results for inter-variety constraint ranking. Table 5: Summary of constraints. CLB

PVG

CMK

MPT

X

—

—

—

Northern Subject Rule, NP vs. Pro —

—

—

—

Existential Constraint Negation

X

—

—

—

Type of subject

X

—

—

Y

158

Sali A. Tagliamonte

Interestingly, the only constraint that is parallel across all four is the Northern Subject Rule. This result jibes well with the fact that these are all Northern Englishes. Portavogie, Cumnock and Maryport also share the existential constraint and the negation effect. Portavogie and Cumnock share all the constraints. Cullybackey, stands apart from the rest on three counts. Mayport on one. Table 6 summarizes the findings for the inter-variety contrast analysis of these constraints. Table 6: Summary of contrast. CLB

PVG

CMK

MPT

Existential Constraint

strong

moderate

moderate

strong

Northern Subject Rule, NP vs. Pro

very strong

very strong

very strong

moderate

Negation

weak

weak

weak

weak

Type of subject

strong

moderate

moderate

strong

Table 6 reveals that the negative effect has the same strength across all dialects. Portavogie and Cumnock share contrast effects across the board whereas Cullybackey and Maryport differ. Surprisingly, the parallels across dialects are not intuitively obvious. While you might think that two dialects in Northern Ireland might pattern together, they often do not. Instead, we find that Portavogie and Cumnock tend to pattern together. Cullybackey stands apart on two key dimensions: 1) existential subjects have default agreement only moderately, unlike any other location studied; and 2) the negation constraint operates with a ranking antithetic to the other Northern Ireland community, in Cullybackey negative constructions get standard agreement. As it happens, some dialects in the United Kingdom and Northern Ireland have an alternative pivot for regularization. Instead of a paradigm leveling to was, were occurs across the board. This is often referred to in the literature as were regularization (e.g. Schilling-Estes and Wolfram 1994), as in (10). (10)

a. There were a wee alarm-clock sat on the window. (CLB/å) b. I were just thinkin’ that. (CLB/î) c. She were a great worker mi mother. (CLB/³)

Variation as a window on universals

159

Figure 17 shows the distribution of were regularization across communities. singular ‘were’

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

plural existential ‘were’ (standard) standard ‘were’ elsewhere

Cullybackey Portavogie

Cumnock

Maryport

Figure 17: Distribution of ‘were’ regularization across communities.

Figure 17 reveals that all the dialects have a certain amount of were regularization; however, of all the dialects Cullybackey has the most. I suggest that the frequent occurrence of were outside of its standard contexts of use influences the use of were generally. Supportive to this hypothesis is the fact that for all the dialects the rate of were in existentials patterns along with the rate of were in non-standard contexts. Where were is more frequent in regularized contexts, it is also more frequent in existentials. This may also explain the high rate of standard agreement in negatives in Cullybackey since 93 % of the negative constructions in the data are also existential, as in (11). (11)

There were nae cars to bring them in them days. (CLB/å)

In other words, in some varieties the pivot for regularization is towards were not was (Schilling-Estes and Wolfram 1994). This influences the use of were for existentials despite the strong tendency for default agreement in this context in other varieties. Further, because existentials tend to be negative, this makes negative constructions to be marked with were as well. This reveals how much the patterning of one aspect of the grammatical context influences the patterning of another and highlights the

160

Sali A. Tagliamonte

importance of seeking out and distinguishing linguistic features according to the systems of grammar with which they intersect. Another dilemma remains. Why are there so many differences between the two Northern Ireland communities, Portavogie and Cullybackey, but parallels between Portavogie and Cumnock, the Lowland Scots location? Here, external factors come to the fore. Although Portavogie and Cullybackey are both small fishing villages with historical links to Scotland, only Portavogie has had continuing links to the Scottish coast. Moreover, a large study in biological anthropology, a method which measures the similarity between two populations based on a model of genetic relationship, found substantive links between Portavogie and locales in Scotland due to the shorter crossing between ports (Smith, Hepburn, and MacRaild 2003–2006). This evidence conveys information about relationship. Interpreting it, along with the linguistic parallels, provides a partial explanation for why these dialects share so many aspects of default agreement with each other. They have a common and enduring history. This case provides a good example of how internal factors and external social and other factors converge to explain linguistic phenomena.

7.

Discussion

This paper has studied default agreement by analyzing its variable occurrence in preterit contexts of the verb to be, i. e. the use of was in contexts where were is standard. The data come from spoken data from 13 different varieties of English distinguished by levels of geographic scale as well as socio-historical and cultural characteristics. The findings are based on thirty logistic regression analyses, an analysis of constraint hierarchy, an analysis of strength of the individual constraints (i. e. contrast), and an analysis of geographic scale. Default agreement occurs in all these varieties just as it has been reported in many different places all over the world as well as in the history of English. Thus, by incidence alone default agreement can be considered to have “a very wide areal and/or social reach” using the terminology of Kortmann and Szmrecsanyi (this volume), since it occurs in 70–78 % of the varieties they surveyed. However, the analyses of constraints and contrast have uncovered much more complexity (and difference) than previously reported. Furthermore, consistent with Horvath and Horvath (2003), the measure of constraint ranking alone has been shown to hide important aspects of the variation unless contrast is considered as well. When the strength of the relationship between categories within a given hierarchy is measured the analysis uncovers variability

Variation as a window on universals

161

within universality or “sub-regularity” (see Joseph, this volume). While a specific constraint may apply across the board, its strength as well as its pattern across varieties may differ dramatically. The question is whether there is any general principle that can explain all this – why do some patterns operate globally and others only locally? I suggest the answer lies in the nature of the contrast. It is unmistakable that the grammatical context is critical. Default agreement with existential constructions, i. e. there was, is pervasive. Indeed, the contrast between existential constructions and all other grammatical contexts is the foremost scale independent effect. Default agreement with plural noun phrase subjects is also widespread, i. e. the boys was, yet it is not nearly as frequent as in existential constructions. In contrast, the use of default agreement in (personal) pronoun contexts, i. e. you was, we was, they was, appears to be a feature of certain rural, isolated, or peripheral communities. Does this qualify it as a vernacular universal as opposed to a universal? Default agreement in this context is extremely erratic across varieties, confined and particular. Moreover the restricted use of this feature in rural locales and the fact that these contexts do not exhibit scale independent patterns suggests that it is an obsolescent feature. If so, it is definitely a vernacular feature, but only primitive in the sense of ‘old’. In the contexts where default agreement is robust, namely with existentials and plural noun phrases, the constraints and contrast effects are telling. Where there is a binary opposition, such as negative vs. affirmative, the variation reveals a constraint ranking that may go one way or the other. For some varieties, default agreement occurs more frequently with negatives, i. e. wasn’t than affirmatives. In other varieties, default agreement occurs more frequently in affirmatives, i. e. was, than negatives, i. e. weren’t. However, the contrast analysis for the negative constraint shows scale independence. In other words, the constraint itself exists nearly everywhere; it is only the way it patterns that differs. In constraints that have several categories within the hierarchy pairs of relationships are often strongly ordered and it is the 3rd or 4th category that varies (see also Horvath and Horvath 2003). For instance the use of default agreement with 2nd person singular you and 1st plural we is erratic. However opposition pairs like noun phrases vs. personal pronouns and existential vs. non-existential are more scale independent, and by the criteria set at the outset, this gives them status as universals. It also appears that constraints are multi-layered. Looking deeper into the nature of the subject, for example, patterns of default agreement were uncovered which systematically differentiated types of pronouns. Even though default agreement is also diminishing in these contexts, the extant

162

Sali A. Tagliamonte

patterns tenaciously expose locality. While one of the Northern Ireland dialects (Cullybackey) displayed the constraint ranking reported by Henry (1995) for Belfast, the other variety (Portavogie) did not and instead patterned with a Scots dialect. Moreover, three regional differences were found. This demonstrates that even among closely related dialects there are particular ‘weak spots’ in the grammar that avail themselves to (local) variation. Not the core contrasts, like existential vs. other, but the more peripheral categories – such as the relative strength of pronouns (strong vs. weak), the composition of the determiner phrase, or the ranking for negative effects. These do indeed appear to involve aspects of structure that permit deviations on a theme. The way forward is to pursue the formal apparatus that can explain all these patterns. Are they the result of parameter setting as envisaged by Henry (1995). Do they derive from feature specifications of lexical items as laid out by Adger (2006) or are they the result of competing grammars in the spirit of Kroch (1989; 1999)? While a definitive answer to such questions must be left for another phase of research, the findings exposed here permit a number of speculative observations. Over-arching syntactic differences that arise from the re-arrangement of phrase structure (i. e. existential vs. canonical SVO word order) appear to impact universally on English dialects. Thus, if there is a language universal to be found in the study of default agreement this feature shows the most promise. Future research on languages other than English will elucidate this possibility. Similarly, the widespread and consistent contrast between rich and complex subjects as opposed to personal pronouns also seems to bear the hallmarks of a universal, as does the contrast between negative and affirmative. In both cases, however, the internal composition of the structures involved and the way they operate exhibit local contrasts. Such differences may be explained by the nature of variation itself. As recently been noted by Adger and Smith (to appear) certain types of variation involve feature specification of function categories while others involve underspecification in the mapping between these categories and morphological forms. This may be the way forward in explaining why some constraints apply on a global level while others diverge under local conditions. Indeed, the approach to default agreement adopted by Adger and Smith (e.g. Smith and Tagliamonte 1998; Smith 2000; Adger and Smith 2005) based on a single community – Buckie, a variety of Scots spoken on the far north shore of Scotland – provides a compelling model to test across the varieties studied here. In particular, the widely diverging frequencies of default agreement as well as the contrasting grammatical patterns across communities pose a challenge to the explanatory power of their analysis.

Variation as a window on universals

163

In sum, so-called universal features may operate globally but manifest differently in varying scenarios. Variation provides a window to the nature of the underlying mechanisms via the operation of constraints and their strength across varieties. Variation reveals the core contrasts (the universals) but the constraints and their strength show the peripheral patterns (the local deviations). This is how variation provides a unique window on universals. It reflects the constellation of influences in the local setting, however close examination of the patterns and contrasts from a comparative perspective exposes where the universal attribute holds firm.

Abbreviations In Scotland: BCK = Buckie, CMK = Cumnock; in Northern Ireland: CLB = Cullybackey; PVG = Portavogie; in England: MPT = Maryport, WHL = Wheatley Hill, YRK = York; DVN = Tiverton (Devonshire); SMT = Wincantin (Somerset); in Canada: TOR = Toronto; NPR = North Preston; GYE = Guysborough Enclave; GYV = Guysborough Village.

164

Sali A. Tagliamonte

References Adger, David 2006 Combinatorial variability. Journal of Linguistics 42(3): 503–530. Adger, David and Jennifer Smith to appear Variation in agreement: A lexical feature based approach. Lingua. Adger, David and Jennifer Smith 2005 Variation and the minimalist program. In: Leonie Cornips and Karen Corrigan (eds.), Syntax and Variation: Reconciling the Biological and the Social, 149–178. Amsterdam: Benjamins. Anderwald, Lieselotte 2002 Negation in Non-Standard British English. London/New York: Routledge. Breivik, Leiv Egil 1983 Existential there: A Synchronic and Diachronic Study. Bergen: Department of English, University of Bergen. Britain, David and Andrea Sudbury 2002 There’s sheep and there’s penguins: Convergence, ‘drift’ and ‘slant’ in New Zealand and Falkland Island English. In: Mari C. Jones and Edith Esch (eds.), Language Change: The Interplay of Internal, External and Extra-Linguistic Factors, 211–240. Berlin: Mouton de Gruyter. Brunner, Karl 1963 An Outline of Middle English Grammar. Oxford: Blackwell Publishers. Chambers, J. K. 1995 Sociolinguistic Theory: Linguistic Variation and its Social Significance. Oxford: Blackwell Publishers. Chambers, J. K. 2000 Universal sources of the vernacular. In: Ulrich Ammon, Peter H. Nelde, and Klaus J. Mattheier (eds.), Special Issue of Sociolinguistica: International Yearbook of European Sociolinguistics, 11–15. Tübingen: Max Niemeyer Verlag. Chambers, J. K. 2001 Vernacular universals. In: Josep M. Fontana, Louise McNally, Teresa M. Turell and Vallduvi Enric (eds.), Proceedings of ICLaVE!, The First International Conference on Language Variation in Europe, 52–60. Barcelona: Universitate Pompeu Fabra. Chambers, J. K. 2003 Sociolinguistic Theory: Linguistic Variation and its Social Significance. Malden/Oxford: Blackwell Publishers. Chambers, J. K. 2004 Dynamic typology and vernacular universals. In: Bernd Kortmann (ed.), Dialectology Meets Typology: Dialect Grammar from a CrossLinguistic Perspective, 127–145. Berlin: Mouton de Gruyter.

Variation as a window on universals

165

Cheshire, Jenny 1982 Variation in an English Dialect: A Sociolinguistic Study. Cambridge: Cambridge University Press. Cheshire, Jenny, Viv Edwards and Pamela Whittle 1989 Urban British dialect grammar: The question of dialect levelling. English World-Wide 10(2): 185–225. Christian, Donna, Walt Wolfram and Nanjo Dube 1988 Variation and Change in Geographically Isolated Communities: Appalachian English and Ozark English. Tuscaloosa, Alabama: American Dialect Society. Curme, George O. 1977 A Grammar of the English Language. Essex, Connecticut: Verbatim. Feagin, Crawford 1979 Variation and Change in Alabama English: A Sociolinguistic Study of the White Community. Washington, D. C.: Georgetown University Press. Forsström, Gösta 1948 The Verb ‘to be’ in Middle English: A Survey of the Forms. Lund: C.W.K. Gleerup. Fries, Charles Carpenter 1940 American English Grammar. New York: Appleton, Century, Crofts. Henry, Alison 1995 Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Horvath, Barbara M. and Ronald J. Horvath 2003 A closer look at the constraint hierarchy: order, contrast, and geographical scale. Language Variation and Change 15(2): 143–170. Ihalainen, Ossi 1994 The dialects of England since 1776. In: Robert Burtchfield (ed.), The Cambridge History of the English Language. English in Britain and Overseas: Origin and Development, 197–270 (Volume 5). Cambridge: Cambridge University Press. Jespersen, Otto H. 1940 A Modern English Grammar on Historical Principles: Part V: Syntax. London: George Allen and Unwin. Klemola, Kaarlo Juhani 1996 Non-standard periphrastic do: A study of variation and change. Ph.D. dissertation, University of Essex. Kroch, Anthony s. 1989 Reflexes of grammar in patterns of language change. Language Variation and Change 1(3): 199–244. Kroch, Anthony S. 1999 Syntactic change. In: Mark Baltin and Chris Collins (eds.), Handbook of Syntax. Malden: Blackwell Publishers.

166

Sali A. Tagliamonte

Labov, William 1972 Sociolinguistic Patterns. Philadelphia: University of Pennsylvania Press. Labov, William 2007 Transmission and Diffusion. Language 83(2): 344–387. McCafferty, Kevin 2003 The northern subject rule in Ulster: How Scots, how English? Language Variation and Change 15(1): 105–139. Meechan, Marjory and Michele Foley 1994 On resolving disagreement: Linguistic theory and variation – There‘s bridges. Language Variation and Change 6(1): 63–85. Montgomery, Michael B. 1994 The evolution of verbal concord in Scots. In: Alexander Fenton and Donald A. MacDonald (eds.), Studies in Scots and Gaelic: Proceedings of the 3rd International Conference on the Languages of Scotland, 81–95. Edinburgh: Canongate Academic Press. Montgomery, Michael B. 1995 The linguistic value of Ulster emigrant letters. Ulster Folklife 41: 1–16. Montgomery, Michael B. 1997 Making transatlantic connections between varieties of English. Journal of English Linguistics 25(2): 122–141. Montgomery, Michael B. and Curtis Chapman 1992 The pace of change in Appalachian English. In: Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.), History of Englishes: New Methods and Interpretations in Historical Linguistics, 624–639. Berlin: Mouton de Gruyter. Montgomery, Michael B., Janet M. Fuller and Sharon DeMarse 1993 “The black men has wives and sweet harts [and third person plural -s] jest like the white men”: Evidence for verbal -s from written documents on 19th-century African American speech. Language Variation and Change 5(3): 335–357. Murray, James A. H. 1873 The Dialect of the Southern Counties of Scotland: Its Pronunciation, Grammar and Historical Relations. London: Philological Society. Pietsch, Lukas 2005 Variable Grammars: Verbal Agreement in Northern Dialects of English. Tübingen: Niemeyer. Poplack, Shana and Sali A. Tagliamonte 1991 African American English in the diaspora: Evidence from old-line Nova Scotians. Language Variation and Change 3(3): 301–339. Poplack, Shana and Sali A. Tagliamonte 2001 African American English in the Diaspora: Tense and Aspect. Malden: Blackwell Publishers.

Variation as a window on universals

167

Pyles, Thomas 1964 The Origins and Development of the English Language. New York: Harcourt, Brace and World, Inc. Sankoff, David, Sali A. Tagliamonte and Eric Smith 2005 Goldvarb X. Department of Linguistics, University of Toronto, Toronto, Canada. http://individual.utoronto.ca/tagliamonte/Goldvarb/GV_index.htm. Schilling-Estes, Natalie and Walt Wolfram 1994 Convergent explanation and alternative regularization patterns: were/weren’t leveling in a vernacular English variety. Language Variation and Change 6(3): 273–302. Schreier, Daniel 2002 Past be in Tristan da Cunha: The rise and fall of categoricality in language change. American Speech 77(1): 70–99. Siemund, Peter 2009 Linguistic universals and vernacular data. In: M. Fillpula, J. Klemola and H. Paulasto (eds.), Language Contacts: Evidence from Varieties of English and Beyond. London: Routledge, 321–346. Smith, Jennifer 2000 Synchrony and diachrony in the evolution of English: Evidence from Scotland. D.Phil. dissertation, University of York. Smith, Jennifer and Sali A. Tagliamonte 1998 ‘We were all thegither... I think we was all thegither’: was regularization in Buckie English. World Englishes 17(2): 105–126. Smith, Malcolm T., Anthony Hepburn and Donald MacRaild 2003–2006 The isonymic analysis of historical data; Irish migrant in Britain, 1851–1901. Economic and Social Science Council of the United Kingdom (ESRC) Research Project. Tagliamonte, Sali A. 2006 Analysing Sociolinguistic Variation. Cambridge: Cambridge University Press. Tagliamonte, Sali A. 2008 There was universals; then there weren’t: A comparative sociolinguistic perspective on ‘default singulars’. In: Markku Fillpula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals versus Contact Induced Change. Oxford: Routledge. Tagliamonte, Sali A. and Jennifer Smith 2000 Old was; new ecology: Viewing English through the sociolinguistic filter. In: Shana Poplack (ed.), The English History of African American English, 141–171. Oxford/Malden: Blackwell Publishers. Tagliamonte, Sali A., Jennifer Smith and Helen Lawrence 2004 Disentangling the roots: The legacy of British dialects in cross-variety perspective. In: Markkuj Filppula, Juhani Klemola, Marjatta Palander, und Esa Penttilä (eds.), Papers from Methods XI Joensu, Finland, 5–9 August 2003. Amsterdam: John Benjamins.

168

Sali A. Tagliamonte

Traugott, Elizabeth Closs 1972 A History of English Syntax; A Transformational Approach to the History of English Sentence Structures. New York: Holt, Rinehart & Winston. Visser, Fredericus T. 1963–73 An Historical Syntax of the English Language. Leiden: E. J. Brill. Visser, Fredericus T. 1970 An Historical Syntax of the English Language. Leiden: E. J. Brill. Wales, Katie 2006 Northern English: A Social and Cultural History. Cambridge: Cambridge University Press. Walker, James A. 2007 “There’s bears back there”: Plural existentials and vernacular universals in (Quebec) English. English World-Wide 28(2): 147–166.

Part 2 Contact-induced variation

Gender and contact – a Natural Morphology perspective on Scandinavian examples1 Hans-Olav Enger Abstract Traditionally, it has been claimed that contact leads to reduction, and this has been suggested to hold true also for Scandinavian. Indeed, it holds true both for lexical gender (section 2) and for declensions (section 5) in Norwegian. However, it is too simple to talk of the Scandinavian gender system in terms of reduction only (sections 3 and 4). In the gender system, there are also cases of increase rather than reduction. Pronominal gender is not reduced; in fact, the opposite, and in some ways, gender has become more central in Scandinavian (sections 3 and 5). Although there has been reduction of gender along the paradigmatic axis, this is not the full story, for there is also a syntagmatic aspect to gender. The idea of gender reduction in contact should be restricted to lexical gender, and the way the reduction works, has to do with structural conditions also. ‘Numerical reduction’ and ‘simplification’ are not necessarily synonymous (section 4), and numerical increase is not necessarily synonymous with increased complexity. Within Natural Morphology, we can refine the main idea so as to keep its spirit: motivation is a better yardstick for simplicity than is numerical reduction. Keywords: contact, gender, lexical declension, Natural Morphology, Norwegian, paradigmatic, pronouns, reduction, Scandinavian, syntagmatic

1 Thanks to the University of Hamburg and Sonderforschungsbereich 538 for financing my stay in Hamburg in September 2007. Thanks also to Philipp Conzett, Tore Nesset, Peter Siemund, Hilde Sollid, Arne Torp, an audience in Hamburg and an anonymous referee for helpful comments and suggestions.

172

Hans-Olav Enger

1.

Introduction

1.1.

Basic ideas

A traditional claim in linguistics is that there tend to be ‘simplification processes’ – also known as ‘reduction’ – at work in the inflectional morphology of varieties that are exposed to much contact (including Creoles). The idea is, in a nutshell, that more contact will lead to simpler inflection (cf. e.g. Braunmüller 1984; Andersen 1988; Trudgill 1986, 1992; Kusters 2003, 2008; Hudson 1996; Sandøy 1998; Enger 1999; Koefoed and van Marle 2004; McWhorter 2001, 2008). The claim is thus that there is (or tends to be) a relation between the structure of a language and the sociohistorical conditions this language has been exposed to. For example, Old Norse had person agreement between the subject and the finite verb. Modern Norwegian and Icelandic both stem from Old Norse. Icelandic retains person agreement, Modern Norwegian does not. Norwegian has been exposed to more contact with other languages, notably Low German and Danish, than has Icelandic. It is assumed that the loss of agreement has to do with contact (see Kusters 2003 for further discussion). More generally, the morphology of Icelandic and Faroese is more intricate than that of Danish, Swedish and Norwegian, and this is often linked to the mainland languages having been exposed to more foreign influence, bilingualism, and imperfect language-learning by adult speakers. On a general basis and with no particular reference to Scandinavian, Koefoed and van Marle (2004: 158) state that “[m]orphology, particularly inflection, appears to be highly vulnerable to language contact”. The claim that contact leads to simplification is often substantiated by examples from conjugation and declension. In the Scandinavian languages, declension is closely linked to gender (cf. Enger 2004b). Therefore, it is not surprising that attempts have been made to link contact and ‘gender reduction’ as well (e.g. by Thomason and Kaufman 1988; Jahr 2008; Nesse 2002; Kusmenko 2000). However, declensional reduction does not by necessity entail gender reduction; there has been much declensional simplification that does not relate directly to gender (see section 5 below for an example). The historical development of the pronouns, as discussed in section 4 below, also shows that contact does not necessarily lead to gender reduction.

Gender and contact

173

The empirical focus for this paper is on Norwegian in particular. I shall argue that: ± The predicted reduction in contact varieties does indeed manifest itself in Scandinavian gender as well (as observed in previous literature). ± However, reduction along the paradigmatic axis is not everything, there is also a syntagmatic aspect to gender, which deserves more attention. ± Moreover, the contact varieties show that gender has become more important in several respects. There is an increased inventory of pronouns, and if simplification is synonymous with decreased inventory, then this is a problem. ± Simplification and increased complexity is not only conditioned by contact, but also by structural conditions. ± Simplification is not necessarily synonymous with numerical reduction, and numerical increase is not necessarily synonymous with increased complexity. ± The idea of gender reduction in contact should be restricted to lexical gender. Two quotations from two different handbook articles underline the need for an article along present lines. Firstly, Aikhenvald (2004: 1042) claims that “[b]oth expansion and reduction of gender and noun class systems very often happen through diffusion […]”. My aim is to try and specify this statement a bit: in Scandinavian, the expansion and reduction associated with diffusion (or contact) do happen, but not in the same areas of the grammar. Secondly, Muysken (2004: 1655) notes the need for “a much more differentiated view of the process of morphological reduction”. It is to be hoped that this paper can contribute to such a view.

1.2.

Reduction and simplification

In the literature on the Scandinavian languages and dialects, there are numerous observations on ‘gender reduction’ linked to contact (cf. section 2 below). Similar observations are found for other languages. For example, Boretzky (2004: 1650) presents an example of the reduction of number of genders in certain dialects of Latvian, and he suggests that this is due to contact with Finnish. Bechert (1982) argues that bilinguals play an important role for the reduction of gender systems in Germanic. The idea of gender reduction because of contact is a variant of the very traditional idea of ‘simplification’ due to contact. This idea is also found

174

Hans-Olav Enger

elsewhere in Scandinavian language histories. For example, in North Swedish and East Norwegian, there is a consonant phoneme /ˆ/. It is often said to be phonetically the same sound as found in Hindi, i. e. a retroflex flap, but this is questioned on articulatory grounds by Simonsen, Moen and Cowen (2000). Be that as it may, the sound is rare in other parts of Europe.2 Intriguingly, /ˆ/ is found in all the dialects surrounding Stockholm, but not in the urban dialect of the city. By traditional accounts, the absence of /ˆ/ in Stockholm has to do with contact with speakers of other languages, notably Low German. There were many people from the Hansa in the city, but very few in the surrounding areas./ˆ/ is already in 1678 called “impossible for foreigners” (cf. Wessén 1969: 165–166). Thus, the idea of “simplification due to contact” has not been restricted to inflection. On grounds that partly have to do with computer simulation, Nettle (1999: 134) claims that “marked structures are more likely to persist in small groups”. While this claim is not synonymous with the claim that contact can lead to reduction, it is clearly related, and all the more relevant here because Nettle draws on a very different kind of evidence (computer simulations). Before we go on, it is important to note that contact does not necessarily lead to gender reduction. An example is the Swedish dialect spoken on the island of Nuckö in Estonia from approximately 1300 to 1943/44 (cf. Lagman 1971). While surrounded by Estonian for centuries, this dialect was extremely conservative in its gender system, more than almost all Norwegian dialects (which were less exposed to contact): it retained threegender agreement in the plural of adjectives. Swedish spoken in Sweden by and large did not. Now, Nichols (2003: 303) takes gender to be a feature that is “somewhat recessive, of high stability only when reinforced by gender systems in neighboring languages”. Recessive features tend to become less frequent over time, according to Nichols. Interestingly, Nichols’ claim is similar to a claim made independently by Nesse (2002: 229): “Contact between a Scandinavian dialect and one or more genderless, non-related languages leads to the dissolution of the gender system” (translation Duke 2008). Duke labels Nesse’s hypothesis “bold, problematic and stimulating”. Considering how conservative the gender system of Nuckö Swedish has been, the hypothesis seems too strong. Similarly, Swedish dialects spoken in Finland have tended to retain the three gender opposition in the singular

2 It is found in one village (Pigna) in Northern Italy, however.

Gender and contact

175

until very recently (cf. e.g. Rabb 2007) and they have certainly been no less conservative than Swedish spoken in Sweden, despite intensive contact with Finnish for many centuries (see also section 6). Thus, contact does not have to lead to reduction. Still, contact can open for reduction. An obvious question is why. In very general terms, where there is contact, there is competition between linguistic forms. In that case, it is not surprising that forms that in some respects are “simpler” or less “marked” should tend to win. This is a very simplistic characteristic, but it is sufficient for present purposes.

1.3.

Theory

In this paper, I am applying Natural Morphology, in particular the version espoused by Wurzel (1984, 1986, 2000) – more accessible to the Englishreading public through Carstairs-McCarthy’s (1992: ch. 8) brilliant survey. This framework is admittedly not en vogue today, but Andersen (2005: 163) calls Natural Morphology “currently the only systematic, coherent framework for the description and explanation of morphological change”. It is a framework that allows us to link many observations rather directly to the Neogrammarian tradition, which remains central in the writing of language histories. Also more recent publications (e.g. Gaeta 2007; Enger 2007) indicate that Natural Morphology still can be used profitably. There is a risk in linguists becoming “dedicated followers of fashion”, searching prematurely for a “Grand Theory of Everything”, and Natural Morphology is relevant for present purposes, for several reasons. Firstly, the theory does not lead us to expect overall simplification. On the contrary, Natural Morphologists, such as Wurzel (2000) and related theorists (such as Vennemann 1993), have repeatedly emphasized that simplification with respect to one parameter can often lead to more complexity with respect to another parameter. This does not necessarily amount to a defense of an aprioristic assumption that all languages somehow have to be equally complex. Nonetheless, it would be unwise to neglect the old insight that a simplification in one field often leads to changes in another field, and that these other changes do not have to be simplifications themselves. Secondly, Natural Morphology helps us formulate the idea of simplification as reduction in lexical marking (Wurzel 1984). For example, when the formerly strong verb backen ‘bake’ in German becomes weak, the verb no longer needs to be marked as [+ strong] in the lexicon. The notion of simplification is also used extensively in much sociolinguistics and his-

176

Hans-Olav Enger

torical linguistics (e.g. Trudgill 1986; Thomason and Kaufman 1988), but sometimes in such a way as to become imprecise and thus objectionable to some linguists. Thus, the only reference to the word ‘simplification’ I have found in the relatively recent Handbook of Historical Linguistics is in a contribution by Hale (2003: 352), who says that “it is not clear that notions such as ‘optimization’ and ‘simplification’ have any interesting role to play in diachronic linguistics”. A better working hypothesis, in my view, is that although the notion of complexity is problematic, it has a role to play. Indeed, concepts as ‘simplification’ and ‘complexity’ appear to undergo something of a revival, occurring in the title of recent books as Kusters (2003), Miestamo et al. (2008). Many of the changes in Scandinavian gender that sociolinguists as Thomason and Kaufman (1988) discuss under the heading of “simplification” could count as ‘increase in naturalness’ (reduction of lexical marking) in a Natural Morphology approach. Thus, there should be room for dialogue. Thomason and Kaufman (1988: 315) see changes in the gender systems in standard Swedish and Danish in terms of simplification.3 There is clearly much truth in this view, as we shall see below, but we shall also see that this approach may need some refining. In particular, contact varieties such as standard Swedish and standard Danish do not only display ‘reduction’ in their gender system, if we consider other sides of the system, in particular pronouns, cf. section 3 below. Indeed, Thomason and Kaufman (1988: 87) admit that “[d]eciding whether a given change simplifies the grammar, complicates it, or has neither effect is not always easy”. Natural Morphology is one possible approach to this problem.4

3 However, it is somewhat disappointing to see Thomason and Kaufman make references to “Landsmaal”. This is admittedly a detail, but the label went out of official Norwegian use in 1929 (cf. Vikør 1993). Thomason and Kaufman (1988: 318) claim that “Except for gender […] agreement on noun modifiers and the definite form of adjectives Danish and Swedish are every bit as analytical as English”, emphasis original. I find it hard to agree. While neither standard Danish nor standard Swedish retain any person or number agreement (while standard English may be said to retain a vestige), both Danish and Swedish have a ‘synthetic’ s-passive, and they have definiteness marking on nouns. Both are non-analytic features that English lacks. 4 In a discussion of Natural Morphology and language contact, Stolz (1987) has emphasized the importance of contact situations for the theory, focusing on the problems that Natural Morphology meets with. Stolz’s points are well taken, but I hope to show in this chapter that there are certain advantages to the theory as well.

Gender and contact

1.4.

177

Plan

The paper is structured as follows. Section 2 is devoted to a case of reduction in Scandinavian gender; in the older literature, it has been accounted for by ‘sound law’, in the more recent literature, a more sociolinguistic account involving contact is usually favored. While there clearly is much to be said for this more recent approach, some problems with the focus on gender reduction are pointed out in section 3. The idea is clearly justified on paradigmatic grounds, but syntagmatically, it is not obvious that gender is reduced. Pronominal gender is not reduced, rather expanded – especially in contact varieties. It is thus too simple to talk of the Scandinavian gender system in terms of reduction or simplification only. We go on to refine the idea of reduction and eliminate the exception in section 4. This is not an ad hoc move, but in line with the framework of Natural Morphology (Wurzel 1984). Section 4 also shows that the two concepts ‘numerical reduction’ and ‘simplification’ are not necessarily synonymous. While this is well known to Natural Morphologists, my message may hopefully be useful to some sociolinguists and historical linguists. In section 5, we look more closely at the interaction between gender and declension and go into declension in some detail. Contact and simplification appear to be relevant here as well. Section 6 presents a warning against assuming automatically that phonological change is sufficient to explain changes. In section 7, the paper is summarized.

1.5.

A warning to critics

A central claim in this paper is that there is a relation between language contact and structural reduction (cf. 1.1). This claim is often questioned on the grounds that counter-examples exist or that they are too numerous. This concern is understandable, but it is not sufficient grounds for dismissing the hypothesis. Firstly, the logical alternative to the hypothesis that there is a relation between A and B is the hypothesis that there is no relation between A and B. In other words: linguists who really wish to dismiss our starting-point should, for example, try to explain why gender reduction and related simplification processes should happen in exactly those Norwegian dialects where it does occur (see 2 and 5 below) – without any kind of reference to social factors. Such a project is unlikely to succeed, in my view. Kusters (2003) presents evidence that verbal inflection is sociolinguistically sensitive and that changes in societal structure are accompanied by changes in

178

Hans-Olav Enger

verb inflection in Quecha, Arabic, Swahili and North Germanic. When the evidence for a relation comes from so different languages, the claim that there is no relation becomes less appealing. There are too many counterexamples against the hypothesis that there is absolutely no relation between contact and reduction, and I do not think many linguists really want to commit themselves to such a view. Secondly, tendencies are respectable in other sciences. Some critics apparently wish to see a list of examples and counter-examples (see e.g. Stolz 1987: 41). This is a legitimate call for further research, but not for rejection. The only quantitative studies that I am aware of, Parkvall (2008), who discusses creoles, and Szmrecsanyi and Kortmann (2009, see also Szmrecsanyi and Kortmann, this volume), whose concern is with different varieties of English, present evidence that support the relation between contact and reduction assumed here. Thirdly, Bauer and Hay (2007) have recently found a surprising but statistically significant correlation between the size of the phoneme inventory of a language and the number of speakers of that language: larger languages have more phonemes (!). When such correlations are actually uncovered these days, it is premature to dismiss the idea of a relation between language contact and structural reduction.

2.

Reduction due to contact in the Scandinavian gender system5

In the literature on Scandinavian diachrony, gender reduction features prominently. Table 1 shows the contrast between Old Norse as presumably spoken in Norway around 1200 and the Bergen dialect (of Norwegian) spoken today.

5 This section draws heavily on arguments made by Jahr (2008), Nesse (2002), Kusmenko (2000).

Gender and contact

179

Table 1: Reduction of gender – from Old Norse to Bergen. determiner + INDEF.SG.NOM Old Norse

einn leikr

ein bók

eit barn

Bergen

/en le:k/

/en bu:k/

/et ba:‰n/

‘a game’

‘a book’

‘a child’

POSS.1SG

+ INDEF.SG.NOM

Old Norse

minn leikr

mín bók

mitt barn

Bergen

/mi:n le:k/

/mi:n bu:k /

/mit ba:‰n/

DEF.SG

Old Norse

leikrinn

bókin

barnit

Bergen

/le:ken/

/bu:ken/

/ba:‰ne/

Genders are usually understood as agreement classes, in other words as “classes of nouns, reflected in the behavior of associated words” (Corbett 1991: 1, quoting Charles Hockett). The “associated words” of interest in Table 1 are the so-called determiners (also known as indefinite articles), the function words preceding the noun itself. Old Norse has einn (masculine), ein (feminine), eit (neuter), Bergen has en, et. Given the data in 1, Old Norse has three genders and Bergen two, and this is the background for the traditional label gender reduction. The possessives support this picture: Old Norse has three different ones, minn (masculine), mín (feminine) and mitt (neuter), while Bergen only has two, /mi:n/ and /mit/. The suffixes in the definite singular in Table 1 present an interesting analytical problem, both in Old Norse and in Modern Scandinavian. By the definition of genders as “classes of nouns, reflected in the behavior of associated words” (Corbett 1991: 1), the definiteness suffixes – such as Old Norse -inn or Bergen -en – are not exponents of gender in the strict sense; they are not found in associated words, but on the noun itself. Nevertheless, these suffixes are strong indicators of gender. If a noun ends in -en in the definite singular in the Bergen dialect, it is a common gender noun. If a noun ends in -et in the definite singular, it is a neuter noun. So even if we linguists should not treat definiteness suffixes in isolation as criterial for gender, it is nevertheless defensible to treat the definiteness suffixes as indicative of gender (see Enger 2004b for further justification).

180

Hans-Olav Enger

Now, a classic question in the study of Norwegian is why the gender reduction has taken place in Bergen, since it has not taken place in any other traditional dialect of Norwegian. Compare Table 2. Table 2: No reduction in gender – from Old Norse to Oslo. determiner + INDEF.SG.NOM Old Norse

einn leikr

ein bók

eit barn

Oslo

/en le:k/

/æi bu:k/

/et ba:£/

POSS.1SG

+ INDEF.SG.NOM

Old Norse

minn leikr

min bók

mitt barn

Oslo

/mi:n le:k/

/mi: bu:k/

/mit ba:£/

DEF.SG

Old Norse

leikrinn

bókin

barnit

Oslo

/le:ken/

/bu:ka/

/ba:£e/

Most Norwegian dialects and the written language of Nynorsk6 are threegender systems of the type illustrated by Oslo in 2. On the other hand, Bergen is not unique in Scandinavia; the urban dialects of Stockholm and Copenhagen are two-gender systems of the type shown in 1, as are the written standard languages Swedish and Danish, and (optionally) the written Bokmål standard of Norwegian. So why has the Bergen dialect developed the way it has? Our Neogrammarian predecessors would posit sound laws. That is, they would claim that there was a regular phonological development responsible for the development -in>-en in the final syllable of the feminines in Bergen; a sound law that differed from those posited for other dialects. It is not necessary to go into all the Nordicist details here; suffice it to say that phonological change alone seems a less convincing explanation, as shown by more sociolinguistically oriented scholars, such as Jahr (e. g. 2008), Pedersen (1999), Kusmenko (2000) and Nesse (2002, 2005), among others. Another argument in their favor is that the Stockholm dia-

6 There are two written standards of Norwegian, Bokmål and Nynorsk. Briefly, Bokmål has been developed from Danish, Nynorsk is based on the dialects. See e.g. Vikør (1993) for a more comprehensive description.

Gender and contact

181

lect and the Copenhagen dialect are like Bergen (and not like Oslo) in showing gender reduction. Bergen, Stockholm and Copenhagen were central cities for the Hansa in the late Middle Ages (while Oslo was less significant). It seems too much of an accident that separate sound laws should be responsible for eliminating the feminine gender in all these three large Hanseatic cities and nowhere else. Boretzky (2004: 1649) puts the general principle nicely: “Wo […] eine innersprachliche Erklärung mit einer kaum wahrscheinlichen Häufung von Zufällen rechnen müßte, sollten wir […] Entlehnung annehmen” [Where an internal account has to operate with an improbable heaping of accidents, we should assume loan/ contact.]. Thus, the more recent and more sociolinguistically oriented generation of scholars has assumed that the primary cause of the gender reduction for the Bergen dialect (Jahr 2008, Nesse 2002, 2005) and for Copenhagen (Pedersen 1999) is not sound law, but contact (with Low German).7 The process of reducing three genders to two could have followed other paths, as other languages show (cf. Priestley 1983; Nichols 2003). For example, the neuter might have been eliminated, as in French. The usual case in Scandinavian, however, is that if anything should merge, it is the masculine and the feminine. This is overlooked even by Nichols (2003: 300), who claims that Indo-European languages either have a three-way masculine/feminine/neuter opposition (as in German) or a two-gender system with masculine and neuter collapsed into one (as in French), thus overlooking the Scandinavian model of simplification.8 Within Germanic, 7 Recently, Heide (2003) has tried to defend a primarily phonological account, but not entirely successfully (cf. Nesse 2005). 8 It has been argued that there is an important difference in the way threegender systems and two-gender systems work in North Germanic. According to Braunmüller (1999: 43), there is, in Icelandic and Faroese, no vacillations in gender comparable to German der/das Radio, Danish en/et circus (where the same noun can belong to different genders depending on idiolect, for example): “One (internal grammatical) reason for this is the fact that gender in a highly inflectional language, such as Icelandic, forms an integral part of its numerous inflectional classes […] and can therefore not be separated from other morphological features such as case or number”. The problem is that such ‘vacillation’ is attested from Old Norse. For example, Old Norse lóð ‘harvest, crop’ can be either a feminine or a neuter. It is difficult to see why Old Norse should be less ‘highly inflectional’ than Icelandic. Braunmüller (1999: 42) claims that “Only in those languages which have either lost a great deal of their case markers or show many merged inflectional suffixes can gender additionally be used to keep homonyms apart”. Examples are

182

Hans-Olav Enger

the masculine-feminine merger is found also in Dutch and Frisian (cf. Audring 2010). Presumably, the causes for the Scandinavian pattern of merger are both formal and semantic. The Old Norse masculine and the feminine are closer to one another, both in form and in content, than either is to the neuter. Neuter nouns have no indefinite plural suffix in Old Norse, masculines and feminines do; they even share some suffixes. While a common plural suffix for a masculine is -ar and for a feminine -ir, there are also masculines that get -ir and feminines that get -ar. The neuter is tendentially the inanimate gender; the masculine and feminine are tendentially animate. An important point to be noted here is that even if both the French and the Scandinavian pattern may be described as “simpler” than the Indo-European one, the reasons for either simplification can hardly be found in contact alone; it has to be related to structural conditions. From the formal point of view, the merger of masculine and feminine is a loss of the feminine. This may be an example of the well-known diachronic tendency that “the rich get richer”; the masculine is the largest gender in Old Norse (Trosterud 2006; Conzett 2007). As pointed out in connection with Table 2 above, there are varieties of Scandinavian that did not get the same gender reduction as Bergen. This includes dialects such as Oslo, cf. Table 2. I have to repeat the caveat from 1.2 above: contact does not always lead to gender reduction, and the entailment does not go both ways. Another Hanseatic centre in Eastern Norway, Tønsberg, illustrates this, since the dialect there retains three genders. When cities like Oslo and Tønsberg did not get gender reduction the way Bergen and Stockholm did, it seems reasonable to assume that a) contact does not necessarily have to trigger reduction (cf. 1.1), b) the historical/sociological facts on the ground were different (there were more Germans in Bergen, relatively speaking), c) the structural factors need not have been identical. On the other hand, the fact that reduction from three to two genders is found also in contact varieties in Northern Norwegian (cf. Bull et al. 1986;

given from Danish and German (including der Gehalt ‘the.MASC content’ vs. das Gehalt ‘the.NEUT salary’). Unfortunately, parallel examples can be found in Old Norse, such as the feminine ól ‘belt’ vs. the homonymous neuter ól ‘witch’. Thus, one may doubt whether the differences between three-gender systems and two-gender systems in North Germanic are quite as significant as Braunmüller (1999) takes them to be.

Gender and contact

183

Bull 1995) also makes it hard to believe that there is no relation between contact and gender reduction, as Jahr (2008) points out. The written standards of Danish and Swedish show the same kind of ‘simplification’, in that they only have two lexical genders. On the face of it, this fact is neither surprising nor interesting, since these varieties have a close relation to the languages of Copenhagen and Stockholm, respectively. On the other hand, there may be an additional reason which has more to do with the main theme of this paper: standard varieties and written languages normally imply “a high degree of dialect contact”, as observed by Szmrecsanyi and Kortmann (this volume). They would thus be likely to display signs of reduction and simplification.

3.

Problems with the focus on reduction

As section 2 has shown, it is perfectly legitimate to talk of reduction of the gender system in dialects in Bergen, Copenhagen and Stockholm, and in the written standards of Danish and Swedish, compared to Old Norse. In the following, we shall nevertheless look at some objections. They have to do with the syntagmatic, rather than paradigmatic, aspects of gender. The label ‘reduction’ is more appropriate for the paradigmatic aspects, which have been the focus of much previous research, than for the syntagmatic aspects. The distinction between syntagmatic and paradigmatic reduction is taken from Duke (2009). Firstly, while the definite singular is a “cell” in the Old Norse noun paradigm just as much as in the modern Norwegian one,9 it was much less frequent in Old Norse. It is more frequent and more central today. The indefinite singular without determiner is used more often in Old Norse than in Modern Scandinavian. Compare the examples in Table 3. For expository reasons, Swedish is used as an example in Tables 3 and 4, even if Swedish does not descend from Old Norse, strictly speaking. Like the Bergen dialect, standard Swedish has only two genders, it is a variety that has been exposed to much contact.

9 At least by standard views. Faarlund (2009) takes the Old Norse definiteness marker to be a clitic, not an affix. On such a view, the definite singular is not a member of the Old Norse noun paradigm.

184

Hans-Olav Enger

Table 3: The definite form has become more central. Old Norse

Swedish

Konung drap bóndi

Bonden dräpte konungen

king (ACC) killed farmer (NOM)

farmer (DEF.SG) killed king (DEF.SG)

‘(a/the) farmer killed (a/the) king’

‘the farmer killed the king’

In other words, Old Norse uses indefinite noun phrases in many cases where Modern Scandinavian has definite noun phrases. Table 3 shows that the Swedish example contains two definiteness suffixes that indicate gender, the corresponding Old Norse example contains none. It is possible to utter many Old Norse sentences without using the definiteness suffix that indicates gender; this is less frequently an option in Modern Scandinavian. In this respect, gender has not been reduced. If anything, it has become more important. Secondly, while the indefinite singular in Old Norse typically is used without an article, it is common in Modern Scandinavian to use it with an article. Compare the examples in Table 4. Table 4: “Naked” NPs are less common in Modern Scandinavian. Old Norse

Swedish

Bóndi sá mann. ‘(A/The) farmer saw (a/the) man.’

Bonden såg en man. ‘The farmer saw a man.’ Bonden såg mannen. ‘The farmer saw the man.’ En bonde såg en man. ‘A farmer saw a man.’

There are three possible modern Swedish translations of the Old Norse sentence in 4. The Old Norse example does not reveal gender of either NP straightaway; the Swedish examples contain the indefinite article en and the definiteness suffix -en; both indicate common gender (formally identical to the old masculine). So, the indefinite NP in Modern Scandinavian will reveal the gender of the head noun more often than does its Old Norse cognate. This is not reduction of gender, either. Gender has been reduced paradigmatically – Swedish behaves like Bergen in terms of paradigmatic gender reduction (cf. section 2) – but Tables 3 and 4 have shown that gender has not been reduced syntagmatically. In order to produce syntactically correct sentences, speakers have to use the gender of the

Gender and contact

185

noun more often in modern Swedish than in Old Norse. In this way, gender has become more central for the grammar; focusing on reduction should not blind us to this fact. We now turn to a third syntagmatically based objection against the focus on reduction. This objection has to do with pronouns. Consider Table 5. Table 5: Increase in pronominal gender – Old Norse and Swedish sentences compared. Old Norse Swedish

Ek sá konung. Hann … Jag såg kungen. Han …

‘I saw (the/a) king. He …’ ‘I saw the king. He …’

Old Norse Swedish

Ek sá hest. Hann … Jag såg hästen. Den …

‘I saw (a/the) horse. [He] …’ ‘I saw the horse. It …’

Old Norse Swedish

Ek sá drottning. Hon … Jag såg drottningen. Hon …

‘I saw (a/the) queen. She …’ ‘I saw the queen. She …’

Old Norse Swedish

Ek sá bók. Hon … Jag såg boken. Den …

‘I saw (a/the book). [She] …’ ‘I saw (a/the) book. It …’

Old Norse Swedish

Ek sá barn. þat … Jag såg barnet. Det …

‘I saw (a/the) child. It …’ ‘I saw (a/the) child. It …’

The upshot of this is that Old Norse has three genders in the pronouns, just as in the so-called determiners and possessives in Table 1. By contrast, Modern Swedish has four genders in the pronouns (han, hon, den, det), one more than Old Norse. That is not a reduction, it is an increase. This is not unique to Swedish. A new pronoun den has found its way also into three of the four written languages – Danish, Swedish and Norwegian Bokmål – and into the speech of many Scandinavians. The traditional three-gender system in the pronouns is retained in Norwegian Nynorsk10 and in the speech of many Norwegians (and traditional dialects also in Denmark and Sweden). There is a considerable overlap between the innovations in 1 and 5 (cf. section 4 below). In other words, many varieties that have had reduction from three to two lexical genders also have had their pronominal genders expanded from three to four (e.g. standard Danish). The two do not necessarily correlate, however. (The pronominal gender system in traditional Bergen dialect is different, for example.) 10 At least in the normative recommendations issued from the Norwegian language council (Språkrådet). In actual usage, there seems to be vacillation, presumably partly due to influence from Bokmål, but perhaps also for structural reasons (cf. Enger 2004a).

186

Hans-Olav Enger

In much traditional literature on Scandinavian, pronouns are not seen as manifestations of agreement and, accordingly, not as gender. That may be the reason for the focus on reduction; gender has been seen in predominantly paradigmatic terms. In a relatively recent introduction to morphology, for example (Lie 2006: 31), gender is presented as ‘main inflectional class’, and it is not mentioned that genders are established on a syntactic basis. In more recent linguistic literature, however, pronouns are usually taken to be agreement (cf. Lehmann 1982; Corbett 1991, 2006; Barlow 1999; Jobin 2004; Audring 2006). Admittedly, pronouns are not prototypical (or canonical) agreement, since they also have deictic functions. Nevertheless, Table 5 shows agreement classes of pronouns, i. e., gender. By this more recent understanding, the varieties found in the old Hanseatic cities – Bergen, Stockholm, Copenhagen – have not had gender reduction in every respect. Stockholm has one gender less in the determiner – and that can surely be called “reduction in contact” – but one gender more in the pronouns. So is the development in the pronouns as shown in Table 5 a counter-example against the idea of simplification in contact? In the next section, I shall suggest that the answer is “not quite”.

4.

Refining the idea of simplification in contact: referential versus lexical gender

As noted in section 2, genders are usually defined as classes of nouns, reflected in the behavior of associated words, i. e., as agreement classes. Yet one and the same noun can be reflected differently in the behavior of associated words in different contexts. Compare the following examples in Table 6 from Nynorsk, which has a fully-fledged three gender system, essentially of the Old Norse type: Table 6: Two kinds of gender: lexical vs. referential. Eg kjenner ein lege. ‘I know a (MASC) doctor.’ Han er liten ‘He is small (MASC).’ Den nye legen er liten. ‘The new doctor is small (MASC).’ Eg kjenner ein lege. ‘I know a (MASC) doctor’ Ho er lita. ‘She is small (FEM).’ Den nye legen er lita. ‘The new doctor is small (FEM).’

The noun lege can only be combined with the masculine determiner (article) ein. Combining the noun lege with the feminine determiner ei (*ei lege) is simply wrong. If the doctor is a man, we find examples of pro-

Gender and contact

187

nominalization such as legen – han ‘the doctor – he’. If the doctor is a woman, however, we find examples as legen – ho ‘the doctor – she’. In other words, the choice of pronominal gender is a real choice, and it depends on semantic and pragmatic factors. In these two respects, it differs from the choice of determiner, which is not a real choice, and depends on lexical/grammatical factors. The sort of gender agreement illustrated by legen – ho is called “referential” by Dahl (1999a) and “semantic” by Corbett (1991). The sort of gender agreement illustrated by ein lege is called “lexical” by Dahl and “syntactic” by Corbett. A third set of terms is ad sensum vs. ad formam. The differences in terminology need not concern us; the point is that there are two different kinds of gender agreement – even if they are related and the difference is a gradual one. We shall use the terms lexical and referential gender. A well known German equivalent would be this: das Mädchen ‘the.NEUT girl’ could be referred to using either es ‘it (NEUT)’ or sie ‘she (FEM)’. Using es is lexical gender, using sie is referential gender. In view of the data in Table 5 above, it seems sensible to suggest that the idea of gender reduction in contact should be restricted to lexical gender. The reasoning is very simple. Lexical gender is exactly that; it is lexical, which means that it can to some extent be irregular (see e.g. Dahl 1999a; Fraurud 1999).11 A central hypothesis in Natural Morphology as advocated by Wurzel (1984) is that diachronic change is reduction of irregularities, i. e., reduction of lexical marking (cf. section 1.3 above). Thus, if contact entails reduction of irregularity also in the field of gender, this should primarily hit lexical gender. The semantics of referential gender is more ‘transparent’. (In Scandinavian, it has to do primarily with animacy and sex). And indeed, this is exactly what has happened, as a comparison of Table 1 and Table 5 makes clear. While the use of the determiner in Scandinavian has to obey the lexical gender of the noun, the pronoun can be referential. This is in line with Corbett’s (1979, 1991) agreement hierarchy (cf. also Audring 2006). The semantics behind the pronominal genders in Swedish is rather transparent; the semantic rationale behind the pronominal gender system of Old Norse is arguably more ‘opaque’. It is not obvious, from the semantic point

11 Recent research on gender has, partly as an understandable reaction against the older idea that gender was irregular, perhaps over-emphasized the regularity of gender somewhat.

188

Hans-Olav Enger

of view, why queens should be grouped with books, the way they are in the Old Norse examples in Table 5. The Old Norse gender system is thus a system with much formal motivation, but it is also a system with complex interaction of semantic, morphological and phonological information (cf. e.g. Trosterud 2006; Conzett 2007). From the semantic point of view, the rules for pronominal agreement classes are more transparent in modern Swedish than in Old Norse as in Table 5. Thus, a numerical increase in classes does not necessarily imply an increased burden on the language learner.12 The increased number of pronominal gender classes has reinforced rather than obliterated the rationale behind the classes, and in gender systems, reinforcement often occurs in the pronouns (cf. Wurzel 1986 and Jobin 2004). It is thus not arbitrary that the change should happen here. In his version of Natural Morphology, Wurzel (1984) emphasizes motivation in terms of phonology and semantics: to the extent that a word’s grammatical properties are motivated by its phonology and semantics, they do not represent an extra burden on the memory – to know a word is, quite simply, to know its phonology and semantics.13 The new pronominal genders in Table 5 are strongly motivated by semantics and do not lead to any increase in lexical marking. Restricting the hypothesis of gender reduction in contact to lexical gender reduction is thus not an arbitrary move to escape from a potentially unpleasant counter-example; it is in line with a general approach. The idea that numerical increase does not equal increased complexity gets support from an example from the verb inflection: it is often said that on the way from Old Norse to modern Mainland Scandinavian, the morphology is reduced or simplified (cf. e.g. Delsing 2002). Yet an entirely new verb conjugation has come up in Norwegian and Swedish. This class can be referred to as ‘short verbs’, such as nå ‘reach’, and in Swedish it arises before 1500. Nobody has ever suggested that this is a counter-example to the putative reduction or simplification, however, presumably because the new conjugation has a clear motivation. In Natural Morphology terms, the motivation is extra-morphological; the class of short verbs does not com-

12 Corbett (1991: 317) observes the logical converse; numerical reduction of a gender system may make its semantics less transparent. 13 This idea is not unique to Natural Morphology; see, for example, Lyons (1977: ch. 13).

Gender and contact

189

pete with other classes (cf. Enger 2007: 293–294 for further discussion).14 The example supports the idea that numerical increase of classes does not necessarily equal increased complexity. Therefore, the increase in pronominal gender from Old Norse to Swedish (Table 5) does not constitute a problem for the contact-cum-simplification-hypothesis. On the contrary: if anything, it supports the hypothesis. The fact that the pronominal gender system expands can be linked to contact, and the semantic rationale is more obvious behind the Swedish pronouns than behind the Old Norse one. It is often assumed that the spreading of the fourth pronominal gender, den, has to do with contact, but the issue is debated (cf. e.g. Davidson 1999; Enger 2004a). On the basis of other languages, Aikhenvald (2004: 1036) observes that a “grammatical system of gender assignment can move towards a more ‘semantically oriented’ one”. Her example is “Cantabrian Spanish”. If we accept Braunmüller’s (1999) analysis of North Germanic gender (despite the small reservations in footnote 8), Scandinavian is another case in point.

5.

Genders and declensions

We now turn to Scandinavian plural declensions, which are interrelated with gender in an interesting way.15 One way of simplifying the relation between plural declension and gender is to make it more transparent, more 1:1. In Old Norse, lexical gender can be used to predict the plural declension (and vice versa), though not very reliably. If learners of Old Norse know that the noun hestr ‘horse’ is a masculine, they can infer that it will probably have the plural suffix -ar, which it does; if they know that the noun gestr ‘guest’ is a masculine, they can likewise infer that it will probably have the plural suffix -ar, which it does not (the correct indefinite plural is gestir). Compare

14 The members of the new class have a stem that ends with the stressed vowel; other verbs have one or more consonants after the stressed vowel in the stem. The members of the new class have no suffix in the infinitive (which is why they are called ‘short’), the ‘canonical verb’ in Norwegian does. Compare the ‘short verbs’ nå /no:/ ‘reach’, spå /spo:/ ‘foretell, prophesy’, fli /fˆi:/ ‘groom’ vs. the more prototypical danse /danse/ ‘dance’, bygge /byge/ ‘build’, kjøpe /çø:pe/ ‘buy’. 15 Genders and declensions are often related, not only in Scandinavian. See Corbett (1991), Aikhenvald (2004) and Wurzel (1986).

190

Hans-Olav Enger

Table 7, which also illustrates the diachronic development from Old Norse to one modern dialect. The important point is the plural forms that have been boldfaced in the modern Porsgrunn dialect of Norwegian. By regular phonological development, Porsgrunn should have had e.g. /jester/ and /ræimær/. But it does not. The two plural forms that Porsgrunn actually have here, /jestær/ and /ræimer/, show plural suffixes that are not predicted by sound law. The standard account is that the plural suffix /er/ was felt by speakers to be typical of feminine nouns, and, since speakers knew that /jest/ was not a feminine, but a masculine, they gave /jest/ the typical masculine suffix. Conversely, the noun /ræim/ ‘belt’ was felt to have a suffix atypical of feminine nouns, and it was given the more typical feminine suffix.16 Compare Table 7, where the more recent, analogical forms are boldfaced. Table 7: Improving the relation between gender and declension. Old Norse MASC

FEM

Porsgrunn dialect of Norwegian

NOM SG

NOM PL

SG

PL

bátr

bátar

boat-boats

bo:t

bo:tær

hestr

hestar

horse-horses

hest

hestær

gríss

grísir

pig-pigs

gri:s

gri:sær (e process (Kolsrud 1951: 96). The same holds for the dialect of Valdres. It has adjectival -a (Aasen 1965: 159), but it has otherwise the same amount of a>e as other Eastern dialects. Since these dialects indicate that feminine agreement on plural adjectives can be retained despite the standard phonological development of unstressed final vowels, the phonological change cannot be the sole cause of the demise of the feminine plural agreement. The argument so far indicates that an entirely phonological account of the loss of the feminine plural agreement is not a sufficient explanation. We may now go one step further. Why should exactly these two dialects have retained parts of the gender agreement in the plural of adjectives? If any Norwegian dialect is taken to be peripheral and isolated, Setesdalen is.

Gender and contact

195

Valdres is also rather isolated among Eastern Norwegian dialects. This is an indirect argument that contact has mattered in the loss of gender agreement on the plural of adjectives. Phonology is not all there is to the issue.20

7.

Summary of the main claims

The traditional idea that contact can lead to simplification receives support from lexical gender (section 2) and from declensions (section 5) in Norwegian. So, the expected reduction in contact varieties does indeed manifest itself also in Scandinavian gender. However, it is too simple to talk of the Scandinavian gender system in terms of reduction only (sections 3 and 4). Also in the gender system, there are cases of increase rather than reduction. Pronominal gender is not reduced; in fact, quite the opposite, especially in some of the ‘contact varieties’. In some other ways, gender has also become more central in Scandinavian (sections 3 and 5). Reduction along the paradigmatic axis is not everything, there is also a syntagmatic aspect to gender, which deserves more attention than it has been given. The idea of gender reduction in contact should be restricted to lexical gender, and the way the reduction works, has to do with structural conditions. ‘Numerical reduction’ and ‘simplification’ are not necessarily synonymous (section 4), and numerical increase is not necessarily synonymous with increased complexity. Within the framework of Natural Morphology, we can refine the main idea so as to keep its spirit: motivation is a better yardstick for simplicity than is numerical reduction. More generally, we

20 A more indirect and weaker argument in support of this view might go as follows. A priori, one might have found neutralization of gender in dialects anywhere in the core of the Scandinavian area. If ‘sound law’ was all there was to it, this should not be precluded. In other words: according to the hypothesis that ‘phonology is everything’, the geographical distribution is haphazard. Yet the two cases where we do find complete neutralization of gender also in the singular, is in Finland (Österbotten, cf. above) and Western Denmark (Western Jutland). These areas are on the outskirts of the area where Scandinavian is spoken, close to Finnish and German, respectively. We do not find gender neutralization in the singular in e.g. Southern Norway or Central Sweden, areas that traditionally have been less bilingual than these ‘fringe areas’; it is hard to believe that this geographical distribution is an accident only (compare ‘Boretzky’s principle’ in section 2 and Kusmenko 2000).

196

Hans-Olav Enger

should not abandon the idea that there may be some link between social factors and linguistic ones.

Abbreviations 1 = first person; ACC = accusative; FEM = feminine; INDEF = indefinite; MASC = masculine; NEUT = neuter; NOM = nominative; PL = plural; POSS = possessive; SG = singular.

Gender and contact

197

References Aasen, Ivar 1965/1864 Norsk Grammatik [Norwegian Grammar]. Oslo: Universitetsforlaget. Aikhenvald, Alexandra 2004 Gender and noun class. In: Geert E. Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds.), Morphologie/Morphology: Ein internationales Handbuch zum Flexion und Wortbildung, 2. Halbband, 1031–1045. Berlin: Mouton de Gruyter. Andersen, Henning 1988 Center and periphery: Adoption, diffusion and spread. In: Jacek Fisiak (ed.), Historical Dialectology, 39–83. Berlin: Mouton de Gruyter. Andersen, Henning 2005 On the handbook of historical linguistics. Diachronica 22: 155–176. Audring, Jenny 2006 Pronominal gender in spoken Dutch. Journal of Germanic Linguistics 18: 85–116. Audring, Jenny 2010 Deflexion und pronominales Genus. In: Damaris Nübling, Antje Dammel and Sebastian Kürschner (eds.) 693–718, Kontrastive Germanistische Linguistik. Hildesheim: Olms. Barðdal, Jóhanna 2009 The development of case in Germanic. In: Jóhanna Barðdal and Shobhana L. Chelliah (eds.), The Role of Semantic, Pragmatic and Discourse Factors in the Development of Case, 123–159. Amsterdam: Benjamins. Barlow, Michael 1999 Agreement as a discourse phenomenon. Folia Linguistica XXXIII: 187–211. Bauer, Laurie and Jennifer Hay 2007 Phoneme inventory and population size. Language 83: 388–401. Bechert, Johannes 1982 Grammatical gender in Europe: An areal study of a linguistic category. Papiere zur Linguistik 26: 23–34. Boretzky, Norbert 2004 Morphologische Entlehnung und Lehnübersetzung. In: Geert E. Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds.), Morphologie/Morphology: Ein internationales Handbuch zum Flexion und Wortbildung, 2. Halbband, 1644–1653. Berlin: Mouton de Gruyter. Braunmüller, Kurt 1984 Morphologische Undurchsichtigkeit – ein Charakteristikum kleiner

198

Hans-Olav Enger

Sprachen. Kopenhagener Beiträge zur Germanistischen Linguistik 22: 48–68. Braunmüller, Kurt 1999 Gender in North Germanic: A diasystematic and functional approach. In: Barbara Unterbeck and Matti Rissanen (eds.), Gender in Grammar and Cognition, 25–55. Berlin: Mouton de Gruyter. Bull, Tove, Jorid Hjulstad Junttila and Aud-Kirsti Pedersen 1986 Nominalfrasen i skibotnmålet [The NP in the Skibotn dialect]. Norsk Lingvistisk Tidsskrift 4: 60–72. Bull, Tove 1995 Language contact leading to language change: The case of Northern Norway. In: Jacek Fisiak (ed.), Linguistic Change under Contact Conditions, 15–35. Berlin: Mouton de Gruyter. Carstairs-McCarthy, Andrew 1992 Current Morphology. London: Routledge. Christoffersen, Marit 1990 Kristiansand. In: Ernst Håkon Jahr (ed.), Den store dialektboka [The Big Dialect Book], 187–193. Oslo: Novus. Conzett, Philipp 2007 Genus i norrønt og nynorsk [Gender in Old Norse and Nynorsk]. MA thesis. Tromsø: University of Tromsø. Corbett, Greville G. 1979 The agreement hierarchy. Journal of Linguistics 15: 203–224. Corbett, Greville G. 1991 Gender. Cambridge: Cambridge University Press. Corbett, Greville G. 2006 Agreement. Cambridge: Cambridge University Press. Dahl, Östen 1999a Animacy and semantic gender. In: Barbara Unterbeck and Matti Rissanen (eds.), Gender in Grammar and Cognition, 99–117. Berlin: Mouton de Gruyter. Dahl, Östen 1999b Elementary gender distinctions. In: Barbara Unterbeck and Matti Rissanen (eds.), Gender in Grammar and Cognition, 577–595. Berlin: Mouton de Gruyter. Dalen, Arnold 1990 Talemålet i Trondheim [The dialect in Trondheim]. In: Ernst Håkon Jahr (ed.), Den store dialektboka [The Big Dialect Book], 205–211. Oslo: Novus. Davidson, Herbert 1999 Hanhonden: Genusutvecklingen i svenskan under nysvensk tid [Hesheit: The development of gender in Swedish after 1500]. Lund: Lund University Press. Delsing, Lars-Olof 2002 The morphology of Old Nordic II: Old Swedish and Old Danish. In:

Gender and contact

199

Oskar Bandle, Kurt Braunmüller, Ernst Håkon Jahr, Allan Karker, Hans-Peter Naumann and Ulf Teleman (eds.), The Nordic Languages, 925–940. Berlin: Mouton de Gruyter. Duke, Janet 2009 Gender reduction and loss in Germanic: Three case studies. In: Damaris Nübling, Antje Dammel and Sebastian Kürschner (eds.) to appear. Kontrastive Germanistische Linguistik. Hildesheim: Olms. Endresen, Rolf Theil 1990 Vikværsk [The Viken Dialect]. In: Ernst Håkon Jahr (ed.), Den store dialektboka [The Big Dialect Book], 89–101. Oslo: Novus. Enger, Hans-Olav 1999 Markedness conflicts: Sociolinguistic speculations. Sprachtypologie und Universalienforschung 52: 380–396. Enger, Hans-Olav 2004a Tre endringer i det skandinaviske genussystemet [Three Changes in the Scandinavian Gender System]. Arkiv för nordisk filologi 105: 125–146. Enger, Hans-Olav 2004b On the relation between gender and declension: A diachronic perspective from Norwegian. Studies in Language 28: 51–82. Enger, Hans-Olav 2007 The no blur principle meets Norwegian dialects. Studia Linguistica 61: 278–309. Faarlund, Jan Terje 2009 On the history of definiteness marking in Scandinavian. Journal of Linguistics 45, 617–639. Fiva, Toril 1990 Bodødialekten [The Bodø dialect]. In: Ernst Håkon Jahr (ed.), Den store dialektboka [The Big Dialect Book], 211–217. Oslo: Novus. Fraurud, Kari 1999 Proper names and gender in Swedish. In: Barbara Unterbeck and Matti Rissanen (eds.), Gender in Grammar and Cognition, 167–221. Berlin: Mouton de Gruyter. Gaeta, Livio 2007 Is analogy economic? In: Fabio Montermini, Gilles Boyé and Nabil Hathout (eds.): Selected Proceedings of the 5th Decembrettes: Morphology in Toulouse, 20–33. Somerville, MA: Cascadilla. Hale, Mark 2003 Neogrammarian sound change. In: Brian Joseph and Richard D. Janda (eds.), Handbook of Historical Linguistics, 343–368. Oxford: Blackwell. Haspelmath, Martin 2006 Against markedness (and what to replace it with). Journal of Linguistics 42: 25–70.

200

Hans-Olav Enger

Heide, Eldar 2003 Felleskjønnet i bergensk [The common gender in the Bergen dialect]. Maal og Minne: 23–35. Hudson, Richard 1996 Sociolinguistics. Cambridge: Cambridge University Press. Huldén, Lars 1973 Genussystemet i Karleby och Nedervetil [The gender system in Karleby and Nedervetil]. Folkmålsstudier 22: 47–82. Jahr, Ernst Håkon 2008 Historical sociolinguistics: The role of Low German language contact. In: Geirr Wiggen, Tove Bull and Marit Christoffersen (eds.), Språkhistorie og språkkontakt/Language History and Language Contact [Festschrift], 246–258. Oslo: Novus. Jobin, Bettina 2004 Genus im Wandel. Studien zu Genus und Animatizität anhand von Personenbezeichnungen im heutigen Deutsch mit Kontrastierungen zum Schwedischen. Stockholm: Almqvist and Wiksell. Knudsen, Trygve 1967 Kasuslære, bind I [Case, Volume 1]. Oslo: Universitetsforlaget. Koefoed, Geert and Jaap van Marle 2004 Fundamental concepts. In: Geert E. Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds.), Morphologie/Morphology: Ein internationales Handbuch zur Flexion und Wortbildung, 1574–1590. Berlin: Mouton de Gruyter. Kolsrud, Sigurd 1954 Nynorsken i sine målføre [The dialects of New Norwegian]. Oslo: Universitetsforlaget. Kusmenko, Jurij K. 2000 Entwicklung des Genussystems in den skandinavischen Sprachen. In: Fritz Paul (ed.), Arbeiten zur Skandinavistik. 13. Arbeitstagung der deutschsprachigen Skandinavistik, 469–479. Frankfurt a.M: Peter Lang Verlag. Kusters, Wouter 2000 Morphological simplification: More than erosion. In: Dicky Gilbers, John Nerbonne and Jos Schaeken (eds.), Languages in Contact, 225–230. Amsterdam: Rodopi. Kusters, Wouter 2003 Linguistic Complexity: The Influence of Social Change on Verbal Inflection (LOT 77). Leiden: LOT (Netherlands Graduate School of Linguistics). Kusters, Wouter 2008 Complexity in linguistic theory, language learning and language change. In: Matti Miestamo, Kaius Sinnemäki and Fred Karlsson (eds.), Language Complexity: Typology, Contact, Change (SCLS 94), 3–22. Amsterdam: Benjamins.

Gender and contact

201

Lagman, Herbert 1971 Svensk-estnisk språkkontakt: Studier över estniskans inflytande på de estlandssvenska dialekterna [Swedish-Estonian Language Contact: Studies in the Estonian influence on the Swedish Dialects of Estonia] (Acta Universitatis Stockholmiensis, Stockholm Studies in Scandinavian Philology, New Series 8). Stockholm: Almqvist and Wiksell. Langacker, Ronald W. 1987 Foundations of Cognitive Grammar: Theoretical Prerequisites. Stanford: Stanford University Press. Lehmann, Christian 1982 Universal and typological aspects of agreement. In: Hansjakob Seiler and Fritz Stachowiak (eds.), Apprehension, Teil II, 201–268. Tübingen: Gunter Narr. Lie, Svein 1990 Indre Østlandet. In: Ernst Håkon Jahr (ed.), Den store dialektboka [The Big Dialect Book], 101–116. Oslo: Novus. Lie, Svein 2006 Norsk morfologi [Norwegian Morphology]. Oslo: Novus. Lyons, John 1977 Semantics. Cambridge: Cambridge University Press. McWhorter, John 2001 The world’s simplest grammars are creole grammars. Linguistic Typology 5: 125–166. McWhorter, John 2008 Why does a language undress? In: Matti Miestamo, Kaius Sinnemäki and Fred Karlsson (eds.), Language Complexity: Typology, Contact, Change (SCLS 94), 168–190. Amsterdam: Benjamins. Muysken, Pieter 2004 Pidginization, creolization and language death. In: Geert E. Booij, Christian Lehmann, Joachim Mugdan and Stavros Skopeteas (eds.), Morphologie/Morphology: Ein internationales Handbuch zum Flexion und Wortbildung, 1653–1661. Berlin: Mouton de Gruyter. Nesse, Agnete 2002 Språkkontakt og tysk i hansatidens Bergen [Language Contact and German in Bergen in the Hanseatic Age]. Oslo: Novus. Nesse, Agnete 2005 Boken – han og kua – den: om endringer i norske genussystem [The book – he and the cow – it: on changes in Norwegian gender systems]. Maal og Minne: 136–146. Nettle, David 1999 Is the rate of linguistic change constant? Lingua 108: 119–136. Nichols, Johanna 2003 Diversity and stability in language. In: Richard Janda and Brian

202

Hans-Olav Enger

Joseph (eds.), Handbook of Historical Linguistics, 283–311. Oxford: Blackwell. Parkvall, Mikael 2008 The simplicity of creoles in a cross-linguistic perspective. In: Matti Miestamo, Kaius Sinnemäki and Fred Karlsson (eds.), Language Complexity: Typology, Contact, Change (SCLS 94), 265–285. Amsterdam: Benjamins. Pedersen, Karen Margrethe 1999 Genusforenklingen i københavnsk [The Gender Simplification in the Copenhagen dialect]. Danske Folkemål 41: 79–105. Plank, Frans 1991 Of abundance and scantiness in inflection. In: Frans Plank (ed.), Paradigms: The Economy of Inflection, 1–39. Berlin: Mouton de Gruyter. Priestley, Tom M. S. 1983 On ‘drift’ in Indo-European gender systems. Journal of Indo-European Studies 11: 339–363. Rabb, Viveca 2007 Genuskongruens på reträtt [Gender agreement retreating]. Dissertation, Åbo Akademi. Sandøy, Helge 1998 The diffusion of a New Morphology in Norwegian dialects. Folia Linguistica XXXII: 83–100. Simonsen, Hanne Gram, Inger Moen and Steve Cowen 2000 Retroflex consonants in Norwegian: Are they really? Evidence from EMA and EPG. In: The 5th Seminar on Speech Production, Models and Data, 113–116. Kloster Seeon. Skjekkeland, Martin 2005 Dialektar i Norge [Dialects in Norway]. Kristiansand: Høyskoleforlaget. Stolz, Thomas 1987 Natürlichkeit und Interferenz: Fallbeispiele aus dem morphologischen Bereich in Sprachkontaktsituationen [Naturalness and Interference: Case Studies from Morphology in Contact Situations]. Papiere zur Linguistik 36: 17–44. Szmrecsanyi, Benedikt and Bernd Kortmann 2009 Between simplification and complexification: Non-standard varieties of English around the world. In: Sampson, Geoffrey, David Gil and Peter Trudgill (eds.), Language Complexity as an Evolving Variable, 64–79. Oxford: Oxford University Press. Szmrecsanyi, Benedikt and Bernd Kortmann (this volume) Parameters of morphosyntactic variation in World Englishes: prospects and limitations of searching for universals.

Gender and contact

203

Thomason, Sarah Grey and Terence Kaufman 1988 Language Contact, Creolization, and Genetic Linguistics. Berkeley, California: University of California Press. Torp, Arne 1982 Norsk og nordisk før og nå [Norwegian and North Germanic past and present]. Oslo: Universitetsforlaget. Trosterud, Trond 2006 Gender assignment in Old Norse. Lingua 116: 1441–1463. Trudgill, Peter 1986 Dialects in Contact. Oxford: Blackwell. Trudgill, Peter 1992 Dialect typology and social structure. In: Ernst Håkon Jahr (ed.), Language Contact: Theoretical and Empirical Studies (TiL SaM 60), 195–212. Berlin: Mouton de Gruyter. Vennemann, Theo 1993 Language change as language improvement. In: Charles Jones (ed.), Historical Linguistics, 319–345. London: Longman. Vikør, Lars 1993 The Nordic Languages. Oslo: Novus. Wessén, Elias 1969 Svensk språkhistoria 1 [Swedish Language History I]. Stockholm: Almqvist and Wiksell. Wetås, Åse 2008 Kasusbortfallet i mellomnorsk [The loss of case in Middle Norwegian]. Dissertation, University of Oslo. Wurzel, Wolfgang Ullrich 1984 Flexionsmorphologie und Natürlichkeit. Berlin: Akademie-Verlag. Wurzel, Wolfgang Ullrich 1986 Die wiederholte Klassifikation von Substantiven. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 39: 79–96. Wurzel, Wolfgang Ullrich 2000 Morphological properties in the lexicon: Diachronic evidence. Acta Linguistica Hungarica 47: 345–356.

Universals of structural borrowing Yaron Matras Abstract The chapter is devoted to the questions whether certain categories are more easily borrowed than others, whether there are inherent semanticpragmatic functions that make some categories more borrowable than others, and what this tells us about the motivations behind borrowing. Rather than view borrowing as a transfer of structure from one system into another, I view it as the removal of an invisible demarcation line that separates subsets within the linguistic repertoire (or the speaker’s ‘languages’). I take the perspective of language processing by speakers in multilingual settings, and argue that there is a functional advantage to compromising this mental demarcation line. Based on recent sampling of languages in contact, I review a set of borrowing hierarchies and identify a number of motivations that prompt speakers to merge the representation of categories across their bilingual repertoire, leading to identical cross-system representation of those categories, or ‘structural borrowing’. Keywords: contact, contrast, borrowing, discourse, gaps, hierarchies, interaction, language processing, prestige, presupposition

1.

Introduction

The view that there are no absolute constraints on structural borrowing (cf. Campbell 1993) is probably not widely contested. Nonetheless, many authors have proposed that borrowing is not entirely arbitrary, either, and that some factors facilitate, while others may impede the transfer of wordforms, morphemes, and even structural organization patterns from one language to another. Generalizations have often tended to focus on the status of word classes or grammatical categories. Thomason and Kaufman’s (1988) frequently cited borrowing scale is an example of an attempt at a generalization that says, essentially, that some word classes or types of structure are more easily borrowed than others. Ease of borrowing is interpreted as the likelihood that borrowing will occur earlier in the history of contact and

Universals of structural borrowing

205

hence that it will require less intensive contact, and that borrowing of the relevant category will therefore be attested more frequently in contact situations cross-linguistically. Thomason and Kaufman do not, however, offer an explanation as to why and how word class affiliation facilitates or inhibits borrowing. More explicit suggestions of factors that facilitate borrowing have listed structural autonomy and semantic transparency (Moravcsik 1978; Field 2002; Johanson 2002), low paradigmaticity (van Hout and Muysken 1994), operation at the level of a higher or broad unit of speech (discourse > paragraph > sentence > phrase > word; Stolz and Stolz 1996; Ross 2001), and ‘pragmatic detachability’ or automatized, gesture-like interventions that are subject to more routine and less analytic reflection (Matras 1998). Most of these observations were based on individual case studies. Ross (2001) for example discusses a contact situation involving two languages in Melanesia, Stolz and Stolz (1996) discuss a sample of diverse Mesoamerican languages in contact with Spanish, and Matras (1998) discusses various dialects of Romani in contact with diverse European languages, as well as a small sample of diverse languages under the historical sphere of influence of Arabic. The purpose of the present contribution is to review a set of borrowing hierarchies that are based on two samples of a somewhat larger scale. The first is the extended sample of some seventy-five dialects of Romani discussed in Elšík and Matras (2006). With an assumed time-depth separation of up to 500–600 years (see Matras 2002), these dialects represent a similar recipient language, spoken under more-or-less uniform sociolinguistic conditions, in contact with a sample of up to twenty-five different European donor languages, among them Basque, Finnish, Welsh, Hungarian, Greek, Romanian, and Turkish. The second is a comprehensive crosslinguistic survey of grammatical borrowing in the form of individual contributions to a volume by Sakel and Matras (2007), each focusing on a particular language in contact.1 General tendencies in the sample were evaluated in Matras (2007). 1 Tasawaq (M. Kossmann), K’abeena (J. Crass), Likpe (F. Ameka), Katanga Swahili (V. de Rooij), Khuzistani Arabic (M. Shabibi), Domari (Y. Matras), Kurmanji (G. Haig), NE and Western Neo-Aramaic (G. Khan, W. Arnold), Macedonian Turkish (ù. Tufan), Kildin Sami (M. Rießler), Yiddish (G. Reershemius), Rumungro (V. Elšík), Manange (K. Hildebrandt), Indonesian (U. Tadmor), Biak (W. van den Heuvel), Vietnamese (M. Alves), Jaminjung (E. Schultze-Berndt), Rapanui (s. Fischer), Nahuatl (A. Jensen), Yaqui (Z. Estrada and L. Guerrero), Otomi (E. Hekking and D. Bakker), Purepecha (C. Chamoreau),

206

Yaron Matras

The prototypical conditions of contact as represented by these samples include prolonged use of a pair of languages – a donor and a recipient language – in the same society, and at least some degree of bilingualism. In the overwhelming number of cases, speakers of the recipient or borrowing language – the language under investigation – are almost all bilingual and have been bilingual for more than one generation, and bilingualism tends to be unidirectional, that is, speakers of the donor language do not normally or invariably learn the recipient language. In most cases, there is some degree of domain separation between the languages, each language being used by speakers for different purposes in different settings. The diglossic conditions are not uniform, however. Some languages, such as Yiddish and Maltese, have a written tradition and are used in institutions. Others, like Indonesian and Vietnamese, even enjoy the status of principal state languages. A number of contact settings fall outside the scope of the present investigation. Languages that are formed as a means of communication in response to a contact situation, namely pidgins and Mixed Languages, are not taken into consideration in the sample. This is because the conditions of contact are radically different: there is no stability of language use in particular domains. Instead, pidgins show makeshift communication while Mixed Languages show deliberate language distortion or at least conscious mixing. While extra-linguistic variables have been taken into account in the evaluation, the principal framework of comparison involved an assessment of the contact behavior of different functional categories. ‘Category’ represents a functional notion. Categories are understood here to be operational devices that trigger mental processing activities in communicative interaction: Nouns name objects, interjections direct attention to emotive evaluations of the speech situation, connectors establish links between the processing of individual propositions, word order serves as a map to organize information at the utterance level, and so on. Their concrete representation in a given language is through a structural form, which may or may not have cross-linguistic equivalents. The challenge of describing universals of borrowing is to identify general trends in the sample, which may then serve as predictions for the borrowing behavior in languages in general. A further challenge is to explain

Quechua and Guarani (J. Rendón), Hup (P. Epps), Moseten (J. Sakel), as well as Wichi (A. Vidal), Maltese and Chamorro (T. Stolz).

Universals of structural borrowing

207

these generalizations. From a functional perspective, explanations must be sought in terms of the relations between the function that defines a category, and the process that is defined as ‘borrowing’. Seen through the prism of the individual who is communicating in a multilingual setting, i. e. the point of initiation of any contact-induced language change, borrowing means replication of a form or structure that is normally reserved for communication in language L-X (that is, in the set of interaction contexts that are characterized by the use of set X from within the repertoire of linguistic structures; see Matras 2009), while communicating in language L-Y (that is, in an interaction context that is not identified as part of set X, but as set Y). This characterizes the process of borrowing irrespective of the social circumstances that give rise to bilingualism, that is, whether communication in a second language that is being acquired is influenced by the structures of the first language, or whether communication in the native language is being re-shaped as a result of the influence of a second language. What interests us is that in both cases speakers are in some way or other reluctant or unable to maintain a strict separation line within their overall repertoire of linguistic structures. The phenomena described by Thomason and Kaufman (1988) as “shift-induced interference” (substrate influence) and as “borrowing” in the stricter sense (so-called prestige-related influence) are treated here as essentially one and the same: both derive from a bilingual speaker’s motivation to compromise language-separation for a given structure, and to employ that structure irrespective of context-based language choice. The explanatory model of borrowing that is pursued here seeks to account for the connection between category – the triggering of a particular linguistic-mental processing activity – and the failure to keep apart subsets of the repertoire of linguistic structures when communicating in different settings or contexts. The questions that we ask are these: 1. Are certain categories more easily borrowed than others? 2. What are the inherent functions that make some categories more borrowable? 3. What can we learn from this about the motivations behind borrowing? A note is in order concerning exceptions to postulated hierarchies. Counterexamples have sometimes been offered in the literature in order to prove that postulated generalizations are wrong. But it would seem counter-productive to ignore tendencies that are followed by a substantial group of languages within a sample, merely because they may not be followed by all languages or because they might be contradicted in one or

208

Yaron Matras

two instances. Where there appears to be a motivation behind trends, one that is beyond pure coincidence, then these trends deserve our attention. Quite often, it is the counterexample that can be explained as resulting from a local, language-particular constraint that impedes the realization of common patterns in a particular instance. Our interest is thus in describing and explaining meaningful trends, not in discovering laws that are on a par with the physical laws of nature.

2.

Structural factors that facilitate borrowing

The borrowing hierarchies depicted under (1) have been suggested in the literature (cf. Moravcsik 1978; Field 2002), and are confirmed by our own two samples as well: (1)

Structural factors facilitating borrowing a. nouns > non-nouns, function words b. free morphemes > bound morphemes c. derivational morphology > inflectional morphology d. agglutinating affix > fusional affix

The factors under (1) can be summarized as pertaining to the structural autonomy and the semantic transparency of morphemes. Elements that occupy a higher position on the scale are more likely to occur in speech as autonomous or relatively autonomous entities with a consistent meaning (cf. Johanson 2002). Those that are lower on the scale tend to have a meaning that is more abstract or more contextually bound and are more likely to accompany other parts of speech than to occur on their own. Structural autonomy as represented by the higher-ranking values under (1) suggests that the meaning of the item is less dependent on a particular environment. This facilitates the replication of an item in a different structural environment. Semantic and structural autonomy also allow greater pragmatic independence, which in turn facilitates the embedding of higher-ranking items into speech acts independently of the selected language of the interaction. The higher borrowing likelihood of nouns is sometimes regarded as a purely structural constraint, triggered by the tendency of verbs toward greater morphological complexity. In fact, nouns are more numerous in the lexicon, due to the need to label a greater number of concepts and physical objects compared to activities, states and events; cross-cultural

Universals of structural borrowing

209

contact and the resulting expansion of communicative interaction domains arguably create a greater need to enrich the nominal lexicon.2 The predictions made under (1) are confirmed by the samples in several ways. First, higher-ranking items are more likely to occur in the corpus as a whole and in any given language, than lower-ranking items. Second, the hierarchies are implicational, in that the occurrence of a lower-ranking item in any given language implies the occurrence of a higher-ranking item in that same language (i. e. if a language borrows inflectional morphology, it also borrows derivational morphology). Interpreted against the background of the actual mechanism of borrowing as described in the previous section, this provides us with an insight into speakers’ confidence to carry-over linguistic structures that are associated primarily with a set X of interaction contexts into interaction contexts of set Y, or in other words: their confidence to employ structures from their linguistic repertoire irrespective of contextual constraints. This confidence grows gradually in accordance with the ease with which a structural form can continue to be associated with its referential meaning even outside its original structural environment and usage context.

3.

Motivations for borrowing

In the previous section I interpreted factors that facilitate borrowing as factors that license bilingual speakers to generalize a particular structure or form across their linguistic repertoire and to use it irrespective of the constraints on selection of structures imposed by the context or setting of the interaction. In this section I discuss several groups of borrowing hierarchies that provide us with insights not just into the factors that facilitate borrowing, but into those that motivate and promote it in the first place. Recall once again that borrowing is defined here (following Matras 2009) as the generalization of an item within the multilingual linguistic repertoire, and hence the partial or complete removal of interaction-setbased constraints on its occurrence. When addressing the question of motivation for borrowing, we are thus seeking an answer to the question:

2 The integration of verbs may itself require additional, specialised morphology, due to the cognitive complexity of the verb as both a label for an activity and as the anchor of the predication in the utterance. Borrowed verbs often require a morphological extension indicating their ‘verbness’ (cf. Matras 2007a, 2008; see also Wichman and Wohlgemuth 2008; Muysken 2000).

210

Yaron Matras

when are bilingual speakers more likely to generalize a single item across the linguistic repertoire in such a fashion, and why? The when part of the question translates into our interest in the behavior of different grammatical categories, and thus it means essentially ‘in pursuit of which linguistic-mental processing tasks’. The why part suggests that there is, under certain circumstances, a functional advantage to compromising the mental demarcation line that separates subsets within the linguistic repertoire. A trivial example is the case of so-called ‘cultural loans’: e. g. lexical content-words that denote an object or concept hitherto unfamiliar to the receiving society, terminology related to institutions that are the property of the neighboring culture, and so on. The commonplace interpretation of borrowing here is that such terms fill lexical gaps in the recipient language. But semantic gaps can also be filled by neologisms or meaning extensions. Borrowing of word-forms, we might hypothesize, is motivated by the treatment of the respective entities as unique referents which cannot simply be re-labeled; replication of the original label allows activation of the world of associations that accompanies the term in its original interaction context. The outcome is a single structure that represents the relevant concept in the entire repertoire, irrespective of language or choice or repertoire subset in a particular interaction context. Grammatical operations pose a greater challenge. While some studies have opted to make use of the term ‘gap’ to refer to cases where languages in contact possess different structural means for expressing similar relations, it is clear that this particular notion of ‘gap’ is very different from the former. Here, it is not the absence of the abstract relation or conceptualization – for example contrast or repetition – that motivates borrowing, but a motivation to employ the same or similar structural means throughout the repertoire when expressing these relations. This brings us once again to the question of when and why. Following up on the argument developed in Matras (1998) and expanded in Matras (2009), I suggest that around certain types of operations, the need to constantly select among parallel structures overburdens the speaker and serves as a potential trigger for malfunctions of the processing mechanism that guides the speaker to select the form from the ‘correct’ (or active) language and to inhibit selection of the form from the ‘incorrect’ language (cf. Green 1998). Malfunctions thus result in the production of a form with the appropriate meaning, but from the non-appropriate language. To be sure, not all bilingual speech production errors of this kind are likely to lead to language change. But in settings in which there is a) widespread bilingualism, such that malfunctions do not actually hinder com-

Universals of structural borrowing

211

munication, and b) lax normative control over speech, usually owing to the absence of institutionalized language norms and possibly also related to a lower linguistic self-esteem, tolerance, in the first instance, of errors of this kind, and their eventual propagation can lead to the substitution of inherited forms by forms from the contact language – or ‘borrowing’. Traditionally, speakers’ (cognitive) inability to keep apart structures from their two languages – linguistic interference – has been recognized primarily with respect to influence of L1 on communication in L2, while instances of L2 insertions into L1 have been regarded as code switching or code mixing. I argue here, as in earlier work (Matras 1998, 2000), that there is a type of interference that leads to L2 influence on L1 in the form of non-voluntary L2 insertions, which result from lapses in control over the language selection and inhibition mechanism. The issue of control over word-form selection is only one of several functional factors that appear to be of relevance in motivating and facilitating borrowing. In the next few sections I review some of these factors, toward an integrated picture of the complexity of motivations that are involved in the process of borrowing.

3.1.

Domain specialization and non-routine tasks

A traditional, perhaps even ‘classic’ explanation for the motivation behind borrowing is prestige: it is assumed that the use of word-forms from a language that is associated with cultural and technological progress or simply with social and economic power – dubbed the ‘dominant’ language – serves to signal the speaker’s competence in this language, which in turn is associated with upwards social mobility. Speakers will, according to this assumption, integrate word-forms from the dominant language into their speech in the language of the socially weaker group as a token of higher social standing, power, or competence. The key to understanding this kind of relationship between languages is the diglossic setup or patterns of domain-specialization of languages in the community. A tension potential emerges between the context-specialization of the dominant or donor language and the association of certain activity domains with that language, and routine tasks that support the retention of old, inherited forms in the in-group or recipient language. In this kind of relationship, items that are less subject to routine employment prove more susceptible to new fashions and are more likely to be represented by borrowings that serve as tokens for social accommodation. On other hand, those that are more tightly embedded into routine interaction

212

Yaron Matras

are protected by the routine, by the intimacy of their meaning, and apparently also by their frequency of occurrence. As illustrated by the kinship term systems of both English and Maltese (Figures 1 and 2), concepts representing closest kin are less likely to be borrowed. Those representing somewhat more remote kin tend to be of Romance origin, both languages having been influenced at some stage during their history by contact with a dominant Romance-speaking society. Figure 1: Close kin and remote kin in English (shaded cells represent borrowed terms). grand-(parents) aunt

mother

cousin

sister

niece

father brother

EGO

daughter

uncle cousin

son

nephew

grand-(children)

Figure 2: Close kin and remote kin in Maltese (shaded cells represent borrowed terms) (data from Borg and Azzopardi-Alexander 1997: 339–340). nanna ‘grandmother’ zija ‘aunt‘

nannu ‘grandfather’

omm ‘mother’

kuƥina ‘cousin’ oƫt ‘sister’ neputija ‘niece’ bint ‘daughter‘

missier ‘father’ EGO

ziju ‘uncle‘

ƫu- ‘brother’ kuƥin ‘cousin’ (i)bn- ‘son‘

neputi ‘nephew’

neputija ‘granddaughter’ neputi ‘grandson’

A further domain in which hierarchical splits may be connected to the context-related roles of the languages involved is the borrowing of numerals. Many languages borrow numerals strictly in contexts involving economic activity and institutional interaction, where the donor language is the language associated with economic or institutional interaction domains. The Betawi language of Indonesia, for example, borrows a complete set of numerals from Hokkien Chinese, but these are used only when referring to sums of money (Tadmor 2004). Chinese numerals are used in formal contexts in Vietnamese (Alves 2007). Romani speakers generally use numerals from the majority language when citing dates, something that is performed primarily in the context of official institutions.

Universals of structural borrowing

213

The diglossic relationship between the languages may also be reflected in the hierarchy of borrowing of individual numerals. Lower numerals are generally protected through their association with routine counting tasks in more familiar and everyday environments, and are less likely to be borrowed. Higher numerals by contrast are more likely to be associated with more complex calculations in more formal transactions, and are more likely to be borrowed, though the numerals ‘ten’ and ‘one hundred’ tend to be more stable as salient quantities. The Pharas dialect of Asia Minor Greek borrows the Turkish terms for ‘seventy’, ‘eighty’, and ‘ninety’ (Dawkins 1916: 171). Some languages, such as Moseten, Quichua, and Nahuatl, borrow only numerals above ‘ten’. Languages that borrow numerals between ‘five’ and ‘ten’ include Swahili, which has sita, saba, nane, and tisa (‘six-nine’) from Arabic, and Romani, which has efta, oxto, enja (‘seven-nine’) from Greek; both also borrow numerals above ‘twenty’. Other languages that borrow numerals above ‘five’ are Domari, Tasawaq, Otomi, Guarani, Purepecha, Yaqui, and Kildin Saami. Borrowing of numerals under ‘five’ is attested in Jaminjung, which has no native numerals above ‘three’ (Schultze-Berndt 2007), and in other Australian languages with similar, very basic and arguably non-numerical quantification systems. Chamorro has replaced its entire earlier system of numerals, which was still attested in the early twentieth century, with Spanish loans (Topping 1973: 166–167). Thai uses Chinese numerals above ‘two’, and Khmer uses Thai numerals (of Chinese origin). The most commonly borrowed numerals represent abstract figures. Standard Indonesian borrows the numerals laksa ‘ten thousand’ and juta ‘million’ from Sanskrit, and milyar ‘billion’ and nol ‘zero’ from Dutch (Tadmor 2004, 2007). Tasawaq borrows its lower numerals from Arabic, but its word for ‘hundred’ from Tuareg (Kossmann 2007); Romani in central Europe shows words for ‘thousand’ from Romanian (miji) and from Hungarian (ezera), Vietnamese uses a Chinese word for ‘ten thousand’ (Alves 2007), English borrows its terms for zero, million and billion, Turkish has sıfır from Arabic for ‘zero’ and European milyon for ‘million’, and K’abeena has zeeruta, from Italian via Amharic, for ‘zero’ (Crass 2007). The effect of the adoption of word-forms from a dominant or prestige language in non-routine interaction contexts can be summarized for kinship terms and numerals through the following borrowing hierarchies: (2)

remote kin > close kin

(3)

numerals in formal contexts > numerals in informal contexts

(4)

higher (and abstract) numerals > lower numerals

214

Yaron Matras

It is noteworthy that the ‘routine-proximity constraint’, as we might refer to it, is not just limited to these categories. A further structural borrowing hierarchy is observed for expressions of local relations (adpositions and local adverbs): (5)

peripheral local relations > core local relations

Under peripheral relations we find in particular those that are cognitively more complex, as they involve the creation of a conceptual relationship with reference to two or more rather than just one named object (‘between’), or to movement rather than situation (‘around’, ‘across’), or to a relation that is qualified (by attitude) rather than just spatial (‘against’, ‘for the benefit of’). Delimitational relations, which are more abstract and which deal with the negative processing of presuppositions concerning sets (e.g. ‘without’, ‘instead of’, ‘except’) all fall under the highly borrowable, conceptually more complex prepositions. In the latter case, a connection can also be established with the feature ‘contrast’ and the factor ‘speaker control’, which I deal with in the next section. Romani dialects tend to borrow prepositions like pretiv/protiv (Slavic) ‘against’, is (Slavic) and fon (German) ‘from’, za (Slavic) and bis (German) ‘until’, de (Romanian) ‘since’, bez/brzo (Slavic), xoris (Greek), utan (Swedish), and oni (German) ‘without’, vmesto/namesto (Slavic) ‘instead’, osven, skluchenje, kromje (Slavic), in loc dΩ (Romanian) ‘instead’, and ektos (Greek) ‘except for’. Spanish de ‘from’, por ‘for’, and para ‘for the purpose of’ and entre ‘between’ are among the frequently-attested borrowings in South- and Central American languages (see e.g. Stolz 1996), as opposed to, for example, prepositions with more ‘core’ meanings such as ‘at’, ‘on’, ‘above’, or ‘with’. Chamorro has desde ‘from’, asta ‘until’, sin ‘without’, pot ‘in order to’, and kontra ‘against’ (Topping 1973: 126–129) and Maltese borrows kontri ‘against’ and faƛƛata ‘opposite’ from Italian.

3.2.

Control over form selection in speech production

Above I alluded to the model of bilingual speech production, which considers the bilingual’s linguistic competence as constituting one full repertoire of forms and structures rather than two language systems. Individual structures and word-forms are accompanied in the repertoire by the speaker’s awareness of constraints on the appropriateness of their usage in individual settings and contexts. These constraints guide the speaker in selecting forms and structures within the repertoire, in separating subsets

Universals of structural borrowing

215

within the repertoire following more or less stable situational and contextual triggers, as well as in occasionally mixing subsets, i. e. making use of the repertoire as a whole in certain types of communicative interaction and often for particular communicative purposes (cf. Grosjean’s 2001 notion of “bilingual mode”). The speaker is assisted by a mechanism that monitors the extra-linguistic conditions of the interaction and facilitates the selection of appropriate structures and word-forms, and inhibits the selection of others (cf. Green 1998). It is hypothesized here that this mechanism is, like other aspects of speech production and especially of lexical retrieval, sensitive to a set of factors which we might refer to in a pre-theoretical sense as the speaker’s ‘state of mind’. They include pathological conditions such as fatigue or impairment, but are also sensitive to moments of high tension at the level of the speech interaction and its management. Specifically, the proper functioning of the selection mechanism might be disturbed by the need to divert attention away from it and onto an increased effort to reassure the listener of the speaker’s assertive authority. Such instances typically occur when the speaker anticipates a clash between the speaker’s delivery of propositional content, and the listener’s expectations or a shared presupposition basis. It is then possible, though in no way inevitable, that the speaker may lose control over the selection and inhibition mechanism. The result will be the selection of a structure or word-form that is functionally adequate as far as its meaning is concerned, but contextually inappropriate as it belongs to a subset of the repertoire that is not normally activated in the set of contexts to which the ongoing interaction belongs. The result is, in other words, a bilingual speech production error. Note that the notion of ‘error’ is, of course, a matter of (normative) perception for the speaker and listener. My point is precisely that in some interaction settings, unintentional selections of this kind might ultimately cease to be perceived as errors, and become accepted and even propagated as legitimate and fully functional variants, and might over time even replace any counterpart expressions that had been inherited by the recipient language. However, their initiation is not a result of either the prestige effect, nor does it represent a link between the content and meaning of the expression and the context-specialization of the donor language in the overall diglossic setup, as in the cases discussed in the previous section. What evidence is there for the link between long-term language change (borrowing) and bilingual speech production errors? The first piece of evidence is the existence of a corpus of documented speech production errors that involve specifically the class of items that can be said to repre-

216

Yaron Matras

sent potential clashes between speaker’s message and listener expectations (see Matras 1998, 2000, 2009 for a discussion of relevant examples). Among the documented examples are cases where the speaker falls back to an expression in a language that is not even accessible to the listener, illustrating that such slips are not necessarily licensed by a bilingual mode, and are not necessarily produced as tokens of prestige; indeed, they are often counter-strategic and may, if not repaired, lead to a momentary breakdown in the communication. There is separate evidence from case-studies confirming that the occasional infiltration of items such as discourse markers, modality markers and connectors from a dominant language may, over time, become propagated within the speech community and lead to long-term borrowing (see e.g. Salmons 1990; Brody 1995; Fuller 2001; Del Negro 2005). Once again it is necessary to emphasize that the origin of these borrowings in a dominant language does not mean that the trigger for their first occurrence is inherently connected to any association between the donor language and the meaning conveyed by the borrowed expression. As indicated above, bilingual slips or production errors occur in various directions, among them directions that are dysfunctional from a communicative viewpoint and therefore unlikely to lead to long-term change. Those instances of production errors that may lead to change are those where the error is both understood and accepted by the interlocutors, and thus allowed to occur over and over again without normative intervention. Such situations typically involve unidirectional bilingualism where the language of a socially less powerful group is the recipient language, and that of a dominant elite the donor language. The diglossic setup is thus vital for the successful propagation of the innovation, not for its initiation. Finally, evidence concerning the selection malfunction hypothesis comes from the hierarchical arrangement of borrowability in paradigms of certain functional elements: (6)

contrast > disjunction > addition (‘but’ > ‘or’ > ‘and’; ‘only’ > ‘too’; concessive > most other subordination markers; ‘except’, ‘without’, ‘instead of’ > most other adpositions)

(7)

discourse markers (including fillers, tags, interjections) > focus particles, phasal adverbs > other function words

Universals of structural borrowing

217

The susceptibility of contrast (in the sense defined by Rudolph 1996, i. e. adversative and concessive relations) to borrowing is attested in implicational hierarchies that summarize the borrowing behavior of connectors, focus particles, subordination markers, and adpositions. Attestations of borrowings supporting these generalizations are numerous. Arabic-derived markers for contrast (ama/amma/lƗkin/lakini etc., all meaning ‘but’) are found across a vast area from west Africa to the Caucasus and on to Southwest Asia, including in Hausa, Ful, Somali, Swahili, Lezgian, Turkish, Uzbek, Hindi and Punjabi. Many of these languages also use the Arabic-derived disjunction marker ya, and some also an Arabic-derived addition marker u/w (cf. Matras 1998, 2000). Romani dialects always borrow ‘but’ from the contemporary or recent contact language (e.g. Slavic no, po and ali/ale, Hungarian de, Turkish ama, Greek ala, German aber). Many Romani dialects also borrow ‘or’ and ‘and’. These latter two are borrowed frequently, but are sometimes retained from an older contact language. Ajia Varvara Romani in Athens for instance has ja ‘oder’ from its recent contact language Turkish, but ala ‘but’ from its current contact language Greek; Finnish Romani has elle ‘or’ from Swedish, but mut ‘but’ from Finnish; Manush Romani in France has German un ‘and’ and otar ‘or’, but French-derived me ‘but’, and so on (cf. again Matras 1998 and 2002, as well as Elšík and Matras 2006). Stolz and Stolz (1996, 1997) and Stolz (1996) document the high frequency in Central American and Pacific languages of Spanish-derived pero ‘but’, o ‘or’, porque ‘because’, bueno ‘well’, ni ‘neither’, and sino ‘but, however’, and Stolz (2007) identifies the Italian loans allora ‘well, and then’ dopo, poi ‘and then’, dunque ‘thus, and so’, as well as the connectors però and ma ‘but’, o ‘or’, and e ‘and’ in Molise Slavic, Italo-Albanian, Italo-Greek, and Maltese. Of the sample languages covered in the volume by Matras and Sakel (eds. 2007; see evaluation in Matras 2007), a significant number borrow all three coordinating connectors ‘but’, ‘or’ and ‘and’: Domari, Moseten, Nahuatl, Kurmanji, Rapanui, Indonesian, Quechua, Otomi, Guarani, Kildin Saami, and Western Aramaic; languages that borrow only ‘but’ and ‘or’ are Tasawaq, Purepecha, Vietnamese, Rumungro, K’abeena, and Likpe. Macedonian Turkish borrows Macedonian i ‘and’ as well as ili ‘or’ and a ‘or, whereas’, but retains Turkish ama ‘but’, which is identical to Macedonian ama (cf. Matras 2004). Jaminjung (Schultze-Berndt 2007) uses the borrowed Kriol ani ‘but/only’ alongside its native bugu, while only borrowed forms are used for addition and disjunction. There does not seem to be any dimension of social meaning that is attached to these hierarchies, that is, any reason why the expression of contrast should be linked more tightly with interaction in the donor lan-

218

Yaron Matras

guage, than expressions of disjunction or addition. The link is therefore not referential. Nor is there, in the cross-linguistic samples that have been analyzed, any consistent cultural or structural feature that would explain the consistent borrowability of contrastive expressions, not just in one domain of structure (connectors), but also in others. We are therefore left with the impression that contrast itself is a property that motivates borrowing. The link between contrast and borrowability runs directly via the link between contrast and a clash between speaker- and hearer-sided expectations. In anticipation of the clash, the speaker engages in an increased mental effort to win over the hearer’s confidence. It is this increased engagement of the speech production mechanism that (potentially) takes its toll on the ability to effectively manage the selection and inhibition mechanism, increasing the likelihood of selection malfunctions. Once again, most malfunctions may, in some multilingual settings, remain one-off occurrences, subject to repair by the speaker or to correction by the listener or both, and are prevented from frequent re-occurrence through the presence of normative linguistic values propagated by community institutions. Those malfunctions that are most likely to go unnoticed, non-repaired and uncorrected and to be repeated and ultimately even propagated are those where there is full acceptance of bilingualism, or where a bilingual minority has the prestige to act as catalyst of language change, and where innovation is not hindered by a strong normative, interventionist attitude toward language. On a par with the semantic-pragmatic feature of contrast, we find the wholesale category of discourse markers, including sequential markers (‘and then’, ‘and so’, etc.), fillers, tags, interjections and the like. Their function is intrinsically to monitor and direct the listener’s participation in the interaction, usually by processing vulnerable points in the interaction, where the speaker’s assertive authority might be at stake (hesitations, argumentation, or other need for reassurance). Discourse markers are similarly subject to bilingual speech production errors (see Matras 2000), and are widespread borrowings especially in situations of unidirectional bilingualism among oral minority languages, but also elsewhere. Their specialized use as interaction-regulating operators with reduced semantic autonomy makes them pragmatically detachable from the body of lexical items that is more easily associated with a particular subset of the linguistic repertoire, thereby overriding semantic transparency as a property that acts to facilitate borrowing and illustrating the primacy of functional motivations over structural factors in the borrowing process.

Universals of structural borrowing

219

Lower ranking than discourse markers, but still relatively high on the borrowability scale are focus particles (‘even’, ‘too’, ‘only’, etc.) and phasal adverbs (‘still’, ‘already’) (see Matras 1998). These generally contain more explicit semantic content than discourse markers, and are more easily perceived as lexical entries that are attributable to a lexical subset (language system) within the repertoire. Nonetheless, they too process primarily the relations between the speaker’s proposition, and the presuppositional domain. A similar intensity-factor is involved in the operations triggered by such expressions as around contrast or disjunction: here too, the speaker is on guard to protect assertive authority against the listener’s potential or anticipated reluctance to accept intervention in expectations and the presuppositional domain. The position of focus particles and phasal adverbs on the scale, following that of discourse markers, is a product of the empirical observation derived from the samples that a language that has borrowed items belonging to this class, will have normally also borrowed discourse markers from its principal or relevant contact language. The scalar vulnerability of these classes of elements to borrowing exposes the void that is represented by terms such as ‘uninflected function words’ as far as the reality of contactinduced change is concerned. Such wholesale category is sometimes cited as being high on the borrowability scale (see e.g. Thomason and Kaufman 1988), though in reality the group of uninflected function words is not at all uniform, and individual categories behave not in accordance with an overall structural categorization, but rather following their distinct functional, semantic-pragmatic properties as triggers of linguistic-mental processing tasks in discourse.

3.3.

Control and the speaker’s assertive authority

Mention was made above of categories that are vulnerable to selection malfunctions as a result of the special burden which they impose on the speaker in instances of potential clash between the speaker’s presentation of propositional content and the listener’s expectations as derived from the presuppositional domain. Such instances in discourse might be regarded for the sake of the argument as high-risk operations, where the speaker risks assertive authority and the cooperative relation with the listener. To the high risk operations listed above we can add grammatical markers that highlight that the speaker is basing a proposition on a relatively insecure basis of knowledge, thereby relying more strongly on the listen-

220

Yaron Matras

er’s cooperation and supplementary extension of the presuppositional base to accommodate uncertain assertions. This is reflected in the following borrowability hierarchies: (8)

indefinites > interrogatives > deixis, anaphora

(9)

modality > aktionsart > future tense > other tense/aspect

(10)

obligation > necessity > possibility > ability > desire

(11)

superlative > comparative

(12)

concessive, conditional, causal, purpose > other subordinators

(13)

factual > non-factual complementizers

Indefinites belong to the explicit presupposition-processing apparatus: the speaker is entrusting the hearer with the procedure of retrieval or imaginary substitution or supplementation of the relevant information within a specified ontological domain. Thus, with an expression like ‘anywhere’ the speaker is effectively laying down responsibility for specifying relevant information, while with an expression like ‘somebody’ the specific information is being withheld. Like other high risk operators, indefinites are likely candidates for borrowing, and numerous cases of borrowed indefinite markers as well as word-forms are attested. Turkish for instance has borrowed Persian hiç ‘no, any’, as well as Arabic úey ‘thing’ in bir úey ‘something’, hiç bir úey ‘nothing’, as well as Arabo-Persian bazi ‘some’ and bazen ‘sometimes’. Maltese employs Italian kwalunkwe ‘any’. All Romani dialects borrow indefinite markers, and many indefinite word-forms are also borrowed. Borrowed markers include Slavic (v)sako, ni, bilo, Romanian-derived vare, Turkish-derived hiþ, her, bazi, Hungarian vala, and more. Borrowed indefinite word-forms are numerous and include Slavic ništo ‘something, nothing, anything’, Polish zawsze ‘always’, Romanianderived mereu ‘always’, Hungarian-derived šoha ‘ever, never’ and mindig ‘always’, Greek-derived þipota ‘nothing’ and kathenas ‘somebody’, and many more (cf. Elšík and Matras 2006). Spanish-derived indefinites are common among the indigenous languages of Central and South America, such as Otomi kada kyen ‘everybody’, en kwalkyer parte ‘anywhere’, nunka ‘never’, syempre ‘always’, and many more (Hekking and Bakker 2007), and Guarani borrows the Spanish indefinites alguno ‘somebody’, toda ‘everybody’, and others. Modality is the conditioning of the speaker’s secure knowledge, and thus represents a reduction in the speaker’s assertive authority and so a risk factor in the maintenance of stability of the roles of speaker and lis-

Universals of structural borrowing

221

tener in the interaction. Modality is followed by Aktionsart – the subjective-evaluative perspective on an action/event. Tense and aspect follow, and it is indicative that the tense-aspect category that appears most susceptible to contact-induced structural change is that of Future, where the speaker’s epistemic authority is once again at its weakest. The hierarchy of modality-aktionsart-tense/aspect (9) is matched rather closely by the borrowability hierarchy of modal expressions, depicted in (10). The greater the involvement of an external force in determining the modality of the proposition, the weaker the speaker’s control over the truth and accuracy of that proposition. Both Elšík and Matras (2006) and Matras (2007) for the cross-linguistic sample establish the greater borrowability of superlative markers over comparative markers (11). The comparative procedure activates presupposed relevant knowledge about an object of comparison; this activation already puts it on the side of operations in which the speaker has to rely heavily on the listener. The superlative goes a further step beyond, setting the object of reference apart from the presupposed set of relevant objects, creating a demarcation or delimitation line. Superlative relations thus fall within the broad domain of contrast, exemption and restriction, which are particularly contact-susceptible. Romani dialects recruit comparative and superlative markers among the preposed, unbound or semi-bound markers of their various contact languages, such as Slavic po, Romanian mai, and Turkish daha for the comparative, Slavic naj, Hungarian leg, Turkish en for the superlative. A bound comparative/superlative marker eder had been borrowed into Proto-Romani from Iranian, and continues to be used in many dialects. Sinti Romani for example uses it for the comparative – thus sik ‘fast’, sikedΩr ‘faster’ – and adopts the German superlative, including both its inflection ending and an accompanying preposition: am sikestΩ ‘fastest’, (dialectal) German am schnellste(n). The Spanish comparative and superlative marker más has entered several Central American languages, e.g. Nahuatl, Otomi, Yucatec Maya, Tzutujil, and others, as well as Austronesian languages like Tagalog, Hiligaynon, and Chamorro (Stolz and Stolz 1997). Speakers of Kurmanji, Adyghe, and Aramaic in Turkey often use the Turkish particles daha and en, which are also attested in Asia Minor Greek. Indonesian borrows a Javanese superlative particle paling. Yiddish has grojs ‘big’, grejser ‘bigger’, but same grojs ‘biggest’, using the Russian-derived marker same, which is also borrowed as a superlative marker into Kildin Saami. Domari even adopts Arabic word-forms for the comparative and superlative, resulting in a suppletive paradigm of comparison/suppletion for all indigenous adjectives in the language.

222

Yaron Matras

Some of the most frequently borrowed subordinating conjunctions express concessive relations, causal relations, purpose, and conditionality (as in the hierarchy depicted under 12). These are once again the domains in which the relations between states of affairs depicted in the proposition are potentially controversial or beyond the speaker’s domain of secure knowledge. Finally, factual complementizers appear to be more borrowing-prone than non-factual complementizers (hierarchy 13). At first glance, this appears to contradict our impression from other category domains, where the speaker’s weak epistemic authority correlates with greater borrowability. But in fact this very same condition is met in factual complementation: There is a requirement to enforce the integrated interpretation of two separate events, which makes the proposition more difficult to accept, while non-factual complements typically portray a single, integrated event (cf. Givón 1990). Romani dialects, irrespective of location, generally distinguish the two types of complementizer, as do most languages of the Balkans. The original Romani factual complementizer kaj is often replaced, in the respective dialects, by Greek oti, by Bulgarian þi, by Romanian-derived ke, by Italian ke, and by Hungarian-derived hodž/hod/hoi; the non-factual complementizer (inherited te) is virtually never replaced. Other languages, too, tend to borrow just the factual complementizer. Khuzistani Arabic borrows Persian ke in factual clauses, and Likpe borrows an Ewe marker bΩғ (Ameka 2007).

3.4.

From discourse organization to word-level

A number of hierarchies confirm the greater susceptibility to borrowing of grammatical devices that operate at the discourse level, followed by those at the sentence and clause level, and eventually by those at the phrase and word level, as observed by Stolz and Stolz (1996) and Ross (2001). The high position of discourse operations on the scale represents what Salmons (1990) described as the convergence of communication management strategies: (14)

discourse markers > focus particles, phasal adverbs > other function words

(15)

prosody > segmental phonology

(16)

greetings and similar formulae > question particles, conditional particles, modality particles > negation and other content particles

Universals of structural borrowing

223

Above I already discussed the position of discourse markers in relation to other so-called uninflected function words. The position of prosody remains largely uninvestigated, but with agreement that suprasegmental features are much more volatile in contact situations (cf. Matisoff 2001; Queen 2001). Romani dialects once again provide a useful sample, showing overwhelming tendencies to converge in key prosodic features with their contact languages.3 A further discourse-level borrowing phenomenon is the use of greetings. Here too, we find considerable volatility in bilingual contexts. German dialects have the forms ade, tschü, tschu and tschüß deriving from French adieu ‘goodbye’. The Arabic greeting marh.aba appears in Hausa maraba and Turkish merhaba, among many others. Colloquial Hebrew has hi and bye from English, ahlan from Arabic, and the combination yallah bye! ‘bye, then!’ from Arabic and English, which in turn is borrowed into Palestinian Arabic. Urdu has Arabic shukriya for ‘thank you’, as does Turkish (teúekkür), though middle-class, urban Turkish copies middleclass urban Persian in using French derived merci. The reflexive-imperative hadi is common throughout the Balkans. Following on the scale, after greetings – which themselves constitute speech acts – we find speech act markers, such as question and conditional particles, as well as modal particles (cf. borrowing of Turkish question particles into minority languages such as Kurdish, Adyghe, and Romani; or use of Slavic and Turkish conditional particles in various Romani dialects). These are generally more prone to borrowing than particles that modify content or meaning, e.g. negation or temporal particles. The motivation to borrow discourse operators of these types arises from their gesture-like and automatized use that is sensitive primarily to changes in the overall interaction constellation and so susceptible toward the more intuitive need to monitor and control the interlocutor’s participation and attitude to the speaker and the interaction as a whole, rather than form part of the propositional content of utterances that are subject to more analytical planning and control on the part of the hearer. This automatized, intuitive reaction is prone to escaping the speaker’s control over the selection and inhibition mechanism that regulates the choice of context-appropriate structures within the linguistic repertoire. Essentially, then, the procedure by which discourse-regulating elements are subjected to production malfunctions is similar to the process that leads to fusion

3 Cf. the Manchester Romani Project online phrase documentation for Romani dialects: http://romani.humanities.manchester.ac.uk/.

224

Yaron Matras

around high-risk operators, as described above, the point in common being their proneness, for distinct reasons, to escape analytical control. The discourse-clause-phrase continuum of borrowing is represented at the level of syntactic convergence, as well: (17)

clause linking > word morphology

(18)

possessive construction > attributive construction

(19)

existential (copula) predication > lexical predication

Linguistic areas or cases of convergence of even just two languages will rarely show convergence at the level of word morphemes unless they also show convergence at the level of clause combining. This holds for groups of languages that show similarities in the system of case representation, such as the languages of Sri Lanka (cf. Bakker 2005), the Balkans, Siberia (Anderson 2005), or Arnhem Land (Heath 1978), or in the system of tense-aspect representation, such as the languages of Anatolia, South Asia, or the Amazon Basin (Aikhenvald 2002): in all cases and in others, these manifestations of morphological convergence are accompanied by similar clause combining strategies. At a somewhat more subtle level, convergence appears to begin with relations among potentially more autonomous entities than among those that are more closely interdependent. Evidence from a variety of languages such as Domari, Macedonian Turkish, Khuzistani Arabic, and Romani suggests that representation of nominal attribution is more likely to undergo structural convergence than that of adjectival attribution. The existential predication, which invariably represents a linear arrangement of at least two arguments, is more prone to convergence than the potentially less complex lexical predication (cf. Matras 2008).

4.

Conclusions: re-assessing borrowing

Before reviewing the patterns that emerge from the above survey of borrowing universals, let us remind ourselves once more what, precisely, is understood as the psychological mechanism of borrowing. In diachronicetymological terms, borrowing denotes that the origin of a word-form or construction is from a donor language that has been in contact with the recipient language sometime in its history. It is, in other words, an innovation that owes its trigger to a particular sociolinguistic setting, one in which at least some speakers of the recipient language have had at least some

Universals of structural borrowing

225

contact with the donor language (or an interim donor language; cf. borrowing of Arabic words into English via Spanish and French, or via Turkish and French, etc.), and where this sector of speakers was in a position to propagate the replication of the structure in question throughout a significant part of the speech community, whether bilingual or monolingual. This, more or less, amounts to the generally agreed on diachronic characterization. In order to gain a deeper insight into the actual triggering of the borrowing process, we need to zoom in on the initiation phase itself. Here, we are dealing, prototypically at least, with a bilingual setting, that is, with a setting where speakers interact in a variety of contexts and situations and in which their choice of linguistic structures reflects the requirements and expectations of those contexts and situations. The speakers on whom we are focusing have, in other words, a complex and multi-layered repertoire of linguistic structures, from which they select in order to accommodate to the expectations of distinct audiences in a variety of interaction contexts (relevant factors being addressee, setting, topic, and stylistic contrasts). One might describe this state of affairs as the creation of a mental demarcation boundary within the linguistic repertoire, or even better as the activation of a kind of selection mechanism that controls the selection of appropriate forms and inhibits the selection of others (cf. Green 1998; Paradis 2004). Borrowings have been referred to as “transfers” or “interferences” (Weinreich 1953), not without reason. They emerge as carry-overs from the set of structures normally reserved for a certain set of interaction contexts, into a different set of interaction contexts. In order to understand what can be borrowed with what degree of relative ease, we must first understand what borrowing entails, and why it occurs. The question we address first is therefore: why should bilinguals act against the expectations regarding appropriate structure selection in a particular interaction context, and disregard the mental demarcation line among repertoire components or neutralize the mechanism that selects only appropriate structures and inhibits those that are not appropriate in the ongoing interaction? Most explanations of borrowing have highlighted the aspects of gaps and prestige. In direct response to our question “why borrow?” the first – the gap hypothesis – postulates that speakers have essentially no choice in pursuit of certain communicative goals but to cross the mental demarcation lines and disregard context-appropriateness. Such communicative goals include making reference to concepts or objects that are normally present only in some settings but not in others, and consequently lack

226

Yaron Matras

well-established reference devices or labels in these settings. These are the notorious cultural loans. The gap hypothesis might be extended to the structural domain to account, for instance, for the extension of the phoneme inventory in response to the need to integrate new words that contain new phonemes. Note however that borrowing is not the only option in any of these cases. It is possible to integrate loanwords phonologically, rather than extend the phonological system to accommodate them; and it is possible to coin new words for new objects rather than import word-forms from another language. Borrowing thus entails a choice in favor of a strategy that puts the effective use of the linguistic repertoire in its entirety in pursuit of certain communicative goals above the maintenance of demarcation boundaries within the repertoire and the audience-oriented loyalty that such boundaries serve to flag. It is here that the role of prestige becomes of interest. It is often suggested that speakers prefer to use items (word-forms, constructions) from a language that symbolizes power, dominance, and success in order to capitalize on the symbolic association of their speech with that of a dominant population sector. The most obvious indication that powerful languages attract prestige is the willingness of speakers all around the globe to give up loyalty to the languages of smaller groups, nations, or the family, and to conduct most interactions just in the more prestigious language(s). Nonetheless, the process of borrowing does not, of course, inevitably lead to language shift, and it would be wrong to interpret borrowing as invariably a first stage in the process of language shift. Borrowing is in many ways, in fact, the opposite: a bundle of strategies that ensures that the weaker, recipient language can continue to be used and to flag group-loyalty in spite of the growing reliance on a dominant, donor language for a large number of communicative purposes. Moreover, both the ‘gap hypothesis’ and the ‘prestige hypothesis’ fail to explain why speakers should prioritize certain structures for borrowing based on semantic and grammatical parameters rather than on parameters that are related to social and referential meaning. It is therefore necessary to re-assess the role of both referential gaps and social prestige in the process of borrowing. Let us return to the question of where and why a speaker should choose to generalize a single structure from within the complex linguistic repertoire, rather than to ensure separation of repertoire subsets and thereby act consistently in flagging loyalty to two separate linguistic identities. Phrased somewhat differently, we are looking for the conditions under which the convenience of having just a single mode of expression might override the principle of overt accommodation to the distinct inter-

Universals of structural borrowing

227

action context. Gap in this perspective means that the speaker prefers the convenience of a single representation over the creativity that is needed to accommodate to an audience by inventing a new term; and prestige means that, for the sake of having the convenience of using just a single form to represent a concept or grammatical operation, the speaker is exploiting the status of one of the languages in the multilingual setting and expecting the listener to accept the generalization of forms from this language in all interaction contexts. At this point we can return to the borrowing hierarchies discussed above. By examining values of individual categories and category classes in respect of their borrowing behavior, we have been able to isolate various language-internal properties that facilitate or even promote borrowing. Of those, some have to do with the content of the referent: tight associations between the meaning of a word and a particular context in which it is used may result in associating that word with a particular language that is the preferred language of negotiating activity in that particular context. Borrowing is in such instances a reflection of the diglossic specialization of languages in the multilingual repertoire. We have seen that this feature applies to unique referents as well as to specialized professional and cultural terminology. On the other hand, association with proximate and intimate routines may act as a constraint against borrowing. In all cases, borrowing is facilitated by structural detachability and semantic transparency, but these are not a pre-condition for borrowing. In general, interaction-level management of the discourse acts as a motivating factor toward the fusion or non-separation of structures in the repertoire and the generalization instead of just one item or set of items for the function in question, regardless of interaction context. It was argued that this non-separation around devices that manage the communicative interaction facilitates effective management, by removing the need to select among structures in instances that require a high concentration effort. At the top of the scale are those devices that process an anticipated clash between the speaker’s speech activity and propositional content, and the listener’s expectations. But in general we saw that the level of discourse interaction is particularly prone to convergence. Control and management of the interaction can be said to constitute a more automatized, instinctive or gesture-like operation, and is less subject to analytical control. While interaction-level operators may appear in some linguistic systems as less complex, uninflected, or clause-peripheral, it is essentially their language-processing function that determines their behavior, not their formal shape. It is quite possible that control over interaction-level devices is subject to a different set of operational proce-

228

Yaron Matras

dures at the neurophysiological level, as indicated by the behavior of modal particles in the speech of aphasia patients, or by the production of intonation and prosody (cf. Menn and Obler 1990; Friederici 2001; McMahon 2005). We might even speculate further that those components of speech that are designed to grab the hearer’s attention and direct the hearer emotionally through the discourse represent a very archaic layer in the architecture of the language faculty itself, one that is more gestureoriented than analytical. In this respect, borrowing hierarchies might help us gain insights into the more general hierarchical nature of grammar. I have re-interpreted the notions of ‘gap’ and ‘prestige’, which are widespread in the discussion of structural borrowing, and used them to describe the speaker’s motivation to generalize just one structure for one function, referent, or grammatical operation in the linguistic repertoire, and the speaker’s ability to do so without loss of face and without causing a breakdown in the communication. The borrowing of at least some operations displays a link with the speaker’s control over the language production mechanism, and it is here that the question was raised whether this link is in turn a by-product of the evolution of speech and the neurophysiological architecture of grammar. If this were indeed the case, then we might need to face a somewhat embarrassing, and politically very incorrect conclusion, namely that the speech mechanism is set up to cater in the first instance for monolingual speech production. Multilingual speech – the need to select among structures according to their context-appropriateness – is a facility that is controlled by an analytical skill. When this skill breaks down, the speaker ends up reverting to a state of affairs of nondifferentiation of linguistic systems, one that is typical of the more simplex repertoires of monolinguals.

Universals of structural borrowing

229

References Aikhenvald, Alexandra Y. and R. M. W. Dixon (eds.) 2001 Areal Diffuson and Genetic Inheritence: Problems in Comparative Linguistics. Oxford: Oxford University Press. Aikhenvald, Alexandra Y. 2002 Language Contact in Amazonia. Oxford: Oxford University Press. Alves, Mark 2007 Sino-Vietnamese grammatical borrowing: An overview. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in CrossLinguistic Perspective, 343–361. Berlin: Mouton de Gruyter. Ameka, Felix K. 2007 Grammars in contact in the Volta basin (West Africa): On contactinduced grammatical change in Likpe. In: Alexandra Yurievna Aikhenvald and R. M.W. Dixon (eds.), Grammars in Contact: A Cross-linguistic Typology, 114–142. (Explorations in Linguistic Typology, Volume 4). Oxford: Oxford University Press. Ameka, Felix K. 2007 Grammatical borrowing in Likpe (Sİkpİlé). In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 107–122. Berlin: Mouton de Gruyter. Bakker, Peter 2006 The Sri Lanka Sprachbund: The newcomers Portuguese and Malay. In: Yaron Matras, April Mc-Mahon and Nigel Vincent (eds.), Linguistic Areas. Convergence in Historical and Typological Perspective, 135–159. New York: Palgrave Macmillan. Borg, Alexander and Marie Azzopardi-Alexander 1997 Maltese. London: Routledge. Brody, Jill 1995 Lending the ‘Unborrowable’: Spanish discourse markers in indigenous American languages. In: Carmen Silva-Corvalán (ed.), Spanish in Four Continents. Studies in Language Contact and Bilingualism, 132–147. Washington D. C.: Georgetown University Press. Campbell, Lyle 1993 On proposed universals of grammatical borrowing. In: H. Aertsen and R. J. Jeffers (eds.), Historical Linguistics 1989, 91–109. Amsterdam: Benjamins. Crass, Joachim 2007 Grammatical borrowing in K’abeena. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 91–105. Berlin: Mouton de Gruyter. Dal Negro, Slivia 2005 Lingue in contatto: Il caso speciale dei segnali discorsivi. In: Giorgio Banti, Antoietta Marra and Edoardo Vineis (eds.), Atti del 4q

230

Yaron Matras

congresso di studi dell’Associazione Italiana di Linguistica Applicata, 73–88. Perugia: Guerra Edizioni. Dawkins, Richard MacGillivray 1916 Modern Greek in Asia Minor. Cambridge: Cambridge University Press. Elšík, Viktor and Yaron Matras 2006 Markedness and Language Change: The Romani Sample. Berlin: Mouton de Gruyter. Field, Frederic W. 2002 Linguistic Borrowing in Bilingual Contexts. Amsterdam: Benjamins. Friederici, Angela 2001 Syntactic, prosodic, and semantic processes in the brain: Evidence from event-related neuroimaging. Journal of Psycholinguistic Research 30: 237–250. Fuller, Janet 2001 The principle of pragmatic detachability in borrowing: Englishorigin discourse markers in Pennsylvania German. Linguistics 29: 351–369. Givón, Talmy 1990 Syntax: A Functional-Typological Introduction. Volume II. Amsterdam: Benjamins. Green, David 1998 Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition 1: 67–81. Grosjean, François 2001 The bilingual’s language modes. In: Janet L. Nicol (ed.) One Mind, two Languages. Bilingual Language Processing, 1–22. Oxford: Blackwell. Heath, Jeffrey 1978 Linguistic Diffusion in Arnhem Land. Canberra: Australian Institute for Aboriginal Studies. Hekking, Ewald and Dik Bakker 2007 The case of Otomi: A contribution to grammatical borrowing in cross-linguistic perspective. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 435–464. Berlin: Mouton de Gruyter. Johanson, Lars 2002 Structural Factors in Turkic Language Contacts. Richmond: Curzon. Kossmann, Maarten 2007 Grammatical borrowing in Tasawaq. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 75–89. Berlin: Mouton de Gruyter. Matisoff, James A. 2001 Genetic versus contact relationship: Prosodic diffusibility in SouthEast Asian languages. In: Alexandra Y. Aikhenvald and R. M. W.

Universals of structural borrowing

231

Dixon (eds.), Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, 291–327. Oxford: Oxford University Press. Matras, Yaron and Jeanette Sakel 2007 Investigating the mechanisms of pattern-replication in language convergence. Studies in Language 31: 829–865. Matras, Yaron and Jeanette Sakel (eds.) 2007 Grammatical Borrowing in Cross-linguistic Perspective. Berlin: Mouton de Gruyter. Matras, Yaron 1998 Utterance modifiers and universals of grammatical borrowing. Linguistics 36: 281–331. Matras, Yaron 2000 Fusion and the cognitive basis for bilingual discourse markers. International Journal of Bilingualism. 4:4: 505–528. Matras, Yaron 2002 Romani: A Linguistic Introduction. Cambridge: Cambridge University Press. Matras, Yaron 2004 Layers of convergent syntax in Macedonian Turkish. Mediterranean Language Review 15: 63–86. Matras, Yaron 2007 The borrowability of grammatical categories. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 31–74. Berlin: Mouton de Gruyter. Matras, Yaron 2009 Language Contact. Cambridge: Cambridge University Press. Matras, Yaron, April McMahon and Nigel Vincent (eds.) 2005. Linguistic Areas. Convergence in Historical and Typological Perspective. Houndmills: Palgrave. McMahon, April 2005 Heads I win, tails you lose. In: Philip Carr, Jacques Durand and Colin J. Eden (eds.), Headhood. Elements, Specification and Contrastivity, 255–275. Amsterdam: Benjamins. Menn, Lise and Lorraine K. Obler 1990 Cross-linguistic data and theories of agrammatism. In: Lise Menn and Lorraine K. Obler (eds.), Agrammatic Aphasia: A Cross-language Narrative Sourcebook, 1369–1389. Amsterdam: Benjamins. Moravcsik, Edith 1978 Universals of language contact. In: Joseph H. Greenberg (ed.), Universals of Human Language, 94–122. Stanford: Stanford University Press. Muysken, Pieter 2000 Bilingual Speech. A Typology of Code-mixing. Cambridge: Cambridge University Press.

232

Yaron Matras

Paradis, Michel 2004 A Neurolinguistic Theory of Bilingualism. Amsterdam: Benjamins. Queen, Robin M. 2001 Bilingual intonation patterns: Evidence of language change from Turkish-German bilingual children. Language in Society 30: 55–80. Ross, Malcolm 2001 Contact-induced change in Oceanic languages in north-west Melanesia. In: Alexandra Y. Aikhenvald and R. M.W. Dixon (eds.), Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, 134–166. Oxford: Oxford University Press. Rudolph, Elisabeth 1996 Contrast. Adversative and Concessive Relations on Sentence and Text Level. Berlin: De Gruyter. Sakel, Jeanette and Yaron Matras 2007 Modelling contact-induced change in grammar. In: Thomas Stolz, Rosa Salas Palomo and Dik Bakker (eds.), Aspects of Language Contact. New Theoretical, Methodological and Empirical Findings with Special Focus on Romanisation Processes, 63–88. Berlin: Mouton de Gruyter. Salmons, Joseph 1990 Bilingual discourse marking: Code switching, borrowing, and convergence in some German-American dialects. Linguistics 28: 453–480. Schultze-Berndt, Eva 2007 Recent grammatical borrowing into an Australian Aboriginal language: The case of Jaminjung and Kriol. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-linguistic Perspective, 363–386. Berlin: Mouton de Gruyter. Stolz, Christel and Thomas Stolz 1996 Funktionswortentlehnung in Mesoamerika. Spanisch-amerindischer Sprachkontakt Hispanoindiana II. Sprachtypologie und Universalienforschung 49: 86–123. Stolz, Christel and Thomas Stolz 1997 Universelle Hispanismen? Von Manila über Lima bis Mexiko und zurück: Muster bei der Entlehnung spanischer Funktionswörter in die indigenen Sprachen Amerikas und Austronesiens. Orbis 39: 1–77. Stolz, Thomas 1996 Grammatical Hispanisms in Amerindian and Austronesian languages. The other kind of Transpacific isoglosses. Amerindia 21: 137–160. Stolz, Thomas 2007 Allora. On the recurrence of function-word borrowing in contact situations with Italian as donor language. In: Jochen Rehbein,

Universals of structural borrowing

233

Christiane Hohenstein and Lukas Pietsch (eds.), Connectivity in Grammar and Discourse, 75–99. Amsterdam: Benjamins. Tadmor, Uri 2004 Function loanwords in Malay-Indonesian (and some other Southeast Asian languages). Paper presented at the Loan Word Typology Workshop, Max-Planck-Institute for Evolutionary Anthropology, Leipzig, 1–2 May 2004. Tadmor, Uri 2007 Grammatical borrowing in Indonesian. In: Yaron Matras and Jeanette Sakel (eds.), Grammatical Borrowing in Cross-Linguistic Perspective, 301–328. Berlin: Mouton de Gruyter. Thomason, Sarah Grey and Terrence Kaufman 1988 Language Contact, Creolization and Genetic Linguistics. Berkeley: University of California Press. Topping, Derek M. 1973 Chamorro Reference Grammar. Honolulu: University of Hawaii Press. Van Hout, Roland and Pieter Muysken 1994 Modelling lexical borrowability. Language Variation and Change 6: 39–62. Weinreich, Uriel 1953 Languages in Contact. The Hague: Mouton. Wichmann, Søren and Jan Wohlgemuth 2008. Loan verbs in a typological perspective. In: Thomas Stolz, Rosa Salas Palomo and Dik Bakker (eds.), Aspects of Language Contact. New Theoretical, Methodological and Empirical Findings with Special Focus on Romanisation Processes, 89–122. Berlin: Mouton de Gruyter.

Part 3 Methodological issues of variation research

Variation and reproducibility in linguistics Walter Bisang Abstract Variation and its adequate theoretical integration is of crucial importance for linguists interested in the grammar of individual languages as well as for linguists dealing with cross-linguistic generalizations. The present paper will discuss both types of research interests from the perspective of reproducibility and criteria of falsification (Popper [1959] 2002) and from the perspective of integrative functionalism (Croft 1995, 2000). As will be shown, the combination of these two perspectives has the following three consequences for the assessment of the adequacy of linguistic theorizing: (i) Functional and social factors must be applied to all linguistic theories. (ii) Given the properties of these factors, exact reproducibility is not possible. (iii) Both factors operate at the level of grammatical descriptions of individual languages as well as at the level of typological generalizations (and formal explanations in terms of Universal Grammar). Keywords: falsification, innovation vs. propagation, integrative functionalism, linguistic typology, methods of data gathering, reproducibility, sampling (of languages), sociolinguistic factors, text corpora, universal grammar

1.

Setting the stage

1.1.

The link between variation and reproducibility

Variation and its limitations are fundamental notions of linguistics. Researchers who are interested in the grammatical structures of individual languages look at variation from the perspective of what is used and accepted by native speakers of certain languages and what is rejected, while researchers looking for generalizations of a universal dimension are interested in what is at the roots of cross-linguistic variation. In each case, the validity of regularities and generalizations claimed by linguists crucially depends on reproducibility, i. e., on certain factors that are necessary to define a speech situation in which the production of a particular gram-

238

Walter Bisang

matical structure is to be expected. In such an experimental setting, the factors that define the speech situation are used to reduce variation among theoretical predictions. The very same factors do not only reduce variation, they also trigger it. Thus, the study of reproducibility and the study of variation are based on the same factors – the factors that are responsible for variation are also the factors that need to be considered when testing reproducibility. The perspective of variation and its consequences for reproducibility will provide the basis for a number of general thoughts about different linguistic approaches and about the foundations on which they are established. These thoughts will be laid out in this section as follows. Subsection 1.2 will briefly discuss the theoretical background. The role of reproducibility in functional typology and the insufficiency of sampling criteria will be presented in subsection 1.3. Based on the overall relevance of the speech situation and its properties for defining reproducibility conditions, subsection 1.4 will introduce the three hypotheses whose validity will be argued for in the remainder of this paper. Finally, subsection 1.5 will describe the structure of the paper as a whole.

1.2.

Theoretical background

As is well known, reproducibility is a fundamental part of Popper’s philosophy of science. In his hypothetical-deductive approach, the scientist has to start deductively from a theory consisting of basic statements. Such a theory cannot be verified by experience, it can only be falsified or “singled out by means of empirical tests, in a negative sense: it must be possible for an empirical scientific system to be refuted by experience” (Popper [1959] 2002: 18). A basic statement, i. e., “a statement which can serve as a premise in an empirical falsification; in brief, a statement of a singular fact” (Popper [1959] 2002: 21), must be observable and it must be reproducible to grant inter-subjective testability by observation. As Popper ([1959] 2002: 66) puts it, “a few stray basic statements contradicting a theory will hardly induce us to reject it as falsified. We shall take it as falsified only if we discover a reproducible effect which refutes the theory”. As can be seen from the above statement, Popper’s ([1959] 2002) approach is based on deduction. This is no problem for formal approaches, since their hypotheses on the properties of innate Universal Grammar are deductive, no matter whether Universal Grammar is described in terms of parameters and modules (Chomsky 1986), merger and movement (Chomsky 1995) or recursion (Hauser, Chomsky and Fitch 2002; Fitch, Hauser,

Variation and reproducibility in linguistics

239

and Chomsky 2005). In contrast, functional typological approaches clearly start out from inducing their generalizations from the comparison of data from a set of languages (language sample). In spite of this, generalizations in functional typology are open to falsification and reproducibility if they are reinterpreted in terms of basic statements. Thus, absolute implicational universals (Greenberg 1963) and hierarchies (e.g. animacy hierarchy, Silverstein 1976) as well as functional explanations from parsing (Hawkins 2004), iconicity vs. economy (Haiman 1983) or discourse (Givón 1979; DuBois 1987) can be understood as potentially falsifiable theoretical statements. Even though the two types of formal and functional generalizations are very different, both of them can be understood as factors that define the boundaries of variation. If we follow Newmeyer (2005: 127), who presents the following quotation, the two types of generalizations may even be in a complementary relation: For the traditional formalist, it is actually desirable for some linguistic patterns, especially those that are gradient, to be explained by functional principles. The remainder, once language processing influences are factored out, might be a simpler, cleaner, and more accurate picture of the nature of the innate language faculty and its role in delimiting the set of possible human languages (Frisch 1999: 600).

1.3.

Reproducibility in functional typology

If functional approaches make generalizations whose validity depends on reproducibility, then this leads to the next question about the factors that are necessary to guarantee reproducibility. To answer this question, let’s first look at different levels where issues of reproducibility arise. The levels of writing adequate grammars of individual languages and of discovering typological generalizations have already been mentioned at the very beginning of subsection 1.1. Haspelmath and Siegmund (2006) add two more of them and thus distinguish four levels: (1)

Haspelmath and Siegmund (2006: 74): (i) Replicability of speaker behaviour under identical experimental conditions (acceptability judgments, elicitation of narratives). (ii) Replicability of descriptions of the same language (either the same corpus: will two linguists arrive at the same description of Gothic on the basis of the Wulfila corpus?; or the same speech community: will

240

Walter Bisang two fieldworkers doing fieldwork in the same village arrive at the same grammars?). (iii) Replicability of typological coding based on identical descriptions (will two typologists extract the same information from the same grammars?). (iv) Replicability of typological generalizations based on different samples (drawn from the same universe of languages).

Concentrating on level (1iv), Haspelmath and Siegmund (2006) replicate Greenberg’s (1963) results by using a range of different 30-language samples. As they conclude from this exercise, “prospects for replication are actually quite good” (Haspelmath and Siegmund 2006: 81). Then, they go on arguing that “the impact of [...] different sampling methods has perhaps been overestimated, at least for correlations between features” (Haspelmath and Siegmund 2006: 81). The criterion on which the reproducibility of Greenberg’s (1963) wordorder correlations is confirmed is language sampling. Of course, this criterion can be used for reproducibility in functional typology and it has not only been used by Haspelmath and Siegmund (2006) but also by other contributors to a debate on reproducibility published in Linguistic Typology 10.1 (2006) (also cf. Harris, Hyman and Staros 2006). As will be shown in the next subsection and in subsection 3.2, the selection of individual languages into language samples is not fully sufficient for confirming the reproduciblity of typological generalizations. 1.4.

Three hypotheses based on the relevance of speech situations to reproducibility

Reproducibility of linguistic data ultimately depends on speakers who have to select the most appropriate linguistic structure in a particular speech situation, which is the testing ground for reproduction (cf. Eide and Sollid, this volume, on exceptions to the verb-second rule in Norwegian and the selection of contextually adequate grammatical solutions). For that reason, hypotheses about reproducibility and its applicability to linguistic generalizations must be based on the factors that are used to define speech situations. In his evolutionary approach to language change, Croft (2000) offers the following two factors that are responsible for the selection of linguistic structures: ± Functional factors or factors of form-function mapping. ± Social factors.

Variation and reproducibility in linguistics

241

Functional factors create variation via the difficulty of the task faced by the speaker to comply with a large number of rules almost simultaneously. Croft (2000) will discuss these factors in terms of strategies that apply to processes of grammaticalization. Social factors are responsible for variation because different structures may be associated with different social settings and the speaker has to select the one that is most appropriate for the actual speech situation in question. Given the relevance of functional and social factors for testing the reproducibility of a particular linguistic structure, they will be of crucial importance for the three hypotheses listed below. Each of them will be discussed in detail later in the paper (on hypotheses 1 and 2, cf. section 2; on hypothesis 3, cf. section 3). ± Hypothesis 1: The functional and the social factors must be applied to all linguistic theories. ± Hypothesis 2: Given the properties of the functional and the social factors, exact reproducibility is not possible. ± Hypothesis 3: The functional and the social factors do not only operate at level (1i), they also have their repercussions at levels (1ii) and (1iv). At level (1iv), the functional factor is of minor significance. As has been pointed out at the begining (subsection 1.1), the present paper will be of a rather general nature. It will look at a number of consequences for the foundations of different linguistic approaches from the perspective of variation and reproducibility. As a corollary, it will also provide new insights into differences and similarities between these approaches. If exact reproducibility is impossible as stated in hypothesis 2, this casts of course its shadow on the robustness of linguistic generalizations if it does not lead to the outright rejection of reproducibility as a criterion for linguistic theories. As I will argue in section 4, linguistics should not give up reproducibility.

1.5.

A short overview of the structure and the content of the paper

The overall structure of this paper follows the levels that are relevant to reproducibility presented in (1). Section 2 will discuss reproducibility at the level of speaker behaviour (1i). Its first subsection (2.1) will start out

242

Walter Bisang

from reproducibility in formal linguistics (ideal speaker-listener; Chomsky 1965) and functional linguistics (Natural Language User; Dik 1997) and will show that social factors are not only missing in Chomsky’s approach but also in Dik’s Functional Grammar. The remainder of this subsection will deal with different criteria for defining social factors. Subsection 2.2 will argue on the basis of Integrative Functionalism as defined by Croft (1995, 2000) that reproducibility is relevant to whatever theoretical approach and will thus confirm hypothesis 1. It will also explain in what way and why the functional and the social factors do not allow exact reproducibility as claimed in hypothesis 2. Section 3, which will support hypothesis 3, will take up the level of individual grammars, i. e. reproducibility of “descriptions of the same language” (1ii) in subsection 3.1 and the level of typological generalizations (1iv) in subsection 3.2. The question of whether two different typologists “will extract the same information from the same grammars” (1iii) certainly is another important criterion but it will not be addressed in this paper because it is not related to reproducibility in terms of speaker behaviour. Finally, the paper will end with a summary and some general remarks on the relevance of reproducibility for linguistics in section 4.

2.

Behaviour under identical experimental conditions— the factors that are responsible for variation

2.1.

From form to function to the social

Reproducibility depends on speaker behaviour under certain conditions. To understand more about speakers and how they behave, let’s start with a look at how they are conceptualized by two rather divergent theoretical approaches, Chomsky’s (1965) ideal speaker-listener and Dik’s (1997) Natural Language User (NLU). The ideal speaker-listener is defined by his competence, i. e., his knowledge of the language and his ability to use it in performance: “Linguistic theory is concerned primarily with an ideal speaker-listener, in a completely homogeneous speech-community, who knows its language perfectly and is unaffected by such grammatically irrelevant conditions as memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic) in applying his knowledge of the language in actual performance.” (Chomsky 1965: 3). From the perspective of reproducibility, such an abstract definition cannot be implemented into an experiment with concrete informants. The ideal speaker-listener simply does not act in a concrete context or speech situa-

Variation and reproducibility in linguistics

243

tion with individuals with their specific mental states and their specific social positions. From such a perspective, Chomsky’s ideal speaker-listener is context-free and thus beyond reproducibility as far as social factors and factors of form-function mapping are concerned. In contrast to Chomsky’s ideal speaker-listener, the Natural Language User as discussed in Dik’s (1997) Functional Grammar is defined by what properties he needs for successful communication in a broader sense. He does not only have his linguistic capacity but a number of other capacities: (2)

The capacities of Natural Language User (NLU) (Dik 1997: 1–2):

(i) A linguistic capacity: NLU is able to correctly produce and interpret linguistic expressions of great structural complexity and variety in a great number of different communicative situations. (ii) An epistemic capacity: NLU is able to build up, maintain, and exploit an organized knowledge base; he can derive knowledge from linguistic expressions, file that knowledge in appropriate form, and retrieve and utilize it in interpreting further linguistic expressions. (iii) A logical capacity: provided with certain pieces of knowledge, NLU is able to derive further pieces of knowledge, by means of rules of reasoning monitored by principles of both deductive and probabilistic logic. (iv) A perceptual capacity: NLU is able to perceive his environment, derive knowledge from his perceptions, and use this perceptually acquired knowledge both in producing and interpreting linguistic expressions. (v) A social capacity: NLU not only knows what to say, but also how to say it to a particular communicative partner in a particular communicative situation, in order to achieve particular communicative goals.

Such a definition of speakers and listeners as Natural Language Users provides a better basis for designing linguistic experiments in terms of reproducibility because it integrates more factors that may determine a speaker’s actual production of an utterance in a concrete situation. As Dik (1997) points out quite clearly, language and its use is embedded in a broader psychological and social context: “[...] linguistic expressions can be understood properly only when they are considered as functioning in settings, the properties of which are codetermined by the contextual and situational information available to speakers and addressees. Language does not function in isolation: it is an integrated part of a living human (psychological and social) reality.” (Dik 1997: 6). Even though Dik (1997)

244

Walter Bisang

mentions the relevance of the social reality, his Functional Grammar actually concentrates on psychological and cognitive factors, but rarely addresses the social perspective. This does not come as a surprise if one assigns Functional Grammar to external functionalism in Croft’s (1995) classification of linguistic theories (cf. subsection 2.2). What still remains to be done is to elaborate on the highly complex issue of the social factors that determine speaker behaviour. This is to say that an adequate account of the reproducibility of linguistic structures in concrete speech situations has still to be provided. For that purpose, let’s start with the fundamental insight that languages are not structurally homogeneous. This was pointed out at the beginning of the seventh chapter (“Language as a historical product: Drift”) of Sapir’s (1921) famous book Language. An Introduction to the Study of Speech: Every one knows that language is variable. Two individuals of the same generation and locality, speaking precisely the same dialect and moving in the same social circles, are never absolutely at one in their speech habits. A minute investigation of the speech of each individual would reveal countless differences of detail – in choice of words, in sentence structure, in the relative frequency with which particular forms or combinations of words are used, in the pronunciation of particular vowels and consonants and of combinations of vowels and consonants, in all those features, such as speed, stress, and tone, that give life to spoken language. In a sense they speak slightly divergent dialects of the same language rather than identically the same language (Sapir 1921: 147).

Sapir (1921) addresses the lack of structural homogeneity within a language as far as variation within its dialects and, ultimately, among its individual speakers is concerned. Even speaker-specific varieties that strongly deviate from a dialect will not be assigned to another dialect because they are, in terms of Sapir (1921: 148), “not refereable to another norm than the norm of their own series” (Sapir orders speaker variation into “a very finely intergrading series clustered about a well-defined center or norm”). Later, Chambers and Trudgill (1998: 9–12) will show along a similar way of argumentation that the difference between language and dialect can only be defined in terms of the social criterion of how speakers of a certain variety situate themselves with regard to other varieties. If they perceive their variety as an autonomous variety it is a language, but if they refer to their variety as a part of another autonomous variety (heteronomy) they speak a dialect of that more general variety. The criteria of structural similarity and mutual intelligibility are of secondary importance. Given the social background of their definition, dialects are also treated under the heading of social factors in this paper. In addition to their social definition, dialects are also characterized by the geographic

Variation and reproducibility in linguistics

245

criterion of areality. Even though the speakers of a dialect are not socially homogeneous they form a community that is defined by a smaller geographic area which offers enough chances for contact and thus for a certain degree of structural closeness. The division of different varieties into the two levels of dialect and language has its traditional roots in Europe but it is by no means universal. The Chinese cultural tradition distinguishes between the three levels of tuhua ‘local dialect’, fangyan ‘dialect’ and yuyan ‘language’ (Norman 1988). This indicates that maybe a more fine-grained terminology is needed for a more general account of the geographic and contact-based world-wide distribution of linguistic variation. The lack of structural homogeneity does not stop at the level of idiolects, i. e., varieties spoken by individuals, since even the speech production of individuals can be subject to variation. This will be discussed briefly from the perspective of the social status of the participants in a speech situation, their social identity and the communities of practice they are involved in (on functional factors and the lack of structural homogeneity of individuals, cf. subsection 2.2). The social status of the members of a speech situation can be described in terms of power and prestige. Power is defined by Brown and Gilman (1960) as the ability of a person to control the behaviour of another person. Prestige is correlated to power inasmuch as higher power is often associated with higher esteem in a society. However, there are also examples like Greek culture which had higher prestige in the socially dominant Roman society than Roman culture itself (cf. “covert prestige” as discussed by Trudgill (1972) on the example of British English spoken in Norwich). No matter how these rather complex notions may be defined in detail, speakers tend to have different degrees of power and prestige in different speech situations. As a consequence, they may use different linguistic structures depending on the distribution of power and prestige in a given situation. Identity is relevant because the group by which an individual wants to be recognized has its impact on the linguistic structure that will be used (LePage and Tabouret-Keller 1985). This maxim of behaviour is of particular importance in contexts of bilingualisms and language contact but it also applies more generally to any situation in which the use of certain varieties is associated with certain social groups. Since speakers may want to be associated with different social groups in different situations, they may also use different linguistic structures according to their actual selfconceptualization. Additional evidence of linguistic variation within individuals comes from the perspective of language as social practice (Foley 1997). Speakers

246

Walter Bisang

communicate within communities of practice (Eckert 2000) that consist of individuals that participate informally in a shared endeavour or enterprise. The permanent interaction among the members participating in such a common endeavour may lead to linguistic structures that are characteristic of a certain community of practice. Since one and the same individual is normally involved in more than one community of practice, the linguistic properties of its utterances will vary according to those structures that are characteristic of the communities to which that individual belongs. Since not even the speech of individuals is homogeneous, a number of social factors must be reflected to guarantee reproducibility of linguistic data within a given speech situation. Apart from such common social criteria as age, gender and educational background, it will be necessary to look at the social status of the participants in a speech situation, the construction of their identity in that situation, as well as the communities of practice they are involved in. The last point implies that it is necessary to know the social networks (Milroy and Milroy 1985) they interact with. Moreover, the language-as-social-practice approach relativizes general sociological categorizations such as gender and social class (against essentialism) because they are the result of how individuals negotiate their position within their social environment. From a more general perspective, social theories do not seem to be easily accessible to reproducibility tests in general. This is true at least for invisible-hand theories such as the one presented by Keller (1990, 1994), which is of central importance for Croft’s (1995, 2000) Integrative Functionalism (subsection 2.2). In this view, language is an invisible-hand product of individual speakers and their behaviour. If speakers follow their speech-act maxims at a micro-level, they will produce effects at a macrolevel that cannot be predicted from what is going on on the micro-level. This implies that no criteria can be formulated which fully guarantee reproducibility.

2.2.

Integrative functionalism (Croft 1995, 2000) and the factors that trigger variation

Integrative functionalism is one of the three types of linguistic theories distinguished by Croft (1995), the other two are formal linguistics and external functionalism (cf. table 1). Formal linguists claim the existence of an innate purely syntax-oriented language faculty, i. e. Universal Grammar, which is self-contained. External functionalists deny innateness and

Variation and reproducibility in linguistics

247

they have a broader concept of grammar that goes beyond syntax. Dik’s (1997) Functional Grammar as discussed in subsection 2.1 can be seen as an example of external functionalism. What external functionalists share with formal linguists is the assumption that grammar is in some way selfcontained, i. e., its structure is not influenced by the social, historical or cultural facts the speakers of a language are involved in. Finally, integrative functionalists refute self-containedness and thus differ from formal linguists and external functionalists. Starting out from the existence of language-internal variation for expressing one and the same content, they assume that individual speakers also select the variant they are going to use in a given situation according to social criteria, i. e., they do not only apply cognitive-functional principles but they also follow grammar-external criteria. In particular, speakers select from a number of different constructions that can be used for expressing the concept they wish to express on the basis of social criteria. Table 1: Summary of the three different basic approaches to language as discussed by Croft (1995). Formal linguistics:

innateness of a purely syntax-oriented language faculty;

External functionalism:

grammar (syntax and other aspects of grammar) is self-contained;

Integrative functionalism: grammar (syntax and other aspects of grammar) is not self-contained.

The reproducibility of a linguistic structure depends on the definition of a speech situation that guarantees the use of that structure whenever that situation occurs. Since the selection of a certain linguistic structure may also be determined by social factors, the definition of a speech situation for the purpose of a reproducibility test must cover language-internal as well as language-external factors. For formal theories, this means that the impact of Universal Grammar cannot be observed in its essence without simultaneously subtracting the impact of performance with its cognitive and social properties. The same also applies to external functionalism, which also includes research in the sense of (1). Generalizations described in terms of motivations such as discourse, parsing or competing motivations (economy vs. iconicity) are mixed up with social factors in an utterance that need to be distinguished. As a consequence, reproducibility must be based on an integrative perspective no matter what theory is to be tested for reproducibility (hypothesis 1).

248

Walter Bisang

As Popper ([1959] 2002) clearly states, the criteria for setting up a test situation that yields reproducible effects must be based on the theory to be falsified, i. e., what can be falsified by a given theory must be formulated in that theory (“Theory dominates the experimental work from its initial planning up to the finishing touches in the laboratory.” Popper [1959] 2002: 90). From such a perspective, the integration of social factors into theories that exclude them from their premises looks like a contradiction to Popper ([1959] 2002). The problem with this criticism is that social factors are always somehow present in the experimental setting, i. e., the speech situation with its social properties is always there and has its influence on the linguistic structure to be selected, irrespective of whether social factors are part of the theory to be tested or not. Croft’s (1995, 2000) integrative functionalist approach has been developed in the context of language change and is based on the observation that an individual divergence from the norm by one speaker in one utterance cannot be addressed as an instance of language change because it lacks successful diffusion in a speech community. For that reason, it is necessary to distinguish between innovations, i. e. unique deviations from the norm, and propagations, i. e. instances of successful dissemination of innovations. In Croft’s (1995, 2000) view, innovations are the result of language internal functional principles of utterance production or processes of grammaticalization at the level of form-function mapping, while propagation or the successful dissemination of innovations is exclusively governed by social factors: (3)

Croft’s (1995: 524) hypotheses on the factors determining innovation and propagation: a. Innovation will occur as the result of functional factors, specifically novel resolutions of competing functional motivations (stretching of the conventional rules). b. Propagation of innovations is determined by social factors.

As I pointed out in Bisang (2001), propagation is not only subject to sociolinguistic factors, it also depends on cognition. Innovations, which have certain cognitively favourable properties (e.g. in terms of parsing or economy) or fit into already existing construction patterns, are more likely to be selected and disseminated into a speech community. In spite of this, I agree with Croft that social factors are ultimately stronger than cognitive factors because they can promote a cognitively less suitable structure if that structure is socially preferable.

Variation and reproducibility in linguistics

249

In the case of reproducibility, social factors and factors of form-function mapping are of equal importance because both of them are responsible for the structural properties of an utterance in a given speech situation. As will be briefly argued below, both of these factors consist of a multitude of variables that entail the impossibility of fully eliminating the risk of divergent results if they are used to test reproducibility within identical or near identical speech situations (hypothesis 2). The form-function mapping normally leads to rather stable grammatical structures, i. e., the probability that the same grammatical forms are used to express the same concept is very high. But there is a “potential indeterminacy or ambiguity in the attribution of semantic components to syntactic components in an utterance” (Croft 2000: 118). Thus, there is always a certain probability that speakers or listeners analyse the formmeaning mapping within a grammatical construction in a way that deviates from established rules. These changes in the form-function mapping are due to four different types of reanalysis that Croft (2000: 121–143) calls hyperanalysis, hypoanalysis, metanalysis and cryptanalysis. In addition to these functional factors, which operate within individual languages, changes in the form-function mapping may also be due to contact-specific crosslinguistic functional factors such as interference and intraference. Finally, Croft (2000: 156–165) mentions mechanisms of grammaticalization as probably the most important source of changes in the form-function mapping. Joseph (this volume) even sees the origin of linguistic universals in problem solving strategies at a small-scale level that can spread into major generalizations. The ultimate reasons for all the above processes of innovational changes are the difficulty of the task to handle a large number of rules almost simultaneously and neural properties of the brain: Most of the time, the grammatical forms produced are essentially the same as those produced before, albeit in novel combinations, in novel meanings in context, and also with variable pronunciations [...]. The innovations may be due to random low-level neural processes, as would be modeled by interactive activation networks; or by higher-level restructuring of the knowledge of form-function mappings in the grammar. Either way, innovations result from speakers attempting to conform to convention (Croft 2000: 118).

Since there is a chance of innovation with each utterance, there is always a certain probability that even in an idealized test setting of absolutely identical speech situations the linguistic output varies. Thus, a one-hundred-percent reproducibility rate is never guaranteed. Many social factors are difficult to define and they are subject to change in the course of time. Both of these facts make it hard to replicate

250

Walter Bisang

social factors for testing reproducibility. While it is possible to define the social status of the participants in a speech situation with some precision (but cf. problems from the perspective of social practice in subsection 2.1), it is much more difficult to provide details on how to specify criteria of identity and to describe membership in a community of practice and the exact network structure in which it is reflected. And even if it is possible to convene the same individuals to the same speech situation for the purpose of reproducibility, some of them may have changed their social position or they may be in a different mental state with different views about their identities and their speech-community memberships.

3.

From grammars to typology

3.1.

Grammars

This subsection starts with a general description of different methods of collecting data. Intuition-based, corpora/text-based and usage-based approaches will be examined with regard to the extent to which their data are reproducible (on a comparison of usage-based and rule-based approaches and their value in accounting for the reduction and the deletion of unstressed vowels in Dutch, see Hinskens, this volume). As will become evident, the social factors and the factors of form-function mapping also apply at the level of individual grammars (1ii) (hypothesis 3). Since grammatical descriptions of individual languages are the basic source for collecting typological data, the end of this subsection will address the specific problems of reproducibility in grammars of languages that are rarely addressed in the literature. The distinction between intuition-based data gathered by elicitation and text-based data extracted from text corpora has achieved textbook status in linguistic discussions on methodology. As will be shown, both methods have their difficulties integrating social factors. Elicitation is the main source of information in formal approaches. The cognitive-categorial knowledge of language, i. e. Universal Grammar, is primarily accessible through the intuitions of linguists and other informants about sentences that are well formed or grammatical in their mother tongue. For a long time, acceptability judgments were mainly based on the binary distinction between [+grammatical] and [-grammatical]. In this context, Belletti and Rizzi’s (1988) seven-point scale for acceptability may be taken as a rather positive exception even though it has often been criticized for its lack of an explicit definition of degrees of grammaticality. As was pointed out by

Variation and reproducibility in linguistics

251

Schütze (1996: 43), “[T]he notion of relative and absolute badness of particular violations is ad hoc, and is used in just those cases where it is convenient”. More recent methodological approaches like Sorace and Keller (2005) and Featherston (2007) try to improve intuition-based judgments by integrating them into experimental settings. Sorace and Keller (2005) operate with acceptability scales in which informants have to evaluate the degree of unacceptability of linguistic examples. As their experiments show, there is a difference between hard constraints “that cause strong unacceptability when violated” and soft constraints “whose violation causes only mild unacceptability” (Sorace and Keller 2005: 1503). These two types of constraints differ with regard to their immunity to context effects and developmental optionality in first and second language acquisition. In addition, their degree of strength is universal, i. e., there are no constraints that are soft in one language while they are hard in another. Unfortunately, Sorace and Keller (2005) do not say much about the social properties of their informants. This leaves open the question of whether the data are really hard or whether they are due to the selection of a socially rather homogeneous group of informants. Apart from that, the experiments are only based on the two languages of English and Modern Greek. Featherston (2007) also supports experimental data gathering and different degrees of well-formedness. In his view of what is a minimum standard, linguistic data must be checked by a group of twenty to thirty informants and they must work with about ten different lexical forms (multiple lexical variants) that show the same effect. While Sorace and Keller (2005) do not address social factors, Featherston (2007) does. Referring to Chomsky’s (1965) definition of the ideal speaker-listener (cf. subsection 2.1), he points out the necessity of controlling such phenomena as individual deviations from the norm, personal quirks in usage and dialect variation. As he argues, the use of multiple informants can resolve this problem: “There is no such thing as an ideal speaker-listener in one person, but the mean of a reasonably sized sample has exactly the characteristics one would expect an ideal speaker-listener to have.” (Featherston 2007: 29). As was pointed out in the quotation from Chomsky (1965: 3) in subsection 2.1, the concept of the ideal-speaker-listener is an abstraction that is based on “a completely homogeneous speech-community”. Since Featherston’s (2007) approach does not consider social factors that may trigger the preference of a certain linguistic structure over another one, nothing can be said to what extent the “mean of a reasonably sized sample” really represents a socially homogeneous speaker-listener. As long as it is not experimentally checked, it is simply impossible to know to what extent socially induced variaton would affect formal generalizations.

252

Walter Bisang

Even if the social perspective is disregarded, it is highly questionable whether Featherston’s (2007) statistical “mean” can be immediately associated with processes that take place in an individual’s brain with its innate language capacity. The statistical results produced by Featherston’s (2007) sampling method may be due to such a large number of different processes in the brain that no reliable conclusions can be drawn. In fact, strict grammaticality judgments that reflect the manifestation of UG in an individual’s cognition are no longer accessible in such a statistical approach. Finally, there are two more problems of very different order that apply to both papers, that of Sorace and Keller (2005) and that of Featherston (2007). As has been shown in subsection 2.1 and as has been pointed out by sociolinguists like Labov (1966), one and the same speaker does not always use the same structures to express a certain concept. This type of knowledge is not discussed at all in the above proposals for the improvemt of the quality of data gathering. The second problem is even more fundamental since it addresses the questions of how reliably acceptability judgments reflect grammar. Bornkessel-Schlesewsky and Schlesewsky (2007) point out that there is no neurological evidence for a direct correlation between judgments and grammatical knowledge: “In fact, using acceptability ratings as predictors for neuronal activity does not reveal a correlation between these ratings and any region within the neural network associated with language processing [...].” (Bornkessel-Schlesewsky and Schlesewsky 2007: 329). The representatives of corpus-based methods heavily criticize the exclusively intuition-based methodology of formal linguistics. The following quotation is a good example: [G]enerative grammar has produced many explanatory hypotheses of considerable depth, but is increasingly failing because its hypotheses are disconnected from verifiable linguistic data. Issues of frequency of usage are by design made external to matters of syntax, and as a result categorical judgments are overused where not appropriate, while a lack of concern for observational adequacy has meant that successive versions have tended to treat a shrinking subset of data increasingly removed from real usage (Manning 2003: 296).

In Manning’s (2003) view, formal approaches claim too much because, on the one hand, they miss statistically rare cases that can be found by going through large linguistic corpora and, on the other hand, they do not explain enough because they do not account for weak constraints that specify how individuals select from a number of different structural options. While it is certainly true that intuition strongly tends to overlook

Variation and reproducibility in linguistics

253

a number of structural options within a certain grammatical domain that can be discovered by the analysis of corpora, the extent to which corpora can contribute to the understanding of how speakers select the structures they actually use in a given speech situation is rather questionable. As we have seen in subsection 2.2, the form-function mapping may well be subject to unmotivated and unpredictable innovations. In addition and more importantly, corpora do not support any statements concerning selection processes based on social factors simply because the texts they offer do not provide any details about the social background of their authors and the social situation for which they have been produced. Usage-based approaches even deny the existence of a clear distinction between knowledge of language and use of language: “[E]mergent structures are unstable and manifested stochastically [...]. From this perspective, mental representations are seen as provisional and temporary states of affairs that are sensitive, and constantly adapting themselves, to usage. ‘Grammar’ itself and associated theoretical postulates like ‘syntax’ and ‘phonology’ have no autonomous existence beyond local storage and real-time processing [...].” (Bybee and Hopper 2001: 2–3). This type of usage-based approach is radically corpus-based in the sense that it only accepts data that are attested in concrete texts and explicitly excludes elicited data. As a consequence, linguistic generalizations are only valid if they are derived from texts. There can be no doubt that such a restriction to empirical facts considerably increases the authenticity of the data. The problem with this usage-based empiricism is reproducibility. With the exception of certain ritual texts that belong to the cultural memory of a speech community and are known by heart by at least some of its members, texts are unique in the sense that they cannot be reproduced. They have been uttered once for a specific purpose and that’s it – it is extremely unlikely that they will be produced in identical form for a second time. A look at different methods of data gathering from the perspective of reproducibility reveals interesting similarities between radical contextfree and radical usage-based approaches, since both of them defy reproducibility. At the level of idealization, Chomsky’s (1965) ideal speakerlistener is not anchored in a concrete speech situation and thus does not allow the establishment of an experimental setting that has the potential of inducing an identical linguistic structure. At the level of strict empiricism, the examples that constitute the data basis are part of a text that is unique and thus is beyond reproducibility. Between these two poles, there are many approaches that can be studied in the light of reproducibility. Unfortunately, all of them are problematic if it comes to the integration of the social factors that play an important role for the definition of the

254

Walter Bisang

speech situation in which a certain linguistic structure is expected to be reproduced. The methodological approaches discussed so far are developed from languages with a rich tradition of analysis like English. Of course, it is also possible to apply them to hitherto undescribed languages, at least in principle. The problem is that linguists who produce a grammar of a language that has not been described so far work alone or sometimes in a small team and have to collect all their data in fieldwork. Thus, they simply do not have the capacity to meet all the methodological conditions required by any of the above approaches. To mention just a few rather trivial examples, the text corpus is always too small to seriously compete with say corpora of English, there is not enough time to check each sentence with twenty to thirty informants (cf. Featherstone 2007), systematic studies on the social basis of variation are too time-consuming and experimental studies may be done for a few grammatical phenomena but certainly not for the whole range of structures described in a fully fledged reference grammar. To minimize wrong analyses and to optimize the discovery of the categories and rules that constitute the grammar of a language, grammarians usually combine text-based research and elicitation. Given the limitations of the resources available, this method is very suitable to show whether a certain regularity inferred from text analysis is really stable and thus increases the chance that it is reproducible. Of course, the general problems related to the social factors and to the stochastic nature of the formfunction mapping (cf. subsection 2.2) cannot be eliminated. The social factor produces some effects that are important for the role of individual grammars in typological studies and will be discussed before this subsection will be concluded. Languages are usually characterized by different varieties that depend on social stratification and areal division into dialects (cf. subsection 2.1). Since an individual grammarian will not be able to cover all the varieties of the language to be described, there are basically two options. The grammar is limited to a single variety, say a single dialect of that language, or it represents some compromise across a number of varieties. The first method based on a single variety is implicitly applied by a large number of grammarians who only mention a small number of informants with whom they cooperated closely and regularly in their acknowledgments. It has the advantage that the grammar reflects a more or less homogeneous variety as it may be produced by an individual’s brain and as it is needed for cognitive models. From a typological perspective, such a grammar has the disadvantage that it may not present the whole structural variation that actually exists in the language to be

Variation and reproducibility in linguistics

255

described. The second method based on combining different varieties may convey a broader range of language-internal variation and thus be typologically more attractive but too much compromising across varieties may detract from what can plausibly be attested in an individual’s brain. Linguistic variation is also related to the spoken/written dimension. This dimension partially overlaps with social factors in the sense that there is usually a certain variety of a language that provides the written standard and excludes other varieties from achieving the same status. In addition, reading and writing depend on specific cognitive abilities, which may favor certain linguistic structures to the detriment of others. They may even generate additional structures that are not present in the vernacular languages. Grammars that are based on written varieties may thus miss a considerable number of structures that did not make their way into the written form of a language or they may overemphasize certain structures that are due to the cognitive properties of reading and writing. Finally, grammar writers come from a literate background and they may unconsciously transfer that perspective onto the data they select when they describe a language whose speakers have no writing tradition. It may thus come as no surprise that many grammars show a preference for narrative texts over conversations. As was pointed out recently by Foley (2003), “[i]t is truly remarkable how most text collections in the languages of traditional communities around the world have such a high percentage of narratives, sometimes only these”.

3.2.

Typology

Functional and social factors must also be taken into consideration at the level of typology (1iv) (hypothesis 3). The case of form-function mapping will only be mentioned briefly because its overall impact on typological generalizations seems to be minimal. In spite of this, it cannot be excluded that some neural properties of the brain or the difficulty to comply with a large number of rules almost simultaneously may trigger a structure which defies typological generalizations in a single speech situation in a language which normally fully harmonizes with typological findings. If this happens, this is of course an instance of failed reproducibility. Of more potential importance are social factors. They may have an impact on the reproducibility of cross-linguistic generalizations in at least two ways. The first has to do with the definition of language vs. dialect, the second is linked to the problem of whether the distribution of types in the present population of the world’s languages is neutral (Maslova 2000).

256

Walter Bisang

The fact that the degree of structural divergence across varieties cannot be predicted from their designation as languages or dialects (cf. subsection 2.1 on the definition of language vs. dialect) has its consequences for variety samples as well as for probability samples (for the terminology, cf. Rijkhoff et al. 1993 and Rijkhoff and Bakker 1998; for the discussion of this problem, cf. Bisang 2004). For a variety sample as the one developed by Rijkhoff et al. (1993) and Rijkhoff and Bakker (1998) the problem starts from the fact that it is far from clear for a number of geographic areas whether the different varieties spoken in them are to be classified as languages or as dialects. What quite often seems to be the case is a situation similar to a dialect continuum with a decreasing number of shared features from one end of that continuum to the other but with insufficient information concerning the language loyalties within the different varieties. In such a situation, assumptions about how many varieties are dialects and how many of them are subsumed under one language are rather arbitrary but they may well have an impact on the size of language families and thus on the overall statistical balance of the variety sample. From the perspective of reproducibility the interesting question in this context is whether and to what extent another variety sample based on partially different dialect/language distinctions will produce a similar range of variation. This question can also be asked with regard to the overall assumption that the complexity of the genetic tree structure within a family (for details, cf. the formula for calculating the diversity value of a family in Rijkhoff and Bakker 1998: 271) is a faithful indicator of structural diversity. Intuitively, I would not deny that genetic relations are of some importance. However, given the lack of a direct correlation of structural similarity at the lower level of dialect vs. language, one has to ask to what extent such a correlation may hold at the higher level of family relations across languages (even in an ideal situation which disregards the potential of contact-induced change across family boundaries like the one suggested by Rijkhoff and his colleagues). A probability sample, as the one developed by Dryer (1992), has to face the problem that it may not cover the entire potential of structural variance within a genus. As was pointed out in subsection 3.1, a large number of fieldworkers writing a grammar for good reasons somehow concentrates on a certain variety. As a consequence, there is a considerable potential for structural variance that remains inaccessible to the typologist. A similar problem also applies to well-studied languages such as English but for other reasons. In these cases, typologists basically look at the standard varieties without being aware of the sometimes considerable structural variance attested within the dialects of that language (for more

Variation and reproducibility in linguistics

257

on this topic cf. the volume edited by Kortmann 2004). Since the standard variety also corresponds to the written variety, this problem is also partially related to the spoken/written dimension. Nobody can say today to what extent this will affect the overall statistics of presently existing probability samples. But let’s assume for a moment that for each grammar of a language used in a given probability sample there is another grammar describing another variety of the same language and that this other grammar will be used in a test of reproducibility. I won’t dare any predictions. Maslova (2000) rightly points out that typologists cannot say whether the distribution of types in the present population of languages is neutral, i. e. unbiased by socio-cultural factors of contact-induced transfer, because they lack information on the death and birth of languages and the types attested in them and because they lack enough diachronic data on type shifts. As a consequence, there is no way to “distinguish between general distributional universals and accidental statistical properties of the current language population” (Maslova 2000: 307). To overcome this problem, Maslova suggests a method that integrates estimates of transition probability from one type to another type within a certain period of time (T stands for “Type”): It is clear [...] that if we want to estimate the probability of a shift Ti —> Tj we must compare two quantities: the number of languages which have undergone this shift and the number of languages which have retained type Ti or shifted to another type within the same time interval. Accordingly, in order to estimate a transition probability p(Ti —> Tj) for some time interval t, one would need a sample of languages which can be assumed to have been in a state Ti t years ago. The current frequency of type Tj in this sample would give an estimate of the transition probability p(Ti —> Tj) (Maslova 2000: 329).

This method tacitly assumes that the relative time interval needed for type shifts is in principle predictable due to some cognitive reasons. However, if the propagation of a type shift can also be part of social factors such as speaker’s attitudes towards different varieties involved in situations of contact or due to maxims of language use (Keller 1990, 1994) the time interval within which a type shift takes place is potentially arbitrary, i. e., cognitive or discourse motivations alone do not allow fully reliable generalizations about what tends to be a short-term shift and what tends to be a long-term shift. In a reproducibility test with two different samples covering the same type shift from Ti to Tj, a non-identical result concerning transition probability is thus to be expected at least to some extent because of the social factor. In addition, the data on type shift are necessarily based on written sources as soon as we go back in time. This leads at least to the following two complications. A given type may still have been present at a certain period

258

Walter Bisang

but it may not be attested because its presence was limited to non-written varieties. A shift from one type to another may be due to the fact that a new generation of literate speakers belonged to another dialect group. The only situation that is of potential theoretical value is the situation in which the statistical distribution of the values for the shifts Ti —> Tj within a sample is random. In such a case, it is relatively plausible that the statistical distribution of the two types of Ti and Tj is neutral. The problem with this conclusion is that the number of languages on which there is data available with the relevant diachronic time-depth is relatively small in comparison with the vast majority of languages that are only attested synchronically. Thus, it cannot be fully excluded that a random distribution of type shifts within a sample is accidental because of the limited number of languages that are accessible.

4.

Conclusion

The present paper has confirmed the three hypotheses presented in subsection 1.4. Since reproducibility is always tested in concrete speech situations, the functional and social factors defined in section 2 are relevant for all linguistic approaches (hypothesis 1). Even if social factors do not matter to a theory, they cannot be eliminated from the experimental setting because they are always part of it. Given the stochastic nature of the form-function mapping and the difficulties to provide exact definitions for many social factors, waterproof reproducibility is unattainable for language data and thus for linguistic findings in general (hypothesis 2). Finally, the functional and the social factors that determine the speech situation (1i) also operate at the level of individual grammars (1ii) and cross-linguistic generalizations (1iv) (hypothesis 3). Experimental settings for grammaticality judgments in terms of intuition-based elicitation are always influenced by functional and social factors, while corpora cannot provide data that guarantee reproducibility because their texts are not classified according to social properties of their authors and the situation in which they are used. In the case of typological generalizations, the social factors are more significant than functional factors. The problem of the language/dialect-distinction may have its impact on the overall statistical balance within a variety sample and it may lead to the inaccessibility of language-internal variation if a grammar concentrates on a particular variety. Finally, the impact of social factors on processes of type shift (Maslova 2000) is often unknown. It is thus questionable to what extent sampling of diachronic changes within languages can produce reliable

Variation and reproducibility in linguistics

259

insights into the statistical neutrality of the diffusion of types within the population of the world’s languages. To conclude this section, I would like to add a remark on the status of reproducibility and on the problem that exact reproducibility is not possible in linguistics. It is probably needless to say that the relevance of falsification (and reproducibility) is far from being uncontroversial in linguistics. Chomsky for instance argues against its relevance: People talk about Popper’s concept of falsification as if it were a meaningful proposal to get rid of a theory: the scientist tries to find refuting evidence and if refuting evidence is found then the theory is given up. But nothing works like that. If researchers kept to those conditions, we wouldn’t have any theories at all, because every theory, down to basic physics, is refuted by tons of evidence, apparently. So, in this case, what would refute the strong minimalist thesis is anything to look at. The question is, as in all these cases, is there some other way of looking at the apparently refuting phenomena, so as to preserve or preferably enhance explanatory power, where parts of the phenomena fall into place and others turn out to be irrelevant, like most of the phenomena of the world, because they are just the results of the interactions of too many factors? (Chomsky 2002: 124)

If a theory makes such a strong claim as the existence of an autonomous and innate language faculty, it should also care for presenting some criteria of reproducibility. This means that it should at least in some limited domains try to disentangle the different factors (e.g. cognitive and social factors) involved in the production (and the perception) of utterances in a way that clearly points at the existence of something substantial that is being left for Universal Grammar to explain. Of course, this implies the integration of performance at least as far as the impact of the factors that determine it are concerned (also cf. the quotation of Frisch 1999 in subsection 1.2). Even if exact reproducibility is a problem in linguistics, I think it is necessary to maintain reproducibility as a criterion for evaluating linguistic generalizations in principle because it helps assessing their overall status and keeps linguists from premature conclusions concerning the properties of language. Although nothing absolutely certain can be said about the reliability of linguistic generalizations I basically share Newmeyer’s (1998) optimism when he states that “[t]he best strategy is to assume, for want of evidence to the contrary, that the more robust-seeming of them [i. e., typological generalizations; W. B.] are valid and to persue whatever combination of formal and functional explanatory mechanisms seem appropriate to explaining them.” (Newmeyer 1998: 364). From what I have said in this paper, I would of course also say the same about the robustness of formal theories.

260

Walter Bisang

References Belletti, Adriana and Luigi Rizzi 1988 Psych-Verbs and T-Theory. Natural Language and Linguistic Theory 6: 291–351. Bisang, Walter 2001 Areality, grammaticalization and language typology. On the explanatory power of functional criteria and the status of Universal Grammar. In: Walter Bisang (ed.), Language Typology and Universals, 175–223. Berlin: Akademie-Verlag. Bisang, Walter 2004 Dialectology and typology – an integrative perspective. In: Bernd Kortmann (ed.), Dialectology Meets Typology. Dialect Grammar from a Cross-Linguistic Perspective, 11–45. Berlin: Mouton de Gruyter. Bornkessel-Schlesewsky, Ina and Matthias Schlesewsky 2007 The wolf in sheep’s clothing: Against a new judgement-driven imperialism. Theoretical Linguistics 33: 319–333. Brown, Roger and Gilman, Albert 1960 The pronouns of power and solidarity. In: Thomas A. Sebeok (ed.), Style in Language, 253–275. Cambridge, M.A.: MIT Press. Bybee, Joan L. and Paul J. Hopper 2001 Introduction to frequency and the emergence of linguistic structure. In: Joan L. Bybee and Paul J. Hopper (eds.), Frequency and the Emergence of Linguistic Structure, 1–24. Amsterdam: Benjamins. Chambers, J. K. and Peter Trudgill 1998 Dialectology. 2nd edition. Cambridge: Cambridge University Press. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, Mass.: MIT Press. Chomsky, Noam 1986 Knowledge of Language. Its Nature, Origin, and Use. New York: Praeger. Chomsky, Noam 1995 The Minimalist Program. Cambridge, Mass.: MIT Press. Chomsky, Noam 2002 On Nature and Language. Cambridge: Cambridge University Press. Croft, William A. 1995 Autonomy and functional linguistics. Language 71: 490–532. Croft, William A. 2000 Explaining Language Change. An Evolutionary Approach. Edinburgh: Pearson Education. Dik, Simon 1997 The Theory of Functional Grammar. Part 1, the Structure of the Clause. 2nd, revised edition, Kees Hengeveld (ed.). Berlin: Mouton de Gruyter.

Variation and reproducibility in linguistics

261

Dryer, Matthew S. 1992 The Greenbergian word order correlations. Language 68: 81–138. DuBois, John W. 1987 The discourse basis of ergativity. Language 63: 805–855. Eckert, Penelope 2000 Linguistic Variation as Social Practice: The Linguistic Construction of Identity in Belten High. Oxford: Blackwell. Featherston, Sam 2007 Data in generative grammar: The stick and the carrot. Theoretical Linguistics 33: 269–318. Fitch, Tecumseh W., Marc D. Hauser and Noam Chomsky 2005 The evolution of the language faculty: Clarifications and implications. Cognition 9: 179–210. Foley, William A. 1997 Anthropological Linguistics: An Introduction. Oxford: Blackwell. Foley, William A. 2003. Genre, register and language documentation in literate and preliterate communities. Paper presented at the Workshop on Endangered Languages and Language Documentation, March 1st, 2003, SOAS, London. Sydney: Mimeo. Frisch, Stefan 1999 Review of linguistic structure and change: An explanation from language processing by T. Berg. Journal of Linguistics 35: 597–601. Givón, Talmy 1979 On Understanding Grammar. New York: Academic Press. Greenberg, Joseph H. 1963 Some universals of grammar with particular reference to the order of meaningful elements. In: Joseph H. Greenberg (ed.), Universals of Grammar, 73–113. Cambridge, Mass.: MIT Press. Haiman, John 1983 Iconic and economic motivation. Language 59: 781–819. Harris, Alice C., Larry M. Hyman, and James V. Staros 2006 What is reproducibility? Linguistic Typology 10.1: 69–73. Haspelmath, Martin and Sven Siegmund 2006 Simulating the replication of some of Greenberg’s word order generalizations. Linguistic Typology 10.1: 74–82. Hauser, Marc D., Noam Chomsky and Tecumseh W. Fitch 2002 The faculty of language: What is it, who has it, and how did it evolve? Science 298: 1569–1579. Hawkins, John A. 2004 Efficiency and Complexity in Grammars. Oxford: Oxford University Press. Keller, Rudi 1990 Sprachwandel. Tübingen: Francke.

262

Walter Bisang

Keller, Rudi 1994 On Language Change: The Invisible Hand in Language. London: Routledge. (Translation and expansion of Keller 1990). Kortmann, Bernd (ed.) 2004 Dialectology Meets Typology. Dialect Grammar from a Cross-Linguistic Perspective. Berlin: Mouton de Gruyter. Labov, William A. 1966 The Social Stratification of English in New York City. Washington, D. C.: Center for Applied Linguistics. LePage, Robert B. and Andrée Tabouret-Keller 1985 Acts of Identity. Cambridge: Cambridge University Press. Manning, Christopher D. 2003 Probabilistic syntax. In: Rens Bod, Jennifer Hay and Stefanie Jannedy (eds), Probabilistic Linguistics, 289–341. Cambridge, Mass.: MIT Press. Maslova, Elena 2000 A dynamic approach to the verification of distributional universals. Linguistic Typology 4.3: 307–333. Milroy, James and Lesley Milroy 1985 Linguistic change, social network and speaker innovation. Journal of Linguistics 21: 339–384. Newmeyer, Frederick J. 1998 Language Form and Language Function. Cambridge, Mass.: MIT Press. Newmeyer, Frederick J. 2005 Possible and Probable Languages. A Generative Perspective on Linguistic Typology. Oxford: Oxford University Press. Norman, Jerry 1988 Chinese. Cambridge: Cambridge University Press. Popper, Karl 2002 The Logic of Scientific Discovery. London: Routledge Classics. [Original [1959] version: 1935. Logik der Forschung. Wien: Springer]. Rijkhoff, Jan, Dik Bakker, Kees Hengeveld and Peter Kahrel 1993 A method of language sampling. Studies in Language 17.1: 169–203. Rijkhoff, Jan and Dik Bakker 1998 Language sampling. Linguistic Typology 2: 263–314. Sapir, Edward 1921 Language: An Introduction to the Study of Speech. New York: Harcourt Brace. Schütze, Carson T. 1996 The Empirical Base of Linguistics. Chicago: University of Chicago Press. Silverstein, Michael 1976 Hierarchy of features and ergativity. In: R. M. W. Dixon (ed.), Gram-

Variation and reproducibility in linguistics

263

matical Categories in Australian Languages, 112–171. Canberra: Australian Institute of Aboriginal Studies. Sorace, Antonella and Frank Keller 2005 Gradience in linguistic data. Lingua 115: 1497–1524. Trudgill, Peter 1972 Sex, covert prestige, and linguistic change in the urban British English of Norwich. Language in Society 1: 179–196.

Parameters of morphosyntactic variation in World Englishes: prospects and limitations of searching for universals1 Bernd Kortmann and Benedikt Szmrecsanyi Abstract This paper contributes to an objectification of Chambers’ notion of vernacular universals and somewhat relativizes the predominantly critical opinions that have been voiced concerning this notion in the recent literature (especially in Filppula, Klemola, and Paulasto 2009). Focusing on the morphosyntax of non-standard varieties of English around the world (including English-based Pidgins and Creoles), it highlights the advantages this notion may offer to dialectological and variationist studies and, more broadly, to language typology. Our overall argument runs as follows: the notion of vernacular universals will be most useful (and least controversial) if we do not explore the alleged absolute universality of individual features, but rather the universality of “conspiracies” of morphosyntactic features and strategies for coding certain types of grammatical information which can be identified for individual types of varieties in any language (e.g. analytic vs. synthetic coding strategies in L1 as opposed to L2

1 This paper is dedicated to Markku Filppula (University of Joensuu, Finland) on the occasion of his 60th birthday. Markku has not only significantly advanced our knowledge about Irish English and its dialects, but also about Celtic-English language (and dialect) contacts since the Middle Ages and, most recently, about the relationship between language contact and vernacular universals. He has always understood to bring together interesting people on timely research areas for workshops and conferences in Joensuu, never failing to serve as the perfect host and to open the eyes of his guests for the beauties of Finland. As the Director of LANGNET, the Finnish Graduate School in Language Studies, he has been responsible in no small part for the excellent international standing that Finnish Ph.D. students enjoy around the globe. The authors would like to express their warmest thanks to the Freiburg Institute for Advanced Studies (FRIAS) for having given them the unique opportunity of coauthoring this paper in an intellectually most stimulating environment and an atmosphere of joyful tranquillity.

Parameters of morphosyntactic variation in World Englishes

265

varieties, low-contact vs. high-contact varieties, spontaneous spoken vs. written varieties). Keywords: analyticity, complexity, grammaticity, Greenberg, language typology, L2-complexity, L2-simplicity, morphosyntax, syntheticity, universals (absolute, statistical, implicational, vernacular), variety types

1.

Introduction

Anyone interested in linguistic typology, on the one hand, and the partnership between cross-linguistic and language-internal variation (or: macro- and microparametric variation), on the other hand, spontaneously sympathizes with the idea of searching for universals. Thus the idea of searching for universals in vernaculars, of English and other languages, as suggested by Jack Chambers in various papers (2000, 2001, 2003, 2004), also has something immediately appealing about it. However, in Filppula, Klemola, and Paulasto (2009) and in some papers in the present volume, Chambers’ notion of ‘vernacular universals’ has given rise to a controversy, with often quite critical evaluations of the concept pointing to fundamental problems or even apparent flaws. It is against the backdrop of this critical discourse that the present paper needs to be seen. Its aims are twofold: it tries to (a) contribute to an objectification of vernacular universals, thus somewhat relativizing the predominantly critical opinions which have been voiced concerning this notion in the recent literature and giving them their appropriate place, and (b) point out the advantages and value this notion may have for dialectological and variationist studies and, more broadly, for language typology. The focus in doing so will solely be the morphosyntax of non-standard varieties of English around the world (L1, L2, English-based Pidgins and Creoles). In section 2 some pros and cons of Chambers’ notion will be weighed against each other, both in the light of rather general considerations and of recent and ongoing empirical research by ourselves. The ultimate idea this section will lead up to is that there are two lines of understanding vernacular universals, which as a result inform two strands of research into large-scale morphosyntactic patterns across a large number of varieties of English, and that these, in turn, may help to give Chambers’ notion some more credit than it has received in recent accounts. While section 2 is largely concerned with individual or sets of two or three morphosyntactic features (i. e. the outcome of what we suggest to call ‘feature-based approaches’), sections 3 and 4 will present some results

266

Bernd Kortmann and Benedikt Szmrecsanyi

from our current research on large-scale cross-varietal patterns in entire coding strategies of varieties of English around the world (i. e. ‘strategybased approaches’). Our overall argument in this paper will be that the notion of vernacular universals will be most useful (and least controversial) if we do not explore the alleged universality of individual features (in the sense of absolute universals), but rather the universality of “conspiracies of morphosyntactic features” and strategies for coding certain types of grammatical information which can be identified for individual types of varieties in any language (e.g. L1 as opposed to L2 varieties, low-contact vs. high-contact varieties, spontaneous spoken vs. written varieties). In section 5, finally, we will sketch how the approach we are suggesting for determining different degrees of syntheticity and analyticity in the coding of grammatical categories among (spoken and written) varieties of English also holds potential for language typology. The systematic difference in preferred grammar coding strategies (synthetic vs. analytic) which can be identified for spontaneous spoken and written varieties in English respectively also seems to hold, as pilot studies suggest, across languages. This again would argue in favor of looking at entire coding strategies and large feature bundles rather than isolated morphosyntactic features when trying to determine candidates for structural properties which are universal in spontaneous spoken (including vernacular) language.

2.

Weighing the pros and cons of vernacular universals

Chambers’ well-known claim concerning the socio-dialectological notion of vernacular universals (or: vernacular roots) is that these universals comprise “a small number of phonological and grammatical processes [that] recur in vernaculars wherever they are spoken […] not only in working class and rural vernaculars, but also in […] pidgins, creoles and interlanguage varieties” (2004: 128). He goes on to argue that the putative ubiquity of such features, not just in varieties of English but in vernaculars of all languages, is unlikely to be due to sociolinguistic diffusion. Rather, they must be “primitive features of vernacular dialects” (Chambers 2003: 243), that is, unlearned and thus innate. Chambers (2004: 129) lists the following four morphosyntactic candidates for vernacular universals: 1. Conjugation regularization, or leveling of irregular verb forms: e.g. John seen the eclipse, Mary heared the good news. 2. Default singulars, or subject-verb non-concord: e.g. They was the last ones.

Parameters of morphosyntactic variation in World Englishes

267

3. Multiple negation, or negative concord: e.g. I don’t/ain’t know nothing. 4. Copula absence, or copula deletion: e.g. She smart, We going as soon as possible. Below we will address some pros and, especially, cons that have been voiced in the recent literature, and will seek to put these in perspective. Criticism has been leveled at Chambers’ notion of vernacular universals primarily from two angles: language typology and socio-/variationist linguistics. Let us first examine the problems that typologists (may) have with this notion.

2.1.

Some general considerations

From a typological angle there are at least five points worth making: (i) The fundamental problem typologists have with the notion of vernacular universals is that a special status is claimed for universals of vernaculars (as opposed to universals of non-vernacular varieties, especially written standard varieties of languages). Rather, the candidates for vernacular universals should form a proper subset of those universals (be they statistical or not, implicational or not) which have been identified in decades of solid, methodologically increasingly refined typological research exploring large-scale cross-linguistic variation of individual parameters of variation. Universals proper, the argument basically runs, apply to all languages and all their varieties equally, thus there is no need for postulating an extra-set of vernacular universals. In our view, there is a lot to be said in favor of this line of argumentation. At the same time, however, it needs to be acknowledged that typological studies of the last 40 years have not really covered all the linguistic variation existing on the globe. Especially for languages with a long literary tradition (such as the synchronically and historically extensively welldescribed major European languages), there has been a bias towards the written standard varieties. Spontaneous spoken varieties have largely, until recently almost entirely, been left out of consideration.2 Judged on a global scale, this leads to the following major methodological apples-andoranges problem in language typology, which as yet has hardly been 2 For example, there is no mention of non-standard varieties in the special issue “Whither Linguistic Typology – an und für sich and in relation to other types of linguistic pursuits?” of the journal Linguistic Typology 11.1 (2007: 1-306), which celebrates the tenth anniversary of the journal.

268

Bernd Kortmann and Benedikt Szmrecsanyi

acknowledged as such by its practitioners. For the majority of languages in the world, especially for those lacking a literary tradition or even a codified written representation, spoken language data form the basis for typological descriptions and generalizations, while for most European languages written data taken from the standard varieties form the object of typological study. So linguistic typology would definitely benefit from the systematic inclusion of vernaculars, at least for languages whose written standards have exclusively been taken to represent them in, for example, such magnificent typological undertakings as the World Atlas of Language Structures (WALS; Haspelmath et al. 2005). This would offer the opportunity to make the methodology of typology and the resulting typological claims (e.g. when formulating generalizations in the form of universals) more watertight. As has variously been argued and shown for English, for example, Standard British English is the “odd man out” in several domains of morphosyntax (e.g. negation, agreement, relativization, reflexives) compared with the vast majority of (standard and) non-standard varieties of English (cf. e.g. Kortmann et al. 2005). Thus taking Standard British English as the representative of English in typological studies may crucially distort the picture. For relative clauses, for instance, the most frequent strategy in non-standard varieties of English spoken in the British Isles is the relative particle strategy (notably that and what as in The man that/what came in) and not the relative pronoun strategy, as claimed in the WALS. For putting the typological picture right it is especially non-standard features with a wide areal and/or social reach which will be most useful (cf. Auer 2004), i. e. exactly those which qualify as candidates for Chambers’ vernacular universals. Moreover, such candidates for vernacular universalhood may well turn out to be veritable candidates for universals proper (like the prototypical Greenbergian universal, i. e. [statistical] implicational universals). (ii) Typologists criticizing the notion of vernacular universals should also acknowledge the following: (a) So far, it has been mainstream typological practice not to include pidgins and creoles in their samples. Only now is there a research initiative headed by the Max-Planck Institute of Evolutionary Typology at Leipzig (Germany) which is working on a World Atlas of Pidgin and Creole Structures, following the model of the Leipzig-made WALS, and which ultimately pursues the question whether pidgins and creoles represent a language type of its own; (b) more generally, with its strict focus on the structure of genetically, areally, and historically unrelated languages, linguistic typology has largely left out of account the role of language contact in the formation of individual languages, and

Parameters of morphosyntactic variation in World Englishes

269

even more so the role of socially driven diffusion of individual language structures and linguistic types across languages (cf. the illuminating paper by Bisang 2004). By contrast, anyone interested in putting to test and possibly giving (more) substance to the notion of vernacular universals is bound not to neglect aspects such as those in (a) and (b). (iii) Typologists discussing and trying to evaluate Chambers’ notion in an open-minded and objective manner should not restrict vernacular universals to absolute universals, but should, for example, also inquire into the possibility of identifying implicational universals (or rather implicational tendencies). No typologist seriously expects to find, across languages, absolute universals in grammar beyond that mere handful mentioned as likely candidates in textbooks. So why should the fact that a given candidate for a vernacular universal does not occur in 100 % of all vernaculars in a given language, let alone across languages, be turned against the notion of a vernacular universal? Even if this may be a problem for Chambers himself, since – as someone thinking of universals in a formalist sense, i. e. as part of Universal Grammar, or outgrowths of the bioprogram – he formulates his criteria for vernacular universals in a most sweeping way, for a typologist it would be amazing already if a given feature can be observed in, say, 70 % or more of the languages of the world. From the point of view of the typologist investigating the world-wide spread of morphosyntactic features in varieties of English (in our study, 76 morphosyntactic features in 46 varieties of English; for details, see section 3), it therefore (a) does not come as a surprise that not a single feature is found in all varieties, and (b) it is rather astonishing to learn that the five most widely found features of English morphosyntax worldwide occur in 80–89 % of all varieties of English. By contrast, all four of Chambers’ candidates for vernacular universals (or rather ‘angloversals’) are documented in fewer varieties (70–78 %). Even multiple negation is found in no more than 76 % of the 46 varieties. The lesson, then, to be learnt from these findings is that we should interpret Chambers’ vernacular universals as having a very wide (areal and/or social) reach in individual languages, and possibly across languages, but not as absolute universals. 100 % scores are possible only if we consider subsets of nonstandard varieties. Thus it is possible to identify morphosyntactic features, including all of Chambers’ four top candidates for vernacular universals, which are part of all non-standard varieties of North America (US and Canada; something we have elsewhere suggested to call ‘areoversals’; cf. Szmrecsanyi and Kortmann 2009b), or in the vast majority of individual variety types (e.g. in L2 varieties of English, so-called ‘vario-

270

Bernd Kortmann and Benedikt Szmrecsanyi

versals’3). We will return to the importance of variety types repeatedly from here on. (iv) Implicational universals constitute the prototype of universals in functional typology. So when putting to test the notion of vernacular universals, it should be explored for them, too, whether ‘biconditional implications’ (alternatively: ‘equivalences’) or one-way implications (also known as ‘preferences’) can be identified. On the basis of the 46 nonstandard varieties of English we investigated, such dependencies can clearly be confirmed. Here are a few examples (for details cf. Szmrecsanyi and Kortmann 2009b, c). With respect to biconditional implications, the bulk of varieties of English have both ain’t as the negated form of be and ain’t as the negated form of have, or they have none of these options. In a similar vein, non-coordinated subject pronoun forms in object function (Did you get he out of bed?) and non-coordinated object pronoun forms in subject function (Us say ‘er’s dry) tend to go together, as is the case for habitual do (He does catch fish every day) and do as an unstressed tense and aspect marker (The man what did say that…). Likewise, if a variety exhibits the relative particle at, it will have the relative particle as, and vice versa. In short, pairings like the above are best seen as feature bundles instead of pairs of independent features. Such biconditional implications also depend on the variety type: so, for instance, among L1 varieties we find a correlation such that when in a variety past forms of irregular verbs can replace participle forms, unmarked forms are also possible, and vice versa. Both among L2 varieties and among English-based pidgins and creoles (but not in L1 varieties), we observe that if a variety has one use of ain’t (either for be, have, or as a generic negator), it will also have the other two uses of the form. As regards ‘one-way implications’, any variety in our sample that attests would in if–clauses also displays loosening of the sequence of tense rule, but not necessarily vice versa. Moreover, a variety will not have the relative particle as (There’s the man as kicked me in the face) unless that variety also has (i) the relative particle what (There’s the man what kicked me in the face) and (ii) gapping or zero-relativization in subject position (There’s the man ____ kicked me in the face). And once again, distinguishing between variety types is instructive, as the following two examples 3 Compare Kortmann and Szmrecsanyi (2004) for an extended empirical discussion of unrestricted angloversals, i. e. the most frequent features in varieties of English across the board (pp. 1153–1160), unrestricted areoversals (pp. 1160–1184), and unrestricted varioversals (pp. 1184–1194).

Parameters of morphosyntactic variation in World Englishes

271

show: an L1 variety, unlike L2 and Pidgin/Creole varieties, needs habitual do in its inventory in order to attest do as a tense and aspect marker, and in L2 varieties, the possibility of regularized reflexives-paradigms appears to necessitate generic he/his for all genders (as in The car, he’s broken). (v) We saw in (iii) that Chambers’ candidates for vernacular universals do not even stand up to the status of vernacular angloversals, i. e. features found in 100 % of all non-standard varieties of English. But even if we did find absolute, unrestricted angloversals, there is much room for doubt, to put it mildly, that counterparts of these vernacular universals for English were to be found in vernaculars of other languages, too, as Chambers (2004: 129) assumes: I have listed the vernacular universals with their English names and illustrated them with English examples. This is misleading. In so far as these processes arise naturally in pidgins, child language, vernaculars, and elsewhere, they are primitive features, not learned. As such they belong to the language faculty, the innate set of rules and representations that are the natural inheritance of every human being. They cannot be merely English. They must have counterparts in the other languages of the world that are demonstrably the outgrowths of the rules and representations in the bioprogram.

Considerable room for doubt there is at least as long as in searching for vernacular universals the focus is on individual, highly specific features like those given by Chambers as promising candidates at the beginning of section 2. Take, for example, copula absence, as in She___ smart or He __ a sailor. Cross-linguistically, it turns out that zero copulas for such predicate nominals are impossible in the majority of languages (211 vs. 175 where it is possible) in the sample investigated by Stassen in WALS Map 120 (see online version at wals.info). If zero copulas were part of the language faculty, there should be more languages that have them. Since this is evidently not the case, zero copulas cannot claim the status of “promising candidate for a vernacular universal”. Something similar can be observed for subject-verb non-concord (or default singulars). In order to exhibit subject-verb non-concord in any non-vacuous way, a language needs structural means to display subject-verb concord (at least historically). Yet many languages – for instance, highly isolating languages such as Vietnamese – do not show agreement at all; so vernacular Vietnamese showing non-concord is an utterly unremarkable fact. Along similar lines, the other of Chambers’ candidates for vernacular universals, too, could rather easily be attacked from a typological angle (with the possible exception of multiple negation).

272

Bernd Kortmann and Benedikt Szmrecsanyi

In short, what we mean to suggest here is that a feature like subjectverb non-concord is rather typical of languages that, like English, have some inflectional morphology but are in the process of getting rid of what remains. But what is happening in non-standard varieties of English and, possibly, languages belonging to the same morphological type as English, does not necessarily apply to vernaculars of inflectional or agglutinating languages such as Finnish and Hungarian (which have a great deal of grammatical agreement), or Turkish. A feature such as subject-verb nonconcord may be better referred to as a vernacular ‘typoversal’ – a feature, in other words, which is typical of vernaculars of inflecting languages. It is unlikely that the ‘language faculty’ would provide for special rules and representations applying to English-like vernaculars; zero copulas and default singulars are just too conditioned on the linguistic type of English to be cross-linguistically “universal”. At the same time, notice that it is not only loss of agreement or loss of redundancy that we can observe in vernaculars. Individual vernaculars have, and can indeed be shown to currently develop, a more elaborate inflectional morphology or, for example, agreement system than the standard variety has (e.g. de Vogelaer, Neuckermans and Vanden Wyngaerd 2002; Wagner 2004; Haser and Kortmann 2009). Given these observations, we have argued (Szmrecsanyi and Kortmann 2009b) that a candidate feature for a vernacular universal should, at a minimum, fulfill the following criteria: ± The candidate feature should be attested in a vast majority of a given language’s vernacular varieties. ± The candidate feature should neither be patterned geographically nor according to variety type (in the case of English: L1 (high- vs. low-contact), L2, or pidgin/creole). ± For the sake of cross-linguistic validity, the candidate feature should not be tied to a given language‘s typological make-up (inflectional, agglutinating, etc.). ± The candidate feature should be cross-linguistically attested in a significant number of the world‘s languages (especially among the vast number of languages without a literary tradition).

2.2.

The sociolinguistic and variationist perspective

It may be recalled that Chambers, in his account of vernacular universals, draws a major divide between Standard English(es), on the one hand, and non-standard varieties of English, ranging from traditional dialects to pid-

Parameters of morphosyntactic variation in World Englishes

273

gins and creoles, on the other hand. It is exactly this divide, along with the notion of vernacular universals in general, which Trudgill takes issue with in his contribution (“Vernacular universals and the sociolinguistic typology of English dialects”) to the vernacular universals debate in the volume by Filppula, Klemola, and Paulasto (2009). According to Trudgill, not enough vernacular universals have been found to make the concept fruitful and, more fundamentally, the “true typological split” among varieties of English lies elsewhere, not between vernacular and non-vernacular varieties. The major split – according to him – rather lies between highcontact and low-contact varieties of English. The former include (a) koineized non-standard urban varieties in the British Isles and colonial varieties of North America, Australasia and South Africa; (b) indigenized non-native L2 varieties like Indian English or Nigerian English; (c) shift varieties like Irish English and Welsh English; (d) English-based pidgins and creoles (as extreme cases resulting from language contact); and, notably, (e) Standard English(es). As low-contact varieties Trudgill identifies the traditional dialects of English, which are largely located in the British Isles, but also include Appalachian English or Newfoundland English. For the purposes of the present paper, the crucial point of Trudgill’s account is this: he argues that the grammars of high-contact varieties are characterized by processes of simplification as opposed to processes of complexification to be observed in the grammars of low-contact varieties. Three points follow from Trudgill’s “true typological split” among varieties of English and the way in which he motivates it. First, he attributes the central role in his typology to language contact (and along with it adult language acquisition), highlighting its impact on the structural properties of languages and their varieties (for similar views cf., for example, the papers by Siemund and Winford in Filppula, Klemola and Paulasto 2009). Second, implicitly, there is an even more fundamental split underlying Trudgill’s division of varieties of English (and essentially, given a similar socio-cultural background, the varieties of any language) into high- and low-contact varieties, namely the one between spontaneous spoken and written varieties – a point always worth remembering in dialectology, variationist studies, and especially in language typology. Third, Trudgill does not base his claim on individual morphosyntactic features but on what we prefer to call overall coding strategies (e.g. inflectional coding of grammatical information as a complex[ifying] strategy and analytic, or even zero, marking as simplifying strategies). Trudgill’s hypothesis concerning the significance of variety types for such overall grammar coding strategies as simplicity/simplification and complexity/complexification will be tested – and ultimately confirmed – in

274

Bernd Kortmann and Benedikt Szmrecsanyi

section 4. Systematic differences in the grammar coding strategies used in spontaneous spoken as opposed to written registers in English (and other languages) will be tracked in section 5. Before, however, it will be demonstrated in section 3 that even on a more general level, i. e. when not specifically exploring different degrees of morphosyntactic complexity and simplicity, the morphosyntax of non-standard varieties of English around the world clearly patterns according to variety type. Moreover, it will be shown that the clustering of varieties according to variety type has more explanatory power than geography.

3.

Strategy-based approaches I: the significance of variety types

In this section, we will briefly demonstrate that in accounting for shared morphosyntactic features and, indeed, feature bundles and entire coding strategies, it is the variety type (L1 vs. indigenized non-native L2 vs. pidgins and creoles) that is of towering importance. The following is a brief sketch of previous research (cf. Szmrecsanyi 2009; Szmrecsanyi and Kortmann 2009b, 2009c) where we used Principal Component Analysis to see the wood for the trees in the survey data we had collected in Kortmann and Szmrecsanyi (2004). Let us first review the nature of the survey data and the composition of our sample of varieties. The source for our survey data, i. e. the classic data type in typological and dialectological research, is what we have informally come to call The World Atlas of Morphosyntactic Variation in English, i. e. the survey of morphosyntactic features underlying the interactive maps on the CDROM accompanying the Handbook of Varieties of English (Kortmann et al. 2004) and subjected to a first close examination in Kortmann and Szmrecsanyi (2004). For this survey, material was collected from (often nativespeaker) experts on 76 non-standard morphosyntactic features from 46 (exclusively spoken) non-standard varieties of English around the world (for details of the survey procedure cf. Kortmann and Szmrecsanyi 2004: 1142–1145). The features in the survey are numbered from 1 to 76 and cover 11 broad areas of morphosyntax: pronouns, the noun phrase, tense and aspect, modal verbs, verb morphology, adverbs, negation, agreement, relativization, complementation, discourse organization and word order. This, for example, is the complete set of agreement features in the survey (including the feature numbering): 53.

invariant present tense forms due to zero marking for the third person singular (e.g. So he show up and say, What’s up?)

Parameters of morphosyntactic variation in World Englishes

275

54.

invariant present tense forms due to generalization of third person -s to all persons (e. g. I sees the house.)

55.

existential/presentational there’s, there is, there was with plural subjects (e. g. There’s two men waiting in the hall.)

56.

variant forms of dummy subjects in existential clauses (e. g. they, it, or zero for there)

57.

deletion of be (e. g. She ___ smart)

58.

deletion of auxiliary have (e. g. I ___ eaten my lunch.)

59.

was/were generalization (e. g. You were hungry but he were thirsty, or: You was hungry but he was thirsty.)

60.

Northern Subject Rule (e. g. I sing [vs. *I sings], Birds sings, I sing and dances.)

The 46 varieties are taken from all seven anglophone world regions (the British Isles, America, Caribbean, Australia, Pacific, South/Southeast Asia, Africa). Table 1 provides a breakdown of the 46 varieties by variety type (20 L1 varieties, 11 L2 varieties, 15 English-based pidgins and creoles), which for the L1 varieties includes Trudgill’s split between high-contact L1 varieties (12 out of 20) and low-contact varieties (8 out of 20; more on the high- vs. low-contact distinction in section 4.1): Table 1: Varieties sampled in the World Atlas (Kortmann et al. 2004). varieties

variety type

Orkney and Shetland, North, Southwest and Southeast of England, East Anglia, Isolated Southeast US E, Newfoundland traditional L1 E, Appalachian E Scottish E, Irish E, Welsh E, Colloquial American E, Ozarks E, Urban African-American Vernacular E, Earlier Africanhigh-contact American Vernacular E, Colloquial Australian E, Australian L1 Vernacular E, Norfolk, regional New Zealand E, White South African E Chicano E, Fiji E, Standard Ghanaian E, Cameroon E, East African E, Indian South African E, Black South African E, Butler E, Pakistan E, Singapore E, Malaysian E

L2

Gullah, Suriname Creoles, Belizean Creole, Tobagonian/ Trinidadian Creole, Bahamian E, Jamaican Creole, Bislama, Solomon Islands Pidgin, Tok Pisin, Hawaiian Creole, Aboriginal E, Australian Creoles, Ghanaian Pidgin E, Nigerian Pidgin E, Cameroon Pidgin E

P/C

276

Bernd Kortmann and Benedikt Szmrecsanyi

0.8

When applying Principal Component Analysis (PCA) to this data set in order to explore the distribution of the 76 morphosyntactic features across the 46 varieties of English, we find that co-presence and co-absence patterns point to two highly explanatory dimensions (accounting for 38.4 % of the total variance, which is quite remarkable). As can be seen in Figure 1, our pre-established 46 varieties cluster very nicely according to whether they are L1 varieties (represented by squares), L2 varieties (represented by triangles), or English-based pidgins and creoles (represented by circles) – and indeed better than geographically. Thus variety type turns out to be the better predictor of overall similarity or distance between individual varieties than the world region where they are spoken.

Norfolk

NigP

BelC

0.6

0.7

GhP

Bislama

ButlE

0.5

SolP

1: pidgins

HawC SgE JamC

AbE

– 0.3 – 0.2 – 0.1 0 – 0.38

3: L2 varieties FijiE

SurCs

Tob/TmC Gullah EAfE CamP

GhE

CamE

PakE Urban AAVE CollAusE

BahE

Earlier AAVE IrE

CollAmE OzE AusVE NZE NfldE North ScE

WelE

AppE WhSAFE

L1 L2 P/C

ChcE

InSAfE

MalE BlSAfE

0.1

component 2 0.2 0.3 0.4

2: creoles

AusCs

TP

Orkney Shetland East Anglia Southwest

4: L1 varieties

IsSE

Southeast

– 0.28

– 0.18

– 0.08

0.02

0.12

0.22

0.32

0.42

0.52

0.62

0.72

component 1

Figure 1: Visualization of principal components of variance in the 76 × 46 database. Dotted boxes indicate group memberships (cf. Szmrecsanyi and Kortmann 2009b,c).

This significant result of our global study on English leads us to hypothesize that the importance of variety type in accounting for similarities and differences in the morphosyntax of non-standard varieties will be found confirmed when exploring morphosyntactic similarities and differences and, inspired or challenged by Chambers, looking for universal features in vernaculars of other languages, too. Moreover, we hold that the relevant feature bundles for each of the variety types (i. e. at its crudest: L1 vs. L2

Parameters of morphosyntactic variation in World Englishes

277

vs. pidgins/creoles), regardless of what the individual features look like for a given language and its non-standard varieties, ultimately all “conspire” and jointly instantiate an overall coding strategy which is constitutive of the variety type at hand. These overall coding strategies can be seen to be instantiated by the two dimensions (components 1 and 2) that PCA yields in Figure 1. How, by the way, are these dimensions to be interpreted? Encouraged by an independent study (cf. Kortmann and Szmrecsanyi 2009) in which we explore different degrees of morphosyntactic complexity in non-standard varieties of English on the basis of exactly those survey data used for the PCA above, we are inclined to interpret component 1 as increasing degrees of L2-acquisition difficulty and component 2 as increasing degrees of transparency (i. e. regularity for synthetic markers of grammatical information). Notice, now, that in a different large-scale cross-varietal study (Szmrecsanyi and Kortmann 2009a), we explore two other global coding strategies (namely, analytic vs. synthetic coding of grammatical information) on the basis of naturalistic corpus data for a smaller set of spontaneous spoken varieties (all L1 or L2) – and again, variety type turns out to be the best predictor for a given variety’s morphosyntactic profile. This particular study will be summarized in the following section.

4.

Strategy-based approaches II: analyticity vs. syntheticity in varieties of English around the world

Our findings in this section are based on the dataset and method utilized in Szmrecsanyi and Kortmann (2009a). A total of three frequency-based metrics will be introduced here, which are applied to 15 varieties of English on the basis of naturalistic corpus data. The bulk of our corpus data stems from two major digitized speech corpora (the Freiburg Corpus of English Dialects (FRED; cf. Hernandez 2006; Anderwald and Wagner 2007; Kortmann and Wagner 2005) and the International Corpus of English (ICE; cf. Greenbaum 1996). From these two corpora we sampled 12 spoken varieties: two high-contact varieties (Welsh English and New Zealand English), five British low-contact varieties (Southeast England + East Anglia, Southwest England, English Midlands, English North, Scottish Highlands) and five L2 varieties (Hong Kong English, Philippine English, Singapore English, Indian English, Jamaican English). In addition, we used data from the Northern Ireland Transcribed Corpus of Speech (NITCS) to represent Northern Irish English, another high-contact variety. Purely for benchmarking purposes, we also included data

278

Bernd Kortmann and Benedikt Szmrecsanyi

from two high-contact standard varieties of British English (from the ICE-GB) and American English (from the Corpus of Spoken American English). Note that pidgins and creoles are not represented in these samples. Before turning to the frequency-based metrics, it is necessary to make clear (a) on what basis we have classified a given variety of English as belonging to one of three variety types high-contact L1, low-contact L1, and L2 (section 4.1), (b) what is understood by analyticity and syntheticity in our study (section 4.2), and (c) how we calculate degrees of analyticity and syntheticity in our quantitative analysis (section 4.3).

4.1.

Classification of varieties

Our classification of varieties (in this sample and also in Table 1 above) into L2 varieties and, much more controversial, high- vs. low-contact L1 varieties rests on the following assumptions (on this issue, cf. also Szmrecsanyi 2009). L2 varieties of English are non-native, indigenized varieties that do not have significant numbers of native speakers but that nonetheless have prestige and important normative status in certain political communities (e. g. Indian English or Jamaican English). As for L1 varieties, the distinction between high- and low-contact varieties goes back to Peter Trudgill (2009), who does not, however, provide hard-and-fast criteria for classifying a given L1 variety of English as either high- or low-contact. As outlined in section 2.2, Trudgill’s idea, in a nutshell, is that contact implicates adult language learning, which in turn implicates simplification of grammars, especially by reduction of inflectional morphology (for a recent similar view, cf. Wunderlich 2008: 252). The resulting simplicity in different domains (see Trudgill 2009 for an overview) is what sets apart high-contact from low-contact varieties as synchronic groups. In this spirit, we operationally define ‘high-contact L1 varieties’ as varieties that fall into one of the following categories:4 ± Transplanted L1 Englishes or colonial (standard) varieties (cf. Mesthrie 2006: 382), i. e. varieties whose genesis is such that thanks to settlement colonization in the course of the past 400 years, settlers with diverse linguistic and/or dialectal backgrounds – with all the dialect and lan-

4 Note however that these categories are not mutually exclusive. For instance, New Zealand English is a transplanted L1 variety, yet it also serves as a standard variety.

Parameters of morphosyntactic variation in World Englishes

279

guage contact that this implicates – formed new indigenized English dialects that have had native speakers from early on. Examples include New Zealand English and Australian English. ± Language-shift Englishes, i. e. varieties “that develop when English replaces the erstwhile primary language(s) of a community” and that have “adult and child L1 and L2 speakers forming one speech community” (Mesthrie 2006: 383). We also include in this department what we call shifted varieties, i. e. varieties that used to be genuine language-shift varieties within the past 400 years but which do not now have significant numbers of L2 speakers any more. A prime example is Irish English (cf. Mesthrie 2006: 383). ± Standard varieties, such as Standard British English, the genesis of which, according to Trudgill (2009), always implicates a high degree of dialect contact. Hence, high-contact L1 varieties, in our diction, are essentially ‘New Englishes’ (cf. Pride 1982; Platt, Weber and Ho 1984) minus non-native, indigenized L2 varieties minus English-based pidgins and creoles plus standard varieties. Varieties that do not fall into one of the above categories will be considered ‘low-contact L1 dialects of English’, i. e. traditional nontransplanted regional dialects which are “long-established mother tongue varieties” (Trudgill 2009). This classification scheme gives rise to the following categorization of the 15 varieties studied in this part of the paper: ± Non-native, indigenized L2 varieties: Indian E; Jamaican E; Hong Kong E; Philippines E; and Singapore E. ± High-contact L1 varieties of English: New Zealand E (by virtue of being a transplanted English and a standard variety); Standard (colloquial) BrE + Standard (colloquial) AmE (by virtue of being standard varieties); Scottish Highlands E (by virtue of being a shifted or even shift variety, with English having been introduced not before the 18th century; cf. Mesthrie 2006: 388); and Welsh E (again, by virtue of being a shift variety, with English not really having spread before the 19th or even 20th century). ± Low-contact L1 dialects of English: the traditional dialects spoken in the Southwest of England; the traditional dialects spoken in the Southeast of England; the traditional dialects spoken in the English Midlands; and the traditional dialects spoken in the North of England.

280

Bernd Kortmann and Benedikt Szmrecsanyi

4.2.

Defining analyticity and syntheticity

The terms ‘analytic’ and ‘synthetic’ have a long history (cf. Schwegler 1990, chapter 1 for an excellent overview of the rich history of thought in this area). Unfortunately, both terms also “are used in widely different meanings by different linguists” (Anttila 1989: 315), which is why a concise definition is necessary at this point. Following Szmrecsanyi and Kortmann (2009a) and Szmrecsanyi (2009), we are interested, first, in the coding of grammatical information only, which is why lexical analyticity and syntheticity will not enter into consideration here. Second, our idea of grammatical analyticity and syntheticity is a strictly formal one (and not a semantic one), which roughly follows Danchev’s notion that “[f]ormal analyticity [...] implies that the various meanings […] of a given language unit are carried by […] free morphemes, whereas formal syntheticity is […] characterized by the presence of [at least] one bound morpheme” (1992: 26). We thus operationally define grammatical analyticity and syntheticity as follows: 1. Formal grammatical ‘analyticity’ comprises all those coding strategies where grammatical information is conveyed by free grammatical markers, which we in turn define in a fairly standard way as closedclass (or: function) word tokens that have no independent lexical meaning and thus belong to one (or more) of the following word classes: determiners (e. g. who), pronouns (e. g. he), prepositions (e. g. in), conjunctions (e. g. and), infinitive markers (e. g. to), so-called primary verbs (be, have, do), modal verbs (e. g. can), and negators (e. g. not). 2. Formal grammatical ‘syntheticity’ comprises all those coding strategies where grammatical information is conveyed by bound grammatical markers, such as verbal, nominal, and adjectival inflectional affixes (e. g. past tense -ed, plural -s, comparative -er, and so on), the Saxon genitive (e. g. Tom’s house) as a clitic, as well as allomorphies such as ablaut phenomena (e. g. past tense sang), i-mutation (e. g. plural men) and other non-regular yet clearly bound grammatical markers. Our model of morphological analysis is thus, at base, an item-and-process model (Hockett 1954: 396) where grammatically marked forms are derived from simple forms via some sort of process – in our diction, via adding some sort of overt grammatical, not necessarily segmentable bound grammatical marker, be it a (regular) grammatical affix, a stem vowel change, and the like.

Parameters of morphosyntactic variation in World Englishes

281

3. Derived from these two notions is a third one, which we choose to label ‘grammaticity’ and which comprises the sum of the former two, i. e. all coding strategies where grammatical information is conveyed either by free or bound grammatical markers.

4.3.

The frequency-based metrics

Inspired by Joseph Greenberg‘s (1960) paper, “A Quantitative Approach to the Morphological Typology of Language”, we conducted a morphological/grammatical-functional analysis of random samples spanning 1,000 de-contextualized tokens (word forms) for every one of the geographic varieties sampled (see Szmrecsanyi 2009 for more detail on the robustness of the method). For each token in the database, we established ± whether the token bears a bound grammatical marker (fusional or suffixing), as in sing-s or sang; and/or ± whether the token is a free grammatical marker, or a so-called function word, belonging to a closed grammatical class as defined in section 4.2. On the basis of this analysis, we established three indices: a ‘syntheticity index’ (measuring the text frequency of bound grammatical markers per 1,000 words of running text), an ‘analyticity index’ (gauging the text frequency of free grammatical markers per 1,000 words of running text), and an overall ‘grammaticity index’ (the sum of the former two indices). The syntheticity and analyticity indices have a lower bound of 0 and an upper bound of 1,000; the grammaticity index has a lower bound of 0 and an upper bound of 2,000. Thus a, syntheticity index value of 100 would indicate that a variety attests, on average, 100 bound grammatical markers per 1,000 words of running text; an analyticity index value of 500 would indicate that a variety attests, on average, 500 free grammatical markers (i. e. function words) per 1,000 words of running text; and a grammaticity index value of 600 would indicate that a variety attests, on average, 600 overt grammatical markers, bound or free, per 1,000 words of running text.

4.4.

Results and discussion

The results of this exercise in index calculation can be summarized as follows. There is a strikingly consistent hierarchy that governs grammaticity levels: traditional L1-vernaculars > high-contact L1-vernaculars > L2-vari-

282

Bernd Kortmann and Benedikt Szmrecsanyi

eties. Hence, traditional L1-varieties exhibit the highest degree of grammaticity, L2-varieties the lowest degree, and high-contact L1-varieties cover the middle ground. This hierarchy dovetails nicely with claims, as pointed out in section 2.2 above, that a history of contact and adult language learning can eliminate certain types of redundancy, especially those found in grammatical marking (cf. for example McWhorter 2007; Siegel 2004, 2008; Trudgill 2001 and, especially, Trudgill 2009; Wunderlich 2008). L2-varieties in particular seem to follow this strategy most radically: our results suggest that L2-speakers do not generally opt for “simple” features instead of “complex” features, but rather appear to prefer zero marking over explicit marking, be it (presumably) L2-easy or complex. The interplay between syntheticity and analyticity is visualized by way of a scatterplot in Figure 2. The dialects of the Southeast and East Anglia (in the top right corner) and Hong Kong English (in the bottom left corner) are the extreme cases in our dataset. The former are both highly analytic and above-average synthetic varieties, while Hong Kong English exhibits the lowest figures for either type of grammaticity. In general, the traditional L1 vernaculars are to be found in the upper right half of the diagram (i. e. they exhibit above-average grammaticity), whereas L2-varieties are located in the lower left half, being neither particularly analytic nor synthetic. High-contact L1-varieties cover the middle ground, along with the two standard varieties (colloquial American English and spoken British English) and, interestingly, the Southwest of England. The dotted trend line merits particular attention in Figure 2: as a rather robust (R2 = 0.40) statistical generalization, this line indicates that on the inter-variety level, there is no trade-off between analyticity and syntheticity. Instead, analyticity and syntheticity correlate positively such that a variety that is comparatively analytic will also be comparatively synthetic, and vice versa. Once again, in terms of L2-varieties this is another way of saying that these tend to opt for a coding strategy of less overt marking, rather than trading off synthetic marking for analytic marking, which is purportedly L2-easy:

Parameters of morphosyntactic variation in World Englishes

283

520 Southeast (+ East Anglia)

510 500 490

Midlands spStBrE ScH North IndE NIrE PhE WelE JamE SgE Southwest NZE collStAmE

analyticity index

480 470 460 450 440 430 420

HKE

410 400 60

70

80

90

100

110

120

130

140

syntheticity index

Figure 2: Analyticity by simplicity. Dotted trend line represents linear estimate of the relationship (cf. Szmrecanyi and Kortmann 2009a).

5.

Strategy-based approaches III: analyticity, syntheticity, and the typological relevance of the distinction between spoken and written language

On the basis of the general method outlined in section 4, Szmrecsanyi (2009) surveys genre (or: register) variation in the British National Corpus with regard to analyticity and syntheticity and finds instructive differences between spoken and written genres. It turns out that, for one thing, spoken texts are significantly more analytic than written texts: the typical spoken text exhibits in excess of 50 more analytic markers per 1,000 words of running text than the typical written text. Secondly, written texts are significantly more synthetic than spoken texts, in that the former exhibit, on

284

Bernd Kortmann and Benedikt Szmrecsanyi

average, in excess of approximately 30 more synthetic markers per 1,000 words of running text than the latter. Furthermore, spoken texts exhibit significantly more grammaticity than written texts, and as far as the scope of variability is concerned, variability among written texts is more sizable than among spoken texts. In sum, variability in the analyticity-syntheticity dimension is endemic in stylistic varieties of English.

Italian

550 English 500

German

analyticity index

English 450

Russian

Italian German

400

350 spoken written

Russian

300 100

200

300

400

500

600

700

syntheticity index

Figure 3: Analyticity vs. syntheticity in written and spoken standard varieties of four European languages.

The crucial link of this sort of register variation study in English to language typology is the following. We have conducted preliminary investigations – utilizing the methodology detailed above – into written and spoken registers of four European languages (English, Italian, German, and Russian), and the results (cf. Figure 3) show that spoken-written differ-

Parameters of morphosyntactic variation in World Englishes

285

ences can be substantial, and indeed more substantial than in English. While the overall pattern of variability seems to be the same throughout (i. e. spoken registers are less synthetic and more analytic than written registers), Figure 3 makes amply clear that, intriguingly, intra-lingual register differences can be more pronounced than inter-lingual differences. For example, spoken Italian is more similar to spoken English than it is to written Italian. Against the backdrop of the overall argumentation in this paper, we take the findings of our small set of pilot studies represented in Figure 3 as support for our view that instead of looking for (typically rather more than less) specific individual features as candidates for vernacular universals, a strategy focusing on universal coding strategies dependent on variety type (in this case: spontaneous spoken vs. written varieties) holds considerably more promise.

6.

Conclusions

In this paper we have sought to present a more balanced picture of Chambers’ notion of vernacular universals. Limitations of his original, admittedly both rather crude and sweeping understanding of vernacular universalhood there are many, but ultimately, we believe, there are plenty of prospects as well. Chambers’ original notion may have been found unconvincing in numerous respects. Yet at the same time, it has undoubtedly proved to be a useful concept in triggering a fascinating controversy and promising line of research in dialectology, contact linguistics, and the study of morphosyntactic variation in English, which will undoubtedly be continued, and may even find its way into language typology. We hope that we have shown along which lines the notion of vernacular universals needs to be downsized, as it were, adjusted and refined in order for it to be turned into an even more fruitful concept that, ultimately, will help us (dialectologists, creolists, variationists, second language acquisition specialists, language historians, and typologists) to truly see the wood for the trees – which is exactly, in our view, what Chambers had in mind. Ways of making Chambers’ original notion a less controversial and more fruitful concept include the following: first, taking seriously the constraints we formulated at the end of section 2.1; second, operationalizing it in ways as presented in sections 3 to 5; third, admitting standard ways of formulating universals in language typology to the search and formulation of vernacular universals, notably by looking not just for absolute, but also for statistical (nonimplicational as well as implicational) universals; fourth, keeping in mind

286

Bernd Kortmann and Benedikt Szmrecsanyi

the broad picture, i. e. looking for universal coding strategies which – lastly – are distinctive of individual variety types, regardless in which language we encounter them. In Freiburg, we will continue to pursue this line of research by significantly broadening the feature catalogue to be investigated, widening the range of varieties (starting with naturalistic corpus data for pidgins and creoles, but also (adult) learner varieties of English and other languages) and the range of languages such that spontaneous spoken and written varieties can be systematically compared with regard to overall coding strategies in grammar. This research agenda is likely to contribute to a better understanding of those basic principles that underlie the structural variation we can observe within and across languages, adding a crucial new and truly integrative facet to existing research into micro- and macroparametric variation.

Parameters of morphosyntactic variation in World Englishes

287

References Anderwald, Lieselotte and Susanne Wagner 2007 The Freiburg English Dialect Corpus (FRED): Applying corpuslinguistic research tools to the analysis of dialect data. In: Joan Beal, Karen Corrigan and Hermann Moisl (eds.), Using Unconventional Digital Language Corpora. Volume I: Synchronic Corpora, 35–53. Basingstoke: Palgrave MacMillan. Anttila, Raimo 1989 Historical and Comparative Linguistics. Amsterdam/Philadelphia: Benjamins. Auer, Peter 2004 Non-standard evidence in syntactic typology – Methodological remarks on the use of dialect data vs. spoken language data. In: Bernd Kortmann (ed.), Dialectology Meets Typology, 69–92. Berlin/ New York: Mouton de Gruyter. Barbiers, Sjef, Leonie Cornips and Susanne van der Kleij (eds.) 2002 Syntactic Microvariation. Amsterdam: SAND. (http://www. meertens.nl/books/synmic/). Bisang, Walter 2004 Dialectology and typology: An integrative perspective. In: Bernd Kortmann (ed.), Dialectology Meets Typology, 11–45. Berlin/New York: Mouton de Gruyter. Chambers, J. K. 2000 Universal sources of the vernacular. In: Ulrich Ammon, Klaus J. Mattheier and Peter H. Nelde (eds.), The Future of European Sociolinguistics, 11–15. (Special issue of Sociolinguistica: International Yearbook of European Sociolinguistics 14.) Tübingen: Niemeyer. Chambers, J. K. 2001 Vernacular universals. In: Josep M. Fontana, Louise McNally, M. Teresa Turell and Enric Vallduví (eds.), ICLaVE 1: Proceedings of the First International Conference on Language Variation in Europe, 52–60. Barcelona: Universitat Pompeu Fabra. Chambers, J. K. 2003² Sociolinguistic Theory: Linguistic Variation and Its Social Implications. Oxford, UK/Malden, US: Blackwell. Chambers, J. K. 2004 Dynamic typology and vernacular universals. In: Bernd Kortmann (ed.), Dialectology Meets Typology, 127–145. Berlin/New York: Mouton de Gruyter. Danchev, Andrei 1992 The evidence for analytic and synthetic developments in English. In: Matti Rissanen, Ossi Ihalainen, Terttu Nevalainen and Irma Taavitsainen (eds.), History of Englishes: New Methods and Inter-

288

Bernd Kortmann and Benedikt Szmrecsanyi

pretations in Historical Linguistics, 25-41. Berlin/New York: Mouton de Gruyter. De Vogelaer, Gunther, Annemie Neuckermans and Guido Vanden Wyngaerd 2002 Complementizer agreement in the Flemish dialects. In: Sjef Barbiers, Leonie Cornips and Susanne van der Kleij (eds.), Syntactic Microvariation. Amsterdam, 97–115. Available online at http://www. meertens.nl/ books/synmic/. Filppula, Markku, Juhani Klemola and Heli Paulasto (eds.) 2009 Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond. London/New York: Routledge. Gil, David, Peter Trudgill and Geoffrey Sampson (eds.) 2009 Language Complexity as a Variable Concept. Oxford: Oxford University Press. Greenbaum, Sidney 1996 Comparing English Worldwide: The International Corpus of English. Oxford/New York: Clarendon Press/Oxford University Press. Greenberg, Joseph H. 1960 A quantitative approach to the morphological typology of language. International Journal of American Linguistics 26: 178–194. Haser, Verena and Bernd Kortmann 2009 Agreement in English dialects. In: Andreas Dufter, Jürg Fleischer and Guido Seiler (eds.), Describing and Modeling Variation in Grammar, 271-296. Berlin/New York: Mouton de Gruyter. Haspelmath, Martin, Matthew s. Dryer, David Gil and Bernard Comrie (eds.) 2005 The World Atlas of Language Structures. Oxford: Oxford University Press. http://wals.info/index. Hernández, Nuria 2006 User’s guide to FRED. Available online at http://www.freidok.unifreiburg.de/volltexte/2489. Freiburg: English Dialects Research Group. Hockett, Charles F. 1954 Two models of grammatical description. Word 10: 210–231. Kortmann, Bernd (ed.) 2004 Dialectology Meets Typology. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd, Tanja Herrmann, Lukas Pietsch and Susanne Wagner 2005 A Comparative Grammar of British English Dialects: Agreement, Gender, Relative Clauses. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd, Edgar Schneider, Kate Burridge, Rajend Mesthrie and Clive Upton (eds.) 2004 A Handbook of Varieties of English, Volume 2: Morphosyntax. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd and Benedikt Szmrecsanyi 2004 Global synopsis: Morphological and syntactic variation in English. In: Bernd Kortmann et al. (eds.), A Handbook of Varieties of Eng-

Parameters of morphosyntactic variation in World Englishes

289

lish, Volume 2: Morphosyntax, 1142–1202. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd and Benedikt Szmrecsanyi 2009 World Englishes between simplification and complexification. In: Lucia Siebers, and Thomas Hoffmann (eds.), World Englishes: Problems – Properties – Prospects, 265–285. Amsterdam/Philadelphia: Benjamins. Kortmann, Bernd and Susanne Wagner 2005 The Freiburg English Dialect project and corpus. In: Bernd Kortmann, Tanja Herrmann, Lukas Pietsch and Susanne Wagner (eds.), A Comparative Grammar of British English Dialects: Agreement, Gender, Relative Clauses, 1–20. Berlin/New York: Mouton de Gruyter. McWhorter, John 2007 Language Interrupted: Signs of Non-Native Acquisition in Standard Language Grammars. Oxford: Oxford University Press. Mesthrie, Rajend 2006 World Englishes and the multilingual history of English. World Englishes 25: 381–390. Platt, John Talbot, Heidi Weber and Mian Lian Ho 1984 The New Englishes. London: Routledge & Kegan Paul. Pride, John B. (ed.) 1982 New Englishes. Rowley, Mass.: Newbury House. Schwegler, Armin 1990 Analyticity and Syntheticity: A Diachronic Perspective with Special Reference to Romance Languages. Berlin/New York: Mouton de Gruyter. Siegel, Jeff 2004 Morphological simplicity in Pidgins and Creoles. Journal of Pidgin and Creole Languages 19: 139–162. Siegel, Jeff 2008 The Emergence of Pidgin and Creole Languages. Oxford: Oxford University Press. Siemund, Peter 2009 Linguistic universals and vernacular data. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond. London/New York: Routledge. Szmrecsanyi, Benedikt 2009 Typological parameters of intra-lingual variability: Grammatical analyticity vs. syntheticity in varieties of English. Language Variation and Change 21(3): 319–353. Szmrecsanyi, Benedikt and Bernd Kortmann 2009a Between simplification and complexification: Non-standard varieties of English around the world. In: David Gil, Peter Trudgill and

290

Bernd Kortmann and Benedikt Szmrecsanyi

Geoffrey Sampson (eds.), Language Complexity as a Variable Concept, 64–79. Oxford: Oxford University Press. Szmrecsanyi, Benedikt and Bernd Kortmann 2009b Vernacular universals and angloversals in a typological perspective. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, 33–53. London/New York: Routledge. Szmrecsanyi, Benedikt and Bernd Kortmann 2009c The morphosyntax of varieties of English worldwide: A quantitative perspective. Special issue The Forests Behind the Trees. Lingua 119: 1643–1663. Trudgill, Peter 2001 Contact and simplification: Historical baggage and directionality in linguistic change. Linguistic Typology 5: 371–374. Trudgill, Peter 2009 Vernacular universals and the sociolinguistic typology of English dialects. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, 304-322. London/New York: Routledge. Wagner, Susanne 2004 ‘Gendered’ pronouns in English dialects – a typological perspective. In: Bernd Kortmann (ed.), Dialectology Meets Typology: Dialect Grammar from a Cross-Linguistic Perspective, 479–496. Berlin/New York: Mouton de Gruyter. Winford, Donald 2009 The interplay of ‘universals’ and contact-induced change in the emergence of new Englishes. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, 206–230. London/New York: Routledge. Wunderlich, Dieter 2008 Spekulationen zum Anfang der Sprache. Zeitschrift für Sprachwissenschaft 27: 229–265.

Comparing varieties of English: problems and perspectives Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund Abstract This contribution presents and discusses a number of different, mutually complementary perspectives and approaches to the comparative study of the grammar of English varieties. As a recurrent theme of all these approaches, we address the question of how analyses of recurrent crossdialectal grammatical patterns can contribute to the understanding of the underlying causes of variation, focusing on hypotheses involving language contact, language-internal dialect contact and multilingualism as causative factors. Case studies exemplifying the different approaches are taken from the study of pronominal systems, tense/aspect use in learner Englishes and contact varieties of English, and the syntax of embedded questions in world-wide varieties of English, among others. We argue that a mere comparison of surface forms may entail severe problems since identical or similar surface forms may hide differences in terms of function, distribution and historical development. Keywords: contact varieties, dialect typology, grammatical variation, inversion, perfect, pronominal gender, shift varieties, tense and aspect

1.

Introduction

English has diversified into many well-documented varieties around the world. We here take documentation to mean either descriptions of individual varieties (Irish English (IrE), Singapore English (SingE), etc.), as they can be found in various handbooks (Burchfield 1994; Kortmann and Schneider 2004, 2008; Trudgill and Hannah 2002, etc.), or systematic corpus material such as the ICE-family of English corpora (International Corpus of English) as well as various more specific corpus initiatives (the Freiburg English Dialect Corpus (FRED), the Hamburg Corpus of Irish English (HCIE), etc.).

292

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

Studying and comparing these varieties has shown that many of them share specific morpho-syntactic properties that are not found in the standard varieties. For example, the inversion of subject and auxiliary in embedded interrogative clauses is found in Irish English, Singapore English and Indian English (IndE) (among others); the overuse and underuse of definite articles is reported for Singapore English, Jamaican English (JamE) and several African Englishes (and others); the progressive aspect is attested with stative verbs in very many non-standard varieties and the same holds for non-standard patterns of subject-verb agreement; copula drop occurs in nearly all English-based creoles and pidgins. The use of the simple past tense and the present tense in contexts where Standard English requires the present perfect has been attested for many second-language varieties as well as learner Englishes. Many more relevant examples could in principle be listed here. The puzzling fact about these empirical findings is that different varieties of English share non-standard morpho-syntactic features even though no obvious historical connection can be reconstructed between them. It is virtually impossible to find connections in the socio-cultural histories of, say, Irish English and Indian English, apart perhaps from the general fact that these Englishes are colonial varieties, i. e. the result of British colonial expansion and the subsequent transplantation of English to these new territories. Even though it would be conceivable that the shared non-standard features that we can observe today across varieties of English find their origin in archaic dialectal forms that were exported to the colonies, such a reconstruction is unlikely for the simple fact that morpho-syntactic features like embedded inversion or copula drop simply do not – or only very sporadically – occur in the traditional dialects of the British Isles. Having said that, we hasten to add that such a reconstruction may be plausible for other features, notably non-standard patterns of subject-verb agreement (cf. Tagliamonte 2002). It is against the background of such observations that hotly debated notions such as “vernacular universals” (Chambers 2004) and “angloversals” (Mair 2003) need to be seen and evaluated. Intuitively, it appears to make a lot of sense to invoke some sort of universalist explanation for common properties of disconnected linguistic systems that fail to be motivated by the ingredients that fed these systems. The term ‘vernacular universal’ addresses the fact that most of these shared properties can be found in spoken varieties, i. e. vernacular forms, and postulates that they are the result of using language in this particular mode. According to Chambers’ hypothesis, there are natural principles, ultimately grounded in the human linguistic endowment, that are active in all kinds of vernacu-

Comparing varieties of English: problems and perspectives

293

lar varieties (in principle, of all languages), as opposed to standardized high-prestige varieties. All vernaculars, including traditional dialects, pidgins, creoles, and contact varieties, can therefore be expected to share a significant number of common features. A more nuanced approach seeks cross-dialectal patterns not so much on the basis of an undifferentiated concept of “vernaculars”, but on the basis of a typology of vernaculars according to the nature of their speakers’ social networks. Thus, Mair’s concept of “Angloversals” sees the common denominator in language contact and (imperfect) language acquisition. Peter Trudgill, in a series of publications (1995, 1996, 2002 and others), identifies “low contact” versus “high contact” varieties as a principal dimension along which dialectal features vary in characteristic ways. And indeed, many shared properties occur in contact and shift varieties of English. We consider the discussion of vernacular universals and angloversals a very useful one since it has helped to focus and sharpen the analysis of varieties of English tremendously. However, we also believe that any such sweeping generalization will attract severe problems since they necessarily have to pass over many intricate empirical problems that keep manifesting themselves in the data. For one thing, varieties of English are not uniform and well defined objects of study and do not lend themselves to a comparison in any straightforward sense. They are the result of complicated processes of language contact, language shift, language acquisition and change and in many cases are right in the middle of such processes. Moreover, they display a high degree of internal differentiation across individual speakers and speaker groups, resulting in a continuum of basilectal to acrolectal forms. It simply does not seem plausible to assume that this complexity can be captured by some simple generalization. There are also fundamental theoretical problems associated with notions such as ‘vernacular universals’ and ‘angloversals’ since – taken at face value – they are no more than extremely loose generalizations across relatively few recurrent surface structures. They abstract away from the frequency of a phenomenon and its status in the overall linguistic system. As we see it, these generalizations are fundamentally flawed in their assumption that similar or even identical surface structures are necessarily exponents of the same phenomenon. In spite of all these problems, we do of course acknowledge that there are important similarities across varieties of English that need to be taken seriously and that deserve to be properly analyzed. In addition, we do share the conviction that it makes sense to compare the specific morphosyntactic properties found across varieties of English and that there is

294

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

something interesting to compare in the first place. At the same time, however, we believe that there is more than one way of doing such a comparison and that the identification of shared surface structures is just one of several alternatives. The main goal of our contribution, therefore, is to explore alternative and complementary dimensions of comparing varieties of English. Besides the familiar approach of charting shared morpho-syntactic features, we will here discuss three alternative bases of such a comparison, namely typological hierarchies, the proximity of a variety to some reference variety and the source or origin of a non-standard grammatical phenomenon. We will illustrate these dimensions by providing detailed comparisons of narrow grammatical domains, taken from the areas of pronominal gender, reflexive marking, tense and aspect, and clause structure.

2.

Shared morpho-syntactic properties

A number of quantitative approaches to the multidimensional comparison of shared morpho-syntactic features have been explored by B. Kortmann and B. Szmrecsanyi in a recent series of papers (this volume, as well as Kortmann and Szmrecsanyi 2004; Szmrecsanyi and Kortmann 2009a, 2009b, 2009c, 2009d; Szmrecsanyi 2009).1 The explanatory hypotheses motivating their work are mostly inspired by Trudgill’s suggestions about “high-contact” versus “low-contact” varieties as the principal typological dividing line among World Englishes. Accordingly, Szmrecsanyi and Kortmann look for differences across four general types of varieties, whose profiles should be related to the social typology of density of networks invoked by Trudgill: “traditional L1” varieties (including traditional rural dialects), “high-contact L1” (including modern urban and standard varieties as well as overseas varieties with significant mixed settler input), L2 varieties, and pidgins and creoles. In what they call a “feature-based approach”, the authors offer global bird’s-eye comparisons across varieties based on a pool of 76 morphosyntactic features, including pronominal gender, reflexives, negation, agreement, relativization and many others. For example, a non-standard feature like multiple negation can be found in Newfoundland English, Gullah, Black South African English, Irish English, New Zealand English,

1 We keep the discussion here extremely short since the authors give an overview of their approach in a separate paper in this volume.

Comparing varieties of English: problems and perspectives

295

component 2

among others. Double modals, to give another example, occur in Scottish English, Nigerian Pidgin English, Appalachian English as well as some other varieties. Owing to the very broad empirical scope, the analysis can of course not go into much qualitative and quantitative detail; instead, the data consists of rough qualitative judgments regarding absence vs. presence of a given feature in the grammar of a variety. To this they add an assessment, albeit a very rough one, classing the phenomenon as “frequent” or “infrequent”. For instance, the so-called ‘a-prefixing’ exists in Irish English, but, there, it apparently occurs only rarely, while it is frequent in Appalachian English. It is obvious that this approach is a great improvement over descriptions of isolated varieties.

Pidgin/Creole Englishes L2-Englishes

L1-Englishes

component 1

Figure 1: Variety types as clusters of morpho-syntactic (adapted from Szmrecsanyi and Kortmann 2009a: 46).

features

Szmrecsanyi and Kortmann (2009a) and also Kortmann and Szmrecsanyi (this volume) show that a systematic comparison of shared morpho-syntactic properties yields clusters of variety types such that bundles of nonstandard features become valuable predictors for the kind of variety involved. This they call the “strategy-based approach” since varieties are compared based on the major coding strategies used. Figure 1 shows the result of such a comparison, where the authors interpret component 1 as “complexity” (syntheticity) and component 2 as “analyticity”. To be sure, there is considerable overlap since some properties are shared by Native (L1) Englishes, L2 Englishes and Pidgin/Creole Englishes or between certain subsets. To give some examples, the use of me instead of I in coordinate subjects is found in all variety types. A wider

296

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

range of the progressive is attested in L2 and L1 varieties. Finally, the lack of number distinctions on reflexives and the lack of inversion in wh-questions is reported for both L2 varieties as well as Pidgin and Creole Englishes. Elsewhere, Szmrecsanyi and Kortmann (2009b) compare varieties according to several indices of grammatical complexity, concluding among other things that “ornamental rule/feature complexity” (in the sense of McWhorter 2001) is clearly correlated with low-contact status. In another test, they fail to find expected effects of a global decrease of “outsider complexity” (a concept borrowed from Kusters 2003, i. e. features supposedly difficult in second-language acquisition) in L2 varieties.

3.

Typological hierarchies

An interesting alternative to charting differences and similarities between and across varieties of English is furnished by typological hierarchies, which can be understood as ordered relations of cognitive domains or spaces that may then be accessed or covered by linguistic forms in a systematic and predictable way. A prominent example of such a typological hierarchy, which will here be used for expository purposes, is the hierarchy of individuation (also known as ‘animacy hierarchy’). This hierarchy orders material and also immaterial entities by virtue of their inherent degree of individuation or level of boundedness, which, in turn, correlates with well-known types of noun classes such as concrete nouns, abstract nouns, count nouns, mass nouns and the like. Figure 2 shows the formulation of the hierarchy of individuation found in Sasse (1993). PROPER NAMES

HUMANS

ANIMALS

INANIMATE TANGIBLE OBJECTS

ANIMATES PROPER NOUNS

ABSTRACTS

MASS NOUNS

INANIMATES COMMON NOUNS

COUNT NOUNS

MASS NOUNS

Figure 2: The hierarchy of individuation (Sasse 1993: 659; Siemund 2008: 4).

Comparing varieties of English: problems and perspectives

297

Siemund (2008) shows that the distribution of third person pronouns across varieties of English follows the hierarchy of individuation and can be predicted from it in reasonably precise ways. For example, the distribution of he and she (i. e. the animate pronouns) in contrast to it (the inanimate pronoun) as found in the English Standard Varieties can be mapped onto the hierarchy of individuation in a straightforward way since it is based on a simple cut-off point located roughly between higher and lower animals (or domesticated and wild animals). Siemund (2008) also shows that extensions of animate pronouns into the inanimate domain (ships, cars, sun, moon, musical instruments, etc.) mainly concern tangible and individuated objects; in other words they are located to the right of the cut-off point.

Nouns

[+ count, + concrete]

[+ animate]

[+ female]

[– female]

she

he

[– animate]

[– count (mass)] [– concrete (abstract)]

he

it

Figure 3: Pronominal gender in the traditional dialects of Southwest England (Siemund 2008: 62).

Some non-standard varieties of English place the cut-off point between animate and inanimate pronouns even farther to the right on the hierarchy of individuation. In the traditional dialects of Southwest England systems of pronominal gender corresponding to Figure 3 can be found. There, the opposition between animate and inanimate pronouns basically yields a contrast between count nouns and mass nouns. Even though this system is on the verge of extinction today, it is well documented in 19th century sources and also survives in the English vernacular spoken in Newfoundland. A few examples are provided below in (1).

298 (1)

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund a. [What’s the matter with your hand?] Well, th’old horse muved on, and the body of the butt valled down, and he [the hand] was a jammed in twixt the body o’ un and the sharps (bran-pollard). (Siemund 2008: 43) b. Thick there cask ‘ont hold, tidn no good to put it [the liquid] in he [the cask]. (Siemund 2008: 46)

PROPER NAMES

HUMANS

ANIMALS

animate pronouns

INANIMATE TANGIBLE OBJECTS

ABSTRACTS

MASS NOUNS

inanimate pronouns

Figure 4: Predicted systems of pronominal gender (Siemund 2008: 142).

Typological hierarchies are powerful concepts because they make clear and testable predictions. The hierarchy of individuation predicts the systems of pronominal gender shown in Figure 4, but would exclude pronominal gender systems in which the pronominal forms do not cover adjacent areas on the hierarchy or in which some intermittent areas are left uncovered. The approach just outlined for the analysis of pronominal gender can be extended to other grammatical phenomena, such as reflexive marking, tense marking and perhaps also agreement. As far as reflexive marking is concerned, we just want to draw attention to the well-known person hierarchy [3 person < 2 person < 1 person], which states that if a language has a dedicated reflexive marker for the first person, it also has such markers for the second and third person, but not vice versa. Varieties of English conform to this generalization, as there is no variety of English in which the reflexive marker is optional only in the third person, but there are many attested varieties that allow a simple pronoun in reflexive function in the first person (e. g. I brew me a cup of coffee). The bottom line of looking at varieties of English from the point of view of typological hierarchies is that a construct independently motivated, by other languages and in the ideal case also by language-external evidence, that has the status of a universal or at least of a robust general-

Comparing varieties of English: problems and perspectives

299

ization provides some structured space of variation in which the grammatical phenomena found across varieties of English take up predictable positions. The approach outlined should in principle be extendable to other typological hierarchies, such as the accessibility hierarchy, definiteness hierarchy, number hierarchy and perhaps also the sonority hierarchy. Last but not least, these hierarchies make predictions about expectable diachronic developments.

4.

Proximity to reference variety

Another dimension along which varieties of English can be compared is their proximity or distance to a reference variety. As has been noted in the current literature (Sharma 2001; Mbangwana 2004), a reference variety provides a useful, and in many cases necessary point of comparison as it allows a researcher to establish what native and non-native varieties may have in common and to what extent they may diverge from one another. In addition, “capturing the native variability and contrasting it with that attested in non-native Englishes is crucial to any insightful analysis of non-native usage” (cf. Sharma 2001: 347).2 The major purpose of this section is to demonstrate how language internal variation attested in non-native varieties of English can be studied within this framework. More specifically, we shall take a closer look at the morphosyntactic variable of the present perfect, which has been extensively discussed in the literature on the native varieties of English (Elsness 1997; Tagliamonte 1996, 1997, 2000; van Herk 2008). We here focus on the present perfect because, as noticed in Tagliamonte (1996: 351), “the sheer complexity and abundance of grammatical apparatus concentrated in this area of the grammar make it an excellent site for examining the differences and similarities amongst related varieties”. To start with, the present perfect has traditionally been described as a verb form used in at least three dominant semantic contexts (cf. Jespersen 1924; Zandvoort 1932; Bauer 1970; Fenn 1987; Winford 1993; Tagliamonte 2000). These contexts are (i) resultative contexts, (ii) extended-now or

2 A reference variety is, however, not the only possible yardstick against which language data can be measured and compared. Depending on the research design as well as its objectives, non-native data can be further juxtaposed with other units of comparison such as a community, a region, a country or even a continent (cf. Tagliamonte, this volume).

300

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

continuative contexts and (iii) experiential or existential contexts. They are illustrated in (2) through (4). (2)

He has broken his arm. / She has eaten the pizza all on her own. (Siemund 2004: 414)

(3)

I have lived in Hamburg since 2001.

(4)

a. I have never been to Russia. b. I have played tennis, but not very often. (Siemund 2004: 414)

Present perfect forms in resultative contexts suggest that an action that took place in the past results in a change of state at the moment of utterance. Their occurrence in extended-now contexts, in contrast, is taken to describe a situation that started in the past and continues into or persists in the present. Used in existential contexts, finally, the present perfect is interpreted in such a way that a situation has taken place once or several times during a period of time leading up to the present. In addition, some researchers single out contexts of recent past and its subtype hot-news contexts as separate contexts requiring the present perfect in Standard English (Leech 1971; McCawley 1971; Comrie 1976; Brinton 1988; Huddleston and Pullum 2002; Siemund 2004; Radden and Dirven 2007), as exemplified in (5).3 (5)

The Prime Minister has resigned recently.

Non-native varieties of English exhibit a plethora of forms surfacing in contexts under study here, such as lone past participles, three verb clusters, past perfect forms, be perfects, and so on. There are, however, three variants which are particularly robust in these linguistic environments. These forms are the have perfect, the simple past tense, otherwise known as the preterite, and – to some extent – the present tense. Consider, for instance, the distribution of variants across various sub-varieties (or lev-

3 Some scholars argue that semantically the present perfect can be analysed in terms of several independent meanings or readings (see Leech 1971; McCawley 1971; Comrie 1976; Dahl 1999; Dahl and Hedin 2000; Lindstedt 2000) and, therefore, must be understood as an ambiguous verbal category. Others (McCoard 1978; Inoue 1979; Meyer 1992, 1995; Klein 1994; Musan 2002; Radden and Dirven 2007) postulate one central meaning for this verb form claiming that all other readings are pragmatic realizations of the prototypical meaning. The opinions are, however, divided as to what meaning should be considered the fundamental property of the present perfect.

Comparing varieties of English: problems and perspectives

301

els) of Indian English4 and compare it to that attested in Standard English English (StEngEng).5 The relevant distributions are shown in Figure 5.6 other

present

preterite

HAVE-perfect

Basilectal IndE

Mesolectal IndE

Upper mesolectal IndE (ICE)

Acrolectal IndE

StEngEng

0%

20 %

40 %

60 %

80 %

100 %

Figure 5: Distribution of tenses across levels. 4 The distributions presented in Figure 5 are based on data that were collected during a field trip to India by Julia Davydova in 2007. It should be mentioned for the sake of clarity that all of the informants on Indian English had Hindi as their native tongue and varied with respect to their length of exposure to the target language in a classroom setting. All of the acrolectal speakers presented in this study received instruction in English for more than 15 years, whereas mesolectal speakers were exposed to English for more than 5 but less than 15 years. Finally, basilectal speakers received minimal instruction in English over the time period of no longer than five (in some cases three) years. 5 The term ‘Standard English English’ is used to refer to the variety of English spoken by educated native speakers in England. The analysis of this variety draws on data from the London Lund Corpus (LLC) of Spoken English (Svartvik 1990). 6 In Figures 5 through 7 we look at finite forms excluding modal and non-finite uses of the present perfect, as in He may have arrived already. See also Davydova (2009) for a more detailed discussion of the variable context.

302

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

The comparison reveals that the acrolectal variety of Indian English almost replicates the distribution of the present perfect and other overt forms in the native variety of English studied here. Another striking observation is that the percentage of the have perfect gradually diminishes as we move from acrolect through upper-mesolect and mesolect to basilect. At the same time, the percentage of the preterite, the present tense and other non-standard variants increases dramatically towards basilect. It would appear, then, that the use of the present perfect is a function of the degree of contact to the target language. Learners that have been exposed to the medium of English for the longest period of time exhibit patterns that most closely approximate the standard. In contrast, learners that had the least extensive contact with English, i. e. basilectal speakers, deviate most strongly in their use of the present perfect from Standard English English. Another angle of comparison is provided by the semantic contexts themselves.7 Thus, Davydova (2008) shows that the have perfect is closely associated with resultative and extended-now contexts, whereas in experiential and recent past contexts this association is less robust. Interestingly, many non-native varieties of English reduplicate (albeit to a lesser extent) this distributional pattern, as shown in Figure 6. IrE

IndE

StEngEng

120 % 100 % 80 % 60 % 40 % 20 % 0% RES

EXT-NOW

EXP

RECPAST

Figure 6: Present perfect across semantic contexts. 7 Further statistical analyses can reveal which factors, both linguistic and nonlinguistic, promote and inhibit the variants under study, in what order and with what impact (cf. Tagliamonte, this volume).

Comparing varieties of English: problems and perspectives

303

Figure 6 also demonstrates that in Standard English English the have perfect occurs most frequently in resultative and extended-now contexts, whereas in experiential and recent past contexts it is less frequent. Indian English and Irish English by and large follow this distributional pattern, although both varieties exhibit a smaller overall percentage of the have perfect. How can we account for these striking similarities attested between Standard English English and non-native varieties of English? The distributional patterns presented above are best explained in terms of the semantics of the present perfect. It has been argued elsewhere (Davydova 2008) that resultative and extended-now contexts are most explicit with respect to the expression of the semantic component of current relevance, inherent in all present perfect types (cf. Huddleston and Pullum 2002: 143; Siemund 2004: 414). Indeed, the resultative present perfect has been claimed to be a central manifestation of the perfect in, for instance, Comrie (1976) and Siemund (2004). Furthermore, current relevance needs very little motivation in extended-now contexts as the latter refer to events that still obtain at the moment of utterance, having started in the past. In contrast, in contexts of recent past and in experiential contexts the notion of current relevance can be claimed to be less vivid. In fact, the only way to link the past event to the moment of utterance in contexts of recent past is to say that the event is very recent. To give an example, the proposition The Prime Minister has resigned this year is not necessarily associated with any change of state at speech time. In other words, a recent (or a fairly remote) event has no immediate repercussions for the speakers at the moment of utterance. In contrast, an event immediately preceding speech time frequently involves a change of state in the present, which in its turn might have direct consequences for the participants of the situation (cf. Dahl and Hedin 2000). Similarly, the only way to justify the semantic component of current relevance in experiential contexts is to describe it in terms of repeatability, i. e. by saying that the past situation or event logically applies to a time span which is not over yet (e. g. I have been to America three times in my whole life.) and can thus be repeated.8 It is this unspecified connection to

8 That experiential meaning can be described in terms of repeatability is explicit in Brinton (1988: 10), who states that the perfect of experience “refers to a situation which has occurred once or repeatedly before the present”. Moreover, Dahl and Hedin (2000) claim that the experiential perfect obeys the socalled “repeatability constraint” i. e. it is typically not used to refer to a particu-

304

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

the present moment which basically accounts for this use type in Standard English. The overall conclusion is that the semantics of present perfect contexts constrain variation of the linguistic variable under study, hence accounting for the close similarities attested across varieties of English. The final angle of comparison considered in this section is presented by the distributional patterns of variants competing in present perfect contexts in the upper mesolectal area of the acrolect/basilect continuum. Figure 7 depicts the distribution of tensed forms across upper mesolectal varieties of English. other

present

preterite

HAVE-perfect

Mesolectal IndE

Upper mesolectal EAfrE IndE Upper mesolectal SingE (ICE) Upper mesolectal IndE (ICE) Upper mesolectal IrE (HCIE) 0%

10 % 20 % 30 % 40 % 50 % 60 % 70 % 80 %

Figure 7: Distribution of tensed forms across varieties.

Upper mesolectal varieties of Irish English, Singapore English, Indian English and East African English (EAfrE) exhibit striking similarities in the distribution of the present perfect and other verb forms in present

lar or unique event such as the death of a specific person that occurred only once in the past.

Comparing varieties of English: problems and perspectives

305

perfect contexts. What is conspicuous is the robust occurrence of the have perfect. The second most frequent variant in the data is the preterite followed by the present tense. Other competing variants are bare past participles, be perfects and bare verb stems. On the whole there is surprisingly little influence from the substrate in this domain since the have perfect is the dominating form in all upper mesolectal varieties of English. These striking similarities can be ascribed to a relatively long exposure to the target language on the part of non-native speakers represented in the corpora under study. All in all, the amount of contact with English is one of the significant factors constraining variation of linguistic items in present perfect contexts across different varieties of English. Summing up, studying a linguistic variable across varieties allows the analyst to measure the extent to which the distribution of a linguistic variant in one variety approximates the distribution of the same linguistic variant in another variety of the language. On the one hand, this method permits a description of pervasive regularities in the distribution of surface variants across different varieties of English, thus shedding some light on the processes of interdialectal variation that might be considered to be universal in their nature. On the other hand, it allows us to pinpoint distinct non-standard phenomena that are in need of an explanation.

5.

Sources of non-standard phenomena

A final important perspective in comparing varieties of English is to investigate commonalities and differences in the historical sources of nonstandard phenomena. Often, a range of different causal explanations is available for the emergence of a given feature in a certain variety, and it is in principle possible that a feature that is shared in a superficially similar or even identical form across different varieties may have arisen from different sources or through different mechanisms in each. A careful analysis of the differences and commonalities across varieties can then help to disentangle the chains of causation, either in order to clarify the history of one of the varieties in question, or to assess the general role and import of a given causal principle. The typical explanatory approaches invoked in such analyses are discussed in the following. Contact with another language or languages is frequently considered as the source of non-standard features in varieties of English. Particularly noteworthy research traditions regarding these exist with respect to the so-called Celtic Englishes (see for instance the contributions in Tristram

306

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

1997, 2000, 2003, 2006; Filppula, Klemola, and Pitkänen 2002; Filppula, Klemola, and Paulasto 2008; specifically on Irish English see Filppula 1999 and Hickey 2007). Similar issues arise in the analysis of overseas Englishes. In this field, a common question is to what extent non-standard features can be linked causally to contact influence, often in situations where alternative scenarios of independent causation seem quite possible. A more refined perspective will have to distinguish between different types of contact influence: while some features can be explained as structural transfer, i. e. replication of structures present in the contact language, others are more amenable to an explanation in terms of effects of imperfect acquisition that are independent of the structures in the contact language, for instance effects of simplification, over-generalization, or preference for typologically more ‘natural’ structures. This is another area where cross-variety comparison is of crucial importance: unlike effects of the first kind, effects of the second type should display similar distributional patterns across different contact varieties, sensitive not to the language(s) involved but only to the social parameters under which contact has taken place. Some of the complexities involved in detecting the sources of shared or partially shared features across varieties can be seen in Hilbert’s (2008) study of inversion in embedded interrogatives, in which patterns that superficially appear to be similar across a range of English non-standard varieties are shown to be conditioned by different internal factors. Hilbert’s study is concerned with cases of embedded interrogative inversion as illustrated in (6) through (8). (6)

I wonder what is he like at all. (IrE, Filppula 1999: 168)

(7)

A lot of them ask me why don’t I go into teaching. (IndE)

(8)

So I was beginning to wonder what would I be doing there. (SingE, ICE-Sing s1a-047)

Such cases of embedded inversion can be found in several varieties of English, such as Irish English, Indian English, Singapore English, Jamaican English, Tanzanian English, Kenyan English and a range of other socalled L2-Englishes in Asia, Africa and the Caribbean; but also in African American English as well as colloquial American English (cp. the relevant chapters in Kortmann and Schneider 2004). While the shared occurrence across genetic relationships and geographical regions has made this phenomenon one of the prime candidates

Comparing varieties of English: problems and perspectives

307

for ‘Angloversal’, ‘vernacular universal’, or ‘vernacular Angloversal’ status, explanations as to the origin of the phenomenon differ from variety to variety. Inversion in vernacular Indian English is usually left unexplained (cf. Bhatt 2004: 1020), since main clause interrogatives in this variety have a tendency to occur in uninverted word order as in (9), resulting in the problem that the pattern of interrogative inversion in this variety seems to be the “mirror image” (Bhatt 2004: 1020) of that of Standard Indian English. (9)

Where he went? (IndE)

Substrate influence has mainly been offered as an explanation for embedded inversion in Irish English (cf. Filppula 2004: 179). McDavid and Card (1973) argue for overgeneralization of main clause inversion to embedded interrogatives. Whereas both explanations are plausible for Irish English, neither substrate influence nor simple overgeneralization of the main clause inversion rule could account for the occurrence of the superficially identical phenomenon in Indian English or most of the other varieties in question. Following the hypothesis that shared internal conditioning is an indicator of common origin of the feature across varieties, Hilbert (2008) shows that patterns of interrogative inversion in Irish English display the same tendency as the ones in other L2-varieties of English for embedded constituent interrogatives. The study is based on the spoken sections of the respective ICE (International Corpus of English) corpora. For Irish English, the Hamburg Corpus of Irish English (HCIE) is used, since even the spoken sections of ICE-Ireland do not include non-standard inversion in embedded interrogatives. In all these varieties, embedded inversion occurs primarily when be is the finite verb of the embedded clause (Figure 8). Other verbs (such as modals), and full verbs requiring periphrastic do for inversion, occur preferably with uninverted word order, identical to embedded interrogatives in Standard English, as shown in (10) and (11). (10)

You know what is Casanova? (IndE, ICE-Ind s1a-38)

(11)

Have you measures how much how high is your cholesterol? (SingE, ICE-Sing s1a-13)

308

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

others

DO + main verb

BE (as main Verb)

JamE

EAfrE

IrE

SingE

IndE

0

20

40

60

80

100

120

140

Figure 8: Inversion in embedded constituent interrogatives (according to verb type; absolute numbers).

$OVRLQWKHH[WHQWWRZKLFKHPEHGGHGFRQVWLWXHQWLQWHUURJDWLYHVDUH LQYHUWHGDWDOO,ULVK(QJOLVKFDQEHORFDWHGZLWKLQWKHRWKHUYDULHWLHV LQTXHVWLRQ 7DEOH  Table 1: Absolute occurrence of inversion in embedded constituent interrogatives. IndE

SingE

IrE

EAfrE

JamE

+INV

130

67

75

45

30

–INV

507

384

350

261

336

Total

637

451

425

306

366

The prominence of be in inverted embedded constituent interrogatives favors the hypothesis that non-standard inversion is based on the fusion of the interrogative word with be (and other frequent verbs), rather than

Comparing varieties of English: problems and perspectives

309

the productive application of the inversion rule to embedded interrogatives, which is suggested by the use of the term “overgeneralization” in studies not comparing internal conditioning. This hypothesis is supported by two further pieces of evidence. Firstly, we find the fusion of the interrogative word and be even in cases where the embedded clause occurs uninverted, repeating the main verb and thus resulting in double marking, as in (12). (12)

a. Don’t understand how’s it is coming up. (SingE, ICE-Sing s1a-053) b. let me know how is all the boys and girls is and how they are getting along (IrE, HCIE) c. I want to [k]now how is David John Taylor is getting along and Paddy Heron also (IrE, HCIE)

Such cases occur in all of the varieties in question, including Irish English. Hilbert (2008) shows that the fusion of the interrogative word and the inverted verb can be predicted from the frequency of co-occurrence, that is, the more frequently a given verb co-occurs with a given interrogative word (and for that matter also with a given subject), the more likely it is to be inverted. Secondly, the same co-occurrence of inversion and be can be observed in main clause constituent interrogatives (Hilbert 2008), not including, however, Irish English, as uninverted main clause constituent interrogatives do not occur in this variety (Table 2). Table 2: Absolute occurrence of inversion in main clause constituent interrogatives. IndE

SingE

JamE

EAfrE

IrE

+INV

552

1050

732

767

140

-INV

78

91

201

46

0

Other

26

392

46

49

1

Total

656

1533

979

860

141

To give a brief first summary, the data suggest that Irish English patterns with other L2 varieties with respect to the internal factors conditioning the occurrence of non-standard inversion in embedded constituent interrogative clauses. This favors the conclusion that Irish English embedded inversion, at least for this type of interrogative clause, is primarily a substrate-independent learner phenomenon, rather than L1-influence, even

310

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

though not in the form of simple overgeneralization as suggested by McDavid and Card (1973). Total

– INV

+ INV

IrE

EAfrE

JamE

SingE

IndE 0

50

100

150

200

250

300

350

Figure 9: Inversion in embedded polar interrogatives (absolute numbers).

For embedded polar interrogatives, however, the picture looks very different. Firstly, the absolute number of non-standard inverted interrogatives is much higher in the Irish English data than in the other varieties in question (Figure 9). 21 % of all embedded polar interrogatives are inverted in Irish English, whereas the highest inversion ratio in the other varieties is only 8 % (the two East African varieties taken together); and also higher than the inversion ratio for embedded constituent interrogatives in Irish English (also 8 %). Thus, inversion in embedded polar interrogatives seems to be a much more common option in Irish English than in the other L2-varieties. The latter prefer subordination with if or whether, as in (13). If inversion is ever used in these varieties, it is with be, as in (14). (13)

She asked if I would like to have dinner at her place. (IrE)

(14)

Anyway I ask is that Barry. (SingE, ICE-Sing s1a-092)

Comparing varieties of English: problems and perspectives

311

In Irish English, in contrast, we find inversion with all types of verbs (cf. (15)): (15)

a. You ask me is Hamilton home from new Zealand. (IrE, HCIE) b. … who asked me would I go with him as a servant. (IrE, HCIE) c. Let me know did your sister send for Thomas. (IrE, HCIE)

others

BE (as main verb)

DO + main verb

IrE

EAfrE

JamE

SingE

IndE 0

5

10

15

20

25

30

Figure 10: Inversion in embedded polar interrogatives (according to main verb; absolute numbers).

In absolute numbers, Irish English displays a relatively even distribution of inversion across all verb types (Figure 10), which suggests that inversion in this clause type is applied as a productive rule rather than being the accidental result of some chunking process. The other varieties investigated show similar distributions for embedded constituent interrogatives, with the restriction, however, that the overall occurrence of embedded inversion in polar interrogatives is very low (Table 3).

312

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

Table 3: Inversion in embedded polar interrogatives (according to main verb). IndE

SgE

JamE

EAfrE

IrE

DO + main verb 1

1

0

1

26

BE

3

5

1

2

18

others

1

1

1

1

28

The analysis of apparent similarities between embedded inversion in a range of varieties of English shows that a large proportion of the regularities observed in embedded inversion is primarily governed by mechanisms rooted in substrate-independent learner strategies, rather than direct structural influence from the learner’s L1. The differences in internal conditioning that were shown to set apart Irish English from the other L2 varieties, and that suggest different sources of the construction, are for the most part differences in the extent of generalization and conventionalization associated with the pattern: while in most of the varieties investigated here embedded inversion is linked to the type of question and to the type of verb involved, and appears to be based on certain high-frequency lexical collocations, Irish English embedded inversion in polar (as opposed to constituent) interrogatives displays a much more regular and productive pattern. To what extent this difference is caused by specific properties of the substrate language, or by differences between the sociohistorical conditions of language contact in the case of Irish English and other contact situations, remains to be explored. The hypothesis of different sources for polar and constituent interrogatives in Irish English is supported by the fact that the two clause types are rated differently in grammaticality judgment tests, with embedded polar interrogatives being much more readily accepted than embedded constituent interrogatives (Henry 1995: 106; Filppula 1999: 168). As for structural substrate features, however, it must be borne in mind that Irish, being a consistent VSO language, offers no direct structural parallel to inversion as a specific grammatical marking device: all clauses, independent of clause type, are “inverted” (i. e. V-initial). While this might conceivably serve as an explanation for a general tendency of stronger relative preference for V-initial structures, the question in what sense the use of embedded inversion in Irish English can genuinely be characterized as an instance of structural replication from Irish is a complex one and cannot be fully discussed within the scope of this paper.

Comparing varieties of English: problems and perspectives

313

The case of inversion in Irish English also throws into relief the need for a more refined general perspective on different types of contact influence. A ready intuition that wants to link some non-standard features with apparent counterparts in contact languages has often been in competition with a more skeptical stance, whose proponents point to alternative, language-internal possibilities of causation, and to the general empirical difficulties of proving or disproving contact explanations with any amount of reliability. The competition between substratal and retentionist (or superstratal) hypotheses has thus been a perennial problem in contact dialect studies. However, what the case of inversion in indirect questions illustrates is that contact effects may themselves be of several quite different kinds. While some features can be explained as direct structural transfer, i. e. more or less identical replication of specific structures present in the contact language, others are more amenable to an explanation invoking effects of imperfect acquisition that are independent of the structures in the contact language. These may include effects of simplification, overgeneralization, or preference for more typologically ‘natural’ structures. As the case of inversion shows, cross-variety comparison can provide crucial evidence here: while effects of the first kind should occur only in specific contact situations dependent on the contact language(s) involved, effects of the second type should display uniform distribution patterns across different contact varieties, sensitive not to the language(s) involved but only to the social parameters under which contact has taken place. Substantiating hypotheses of contact-induced change in English varieties is often made difficult by historical uncertainties about the precise shape of the original varieties that provided the linguistic input to the contact situations. In the case of Irish English, the problem is exacerbated by the fact that the input was itself highly heterogeneous, and languageinternal (cross-dialect) contact must have played an important role in the genesis of present-day Irish English in addition to contact with Irish itself. An empirical approach to these questions requires an even more finegrained geographic/diachronic comparison, not only between Irish English and other contact varieties, but also within Irish English itself. Such an approach is explored by Pietsch (2009a, 2009b), where the observation of multiple grammatical variables together through historical corpus data is used in order to detect the socio-historical locus of innovations within Irish English. Observation of how various nonstandard features cluster together in the linguistic profiles of writers from different time periods and areas in 18th and 19th century Ireland can provide important hints not only about the socio-linguistic situations characteristic of particular

314

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

sectors of Irish society at their respective time periods, but also – indirectly – about the causal factors in the linguistic development of each variable. This is especially valuable, under a contact-linguistic perspective, where linguistic innovations can be related to demographical information about the progress of language shift through different parts of the country.

Figure 11: Embedded inversion in the Hamburg Corpus of Irish English.

Thus, it can be shown that the occurrence of subject-verb inversion in indirect question clauses in historical Irish English is strongly linked to the south-west of Ireland, those parts of the country where relatively late but rapid language shift from Irish to English went on and affected large parts of the population during the early and mid-19th century. Figure 11 shows in which locations of Ireland embedded inversion sentences were found in the Hamburg Corpus of Irish English (the size of the dots represents the number of examples attested in each location). Table 4 and Figure 12 rep-

Comparing varieties of English: problems and perspectives

315

resent in more detail the distribution of these tokens across three subareas, as well as two time periods (before the mid-19th century, and after), and across the two sentence types discussed earlier: wh-questions versus polar questions. Note that the division into three geographical areas was not given a priori, and does not correspond to any preexisting geographical or political division, but was derived from the cumulative evaluation of the distribution patterns of multiple linguistic variables in the corpus. The northern area corresponds roughly to that identified as “Ulster English” in traditional dialectological work (cf. Barry 1981). The boundary of the area marked as “south-west” roughly corresponds to the line along which native Irish was receding during the 19th century. Very roughly speaking, to the south-west of this line, Irish still tended to be spoken by about half the population or more around the beginnings of the 19th century, according to census data documented in FitzGerald (1984). Table 4: Inverted embedded questions in the Hamburg Corpus of Irish English (frequency per 10,000 words). period 1720–1860

region

polar questions

wh questions

N

freq

N

freq

South-West

48,141

13

2.7

16

3.3

South-East

67,131

5

0.7

3

0.5

166,114

0

0

2

0.1

South-West

40,490

31

7.7

37

9.1

South-East

62,063

14

2.3

5

0.8

112,520

5

0.4

9

0.8

Ulster 1860–1910

word count

Ulster

From the data in the table and figure, the following observations can be drawn: all areas and both sentence types show a marked increase of inverted structures from the earlier to the later sub-period.9 Most notable, however, are the geographical differences. While embedded inversion is found in all parts of the country, it is decidedly most frequent in the data from the southwest, followed only at some distance by that from the southeast, and very sparse attestation in Ulster. Moreover, inversion in 9 This is likely to represent a genuine linguistic change only to some extent; parts of the observed effect may also be caused by a stronger representation of non-standard speech in the corpus, caused for sociolinguistic reasons involving an increase of accessibility of writing to speakers from the lower social strata.

316

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

polar question contexts, the sentence type that is typologically most marked according to the comparison with the other worldwide varieties, is virtually absent from Ulster in this historical data. 1860 – 1910

1720 – 1860

wh-Ulster polar Ulster wh-SE polar SE wh-SW polar SW 0

2

4

6

8

10

Figure 12: Inverted embedded questions in the Hamburg Corpus of Irish English (SE = Southeast; SW = Southwest; frequency per 10,000 words).

The concentration of embedded inversion use in the south-west, the most active zone of ongoing language shift and bilingualism at that time, points towards a contact-related origin scenario, possibly one where imperfect L2 acquisition involving grammatical simplification strategies and overgeneralization played a larger role than specific substrate-based transfer.

6.

Summary and conclusion

As we have demonstrated with the examples presented here, varieties of English can be compared along many different dimensions. Such comparisons, if conducted under the right angle, can reveal highly interesting systematic patterns in the differences and similarities displayed between

Comparing varieties of English: problems and perspectives

317

varieties. English varieties present anything but a chaotic picture of random distribution; instead, their internal differentiation allows meaningful conclusions and generalization about their genesis and the forces that shaped them. Most importantly, however, careful analysis not only reveals to what extent varieties display interesting similarities, but also, on closer inspection, shows how much real difference can sometimes be hidden below apparent surface similarity. Identical or near-identical surface structures may often hide phenomena that are structurally and historically quite different: what looks the same may not in fact be the same. This finding may have important repercussions for the ongoing discussion about notions of “vernacular universals” and “angloversals”. If universal patterns of similarity across all varieties, or across certain classes of varieties, are to be related to causative hypotheses in a meaningful way, it is crucial that each of the phenomena in question is also analyzed carefully within each of the varieties it occurs in, with a view to detecting its sources and defining its conditioning factors in each. Comparative studies that have looked globally at a large range of varieties from an extreme bird’s-eye perspective have contributed much to the formation of interesting research hypotheses, which now help to shape the questions that have to be asked in relation to individual varieties. We feel, however, that much more remains to be done in the field of more detailed analysis of individual varieties, or detailed comparative analysis of small individual domains of grammar across varieties, until conclusive answers about the nature of the universal patterns can be given.

Abbreviations EAfE = East African English; EXP = experiential context; EXT-NOW = extended-now context; FRED = Freiburg English Dialect Corpus; HCIE = Hamburg Corpus of Irish English; ICE = International Corpus of English; IndE = Indian English; IrE = Irish English; JamE = Jamaican English; LLC = London Lund Corpus; RECPAST = context of recent past; RES = resultative context; SingE = Singapore English; StEngEng = Standard English English.

318

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

References Barry, Michael V. 1981 The southern boundaries of Northern Hiberno-English speech. In: Michael V. Barry (ed.), Aspects of the English Dialects in Ireland, 52–95. Belfast: Queen’s University, Institute of Irish Studies. Bauer, Gero 1970 The English ‘Perfect’ reconsidered. Journal of Linguistics 6: 189– 198. Bhatt, Rakesh M. 2004 Indian English: Syntax. In: Bernd Kortmann, Kate Burridge, Rajend Mesthrie, Edgar W. Schneider and Clive Upton (eds.), A Handbook of Varieties of English, 1016–1030. Berlin: Mouton de Gruyter. Brinton, Laurel J. 1988 The Development of English Aspectual Systems. Cambridge: Cambridge University Press. Burchfield, Robert (ed.) 1994 The Cambridge History of the English Language. English in Britain and Overseas: Origins and Developments (Volume 5). Cambridge: Cambridge University Press. Chambers, J. K. 2004 Dynamic typology and vernacular universals. In: Bernd Kortmann (ed.), Dialectology Meets Typology, 127–145. Berlin: Mouton de Gruyter. Comrie, Bernard 1976 Aspect. Cambridge: Cambridge University Press. Dahl, Östen 1999 Perfect. In: Keith Brown and Jim Miller (eds.), Concise Encyclopaedia of Grammatical Categories, 290–91. Oxford: Elsevier. Dahl, Östen and Eva Hedin 2000 Current relevance and event reference. In: Östen Dahl (ed.), Tense and Aspect in the Languages of Europe, 385–402. Berlin: Mouton de Gruyter. Davydova, Julia 2008 Preterite and present perfect in Irish English: Determinants of variation. Arbeiten zur Mehrsprachigkeit, 84, Series B. Universität Hamburg. Davydova, Julia 2009 The present perfect: A corpus-based study of variation across nonnative Englishes. Doctoral thesis, University of Hamburg. Elsness, Johan 1997 The Perfect and Preterite in Contemporary and Earlier English. Berlin: Mouton de Gruyter.

Comparing varieties of English: problems and perspectives Fenn, Peter 1987

319

A Semantic and Pragmatic Examination of the English Perfect. Tübingen: Gunter Narr Verlag. Filppula, Markku 1999 The Grammar of Irish English: Language in Hibernian Style. London: Routledge. Filppula, Markku 2004 Irish English: Morphology and syntax. In: Bernd Kortmann, Kate Burridge, Rajend Mesthrie, Edgar W. Schneider and Clive Upton (eds.), A Handbook of Varieties of English, 73–101. Berlin: Mouton de Gruyter. Filppula, Markku, Juhani Klemola and Heli Paulasto 2008 English and Celtic in Contact. London: Routledge. Filppula, Markku, Juhani Klemola and Heli Pitkänen (eds.) 2002 The Celtic Roots of English. Joensuu: Joensuu University Library. FitzGerald, Garret 1984 Estimates for baronies of minimum level of Irish speaking amongst the successive decennial cohorts, 1771–1781 to 1861–1871. Proceedings of the Royal Irish Academy, Section C 84: 117–155. Henry, Alison 1995 Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: Oxford University Press. Hickey, Raymond 2007 Irish English: History and Present-Day Forms. Cambridge: Cambridge University Press. Hilbert, Michaela 2008 Interrogative inversion in non-standard varieties of English. In: Peter Siemund and Noemi Kintana (eds.), Language Contact and Contact Languages, 261-289. Amsterdam: Benjamins. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Inoue, Kyoko 1979 An analysis of the English present perfect. Linguistics 17: 561–589. Jespersen, Otto H. 1924 The Philosophy of Grammar. London: Allen. Klein, Wolfgang 1994 Time in Language. London: Routledge. Kortmann, Bernd and Benedikt Szmrecsanyi 2004 Global synopsis: Morphological and syntactic variation in English. Volume 2: Morphology, syntax. In: Bernd Kortmann, Kate Burridge, Rajend Mesthrie, Edgar W. Schneider and Clive Upton (eds.), A Handbook of Varieties of English, 1142–1202. Berlin: Mouton de Gruyter.

320

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

Kortmann, Bernd and Edgar Schneider (eds.) 2004 A Handbook of Varieties of English. Berlin: Mouton de Gruyter (two volumes). Kortmann, Bernd and Edgar Schneider (eds.) 2008 Varieties of English. Berlin: Mouton de Gruyter (four volumes). Kusters, Wouter 2003 Linguistic Complexity: The Influence of Social Change on Verbal Inflection. Utrecht: LOT. Leech, Geoffrey N. 1971 Meaning and the English Verb. London: Longman. Lindstedt, Jouko 2000 The perfect – aspectual, temporal, evidential. In: Östen Dahl (ed.), Tense and Aspect in the Languages of Europe, 365–383. Berlin: Mouton de Gruyter. Mair, Christian 2003 Kreolismen und verbales Identitätsmanagement im geschriebenen jamaikanischen Englisch. In: Elisabeth Vogel, Antonia Napp and Wolfram Lutterer (eds.), Zwischen Ausgrenzung und Hybridisierung, 79–96. Würzburg: Ergon. Mbangwana, Paul 2004 Cameroon English: Morphology and syntax. In: Bernd Kortmann, Kate Burridge, Rajend Mesthrie, Edgar W. Schneider and Clive Upton (eds.), A Handbook of Varieties of English, 898–908. Berlin: Mouton de Gruyter. McCawley, James D. 1971 Tense and time reference in English. In: Charles J. Fillmore and D. Terence Langendon (eds.), Studies in Linguistic Semantics, 96–113. New York: Holt, Rinehart and Winston. McCoard, Robert W. 1978 The English Perfect: Tense-Choice and Pragmatic Inferences. (NorthHolland Linguistic Series 38.) Amsterdam: North-Holland. McDavid, V. G. and Card, W. 1973 Problem areas in grammar. In: A. L. Davis (ed.), Culture, Class, and Language Variety: A Resource Book for Teachers, 2nd edition, Urbana, IL: National Council of Teachers of English. McWhorter, John 2001 The world’s simplest grammars are creole grammars. Linguistic Typology 6: 126–166. Meyer, Matthias L. G. 1992 Das englische Perfekt: Grammatischer Status, Semantik und Zusammenspiel mit dem Progressive. (Linguistische Arbeiten 277.) Tübingen: Niemeyer. Meyer, Matthias L. G. 1995 Past tense and present perfect in the Brown and the LOB corpora.

Comparing varieties of English: problems and perspectives

321

In: Wolfgang Riehle and Hugo Keiper (eds.), Anglistentag, 1994. Graz. Proceedings (8): 201–228. Tübingen: Niemeyer. Musan, Renate 2002 The German Perfect: Its Semantic Composition and its Interaction with Temporal Adverbials. (Studies in Linguistics and Philosophy 78.) Dordrecht: Kluwer. Pietsch, Lukas 2009a Hiberno-English medial-object perfects reconsidered: A case of contact-induced grammaticalisation. Studies in Language 33: 528– 568. Pietsch, Lukas 2009b English in Ireland: Grammar in Language Contact. Post-doctoral thesis, University of Hamburg. Radden, Günter and René Dirven 2007 Cognitive English Grammar. Amsterdam: Benjamins. Sasse, Hans-Jürgen 1993 Syntactic categories and subcategories. In: Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), Syntax: Ein internationales Handbuch zeitgenössischer Forschung, 646–686. Berlin: Mouton de Gruyter. Sharma, Devyani 2001 The pluperfect in native and non-native English: A comparative corpus study. Language Variation and Change 13: 343–373. Siemund, Peter 2004 Substrate, superstrate and universals: Perfect constructions in Irish English. In: Bernd Kortamann (ed.), Dialectology Meets Typology: Dialect Grammar from a Cross-Linguistic Perspective, 401–434. Berlin: Mouton de Gruyter. Siemund, Peter 2008 Pronominal Gender in English: A Study of English Varieties from a Cross-linguistic Perspective. London: Routledge. Svartvik, Jan 1990 The London-Lund Corpus of Spoken English: Description and Research. Lund: Lund University Press. Szmrecsanyi, Benedikt 2009 Typological parameters of intralingual variability: Grammatical analyticity vs. syntheticity in varieties of English. Language Variation and Change 21 (3): 319–353. Szmrecsanyi, Benedikt and Bernd Kortmann 2009a Vernacular universals and angloversals in a typological perspective. In: Markku Filppula, Juhani Klemola and Heli Paulasto (eds.), Vernacular Universals and Language Contacts: Evidence from Varieties of English and Beyond, 33–53. London: Routledge.

322

Julia Davydova, Michaela Hilbert, Lukas Pietsch and Peter Siemund

Szmrecsanyi, Benedikt and Bernd Kortmann 2009b The morphosyntax of varieties of English worldwide: A quantitative perspective. Lingua 119 (11): 1643–1663. Szmrecsanyi, Benedikt and Bernd Kortmann 2009c Between simplification and complexification: Non-standard varieties of English around the world. In: Geoffrey Sampson, David Gil and Peter Trudgill (eds.), Language Complexity as a Variable Concept, 64–79. Oxford: Oxford University Press. Szmrecsanyi, Benedikt and Bernd Kortmann 2009d World Englishes between simplification and complexification. In: Lucia Siebers and Thomas Hoffmann (eds.), World Englishes: Problems, Properties, Prospects, 265–285. Amsterdam: Benjamins. Tagliamonte, Sali 1996 Has it ever been ‘Perfect’? Uncovering the grammar of early Black English. York Papers in Linguistics 17: 351–396. Tagliamonte, Sali 1997 Obsolescence in the English Perfect? Evidence from Samaná English. American Speech 72(1): 33–68. Tagliamonte, Sali 2000 The grammaticalisation of the present perfect in English. In: Olga Fischer, Anette Rosenbach and Dieter Stein (eds.), Pathways of Change. Grammaticalization in English, 329–354. Amsterdam Benjamins. Tagliamonte, Sali 2002 Comparative sociolinguistics. In: Jack Chambers, Peter Trudgill, Natalie Schilling-Estes (eds.), Handbook of Language Variation and Change, 729–763. Oxford: Blackwell. Tristram, Hildegard (ed.) 1997 The Celtic Englishes. Heidelberg: Winter. Tristram, Hildegard (ed.) 2000 The Celtic Englishes II. Heidelberg: Winter. Tristram, Hildegard (ed.) 2003 The Celtic Englishes III. Heidelberg: Winter. Tristram, Hildegard (ed.) 2006 The Celtic Englishes IV. Potsdam: Universitätsverlag. Trudgill, Peter 1995 Grammaticalisation and social structure: Nonstandard conjunctionformation in East Anglian English. In: Frank R. Palmer (ed.), Grammar and Meaning: Papers in Honour of Sir John Lyons, 136– 147. Cambridge: Cambridge University Press. Trudgill, Peter 1996 Dialect typology: Isolation, social network and phonological structure. In: Gregory R. Guy, Crawford Feagin, Deborah Schiffrin and John Baugh (eds.), Towards a Social Science of Language, Volume 1: Variation and Change in Language and Society, 3-22. Amsterdam: Benjamins.

Comparing varieties of English: problems and perspectives

323

Trudgill, Peter 2002 Linguistics and social typology. In: Jack Chambers, Peter Trudgill and Natalie Schilling-Estes (eds.), Handbook of Language Variation and Change, 707–728. Oxford: Blackwell. Trudgill, Peter and Jean Hannah 2002 International English: A Guide to Varieties of Standard English. London: Arnold. Van Herk, Gerard 2008 The Letter Perfect: The Present Perfect in early African American correspondence. English World-Wide 29(1): 45–69. Winford, Donald 1993 Variability in the use of perfect have in Trinidadian English: A problem of categorical and semantic mismatch. Language Variation and Change 5: 141–188. Zandvoort, R.W. 1932 On the perfect of experience. English Studies 14(1): 11–20.

Part 4 Variation and linguistic theory

Norwegian main clause declaratives: variation within and across grammars Kristin Melum Eide and Hilde Sollid [I]t remains to be emphasized that linguistic diversity begins next door, nay, at home and within one and the same man. […] What we heedlessly and somewhat rashly call ‘a language’ is the aggregate of millions of such microcosms many of which evince such aberrant linguistic comportment that the question arises whether they should not be grouped into other ‘languages’. André Martinet: Preface to Weinreich (1953)

Abstract Norwegian is generally considered a strict verb second (V2) language with a few exceptions. However, there is considerable variation across registers and genres. This paper investigates variation in main clause declaratives, focusing on the placement of the finite verb. The observed variation is explained using an integrative approach that reconciles generative and sociolinguistic theories of linguistic variation. The basic assumption is that each speaker is multilingual and has a set of parallel grammars available. In the case of Norwegian main clause declaratives, we argue that the available grammars are “verb raising” and “non-verb raising”. As with vocabulary items, the speaker chooses the sub-grammar considered most appropriate in a given social context. Keywords: main clause declaratives, multilingualism, parallel grammars, verb raising, verb second, verb third

1.

Norwegian declaratives and verb second

Norwegian, like other Mainland Scandinavian languages, is considered a verb second (V2) language (cf. e. g. Faarlund, Lie and Vannebo 1997; Holmberg and Platzack 1995; Vikner 1995) since the finite verb occupies the second position in main clause declaratives, regardless of the properties of the constituent occupying the first position. There are exceptions to this,

328

Kristin Melum Eide and Hilde Sollid

but they are usually not considered sufficiently robust, significant, or theoretically interesting to weaken the generalization that Norwegian declaratives are subject to a V2 requirement. However, when considering a range of registers and genres, both spoken and written, we find considerable variation in Norwegian main clauses, even when we restrict our attention to declaratives and focus on the placement of the finite verb. What the textbooks label ‘rare occurrences’ of exceptional word order turn out to be astonishingly frequent, especially in certain types of discourse. Hence, the apparently unassailable V2 rule on closer inspection turns out to vary across speakers, dialects, modes and genres. This variation in Norwegian main clause declaratives is the focal point of this paper. Our main objective is to adequately describe and explain the observed variation in Norwegian main clause declaratives. The study reported in this paper is part of a larger project (ParaGram) aiming to account for intra-speaker variation with an integrated theory of syntactic variation. Our approach combines generative grammar of the Principles and Parameter brand with sociolinguistic accommodation theories to shed light on the linguistic variation observed in the same individual in different contexts. The theoretical bridge between our two frameworks is the fundamental assumption that each speaker is multilingual: in addition to a set of vocabularies, each speaker has a set of parallel grammars from which he selects the sub-grammar most appropriate in a given context and speech act. In section 2, we discuss the notion of parallel grammars, invoked here to account for intra-speaker variation. We discuss how our respective frameworks have accommodated intra-speaker variation and various brands of multilingualism and show that Norwegian speakers range over a number of sub-grammars (section 2.2). In section 2.3, we present Roeper’s (1999) theory of Universal Bilingualism, which can account for the coexistence of several sub-grammars in one mind. The notion of ‘a grammar’ is the topic of section 2.4 and, in section 2.5 we use accommodation theories to explain how speakers select between available grammars. Section 3 discusses the empirical focal point of the paper, finite verb placement in Norwegian main clause declaratives. We present the corpus and the interviews in section 3.1, provide an overview of the findings in section 3.2, and offer an analysis of the data in section 3.3 Section 4 concludes the paper.

Norwegian main clause declaratives: variation within and across grammars

2.

Parallel grammars

2.1.

Intra-speaker variation in the generativist framework and in sociolinguistics

329

The generative framework has not dealt extensively with intra-speaker variation. For decades, the leading approaches have been guided by this much-quoted passage from Chomsky (1965: 3): “Linguistic theory is concerned with an ideal speaker-listener in a completely homogeneous speech community.”1 This entails that some data are seen as belonging to the core and others as belonging to the periphery. Some data are seen as belonging to a speaker’s core grammar; others, such as relics, borrowings, and various code-switching situations, are seen as clutter impeding any theoretically significant description of the primary, core grammar of the speaker.2 Even certain grammars constituting a potential single non-variant mental grammar of an individual are described as belonging to the periphery because they are considered ‘impure’ and thus irrelevant; cf. Chomsky (1986: 17): We exclude, for example, a speech community of uniform speakers each of whom speaks a mixture of French and Russian (say an idealized version of the 19th century Russian aristocracy). The language of such a speech community would not be “pure” in the relevant sense because it would not represent a sin-

1 See also the discussion on the ideal speaker-listener and the consequences of this construct for reproducibility and falsification in Bisang (this volume). 2 This fact leads an anonymous reviewer to ask why we do not “have a look at contemporary sociolinguistic and dialectological models that have been more ‘successful’ with respect to intra-lingual (including intra-individual) variation … and the way non-generative grammatical theories (for instance construction grammar, cognitive grammar) treat syntactic variation.” This is certainly a legitimate point, and the format of this paper is partly responsible for our narrow theoretical focus. Another important reason for not bringing in other theories is the fact that we discuss the frameworks we specialize in, not the frameworks that prima facie would seem to have the best chance of accounting for this type of variation. In our view, it is a fruitful exercise to attempt to reconcile the two frameworks of our choice since most linguists would agree that there is a traditional entrenchment between generativist approaches and sociolinguistics. Our integrated approach allows us to show that the insights of the two frameworks are compatible, which allows for the accumulation of new and productive linguistic insights.

330

Kristin Melum Eide and Hilde Sollid

gle set of choices among the options permitted by UG but rather would include “contradictory” choices for certain of these options.

Any serious description of linguistic systems involves some idealization of the subject matter since authentic data contain slips-of-the-tongue and playful turns of phrase, among other things. However, one line of criticism of the generative framework has maintained that it discards data that ought to belong to the core, and that the border between core and periphery in some mysterious way seems to coincide with the line between what can and cannot be explained by the current version of the theory. Sociolinguistics, at least quantitative variationist approaches, could be rightfully criticized for playing down the theoretical significance of the individual speaker, as the focus has been on sociolinguistic surveys and ordered heterogeneity in speech communities. Schilling-Estes (2004: 163– 164) notes: “In [traditional variationist approaches] the focus has been on correlations between aggregate patterns of language use and social categories such as age, sex and race, that traditionally have been taken to be relatively fixed, at least for the purposes of analysis.” In contemporary sociolinguistic variationist studies, the individual speaker and intra-individual variation have become the centre of attention, cf. the third wave in sociolinguistics (Eckert 2005). This development has its roots in earlier studies of style (Labov 1966), speech accommodation (cf. Giles 1973; Giles and Powesland [1975] 1997), code switching (Blom and Gumperz [1972] 2000), and audience design (Bell 1984). These and other sociolinguistic approaches to linguistic variation provide substantial empirical support for our fundamental assumption of the multilingual speaker.

2.2.

The multilingual speaker

Over the last decade, there have been multiple studies in microcomparative syntax, particularly among European generativists working within Principles and Parameters (P&P) theory. Microcomparative research investigates the similarities and differences between closely related linguistic systems, such as geographically adjacent dialects of a national language. As pointed out by Eide and Åfarli (2007), examining dialectal systems inevitably leads to investigations of intra-speaker variation since in modern society any speaker with a viable dialect is also a speaker of one or more standards of the national language. This implies a special kind of bilingualism, the use of two closely related grammars by the same individual. Recent studies (Bentzen and Vangsnes 2007; Vangsnes 2006;

Norwegian main clause declaratives: variation within and across grammars

331

Westergaard and Vangsnes 2005; and the publications of the Dutch SAND project) have attempted to describe this variation by using current versions of P&P theory in spite of the traditional theoretical bias of this framework towards a monolingual speaker. Voices in the generativist camp have been advocating that P&P theory reflect the fact that humans are designed to learn more than one language; cf. e. g. Cook and Newson (2007: 223–224): “The issue is really whether it is proper to set this universal bilingualism to one side in linguists’ descriptions of competence or whether it should in effect form the basis of the description from the beginning. […] UG [Universal Grammar] Theory has to account for this universal ability to have two, possibly conflicting, grammars at the same time.” Although the “emphasis on monolingualism has simply been taken for granted by those working within the UG Theory” and “the only true knowledge of the language is taken to be that of the adult monolingual native speaker” (Cook and Newson 2007: 221), it would be a gross misrepresentation of P&P theory to claim that its defendants are reluctant to recognize bilingualism and multilingualism. Chomsky himself readily admits that “everyone grows up in a multilingual environment” and states explicitly that “[w]hatever the language faculty is it can assume many different states in parallel” (Chomsky 2000: 59). Moreover, he appreciates that dialects and styles constitute linguistic systems in their own right; cf. the following dialogue in Chomsky (1988: 187) between Chomsky and an interlocutor: Question: A child can learn two languages simultaneously, one in the house and the other on the street. Does this mean that the child relates the position of the switches [parameters] to the environment? Chomsky: This is a very important question which I have been pretending all along does not arise. The question is a very mysterious one. […] The child learns different languages, say Spanish at home and English in the streets. But in fact, the problem is really more general, because every human being speaks a variety of languages. We sometimes call them different styles or dialects, but they are really different languages, and somehow we know when to use them, one in one place and another in another place. Now each of these different languages involves a different switch setting.

Chomsky (1988: 187) adds that this multicompetence implies the existence of a theory of society and language in the language-acquiring child: “Somehow, young children have a theory of society and a theory of lan-

332

Kristin Melum Eide and Hilde Sollid

guage, and they are able to link them up in some fashion to indicate that you speak this language in this social situation.” Among the potentially far-reaching consequences of these observations is the plausible hypothesis that every individual masters a range of linguistic systems, i. e. grammars. (We return to the notion of grammar in section 2.4) This hypothesis is corroborated by the different syntactic rules that apply to different styles, contexts, or dialects. For example, poetic syntax is often seen as a language somehow exempt from the ordinary syntactic rules of the language. However, even in these special settings, the syntactic system is subject to rules, even though these rules are different from those applying to other written and spoken standards. In fact, the syntax of Norwegian hymns (example 1a, Barstad 2000) has much in common with German word order, with verb-final structures in embedded (and even some main) clauses and the reflexive object shifted to the left of the verb.3 Different rules apply in the corresponding clauses in prose, as (1b) shows. (1)

a. Om du if you

deg REFL

skjuler når angsten meg hide when fear.DEF me

plager … torments

b. Om du skjuler deg når angsten plager meg … if you hide REFL when fear.DEF torments me ‘If you conceal yourself when fear torments me.’

One might be inclined to conclude that this is a result of the fact that many traditional hymns used in the Norwegian church were translated from German. However, the origin of traditional hymns sung centuries ago is no explanation for the fact that modern Norwegian language users who have scarcely heard a word of German readily accept a German-like word order as grammatical in modern hymns, but reject that same word order in all other contexts. To give rise to these unanimous judgments, the syntactic rules of hymns must be internalized as a separate system. This system is also a grammar of Norwegian albeit a different one, one confined to hymns (cf. Eide and Åfarli 2007).

3 There is a well-known rule of ‘object-shift’ in Scandinavian syntax, shifting the object to the left of middle-field adverbials such as negation. However, it does not apply in constructions like (1) in standard syntax.

Norwegian main clause declaratives: variation within and across grammars

333

The average speaker of Norwegian will master at least one written norm (and usually two)4 and have productive competence in at least one spoken norm, but sometimes several. In addition, Norwegian speakers have receptive competence in a range of spoken variants. There are also distinct rules – and grammaticality judgments – for a number of other contexts. For instance, there seems to be a special headline syntax, cf. Fjeldstad (2000). The norms of ordinary written Norwegian require a visible subject in finite clauses (2a) while headlines frequently allow ‘subject drop’ (2b and 2c). A similar subject drop exists in diary syntax in languages usually disallowing subject drop in finite clauses; cf. Haegeman and Ihsane (2001). This is also true for Norwegian, as in (2d). The phenomenon is better described as ‘topic drop’, since it happens only in root clauses and only when the subject is the topic. Topic drop, although frequent in languages such as Mandarin Chinese, is not considered grammatical in the written standards of Norwegian. However, it is quite frequent in spoken informal language (2e); in this register, even the auxiliary may be left out (2f), unlike in more formal spoken styles or formal written prose. (2)

a. *(Israel) fortsetter bombingen. Israel continues bombing.DEF ‘Israel continues the bombing.’ b. Fortsetter bombingen. Fjernet bevis på massacre. continues bombing.DEF removed evidence of massacre ‘Continues the bombing. Removed evidence of massacre.’ c. Bekjemper krim med fights crime with ‘Fights crime with words.’

ord. words

d. Traff HAM met HIM ‘Met him again today.’

i dag. today

igjen again

e. [Den filmen] så æ [that film.DEF] saw I ‘That film I saw yesterday.’ f.

(Fjeldstad 2000)

i går. yesterday

Sett'-n Ola nå seen-CL.MASC Ola PTL ‘anything’ ‘Have you seen Ola (at all/anywhere)?’

du? you?

4 Norway has two written standards, Bokmål and Nynorsk, and Norwegians are trained in both in school. All official clerks are required to be fluent in both standards, and all official papers are supposed to exist in both standards.

334

Kristin Melum Eide and Hilde Sollid

Poems other than hymns are also characterized by a set of syntactic rules different from those applying to other spoken and written registers. This is not proof of anarchy in syntax; compare (3a), which is grammatical in a poetic context (a nursery rhyme) to (3b), which would be considered ungrammatical in all contexts. (3)

a. tre three

katter cats

små, small,

så so

søte cute

PTL



b. *katter små tre, så cats small three, so ‘Three small cats, so (very) cute.’

søte cute

PTL



What we call festive syntax (cf. Eide and Åfarli 2007) is also found in birthday greetings on local radio shows or in the newspaper. These contexts display non-V2 word order, while the corresponding contexts in a written norm, and even the standard and local spoken norm, would require V2. Moreover, the non-V2 word order does not serve purposes of rhyme or rhythm since this is prose. It seems that non-V2 word order is somehow considered more festive than V2. (4)

Den 8.8. Stina fyller 8 år. Hurra for Mia the 8.8 Stina is 8 years. Hooray for Mia som på Elveng bor! who on Elveng lives! ‘The 8th of August Stina is 8 years old. Hooray for Mia living in Elveng!’

As mentioned earlier, most speakers of dialects are bilingual: they have competence in a standard norm in addition to their local dialect. The local dialects are subject to their own syntactic rules, which can be quite different from those of the standard norms. To illustrate, the main clause whquestion in (5a) with a non-V2 structure (from a dialect in Nordmøre) would be ungrammatical as a main clause in the standard written norm (5b). (5)

a. Kåin du lika best? Who you like best? ‘Who do you like better?’ b. *Hvem du liker best? Who you like best? ‘Who do you like better?’

(Åfarli 1986)

Norwegian main clause declaratives: variation within and across grammars

335

The examples in (1)–(4) raise the question of whether these special registers require only receptive competence (for example, to recognize a headline as a headline) or whether they are productive registers used in everyday dialectal speech, as in (5a). The distinction between receptive and productive competence might be the demarcation line between core and peripheral grammar, cf. section 2.4. This is at its core a methodological question. A plausible hypothesis would be that most speakers of Norwegian perceive these grammars as poetry, lyrics, and birthday greetings, and that some (but not all) speakers of Norwegian also regularly produce those grammars in the right context albeit not every day. Furthermore, we would expect the frequency of these registers to vary during the speaker’s lifespan: for example, speakers produce more nursery rhymes when they are around small children on a regular basis. Determining whether a given grammar, say verb-final birthday greetings, is receptive or productive for a speaker is problematic: receptive competence has the potential to become productive and productive competence has the potential to become receptive. Thus, even grammars that seem receptive, i. e. seem to belong to the peripheral grammar, must be accounted for in a theory of grammar and grammatical variation. We elaborate on this in section 2.5. The list of examples of special context syntax is a lot longer, but the examples in (1)–(5) suffice to illustrate our point – in the context of specific styles, dialects, and registers, speakers perceive as grammatical structures deemed ungrammatical in other contexts. Moreover, these styles and registers do not display anarchy of syntactic rules; their syntactic systems are simply subject to a different set of rules. Finally, the shift between different styles regularly leads to different grammaticality judgments. This almost unanimous shift of grammaticality judgments suggests that these styles are, in Chomsky’s (1988) words, “really different languages”, i. e. different internalized grammars, each with its own syntactic rules.

2.3.

Universal Bilingualism

What kind of formal grammatical theory might account for intra-individual variation? In our opinion, Roeper’s (1999) theory of Universal Bilingualism takes us in the right direction. Roeper argues that the concept of “grammar” should be understood as a much more “local” notion, relativized to classes of elements in a speaker’s vocabulary. This implies that the number of grammars internalized in an individual is not confined to a set of distinct national languages, or a number of national languages plus a number of dialects. Instead, what we consider a national language or a

336

Kristin Melum Eide and Hilde Sollid

dialect actually consists of a number of languages, a number of simultaneously internalized grammars.5 Roeper (1999: 170) claims that a person has numerous grammars: every lexical class with rules that are incompatible with another class should constitute a separate grammar. It sounds unwieldy and implausible to argue that a person has a dozen grammars. The essence of this assertion may, nonetheless, be true. It implies that the notion of a grammar should change to a more local conception.

Roeper supports this claim with the observation that although English main verbs usually do not take part in inversion, in narratives verbs denoting acts of speaking regularly show up with verb second (V2) word order (quotative inversion). (6)

a. “I am melting,” shrieked the witch. b. “What are you two fine ladies up to tonight”, says Mike.

This can be explained if we assume that English speakers have two separate grammars for V2: one allowing V2 for main verbs of a specific semantic class and another disallowing V2 for all other main verbs. This is an instance where “two properties exist in a language that are not stateable within a single grammar” (Roeper 1999: 169) since the facts give rise to contradictory settings for a syntactic rule regarding V2 and main verbs. Stages in language acquisition and diachronic change are explained in this theory along the same lines; apparent optionality is recast as coexisting and competing linguistic subsystems. P&P-based theories do not cope well with optionality or gradualism of any brand. Instead, Roeper (1999: 169) says: “I proceed from the assumption that wherever one finds a continuum, or historical gradualism, a more refined level of analysis will reveal discrete phenomena.”

2.4.

So what is ‘a grammar’?

This question raises several important issues. Firstly, there is a theoretical and terminological issue: what constitutes ‘a grammar’ in this view? If Universal Bilingualism has merit, any and every data set from any speaker

5 See also Mensching and Remberger (this volume), in particular the discussion in their sections 2 and 5 on the locus of parametric variation.

Norwegian main clause declaratives: variation within and across grammars

337

will contain elements and structures from a set of sub-grammars. Each sub-grammar in principle constitutes its own linguistic system with choices and settings that may contradict those of another sub-grammar in the same speaker. Thus, what we dub a grammar, a dialect, or a language, on this view, is a somewhat arbitrarily chosen set of sub-grammars, where each set may be idiosyncratic to a speaker. In reality, this set of grammars constitutes an idiolect. This is comparable to the term ‘vocabulary’. No single speaker of Norwegian will actively employ all and only words that belong to the vocabulary of standard Norwegian. Sometimes he borrows words from other languages; sometimes he uses invented words or archaic and obscure dialectal words that existed as productive items only in the dialect of his grandparents. His idiosyncratic selection of vocabulary items constitutes a set of subsets, where each subset belongs to a specific register or style. Moreover, his selection of vocabulary items is not likely to overlap exactly with that of any other person, which is why we dub it idiosyncratic.6 If the grammars of individuals are also a set of subsets, it would be impossible to construct a single grammar for an individual, where the grammar generates all the individual’s sentences and constructions and consists of all and only contingent and consistent choices and settings. This would be comparable to putting together a coherent picture using puzzle pieces belonging to an unknown number of puzzles, with the requirement that all pieces be used. This would probably be a frustrating experience as well as a fruitless task.7 There is another matter that can strongly affect our methodological approach to the data. Given the assumption that several languages, i. e. grammars, can co-exist in the human brain, those grammars may influence each other. This raises the issue of possible “contamination” of grammars and grammaticality judgments. How can we know whether our judgments are “pure” in the sense that they stem from a single and contingent set of grammatical rules? Cook (2003) shows that this type of influence of one (L2) grammar on another (L1) does indeed occur. Some French speakers 6 Cf. also Eide and Åfarli (2007), quoting Kayne (1996: xiv-xv): “It is possible to arrive at a much more radical reevaluation based on the following question: Can anyone think of another person with whom they agree 100 percent of the time on syntactic judgements (even counting only sharp disagreements)? […] [I]t is entirely likely that no two speakers of English have exactly the same syntactic judgements. In which case there must be many more varieties of English than is usually assumed.” 7 Cf. also Eide and Åfarli (2007) for this analogy.

338

Kristin Melum Eide and Hilde Sollid

who speak English reject French sentences in the middle voice, while those who do not speak English accept them. This shows that under the influence of the L2, English, French speakers develop a dislike for constructions like (7a) in their native language since the corresponding construction is ungrammatical in English; cf. (7b) (from Cook and Newson 2007: 239). (7)

a. Un tricot de laine se lave a sweater of wool REFL wash ‘A wool sweater washes in cold water.’

à in

l’eau water.DEF

froide. cold

b. *A wool sweater washes in cold water.

Cornips (2006) reports similar patterns for Dutch dialects. She found that there are “intermediate standards” and shades of gradation between standard and dialect in production and also in grammaticality judgments. In one of her studies, an informant occasionally uses a strongly dialectal trait, the article et before the proper name; this informant is more likely to leave out the article when talking to a speaker of the standard Dutch variant (which does not employ this type of article). When talking to a speaker of his own local dialect, he uses the article much more frequently (i. e. accommodation, more on this in section 2.5). (8)

a. Der

Wim dach dat ich Ø Wim thought that I Ø han will geve. have will give ‘Wim thought I wanted to give a book to Els.’ DET

b. Der

Wim menet dat ich et Wim thought that I ART probeerd ha kado te geve. tried have gift to give ‘Wim thought I tried to give a book to Els.’

DET

Els Els

e a

boek book

Els Els

e a

boek book

A Universal Bilingualism approach to these phenomena would assume that this speaker has several subsets of grammars, where each subset can be described as one variety. The standard Dutch grammar does not employ the relevant article, but the local dialect does. Crucially, there are intermediate variants displaying traits from both grammars, where some sets of articles are used and others are left out. If each speaker has several grammars and one internalized grammar can influence others, how can we be sure which grammar we are looking

Norwegian main clause declaratives: variation within and across grammars

339

at in each case? How can we trust grammaticality judgments if the speaker cannot be sure from which of his internalized grammars he is reporting? Is it a productive or a receptive grammar? Determining what constitutes a grammar in the relevant sense, just like determining what constitutes a person’s sub-vocabularies, is not an easy task. It is tempting to assume that if two constructions8 are synonymous, with no differences in semantics, lexical items, information structure, or pragmatics, but have different word orders and different syntactic features, i. e. contradictory settings, we are dealing with two different grammars. However, it is debatable whether two constructions are ever completely synonymous on all these variables, especially if all types of connotations go into the equation; cf. Cheshire (2005) for a summary. Thus, in this paper we add a theory-instrumental dimension to this question: if two constructions are not stateable within a single system given the restrictions embedded in current P&P approaches, we are dealing with two different grammars. We augment these two sets of (synonymy and theory-restricted) guidelines with a third, even more crucial feature. If informants clearly use the construction in question as a sociolinguistic resource, we are dealing with an element belonging to a separate grammar. This is because of an unassailable logic hidden in the concept of a sociolinguistic resource: to use any variable as a sociolinguistic resource, a real opposition between two variables must be implied. Where no alternative is available, the sociolinguistic effect of using a certain construction must be zero. To illustrate this claim with an example, each Norwegian dialect employs its own set of demonstratives. In the southeastern Norwegian dialects close to the written standard Bokmål, the common demonstrative denne [this.DEF.NON-NEUTER] ‘this one’ is used for a proximal masculine noun. In the central Norwegian dialect Trøndersk, the corresponding demonstrative is herren [here.DEF.MASC] ‘this one’. These two elements are clearly constructed on the basis of different morphological rules and belong to two different linguistic systems. In a dialogue between two speakers one of whom speaks Trøndersk and the other southeastern Norwegian, if these two demonstratives were the only possible demonstratives for a proximal masculine noun, no sociolinguistic effect would arise whenever the speaker of Trøndersk used the demonstrative herren because

8 In this paper, we use the term construction in an informal way, as the visible output of a set of syntactic operations. We do not intend the term to encode the chunks of fixed form-meaning pairs employed in Construction Grammar and similar frameworks.

340

Kristin Melum Eide and Hilde Sollid

he would have no choice; this would be the only alternative to refer to a proximal masculine noun in his dialect. Similarly, for a speaker of southeastern Norwegian, no flag would be raised when he used the demonstrative denne since this would be the only possible demonstrative for a proximal masculine noun. As it happens, the aforementioned dialects do in fact make another alternative available, the demonstrative den her [the.DEF. NON-NEUTER here] ‘the one here’. This demonstrative is constructed by morphological rules different from those underlying herren and denne; that is, den her requires a different demonstrative-building sub-grammar. Crucially, the demonstrative den her and the corresponding morphological rules constructing it are common to both dialects, a fact likely to be known to or at least easily detectable by both speakers. Thus, the speaker has a choice. He can chose to use the demonstrative den her, which would be the unmarked variant in this context because it exists in both dialects, or he can use the demonstrative with the most dialect-specific feature (herren/denne). The speaker’s choice may signal his intention to be similar to or different from his interlocutor. Had the common denominator demonstrative den her not existed, no such choice would be available and no sociolinguistic effect could be read into the use of herren or denne as a demonstrative.

2.5.

The choice between grammars

As acknowledged earlier, the assumption of Universal Bilingualism is the debate on core vs. periphery within the generative framework in its extreme: what structures belong to the core or primary grammar? In the present view, there is no single core grammar, but a register of grammars, and the individual is free to choose the one that best suits the communicative purposes in a given situation. The choice is dependent on both linguistic and social factors. Roeper (1999: 171) is explicit about the need for social factors to explain this choice: “[…] we will continue to make vague reference to ‘social factors’ as an expression intended to cover a myriad group of factors which may determine the use of grammar but are not expressible in grammatical notation.” We agree that social factors are not expressible in grammatical notation, and in our integrative approach we emphasize the speakers’ ability to link grammars to a theory of society. In our opinion, the basic insights from speech accommodation theory (cf. Giles 1973; Giles and Powesland [1975] 1997) provide the needed complement to Universal Bilingualism in explaining the speakers’ choice between grammars when there is an

Norwegian main clause declaratives: variation within and across grammars

341

option. The insights of accommodation theory are discussed and adopted within sociolinguistic approaches to language variation and style (Bell 1984, 2001; Coupland 2007; Eckert and Rickford 2001; Trudgill 1986). At the heart of accommodation theory is the assumption that a speaker is able to change his speech to converge towards or diverge from his interlocutors. It is also proposed that a speaker can maintain the established style of interaction despite a speech partner’s attempt to converge or diverge. This short-term accommodation ability is not restricted to the choice of vocabulary or sound patterns,9 but is also relevant to the choice of grammatical variant. Ganuza (2008), for instance, found that adolescents in multilingual urban settings in Sweden use inversion and noninversion in declaratives, questions, and subordinated clauses in a socially significant manner. Accommodation is thus motivated by an (un)conscious wish to achieve solidarity with and approval by present or nonpresent receivers, or to dissociate oneself from present or non-present receivers, cf. Bell’s (2001) theory on audience and referee design.10 Short-term linguistic accommodation is not exclusive to adolescents or adults. Andersen (1990) discusses the sociolinguistic skills of children between the ages of four and seven. What we find particularly significant are dialect shifts in role-playing games performed by Norwegian children from the age of three. This phenomenon has not received much attention in research, so the details of this linguistically complex game are not fully understood. However, on the surface, children perform astonishing, rapid shifts between a local and a non-local variety. When they are planning the roles and the events of the role-play, they typically use their mother dialect. When they enter the roles, they typically shift and use a second variety, a standard or non-local variety that they may know from children’s

9 Trudgill (1986) differentiates between short- and long-term accommodation. Short-term accommodation is related to face-to-face interactions and does not necessarily result in a permanent change in the speaker’s speech. Long-term accommodation, on the other hand, permanently affects the speaker’s variety. See also Auer and Hinskens (2005). Here, we focus on short-term accommodation. 10 Different researchers emphasize different aspects of the accommodation process. Giles and Powesland ([1975] 1997) and Bell (2001) highlight the importance of the audience and the often-absent reference group, while ShillingEstes (2002, 2004) and Coupland (2007) underline speaker agency. The choice of variant is also dependent on the type of interaction, the discourse topic, and the social context (Giles and Powesland [1975] 1997).

342

Kristin Melum Eide and Hilde Sollid

television programs. These dialect shifts do not automatically become the unmarked way of speaking in later years. Instead multidialectism is rather limited, cf. section 3.1. The ability to shift from one variety to another, from one style to another, is thus considered fundamental, and the speaker becomes able to relate a specific way of speaking to a specific context. This leads to another important question: how does grammar get a social meaning? Following Eckert (2008), we argue that this semiotic process relies on a dynamic relationship: in using a specific linguistic resource speakers (subconsciously) call the attention to this resource, and other speakers (subconsciously) perceive this resource in relation to other resources. This leads to an opposition between linguistic resources. In interactions, the speakers produce and reproduce this opposition and pave the way for the development of the indexical social meaning attached to linguistic items (Eckert 2008; Silverstein 2003). The social meaning of linguistic items is thus dynamic: the form-meaning relationship always has the potential to change. Hypothetically, under given social conditions, even the highly marked grammar of festive syntax has the potential to spread to other registers or styles.

3.

Norwegian main clause declaratives

Section 2.2 presented data from genres such as birthday greetings, diaries, hymns, headlines, poetry and dialects, illustrating that speakers of Norwegian have several different grammars restricting subject-verb inversion and more generally the placement of the finite verb. Returning now to our empirical focal point, in what follows we report on our study of Norwegian declaratives in a data corpus of spoken Norwegian collected through research interviews. We gathered 2015 main clause declaratives from eight interviews with respondents from two dialect areas (Tromsø, Northern Norway, and Oslo, South-Eastern Norway). In section 3.1, we outline the research interviews and discuss some features of the corpus. In section 3.2, we present the general pattern for Norwegian main clause declaratives, and in section 3.3, we offer an analysis of the data.

3.1.

The research interviews

We examined 2015 main clause declaratives produced by eight respondents taking part in research interviews where a researcher or field assis-

Norwegian main clause declaratives: variation within and across grammars

343

tant is interviewing one respondent at the time.11 The four interviews from Tromsø constitute the main set of data and the four interviews from Oslo serve as the control.12 Comparing data from two different dialect areas allows us to uncover potential dialectal differences in the material. We categorize the declaratives in the corpus according to the type of phrase in the initial position and the placement of the finite verb. In general, Norwegian dialects are not confined to isolated rural communities. People use their local dialects for all purposes, even when speaking in front of Parliament or appearing in the media. Norwegians are exposed to multiple dialects because of migration within the country; they are also exposed to two written standards through the education system.13 This multilinguistic context does not however amount to extensive bi- or multidialectism in Norway. Rather, people have receptive competence in most varieties of Norwegian, and they may use other dialects for limited purposes. General productive competence in two or more dialects is certainly one possible outcome in a dialect contact situation, but today people usually maintain their dialect and accommodate towards the other dialects.14 11 Defining a main clause declarative is not always straightforward, and we readily admit that our decisions are not theory neutral. Besides clauses approaching written standard declaratives, we include restarted declaratives and declaratives where the verb changes after a restart. Each start-plus-correction counts as one example. Secondly, we include declaratives with a cut-off point either right after a verb or after a verb plus an argument. These cut-offs are categorized according to the properties of the phrase in the first position and the position of the finite verb. The five hours of recordings yielded a total of 2193 sentences (interviewer’s contribution discarded). We excluded main clause questions, imperatives, ellipses without a finite verb and elliptic subordinated clauses not embedded in a main clause. The resulting corpus consists of 2015 main clause declaratives. 12 The Tromsø data is from Hilde Sollid’s UPUS-project Linguistic variation in Tromsø (2006-2007) and the Oslo data from the NoTa-Oslo corpus http:// www.hf.uio.no/tekstlab/. (NoTa-Oslo is an acronym for Norsk Talemålskorpus – Oslodelen [Norwegian speech corpus – Oslo part]) We use this opportunity to thank Kristin Hagen at Tekstlaboratoriet for her help. 13 Note that there is no reference to the respondent’s language of instruction. 14 This used to be different, and dialect shift or bidialectism was quite common among people from Northern Norway who moved to Oslo prior to the “dialect wave”, i. e. before the 1970s. The picture is also different in language contact situations in Norway. Speakers of minority languages usually get Norwegian as their main language, and Norwegian minority policy traditionally favours language shifts.

344

Kristin Melum Eide and Hilde Sollid

To ensure that the data are comparable, we chose informants of the same age (30–45) with higher education; we also chose an equal number of males and females. Other than these two features, there was no attempt to match the social profiles of the informants. The eight research interviews are part of two larger datasets the main purpose of which is to study linguistic variation. The research interviews are structured. The NoTa-Oslo interviews centre on the residences of the speaker. The UPUS-Tromsø interviews are metalinguistic: the respondents were asked about their dialect and dialect use. This is known to trigger dialectal speech, but due to the semi-formal setting of the research interviews we expect the informants to (sub-)consciously monitor their speech and perhaps strive towards a more formal speech style. As we observe in the respondents’ oral language repeated features alien to the written standards, we ascribe these features to the spoken repertoires of the speakers. At this point, we cannot firmly say whether the features are part of a stylistic pattern or an act of stylizing. The interviews are quite serious, however, and the semi-formal setting is likely to limit the frequency of non-standard features. We also expect the speakers to refrain from using poetic syntax, for example, as the interview setting is not expected to promote the use of a poetic register. As Ganuza (2008) shows, the number of non-inversion in Swedish main clause declaratives is higher in informal, peer-to-peer settings than in formal contexts. The presence of non-standard features in the data is thus suggestive of features in the speakers’ interview repertoire.15

3.1.1. Tromsø versus Oslo The number of declaratives from the Tromsø dialect is higher even though the number of informants from Tromsø and Oslo is equal (1542 vs. 473 declaratives). This is easily explained: in the UPUS-Tromsø data the interviews last 45–80 minutes, while in the NoTa-corpus they are approximately 10 minutes. The interviews also have different structures and topics; in addition, there are two different interviewers in the data sets. Thus, we 15 In this study, we do not fully explore the potential for analyzing intra-individual variation and accommodation. That would require a more narrow investigation of the four main interviews similar to Schilling-Estes (2004), who combines microanalysis of a one-hour interview with a macro level survey of sociolinguistic variation.

Norwegian main clause declaratives: variation within and across grammars

345

cannot be sure whether the relationship between the main and control data would be different if the interviews were equally long, had equal structure, and used the same interviewer. Nevertheless, as we see in table 1 the distribution of initial position phrase types in the Tromsø and Oslo data is not very different (the difference is not statistically significant according to an unpaired t-test), and the distribution is not very different from that in other Scandinavian corpus studies (e. g. Ganuza 2008; Jörgensen 1976). We thus conclude that the Tromsø and Oslo data are comparable for our purposes.

3.2.

Norwegian main clause declaratives: overall picture

Norwegian is classified as an SVO-language: the basic word order is subject-verb-object, cf. (9a). This is also the most frequent word order pattern in our data; over 63 % of the 2015 declaratives are in this category (Table 1). Subject-initial declaratives, however, do not reveal much about subject-verb inversion patterns (at least not in clauses lacking a sentence adverb or negation). A language can be SVO without being V2; English is a case in point. True V2 languages reveal themselves through clauses where a non-subject phrase is topicalized. Topicalization causes a nonsubject phrase to appear in the left-most position of the clause; in such clauses, the finite verb is required to be in the second position, as in (9b). Absence of subject-verb inversion in such constructions gives rise to ungrammaticality in most varieties of Norwegian, both written and spoken, as in (9c). The data in (9) are thus the empirical cornerstone in the analysis of Norwegian as a V2 language. (9)

a. Gunnar spiser fisk Gunnar eats fish ‘Gunnar eats fish every Monday.’

hver every

mandag. Monday

b. Fisk spiser Gunnar Fish eats Gunnar ‘Fish Gunnar eats every Monday.’

hver every

mandag. Monday

c. *Fisk Gunnar spiser Fish Gunnar eats ‘Fish Gunnar eats every Monday.’

hver every

mandag. Monday

In our data, about 28 % of the 2015 declaratives display topicalization. The most frequent types of topicalization are those involving adverbials (22 %), as in (10a), and objects (5.2 %), as in (10b); topicalized subject

346

Kristin Melum Eide and Hilde Sollid

predicatives, as in (10c), are rare (0.6 %). We also found occasional topicalization out of prepositional phrases and subordinated clauses (0.4 %), as in (10d). (10)

a. Der hadde æ besteforeldran mine. There had I grandparents.DEF mine ‘There I had my grandparents.’ (Tromsø) b. Tyggegummipapiret tørket vi av på radiatorene her. Gum-paper.DEF swiped we off on radiators.DEF here ‘The gum paper we swiped off on these radiators.’ (Oslo) c. Og stille og rolig er det. And calm and quiet is it ‘And calm and quiet it is.’ (Oslo) d. ‘Meddag’ trur æ åsså værtfall faderen sa. Meddag believe I also at-least father.DEF said ‘‘Meddag’ I believe also that at least my father said.’ (‘meddag’ = dinner) (Tromsø)

As shown in section 2.2, topic drop occurs in Norwegian headline syntax, but is not confined to this genre; it is quite common in written prose (mimicking oral discourse) and in spoken dialogues (Faarlund, Lie and Vannebo 1997). In our data, 8.3 % of the declaratives have topic drop where the topic equals the subject (11a); we also found examples where the topic is a non-subject (11b). Both types of topic drop yield V1 word order. (11)

a. hører at di snakker finere på vestkanten. hear that they speak posher on west-end.DEF ‘[I] hear that they speak posher in the west end.’ (Oslo) b. va ho å kjøpte hus der. was she and bought house there ‘[Then] she went and bought a house there.’ (Tromsø)

Table 1 summarizes the distribution patterns of initial phrase types in the declaratives in our corpus. It can be readily observed that speakers from both Tromsø and Oslo show a strong preference for subject initial declaratives, while topicalizations and finite verb initial declaratives are a lot less frequent.

Norwegian main clause declaratives: variation within and across grammars

347

Table 1: Main clause declaratives in Tromsø and Oslo: Phrase in the initial position. Initial position

Tromsø

Oslo

Total

Total (N=Number)

N=1542

N=473

N=2015

Subject initial

N=982 (63.6 %)

N=292 (61.7 %)

N=1274 (63.2 %)

Topicalization

N=434 (28.1 %)

N=138 (29.1 %)

N=572 (28.3 %)

Finite verb initial

N=126 (8.1 %)

N=43 (9 %)

N=169 (8.3 %)

Table 1 focuses on the initial element of the clause and does not tell us anything about where the finite verb is, except for 8.3 % of the declaratives where the verb is the initial element of the clause. Examining our data for finite verb placement produces the results in Table 2. Table 2: Main clause declaratives in Tromsø and Oslo: Position of the finite verb. Finite verb

Tromsø

Oslo

Total

Total (N=Number)

N=1542

N=473

N=2015

V2

N=1267 (82.1 %)

N=401 (84.7 %)

N=1668 (82.7 %)

V3

N=149 (9.6 %)

N=29 (6.1 %)

N=178 (8.8 %)

V1

N=126 (8.1 %)

N=43 (9 %)

N=169 (8.3 %)

We can see that V2 declaratives are very frequent. Tables 1 and 2 together reveal that the typical Norwegian declarative is a subject-initial V2 clause. This is no surprise. What is surprising, however, is the number of other patterns attested in our corpus. We consider as the most important finding that as much as 17 % of the declaratives are non-V2 declaratives: 8.8 % are V3 declaratives and 8.3 % are finite verb initial declaratives, topic drop declaratives where the finite verb in the first position (V1). Non-V2, which occurs in 17 % of all tokens, cannot possibly be considered a marginal phenomenon. We conclude that the V2-rule is not an absolute rule, at least not in spoken interviews, which constitute the corpus in the present study.

348

Kristin Melum Eide and Hilde Sollid

3.3.

Analyzing the data: how many grammars?

To decide whether the different constructions in our data should be ascribed to different sub-grammars, i. e. separate linguistic systems, we utilize a strongly modified version of a fairly widespread P&P analysis of V2, the verb raising analysis (Den Besten 1983). In the traditional version of this analysis, the V2 effect derives from a feature of a syntactic head associated with assertion and finiteness, attracting the finite verb to the second position of the clause in main clause declaratives. This accounts for the generalizations that a) only the finite verb moves past negation in V2 clauses, as in (12a) and (12b); b) the non-finite verb does not move past negation (12c); and c) embedded clauses headed by a complementizer are usually not V2, since the syntactic head attracting the verb in main clauses is occupied by the complementizer in subordinate clauses, preventing the verb from moving (12d). In these clauses, even a finite verb will not move past negation. (12)

a. Jeg drikker ikke øl på fredager. I drink not beer on Fridays ‘I don’t drink beer on Fridays.’ b. På fredager drikker jeg ikke øl. On Fridays drink I not beer ‘On Fridays I don’t drink beer.’ c. Jeg ikke drikke øl på en fredag? Særlig.16 I not drink beer on a Friday? As-if. ‘I not drink beer on a Friday? As if.’ d. … ettersom jeg ikke drikker øl på fredager. … since I not drink beer on Fridays ‘… since I don’t drink beer on Fridays.’

What we categorized as V2 declaratives in Table 2 are both subject initial declaratives and topicalizations. In clauses employing topicalization, determining whether or not the finite verb is in V2 is straightforward: the V2 position is between the topicalized phrase and the subject (12b). It is a lot less straightforward to decide whether subject initial declaratives employ verb raising, unless the clause contains a sentence adverbial like clausal negation, as in (12a). Subject initial declaratives without sentence adverb or negation reveal a lot less about V2; the word order in these

16 The example is from Eide (2009).

Norwegian main clause declaratives: variation within and across grammars

349

cases could result from the cross-linguistically much more general SVO pattern instead of the V2 rule. In our data, there are 1219 subject initial V2 declaratives and 58 % lack a sentence adverb or negation. However, we will assume, with the mainstream P&P analysis of V2, that verb raising occurs even in these subject initial declaratives although it is string-vacuous (does not affect word order).17 Thus, all these structures can be accommodated by the same grammar without any contradictory settings; this grammar applies to the spoken register. The same can be argued for the topic drop V1 structure in (11a) and (11b). This is a feature of a sub-grammar utilized in informal spoken discourse, the register we are considering here. Thus, we can implement this feature of our grammar as a zero topic, i. e. this specific grammar allows the realization of the (pragmatically salient) topic as an element that is phonologically zero. The interesting part of our material resides in the V3 structures. Various exceptions to the V2 rule, dubbed exceptional V3, have been acknowledged for Norwegian. They are usually described either as limited to very specific construction types or associated with specific words and phrases. Faarlund, Lie and Vannebo (1997) note that adverbs such as bare (‘only’), mon (‘perhaps’), kanskje (‘perhaps’), and liksom (‘like’) sometimes give rise to exceptional V3, as in (13a) from Faarlund, Lie and Vannebo (1997: 821). Various left dislocations fall under a similar category, as in (13b) from the newspaper Nettavisen (July 2007). (13)

a. Vi berre We just ‘We just danced.’

dansa. danced

b. Seieren til tross, sveitseren er ikke fornøyd Victory.DEF inspite, Swiss.DEF is not pleased med elektronikken i profesjonell tennis.18 with electronic.DEF in professional tennis ‘In spite of the victory, the Swiss are not satisfied with the electronics in professional tennis.’

17 There have been several proposals over the years that subject initial declaratives in Germanic languages give rise to a structure different from non-subject initial V2-clauses. Van Craenenboeck and Haegeman (2007) revise the different analyses arguing for and against verb raising, respectively, and conclude that only the verb raising analysis can account for their new set of data. 18 We thank Curt Rice for this example.

350

Kristin Melum Eide and Hilde Sollid

These exceptions to the V2 rule are not confined to spoken Norwegian; some of them have a certain archaic flavor. For instance, (13b) is fine in writing, but seems less felicitous in spoken Norwegian; the adverb mon mentioned earlier would hardly ever be used by anyone in spontaneous speech unless they were over 80 and living in a wealthy neighborhood. Thus, these exceptional V3 are not typical of the V3 declaratives found in our corpus, where the informants are much younger and the setting less formal. A different construction, the så-construction, is much more frequent in the V3 declaratives in our data. This construction has received some attention in the literature. Faarlund, Lie and Vannebo (1997: 817) analyze så as an adverbial occurring in left dislocation structures, where så marks the left edge of the main clause, with a topic to its left, as in (14). (14)

I går [så spiste ikke Gunnar yesterday then ate not Gunnar ‘Yesterday, (then) Gunnar didn’t eat fish.’

fisk]. fish

We suggest a different approach. Building on the observations in Østbø (2006), we assume that så is a clause internal particle, filling the position of a syntactic head between the topic – which we assume to be clause internal as well – and the verb position.19 In our analysis (cf. Sollid and Eide 2007), these structures are different from the proper left dislocation structures in (13b): structures like those in (14) are ordinary topicalization structures, albeit with an extra particle between the topic and the V2 position, which is demoted to a V3 position. How can we fit these facts into the single sub-grammar we have been developing? Does this amount to the same grammar or to a different sub-grammar? This is where we need to undertake the previously announced strong modification of the traditional V2-as-verb-raising analysis. We propose that the V3 structure in (14) is the maximal realization of a Norwegian

19 It should be noted here that this is not necessarily the case for all types of constructions where så occurs. There is a range of constructions where the particle så plays a prominent part; så sometimes acts as an ordinary temporal adverbial, sometimes as a complementizer, sometimes as a degree adverbial, etc. The construction we refer to as the så-construction has some special properties (cf. Østbø 2006) only some of which are shared with the conditional and temporal constructions we find in many other languages, e. g. the German conditional and temporal constructions employing dann; however, the type of såconstruction we are discussing here is also found in Swedish and Danish.

Norwegian main clause declaratives: variation within and across grammars

351

main clause declarative, with all positions occupied. In our proposal, the structure needed to account for (14) is the underlying structure for virtually all main clause declaratives; in many cases, the particle position – the one occupied by så in (14) – is not filled or is filled by an element that is phonologically zero. We believe this is the so-called Wackernagel position found in many Indo-European languages,20 and we argue that this position has agreement-like properties in Norwegian. Schematically, this gives rise to an underlying structure for Norwegian main clause declaratives as in (15), where some of the slots are optionally left empty (or sometimes, phonologically zero). NEG is negation, SA is sentence adverb. (15)

Topic Æ ParticleAGR Æ Finite VerbÆ NEG/SAÆ SubjectÆNEG/SA

It may seem like overkill to propose this structure for all main clause declaratives to account for one particular construction type, i. e. the såconstruction. This representation implies that all Norwegian main clause declaratives are underlyingly V3, entailing the radical claim that Norwegian is a V3 language. However, this may be justified. Firstly, the så-construction is not as marginal as often portrayed; it is very frequent in spoken discourse and increasingly spreading also to written discourses (e. g. newspapers). Recall from Table 1 that 28.3 % of all declaratives in this corpus are topicalization structures. As mentioned earlier, the most frequent topicalization type is topicalization of adverbials, found in 22 % of all declaratives. On closer inspection, we discover that 120 examples of 445 the topicalized adverbials (27.1 %) are V3 structures; the så-construction represents a large part of these 120 tokens. The particle så appears after a fronted subordinated sentence (16a), a fronted adverb or PP (16b). There are also examples where the adverb da (‘then’) behaves as så (16c).

20 The format of this paper prevents us from going into all the arguments for assuming that this is the Wackernagel position, but let us at least quote Cardinaletti and Starke (1999: 195) who assume that weak and clitic elements must be licensed in fixed syntactic positions (identified as the ‘Wackernagel position’ in German). Also, weak and clitic elements need a prominent antecedent in the discourse or clause (like these particles), they are ‘less referential’ than strong pronouns, and they undergo phonological restructuring (in our case, they lean to a host to their left). All characteristics apply to the clitic particle discussed here. Also, Wackernagel (1892: 342) himself says: “Die Vorliebe […] für die zweite Stelle im Satz gehört nun aber in einen grösseren Zusammenhang hinein.”

352 (16)

Kristin Melum Eide and Hilde Sollid a. sånn når de blei etablert så va de mange like when it got established then was it many læga som bodde dær doctors who lived there ‘When it got established, there were many doctors who lived there.’ (Tromsø) b. men i dag så e de jo but to-day then is it PTL ‘But today it is totally different.’ (Tromsø)

helt annerledes totally different

c. men da æ va yngre da va de mer but when I was younger then was it more sånn at man … like that you … ‘But when I was younger it was more like that you …’ (Tromsø)

This particle+V3 structure is by no means limited to adverbial topicalization structures. We also find V3 in subject initial declaratives, a construction with striking similarities to the så-construction. Here, a pronominal particle appears between the subject and the finite verb (17a);21 it agrees with the topic, in this case the plural nominal topic.22 (17)

a. å [alle artistan [dem har jo ei vanvitti and all artists.DEF they have yes a huge platesamling sant record-collection right ‘And all the artists have like a huge record collection, right.’ (Tromsø)

If this analysis has merit, all V1, V2, and V3 structures discussed so far can be accommodated within the same sub-grammar because no structure

21 Dutch has a very similar construction discussed by Weerman (1988: 58): Jan die heft een auto gekocht ‘Jan that-one has a car bought.’ Weerman analyses this construction exactly the way we analyse the Norwegian construction in (17a). 22 In a forthcoming paper, we analyse expletive presentational constructions in the same V3 clausal format as (15), with a phonologically zero topic. The agreeing particle in these cases is the expletive. (i) Topic Æ ParticleAGR Æ Finite verb Æ NEG/SA Æ subject (ii) Ø Æ EXPLETIVE Æ COPULA Æ NEG/SA Æ Ø

Norwegian main clause declaratives: variation within and across grammars

353

gives rise to contradictory settings for the rule of verb raising. The apparent optionality in this sub-grammar is accounted for by assuming that syntactic heads and clause initial topics can be realized as phonologically zero, which is well within mainstream reasoning in current P&P theory.23 A V3 structure superficially similar to those in (16) and (17) is documented in ethnolects in Northern Norway and in multiethnolects in Oslo (cf. Opsahl 2009; Sollid 2005; Svendsen and Røyneland 2008). In Northern Norway, we regularly find non-V2 in declaratives (cf. 18ab from Sollid 2005: 140, 216, and 18c from the present study).24 (18)

a. Når vi kom dit, så vi fikk alt som vi when we arrived there, then we got all that we ville ha. wanted have. ‘When we arrived there, we got all that we wanted to have.’ b. Før vi hadde jo lov å sette earlier we had PTL permission to place yngel i elva. spawn in river-DEF. ‘Earlier we were allowed to place spawn in the river.’

ut out

c. så derfor vi flytta litt rundt omkring.25 so therefore we moved a-little around. ‘Therefore we moved around a little.’ (Oslo)

On closer inspection, however, the similarity between (16), (17), and (18) does not hold: the declaratives in (16) and (17) have subject-verb inversion, i. e. verb raising, while the declaratives in (18) do not. Unlike (16) and 23 The exact conditions triggering the realizations of these particles deserve a paper on their own, and at present we do not know much about them. We know that the particle is more likely to be realized, i. e. visible, when the topic is heavy and complex (cf. Sollid and Eide 2007). We also know that the topic may be realized as zero only when the referent is contextually salient (and on our analysis, also when the particle is an expletive). However, much work remains to be done to formulate the principles allowing for a zero realization of these topics and particles. 24 Non-inversion in main clause wh-questions, on the other hand, is well documented in a range of Norwegian dialects (Åfarli 1986; Nilsen 1996; Vangsnes 2006; Westergaard 2005). 25 The så in this example is not the clitic-particle så, but a type of adverb occurring with derfor; så derfor ‘so therefore’ is probably some type of collocation. Thus, (18c) still conforms to the rule in (19).

354

Kristin Melum Eide and Hilde Sollid

(17), the clauses in (18) are based on a rule requiring the verb to remain in its base position to the right of the subject. Thus, the finite verbs in (18b) and (18c) show up in the third position and in (18a) they are in the fourth position. As with particles in verb raising contexts, we argue that the presence or absence of particles in the non-verb raising contexts in (18) can be accounted for with the same sub-grammar.26 The underlying structure for the declaratives in (18) obeys a different rule from the ones in (16) and (17); compare the structure in (19) to the one in (15). The defining trait is that the verb occurs to the right of the subject (i. e. non-inversion). (19)

Topic Æ ParticleAGR Æ (SA)Æ SubjectÆ(SA) ÆFinite Verb (SA)27

Lack of subject-verb inversion is well documented in Mainland Scandinavian as a second language (Brautaset 1996; Hyltenstam 1978), ethnolectal varieties of Scandinavian languages in Stockholm, Göteborg, and Malmö in Sweden (Ganuza 2008), and Köbenhavn (Quist 2008). Our study does not include second language speakers of Norwegian, or speakers who grew up in environments similar to those of the respondents in the other Scandinavian studies. This kind of non-V2 nevertheless exist in our data, cf. example (18c). This is the type of construction where we need to postulate a different sub-grammar. If all main clause declaratives in Norwegian obey the restriction of verb raising (as has been argued for the constructions considered so far), the lack of such inversion in (18) gives rise to contradictory settings with respect to this rule (recall Roeper’s 1999 discussion of V2 in English in section 2.3). This is where we revert to the theory-instrumental restriction on grammars discussed in section 2.4. Moreover, the structures without inversion in (18) are seemingly quite synonymous with the corresponding structures employing inversion, so even this criterion on different sub-grammars seems to be fulfilled. The most convincing criterion to support the hypothesis that non-inversion declaratives belong to a different sub-grammar, in our view, stems from the fact that language users clearly utilize these non-inversion structures as sociolinguistic resources. As discussed above, employing a construction as a sociolinguistic resource entails that that construction stands in opposition to another (usually less marked) structure. This is what Ganuza (2008: ii) concludes about these structures for her informants: 26 In this context, the presence or absence of så does not lead to different grammaticality judgments in the ethnolects that have non-inversion as an option, cf. Sollid (2005). 27 SA=sentence adverbial can also be negation, compare to (15).

Norwegian main clause declaratives: variation within and across grammars

355

The use of non-inversion appears to be part of some adolescents’ spontaneous language use in certain contexts. More importantly, however, the results suggest that some adolescents employ non-inversions as an active linguistic resource to express their identification with the multilingual environment and the different varieties of Swedish spoken there, to show solidarity with peers, to contest official school discourses, and to play around with linguistic stereotypes.

The same conclusion can be drawn from Sollid’s (2005) study of noninversion in Northern Norway (cf. also Sollid and Eide 2007).28 Speakers eager to identify with their local dialect and the dialect of their parents and grandparents are much more likely to judge non-inversion main clause structures as grammatical than speakers less eager to identify with their local dialect. Again, this supports our analysis of non-inversion V3 structures as belonging to a different sub-grammar used by the speaker to express in-group identity and solidarity in a given speech event.29

4.

Summary

In this paper we discussed variations in the placement of the finite verb in Norwegian main clause declaratives, aligning our discussion with Universal Bilingualism. We conclude that most of the variation found in Norwegian declaratives can be accommodated within the same register or subgrammar. One V3 construction, however, the non-inversion V3 structure, belongs to a different subsystem used by speakers as a sociolinguistic resource (cf. also Ganuza 2008).

Abbreviations = agreement; ART = article; CL = clitic pronoun; DEF = definite article or definite suffix; DET = determinator; MASC = masculine; NEG = negation; PTL = particle; REFL = reflexive pronoun; SA = sentential adverbial. AGR

28 See also the discussion on the relation between the structure of a language and its socio-historical conditions in Enger (this volume: section 1). 29 Of course there might also exist other reasons why these V3 or lack-of-inversion structures show up in certain contexts for bilingual speakers, cf. the discussion in Matras (this volume, section 3): “[T]he need to constantly select among parallel structures overburdens the speaker and serves as a potential trigger for malfunctions of the processing mechanism […]”

356

Kristin Melum Eide and Hilde Sollid

References Andersen, Elaine Slosberg 1990 Speaking With Style. The Sociolinguistic Skills of Children. London/ New York: Routledge. Auer, Peter and Frans Hinskens 2005 The role of interpersonal accommodation in a theory of language change. In: Peter Auer, Frans Hinskens and Paul Kerswill (eds.), Dialect Change. Convergence and Divergence in European Languages, 335–357. Cambridge: Cambridge University Press. Åfarli, Tor A. 1986 Some syntactic structures in a dialect of Norwegian. Working Papers in Linguistics. University of Trondheim. Barstad, Åse Kristin 2000 Salmers syntaktiske struktur: En analyse av leddstillingen i norske salmevers [The syntactic structure of hymns: an analysis of word order in Norwegian hymns]. Master thesis, Trondheim: NTNU. Bell, Alan 1984 Language style as audience design. Language in Society 13/2: 145– 204. Bell, Alan 2001 Back in style: Reworking audience design. In: Penelope Eckert and John R. Rickford (eds.), Style and Sociolinguistic Variation, 139–169. Cambridge: Cambridge University Press. Bentzen, Kristine and Øystein A. Vangsnes (eds.) 2007 Scandinavian dialect syntax 2005, special issue of Nordlyd – Tromsø University Working Papers in Language & Linguistics. Tromsø: The University Library of Tromsø. Blom, Jan-Petter and John J. Gumperz [1970] 2000 Social meaning in linguistic structure: Code-switching in Norway. In: Li Wei (ed.), The Bilingualism Reader, 111–136. London/New York: Routledge. Brautaset, Anne 1996 Inversjon i norsk mellomspråk [Inversion in Norwegian interlanguage]. Oslo: Novus forlag. Cardinaletti, Anna and Michal Starke 1999 The typology of structural deficiency: A case study of the three classes of pronouns. In: Henk van Riemsdijk (ed.), Clitics in the Languages of Europe, 145–233. Berlin: Mouton de Gruyter. Cheshire, Jenny 2005 Syntactic variation and spoken language. In: Leonie Cornips and Karen P. Corrigan (eds.), Syntax and Variation. Reconciling the Biological and the Social, 81–106. Amsterdam: Benjamins. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.

Norwegian main clause declaratives: variation within and across grammars

357

Chomsky, Noam 1986 Knowledge of Language: Its Nature, Origin, and Use. New York: Praeger. Chomsky, Noam 1988 Language and Problems of Knowledge. Cambridge, MA: MIT Press. Chomsky, Noam 2000 The Architecture of Language. Oxford: Oxford University Press. Cook, Vivian 2003 The Effects of the Second Language on the First. Clevedon: Multilingual Matters. Cook, Vivian and Marc Newson 2007 Chomsky’s Universal Grammar: An Introduction. Malden: Blackwell. Cornips, Leonie 2006 Intermediate syntactic variants in a dialect–standard speech repertoire and relative acceptability. In: Gisbert Fanselow, Caroline Fery, Matthias Schlesewsky and Ralf Vogel (eds.), Gradience in Grammar, 85–105. Oxford: Oxford University Press. Coupland, Nikolas 2007 Style. Language Variation and Identity. Cambridge: Cambridge University Press. Den Besten, Hans 1983 On the interaction of root transformations and lexical deletive rules. In: Werner Abraham (ed.), On the Formal Syntax of the Westgermania, 226–273. Amsterdam: John Benjamins. Eckert, Penelope 2005 Variation, convention, and social meaning. Paper presented at the Annual Meeting of the Linguistic Society of America. Oakland CA. Jan. 7, 2005. Downloaded 22.05.2009 from: http://www.stanford. edu/~eckert/thirdwave.html Eckert, Penelope 2008 Variation and the indexical field. Journal of Sociolinguistics 12/4: 453–476. Eckert, Penelope and John R. Rickford (eds.) 2001 Style and Sociolinguistic Variation. Cambridge: Cambridge University Press. Eide, Kristin M. 2009 Finiteness. The haves and the have-nots. In: Artemis Alexiadou, Jorge Hankamer, Thomas McFadden, Justin Nuger & Florian Schäfer (eds.), Advances in Comparative Germanic Syntax. Amsterdam: Benjamins. Eide, Kristin M. and Tor A. Åfarli 2007 Dialektsyntaks og parallelle grammatikker [Dialect syntax and parallel grammars]. Proceedings from the 8th Nordic Dialectology Conference, Ord og Sag, Århus: Peter Skautrup Centret for Jysk Dialek-

358

Kristin Melum Eide and Hilde Sollid

tforskning. (English version forthcoming in Hrafnbjargarson, Vangsnes, and Østbø (eds.), Nordlyd, Tromsø Working Papers in Language and Linguistics.) Faarlund, Jan Terje, Svein Lie and Kjell Ivar Vannebo 1997 Norsk referansegrammatikk [Norwegian reference grammar]. Oslo: Universitetsforlaget. Fjeldstad, Marthe 2000 ‘Drept av krimliga’. En syntaktisk og pragmatisk analyse av fragmentariske avisoverskrifter i norsk [‘Killed by crime league’. A syntactic and pragmatic analysis of fragmented headlines in Norwegian]. Master thesis, Trondheim: NTNU. Ganuza, Natalie 2008 Syntactic variation in the Swedish of adolescents in mutilingual urban settings. Subject-verb order in declaratives, questions and subordinate clauses. Ph.D. dissertation, Centre for Research on Bilingualism, Stockholm University. Giles, Howard 1973 Accent mobility: A model and some data. Anthropological Linguistics 15: 87–105. Giles, Howard and Peter Powesland [1975] 1997 Accommodation theory. In: Nikolas Coupland and Adam Jaworski (eds.), Sociolinguistics. A Reader and a Coursebook, 232–239. Hampshire, NY: Palgrave. Hageman, Liliane and Tabea Ishane 2001 Adult null subjects in the non-pro-drop languages: Two diary dialects. Language Acquisition 9: 329–346. Holmberg, Anders and Christer Platzak 1995 The Role of Inflection in Scandinavian Syntax. Oxford: Oxford University Press. Hyltenstam, Kenneth 1978 Variability in interlanguage syntax. Working Papers, 18. Lund: Lund University, Department of linguistics. Jörgensen, Nils 1976 Meningsbygnaden i talad svenska [Sentence structure in spoken Swedish]. Lund: Lundastudier i nordisk språkvetenskap. Kayne, Richard 1996 Microparametric syntax: Some introductory remarks. In: James R. Black and Virginia Motapanyane (eds.), Microparametric Syntax and Dialect Variation, ix–xviii. Amsterdam: Benjamins. Labov, William 1966 The Social Stratification of English in New York City. Washington, D. C. : Center for Applied Linguistics. Nilsen, Hilde 1996 Koffer dæm sier det? Spørresetninger i Nordreisadialekten [Why

Norwegian main clause declaratives: variation within and across grammars

359

they say that? Interrogatives in the Nordreisa dialect]. Master thesis, Faculty of languages and literature, University of Tromsø. Opsahl, Toril 2009 Enkelt og tøft: Non-V2 i dekalarativsetninger med topikaliserte elementer hos ungdommer i multietniske miljøer i Oslo [Simple and tough: Non-V2 in declarative clauses with topicalized elements from youngsters in multiethnic environments in Oslo]. Part of doctoral thesis “Egentlig alle kan bidra!” – en samling sosiolingvistiske studier av strukturelle trekk ved norsk i multietniske ungdomsmiljøer i Oslo. [“Actually everyone can contribute!” A collection of sociolinguistic studies on structural traits in Norwegian spoken in multiethnic young environments in Oslo]. Doctoral Thesis, Oslo University. Quist, Pia 2008 Sociolinguistic approaches to multiethnolect: Language variety and stylistic practice. International Journal of Bilingualism 12: 43–61. Roeper, Tom 1999 Universal bilingualism. Bilingualism 2/3: 169–186. Schilling-Estes, Natalie 2002 Investigating stylistic variation. In: J. K. Chambers, Peter Trudgill, and Natalie Schilling-Estes (eds.), The Handbook of Language Variation and Change, 375–401. Malden/Oxford/Victoria: Blackwell Publishing. Schilling-Estes, Natalie 2004 Constructing ethnicity in interaction. Journal of Sociolinguistics 8/2: 163–195. Silverstein, Michael 2003 Indexical order and the dialectics of social life. Language and Communication 23: 193–229. Sollid, Hilde 2005 Språkdannelse og –stabilisering i møtet mellom kvensk og norsk [Language creation and stabilization in the encounter between Kven and Norwegian]. Oslo: Novus. Sollid, Hilde and Kristin M. Eide 2007 On verb second and the så-construction in Mainland Scandinavian L2. In: Marit Westergaard and Merete Andersen (eds.), Proceedings from Workshop on Language Acquisition, 22nd Scandinavian Conference of Linguistics. Nordlyd: 34.3:7–28. Svendsen, Bente Ailin and Unn Røyneland 2008 Multiethnolectal facts and functions in Oslo, Norway. International Journal of Bilingualism 12: 63–83. Trudgill, Peter 1986 Dialects in Contact. Oxford: Basil Blackwell.

360

Kristin Melum Eide and Hilde Sollid

Van Craenenbroeck, Jeroen and Liliane Haegeman 2007 The derivation of subject-initial V2. Linguistic Inquiry 38–1: 167– 178. Vangsnes, Øystein A. 2006 Microparameters for Norwegian wh-grammars. Linguistic Variation Yearbook 5: 187–226. Vikner, Sten 1995 Verb Movement and Expletive Subjects in Germanic Languages. Oxford: Oxford University Press. Wackernagel, Jacob 1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen I, 333–435. Weerman, Fred 1988 The V2 Conspiracy. A Synchronic and Diachronic Analysis of Verbal Positions in Germanic Languages. Dordrecht: Foris publications. Weinreich, Uriel 1953 Languages in Contact. Findings and Problems. The Hague/Paris/ New York: Mouton. Westergaard, Marit R. 2005 Optional word order in wh-questions in two Norwegian dialects: A diachronic analysis of synchronic variation. Nordic Journal of Linguistics 28/2: 269–296. Westergaard, Marit and Øystein A. Vangsnes 2005 Wh-questions, V2, and the left periphery of three Norwegian dialect types. Journal of Comparative Germanic Linguistics 8: 119–160. Østbø, Christine Bjerkan 2006 The Norwegian function word så and Norwegian CP syntax. Talk at Workshop on Inversion and verb Movement, University of Tromsø. Handout: http://uit.no/castl/V2workshop/5.

Syntactic variation and change in Romance: a Minimalist approach Guido Mensching and Eva-Maria Remberger Abstract The main objective of this article is to expose the fundamental ideas underlying a project on the basic syntax of the major Romance languages and to illustrate the appropriate formalisms within which their syntactic description and analysis might be implemented. The project is theoretically situated in the context of Generative Grammar, in particular recent developments of the Minimalist Program. One of its main assumptions is that lexical items as well as grammatical items (i. e. the functional categories) are stored in the mental lexicon, and that parametric variation in syntax is exclusively dependent on the feature composition of functional categories. The article offers some minimalist case studies in comparative Romance syntax, taking into account the modern as well as the historical varieties. It shows that syntactic change, synchronic variation and apparent optionality can indeed be modeled in an explanatorily satisfying way within the proposed formalism. Keywords: basic word order, diachronic variation, functional categories, information structure, lexicon, Minimalist Program, null subjects, parametrization, Romance, syntactic variation

1.

Introduction

1.1.

Aims and structure of this article

The study of language variation has a longstanding tradition within Romance philology. It was the starting point of Romance linguistics in the 19th century: in their effort to explain the historical development of the different Romance languages, scholars looked at the group of Romance languages as a whole. Since the second half of the 20th century, linguistics has achieved many insights into language, partly due to the cognitive turn. Modern Generative Grammar, e. g., has quite an elaborated theory on how syntax might work and how it interacts with other components of

362

Guido Mensching and Eva-Maria Remberger

human language. However, at least with respect to Romance, the original broad comparative view has disappeared, i. e. there has been no attempt to elaborate a syntax of all Romance languages, using the formalisms of Generative Grammar. Instead, we either find detailed studies on one language or detailed studies on various languages that are restricted, however, to one phenomenon. The aim of this article is to introduce a project named “Components of Romance Syntax”.1 Since the project had just begun when this article was finished, we will describe the basic framework and some first results, which will of course have to be refined and reviewed during the next years. In line with the subject of this book, we try to show how syntactic variation and change can be understood and formalized within our project. The general idea of the project is to write a kind of “basic syntax” of the Romance languages, which focuses on word order in the fundamental clause types and core constructions in ideally all Romance languages. That is to say, instead of an in-depth-study of either one Romance language or of one phenomenon across several Romance languages, the idea is to go breadth-first and, eventually, depth-second. The result of this kind of research should show 1. what the Romance languages have in common and 2. in what they diverge, and, most important, the underlying properties of 1 and 2. As will become clear in the course of this article, the minimalist framework is particularly appropriate for this purpose, due to its inherent lexicalism and basic principles such as economy and simplicity. One of the main points that we will try to show is that, in order to make the syntax of languages easily comparable with the aid of the minimalist framework, a unitary formalism of description and parametrization should be used. It will be shown that most differences between languages can be captured in the lexicon, where lexicon is understood in a rather broad

1 This article describes the project “Bausteine romanischer Syntax” (Freie Universität Berlin/University of Konstanz), which has been funded by the Deutsche Forschungsgemeinschaft (DFG) starting from November 2008. Most parts of this article have been written in phases previous to the project launch and correspond to the original ideas that motivated the project. We thank the DFG for their support and, in addition, Anja Weingart, Ion Giurgea and Lina Mavroudi for their revision of this article, as well as Peter Siemund for giving us the opportunity to present our project in this volume. Some of the ideas in section 4 have been inspired by a project internal paper by Ion Giurgea (2009) and by the discussion of the talk at the Hamburg workshop.

Syntactic variation and change in Romance: a Minimalist approach

363

sense as an inventory that includes both lexical and grammatical items.2 Modern Minimalist Generative Grammar has thus been approaching other generative frameworks, such as Head Driven Phrase Structure Grammar (HPSG). It is one of the aims of this paper to show that the use of such formalism can enrich the minimalist approach to syntax and to syntactic variation and change. This article is organized as follows: in section 2, we introduce the Minimalist Program and the idea to parameterize the differences between languages through the lexicon. In section 3, we elaborate on some important differences in word order in a few Romance languages (cf. 3.1), taking into account, in a second step (cf. 3.2), also their medieval stages. Section 4 will deal with some current points of discussion within minimalist theory, in particular, information structure. We will also initiate a discussion on the theory of the minimalist lexicon, and, finally, in the last section (5), return to the issue of how syntactic variation can be modelled within this framework. Before we start, we will somewhat further elaborate on the necessity of our proposal for an in-breadth comparison of all Romance languages.

1.2.

Comparative Romance syntax: state of the art

Describing and explaining the syntax of the Romance languages has been an important theory-driving factor within Generative Grammar. Many relevant syntactic phenomena were introduced into the discussion through the investigation of Romance and have had strong impulses for theoryinternal developments, such as the discovery of unaccusativity (cf. Perlmutter 1978; Rizzi 1982; Burzio 1986), the Split-Infl-hypothesis (cf. Pollock 1989), or the “cartographic” approach within the generative model (cf. Rizzi 1997; Cinque 1999). However, there is no comprehensive work that investigates all Romance languages, focussing on their most important syntactic phenomena within a uniform theoretical framework, and cataloguing them by using a consistent formalism. The beginning of Romance linguistics – and of linguistics in general – was marked by a strong interest in language comparison. The three vol-

2 The distinction between “content words” and “function words” is critically reviewed e. g. in Corver and Riemsdijk (2001), where a whole series of socalled “semi-lexical categories” (nominal, verbal, adjectival and particle-like elements) is discussed.

364

Guido Mensching and Eva-Maria Remberger

umes of Friedrich Diez’s Grammatik der romanischen Sprachen (1826–44) comprehend the phonetics, the morphology and the syntax of six Romance languages (French, Occitan, Italian, Romanian, Spanish, and Portuguese). Of course, a linguistically founded description of the Romance languages was not possible before the beginning of structuralism. But at that moment, the original comparative claim of Romance linguistics as a broad discipline encompassing all Romance languages was taking a backseat. Even within typology, there is no study which would be comparable – with respect to its large-scale, cross linguistic claim – with the huge comparative enterprises of the older days (also see Meyer-Lübke’s Grammatik der Romanischen Sprachen of 1890–1902). Certainly, some typologically classifying sketches exist: Muljaþiü (1967) attempts a classification of the Romance languages using 40 specific binary features, which focus, however, on sound change phenomena and include morpho-syntax to a lesser degree. Renzi (1989) analyzes some sample cases in order to argue that the Romance languages can actually be typologically characterized and classified only within the dimensions of diachronic, diatopic, and diastratic variation. A more recent sketch aiming at a pan-Romance overview can be found in Sörés (1995), which is based on the Greenbergian word-order typology. Of all these studies only the last one is designed as a purely syntactic and synchronic enquiry, but, admittedly, in the shape of a reduced size survey. This situation is similar for formal approaches, such as the theoretical framework of Generative Grammar. As in other lines of research, here, too, we mainly find detailed studies that focus either on one language or one type of phenomenon in more than one (but rarely all) Romance languages. Examples of the first type are Kayne’s (1975) famous book on French syntax and some comprehensive works on Italian (Rizzi 1982; Burzio 1986), Romanian (Dobrovie-Sorin 1994) and Sardinian (Jones 1993). We could also mention some introductions, such as Graffi (1994) or Donati (2008) on Italian, Zagona (2002) on Spanish, Jones (1996), Tellier (1997) and Laenzlinger (2003) on French, as well as large-scale grammars on individual languages with a generative background, though without an explicit formalization (e. g. the Grande Grammatica Italiana di Consultazione initiated by Renzi 1988, also see Bosque and Demonte 1999 for Spanish, Solà et al. 2002 for Catalan). The second type, i. e. cross-linguistic studies on one phenomenon, is much less frequent. Examples are Zanuttini (1997) on negation in Romance or Kaiser (2002) on word order change in Romance and Mensching (2000) on certain infinitive constructions. This kind of study is often not based on Generative Grammar in a strict sense (cf., e. g., Loporcaro

Syntactic variation and change in Romance: a Minimalist approach

365

1998 on participle agreement within Relational Grammar or Monachesi 2005, The Verbal Complex in Romance, who uses the framework of HPSG). It might appear at first sight that the work of Harris and Vincent (1990) is an exception, since the authors dedicated language-specific articles to short sketches of all subdomains of the Romance language systems; but with respect to syntax its conception as a one-volume work is not able to offer more than a founded outline without explicit and detailed formalizations. Two publications which are mostly comparable to the aims of our project would be the introductions by Müller and Riemer (1998) and Gabriel and Müller (2008); however, they are too basic and mostly directed to students. Furthermore, we can mention a huge series of studies that have been made within the context of the Syntactic Atlas of Italy (Atlante Sintattico d’Italia – ASIt; cf. http://asis-cnr.unipd.it/), such as Benincà (2001) and Poletto (2000). These publications discuss a great number of Romance varieties, but are usually limited to Italo-Romance. In addition, the more recent publications mostly discuss the position of functional categories following cartographic approaches, such as Rizzi (1997) and Cinque (1999), and they do not aim at a feature-based implementation. This also applies to the “Cartography Project” (cf. Cinque 2002). Although most of the articles appearing in the four volumes that have been published so far discuss several Romance languages, here, too, an important focus lies on the Italo-Romance varieties.

2.

Foundations of the Minimalist Program in the light of syntactic variation

In the Minimalist Program (Chomsky 2000, 2005, among others) syntax in a narrow sense is a set of universal properties, which are uniform across the languages of the world, whereas parameters are seen as restricted to the lexicon (Chomsky 2001b: 4). The main syntactic mechanism is called the ‘Merge’ operation. Merge can either apply to two items taken from the lexicon (cf. 1a) or to the syntactic objects already formed by Merge (cf. 1b). Both are more properly referred to as External Merge (cf. Chomsky 1995a, 2001b).

366

Guido Mensching and Eva-Maria Remberger

(1)

External Merge



D

E C

A A lex. item

B lex. item

A

C lex. item A lex. item

LEXICON

B lex. item

LEXICON

Another way for Merge to operate is to take no new material at all from the lexicon but to use the material that already exists in the tree via a copy mechanism (Internal Merge). This yields the displacement property of human syntax, previously called Move-Į (Chomsky and Lasnik 1977; Chomsky 1981). (2)

Internal Merge A A

B A

B

COPY

The item that is immediately merged to a head is called a complement, whereas all other items belonging to the same phrase are called specifiers (Chomsky 1993 et seq.; Donati 2006), which means that one head can have multiple specifiers.3 The lexical items are not introduced directly from the lexicon but through a set of items called numeration, which contains the lexical mate-

3 The status of adjuncts is controversial within this framework (see Kayne 1994; Chomsky 2000).

Syntactic variation and change in Romance: a Minimalist approach

367

rial which will be used by Merge. For example, the numeration in (3a) will yield the tree in (3b):4 (3)

a. N = {Jean [N], livres [N], achète [V], des [D]} b. V N Jean

V D

V ACHETD des

N livres

The status of a node as maximal projection or a head is implicitly known in the structure. At least for the Romance languages,5 we can assume that a head is always to the left of its complement. E. g., in the structure in (3b), [D un] and [V ACHET-] are heads, since they are on the left and cannot be substituted by a phrase. In order to make the trees more readable, we will in what follows adopt the older X-bar-notation. The structure in (3b) thus represents a verbal phrase (VP=V’’). A VP such as (3b) encodes lexical content that describes the action performed and its actors. The temporal setting is still missing here. Furthermore, in order to yield a grammatical sentence, the verb has to agree in person and number with the subject. In standard minimalist theory, these properties are accounted for by the insertion of a Tense head T°, which comes together with features other than tense: a feature called [EPP],6 which is an instruction for the computational system to project a specifier, and, second, unvalued person, number (and maybe gender7) features (also called ij-features). Since unvalued features are uninterpretable, the need to value these features triggers a search into the already 4 The phrase des livres is an object of type D(eterminer) according to the DPhypothesis (cf. Abney 1987). 5 For all languages according to Kayne (1994). 6 From the Extended Projection Principle (Chomsky 1981), according to which every sentence has a subject. The existence of [EPP]-features in general and their obligatoriness in the T-head are not uncontroversial (cf., among many others, Boeckx 2000; Lavine and Freidin 2002; Bailyn 2004). 7 In what follows, for simplicity, we will usually not represent gender features.

368

Guido Mensching and Eva-Maria Remberger

existing structure for an item with matching valued features. The unvalued ij-features are therefore called a “probe” in the framework introduced by Chomsky (2000); cf. (4a):8 (4)

The probing mechanism a. b.

$ %  !"       

 

%    !  

  $  !# !   

$

%  !"      !

 $

%   !# !    !    

As shown in (4b), the probe searches the tree, and, in the standard case, detects the subject and values its variables with its features – this process is called Agree. A constituent that has successfully taken part in a probing process is licensed for movement (we represent this by the symbol μ). In (4b), the subject will actually be copied to the specifier position that has been created by the EPP feature (internal merge). The verb, which still lacks ij-features and tense, moves to T° in most Romance languages, presumably for morphological reasons. The property of a head to attract another head will be formalized, in what follows, by a Head Attraction Feature (HAF), which is not yet represented in (4) for reasons of simplicity.9 In order to show how the parametrization of the Romance languages can be expressed, we first have to extend the numeration in (3a). As shown in (5), the lexical items in the lexicon come with a set of features; in addi-

8 We use variables for expressing unvalued features. 9 The HAF was introduced by Pomino (2008) as a neutral term to formalize the trigger for head-movement, the nature of which is still controversial in minimalist theory.

Syntactic variation and change in Romance: a Minimalist approach

369

tion, the functional category T° is also part of the numeration and hence stems from the lexicon.10 (5)

{‡ [T], | [T= present] [P=X] [N=Y] [EPP] [HAF]

Marie [N], | [P=3] [N=sg] [Case = X]

ACHET-[V], des [D], | |

[indef]

[N=pl]

livres [N]} | [P=3] [G=m] [N=pl] [Case = X]

Strictly speaking, not all features, values and variables come from the lexicon. The values of some features (those in italics), like the values of tense and number, are added when the entry enters the numeration, according to the content that the speaker wishes to transmit (optional or variable features, cf. Chomsky 1995: 231; Zwart 1997: 170 et seq.; see section 4.2 for further discussion). Another extension to the basic framework just sketched concerns the functional categories. Apart from T there is the so called ‘little’ v, which projects a “little” vP between the TP and the VP (see section 3.1, number 16a, for an illustration). One of the motivations of these assumptions (dating back to Larson 1988 and adopted by Chomsky 1995) is the fact that, for ditransitives, we need both the specifier and the complement position of VP for the direct (DO) and the indirect object (IO) respectively. The subject is merged in the specifier of vP, from where it moves to [Spec, TP]: the verb is attracted first to the head of v°, before it moves on to T.11 In order to provide a first example of lexical parametrization in the present framework, we will now have a brief look at the complementizer system (CP). Similar to what has been said about the features of T°, we can argue that, for wh-questions, many languages have an empty complementizer that contains an appropriate probe, an EPP feature and a HAF. See the French data in (6): (6)

a. Où vas-tu? Where go.2SG-you? ‘Where do you go?’

10 = semantic features; ș = theta (role). 11 The head v° contains a probe, the effects of which are visible on the verb only in languages with object agreement.

370

Guido Mensching and Eva-Maria Remberger b. Où est-ce que where is-this that ‘Where do you go?’

tu you

vas? go.2SG?

The basic idea to account for these examples is sketched in (7):12 (7)

Derivation for (6b) CP Spec



C° [Op=X] [EPP] [HAF]

TP

tu vas

où μ [Op = wh]

Roughly following Chomsky (2000), we can say that the wh-element, in this case the interrogative element où has an operator feature (Op) valued with [wh], whereas the empty functional element in the complementizer position (C°) has an unvalued operator feature, i. e. a probe, which finds the wh-element as a goal. The Agree process marks où as mobile (P), which allows it to move to the specifier created by the [EPP]-feature. The verb incorporated into T° (by virtue of its HAF) is then attracted to C°, which has a HAF of its own. In addition, the lexicon can also contain items that only serve to check either HAFs or [EPP]-features, usually known as expletive elements. Thus, we can argue that est-ce que in (6b), in the simplest analysis, is just an expletive element (Expl) that can check the HAF of the empty interrogative complementizer, as an alternative to checking of HAF by verb movement.13 12 All this is highly simplified for expository reasons, among others because of the status of the pronoun tu as a clitic. 13 The label Expl° is provisional here; in addition, such an analysis must determine a feature that makes sure that this element can only be inserted in C and only in interrogative contexts; if such a feature is semantic in nature, the term

Syntactic variation and change in Romance: a Minimalist approach

371

If we look at the multiple possibilities of forming wh-interrogatives in French, we see that, apart from the two options shown in (6), in colloquial French, the verb is not necessarily moved to C, as in (8a). Other colloquial alternatives are not to move either the verb or the wh-element; cf. (8b). (8)

a. Où tu where you ‘Where do you go?’

vas? go2SG?

b. Tu vas you go.2SG ‘Where do you go?’

où? where?

A way to formalize this is to make the [EPP]/HAF complex optional, and within this complex, to define the HAF as an option. This would allow both the derivation in (8a) (no [HAF]-, but an [EPP]-feature, since où is moved) and the derivation in (8b) (neither HAF nor [EPP]-feature). The solutions *Est-que tu vas où and *Vas-tu où would still be prevented from becoming grammatical, since there is no lexical entry with only HAF, but no [EPP]. A lexicon fragment of French that would be able to account for the options in (6) and (8), excluding, at the same time, all ungrammatical solutions, would be as stated in (9), where round brackets indicate optionality: (9)

(

‡ [C] [Op = X] [EPP] ([HAF])

est-ce-que [Expl°]

)

What we have just illustrated is language internal variation, in this case between different registers such as different stylistic levels of both written and spoken French. In what follows, we will also consider syntactic variation between different Romance languages.

expletive should rather be avoided. An alternative would be to consider est-ce que as an overt counterpart to the phonologically empty C° which lacks a HAF. For monomorphematic analyses of the type presented here, cf., e. g. Obenauer (1981), in contrast to decomposing analyses such as the one by Munaro and Pollock (2005). See Grimaldi (2008) for an overview and discussion.

372

Guido Mensching and Eva-Maria Remberger

3.

Parameterizing Romance

The hypothesis that we are pursuing here entails the idea that the most important syntactic differences between the Romance languages can be expressed by a lexical parametrization in the sense of (9). In this section, we will further elaborate on lexical parametrization, taking some basic word order patterns of French, Spanish and Italian as an example. As far as intra-language variation is concerned, varying word-order patterns that express the same semantic contents often go along with varying semantic and pragmatic interpretations (e. g. information structure). In this section, we abstract away from this important property, which will be a point of discussion in section 4.

3.1.

Basic Romance word order

To begin with, look at these examples:14 (10)

Italian a. Gianni mangia una mela. Gianni eat.3SG an apple ‘Gianni eats an apple.’

Spanish a.’ Juan come una manzana. Juan eat.3SG an apple ‘Juan eats an apple.’

b. Mangia una mela. eat.3SG an apple. ‘He/she/it eats an apple.’

b.’ Come una manzana. eat.3SG an apple ‘He/she/it eats an apple.’

c. Piove. rain.3SG ‘It rains.’

c.’ Llueve. rain.3SG ‘It rains.’

French a.’’ Jean mange Jean eat.3SG ‘Jean eats an apple.’

une an

b.’’ *Mange une pomme. eat.3SG an apple ‘He/she/it eats an apple.’

pomme. apple Il/elle mange une pomme. he/she eat.3SG an apple. ‘He/she/it eats an apple.’

14 In most varieties of Spanish and depending on context, the examples might sound better in the continuous form. We use the simple present here for the sake of simplicity.

Syntactic variation and change in Romance: a Minimalist approach c.’’ *Pleut. rain.3SG ‘It rains.’

373

Il pleut. it rain.3SG ‘It rains.’

In Italian, Spanish and French, we have SVO order, but in Italian and Spanish the subject can remain unexpressed. This is also the case with meteorological verbs like in (10c, c’). Here, French uses the expletive pronoun il, cf. (10c’’), similar to English and German. The formalization is rather trivial at first sight – Spanish, Italian and French have the same entry for finite T° (see (5) in section 2), with an [EPP]-feature that projects a specifier (the “classical subject position”), but the lexicon of Italian and Spanish has, in addition, an empty nominal element, traditionally known as pro.15 It has variable ij-features, which, in sentences such as (10b, b’), are contextually determined before entering the numeration. In sentences like (10b, b’), pro is inserted into the subject base-position [Spec, vP], where it is found by the probe in T. It is then raised to the specifier position of T in order to fulfil the [EPP]-feature. For the meteorological verb ‘to rain’, we can argue that in null-subject languages, pro is directly inserted into T as an expletive element to check the [EPP]-feature.16 The simplest assumption is that expletive pro17 has no ij-features at all in this case, which have thus to be marked as optional in the lexicon. In French, the overt expletive pronoun il (cf. 10b’’) seems to fulfil the same function, but should have fixed ij-features: (11)

Lexicon of

Spanish/Italian ‡ [T] [P=X] [N=Y] [EPP] [HAF]

French ‡ [T] [P=X] [N=Y] [EPP] [HAF]

‡pro [D] [P=X] [N=Y]

il [D] [P=3] [N=sg]

15 Roberge (1990) offers a general solution, valid for all languages, for the “recoverability of null arguments” (and thus also pro), which relies on the concept of government within the older Government and Binding framework. 16 There are several kinds of expletives, such as, for English, “extraposition it”, “weather it”, and impersonal there (cf. Svenonius 2002), which differ in their syntactic properties; see Bayer and Suchsland (1997) for a complete typology. For “weather it”, see note 18. 17 For some recent ideas on expletive pro, see Cardinaletti (2004), Rizzi and Shlonsky (2007), Alboiu (2009).

374

Guido Mensching and Eva-Maria Remberger

Some of the structures assumed for (10) can roughly be summarized by the following labelled bracketing notation (the cancelled items are deleted copies):18 (12)

a. [TP pro mangia [vP pro mangia [VP mangia una mela]]] b. [TP pro piove [VP piove]] c. [TP il pleut [VP pleut]]

The empty element pro is often advocated to explain inversion structures, such as in the Spanish sentence with VSO-order, cf. (13) as opposed to (10a’): (13)

Come Juan eat.3SG Juan ‘Juan eats an apple.’

una an

manzana. apple

Recall that the probing process marks an element as potentially movable, but it may stay in its place if the [EPP]-feature can be fulfilled otherwise. In the present case, we assume (following Chomsky 1995: 274) a derivation such as (14), where pro checks the [EPP]-feature in T°, allowing the subject to stay in its base position. (14)

a. [TP pro come [vP Juan come [VP come una manzana]]]

However, whereas in Spanish there seems to be hardly any syntactic restriction, Italian verb-subject order is mainly restricted to unaccusative verbs, as are the analogous cases in French, with the expletive il. In addition, in Spanish and Italian, the verb agrees with the subject, whereas, in French, the verb apparently agrees with the expletive il. (15)

Spanish Italian a. Llegaron varias personas. a.’ Arrivarono alcune persone. arrived.3PL several persons arrived.3PL several persons ‘There arrived several persons.’ ‘There arrived several persons.’

18 For (12b,c), we provisionally assume here that this kind of defective verb comes from the lexicon with 3rd person singular features. Note that Chomsky (1981: 323–325) has argued that “weather it” is not an expletive, but a quasiargument (cf. Svenonius 2002).

Syntactic variation and change in Romance: a Minimalist approach b. Comió Juan una manzana. ate.3SG Juan an apple ‘Juan ate an apple.’

375

b.’ *?Mangiò Gianni una mela.19 ate.3SG Gianni an apple ‘Gianni ate an apple.’

French a.’’ Il arriva plusieurs personnes. it arrived.3SG several persons ‘There arrived several persons.’ b.’’ *Il mangea Jean it ate.3SG Jean ‘Jean ate an apple.’

une an

pomme. apple

To explain these facts, we start from the idea that an unaccusative verb also selects a little v.20 But crucially, since an unaccusative verb does not subcategorize for an external theta-role, it does not lexically project a specifier, according to standard assumptions; cf. the illustration in (16): (16)

a. vP of transitive verbs

b. vP of an unaccusative verb like to go vP 

VP

vP DP (Subj.)



DP3 (Subj.)

V´ V



VP

PP (IO)



DP2 (DO) V

DP1 (IO)

19 Such cases are grammatical in Italian only when the subject is focalized. In previous interpretations, the object was interpreted to be “marginalized” (and distressed), but these cases are seen as instances of a subject in situ construction by Cardinaletti (2001). See section 4.1 for further discussion. 20 According to Chomsky (1995), unaccusative verbs do not have a vP-stratum. But see Radford (2004: 273 et seq.), Remberger (2006), Kallulli (2007), among many others.

376

Guido Mensching and Eva-Maria Remberger

However, in Remberger (2006), it is assumed that unaccusative v does project a specifier, contrary to the illustration in (16b). Let us assume here that this little v has an [EPP]-feature, at least as an option, in Italian and French. We can then argue that the expletive element must be inserted from the numeration at this early stage.21 In addition, for French, we have already assumed in (11) that the ij-features of il are fixed to 3rd person singular (cf. the agreement facts in (15a”)). When T is merged, the probe will find the features of the expletive and value its own features accordingly: (17)

Derivation for constructions like (15a”) a. b. % &  !  

 #



 #%         #&     

 !"$     

%

&  !              

# #% & 

 "$     

 

The pronoun is then raised to the specifier of T in order to check T‘s [EPP]-feature.22

21 Similar ideas are developed independently in López (2007) for English there vs. Germanic es/thadh and French il. For the general assumption that expletives can be merged in a low position, see, among others, Rizzi and Shlonsky (2007), Moro (1997), Kallulli (2008), Epstein and Seely (2006), den Dikken (1995), Groat (1999), Richards (2004). 22 The case is rather more complicated, among others because of the definiteness effect, which can be found in French and many other languages. Intuitively, it might seem that a definite element (il) is inserted into the indefinite structure in order to make the derivation converge, maybe as a kind of existential oper-

Syntactic variation and change in Romance: a Minimalist approach

377

Turning to Italian now, where the verb agrees with the associate, we can start from the same idea of an [EPP]-feature on the unaccusative v. Since Italian has a pro in its lexicon, it can be merged in [Spec, vP] in order to satisfy the [EPP]-features. Since, in this case, pro enters the derivation without values,23 which means, as an expletive, it acts itself as a probe24 and finds the complement of the verb. The values of pro are set accordingly. In the next step, the higher probe in T will detect these values, so that we end up with a verb that agrees with the complement, cf. (18): (18)

Derivation for constructions like (15a’) a. b. 

( ) !"#$  

&

 '!   

&) !"      !" 

! &(  $%!"#    !

(

) !"#$  !!   '!  #   !

& &(   %!"#    !

 

 

 



Thus, for Italian, we can assume exactly the same entries for unaccusative v as in French, the syntactic differences being yielded uniquely by the difference of the features contained in il and pro, respectively:

ator. For discussion, see, among many others, Belletti (1988), Diesing (1992), Mikkelsen (2002), Moon (2005), Remberger (2009). 23 See Groat (1995) for similar ideas concerning the difference between English there and it. In our approach (recent minimalism), the lack of ij-features is not a complete lack, but only a lack of values. 24 For the idea that an expletive element may work as a probe cf., e. g., Chomsky (2001a: 14) and Epstein and Seely (2006: 176), Alboiu (2009).

378 (19)

Guido Mensching and Eva-Maria Remberger Lexicon of

Italian ‡pro [D] [P=X] [N=Y]

French il [D] [P=3] [N=sg]

In contrast to our first hypothesis in (11), pro in Italian always has ij-features with open values in its lexical entry. These must be valued before the derivation in case of referential pro. If it does not get the values before the derivation starts (expletive use), it acts itself as a probe, thus getting the values in the course of the derivation. French il has fixed values in the lexicon; it thus behaves like a “normal” DP, in the sense that il is the goal of the probe in T°. In what follows, we will argue that, in Spanish, pro is always inserted in [Spec, TP] and never in [Spec, vP]. We will stipulate something along the lines of the following provisional lexicon fragments:25 (20)

Lexicon of

Spanish

Italian [v] unacc. ([EPP]) [HAF]

[v] unacc. [HAF]

pro[D] [P=X] [N =Y]

French

il[D] [P=3] [N=sg]

To ultimately account for the word order facts in (15), we have to consider Case assignment. According to the basic minimalist assumptions, a finite T°-node in non-ergative languages assigns nominative Case to the goal of its probe after a successful Agree-operation.26 Let us assume that, if an

25 French and Italian also share the property of selecting the auxiliary BE instead of HAVE with unaccusative verbs, in contrast to Spanish (where the auxiliary is always HAVE). This might be connected to the differing lexical entries for unaccusative v in the languages at issue. 26 Following, e. g., Belletti (1988), Groat (1995, 1999), Lasnik (1999), Epstein and Seely (2006), expletives need Case, against Chomsky (1995 et seq.). For Cardinaletti (1997), there are languages with expletives that need Case and others which have expletives with no Case requirement. See also Bayer and Suchsland (1997). For expletives, Iwakura (2002) (as quoted by Kirby 2005: 15) shows that English expletive there appears in typical Case positions.

Syntactic variation and change in Romance: a Minimalist approach

379

expletive needs Case, it will get nominative Case from T in our derivation.27 In fact, our approach is based on the assumption that pro always needs Case in Italian and French, even when it is used as an expletive. Thus, when the verb is transitive, an expletive pro cannot be introduced into the structure: it cannot be inserted into a second specifier (see section 2) of vP because transitive v has no [EPP]-feature. And it cannot be inserted into [Spec, TP], because no higher functional category above T could assign Case and the sentence will be ungrammatical. Thus, with the subject remaining in situ, there would be no way to check the [EPP]-feature of T. On the contrary, the Spanish facts (allowing VSO) will follow if the Case feature of pro is optional. If pro is used as an argument, the Case feature must be realized.28 If it is non-argumental (expletive), the numeration need not contain a Case feature, and, in fact, only the derivation without a Case feature on pro will succeed. Thus, an argumental DP can remain in [Spec, vP] of a transitive verb, and pro can occupy [Spec, TP] in order to satisfy the [EPP]-feature, leading to VSO order (cf. the structure in (14)). The lexical entries for pro and il in (19) and (20) will thus have to be modified as follows: (21)

Lexicon of

Spanish ‡pro [D] [Pers=X] [Num=Y] ([Case=Z])

Italian ‡pro [D] [Pers=X] [Num=Y] [Case=Z]

French il [D] [Pers=3] [Num=sg] [Case=Z]

27 Note that the Case of the associate would then be a problem. We might adopt Belletti’s (1988) theory on a special partitive Case assigned by unaccusative verbs (or unaccusative v in our framework). For Italian, we could also think of the multiple-agree mechanism (cf. Hiraiwa 2001; Chomsky 2001) or Alboiu’s (2009) idea that expletive pro is itself a Case assigner. 28 See Jones (1988) for restricting the Case-filter to argument NPs. Whereas this predicts that argument NPs must be assigned Case, it does not follow that nonargument NPs are barred from being assigned Case.

380

Guido Mensching and Eva-Maria Remberger

3.2.

The diachronic dimension

In what follows we would like to give some further examples of how the same formalism can be used to explain diachronic variation. We will discuss some word order patterns in the medieval stages of the Romance languages following an idea presented in Mensching (2003). A more recent alternative approach (Poletto 2008) will be introduced and briefly discussed in section 4. If we look at the medieval stages of the Romance languages, several details catch the eye; cf. the examples in (22) to (24): (22)

a. E

esto

fiz

yo porque

tomases

ejiemplo. (Conde Lucanor, enx. 2) and this did.1SG I because took.SUBJ.2SG example ‘And I did this for you to have an example.’

b. Questo tenne lo re a grande maraviglia. this held.SG the king to great miracle ‘The king considered this a great miracle.’ (Novellino 7, cf. Mensching 2000: 47) (23)

a. porque ella non avia las cartas resçebidas because she not had.3SG the letters.F.PL received.F.PL ‘because she had not received the letters’ (L. de Buen Amor, I 191a, cf. Batllorri, Sánchez and Suñer 1995: 204) b. avrebbono a Alessandro e forse alla donna fatta have.COND.3PL to A. and maybe to-the woman done villania affront ‘they would have affronted A. and perhaps the lady, too’ (Boccaccio, Dec. 2,3, cf. Mensching 2003: 184)

(24)

a. las ventas e compras de tu engañosa feria no the sales and purchases of your fraudulent fair not prósperamente sucedieron successfully happened.3PL ‘the sales and purchases of your fraudulent fair were not performed with success’ (Celestina 21, cf. Mensching 2003: 184) b. e loro ordinatamente disse come era avvenuto and to-them orderly told.3SG how was.3SG happened ‘and he told them in detail how it had happened’ (Boccaccio, Dec. 2,5, cf. Mensching 2003: 184)

Syntactic variation and change in Romance: a Minimalist approach

381

The examples in (22) show a direct object in initial position, while the subject remains postverbally. As is generally known in the literature, the fronted constituents could be almost of any type, e. g. direct objects, indirect objects or adverbs. Usually, these summarize or take up information already given in the context and are thus to be seen as topics, as in the contexts from which the two examples in (22) have been taken (see section 4.1 for further discussion concerning information structure). Importantly, topicalized constituents demand for a clitic left dislocation (CLDL) structure in the modern Romance languages, and it is usually thought that they leave the TP and are ultimately located in the CP domain. We think that this difference indicates an important parameter that separates the medieval from the modern stages of the Romance languages. The second property, illustrated in (23), refers to short-distance scrambling (cf. among others, Haider and Rosengreen 1998), i. e. movement of a constituent to a lower part of the clause, i. e. not as high as TP or CP. Since, according to basic assumptions of Generative Grammar, indirect objects are merged on the right hand side of V (cf. the structure in (16a)), the indirect object in (23b) must be supposed to have moved to the left of the verb. Finally, the examples in (24) show that the finite verb itself occupies a position lower than in the modern stages of the language, if we follow the standard assumption that the type of adverb in the example occupies a place at the left edge of vP. In Mensching (2003) it was argued that the first and the second property, i. e. XP-V-remainder-order and short scrambling can be explained by one and the same parameter. Let us start with XP-V-remainder-order. For reasons which cannot be discussed here (but see Mensching 2000, Kaiser 2002, among many others), one of the solutions for the Old Romance structure is the idea (similar to that of Diesing 1990 for Yiddish), that the Old Romance languages could move any XP to [Spec, TP], a place held to be reserved for subjects in English and modern varieties of Romance. If we adopt this view, the sentence in (22a) would have the following structure:

382

Guido Mensching and Eva-Maria Remberger

(25)

Movement to [Spec, TP] in Old Romance    j





  i  

 v ti

 ti  tj

The idea is that the direct object moves to [Spec, TP] and can check the [EPP]-feature in T°, which would have the same description as in the modern stages of the languages at issue. Since T° has only one [EPP]-feature, the subject will remain in its base position. But why should that be if the feature composition of T° was already the same as today? It would not be desirable in the Minimalist Program to specify something like a ‘subject related’ [EPP]-feature for the modern stages and a ‘general’ [EPP]feature in the old stages of Romance. In addition, there is a severe theoretical problem: in the tree structure in (25), the subject is nearer to T° than the object, which would predict that, due to economy reasons, the subject and not the object would be chosen for [EPP]-driven movement. A solution to both problems would be as follows: little v generally had an optional [EPP]-feature in the old Romance languages. If this feature enters the numeration, a non-subject constituent is moved to a second (i. e. an outer) specifier of v. Since, in this case, both the subject and the object would occupy specifiers of v, they would be equidistant (cf. among others, Ura 1996; Chomsky 2000) and either of them could be attracted by the [EPP]-feature in T. Interestingly, this hypothesis is actually borne out by the data: as argued in Mensching (2003), the scrambling phenomenon in (23a) can be interpreted as in (26):

Syntactic variation and change in Romance: a Minimalist approach (26)

383

Movement to [Spec, vP] in Old Romance vP v´

DPi las cartasi DP ella

v´ v [EPP] resçebidasj V tj

VP ti

In the example, the subject can be attracted by T in a further step in order to check the [EPP]-features in T°, but, because of equidistance, it could virtually also be the object, which would lead to structures like those in (22). Finally, the adverb position shown in (24) has to be interpreted as a property independent of the two characteristics just explained. As we saw in section 2, a head attraction feature (HAF) is postulated for the modern Romance languages that we have been examining. The background of this feature is the idea (dating back at least to Pollock 1989) that some languages (such as Italian and Spanish) show verb movement to T whereas others (like English) do not. Since certain adverbs are argued to be located in an adjunct position at the left edge of VP (or vP in the modern notation), lack of verb movement in English will explain the word order adverb-verb, in contrast to the inverse order in Romance. As argued by Pollock, French is a special case, because verb movement to T° is only present in finite clauses, which leads, among others, to different adverb positions in finite and non-finite clauses. Let us now assume that the adverbs in the Old Romance examples in (24) are vP-adverbs. We could then say that, in the Old Romance examples, verb movement to T° has not occurred. We will assume that, actually, v-to-T movement was optional. In our approach, we will formalize this by making the HAF optional:

384

Guido Mensching and Eva-Maria Remberger

(27) Old Romance

Mod. French

Mod. Italian [T] + fin. [P=X] [N=Y] [EPP] ([HAF])

[T] + fin. [P=X] [N=Y] [EPP] ([HAF])

v [P=X] [N=Y] ([EPP]) ([HAF])

Mod. Spanish

vunacc. [P=X] [N=Y] ([EPP]) [HAF]

v [P=X] [N=Y] [HAF]

vtrans. [P=X] [N=Y] [HAF]

The table shows the advantages of the lexicalist-minimalist approach. Synchronically, within the modern stages of the languages at issue, the diverging syntax is the result of the different feature composition of the two functional categories. In addition, the modern stages can be derived from the common feature composition of Old Romance, as follows: ± The optional HAF of T became obligatory in Italian and Spanish. In French, the feature became obligatory for finite T and non-available for non-finite T.29 ± The optional [EPP]-feature in v persisted in Italian and French unaccusative v and has disappeared from the lexicon of Spanish. This clearly shows that our approach is fundamentally based on optionality, which is therefore one of the issues to be discussed in the next section.

29 The latter is not included in (27). See Mensching and Remberger (2006) for different types of Romance non-finite T.

Syntactic variation and change in Romance: a Minimalist approach

4.

385

Some problems and extensions: the lexicon and information structure

The validation and revision of the ideas sketched in section 3 in a very simplified way will largely depend on the other Romance languages to be examined and on the integration of further theoretical developments (e. g. phase theory, cf. Chomsky 2001a, 2005). But there are several problems both in the overall approach and in the tentative implementation of section 3. They will have to be tackled within the project and are briefly discussed here.

4.1.

Information structure

One major problem is the relationship between syntactic principles, lexical parametrization and information structure. For example, Spanish VSO can yield (among others) a thetic expression, whereas in Italian, this structure very rarely occurs and usually implies that the subject is focussed. If we want to maintain the theory sketched in 3.1, which actually prohibits that the subject remains in situ in Italian transitive structures, we would have to suppose that it has moved, maybe to a Focus-head between TP and vP, as the one assumed by Belletti (2004).30 This raises the issue if it is desirable to adapt a cartographic framework (in the sense of Rizzi 1997, Poletto 2008, among many others), which includes several special functional categories concerning information structure. An alternative view, which will be sketched in what follows, is to suppose that information structural features can be part of functional heads in general. The most obvious problem in section 3 is the fact that the optional [EPP]-feature of little v, supposedly generalized in Medieval Romance, has persisted until today in French and Italian unaccusative v. But we had argued in 3.2 that, in Old Romance, this feature was ultimately responsible for the word orders shown in (22) and (23), with non-subject topic constituents being allowed both in [Spec, TP] and in an outer specifier of vP. According to our assumption in 3.1, the same feature explains the insertion of an expletive element in a low position ([Spec, vP]). But we would then expect, of course that Italian and French unaccusative con30 Against Cardinaletti (2001), see note 19. If this is really a subject in situ construction, our theory on empty expletives is, evidently, at stake. We are currently comparing Italian, Spanish, Romanian and Portuguese data for that issue.

386

Guido Mensching and Eva-Maria Remberger

structions still show structures like those in (22, 23). Our hypothesis is that this problem can be resolved by including features for information structure. According to Poletto (2008), who treats Old Romance examples such as those in (22) and (23) in a split-CP-approach, the preposed constituents both in the low position (cf. (23)) and in the high position (cf. (22)) can either be topics or foci, at least in Old Italian.31 Let us try to apply the probe-goal approach that we have been using in our framework, tentatively extending the operator approach sketched at the end of section 2 to topic and focus movement.32 Let us assume that both v° and T° could have an unvalued operator feature in Old Romance, modifying the entries for Old Romance in (27) as follows: (28)

a. v [P=X] [N=Y] [Op=X] | ([EPP]) ([HAF])

b. T [P=X] [N=Y] [Op=X] | [EPP] ([HAF])

There are actually two probes in each of the two entries. Crucially, we must assume that the [EPP]-features are linked to the operator probes and not to the ij-probes, which will thus only serve for Case checking and agreement but not for movement.33 If a DP is marked as [Op=topic] or [Op=focus] in its base position, it can thus be attracted to the outer specifier of the vP first, from where it is visible for the probe in T. The ij-probe in T° will agree with the subject in the inner specifier of vP, which is, how31 According to Poletto (2008), there is a series of low information structure related heads between vP and TP, which is exactly mirrored in a high position (the split-CP). She assumes a parameter along the following lines: if a head of category X attracts movement in the low informational structural field, a head of the same category X has the same property in the corresponding high field. 32 It remains to be seen if the assumption of operator features for both focus and topic is really appropriate. According to Rizzi (1997), topic movement is not operator movement, but there is also the opposite view, see Fanselow (2003) for the discussion of several accounts that include topic movement within operator movement. The generalized operator approach also leaves open the question of why wh-movement is still possible in Modern Romance. 33 See, e. g., Miyagawa (2005) for a similar proposal concerning “focus-prominent languages”. In this account, the [EPP]-feature on T is linked to a focus-feature in these languages.

Syntactic variation and change in Romance: a Minimalist approach

387

ever, not attracted to [Spec, vP], and, accordingly, not to [Spec, TP] either, for reasons of distance. Let us now reconsider the modern Italian and French case, based on the lexical entry in (20): (29)

vunacc [P=X] [N=Y] | [EPP] [HAF]

Here the [EPP]-feature is linked to the ij-probe. Even if some constituent other than the subject was marked as topic or focus, it could not be attracted to [Spec, vP]. Viewed in a superficial way, we could say that topic and focus marking in terms of operator features connected to [EPP]-features was done in T, v,34 and probably C in Old Romance, whereas it is limited to C nowadays. With respect to left peripheral topics, the modern structures are clearly different from the medieval ones, cf. (30): (30)

a. Old Spanish

E esto fiz yo.35 and this did.1SG I ‘And this is what I did.’ (topic reading)

b. Modern Spanish

Y esto loi hice yo. and this CL did.1SG I ‘And this is what I did.’

If we consider the clitic in (30b) as something like an overt intermediate copy of the moved element (cf., e. g., Grohmann 2003) that serves – in rather informal terms – to bridge a long distance (here from VP to CP), its lack in Old Romance examples such as those in (22) is explained because here movement only leads up to TP.36 34 In v, if Poletto’s (2008) view that the scrambled elements in examples such as (23) could either be topics or foci in Old Italian can be generalized for Old Romance. 35 Here, too, we follow Poletto (2008) in assuming that the left peripheral item could either be topic or focus, depending on context. 36 For left-peripheral focus, the issue is rather more complicated and more language specific. The subject precedes the verb in Italian (cf. Rizzi 1997: 299), but follows it in Spanish (Zagona 2002: 250). This and many other issues will have

388

Guido Mensching and Eva-Maria Remberger

Questions remain, of course, for the low focus positions within the vPVP-complex. It seems, at first sight, that all lexical material that has not moved out of vP-VP can potentially bear focus. This may be implemented by allowing information structure to be assigned postsyntactically on the Logical Form, cf. Vallduví (1992a, 1992b), and the following illustration in Miyagawa (2005): (31)

[TP ... topic

[vP

]

focus

Although this could be thought of as a kind of universal default, it cannot explain language specific behavior: whereas an Italian VSO structure only allows focus on the subject, the same structure in Spanish allows either the object, or the subject or the whole predicate to be focussed. An alternative to this approach can be to assume operator features that are not linked to an [EPP]-feature, so that, e. g., in-situ focus can be explained (cf., e. g., Alboiu 2004 for a similar approach).37

4.2.

Considerations on the minimalist lexicon

As we have tried to show, alternative lexical entries for grammatical properties can not only explain syntactic differences between languages but also language internal variation. This idea is a direct consequence of the minimalist framework, in which the computational system of syntax is thought of as an optimal system, which does not allow optionality (i. e. more than one syntactic solution starting from one numeration). If two structures (like SVO and VSO) are allowed in one language to express the same semantic content, there must be two different numerations, e. g. one with and one without an expletive pro. For other kinds of variation, we must suppose that the lexicon contains several alternative entries for one

to be considered, such as: 1. VOS-order. The approach sketched in 3.1 would be compatible with several theories that do not consider the subject as an in situ-subject in VOS structures (cf. Ortega-Santos 2006); e. g. the remnant movement approach (cf. Kayne and Pollock 2001). 2. The different types of focus and topic and other information structural categories (see Féry and Krifka 2008, among many others). 3. The possibility of Italian to have VSO order when the subject is focalized (see above). 37 Alboiu (2004) uses a concrete [Foc]-feature instead of an abstract operator feature. This kind of solution might be preferable, at least for topics; cf. note 32.

Syntactic variation and change in Romance: a Minimalist approach

389

and the same functional category. In particular, for “little” v, this idea has been quite extensively discussed in the literature, e. g. in terms of “flavors of v” (Embick 2004; Kallulli 2007). But whereas alternative or optional entries in the lexicon thus belong to the standard minimalist assumptions, the issue of optional features is not so clear. Chomsky (1995) originally distinguished between intrinsic features (such as gender of inanimate nouns) and optional features (such as number of nouns). Using a different terminology, “[t]he value of intrinsic features is fixed, the value of variable features is determined arbitrarily as the lexical item enters the numeration” (Zwart 1997: 170–171). These notions will have to be reformulated in the present framework (Chomsky 2000 et seq.), in which several types of variable features have to be distinguished: 1. variable features that act as probes, i. e. their values are fixed during the derivation, and, 2. variable features like those at issue here, the values of which are fixed according to context before the derivation. These two types refer to attribute-value pairs. Yet another case of optional features (3.) would be optional [EPP]features, stipulated by Chomsky (2001a) in order to explain object shiftconstructions. We have implicitly used these types, although a reinterpretation of the original “optional features” of Chomsky (1995) (our type 2, which we applied to expletive pro in section 3.1) has not been addressed in the literature and will need some further discussion in future work. In particular, the question remains of where these features come from. Is there an inventory of variable features as part of the lexicon or are they located in another part of the language system?38 The strict minimalist approach, which considers narrow syntax as uniform across languages and parameters restricted to the lexicon may also appear as problematic in principle, because it blurs the difference between lexical and grammatical information. Some of the classical arguments for separating grammar and lexicon are summarized by Contini-Morava and Yishai (2000) under the key-words of different types of notional content, optional versus obligatory elements, open class versus closed class, difference between form and content. Another view (cf. Bloomfield 1933) is to consider the lexicon as a list of irregularities or exceptions. This view is mainly the one that has been adopted by Generative Grammar; cf. also the dichotomy “computation” (grammar) vs. “storage” (lexicon) (ContiniMorava and Yishai 2000: xi). In minimalism, the latter dichotomy has been 38 The same question arises for the features of type 3 (optional [EPP]-features), at least in Chomsky’s original approach. In our approach, this problem does not occur, because optional [EPP]-features are explicitly encoded in the lexicon.

390

Guido Mensching and Eva-Maria Remberger

taken in a very strict sense. Consider the parameters of the Principles-andParameters framework (Chomsky 1981). The null-subject parameter (cf. Rizzi 1982), e. g., stated that the existence of the empty category pro was related to other properties such as free inversion, absence of that-trace effect, or the absence of overt expletive pronouns. Note that the null-subject (cf. Rizzi 1982; Gilligan 1987; Roberts 2007) parameter was supposed to be a grammatical principle, which contained, however, references both to form (e. g. inversion) and to content (pro, expletives). In minimalism, it seems that the content-elements are strictly separated from the grammar (in the sense of narrow syntax). These elements are held to be language specific and are thus argued to be part of the lexicon. Apart from the content-character, the “grammatical” items of the minimalist lexicon share other properties of “real” lexical items: they are optional in the sense that there are often several versions of the same functional category (different versions or “flavors” of T, v, etc.). The open versus closed dichotomy is not as clear as that. The set of functional items of a given type (e. g., T) is relatively closed in the sense that, whereas, say, the number of nouns in a given language grows almost daily, the number and the shape of Ts remain stable for some centuries. And, whereas it is still a matter of debate whether the semantic features of lexical items are universal or not, functional elements, at least according to the minimalist view, are composed of a relatively small and fixed universal set of features (cf. Chomsky 2001a). Thus, to subsume lexemes and functional categories in one and the same lexicon is maybe too much of a simplification, and we might decide to distinguish a lexicon proper from another inventory which contains functional elements (which we might call a “grammaticon”). These issues are rarely addressed in minimalist studies and must clearly be discussed within our project. To return to the example of the pro-drop parameter, it has been shown that most of the postulated corollaries are problematic (cf., e. g., Gilligan 1987). However, the corollary between the existence of pro and the non-existence of overt expletives has been argued to hold universally.39 In principle this might well follow from the lexical approach: since pro has unvalued features, it can, in principal, be inserted without concrete values. In this case it cannot be referential and is thus an expletive element. Once the lexicon of a given language has this expletive pro, the development of an overt expletive may be blocked because of

39 See, among other, Hinzelin and Kaiser (2005). Questions remain about the expletive egli in Fiorentino, but note that it is not clear whether modern Fiorentino really is a null-subject language.

Syntactic variation and change in Romance: a Minimalist approach

391

something like a lexical blocking effect: this lexical effect is similar like the one that blocks, say, the lexicalization of a word like stealer because of the existence of the word thief. But since lexical items can have synonyms, it is not excluded that stealer enters the lexicon, once both words would have acquired different nuances of meaning. But an expletive element has no meaning, and therefore it might be seen as extremely difficult (though maybe not impossible) for an overt synonym of an expletive pro to be adopted by a language.

5.

Final considerations on syntactic variation

In the preceding sections we have outlined the basic idea that syntax (in a narrow sense) works alike in all languages. Since syntax is feature driven and the features (stemming from the lexicon) vary from one language to another, different word order and agreement facts are produced. We have exemplified this for several modern and past varieties of Romance. But we have seen that our model is also applicable to inner-language variation.40 Several scenarios are to be distinguished: 1. Two different dialects (in the sense of diatopic varieties) spoken by two different speakers: there is not much to say about this; the lexicon of each speaker will be different, and the case is therefore the same as the comparison between different languages (of one group) that we have been examining. Our formalism has a clear advantage for dialect studies in general, because grammatical differences are captured in lexicon entries with a unified format. The Abstand between two varieties thus becomes measurable. 2. Two or more varieties spoken by one speaker. Here we have to distinguish – in theory – several subcases: a) The speaker has actually two or more lexicons. This is the standard case in bilingual individuals. Evidence for this comes, among others, from code-switching phenomena. The feature- and value-sets contained in the functional categories of either language are different, so that they cannot possibly be combined in one lexicon. E. g., as has been shown by González (2005) and González and Struckmeier (2008), German has a gender feature [gender = neuter], which

40 For the following, see Mensching (2005). A totally different (LF-based) approach is proposed by Adli (2006).

392

Guido Mensching and Eva-Maria Remberger

Spanish has not.41 The intra-clausal code-switching of a bilingual speaker thus consists in the fact that the numeration is fed from two different lexicons. This can lead to incompatibilities which exclude certain combinations in code-switching. For dialectal variation within one speaker (e. g. one who speaks a dialect and a standard variety), the same might be the case if the dialects are so divergent (see above, in particular with respect to the features and/or values allowed) that two lexicons have to be stipulated. If the features and values are compatible, it might be the case described in b) (one lexicon). b) We may suppose that different registers and sociolects used by one speaker are stored in one and the same lexicon and that the entries can have diasystematic labels. This may be the case of different registers of French (standard French, “spoken” French on different diastratic levels, different versions of regional French). This is not difficult to see for items with lexical content, e. g.: (32)

a. N, bourgmestre | Belgian […]

N, maire | Standard […]

b. N, bagnole | Argot […]

N, voiture | Standard […]

Now, consider the simplified entry for interrogative C° from (9) above: (33)

(

‡ [C] [Op = X] [EPP] [HAF]

)

If functional categories such as T, C, etc. are also stored in the lexicon, it would be predicted that they can have variety labels such as those of the “purely” lexical entries in (32):42 41 So called “neuter pronouns” such as ello, esto, aquello are maybe “semantically neuter” but cannot be argued to contain a grammatical gender feature with the value neuter. 42 The labels are based on the discussion in Kleineidam (1986: 139–145). They may not be sufficient, but it is not our purpose here to correctly describe the

Syntactic variation and change in Romance: a Minimalist approach (34)

393

a. [C] | spoken (current) [Op = X]

b. [C] | spoken (familiar) [Op = X] [EPP]

c. [C] | standard [Op = X] [EPP] [HAF]

All these considerations show (as do those in the previous section) that the minimalist ideas on the lexicon will have to be discussed in great detail if we really want to explain syntactic variation by means of the Minimalist Program. This includes the integration of the insights of modern variational linguistics, some of which are reflected in the other contributions of the present volume. For example, the distinction between diatopic varieties and registers is not always easy to be made, and there seem to exist gradual transition zones between the diasystems, so that it remains to be seen if the distinctions made above (e. g., between one or several lexicons) are really possible. It would surely be too ambitious to try to resolve all these problems within the project that we have been describing in this article, but they will have to be addressed within a theoretical framework that shifts parameters to the lexicon. By these last comments, we hope to have shown that a Minimalist formalization as proposed in this paper is not only able to describe the differences in syntactic behavior among the Romance languages, but also that the general ideas of the Minimalist Program are far from being incompatible with variational approaches to language.

pragmatics and sociolinguistics of the French constructions.

394

Guido Mensching and Eva-Maria Remberger

Abbreviations 1 = 1st person; 2 = 2nd person; 3 = 3rd person; C = complementizer; CL = clitic; CLDL = clitic left dislocation; COND = conditional; CP = complementizer phrase; D = determiner; DFG = Deutsche Forschungsgesellschaft; DO = direct object; DP = determiner phrase; EPP = extended projection principle; Expl = expletive; F = feminine; fin = finite; Foc = focus; G = gender; HAF = head attraction feature; HPSG = head driven phrase structure grammar; indef = indefinite; IO = indirect object; lex = lexical; LF = logical form; m = masculine; mod = modern; N = noun; N = number; Num = number; O = object; Op = operator; P = person; Pers = person; PL/ pl = plural; SEM = semantic component; SG/sg = singular; Spec = specifier; Subj = subject; SUBJ = subjunctive; T = tense; TP = tense phrase; trans = transitive; unacc = unaccusative; V = verb; v = “little” v; VP = verb phrase; vP = “little” v phrase; ș = theta role; μ = mobile / licensed for movement; ij = agreement features.

Syntactic variation and change in Romance: a Minimalist approach

395

References Abney, Steven 1987 The English noun phrase in its sentential aspect. Ph.D. dissertation. Cambridge, Mass.: MIT Press. Adli, Aria 2006 French wh-in-situ questions and syntactic optionality: Evidence from three data types. Zeitschrift für Sprachwissenschaft 25: 163– 203. Alboiu, Gabriela 2004 Optionality at the interface: Triggering focus in Romanian. In: Henk van Riemsdijk and Anne Breitbath (eds.), Triggers, 49–75. (Studies in Generative Grammar 75.) Berlin/New York: Mouton de Gruyter. Alboiu, Gabriela 2009 Null expletives and case: The view from Romance. In: Pascual José Masullo, Erin O’Rourke and Chia-Hui Huang (eds.), Romance Linguistics 2007: Selected Papers from the 37th Linguistic Symposium on Romance Languages (LSRL), Pittsburgh, 15–18 March 2007, 1–18. Bailyn, John F. 2004 Generalized inversion. Natural Language and Linguistic Theory 22: 1–49. Bayer, Josef and Peter Suchsland 1997 Expletiva und leere Subjekte im Deutschen. Groninger Arbeiten zur germanistischen Linguistik 41: 12–38. Belletti, Adriana 1988 The case of unaccusatives. Linguistic Inquiry 19: 1–34. Belletti, Adriana 2004 Aspects of the low IP area. In: Luigi Rizzi (ed.), The Structure of CP and IP: The Cartography of Syntactic Structures. Volume 2, 16–51. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press. Benincà, Paola 2001 The position of Topic and Focus in the left periphery. In: Guglielmo Cinque and Giampaolo Salvi (eds.), Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, 243–297. Amsterdam: Elsevier. Bloomfield, Leonard 1933 Language. New York: Holt, Rinehart and Winston. Boeckx, Cedric 2000 EPP Eliminated. Ms., University of Connecticut. Bosque, Ignacio and Violeta Demonte (eds.) 1999 Gramática descriptiva de la lengua española. 3 Volumes. Madrid: Espasa Calpe.

396

Guido Mensching and Eva-Maria Remberger

Burzio, Luigi 1986 Italian Syntax. A Government and Binding Approach. (Studies in natural language and linguistic theory, 1.) Dordrecht: Reidel. Cardinaletti, Anna 2001 A second thought on emarginazione: “Destressing” vs “Right Dislocation”. In: Guglielmo Cinque and Giampaolo Salvi (eds.), Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, 117–135. Amsterdam: North Holland. Cardinaletti, Anna 2004 Toward a cartography of subject positions. In: Luigi Rizzi (ed.), The Structure of CP and IP: The Cartography of Syntactic Structures. Volume 2, 115–166. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press. Chomsky, Noam 1981 Lectures on Government and Binding. The Pisa Lectures. (Studies in Generative Grammar 9). Dordrecht: Foris. Chomsky, Noam 1995a Bare phrase structure. In: Gert Webelhuth (ed.), Government and Binding Theory and the Minimalist Program. Principles and Parameters in Syntactic Theory, 383–439. Cambridge: Blackwell. Chomsky, Noam 1995b The Minimalist Program. (Current Studies in Linguistics 28.) Cambridge, Mass.: MIT Press. Chomsky, Noam 2000 Minimalist inquiries: The framework. In: Roger Martin, David Michaels and Juan Uriagereka (eds.), Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. Chomsky, Noam 2001a Derivation by phase. In: Michael, J. Kenstowicz (ed.), Ken Hale. A Life in Language, 1–52. Cambridge, Mass. [u. a.]: MIT Press. Chomsky, Noam 2001b Beyond Explanatory Adequacy. Ms. MIT. Chomsky, Noam 2005 On Phases. Ms. MIT. Chomsky, Noam and Howard Lasnik 1977 Filters and control. Linguistic Inquiry 8: 425–504. Cinque, Guglielmo 1999 Adverbs and Functional Heads: A Cross-linguistic Perspective. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press. Cinque, Guglielmo (ed.) 2002 Functional Structure in DP and IP. The Cartography of Syntactic Structures. Volume 1. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press.

Syntactic variation and change in Romance: a Minimalist approach

397

Contini-Morava, Ellen and Yishai Tobin (eds.) 2000 Between Grammar and Lexicon. (Current Issues in Linguistic Theory 183.) Amsterdam: John Benjamins Publishing Company. Corver, Norbert and Henk van Riemsdijk 2001 Semi-lexical categories. In: Corver, Norbert and Henk van Riemsdijk (eds.), Semi-lexical Categories. The Function of Content Words and the Content of Function Words, 1–19. Berlin/New York: Mouton de Gruyter. den Dikken, Marcel 1995 Binding, expletives and levels. Linguistic Inquiry 26: 347–354. Diesing, Molly 1990 Verb movement and the subject position in Yiddish. Natural Language and Linguistic Theory 8: 41–79. Diesing, Molly 1992 Indefinitess. (Linguistic Inquiry Monograph 20.) Cambridge, Mass.: MIT Press. Diez, Friedrich 1826–1844 Grammatik der romanischen Sprachen. 3 Volumes. Bonn: Weber. Dobrovie-Sorin, Carmen 1994 The Syntax of Romanian. Comparative Studies in Romance. (Studies in Generative Grammar 40.) Berlin/New York: Mouton de Gruyter. Donati, Caterina 2006 Labels and Merge. Talk delivered at the InterPhases conference, University of Cyprus, May 18–20, 2006, Ms. Donati, Caterina 2008 La sintassi. Regole e strutture, Bologna: Il Mulino. Embick, David 2004 On the structure of resultative participles in English. Linguistic Inquiry 35: 355–392. Epstein, Samuel D. and T. Daniel Seely 2006 Derivations in Minimalism. Cambridge: Cambridge University Press. Fanselow, Gisbert 2003 Free constituent order: A minimalist interface account. Folia Linguistica 37: 191–231. Féry, Caroline and Manfred Krifka 2008 Information structure. Notional distinctions, ways of expression. In: Piet van Sterkenburg (ed.), Unity and Diversity of Languages, 123– 136. Amsterdam: John Benjamins. Gabriel, Christoph and Natascha Müller 2008 Grundlagen der generativen Syntax: Französisch, Italienisch, Spanisch. (Romanistische Arbeitshefte, 51). Tübingen: Niemeyer.

398

Guido Mensching and Eva-Maria Remberger

Gilligan, Gary 1987 A Cross-Linguistic Approach to the Pro-Drop Parameter. PhD dissertation, University of Southern California, Los Angeles. Giurgea, Ion 2009 Theoretical proposals based on Romanian and Italian data. Ms, University of Konstanz. González, Kay 2005 Die Syntax des Code-Switching. Esplugisch: Sprachwechsel an der Deutschen Schule Barcelona. Doctoral Dissertation: Universität zu Köln. González, Kay and Volker Struckmeier 2008 Asymmetrien im Code-Switching: Eine DM-Lösung zur Partizipselektion. In: Eva-Maria Remberger and Guido Mensching (eds.), Romanistische Syntax – minimalistisch, 83–102 (Tübinger Beiträge zur Linguistik 506.) Tübingen: Narr. Graffi, Giorgio 1994 Sintassi. Bologna: Il Mulino. Grimaldi, Lucia 2008 Eine minimalistische Analyse französischer est-ce que-Interrogative. Unpublished Ms., Freie Universität Berlin. Groat, Erich M. 1995 English expletives: A minimalist approach. Linguistic Inquiry 26: 354–365. Groat, Erich M. 1999 Raising the case of expletives. In: Samuel Epstein and Norbert Hornstein (eds.), Working Minimalism, 27–43. Cambridge, Mass.: MIT Press. Grohmann, Kleanthes 2003 Prolific Domains. On the Anti-Locality of Movement Dependencies. Amsterdam: John Benjamins. Haider, Hubert and Inger Rosengren 1998 Scrambling. (Sprache und Pragmatik, 49.) Lund: University of Lund. Harris, Martin and Nigel Vincent (eds.) 1990 The Romance Languages. London: Routledge. Hinzelin, Marc-Olivier and Georg Kaiser 2005 Das neutrale Pronomen ‚ello‘ im dominikanischen Spanisch und die Nullsubjekteigenschaft. (Konstanzer Arbeitspapiere des Fachbereichs Sprachwissenschaft 116). Konstanz: Universität Konstanz. Hiraiwa, Ken 2001 Multiple agree and the defective intervention constraint in Japanese. In: Ora Matushansky et. al. (eds.), The Proceedings of the MITHarvard Joint Conference, 67–80. Cambridge, Mass.: MIT Working Papers in Linguistics.

Syntactic variation and change in Romance: a Minimalist approach

399

Iwakura, Kunthiro 2002 A minimalist approach to expletive constructions in English. English Linguistics 19: 186–21. Jones, Michael Allan 1988 Cognate objects and the case Filter. Journal of Linguistics 24: 89–110. Jones, Michael A. 1993 Sardinian Syntax. London/New York: Routledge. Jones, Michael A. 1996 Foundations of French Syntax. (Cambridge Textbooks in Linguistics.) Cambridge: Cambridge University Press. Kaiser, Georg 2002 Verbstellung und Verbstellungswandel in den romanischen Sprachen. (Linguistische Arbeiten; 465). Tübingen: Niemeyer. Kallulli, Dalina 2007 Rethinking the passive/anticausative distinction. Linguistic Inquiry 38: 770–780. Kallulli, Dalina 2008 There is secondary predication in there-existentials. In: Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics, 279–287. Somerville, Mass.: Cascadilla Proceedings Project. Kayne, Richard 1975 French Syntax: The Transformational Cycle. Cambridge, Mass.: MIT Press. Kayne, Richard 1994. The Antisymmetry of Syntax. (Linguistic Inquiry Monograph 25.) Cambridge, Mass.: MIT Press. Kayne, Richard and Jean-Yves Pollock 2001 New thoughts on stylistic inversion. In: Hulk, Aaafke and Jean-Yves Pollock (eds.), Subject Inversion in Romance and the Theory of Universal Grammar, 107–162. (Oxford Studies in Comparative Syntax.) New York/Oxford: Oxford University Press. Kirby, Susannah 2005 Semantics or subcases? The acquisition of referential vs. expletive it. Ph.D. dissertation, University of North Carolina at Chapel Hill. [http://www.unc.edu/~tofu/kirbythesis.pdf] Kleineidam, Hartmut 1986 Fremdsprachengrammatik. Analysen und Positionen: Beiträge mit dem Schwerpunkt Französisch. (Tübinger Beiträge zur Linguistik 277.) Tübingen: Narr. Laenzlinger, Christopher 2003 Initiation à la Syntaxe formelle du français. Le modèle Principes et Paramètres de la Grammaire Générative Transformationnelle. (Sciences pour la communication 70.) Bern: Lang.

400

Guido Mensching and Eva-Maria Remberger

Larson, Richard 1988 On the double object construction. Linguistic Inquiry 19: 335–391. Lavine, James and Robert Freidin 2002 The subject of defective t(ense) in Slavic. Journal of Slavic Linguistics 10: 253–89. López, Luis 2007 Locality and the Architecture of Syntactic Dependencies. London: Palgrave-MacMillan. Loporcaro, Michele 1998 Sintassi comparata dell‘accordo participiale romanzo. (Linguistica 17.) Torino: Rosenberg Sellier. Meyer-Lübke, Wilhelm 1890–1902 Grammatik der romanischen Sprachen. 3 Volumes. Leipzig: Reisland. Mensching, Guido 2000 Infinitive Constructions with Specified Subjects. Oxford: Oxford University Press. Mensching, Guido 2003 Minimalistische Syntax der romanischen Sprachen. In: Elisabeth Stark and Ulrich Wandruszka (eds.), Syntaxtheorien: Modelle, Methoden, Motive, 117–191. Tübingen: Narr. Mensching, Guido 2005 Variación sintáctica, lingüística de corpus y gramática generativa: Teorías, métodos y problemas. In: Gabriele Knauer and Valeriano Bellosta von Colbe (eds.), Variación sintáctica en español: un reto para las teorías de la sintaxis, 13–33. (Linguistische Arbeiten 494.) Tübingen: Niemeyer. Mensching, Guido and Eva-Maria Remberger 2006 Probes: Lack of agreement in Romance. In: Joâo Costa and Maria Cristina Figueiredo (eds.), Studies on Agreement. (Linguistics Today 86.) Amsterdam: John Benjamins, 173–201. Mikkelsen, Line 2002 Reanalyzing the definiteness effect: Evidence from Danish. Working Papers in Scandinavian Syntax 69: 1–75. Miyagawa, Shigeru 2005 On the EPP. In: Martha McGinnis and Norvin Richards (eds.), Perspective on Phases, 201–236. MIT: MIT Working Papers in Linguistics. Moon, Gui-Sun 2005 The definiteness effect in expletive there-constructions from the perspective of semantic interpretation. Studies in Generative Grammar 15: 265–274. Monachesi, Paola 2005 The Verbal Complex in Romance. A Case Study in Grammatical Interfaces. (Oxford Studies in Comparative Syntax) Oxford/New York: Oxford University Press.

Syntactic variation and change in Romance: a Minimalist approach

401

Moro, Andrea 1997 The Raising of Predicates: Predicative Noun Phrases and the Theory of Clause Structure. (Cambridge Studies in Linguistics 80.) Cambridge: Cambridge University Press. Muljaþiü, Zarko 1967 Die Klassifikation der romanischen Sprachen. Romanistisches Jahrbuch 18: 243–253. Müller, Natascha and Beate Riemer 1998 Generative Syntax der romanischen Sprachen: Französisch, Italienisch, Portugiesisch, Spanisch. (Stauffenburg Einführungen.) Tübingen: Stauffenburg. Munaro, Nicola and Jean-Yves Pollock 2005 Qu’est-ce-que (qu)-est-ce-que? A case study in comparative Romance interrogative syntax. In: Guglielmo Cinque and Richard Kayne (eds.), The Oxford Handbook of Comparative Syntax, 542– 606, Oxford: Oxford University Press. Obenauer, Hans-Georg 1981 Le principe des catégories vides et la syntaxe des interrogatives complexes. In: Langue Française 52: 100–118. Ortega-Santos, Iván 2006 On postverbal subjects, PF and the copy theory: The Spanish case. In: Proceedings of the Joint Meeting of The Hispanic Linguistics Symposium and The Conference on the Acquisition of Spanish and Portuguese as First and Second Languages, 56–66. Somerville, Mass: Cascadilla University Press. Perlmutter, David M. 1978 Impersonal passives and the unaccusative hypothesis. In: Proceedings of the Fourth Annual Meeting of the Berkeley Linguistics Society, 157–189. University of California: Berkeley Linguistics Society. Poletto, Cecilia 2000 The Higher Functional Field: Evidence from Northern Italian Dialects. (Oxford Studies in Comparative Syntax). Oxford/New York: Oxford University Press. Poletto, Cecilia 2008 Die linke Peripherie der “unteren Phase”: OV-Stellung im Altitalienischen. In: Eva-Maria Remberger und Guido Mensching (eds.), Romanistische Syntax – minimalistisch, 131–156 (Tübinger Beiträge zur Linguistik 506.) Tübingen: Narr. Pollock, Jean-Yves 1989 Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry 20: 365–424. Pomino, Natascha 2008 Spanische Verbalflexion: Eine minimalistische Analyse im Rahmen der Distributed Morphology. (Linguistische Arbeiten 523.) Tübingen: Niemeyer.

402

Guido Mensching and Eva-Maria Remberger

Radford, Andrew 2004 English Syntax: An Introduction. (Cambridge Textbooks in Linguistics.) Cambridge: CUP. Remberger, Eva-Maria 2006 Hilfsverben. Eine minimalistische Analyse am Beispiel des Italienischen und Sardischen. (Linguistische Arbeiten 504.) Tübingen: Niemeyer. Remberger, Eva-Maria 2009 Null subjects, expletives and locatives in Sardinian. In: Georg Kaiser, Eva-Maria Remberger (eds.), Null Subjects, Expletives and Locatives in Romance, 231–261. (Konstanzer Arbeitspapiere des Fachbereichs Sprachwissenschaft 123.) Konstanz: [http://kops.ub. uni-konstanz.de/volltexte/2009/7860/] Renzi, Lorenzo 1989 Wie können die romanischen Sprachen typologisch charakterisiert werden? Vox Romanica 48: 1–12. Renzi, Lorenzo (ed.) 1988 Grande Grammatica Italiana di Consultazione. Volume 1. La Frase. I sintagmi nominale e preposizionale. Bologna: Il Mulino. Richards, Marc 2004 Object Shift and Scrambling in North and West Germanic: A Case Study in Symmetrical Syntax. Ph.D. dissertation, University of Cambridge. Rizzi, Luigi 1982 Issues in Italian Syntax. (Studies in Generative Grammar 11.) Dordrecht: Foris. Rizzi, Luigi 1997 The fine structure of the left periphery. In: Liliane Haegeman (ed.), Elements of Grammar: Handbook in Generative Syntax, 281–337. (International Handbooks of Linguistics 1.) Dordrecht: Kluwer. Rizzi, Luigi and Ur Shlonsky 2007 Strategies of subject extraction. In: Hans-Martin Gärtner and Uli Sauerland (eds.), Proceedings of the Workshop on Interfaces + Recursion = Language? Chomsky‘s Minimalism and the View from Syntax-Semantics, 115–160. Berlin/New York: Mouton de Gruyter. Roberge, Yves 1990 The Syntactic Recoverability of Null Arguments. Kingston: McGillQueen‘s University Press. Roberts, Ian (ed.) 2007 Comparative Grammar: Critical Concepts in Linguistics. Volume 2: The Null-Subject Parameter. London/New York: Routledge. Solà, Jaume, Maria-Rosa Lloret, Joan Mascaró and Manuel Pérez Saldanya (eds.) 2002 Gramàtica del català contemporani. Volume 2: Sintaxi. Barcelona: Editorial Empúries.

Syntactic variation and change in Romance: a Minimalist approach

403

Sörés, Anna 1995 Rapport génetiques et typologiques dans l‘étude synchronique des langues romanes. Revue Romane 30: 4–79. Svenonius, Peter 2002 Introduction. In: Peter Svenonius (ed.), Subjects, Expletives, and the EPP, 1–25. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press. Tellier, Christine 1997 Éleménts de syntaxe du français. Méthodes d‘analyse en grammaire générative. Montréal: Presses de l‘Université de Montréal. Ura, Hiroyuki 1996 Multiple feature-checking: A theory of grammatical function splitting. Ph.D. dissertation. Cambridge, Mass.: MIT Working Papers in Linguistics. Vallduví, Enric 1992a Focus constructions in Catalan. In: Christiane Laeufer and Terrell A. Morgan (eds.), Theoretical Analyses in Romance Linguistics, 457–479. Amsterdam/Philadelphia: John Benjamins. Vallduví, Enric 1992b The Informational Component. New York: Garland. Zagona, Karen 2002 The Syntax of Spanish. Cambridge: Cambridge University Press. Zanuttini, Raffaella 1997 Negation and Clausal Structure: A Comparative Study of Romance Languages. (Oxford Studies in Comparative Syntax.) Oxford/New York: Oxford University Press. Zwart, Cornelius Jan-Wouter 1997 Morphosyntax of Verb Movement: A Minimalist Approach to Dutch Syntax. Dordrecht: Kluwer.

A localistic approach to universals and variation Brian D. Joseph Abstract A basic premise adopted here is that language is important to both linguists and naïve speakers but in different ways and for different reasons. While linguists can take a broad view of human language, speakers in general take a very local stance towards language, one that is shaped by and extends largely to just their native language(s). With regard to universals, a generally functional and cognitive approach to language universals is advocated here, though with some role for innateness. Universals, it is claimed, have to derive from what speakers do. Speakers are able to generalize, and do so quite well, but they do so only locally, with their generalizations ranging over relatively small sets of data. Incomplete generalization can lead to variation, so a natural question is how local generalizations might lead to universals. The resolution of the conflict between localistically inclined speakers and the existence of universals that derive from their localistic practices comes from the recognition that oftentimes speakers, even of different languages, are reacting to the same sorts of stimuli and are coming at those stimuli with similar cognitive preferences (e. g. for patterning). Examples of local and incomplete generalizations from numerous languages are provided in support of this view. Keywords: analogy, Bulgarian, functionalism, generalization, Greek, innateness, Latin, local generalization, Macedonian, naïve speaker, paradigm, Sanskrit

1.

Introduction

Linguists and language speakers have some overlapping interests and goals; on an obvious level, both groups, whether wittingly or not, see language as something important. Speakers use their language for important purposes, including communicating ideas, being expressive, showing solidarity, marking identity, and so on. Linguists see all of these functions as highly revealing of something important about the human condition and the role that language plays in human life. And, the fact that language is a

A localistic approach to universals and variation

405

structured system is clearly important to speakers, since structure allows for predictability and thus contributes to processing and decoding, and is of paramount importance to linguists, for whom structure offers a basis for scientific examination. But linguists and speakers also have different interests and different perspectives, and nowhere do these differences become clearer than in the realm of language universals. Linguists, by virtue of an interest in the general phenomenon of language, are inevitably drawn to consider what all languages have in common and the ways in which they differ; the quest for linguistic universals is part and parcel of the quest to understand what language is all about. To that end, linguists have a wide variety of tools at their disposal, in the form of dictionaries, grammatical descriptions, historical accounts, comparative evidence from related languages, typological evidence from other languages, experimental results, access to large-scale corpora, and the like, that they use all the time as part of the active examination of language in the abstract. By contrast, the typical “naïve” speaker does not have access to, or certainly does not avail him- or herself of such tools. And there is no reason why speakers should concern themselves with such resources. For speakers, the particular, rather than the universal, is what is crucial. Speakers’ focus is on their language, and whether it fits into a larger schema in some way or another is largely irrelevant for the uses to which they want to put their language. Moreover, speakers are generally interested in preserving a feeling that their language is unique and the use of strategies for the active differentiation of one group’s language from that of another group is well documented. Thus speakers can happily exist and be effective language users without any sense of the universal that might be found in language more generally and which linguists so earnestly seek to elucidate. If speakers are not interested in universals and in a certain sense do not need them, one has to wonder where cross-linguistic parallels of the sort that lead linguists to posit linguistic universals come from. One standard answer is that they are innate and thus any particular language conforms in some way to the dictates of universal grammar by virtue of its speakers’ biological endowment. Especially important in this regard is the notion that the biological endowment helps to shape the language learning – or rather, language acquisition – process; universals thus are instantiated in particular languages through this “guided” acquisition process. Another, essentially countervailing, perspective takes universals as deriving from the functions to which language is put by speakers in their interactions with other speakers; the idea is that similar functions and goals

406

Brian D. Joseph

lead speakers of different languages to similar “solutions” to the linguistic problems attendant with getting their message across to their interlocutors. In either case, what seems to matter for universals is what speakers bring to the enterprise of language learning and language use. If universals are due to innateness, then their realization in actual languages is a matter of how speakers apply them in their acquisition of their own language, and if they are due to functional factors, then their realization is a matter of how speakers use their language as they talk to, communicate with, express themselves to, and in essence perform before, other speakers. These basic facts mean that one way or the other, our understanding of universals rests squarely on our understanding of what speakers do and what they are capable of. Exploring these capabilities thus becomes of paramount importance. Given the premise stated at the outset, namely that speakers and linguists have overlapping but ultimately different goals and overlapping but ultimately different takes on language, an interesting exercise presents itself, namely that of seeing how to balance the universal, representing linguists’ interests, with the particular, representing speakers’ interests. This balancing act further has to take variation into consideration: even if there are universals of language, there has to be room for variation across languages, as they are clearly not all identical, and linguists look to interlanguage differences as a source of insight into the nature of universals (e. g., are they valid for only certain types of languages, are they statistical tendencies rather than absolutes, etc.). But there also has to be room for variation within languages, and it is here that the speaker’s interest in the particular comes into play as well, since intra-language differences often hinge on particularities of detail, whether of a phonetic or a grammatical or a lexical nature, that are salient and noticeable to native speakers. A key question, therefore, is whether it is possible to derive universals and still have variation. I argue here that the answer is yes, and that the answer lies in recognizing the relationship between the particular and the general.

2.

What is a generalization?

In a certain sense, a generalization is an expression of something universal, in that it extends beyond a very localized and highly particular starting point. For instance, to take an example from phonology, covering both sound changes and the synchronic phonological rules that arise from

A localistic approach to universals and variation

407

them, it is often the case that sound change starts small, so to speak, being found at first just in a highly restricted and localized environment where the change is phonetically natural. A case in point is word-final devoicing, as found in Russian, Turkish, and German, among many other languages. Like many if not most practicing historical linguists, we can assume a view of sound change that may be called “Neogrammarian”,1 whereby sound change is taken to be regular, in the sense that all tokens of a sound that occur in the conditioning environment for a sound change will undergo that change, and also conditioned only by phonetic factors. Any such change in the realization of a word can thus be labeled with a technical term referring to sound change in a very strict sense – what may be called “sound change proper” – and must have a phonetic motivation; in this way it would be opposed to other changes in the pronunciation of a word induced, for instance, by morphological factors, as a result of analogically based generalizations, and thus with a more psychological or cognitive motivation. In the case of word-final devoicing, one can argue (see Hock 1976; Joseph 1999) that it is not particularly phonetically motivated, inasmuch as word-finality has no consistent phonetic cue associated with it; rather word-finality is defined at a higher level of analysis, at the level of words rather than of sounds themselves, and thus at the morphological/ syntactic level. From the perspective of Neogrammarian sound change, therefore, devoicing in word-final position could not be a primary event of change; rather, one has to posit that word-final devoicing in a given language had its start in a phonetically natural position and that it spread, i. e. was generalized, from that position to phonetically less natural positions. In particular, following Hock (1976), who draws on Sanskrit evidence, we can say that devoicing in general began in utterance-final position, as that is a context where, due to the silence that follows the end of an utterance, with the vocal folds in a resting position, devoicing is phonetically motivated; word-final devoicing would thus represent a generalization from devoicing in utterance-final contexts to devoicing in word-final contexts. Such a generalization is analogical in nature, with word-finality being influenced by – analogizing to – utterance-finality, based on the fact that for the most part, utterance-ends coincide with word-ends. In this way, therefore, a highly restricted starting point of utterance-final position

1 So called after the (mostly) Leipzig-centered school of historical linguistics in the mid-to-late 19th century, associated with such luminaries as Karl Brugmann.

408

Brian D. Joseph

leads to a more general application and widespread realization of a phonological (i. e., no longer a purely phonetic) process.2 A similar view emerges from the domain of the development of grammar, in that grammatical material with a general application often can be shown to arise from very particular combinations. Jasanoff (1978), for instance, has explained the Latin imperfect in -bƗ-, a formation found for all verbs in all of the regular conjugational classes of Latin and absent only in a few irregular verbs (such as sum ‘be’), as originating in a periphrastic formation, one that has a direct parallel in Vedic Sanskrit. The periphrasis in question consisted of a noun in the instrumental case with a past tense form of the verb *bhuH- ‘be; become’,3 so that an imperfect such as ‘I was laughing’ was originally something like ‘I was with-a-laugh’. Jasanoff posits that the periphrasis was reanalyzed as a monolectal verbal form (thus, “univerbated”), with -bƗ-, deriving from *bhuH-Ɨ-, treated as an inflectional suffix added to a verb stem. Since the instrumental case ending in Proto-Indo-European was *-eH1, an ending which yielded *-Ɲ in Latin, this account means that the imperfect originated in the -Ɲ-stem verbs, that is, in the second conjugational class, and spread from there to the other classes. This very general imperfect tense formation therefore started in a restricted part of the verbal system. Examples of this sort can be multiplied easily. The fully general and completely exceptionless use of -m as a first person singular marker in the present tense in present-day Macedonian, Slovene, and Slovak, for instance, as discussed most recently by Janda (1996), represents the generalization of a verbal ending that in Proto-Slavic was restricted to just five

2 Kiparsky (2008: 45–48) takes a somewhat different view of word-final devoicing, seeing it as a purely phonological phenomenon right from the start, a view that is not entirely compatible with that taken here. That is, Kiparsky seems to have no qualms about starting with a sound change that makes reference to a non-phonetically defined environment like word-finality, attributing the suppression of certain features in codas to “those features [being] perceptually less salient in those positions” (p. 46); by contrast, I would want to ask what it is that makes for diminished perceptual salience – presumably this is primarily an articulatory and/or acoustic phonetic fact, with the phonological correlate being secondary to that. 3 I use the symbol “H” for a so-called laryngeal consonant of Proto-Indo-European that is of indeterminate quality, as in this root; where the quality of the laryngeal can be determined, due to its vowel-colouring properties, for instance, I use a subscript number. Thus *H1 (below) is the “first” or e-coloring laryngeal.

A localistic approach to universals and variation

409

verbs, as it is in Old Church Slavonic.4 Similarly, the full perfect system of Modern Greek, with a present perfect, a past perfect, and a future perfect, in indicative, subjunctive, and imperatival moods, along with a participial as well, seems to have originated in just the past perfect; as argued by Joseph (1983: Ch. 3) and Joseph (2000), following Thumb (1912), the first place one sees a perfect formation of the Modern Greek type (consisting of forms of the verb ‘have’ (stem ex-) and a remnant of the older infinitive, now in -(s)i), is in the past perfect, so that the fully elaborated contemporary system was built up from that limited starting point. Such is the case too with the suffix -ús-, used with end-stressed verbs in Modern Greek to mark the imperfect tense (past imperfective) for all person-and-number forms, as it is best taken as having begun in a reanalysis of the third person plural form and to have been generalized from there.5 The overall lesson to be drawn from such examples is that generally applicable material – what can be recognized as the “stuff” of generalizations, the substance that generalizations are built on – starts out as highly particularized as to its originating context. Speakers therefore presumably build up from particular instances and extend them into ever-larger domains, generalizing the context of occurrence and distribution for particular elements, whether sounds or grammatical markers.6 What this means, however, is that if any domain is missed in the generalization process, what will result is some sort of variation in the realization of the sound or grammatical category being generalized. Incomplete 4 The five verbs in question are byti ‘be’ (1SG esmɶ), dati ‘give’ (1SG damɶ), Čsti ‘eat’ (1SG Čmɶ), imČti ‘have’ (1SG imamɶ), and vČdČti ‘know’ (1SG vČmɶ). Some of the modern Slavic languages, e. g. Russian, show a similarly restricted distribution for 1SG -m; see below for more on this phenomenon in other Slavic languages. 5 That is, the third person plural of a verb-stem in -e- would have ended, in Ancient Greek, in -oun (from stem-final -e- + 3PL ending -on); that ending was extended with the ending -sa-n proper to another past tense formation (the “sigmatic aorist”), giving -oûsan (as if from /-e-o-sa-n/) and this was then reanalyzed as involving an imperfect marker -oûs- (becoming -ús- in Modern Greek by regular sound change) with an ending -an. From that humble beginning, it spread into all other persons and numbers, giving now in Modern Greek, for instance, fil-ús-a ‘I was kissing’, fil-ús-ate ‘you (all) were kissing’, etc. 6 This is assuming of course that the extension process even takes place; it need not, and the conditions that lead speakers to generalize in some instances but to leave a feature in its original restricted environment in others, while an interesting object of study in their own right, are beyond the scope of the present discussion.

410

Brian D. Joseph

generalization leads to there being some loose ends, so to speak, in the grammar viewed more widely. For example, there are speakers of American English for whom word-final devoicing seems to be restricted to occurrence just in vocatives,7 so that a shout to someone named David or Jacob can be realized as [de+v+t] and [d