New Trends in Nordic and General Linguistics 9783110346978, 9783110343861

Citation preview

Martin Hilpert, Jan-Ola Östman, Christine Mertzlufft, Michael Rießler, Janet Duke (Eds.) New Trends in Nordic and General Linguistics

linguae & litterae

Publications of the School of Language & Literature Freiburg Institute for Advanced Studies Edited by Peter Auer, Gesa von Essen, Werner Frick Editorial Board Michel Espagne (Paris), Marino Freschi (Rom), Ekkehard König (Berlin), Michael Lackner (Erlangen-Nürnberg), Per Linell (Linköping), Angelika Linke (Zürich), Christine Maillard (Strasbourg), Lorenza Mondada (Basel), Pieter Muysken (Nijmegen), Wolfgang Raible (Freiburg), Monika Schmitz-Emans (Bochum)

Volume 42

New Trends in Nordic and General Linguistics

Edited by Martin Hilpert, Jan-Ola Östman, Christine Mertzlufft, Michael Rießler and Janet Duke

ISBN 978-3-11-034386-1 e-ISBN [PDF] 978-3-11-034697-8 e-ISBN [EPUB] 978-3-11-026793-8 ISSN 1869-7054 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. © 2015 Walter de Gruyter GmbH, Berlin/Munich/Boston Typesetting: epline, Kirchheim unter Teck Printing: Hubert & Co. GmbH & Co. KG, Göttingen ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Contents Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, Janet Duke Introducing new trends in Nordic linguistic research  1

Section 1: Language contact Kurt Braunmüller Competing tendencies in Germanic pronominal and deictic systems: The most general principle will prevail  11 Pavel Iosad “Pitch accent” and prosodic structure in Scottish Gaelic: Reassessing the role of contact  28 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn Implications of language contact: Evaluating the appropriateness of borrowings in written Icelandic  55 John Weinstock At the frontier: Sámi linguistics gets a boost from outside  68

Section 2: Phonology Natalia Kuznetsova Two phonological rarities in Ingrian dialects  91 Kristján Árnason Analysing phonological variation in Faroese  118

Section 3: Morphosyntax Jeremy Bradley Mari converb constructions – Interpretation and translation  141 Ulla Stroh-Wollin Han and hon – Anaphoric pronouns in Early Scandinavia  162

VI 

 Contents

Margrét Jónsdóttir From accusative to dative (via nominative): The case of fjölga ‘increase’ and fækka ‘decrease’ in Icelandic  181

Section 4: Syntax Karl Erland Gadelii A generative interpretation of Diderichsen’s positional grammar  205 Henrik Rosenkvist Evidence for a syntactic Parameter at work in Övdalian  224 Ida Larsson and Janne Bondi Johannessen Embedded word order in Heritage Scandinavian  239

Section 5: Grammaticalization Andres Karjus Through the spyglass of synchrony: Grammaticalization of the exterior space in the Eastern Circum-Baltic  267 Helle Metslang, Karl Pajusalu and Külli Habicht Conjunctive markers of polar questions in Estonian  283

Index  307

Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, and Janet Duke

Introducing new trends in Nordic linguistic research 1 A fresh start This book started with the idea of reviving a broken tradition. The conference series International Conference of Nordic and General Linguistics (ICNGL) had been organized at regular intervals between 1969 and 1998, but had disappeared from the linguistics conference circuit. The conference series had served as a common forum for linguists from the Nordic countries and international linguists working on Nordic languages, while at the same time being open to general linguistics, practiced by anyone, on any topic. Eliasson (2010) chronicles the conference series and contextualizes it in the research landscape of Nordic linguistics. To us, the editors of this book, the concept of this conference series was too attractive to be simply dropped; we thought it a worthwhile endeavor to re-invigorate the tradition, thereby creating a space for a community of researchers with a common, explicit or implicit interest in the Nordic languages, but with a wide spectrum of theoretical and methodological approaches, and without any restriction to a closed set of languages to be investigated. What we had in mind was as much a reconnection with the past as also a forward-looking enterprise. We wanted to see what issues were currently being worked on in the Nordic linguistics community in order to get a sense of the direction in which Nordic linguistics is moving. We also aimed for a broad interpretation of the adjective Nordic, including not only the languages of Denmark, Finland, Iceland, Norway, and Sweden, but also those of Greenland, the Faroe Islands, Åland, Estonia, Latvia, Lithuania, Russia, and other countries around the North Sea, the Baltic Sea, and the Barents Sea. Given that the ICNGL conference had been organized outside the Nordic countries before (Austin, TX in 1976), it seemed not too much of a stretch to re-start the tradition in Freiburg, in the southwest of Germany. In 2012 then, it was time for a fresh start. With due permission from the Nordic Association of Linguists (NAL), the 11th International Conference of Nordic and General Linguistics (ICNGL11) was organized at the University of Freiburg (Albert-Ludwigs-Universität Freiburg) by the editors of this book together with Peter Auer, the then-director of the linguistics section of the Freiburg Institute for Advanced Studies (FRIAS), and our conference secretary Elizabeth Zima and

2 

 Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, and Janet Duke

conference assistant Diane Zille. We were further helped by the members of the advisory board of the conference, Jan Anward, Kurt Braunmüller, Tove Bull, Östen Dahl, Stig Eliasson, Johanna Laakso, and Damaris Nübling. An important sponsor that helped us transform our idea into an actual event was the Deutsche Forschungsgemeinschaft (DFG). Nordic and linguistics scholars and students alike in Freiburg gladly offered their assistance at the conference site. And for the practical preparation of this volume we have been assisted by Iris Perkmann. But the administrative side of a conference is merely a facilitator for the real thing. The success of a conference is measured by its academic man- and womanpower, by those who attend and present new research results. By the testimony of those who were present, the conference was a great success and we were pleased to see that our sense of what the field needs was indeed shared by the linguistics community at large. The conference drew a large number of participants from the Nordic countries as well as from other parts of the world, and there were 90 academic presentations altogether at the conference. Among the presentations were six invited plenary talks, by Lars-Olof Delsing, Frans Gregersen, Auli Hakulinen, Martin Hilpert, Anneli Sarhimaa, and Sarah Thomason. Overall, the topics of discussion showed a strong focus on Nordic linguistics, but there were presentations on a good many other languages as well. Making a point of going outside of the Nordic countries to engage practitioners of Nordic and general linguistics more broadly proved to be a very welcomed and appreciated step. In fact, the success of the conference was so overwhelming that it was decided to continue the conference series further. The twelfth conference in the series will thus be organized in Helsinki in 2016.

2 The volume The contributions in this volume have been selected from among the presentations given at that conference. They have gone through the process of external peer reviewing, and we can only regret that we were not able to include more of the papers that were submitted to us after the conference. The contents of the book mirror the diversity that we experienced at the conference, and that characterizes Nordic linguistics in the early twenty-first century. In fact, the very notion of diversity is, and has always been, at the center of Nordic linguistic research, both with respect to empirical data and with respect to the many-foldedness of the theoretical and methodological approaches that are practiced side by side. In this volume, analyses of phonological, morphological, and syntactic phenomena are found next to studies on language variation and change, language contact, grammaticalization, and typological comparisons. The reader is taken



Introducing new trends in Nordic linguistic research 

 3

all around the Baltic Sea, into Russia, to Iceland and the Faroe Islands, and even to speakers of Heritage Scandinavian in North America. Nonetheless, in this set of contributions there are recurring common threads that represent issues that have also in the past informed Nordic linguistics. Several chapters in the volume deal with the conditions and consequences of language contact, other chapters reflect on the social forces that shape linguistic variation, and others still take language diachrony into account in their explanation of synchronic phenomena. The studies in this volume reflect the topics and theories that are traditionally associated with Nordic linguistics, while at the same time pointing the way towards new developments. We find papers both in the traditional Nordic temperament, like studies in the spirit of Diderichsen and in traditional structuralist approaches of various kinds, as well as papers in frameworks like generative minimalism. We invite readers to sample the offerings of this volume and to discover a tradition that is not only very much alive, but that is eager to move on.

3 The contributions The volume is structured into five sets of chapters that reflect different thematic foci. The first and largest group of studies is dedicated to language contact, which was the central conference theme for ICNGL11. The chapters that follow are ordered in accordance with the levels of linguistic structure that are addressed: phonology, morphosyntax, and syntax, respectively. The volume is concluded with a number of chapters addressing processes of grammaticalization. Each set of approaches has contributions on a variety of different languages, illustrating how common points of interest reach across linguistic and political boundaries.

3.1 Language contact The opening contribution by Kurt Braunmüller compares the pronominal systems of old Germanic languages with the aim to study the possible consequences of language contact, notably simplification, increase of semantic transparency, and the loss of equivalent and hence ‘superfluous’ forms. The specific case studies that inform the discussion include the integration of Scandinavian third person plural pronouns into the Old English pronominal system, enclitic definite marking in Norwegian and Swedish, and the shift in Mainland Scandinavian from definite articles beginning in h- to definite articles beginning in d-. The chapter concludes by suggesting that grammatical terminology can be

4 

 Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, and Janet Duke

fruitfully reconsidered if the perspective of the second language learner is taken into account. The project of ‘rethinking grammar from an L2-perspective’ has profound consequences for grammar writing, as it breaks away from what Braunmüller characterizes as a fossilized view of grammar that is based on the categories of Classical Latin and Greek. Pavel Iosad addresses pitch accent and prosodic structure in Scottish Gaelic, which shows intriguing parallels to pitch accent contrasts in Swedish and Norwegian, and which thereby provokes the question of whether language contact had a role to play in the emergence of this phenomenon. Using criteria set forth in Thomason (2001), the paper reviews the evidence for and against a common origin, ultimately arguing in favor of parallel developments in Scottish Gaelic and North Germanic, as opposed to an explanation in terms of language contact. This interpretation is supported by the observation that pitch accent can be seen as the outcome of prosodic patterns that are independently attested in various branches of Celtic, not all of which were in contact with North Germanic. Ari Páll Kristinsson and Amanda Hilmarsson-Dunn study Icelandic speakers’ attitudes towards borrowings, examining the effects of age, profession, and text genre. While speakers of Icelandic are commonly reputed to harbor a certain linguistic conservatism, the results of this study suggest a more differentiated picture. Whereas speakers of all age groups view texts with borrowings as inappropriate in formal prestigious genres, the same speakers find them appropriate for genres of electronically mediated communication. That is, sensitivity to genre conventions is shown to be a more reliable explanation for the informants’ behavior than what has traditionally been characterized as a general ideology against borrowed items in Icelandic. The authors further show that exposure to different genres leads to different attitudes in younger speakers than in older speakers. John Weinstock broadens the discussion of language contact towards prehistoric developments, offering a reflection on the early development of the Sámi languages. The paper re-examines a proposal made by Aikio (2004), which locates the emergence of Proto-Sámi near the Gulf of Finland at the onset of the Iron Age. Weighing different pieces of anthropological, archaeological and genetic evidence, the study concludes that Aikio’s hypothesis remains unrefuted, as it can be brought into accord with the available linguistic evidence.

3.2 Phonology In her chapter on the endangered Finno-Ugric language Ingrian, Natalia Kuznetsova addresses two phonological rarities. The first of these is a case of con-



Introducing new trends in Nordic linguistic research 

 5

sonant gradation, a phenomenon that is quite common across Finnic and Sámi languages, but which is special in Ingrian because a three-way quantity contrast can be observed in this language. As a second phenomenon, the study discusses reduced voiceless vowels. Vowels of this kind are typologically rare, and the few cases that are discussed in the literature differ qualitatively from those that are found in Ingrian. The chapter offers a description of these two phenomena and continues with a thorough discussion of their evolution and maintenance. Kristján Árnason studies phonological variation in Faroese, which differs sociolinguistically from both continental Europe and from Iceland, with which it shares tendencies of linguistic conservatism. For a long time, the Faroese community had been diglossic, with Danish serving as the H-variety for use for clerical and administrative purposes. An indigenous written standard of Modern Faroese was only created in the 19th century; that is, the linguistic situation and the historical development of Faroese is in many respects like the development of Norwegian. Analyzing variables such as final syllable reduction and final vowel deletion in present-day spoken Faroese, Árnason identifies patterns that differ from traditional descriptions. These findings are discussed with reference to the phenomena of dialect levelling and koineization.

3.3 Morphosyntax In a chapter that illustrates the outward-looking nature of the ICNGL conference, Jeremy Bradley offers a description of converb constructions in the Finno-Ugric language Mari, which is spoken in the Volga basin. The discussion focuses on how these complex verbal predicates map onto different meanings, such as simultaneous action, cause and effect, directed motion, as well as convey different aspectual meanings. The study further raises the issue of how converb constructions are to be translated into languages that lack comparable morphosyntactic structures. Ulla Stroh-Wollin studies the diachrony of the anaphoric pronouns han(n) ‘he’ and hon (hun) ‘she’. These forms are exclusive to North Germanic and their origins are the subject of a long-standing debate. Taking another look at the history of potential etymologies for han, the study revisits a highly influential account by Kock (1908), who viewed the pronoun as a Proto-Nordic innovation, and who explicitly argued against a reconstruction from Indo-European (cf. Greek kēnos). Stroh-Wollin argues that Kock’s proposal is less plausible than the IE proposal, which is based on regular sound changes. It is suggested that Kock’s proposal owed its popularity to the predominant assumption of the early 20th century that Proto-Germanic was relatively homogenous. It therefore, at the time,

6 

 Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, and Janet Duke

seemed unlikely that these pronouns could have survived only in the North Germanic branch. The chapter by Margret Jónsdóttir addresses the Icelandic “dative sickness” (Thráinsson 2007), which describes the diachronic drift from accusative to dative subjects documented in many Icelandic verbs, which can take both dative and accusative subjects. Since non-nominative subjects have been lost in Mainland Scandinavian as well as in English, it is commonly thought that the diachronic drift finds its natural end in nominative subjects. Jónsdóttir’s study shows that the issue is more complex. The chapter focuses on the verbs fjölga ‘increase (in number)’ and fækka ‘decrease (in number)’, which have changed in the history of Icelandic from having subjects in the accusative to having them in the dative, but with nominative appearing as an intermediate stage. The study discusses possible explanations for this development, among them a semantic connection between the causative and anticausative uses of fjölga and fækka.

3.4 Syntax Karl Erland Gadelii takes up a theoretical model that has a long tradition in the syntactic analysis of the Nordic languages, namely Diderichsen’s positional scheme (Diderichsen 1946). Gadelii argues that this topological tool for the analysis of clausal syntax not only has a useful role to play in language teaching and grammatical description, where it remains in current use in various guises, but that the scheme holds implications for syntactic theory, a point that is not widely acknowledged. In order to support this claim, Gadelii shows the need to augment Diderichsen’s original syntactic template by an additional slot and demonstrates the applicability of this augmented scheme for a range of examples discussed in the generative-minimalist literature. The chapter furthermore notes a set of problematic examples and suggests intricate avenues for further study. Henrik Rosenqvist takes a principles and parameters approach to the analysis of Övdalian (Swe. älvdalsmålet), which is spoken in North-Western Dalarna, Sweden. Within this approach, the setting of a parameter is thought to influence several syntactic structures at the same time, thereby constraining what structures may be found alongside each other in the same language. For instance, it has been argued for the Scandinavian languages that syntactic constructions such as null expletive subjects, transitive expletives and stylistic fronting appear in languages that show verb raising in embedded clauses and that have rich inflectional morphology (Holmberg and Platzack 1995). However, a recurrent problem in research addressing this issue has been that dialectal variability has made broad generalizations impossible. Rosenqvist argues that referential null



Introducing new trends in Nordic linguistic research 

 7

subjects in Övdalian do indeed provide evidence that there is a specific syntactic parameter at work that influences these structures. The final chapter directly addressing syntactic phenomena is the contribution by Ida Larsson and Janne Bondi Johannessen, which takes the reader across the Atlantic and offers a discussion of embedded word order in Heritage Scandinavian, i. e. the language of American Scandinavians who were born in America. On the basis of authentic recordings it is shown that Heritage Scandinavian differs from European Scandinavian with respect to word order in syntactically subordinate contexts, specifically as regards the order of adverb and finite verb in consecutive clauses, indirect questions, and relative clauses. The study considers several explanations for these phenomena. The possible explanations in terms of language contact and language attrition are rejected as the authors settle for an explanation in terms of generational change through incomplete language acquisition.

3.5 Grammaticalization The contribution by Andres Karjus is concerned with the development of lative markers that encode an outward direction. The paper focuses on four Eastern Circum-Baltic languages, specifically the Finnic languages Estonian and Võro and the Indo-European languages Latvian and Lithuanian. In these languages, a lexical item with the meaning ‘field’ has given rise to different lative markers. Karjus takes a synchronic approach, using questionnaire data to explore the semantic spectrum that these markers cover in present-day usage. The responses reveal cross-linguistic differences in the semantic breadth of the respective markers, which suggests that in some languages, the markers have grammaticalized to a greater extent than in others. The chapter demonstrates that quantitative comparisons of semantic variation can be usefully applied to the cross-linguistic study of grammaticalization. Helle Metslang, Karl Pajusalu, and Külli Habicht also deal with a Finnic language; their study addresses interrogative particles in Estonian that have developed out of additive, contrastive, and adversative conjunctions. In Estonian, polar questions are typically formed by means of sentence-initial or sentence-final particles. In relation to this, the chapter discusses several markers in which questions are introduced by elements that can also serve as conjunctions. The study explains the grammaticalization of coordinative interrogative particles with reference to the theory of on-line syntax (Auer 2009). When speakers use a conjunction to introduce an increment to an interlocutor’s speech, the follow-up proposition can be interpreted as an implicit request for confirmation

8 

 Martin Hilpert, Michael Rießler, Jan-Ola Östman, Christine Mertzlufft, and Janet Duke

or non-confirmation from the interlocutor. In this way, constructions with initial conjunctions may become conventionally associated with interrogative meaning.

4 The way ahead If we take this sample of contributions from the Freiburg ICNGL conference as a representation of Nordic linguistics as it currently lives and breathes, what can we say about its future? We can observe that there are traditions and topics of interest that continue to be especially relevant in a Nordic context, such as language contact, the comparison of closely related languages and language varieties, and the diachrony of linguistic variation. To us it seems that a promising development visible in the contributions of this volume is the continuation of research on these issues, as the discipline broadens its geographical scope and as newly-developed methodological tools and theoretical concepts are adopted and applied to issues in Nordic linguistics. How this development will unfold in practice is a matter that presentations at future ICNGL meetings will undoubtedly reveal.

References Aikio, Ante (2004): An essay on substrate studies and the origin of Saami. In: Irma Hyvärinen, Petri Kallio and Jarmo Korhonen (eds.), Etymologie, Entlehnungen und Entwicklungen: Festschrift für Jorma Koivulehto zum 70. Geburtstag, 5–34. Helsinki: Société Néophilologique. Auer, Peter (2009): On-line syntax: thoughts on the temporality of spoken language. Language Sciences 31: 1–13. Diderichsen, Paul (1946): Elementær dansk Grammatik. Copenhagen: Gyldendal. 2nd edition 1957, 3rd edition 1962. Eliasson, Stig (2010): The Nordic Association of Linguists: The preparatory phase and the first thirty years (1977–2006). In: Hans Götzsche (ed.), Memory, mind and language, 4–54. Newcastle upon Tyne, UK: Cambridge Scholars Publishing. Holmberg, Anders and Christer Platzack (1995): The Role of Inflection in Scandinavian Syntax. Oxford: Oxford University Press. Kock, Axel (1908): Etymologiska anmärkningar. Arkiv för nordisk filologi 24: 179–198. Thomason, Sarah (2001): Language Contact: An Introduction. Washington D. C.: Georgetown University Press. Thráinsson, Höskuldur (2007): The Syntax of Icelandic. Cambridge: Cambridge University Press.

Section 1: Language contact

Kurt Braunmüller

Competing tendencies in Germanic pronominal and deictic systems: The most general principle will prevail Abstract*: The pronominal systems of old Germanic languages vary considerably both in their inflexional forms and in the forms of autonomous pronouns due to transfer from genetically closely related dialects. When language contact between related languages occurs (e. g. Scandinavian > Old English or Middle Low German > Scandinavian), complexity does not necessarily increase. The results, most of them due to L2 acquisition by adults, may be (A) simplification, e. g. a reduction to basic referential deictic functions in discourse, (B) increase of semantic transparency/iconicity or (C) disregard of equivalent forms, which obviously were considered ‘superfluous’ by L2-learners, and preference of other forms with the same deictic function. The respective instances to be discussed are: (A) Transfer and later integration of Scandinavian third person plural pronouns into the Old English personal pronominal system. These borrowings should, however, not be regarded as beneficial suppletion, establishing a maximum of contrast between the singular (he/his/ him/hine, masc.) and the new third person masc. & fem plural forms (þe[a]i/þeira/ þe[a]im/ < þei[æ]r/þei[æ]r/þe[æ]im, replacing older hīe/hira/him). (B1) Obligatory enclitic [+definite] marking of definite NPs in Norwegian and Swedish on the one hand and structural convergence between German and Danish complex NPs on the other. (B2) Retention of the t-definite article (tann, tað) in colloquial Faroese due to bilingualism with Danish – though all other Old Norse demonstrative pronouns on þ- merged with the other deictic h-forms (þetta > hetta; har ‘there’; her ‘here’; cf. also the other definite article hin, hitt, enclitic -in, -ið). (C) Loss of the h-definite articles in Mainland Scandinavian in favour of the functionally equivalent forms beginning with d- (den, det; de, dem), which has its causes both in inter-lingual functional identification and substantial phonetic equivalence to Middle Low German de ‘the’. In all these cases, simplification, transparency or disregard of equivalent forms prevailed.

* I would like to thank the two reviewers and the editors, esp. Martin Hilpert, for helpful comments and suggestions.

12 

 Kurt Braunmüller

1 Introduction When looking at the presentation of pronouns and pronominal systems in historical grammars of Germanic, one will realise that they are organised and displayed in divergent ways. Some grammars, such as Krahe and Meid’s survey (1969, II: 51–74), begin with “The [!] pronoun”: the personal pronouns of the 1st and 2nd person, followed by the reflexive pronouns and finally the 3rd person pronouns. No definitions are provided as to why the personal pronouns of the first two persons (singular/plural) should be considered ‘pronouns’. Although it is common grammatical terminology to speak of personal deictic forms such as German ich / Swedish jag ‘I’ and German/Swedish du ‘you’ as ‘pronouns’, it should be noted that deictic terms for the 1st and 2nd person are in many respects quite unlike pronouns. They represent, among other things, semantically unspe­ ci­fied referential markers for the speaker(s) and the addressee(s) and do not refer back to any (syntactic) antecedent such as he, she or it/they. Prokosch (1939: 266–285), still one of the leading reference surveys in English, starts out with the 3rd person pronouns in three genders (cf. Latin is, ea, id ‘her, she, it’), then presents the anaphoric, relative, interrogative, indefinite and, finally, the personal pronouns (of the first and second person). This is precisely the reverse order, obviously based on predominantly syntactic criteria. The degree of co-reference decreases from the 3rd person pronouns, via relative and indefinite to the personal pronouns (1st and 2nd person). Hirt (1932: 70–85) proceeds in a similar way as Krahe and Meid (1969), but decompo­ses these ‘pronouns’ of whatever kind into (a) stems and (b) inflexions, which paves the way for the so-called corresponding (or genitival) adjectives, better known as possessive pronouns. Then the gender-marked pronouns are dealt with (sc. the pronouns of the 3rd person, occurring as masculine, feminine and neuter in form), followed by the interrogative, indefinite, demonstrative pronouns (and their inflexions), concluding with the composed pronouns. These three approaches clearly show that the category of ‘pronouns’ is a mixture of rather divergent parts-of-speech with predominantly morphological/ inflectional and syntactic/anaphoric features. None of them is, however, defined in an explicit way by the previously mentioned criteria. All these ‘pronouns’ are obviously regarded as a natural class or, at least, as inherited morphological categories that need not be defined or ordered according to certain perspicuous criteria. They seem to be pre-existing, pre-theoretical and are, therefore, non-questionable. However, as is well-known due to efforts within structural linguistics during the last de­cades, ‘pronouns’ represent a very heterogeneous linguistic category based on morphological, syntactic and some other not explicitly specified crite-

Competing tendencies in Germanic pronominal and deictic systems 

 13

ria, such as the genitive of possession (cf. the possessive pronouns such as my/ mine or your/yours). As already pointed out, the pronouns of the first and second person are, syntactically speaking, by no means ‘pronouns’ that refer ana- or cataphorically to any syntactic constituents but represent semantically unspecified role variables (Braunmüller 1977:  119–125) for the speaker/s (I, we: a. ‘I + you’, b. ‘I + he/they’) or the addressee/s (you). The common de­nominator might have been some inflectional similarities or paradigmatic intersections with other pronouns of the third person (cf. German mein-, dein-, sein- or Swedish min, din, sin/ sitt ‘my, yours, his’). This mixture of divergent grammatical criteria and, as a consequence, the cross-classifications and intersections between morphology and syntax might have caused problems for learners, especially for foreign language learners, who have to find out the systematic and accidental differences between these related parts-of-speech and their overlapping functions. They have, for example, to discover whether or to what extent it is possible to use third person pronouns (German sie, Scandinavian de/dem ‘they’ [± nom.]) and the referentially equivalent demonstrative pro­nouns (German die, Scandinavian de/dem ‘these’ [± nom.]) in the same syntactic frames (cf. the use of which and that as relative pronouns in English; cf. also Diessel 1999: 120–127). As the Scandi­navian examples show, these forms might be formally identical. But why is that the case? Is this syncretism accidental or rather the result of a functionally motivated change? In the following sections we will present an explanation for this kind of merger, based on observations from Germanic language history, language contact and language acquisition. At the end it should have become clear why the presentations in some historical grammars of Germanic (dialects) are so heterogeneous and inconsistent in their classifications or categorical definitions. L2-languages learners will turn out to be the key for a deeper understanding of functional equivalence between some pronominal categories and the ways to simplify or ra­ther unify parts of the pronominal systems in the Germanic languages (see McWhorter 2007 for a broader survey).

2 Pronouns and language contact The pronominal systems belong to the core of grammar and are therefore rarely affected by (external) borrowing (cf. Law 2009: 214–218). But the Old English personal pronominal sys­tem seems to belong to one of these few exceptions since Old English has borrowed the third person plural pronouns from Ancient Nordic during the Viking era – as it has been claimed by the ruling doctrine: Old English hī(e); hira; him > Middle English þeʒʒ, þe, þai; þe/air; þai/em. However, the Mainland Scandinavian demonstrative pronoun systems seem to be af­fected

14 

 Kurt Braunmüller

by language contact as well. A homophonic relation to the Low German demonstrative pronouns has been established, e. g. Old Swedish þæn, þē, þæt; þē(r), þā(r), þø̄n/þé [nom.sg.; nom.pl.] (cf. Bergman 1972: 51–52), pronounced as: [ðæn, ðe(:), ðæt; ðe:(r), ða:(r), ðø:n/ðe:] ≈ Middle Low German dē/di(e), dat; dē/di(e) [sing.; plural]. The modern (Mainland) Scandinavian languages have taken over the borrowed obstruent from Low German,1 a phenomenon which, again, reflects a case of simplification in pronunciation from a (voiced) fricative to a (voiced) obstruent: den, det; de, dem. These two examples clearly show that there are strong tendencies to abo­lish the inherited 3rd person pronouns in favour of (simple) demonstrative pronouns.2 This development towards rendering personal reference by using demonstratives is, however, much older. It goes back to Ancient Germanic and can be observed in Ancient/Ru­nic Nordic: is/eR > Old Norse (ON) hann ‘he’, which is part of the Indo-European (IE) pronominal system: IE *kw- (cf. Latin quis, quid) > Germanic *χ- > h- /*χw- > hw-. Moreover, the neuter singular and the whole plural be­came substituted by other, competing deictic pronouns, which are characterised by a dental in the onset: IE *te-/to- (cf. Old Greek τó(δε) ‘that’; Germanic *þe-/þo-, in weak stress positions realised as [ðe-/ðo-]). The nominative singular paradigm of Old Norse still re­flects the adoption of both demonstrative roots, h- and þ-, into the paradigm of the personal pro­nouns: han, hon; þat ‘he, she; it’, whereas only the þ-root occurs in the plural: þeir, þæ¯r, þau ‘they’. A parallel develop­ ment can be observed in Modern (colloquial) German, where díe often replaces the person­al pronoun sie ‘they’ – parallel to the integration of the Scandinavian pronouns into the Old English pronominal system mentioned above (more about these replacements in section 4.1). The result of all these developments can be seen as a levelling of the (superficial) dif­fe­rences between exophoric/direct and discourse deixis. The aim of this

1 This observation leads to the more crucial issue of whether the phonological change from /ð/ to /d/ in Main­land Scandinavian is, as I suppose, in general due to language contact with (Low) German (viz. simplification by L2-learners of Swedish and Danish) or whether this development has to be considered an internally mo­tivated tendency (cf. Sapir’s drift) towards obstruents which are, in some respect, simpler to pronounce than fri­catives. The most obvious counter-example in Germanic is English however: it still keeps both voiced and non-voiced dental fricatives in almost all of its varieties and dialects. Icelandic, which shows the same fricatives as English, has to be treated as a special case due to a rigorous language preserving policy, but Faroese fits nicely into the L2-simpli­fi­cation hy­po­thesis since it has no dental fricatives any more: Old Norse þ/Þ /θ/ > Faroese /t, (h)/ and ON ð /ð/ > Far. /∅; j,v,w; (g)/. 2 This development is part of a more general tendency also observed in other languages, such as the Ro­mance languages. Our investigation is, however, limited to the Germanic language family.

Competing tendencies in Germanic pronominal and deictic systems 

 15

drift is to unify the modes of (deictic) reference.3 When referring to third persons or objects, a unification of the modes of reference takes place, the aim of which is to achieve more generalisation and sim­plification in reference, thereby removing differences between direct and discourse refe­rence.

3 Redundancy, polyfunctionality, and variation The question is now why there is so much redundancy and duplicity of equivalent functions and cate­gories in grammar. Why have we active and passive voice (and, as in the case of the Mainland Scandinavian languages, even several types of passives), converse verbs such as to give ↔ to receive, different demonstrative pronouns (this, that), which are equivalent in func­tion to a certain extent; or more generally, why is there so much variation and stratification in lan­guage? The answer must be that any linguistic system represents some sort of a common deno­mi­nator of several more or less diverging subsystems that may be considered varieties of a greater overall system. Which forms will prevail, and in which con­texts, is still an open ques­tion and can be a motive of linguistic change. Variation and flexi­bility in rephrasing one’s message presupposes diffe­rent means to ex­press equivalent functions and similar meanings.4 Without variation and redundancy, language change triggered by lan­guage contact, e. g. by re­plication (Heine and Kuteva 2005, ch. 2), would be much more difficult. Language contact

3 As shown otherwise (see Braunmüller 2008: 360–368), demonstratives – much like the socalled weak adjective declension (Proto-Germanic *-an, Modern German -en) with its individualising function – have been subjec­ted to ‘grammatical de­generation’: They obviously have to be reinforced by adding a combination of vari­ous deictic elements (cf. Danish den, den her/der, den der over, denne etc., or German der, der da/dort, der da drüben, die­ser ‘this one here/over there’, to name some of the most fre­quent deictic expressions in two Ger­ma­nic languages for multiple/re­pe­titive deixis). Diessel’s investigation (1999: 2) focuses rather on the various distinctions connec­ted with the para­meter ‘proximity’: [near] vs. [distance] to the deictic centre. 4 For example: the use of the passive mode allows the speaker to neglect or to suppress the mentioning of the agent. Dif­ferent passive forms (inflexional vs. periphrastic) may allow speakers to distinguish between actions that are e. g. performed ‘habitually’ or ‘non-habitually/result oriented’ (as in Danish in the present tense: tømmes vs. bliver tømt ‘gets emptied’. This feature is, however, often overlooked by L2-learners). Emphasis on specific objects may be expressed morphologically by left-dislocation and non-nominative inflexional markers or syntactically by cleft-sentences. In addition, the use of lexical synonyms (of various origins) or the change of the spea­ker’s perspective by using converse verbs/adjective may also be referred to as instances for redundancy and variation in speech. All these examples document the principle of diverging but (semantically) overlapping subsystems.

16 

 Kurt Braunmüller

adds, at least in the beginning, nothing more than variation (and thus more complexity) to a non-unified system. Which forms, and sometimes also which functions, disappear in the long run, is a subordinate issue. Due to both redundancy in grammar and inde­ter­mi­nacy (or vagueness) in grammatical representations, similar or equivalent grammatical functions can replace each other without any severe consequences for the functioning of the system as a whole.

4 R  eplacing personal pronouns by demonstratives in Germanic languages In the following, four cases will be investigated: (a) the replacement of the Old English personal pronouns in the plural paradigm by deictic pronouns, (b) the obligatory enclitic [+definite] marking of the definite noun phrases (NPs) in Norwegian and Swedish on the one hand and the structural convergence of Danish and German definite NPs on the other, (c) the retention of both the t- and h-defi­ nite articles in Modern Faroese due to Faroese-Danish bilingualism, and finally (d) the loss of the h-based definite article in Mainland Scandinavian languages and the preference of translinguis­tic convergences with Middle Low and, later, High German forms (de-articles).

4.1 T  he replacement of Old English personal pronouns by their Scandinavian equivalents The generally received view, which has never really been challenged, says that the Old Eng­lish per­sonal pronouns in the plural became replaced by their Ancient Nordic counterparts due to intense language contact (cf. English they, them; their < Old Scandinavian þeir, þeim; þeir). The purpose of this restructuring of the Old English pronominal system was to avoid homo­nyms. Some Old English forms of the singular were quite similar to their counterparts in the plural: masc.nom.sg. he, fem.nom.sg. hīo, hēo, neuter.nom./acc.sg. hit vs. nom.pl. hīe, hī ... hēo etc. (cf. Brunner 1965: 261–262/§ 334 or Pilch 1975: 121/§ 28). Werner’s investigation (1991: 384), one of the most recent contributions to this issue, comes to the fol­lowing conclusion: “The speakers of English preferred, however, to use the distinct expres­sion from the other language, thus avoiding homophony and category-neutralization.” This explanation may be regarded as a case of beneficial suppletion. But Werner has also no­ticed the occurrence of an already blended paradigm in Ancient Nordic: the neuter.sg. form (ON hit) and all plurals (ON þeir, þæ̅r, þau; þeira, þeim) have actually been taken from the para-

Competing tendencies in Germanic pronominal and deictic systems 

 17

digm of the de­monstrative pro­nouns. The conclusion to be drawn from Werner’s explanation is the following: Should not, in the end, the spea­kers of Old English have been very grateful to the Scandinavi­an invaders, the Vikings, be­cause they prevented their Old English pronominal system from a collapse by homophony? More generally speaking, is there something like a ‘beneficial sup­pletion’? Let us take a step back in order to come to a more plausible explanation. The contact be­tween the Anglo-Saxons and the Scandinavians can undoubtedly be classified as a case of dia­lect contact, viz. as contact between mutually intelligible varieties (cf. Braunmüller 2002: 1032). In addition, the Scandina­vi­an invaders were mostly illiterates and learned those Old Eng­lish dialects they were faced with, if at all, incompletely as an L2 acquired during adulthood, viz. with simplifications and deficiencies. Gram­matical correctness and the use of an elaborated style were not of importance in oral face-to-face communication at that time. They used de­mon­strative pronouns for any direct or indirect (discourse) reference – as their ancestors had done centuries before, when they replaced the Ancient Nordic forms is/er by another demon­stra­tive pronoun hit (cf. section 2) – for whatever reasons. Now, the Norse speaking people sim­pli­fied the Old English pronominal system due to imperfect language learning and made this sys­tem – unwittingly – more transparent, comparable to the so-called Invisible Hand Principle (see Keller 1990: 91–105). Later on, the mixed English-Scandinavian local population kept these deictic pro­nouns when referring to plural entities, as in present-day colloquial German: Die ha­ben es getan ‘They have done it’ [die = subject] or Ich habe die aber heute noch nicht gesehen ‘But I haven’t seen them today yet’ [die = object].5 The use of demonstratives for third person references and for references to objects thus became unified and the formal differences that have to be ob­ served when referring to things or persons were equalised. The result was a much simpler system of reference by pronouns since the distinction between personal and deictic reference had been neutralised in these instances.

5 Thus, focus, stress or case does not play a role here. Third person demonstratives can replace third person personal pronouns in colloquial speech – and without any semantic or referential differences. die, mentioned in the second example, may also represent an object constituent in the third person singular feminine (‘her’), which ad­ditionally supports our hypothesis concerning the functional equivalence of third person pronouns and demonstratives. For a more comprehensive cross-linguistic perspective see Diessel (1999: 95–100).

18 

 Kurt Braunmüller

4.2 T  he use of double definiteness marking in Swedish and (modern) Norwegian Dahl (2009) explains the use of double definiteness marking in Swedish and Norwegian by the intersection of two waves, where preposed definite articles came from West Ger­manic varieties whereas the enclitic definite articles came from eastern varieties. (Why the definiteness markers became cliticised in Scandinavian still remains an open issue.) In all western Germanic languages and dialects, the definiteness marker in the NP occurs to the left of the noun/NP to be modified, as e. g. in English the red car or in German das rote Auto. Interestingly, most Jutish dialects behave like any West Germanic variety since they also use preposed definite articles æ (rö’) bil ‘the (red) car’ (the indefinite counterpart would have been en [emphasised jen’ ‘one/somebody’]). Whether western Jutish dialects were actually of West Germanic origin before they came under the roof of Danish, a Scandinavian language, is a question that cannot be dealt with here. However, there are several grammatical features in favour of this view – West Jutish as an Ingvaeonic/North Sea Germanic dialect –, such as the loss of (a) morphosyn­tac­tic gender mar­k­ing and (b) unstressed vowels due to apocope, (c) the loss of many in­flexional mark­ers – all cases are simplifications –, (d) the emergence of a semantic gender system, which shows some paral­lels to English; or (e) different sentence intonation patterns which show parallels to West Ger­ma­nic varieties (cf. e. g. Nielsen 1959: 39–49). All these observa­tions seem to support the assumption that Jutish dia­lects could have been of non-Scandinavian origin. The opposite hypothesis would be that these western Jutish dialects became heavily influenced by North Sea Germanic dialects and be­came thus more and more similar to any western type of a Germanic language. (We will come back to this point.) The reverse principle is represented by the Scandinavian languages. In their most archaic Old Norse form, we still find non-preposed definite articles in proper names in the Mainland Scandinavian varie­ties, such as in Swedish Nordiska museet ‘the Nordic museum’, yet quite frequently in Insular Scan­dinavian, such as in Icelandic röði bíllinn ‘the red car [nom.]’. The Swedish/Nor­wegian corresponding NPs would be den röda/røde bilen. In mo­d­ern Norwegian, but not in its more traditional/conservative variety Riksmål, a typologically interesting construction occurs when possessive pronouns are used, as in Bokmål-Norwegian den røde bilen min vs. Swedish min röda bil ∅ ‘my red car’. The Norwegian form is structurally very similar to the corresponding NP in Italian la mia macchina rossa ‘my red car’, where both possessive and de­mon­strative pronouns co-occur, thus emphasising the uniqueness of the object referred to: ‘this my (red) car’. (For more details on the internal structure of the NP in Scandinavian, see the surveys in Braunmüller 2007.)

Competing tendencies in Germanic pronominal and deictic systems 

 19

In contrast to Dahl (2009), I would like to consider this double marking of definiteness a case of simplification that follows the principle: ‘Once a definite noun, always a definite noun’. The contrast to Danish is obvious: den røde bil ∅ ‘the red car’, like e. g. in German das rote Auto on the one hand, and the bilen ‘carthe/the car’ without any attributive adjective on the other (more on this typological vacillation in the next paragraph). ‘Simplification’ is the appropriate term for this development in Swedish and Norwegian because the definiteness marking in the NP is not subject to any change, irrespective of whether or not the noun in the NP contains an adjective. In Danish, the selection of an attribute in the NP triggers, as already mentioned, a reverse order for definiteness marking from postposed definiteness -en like in bilen ‘car-the’/the car’ to preposed definiteness marking den (den røde bil ∅ ‘the red car’). Norwegian which, in its most frequently used variety Bokmål, has according to Trudgill (1992: 21–22) developed as a creoloid, viz. from predominantly written Danish towards the local Norwegian dialects, which show in many instances the same structures as the Swedish dialects.6 But why do we find changing patterns in Danish definite NPs? What is the reason for this typological inconsistency and vacillation in definiteness marking? Why has no levelling taken place like in Swedish or Norwegian? At first glance, this vacillation seems to make defi­nite reference more complicated than necessary. Therefore, at least many German L2-lear­n­ers of Danish prefer NPs like *den bil to bilen ‘the car’. This construction represents an NP that can easily be expanded by an attributive adjective into a more complex NP like den røde bil (like in any other West Germanic language). On the other hand, double definite­ ness marking as de­fault in Swedish is not a desirable construction either, as far as L2-learn­ers are con­cerned. Most German students tend to omit the second definiteness marker in Swedish, which leads to an ungrammatical NP: *den röda bil-∅7. In Norwegian, this could be be accept­able but it may change the style of

6 The reason for this is the common dialectal ground, a continuum, which can be made responsible for the adap­ta­tions towards the local usage on the Scandinavian peninsula. To give some details: Where (written) Da­nish has voiced or deleted obstruents in the interlude and the coda, Norwegian has (now) non-voiced obstruents (cf. Nor­we­gian kake [-k-] < Da­nish kage ‘cake’ [-∅-] or Norwegian ut [-t] < Danish ud ‘out’ [-ð]), or -e as a plural allo­morph was sub­sti­tuted by an -er allomorph (cf. Danish heste > Norwegian hester ‘horses’; Swedish and New Norwegian have hestar) etc. (There are many more ad­just­ments and simplifications.) Even hybrid constructions occur: gratulerer med bursdagen < Danish til lykke med fødselsda­gen < Swedish gratulerar på födelsedagen ‘[all the best on the occa­sion of your birth­day]’. 7 … unless a restrictive relative clause follows immediately: den röda bil ∅ som står där borta ... ‘the red car__ that stands over there’.

20 

 Kurt Braunmüller

the linguistic form (den røde bil sounds more tradition­al, more conser­vative; it stands for a marked variety in modern Norwegian). An explanation, however, is at hand when we have a look at Danish language history, which has heavily been influenced by Low German and later on High German speakers dur­ing the era of the Hanseatic League in the late Middle Ages and in early Modern Times (the age of the Reformation) and beyond. As a matter of fact, bilinguals tend to use isomorphic (here: syntactic) structures wherever possible (cf. e. g. Nettle 1999: 137 and Muysken 2000, ch. 5). For (Low) German L2-speakers of Danish, NPs with a German/West Ger­manic word order are therefore easier to produce than those with the Old Scandinavian order of constitu­ents in the NPs, viz. without any pre­posed definiteness marker: [∅ adjective + noundef.mar­ker]NP, such as in Old Norse stóri maþrinn ‘big man-the’ with a postposed definiteness marker. Bilingual merchants and craftsmen of German descent thus preferred structures that were si­milar or ‘congruently lexicalised’ (Muysken 2000, ch. 5) with the structures in their own language. They therefore dropped the Old Scandinavian enclitic definiteness marking in complex NPs, i. e. NPs containing attributes. Whether they also used this type of construction in NPs without attri­ butive adjectives, however, cannot be tested since we lack sources for oral communication. Additional support for the preference of the ‘German NP type’ in Danish comes from western Danish, especially from West Jutish dialects, which do not use the ‘Scandinavian NP type’ at all (as already mentioned). Thus, in the contact zone on the Jutish peninsula the stru­tures of definite NPs were identical and had a clear preference for the German/West Germanic type of NP. Whether the enclictic article in non-Jutish varieties of Danish was dismissed as irrelevant for communication or whether the German L2-speakers sooner or later learned to use it, cannot be decided due to the lack of evidence from (informal) sources outside Jutland. As far as Jutland is concerned, we only have two (important) law texts, the Jutish Law (1241) and the Town charter of Flensburg (1284), both showing West Germanic definiteness marking – if overt definiteness markings occur at all.8

8 The Jutish Law has in most cases no definite articles at all (cf. the famous first sentence in the preface: MEth logh schal land byggæs. ‘The country shall be built by __ law.’; vs. later: Thæn logh thær koning gyuær … ‘That law the king gives …’). Whether this fact might be considered an archaism or is due to Latin as a mo­del for law texts in general does not play a role for our argumentation. If definite articles occur, they only appear as pre-posed ar­ticles or rather as (simple) de­monstrative pronouns. The Town charter of Flensburg (Danish version) shows the same picture: tha mu­ghæ the frændær ei takæ the børn ‘then the relatives must not take the children’. In Modern Standard Danish it would be frændene ‘the rela­tives’ and bør­nene ‘the children’, respectively.

Competing tendencies in Germanic pronominal and deictic systems 

 21

4.3 The retention of the t-article in Modern Faroese There are two inherited definite articles in Modern Faroese, (a) the hin, hitt-article ‘the [nom.sg.masc./fem.; -.neutr.]’ that is also to be found in Modern Icelandic (hinn, hin, hið ‘the [nom. sg.masc.; -.fem.; -neutr.]’), and (b) the tann, tað-article which can be traced back to Old Norse sá, sú, þat (in the nominative singular) and þeir, þær, þau (in the nomina­tive plural, and with the same meaning as in Icelandic, which are not used as definite articles, but still as demon­strative pronouns). Both articles are preposed and occur only in NPs with attributes, mostly preposed adjectives. The existence of two definite articles is, how­ever, quite rare in Germanic languages. As far as I know, they occur only (a) in Faroese, (b) in North Frisian, in the western Fering dialect spoken on the island of Föhr (see Ebert 1971), and (c) in the dialect of Amern, today part of the community of Schwalmtal and possibly in the surrounding Rhine dialects as well (Amern is situated in the Low Rhine area; see Heinrichs 1954). But there might be other further southern German dialects as well. The reason for the retention of two formally different ways of definiteness marking, which express minor differences in meaning and reference, not to be discussed here, lies in the very strong Faroese-Da­n­ish bilingualism, which has now lasted for far more than half a millenni­um. Every native spea­ker of Faroese learns Danish at a very early age, in many cases as soon as they attend kindergarten or watch (Danish dominated) television. Later on, Danish becomes an obligatory subject in school from the third grade onwards, where a very high, nativelike proficiency of the other national language Danish is considered mandatory. We therefore have to treat all native speakers of Faroese as very early sequential bilinguals. And as a consequence of this fact, there are no monolingual speakers of Faroese on these islands.9 In the two official languages used on the Faroe Islands, we find both the West Nordic h-and t-arti­cles as definiteness markers (in the Faroese language) on the one hand, and the East Nordic d-article as the default definiteness marker (in the Danish language). In addition, an inflected enclitic definite article occurs in Faroese (-in, -ið in the nominative singular: masc./fem.; -neutr.) like in (almost) all Nordic/Scandinavian languages, where we have -en, -et as equivalent expressions. All these inflexional forms have been derived from the h-articles but have lost their consonantal onsets (h-) due to weak stress, clitisation and finally uni­ verbation. The question is why the t-article has survived in Modern Faroese even

9 More details about the consequences of the strong Faroese-Dan­ish bilingualism can be retrieved from Petersen 2010.

22 

 Kurt Braunmüller

though it is not used in this function in the genetically very closely related Icelandic language. Both forms have undoubtedly the same origin as (simple) demonstrative pronouns, as mentioned above. I advance the view that the most plausible answer to this question is very simple and has much to do with the ubiquitous bilingualism on the Faroe Islands and the frequent code switching between the two national languages on the Faroe Islands, Faroese and Danish. The Faroese t-article and its inflected forms (tann, tað …) have their morphophonologically perspicuous counterparts in the Danish d-articles (den, det; de, dem), which did not only keep the Faroese t-article in function but which has paved its way towards functioning as a definite article – in clear opposition to the development in Icelandic, where the corresponding þ-demonstrative pronouns have ne­ver been used as simple definite articles. Both forms in Faroese and Danish have similar dentals in the onset (t- and d-). In other words, interlingual diamorphs with the same grammatical function support not only frequent code switching but gave also rise to the re-categorisation of a demonstrative pronoun becoming a definite article and, later, its survival as a second definiteness marker. But the Faroese h-article could not be ignored or abolished because of the retention of the t-article since it has turned out to be necessary for the internal West Nordic communication between Icelanders and the inhabitants of the Faroes. In these two languages, the h-articles are almost the same in form and function (cf. Faroese hin, hitt … and Icelandic hinn, hin, hið ...). The conclusion to be drawn from this development is a paradox: two different forms of marking definiteness make life with two different (genetically more or less closely rela­ted) lan­guages easier. Seen from the perspective of bilinguals that use both varieties side by side in their daily lives, this greater complexity in referential definiteness in the NP actually turns out to be a simplification because it establishes more parallel structures between the two languages involved.

 he loss of Old Norse hinn, hin, hit as definite articles 4.4 T in mainland Scandinavian As already mentioned (in section 2), there were originally two morphological candidates for recategorisation as definite articles, the (simple) demonstrative pronouns on h- (< IE *kw-) and þ- (< IE *te-/to). But why did the h-article disappear in the Mainland Scandinavian languages and why did the þ-/d-article prevail in these languages? Again, the answer is to be found in the long lasting era of language contact between the Scandinavians in Denmark, Sweden and Norway and the speakers of Low German in the late Middle Ages and beyond. L2-learners of Old Swedish

Competing tendencies in Germanic pronominal and deictic systems 

 23

and Middle Danish/Norwegian who had Low German as their first language preferred those definite article forms with the most isomorphic relationship to their L1, which is the þ-/d-article, since the definite article in Low German has d- in the onset (de man, de fru … ‘the man, the woman’). Due to incomplete L2-acquisition they simplified the fricative [ð-] in the Scandinavian þ-article (and beyond; see fn. 1) and replaced it with the plosive [d-], which is easier to pronounce and almost homophonic to the definite article in their mother tongue, Low German. A related kind of phonetic simplification is well-known from overseas varieties of English, e. g. English in India, where the [here: unvoiced] fri­cative [θ-] is often rendered as a voiced [d-]; thirty and dirty became thus identical in pronunciation: [d-]. The h-article had its (stylistic) domain in the written language and did, therefore, not occur very fre­quently in oral communication and, thus, did not become part of the (simplified) L2-language system that was acquired by the Hanseatic merchants and craftsmen. Later on, the mixed population ignored the h-article in oral communication more and more, with the re­sult that this article either disappeared in colloquial Scandinavian or became a highly mar­ked va­riety. One may describe this kind of change – paradoxically – as ‘language change without language change’,10 which means that only the usage, the linguistic norm has changed and not the whole language system as such. The h-articles have survived in modern Main­land Scandinavian varieties but they occur rarely and only in highly marked structures. Again, language contact and different means to express more or less the same grammatical function gives way for a smooth change in language: This kind of change demands least effort and can occur anywhere without being noticed as language change at once. Only in the long run does one see that the marking of certain grammatical structures has changed.

5 Summary and conclusions It has been shown that the category ‘pronouns’ have unluckily been lumped together into only one (heterogeneous) grammatical category, which obscures the divergent syntactic and refe­rential functions of the morphological entities. This form of classification has led to wrong conclusions about the nature and the function of ‘pronouns’ in general. Some of them, the so-called pronouns of the 1st and

10 Cf. Bernd Heine’s dictum (p. c.): ‘word order change without word order change’, which means that the new word order already was part of the older system but became now the default or dominant or­der, whereas the old word order faded out or became highly marked.

24 

 Kurt Braunmüller

2nd person, have no co-referential functions at all and can thus not be considered ‘pro’-nouns or pro-NPs (in the sense of: aliquid stat pro aliquo). Anaphoric or 3rd person pronouns can easily be replaced by demonstrative pronouns. They have basic referential functions in common. The (occasional) abolishment of the 3rd per­son pronouns therefore leads to grammatical simplification and to more referential tran­spa­ren­cy, which seems to play an important role in language contact and language change, espe­cial­ly when second language learn­ing is involved. Splitting up pronouns into (a) direct or exophoric pronouns and (b) discourse deictic or anaphoric pronouns has to be regarded as a rather artificial grammatical classification, which is not shared by all speakers. Schooling and adherence to (grammatical) traditions ob­vi­ous­ly play an im­portant role. Modern linguistic analyses have, however, not yet succeeded in re­pla­cing this an­cient classification by a more functionally based one. In many cases, (di­rect/ exophoric) 3rd per­son pronouns are considered redundant. They compete with demonstrative pronouns and other deictic elements, which may give rise to the disappearance of one of these forms due to their equiva­lent functions. In most conflicting cases, the 3rd person pronouns will be neglected and downgraded in their default value. Therefore, we should not consider this as some sort of beneficial suppletion, as far as the emergence of the modern English personal prono­minal system is concerned, but rather denote it as a case of simplification or unification, caused by adult L2-learners. Seen from this perspec­tive, the change in the Old English prono­minal system can be regarded as a ‘natural’ linguistic development, where two competing forms with equivalent gram­matical/referential functions became simplified, underlining our hypo­thesis which says that “the most general principle [of reference] will prevail”. Simplification due to interparadigmatic alignment (cf. “once a definite noun, always a de­finite noun”) could be demonstrated as the leading principle in modern Swedish and Nor­wegian, where the vacillation in marking the noun in an NP with a definiteness marker has been abandoned. This principle leads to language-internal simplification. Language-external simplification due to language contact has been demonstrated in the case of the Danish NP. The result of intense contact between Low German and Jutish speakers on the one hand and (insular) Danish speakers on the other was to achieve isomorphism in the structure of complex NPs (sc. NPs containing adjectives). Moreover, L2-lear­ners of Danish obviously followed the principle which says that “one definiteness marker in an NP is sufficient”. This fact has been corroborated by the frequent mistakes of (German) L2-lear­ners when learning Swedish or Norwe­gian. Retaining two morphologically equivalent forms in a language may make sense when these two forms help to bridge the gap in (frequent) code-switch-

Competing tendencies in Germanic pronominal and deictic systems 

 25

ing between genetically close­ly related varieties (here: Faroese and Danish). One definite article, the t-article, gets sup­ported by the daily use of the Danish language which shows an almost homophonous definiteness marker in the NP with a d- in the onset. The other definite article, the h-article, did not become superfluous because it continues to guarantee the unimpeded communication with the closest sister language to Faroese, Icelandic.11 Bilateral language con­tact preserves two pro­nouns that are identical in function, but both forms are needed because they facilitate the ancient link to Icelandic as well as the traditional political link to Denmark and the Danish language and cul­ture. To cut a long story short, traditional grammatical classifications can make the descrip­tion of a language change more complicated than necessary. The heterogeneous classifications found in historical grammars of Germanic (and its dialects) are not very helpful in under­standing the functions of pronouns or of the interrelations to other equivalent parts-of-speech, such as demonstrative pronouns and deictic expressions in general. A cross-linguistic perspective supports the critique presented here: Many languages have no special 3rd person pro­nouns (Dixon 2010, II: 189 mentions for example Korean) or only pronouns that are re­lated to all or [± remote] demonstratives (for more details see The World Atlas of Language structures Online, http://wals.info/feature/43A). They use deictic constructions or compounds instead. This fact shows that 3rd person pronouns are not essential for grammatical re­fe­rence to persons, objects or for textual (cata- or anaphoric) reference. If both categories, the 3rd person pro­nouns and the demonstrative pronouns, co-occur side by side, the demonstrative pronouns are the more general and probably the more universal ones, and can therefore not be disregarded. On the other hand, the use of two equivalent forms in similar contexts may give way to gram­matical restructurings, as we have seen in the case of the realignment of the Old English pronominal system. But duplicity in form and function may open new perspectives in language contact situations, where bilingual speakers can sim­plify their two grammatical systems in selec­ting those forms which are closer to their other language or which shows more formal or overt cor­respon­dence with their other language. The result, again, is sim­pli­fi­cation and more transparency in the referential systems. As a rule of thumb, L2-learners can be used as the best ‘reagents’ for what categories are in fact grammatically redundant. These categories or forms may

11 Some inhabitants of the Faroe Islands see themselves as some sort of ‘younger’ brothers and sisters of their neighbours in the North. Some of them enjoy e. g. reading Old Nordic/Icelandic literature. Other ones, though a minority, try to follow the puristic tendencies when looking for an equivalent, more ‘genuine’ Western Nordic term in their own language.

26 

 Kurt Braunmüller

sooner or later be ignored and will, in the long run, become marked or considered more or less redundant. But L2-bilinguals will preserve all structures which are relevant for communication.12 One can use their way of acquiring other languages as a kind of a litmus test of what is mandatory to acquire be­cause it is essential for communication, and what may obviously be considered redun­dant, for whatever reasons, and can thus, in principle, be neglected. Moreover, it might become a fruitful new per­spective for further grammatical research to re­consider grammatical terminology, such as the defini­tion(s) of ‘pronouns‘ (see section 1), not only from a (monolingual) na­tive speaker but also from a (bilingual) learner’s perspec­tive. Which gram­ma­tical forms and categories are, at least in some aspects, functionally equivalent, which have overlap­ping func­ tions, depending e. g. on the speaker‘s/observer‘s perspec­tive or on information structure, and which have to be kept strictly apart because they repre­sent absolutely distinct categories and functions? Additionally, this idea may lead to a rethinking of grammar from an L2-perspective, which is no longer based on a fossilised view of grammar that is based on the traditional school lan­guages (Classical) Latin and Greek, but which has never been challenged. With his view in mind, there might be a chance to gain deeper insights into the system of languages when both the descrip­tive (typological) principles of grammatical codification and the foundations of the learner’s – and more generally speaking the bi-/multi­lingual’s – knowledge of grammar are brought together.

References Bergman, Gösta (1972): Kortfattad svensk språkhistoria. En översikt över det svenska språ­kets utveckling från de äldsta nordiska runinskrifterna fram till vår egen tid. Stockholm: Prisma. 3rd edition. Braunmüller, Kurt (1977): Referenz und Pronominalisierung. Zu den Deiktika und Proformen des Deutschen. Tübingen: Niemeyer. Braunmüller, Kurt (2002): 115. Language Contacts During the Old Nordic Period I: With the British Isles, Frisia and the Hanseatic League. In: Oskar Bandle et al. (eds.), The Nordic languages. An International Handbook of the History of the North Germanic Languages, 1028–1039. Berlin and New York: De Gruyter. Braunmüller, Kurt (2007): Die skandinavischen Sprachen im Überblick. Tübingen and Basel: Francke. 3rd edition.

12 McWhorter (2007: 10) would say that “[s]implification occurs, but is counterbalanced by new emergen­ces”, but this does not happen necessarily, as demonstrated here.

Competing tendencies in Germanic pronominal and deictic systems 

 27

Braunmüller, Kurt (2008): Observations on the Origins of Definiteness in Ancient Germanic. Sprachwissenschaft 33: 351–371. Brunner, Karl (1965): Altenglische Grammatik nach der angelsächsischen Grammatik von Eduard Sievers. Tübingen: Niemeyer. 3rd edition. Dahl, Östen (2009): Increases in Complexity as a Result of Language Contact. In: Kurt Braunmüller and Juliane House (eds.), Convergence and Divergence in Language Contact Situations, 41–52. Amsterdam and Philadelphia: John Benjamins . Diessel, Holger (1999): Demonstratives: Form, Function, and Grammaticalization. Amsterdam and Philadelphia: John Benjamins. Dixon, R. M. W. (2010): Basic Linguistic Theory, vol. II: Grammatical Topics. Oxford and New York: Oxford University Press. Ebert, Karen (1971): Referenz, Sprechsituation und die bestimmten Artikel in einem nordfriesischen Dialekt (Fering). Bredstedt: Nordfriisk Instituut. Heine, Bernd and Tania Kuteva (2005): Language Contact and Grammatical Change. Cambrigde and New York: Cambridge University Press. Heinrichs, Heinrich Matthias (1954): Studien zum bestimmten Artikel in den germanischen Sprachen. Gießen: Schmitz. Hirt, Hermann (1932): Handbuch des Urgermanischen. Volume 2: Stammbildungs- und Flexionslehre. Heidelberg: Winter. Keller, Rudi (1990): Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. Krahe, Hans and Wolfgang Meid (1969): Germanische Sprachwissenschaft. Volume 2: Formenlehre. Berlin: De Gruyter. 6th edition. Law, Danny (2009): Pronominal Borrowing Among the Maya. Diachronica 26: 214–252. McWhorter, John (2007): Language Interrupted. Signs of Non-Native Acquisition in Standard Language Grammars. Oxford and New York: Oxford University Press. Muysken, Pieter (2000): Bilingual Speech. A Typology of Code-Mixing. Cambridge and New York: Cambridge University Press. Nettle, Daniel (1999): Linguistic Diversity. Oxford and New York: Oxford University Press. Nielsen, Niels Åge (1959): De jyske dialekter. Copenhagen: Gyldendal. Petersen, Hjalmar P. (2010): The Dynamics of Faroese-Danish Language Contact. Heidelberg: Winter. Pilch, Herbert (1970): Altenglische Grammatik. Dialektologie, Phonologie, Morphologie, Syntax. München: Hueber. Prokosch, E[duard] (1939): A Comparative Germanic Grammar. Baltimore: University of Pennsylvania. Trudgill, Peter (1992): Introducing Language and Society. London: Penguin. Werner, Otmar (1991): The Incorporation of Old Norse Pronouns into Middle English. Suppletion by Loan. In: P. Sture Ureland and George Broderick (eds.), Language Contact in the British Isles. Proceedings of the Eighth International Symposium on Language Con­tact in Europe, Douglas, Isle of Man, 1988, 369–401. Tübingen: Niemeyer.

Pavel Iosad

“Pitch accent” and prosodic structure in Scottish Gaelic: Reassessing the role of contact1 Abstract: This paper considers the origin of “pitch accents” in Scottish Gaelic with a view to evaluating the hypothesis that this feature was borrowed from North Germanic varieties spoken by Norse settlers in medieval Scotland. It is shown that the “pitch accent” system in Gaelic is tightly bound with prosodic structure (more precisely syllable count), certainly diachronically, and probably (at least in some varieties) synchronically. Gaelic “pitch accent” is argued to be a plausible internal development, parallel to similar phenomena in other branches of Celtic (specifically in Breton), as well as in Germanic. This conclusion may appear to undermine the contact hypothesis, especially in the absence of reliable written sources; nevertheless, a certain role for Norse-Gaelic contact in the appearance of the pitch accent system cannot be completely excluded.

1 Introduction Several varieties of Scottish Gaelic have attracted interest in both descriptive and theoretical literature due to an interesting feature of their prosody, which involves contrasting laryngeal activity (pitch and glottalization) in segmentally identical forms. As an example, consider the difference between [ˈtuan] ‘hook’ (written dubhan) and [ˈtuan] ‘song’ (duan). As documented by Ladefoged et al. (1998), the first of these is realized in the dialect of Bernera (off the coast of Lewis in the Outer Hebrides) with a rising-falling contour, while the second one has a rising

1 A version of this paper was presented at the 11th International Conference of Nordic and General Linguistics at the University of Freiburg. I would like to thank the audience in Freiburg, in particular Kristján Árnason, Hans Basbøll, Aðalsteinn Hákonarson, Michael Rießler, Michael Schäfer, Sarah Thomason, and Allison Wetterlin, for questions and discussion. The paper has also benefited from comments, suggestions, and assistance by Ricardo Bermúdez-Otero, Björn Köhnlein, Iain MacPherson, Bruce Morén-Duolljá, and Haukur Þorgeirsson. Thanks also to two anonymous reviewers for incisive comments which have greatly improved both content and presentation of the paper. All errors and shortcomings remain entirely mine.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 29

contour: an alternative way of describing the difference is analysing [ˈtuan] ‘hook’ as having an early pitch peak and [ˈtuan] ‘song’ as having a late one. An essentially identical contrast is found between words with and without a certain type of epenthetic vowel, traditionally called the svarabhakti vowel (underlined for clarity in the remainder for this paper). For instance, in the dialect of Bernera, the word [ˈpalˠak] ‘skull’ (ballag) has a high tone on the first vowel, comparable to the early peak of [ˈtuan] ‘hook’, while [ˈpalˠak] ‘belly’ (balg) has a rising pattern throughout both vowels, similar to the late peak in [ˈtuan] ‘song’. Descriptively, this contrast is extremely similar to “pitch accent” contrasts found in the North Germanic varieties of Sweden and Norway (e. g. Bruce 1977; Riad 1992; Riad 1998; Lorentz 1984; Kristoffersen 2000; Lahiri et al. 2005) and in the West Germanic varieties of Limburg and the Franconian area (Gussenhoven and van der Vliet 1999; Liberman 2000; Gussenhoven 2000; Gussenhoven and Peters 2004; Peters 2008; Köhnlein 2011). It has occasionally been treated in the literature as such, for instance by MacAulay (1993) and Ternes (2006). In the remainder of this paper, I will indicate early-peak words using the symbol ¹, and late-peak words using the symbol 2, recalling the tradition of “accent 1” and “accent 2” found in studies of the Germanic languages (see also Ternes 2006 for this convention). Historically, the contrast is reconstructed as one of monosyllabic vs. disyllabic forms: [1ˈtuan] ‘hook’ corresponds to Old Irish dubán, while [2ˈtuan] ‘song’ goes back to monosyllabic duan with a diphthong. Similarly, in pairs such as [1ˈpalˠak] ‘skull’ and [2ˈpalˠak] ‘belly’ the late-peak words are descended from monosyllables, with the second vowel being epenthetic (Old Irish bolg ‘sack’).2 The parallel is strengthened further by the existence of varieties where the same contrast is reflected in glottalization rather than pitch contours. In North Germanic, the prime example of this is Danish stød, while in southern dialects of Scottish Gaelic (Holmer 1938; Ternes 1980) we find pairs such as [ˈpoʔɔ] ‘underwater rock’ (from Norse boði) vs. [ˈpoː] ‘cow’ (Old Irish bó), which correspond to Lewis [1ˈpoː] and [2ˈpoː] respectively. From a historical perspective, some authors, notably Borgstrøm (1974), have suggested that the development of the “pitch accent” system in Scottish Gaelic, along with other phonetic and lexical features of the language (e. g. Marstrander 1932; Hansson 2001; Stewart 2004), is a contact phenomenon due to language shift from Norse to Gaelic that was assumed to occur as the Norse-occupied areas of the Highlands and the Isles became part of the kingdom of Scotland. Others,

2 The literature on this epenthesis is extensive; some examples are Ó Baoill 1980; Clements 1986; Carnie 1994; Ní Chiosáin 1991, 1999; Hind 1996; Bosch and de Jong 1997; Smith 1999.

30 

 Pavel Iosad

such as Ternes (1980, 2006) and Eliasson (2000), have taken a more cautious line, preferring to see the commonalities in vaguer “areal” terms, with occasional references to the possibility of a common substrate (Wagner 1964). In this paper I explore the possibility of analysing the “pitch accent” as an independently occurring phenomenon. I review the phonological evidence that supports analysing the contrast as one of the number of syllables (Oftedal 1956; Ladefoged 2003) and argue that the “pitch accent” found in Scottish Gaelic requires the confluence of two factors: contrastive prosodic (including syllabic) structure in the surface phonology and a phonological computation that may associate tones and laryngeal features with boundaries and heads of certain prosodic constituents (including syllables and morae). I show that both of these can arise without significant external influence, focusing in particular on an example from a Breton dialect, where a “pitch accent” system appears to have arisen through the lexicalization of prosodic structure and associated tones, in a development parallel to that hypothesized by Riad (1998) for North Germanic but not clearly connected to language contact. I conclude by reflecting on the consequences of this approach for the bigger picture of Norse – Gaelic contact in early medieval Scotland.

2 The syllabic analysis in Scottish Gaelic In the remainder of this paper I adopt the hypothesis that the contrast in “pitch accents” in at least some varieties of Scottish Gaelic is a function of their surface prosodic structure. In this section I review the evidence for this analysis. A minor point of terminology is in order: in the remainder of this paper I will refer to rising-pitch words such as [2ˈpalˠak] ‘belly’ (balg) as “monosyllabic”, and to early-peak words such as [1ˈpalˠak] ‘skull’ as “disyllabic”.

2.1 Pitch contours One type of evidence involves the pitch contours themselves, at least in certain dialects such as the Hebridean varieties studied by Oftedal (1956) and Ladefoged et al. (1998) As noted above, monosyllabic words receive a rising pitch, while disyllabic ones have an early pitch peak followed by a long fall. A very plausible analysis for these facts is suggested by Ladefoged (2003), who proposes that stress in Scottish Gaelic is by default associated with a disyllabic LH L contour. In monosyllabic words, including both straightforwardly monosyllabic ones such as [2ˈpoː] ‘cow’ and more complex cases with epenthetic vowels of the type [2ˈpalˠak]

“Pitch accent” and prosodic structure in Scottish Gaelic 

 31

‘belly’, the second L tone has no syllable to associate with, and is therefore absent from the surface representation. This results in the rising pitch contour. Conversely, in disyllabic words – both those that are straightforwardly so, such as [2ˈpalˠak] ‘skull’, and cases such as [2ˈpoː] ‘underwater rock’ that require some device to distinguish them from monosyllabic items with identical segmental content – the entire LH L contour surfaces as expected.

2.2 Rhyme palatalization and the analysis of epenthesis An important piece of evidence concerns a palatalization process which is used as the exponent of certain grammatical categories, such as the genitive singular in some declension classes. It involves the palatalization of a final consonant or consonant cluster and fronting and/or raising of the preceding (short) vowel; however, normally it does not affect consonants preceding the vowel involved. This can be seen in the following examples from the Lewis dialect (Oftedal 1956); here and elsewhere in this paper I write the affected segments in bold face: (1) a. Long vowel unaffected: (i) [ˈɔːr] òr ‘gold’ (ii) [ˈɔːðʲ] òir ‘gold (gen. sg.)’

b. Palatalization stops at short vowels: (i) [ˈslˠɔʰk] sloc ‘pit’ (ii) [ˈslˠuʰkʲ] sloic ‘pit (gen. sg.)’ (iii) *[ʃluʰkʲ]

Crucial examples, as pointed out by Smith (1999), among others, involve the contrast between true disyllabic words and monosyllabic words with epenthesis. The following examples, also from the dialect of Leurbost, demonstrate this clearly. (2) a. Disyllabic words (i) [1ˈpalˠəx] balach ‘boy’ (ii) [1ˈpalˠiç] balaich ‘boy (gen. sg.)’ (iii) *[1ˈpilʲiç]

b. Monosyllabic words with epenthesis (i) [2ˈpalˠak] balg ‘bellows’ (ii) [2ˈpulukʲ] builg ‘bellows (gen. sg.)’ (iii) *[2ˈpalˠukʲ]

The treatment of these facts in the literature has often been connected with the contention by Borgstrøm (1940) that speakers report a difference in syllabification between mono- and disyllabic words of the type balg/ballag. According

32 

 Pavel Iosad

to Borgstrøm (1940), speakers show the expected V.CV syllabification in disyllabic words such as [1ˈpalˠak] ‘skull’ (ballag) but the unexpected VC.V boundary in words like [2ˈpalˠak] ‘bellows’ (balg). However, there are good reasons to believe that the placement of the syllable boundary is not the relevant phonological difference. A cautionary tale in this respect is provided by Ní Chiosáin et al. (2012), who show that Irish speakers report VC.V syllabifications in contexts where such a placement of syllable boundaries does not correspond to clear differences in “core” phonological properties such as alternation-related behaviour. This suggests (if more evidence were needed) that speaker judgements do not necessarily correspond to meaningful phonological distinctions. More pertinently, Hind (1996) and Hall (2006) have separately argued that the “epenthesis” process which appears to result in the insertion of a vowel between a sonorant and a following segment is different in kind from epenthesis understood as the insertion of a root node; see also work by Levin (1987). They argue that the epenthesis (at least in those dialects where it is described as involving a full copy of the pre-sonorant vowel) represents an obligatory realignment of the vocalic gesture extending beyond the sonorant (for another recent discussion of the theoretical importance of dissociating vocalic and consonantal gestures, see the work by Operstein 2010). Crucially, this account does not require the insertion of a second root node by the phonological computation.3 From a phonological perspective, the most parsimonious analysis of these facts involves viewing monosyllabic words such as [2ˈpalˠak] ‘belly’ as containing a single vowel in the surface representation: in other words, for the purposes of phonological computation the difference between the words for ‘belly’ and ‘skull’ is represented as the difference between [ˈpalˠk] and [ˈpalˠak]. This analysis immediately provides an elegant way of unifying the behaviour of disyllables and monosyllables for the purposes of rhyme palatalization. Specifically, we can analyse rhyme palatalization as involving the nucleus of the final syllable and any following consonants. This is very clear in the case of disyllabic [1ˈpalˠɪç] from [1ˈpalˠəx] ‘boy’. The behaviour of monosyllabic [2ˈpulʲukʲ] ‘bellows (gen. sg.)’ is explainable if we treat it as phonological [2ˈpulʲkʲ], from non-palatalized [ˈpalˠk].

3 An anonymous reviewer points out that this proposal is hardly unprecedented: many languages with complex consonant clusters break these up using schwa-like vocalic segments which do not behave like phonological vowels; examples are Imdlawn Tashlhiyt Berber (Dell and Elmedlaoui 1985; Ridouane 2008) and (for some speakers) Syrian Arabic (Gouskova and Hall 2009; Hall 2013).

“Pitch accent” and prosodic structure in Scottish Gaelic 

 33

2.3 Syncope Another piece of evidence offered by Smith (1999) concerns the behaviour of epenthetic vowels in syncope. According to Smith (1999), syncope affects medial syllables in order to optimize prosodic structure in terms of lapses, resulting in alternations such as the following: (3) a. [ˈopəðʲ] b. [ˈoprəx] c. *[ˈopəðʲəx]

obair oibreach

‘work’ ‘work (gen. sg.)’

However, this syncope does not apply to vowels resulting from epenthesis: (4) a. [1ˈpalˠəx] b. [2ˈvalˠaxu] c. *[2ˈvalˠxu]

balach ‘boy’ a bhalachaibh ‘boy (voc. pl.)’

In the analysis offered by Smith (1999), epenthetic vowels in words such as [2ˈvalaxu] are root nodes inserted by the phonological component. These root nodes exhibit special behaviour, in that they remain invisible to top-down prosodic rules regulating syncope. Smith (1999) suggests that they remain affiliated to the initial syllable by proposing an elaborate syllable-internal structure, whereby the svarabhakti vowel is seen as projecting a non-maximal syllable contained inside the initial one ([va[la]σ]σ[xu]σ). Under the assumptions outlined in section 2, this elaborate structure is not needed, because the svarabhakti vowel is simply not manipulated by the phonology: the surface representation of example (4-b) is [valxu]. This allows us to express the generalization that the vowels are invisible to the prosodic phonology, since they are simply absent from the surface-phonological representation.4

4 Note, however, this particular analysis is only applicable to dialects such as that of Leurbost, where the svarabhakti vowels are complete copies of preceding nuclei or are simply exponents of the same vocalic gesture (Hind 1996; Hall 2006). Other varieties, such as Barra Gaelic (Borgstrøm 1937; Clements 1986; Nevins 2010) are described as exhibiting a mismatch in quality between the underlying and the svarabhakti vowel whenever the sonorant is palatalized: the genitive singular of [2ˈpalˠak] ‘bellows’ is said to be [2ˈpulʲikʲ] (Borgstrøm 1937: § 263). If this vowel is indeed identical to lexical [i], it is not impossible that something like the analysis offered by Smith (1999) is required for these varieties.

34 

 Pavel Iosad

2.4 Glottal stop insertion The data discussed in sections 2.1–2.2 do not provide conclusive evidence for a contrast in syllable affiliation, but rather only in syllable count (cf. Vaux 2003): the crucial differences are between mono- and polysyllabic words, not between V.CV and VC.V syllabifications. In this section I leverage data from southern varieties of Scottish Gaelic to argue that the interaction of glottal stop insertion with other processes active in Scottish Gaelic phonology provides the best evidence for viewing the contrast as one of syllable structure.5 As noted above, southern varieties of Scottish Gaelic, such as the dialects of Islay (Holmer 1938) and Tiree (Ternes 1980) show the insertion of a postvocalic glottal stop where Hebridean varieties have “accent 1”, which I have argued to reflect disyllabicity. Smith (1999), building on proposals by Clements (1986), suggests that insertion of [ʔ] is used in Islay Gaelic to achieve bimoraicity of stressed syllables when no consonant is available to serve as a (moraic) coda. This is seen in the following examples (numbers refer to pages in Holmer 1938): (5) a.

No glottal stop insertion in heavy syllables (i) [ˈtʰrɑμiμ] tràigh ‘shore’ (227) (ii) [ˈkʰlʲuːμμ] cliù ‘fame’  (46) (iii) [ˈpjɔːμμ] beò ‘alive’ (127)



b.

Glottal stop insertion is subminimal monosyllables (i) [ˈtʰʲeμʔμ] teth ‘hot’ (221) (ii) [ˈmɛμʔμ] math ‘good’ (188) (iii) [ˈkruμʔμ] gruth ‘curds’ (177)



c.

Glottal stop insertion in polysyllables (i) [ˈpɑμʔμlɑx] balach ‘boy’ (125) (ii) *[ˈpɑμlɑx] (iii) [ˈkoμʔμur] gobhar ‘goat’ (176)

Crucially, glottal stop insertion is not found before sonorants when these are followed by svarabhakti vowels: (6) a. [ˈmarəv] b. *[ˈmaʔrəv]

marbh

‘dead’ (189)

This is explained if the correct surface representation in (6) is [ˈmaμrμv], with a moraic coda consonant obviating the need for glottal stop insertion. Thus, glottal

5 Thanks to an anonymous reviewer for raising many of the concerns discussed in this section.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 35

stop insertion can be viewed as a device to provide a light stressed syllable with a mora (Smith 1999). If this account is correct, the minimal pair [ˈpoʔɔ] ‘underwater rock’ vs. [ˈpoː] ‘cow’, given by Ternes (1980) for Tiree Gaelic, can be explained if we assume the former to be stored as lexically disyllabic (/(po)σ(o)σ/), the same solution that is applicable to Hebridean [1ˈpoː] ‘underwater rock’ vs. [2ˈpoː] ‘cow’. Alternative analyses of these data are possible. In particular, the distribution of the glottal stop in examples (5-c-i) to (5-c-iii) could be static (i. e. lexically determined), while in example (5-c-iii) the insertion of the glottal stop could be viewed as breaking the hiatus and thus as involving an onset [ʔ]. However, it can be demonstrated that the glottal stop is both derived in the phonology and affiliated to the preceding syllable.

2.4.1 Glottal stop insertion is phonological To demonstrate that glottal stop insertion is phonological, we turn to alternations involving closed and open syllables: we expect that contexts in which the stressed syllable is closed do not involve glottal stop insertion, since the coda consonant can project the second mora, while an open syllable should be associated with the epenthetic glottal stop. The literature provides three sources for such alternations: inflection, syncope, and resyllabification. In stems of the form CVC, the first syllable is closed when no suffix follows (a relatively frequent occurrence, see e. g. Adger 2010) but open when followed by a vowel-initial suffix. It appears that glottal stop insertion operates in line with expectations in this case in Islay Gaelic (the pronouns appear to be clitics which do not influence syllabification). The examples are from Holmer (1938: § 95). (7) a. Open syllables, glottal stop inserted (i) [ˈkʰuμʔμ.riç mi] cuiridh mi (ii) [ˈxuμʔμ.rə tu] chuireadh thu

b. Closed syllables, no glottal stop (i) [ˈxuμrμ mi] chuir mi (ii) [ˈxuμrμ u] chuir thu

‘I will put’ ‘you would put’

‘I put (past)’ ‘you put (past)’

Another process creating these alternations is the syncope referred to above in example (2), as seen in the following example (Holmer 1938: 156):6

6 Some of these examples are also discussed by Smith (1999), who also adduces cases such as [ˈtʰɯʔrəm] ‘dry’ (tioram), [nəs-ˈtʰɯrəma] ‘drier’ (nios tiorma), which are consistent with the approach that sees glottal stop insertion as driven by minimality, but the argument hinges on the

36 

 Pavel Iosad

(8) a. (i) [ˈtɔμʔμ.rəs] (ii) [ˈtɔμrμ.ʃən]

dorus doirsean

‘door’ ‘doors’

b. (i) [ˈpɑμʔμ.lʲə] (ii) [ˈpɑμlμ.tʲən]

baile bailtean

‘place’ ‘places’

Finally, relevant data from another southern dialect (that of Jura) are provided by Jones (2000). In that variety, we find glottal stops in open syllables, as in [ˈpɑʔlˠɑx] ‘boy’, [rɑʔtan] ‘rat’, [ˈtʰɑʔpɑtʲ] ‘quarrel’. However, we also find that a glottal stop can be inserted in what is lexically a VC-final stem in a phrasal context: (9) a. b. c.

[ɣɛʔn ɑ] dh’fhan e [stɑʔt əŋ kʰɑːr] stad an càr [koʔp ənʲ ɛːn] gob an eun

‘he stayed’ ‘stop the car’ ‘the bird’s beak’

When these words appear in a non-prevocalic context, the glottal stop is not obligatory: (10) a.

[fɛn lɛm]

fan leam

‘stay with me’

This suggests that glottal stop insertion in contexts such as those in example (9) is driven by the postlexical resyllabification of the word-final consonant as an onset before the following vowel, which leaves the stressed syllable light. This further confirms that the glottal stop insertion is a phonological process driven by the stress-to-weight principle (e. g. Prince 1990; McGarrity 2003; Bye and de Lacy 2008).7

2.4.2 The glottal stop is a coda Having established that glottal stop insertion in Scottish Gaelic is a phonological process, we are in a position to reconsider its relevance for the issue of contrastive

analysis of svarabhakti here, so I do not discuss it further to avoid circularity. 7 Jones (2000) states that this rule is not “fully regular” (gu léir cunbhalach) and that glottal stop insertion appears possible in preconsonantal contexts as well: [ˈɣɛʔn mi] ‘I stayed’ (dh’fhan mi). It seems plausible, however, that the variable application of the rule in preconsonantal contexts could be seen as a further development of the system sketched here. It is noteworthy that the overapplication is said to happen before the sonorants [n l r], i. e. precisely those segments that enter a “fortis/lenis” contrast in Scottish Gaelic (and Irish), which is often treated in moraic terms. This glottal stop insertion could then be seen as a process parallel to the lengthening and/ or diphthongization of vowels before fortis sonorants, analysed as a compensatory process due to the delinking of the mora from the sonorant (see e. g. Ní Chiosáin 1991: § 4).

“Pitch accent” and prosodic structure in Scottish Gaelic 

 37

syllabification. As discussed above, this analysis of glottal stop insertion allows us to account for minimal pairs such as Tiree [ˈpoʔɔ] ‘underwater rock’ vs. [ˈpoː] ‘cow’ (and therefore possibly Hebridean [1ˈpoː] vs. [2ˈpoː]) in terms of stored prosodic structure, i. e. as underlying /(po)σ(o)σ/ vs. /poo/. All of the arguments adduced in example (6) concern glottal stop insertion in preconsonantal position. Crucial cases such as those in [ˈpoʔɔ] vs. [ˈpoː] (and parallel examples) all involve prevocalic glottal stops. These could, in principle, be analysed as hiatus-breaking devices with an onset glottal stop. Although this would involve two different accounts for what appear to be parallel processes, such an analysis might allow us to dispense with contrastive syllabic structure. In this case, ‘underwater rock’ would involve a mapping /poo/ → [.po.ʔo.] and ‘cow’ would be stored with a long vowel. This scenario involves only storing moraic structure, which is uncontroversial (e. g. Morén 2001). However, if it could be shown that the default response to hiatus is not glottal stop insertion, then forms such as [poʔɔ] cannot be accounted for in this way. As observed by Smith (1999), the response is contraction rather than consonant epenthesis. This evidence comes from the interaction of syncope and glottal stop insertion. As discussed above, some suffixes trigger a deletion of the second syllable in polysyllabic stems, cf. example (8-b) above. Crucially, this deletion also affects what I hypothesize to be stored syllabic nodes. Consider the following examples from Holmer (1938); I rewrite his to to make it clear that it is a diphthong. (11) a. (i) [ˈjoʔur] leabhar (ii) [ˈjowriçən] leabhraichean

‘book’ (183) ‘books’ (183)

b. (i) [ˈuʔul] (ii) [ˈuːlən]

‘apple’ (231) ‘apples’ (231)

ubhal ubhlan

Assume ‘book’ is stored as /(jo)σ(ur)σ/. There can be no contraction of the two adjacent vowels, presumably prevented by faithfulness to underlying prosodic structure, since it would force a single output syllable to correspond to two input ones. In a suffixed form which creates a context for syncope, the second syllabic node has to be deleted under pressure from whatever factor drives the syncope, and in this case the vowel /u/ can be subject to the general rules of syllabification active in the language, which force it to be parsed as part of a diphthong. An alternative account assuming that glottal stop insertion is unrelated to the number of syllables in the form has no explanation for why syncope triggers the contraction rule rather than epenthesis of the glottal stop in suffixed forms. Thus, I conclude that the glottal stop in [ˈjoʔur] and [ˈuʔul] (and by extension [ˈpoʔɔ]) cannot be a hiatus breaker, but must rather belong to the coda of a preceding syllable.

38 

 Pavel Iosad

Although in principle we cannot uncritically extend the analysis of glottal stop insertion in southern varieties to Hebridean “pitch accents”, the fact that dialects such as those of Leurbost show entirely parallel examples of the interaction of pitch accents and syncope suggests that we are justified in viewing the pitch accents as deriving from syllabification (numbers show pages in Oftedal 1956): (12) a. (i) [1ˈʎɔːr] (ii) [2ˈʎɔːriçən]

leabhar ‘book’ (70) leabhraichean ‘books’ (175)

b. (i) [1ˈuəlˠ] (ii) [2ˈuːlˠən]

ubhal ubhlan

‘apple’ (76) ‘apples’ (75)

2.5 Interim conclusion If the analysis given above is correct, then a language can acquire the Scottish Gaelic type of “pitch accent” via a confluence of two factors. First, contrastive prosodic structure must be present in surface representations. In the particular case of Scottish Gaelic, it appears that the relevant node in the prosodic hierarchy is, unusually, the syllable. Although it is often claimed that contrastive syllabic structure (in underived forms) should be impossible, I would suggest that the review of the evidence above allows us to make a sufficiently robust claim to that effect.8 Second, once the prosodic structures are in place, even fairly general mechanisms of aligning suprasegmental features such as tones with prosodic pivots (heads and edges) are sufficient to derive pitch accent contrasts. This is the essence of the proposal by Ladefoged (2003) that the dialect of Lewis uses the LH L contour for both “accent 1” and “accent 2” words, with the distinction being

8 In previous research, the facts have been analysed by postulating an empty onset consonant (Clements 1986; Smith 1999), which acts as a proxy for syllable count and ensures that sonorants in svarabhakti words such as [2palˠak] ‘belly’ and glottal stops in cases such as [ˈpoʔɔ] ‘underwater rock’ are parsed as codas (so [.palˠ._ak.], [.poʔ._ɔ.]). However, for this solution to work the qualification that the empty segment should be an onset appears quite crucial: if syllabification is entirely deterministic and driven in large measure by sonority (e. g. Zec 1988; Morén 2001; Gouskova 2004), then it is not at all clear why the empty unprosodified segment is parsed as an onset and does not undergo some other process (such as deletion or coalescence). Designating a segment as an onset is tantamount to storing at least a syllabic treelet if not a full syllable: it adds a poorly motivated element to the surface representation without resolving the conundrum of stored syllable structure.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 39

derived only by the difference in prosodic structure rather than by a lexical difference in tonal melodies. In the next section I will show that both components of the Scottish Gaelic recipe – lexically specified prosodic structure and the realization of what are phonologically prosodic contrasts by pitch differences – are independently found in the Celtic languages. I focus on Irish for exceptional prosodic structure and on a Breton dialect for the relationship between pitch and prosodic constituency.

3 Exceptional prosodic structure in Irish As shown in section 2, nontrivial prosodic structure is reflected in Scottish Gaelic in a number of ways apart from the pitch accent contrast. Here I consider data from related varieties which demonstrate a very similar pattern of rhyme palatalization, despite not having the other robust cues to surface prosodic structure. Unlike Scottish Gaelic, in varieties of Irish the epenthetic vowel that is inserted following sonorants that precede certain consonants “counts” for the purposes of prosodic structure, in particular for the projection of feet. First, as demonstrated by Ní Chiosáin (1999), Irish prefers words to contain a binary non-final foot, and svarabhakti in Irish is only allowed when it contributes to improving foot structure. Conversely, it is blocked when the word has sufficient segmental material to build the needed structure without the insertion of an extra root node. Second, as discussed by Green (1997), in some dialects of Irish stress falls within a three-syllable window at the left edge of the word. Barring some irrelevant complications, long vowels attract stress; in the absence of long vowels in the three-syllable window, stress falls on the initial syllable. Crucially, epenthetic vowels “count” for the purposes of establishing the number of syllables, because they are able to push a long vowel outside the three-syllable window: in a word of the form LLLH stress is initial even if one of the vowels in the light syllables is epenthetic. Thus, learners of Irish lack some important cues to surface prosodic structure that are available to learners of Scottish Gaelic, such as the phonological invisibility of epenthetic vowels, pitch accents, and glottalization. Nevertheless, Irish dialects preserve at least one phenomenon associated with exceptional prosodic structure, namely rhyme palatalization. For concreteness, I consider here the dialect of Corca Dhuibhne (Dingle), a Munster variety spoken in the south-west of Ireland and described in detail by Ó Sé (2000). Given that the epenthetic vowel in Irish is always [ə] or [ɪ], it would be relatively difficult for speakers to recover the unusual surface prosodic structure found in cases of svarabhakti: there does not appear to be a phonetic factor that

40 

 Pavel Iosad

would allow the speakers to identify whether a given instance of [ə] is derived via svarabhakti or comes from some other source. Nevertheless, the dialect retains traces of at least one cue for this prosodic structure. In monosyllables, we find the expected pattern whereby palatalization and fronting/raising affects only the rhyme of the syllable, as in the following examples: (13) a. (i) [ˈbrov] (ii) [ˈbrivʲ]

brobh broibh

‘rush’ ‘rush (gen. sg.)’

b. (i) [ˈknuk] (ii) [ˈknikʲ]

cnoc cnoic

‘hill’ ‘hill (gen. sg.)’

If Munster Irish reproduced the historical pattern faithfully, we would expect disyllables, such as [ˈsoləs] ‘light’ (solas), to exhibit the raising effects only in the second syllable (cf. Old Irish genitive singular soluis rather than *suilis); conversely, in monosyllables such as [ˈboləɡ] ‘belly’ (bolg) we would expect irregular fronting and/or raising of both vowels. Historical monosyllables often exhibit the expected behaviour: (14) a. (i) [ˈboləɡ] (ii) [ˈbilʲɪɡʲ]

bolg builg

‘belly’ ‘belly (gen. sg.)’

b. (i) [ˈlʲanəv] leanbh ‘child’ (ii) [ˈlʲinʲɪvʲ] linbh ‘child (gen. sg.)’

However, the raising of both vowels and palatalization of the medial sonorant can also affect historical disyllables, where this alternation may vary with the historically correct pattern. (15) a. (i) [ˈsoləs] (ii) [ˈsolɪʃ] (iii) [ˈsilʲɪʃ]

solas solais

‘light’ ‘light (gen. sg.)’

b. (i) [ˈdorəs] (ii) [ˈdirɪʃ]

doras dorais

‘door’ ‘door (gen. sg.)’

Clearly, the Munster Irish forms do not merely descend from Old Irish unepenthesized ones, because we find examples such as those in (15), where the historical pattern associated with monosyllables is extended to disyllables. This suggests that the generalizations associated with the unusual prosodic structure of svarabhakti words may have survived the obliteration of the other cues. Thus, surface prosodic structure can be a robust enough element of the system, and

“Pitch accent” and prosodic structure in Scottish Gaelic 

 41

therefore it may exist in the absence of a pitch accent. Crucially, the existence of this surface structure is not at all contingent on language contact: from a historical perspective, it simply represents the reflex of the Old Irish syllable count.9 In the next section I consider some cases that have been analysed as contrasts in surface prosodic structure realized as contrasts in pitch.

4 Prosodic structure contrasts and pitch A crucial distinction between Munster Irish and Scottish Gaelic under the present proposal is that the latter does not just retain surface prosodic structure but also reflects this contrast using in the assignment of pitch accents. In this section I consider data from another Celtic variety – the Breton dialect of Bothoa – where a “pitch accent” contrast also appears amenable to an analysis in terms of surface prosodic structure. Before I present the Breton analysis, I briefly discuss the relationship between prosodic structure and pitch accent contrasts in Germanic languages.

4.1 North and West Germanic In the preceding sections I have treated Germanic (especially North Germanic) as a “prototypical” pitch accent system, following much of the literature in the field. However, the analysis of Germanic varieties as exhibiting lexical tones, although quite widespread (see, for instance, Bruce 1977; Lorentz 1984; Riad 1992; Wetterlin 2010 for North Germanic and Gussenhoven 2000, 2004; Peters 2008 for West Germanic), is not entirely uncontroversial. In particular, Morén (2003, 2008) treats both the Swedish and Norwegian tonal accents and the Danish stød as involving the top-down association of laryngeal activity with surface prosodic structure, while Köhnlein (2011) offers an analysis of a West Germanic system along these lines. In itself, the proposition that some of the tonal picture in Germanic derives from some sort of default assignment of tone associated with a relevant intona-

9 This is not to say Corca Dhuibhne Irish today exhibits the same sort of surface structure as that posited for Hebridean dialects. The phonological irregularity and the lexical specificity of the alternation suggest that it is quite advanced in terms of the life cycle of phonological processes (e. g. Bermúdez-Otero 2007). Since synchronic analysis of the Irish patterns is not the focus of this paper, I leave the question aside here.

42 

 Pavel Iosad

tional structure is not entirely new: it is a staple of approaches that view the contrast between the two types of accents as privative (see e. g. Lahiri et al. 2005 for an overview). However, this top-down assignment of intonational tones at the higher levels of the prosodic hierarchy such as the Intonational Phrase is usually assumed to coexist with a lexical contrast in tonal melodies; disentangling the two is an important area of enquiry (e. g. Bruce 1977; Gussenhoven and van der Vliet 1999). By contrast, approaches such as those of Morén-Duolljá (2013) and Köhnlein (2011) imply that also the lexically specific contours derive not from tonal contrasts but rather via general algorithms regulating the assignment of tones at lower levels of the hierarchy such as the foot or even the mora. For example, Köhnlein (2011) treats the contrast between “accent 1” and “accent 2” in the Arzbach dialect as one between monosyllabic binary feet, consisting of a single bimoraic syllable (H), and disyllabic uneven trochees (HL). He argues that the difference in tonal curves (in declarative contexts) derives from the fact that low tones are dispreferred on head morae. Given the foot structure he proposes for that dialect, either one or both of the morae in a stressed (bimoraic) syllable are treated as “head morae”, and this derives the pitch distinction. No storage of tone in the lexicon is required under this analysis, but foot structure must be specified (either as part of the underlying representation or through morphological processes). This analysis allows Köhnlein (2011) to combine completely general mechanisms for associating tone with prosodic structure and independently specified constituency, just as proposed by Ladefoged (2003) for Scottish Gaelic. However, the case might still not be watertight, in that, as far as the data are described by Köhnlein (2011), there does not appear to be any non-tonal evidence corroborating the prosodic structure he proposes. In the next section I consider the data from the Breton dialect of Bothoa, where such evidence is available.

4.2 Celtic In this section I sketch some relevant aspects of the prosodic system of the Breton dialect of Bothoa as described by Humphreys (1995). Specifically, I argue that the contrast between the two “accents” identified by Humphreys (1995) is best analysed as a contrast between monopedal structures and structures specified as having two feet. Crucially, the behaviour of words showing the accent which I hypothesize to reflect the presence of two feet is completely in line with the behavior of other words containing multiple feet in the language. For reasons of

“Pitch accent” and prosodic structure in Scottish Gaelic 

 43

space and focus, I omit some of the detailed argumentation here; see Iosad (2012: § 7.3) for a fuller account. Apart from Scottish Gaelic, the Celtic languages are not generally associated with having “pitch accents”. Pitch plays an important role in the phonetic expression of stress in Welsh (e. g. Williams 1985), and at least one scholar has proposed an extensive phonological analysis involving “tonemes” (Pilch 1975). However, the distribution of pitch in Welsh is not driven by lexical factors (Thomas 1967; Rhys 1984), and thus it is not a “pitch accent” system: in terms of stress typology, Welsh exhibits a relatively orthodox system enforcing penultimate stress by default. Closely related to this fact is the restriction on weight: long vowels in Welsh are restricted to stressed syllables.

4.2.1 Bothoa Breton: The data Breton is closely related to Welsh, and the stress system of many dialects is all but identical to the Welsh one, with consistent penultimate stress. A notable exception is found in the dialects found in the south-east of the Breton-speaking area (commonly called Vannetais). Stress in Vannetais dialects falls on the final syllable, and further developments of the Vannetais pattern are described by Jackson (1967), Falc’hun (1981), and Plourin (1985). The hamlet of Bothoa lies in the far east of the Breton-speaking region, near the border of what is traditionally considered to be Vannetais territory. For our purposes, the most interesting aspect of this variety is its prosodic system, which, unusually for Brythonic languages in general, puts relatively few restrictions on the distribution of stress and weight within words. In particular, Bothoa Breton allows more than one long vowel per word. In addition, the placement of stress is entirely lexical, and it is not tied to word edges: adding more syllables to a word within a paradigm does not lead to stress shift. Similarly, whereas in Welsh and many Breton varieties long vowels may only appear in stressed syllables, in Bothoa Breton vowel length is essentially lexical: if a long vowel is present in the underlying representation of a morpheme, it will surface as long irrespective of whether it bears main stress. Another important aspect of Bothoa Breton prosodic structure that appears deviant from a Celtic perspective is the fact that it allows more than one stress per word, i. e. it has both main and secondary stress. In principle, both short and long vowels may bear both types of stress, as the following examples illustrate: (16) a. [ˌbyːˈeːəw] buhezioù ‘saints’ lives’ b. [ˌʃyːˈbadər] skubadur ‘swept rubbish’ c. [ˌdisˈpako] dispakañ ‘to unpack’

44 

 Pavel Iosad

Finally, and most importantly for our purposes, Humphreys (1995) describes the existence of a contrast between two classes of words in terms of their pitch patterns. In the normal case, stress in all-light disyllables falls on the initial syllable; phonetically, the initial syllable is the locus of both a pitch accent and increased duration. In a minority of lexical items, however, the vowels are said to be of about equal duration, and the second syllable hosts a significant rise in pitch “strikingly reminiscent of Welsh accentuation” (“[qui] rappelle d’un façon assez frappante l’accentuation du gallois”). Humphreys (1995) writes the latter words as containing two stresses, reflecting an analysis that I will show to be essentially correct. (17) a. Default accentuation [ˈparuz̥] parrez ‘parish’

b. Additional pitch accent [ˈdaˌvad̥] dañvad ‘ewe’

From now on, I will refer to words such as those in (17-a) as “single-peaked” words and to words such as those in (17-b) as “double-peaked”. In the remainder of this section I will show that Humphreys (1995) is correct in identifying the source of the pitch accent as lexically stored prosodic structure, and more specifically as foot structure.

4.2.2 Bothoa Breton: The analysis In order to establish the correctness of this hypothesis, we turn to a consideration of the behaviour of words in which the existence of multiple feet should be uncontroversial, namely those containing multiple stresses. The majority of such words fall into two categories: words with multiple long vowels and words with “stressed suffixes”. In the former case, the generalization is quite simple: all long vowels in Bothoa Breton bear (at least secondary) stress: (18) a. [ˌhyːˈaːl] hual ‘hindrance’ b. [ˌziːȷ̃aˈtyːr] sinatur ‘signature’ c. [ˌʧɒːˈdiːʒən] teod-ejen ‘plantain’

In the latter case, Humphreys (1995) identifies a class of suffixes that receive stress despite having a short vowel, as in the examples in (19). (19) Stressed affixes a. (i) [ˈʃyːb-ad̥] (ii) [ˌʃyːˈb-adər]

skubañ skubadur

‘to sweep’ ‘swept rubbish’

“Pitch accent” and prosodic structure in Scottish Gaelic 

 45

b. (i) [ˈdesk-o] deskiñ ‘to study’ (ii) [ˌdesˈk-adəræz̥] deskadurezh ‘teaching’

If we follow the standard assumption that stress is the phonetic correlate of being the head of a foot, we can conclude that in Bothoa Breton multiple feet may have two sources: long vowels, where the rationale for foot-building is presumably phonological (the weight-to-stress principle), and lexical specification of suffixes for foot structure. For our purposes, the crucial fact is that whatever the sources of this foot structure are, main stress (i. e. the head foot of the word) is always the rightmost stress-attracting element, as can be confirmed by inspection of the relevant forms in Humphreys (1995). These data also show that the canonical foot type in Breton is the moraic trochee, i. e. either a (H) syllable (normally only long vowels count for weight) or a (LL) disyllabic sequence. Degenerate feet (i. e. those consisting of a single syllable with a short vowel) are also found in the language, albeit in strictly circumscribed conditions: they are never built by the phonological computation, but only preserved when they are part of the input to the phonological component, either because they are lexically specified (at the first cycle) or because they are output by a previous cycle (see Iosad 2012: § 7.3.2 for detailed argumentation to this effect). With these findings in mind, we turn back to the contrast between single-peaked and double-peaked words. In describing double-peaked disyllables, Humphreys (1995) admits that the choice of writing them as [σ́σ̀], with main stress preceding secondary stress, is essentially arbitrary, as the two syllables have similar levels of phonetic prominence (with pitch playing a prominent role in the expression of the contrast). However, he is consistent in writing the suffixed forms of double-peaked words with main stress on the second syllable: (20) a. (i) [ˈdaˌvad̥] dañvad (ii) [ˌdaˈvadəw] deñved

‘ewe’ ‘sheep (pl.)’

b. (i) [ˈlaˌɡad̥] (ii) [ˌlaˈɡadən]

‘eye’ ‘bud’

lagad lagadenn

This “flip” can be explained if we assume that the relevant lexical items are in fact stored with foot structure: thus, ‘ewe’ is stored as /(da)Ft(vad)Ft/. According to Humphreys (1995: 66), words pronounced in isolation tend to have a rising intonation, with an especially abrupt rise on the final syllable. This seems to indicate a H% or LH% boundary tone on some prosodic constituent (since no data are provided on phrase-level intonation, it is difficult to identify which level of the prosodic hierarchy introduces this tone). If we also assume that foot heads may

46 

 Pavel Iosad

be associated with some pitch accent, we can expect that unsuffixed forms such as [ˈdaˌvad̥] with two degenerate feet, when pronounced in isolation, will show extensive tonal specification. This will be particularly true of the final syllable in double-peaked words, which will host both the tone associated with foot heads and the boundary tone(s) of the higher-level constituents, explaining the “Welshlike” abrupt rise. In suffixed forms of the relevant words, there is enough segmental material to build a binary foot for the second stress, resulting in a pattern which is more readily recognizable as similar to main stress in other words with multiple stresses such as those in (19). Crucially, these suffixed forms show the same right alignment of main stress as that seen in the language otherwise. Moreover, the consistent right-headedness in such words finds a parallel in the behaviour of “stressed affixes”: almost all affixes identified by Humphreys (1995) as consistently attracting main stress from a long vowel (as in (19)) are at least two morae long (i. e. they contain at least two syllables with short vowels or a long vowel). This evidence allows us to conclude that the “pitch accents” identified in the Bothoa dialect of Breton by Humphreys (1995) are highly likely to be artefacts of lexical specification of prosodic structure coupled with general mechanisms for the assignment of tone: essentially the same analysis as that proposed in section 2 for Scottish Gaelic. In the next section I briefly describe the possible origins of the Bothoa prosodic system.

4.2.3 Bothoa Breton: The history The pattern in Bothoa Breton is interesting not just synchronically, but also with respect to its origins. Common to Scottish Gaelic, North Germanic, and West Germanic systems discussed above is the historical relationship between exceptional prosodic structure and (apparent) changes in syllable count. In Scottish Gaelic, the appearance of a “pitch accent” contrast is at least partially due to the rise of apparently disyllabic words through epenthesis. In North Germanic, the rise of pitch accents has been variously related to epenthesis in previously monosyllabic words (e. g. Oftedal 1952) or to syncope (Riad 1998). In West Germanic the rise of the pitch accent is clearly related to apocope and the (partial) neutralization of a contrast between disyllabic and monosyllabic forms in the direction of the latter (Gussenhoven 2000, 2004; Köhnlein 2011). The origin of the Bothoa Breton

“Pitch accent” and prosodic structure in Scottish Gaelic 

 47

pattern is also related to changes in syllable count, being due to a scenario that is fairly similar to that proposed for North Germanic by Riad (1998).10 As briefly discussed above, Bothoa Breton shares many properties with the Vannetais dialects spoken in the south-east of the Breton-speaking area. Among the peculiar features of these dialects is final stress. However, this system has undergone a number of further developments (Falc’hun 1981; Plourin 1985). Among these, of particular interest to us are stress retraction and syncope. First, Falc’hun (1981: 271–274) notes a number of examples of trisyllabic words with initial stress, as in hwérezet ‘sisters’ instead of the expected hwerezét. He interprets this in terms of a secondary stress appearing in order to repair the initial lapse (sequence of unstressed syllables) and then taking over as the main stress in the word: σσσ́ → σ̀σσ́ → σ́σσ. A second development found in Vannetais dialects that are affected by this retraction is syncope, i. e. deletion of medial syllable: hwérzet or hwárzet ‘sisters’. Bothoa Breton appears to have partaken of at least some of these developments: as recorded by Humphreys (1995), the plural of ‘sister’ in the dialect is [ˈhwɛːrzəd̥] (or [ˈhwɛːrəzəd̥]). Neither process is completely regular in this variety: as discussed above, stress in Bothoa Breton is relatively free (or at least not fixed on the initial syllable), whereas syncope appears to be a variable process that has not yet been completely phonologized (see Iosad 2012 for details). Nevertheless, it appears plausible that the ultimate source of the “two degenerate feet” pattern is the historically intermediate representation σ̀σσ́ hypothesized by Falc’hun (1981). The structure can be derived if we assume that the correct foot structure in this case was (σ̀σ)(σ́). Following syncope, which affected the weak branch of the initial foot, the foot structure itself could remain intact, giving (σ̀)(σ́). This development is similar to that hypothesized for Proto-Nordic by Riad (1998), who suggests syncope in (H́L)(H́) structures created a suboptimal (H́)(H́) configuration with stress clash. The difference between Bothoa Breton and Proto-Nordic under this interpretation is that the former tolerates the clash and treats both syllables as metrically strong positions, as suggested above, while in Proto-Nordic the stress clash was removed by reinterpreting the second pitch peak as a lexical tone associated with the initial syllable (HL*+HL) rather than as a way to mark an ictus. In the particular case of Bothoa Breton, it is clear that the words exhibiting the “double-peaked” pattern do not necessarily go back to historical trisyllables: in this class, we find both historical disyllables ([ˈdaˌvad̥] ‘ewe’, Welsh dafad) and what are obviously recent borrowings such as [ˌlasˈtikən] ‘rubber band’ (French

10 Thanks to an anonymous reviewer for highlighting this similarity.

48 

 Pavel Iosad

élastique). Nevertheless, I suggest that historically the double-peaked accent must represent a Breton-internal development, probably with a role for dialect contact. While I cannot speculate on the precise sequence of events in the specific case of the Bothoa dialect, it appears quite likely that the ultimate origin of the pattern with two degenerate feet still lies within the prosodic system of Breton itself rather than in contact with some other language. Thus, a system of tones marking (sometimes lexically determined) prosodic constituents can arise in Celtic without an external influence.

5 Reconsidering contact between Nordic and Gaelic In this paper I have argued that the “pitch accent” system described for some Scottish Gaelic dialects is best viewed as the outcome of a combination of two distinct patterns, namely the existence of lexically specified prosodic structure and general rather than lexically specific mechanisms for the assignment of tone. I have also shown that both of these features are independently attested in various branches of Celtic. This claim undermines the otherwise plausible suggestion by scholars such as Borgstrøm (1974) that the “pitch accents” of Scottish Gaelic are, like other features of the Scottish Gaelic sound system (Marstrander 1932; Hansson 2001), the outcome of contact with North Germanic languages. We can take as our starting points the criteria for the plausibility of a contact explanation identified by Thomason (2001, 2010) as follows: 1. The existence of contact between the two languages; 2. Diverse shared features in the two languages, preferably from at least two different subsystems; 3. The existence of the shared feature(s) in the “source” language prior to contact; 4. The absence of the shared feature(s) in the “recipient” language prior to contact. The first criterion is unproblematic, as contact between Goidelic and North Germanic speakers is amply documented by historical and literary sources (e. g. Woolf 2007) and supported by the existence of numerous North Germanic loanwords in Scottish Gaelic11 (e. g. Stewart 2004) and by genetic studies (e. g. Helgason et al. 2000). The existence of shared features appears uncontroversial,

11 Lexical borrowing in the other direction appears more limited, with a concentration in personal names and (especially in the case of Iceland and the Faroe Islands) toponyms.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 49

although, interestingly, they are largely concentrated in the sound system: North Germanic is not commonly assumed to have contributed to either the morphology or the syntax of the Goidelic languages in a meaningful way (cf. Eska 2010). The existence of lexically specific pitch accents in North Germanic in the relevant period is more problematic. Although most continental North Germanic languages all have either “pitch accent” or pitch accent-like phenomena such as stød, it is notably absent in the insular languages Icelandic and Faroese (it is probably not knowable where it was found in Norn), which makes it uncertain whether the dialects spoken by the North Germanic inhabitants of the British Isles possessed the distinction. On the other hand, at least in the case of Icelandic, it appears possible that Old Norse syllable counts did survive for some time after epenthesis disrupted them, and it is possible that the distinction was expressed by pitch movements (e. g. Ottósson 1986; Myrvoll and Skomedal 2010; Þorgeirsson 2013). Moreover, it has been proposed (see especially Riad 1998, 2003) that the tonal specifications themselves, if not necessarily the pitch accent system as we know it, go back to Proto-Nordic, in which case the problem disappears (although see e. g. Bye 2011 for a different view). However, this paper does cast doubt on whether the fourth criterion is in place for contact between Gaelic and North Germanic. If the analysis presented in section 2 is correct, Gaelic “pitch accents” must have arisen in connection with the appearance of epenthesis and the deletion of voiced fricatives which swelled the ranks of “hiatus words”.12 In terms of dating, both of these processes would appear to be roughly contemporaneous with the rise of “lexical” pitch accents and stød in most of North Germanic (after about 1100). Essentially, the precursors for pitch accents as they appear in contemporary Scottish Gaelic (that is, the tones associated with the relevant prosodic constituents) were plausibly present at an early stage and do not need a contact explanation. This does not mean that I necessarily advocate the internal explanation over the one based on contact. As emphasized by Thomason (2001, 2010), multiple causation of language change is an extremely frequent phenomenon. Given the close similarity and rough contemporaneity of the appearance of lexically specific “pitch accents” in Gaelic and North Germanic, it is not at all implausible that the developments were parallel in the two languages, and perhaps reinforced by this closeness.

12 Some of the “hiatus words” were in fact disyllabic in Old Irish, e. g. fïach ‘raven’ with hiatus vs. fíach ‘obligation’ with a diphthong. It appears impossible to determine whether this contrast was implemented (primarily) by pitch in Old Irish, although presumably pitch at least participated in its expression.

50 

 Pavel Iosad

Gauging the probability of each scenario requires a better understanding of the historical circumstances of the contact (which is unfortunately quite difficult, see Woolf 2007) to reach a better understanding of the sociolinguistic setting (see e. g. Trudgill 2010, 2011 on its importance). It has often been assumed that the contact-based features of Scottish Gaelic such as preaspiration and “pitch accent” are due to language shift from Norse to Gaelic; this is also the conclusion of the study of lexical borrowings by Stewart (2004). There are some weaknesses in this approach, however. First, the status of contact as the source of preaspiration has also been put into doubt, on the basis of both phonetic (Ó Baoill 1980; Ní Chasaide 1986; Ó Murchú 1985) and dialectological (Ó Maolalaigh 2010) evidence. Second, contrary to the conclusions of Stewart (2004), Cox (2010) suggests that the pattern of Norse lexical borrowings in Gaelic indicates prolonged contact rather than abrupt language shift. Third, despite claims by earlier scholars such as Oftedal (1956) that Norse predominates in Hebridean toponymy, Cox (2002) argues that these numbers are significantly overestimated; this could also constitute evidence against the language-shift approach to contact between Norse and Gaelic. Thus, it is clear that further study is required to determine the type and scenario of language contact between Norse and Gaelic in Scotland. In this paper, I have shown that despite some compelling synchronic and diachronic similarities between the “pitch accent” systems of North Germanic and Gaelic, the hypothesis that their ultimate cause lies in contact should not be taken for granted.

References Adger, David (2010): Gaelic Morphology. In: Michelle Macleod and Moray Watson (eds.), The Edinburgh Companion to the Gaelic Language, 283–303. Edinburgh: Edinburgh University Press. Bermúdez-Otero, Ricardo (2007): Diachronic Phonology. In: Paul de Lacy (ed.), The Cambridge Handbook of Phonology, 497–518. Cambridge: Cambridge University Press. Borgstrøm, Carl Hjalmar (1974): On the Influence of Norse on Scottish Gaelic. Lochlann 6: 91–107. Borgstrøm, Carl Hjalmar (1937): The Dialect of Barra in the Outer Hebrides. Norsk tidsskrift for sprogvidenskap 8: 71–242. Borgstrøm, Carl Hjalmar (1940): The dialects of the Outer Hebrides. Oslo: Norwegian Universities Press. Bosch, Anna R. K. and Kenneth de Jong (1997): The Prosody of Barra Gaelic Epenthetic Vowels. Studies in the Linguistic Sciences 27: 1–15. Bruce, Gösta (1977): Swedish Word Accent in Sentence Perspective, Lund: Gleerup. Bye, Patrik (2011): Mapping Innovations in North Germanic Using GIS. Oslo Studies in Language 3(2): 5–29.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 51

Bye, Patrik and Paul de Lacy (2008): Metrical Influences on Fortition and Lenition. In: Joaquim Brandão de Carvalho, Tobias Scheer and Philippe Ségéral (eds.), Lenition and Fortition, 173–206. Berlin and New York: Mouton de Gruyter. Carnie, Andrew (1994): Whence Sonority? Evidence from Epenthesis in Modern Irish. In: Tony Bures, Andrew Carnie, Heidi Harley (eds.), MIT Working Papers in Linguistics 21: Papers on Phonology and Morphology. Cambridge, MA: Distributed by Dept. of Linguistics and Philosophy, Massachusetts Institute of Technology. Clements, G. Nick (1986): Syllabification and Epenthesis in the Barra Dialect of Gaelic. In: Koen Bogers, Harry van der Hulst, Maarten Mous, and John Massie Stuart (eds.), The Phonological Representation of Suprasegmentals, 317–336. Dordrecht: Foris Publications. Cox, Richard A. V. (2010): Faclan-iasaid às an Lochlannais ann an Gàidhlig na h-Albann. In: Richard A. V. Cox and Gillian Munro (eds.), Cànan & Cultar / Language & Culture, 1–6. Edinburgh: Dunedin Academic. Cox, Richard A. V. (2002): The Gaelic Place-Names of Carloway, Isle of Lewis. Their Structure and Significance. Dublin: School of Celtic Studies, Dublin Institute for Advanced Studies. Dell, François and Mohamed Elmedlaoui (1985): Syllabic Consonants and Syllabification in Imdlawn Tashlhiyt Berber. Journal of African Languages and Linguistics 7: 105–130. Eliasson, Stig (2000): Typologiska och areallingvistiska aspekter på de nordeuropeiska språkens fonologi. In: Ernst Håkon Jahr (ed.), Språkkontakt: innverknaden frå nedertysk på andre nordeuropeiske språk, 21–70. København: Nordisk Ministerråd. Eska, Joseph F. (2010): Contact and the Celtic Languages. In: Raymond Hickey (ed.), The Handbook of Language Contact, 539–549. Chichester: Wiley-Blackwell. Falc’hun, François (1981): Perspectives nouvelles sur l’histoire de la langue bretonne, Paris: Union générale d’éditions. Gouskova, Maria (2004): Relational Hierarchies in Optimality Theory: The Case of Syllable Contact. Phonology 21: 201–250. Gouskova, Maria and Nancy Hall (2009): Acoustics of Epenthetic Vowels in Lebanese Arabic. In: Steve Parker (ed.), Phonological Argumentation: Essays on Evidence and Motivation, 203–226. London: Equinox. Green, Anthony Dubach (1997): The prosodic structure of Irish, Scots Gaelic, and Manx. PhD thesis, Ithaca, NY. Gussenhoven, Carlos (2000): On the Origin and Development of the Central Franconian Tone Contrast. In: Aditi Lahiri (ed.), Analogy, Leveling, Markedness, 213–260. Berlin and New York: Mouton de Gruyter. Gussenhoven, Carlos (2004): The phonology of tone and intonation. Cambridge: Cambridge University Press. Gussenhoven, Carlos and Jörg Peters (2004): A Tonal Analysis of Cologne Schärfung. Phonology 21: 251–285. Gussenhoven, Carlos and Peter van der Vliet (1999): The Phonology of Tone and Intonation in the Dutch Dialect of Venlo. Journal of Linguistics 35: 99–135. Hall, Nancy (2006): Cross-Linguistic Patterns of Vowel Intrusion. Phonology 23: 387–429. Hall, Nancy (2013): Acoustic Differences between Lexical and Epenthetic Vowels in Lebanese Arabic. Journal of Phonetics 41: 133–143. Hansson, Gunnar Ólafur (2001): Remains of a Submerged Continent: Preaspiration in the Languages of Northwest Europe. In: Laurel J. Brinton (ed.), Historical Linguistics 1999, 157–173. Amsterdam: John Benjamins.

52 

 Pavel Iosad

Helgason, Agnar, Sigrún Sigurðardóttir, Jayne Nicholson, Bryan Sykes, Emmeline W. Hill, Daniel G. Bradley, Vidar Bosnes, Jeffery R. Gulcher, Ryk Ward, and Kári Stefánsson (2000): Estimating Scandinavian and Gaelic Ancestry in the Male Settlers of Iceland. The American Journal of Human Genetics 67: 697–717. Hind, Kevin (1996): The Structure of Epenthesis in Gaelic. Journal of Celtic Linguistics 5: 91–119. Holmer, Nils M. (1938): Studies on Argyllshire Gaelic, Uppsala: Almqvist & Wiksell. Humphreys, Humphrey Lloyd (1995): Phonologie et morphosyntaxe du parler breton de Bothoa en Saint-Nicolas-du-Pélem. Brest: Ar Skol Vrezoneg. Iosad, Pavel (2012): Representation and Variation in Substance-Free Phonology: A Case Study in Celtic. PhD thesis, Tromsø. Jackson, Kenneth Hurlstone (1967): A Historical Phonology of Breton. Dublin: Dublin Institute for Advanced Studies. Jones, George (2000): Beagan mu’n stad ghlotasach ann an Gàidhlig Ceann a Deas Earraghaidheil. Scottish Gaelic Studies 20: 201–211. Köhnlein, Björn (2011): Rule Reversal Revisited: Synchrony and Diachrony of Tone and Prosodic Structure in the Franconian Dialect of Arzbach. PhD thesis, Leiden. Kristoffersen, Gjert (2000): The phonology of Norwegian. Oxford: Oxford University Press. Ladefoged, Peter (2003): Commentary: Some Thoughts on Syllables – An Old-Fashioned Interlude. In: John Local, Richard Ogden, and Rosalind A. M. Temple (eds.), Papers in Laboratory Phonology VI, 269–276. Cambridge: Cambridge University Press. Ladefoged, Peter, Jenny Ladefoged, Alice Turk, Kevin Hind, and St. John Skilton (1998): Phonetic Structures of Scottish Gaelic. Journal of the International Phonetic Association 28: 1–41. Lahiri, Aditi, Allison Wetterlin, and Elisabeth Jönsson-Steiner (2005): Lexical Specification of Tone in North Germanic. Nordic Journal of Linguistics 28: 61–96. Levin, Juliette (1987): Between Epenthetic and Excrescent Vowels. In: Megan Crowhurst (ed.), Proceedings of the Sixth West Coast Conference on Formal Linguistics, 187–201. Stanford, CA: Stanford Linguistics Association. Liberman, Anatoly (2000): Schärfung/stoottoon and Trägheitsakzent/sleeptoon in the Rhein– Limburg Area and their Scandinavian Analogues. In: Carol F. Justus and Edgar C. Polomé (eds.), Language Change and Typological Variation: In Honor of Wilfred P. Lehmann on the Occasion of his 83rd Birthday. Volume 2: Grammatical Universals and Typology, 275–298. Washington, D. C.: Institute for the Study of Man. Lorentz, Ove (1984): Stress and Tone in an Accent Language. In: Claes-Christian Elert, Iréne Johansson, and Eva Strangert (eds.), Nordic Prosody III, 165–178. Umeå and Stockholm: Almqvist & Wiksell. MacAulay, Donald (1993): The Scottish Gaelic Language. In: Donald MacAulay (ed.), The Celtic languages, 137–248. Cambridge: Cambridge University Press. Marstrander, Carl (1932): Okklusiver og substrater. Norsk tidsskrift for sprogvidenskap 5: 258–304. McGarrity, Laura W. (2003): Constraints on Patterns of Primary and Secondary Stress. PhD thesis, Bloomington, IN. Morén, Bruce (2001): Distinctiveness, Coercion, and Sonority: A Unified Theory of Weight. London and New York: Routledge. Morén, Bruce (2003): Danish Stød and Eastern Norwegian Pitch Accent. The Myth of Lexical tones. Presentation at the 13th Manchester Phonology Meeting. Morén, Bruce (2008): Using the Prosodic Hierarchy to Account for North Germanic Tones. Presentation at the Workshop on the Prosodic Hierarchy, Tromsø.

“Pitch accent” and prosodic structure in Scottish Gaelic 

 53

Morén-Duolljá, Bruce (2013): The Prosody of Swedish Underived Nouns: No Lexical Tones Required. Nordlyd 40(1): 196–248. Myrvoll, Klaus Johan and Trygve Skomedal (2010): Tonelagsskilnad i islendsk i Tridje grammatiske avhandling. Maal og Minne 1: 68–97. Nevins, Andrew (2010): Locality in Vowel Harmony. Cambridge, MA: MIT Press. Ní Chasaide, Ailbhe (1986): Preaspiration in Phonological Stop Contrasts: An Instrumental Phonetic Study. PhD thesis, Bangor. Ní Chiosáin, Máire (1991): Topics in the Phonology of Irish. PhD thesis, Amherst. Ní Chiosáin, Máire (1999): Syllables and Phonotactics in Irish. In: Harry van der Hulst and Nancy Ritter (eds.), The Syllable: Views and Facts, 551–575. Berlin and New York: Mouton de Gruyter. Ní Chiosáin, Máire, Pauline Welby, and Robert Espesser (2012): Is the Syllabification of Irish a Typological Exception? An Experimental Study. Speech Communication 54: 68–91. Ó Baoill, Dónall P. (1980): Preaspiration, Epenthesis and Vowel Lengthening: Interrelated and of Similar Origin? Celtica 13: 79–108. Ó Maolalaigh, Roibeard (2010): The Sound of Silence. Some Structural Observations on Preaspiration in Scottish Gaelic. In: Abigail Burnyeat, Thomas Owen Clancy, Wilson McLeod, Roibeard Ó Maolalaigh, and Domhnall Uilleam Stiùbhart (eds.), Bile ós Chrannaibh: a Festschrift for William Gillies, 365–404. Ceann Drochaid: Clann Tuirc. Ó Murchú, Máirtín (1985): Varia VIII. Devoicing and Pre-aspiration in Varieties of Scots Gaelic. Ériu 36: 195–198. Ó Sé, Diarmuid (2000): Gaeilge Chorca Dhuibhne. Baile Átha Cliath: Conradh na Gaeilge. Oftedal, Magne (1952): On the Origin of the Scandinavian Tone Distinction. Norsk tidsskrift for sprogvidenskap 16: 201–225. Oftedal, Magne (1956): The Gaelic of Leurbost, Isle of Lewis. Oslo: Norwegian Universities Press. Operstein, Natalie (2010): Consonant Structure and Prevocalization, Amsterdam: John Benjamins. Ottósson, Kjartan G. (1986): Indicier på tonaccentdistinktion i äldre isländska. Íslenskt mál 8: 183–190. Peters, Jörg (2008): Tone and Intonation in the Dialect of Hasselt. Linguistics 46: 983–1018. Pilch, Herbert (1975): Advanced Welsh Phonemics. Zeitschrift für celtische Philologie 34: 60–102. Plourin, Jean-Yves (1985): L’accentuation en Haute-Cornouaille et en bas-vannetais. La Bretagne linguistique 1: 103–115. Prince, Alan S. (1990): Quantitative Consequences of Rhythmic Organization. In: Michael Ziolkowski, Manuela Noske, and Karen Deaton (eds.), Parasession on the Syllable in Phonetics and Phonology, 355–398. Chicago: Chicago Linguistic Society. Rhys, Martin, (1984): Intonation and the Discourse. In: Martin J. Ball and Glyn E. Jones (eds.), Welsh Phonology: Selected Readings, 125–155. Cardiff: University of Wales Press. Riad, Tomas (1992): Structures in Germanic Prosody. PhD thesis, Stockholm. Riad, Tomas (1998): The Origin of Scandinavian Tone Accent. Diachronica 15: 63–98. Riad, Tomas (2003): Diachrony of the Scandinavian Accent Typology. In: Paula Fikkert and Haike Jacobs (eds.), Development in Prosodic Systems, 91–144. Berlin and New York: Mouton de Gruyter. Ridouane, Rachid (2008): Syllables without Vowels: Phonetic and Phonological Evidence from Tashlhiyt Berber. Phonology 25: 321–359.

54 

 Pavel Iosad

Smith, Norval (1999): A Preliminary Account of Some Aspects of Leurbost Gaelic Syllable Structure. In: Harry van der Hulst and Nancy Ritter (eds.), The Syllable: Views and Facts, 577–630. Berlin and New York: Mouton de Gruyter. Stewart, Thomas W. (2004): Lexical Imposition. Old Norse Vocabulary in Scottish Gaelic. Diachronica 21: 393–420. Ternes, Elmar (1980): Scottish Gaelic Phonemics Viewed in a Typological Perspective. Lingua 52: 73–88. Ternes, Elmar (2006): The Phonemic Analysis of Scottish Gaelic, Based on the Dialect of Applecross, Ross-shire, Dublin: Dublin Institute for Advanced Studies, School of Celtic Studies. Thomas, Ceinwen H. (1967): Welsh Intonation: A Preliminary Study. Studia Celtica 2: 8–28. Thomason, Sarah G. (2001): Language Contact: An Introduction, Edinburgh: Edinburgh University Press. Thomason, Sarah G. (2010): Contact Explanations in Linguistics. In: Raymond Hickey (ed.), The Handbook of Language Contact, 31–47. Oxford: Wiley-Blackwell. Trudgill, Peter (2010): Contact and Sociolinguistic Typology. In: Raymond Hickey (ed.), The Handbook of Language Contact, 299–319. Oxford: Wiley-Blackwell. Trudgill, Peter (2011): Sociolinguistic Typology. Social Determinants of Linguistic Complexity, Oxford: Oxford University Press. Vaux, Bert (2003): Syllabification in Armenian, Universal Grammar and the Lexicon. Linguistic Inquiry 34: 91–125. Wagner, Heinrich (1964): Nordeuropäische Lautgeographie. Zeitschrift für celtische Philologie 29: 225–298. Wetterlin, Allison (2010): Tonal Accents in Norwegian. Phonology, Morphology and Lexical Specification. Berlin and New York: Mouton de Gruyter. Williams, Briony (1985): Pitch and Duration in Welsh Stress Perception. The Implications for Intonation. Journal of Phonetics 13: 381–406. Woolf, Alex (2007): From Pictland to Alba. Scotland, 789–1079. Edinburgh: Edinburgh University Press. Zec, Draga (1988): Sonority Constraints on Prosodic Structure. PhD thesis, Stanford, CA. Þorgeirsson, Haukur (2013): Hljóðkerfi og bragkerfi: Stoðhljóð, tónkvæði og önnur úrlausnarefni í íslenskri bragsögu ásamt útgáfu á Rímum af Ormari Fraðmarssyni. PhD thesis, Reykjavík.

Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

Implications of language contact: Evaluating the appropriateness of borrowings in written Icelandic

Abstract: Evidence suggests that lexical purism is an integral part of Icelandic language ideologies and linguistic practices. This paper reports an investigation into how the purist Icelandic speech community evaluates lexical borrowings that do make up a part of its linguistic repertoire. In particular it was investigated whether there were generational differences in this respect by comparing the ways in which 18–21 years old students vs. a group of older adults (teachers) evaluate written texts containing such borrowings with how they evaluate parallel texts with no borrowings, for seven different genres. The results of the study show that both age groups consider it inappropriate to use borrowings in texts in formal prestigious genres, such as dissertations and books, while the use of borrowings is deemed by both cohorts to be more appropriate in informal genres such as blogs and Facebook.

1 Introduction Throughout most of Iceland’s history, the majority of Icelandic speakers in Iceland were never exposed to long-term language contact situations, the exceptions being when they were exposed to Danish when a small number of Danes came to stay permanently in Iceland in the 19th and early 20th centuries and, in World War II, when Iceland was occupied by English-speaking Allied forces. However, in late modernity, global media and migration have given rise to language and culture contacts on an unprecedented scale. These contacts have influenced Icelandic, primarily at the lexical level. As language contact situations occur in the history of a language, lexical additions can come about and surface in a number of ways, for example as semantic loans, i. e. “translations” of certain phrases or metaphors. The most easily recognizable lexical product of language contact is the borrowing of lexemes as linguistic forms. Such borrowings in Modern Icelandic, mostly from English and to a lesser extent from Danish, are the subject of this paper. The paper reports and discusses results of one part of a research project that the present authors carried out in 2011, funded by the University of Iceland

56 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

Research Fund, into the evaluation of standard and non-standard language use in written Icelandic (Kristinsson and Hilmarsson-Dunn 2013). Our research as a whole included not only speaker evaluation of borrowings but also the participants’ evaluations of standard vs. non-standard grammar features, and aspects of the combination of grammar and vocabulary evaluations. In this paper however, we focus on only one of these areas, i. e., vocabulary, and borrowings in ­particular. The aim is to investigate the question of how the allegedly purist Icelandic speech community evaluates lexical borrowings (from English and Danish) that make up a part of its linguistic repertoire. To fulfil this aim, we compared the ways in which students vs. teachers evaluate written texts containing such borrowings to how they evaluate parallel texts with no borrowings, for seven different genres. Section 2 below contains an overview of earlier research, section 3 describes the methodology of our investigation, section 4 provides and discusses the results and, finally, section 5 contains some concluding remarks.

2 Review of the literature The traditional perception that the Icelandic speech community has negative attitudes towards borrowings is corroborated by recent empirical investigations that suggest that lexical purism is indeed an integral part of shared beliefs of the Icelandic speech community as to language ideologies and linguistic practices. This assumption is, above all, based on investigations that have been carried out in the 2000s as part of a large common Nordic research project: Moderne importord i språka i Norden, ‘Modern borrowings in the Nordic languages’ (cf. e. g. www.nord. helsinki.fi/importord/projekt.htm). This included research into attitudes towards the use of lexical borrowings from English, such as in a common Nordic opinion poll, that was carried out in 2002 and was followed up by a number of in-depth interviews. The poll in Iceland showed that the more educated the respondents were, the more strongly they supported the prevailing language policy of inventing Icelandic neologisms instead of using borrowings from English (Árnason 2006: 29). This is an indication that purist language policies enjoy elite support in Iceland, at least in relation to explicitly stated attitudes. At the same time, and not surprisingly, more highly-educated respondents in Iceland claimed to use English on a daily basis, compared to less-educated respondents (Árnason 2006: 28). Thus, a typical educated Icelandic speaker is likely to read, write and speak English regularly for personal or professional purposes, while at the same time seeing herself/himself as a language purist supporting the traditional Icelandic

Evaluating the appropriateness of borrowings in written Icelandic 

 57

language ideology of using “pure” Icelandic instead of borrowings from English, at least in more formal communication situations. It should be added that, generally speaking, borrowings such as bensín ‘gas, petrol’ that have no “Icelandic” counterparts, are viewed less negatively than borrowings such as seiva ‘save’ that compete with words of Icelandic origin, e. g. vista ‘save’. Also, in general, borrowings that have been adapted to Icelandic orthographically, phonologically, and grammatically, e. g. krem ‘cream’, are viewed less negatively than borrowings such as paintball ‘paintball’ that show few or no signs of linguistic adaptation. For those media genres that are expected to use standard language, such as newspaper editorials and radio news, any considerable use of borrowings would traditionally be perceived as absolutely unacceptable in Iceland. The Nordic opinion poll mentioned above showed no significant difference between the answers of the youngest and the oldest Icelanders (Árnason 2006: 26) when asked for their opinions about the use of ‘everyday language’ in the media. All generations preferred the use of standard grammar and standard vocabulary in the media (in this case TV and radio), rather than non-standard or ‘everyday’ language. In a follow-up to the opinion poll, Óladóttir (2009) carried out a qualitative investigation, by means of in-depth interviews in 2002 and found that her 27‒36 years old respondents expressed the view that the use of borrowings and other non-standard language features undermines the credibility of the content of spoken and written texts in Icelandic (Óladóttir 2009: 121), and that the newspaper Morgunblaðið (at that time Iceland’s largest paper), as well as the Icelandic State Broadcasting Service, radio and television, should serve as models for language use (Óladóttir 2009: 119). Óladóttir (2009) concluded that the general perception is clearly that more formal situations and texts require the avoidance of borrowings (Óladóttir 2009: 121). Thus, it is not surprising that mainstream Icelandic news media, for example, avoid the use of such words. The outcome of a matched-guise study of subconscious language attitudes in Iceland (Ewen and Kristiansen 2006: 35) is very interesting in this context. This investigation was also part of the ‘Modern borrowings in the Nordic languages’ project mentioned above. The plan was to make the Icelandic respondents accept the idea that the two different texts that they listened to belonged to the radio news genre. One of the texts contained borrowings from English, such as ímeil ‘e-mail’, hakkari ‘hacker’, kreditkort ‘credit card’, laptop ‘laptop’, dánlóda ‘download’, seiva ‘save’. The other one contained Icelandic neologisms: tölvupóstur ‘e-mail’, tölvuþrjótur ‘hacker’, greiðslukort ‘credit card’, fartölva ‘laptop’, hlaða niður ‘download’, vista ‘save’ (Ewen and Kristiansen 2006: 35). Discussions with the participants in the experiments showed that a number of them found it problematic that the texts included so many borrowings since the text was supposed

58 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

to belong to the genre radio news (Ewen and Kristiansen 2006: 36). These data corroborate our assumption that borrowings are not evaluated as appropriate for a formal genre such as radio news. While Ewen and Kristiansen (2006) are reluctant to claim that their results reveal subconscious attitudes of the Icelandic participants, since some of them clearly seemed to realise what the aim of the experiment was, they could nonetheless gather important data in their experiment (Ewen and Kristiansen 2006: 36). It became evident that the “pure Icelandic” guise scored more positively than English-coloured speech for all personality traits involved: ambitious, intelligent, independent, pleasant, interesting, relaxed, trustworthy, efficient (Ewen and Kristiansen 2006: 39). One of the interesting aspects of the Nordic ‘Modern borrowings in the Nordic languages’ study was that it was able to compare results for Icelandic with the other Nordic speech communities. It was found that influence from English is generally perceived as a threat in the Nordic countries (cf. e. g. Kristiansen 2010). This was especially true for Icelanders who scored highest for (explicit) negative attitudes towards borrowings from English in the study (Kristiansen 2010). One of the conclusions of the Nordic research project as a whole was a ranking of the Nordic communities from “maximum purism” to “maximum liberalism”, with Iceland at the former extreme and Denmark at the latter (Vikør 2006: 27), notably for “conscious” attitudes. In other words, Icelanders are the most negative of all the people in the Nordic countries towards foreign language influence. In addition to data on attitudes, investigations in the ‘Modern borrowings in the Nordic languages’ project into language practices in specific genres showed extremely low frequencies of borrowings in Icelandic newspapers (borrowings from any language, not only from English), compared to the other Nordic languages (Greadler and Kvaran 2010). Overall frequency of borrowings in Icelandic newspaper language, in a corpus from 1975 and 2000, is 17 words per 10,000 words of running text, while, for example, the number is 84 for Danish newspapers, and 111 for Norwegian newspapers (Greadler and Kvaran 2010: 33). In the context of the present paper, it is of particular interest that there are clear differences between different genres of newspaper language in this respect. This is particularly the case for the Icelandic material from the year 2000: advertisements contained 6–7 times more borrowings on average than the editorial texts (Greadler and Kvaran 2010: 34–35). This suggests a negative correlation between degree of formality and frequency of borrowings in the Icelandic newspapers. However, while public debates among Icelanders tend to address English language influence negatively when it comes to borrowings from English in Icelandic language use, Icelanders are relatively tolerant as to the status of Icelandic vis à vis English in various domains of use. They consider English to be a prestigious language, it is the first foreign language they learn at school, it is the most

Evaluating the appropriateness of borrowings in written Icelandic 

 59

common language of entertainment and new media, and is increasingly used as a language of instruction in post-secondary institutions, and in academic writing in Iceland. A similar observation on Icelandic sociolinguistic and language ideology discourse has been made by Leonard and Árnason (2011: 93): Although the main issue is a status problem and centres around the value and domain of Icelandic vs. English, in public debate and folk-linguistics the discussion is mostly about form: preserving the pure form of the standard, strengthening it and getting rid of the loanwords.

According to a number of observations (e. g. Helgason 1954; Halldórsson 1971; Kvaran and Svavarsdóttir 2002; Svavarsdóttir 2003), the use vs. the avoidance of lexical borrowings is clearly a feature of variation in different registers of spoken and written Icelandic. In fact, there has been evidence of this since at least about 1700 (Kristinsson 2004:  55). In Icelandic, (near) synonyms are quite common where one word is a borrowing and the other is not a borrowing. This situation is very much the product of two competing forces in Icelandic language culture, i. e. the acceptance and adaptation of foreign linguistic forms vs. the tendency to eradicate these forms. Often a borrowing from a foreign language has come into use, and subsequently or simultaneously one or more “native” words are introduced that “compete” with the borrowing. Of course, speakers have the ability to switch between different registers and can adapt the use and non-use of borrowings to that which they find suitable for any given language situation. Thus, Icelandic speakers and writers often choose between using borrowings and using “native” words. Examples include: paintball ‘paintball’ vs. litbolti ‘paintball’; seiva ‘save’ vs. vista ‘save’, etc. Some borrowings have no equivalent Icelandic neologisms (e. g. bensín ‘gas, petrol’, sósíalisti ‘socialist’), and thus the borrowings are accepted as normal lexical choices for any register. Likewise, it is also very common that Icelanders use a neologism or an older Icelandic word exclusively, e. g. viðtal ‘interview’, ráðstefna ‘conference’, i. e. both for the most formal and for the most informal of communication situations. However, there are, as we have already mentioned, many instances of word pairs, such as seiva–vista above. The use of borrowings is a principal characteristic of non-standard Icelandic (Leonard and Árnason 2011: 94), and we find it interesting to look into the patterns of use of parallels such as seiva–vista, above, and the way the use of such pairs reveals speaker evaluation of the appropriateness of borrowings for different genres.

60 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

3 Methodology Our investigation was carried out by means of a questionnaire. Firstly the participants ‒ students and teachers ‒ were asked about their reading habits, i. e. how often they read printed newspapers, news on websites, novels, books other than novels, blogs, Facebook, and e-mail. These questions were included because reading habits may influence how people evaluate written registers. Then the participants were asked to evaluate different versions of the “same” text, i. e. one which contained 20th century borrowings from English and Danish, along with some colloquialisms, while the other contained Icelandic lexical equivalents, such as, neologisms. Quantitative and qualitative data were collected on how the two different generations of participants evaluated the suitability of the two different text versions in relation to seven written genres, covering a spectrum from rather formal (e. g., reports) to rather informal ones (e. g., Facebook). The reasons for choosing the participants and the genres are as follows: Firstly, there is the question of why different age groups were chosen for comparison. The results of the opinion poll mentioned in the previous section (carried out a decade ago), suggest that while Icelanders on the whole exhibit negative attitudes towards English borrowings, there may be some generational differences in attitudes to English, i. e. people under 30 were more likely than older people to claim that they used the English borrowing ímeil ‘e-mail’, instead of the neologism tölvupóstur; and those same people did not agree as strongly as the older participants with the claims ‘too many English words are used today’ and ‘new Icelandic words should be coined to replace the English words entering the language’ (Árnason 2006). For example, about 82 % of the Icelandic respondents in the age group 60+ agreed or agreed strongly with the claim that too many English words were used in Icelandic, while the percentage in the age group younger than 30 years old was 52 % (Árnason 2006:  24). In inter-Nordic comparisons, to be sure, these youngest participants in Iceland reveal relatively purist attitudes even if the score is lower than that for the oldest group of Icelanders. Nevertheless, there was a statistically significant difference between these age groups in Iceland, and this gives rise to the question as to whether there are different evaluations between generations in Iceland as to the appropriateness of using borrowings in different written genres. The hypothesis was that participants of different age groups would evaluate text versions with versus without borrowings in different ways. In our survey we decided to select two age groups, one cohort of 18–21 years old upper secondary school students, and another cohort of upper secondary school teachers. Most of the teachers belonged to the age group of 32–51 years. (Obviously, the two groups represent not only different age groups, but also

Evaluating the appropriateness of borrowings in written Icelandic 

 61

teachers vs. students as social groups.) The total number of participants was 123: 80 students and 43 teachers. The sample of teachers was admittedly rather small for some statistical tests on comparisons between them and the student group. Secondly, there is the question of why we were interested in looking into differences between various written genres. As previous research has shown, there is evidence of the tendency in Icelandic language use that formal situations and written genres require the avoidance of borrowings. Our assumption is that borrowings (recent English borrowings in particular) are among the most salient features of non-standard usage in Icelandic today. We would expect that speakers find borrowings less appropriate for the more prestigious, more planned, more impersonal and more edited written genres, such as textbooks, or newspaper editorials. On the other hand, borrowings might be expected to be considered appropriate in text genres such as Facebook and personal blogs, which are generally less prestigious, less planned, less edited and more personal. With regard to the results of the above-mentioned opinion poll that was carried out in 2002 (Árnason 2006), it seems somewhat problematic that participants were simply asked about the “use” of borrowings from English, i. e. without making a distinction between different communication settings or between speech and writing. In our view, this is a paramount difference in the context of Icelandic sociolinguistics. Óladóttir (2009) reports the results of in-depth interviews that she carried out with 24 people in 2002, who were 27–36 years old at the time. She finds evidence there for the claim that the appropriateness of using borrowings from English is closely connected to communicative settings, formal vs. informal, and the distinction written vs. spoken language. She claims, for example, that in answering a particular question of “using” different words, her respondents would have been (even) more positive towards choosing Icelandic words instead of borrowings from English if the question had been specifically about written language (Óladóttir 2009: 105). One of Óladóttir’s participants claims: “If I were writing I would use húðmjólk and mjólkurhristingur, but in spoken language I would say, at least here, bodylotion and shake” (Óladóttir 2009: 105). On the other hand, Óladóttir (2009) speculates that if the questions had referred exclusively to spoken Icelandic, the participants would have been more “relaxed” as to their attitudes, and that they would have felt a lesser need to use Icelandic words rather than borrowings (Óladóttir 2009: 105). It is evident that borrowings from English in Modern Icelandic informal settings, in particular among younger speakers, are taken as a linguistic resource. This is also the case e. g. in Finnish as shown by Pahta and Taavitsainen (2011: 608). They report evidence for “the increasing use of English in intranational public contexts of discourse as a resource language,” in practices such as: “quoting English phrases and idioms without explanations”, and “playing with

62 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

English segments to create humorous meanings in various printed materials targeted to local audiences” (Pahta and Taavitsainen 2011: 608). However, in Icelandic, such use of English as a resource is only regarded as a norm for the more private spheres of language use. This includes informal conversations among younger people, and informal genres such as Facebook, and similar settings. Bearing the above in mind, we inserted into one of the texts in our study the borrowings stúdíó, mixa, 3D, science fiction musical and sirka. These borrowings have found their way into Icelandic from English and Danish. Another text contained the neologisms hljóðver ‘studio’, hljóðblanda ‘mix’, þrívídd ‘3D’, vísindaskáldsögulegur söngleikur ‘science fiction musical’, and the traditional Icelandic um ‘circa’. The research design only allowed for a limited number of borrowings in the one text version, and broader communicative contexts were not directly addressed. However, since the investigation included a qualitative aspect in addition to the direct questions on the questionnaire, a number of respondents provided valuable additional information about the ways in which they evaluated language use in different settings. The participants were asked to answer questions on their perceptions of the relationship between the registers of the different text versions with certain written genres. The genre options were: report/dissertation, book, printed daily newspaper, online news, blogs, Facebook, e-mail, and, finally ‘none of the above’. Instructions were given to the participants to look carefully at the language use in each text and to tick one or more boxes (□ report/dissertation, □ book, etc.) according to the three following criteria: 1) relative to the language use in the text I would expect to see it in ...; 2) relative to the language use in the text I would consider it appropriate for ...; and 3) I would possibly write (shorter or longer) texts with this kind of language use myself if I were writing .... In what follows, we will focus on the results for the question about considering a text appropriate for a particular written genre.

4 Results and discussion We had hypothesized that speaker evaluations of different texts containing borrowings on the one hand, and no borrowings on the other hand, would be different between the two cohorts in the study, i. e., the group of 18‒21 years old students on the one hand, and the group of older adults, i. e. the upper secondary school teachers, on the other. This is because some of the research described above suggests that younger people are generally more tolerant towards borrowings.

Evaluating the appropriateness of borrowings in written Icelandic 

 63

However, most of the results of the statistical analysis showed that the use of borrowings prompted both students and teachers to perceive such texts as inappropriate in formal written genres (e. g. books, reports, printed newspapers), while both cohorts associated a text with borrowings consistently with more informal ones (e. g. blogs, Facebook). Thus, while generational differences have been reported (cf. above) in general attitudes towards English and borrowings from English, the results of our study show similarities between generations as to the written genres in which it is deemed appropriate to use lexical borrowings from English (and Danish). Our studies corroborate earlier observations, such as those by Helgason (1954), Halldórsson (1971), and Svavarsdóttir (2003), that the Icelandic speech community has shared evaluations as to the relationship between different vocabulary use vis à vis different genres. Table 1 shows the percentages of the participants who evaluated the appropriateness of using borrowings in the seven genres. There were no statistically significant differences between the two cohorts, i. e. the students and the teachers, except for printed newspapers, see discussion below. Thus, for the other six genres, only the percentages for all participants as a whole are shown in Table 1. Note that the percentages in Table 1 don’t add up to 100 since participants could tick more than one option box for each text, and they could also tick a separate box for ‘none of the above’. For example only 1 % of participants claimed that the text with borrowings was appropriate in a report, while 42 % thought that the text without borrowings was appropriate in a report.

considered to be appropriate for:

Text with borrowings

a report a book printed papers

1 % 1 % 10 % students 15 % teachers 3 % 14 % 76 % 36 % 26 %

news on web blogs Facebook e-mail

Text with no borrowings 42 % 44 % 58 % students 51 % teachers 73 % 48 % 22 % 8 % 11 %

Table 1: Evaluation of the appropriateness of two text versions for seven different genres

It is clear then that there is a general perception, both among the 18‒21 years old Icelandic speakers as well as among their teachers, that borrowings are inappropriate features in the more formal written genres.

64 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

However, although most of the results showed no generational differences there was one interesting exception, as indicated in Table 1, i. e. there was a statistically significant difference between students’ versus teachers’ evaluations of the appropriateness of the two text versions for the genre printed newspapers (χ2 = 4.9, df = 1, N = 106, p = .027). Only 3 % of teachers, as opposed to 15 % of students, found the text version which contained lexical borrowings appropriate for printed papers (the percentage for both groups, as a whole, was 10 %, as shown in Table 1). On the other hand, as to the text version with no borrowings, 73 % of teachers found it appropriate for printed papers while “only” 51 % of students did (as shown in Table 1; the percentage for both groups, as a whole, was 58 %). Thus, the teachers seem to have a clear perception that it is particularly appropriate to avoid borrowings for printed newspapers ‒ e. g. compared to news on the web, while for the students the difference between those two genres is irrelevant. We believe that we can offer an explanation for this difference between students and teachers by referring to the background data that we had gathered about different reading habits of the two cohorts. These data showed that the students read printed newspapers significantly less than the teachers do (χ² = 18.15, df = 3, N = 123, p = .000), while there were no significant differences between the groups as to the number of times they read web news. One can surmise that younger people are more inclined to read web-based news than printed news, whereas the teachers use both media to access news. Therefore, it is possible that since the students have had far less exposure to the genre printed newspapers, this may explain why the students treat both genres in the same way, while the teacher cohort makes a distinction between printed papers and web-based news. Despite the difference between the evaluation of printed vs. digital newspapers, it is clear that both cohorts, students and teachers, largely agree that borrowings are less appropriate for the more formal genres than for the more informal genres, i. e., the text versions containing borrowings were evaluated as less appropriate for the more formal genres, and more appropriate for the more informal genres. In short, both cohorts share evaluations, with the exception of printed newspapers, as above. As has already been mentioned, the text version with lexical borrowings also contained a few colloquialisms. Thus, it must be kept in mind that the participants were not necessarily responding only to the borrowings when they were responding to the questions about the appropriateness of the different text versions for different written genres. However, our qualitative data show that it was indeed the case that borrowings were the primary triggering factor for their responses. Examples of comments, that the participants wrote in order to explain why they had chosen to tick the boxes in the questionnaire, include the following (note

Evaluating the appropriateness of borrowings in written Icelandic 

 65

that the Icelandic word slettur, translated as ‘foreignisms’ below, has derogatory connotations): mikið af slettum ‘many foreignisms’, mikið um slettur ‘many foreignisms’, af því það eru slettur ‘because there are foreignisms’, of margar ensku­ slettur ‘too many English foreignisms’, mikið af útlenskum orðum ‘many foreign words’, Nota nokkur ensk orð inn á milli ‘English words are used amongst [Icelandic ones]’, því erlendu orðin og sletturnar minna á talmál ‘because the foreign words and foreignisms remind me of spoken language’. One question posed in the questionnaire was whether students and teachers would themselves write in the registers of the texts that they were presented with. As for the text version with borrowings, 60 % of the teachers ticked the box for ‘none of the options’ which indicates that the majority of teachers is reluctant to claim to use such words in any written texts, not even in the most informal genres. The students, on the other hand, were far more willing to claim that they would possibly use the borrowings when writing in one or more of the informal genres. At the same time, however, no student claimed that he or she would use these borrowings in formal genres, such as books, or reports. While we were unable to find more than one generational difference with regard to the evaluation of the appropriateness of lexical borrowings in formal vs. informal genres, i. e., for printed papers, other parts of our research project did show some generational differences, i. e. our findings suggest that some non-standard Icelandic grammar forms are less problematic to the students than to the teachers for use in the more formal genres (see Kristinsson and Hilmarsson-Dunn 2013).

5 Conclusion The results of the study reported in this paper show that both younger and older Icelandic speakers share the belief that it is appropriate to avoid borrowings in public texts in formal prestigious genres, while the use of borrowings is deemed by both cohorts as more appropriate in informal genres such as blogs and Facebook. These results corroborate previous findings, such as Óladóttir’s (2009) qualitative research, on the relationship between formal situations and avoidance of borrowings in Icelandic. However, as noted above, students were more inclined to claim to use borrowings themselves (notably in informal texts only) than their teachers were, even if a certain proportion of these same teachers had indeed evaluated texts with borrowings as suitable in informal genres. We must bear in mind, of course, that this is not direct evidence of actual use, but only claims by these speakers about their possible language use.

66 

 Ari Páll Kristinsson and Amanda Hilmarsson-Dunn

At first glance, the results of our study, that lexical borrowings are least accepted in both age-groups in the most prestigious genres of written language, may be taken to show the vitality of the protectionist language culture of Icelanders, i. e., that they show the strength of the cross-generational effects that language ideologies have on speaker evaluation of the appropriateness of different language use for different genres. If the findings are analysed in more detail however, there are other interpretation options, e. g. that the decisive factor is not necessarily any anti-foreign language ideologies, but rather some subconscious, mechanical, stylistic preferences, and adherence to what both younger and older Icelanders perceive as a traditional and familiar relationship between registers and genres.

References Árnason, Kristján (2006): Island. In: Tore Kristiansen and Lars Vikør (eds.), Nordiske språkhaldningar. Ei meiningsmåling, 17–39. Oslo: Novus Forlag. Ewen, Halldóra Björt and Tore Kristiansen (2006): Island. In: Tore Kristiansen (ed.), Nordiske sprogholdninger. En masketest, 33‒48. Oslo: Novus Forlag. Graedler, Anne-Line and Guðrún Kvaran (2010): Foreign influence on the written language in the Nordic language communities. International Journal of the Sociology of Language 204: 31–42. Halldórsson, Halldór (1971): Nýyrði frá síðari öldum. In: Baldur Jónsson (ed.), Íslenzk málrækt, 212–244. Reykjavik: Hlaðbúð. Helgason, Jón (1954): Hrein íslenzka og miður hrein. Språkvård. Redogörelser och studier utgivna till språkvårdsnämndens tioårsdag 1954. Skrifter utgivna av nämnden för svensk språkvård 11, 95–114. Stockholm: Svenska bokförlaget. Kristiansen, Tore (2010): Conscious and subconscious attitudes towards English influence in the Nordic countries: evidence for two levels of language ideology. International Journal of the Sociology of Language 204: 59–95. Kristinsson, Ari Páll (2004): Offisiell normering av importord i islandsk. In: Helge Sandøy and Jan-Ola Östman (eds.), “Det främmande” i nordisk språkpolitik. Om normering av utländska ord, 30–70. Oslo: Novus Forlag. Kristinsson, Ari Páll and Amanda Hilmarsson-Dunn (2013): Evaluation of different registers in Icelandic written media. In: Tore Kristiansen and Stefan Grondelaers (eds.), Language (De)standardisation in Late Modern Europe: Experimental Studies, 331–354. Oslo: Novus Press. Kvaran, Guðrún and Ásta Svavarsdóttir (2002): Icelandic. In: Manfred Görlach (ed.), English in Europe, 82–107. Oxford: Oxford University Press. Leonard, Stephen Pax and Kristján Árnason (2011): Language ideology and standardisation in Iceland. In: Tore Kristiansen and Nikolas Coupland (eds.), Standard Languages and Language Standards in a Changing Europe, 91–96. Oslo: Novus Press. Óladóttir, Hanna (2009): Shake, sjeik eller mjólkurhristingur? Islandske holdninger til engelsk språkpåvirkning. Oslo: Novus Forlag.

Evaluating the appropriateness of borrowings in written Icelandic 

 67

Pahta, Päivi and Irma Taavitsainen (2011): English in intranational public discourse. In: Bernd Kortmann and Johan van der Auwera (eds.), The Languages and Linguistics of Europe. A Comprehensive Guide, 605–619. Berlin and Boston: Mouton de Gruyter. Svavarsdóttir, Ásta (2003): Ordbogen og den daglige tale. In: Zakaris Svabo Hansen and Anfinnur Johansen (eds.), Nordiske studier i leksikografi 6, 43–48. Tórshavn: Nordisk forening for leksikografi. Vikør, Lars (2006): Language purism in the Nordic countries. International Journal of the Sociology of Language 204: 9–30.

John Weinstock

At the frontier: Sámi linguistics gets a boost from outside Abstract: The question of the origins of the Sámi people of Northern Scandinavia and Russia has long been contentious. How did they come to speak a Finno-Ugric language? A promising explanation has recently been put forth by Ante Aikio, viz. that speakers of Proto-Sámi merged with and assimilated ancient hunter-gatherers who had long inhabited the interior of the Scandinavian peninsula and whom Aikio calls Palaeo-Europeans. The present paper finds anthropological, archaeological and genetic evidence that helps corroborate Aikio’s theory. There are numerous examples of mutually beneficial contacts between Sámi ancestors and other dwellers on the Scandinavian peninsula; for example, the Sámi produced commodities that were in demand among Germanic (Norwegians, Swedes) peoples living in Scandinavia. The Sámi siida system of relatively small groups opened them to outside influence from Proto-Sámi speakers. Finally, phylogenetic evidence confirms the widespread contacts between the Sámi and others as well as the mutual influence between groups.

1 Introduction The origin of the Sámi1 languages has baffled scholars for more than a century (cf. Weinstock 2010: 30). How did the hunter-gatherer groups of the Scandinavian Peninsula come to speak a Finno-Ugric tongue? Did their ancestors migrate in from the Volga-Ural region of Russia? What sequence of events led to the ethnic differentiation of the Sámi from neighboring peoples? These are but a few of many enigmas that must be addressed if one is to come to grips with Sámi prehistory. Ante Aikio published an eminently plausible proposal in a series of papers (2004, 2006, 2007): according to him, phonological innovations in Pre-Saami (PreS) led to Proto-Saami (PS) near the Gulf of Finland at the onset of the Iron Age (ca. 500 BCE).2 Then, speakers of PS moved north and northwest assimilating Palaeo-Eu-

1 The ethnonym Sámi is also spelled Sami and Saami. I use Sámi unless a particular source has it otherwise. 2 Proto-Saami is the hypothetical shared ancestor of today’s Saami languages. Pre-Saami is an earlier stage thereof, on a par with Pre-Finnic, both of which are descendants of (Proto-)Fin-

At the frontier: Sámi linguistics gets a boost from outside  

 69

ropean3 bands and replacing the latter’s languages with PS throughout the interior of Scandinavia by the beginning of the Common Era. Over the following five centuries or so PS disintegrated into early versions of today’s Sámi languages, plus others that are now extinct. The unknown languages that disappeared left traces in PS and eventually in the Sámi languages. The goal of this paper is to examine Aikio’s theory and see whether there might be anthropological, archaeological and/or genetic evidence to support or refute it. One could conceive of Sámi (pre)history writ large as three periods from the end of the last glaciation in Scandinavia up to the present. Stage one: Palaeo-European languages are spoken throughout the Scandinavian peninsula (and elsewhere); stage two: PS supplants the Palaeo-European languages in Fennoscandia and evolves into the Sámi languages (ca. 500 BCE to ca. 500 CE); and stage three: the Sámi languages in turn are gradually replaced by Finnish, Norwegian and Swedish in many areas, especially in the south and along the coasts (500 CE to the present). Where did these Palaeo-European languages come from? During the Last Glacial Maximum humans vacated Fennoscandia and sought shelter in relatively warmer areas in southern Europe such as the Franco-Cantabrian refuge in the Pyrénées, the Balkan Peninsula and the Black Sea. When the glaciers began to shrink and the climate became more accommodating, flora, fauna and humans returned to the Fennoscandian peninsula thousands of years ago. The humans came through whatever corridors were open. That they were not a uniform population can be inferred from their different origins and seen in genetic profiles of present-day Sámi, which vary considerably from northwest to southeast (cf. Weinstock 2010: 39; Huyghe et al. 2011: 351 suggest that the Sámi are not a homogeneous population). And so late Ahrensburgians moved north along the west coast of Sweden (Hensbacka culture – Schmitt 2007, Schmitt et al. 2009) and the Atlantic coast of Norway (Fosna culture – Bjerck 2008) ~11500 BP or so. Foragers arrived at the coasts of Eastern Finnmark and Finnish Lapland from the western coast of Norway – the Komsa culture (late Ahrensburgian) – and from northwestern Russia (Post-Swiderian) as far back as the Early Mesolithic – ca. 11000 BP or even earlier (Olsen 2003; Rankama and Kankaanpää 2011). Thus, humans have been in Fennoscandia many millennia.4 These were people who spoke the

nic-Saamic. Aikio recently suggested a later date for Proto-Scandinavian loans in Saami. Aikio (2012): 76. 3 In the 2004 paper Aikio calls the unknown languages Palaeo-Laplandic. 4 The Preboreal encounter between settlers of different origins – the Komsa (Ahrensburgian) culture and the Post-Swiderian culture – on the shore of the Varanger Fjord gives an idea of how rapidly humans reached much of Fennoscandia.

70 

 John Weinstock

Palaeo-European languages. Yet nothing is known about these languages other than through contact with languages that have survived.

2 Substrate and Proto-Sámi Substrate influence on intrusive languages is not uncommon but it is often fraught with controversy (cf. Polomé 1990 and Salmons 1992, 2004). To begin with, Thomason and Kaufman (1991: 111) write: “Unless that language is still spoken in some other area(s), or its structure is known from historical documents or by inference from related languages, we cannot study its possible effects on the language that replaced it”. Of course, absolutely nothing is known about the Palaeo-European languages of Fennoscandia, except through their effects on other languages (cf. also Saarikivi 2004a: 189). As Aikio (2004: 7) phrases it: “the extinction of the unknown languages has created an etymological event horizon beyond from which no positive evidence can be retrieved”. The influence of receding languages can come in all aspects of grammar: syntactical, phonetic, lexical; however, with no knowledge of one of the languages in a contact situation, the most likely area of influence is lexical, the aspect of language where there is some hope of success. That is certainly the case with Palaeo-European influence on Proto-Sámi. Contact between the two groups was likely peaceful; the immigrating PS speakers had something to offer the Palaeo-European bands, viz. a viable economic model with furs and other hunting products playing a vital role; and the individual Palaeo-European groups were very small, as we shall see later on. Another issue to keep in mind is that many potential substrate items may have to be rejected because their structural features are too close to the intrusive language. What is needed – as Aikio shows – are words and names that do not fit the regular patterns of the intrusive language. Aikio focuses on lexical items. He has been careful to a fault in creating a rigorous methodological framework to deal with vocabulary that has survived the demise of languages.5 His four criteria are: a) the number of substrate items must be significant, “several hundred basic stems” at a minimum; b) words of unknown origin are mainly in semantic fields liable to substrate influence (e. g. words for animals, topography, weather phenomena and toponyms); c) words in question often show non-native structural features such as phonotactic combinations that are of secondary origin; d) “Some of the words show irregular sound correspondences between languages or dialects, indicating separate borrowings

5 A variation of the Polomé/Salmons criteria (1992: 267).

At the frontier: Sámi linguistics gets a boost from outside  

 71

which reflect different sound substitution strategies or distinct but related source idioms” (Aikio 2004: 9). Moreover, place names must be treated just as ordinary loanwords, because they are likely to be borrowed during language shifts; i. e., “If there is a significant substratum in the lexicon, it is predictable that toponyms have also been extensively borrowed” (Aikio 2004:  9; cf. also Saarikivi 2004a: 209). Aikio’s examples easily fulfill the first two criteria. He lists 161 substrate items – mainly from North Sámi and not including toponyms – a figure that could be substantially increased by including more Sámi languages, plant names, and reindeer vocabulary (Aikio 2004: 12). With regard to c), many words do have phonotactic structures suggesting non-native origin – he has four groups of these; some examples are three-consonant cluster words such as North Sámi (SaaN) gumpe ‘wolf’ < PS *kumppi and the initial cluster sk- in skuolfi ‘owl.’ An example of d) would be inexplicable irregularities between dialects such as reaš’ši ~ reakčá ‘flat seashore with a clay bottom’ (Aikio 2004:. 14). The Sámi lexicon “shows consistent signs of heavy substrate influence” (Aikio 2004:  16). Not surprisingly, Sámi place names include many stems of unknown origin. Aikio includes nearly two hundred names of mountains and highlands in North Sámi alone.6 It might be useful to view Aikio’s theory vis-à-vis Theo Vennemann’s Vasconic substratum ideas for much of prehistoric Europe with Basque as the only survivor from this patchwork of European languages (Vennemann 2003). Vennemann’s explanations cover vast territory and go back to the eighth millennium BCE. And he sees influences on all grammatical levels. This makes many individual proposals difficult to verify or disprove. Aikio on the other hand treats a much smaller, more recent time frame and his theory focuses primarily on lexical matters (ordinary names and toponymy) where success is most likely. Baldi and Page (2006) review Vennemann meticulously.

3 Sámi vs. Others A symbiotic relationship between Sámi (and their ancestors) and Others (Finns, Norwegians, Swedes and their ancestors) has been the norm for thousands of years, lasting until relatively recently when efforts began to assimilate the Sámi into the majority cultures of Scandinavia. There have been frequent contacts

6 Salmons, in an article in the Koivulehto Festschrift where Aikio’s 2004 paper appeared, mentions an alternate possibility for the source of possible substrate lexical items, viz. autochthonous development, though that would, if anything, seem only to affect a small minority of the items Aikio discusses. Cf. Aikio 2004: 19–20.

72 

 John Weinstock

between these groups involving the exchange of goods, ideas, marital partners, genes (admixture), and aspects of language, all confirmed in the anthropological, archaeological, genetic and linguistic records.7 This is not to say that there was no conflict between groups, but conflict was not the rule. Linguistic interference extends in both directions and involves many aspects of grammar and lexicon. Jurij Kusmenko (2008: 36–42) attributes the loss of prefixes in Common Scandinavian (ca. 550–1050 AD) to Sámi interference. Moreover, according to Kusmenko the *s(k) suffix in modern Scandinavian languages may stem from Sámi influence, e. g. setjast ~ setja sig vs. SaaN –sit as in geassásit ‘ to be pulled’ medio-passive vs. geassádit ‘to withdraw’ (Kusmenko 2008:  43–67). Kusmenko also mentions suffixed negation, suffixation of pronouns and much more as possible Sámi influences on Scandinavian during the Common Scandinavian period (Kusmenko 2008: 68–214). He proposes compound prepositions, open syllable consonant lengthening, vowel balance and leveling as possible Sámi influence on the Scandinavian languages during the 12th to 16th centuries (Kusmenko 2008: 217–315). Adam Hyllested (2008: 133–35) discusses the paucity of Sámi loanwords in Old Norse but goes on to add some new etymologies, e. g. ON lundi ‘puffin’ < *londe > SaaN lodde, loddī. Earlier, numerous words were borrowed from PIE (and daughter languages) during previous stages of Finno-Ugric and Uralic continuing up to the present day. According to Pekka Sammallahti, Sámi and Samoyed, for example, share about one hundred stems going back to Proto-Uralic spoken ca. 4500 BCE including ten or so whose ultimate source is PIE (1998: 118). A couple of examples are SaaN borrat ‘to eat’ and SaaN guolli ‘fish.’ He also mentions that there are 550 ordinary Sámi stems that seemingly have no etymology, many of them common words, e. g. SaaN čáhppat ‘black’ and SaaN ravgat ‘to fall.’ Could some of these be substrate items?

4 Pre-Sámi – Proto-Sámi Where might Pre-Sámi have evolved into Proto-Sámi – at least on the basis of the linguistic data? Two layers of Germanic loanwords in Sámi are relevant to Aikio’s theory: 1) the early Iron-Age vowel shift PreS *a > *õ (PS *uo) – in this layer Gmc. *h becomes Sámi *k, e. g. SaaN guos’si ‘guest’ < PS *kuossē < PreS *kansa < PGerm

7 Since the ice sheet melted there have been periodic contacts with groups outside of Fennoscandia, cf. loanwords in Sámi discussed below or trading with metal-producing agricultural societies to the east after 3800 BP.

At the frontier: Sámi linguistics gets a boost from outside  

 73

(Proto-Germanic) *hansō- (> Old English hōs ‘host’); 2) a younger layer of loanwords also shows PreS *a > *õ (PS *uo), but here Gmc. *h > Ø, e. g. SaaN vuoksa ‘depth of fishing net’ < PS *vuopsë < PreS *api/as < NWGerm *hāba-z (cf. Old Norse háfr ‘hoop net’). Loanwords in the first layer are often shared with Finnic, none though from the second layer. Aikio adds 18 new etymologies of lexical items deriving from Gmc., e. g. SaaN deahkki ‘thick meat, muscle’ < PS *teakkē < PreS *tekkā (Eng. ‘thick’), making 43 in all. PreS *a > *õ (> PS *uo) implies that Gmc. *ē > *ā had already occurred in the northern dialects of Proto-Gmc. This contact took place either at the beginning of the Iron Age or shortly thereafter. The Germanic and Baltic languages came in contact with PreS across the waters of the Baltic and the Gulf of Finland. The southern shore of the Gulf of Finland enabled “direct contact with Proto-Finnic” (Aikio 2006: 42) as well. In addition to Germanic influence there was a smaller number of direct loans from Baltic to Sámi such as (e. g., biebmat ‘to feed’), though most Baltic loans in Sámi were mediated by Finnic. The many Pre-Finnic loans in Sámi imply contact near the Gulf of Finland too. Aikio (2006: 45) concludes that Pre-Sámi becomes Proto-Sámi between Germanic and Finnic groups to the south and Palaeo-European bands to the north. He cites additional evidence for the presence of PS in the far south of Finland: “Proto-Sámi emerged somewhere within the area where numerous Sámi substrate toponyms are attested, stretching from southern Finland in the west to the great lakes Ladoga and Onega in the east” (Aikio 2006: 42). This scenario is plausible. There are, all the same, a couple possible problems: Janne Saarikivi (2004b) suggests that Pre-Sámi formed south of the Lakes Ladoga and Onega. This would work if one were to assume that the speakers of Pre-Sámi moved northwest – perhaps via the Karelian Isthmus – to the area north of the Gulf of Finland before the disintegration into Proto-Sámi. Another question: where were the Finnic speakers when the second stratum of lexical items was borrowed by Proto-Sámi from Germanic but not by Finnic? Perhaps the Proto-Sámi speakers had moved further west? This issue is addressed below.

5 Sámi ethnogenesis The frequent contact over many centuries can be demonstrated, but when did the Sámi become Sámi? Sámi language ⟺ Sámi ethnicity [comment: the symbol ⟺ is mathematical, meaning each implies the other, i. e. Sámi language implies Sámi ethnicity and the reverse], at least at some point in time. A number of cultural-historical models have Sámi ethnogenesis occurring toward the end of the last millennium BCE when trappers/foragers began to distinguish between ‘us’ and ‘them,’ the latter often farmers. In other words, ethnic boundaries emerged

74 

 John Weinstock

through social interaction and a concomitant objectification of cultural practices.8 Some scholars place greater emphasis on ethnicity: Knut Odner, Lars Ivar Hansen and Bjørnar Olsen; others on archaeology: Christian Carpelan; and others on linguistics: Ante Aikio. Agriculture arrived in the Baltic between 5500 and 4200 BP and Pre-Finnic farmers moved into Finland near the sea.9 Aikio sees Proto-Sámi developing between the Germanic and Finnic groups with their redistributive economy to the south and the unknown Palaeo-European mobile and egalitarian cultures to the north, i. e. in the southern parts of present-day Finland and Karelia. Following Fredrik Barth (1969), Odner in his 1983 Finner og terfinner offered an explanation for the emergence of Sámi ethnicity which fits quite well with Aikio’s ideas, albeit from a social anthropological perspective. Odner sees the ethnic standardization of the hunter population occurring in Southern Finland; he dates it somewhat later than Aikio, at the beginning of the new era 0–200 CE. At this time, according to Odner, the Baltic/Gulf of Finland was an important transit area for the fur trade from the north. Odner has three phases: 1) the hunter-gatherer population adopts the language of the Finns as a way of signaling solidarity and gaining access to the farmers’ products. Some hunter-gatherers also adopt the lifestyle of the peasants, whereas others adopt their language but remain trappers (Odner from Hansen and Olsen 2007: 32–33); 2) Sámi become dependent on products from outside their region and specialize in producing goods the Finns want, primarily furs and other hunting products. Odner would have this occur in the 3rd or 4th century of the new era; 3) this ethnic package spreads among other foraging groups to the north. With their new Proto-Sámi language and an economic model (furs in exchange for other goods) they head north and assimilate the Palaeo-European peoples already in Fennoscandia.10 The trappers speaking Palaeo-European languages see the advantages of the new model, which they adopt along with the PS language. The trapping population of Northern Fennoscandia actually chooses to adopt Sámi ethnicity because it is economically advantageous for them.11 By reason of this choice they are defined as desirable trading partners and gain access to products controlled by surrounding trapping populations. The economic advantages of participating

8 For a more thorough discussion cf. Olsen 2003: 12–13. 9 There is no general agreement as to whether agriculture came to Scandinavia through cultural or demic diffusion; however, a recent genetic study by Skoglund et al. (2012) found the DNA of a single ~5000-year-old Scandinavian farmer to be most similar to the DNA of extant southern Europeans. 10 The Palaeo-European groups were probably organized in siidas, i. e., small semi-nomadic groups living in relatively discrete territories where they hunted and fished. Cf. further on. 11 Perhaps farming farther inland was undesirable at this point in time.

At the frontier: Sámi linguistics gets a boost from outside  

 75

in the fur trade lead to the rapid spread of Sami ethnicity over much of Northern Fennoscandia. Odner thus suggests that Sami ethnicity did not evolve as a result of migration from the east, but as a result of social and economic contacts between the trapping population and, in the first instance, a Finnish peasant population.12 Odner states (1983: 110): We can imagine that groups which have not yet accepted Sami ethnicity would observe successful transactions between neighboring groups and Finnish traders – transactions that were conditional on signaling Sami ethnicity. By modeling their own social practice accordingly – that is, by adopting Sami ethnicity themselves – they would similarly gain access to the same types of transactions and products [my translation].

Odner rejects the hypothesis that the establishment of Germanic ethnicity along the coast of Northern Norway up to Tromsø can be explained in terms of migration from the south. Hansen and Olsen, however, see things differently. They discuss the social stratification occurring along the coast of Western Norway whereby the original, relatively egalitarian foraging communities become hierarchically organized as the economy is restructured with farming as a central element. By the 3rd century of our era Germanic ethnicity had reached the far north, and Lyngen Fjord (Nord Troms) became the cultural and ethnic boundary between Germanic and Sámi groups: Iron Age sites connected to Sámi settlement are north of Lyngen Fjord: (Germanic) burial mounds, cairns, longhouses, and boathouses are southwest of this border; to the northeast are (Sámi) slab-lined pits, stone structures, scree graves, and circular houses, with many sites just north of this border.13 These mostly coastal developments probably had little effect on the shift to the Sámi languages in the interior of Fennoscandia as proposed by Aikio. They also discuss the cultural differentiation between coastal Sámi in northern Norway (from Sør Troms southward) and in Norrland on the Bothnian coast of Sweden on the one hand and the Sámi in the interior and northeast portions of Fennoscandia. The trapping communities become increasingly interested in the interior and its resources and the relatively extensive contacts with Eastern Russian metal-producing communities, primarily the Ananjino Culture (800–200 BCE).14 This contact was probably based on the trading of hunting products,

12 There were migrations of Finno-Ugric-speaking groups from the east and southeast into the Fennoscandian Shield as far back as the Comb Ceramic or even earlier. Cf. the section on genetic influence. 13 For more insight into the cultural encounter in Northern Norway cf. Hansen and Olsen 2007: 58–93; Olsen 2003: 19–21 and Odner 1992: 25–26. 14 Eastern contacts, of course, go back at least to 1800 BCE as evidenced by asbestos pottery and bifacial arrow points.

76 

 John Weinstock

especially furs, for metals. “Contact between the northern trapping communities and eastern metal-producing peasant communities may have resulted in the ‘discovery’ of distinctive cultural characteristics and differences, thus stimulating new ways of articulation. […] The arrival of uniform terms of material expression, especially Kjelmøy pottery, which was used in much of the Sami area in the last millennium BCE, can be interpreted as a form of objectification or stereotyping of cultural distinctiveness and difference from the others” (2007: 38–40) [my translation]. In other words, Kjelmøy pottery signals ethnic identity. But how does this ethnic identity spread throughout inland areas of Scandinavia and what role does Sámi language play? Hansen and Olsen (2007: 133) suggest that the language may have been a unifying and contrasting trait. They support the idea that “det felles samisk-finske urspråket […] ble splittet opp […] i siste halvdel av 2. årtusen f.Kr.” – “the common Sámi-Finnish proto language […] split up […] in the latter half of the second millennium BC” [my translation], and that this “proto language” may have been spoken throughout much of northern Fennoscandia, including the later “Germanic” coastal areas. They add that the “linguistic parting of ways may have occurred […] during the last millennium BC” (Hansen and Olsen 2007: 134). Though Hansen and Olsen’s dates are in the same general range as Aikio’s they do not put forward the south-to-north dynamic process Aikio has. (Cf. Odner 1992 for his views on earlier versions of Hansen and Olsen’s ideas.) The archaeologist Christian Carpelan (2003: 86–87) suggested that small Proto-Sámi-speaking communities moved north and settled among the original inhabitants of the area, but that the “final” Lappicisation would only have occurred when Kjelmøy Ware spread over Lapland at the beginning of the Iron Age (650 BCE). Aikio finds the possible connection between the spread of Kjelmøy Ware and the expansion of Proto-Sámi intriguing. Kjelmøy ceramics, in use from about 700 BCE–300 CE are quite common in the northern dwelling and iron manufacture places and in much of the Sámi area. Aikio notes that little can be said about the causes of the Proto-Sámi expansion though he alludes to the Sámi exploiting their position between Germanic and Finnic cultures to the south and the Palaeo-European peoples to the north, with fur trade possibly being the key.15 Hence, the chains of events proposed by Odner, Hansen and Olsen, and Carpelan each have aspects that fit quite well with Aikio’s hypothesis.

15  The fur trade was essential for the Cree in Canada.

At the frontier: Sámi linguistics gets a boost from outside  

 77

5.1 Sámi siida system If Proto-Sámi replaced the Palaeo-European languages as sketched above, there may be genetic evidence in today’s Sámi. One would expect the genetic makeup of the Proto-Sámi bands to change gradually as they moved from southeast to north and northwest16 and mixed genes with the previously relatively isolated Palaeo-European groups they subsumed, i. e. gene flow should have occurred along the same axes as language flow.17 Early in the Iron Age these Palaeo-European groups each occupied mostly discrete territories in the interior of Fennoscandia and Kola in which they hunted and foraged.18 This was the so-called siida system, still extant after overhunting depleted the wild reindeer stock many centuries later.19

Fig. 1: Northern and eastern siida areas during the hunting culture20

16 The archaeologist Lou Schmitt points out in a personal communication that early visitors to Fennoscandia used skin hull boats as far back as 9,300 BP. In other words, humans were already quite mobile long before the onset of the Iron Age. 17 A recent (February 8, 2012) note in Cell Press mentions an article by C. Michael Barton and Julien Riel-Salvatore where their computer simulations suggest Neanderthals and early humans mixed genes, resulting in a hybridization of the two species. In other words, hybrid populations still carry Neanderthal genes and mixing occurred despite possible social barriers. 18 “Exploitation of various ecological zones in Northern Sweden is intensified during the Iron Age, with a gradually more limited territorial structure. Large, loosely defined territories are replaced by smaller, more clearly defined trapping areas. This territorial restructuring probably formed an important basis for the historically known division of the Sami area into siida territories” [my translation] (Hansen and Olsen 2007: 93). 19 Cf. Saarikivi’s Karte 17 for the distribution of the PS form of siida: *sijtę ‘(Winter)dorf’ [(winter) village] (2006: 211). 20 The map was prepared by Ørnulv Vorren on the basis of his own investigations and other written sources, cf. Vorren (1989). Reproduced with permission of Norsk Folkemuseum.

78 

 John Weinstock

A closer look at the siida system may help provide justification for Aikio’s model of language replacement. The map above shows the siida boundaries before national borders were set, in other words, during the hunting period (veidekulturen) and well before nomadic reindeer herding on a large scale emerged. Ørnulv Vorren (1989: 14) points out that in Fennoscandia the distances people had to migrate to exploit different food sources – animals and birds on land and in the sea – were considerably shorter than in the rest of the circumpolar north. If one food source failed others were not far off. The territory was divided up along what one today would call ecological principles. Since fishing was important during most periods in the annual cycle, the hunting/gathering areas were organized around rivers and inland lakes that were navigable summer and winter. These areas were the siidas. The boundaries between individual siidas were the watersheds between the watercourses. The Finnish scholar Helmer Tegengren studied in great detail the siidas in Kemi lappmark (northeastern Finland). According to him the typical siida was formed around 8–11 hunters plus their families (cf. Tegengren 1952). The map above shows only some of the Sámi siidas; not included are the areas all the way south to Kokemäki/Kumo in Finland, to Härjedalen in Sweden, and to Hedmark in Norway. Sámi living in a given siida moved seasonally from one part to another depending on where the game was. There is no reason to assume that these siida boundaries varied much for centuries, at least until the wild reindeer no longer played such a key role by the 17th century. If one extrapolates back to the Iron Age the situation was likely quite comparable: the small number of Palaeo-European speakers in each siida was surely easily influenced by the Proto-Sámi bands moving up from the south. And the siidas are clearly one reason that today’s Sámi languages are mutually intelligible only when they are in relatively close proximity to one another. For example, North Sámi speakers can communicate with Lule Sámi speakers just to the south; however, North Sámi and South Sámi are not mutually intelligible due to the greater distance between them. Members of the siidas of old did not have much contact with other speakers of “Sámi” except for those who lived nearby.21

6 Sámi genetic profile This would suggest that Palaeo-Europeans in a given siida had similar genetic profiles before contact with Proto-Sámi. A fine example of this somewhat limited mobility was noted by Einarsdottir et al. (2007: 177) in discussing the genetic

21 And, of course, middlemen or traveling merchants.

At the frontier: Sámi linguistics gets a boost from outside  

 79

population structure of northern Sweden among Sámi, Swedes and Finns: They saw “a tendency towards marriages between individuals born in the same river valley region, at least until the mid 20th century,” a faint reflection of the siidas of old, though with three ethnic groups represented. Many studies have confirmed that the Sámi are genetic outliers in Europe; nevertheless, their genetic makeup is primarily European. Various reasons have been given for this, including long-term isolation, genetic drift, bottlenecks and small population size (cf. Tambets et al. 2004). In two papers from 2010 and 2013 I explained that Sámi mtDNA and Y-chromosomal22 data is characterized geographically by a heterogeneous pattern: the Sámi genetic profile in the northwest is very different from that in the southeast. A number of very recent genome-wide studies confirm that genetic similarity between population pairs correlates with geographic distance (Huyghe et al. 2011; Salmela et al. 2011; Nelis et al. 2009; Lao et al. 2008). Nelis et al. (2009: 8) used PC (principal component) analysis on a mostly Finno-Ugric population in Estonia where there are no geographical barriers. Individuals in the study were grouped by county of birth. The authors found that “the resulting genetic map correlates almost perfectly with the geographic map.” Lao et al. (2008: 1245) studied 23 European subpopulations including Uppsala (Sweden), Førde (Norway) and Helsinki (Finland): “Although the amount of differentiation within the European autosomal gene pool was found to be small, the existing genetic differences nevertheless correlated well with geographic distances.”

7 Caveat Today’s Nordic peoples are defined primarily on the basis of ethnicity, especially mother tongue.23 Genetic analyses of Scandinavian Sámi mostly sample only those who identify themselves as Sámi. This means that all Sámi who were assimilated in the 19th and 20th centuries or earlier are unintentionally excluded in most genetic studies. Not included are the many “Sámi” who were assimilated by the nation-states in the 19th and first half of the 20th century and who left their Sámi ethnicity behind – but not their genes – when they moved away to metropolitan

22 Maternal and paternal DNA respectively. 23 The Sámediggi (Norwegian Sámi parliament) has the following definition for the ca. 70,000 Sámi in Sápmi: “Everyone who declares that they consider themselves to be Sámi, and who either has Sámi as his or her home language, or has or has had a parent, grandparent or g ­ reat-­ grandparent with Sámi as his or her home language, or who is a child of someone who is or has been registered in the Sámi census, has the right to be enrolled in the Sámi census in the municipality of residence.” Cf. http://www.galdu.org/govat/doc/eng_sami.pdf

80 

 John Weinstock

areas such as Helsinki, Oslo or Stockholm: do the genetic analyses of the indigenous and majority populations take this into account and, if so, what role does this factor play in the results? It would seem evident that the extent of admixture in the Nordic countries has been significantly underplayed. This could be an issue too for other indigenous peoples who have been subjected to assimilation.

8 Haplogroup Z Aikio’s (2006: 42) time frame implies that Proto-Sámi emerged in an area “stretching from southern Finland in the west to the great lakes Ladoga and Onega in the east” which enabled contact with Proto-Finnic spoken along the southern shore and at the bottom of the Gulf of Finland. Recent genetic studies suggest a migration into this area at about the same time. Ingman and Gyllensten (2007: 118) showed that mtDNA haplogroup Z is found in Sámi and Northern Asian groups but is virtually absent in Europe, and that “some Sámi [and Finnish] lineages shared a common ancestor with lineages from the Volga-Ural region as recently as 2700 years ago” (Ingman and Gyllensten 2007: 115), especially subhaplogroup Z1a.24 In an earlier study Meinilä et al. (2001: 166) report that Finnish Sámi have a frequency of 4.3 % Z and the population of northern Finland has 3.9 % Z. Huyghe et al. (2011: 347) found “an east Asian contribution to the predominantly European-derived Sámi gene pool” in Finland to be 6 %, though they do not further specify the east Asian population. Salmela et al. (2011: 8), referring to an earlier study (2008) that found more similarity of Eastern Finns to East Asians among North European populations, write: “the observed affinity to Eastern Asia would not be mediated by contacts with Russians but could reflect an ancient eastern influence predating the arrival of Slavic populations to Northeastern Europe in the end of the first millennium.” That this area was inhabited by Finno-Ugrian groups is confirmed by Slavonic chronicles and Scandinavian sagas (Saarikivi 2004a: 192). Saarikivi says that even today there are residents of the Dvina basin who “consider themselves descendants of the čud, a tribe mentioned several times in the historical record” (Saarikivi 2004a: 192).25 He also suggests the area from Beloozero southwest to Tver and Novgorod as somehow central to the formation of the Baltic Finnic and Sámi languages (Saarikivi 2006: 223). Aikio writes: “Various criteria suggest that the ultimate origin of the Sámi language branch

24 A haplogroup is a group of similar haplotypes (combinations of DNA sequences transmitted together) that share a common ancestor. 25 Made famous in the Sámi film Ofelaš (Pathfinder). Cf. also Lehtola 2004: 23.

At the frontier: Sámi linguistics gets a boost from outside  

 81

lies somewhere in present-day Russia” (Aikio 2007: 192). In other words, perhaps Pre-Sámi emanated from there. Ingman and Gyllensten’s dating of 2700 years ago is tantalizingly close to the beginning of the Iron Age and the spread of Kjelmøy pottery throughout Scandinavia.

Fig. 2: Possible route for Pre-Sámi into Finland

Were these carriers of Z1a perhaps the middlemen who not only traded in iron wares, Kjelmøy pottery and furs but who brought their Pre-Sámi language west with them (see map above)?26 Huyghe et al. (2011: 347) find that for the Finnish Sámi, the fixation index, FST, a measure of genetic distance between populations, was smallest when compared with the Russians, smaller even than nearby northern European groups. They found further confirmation of an East Asian contribution to the Sámi gene pool in the Yakuts of Northeast Siberia; however, they do not sample any West Asian groups likely to be closer to the Sámi. Another study assumes that the northern and southern Swedish Sámi populations are descendents of an admixture event between “the ancestral European and ancestral Asian populations” (Johansson et al. 2008). They find the Asian contribution to the Swedish Sámi gene pool to be 13 % as well as recent admixture between the southern Sámi and the non-Sámi Swedish population. Although the frequency of Z in Scandinavia is low, the numbers are revealing: Ingman and Gyllensten give 7.2 % (mtDNA) for Finnish Sámi, 10.9 % for Southern Swedish Sámi who are in traditional Sámi occupations (such as reindeer nomadism), Northern Swedish Sámi 0.7 %, and 0 % for Europe. Lappalainen et al. (2008: 342) have 2.6 % and 1.1 % for Ingrian and Viena Karelian respectively in Russian Karelia.27 There seems to be a cline (geographical gradient) running from higher Z in the south and east to lower

26 Lou Schmitt mentions that ca. 5000 BP members of the Pitted Ware culture based in southern Sweden traded in high quality flint (for axes) and most likely used horses for transport. 27 Ingrian is southeast of the Baltic Sea and Viena Karelian is just above the middle of Karelia.

82 

 John Weinstock

Z further north as might be expected considering where Z came from to Scandinavia and the gradual mixing of genes as Z and its Proto-Sámi carriers slowly moved north and northwest.28

9 Sámi loans from Germanic: A problem? Returning to the question of why Sámi loans from Germanic in the second layer discussed above are not also borrowed by Finnic: in his treatise on Sámi loanwords in Finnish and Karelian, Aikio (2009) mentions the primary dialect split between west and east. He provides distribution maps for 92 words borrowed from Sámi by Finnish and Karelian. A glance at the maps gives one a good picture of the dialect split; Aikio lists 21 items give or take a few borrowed only by the western dialects of Finnish, and 29 or so borrowed only by the eastern dialects and Karelian. One thing stands out, viz. there are relatively fewer Sámi loanwords in the southwestern dialects of Finnish. Moving on to Sámi substrate toponyms in Finland, they are found pretty much everywhere throughout the country, except within about 30 km of the Gulf of Finland coast (Aikio 2009: 219). In other words, there was a Sámi presence in virtually all of Finland at one time. Furthermore, a number of genetic studies have demonstrated dual origins for the Finns. Kittles et al. (1998: 1177) find two major founding Y chromosome lineages in Finland: “Haplotype A/49 (males) may have entered Finland from the Lake Ladoga region, whereas settlers possessing haplotype B/69 traveled across the Gulf of Finland and the Swedish border.”29 Moreover, they say that the expansion time for haplogroup A predated that for haplogroup B by almost 2,000 years, with Uralic speakers arriving from the east ca. 4,000 years ago followed by agriculturalists from Sweden and the Baltic ca. 2,000 years ago. The settlers of the A group provided 53 % of the present Finnish Y chromosomal gene pool. Raitio et al. (2001: 471, 478) find regional differences between eastern and northern areas of Finland vis-à-vis western Finland. Lappalainen et al. (2006: 214) confirm these differences and mention the possibility that “the Finno-Ugric settlement had been an earlier event.” Perhaps the incipient Proto-Sámi speakers were in contact with these two groups at different times.

28  An interesting note is that majority Swedes and Norwegians have small percentages of Z, which can only be due to admixture with Sami or Finns. 29  Their composite haplotypes are constructed of 7 Y-specific microsatellite loci, a restriction site at the Y alphoid satellite DYZ3 locus and a deletion polymorphism at the DYF155W2 locus.

At the frontier: Sámi linguistics gets a boost from outside  

 83

10 Sámi mtDNA The two main maternal haplogroups of the Sámi are V and subhaplogroup U5b1b1, the latter the so-called Sámi motif. After the glaciers melted, V reached Scandinavia from the southwest (the Franco-Cantabrian glacial refuge) and from the southeast (the Volga-Ural region); V is quite common among all Sámi groups, peaking in Northern Sweden at 58.6 %. V is significantly lower among Finns at 6.4 %, 6 % for Karelians and only 4.4 % for Continental Europe. U5b1b1 is mainly found in Northern and Eastern Europe and may have emanated from the Ukrainian glacial refuge.

Fig. 3: Sámi mtDNA (Norga = Norway; Ruoŧŧa = Sweden; Suopma = Finland; Guoládat = Kola peninsula of Russia; Gárjil = Karelia in Russia).

There is a cline running from a high of 56.8 % for Norwegian Sámi to 23.9 % for Southern Swedish traditional Sámi (Ingman and Gyllensten 2007: 117). Finns and Karelians have 6.7 % and 6 % respectively. The next highest figure for U5b1b1 is North Russians at 3 % and only 0.6 % for Continental Europe (2.6 % for U5b). The significant percentage of U5b1b1 for Finns and Karelians may be due to admixture with Sámi at some point in time after the onset of the Iron Age when Proto-Sámi

84 

 John Weinstock

formed in Southern Finland and moved north (cf. Meinilä et al. 2001). Figure 3 shows the percentages of the most important Sámi maternal DNA. Note especially the differences from south to north. Maternal haplogroup Z has already been discussed.

11 Sámi Y Chromosomes The Sámi display mainly three Y-chromosomes: haplogroup I1 shows clinal behavior, going from northeast (lowest) to southwest (highest). The R1a haplogroup frequency is lowest among the Finns and the Sámi, but highest in Middle and North Norway, so it too displays a clinal pattern. Haplogroup N1c (formerly N3) is widespread in the Baltic region and probably reached Fennoscandia from Eastern Europe (Tambets et al. 2004: 677–678). Rootsi et al. (2007) suggest that N1c arrived in Fennoscandia via a counterclockwise route from southern Siberia, possibly altered by founder effects or bottlenecks. There is a clear distinction between the Eastern Finns and the Sámi vs. those living farther to the west with the former having high values of N1c. Lappalainen et al. (2006: 213) find that N3 (N1c) “supports a common component of Finno-Ugric genetic background [among these populations] […] rather than a major Asian component among the Finns or the Balts.” Map 4 below displays Sámi Y-chromosomal DNA. Again note the differences from south to north and east to west. Some of the DNA figures include majority population as indicated in parentheses. Cf. the Y-chromosomal analysis mentioned above under Sámi loans from Germanic: a problem?

12 Conclusions Summing up: Ante Aikio’s theory on the origin of the Sámi language is the best solution to an enduring enigma. When dealing with prehistory where there are no historical documents, the conclusions are nearly always stronger if support can be found in other disciplines, in the case at hand, anthropological conjectures, archaeological relics, siida structure and current genetic patterns among Sámi and other groups in Fennoscandia. Though perhaps not convincing individually, together they carry much more weight. As Pre-Sámi speakers came from the east and their language evolved into Proto-Sámi and who then moved north and northwest over some forty generations beginning in the Early Iron Age, their genetic profile changed as they passed their language on to the people they subsumed.

At the frontier: Sámi linguistics gets a boost from outside  

 85

Fig. 4: Sámi Y chromosomal DNA (Norga = Norway; Ruoŧŧa = Sweden; Suopma = Finland; Guoládat = Kola peninsula of Russia; Gárjil = Karelia in Russia).

References Aikio, Ante (2004): An essay on substrate studies and the origin of Saami. In: Irma Hyvärinen, Petri Kallio, Jarmo Korhonen (eds.), Etymologie, Entlehnungen und Entwicklungen: Festschrift für Jorma Koivulehto zum 70. Geburtstag, 5–34. Helsinki: Société Néophilologique. Aikio, Ante (2006): On Germanic-Saami contacts and Saami prehistory. Journal de la Société Finno-Ougrienne 91: 9–55. Aikio, Ante (2007): The study of Saami substrate toponyms in Finland. In: Ritva Liisa Pitkänen and Janne Saarikivi (eds.), The borrowing of place-names in the Uralic languages, 159–197. Debrecen and Helsinki. Aikio, Ante (2009): The Saami loanwords in Finnish and Karelian. Academic dissertation, Faculty of Humanities of the University of Oulu. Luobbal Sámmol Sámmol Ánte (Aikio, Ante) (2012): An essay on Saami ethnolinguistic prehistory. A Linguistic Map of Prehistoric Northern Europe. Edited by Riho Grünthal & Petri Kallio. Société Finno-Ougrienne, Helsinki 2012: 63–117. Baldi, Philip and B. Richard Page (2006): Review of: Theo Vennemann, Gen. Nierfeld: Europa Vasconica-Europa Semitica. In: Patrizia Noel Aziz Hanna (Ed.), Trends in Linguistics,

86 

 John Weinstock

Studies and Monographs 138, Berlin: Mouton de Gruyter, 2003, pp. xxii + 977. Lingua 116(12): 2183–2220. Barth, Fredrik (1969): Ethnic groups and boundaries. The social organization of culture difference. Oslo: Universitetsforlaget. Bjerck, Hein Bjartmann (2008): Norwegian Mesolithic Trends: A Review. In: Geoff Bailey and Penny Spikins (eds.), Mesolithic Europe, 60–106. Cambridge: Cambridge University Press. Carpelan, Christian (2003): Inarilaisten arkeologiset vaiheet. In: Veli-Pekka Lehtola (ed.), Inari – Aanaar. Inarin historia jääkaudesta nykypäivään, 28–95. Oulu: Inarin kunta. Einarsdottir, Elisabet, Inez Egerbladh, Lars Beckman, Dan Holmberg and Stefan A Escher (2007): The genetic population structure of northern Sweden and its implications for mapping genetic diseases. Hereditas 144: 171–180. Hansen, Lars Ivar and Bjørnar Olsen (2007): Samenes historie fram til 1750. Oslo: Cappelen Akademisk Forlag. 3. opplag. Huyghe, Jeroen R, Erik Fransen, Samuli Hannula, Lut Van Laer, Els Van Eyken et al. (2011): A genome-wide analysis of population structure in the Finnish Sámi with implications for genetic association studies. European Journal of Human Genetics 19: 347–352. Hyllested, Adam (2008): Sámi loanwords in Old Norse. In: Hans Frede Nielsen (ed.): Early and Pre-historic Language Development in North-Western Europe, 131–146. (NOWELE 54/55). Odense: Odense University Press. Ingman, Max and Ulf Gyllensten (2007): A recent genetic link between Sami and the Volga-Ural region of Russia. European Journal of Human Genetics 15: 115–120. Johansson, Åsa, Max Ingman, Steven J Mack, Henry Erlich, and Ulf Gyllensten (2008): Genetic origin of the Swedish Sami inferred from HLA class I and class II allele frequencies. European Journal of Human Genetics 16: 1341–1349. Kittles, Rick A., Markus Perola, Leena Peltonen, Andrew W Bergen, Richard A Aragon et al. (1998): Dual Origins of Finns Revealed by Y Chromosome Haplotype Variation. American Journal of Human Genetics 62: 1171–1179. Kusmenko, Jurij (2008): Der samische Einfluss auf die skandinavischen Sprachen. Ein Beitrag zur skandinavischen Sprachgeschichte. Berliner Beiträge zur Skandinavistik, Band 10. Berlin: Nordeuropa-Institut. Lao, Oscar, Timothy T. Lu, Michael Nothnagel, Olaf Junge, Sandra Freitag-Wolf et al. (2008): Correlation between Genetic and Geographic Structure in Europe. Current Biology 18: 1241–1248. Lappalainen, Tuuli, Satu Koivumäki, Elina Salmela, Kirsi Huoponen, Pertti Sistonen et al. (2006): Regional differences among the Finns: A Y-chromosomal perspective. Gene 376: 207–215. Lappalainen, Tuuli, V. Laitinen, E. Salmela, P. Andersen, K. Huoponen et al. (2008): Migration Waves to the Baltic Sea Region. Annals of Human Genetics 72: 337–48. Lehtola, Veli-Pekka (2004): The Sámi People. Traditions in Transition. Tr. Linna Weber MüllerWille. Anaar and Inari: Kustannus-Puntsi. Meinilä, Maria, Saara Finnilä, and Kari Majamaa (2001): Evidence for mtDNA Admixture between the Finns and the Sámi. Human Heredity 52: 160–170. Nelis, Mari, Tõnu Esko, Reedik Mägi, Fritz Zimprich, Alexander Zimprich et al. (2009): Genetic Structure of Europeans: A View from the North-East. PLoS ONE 4(5): e5472. Odner, Knut (1983): Finner og terfinner: etniske prosesser i det nordlige Fenno-Skandinavia. Oslo: University of Oslo.

At the frontier: Sámi linguistics gets a boost from outside  

 87

Odner, Knut (1992): Ethnicity and Traditions in Northern Fenno-Scandinavia. “Finns and terfinns” ten years after. Acta Borealia 2: 21–36. Olsen, Bjørnar (2003): Belligerent Chieftains and Oppressed Hunters? Changing Conceptions of Interethnic Relationships in Northern Norway during the Iron Age and the Early Medieval Period. In: James H. Barrett (ed.), Contact, Continuity, and Collapse. The Norse Colonization of the North Atlandic. Turnhout: Brepols. Polomé, Edgar C. (1990): Types of Linguistic Evidence for Early Contact: Indo-Europeans and Non-Indo-Europeans. In: Thomas L Markey and John A C Greppin (eds.), When Worlds Collide: Indo-Europeans and Pre-Indo-Europeans–The Bellaio Papers, 267–289. Ann Arbor: Karoma. Rankama, Tuija and Jarmo Kankaanpää (2011): First evidence of eastern Preboreal pioneers in arctic Finland and Norway. Quartär 58: 183–209. Raitio, Mirja, K. Lindroos, M. Laukkanen, T. Pastinen, P. Sistonen et al. (2001): Y-Chromosomal SNPs in Finno-Ugric-Speaking Populations Analyzed by Minisequencing on Microarrays. Genome Research 11: 471–482. Rootsi, Siiri, Lev A Zhivotovsky, Marian Baldovic caron, Manfred Kayser, Ildus A Kutuev et al. (2007): A counter-clockwise northern route of the Y-chromosome haplogroup N from Southeast Asia towards Europe. European Journal of Human Genetics 15: 204–211. Saarikivi, Janne (2006): Substrata Uralica. Studies on Finno-Ugrian Substrate in Northern Russian Dialects. Academic Dissertation. http://ethesis.helsinki.fi/julkaisut/hum/suoma/ vk/saarikivi/substrat.pdf (accessed June 4, 2014). Saarikivi, Janne (2004a): Is there Palaeo-European substratum interference in western branches of Uralic? Journal de la Société Finno-Ougrienne 90: 187–214. Also in Saarikivi 2006. Saarikivi, Janne (2004b): Über das sámische Substratnamengut in Nordrußland und Finnland. Finnisch-Ugrische Forschungen 58: 162–253. Also in Saarikivi 2006. Salmela, Elina, Tuuli Lappalainen, Jianjun Liu, Pertti Sistonen, Peter M. Andersen et al. (2011): Swedish Population Substructure Revealed by Genome-Wide Single Nucleotide Polymorphism Data. PLoS ONE 6(2): e16747. Salmela, Elina, Tuuli Lappalainen, Ingegerd Fransson, Peter M. Andersen, Karin Dahlman-Wright et al. (2008): Genome-Wide Analysis of Single Nucleotide Polymorphisms Uncovers Population Structure in Northern Europe. PLoS ONE 3(10): e3519. Salmons, Joseph (1992): Northwest Indo-European Vocabulary and Substrate Phonology. In: Roger Pearson (ed.), Perspectives on Indo-European language, culture and religion: studies in honor of Edgar C. Polomé. Volume 2. McLean, Va.: Institute for the Study of Man. Salmons, Joseph (2004): How (non-)Indo-European is the Germanic lexicon? … And what does that mean? Irma Hyvärinen, Petri Kallio, Jarmo Korhonen (eds.), Etymologie, Entlehnungen und Entwicklungen: Festschrift für Jorma Koivulehto zum 70. Geburtstag, 5–34. Helsinki: Société Néophilologique. Sammallahti, Pekka (1998): The Sámi Languages. An Introduction. Kárášjohka: Davvi Girji. Schmitt, Lou (2007): The West Swedish Hensbacka from an anthropological point of view, and recent developments concerning the final drainage of the Baltic Ice Lake. In: M. Kobusiewicz and J. Kabaciński (eds.), Studies in the Final Palaeolithic Settlement of the Great European Plain. Poznań: Institute of Archaeology and Ethnology, Polish Academy of Sciences. Schmitt, Lou, Stephan Larsson, Jan Burdukiewicz, John Ziker, Krister Svedhage et al. (2009): Chronological insights, cultural change, and resource exploitation on the west coast

88 

 John Weinstock

of Sweden during the Late Palaeolithic / early Mesolithic transition. Oxford Journal of Archaeology 28(1): 1–27. Skoglund, Pontus, Helena Malmström, Maanasa Raghavan, Jan Storå, Per Hall et al. (2012): Origins and Genetic Legacy of Neolithic Farmers and Hunter-Gatherers in Europe. Science 336(6080): 466–469. Tambets, Kristiina, Siiri Rootsi, Toomas Kivisild, Hela Help, Piia Serk et al. (2004): The Western and Eastern Roots of the Sámi—the Story of Genetic “Outliers” Told by Mitochondrial DNA and Y Chromosomes. American Journal of Human Genetics 74: 661–682. Tegengren, Helmer (1952): En utdöd lappkultur i Kemi Lappmark: studier i Nordfinlands kolonisationshistoria. (Acta Academiae Aboensis, Humaniora 19:4.) Åbo: Åbo Akademi. Thomason, Sarah Grey and Terrence Kaufman (1991): Language Contact, Creolization, and Genetic Linguistics. Berkeley: University of California Press. Vennemann, Theo (2003): Europa Vasconica – Europa Semitica. Berlin and New York: Mouton de Gruyter. Vorren, Ørnulv (1989): Veidekulturens arealfordeling, siidagrenser og ressursutnytting i Samelands nordøstlige strøk. In: Bjørn Aarseth (ed.), Grenser i Sameland. Samiske Samlinger, Bind XIII. Oslo: Norsk Folkemuseum. Weinstock, John (2010): Thoughts About Sámi Prehistory. In: Else Mundal and Håkan Rydving (eds.), Samar som ”den andre”, samar om ”den andre”: Identitet og etnicitet i nordiska kulturmöten, Umeå: Umeå universitet. Weinstock, John (2013): Genetic heterogeneity in Scandinavia: Not only the Sami. L’image du Sápmi II, 148–167. Textes réunis par Kajsa Andersson. Weinstock, John (Forthcoming): Assimilation of the Sami: Its Unforeseen Effects on the Majority Populations of Scandinavia. International Sámi Research Seminar. Scandinavian Studies. Volume 85, Number 4, Winter 2013: 411–430.

Section 2: Phonology

Natalia Kuznetsova

Two phonological rarities in Ingrian dialects1 Abstract: The paper considers two typologically rare phonological phenomena related to the development of sound lengthening and reduction that appear in two dialects of the severely endangered Ingrian language in North-Western Russia. The first is attested in the Soikkola dialect, and involves a phonological ternary quantity contrast of consonants. The second exists in the Lower Luga dialect, and involves a phonological opposition of full modal and reduced voiceless vowels. First, brief phonetic, phonological and typological profiles of these contrasts are given. Furthermore, I analyze their probable further evolution against the typological background of similar phenomena in genetically related and unrelated languages. In the end, I discuss the question of stability and maintenance forces of these contrasts, as well as their significance in light of general prosodic development of the Soikkola and Lower Luga dialects.

1 Introduction 1.1 O  n the phonological and morphonological typology of Finno-Sámi languages Finnic and Sámi complex morphonological systems of grade alternation are well known in linguistic typology. They have developed from purely quantitative and mainly phonetic phenomena, but synchronically these systems of quantitative contrasts and quantity-based alternations are extremely diverse. On one extreme of this continuum there are systems like in Veps, where the original grade alternation was lost completely. Quantitative oppositions of vowels and consonants are marginal in Veps (long phonemes are very rare). Moreover, Veps lacks morphonological stem alternations of long and short consonants (Zaiceva 1981: 17, 35).

1 Supported by the Russian Foundation for Humanities (project 11‑04‑00172а) and the Alfred Kordelin Foundation. I also express my gratitude to professor T.-R. Viitso, who in 2009 drew my attention to the ternary contrast in Ingrian and encouraged phonetic experiments, as well as to M. Coler, E. Markus, M. Muslimov, F. Rozhanskiy, an anonymous reviewer, and the editors of this volume for their valuable remarks and comments.

92 

 Natalia Kuznetsova

On the other extreme are systems like in Livonian. This language has one of the most complicated systems of stem alternations among Finnic languages, where quantitative oppositions, including the ternary ones, are accompanied by glottalization contrasts (Viitso 1975). Sámi languages, genetically close to Finnic, have similar or even more intricate systems. For example, ternary quantity contrasts accompanied by glottalization and pharyngealization oppositions constitute the base for stem alternations in North Sámi dialects (Sammallahti 1998: 39–60; Bals et al. 2007; McRobbie-Utasi 2007: 179–180). For the description of languages on this side of the continuum, a prosodic unit such as the foot is extremely relevant. Hereafter a foot is defined as a sequence of one to three syllables, which are not only linked together into one common stress group, but are also the domain of other prosodic features. The further any Finnic or Sámi variety is removed from the extreme of phonologically and morphonologically complex systems, the lesser is the relevance of the foot for its phonetics and phonology.

1.2 I ngrian phonology and morphonology against the background of Finno-Sámi languages Ingrian2 is situated towards the side of this continuum which has more complex systems. That is, it is closer to Livonian and Sámi than to Veps. At the same time, the level of phonological and morphonological complexity varies considerably within Ingrian dialects. It is precisely the differences in the structure of phonological quantitative contrasts and morphonological quantitative alternations that are the key source for this diversity. In the Lower Luga dialect, quantitative oppositions and alternations exhibit the least complex structure. This dialect differs the most from the other three dialects of Ingrian, as it has undergone heavy contact influence from the neighboring related languages of the Lower Luga area: Votic, Ingrian Finnish and, to some extent, Estonian (Laanest 1966а: 146, 150– 151; Muslimov 2005: 5, 13). In this sense, the Lower Luga dialect could serve as a good example for a widely discussed typological hypothesis regarding the direct correlation between the level of intensity and duration of contact and the level of structural simplicity in the contacting languages. The longer and more actively a language participates in areal contacts, the simpler its structure becomes (Trud­

2 Nowadays, a total of some 100 people born mainly between the 1920s and the 1940s speak the Soikkola and the Lower Luga dialects of Ingrian. Two other dialects, Hevaha and Oredeži are extinct by now.

Two phonological rarities in Ingrian dialects 

 93

gill  2001; Gil  et  al.  2009). The phonological and morphological system of the Lower Luga dialect is also the most innovative of all the Ingrian dialects. The quantitative contrasts and alternations have the most complex structure in the Soikkola dialect. Between these two dialects, there were once the systems of the currently extinct Oredezh dialect (with a simpler system) and the Hevaha dialect (with a more complex system).

1.3 Two typological rarities in Ingrian dialects This paper considers two typologically rare phonological phenomena related to the development of sound lengthening and reduction. The first is attested in the Soikkola dialect, and involves a phonological ternary quantity contrast of consonants. The second exists in southern Lower Luga varieties, and involves a phonological opposition of full modal and reduced voiceless vowels. The paper aims to answer the question how unique or typical as compared to other world languages are the phonetic realization, the phonological structure, the history of emergence and subsequent evolution of these two phenomena. In sections 2 and 3, brief phonetic, phonological and typological profiles of these phenomena are given. Section 4 presents an analysis of their probable further evolution, which is supported by examples from genetically related and unrelated languages. Section 5 is devoted to the question of stability and maintenance forces of these contrasts, and section 6 deals with their significance in light of the general prosodic development of the Soikkola and Lower Luga dialects.

2 The Soikkola dialect: A ternary quantity contrast 2.1 Ternary quantity contrast in the Soikkola dialect The Soikkola dialect manifests a rare phenomenon called a ternary (or threeway) quantity contrast of consonants. This phenomenon is further illustrated in the intervocalic position before a lengthened/long vowel: [ka·nā]3 ‘hen:nom’ vs. [ka·n̆nà] ‘hen:part/ill’ vs. [li·nnà ~ li·ǹnà] ‘town:part/ill’. Figures 1–3 provide examples of spectrograms for each word (pronounced by AJF, a female born in 1933 in a Hammaala village). Pronunciations most easily comparable by their overall length and phrasal position were chosen.

3 Finno-Ugric phonetic transcription is used in the phonetic notation.

94 

 Natalia Kuznetsova

Table 1 shows mean durations of segments in a disyllabic foot nucleus (a sequence ranging from the start of the first vowel to the end of the second one) of each structural type in AJF’s speech. For these measurements, tokens with an intervocalic sonorant (n, l or m) and a second closed syllable were taken from her spontaneous speech.4 For comparative purposes, a structure like [li·nnăn] ­‘town:gen’ containing the longest consonant allophone before a short/reduced vowel was also added to Table 1. Other two durational types of consonants never occur before a short vowel in a disyllabic foot. There is also a phonological transcription for each structural type. For a number of reasons, it is preferable to postulate a binary segmental opposition of consonants and to consider all non-initial vowels as phonologically short. Such a conception serves two purposes: it keeps the phonological and morphonological descriptions consistent, and it highlights the correlation between the reductions and lengthenings of segments within a foot and their existing link to the morphologized stress. The contrast of two “long” consonantal classes and the length contrasts of non-initial vowels are treated as a part of realization of three suprasegmental foot accents, light /′/, heavy /‵/ and extra-heavy /‶/ (cf. Kuznetsova 2009а, 2012a).

Fig. 1: [ka·nā]

4 Note that the duration of V2 in a closed syllable is about 20–30 ms shorter than in an open one. Furthermore, in AJF’s speech, a lengthened vs. short vowel contrast after the long first syllable has in fact been replaced with a short vs. reduced vowel contrast; cf. more data from her in Markus (2010: 46, see Speaker 5) and Kuznetsova (2012a: 49).

Two phonological rarities in Ingrian dialects 

 95

Fig. 2: [ka·n̆nà]

Fig. 3: [li·ǹnà] Table 1: Ternary quantity contrast in AJF’s spontaneous speech structural type

V15, ms σ, ms C/C̄, ms σ, ms V2, ms σ, ms number of tokens

k´anan [ka·nàn] ‘hen:gen’ s’in̄in [ši·n̆nin] ‘blue:nom’ l’’in̄as [li·nnaš] ‘town:in’ l´in̄an [li·nnăn] ‘town:gen’

82 116 117 115

18 30 22 24

74 120 185 161

12 24 26 27

5 A list of abbreviations appears at the end of the article.

109 91 88 64

23 16 16 10

56 50 51 49

96 

 Natalia Kuznetsova

The existence of a ternary contrast in the Soikkola dialect was first attested by Sovijärvi in his experimental research from the 1930s (1944: 12, 14). Kuznetsova (2009а, 2012a) and Markus  (2010, 2011) confirmed Sovijärvi’s findings on the basis on phonetic measurements of speech samples from contemporary Soikkola speakers.

2.2 Positions of the ternary contrast For the stops p, t, k and the fricative s, four main positions (1)–(4) of the ternary contrast, listed below, are distinguished on the basis of phonetic, phonotactic and historical factors. For other consonants, it exists only in two positions, (1) and (3). (1) In VCV2/VC̄V2 nucleus of a dissyllabic foot: [ta·bā ~ ta·bā] ‘kill:imp’ vs. [ta·p̆pà] ‘catch:3sg’ vs. [ta·ppà] ‘kill:inf’; (2) In VRCV2 /VRC̄V2, VVCV2/VVC̄V2 nuclei of a dissyllabic foot: [kar̀·dàš] ‘card:in’ vs. [ka·rtà ~ ka·rt̀à ~ ka·rt‌t̆ à] ‘tin:part/ill’ vs. [ka·rttà ~ ka·rt‌t̆ à] ‘card:part/ ill’, [mā̀·dà] ‘sleep:ips’ vs. [łū·tà ~ łū·t̀à ~ łū·t‌t̆ à] ‘broom:part/ill’ vs. [ł’ūttà] ‘rely:inf’; (3) In VCV2/VC̄V2 nucleus of a trisyllabic foot: [ka·dājà] ‘juniper:part/ill’ vs. [ka·tàjă ~ ka·t̀àjă ~ ka·t‌t̆ àjă] ‘juniper’ vs. ­[ka·ttojă ~ ka·t‌t̆ òjă] ‘roof:pl:part’, [ma·già] ‘sweet:part/ill’ vs. [šu·k‌k ̆ ìa ~ šu·k‌k ̆ ia] ‘1) comb:pl:part; 2) comb:inf’ vs. [šu·kkia] ‘sock:pl:part’; (4) In VRCV2 /VRC̄V2, VVCV2/VVC̄V2 nuclei of a trisyllabic foot: [ka·rdăłł‌ĕ̆ ] ‘card:all’ vs. [ka·rtojă ~ ka·rt̀ŏjă ~ ka·rtt‌ŏ̆ jă] ‘tin:pl:part’ vs. [ka·rtt‌ŏ̆ jă ~ ka·rttŏjă] ‘card:pl:part’ / [ke·rtàmă ~ ke·rt̀àmă ~ ke·rtamă] ‘сучить:prs:1pl’ vs. [ka·rt‌t̆ ŏjă ~ ka·rttŏjă] ‘card:pl:part’ / [so·rttùmă ~ so·rt‌t̆ ùmă ~ so·rttumă] ‘sort:prs:1pl’, [sǖ·dä̆mä̆] ‘feed:prs:1pl’ vs. [rī·tĕlĕ ~ rī·t̀ĕlĕ ~ rī·t‌t̆ ĕlĕ] ‘quarrel:prs:1pl’ / [łā·tìmă ~ łā·t̀ìmă ~ łā·timă] ‘be_going_to_go:prs:1pl’ vs. [ū·t‌t̆ ĕlĕ ~ ū·ttĕlĕ] ‘wait:imp:2sg’ / [sǖ·t‌t̆ ìmä̆ ~ sǖ·ttìmä̆ ~ sǖ·ttimä̆] ‘feed_oneself:prs:1pl’.

The absolute duration of vowels and consonants varies across structures due to a phonetic tendency towards foot isochrony (Kuznetsova 2009а: 63, 2009b: 34; Markus  2010:  48,  50). In this case, the tendency manifests itself such that the more segments a foot contains, the shorter their absolute duration becomes. Under the influence of this tendency, partly different phonetic feature now distinguishes the two “long” classes in positions (2)–(4), compared to (1). In (1), the consonant is always perceived as long both in cases like [ta·p̆pà] and [ta·ppà]. In (2)–(4), the consonants of the third quantitative class are generally heard as long, while the realization of the second class consonants varies between long and short (Laanest 1987: 290–291; Kuznetsova 2009а: 67).

Two phonological rarities in Ingrian dialects 

 97

2.3 Existing terms for the quantitative consonantal classes The three quantitative classes of Soikkola consonants can be referred to in a variety of ways in the literature, depending on whether the authors are Finno-Ugrists or general linguists, and whether they aim to emphasize the concrete phonetic realization or the nature of the phonological opposition of the consonants, or their origin. To facilitate greater clarity of the following discussion, the main terms for these consonants are summarized in Table 2. Table 2: Terms for the consonant classes within the Soikkola ternary contrast Сlass of Etymological Phonetic terms ­consonants terms class 1

short/single

class 2

secondary geminates

class 3

primary ­geminates

Phonological ­interpretation

1) all consonants: short/single(tons); lenis 2) additionally stops and s: (half)voiced [d/d] 1) all consonants: geminates/double/long; fortis — particular types of allophones: short geminates [tt̆ ] ~ half-long/lengthened singletons [t]̀ ; 2) additionally stops and s: voiceless (can also be realized as voiceless singletons [t]) 1) all consonants: geminates/double/long; fortis — particular types of allophones: long geminates [tt] ~ overlong geminates [tt̀ ]; 2) additionally stops and s: voiceless

In Table 2 above, “Class of consonants” can be taken to mean conventional numbers referring to the quantitative classes of consonants (“first”, “second” and “third degree of length or quantity” respectively are also frequent). “Etymological terms” refer to the origin of the consonants in accordance with the standard norms common to Finnic studies. “Primary” are the geminates that presumably existed already in Proto-Finnic, while “secondary” geminates emerged from Proto-Finnic short consonants in certain Finnic languages later. “Phonetic terms” mean the allophonic classes of absolute duration in different structures and positions for each consonantal type. “Phonological interpretation” presents a phonological interpretation of these classes the author of this paper adheres to, cf. section 2.1.

2.4 Typological data on the ternary contrast Modern typological works mention only a handful of languages where the ternary contrast is the most incontrovertibly attested. Three contrastive degrees of con-

98 

 Natalia Kuznetsova

sonantal length are indicated only for languages closely related to Ingrian. The most typologically well known are Estonian and several Sámi languages (Ladefoged, Maddieson  1996:  93; Blevins  2004:  201); a similar phenomenon is also found in Livonian (Viitso 1997: 231). Among languages with three vocalic lengths, the literature typically mentions first and foremost Estonian, as well as Dinka (Kir-Abbaian family; Southern Sudan), Coatlán Mixe and San José Paraíso Mixe (dialects of Lowland Mixe; Mixe-Zoque family; Mexico) and Yavapai (Yuman-Cochimí family; USA) (Ladefoged and Maddieson  1996:  320; Blevins  2004:  201; McRobbie-Utasi 2007: 185–193). A typological claim that only two contrastive segmental lengths can exist in a language was first expressed by N. S. Trubetzkoy and was then reiterated by a number of authors. His idea can be summarized as follows: if there seems to exist a segmental three-way (or more complex) length contrast, the latter is always accompanied by additional phonetic distinctions by other features of the same segments, or by prosodic features (2000 (1939): 206–211). A ternary length contrast cannot therefore be considered the only contrastive phonemic feature and consequently cannot be totally assigned to the segmental level. Experimental phonetic research on Estonian, Livonian and Sámi languages has since completely confirmed this hypothesis. The aforementioned foot concept is crucial for the description of phonetics and phonology of all these languages. In Estonian, the three-way contrast in a disyllabic foot manifests itself in different ratios between the elements of the foot nucleus. Three main quantitative types of a first syllable are in inverse relation with three quantitative types of a second syllable vowel. Three combinations are possible: overlong – short, long – short/lengthened, short – lengthened/long. Moreover, there are four types of ratios between the first syllable vowels and the following consonants: long – short, short – long, half-long – half-long, short – short (cf. detailed measurements in Eek and Meister 2003–2004). Additionally, if the first syllable contains long vowels or diphthongs, there is a difference in pitch movement between Q2 and Q3 feet. These main trends were experimentally substantiated in the early 1960s (Lehiste 1960; Liiv 1961) and further elaborated in numerous experimental works by, most notably, I. Lehiste, A. Eek and E. Meister, D. Krull, E. L. Asu(‑Garcia), and P. Lippus. Later research showed, among other things, the role of intensity and vowel quality as secondary features for length distinctions (Eek 1973; Lippus 2011). There also exists a secondary voicing opposition for the stops and the voiceless alveolar sibilant /s/: class 1 consonant allophones are voiced or half-voiced as opposed to voiceless class 2 and 3 allophones. Similar are Livonian and the Sámi languages. In Livonian, short and halflong consonantal allophones occur only before a lengthened vowel of the second

Two phonological rarities in Ingrian dialects 

 99

syllable, and long consonantal allophones only before a short vowel. Further, glottalization in dissyllables combines only with a long consonantal allophone (Viitso 1997: 231; Lehiste et al. 2008). Phonetic research on Sámi languages also clearly manifests interdependencies between either the first syllable elements or the whole foot nucleus elements. For example, in Skolt Sámi (Eastern Sámi group) there are three types of ratios between the first syllable vowels and the following consonants/consonantal clusters: long – short, short – long, half-long – half-long. The duration of the second syllable vowels does not affect these ratios (McRobbie-Utasi 1996: 29). Research on the North Sámi (Western Sámi group) dialect of Kautokeino shows similar results. Four types of ratios between the first syllable vowels and the following consonants are the same as in Estonian (Bals et al. 2007: 16; the duration of the second syllable vowels was not measured). Detailed experimental research on Inari Sámi (Eastern Sámi group) also established similar kinds of nucleus V1/C and C/V2 ratios. For example, for structures with a short first syllable vowel and an open second syllable the following C/V2 ratios were found: long – short, halflong – long, short – long. In other structures, a bit different ratios were discovered, but in all cases any such correlations were established (Bye et al. 2009). Thus, at least in the case of all languages with a presumed ternary contrast of consonants, the three-way opposition exists at the level of either the syllable or the foot. It is therefore not segmental; cf. similar statements in McRobbie-Utasi  (2007:  194–196). In this sense, Trubetzkoy’s claim still holds. Apparently, many such systems of a complex “three-way quantity contrast” (in terms of McRobbie-Utasi, who opposes it to a segmental “three-way length contrast”) are phonologically most adequately described with the help of suprasegmental (either syllable or foot) accents, with only a binary or even a lack of quantitative phonological contrasts on the segmental level. Among languages with a presumed three-way contrast of vowels, the example of Estonian has been already discussed. As for Mixe, Yavapai and Dinka, there are more reasons to believe that vowel length alone can be a distinctive feature. However, to my knowledge, detailed phonetic research proving the claim has been conducted only on Dinka (Remijsen and  Gilley 2008). The authors show that vowel length correlates significantly neither with the length of the coda (a final consonant in monosyllables), nor with the tonal contour on a word, nor with vowel quality and intensity. Some parallels in the structure and evolution of the ternary contrast in Dinka and Ingrian are discussed in sections 4.1.2 and 5.2.2.

100 

 Natalia Kuznetsova

2.5 The Soikkola ternary contrast against the typological background In the previously discussed languages with a presumed three-way consonant contrast, quantitative consonantal classes are insufficient in themselves to distinguish meanings due to coexisting phonetic contrasts. The Soikkola dialect does not provide a completely pure case of a three-way length contrast either. There are other oppositions that come into play and reinforce the length contrasts. First, as in Estonian, the shortest allophone of the stops and /s/ is voiced or half-voiced, e. g. [ta·bā ~ ta·bā] ‘kill:imp’. The ternary contrast for this group of consonants is therefore not purely quantitative, as the voicing feature also participates in distinguishing between the three consonantal classes. However, for the other members of the consonant inventory, this voicing opposition is irrelevant. The only contrast is that of length. Furthermore, in the VCV2 nucleus, as in [ta·bā], a second syllable vowel undergoes salient prosodic lengthening that is automatic (contrary to all other lengthened non-initial vowels). It is always present in this position. Ingrian does not differ in this respect from Estonian, Votic, Finnish (at least some of its dialects), or Livonian. The duration of this vowel is considerably longer than the one of lengthened vowels after the long first syllable (in [ta·p̆pà], [ta·ppà] and other nucleus types). In many idiolects, it roughly corresponds to the length of an initial long vowel. In this sense, the contrast of a structure with a class 1 consonant vs. with either class 2 or class 3 consonants does not only rely on consonantal length. The length of the following vowel (and possibly also the preceding one, cf. data in Table 1) also plays an important role. However, one feature of the Soikkola ternary contrast makes it unique in comparison to the aforementioned languages. In Soikkola, the contrast of “long” classes 2 and 3 can alone produce minimal pairs, for example: [ta·p̆pà] ‘catch:3sg’ vs. [ta·ppà] ‘kill:inf’; [le·p̆pä̀] ‘rest:imp’ vs. [le·ppä̀] ‘alder:part/ill’; [ła·k̆kà] ‘sweep:imp’ vs. [ła·kkà] ‘hayloft:part/ill’; [šu·k̆kà] ‘1) comb:part/ill; 2) (not) at all’ vs. [šu·kkà] ‘sock:part/ill’.

Phonetic research carried out by Kuznetsova (2009а: 64; checked in 2012a: 48) and Markus (2010, 2011) has shown that at least words with a disyllabic foot nucleus of a structure [VC̄V̀] can be opposed only by two “long” consonantal classes. The durations of nuclear V1 and V2 do not show statistically significant differences (cf. Table 1 and the measurement results in the aforementioned works). Soikkola could thus be said to have a pure ternary length contrast in positions (1) and (3), were it not for the fact that V2 in VCV nuclei is phonetically longer than

Two phonological rarities in Ingrian dialects 

 101

the lengthened second syllable vowels in other nucleus types. Soikkola could also be said to have this contrast in positions (2) and (4), where the second syllable vowels have a comparable duration after all the three quantitative consonantal classes. However, only stops and /s/ participate in a ternary contrast in these positions, and their class 1 allophones are additionally (half‑)voiced. Soikkola Ingrian is still closer to a pure three-way consonantal length contrast case than Estonian, Livonian or Sámi. It also shows that the “long” consonantal classes 2 and 3 can in principle serve as the only distinctive features. This seems an important typological finding that partly overrides Trubetzkoy’s assumption. Phonetically, a ternary vocalic contrast also exists in Soikkola, both in initial and non-initial syllables. There are in total three quantitative vowel types in the non-initial syllables: short/reduced, lengthened and long: [le·ppä̆] ‘alder:nom’ vs. [le·ppä̀] ‘alder:part/ill’ vs. [ta·bā] ‘kill:imp’. However, the long allophone is present only in the VCV2 nucleus, where no other allophonic types are possible. It does not make a phonological opposition with a short and a lengthened type. In the initial syllables, there are also three types of vowel allophones: short, long and overlong (Markus  2011:  111). However, the situation here is similar to the one in Sámi or Estonian. There are four types of ratios between these vowels and the following consonants (ibid.). These ratios are essentially the same as the ones given previously for Estonian: long – long ([vū·ttă] ‘year:part’), short – overlong ([ku·k̀kă] ‘flower:nom’), overlong – short ([sā̀·da] ‘get:inf’), short – short ([ta·bā] ‘kill:imp’). Therefore, the Soikkola threeway contrast of vowels, unlike the consonantal contrast, completely follows a common Finno-Sámi pattern.

3 The Lower Luga dialect: Reduced voiceless vowel phonemes 3.1 C  ontrast of full modal and reduced voiceless vowels in Southern Lower Luga varieties In Lower Luga Ingrian, the reduction of non-initial vowels is the most salient of all the Ingrian dialects. The most innovative varieties in this respect are the southern subdialects immediately adjacent to the Estonian language area. Their system is three steps ahead of the Soikkola system (cf. Kuznetsova 2011: 192). In the first stage, originally short non-initial vowels underwent quantitative reduction and began to elide in rapid speech. Thereafter, some non-initial short vowels underwent qualitative reduction to [ə] and were no longer restored to full vowels. Lengthened non-initial vowels were often pronounced as short vowels.

102 

 Natalia Kuznetsova

Fig. 4: kuk̄ŏ [ku·kkŏ̥]

Finally, in the third stage, a system of reduced vowel phonemes (ĭ, ĕ, ü̆, ö̆, ŏ, ŭ, ə) emerged out of the short non-initial vowels in certain positions. All non-initial lengthened vowels (apart from the second vowel in VCV2 nucleus) were shortened completely. As a result, synchronically there exists a phonological short vs. long vowel contrast in the initial syllables of stem morphemes and a short vs. reduced vowel contrast in the non-initial syllables: tuli ‘fire:nom’ vs. tūlĭ ‘wind:nom’ vs. sūt’i ‘judge:prs:3sg’. Non-initial short vowels are realized as lengthened in VCV2 nuclei ([tu·lì]) and as short in other positions ([sū·t’i] etc.). In no context the three phonetic durational classes of vowels make a ternary opposition, which is similar to the situation in Soikkola presented in the previous section. Reduced vowels of southern Lower Luga varieties are a typological rarity as they are often realized as voiceless ([tū·lı ]̥̆ ). At the same time they render independent phonemes opposed to short non-initial modal vowels (i, e, ü, ö, o, u, a, ä). For example, the Soikkola contrast [ku·kkŏ] ‘rooster:nom’ vs. [ku·k̀kò] ‘rooster. part/ill’ is realized in southern Lower Luga varieties as [ku·kkŏ̥ (~  ku·k̄)] vs. [ku·kko], where the second syllable vowels are the sole carriers of a distinctive function. The pronunciation of these words in the speech of NDP (a male, born in 1924 in Vanakülä) is illustrated in the spectograms in Figures 4 and 5. From an articulatory perspective, the configuration of the vocal tract is the same as during voiced phonation, but instead of the periodical vocal-fold vibration there is either a silent smooth laminar or a turbulent airflow through the (relatively) widely abducted glottis. Depending on a degree of constriction on a scale of decreasing width of glottal opening, voiceless pronunciation can be

Two phonological rarities in Ingrian dialects 

 103

Fig. 5: kuk̄o [ku·kko]

either “nil”, “breathy” or “whisper” (La­ver 1994: 189–191). Acoustically, voiceless vowels do not have F0. The F2 and F3 of breathy and whisper phonations are at the same frequencies as for modal vowels, with an input that is irregular and noisy rather than pulsed. Notably, the Lower Luga reduced vowels are not always pronounced as completely voiceless, especially in slow speech. There can be a short period of modal phonation at the beginning of a vowel. Lower Luga reduced voiceless vowels were first described in Mägiste (1925: 82, 85), and are also mentioned in Ariste (1965, 1969: 173).

3.2 Typological data on voiceless vowels To date, there are no known languages in which the phonological contrast of voiced and voiceless vowels could be incontrovertibly established. Such contrast might exist in the Kewa (Santo Domingo) and Tamaiya (Santa Ana) dialects of Eastern Keresan (Keresan family; USA) and in Malto (Paharia; Dravidian family; India). Moreover, this contrast has been reconstructed for Proto-Keresan. Additionally, Ik (Kuliak family; Easter Uganda), Nishi (Dafla; Sino-Tibetan family; India) and, with lesser certainty, Ute (Southern Paiute) and Comanche (Uto-Aztecan family; USA) are often mentioned as possibly having such a contrast (Mahapatra 1979: 203–205, Ja­kob­son and Waugh 1987: 138–139, with a reference to J. Greenberg; Ladefoged, and Maddieson 1996: 315; Gordon 1998; Blevins 2004: 199–201). Voiceless vowels with “nil” phonation are assumed for Shiwiar (Jivaroan family; Equador) by Kohlberger (2014; oral communication).

104 

 Natalia Kuznetsova

3.3 Lower Luga voiceless vowels against the typological background Blevins (2004: 199) describes two common sources of vowel devoicing. Very short vowels are either devoiced when adjacent to voiceless consonants (obstruents, especially h) or independently; in the latter case usually word- or phrase-finally. In some North Germanic and Sámi languages, vowels can undergo complete or partial devoicing resulting from the preaspiration of the following stops. Skolt Sámi, Icelandic, Faroese, and certain Norwegian and Swedish dialects, which manifest this phenomenon, have even been analyzed by some scholars as having phonologically voiceless vowels, though the assumption is generally contested (cf. Liberman 1982: 90–117; McRobbie-Utasi 1991). Lower Luga voiceless vowels have nothing to do with this case as they emerged according to the second scenario, through reduction. Lower Luga, as the other Ingrian dialects, has a foot structure, though the prosodic unity of a foot is not so salient any longer. Voiceless reduced vowels usually occur in the end ̄ ̥̆] ‘girl:nom’, [a·vàhtŭ̥] ‘be_opened:pst:3sg’, of a final or a non-final foot: [tü·tö [su·k̄ŭ̥r’i:s̄] ‘sugar:in’, [li·sä̀h’ü̥̆mi:ne] ‘be_added:nmlz:nom’ etc. An interesting trait of Lower Luga within the typological background is that such vowels also occur in the non-final second syllable of a trisyllabic foot (cf. however initial-only voiceless vowels in Malto and Shiwiar; Mahapatra 1979: 205; Kohlberger, oral communication). For example: [i·hmısed] ̥̆ ‘man:pl’, [jǟ·hü̥̆tet̄ä:] ‘cool:ips’, [ki·skŏ̥humi:nè] ‘be_torn_off:nmlz’, cf. a detailed chart of the positions possible for voiceless vowels in Kuznetsova (2011: 189, 2012a: 59–60). The main problem in postulating voiceless vowel phonemes when they have developed independently of consonantal environments is that they are typically also very short. This is also the case in Lower Luga Ingrian. Such vowels can be treated either as phonologically voiceless or reduced, and for Lower Luga Ingrian the latter variant seems preferable. The phoneme /ə/, a mid-central vowel, also occurs in those positions where it cannot elide due to phonotactic and speech production restrictions: [ła·m̄əz] ‘sheep:nom’. In such cases, /ə/ is usually pronounced as a reduced but voiced vowel. Therefore, Lower Luga does not provide a completely pure case of voiceless vowel phonemes either.

4 Evolution of rare phonological contrasts 4.1 Evolution of the ternary contrast A cross-linguistic analysis reveals two typical paths for the further evolution of ternary contrasts. These are either the restoring of a binary contrast or the shift of

Two phonological rarities in Ingrian dialects 

 105

a phonological contrast from the segmental to the suprasegmental level. In the following, the Soikkola material is further considered in respect of both variants.

4.1.1 Binarity restoring Remijsen and Gilley (2008: 340–341) made the following claim regarding the fate of the ternary contrast: Languages can develop a three-way vowel length distinction, but the only way forward, in a diachronic sense, is the way back: once a three-way vowel length contrast has developed, it can be maintained in the face of the pressures of reorganisation, but any further increase in quantity contrasts is unlikely, because the phonetic distance between categories may be too small.

Blevins (2004: 202) has formulated a more concrete typological hypothesis: the merger of the long and extra-long categories as a single “long” category is predicted as a common type of sound change. A brief overview on the ternary contrast evolution in Ingrian dialects is given below in light of this assumption. A ternary contrast exists only in Soikkola, and here not even in all of its varieties. The contrast is claimed to be absent from the transitory Soikkola to Lower Luga variety of Koskisenkylä, and in the eastern corner of the Soikkola dialectal area (Laanest 19786: 123). Phonetic measurements indicate that it has nearly or totally disappeared at least from the southernmost Soikkola varieties adjacent to Koskisenkylä, among others from the Saarove (Gordon 2009: 96) and Venakontsa villages (Markus 2010: 42–44, see Speaker 4). Furthermore, in places where the contrast exists in shorter structures, it has started to vanish from longer foot types and multifoot words. For example, in two-foot words, e. g. [ka·tằ mà:] ‘cover:sup’ (“primary” geminate) vs. [ma·k̀àmà:] ‘sleep:sup’ (“secondary” geminate), the durational difference between the geminates became neutralized even in position (1) (Markus 2010: 48–49). Sovijärvi (1944: 12, 14) reported the ternary contrast also after the second and third syllable, but nowadays it is blurred if ever exists. Here, both “primary” and “secondary” geminates are often pronounced just as short consonants, especially sonorants and laryngeal, e. g. [o·n̆nĕt­t̆ ŏmà:ł] ‘unhappy:sg:ad’ (short) vs. [ū·t­t̆ ĕlem̀a: ~ ū·t­t̆ ĕ­le:m̀a ~ ū·t­t̆ ĕle:ma] ‘wait:prs:1pl’ (“primary” geminate) vs.  [ū·t­t̆ ĕłom̀à: ~ ū·t­t̆ ĕłomà:] ‘ждать:sup’ (“secondary” geminate). A three-way contrast is definitely absent from the Lower Luga dialect. According to Laanest (1987: 288–291), it has also disappeared from the Hevaha and

6 This manuscript was later published in Estonian as Laanest (1986).

106 

 Natalia Kuznetsova

Oredezh dialects. The Soikkola dialect, which maintains the opposition, seems to be the most conservative in this respect. The original development of a ternary contrast in Ingrian was linked to the emergence of “secondary” geminates that yielded durational class 2 in modern Soikkola Ingrian (see Table 2). Initially, it was a phonetic phenomenon (Laanest  1978:  123). “Secondary” geminates have presumably emerged no later than in Proto-Ingrian; in position (1) probably even earlier (cf. Laanest 1966а: 23–29; 1978: 136, 320–323; Sovijärvi 1944: 84–90). This assumption seems plausible, especially given the fact that among Finnic languages “secondary” geminates in positions (2)–(4) exist only in Ingrian. If the contrast dates back to Proto-Ingrian, it was then maintained and phonologized in the Soikkola, Hevaha and partly Oredezh dialects (Laanest  1978: 124). Classes 2 and 3 later merged into one “long” category in Hevaha, Oredezh and, to some extent, Soikkola. The situation is different in the Lower Luga dialect. “Secondary” geminates have consistently developed and later merged with “primary” ones only in position (1). Compare, for example, Lower Luga contrast [ta·pā] ‘kill:imp’ vs. [ta·ppà] both ‘catch:3sg’ and ‘kill:inf’ with the Soikkola one given in 2.2. In positions (2)–(4), in those cases where Soikkola has “secondary” geminates there are nearly always short consonants in the Lower Luga dialect. Cf. Lower Luga pairs (from northern varieties): [mā·tà] ‘sleep:ips’ — [łō·tà] ‘broom:part/ ill’, [ka·tàjà] ‘juniper:part/ill’ — [ka·tàjă] ‘juniper:nom’, [pe·rt’ĭllĕ] ‘room:all’ — [ke·rtŏjă] ‘time:pl:part’. In Soikkola, the first word of each pair would contain a short consonant (class 1) and the second word a “secondary” geminate (class 2). It therefore appears that in the Lower Luga dialect Proto-Ingrian “secondary” geminates in positions (2)–(4) have again become short phonemes. Few exceptional cases of “secondary” geminates outside position (1) have been attested in Lower Luga Ingrian, e. g. [tu·lliz] ( [*pa·dà] vs. [*pa·t̆tā] > [pa·Dà] vs. [pa·ttà]. The morphological factor could thus have served an important premise for the merger of the “long” classes, which came to be in the majority of Ingrian varieties.

8 I do not know any full minimal pairs of this kind, though there are at least several quasi-­minimal ̆ pairs, e. g. [tu·łłèt] ‘come:pc:pst:act:pl’ vs. [tu·łłòt] ‘come:prs:3pl’, [mä·nnèt] ‘go:pc:pst:act:pl’ vs. [mä·n̆nö̀t] ‘go:prs:3pl’.

114 

 Natalia Kuznetsova

5.2.2 Morphological load of the vocalic voicing contrast The full voiced vs. reduced voiceless vowel contrast in southern Lower Luga varieties, on the contrary, bears a heavy functional load. It distinguishes between forms in numerous nominal and verbal paradigms (see an example with nom vs. part/ill forms of the word ‘rooster’ in section 3.1). However, even if reduced voiceless vowels disappear, the forms will not become homonymic. The paradigmatic functional load will shift to the final vowel presence vs. absence contrast. Such a situation is observed in more innovative Estonian (see an example with nom vs. gen forms in section 4.2.1). Lower Luga Ingrian is thus different in this respect from Estonian and Dinka, as discussed by Blevins, where numerous pairs of grammatical forms would become homonymic given the ternary contrast disappearance. In Ingrian, a heavy morphological load of the vocalic voicing contrast should not necessarily become an inhibiting factor of sound change.

6 I ngrian phonological rarities in light of general prosodic development of the respective dialects The Lower Luga dialect, forming a Sprachbund with closely related languages, has the most innovative phonology among the Ingrian dialects. The most advanced of these are the southern varieties adjacent to the area where Estonian is spoken. Reduction of non-initial vowels is especially salient in them and, even compared to other varieties of the same dialect, reaches a new developmental stage. In this stage, a contrast of non-initial short and long vowels completely transforms into an opposition of reduced and short vowels. The Soikkola dialect, in turn, has the most conservative Ingrian phonology. Nevertheless, Soikkola (together with the extinct Hevaha and Oredezh dialects) has developed some phonological innovations which are unknown to the Lower Luga dialect. These innovations, in contrast to the prosodic evolution of the Lower Luga dialect, are linked to the more intensive development of vocalic and consonantal lengthening in certain positions. In this sense, the vectors of phonological development in the Soikkola and Lower Luga dialects used to be partly counterdirectional in the stage when the dialects were taking their contemporary phonological shape. The vector of Soikkola dialectal development was highly “aimed” at developing lengthening, while the one of Lower Luga Ingrian was directed towards developing reduction. The analyzed rarities in their phonology serve as a vivid illustration of this. The Soikkola phonological system reaches the utmost point of length development. More than three degrees of length are generally considered impossible for

Two phonological rarities in Ingrian dialects 

 115

human languages. The Lower Luga phonological system manifests the zenith of the vowel reduction development: extra-short voiceless vowels.

Abbreviations 1 1 person 2 2 person ad adessive С short consonant C̄ long consonant R short sonorant V short vowel V̀ lengthened vowel V1 first syllable vowel V2 second syllable vowel VV long vowel or diphthong int. interviewer ms milliseconds σ standard deviation 3 3 person act active all allative

cnd conditional cmp comparative gen genitive ill illative imp imperative in inessive inf infinitive ips impersonal nmlz nominalization nom nominative part partitive pc participle pl plural prs non-past pst past sg singular sup supine

References Ariste, Paul (1965): Etwas über das Ižorische. Sovetskoje finno-ugrovedenije 3: 183–186. Bals, Berit Anne, David Odden, and Curt Rice (2007): The phonology of gradation in North Saami. http://www.ling.ohio-state.edu/~odden/SaamiGradation2007.pdf (accessed June 4, 2014). Blevins, Juliette (2004): Evolutionary phonology: The emergence of sound patterns. Cambridge: Cambridge University Press. Bye, Patrik, Elin Sagulin, and Ida Toivonen (2009): Phonetic duration, phonological quantity and prosodic structure in Inari Saami. Phonetica 66: 199–221. Eek, Arvo (1973): Observations in Estonian palatalization: An articulatory study. Estonian Papers in Phonetics 1973: 18–36. Eek, Arvo and Einar Meister (2003–2004): Foneetilisi katseid ja arutlusi kvantiteedi alalt 1–2. Keel ja Kirjandus 11, 2003: 815–837. 12, 2003: 904–918. 4, 2004: 251–271. 5, 2004: 337–357. Gil, David, Peter Trudgill, and Geoffrey Sampson (eds.) (2009): Language complexity as an evolving variable. Oxford: Oxford University Press. Gordon, Matthew (1998): The phonetics and phonology of non-modal vowels: A cross-linguistic perspective. Berkeley Linguistics Society 24: 93–105.

116 

 Natalia Kuznetsova

Gordon, Matthew (2009): Prominence and gemination in Ingrian. Linguistica Uralica 45(2): 81–100. Jakobson, Roman and Linda R. Waugh (1987): The sound shape of language. In: Roman Jakobson, Selected writings 8. Completion vol. 1: Major works, 1976–1980, 1–316. Berlin, New York and Amsterdam: Mouton de Gruyter. Kohlberger, Martin (2014). Non-nasal nasals: Reinterpreting nasalisation in Shiwiar (Jivaroan). In: Sound Change in Interacting Human Systems: 3rd Biennial Workshop on Sound Change. University of California, Berkeley, May 28–31, 2014. http://linguistics.berkeley. edu/SCIHS/abstracts/4_FridayMorning/Kohlberger.pdf. Kuznetsova, Natalia V. (2009a): Fonologičeskije sistemy ižorskih dialektov. Diss. kand. filol. nauk. St. Petersburg: ILI RAN. Kuznetsova, Natalia V. (2009b): Suprasegmentnaja fonologija soikinskogo dialekta ižorskogo jazyka v tipologičeskom aspekte. Voprosy jazykoznanija 5: 18–47. Kuznetsova, Natalia V. (2011): Typology of prosodic systems in Low Luga Izhorian varieties. In: Peter Austin, Oliver Bond, David Nathan, Monik Charette, and Peter Sells (eds.), Proceedings of Conference on Language Documentation and Linguistic Theory 2, 185–192. London: SOAS. Kuznetsova, Natalia V. (2012a): Prosodika slovoformy v nižnelužskom dialekte ižorskogo jazyka. In: Acta linguistica Petro­politana 8(1), 43–103. St. Petersburg: Nauka. Kuznetsova, Natalia V. (2012b): Vowel reduction in Lower Luga Ingrian: scientific description and “folk” perception. In: Evgeniy V. Golovko, Mikhail L. Lurie, Mehmed  Z. Muslimov, and Sofia  A. Oskolskaya (eds.), ’Folk linguistics’: Language from speakers’ perspective. Paper abstracts for the conference, St. Petersburg, November 19–21, 2012, 85–87. St. Petersburg: Nestor-Istorija. Kuznetsova, Natalia, Elena Markus, and Mehmet Muslimov (2014; in press). Finnic minorities of Ingria: The current sociolinguistic situation and its background. In: Heiko Marten, Michael Riessler, Janne Saarikivi, and Reetta Toivanen (eds.). Cultural and linguistic minorities in the Russian Federation and the European Union. (Multilingual Education 13: Comparative studies on equality and diversity). Laanest, Arvo (1966a): Ižoskie dialekty. Lingvogeografičeskoje issledovanije. Tallinn: Institut jazyka i literatury AN SSSR. Laanest, Arvo (1966b): O vzryvnyh v fonologičeskoj sisteme hevaskogo dialekta ižorskogo jazyka. Sovetskoje finno-ugrovedenije 2: 19–21. Laanest, Arvo (1978): Istoričeskaja fonetika i morfologija ižorskogo jazyka. Diss. kand. filol. nauk. Tallinn: Institut jazyka i literatury AN SSSR. Laanest, Arvo (1986): Isuri keele ajalooline foneetika ja morfoloogia. Tallinn: Valgus. Laanest, Arvo (1987): Zur Phonologie der ingrischen. In: Studien zur Phonologie und Morphonologie der uralischen Sprachen: Eisenstadt, 28. Juni — 1. Juli 1984: Akten der dritten Tagung für uralische Phonologie, 285–293. Wien: Verband der wissenschaftlichen Gesellschaften Österreichs. Ladefoged, Peter and Ian Maddieson (1996): The sounds of the world’s languages. Oxford and Cambridge: Blackwell Publishers. Laver, John (1994): Principles of phonetics. Cambridge: Cambridge University Press. Lehiste, Ilse (1960): Segmental and syllabic quantity in Estonian. In: American Studies in Uralic Linguistics , 21–82. Bloomington: Indiana University. Lehiste, Ilse, Pire Teras, Valts Ernštreits, Pärtel Lippus, Karl Pajusalu, Tuuli Tuisk, and Tiit-Rein Viitso (2008): Livonian prosody. Helsinki: Suomalais-ugrilainen seura.

Two phonological rarities in Ingrian dialects 

 117

Liberman, Anatoly S. (1982): Germanic accentology, Volume 1: The Scandinavian languages. Minnesota: University of Minnesota Press. Liiv, Georg (1961): Eesti keele kolme vältusastmevokaalide kestus ja meloodiatüübid. Keel ja Kirjandus 7: 412–424. 8: 480–490. Lippus, Pärtel (2011): The acoustic features and perception of the Estonian quantity system. Tartu: Tartu University Press. Mägiste, Julius (1925): Rosona (Eesti Ingeri) murde pääjooned. Tartu: E. K.Ü. “Postimehe” trükk. Mahapatra, Bijaya P. (1979): Malto – An ethnosemantic study. Mysore: Central Institute of Indian Languages. Markus, Elena (2010): Primary and secondary geminates in Ingrian. Linguistica Uralica 46(1): 38–52. Markus, Elena (2011): The phonetics and phonology of a disyllabic foot in Soikkola Ingrian. Linguistica Uralica 47(2): 103–119. McRobbie-Utasi, Zita (1991): Preaspiration in Skolt Sami. Simon Fraser University Working Papers in Linguistics 1: 77–87. McRobbie-Utasi, Zita (1996): An acoustic analysis of durational interdependences in Skolt Saami. In: Congressus Octavus Fenno-Ugristarum: Jyväskylä 10.–15. 8. 1995. Pars III: Sessiones sectionum: Phonologia & Morphologia, 28–32. Jyväskylä: Jyväskylän yliopistopaino. McRobbie-Utasi, Zita (2007): The instability of systems with ternary length distinctions: The Skolt Saami evidence. In: Ida Toivonen and Diane Nelson (eds.), Saami linguistics, 167–205. Amsterdam and Philadelhia: John Benjamins. Muslimov, Mehmed Z. (2005): Jazykovyje kontakty v Zapadnoj Ingermanlandii (nižneje tečenije reki Lugi). Diss. kand. filol. nauk. Sankt Peterburg: ILI RAN. Remijsen, Bert and Leoma Gilley (2008): Why are three-level vowel systems rare? Insights from Dinka (Luanyjang dialect). Journal of Phonetics 36: 318–344. Sammallahti, Pekka (1998): The Saami languages: An introduction. Kárášjohka: Davvi Girji. Sidorkevich, Daria (2011): On domains of adessive-allative in Siberian Ingrian Finnish. In: Acta linguistica Petro­politana 7(3), 575–607. St. Petersburg: Nauka. Sidorkevich, Daria (2013). Fonologičeskaja sistema sibirskogo ingermanlandskogo idioma. In: Acta linguistica Petropolitana 9(2), 657–702. St. Petersburg: Nauka. Sovijärvi, Antti (1944): Foneettis-äännehistoriallinen tutkimus Soikkolan inke­rois­murteesta. Helsinki: Suomalaisen kir­jallisuuden seura. Tauli, Valter (1956). Phonological tendencies in Estonian. PhD diss. (Det Kongelige Danske Videnskabernes Selskab. Historisk-filologiske Meddelelser 36(1)). Copenhagen: Blanco Lunos Bogtrykkeri A-S. Trubetzkoy, Nikolai S. (2000) [1939]: Osnovy fonologii. Moskva: Aspekt Press. Trudgill, Peter J. (2001): Contact and simplification: Historical baggage and directionality in linguistic change. Linguistic Typology 5: 371–374. Viitso, Tiit-Rein (1975): Outlines of Livonian phonology. In: Symposion Phonolo­gi­sche Analyse der uralischen Sprachen Berlin 17.–20. September 1974, 83–113. Berlin: Aka­de­mie der Wissenschaften der DDR. Zentralinstitut für Sprachwissenschaft. Viitso, Tiit-Rein (1997): The prosodic system of Estonian in the Finnic space. In: Ilse Lehiste and Jaan Ross (eds.), Estonian prosody: Papers from a symposium: International Sym­posium on Estonian Prosody, Tallinn, Estonia, October 29–30, 1996: Proceedings, 222–234. Tallinn: Institute of Estonian language. Zaiceva, Maria I. (1981): Grammatika vepsskogo jazyka (fonetika i morfologia). Leningrad: Nauka.

Kristján Árnason

Analysing phonological variation in Faroese Abstract: This paper focuses on ways and means of investigating phonological variation in the Faroe Islands in a comparative light and with reference to the history and sociolinguistic conditions of the Faeroes. I will compare the problems of analysing Faroese variation to those involved in studying similar phenomena in Iceland on the one hand and Continental Scandinavia and Europe in general on the other. The sociolinguistic history and present day situation in the Faeroes is fundamentally different from that in Iceland, and there are similarities to (but also differences from) the typical European situation. Using material gathered in the Faeroes in 2008, I will argue that in spite of the different conditions it is possible to make a useful survey of present-day phonological variation in Faroese with the same methodology that was used in a survey carried out in Iceland in the 1980s.

1 Icelandic – Faroese – Norwegian Common structural features of Icelandic and Faroese include (relative) morphosyntactic conservatism, and some phonological characteristics which are due to similar historical changes from Old Norse or Common West Nordic. Sound shifts such as the quantity shift and the widespread diphthongisation of vowels, and what may be called the West Nordic consonant shift, among other things involving preaspiration and prestopping in sonorants, have left their mark on the structure of the languages (cf. Árnason 2011a: 11‒34). A further common characteristic is the lack of word tones (corresponding to the Scandinavian Accent 1 and 2 and the Danish stød), although it seems just as likely that this is not a common innovation, rather an absence of a continental Scandinavian feature. (This is a matter of debate, see e. g. Myrvoll and Skomedal 2010) But the differences between the languages are also considerable. Although both Icelandic and Faroese have preaspiration, there are significant differences in the function and realisation of preaspirarion within the phonological systems, (see Schäfer and Árnason 2012) and the results of diphthongisation differ, the most noticeable feature of Faroese perhaps being the high diphthongs in hvítur [kʰʊiːtʊɹ] ‘white’ and hús [hʉuːs] ‘house, which are absent from Icelandic. A fundamental phonological difference between the two insular languages lies in the fact that Faroese has two different types of syllable, full syllables, and what I have

Analysing phonological variation in Faroese 

 119

called in “restricted syllables”, (also called levis or levissimus). And within the full syllables, which have long or short vowels (V: nuclei or VC nuclei), Faroese shows greater differences in quality between the “long” and “short” vowel paradigms than Icelandic does. There is thus a clear polysystemic structure in Faroese vowels, and it can be argued that there are three vowel systems, one for open full syllables, one for closed full syllables and the third for the restricted syllables (cf. Árnason 2011a: 68‒97). There are also some holistic or over-arching differences in articulatory setting between the two languages. Thus, palatalization has reached its logical conclusion in Faroese, resulting in a palato-alveolar initial consonant in kemur [ʧeːmʊɹ] ‘comes’, compared to the dorso-palatal character of Icelandic kemur [cʰɛːmʏr̥] ‘comes’. And also, Faroese has a palato-alveolar sibilant in skip [ʃiːp] ‘ship’, as well as retroflexion, which among other things means that the element I (‘diffuse spectral composition’) is more prominent in the Faroese consonant system than in the Icelandic one. On the other hand, rounding (corresponding to the element U, ‘flatness or low spectral peak’), is more central in the Faroese vowel system than in the Icelandic one. (See Árnason 2011a: 41‒43 and references for descriptions of such ‘phonological elements’.) This is reflected in, among other things, the fact that MF has front rounded [y] and rounding occurs in all Faroese diphthongs except /ei/ in deyður ‘dead’ and (in some dialects) /ai/ in feitur ‘fat’ (cf. Árnason 2011a: 76‒77). Some Faroese dialect differences are also based on rounding. Thus, we shall see below that one of the most prominent ones is the choice between rounded /ɔi/ in the north and unrounded /ai/ in the south in words like steinur ‘stone’. Since both Iceland and the Faeroes were settled from Norway, Norwegian, especially western Norwegian, is the closest relative of Faroese and Icelandic. And indeed some phonological phenomena similar to those of Icelandic can be found in western Norwegian dialects (cf. e. g. Chapman 1962; Sandøy 2003; Árnason 2012). But due to historical events, there is less phonological common ground, so to speak, in the modern language, but as will be shown in Section 3, the sociolinguistic and historical background of the Faeroes is in some respects similar to that of Norway. Based on the phonological characteristics briefly outlined above, tackling phonological variation in Faroese involves identifying form variants which can be used as variables in a statistical analysis of language use (call this Problem A). The second problem (Problem B) involves making sense of the sociolinguistic climate or ecology of present day Faroese and looking for sensible ways of accounting for the external factors involved. In the hope of eventually grasping the essential characteristics of the modern Faroese situation, I will submit in

120 

 Kristján Árnason

what follows a number of observations, which seem to be essential for any sort of attempt at an analysis, starting with Problem A.

2 Defining variables and variants As is to be expected, phonological differences between Faroese varieties or lects are basically due to historical changes which have been aborted or did not make it to all locations. Among these aborted changes or trends is the diphthongisation of the old long back vowel in bátur ‘boat’, which has given the diphthongal variant [pɔɑːtʊɹ] in southern varieties, but [paːtʊɹ] or [pɑːtʊɹ] in northern parts. And similarly, it is possible to see the different realisation of the intervocalic plosive in this same word, (pre)aspirated in northern parts ([p(ɔ)ɑːʰtʊɹ]), but unaspirated in the south: [pɔɑːtʊɹ], as part of the result of the West Nordic consonant shift. This shift was thus not completed, and the same is true of Icelandic; the incomplete shift has left its mark on Icelandic in the difference between harðmæli and linmæli (cf. Árnason and Thráinsson 2003; Thráinsson and Árnason 1992). (The enhancement of the H element or [spread glottis] feature also leads to preaspiration in other environments, both in Icelandic and Faroese, e. g. in words like hattur ‘hat’, MI [hahtʏr̥a, MF [haʰtːʊɹ].) Both the diphthongisation and aspiration are parts of general phonological trends, which are realised in different ways in the two languages and in different areas within the speech communities. The interesting question is then how, why, and when the phonological differences become significant in linguistic variation and develop into markers of different ways of speaking, to be investigated as sociolinguistic phenomena. Why did the diphthongisation of old /á/ ([ɑː]) lose its impetus before making it to the northern isles? Does this have to do with structural or purely linguistic forces, or is the effect an external or a functional one in a broad sense, to do with social or geographic conditions. (Did the northern islanders (Norðuroyafølk) “decide” that this diphthongal sound, which they may have heard from southerners, was not something that they wanted to adopt?) It seems that in most cases something special is needed in order for a linguistic feature to become focused in this way or its distribution sensitive to socially determined variation; it has to acquire some value as either a marker or indicator in the Labovian sense. But it is perhaps not clear what this “special” thing is exactly. Is it, from the speakers’ point of view, some sort of unconscious or semi-conscious value which the linguistic feature acquires in the social or geographic context, and to which the speaker is sensitive? Are speakers (semi-) consciously aware of the differences? And how would this sort of folk linguistic awareness compare to the “scientific” awareness of the linguists investigating the

Analysing phonological variation in Faroese 

 121

variation? (Is it even possible that some of the variants and varieties which have been identified in dialect studies are constructs of the researchers in cooperation with the informants, as suggested by Auer 2010.) A number of regional phonological characteristics have been identified and reported in the literature on modern Faroese (see e. g. Lockwood 1955; Weyhe 1988; Thráinsson et al. 2004: 338‒367; Árnason 2011b). As already mentioned, one of the most prominent of these is the north-south variation in the pronunciation of the vowel written in words like steinur ‘stone’, as /ɔi/ in the north and /ai/ in the south: [stɔiːnʊɹ] vs. [staiːnʊɹ]. A result of the fact that the dialect differences originate in partially fulfilled trends, many of these variables have very complex definitions from a structural point of view. This is the case, e. g. for the Icelandic devoicing before “hard plosives” in words like hjálpa ‘help’ and velta ‘roll’, and also for the deletion of /g/ between /n/ and /l/ in forms such as englar ‘angels’ (cf. Árnason 1990; Árnason and Thráinssson 2003). And this sort of complexity is well known in other parts of the world when it comes to dialectal variation, cf. e. g. the deletion of /d/ after the sonorants /l/, /n/ or /r/ in German dialects in forms like Hunde ‘dogs’, bald ‘soon’ and werden/geworden ‘become’ (cf. Streck 2012: 79‒136). The phonological force behind this dialect feature (or features, because the isoglosses do not coincide) is a phonological trend, the deletion of /d/ after a sonorant, but the variation which has been documented by dialectologists for more than a hundred years can only be described by referring to individual lexical items or indeed inflectional forms. In fact, in “well established” dialect vs. standard situations, such as Baden-Württemberg in southern Germany, it seems that nowadays most dialect features which are plotted in linguistic atlases are lexicalised, i. e. best described in terms of individual words or word forms. A Faroese case showing this type of thing, i. e. a structurally complex definition of a variable, is the fronting and delabialisation (“brightening”, cf. Árnason 2012) of historical /o:/ in Faroese words like skjótur ‘quick-Masc’ and skjótt ‘quickNEUT’ and ljótur ‘ugly’ and ljótt ‘ugly-NEUT’. Relative to the Old Norse back rounded mid high vowel ó [oː] and the Modern Icelandic [ou], both the short and the long variant have been fronted in parts of the Faroese linguistic area. The long variant is diphthongal like the Icelandic one, but in the northern areas, the first component has been fronted and even delabialized: skjótur /[ʃœuːtʊɹ] /[ʃɛʊːtʊɹ]; ljótur [λœuːtʊɹ]/[λɛʊːtʊɹ]. This brightening does not take place in the southern islands, where the pronunciation is [ʃɔuːtʊɹ] and [λɔuːtʊɹ], and the isogloss for this variable lies across Streymoy north of Tórshavn. For the short correspondent in skjótt and ljótt, the choice is between /œ/ and /ɔ/ ljótt [λœʰtː] vs. [λɔʰtː] ‘uglyNEUT’. And again the front variant is a northern characteristic, and the back colour a southern one. But the important thing to note is that the isogloss for the

122 

 Kristján Árnason

front vs. back variant for the short vowel does not coincide with the isogloss for the long vowel. For the short vowel the isogloss lies south of Tórshavn, whereas, as noted above, the one for the long vowel lies north of Tórshavn. This means that as far as fronting of old West Nordic /o:/ is concerned, Tórshavn speakers use a “mixture” of southern and northern forms. This must mean that the variables have been lexicalised, not only as parts of lexemes, but in the respective inflectional forms of the word skjótur ‘quick’ and ljótur ‘ugly’.

Fig. 1: The Faroe Islands1

Another lexicalised north vs. south variable involves vowels before a velar nasal in words like mangur ‘many’ and blankur ‘bright, shining’; northern varieties have a front vowel: [mɛŋkʊɹ] and [plɛŋ̊kʊɹ] whereas south of Skopunarfjørður

1 Copyright, Umhverfisstofan: http://www.us.fo/Default.aspx?ID=5764.

Analysing phonological variation in Faroese 

 123

these forms have back vowels: [maŋkʊɹ], [plaŋ̊kʊɹ]. A somewhat similar difference is that on Suðuroy, forms showing reflexes of the skerpingin (Verschärfung), by which stops develop in historical hiatus (cf. Árnason 2011a: 31‒33), have back rounded vowels: skógvur ‘wood’ and nógv ‘enough’, i. e. [skɔkvʊɹ] and [nɔkv], as opposed to the more common and northern front and unrounded [jɛkvan] and [nɛkv]. Turning to consonantal structure, a special characteristic of the speech of Suðuroy involves the distribution of the palatal glide /j/ and the palato-alveolar stop /ʧ/ (or perhaps voiced /ʤ/). In most parts of this island, words such as gjógv ‘a crevice’ and geva ‘to give’ have initial /j/, i. e. [jɔkːv], [jeːva], compared to the more common [ʧɛkːv] [ʧeːva]. But conversely, in the variety spoken in the middle of Suðuroy, e. g. in Tvøroyri, these forms, and such forms as have /j/ in other parts (e. g. Jógvan and jól), have a stop /ʧ/. Thus, in this particular variety Jóggvan is pronounced [ʧɔkːvan], and hjarta ‘heart’ and jól ‘Christmas’ are pronounced: [ʧaɻ̥ta] and [ʧɔuːl] in contrast to the more common [jaɻ̥ta] and [jɔuːl]. As a consequence, the words gjørð ‘girdle’ and jørð ‘earth’ are homophonous in most indigenous varieties of Suðuroy, but with two different realisations. In the middle area they both have /ʧ/: [ʧœːɹ], but in most other parts of the island they have /j/: [jœːɹ] (cf. Weyhe 2003: 102, 140‒142, 192‒193). But there are other types of variation which do not seem to be lexicalised in this way. Thus, there is a lot of fluctuation in the realisation of the so-called restricted syllables, such as the final syllables of undir ‘under’, bilur ‘car’ and gamal ‘old’, and the non-initial syllables of trisyllabic forms such as húsinum ‘the house-DAT’. Here, there are varieties where the difference between the three vowel colours (/i/, /u/, /a/) has been neutralised, whereas in other varieties, the difference is still upheld (to some extent). And on top of this, there is a strong tendency in informal styles for reducing the syllables, yielding forms such as [piːlɹ̥] and [untɹ̥] for bílur and undir. This partly stylistic variation must be seen as postlexical in some sense at least, e. g. in being sensitive to speech rate and style, and it is perhaps not localised geographically in the same way as some of the older types of variants, although we will see below that they are subject to sociolinguistic variation and they even figure in the folk linguistic discourse on different ways of speaking. Another variable feature that should probably be characterised as postlexical is the raising and diphthongisation of vowels before hiatus in words like dagar ‘days’ [tiːjaɹ] ‘days’ and boða [puːwa] ‘to preach’. Forms of this type have been described in the handbooks as having mid-low (slightly raised?) vowels before hiatus: [tęːaɹ] and [pǫːa] (cf. Lockwood 1955: 14; Thráinsson et al. 2004: 39‒40). But a traditional variant pronunciation in the dialect of Vágar has diphthongal stem vowels in forms like glaða ‘happy-ACC.FEM’ and laga ‘to make’ hagan

124 

 Kristján Árnason

‘skilled-ACC.MASC’: [lei:ja], [klei:ja] and [hɛijan] (as opposed [lęːa], [klęːa] and [hęːan]) (see Petersen 1996; cf. also Jacobsen 2001). And similarly, words like fáar ‘few’ are in this dialect pronounced as [fɔuːwaɹ], corresponding to [fǫːaɹ] in other varieties, and also forms such as kvøða ‘to chant’ and høga ‘high-ACC.FEM’ can, among older speakers in Vágar, have the form [kʰvʉuːwa] and [hʉuːwa], although due to influence form other areas younger speakers are reported as tending to use more common variants, i. e. [kʰvøːa] and [høːa] (Petersen 1996: 13). But according to my own observations it is common for the first vowel in these forms to be fully raised so that the forms are pronounced as e. g. [kliːja] for glaða ‘happy’, [fuːwaɹ] for fáar ‘few’ (Árnason 2011a: 82‒5). This phenomenon is probably recent and seems not to have been investigated systematically, but the impression is that this is very common indeed among younger speakers.2 I have already mentioned the north vs. south variation of “hard” (in the north) vs. “soft” (in the south) pronunciation in words like pápi [pʰɔɑː(ʰ)pɩ] ‘daddy’, and bátur [pɔɑː(ʰ)tʊɹ] ‘boat’, mostly based on pre-aspiration, and then only following low vowels (see e. g. Weyhe 1997; Helgason 2002; Thráinsson et al. 2004: 345; Árnason 2011a: 118‒21). In the “hard” variant, these sounds have preaspiration, realised as a relatively weak ‘puff of air’ between the vowel and the closure: [pɔɑːʰtʊɹ], [pʰɔɑʰpɩ]. But there is considerable variation in the realisation of Faroese preaspiration, and it seems likely that the strength of aspiration may vary with style, so that this variable might in some sense be postlexical (cf. Schäfer and Árnason 2012). Several observations have been made in the literature regarding regional differences in Faroese intonation. Thus the intonation in Vágar is known to be special, and in folk linguistics this characteristic goes by the name of drynjan

2 There are further phonological dialect differences which cannot be properly discussed here for reasons of space. But it can be mentioned that the peripheral areas, both Suðuroy and Norðuroyggjar, show some special characteristics. Thus, words like fær ‘gets’, far ‘vessel’ and maður ‘man’ are pronounced with a monophthong [fɛːɹ] and [mɛːʋʊɹr] on the northernmost Fugloy, as opposed to the majority pronunciation [fɛaːɹ] and [mɛaːʋʊɹ]. At the other geographical end, a special feature of the dialect of Suðuroy is the monophthongal /e/ in words like læra [leːɹa] ‘to learn’ and fær [feːɹ] ‘gets’ and fer [feːɹ] ‘goes’, which, as shown by the spelling, corresponds historically either to Old West Nordic mid high /e/ or mid-low /æ/. Thus, on Fugloy, there is a merger between historical forms, and a lack of diphthongisation. However, the special features of Suðuroyamál do not affect the number of diphthongs, since the etymological /a/ has the diphthong [ɛaː] in the word far [fɛaːɹ] ‘vessel’. The distinction between fær ‘gets’ and fer ‘goes’ on the one hand and far ‘vessel’ on the other here reflects a historical raising of the old open /æ/, involving a merger of this sound with etymological /e/ instead of /a/, which latter lies behind the situation in the other varieties, including that of Norðuroyar, which have a merger but no diphthongisation of the outcome.

Analysing phonological variation in Faroese 

 125

or drynjing ‘grunting’. Very few speakers can now be identified as having this feature, and it is not too clear what linguistic characteristic is actually involved. According to Hagström (1967: 44) the tonal melody so characterised is an evenly falling curve with lengthened syllables and a clear pronunciation of the endings using the same tonal curve as the preceding syllable. According to Eivind Weyhe (p. c.) the intonational contour may also have involved a HLH pattern at the end of utterances. Hagström (1967) also notes that the intonation in Suðuroy may be different from other places, being characterised by constant alternation between up and down moving melodies. Although this needs further study, it seems that some informants from Suðuroy, taking part in the preliminary survey described below, have intonational patterns that might fit Hagström’s description. Impressionistically, this sounds as though it could be due to frequent use of Low High contours in the middle of utterances, which might analysed as L*H, i. e. pitch accents with relatively late rise associated with stressed syllables. (See Árnason, 2011a: 324‒26 for some further observations.) There is thus a lot to be done when it comes to analysing phonological variation in Faroese. But in spite of the richness in the material, there seem to be good grounds for compiling a list of variables which may then be plotted against external variables, such as sex, age, social background, education and location. And it will be shown below that a similar methodology can be used in analysing the Faroese situation as was used in analysing Icelandic variation, notably in the 1980s (cf. Árnason and Tráinsson 2003). This involved the identification of two types of phonological variables. On the one hand, there were traditional regional variants, which had been investigated earlier by Björn Guðfinnsson in the 1940s (see Guðfinnsson 1946, 1964). Among such variables was the above-mentioned hard speech (harðmæli), the difference between aspiration or no aspiration in internal stop onsets in forms such as bátur ‘boat’, taka ‘to take’ and hlaupa ‘to run’. Another somewhat related variable involves the voicing or not of sonorants before stops in words such as hjálpa ‘to help’, vanta ‘to need’ and heimta ‘to demand’. Still among the traditional Icelandic dialect markers is the so-called and highly stigmatised “slack-jawed speech”, which involves the merger of mid high and mid-low vowels in words such as bitur ‘bitter’ and betur ‘better’ and flugu ‘fly-OBL’ and flögu ‘flake-OBL’. But, just as in the case of Faroese, there are also more recent types of variation to be found in Icelandic, such as the “stop pronunciation”, changing the cluster [xs] to [ks] in forms such as vaxa ‘to grow’ and hugsa ‘to think’ ([vaxsa] > [vaksa], [huxsa] > [huksa]) and the debuccalisation of plosives before /n/ (/l/) in words such as Bjarni ‘a man’s name’: [pjatnɪ] > [pjaʔnɪ]. And somewhat comparable to the reduction of unstressed syllables in Faroese, several types of slurring occur in Icelandic, i. e. in forms such as dagblað ‘newspaper’ [taːɣplaθ] > [taːplaː] and hjólinu ‘the bike-DAT’ [çoulɪnʏ] > [çoulnʏ].

126 

 Kristján Árnason

3 Sociolinguistic and historical aspects Turning to Problem B, the question of how to make sense of the sociolinguistic situation in the Faeroes, it seems that although in some respects the conditions are similar to those in Iceland, there are also some fundamental differences.3 As described in e. g. in Árnason (2011b; see also Leonard 2009; Árnason 2003a) there is a tradition in Iceland of a very high level of “linguistic consciousness” and overt metalinguistic or folk linguistic discussion and a stringent language policy. And it may well be that this high level of linguistic consciousness has been influential in the historical development. The main story regarding Icelandic phonological development is that linguistic innovations were either accepted so that they spread over the whole linguistic area, or they were “turned back”, so that they disappeared again. A good example of the latter is the overt stigmatisation in the 20th century of the above mentioned “slack-jawed speech”, involving the merger of (mostly long) mid-high front vowels /ɪ/ and /ɛ/ and /ʏ/ and /œ/. This has resulted in the virtual extinction of this variant (see Árnason 2003b; Árnason and Thráinsson 2003). The main influence behind this is probably the existence of a relatively stable written norm, which was developed in the 12th century and elaborated e. g. in the Icelandic sagas. This norm has been and still is considered by many to be the “correct” form of Icelandic, to which any variant form must be compared. An important part of the linguistic culture of any society is its folk linguistics. Reflecting the Icelandic linguistic “atmosphere”, many of the most widely-used terms in Icelandic popular linguistic discussion have to do with the “good” or “bad” quality of usage. It is considered to be important to use “correct” or “pure” Icelandic forms (rétt mál, hreint mál). And when it comes to alternative ways of speaking, it is noticeable that there is no term corresponding to dialect in the typical European sense (or for German: Mundart). The Icelandic term mállýska refers primarily to linguistic features or variant formal options, which may be associated with geographic locations, but without the involvement of the social or functional connotation implicit in the typical European conception of dialect as opposed to a standard (cf. Auer 2005). The Icelandic term does not refer to clearly defined local varieties. Another commonly used term is málfar ‘usage’, which can be ‘good’ or ‘bad’ with respect to desirable or undesirable characteristics. And here, negative terms are often used, such as sletta ‘stain, blemish’, referring to foreign (typically Danish or English) forms used in speech or writing,

3 Most of what follows is based on my impressions from informal field work in August 2008 and November 2009. See also an overview in Akselberg 2001.

Analysing phonological variation in Faroese 

 127

and the word málsóði ‘linguistically untidy person’ refers to those who do not take good care in their usage. Turning to the Faeroes, for a long time the community was diglossic in the typical sense of Ferguson (1959) with Danish as the H-variety, used in the church and administration. The indigenous language was used at home and as a language of industry and daily life, i. e. in seamanship and agriculture, and there was no written standard. There was oral culture of course, and it is often said that the language survived in the traditional ballads and folk poetry, and that this had a similar significance for the maintenance of the native Faroese culture as the Icelandic sagas in Iceland. But it is clear that the existence of the ballads did not cause linguistic unity in the same way as the written culture probably did in Iceland; the ballads have been preserved in varying forms, according to where they were written down in the 19th century (cf. Barnes 1978). The endogenous written standard of Modern Faroese was developed in the 19th century with a very conservative orthography, modelled after Icelandic and Old Norse. In many respects, the linguistic situation and the historical background of modern Faroese is thus more like that of Norwegian, mainly due to sociolinguistic history. But in other respects the Faroese situation is more like Iceland (and less like Norway and the rest of Europe). Here isolation and distance from European trends should be mentioned; nowadays both Iceland and the Faeroes have a puristic language policy and a single endogenous standard (unlike e. g. Norway, which has two). Still, as far as phonology is concerned, it is not clear that there is any sort of spoken standard (received pronunciation or Bühnensprache) in the Faeroes. For the most part, staff and interviewees alike use their own idiom on the state radio (see e. g. Weyhe 1988). According to Jacobsen (2006), the majority of the Faroese population subscribes to the puristic language policy. However, the long-standing contact with Danish has left its mark on the form of the Faroese language, and Danish loans are very common. And it may be argued that this has had some fundamental effects on the form of the language, such as the replacement of the native pattern of word stress by a more European one (cf. Árnason 1999, 2011a: 275‒84). One might have expected that, given the “puristic” policy, this Danish influence would be the source of some conflict. But the relations with Danish seem to be rather relaxed, and the coexistence of the vernacular and the former exogenous standard, which is often used as a lingua franca in communication with other Scandinavians, seems to be rather peaceful. Another result of the fact that Danish is used quite a bit as a lingua franca seems to be that the Faroese use English less than Icelanders do (see Akselberg 2001; Petersen 2010; Kristiansen and Vikør (eds.) 2006). In spite of the seemingly similar language policies, there are thus interesting differences between Iceland and the Faeroes when it comes to folk linguistics.

128 

 Kristján Árnason

We saw above that the historical background in the Faeroes is in many respects more like that in Norway. And the typical European conception of dialect both as a relational concept (Coseriu 1980; cf. Auer 2005) and an areal variant of language, or variety, seems to be part of Faroese linguistic culture. Thus, a very commonly-used term in Faroese folk linguistics is bygdarmál, which in fact seems to correspond closely to the normal European term ‘dialect’. In the Faroese dictionary (Føroysk orðabók 1998), the word bygdarmál is defined as “avbrigdi í tungumáli talað á ávísun avmarkaðum øki. T. d. bygd el. landsluti, málføri, ‘a variety of language spoken in a designated, limited place, e. g. community (bygd)’, and as can be seen, the word málføri, which is the same word as the Norwegian term målføre ‘dialect’ is listed as a synonym. In this vein, terms such as Klaksvíkarmál, Vágamál, Havnarmál and Suðuroyarmál are often used to refer to the varieties spoken in the different locations. It is likely that the status of these local varieties was originally defined relative to Danish, but nowadays this dialect/standard Abstand may to some extent have been carried over to the relation between the various dialects or bydgarmál and some sort of standard Faroese as taught in schools. This is markedly different from the typical Icelandic folk linguistic usage, as shown by the fact, mentioned above, that the Icelandic term mállýska refers to formal linguistic features rather than actual varieties. Terms corresponding to the Icelandic “good or bad usage” seem to be less commonly used in Faroese folk linguistic discussion, although the words málbering or málburður, may come close to having a similar meaning to Icelandic málnotkun or málfar, which is often used in evaluating good or bad usage. The gloss given in the Faroese dictionary for málbering is: “háttur at málbera seg, málburður, orðalag, orðafar”, ‘way of speaking, wording, lexis’. And an example, given in the dictionary, of the use of the word in context is: góð/ring málbering ‘good or bad usage’. This reminds one of the Icelandic málfar, but it is not clear to me that this is such a significant concept for the average Faroese, although questions of right or wrong usage inevitably crop up when it comes to learning standard written Faroese, cf. the discussion below. Judging by my informal survey, the Faroese people tend to be conscious of the different dialects or bygdarmál and they are willing to discuss their form and status. And on the whole people are happy with their own variety. As an example, a speaker from Fuglafjørður, when asked about her way of speaking, i. e. Fuglafjarðarmál, responded that this was a good variety (mál), and when asked what was special about it, she said that they use a special accent or stress (dentur is the Faroese term). According to her, this stress is typical of Fuglafjørður (fuglafjarðarligt). And the example she gave was that when you hear a speaker from Fuglafjørður on the radio, he or she will say [nɔiː] instead of [naiː] for ‘no’. This tallies with the fact that the rounded pronunciation of the diphthong written

Analysing phonological variation in Faroese 

 129

ei as in nei ‘no’ or steinur ‘stone’ is a characteristic of northern areas, including Fuglafjørður. There seems thus to be a peaceful awareness of dialect differences and the folk linguistic “analysis”, although not necessarily accurate, corresponds with the facts. But on the other hand the conception of a standard variety seems perhaps to be less clear or less comfortable at least, and it may be less clearly defined, as far as phonology is concerned. As mentioned above, there is no spoken standard (e. g. for the radio), although it is conceivable that the language used in the capital, Tórshavn, may be moving toward such a status. The only sort of established standard variety is the written norm, which is taught in school. But there seem to be problems here, and many people say that it is difficult to master this idiom. Mastering the written school language, duga føroyskt ‘to know (proper?) Faroese,’ as it is called, is not easy for everyone, e. g. when it comes to spelling of the often very weak or reduced endings of words like undir ‘under’ and undur ‘wonder’. This problem, which speakers face in coping with the standard and evaluating for themselves the language of Tórshavn, may be illustrated by a conversation with a speaker from Høyvík, a suburb of Tórshavn (cf. Árnason 2011a: 96‒7). When asked what he thought about Havnarmál, the language of Tórshavn, he said that he thinks that this is a good mál (dialect, idiom). But then he adds that maybe sometimes it has a bit too much of -ir instead of -ur. Illustrating this with the minimal pair undir ‘under’ vs. undur ‘wonder’, he said “these are two meanings”, referring to the different written forms. Another example which he refers to is bilur ‘car’. “This word people in Tórshavn say with -ir; people from other parts of the islands (bygdafólk) say more with -ur”. But when asked about his own speech, he said: “I think I say [piːlɹ̥]”, but then correcting himself saying “[piːlɩɹ]”. The first utterance: [piːlɹ̥] shows the commonly occurring reduction of the last syllable in an utterance (cf. Árnason 2011a: 94‒6). The speaker’s performance and comments strongly suggest that he does not distinguish between the endings -ir and -ur, which is probably true of many other speakers. Since there are varieties which distinguish between these endings (e. g. in Vágar), this is a dialect marker, and the speaker’s comments show awareness of the variation. But there is also a potential effect of influence from the written norm on the variation and development. According to Selås (1996), although the difference between the two endings is neutralised most of the time, there is some indication of a correlation with spelling and grammar. And in general it is likely that problems in the teaching of the standard orthography to speakers who do not make the relevant distinctions, helps to make this feature prominent in popular linguistic thinking.

130 

 Kristján Árnason

All in all, the sociolinguistic story of Faroese is still to be written, but as argued below, there are interesting research questions and hypotheses which can be put forward regarding the future development. Among the effects which may be expected to have an influence are demographic trends and the political and economic situation. One of the scenarios which could be imagined is that the speech of the capital, Tórshavn, as a sort of melting pot, will evolve into a new spoken “standard” variety as a result of dialect levelling (cf. the role of the capital in the emergence of spoken standard varieties in other countries such as Sweden, Denmark, France and England, see Auer 2005). But then it will be interesting to see which variant phonological (or other) forms are chosen as features of the resulting koine. The potential development is an interesting object of research as it takes place in real time.

4 Methodological considerations and a pilot study European dialect geography in the 19th and 20th century basically aimed at identifying traditional dialects (German: Grunddialekte) and drawing dialect maps on the basis of the testimony of conservative older generation, typically male, speakers (NORMS) (cf. Streck 2012 for a recent discussion, see also the overview given in Chambers and Trudgill 1998). On the other hand, as noted e. g. by Dahlstedt (1958) and Benediktsson (1961‒62), the methodology applied in the first sys­ tematic Icelandic dialect study in the 1940s by Björn Guðfinnsson (1946, 1964) was totally different. Instead of older speakers, Guðfinnsson investigated the speech of school children around the age of 13, using a reading task and recording the results on cards, marking the phonological choices made by the young speakers. The results were presented statistically, showing the number and percentage of speakers in each school district which used this or that phonological variant. The variables tested were the traditional phonological variant features of the sort described above. A second Icelandic survey carried out 40 years later under the label RÍN based its methodology partly on Guðfinsson’s methods and was also inspired by later “urban dialectology” (cf. Labov 1994, 2001; Trudgill 1974), i. e. applying a statistical analysis of the performance of speakers in a recorded interview. There were basically two parts to the RÍN-interview, apart from a short introduction and orientation basically eliciting personal information from the speaker (later coded). In the first essential part of the interview, speakers were asked to comment on pictures showing various objects with names containing phonologically interesting words, and in the second part they were asked to read short texts, one, two or

Analysing phonological variation in Faroese 

 131

three according to the location or age of the speaker.4 The whole interview was recorded on tape and the speakers’ performance was then analysed auditively by trained researchers. The scale used for each variable, e. g. presence or absence of aspiration in words such as láta ‘to let’, was from 1 (for an unaspirated utterance) to 2 (for an aspirated utterance), and for unclear or intermediate performances the grade 1.5 could be given, and then the results were analysed statistically. Given the different conditions in different countries or cultures, such as Iceland vs. continental Europe, it is perhaps not surprising that different methods have been used in analysing the variation. So, the difference between the Icelandic approach taken by Björn Guðfinnsson in the 1940s and the typically European approach to dialect research may to some extent be justified by the differences in the object of study. But on the other hand we cannot exclude the possibility that the “facts” are to some extent biased by the researchers or their preconceptions. Thus there might have been some European-like dialect characteristics, which were not likely to emerge from Björn Guðfinnsson’s 1940s survey based on the performance of schoolchildren, and the same could be true of the RÍN survey. And, for that matter, a different approach to linguistic variation in, say Germany, England, or Scandinavia might have led to a different picture than the one presented in traditional dialect atlases. It is possible that the research questions or hypotheses involved were to some extent self-fulfilling. (See Auer 2010 for an interesting discussion and demonstration of this sort of effect in German dialectological work.) Since the Faeroes seem to be typologically somewhere between Iceland and Norway (and the rest of Europe) with regard to the dialect/standard constellations, it is perhaps not obvious which approach should be taken in studying the variation and dialect differences, but it will be argued here that a similar methodology to the one used in the Icelandic RÍN project can be very useful. We have seen that a number of regional phonological characteristics have been identified and discussed in the literature on modern Faroese. Among the most prominent of these is the north-south variation in the pronunciation of the vowel written in steinur ‘stone’, as /ɔi/ in the north and /ai/ in the south. Another north-south variable mentioned above is the choice between /ɑ/or /ɔɑ/ for the stem vowel in bátur ‘boat’: [pɑːtʊɹ] vs. [pɔɑːtʊɹ], the northern variant being the monophthong /ɑ/. And the word bátur also illustrates the variation between presence or absence of aspiration in the postvocalic stop, i. e. [pɔɑːʰtʊɹ] / [pɑːʰtʊɹ]

4 The findings of this study have been reported in a number of publications, e. g. Thráinsson and Árnason (1992) and Árnason and Tráinsson (2003).

132 

 Kristján Árnason

or [pɔɑːtʊɹ] / [pɑːtʊɹ]. Variables based on more recent phenomena such as vowel reduction in restricted syllables and raising in hiatus can also be easily defined.5 Given the present day situation with increased mobility and modernization, it might be expected that some changes in the patterns of variation will take place. It is interesting, for example, to ask what happens in the speech of Tórshavn, which already seems to be a melting pot, and it is also interesting to ask which of the traditional features will survive and be accepted as normal in Tórshavn or may even spread to other locations. In general, the present day distribution of the variables can easily be studied with the RÍN methodology. By way of illustration, I will end with a brief description of a pilot study carried out in 2008.6 In this study, I recorded interviews with 25 Faroese informants, 5 from each of 5 locations. The interview consisted of some brief introductory remarks, the reading of a text, and a brief final discussion about the Faroese language and its situation, varieties or dialects. The recordings were made in Klaksvík (on the island of Borðoy), Fuglafjørður (Eysturoy), Tórshavn (Streymoy), Sandur (Sandoy) and Tvøroyri (Suðuroy), and were analysed auditively by trained researchers (see Karlsson et al. 2012). It should be emphasised that this was not a systematic pre-designed experiment, which would be likely to give any sort of statistically significant results, but it would seem to give some very useful hints about what can be done by way of systematic research. The text had several instances of forms involved in variation of the sort mentioned in Section 2 above. Figure 2 shows three of what might be called traditional variables, namely the choice between /ai/ or /ɔi/ in words like stainur ‘stone’, the choice between monophthongal /ɑ/ or diphthongal /ɔɑ/ in words like bátur ‘boat’, and the choice between preaspiration ([ʰt]) or no aspiration ([t]) in the medial consonant in the same word. The results are shown in Figure 2. The performance of the speakers was graded on a scale from 1 to 2, as shown on the y-axis, showing the combined average (on the scale of 100–200) for the five speakers in each location. As can be seen, the numbers correspond well with traditional accounts of the variation. The northern speakers, from Klaksvík and Fuglafjörður, score relatively high for all features. And the same can in fact be said of the speakers from Tórshavn. The southern speakers, from Sandoy and Tvøroyri in Suðuroy have rel-

5 Unfortunately the text which (beforehand) was chosen for the pilot study reported here only contained two or three (doubtful) hiatus forms. Thus the testimony regarding raising in hiatus is too limited to be of any value. 6 I am grateful to the Nordic Language Variation Network, specially Øystein Vangsnes and Höskuldur Þráinsson for inviting me to join the fieldwork Excursion in August 2008.

Analysing phonological variation in Faroese 

 133

200 190 180 170 160

/ei/

150

/á/

140

/hT/

130 120 110 100

Fuglafj.

Klaksvík

Tórshavn

Sandøy

Tvøroyri

Fig. 2: Three northern variables: /ei/ (steinur: [ai] =100 vs. [ɔi]=200), /á/ (bátur: [ɔɑ]=100, [ɑ] =200) and /hT/ (aspiration) (bátur: [t]=100, [ʰt]=200). Average grade according to location ­­ (5 speakers each).

atively low scores for both the rounded pronunciation of /ei/ and aspiration in bátur ‘boat’, whereas the monophthongal pronunciation of /á/ seems to be relatively common in Sandoy. We are obviously not in a position to draw any solid conclusions from this, but the findings suggest that the traditional variables may turn out to fare differently in the dialect levelling which seems to be taking place. The second example regards the reduction of final syllables, which is indeed a much more complicated phenomenon. Here we have a gradient variable and there is an additional dimension in that beside geographic location, the (morphological and/or phonological) type of ending may behave differently. Figure 3 shows the average grade of reduction for the 25 speakers classified according to ending type, the ending –a, as in kasta ‘to throw’, -ar like in bátar ‘boats’ and –ur like in bátur ‘boat’, -il like in intil ‘untill’ and so on. The numbers suggest that there may be significant differences, but again the limits of the test prevent us from proposing any kind of stringent hypotheses or generalisations. And the same must be said about Figure 4, where we have what might be called a meta-variable, such that the average grade for reduction in all endings is calculated for each group of five speakers from the five locations.

134 

 Kristján Árnason

Reduction (total) 150 145 140 135 130 125 120 115 110 105 100

Total

/ar/

/a/

/uN/

/iN/

/ir/

/u/

/i/

/ur/

/il/

Fig. 3: Final syllable reduction. Average grade (all speakers, all areas) for different ending types: /ar/ (bátar ‘boats’) /a/ (bakka ‘Bank-ACC’), /uN/ (fellum ‘traps-DAT-PL’), /iN/ (eingin ‘nobody’), /ir/ (eftir ‘left’), /u/ (fóru ‘went’), /i/ (úthúsi ‘stable-DAT’), /ur/ (kettur ‘cats’) /il/ (intil ‘until’). “Fully pronounced” = 100; “fully reduced” = 200. Average weakening (meta-variable) 135 130 125 120 Average weakening

115 110 105 100

Fuglafj.

Klaksvík

Sandøy

Tórshavn

Tvøroyri

Fig. 4: Final vowel deletion: average grade (100 = fully pronounced; 200 fully reduced) for all types of syllable and all speakers in each location.

5 Conclusion Although conditions are different and the Faroese linguistic situation is more complex than the Icelandic one, there is good reason to believe that phonological variation in Faroese can be analysed with methods similar to the ones that have been used in Iceland. This would mean defining linguistic variables of various types (sometimes computable as meta-variables). The distribution of the variant forms can be plotted against external variables such as location, age, sex or social class (to the extent that this can be defined for the Faeroes). The pilot

Analysing phonological variation in Faroese 

 135

study reported above was based on phonological variables defined on the basis of variant forms which have been observed in the extant phonological descriptions. Although the data is limited, some interesting research questions can already be formulated. Thus Fig. 2 suggests that three characteristics which have been identified as northern characteristics behave differently in modern speech. Although not traditionally described as typical of Tórshavn speech, the rounded variant [ɔi] in steinur scores highly in the capital. Monophthongal /a/ is also fairly common both in Tórshavn and in Sandoy, but as expected less common there than in its home area (Klaksvík and Fuglafjörður), and even less common in the southernmost Suðuroy. The “hard” (aspirated) variant, typical of the north gets a low score in both of the southern islands (Sandoy and Suðuroy), but gets the same score in Tórshavn as in the northern locations. The question is whether this is indicative of some tendencies in koineization. But it should also be kept in mind that even though data of this sort can be used to give an overview of the present situation and to measure potential tendencies toward dialect levelling, it is likely that an analysis in more traditional terms, with the help of a linguistic atlas based on the elicitation of traditional forms from conservative, native speakers, can give interesting results. And the methodology of cluster analysis or Dialektometrie, cf. Streck (2012), where statistical methods can help to determine the tendencies for co-occurrence of linguistic features, may also provide some insights.

References Akselberg, Gunnstein (2001): Færoysk språk i sosiolingvistisk perspektiv. In: Kurt Braunmüller and Jóggvan í Lon Jacobsen (eds.), Moderne lingvistiske teorier og færøsk, 155–179. Oslo: Nordisk Språkråd. Árnason, Kristján (1990): Conflicting teleologies: Drift and normalisation in the history of Icelandic phonology. In: Henning Andersen and Konrad Koerner (eds.), Historical Linguistics: Papers from the International Conference on Historical Linguistics, 21–36. Amsterdam: John Benjamins. Árnason, Kristján (1999): Icelandic and Faroese. In: Harry van der Hulst (ed.), Word Prosodic Systems in the Languages of Europe, 567–603. Berlin and New York: Mouton de Gruyter. Árnason, Kristján (2003a): Icelandic. In: Ana Deumert and Wim Vandenbussche (eds.), Germanic Standardisations: Past to Present, 245–279. Amsterdam: John Benjamins. Árnason, Kristján (2003b): Language planning and the structure of Icelandic. In: Kristján Árnason (ed.), Útnorður, West Nordic Standardisation and Variation, Papers from a Symposium in Stockholm, October 7th 2001, 193–218. Reykjavík: University of Iceland Press. Árnason, Kristján (2011a): The Phonology of Icelandic and Faroese. Oxford: Oxford University Press.

136 

 Kristján Árnason

Árnason, Kristján (2011b): Standards, varieties and phonological variation in Icelandic and Faroese. In: Maj Reinhammar, Lennart Elmevik, and Lars-Erik Edlund (eds.), Studier i dialektologi och språksociologi: Föredrag vid Nionde nordiska dialektologkonferensen i Uppsala 18–20 augusti 2010, 1–20. Uppsala: Acta Academiae Regiae Gustavi Adolphi. Árnason, Kristján (2012): Chaos and structure in North Atlantic vowel systems. In: Unn Røynedal and Hans-Olav Enger (eds.), Fra holtijar til holting: Språkhistoriske og språksosiologiske artikler til Arne Torp på 70-årsdagen, 27–41. Oslo: Novus Forlag. Árnason, Kristján and Höskuldur Thráisson (2003): Fonologiske dialekttræk på Island: Generationer og geografiske områder. In: Gunnstein Akselberg, Anne Marit Bødal and Helge Sandøy (eds.), Nordisk dialektologi, 151–196. Oslo: Novus. Auer, Peter (2005): Europe’s Sociolinguistic Unity, or: A Typology of European Dialect / Standard Constellations. In: Nicole Delbecque, Johan van der Auwera, and Dirk Geeraerts (eds.), Perspectives on Variation, 7–42. Berlin and New York: Mouton de Gruyter. Auer, Peter (2010): Der Grunddialekt als Konstrukt: Wie Gewährspersonen und Erheber in der direkten Befragung die Daten der Atlasdialektologie konsituieren. In: Dominique Huck and Thiresia Choremi (eds.), Parole(s) et langues(s), espaces et temps. Mélanges offerts à Arlette Bothorel-Witz, 23–36. Strasbourg: Université de Strasbourg. Barnes, Michael P. (1978): Grammatical instability in Faroese ballads and written Faroese. In: Knut Kolsrud, Bjarne Hodne, Ronald Grambo, and Anne Swang (eds.), Tradisjon og samfunn. Festskrift til professor Olav Bø på 60-års dagen 19. mai 1978, 209–235. Oslo: Universitetsforlaget. Benediktsson, Hreinn (1961–62): Icelandic Dialectology: Methods and Results. Íslensk tunga – Lingua Islandica 3: 72–113. Chambers, Jack K. and Peter Trudgill (1998): Dialectology. Second edition. Cambridge: Cambridge University Press. Chapman, Kenneth G (1962): Icelandic-Norwegian Linguistic Relationships. Oslo: Universitetsforlaget. Coseriu, Eugenio (1980): ‚Historische Sprache’ und ‘Dialekt’. In: Joachim Göschel, Pavle Ivić, and Kurt Kehr (eds.), Dialekt und Dialektologie, 106–116. Wiesbaden: Steiner. Dahlstedt, Karl-Hampus (1958): Íslenzk mállýzkulandafræði. Skírnir 132: 29–63. Ferguson, Charles (1959): Diglossia. Word 15: 325–340. Føroysk orðabók (1998): Jóhan Hendrik W. Poulsen, Marjun Simonsen, Jógvan í Lon Jacobsen, Anfinnur Johansen, and Zakaris Svabo Hansen (eds.). Tórshavn: Føroya Fróðskaparfelag. Guðfinnsson, Björn (1946): Mállýzkur I. Reykjavík: Ísafoldarprentsmiðja. Guðfinnsson, Björn (1964): Mállýzkur II. Reykjavík: Heimspekideild Háskóla Íslands. Hagström, Björn (1967): Ändelsevokalerna i färöiskan: En fonetisk-fonologisk studie. Stockholm: Almqvist & Wiksell. Helgason, Pétur (2002): Preaspiration in the Nordic Lanugaes; synhchronic and diachronic aspects. Ph.D dissertation, Department of Linguistics, Stockholm University. Jacobsen, Jógvan í Lon (2001): Um øvutan framburð í tjørnuvíksmáli. Fróðskaparrit 48: 15–20. Jacobsen, Jógvan í Lon (2006): Færøerne. In: Tore Kristensen and Lars Vikør (eds.), Nordiska språkhaldningar. Ei meiningsmåling, 40–60. Oslo: Novus Forlag. Karlsson, Bjarki M., Halldóra Kristinsdóttir, Jón Símon Markússon, Kristín Lena Þorvaldsdóttir, Kristján Árnason, and Sigrún Gunnarsdóttir (2012): Um hljóðkerfislegan breytileika í færeysku. In: Magnús Snædal and Turið Sigurðardóttir (eds.), Frændafundur 7. Fyrirlestrar á íslensk-færeyskri ráðstefnu í Reykjavík 21.–22. ágúst 2010, 73–90. Reykjavík: Hug­­vísindastofnun.

Analysing phonological variation in Faroese 

 137

Kristiansen, Tore and Lars Vikør (eds.) (2006): Nordiske språkhaldningar. Ei meiningsmåling. Oslo: Novus Forlag. Labov, William (1994): Principles of Linguistic Change. Volume 1: Internal Factors. Oxford: Blackwell. Labov, William (2001): Principles of Linguistic Change. Volume 2: Social Factors. Oxford: Blackwell. Leonard, Stephen Pax (2009): Language, Society and Identity in early Iceland. Ph.D. thesis, Oxford University. Lockwood, William Burley (1955): An Introduction to Modern Faroese. Tórshavn: Føroya skúlabókagrunnur. (3. printing 1977.) Myrvoll, Klaus Johan and Trygve Skomedal (2010): Tonelagsskilnad i islendsk. Maal og minne 1: 68–97. Petersen, Hjalmar (1996): Vágamálføri. Fróðskaparrit 44: 5–21. Petersen, Hjalmar (2010a): Føroyskt-danskt málamót. Varðin 76: 3–39. Petersen, Hjalmar (2010b): The Dynamics of Faroese-Danish Language Contact. Heidelberg: Winter. Sandøy, Helge (2003): Språkendringar med eller utan kontakt i Vest-Norden. In: Kristján Árnason (ed.), Útnorður, West Nordic Standardisation and Variation, Papers from a Symposium in Stockholm, October 7th 2001, 81–110. Reykjavík: University of Iceland Press. Schäfer, Michael and Kristján Árnason (2012): A stress test for segmenthood: On the segmental status of preaspiration in Faroese and Icelandic. Paper presented at the International Conference of Nordic and General Linguistics, University of Freiburg, 18. April 2012. Selås, Magnhild (1996): Trykklette endingar i talemålet i Tórshavn. Ei sosiolingvistisk gransking. MA-thesis, University of Bergen. Streck, Tobias (2012): Phonologischer Wandel im Konsonantismus der allemannischen Dialekte Baden-Württembergs: Sprachvergleich, Spontansprache und dialektometrische Studien. Stuttgart: Franz Steiner. Thráinsson, Höskuldur, Hjalmar Petersen, Jógvan í Lon Jacobsen, and Zakaris Svabo Hansen (2004): Faroese, an Overview and Reference Grammar. Tórshavn: Føroya Fróðskaparfelag. Thráinsson, Höskuldur and Kristján Árnason (1992): Phonological variation in 20th century Icelandic. Íslenskt mál 14: 89–128. Trudgill, Peter (1974): The Social Differentiation of English in Norwich. Cambridge: Cambridge University Press. Weyhe, Eivind (1988): Færøske dialekter og deres stilling i dag. In: Andreas Bjørkum and Arve Borg (eds.), Nordiske studiar. Innlegg på den tredje dialektologkonferansen 1986, 139–150. Oslo: Universitetsforlaget. Weyhe, Eivind (1997): Ljóðfrøðileg viðmerking til eitt jólabræv. Málting 21: 26–27. Weyhe, Eivind (2003): Í miðjum grasgarði. Rannsóknir í kvæðauppskriftum úr Suðuroy. Tórshavn: Føroya Fróðskaparfelag.

Section 3: Morphosyntax

Jeremy Bradley

Mari converb constructions – Interpretation and translation1 Abstract: This paper explores the interpretation and translation of aspectual and directional converb constructions – areal phenomena in the Volga basin – in the Finno-Ugric Mari language. In these constructions, a finite number of verbs can serve as modifiers, losing their original semantic content and merely transferring an aspect or a directionality to a second verb. After introducing basic structures, the paper will attempt to provide a holistic overview of these mechanisms, illustrating which verbs can be used in what ways, and how they might best be interpreted and translated. After introducing Mari and its neighbours, the paper explores what the term “converb construction” actually means in different linguistic schools. The structure under consideration is defined and distinguished from other methods of verb serialization available in Mari. Then, the interpretation and translation of the structure at hand is examined. Finally, the findings are summarized and described; factors needing further research are underlined.

1 The Mari language Mari – Cheremis in older sources – is a Finno-Ugric (Uralic) language spoken primarily in the Volga and Ural regions of the Russian Federation. Together with Russian, it enjoys official status in the Mari El Republic, a subject of the Russian Federation roughly the size of Slovenia located some 500–800 kilometres east of Moscow, near the confluence of the Volga and Kama rivers. It is a pluricentric language with two distinct literary norms, Meadow Mari and Hill Mari. Meadow Mari is by far the dominant variant. In the 2010 All-Russia population census, 365,127 people claimed to be Mari speakers, 23,062 of them identified themselves as speakers of Hill Mari (Federal’naâ služba gosudarstvennoj statistiki 2011). This paper will focus on Meadow Mari; Hill Mari examples will be explicitly marked as such.

1 A number of consultations referred to in this paper took place as part of the project “Mari-English Dictionary” (dict.mari-language.com), funded by the Austrian Science Fund FWF (grant P22786–G20).

142 

 Jeremy Bradley

Mari uses a variant of the Cyrillic alphabet slightly different from the Russian alphabet. Mari examples will be transcribed into IPA in this paper; literary references will be transliterated in accordance with the ISO 9:1995 standard. The same procedure will be followed for other languages of the Russian Federation that use the Cyrillic alphabet. While its dominant contact language today is clearly Russian, the Mari language can historically be placed in the Volga-Kama Sprachbund (Helimski 2003: 159; Wintschalek 1993), which includes both Turkic and Finno-Ugric languages. In addition to Mari, the core members of this areal grouping are the Finno-Permic Udmurt language, as well as three Turkic languages: Tatar, Chuvash and Bashkir. Two additional Finno-Ugric languages, Mordvin and Komi, are considered peripheral members. The rough geographic distribution of these languages can be seen in Figure 1. This map shows the political subdivisions of the Russian Federation associated with the nations in question, which only match the actual overlapping settlement areas of the nations to some extent.

Fig. 1: The Volga-Kama Sprachbund in Russia2 The political subdivisions shown here are: Turkic core members (diagonal lines): 1) Tatarstan (Tatar) 2) Chuvashia (Chuvash) 3) Bashkortostan (Bashkir) Finno-Ugric core members (dark dots): 4) Mari El (Mari) 5) Udmurtia (Udmurt) Finno-Ugric peripheral members (light dots): 6) Mordovia (Mordvin) 7) Komi (Komi)

2 Basis: http://commons.wikimedia.org/wiki/File:Russia_-_blank_map_%282008–01 %29.svg

Mari converb constructions – Interpretation and translation 

 143

2 Converb constructions The term “converb construction” is used to label a broad range of phenomena in many different languages. According to one possible definition, a converb is “a nonfinite verb form whose main function is to mark adverbial subordination” (Haspelmath 1995: 3); structures containing such an element can be considered converb constructions. As a SOV-language (Bereczki 1990: 74) with a propensity for embedded nonfinite clauses and a wide range of nonfinite verbal forms – two infinitives, four participles, five gerunds and two productive deverbal nominal derivations, according to one source (Alhoniemi 1985: 131–149, 156–157) – Mari offers many possibilities for verb serialization that are covered by Haspelmath’s definition. Sentences featuring a large number of verbs in a row can easily be found in Mari, as is illustrated in Example 1. Example 1 (1 – Mari – Čhaidze 1960: 13) ‘Erpatɤr tide pogɤnɤmaʃkat mijen tolaʃ ʃonen pɤʃten.’ Erpatɤr tide pogɤn -ɤmaʃ -k -at Erpatyr this meet -NMLZ -ILL -AND n pr v -v.deriv.n -n.case -enc mij -en tol -aʃ ʃon -en pɤʃt -en -Ø. go -GER come -INF think -GER place -PST2 -3SG v -v.adv v -v.inf v -v.adv v -v.tense -v.pers “Erpatyr decided to go to this meeting.”

The focus of this paper will, however, be limited to the phenomenon traditionally described as “converb constructions” in the literature on the Mari language and its neighbours. This phenomenon is not consistent with Haspelmath’s definition, as it does not contain any subordination. Rather, what is at issue are complex predicate constructions consisting of two verbs that are used within a single clause. Structures of this type can be found throughout the Volga-Kama Sprachbund (Wintschalek 1993: 124–129). They also existed in the extinct Uralic (Samoyedic) Kamassian language formerly spoken in Northern Eurasia. While this language was not part of the Volga-Kama Sprachbund, it was like Udmurt and Mari subject to significant Turkic influence (Klumpp 2002). Example 2, taken from the Mari national poet Sergej Čavajn’s novel Èlnet, features two verbal pairings (highlighted in grey) that, if translated literally into English without comment, would lead to confusion.

144 

 Jeremy Bradley

Example 2 (2 – Mari – Čavajn 1967: 211) ‘Lut͡ɕo vaʃkerak iktɤm temen pu, jyɤn koltem, ijezuit-βlak manmɤla, ad majorem dei gloriam.’ lut͡ɕo vaʃke -rak iktɤ -m tem -en pu -Ø, better quick -COMP one -ACC fill -GER give -IMP.2SG av av -av.deg nm -nm.case v -v.adv v -v.mood-pers jy -ɤn kolt -em, drink -GER send -1SG v -v.adv v -v.pers -la, ad majorem dei gloriam. ijezuit -βlak man -mɤ say -PT.PASS -COMP *** Jesuit -PL n -n.num v -v.ad -ad.case *** “Better fill one up for me quickly – I’ll drink it down, as the Jesuits say, ad maiorem dei gloriam.”

In the first marked pairing, one communication partner asks the other to “filling give” (a glass of vodka, presumably), using the Mari verb puaʃII3 “to give” to indicate that the activity in question has a beneficiary – no personal pronoun (“(for) me”) is used in this function, as one might expect in English. In the second pairing, the speaker voices an intention to “drinking send” the assumed shot of vodka. This pairing can only be translated in a meaningful manner if the core semantic content of koltaʃII “to send” is ignored, and a different function is attributed to the word – one of an aspectual modifier in an aspectual converb construction. The abundant usage of the gerund in -n can be explained by the fact that Mari has adopted the usage of converb constructions, which are typical in Turkic languages, and uses the gerund in -n for these. The -n form has a subordinating or coordinating relationship with a second verb, the “main verb”, as is the case in Chuvash converbs, for example. (Bartens 1979: 143)4 [The gerund in -n] is used to create a so-called aspectual converb construction. It contains a verb that, as the main verb, gives the activity an aspectual colouring; the gerund contains the semantic content of the construction. Many verbs are used as aspect givers. They lose

3 Mari verbs belong to one or two conjugation classes. The correct conjugation class cannot be derived from the lexical form, the infinitive. The correct conjugation class will be given in a superscript Roman numeral. 4 “ə̂n-gerundin runsas käyttö selittyy siitä, että tšeremissi on omaksunut turkkilaiskielille luonteenomaisen konverbien käytön ja käyttää tähän ə̂n-tyyppiä. ə̂n-tyypillä on toiseen verbiin, “pääverbiin”, nähden subordinaatio tai koordinaatiosuhde kuten siis esim. tšuvassinkin konverbeillä [...]”

Mari converb constructions – Interpretation and translation 

 145

their lexical meaning either entirely or at least partially. Some studies cite roughly 40 such verbs […] (Alhoniemi 1985: 143–144)5

Similar constructions, in which a gerund is paired with a finite verb that loses its original semantics and serves as an aspectual modifier, can indeed be found throughout the Volga-Kama Sprachbund: Table 1: Aspectual Converb Constructions in the Volga-Kama Sprachbund (Csúcs 1990: 61; Moisio 1992: 407; Poppe 1963: 76; Benzing 1943: 84) Language

Construction

Literal Translation

Meaning

Udmurt Mari Tartar Chuvash

korasa bɨdtɨnɨ kot͡ɕkɤn ʃɤndaʃII jaza baru tipse kajas

(to) cutting end (to) eating place (to) writing go (to) drying depart

to cut down to eat up to continue writing to dry out

The identification of aspectual converb constructions in Mari is anything but trivial, as the aforementioned gerund in -n – the most widely used gerund in Mari – can precede finite verbs in a wide range of functions. The same syntactic construction can have many different meanings. –– The government of a number of Mari verbs requires their constituents to be in gerundial form. Example 3 (3 – Mari – Alhoniemi 1985: 142) ‚Tudo muren moʃta.‘ Tudo mur -en moʃt -a. (s)he sing -GER be.able.to -3SG pr v -v.adv v -v.pers “(S)he can sing.”

–– The gerund in -n can be used to indicate a simultaneous action (relative to the finite verb).

5 „Se esiintyy muodostamassa ns. aspektuaalista konverbirakennetta. Siinä syntaktisena pääverbinä on tekemiselle aspektuaalisen sävyn antava verbi; gerundimuoto ilmoittaa rakenteen semanttisen merkityksen. Aspektuaalisesti käytetään varsin useita verbejä. Aspektuaalisessa konverbirakenteessa ne menettävät joko kokonaan tai ainakin osaksi leksikaalisen merkityksensä. Eräissä tutkimuksissa on mainittu jopa n. 40 tällaista verbiä […]”

146 

 Jeremy Bradley

Example 4 (4 – Mari – Sentence provided by Mari native speaker Èmma Âkimova) ‘Kot͡ɕkɤn ʃint͡ɕena.’ Kot͡ɕk -ɤn ʃint͡ɕ -ena. eat -GER sit -3PL v -v.adv v -v.pers “We are sitting and eating.”

–– In certain phrases, two verbs can be fused to create a joined semantic value; the verbal pairing can denote different relationships between the two verbs – in the example below, one of cause and effect. Example 5 (5 – Mari – Galkin et al. 1990–2005, III: 265) ‚[T]ɤgaj not͡ɕko βurgem dene kɤlmen kolet.‘ [T]ɤgaj not͡ɕko βurgem dene kɤlm -en kol -et. such wet clothing with freeze -GER die -2SG ad ad n pp v -v.adv v -v.pers “You will freeze to death in wet clothing like that.”

–– The gerund in -n can mark previous actions (relative to the finite verb). Example 6 (6 – Hill Mari – Alhoniemi 1985: 143) Tədə pæʃæm əʃten toleʃ. Tədə pæʃæ -m əʃt -en tol -eʃ. (s)he work -ACC do -GER come -3SG pr n n.case v -v.adv v -v.pers ‘(S)he will come when (s)he has done the work.’

–– The gerund in -n can qualify the finite verb it precedes. Example 7 (7 – Mari – Sentence provided by Mari native speaker Èmma Âkimova) Poŋgo kogɤʎetɤm jøraten kot͡ɕkam. Poŋgo kogɤʎ -et -ɤm jørat -en kotɕ͡ k -am. mushroom pie -2SG -ACC love -GER eat -1SG ad n -n.poss -n.case v -v.adv v -v.pers “I love to eat (lit. loving eat) your mushroom pies.”

Mari converb constructions – Interpretation and translation 

 147

–– The gerund in -n can occur in embedded nonfinite clauses with a subject that does not match the subject of the main clause. Example 8 (8 – Mari – Alhoniemi 1985: 143) [Nuno] olɤmbal temɤn pogɤnenɤt. -ɤt. [Nuno] olɤmbal tem -ɤn pogɤn -en [they] bench fill.up -GER assemble -PST2 -3PL n v -v.adv v -v.tense -v.pers [pr] “[They] assembled, filling a bench.” (lit. “[They] assembled, the bench filling itself up.”)

–– If the gerund in -n is coupled with a finite verb of motion, the gerund can carry the main semantic content of the pairing, while the second element of the pairing only marks directionality. Example 9 (9 – Mari – Galkin et al. 1990–2005,VIII: 399) Tunamak pasu ymbat͡ɕ βit͡ɕ-kud kede t͡ɕoŋeʃten kyzɤʃ. Tunam -ak pasu ymbat͡ɕ then -STR field from.over av -enc n pp t͡ɕoŋeʃt -en kyz -ɤʃ -Ø. βit͡ɕ-kud kede five.or.six wood.pigeon fly -GER climb -PST1 -3SG nm n v -v.adv v -v.tense -v.pers “Immediately five or six wood pigeons flew up from the field.”

–– Finally, the gerund in -n can occur in proper aspectual converb constructions: the final verb only adds an aspectual colouring to the pairing; the gerund carries the main semantic content. Example 10 (10 – Mari – Alhoniemi 1985: 144) Lydɤn kajɤʃɤm. Lyd -ɤn kaj -ɤʃ -ɤm. be.afraid -GER go -PST1 -1SG v -v.adv v -v. tense -v.pers “I was startled.”

One of the two morphologically marked past tenses in Mari can be traced back to this syntactic construction: the so-called simple past tense II shown in Table

148 

 Jeremy Bradley

2 below is a fusion of the gerund in -n and the respective form of ulaʃI “to be”. In many dialects, the original periphrastic form is still used in this function (Alhoniemi 1985: 111–112). Table 2: Caption: Periphrastic and Morphologized Forms of the Simple Past Tense II Person

Gerund in -n + Copula

Simple Past Tense II

1Sg 2Sg 3Sg 1Pl 2Pl 3Pl

tolɤn ulam tolɤn ulat tolɤn6 tolɤn ulɤna tolɤn ulɤda tolɤn ulɤt

tolɤnam tolɤnat tolɤn tolɤnna tolɤnda tolɤnɤt

Furthermore, three verbs found in standard Mari owe their existence to the fusion of converb constructions: namijaʃII “to bring” < nalɤn mijaʃII “to taking come”, naŋgajaʃII “to take” < nalɤn kajaʃII “to taking go” (Alhoniemi 1986: 102), and presumably also pureŋgajaʃII “to fall” < puren kajaʃII “to entering go”. It should be noted that namijaʃII and naŋgajaʃII are, in turn, used as modifiers in converb constructions. The pairings nalɤn mijaʃII and puren kajaʃII can be found as parallel forms to the fused variants in contemporary Mari. In spite of the morphologization and lexicalization observed, there are no delexicalized modifiers in Mari. All verbs used as modifiers continue to be used in their original meanings: There are neither gaps in their paradigms, nor can grammatical differences be observed between the usage of a verb in its original semantic function and its usage as a modifier. In contrast to Chuvash, for example – where different nonfinite verb forms are used depending on the semantic nature of a verbal pairing (Benzing 1943: 83) – the correct interpretation of the structure “gerund in -n + second verb” can be based on semantics only. At times, this is comparatively simple. If the second verb’s lexical meaning is not feasible in the sentence in question (e. g. “to drinking send”) and the second verb is known to be a valid aspectual modifier, an aspectual converb construction is the most plausible interpretation. In many cases, however, a known modifier can occur in such a verbal pairing where both its original meaning and its aspectual meaning would be plausible.

6 Mari uses a copula in all persons but the third person singular. Here, a copula is possible, but neither necessary nor usual.

Mari converb constructions – Interpretation and translation 

 149

Example 11 (11 – Mari – Sentence provided by Mari native speaker Èlina Guseva) Avtobusɤm ostanovkɤʃto βut͡ɕen ʃint͡ɕem. Avtobus -ɤm ostanovkɤ -ʃto βut͡ɕ -en ʃintɕ͡ -em. bus -ACC bus.stop -INN wait -GER sit -1SG n -n.case n -n.case v -v.adv v -v.pers ‘I am waiting for the bus at the stop.’

The verb ʃint͡ɕaʃII “to sit” is a known aspectual modifier, used to indicate a long-lasting imperfective activity (Galkin et al. 1990–2005, IX: 139). In Example 11 above, our native Mari informant asserted that the verb here marks this verbal aspect. The verb’s original meaning, however, seems plausible in the given context as well, because sitting is an activity easily combined with waiting at a bus stop. Example 12 below illustrates this same verbal pairing used in a context where arguably the original meaning of the finite verb is still preserved. Example 12 (12 – Mari – Galkin et al. 1990–2005, IX: 444) Prijomɤm βut͡ɕen ʃint͡ɕɤʃe t͡ɕerle-βlak ʃke koklaʃtɤʃt ʃɤβɤge mutlanat. Prijom -ɤm βut͡ɕ -en ʃint͡ɕ -ɤʃe tɕ͡ erle -βlak reception -ACC wait -GER sit -PT.ACT patient -PL n -n.case v -v.adv v -v.ad n -n.num ʃke kokla self middle pr n

-ʃt -ɤʃt ʃɤβɤge mutlan -at. -INN -3PL quietly talk -3SG -n.case -n.poss adv v -v.pers

“The patients sitting and waiting to be received are quietly talking with each other.”

It seems likely that aspectual and directional modifiers make up closed sets, analogous to the set of adverbial particles used in English phrasal verbs. This is not, however, a foregone conclusion: in the Northern Australian Jaminjung language, for example, preverbs – which carry out functions quite similar to modifiers in Mari converb constructions – have been found to be an open class (Schultze-­ Berndt 2003). No existing materials on Mari grammar provide an exhaustive overview of possible modifiers. It is unclear if this is an argument for the openness of this class, or whether it simply is an indication that grammatical descriptions of Mari created to-date are not satisfactory. As mentioned above, Alho Alhoniemi’s Mari grammar claims that some studies cite roughly 40 different aspectual modifiers. Comparable numbers can indeed be found in a range of lexical materials, grammars and scientific publi-

150 

 Jeremy Bradley

cations (e. g. Beke 1911: 21 modifiers; Pengitov et al. 1961: 32 modifiers; Čhaidze 1960: 36 modifiers; Moisio 1992: 27 modifiers). However, it is just the total number of modifiers cited that roughly lines up, not the actual modifiers. In a comprehensive survey of a number of reputable linguistic publications on the Mari language (Alhoniemi 1985; Beke 1911; Pengitov et al. 1961; Čhaidze 1960; Moisio 1992; Galkin et al. 1990–2005; Vasil’ev et al. 2003) we found a startling 136 different verbs supposedly used as modifiers in pairings explicitly marked as converb constructions7, more than three times the estimate given by Alhoniemi. It follows that either Alhoniemi’s estimate is seriously inaccurate or that a large number of verbal pairings were categorized as converb constructions in our sources even though they did not correspond to accepted definitions of converb constructions. Further analysis showed both of these possibilities to be true to some extent. Directionality markers are widely labelled as aspectual markers; even the examples of aspectual modifiers given in Alhoniemi’s grammar include one – the verb puraʃII “to come in”, to which no aspectual function can be ascribed8. Furthermore, a large number of verbal pairings in which the gerund qualifies the finite verb were labelled as converb constructions in our sources (see Example 13). Example 13 (13 – Mari – Vasil’ev et al. 2003: 27) mɤskɤlen peleʃtaʃ mɤskɤl -en peleʃt -aʃ mock -GER say -INV v -v.adv v -v.inf “to say mockingly”

Pairings of this type were rapidly discarded, which radically reduced the number of potential modifiers under consideration. A total of 64 verbs remained that seemed plausible either as aspectual or directional modifiers. A full list is given in the appendix.

7 It is, however, not always entirely clear what a given author means by “converb construction”. 8 Alhoniemi lists 39 different possible modifiers, but only gives usage notes for four of these, not including puraʃII. Thus, he claims that this verb can serve as a modifier, but gives no information on its function. Other sources only cite it as a directionality marker.

Mari converb constructions – Interpretation and translation 

 151

3 Interpretation and translation of converb constructions A directional construction can be said to express the following categories: motion, manner, path, source/goal (Slobin 2000: 109). Figure 2 illustrates how these categories are generally marked in English, Figure 3 how they are marked in Mari. MOTION, MANNER PATH ↓ ↓ VERBfinite SATELLITE ↓ ↓ […] run out

SOURCE/GOAL ↓ N+(adposition, case) ↓ of the house

Fig. 2: English directional construction (Slobin 2000: 109) SOURCE/GOAL ↓ N+(adposition, case) ↓ pørt gɤt͡ɕ9 “out of the house”

MOTION, MANNER ↓ VERBgerund ↓ kurʒɤn ‘running’

PATH ↓ VERBfinite ↓ lektaʃI “to go out”

Fig. 3: Mari directional construction

The same categories are expressed in both languages; it is simply their realization that differs. This must be taken into consideration in the translation of directional converb constructions: rather than translating the directionality-marking verb with a verb in the target language, the mechanisms the latter uses to mark a path should be employed. In the case of English, this is generally an adverbial particle. Indeed, it is often possible to compare one particular directionality marker in Mari to a particular adverbial particle in English, as is illustrated in Table 3. Table 3: puraʃII “to Come in” as a Directionality Marker vs. English Phrasal Verbs with “in” Construction puraʃII

tolɤn tørʃten puraʃII t͡ɕoŋeʃten puraʃII jogen puraʃII

Literal Translation

Meaning

(to) coming come in (to) jumping come in (to) flying come in (to) flowing come in

to come in to jump in to fly in to flow in

9 All examples henceforth were, if not marked otherwise, provided by our native Mari informant Èlina Guseva of Mari State University, Yoshkar-Ola, Russian Federation, as part of the project “Mari-English Dictionary” (dict.mari-language.com).

152 

 Jeremy Bradley

The inventory of directionality markers in Mari includes both intransitive verbs coupled with intransitive verbs of self-motion and transitive verbs coupled with transitive verbs of motion. For example, the verb puraʃII “to come in” is coupled with intransitive verbs (see Table 3), while its transitive counterpart purtaʃII “to bring in”, derived from it through the causative derivational suffix -t (Alhoniemi 1985: 164),10 is coupled with transitive verbs. Table 4: purtaʃII “to Bring in” as a Directionality Marker vs. English Phrasal Verbs with “in” Construction

Literal Translation

Meaning

βyden purtaʃII konden purtaʃII numal purtaʃII ʃyken purtaʃII

(to) leading bring in (to) bringing bring in (to) carrying bring in (to) pushing bring in

to lead in to bring in to carry in to push in

Verbs serving as directionality markers are verbs of motion containing a directionality in their base meaning, e. g. βolaʃII “to descend” (down – intransitive), βoltaʃII “to lower” (down – transitive), lektaʃI “to go out” (out – intransitive), luktaʃI “to lead out” (out – transitive), etc. As there is only a limited number of verbs of motion in Mari, the set of directionality markers can be considered to be a closed one. Translation of aspectual converb constructions is less straightforward. The set of aspectual modifiers, like the set of directional modifiers, seems to be a closed one. However, there is greater variety regarding the semantics of aspectual modifiers in their base meaning: kijaʃII “to lie” (imperfective), lektaʃI “to go” (perfective), koltaʃII “to send” (inchoative), ont͡ɕaʃII “to look” (momentary). The aspectual meaning these verbs transmit as modifiers can be read into the original meanings of these verbs (“sending”, for example, is a quintessentially inchoative activity), but the exact functions of individual modifiers are less predictable than those of directional markers. Likewise, while a range of mechanisms exist in the English language allowing speakers to express various verbal aspects (“I read that book” vs. “I was reading that book when my father entered the room”, “He cried.” vs. “He broke out in tears.”, “You’re stupid.” vs. “You’re being stupid.”, etc.), no systematic correspondences suggest themselves here. A hypothetical translator would have to find a stylistically suitable English counterpart on a case-by-case basis, and in many

10 -aʃ is the ending of the infinitive, and not part of the stem. Thus the derivational suffix, which is connected to the stem directly, precedes it.

Mari converb constructions – Interpretation and translation 

 153

instances the aspectual markers would simply have to be ignored. In many cases, adverbial complements will already mark the aspect sufficiently, as in the English sentence “I worked in Paris for fifteen years”, where there is no need to mark the durative aspect explicitly, because “for fifteen years” already sufficiently communicates this. Table 5: Verbal Pairings Best not Translated Literally Construction

Literal Translation

Meaning

kredal kijaʃII muralten koltaʃII ʃortɤn nalaʃI t͡ɕien ont͡ɕaʃII koʃken pɤtaʃII malen sitaraʃII peled ʃogalaʃI

(to) fighting lie (to) singing send (to) crying take (to) putting on look (to) drying out end (to) sleeping gather (to) blossoming stand up

to fight (durative) to start singing to shed a few tears to try on (a piece of clothing) to dry out completely to have a good sleep to break into blossom

A complete overview of attested modifiers, both aspectual and directional, is given in the appendix.

4 Conclusions and prospects Comparative neglect of this omnipresent aspect of the Mari language is undeniable, regrettable and in line with the systematic disregard of syntax in publications on Uralic languages. While the study of syntax has steered most theoretical linguistics over the last few decades, relatively little progress has been made in our understanding of the Uralic sentence. Data-oriented surveys are rare: […] The lag may be due, in part, to the enormity of the task of understanding the masses of morphophonology typical of most Uralic languages, some of which has been clarified only recently; but it is also doubtless due to the Anglocentricity, then Eurocentricity, of much of the early work in modern syntactic theory. (Abondolo 1998: 33)

Thanks to its 1993 German translation (Alhoniemi 1993), Alho Alhoniemi’s 1985 Finnish-language grammar remains the most extensive and modern resource on Mari grammar that is at least somewhat accessible to the international linguistic community. It explains the phenomenon in one paragraph (quoted in Section 2), provides usage notes on four modifiers, and a total of five example sentences. At the same time, this grammar dedicates 16 pages to derivational suffixes. The meaning of each suffix is given, as are a number of examples as well as informa-

154 

 Jeremy Bradley

tion on the suffix’s productivity and on its respective usage in the language’s two literary standards. 10) M11 H12 -lt (1st conj.) [V → V], reflexive-translative; frequentative: M muʃkɤltaʃ, H mɤʃkɤltaʃ ‘to wash oneself’ (: M muʃkaʃ, H mɤʃkaʃ ‘to wash’), M pot͡ɕɤltaʃ, H pat͡ʃɤltaʃ ‘to open (intr.)’ (: M pot͡ʃaʃ, H pat͡ʃaʃ ‘to open something’), M ont͡ɕɤltaʃ, H anʒɤltaʃ ‘to examine’ (: M ont͡ɕaʃ, H anʒaʃ ‘to look’). Weakly productive. (Alhoniemi 1985: 161)13

Given the ubiquity of converb constructions in Mari, and considering the possible functional similarity between derivational suffixes and aspectual modifiers in converb constructions14, this imbalance seems somewhat arbitrary. Based on a meta-analysis of existing materials on converb constructions in Mari, a first draft of an overview of modifiers in converb constructions comparable to Alhoniemi’s overview of derivational suffixes has been prepared (see appendix), but it is by no means satisfactory. No concrete data is available regarding dialectal variance15, and very little can be said about the productivity of modifiers – that is to say, how wide a range of verbs can be modified with a particular marker. Due to the limited data available, assumptions made regarding certain modifiers seem somewhat anecdotal at this point. In order to obtain a more reliable overview of when which modifiers can be used for what purpose, an extensive corpus analysis and interviews with native speakers of all four (Alhoniemi 1993: 254) Mari dialect groups are needed. Moreover, overlap and interplay between this mechanism and other mechanisms should be analysed. For example, the question of what relations exist between aspectual marking and tense marking still awaits a proper investigation. One aspect in particular that has been neglected so far is the relationship between modifiers in converb constructions and valence. Example 2 showed that

11 Meadow Mari – see Section 1 12 Hill Mari – see Section 1 13 “10) I L -lt (I konj.) [V → V], refleksiivis-translatiivinen; frekventatiivinen: I muškə̑ltaš, L  mə̑škə̑ltaš ‘peseytyä’ (: I muškaš, L mə̑škaš ‘pestä’), I poč́ə̑ltaš, L pačə̑ltaš ‘aueta’ (: I poč́aš, L pačaš ‘avata’), I onč́ə̑ltaš, L anžə̑ltaš ‘tarkastella’ (: I onč́aš, L anžaš ‘katsoa’). Heikosti produktiivinen.” 14 The Mari-Russian Dictionary published in 10 volumes between 1990 and 2005 explicitly equates the perfective converb construction pomɤʒalten koltaʃ (< pomɤʒaltaʃ “to wake somebody up”) with the verbal derivation poməžaltaraʃ, formed from the same stem (Galkin et al. 1990–2005, V: 176). 15 Anecdotes provided to us by our informants suggest that dialectal differences do exist, but we have yet to be able to quantify them.

Mari converb constructions – Interpretation and translation 

 155

the verb puaʃII “to give” can be used to indicate the existence of a beneficiary, even if the sentence in question does not contain an overtly marked indirect object: “filling give one” was translated as “fill one up for me”. An overtly marked secondary object is, however, not precluded by puaʃII (illustrated in Example 14 below), nor does this particular modifier seem to be necessary for the ditransitive usage of generally transitive verbs (see Example 15). Example 14 and 15 (14 – Mari –Newspaper “Marij kalɤk gazet Marij El”, 22 June 2005, Article “Tyl-βüd voʃt erten”) ‘[…] tunɤktɤʃo-βlaklan pørtɤm t͡ɕoŋa.’ […] tunɤktɤʃo -βlak -lan pørt -ɤm tɕ͡ oŋ -a. […] teacher -PL -DAT house -ACC build -3SG […] n -n -n.case n -n.case v -v.pers “[He] is building a house for the teachers.” (15 – Mari – Galkin et al. 1990–2005, VIII: 265) ‘[…] Sem’onɤt͡ɕlan joŋgɤdo pørtɤm t͡ɕsoŋen puenɤt. ‘ ͡ soŋ -en pu -en […] Sem’onɤt͡ɕ -lan joŋgɤdo pørt -ɤm tɕ -ɤt […[ Sem’onɤt͡ɕ -DAT spacious house -ACC build -GER give -PST2 -3PL […] n -n.case ad n -n.case v v.adv v -v.tense -v.pers “A spacious house was built for […] Sem’onɤt͡ɕ […].”

Redundancy of the type seen in Example 15 can also be found in conjunction with directionality markers. Figure 3 showed how the source/goal is expressed with an adposition or a case, while the second verb in a directional converb construction marks the path. The line between source/goal is by no means crisp, however. If the adposition or case already yields information on the path, the usage of directional converb constructions seems to be optional, as illustrated by Example 16 (using a directional modifier) and Example 17 (using no directional modifier): Example 16 (16 – Mari – Sentence provided by Mari native speaker Èmma Âkimova) ‘Kajɤk-βlak ʃokʃo velɤʃ t͡ɕoŋeʃten kajat.’ Kajɤk -βlak ʃokʃo vel -ɤʃ tɕ͡ oŋeʃt -en kaj -at. bird -PL warm region -ILL fly -GER go -3PL n -n.num ad n -n.case v -v.adv v -v.pers “Birds fly off to warmer lands.”

156 

 Jeremy Bradley

Example 17 (17 – Mari – Galkin et al. 1990–2005, IV: 288) ‘Kajɤk-βlak ʃokʃo velɤʃ t͡ɕoŋeʃtenɤt.’ Kajɤk -βlak ʃokʃo vel -ɤʃ tɕ͡ oŋeʃt -en -ɤt. bird -PL warm region -ILL fly -PST2 -3PL n -n.num ad n -n.case v -v.tense -v.pers “Birds flew off to warmer lands.”

It is not clear if and how sentences double-marking certain information (Examples 15, 16) differ from more economic sentences that avoid such redundancy (Examples 14, 17). They certainly do not differ in the Russian translations given in the Mari-Russian dictionary used as a source in this paper (Galkin et al. 1990– 2005). Is there a concrete semantic subtlety that is lost in translation here, or are they functionally equivalent? Within the scope of this particular paper, this question cannot and need not be answered. If a tangible difference exists between the respective sentences, it is presumably not one that can be easily expressed in the likely target languages of translations from Mari. My advice to students of Mari, and translators of Mari, is to confidently consider constructions with and without modifiers as equivalent for their purposes, if the modifier does not seem to be encoding any new information. Be that as it may, a systematic analysis of this phenomenon might yield interesting results. On the one hand, a comparison between the argument structures of converb constructions and of their base lexemes would suggest itself. On the other hand, the likelihood of the occurrence “redundant” modifiers in different contexts, dialect groups, etc. could be analysed. My work to-date on Mari converb constructions has exclusively focused on their interpretation and translation. Advanced L2 learners of Mari might want to use constructions of this sort themselves at some point. The analyses suggested above would also facilitate the creation of didactic materials aiming to convey active Mari language skills to the student. Systematic neglect of this mechanism is as noticeable in available teaching materials as it is in grammatical analyses of Mari.

Appendix: Inventory of modifiers The table below provides an overview of all modifiers, aspectual and directional, for which evidence could be found in our source materials (Alhoniemi 1985; Beke 1911; Pengitov et al. 1961; Čhaidze 1960; Moisio 1992; Galkin et al. 1990–2005;

Mari converb constructions – Interpretation and translation 

 157

Vasil’ev et al. 2003). Verbs are given with a translation of their base meaning; their transitivity is indicated in parentheses (t for transitive, i for intransitive). Then, their function as a modifier is explained and corresponding English adverbial particles are given for directionality markers, if they exist. As mentioned above, some modifiers can only be connected with transitive verbs and some only with intransitive ones. If such information is available, it is provided. A number of verbs have several independent functions as modifiers – e. g. kajaʃII, which can serve both as an aspectual modifier and as a directional modifier. In these cases, a modifier’s functions as aspectual and directional modifier are given in separate lines. Table 6: Modifiers in Mari Converb Constructions Modifier

Literally

As Modifier

βolaʃII

descend, sink (i)

(i) downwards movement, cf. English “down”

βoltaʃII

lower, let down (t)

(t) downwards movement, cf. English “off”

βont͡ɕaʃII

go over, cross (t)

(i) movement across or over, cf. English “across” and “over”

βozaʃI

lie down (i)

ertaraʃII

lead, take (t)

abruptness, finality downwards movement (i/t) durative

ertaʃII

go by, go through (i)

(i) movement through or past, cf. English “past” or “by”

ilaʃII

live (i)

durative

jarnaʃII

tire (i)

activity carried out to exhaustion

kajaʃII

go (i)

kijaʃII

lie (i)

(i) momentary, changing state, inchoative, unexpectedness movement away, cf. English “away (i/t) durative, action taking at one location

kɤɲelaʃI

get up (i)

(i) upwards movement, cf. English “up”

kɤʃkaʃII

throw (t)

kodaʃI

remain, stay (i)

(t) swift, purposeful execution (i/t) storminess, lack of control (i) finality, ostensible results

kodaʃII

(t) finality

kolaʃII

leave, not take with (t) die (i)

koltaʃII

send (t)

kondaʃII

bring (t)

(i/t) finality, momentariness, inchoative, diminutive (t) movement through (t) movement towards the speaker

koʃtaʃI

go, walk (i)

durative, action taking place in many different locations

kudaltaʃII

throw (t)

(i/t) rapidity, unexpectedness, resoluteness

kyzaʃII

climb, rise (i)

(i) upwards movement, cf. English “up”

to the point of exhaustion

158 

 Jeremy Bradley

Modifier

Literally

As Modifier

kyzɤktaʃII

raise, lift (t)

(t) upwards movement, cf. English “to the top”

lektaʃI

go, go out (i)

lijaʃI

become, be (i)

(i/t) completion (i) movement out, cf. English “out” inchoative

luktaʃI

lead out, remove (t)

mijaʃII

come, go (i)

nalaʃI

take (t)

naŋgajaʃI

take away (t)

(t) completion (t) movement out, cf. English “out”, “away” (i/t) durative, increasing effect movement up to something (t, occasionally i) finality (t) movement out, cf. English “out” (t) movement away, cf. English “away”

ojɤraʃII

divide (t)

(t) splitting something, cf. English “apart” and “through”

ont͡ɕaʃII

look (t)

optaʃII

put, place (t)

(i/t) momentariness, attempt, familiarizing oneself with action (t) rapidness, intensity

pɤʃtaʃII

put, place (t)

(t) finality

pɤtaraltaʃI be finished, come to an end (i) pɤtaraʃII finish, complete (t)

(i) finality

pɤtaʃII

end, finish (i)

puaʃII

give (t)

(i) finality, intensity, participation of all members of a group movement in different directions finality, existence of a beneficiary

puraʃII

come in (i)

(i) movement into, cf. English “in”, “into”

purtaʃII

bring in (t)

(t) movement into, cf. English “in”, “into”

saβɤrnaʃII

turn, rotate (i)

finality

seŋaʃII

win, defeat (t)

successful execution

sitaraʃII

(i/t) finality, activity carried out up to a limit

ʃint͡ɕaʃI

gather the needed amount (t) sit down (i)

(i) finality, change of state

ʃint͡ɕaʃII

sit (i)

(i) durative

ʃint͡ɕɤltaʃI

sit around (i)

durative

ʃɤndaʃII

put, seat (t)

(t) momentariness, transition

ʃogalaʃI

stand up (i)

(i/t) finality, activity occurring only once

ʃogaltaʃII

put, place (t)

(t) finality

ʃogaʃII

stand (i)

(i/t) durative

ʃogɤltaʃI

stand in one place (i) durative

ʃualtaʃI

happen (i)

activity carried out up to a limit or goal

ʃuaʃI

reach, arrive (i)

(i) finality

ʃuaʃII

throw (t)

(t) momentariness, rapidness, unexpectedness (t) movement away or off, cf. English “off”, “down”

(t) finality, intensity

Mari converb constructions – Interpretation and translation 

 159

Modifier

Literally

As Modifier

ʃuktaltaʃI ʃuktaʃII

be carried out, finality be executed (i) carry out, execute (t) (t) finality, successful execution

ʃytaʃII

drill, bore (t)

(t) activity perforating something, cf. English “through”

ʃytlaʃII

wear out (t)

(i) activity perforating something, cf. English “through”

temaʃI

fill up (i)

(i/t) finality, action up to limit, large number of participants

temaʃII

fill (t)

(t) action up to a limit, large quantity of object

tolaʃI

come (i)

tolaʃaʃII

try, attempt (i)

(i/t) durative, gradual increase of effect movement up to something marking activity as insufficient, belittling

tøt͡ɕaʃII

try, attempt (i)

activity not carried out properly

t͡ɕɤtaʃII

tolerate (t)

durative

t͡ɕumɤraʃII gather in a pile (t)

(t) gathering object in one place

utaʃII

action carried out abundantly

become superfluous (i)

Glossing Abbreviations Part of Speech

Gloss

Meaning

ad -ad.case

– – -COMP – – -COMP – – -ACC -DAT -ILL -INE – -PL – -1SG […] – -AND -STR

adjective adjective case comparative case adverb adverb degree comparative degree noun noun case accusative dative illative inessive noun number plural noun possessive suffix first person singular […] enclitic particle particle “and” strengthening particle

adv -adv.deg n -n.case

-n.num -n.poss

-enc

160 

 Jeremy Bradley

Part of Speech

Gloss

Meaning

nm -nm.case

– – -ACC – – – – -PT.ACT -PT.PASS – -GER -NMLZ

numeral numeral case accusative postposition pronoun verb adjective form of verb active participle passive participle adverbial form of verb gerund (in –n) derivation (verb to noun) (standard) nominalization infinitive of verb (standard) infinitive verb mood and person imperative second person singular verb person first person singular […] verb tense simple past tense I simple past tense II

pp pr v -v.ad

-v.adv -v.deriv.n -v.inf -v.mood-pers -v.pers

-v.tense

– -INF – -IMP.2SG – -1SG […] – -PST1 -PST2

References Abondolo, Daniel Mario (1998): Introduction. In: Daniel Mario Abondolo (ed.), The Uralic Languages, 1–42. New York: Routledge. Alhoniemi, Alho (1985): Marin kielioppi. Helsinki: Suomalais-ugrilainen seura. Alhoniemi, Alho (1986): Marin kielen lukemisto sanastoineen. Helsinki: Suomalais-ugrilainen seura. Alhoniemi, Alho (1993): Grammatik des Tscheremissischen (Mari). Hamburg: Buske. Bartens, Raija (1979): Mordvan, tšeremissin ja votjakin konjugaation infiniittisten muotojen syntaksi. (Mémoires de la Société Finnoougrienne 170.) Helsinki: Suomalais-ugrilainen seura. Beke, Ödön (1911): Cseremisz nyelvtan. Budapest: Magyar Tudományos Akadémia. Benzing, Johannes (1943): Kleine Einführung in die tschuwaschische Sprache. Berlin: Stollberg. Bereczki, Gábor (1990): Chrestomathia Ceremissica. Budapest: Tankönyvkiadó. Čavajn, Sergej Grigor‘evič (1967): Èlnet. Yoshkar-Ola: Marijskoe knižnoe izdatel‘stvo. Čhaidze, Mihail Pavlovič (1960): Sparennye glagoly v marijskom âzyke. Yoshkar-Ola: Marijskoe knižnoe izdatel‘stvo. Csúcs, Sándor (1990): Chrestomathia Votiacica. Budapest: Tankönyvkiadó.

Mari converb constructions – Interpretation and translation 

 161

Federal‘naâ služba gosudarstvennoj statistiki (2011): Okončatel‘nye itogi Vserossijskoj perepisi naseleniâ 2010 goda. Moscow. http://www.perepis-2010.ru/results_of_the_census/ (accessed October 31, 2013). Galkin, Il’â Savvič and Fedor Ivanovič Gordeev (1990–2005): Slovar’ marijskogo âzyka. Yoshkar-Ola: Marijskoe knižnoe izdatel’stvo. Haspelmath, Martin (1995): The converb as a cross-linguistically valid category. In: Martin Haspelmath and Ekkehard König (eds.), Converbs in Cross-Linguistic Perspective: Structure and Meaning of Adverbial Verb Forms – Adverbial Participles, Gerunds, 1–55. Berlin and New York: Mouton de Gruyter. Helimski, Eugen (2003): Areal groupings (Sprachbünde) within and across the borders of the Uralic language family: A survey. Nyelvtudományi Közlemények 100: 156–167. Klumpp, Gerson (2002): Konverbkonstruktionen im Kamassischen. Wiesbaden: Harrassowitz. Moisio, Arto (1992): Marilais-suomalainen sanakirja. Turku: Publications of the Department of Finnish and General Linguistics of the University of Turku. Pengitov, Nikolaj Tihonovič, Il’â Savvič Galkin and Isanbaev, Nikolaj Isanbaevič (1961): Sovremennyj marijskij âzyk: Morfologiâ. Yoshkar-Ola: Marijskoe knižnoe izdatel’stvo. Poppe, Nicholas (1963): Tatar Manual: Descriptive Grammar and Texts with a Tatar-English Glossary. Bloomington: Indiana University Publications. Schultze-Berndt, Eva (2003): Preverbs as an open word class in Northern Australian languages: Synchronic and diachronic correlates. In: Geert Booij and Ans van Kemenade (eds.), Yearbook of Morphology 2003, 145–177. Dordrecht: Kluwer. Slobin, Dan I. (2000): Verbalized Events: A Dynamic Approach to Linguistic Relativity and Determinism. In: Susanne Niemeier and René Dirven (eds.), Evidence for Linguistic Relativity, 107–138. Amsterdam: John Benjamins. Vasil’ev, Valerian Michajlovič and Zinovij Vasil’evič Učaev (2003): Marijsko-russkij slovar’. Yoshkar-Ola: Marijskoe knižnoe izdatel‘stvo. Wintschalek, Walter (1993): Die Areallinguistik am Beispiel syntaktischer Übereinstimmungen im Wolga-Kama-Areal. Wiesbaden: Harrassowitz.

Ulla Stroh-Wollin

Han and hon – Anaphoric pronouns in Early Scandinavia1 Abstract: The origin of the Scandinavian anaphoric pronouns han(n) “he” and hon (hun) “she” is a matter of dispute. An early hypothesis which derived the stem *hān- from an Indo-European demonstrative kēnos, attested in Aeolic Greek, was rejected at the beginning of the 20th century by the Swedish scholar Axel Kock, on the grounds that it was unlikely that this pronoun could have been preserved in Scandinavia but not in Germanic elsewhere. Instead, it would be a Proto-Scandinavian novation, which replaced an earlier common Germanic pronoun. However, the original hypothesis is formally unproblematic, while Kock’s own etymological proposal poses problems. We may also be less inclined today than a hundred years ago to regard Proto-Germanic as a reality in the sense of it being a language with hardly any variation, which makes Kock’s criticism somewhat outdated. Moreover, it is difficult to understand why an already established pronoun would be replaced by a new one. Thus, we should seriously consider the possibility that han/hon actually is an Indo-European heritage, not a novation.

1 Introduction Among the features that distinguish the Scandinavian languages from the other Germanic languages are the anaphoric pronouns han(n) “he” and hon (hun) “she”. The first instances of these pronouns are found in Viking Age runic inscriptions, but since some forms show umlaut, a Proto-Scandinavian existence can be presumed. It is also possible to reconstruct Proto-Scandinavian forms with high cer-

1 This article is a spin-off from my project “The syntax of the early Scandinavian noun phrase”, financially supported by the Swedish Research Council (project number: 421–2010–1272). I am grateful for the opportunity to deal with the etymology of han and hon, even though the investigation eventually led me to the conclusion that the history of these pronouns cannot shed light on the origin and spreading of the Old Scandinavian demonstrative (h)inn and the Scandinavian definite suffix emanating from it. The opposite outcome would certainly have had more immediate relevance for the issues dealt with in the project, but this does not make the actual result less interesting in a broader perspective.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 163

tainty. For instance, the masculine nominative has developed from PSc. *hānar (> Old Sc. hann) and the feminine nominative from Psc. *hānu (> Old Sc. hōn). On this point there is general agreement. However the origin of the Proto-Scandinavian stem *hān- is less clear and will be discussed in the following. Occasionally I will refer to the Scandinavian he- and she- pronouns by their modern Swedish forms, han and hon respectively, disregarding their transition from grammatical to semantic gender in modern Mainland Scandinavian (cf. modern English he and she). The nominative forms han and hon may represent all (original) cases, and sometimes a simple han denotes the full paradigm, i. e. both han and hon and their corresponding case forms. Pronouns in the third person neuter singular and plural (all genders) are based on a different stem than han and hon and are not encompassed by the paradigmatic label. One reason to have a closer look at the etymology of han is of course the remarkable fact that there is no general consensus about the origin of such a common word. Judging from a 1908 article by the famous Swedish linguist Axel Kock, there might have been agreement at the turn of the last century, namely that han was identical to a Greek demonstrative kēnos (Kock 1908: 186). However, this view is challenged by Kock, who finds it unlikely that this Greek pronoun would have been retained only on Nordic ground, but lost in all other Germanic languages. Instead, Kock assumes that han is a Proto-Scandinavian novation and relates it (or, more precisely, the latter part of it, i. e. PSc. *-ānar) to the Gothic demonstrative jains and German jener “that one/the other one”. Kock’s proposal, which will be further scrutinized below, is not unproblematic from a formal point of view. Nevertheless it was well received by contemporary scholars, especially among Kock’s compatriots; it is expressly adopted by e. g. Adolf Noreen (1913: 180) and Elof Hellquist (in his Swedish etymological dictionary, 1922) as well as in the extensive dictionary of The Swedish Academy (SAOB 1930: 11). Johannes Brøndum-Nielsen (1965: § 566), in his extensive work (in eight volumes, successsively published from 1928 to 1973) on the history of the Danish language, also presents Kock’s etymology as the most probable. Later, non-Swedish, dictionaries, e. g. de Vries (1962), Blöndal-Magnússon (1989) and N.Å. Nielsen (1989), often take a more cautious attitude, but still find it relevant to refer to the etymology proposed by Kock. Martin Syrett in The Nordic Languages (2002,  2:  772, with further references to Brøndum-Nielsen, cf. above, and H. F. Nielsen 1975: 10) also takes it seriously. One reason why Kock’s etymology found so much favour, despite the possible formal concerns, was, I believe, because the premise of his criticism of the prevailing etymology seemed very relevant; differences between the Scandinavian languages on the one hand and other Germanic languages on the other were preferably explained as Proto-Scandinavian novations, not as archaisms, which

164 

 Ulla Stroh-Wollin

follows from the implicit view of Proto-Germanic as a reality in the sense of one single homogeneous language, only later split into varieties. Today however we understand the prehistoric linguistic situation to be far more complicated, which gives us another reason to take an interest in the topic addressed here. Thus the origin of han concerns not only the exact etymology of the *hānstem, but also the issue whether it is necessarily a question of an novation or whether it may be an archaism. Normally only the former possibility is discussed, so e. g. in The Nordic Languages (Syrett 2002: 723): “Since neither pronominal root in *hin- nor *hān- has any immediate parallels elsewhere in Germanic, it is possible to argue that they are both innovations that arose together in the [Ancient Nordic] period”. Earlier linguists also seemingly took for granted that han has replaced a common Germanic anaphoric pronoun, based on the i- stem, still traceable in German er “he”. This view was still put forward by e. g. Elmar Seebold in his book on Germanic personal pronouns from 1984 (Seebold 1984: 64–66). But even more important, it is stated in the long-lived and influential textbooks in language history by (the Swede) Elias Wessén, e. g. in his Icelandic grammar (Wessén 1961: 93, quoted below), which means that it has been taught as a fact to generations of students and has become an accepted truth for researchers in language history. This tradition is certainly firmly rooted in Sweden. The impact of the tradition on this point may also explain its reproduction in later works of Gun Widmark (2001: 34, 2010: 223), who takes the putative novation of han to bear witness to a considerable contact over large distances within Scandinavia during the first centuries A. D. This kind of assumption really emphasizes why we should take interest in the origin of han. If han is a novation, this circumstance may have the kind of consequences Widmark presumes, which of course would be very much worth considering. However do we really know whether han is a novation or not? If it actually is an archaism, perhaps we can draw other, equally interesting conclusions about the prehistory of the Germanic languages. I think it is time to re-open the case of han and to do it with no preconceived opinion as to whether this pronoun is a novation or an archaism. In section 2 below, I will first take a closer look at different proposals for the etymology of han from a formal point of view. Then in section 3, I will turn to the question of what reason and evidence we have to assume a replacement of a prehistoric common Germanic anaphoric pronoun. In section 4, I summarize my conclusions.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 165

2 Proposals for the etymology of han There are two distinct proposals for the etymology of han, these are mentioned briefly in the introduction. According to the first proposal, han is an Indo-European heritage and related to a Greek demonstrative kēnos “that one” (so on the Aeolian Islands; cf. ekeīnos in classical Greek). This etymology is rejected in the following way by Axel Kock in his 1908 article mentioned above: Most linguists today probably regard Isl. hann Old Sw. han identical to Greek [kēnos]. This etymology could be possible from a formal point of view. However it would certainly be most surprising if this Greek [kēnos] was retained only on Nordic ground, but was lost in all other Germanic languages. (Kock 1908: 186; my translation)

According to the second proposal, presented by Kock, han developed from an assumed Proto-Scandinavian *jainar, related to Gothic jains.2 This *jainar is expected to drop its initial j- very early, resulting in an *ainar. This form is, according to Kock, later transformed into *ānar when pronounced “semi-fortis”, after which an initial h- is joined to it, in accordance with other h-initial pronouns. The author mentions, for instance, the demonstrative hi- stem in Gothic, realized in the forms hina, hita and himma. Kock’s article has been influential. However its long-term impact seems to concern primarily the criticism of the kēnos-etymology, whereas the concrete proposal, at least in part, is met with some scepticism, see e. g. the following passage from Elias Wessén’s Icelandic grammar: The word hann, hon is particular to Scandinavian (and has no equivalent in other Germanic languages); it has replaced an older word is: Gothic is, ija, ita (German er, sie, es, English he, she, it), Latin is, ea, id. Obviously there is a connection to Low German he, Dutch hij […], English he [… ,] and [hann] has arisen by means of a fusion of this word and some other pronominal stem; although it is unclear which one. (Wessén 1961: 93; my translation)

Wessén states that han is a novation and he accepts Kock’s idea that the initial h- is an influence from some other pronoun, but the origin of the second part is left open.

2 In his article Kock informs the reader that Swedish linguist Otto von Friesen, according to their personal communication, had independently drawn the same conclusion concerning the etymology of han. Furthermore, it is also noted that basically the same idea had earlier been presented by Isidor Flodström (1883–84: 58–59), which Kock was unaware of until his own article was going to press.

166 

 Ulla Stroh-Wollin

Below I will take a closer look at the two distinct etymologies raised in the debate. I will also discuss the implications of the more open attitude reflected in e. g. the quote from Wessén above. 2.1 The kēnos etymology from a formal point of view The hypothesis that han is derived from a Proto-Indo-European kēnos is, in fact, uncomplicated from a formal point of view (even though Kock’s words, cf. above, implicate some reservation on this point). A sound-by-sound shift gives us the presumed Proto-Scandinavian *hānar, see Figure 1, from which we get Old Sc. hann (by syncope of the short a and assimilation of n and r, followed by a shortening of the long ā). PIE PSc.

k ↓ h

ē ↓ ā

n ↓ n

o ↓ a

s ↓ r

Fig. 1: Sound-by-sound shift from PIE kēnos to PSc. *hānar

Disregarding the ending -s (masc. sing. nom.) of kēnos, correlating to PSc. -r, we start with the stem kēno- with a long ē, which actually is the result of a merger of two primary pronominal stems, ke- and eno-. The initial k- in kēno- would of course, according to Grimm’s Law, be transformed to an h- on Nordic ground. And the shift of the vowels, i. e. long ē > long ā and short o > short a would be equally regular. In sum, the virtues of the kēnos etymology are clear: a start of which we have concrete evidence in (Aeolian) Greek and the presumed result, PSc. *hānar > Old. Sc. hann, when applying the relevant sound laws.

2.2 The etymology proposed by Kock from a formal point of view The etymology proposed by Kock starts with a Proto-Scandinavian *jainar, presumed in relation to Goth. jains. The subsequent steps of the development are the following: *jainar > *ainar > *ānar > *hānar. The loss of the initial j- in this word is uncomplicated, since Proto-Scandinavian j-drop is a well-known phenomenon (cf. Germ. jung – Sw. ung). However the transformation of ai to ā and the addition of a secondary h- are more doubtful steps of the assumed development.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 167

Kock takes the of Goth. jains (and PSc. *jainar) to represent a diphthong, but the phonetic value of in jains was – and is – widely doubted. Normally it is held to represent a short e (Lehmann 1986: 210). Further, a transformation of a diphthong [ai] to [a:], if we count with the possibility that may represent a diphthong, is also questionable. On this point, Kock’s hypothesis relies on the assumption of a “relatively unstressed” pronunciation (“semi-fortis”) of the word, cf. the cardinal number *ainar ‘one’, which retained its diphthong, though with some phonetic drift (*ainar > *æinar > æinn/einn) in all of Scandinavia for a long time. The development of the Proto-Scandinavian diphthong ai in non-primary stress positions is a disputed matter. The debate is extensively accounted for in an article by Elmevik from 1978. The view that the semi-fortis diphthong was contracted to a long ā has been criticised by various linguists and the author rejects it firmly. The addition of the initial h- may be a less crucial point, but this part of the hypothesis also needs some comment. Obviously there is a demonstrative pronominal stem hi- in Proto-Germanic, for which we have concrete evidence in e. g. Goth. hina, hita, himma. We can even observe it in German heute “today” (< hiu tagu “this day”) and in adverbs such as Engl. here, Germ. hier, Sw. här etc. It is also likely that hi- represents the first part of Old Sc. hinn (pre-adjectival definite article and demonstrative with so called jener-deixis), the second part being *ena-, or *en- after syncope. In this context we can ignore whether or not *en(a)- has its origin in *jaina- (as assumed by Kock and others); it could maybe also be directly derived from PIE eno-. However, the result of the fusion of *hi- and *en(a)-, i. e. hin(a)- merits attention, since it retains the quality of the vowel in *hi-. If *hi- is also assumed as the first part of *hānar, it has not affected the phonetic value of the stem vowel at all, which may be unexpected, considering its effect on hinn. An alternative hypothesis is that the h- in *hānar is just inspired by the h- in *hi- and hinn. This possibility cannot be totally ignored, but there is no obvious evidence. Indeed, Kock (1908: 188) assumes that some h-less examples of hann in Old Scandinavian manuscripts possibly reflect an earlier *ānar. However h-dropping is found occasionally, and may affect all sorts of words, so the relevance of scarcely found examples of h-less anaphoric pronouns in medieval texts is very questionable. In this case, runic inscriptions offer a more interesting object of investigation, since they are closer in time to the assumed *ānar and we know that they show h-less forms to a considerable degree. Such an investigation is also much easier to carry out today than in the time of Kock. A simple count of (all case forms of) hann and hōn with and without h- in the register of the words appearing in Swedish Viking Age runic inscriptions (Peterson 1994) shows that as much as c.22 % lack the initial h-. However a count of instances with and without h- of (the infinitive

168 

 Ulla Stroh-Wollin

form of) the verb haggva “cut” gave a similar result: c.21 % without h-.3 The word haggva was chosen for comparison because it is fairly common and because the phonetic context can be assumed comparable with that of hann. It also became clear that the instances without h-, in both cases, were most numerous in the province of Uppland, where h-dropping is a known feature of some dialects. A complementary check in the Scandinavian Runic Text Database on instances of hann and hōn in Denmark and Norway only revealed two more instances of h-less forms, both from Denmark. Thus, the runic inscriptions give no positive evidence to support the idea of a prehistoric h-less anaphoric pronoun. In summary, Kock’s proposal concerning the etymology of han is founded on a number of assumptions with little or no empirical support. Two assumptions are highly problematic: the assumption that represents a diphthong in Goth. jains as well as in the presumed Proto-Scandinavian *jainar and the assumption that this diphthong would be transformed into a long ā. In the event that , as normally assumed for Goth. jains, represents a short e, the transformation into a long ā would be even more spectacular. 2.3 Alternative etymologies? Of course it is possible to have concerns about both the etymologies just scrutinized. For instance the quote from Wessén above reflects a view whereby it is self-evident that han is a novation (which “has replaced an older word is”), but not that Kock’s etymology is correct. Wessén writes of a fusion between a pronoun reflected in e. g. Engl. he and “some other pronominal stem; although it’s unclear which one”. However could there be an alternative etymology? Many (if not all) Indo-European languages share personal pronouns for the first and the second person, but there is no common or widespread personal

3 Runic inscriptions are often damaged and sometimes lost, in the latter case known only by earlier transcriptions. My count is, with one exception, based entirely on my assessment of the instances as they appear in Peterson’s Svenskt runordsregister, which means that the figures are rough approximations. The exception is an instance of on found on the Rök stone (with the signature Ög 136 in SRI (= Sveriges runinskrifter, Stockholm 1900–) and the Scandinavian Runic Text Database), registered as an h-less hann in the Svenskt runordsregister and interpreted the same way in the Scandinavian Runic Text Database. In this context an h-less hann on the Rök stone would have been of great interest since this inscription is very early. However I take the prevailing view of the on in question as highly improbable and I am supported on this point by Bo Ralph (p. c.), who lately has done effortful work on the interpretation of the Rök inscription, see e. g. Ralph (2007) (an article which, however, does not treat the passage of the inscription where the on here discussed is found).

Han and hon – Anaphoric pronouns in Early Scandinavia 

 169

pronoun for the third person. The third person pronouns, or anaphoric pronouns, have different origins in different languages, but it is generally assumed that they derive from demonstratives. For Indo-European languages we expect the origin to be some Indo-European demonstrative. Thus, a good way to start an investigation would be to make an inventory of what demonstrative stems we can find in older Germanic. The most widespread demonstrative is the one we still have in Engl. that, German der, die, das and Sw. den, det etc. This pronoun is actually partly used with anaphoric reference in Scandinavian, namely for neuter and plural referents, and this has been the case for as far back as written records go. Originally, the paradigm had suppletive nominative singular forms (with an initial s-) for the masculine and feminine genders, later replaced by more regular forms (with dental consonants) in Mainland Scandinavian; cf. Runic Sw. sā (masc. sing.) and sū (fem. sing.) vs. Old Sw. þænn and þe respectively. The suppletion seems to be an Indo-European heritage; cf. Sanskrit masculine sá and feminine sā vs. neuter tád (Beekes 1995: 202–204). All Germanic languages have also developed reinforced demonstratives on the basis of the sā pronoun, e. g. Engl. this, Germ. dieser, Sw. denna. Figure 2 below illustrates, with Runic Swedish forms, the expanded use of the inherited demonstrative in Scandinavian. The neuter and plural forms of the demonstrative are used also with anaphoric reference, but the exclusively anaphoric hann- and hōn-forms cover all cases for singular masculine and feminine referents. This distribution is identical in the earliest recordings all over Scandinavia, even though single forms can have slightly different guises in different varieties. Demonstrative pronouns Masc. Fem. Neutr.

Masc.

Sing.: Nom. Gen. Dat. Acc.

sā þæs þæim þænn

sū þæira þæiri þā

þæt þæs þy þæt

hann hans hōnum hann

hōn hænnar hænni hāna

þæt þæs þy þæt

Plur.:

þæir þæira þæim þā

þār þæira þæim þār

þau(n) þæira þæim þau

þæir þæira þæim þā

þār þæira þæim þār

þau(n) þæira þæim þau

Nom. Gen. Dat. Acc.

Anaphoric pronouns Fem. Neutr.

Fig. 2: Runic Swedish forms of demonstrative and anaphoric pronouns. The anaphoric paradigm has adopted the demonstratives for the neuter gender and for the plural (shaded forms), but all singular case forms for the masculine and feminine genders derive from the PSc. *hān- stem.

170 

 Ulla Stroh-Wollin

Obviously, the different forms of hann and hōn are not related to the sā pronoun. Thus, the question is what other demonstrative stems- there might be. In fact, it seems that they have all, more or less, already been identified. Among the pronominal stems mentioned above, we can immediately, on phonetic grounds, dismiss the i- stem from the discussion. The same holds true for an alternative si- stem, found in a few forms within the i- paradigm. The geographical spread of the i- stem is also unclear, even though it is traditionally held to be the first anaphoric pronominal stem in all varieties of Proto-Germanic (cf. next section). However, we undoubtedly have i- stems in Goth. is “he” and can still trace it in Germ. er “he”. We also find it in Latin is “this one”4 and Lithuanian jìs “he” (Lehmann 1986: 207). More relevant here are the *hi- and *ena-stems, which in combination give Old Sc. hin-. The first stem *hi- corresponds to PIE ki-, and we find it also in e. g. Latin cis and Lithuanian šìs “this one” (Lehmann 1986: 183). The second stem, *ena-, corresponds to PIE eno-. Goth. jains is a combination of this stem and a first element of disputed origin (Lehmann 1986: 210). The stem is also found in Lithuanian añs “that one”, Old Prussian tāns “he” and Old Church Slavonic ona “that one, he” (Pokorny, accessed online). Now, in order to get a Proto-Scandinavian stem *hān- we may seek the origin of the h in *hi- or hin- and the origin of the n in *ena-. However it is problematic to determine the origin of the long ā. If we do not accept the etymology proposed by Kock, where a stem *āna- is derived from *jaina-, there is actually no known alternative – which of course is the very reason why Kock devised the idea of relating *hān- to Goth. jains. In summary, the conclusion must be that simply referring to “some other pronominal stem”, as Wessén puts it in his Icelandic grammar (1961: 93, cf. above), is taking the easy way out. It can also be noted that Seebold (1984: 65–66) on the one hand describes han as a combination of the (PIE) roots ke- and eno-, i. e. the components of the classical (Aeolian) Greek kēnos, but on the other hand takes for granted that han is a Proto-Scandinavian novation. This is somewhat confusing. If a novation, han can not build on Proto-Indo-European roots in their original form. We have to take their later Germanic phonetic guise as the point of departure, which has proven problematic. The identification of *hān- as a combination of PIE ke- and eno- is probably only possible if we start with the fused stem present in kēnos, which, in turn, means that this etymology is probably not consistent with the novation hypothesis.

4 The i- stem appears in various ablaut grades, which explains e. g. the feminine form ea in Latin (Prokosch 1939: 274).

Han and hon – Anaphoric pronouns in Early Scandinavia 

 171

3 R  eplacement of a common Proto-Germanic anaphoric pronoun? As already stated, it has been presumed that han replaced an older anaphoric pronoun of Proto-Germanic origin in Scandinavia. Below this hypothesis will be scrutinized from different angles. The first issue concerns the presumption a common Proto-Germanic anaphoric pronoun. Then the idea of a replacement of anaphoric pronouns as such will be discussed. Finally, the specific Scandinavian suppletion within the anaphoric paradigm, i. e. the use of (originally) demonstrative forms besides han and hon, will be illuminated.

3.1 Evidence of a common Proto-Germanic anaphoric pronoun? The presumption that Proto-Germanic originally had only one anaphoric pronoun can be challenged for (at least) two reasons. First, it seems that non-Scandinavian Germanic actually developed anaphoric pronouns from two different stems. Secondly, there is very little empirical evidence that Scandinavian has ever had other anaphoric pronouns than the ones still used. The putative common Germanic anaphoric pronoun is based on the stem i-. We find it, without doubt, in Gothic is “he” and German er “he”, but e. g. the Old English paradigm has initial h- throughout. Figure 3 presents the anaphoric paradigms in Gothic and Old English, following Prokosch (1939:  274, see also Prokosch 1939: 275 for comments on various forms).

Masc.

Gothic Fem.

Neutr.

Masc.

Old English Fem.

Neutr.

Sing.:

Nom. Gen. Dat. Acc.

is is imma ina

si izōs izai ija

ita is imma ita

he his him hine

hēo hiere hiere hīe

hit his him hit

Plur.:

Nom. Gen. Dat. Acc.

eis izē im ins

*ijōs izo im ijos

ija izē im ija

hīe hiera him hīe

hīe hiera him hīe

hīe hiera him hīe

Fig. 3: Anaphoric pronouns in Gothic and Old English

Now, one may ask whether the English pronoun is basically the same as the Gothic, i. e. based on the i- stem with a secondary initial h-, or based on the hi-

172 

 Ulla Stroh-Wollin

stem. Seebold (1984:  66) takes the first possibility as rather certain (“ziemlich sicher”). However, he does not really present arguments for this point of view. It may just follow from a preconceived opinion that Proto-Germanic could have only one anaphoric pronoun. Seebold’s remarkable certainty concerning han as a novation “obviously introduced to replace the archaic forms of the i- pronoun” (Seebold 1984: 64–65, my translation) yields the same impression. The initial h- in the Old English paradigm is according to Seebold (1984: 66) inspired by corresponding forms (“ähnliche Formen”) like Old.Sc. hann, an interesting point of view considering that e. g. Wessén (1961: 93) takes han as the result of a fusion of the h-initial pronoun present in English he (as well as in Dutch hij and Low German he) and some other pronominal stem. The Oxford English Dictionary (1989) takes Engl. he to be derived from the hi-stem, and this is probably also the prevailing assumption in modern linguistics. This possibility is, for some reason, never considered by Seebold. However it is plausible, as we do have empirical evidence that a demonstrative hi- stem existed on Germanic ground. It is, in fact, not difficult to imagine a pattern where south-eastern Germanic varieties developed an anaphoric pronoun based on the demonstrative i- stem (which we also find in demonstratives in surrounding non-Germanic languages), whereas the Germanic varieties in the north-western part of the continent developed an anaphoric pronoun based on the hi- stem, which later spread with the colonizers to England. In summary, it is very plausible that two different anaphoric pronouns developed side by side in continental Germania. If so, we might as well accept the possibility that Scandinavian developed an anaphoric paradigm independently of the processes on the continent. Nevertheless, we will take a closer look at what empirical evidence supports the traditional assumption that an anaphoric pronoun based on the i- stem was also spread in Scandinavia. This evidence is, in fact, poor. There is a possible reminiscence of the masculine singular nominative form of an i- pronoun in Scandinavia. The former relative particle es/er (the forms found in Viking Age runic inscriptions) > Old Sc. er/ær is normally taken to derive from this form, cf. Goth. is; Germ. er. Some instances of es in the runic inscriptions from the Viking Age may also be interpreted as “he”. Occasionally this interpretation may even seem preferable to the relative-clause reading. However, the relative-clause reading is never impossible and in most ambiguous cases, both readings are equally good. There are also several examples already in the runic inscriptions where the relative-clause reading is the only possible alternative, because the relativized constituents are not masculine and/ or singular and/or nominative (Stroh-Wollin 1995). There is no evidence of any other form of a Scandinavian i- pronoun than the putative masculine singular nominative es (> er > er/ær), i. e. out of twenty-four

Han and hon – Anaphoric pronouns in Early Scandinavia 

 173

combinations of gender, number and case, only one is – possibly – attested. On the whole, considering the weak empirical support, it may actually be rather daring to presume that a full paradigm of the anaphoric i- pronoun was once established and spread throughout Scandinavia.

3.2 The replacement hypothesis When discussing the probability, from a theoretical point of view, that han and hon have replaced some forms of the i- pronoun, one has to raise one very fundamental question: Why should a language take on new anaphoric pronouns if there already were perfectly useful ones? There must be some reason other than “just for fun”. I find it also very hard to believe that the *hān- forms was “a Nordic innovation designed to reinforce the distinction between ‘real’ and ‘grammatical’ gender”, as suggested by Martin Syrett in The Nordic Languages (2002: 722). This hypothesis would explain the fact that only eight out of twenty-four possible forms were based on the *hān-stem. However, we see no reflections of any distinction between “real” and “grammatical” gender in the written records of Old Scandinavian; the different forms of hann and hon were in no way reserved for animate referents. The author does not present any support, empirical or typological, for the hypothesis, and there are, to the best of my knowledge, no parallels in other comparable languages.5 As I see it, it is practically impossible to believe in an exchange of anaphoric pronouns without presuming some kind of language contact, most probably between the old population and influential newcomers to a region. This was the prevailing situation in England during the Dane law. The Dane law meant a huge invasion of Vikings to the country, mostly Danish, firmly exercising their power within the held area. And we actually do find an interesting shift of anaphoric pronouns here. English has completely lost the old hi- forms in the plural, cf. the Old English and Modern English forms of the anaphoric pronouns in Figure 4. But we may also note that the forms that won the conquest in this case were not unknown to the old population. The English just had to accept a wider use of the

5 Comparable languages are e. g. contemporary European languages with a similar grammatical system. This is not saying that semantic gender is non-existent; many languages have semantic gender (see e. g. Corbett 1991). Nor am I denying that anaphoric pronouns are semantically distributed e. g. in Modern English – and partly so in modern Mainland Scandinavian.

174 

 Ulla Stroh-Wollin

demonstrative pronouns they had in common with the Scandinavians and which the latter already had adopted for anaphoric reference in the plural.

Masc. Sing.:

Plur.:

Nom. Gen. Dat. Acc. Nom. Gen. Dat. Acc.

he his him hine

Old English Fem. hēo hiere hiere hīe

Modern English Fem. Neutr.

Neutr.

Masc.

hit his him hit

he his

she her

it its

him

her

it

hīe hiera him hīe

they their them

Fig. 4: Anaphoric pronouns in Old English and Modern English

At present we have no evidence of a similar situation that could explain a shift of anaphoric pronouns in prehistoric Scandinavia. The Swedish scholar Gun Widmark touches on the issue, but from the traditional view of a replacement of pronouns as the point of departure. The following passage concerns the Proto-Scandinavian era. Scandinavia tears itself away from the continent, not only by means of phonetic changes, but also by spectacular lexical novations. […] The new pronoun han/hon e. g. replaces common Germanic pronouns. […] The remarkable thing is both that new words of high frequency are being used and that they have conquered the entire speech community of Scandinavia. Such a substantial and fast diffusion is difficult to understand, if at all possible, if it does not have its origin in a group of people dominating the whole area. (Widmark 2010: 223, my translation)

Thus, Widmark takes (among other things) han and hon to testify to a widespread and well organized power elite in prehistoric Scandinavia (see further Widmark 2010, section 4.2). I cannot for the moment assess how realistic that description of the society as such really is. It would certainly be very interesting if linguistics could contribute to our understanding of prehistoric society with regard to this issue. However, as must be clear from the investigation above, I take it as highly questionable whether han/hon can be used as an argument in this case. And, vice versa, as long as we do not have more independent data supporting the hypothesis of a dominating elite, it is highly uncertain if it is at all possible to imagine a very uniform and relatively fast spread of new pronominal forms all over Scandinavia. Furthermore, even if it is, the question of the origin of the novation still remains.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 175

A striking similarity between the Modern English anaphoric paradigm and the anaphoric pronouns in Scandinavian, see Figure 2 above, is the suppletion. In the case of English we know for sure that the suppletion is caused by a partial replacement and it is of course reasonable to suspect a similar explanation as regards the Scandinavian paradigm. However, as none of the forms is an ipronoun, the suppletion does not in fact imply a change according to the traditional view. Rather, it confuses the picture if we have to assume that some forms of a lost paradigm have been replaced by *hān-forms, whereas others have been replaced by the corresponding forms of the demonstrative. The fact that there are no remnants of an i- pronoun in the North Germanic languages does not exclude the possibility that i- pronouns were once used somewhere on Scandinavian ground. They may have later succumbed to the pressure of the words we now recognize as our anaphoric pronouns, for instance if these latter were used by influential immigrants. But it is highly improbable that a full paradigm of anaphoric i- pronouns, well established all over Scandinavia, would have ever been replaced. In a similar way, I have earlier argued that sum first came into use as a relative particle in the provinces around Lake Mälaren in Sweden, where relative clauses early on tended to be unintroduced.6

3.3 A closer look at the Scandinavian suppletion pattern The question of why only eight forms within the Old Scandinavian anaphoric paradigm were based on the *hān- stem is occasionally brought to the fore (e. g. by Syrett 2002: 722), but, interestingly, it is, according to my experience, never discussed why the other sixteen forms had their origin in the demonstrative sā paradigm. Also the extended use of the ordinary Germanic demonstratives was originally a characteristic of the Scandinavian branch (only later exported to England), a fact which may strengthen the suspicion that the Scandinavian anaphoric paradigm developed independently from parallel processes on the continent. It can also be noted in this context that we, besides the neuter and plural forms of the demonstrative, occasionally find the singular nominative forms sā (masc.) and sū (fem.) used with anaphoric reference in Viking Age runic inscrip-

6 We find the first instances of relative sum in runic inscriptions from the 11th century in the Mälar region, from where it spread all over Scandinavia by replacing other relative particles. The replacement was completed in Denmark by c.1400 and in Norway by c.1500. See further: Stroh-Wollin (1995: 99–100); Diderichsen (1941: 159); Lindblad (1943: 115–116).

176 

 Ulla Stroh-Wollin

tions. Thus, we can actually catch a glimpse of an on-going competition between sā and hann and between sū and hōn.7 Table 1 shows the counts of anaphoric sā and sū versus hann and hōn in runic inscriptions from six regions, distributed over five periods of time. The regions are geographically ordered from north to south, i. e. from the provinces north and south of Lake Mälaren in Sweden to Denmark in the south.8 Table 1: The distribution of sā/sū with anaphoric reference versus hann/hōn in Viking Age runic inscriptions from five Swedish provinces and Denmark from the 9th to the 12th century. Skåne, which is now a province of Sweden, is here part of Denmark.

Approx. dating: Region: Uppland Södermanland Östergötland Småland Västergötland Denmark

sā/sū : hann/hōn 800–900 900–1000

3* : 0

0:3

1000–1050

1050–1100

1100–

5 : 25 4 : 27 5:6 1:4 7:1 8:7

0 : 32 0:7

0:2

* All three instances are found in one and the same inscription, Ög N288 (signature used in SRI and the Scandinavian Runic Text Database).

As can be seen in Table 1, there is a clear dominance of hann/hōn in the northern provinces, whereas there is a more even distribution further south. The earliest instances are three sā from Östergötland. All three instances are, however, found in one and the same inscription, the one on the rock of Oklunda (with the signature Ög N288 in SRI and the Scandinavian Runic Text Database), and there are also very early instances of hann from Denmark.

7 There is practically no evidence of corresponding oblique forms of the demonstrative used with anaphoric reference. However one possible instance is the dative form þæi[m] “him?” in an early (c.900) Danish inscription, with the signature DR 230 in DR and the Scandinavian Runic Text Database. Cf. also DR 370 where a genitive (singular or plural) of the strengthened demonstrative sasi/þænni “this” is used with anaphoric reference in the phrase anda þæssa “his/their spirit(s)”. 8 The investigation concerns only nominative forms, cf. footnote 7. The dating is very approximate, based on information given in the Scandinavian Runic Text Database. A few non-dated instances are not accounted for in table 1. The instance of on on the Rök stone (Ög 136), traditionally interpreted hann, is also left out in this investigation, cf. footnote 3.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 177

The overall picture is that the nominative forms hann and hōn compete with sā and sū all over the investigated area and that hann and hōn have been widespread on Scandinavian ground since long. The latter conclusion is also supported by the feminine form hān (instead of hōn) on the island Gotland in the Baltic. The vowel reveals an early presence on the island, as the u-umlaut did not affect the Gutnish variety; the word had probably had the same form as in the mainland varieties, if it was imported after the umlaut period. Now, what kind of conclusions can be drawn from the suppletion of the Scandinavian anaphoric paradigm and the early variation within it? Considering the suppletive paradigm of Modern English, one may of course speculate about a prehistoric encounter between speakers of two Scandinavian varieties, resulting in the surviving mix of forms. It is also possible to imagine that a suppletive paradigm arises if, at some point, only some forms of an inherited kēnos pronoun had survived and the lacunae were filled with forms from the ordinary demonstrative paradigm. The variation accounted for in Table 1 is consistent with both theories. We can never know for sure exactly what happened. But, to sum up, we have empirical evidence that *hān- forms as well as forms from the demonstrative paradigm were used early as anaphoric pronouns, though, as it seems, with some internal competition up to the 11th century. Further, in contrast, we have no clear evidence of a Scandinavian i- pronoun. This latter fact does not exclude the possibility that i- pronouns were once used somewhere on Scandinavian ground (cf. above), but this does not affect the picture on the whole. It is more likely that the *hān- forms were archaic remnants of the kēnos paradigm (surviving at least in some region of Scandinavia), than that they were created to replace other pronominal forms, whether i- pronouns or sā forms.

4 Concluding remarks In this article I have discussed a couple of traditional positions concerning the Scandinavian anaphoric pronoun han/hon. This pronoun is, according to the tradition, a Proto-Scandinavian novation, which replaced an earlier common Germanic pronoun based on the i- stem found in Gothic is, German er. The novation hypothesis is partly founded on a proposal concerning the etymology of han by Axel Kock from 1908. Kock rejects the standpoint, prevailing by that time, that han is to be derived directly from an Indo-European demonstrative, found in the Aeolian Greek kēnos, since he finds it improbable that this word would have been retained only on Nordic ground, but lost in all other Germanic languages. This argument reveals an implicit theoretical standpoint, namely that the Proto-Germanic language was very uniform and hardly had any variation of any

178 

 Ulla Stroh-Wollin

kind. The same way of thinking might also, consciously or not, lie behind the position that also the Scandinavian languages must have had the same anaphoric pronoun as the continental Germanic languages. It is hard to get away from the impression that a common Germanic pronoun has been taken more or less for granted, since the only empirical support for assuming an i- pronoun in Scandinavia is the former relative particle er/ær (< er < es), which presumably has its origin in the masculine singular nominative form of the i- pronoun, and a few instances of es in runic inscriptions, which – possibly – stand for “he”. No other forms of an i- pronoun have been attested in written records. I believe it is, for theoretical reasons, impossible to maintain the view that a full paradigm of the i- pronoun would have been established throughout Scandinavia. Languages regularly develop anaphoric pronouns from demonstratives, when a need for some reason arises, but they do not replace one anaphoric pronoun by another just for the sake of it. Thus, we must imagine that the *hānstem, whatever form it had when it happened, first gained anaphoric use somewhere where it did not compete with an already established alternative. This conclusion has in itself no bearing on the specific etymology of han. However, when scrutinizing the given alternatives, only the kēnos etymology is formally uncomplicated. Kock tries to combine elements of two different pronominal stems found in Germanic, but his proposal is, to say the least, problematic. On the other hand, it seems to be the only possible alternative, if one has to seek the origin in known Germanic pronominal stems. So what conclusions can be drawn? I think the only possible conclusion we may draw concerning the etymology of han is that it might be an Indo-European heritage. It has not been proven that the kēnos etymology is correct; the origin of han may just be wrapped in the obscurity of history. But it is still possible and it is, in fact, unproblematic from the formal point of view to derive han from kēnos. Kock rejected this etymology because he had difficulty seeing how a pronoun could survive only in Scandinavia but nowhere else on Germanic ground. But what if we do not accept the theoretical basis for his suspicion? If we do not take for granted that the Proto-Germanic language was extremely uniformly spread, then we have to take seriously the possibility that han is an archaism rather than a novation. I also want to stress the importance of discussing this possibility, because it may open highly interesting perspectives about the way the Indo-European language spread and developed in this part of Europe. Etymology should be appreciated not only as the science of the history of individual words, but also as the necessary field work of “linguistic archaeology”.

Han and hon – Anaphoric pronouns in Early Scandinavia 

 179

References Beekes, Robert S. P. (1995): Comparative Indo-European Linguistics. Amsterdam and Philadelphia: John Benjamins. Blöndal Magnússon, Ásgeir (1989): Íslensk Orðsifjabók, Reykjavík: Orðabók Háskólans. Brøndum-Nielsen, Johannes (1965): Gammeldansk Grammatik i Sproghistorisk Fremstilling, Vol 5. Pronominer. København: Schultz. Corbett, Grevill (1991): Gender. Cambridge: Cambridge University Press. de Vries, Jan (1962): Altnordisches Etymologisches Wörterbuch. Leiden: Brill. 1st ed. Diderichsen, Paul (1941): Sætningsbygningen i Skaanse Lov. København. DR = Danmarks Runeindskrifter (3 vols: Text, Atlas; Registre) (1941–42), Lis Jacobsen and Erik Moltke (eds.). København: Munksgaard. Elmevik, Lennart (1978): Utvecklingen av urnord. ai i icke huvudtonig ställning. Ett ljudhistoriskt problem i ny belysning. Meijerbergs arkiv för svensk ordforskning 15: 5–85. Flodström, Isidor (1883–84): Härledningen av pronomenet hann. Nordisk tidskrift for filologi 6 (ny række): 58–59. Hellquist, Elof (1922): Svensk etymologisk ordbok. Lund: Gleerups. 1st ed. Kock, Axel (1908): Etymologiska anmärkningar. Arkiv för nordisk filologi 24: 179–198. Lehmann, Winfred P. (1986): A Gothic Etymological Dictionary. Leiden: Brill. Lindblad, Gustaf (1943): Relativ satsfogning i de nordiska fornspråken. Lund: Gleerups. Nielsen, Hans Frede (1975): Morphological and phonological parallels between Old Norse and Old English. Arkiv för nordisk filologi 90: 1–18. Nielsen, Nils Åge (1989): Dansk Etymologisk Ordbog. Ordenes Historie. København: Gyldendal. 4th ed. Noreen, Adolf (1913): Geschichte der nordischen Sprachen. Besonders in altnordischer Zeit. Strassburg: Trübner. Oxford English Dictionary (1989), John Andrew Simpson and Eva S. Weiner (eds.). Oxford: Oxford University Press. 2nd ed. Peterson, Lena (1994): Svenskt runordsregister (Runrön 2). Uppsala: Uppsala universitet. 2nd ed. Pokorny, Julius: Indo-European Etymological Dictionary. Indogermanisches Etymologisches Woerterbuch. http://dnghu.org/indoeuropean.html (accessed June 4, 2014). Prokosch, Eduard (1939): A Comparative Germanic Grammar. Philadelphia: University of Pennsylvania. Ralph, Bo (2007): Rökstenen och språkhistorien. In: Lennart Elmevik (ed.), Nya perspektiv inom nordisk språkhistoria. Föredrag hållna vid ett symposium i Uppsala 20–22 januari 2006, 121–143. Uppsala: Kungl. Gustav Adolfs Akademien för svensk folkkultur. SAOB (Svenska Akademiens Ordbok) = Ordbok över svenska språket utg. av Svenska Akademien 1– (1898–). Lund: Gleerups. Scandinavian Runic Text Database. Department of Scandinavian Languages, Uppsala University. http://www.nordiska.uu.se/forskn/samnord.htm (accessed June 4, 2014). Seebold, Elmar (1984): Das System der Personalpronomina in den frühgermanischen Sprachen. Göttingen: Vandenhoeck & Ruprecht. SRI = Sveriges runinskrifter (14 volumes to date) (1900–). Stockholm: Almqvist & Wiksell. Stroh-Wollin, Ulla (1995): Varför säger vi som? Om inbrytningen av runsvenskt sum i funktionen som relativpartikel. Språk och stil 4: 99–131. Syrett; Martin (2002): Morphological developments from Ancient Nordic to Old Nordic. In: Oskar Bandle, Kurt Braunmüller, Ernst Håkon Jahr, Allan Karker, Hans-Peter Naumann and

180 

 Ulla Stroh-Wollin

Ulf Teleman (eds.), The Nordic Languages. An International Handbook of the History of the North Germanic Languages, Vol. 1, 719–729. Berlin and New York: De Gruyter. Wessén, Elias (1992): Isländsk grammatik. Edsbruk: Akad.-Tryck. First Published Stockholm: Svenska bokförlaget [1961]. Widmark, Gun (2001): Det språk som blev vårt. Ursprung och utveckling i svenskan. Urtid – Runtid – Riddartid. Uppsala: Kungl. Gustav Adolfs akademien för svensk folkkultur. Widmark, Gun (2010): Det nordiska u-omljudet. En dialektgeografisk undersökning. Del 2. Uppsala: Institutionen för nordiska språk vid Uppsala-universitet.

Margrét Jónsdóttir

From accusative to dative (via nominative): The case of fjölga ‘increase’ and fækka ‘decrease’ in Icelandic1 Abstract: In the oldest written sources, the transitive verbs fjölga and fækka were taking accusative as a direct theme object. Very early, accusative object was replaced by the dative case which is still in use. As intransitives, fjölga and fækka were what can be called accusative subject verbs. It means that the subject, which was theme, was in accusative and the finite verb in singular. This use is only found in texts from the sixteenth and seventeenth centuries. Next in line was the nominative subject with the plural form of the verb. The oldest examples are from the seventeenth century; it seems to have replaced the accusative one and was in use until the beginning of the twentieth century. The third step in line is the dative subject with the verb in singular. The oldest sources are from the ­eighteenth century but it was not until the nineteenth century that the dative subject became common; now it is the only one. Furthermore, the use of fjölgast and fækkast, the middle voice forms of fjölga and fækka, is very common in the oldest written sources and until around 1700. Then it mostly disappeared and after that only sporadic examples have been found.

1 Introduction Icelandic has a number of impersonal verbs taking either accusative or dative subjects, with some variation among speakers with respect to the case marking. First, for some speakers, a number of verbs which originally took an accusative subject now takes a dative subject. In Icelandic grammar, this phenomenon is called Dative Sickness or Dative Substitution.2 Secondly, many speakers now use

1 I am very grateful to Höskuldur Thráinsson for his endless help and great support. Also, to three anonymous reviewers for their extensive comments. 2 This topic has been much discussed in the literature and articles are in legions. Here, I only mention Thráinsson (2007) and references cited there, Jónsson and Eythórsson (2005), and Barðdal (2011a).

182 

 Margrét Jónsdóttir

nominative subjects with certain verbs which earlier took oblique subjects. These variations are illustrated in (1). (1) a.

Mig/Mér vantar peninga. me.Acc/Dat needs money.Acc ‘I need money.’ Thráinsson (2007: 224)

b. Bátnum/Báturinn hvolfdi á flóanum. boat.the.Dat/Nom capsized on bay.the ‘The boat capsized on the bay.’ Jónsson and Eythórsson (2005: 225)

Since non-nominative subjects have been lost in Mainland Scandinavian3 and in English, it is standardly believed that the nominative subject case is the natural endpoint in this kind of development, i. e. it is expected that structural case will eventually win out over thematic (or semantically motivated) case.4 In this study, however, I describe an instance of the opposite development of subject case. The subject case for the verbs fjölga ‘increase (in number)’ and fækka ‘decrease (in number)’ has changed in the history of Icelandic from accusative to dative, with nominative as an intermediate stage. For both verbs, the changes occurred during the same period and in the same way. The main question concerns the change in subject case. This will be dealt with theoretically and in the light of other changes occurring in Icelandic. The verbs fjölga and fækka are suffixed (historically), fjölga with -ga and fækka with -kka. Both verbs denote changes of state, and their meanings can be considered opposites. Since the difference in meaning could be described as plus/minus values of the same feature, it could be argued that they are two sides of the same coin. Both changes occurred at more or less the same time. In this study, therefore, exemplification is sometimes based on just one of the two verbs. In the history of Icelandic, it has not been uncommon for causative verbs to change the case they assign to their object; for these verbs, the dative has replaced the accusative. But at the same time, more or less the same happened for their anticausative counterparts, i. e. theme-subject verbs.5 The opposition

3 Oblique subject cases are still preserved in Faroese, cf. Barnes (1986). 4 For further discussion, see among others Allen 1995; Falk 1997, and Jónsson and Eythórsson 2005. It is worth noting that Barðdal (2009, 2011b) was the first to argue against this structural– lexical case dichotomy. Her argumentation is that case marking of verbal arguments in Icelandic is lexically based. 5 The relationship between the causatives fjölga and fækka and the corresponding anticausatives is complex. The semantic properties of the verbs themselves are clear, however, as well as the arguments they take, cf. the subject of the anticausative and the object of the causative verb. Most importantly, according to Zaenen and Maling (1984), the subject with an unaccusative verb

From accusative to dative (via nominative) 

 183

c­ ausative ~ anticausative is illustrated in the following pair where (2a) is a transitive sentence with a nominative subject and an object in the dative; in (2b), the subject is in the dative. (2) a.

Þeir fjölguðu varðmönnunum. they increased guards.the.Dat ‘They increased the watchmen.’

b. Varðmönnunum fjölgaði. guards.the.Dat increased ‘The watchmen became more numerous.’

Icelandic has a number of verbs suffixed with -k(k)a (and with -ga) like grynnka ‘make shallow’, lækka ‘lower, decrease, reduce’, minnka ‘make smaller, reduce’ and many others. As causatives, these verbs govern either the accusative or the dative. Most of them can be used as anticausatives and some take an oblique subject. This possibly sheds an interesting light on the relationship between subject case, object case and thematic roles. Furthermore, -st (passive semantics) can be attached to some of these verbs, such as k(k)a-verbs. We owe some of them to the oldest written sources of the language. From later written sources, it could be discerned that this type of verb mostly disappeared in the middle of the seventeenth century. Furthermore, in later sources, some of a rather informal nature, many examples can be found; their use however is not accepted in grammar books. In section 4, k(k)a-verbs with -st will be discussed, their function is more or less the same as those without the -st. This paper will be organized as follows. In section 2, there is a brief dis­ cussion of the case marking on the objects of fjölga and fækka. In section 3, I will discuss the diachronic development of fjölga and fækka as anticausatives, first with nominative subjects (section 3.1), then with accusative subjects (section 3.2), and finally with dative subjects (section 3.3). Finally, in (section 3.4) there is a short summary. In section 4, I discuss the st-forms of fjölga and fækka and other verbs with the same formation. The results are given in Table 2 in section 5. Section 6 concludes the paper.

is not an underlying object. Furthermore, they say that unaccusative verbs that do not preserve their case are not related to transitive verbs. Accordingly, there must be a connection between the causative fjölga and fækka and the corresponding anticausative.

184 

 Margrét Jónsdóttir

2 Fjölga and fækka as causative verbs In the oldest written sources, the causative verbs fjölga and fækka assigned accusative case to the direct theme object. In (3) there are three examples of this.6 (3) a.

Margfaldlega mun ég fjölga þitt afkvæmi. (1360–1370) many.times will I increase your progeny.Acc ‘I will increase your progeny greatly.’

b. Fækkum meyjarnar, en fjölgum húsfreyjurnar. (around 1500) decrease.1pl virgins.the.Acc but increase.1pl housewives.the.Acc ‘We should reduce the number of the virgins but increase the number of housewives.’

c. nú hefur ... guð þinn fjölgað þig ... (1584; The Bible) now has god your increased you.Acc ‘Now your God has increased your numbers.’

In the case of fjölga and fækka, dative case replaced accusative object marking in the beginning of the eighteenth century, and this is still in use.7 The oldest example of fækka (4) is one century older than that of fjölga. (4) a. Leigukúgildi ... hafa áður ... verið iiii, en vegna vorkunnar hefur landsdrottinn fækkað einu. (1702–1714) rent.units have earlier been 4, but out of pity has landlord.the decreased one.Dat ‘Rent units have earlier been 4, but out of pity the landlord has decreased them by one.’ b. … fjölgar Helgi vígum. (19f) increases Helgi.Nom slayings.Dat ‘Helgi increases the number of slayings.’ c. Kennarinn fjölgar æfingum en fækkar spurningum. (Modern Icelandic) teacher.the.Nom increases exercises.Dat but decreases questions.Dat ‘The teacher increases the number of exercises but decreases the number of questions.’

6 Citations from the oldest written sources up until 1540 are based on the Dictionary of Old Norse Prose = ONP. Citations from after 1540 are nearly all based on Ritmálssafn Orðabókar Háskóla Íslands = ROH, the Written Language Archive. Referring to the age of the examples, f in indicating an item’s date of attestation means that the item dates from the first third of the century given, m means middle of the century, while s refers to the last third of the century. A few examples from the nineteenth and twentieth centuries are from timarit.is; also a few from the internet, all very young. Note that modern spelling is used in all examples, both old and new. 7 Heusler (1932: 115) pointed out that many verbs taking a direct object in the accusative in related languages already took dative objects in Old Icelandic. An anonymous reviewer pointed out to me that Holland (1993: 29) has argued against this. He thinks that dative objects in Icelandic are perhaps inherited; this he supports by presenting examples with dative objects from related languages.

From accusative to dative (via nominative) 

 185

There are more verbs that originally assigned accusative case to their complements but that now take the dative, i. e. ljúka ‘finish’ (lúka in Old Icelandic), loka ‘close’, tapa ‘lose’ and týna ‘lose’. As mentioned above, it is not uncommon in Icelandic for a causative verb to have changed the case it takes, by replacing accusative with dative, cf. i. e. Maling (2002: 38‒40). She points out that the dative theme case is semantically coherent and therefore a regular lexical case (Maling 2002: 97). This claim contradicts Jónsson (1997‒1998: 20‒21) and his explanation of semantic roles and lexical case of subjects, which will be discussed in section 6. Furthermore, it is important to bear in mind that developments of case change can proceed in different directions. There are examples of transitive verbs that have changed with regard to the case of their complements, i. e. samþykkja ‘agree’, which very often took the dative up to the nineteenth century.8 Many verbs with genitive objects have ceased to assign the genitive to their complements, referring to Jónsson and Eythórsson (2011: 219‒222). Also, as far as the subject case is concerned, fluctuation between accusative and dative is old and can be traced back to Old Icelandic, according to Barðdal (2011a: 76). At the same time, as causative verbs replaced the accusative with the dative, more or less the same happened to the anticausative theme-subject verbs, fjölga and fækka. This is the topic of the next section.

3 Fjölga and fækka as anticausative verbs In Modern Icelandic, the verbs fjölga and fækka occur with dative subjects. In earlier stages of the language, on the other hand, they took nominative and accusative subjects. This variation will be discussed in the following subsections.

3.1 Fjölga and fækka with a nominative subject The oldest examples of fjölga and fækka with nominative subjects are from the middle of the sixteenth century; the oldest secure example of fjölga is about a century older than that of fækka. In contrast to the oblique subjects, it is only the nominatives that trigger number and person agreement with the finite verb.

8 Maling (2002: 87) discusses samþykkja and other sam- verbs. Such verbs with a unity meaning govern the accusative.

186 

 Margrét Jónsdóttir

(5) a. af nokkrum vondum bókum svo sem margar ... fjölga daglega. (1541–1550)9 of some evil books such as many.Nom increase.3PL every.day ‘Of some evil books, which become more numerous every day.’ b. kennilýður fjölgaði mjög. (17f) clergy.Nom increased greatly ‘A clergy increased greatly in number. c. en þeir útlensku hvalfangsmenn fækka þá einna mest. (17fm) but they foreign whalers.Nom decrease.3PL then the most ‘But the foreign whalers then decrease the most.’ d. Síðan hestarnir fjölguðu aftur. (1776) since horses.the.Nom increased.3PL again. ‘Since the horses increased again.’ e. ... þó fénaður fækki upp til sveita. (19m) though sheep.Nom decreased up to countryside.Gen ‘Though sheep decreased in the countryside.’

The nominative case was in use until the beginning of the twentieth century. In modern usage, nominative-theme subjects hardly ever occur with fjölga and fækka.10

3.2 Fjölga and fækka with an accusative subject The examples of fjölga and fækka with theme subjects in the accusative can be found in texts from the sixteenth and seventeenth centuries, as illustrated in (6).11 The earliest secure examples of nominative subjects, cf. (5a), and accusative subjects, cf. (6a), both stem from the sixteenth century, the nominative being slightly older. It is noteworthy that example (6b) is from the same source as example (5b), which shows a nominative subject. The examples are just a few pages apart so usage seems to have been vacillating. The most recent of these examples, example (6d), is from an eighteenth century dictionary. As is well known, oblique subjects

9 This example is from ONP. 10 In the dictionary of the modern language, Íslensk orðabók [Icelandic dictionary] (2002), the verbs fjölga and fækka are given with a nominative subject as a second choice after the dative case subject. This use is out of tune with the present author’s style. Note an example from 1966: (i) bátar þess fækkuðu ‘its boats (Nom.) became fewer’. 11 Recall that the verbs fjölga and fækka are both suffixed. Hence these are counterexamples to the claim made by Jónsson (1997–1998: 32–33) that suffixed Icelandic verbs, e. g. verbs with -st or -na, do not take accusative subjects but rather, take subjects in either dative or nominative.

From accusative to dative (via nominative) 

 187

do not trigger number and person agreement with the finite verb; this is most obvious in the case of plurals.12 (6) a. Og þá ed mennina tók að fjölga á jörðu. (1584; The Bible) and then when people.the.Acc began to multiply on earth ‘And when people began to multiply on the Earth.’ b. við þennan vetur þá fjölgaði nokkuð fátæka. (17f) in this winter then increased somewhat poor.people.Acc ‘In this winter the number of poor people increased somewhat.’

c.

Það er hefndarmerki Drottins þá hina góðu fækkar í landinu. (1696) it is vengeance.sign Lord.the.Gen when the good.PL.Acc decrease in land.the ‘It is a sign of the Lord’s anger when the good became fewer in the land.’

d. stríðsmennina fækkaði (1738) warriors.the.Acc.Pl decreased.3SG ‘The warriors decreased (in number).’

As already noted, the earliest secure example of a nominative subject is slightly older than the corresponding example with an accusative subject. This may well be a coincidence. Note that an example of fjölga, supposedly from the late seventeenth century, could be much older.13 Also, the examples of the accusative theme subjects with fjölga and fækka are rather few. Other verbs with this same construction can be found in Old Icelandic, cf. i. e. Nygaard (1906: 14‒16), Sandal (2011: 39), and Ottósson (2013).14 The accusative theme subjects, and those which are still in use, are semantically coherent, connected to natural phenomena, denoting changes in nature or in the landscape. However, this is not always the case, cf. (7):

12 Otherwise, in Icelandic, oblique subjects behave syntactically very much like nominative subjects; cf. e. g. Zaenen, Maling, and Thráinsson (1985); Sigurðsson (1989: 204‒209), and Jónsson (1996: 110–119). 13 Fritzner (1954) refers to an example of accusative subject of fjölga in the saga of Hrólfs saga kraka (The saga of Hrolf kraki). This example is preserved in a manuscript from 1654 and c1700, cf. ONP, but Hrólfs saga kraka is considered to be older, at least from the middle of the fifteenth century, cf. Evans (1993: 304–305). In her thesis, it should be mentioned, Sandal (2011: 56) uses this example. In a younger edition of the saga, it is however in the nominative, cf. ONP. Furthermore, in Cleasby and Vigfússon (1874), there is an example of the same type from Sturlunga saga (The Saga of the Sturlungs) but this is spurious; cf. ONP. 14 The date 2013 refers to a publication in progress but the reference can be found in the manuscript (on pages 44–45 from 2009), which the article is based on. Furthermore, a few of Ottósson’s examples are in the dative.

188 

 Margrét Jónsdóttir

(7) a. þá reið hann mjög og ætlaði sér varla hóf um og þraut hestinn undir honum. (1275) then rode he much and intended himself hardly moderation about and finished horse. the Acc under him ‘Then he rode faster and showed no moderation and his horse gave up under him.’ b. endir þar sögu frá honum. (c1330–1370)15 ends there saga.Acc from him.Dat ‘There ends this saga about him.’

Verbs with an accusative subject have not been productive in the history of the language, according to Jónsson and Eythórsson (2011: 228).16 Among those accusative theme subject verbs are bera ‘carry’ and reka ‘drift’.17 The nominative has replaced or can replace the accusative, however, with many of the older accusative theme subject verbs. (8) a. Bátana rak að landi. boats.the.Acc drifted to shore ‘The boats drifted to the shore.’ (Eythórsson 2000: 188) b. Bátarnir ráku að landi. boats.the.Nom drifted.3PL to shore ‘The boats drifted to the shore.’ (Eythórsson 2000: 188)

The strong status of the nominative may arguably be the reason why fjölga and fækka became personalized and changed their subject case from accusative to nominative. Other conditioning factors might also be envisaged. This will be discussed further in section 6. 3.3 Fjölga and fækka with a dative subject Both fjölga and fækka can govern dative case on a theme-subject. The oldest clear examples are from the beginning of the eighteenth century. Dative is now the only possible case, but it was not until the nineteenth century that the dative subject became common.18

15 The sentence which is from Laxdæla saga – The saga of Laxdæla – is taken from ONP. The example is also found in Sandal (2011: 43) but with a different conjugation of the verb. 16 Furthermore, here, it is appropriate to refer to Viðarsson’s list (2009: 41–42) of accusative theme subject verbs in Old Icelandic. 17 The use of these verbs is what Sigurðsson (2006: 26–28) calls the FATE accusative construction but Barðdal (2011a) talks about verbs of verbs denoting Landscape and Nature. See also Wood (2012: 120–121) and his references. 18 Note, according to Jónsson & Eythórsson (2005: 225), that Dative Substitution, cf. (1), first became widespread around the middle of the nineteenth century; they refer to Halldórsson (1982:

From accusative to dative (via nominative) 

 189

(9) a. Heilögum þínum fækka fer ... (1726)19 holy.Dat your.Dat begin to.decrease ‘Your holy ones begin to decrease.’ b. sauðpeningi hefir fjölgað hér norðanlands. (1859) sheep.Dat have.3SG increased here in.the.North. ‘Sheep have increased here in the North.’ c. Þeim mönnum fjölgaði enn við kosningarnar. (1885) those.Dat men.Dat increased still at elections.the ‘Those men became still more numerous with the elections.’ d. Fólki á vinnumarkaði fækkar og öldruðum fjölgar. (2002) people.Dat on labour.market decrease and elderly.the.Dat increase. ‘People on the labour market have grown less numerous and the elderly more numerous.’

It is noteworthy that the nominative subject and dative subject were in use at the same time. There are even examples of both cases occurring in the same sentence. (10) Nú eru nokkrar líkur til að útlendum ferðamönnum fjölgi hér framvegis, þótt það verði ef til vill ekki enskir ferðamenn sem fjölga. (1892) now there is some probability that foreign travellers.Dat increase here henceforth though it may not perhaps be English travellers.Nom that increase. ‘Now, there is some probability that foreign travellers may increase from now, though it may not be English travellers that increase.’

The reason for this use is not clear. On the one hand there is the change of the subject case from accusative to dative with nominative as a possible intermediate step, temporarily. On the other hand, the reason could be semantic and reflect the status of the theme subject verbs that either have dative subjects or nominative subjects. Many theme verbs, however, show the same behaviour, cf. Eythórsson (2000: 188, 2002: 200). Among them are hvolfa ‘capsize’, cf. (1b), repeated here as (11a) for convenience, and lykta ‘end’. (11) a. Bátnum/Báturinn hvolfdi á flóanum. boat.the.Dat/Nom capsized on bay.the ‘The boat capsized on the bay.’ (Jónsson and Eythórsson 2005: 225) b. Leiknum/Leikurinn lyktaði með ósigri okkar manna. game.the.Dat/Nom ended with defeat our.Gen men.Gen ‘The game ended in defeat for our team.’

This problem will be discussed further in section 6.

181) who claims that examples could be found in Old Icelandic too. For other changes in the nineteenth century case system, see Barðdal (2011a) and Viðarsson (2009). 19 In ROH, this sentence is found under fjölga.

190 

 Margrét Jónsdóttir

3.4 Conclusions The examples of fjölga and fækka with theme subjects in the accusative can be found in texts from the sixteenth and seventeenth centuries, as illustrated in (6).20 The earliest examples of a nominative subject, cf. (5a), and an accusative subject, cf. (6a), both stem from the sixteenth century, the nominative being slightly older. Due to the paucity of accusative examples, which indicates that the con­struction is a historical residue, this may well be accidental and does not warrant any far-reaching conclusions. However, as was discussed in 3.2, cf. (7), there are many examples of other verbs from Old Icelandic that instantiate this construction. As a matter of fact, the construction occurs in many other languages, not only Germanic, cf. Eythórsson (2002: 204‒207) and his references, Ottósson (2013) referring to Hermodsson (1952: 81) and Smirnickaja (1972: 87‒88), but, according to Barðdal (2013: 30‒31), also from Russian, Lithuanian and Latin as well as from Bavarian German.

4 st-forms of fjölga and fækka and other verbs In the history of Icelandic, there are many examples of the fjölgast and fækkast.21 In subsection 4.1, the use of these forms will be discussed. Other verbs can be used in the same way; they will be discussed and compared to fjölgast and fækkast in 4.2.

4.1 Fjölgast and fækkast The use of fjölgast and fækkast is very common in the oldest written sources, cf. ONP. Examples can also be found in ROH; there is only one example of fækkast, cf. (12b). Most examples of fjölgast are from before 1700 but there are also three more examples, one from each century, the eighteenth, nineteenth and twentieth. In other sources, there is only an example from the nineteenth century. The subject

20 Recall that fjölga and fækka are suffixed verbs. Hence these are counterexamples to Jónssons’s (1997–1998: 32–33) claim that suffixed Icelandic verbs do not take accusative subjects, but rather subjects in either the dative or the nominative. 21 It is a matter of definition whether -st is a clitic or a suffix. Ottósson’s (1992: 66, 66–69) arguments for -st as a suffix in Old Icelandic are very convincing. For Modern Icelandic, Wood (2012: 100) argues for a clitic.

From accusative to dative (via nominative) 

 191

is in the nominative and as always, the nominative subject triggers number and person agreement with the finite verb. (12) a. styrktust söfnuðirnir í trúnni og [þeir] fjölguðust daglega að tölu. (1540) strengthened congregations.the.Nom in faith.the and [they.Nom] increased‑ST daily ‘The congregations became firmer in their faith and they increased daily.’ b. Talast og, að dýrgripir af silfri, sem Hólastaður átti, hafi þá fækkast. (17m) it is also said that valuables.Nom of silver which Hólastaður had, had then decreased-ST ‘It is also said, that the valuables of silver which Hólastaður owned had then become fewer.’ c. Hvernig getur fólk í bóndastandinu fjölgast. (18s) how can people.Nom in agricultural.class.the increase-ST ‘How can people in the agricultural class grow in number?’ d. síðan tóku slíkar bækur að fjölgast. (19s) then began such books.Nom to increase-ST ‘Then such books began to grow in number?’

In other sources from the twentieth century, there are few examples of fjölgast and more (18) of fækkast, most of them part of the same proverb. From other sources, most of them younger, sporadic examples have been found, more of fjölgast than fækkast. Some examples are in rather informal texts, cf. (13b,c). (13) a. og fækkast þannig fyrir manni hetjurnar. (1986) and decrease-ST thus for one.Dat heroes.the.Nom.PL ‘Thus one´s heroes decrease in number.’ b.

síðan fór hópurinn að fækkast … (2006) later went group.the.Nom to decrease-ST ‘Later the group diminished.’

c.

Gæsin … er búin að fjölgast mikið. (around 2000) goose.the.Nom is finished increased-ST much ‘The geese have greatly increased in number.’

d.

Á sama tíma hefur nýburum með lága fæðingarþyngd fjölgast. (2005) at same time has neonates.DatPL with low birthweight.Acc increase-ST ‘At the same time neonates with a low birth weight have increased in number. ‘

The oldest examples of nominative theme subjects are from the middle of the sixteenth century; cf. the examples in (5). The middle forms mostly disappeared at the same time. Note that their function, with or without -st, is more or less the same, as can be seen in (14), where example (12a) is repeated as (14a) and (5a) as (14b); the last two examples are from the same source and the same page. There was no discernible difference in usage and function in the late sixteenth century. Gradually the accusative usage disappeared, cf. (6).

192 

 Margrét Jónsdóttir

(14) a. styrktust söfnuðirnir í trúnni og [þeir] fjölguðust daglega að tölu. (1540) strengthened congregations.the.Nom in faith.the and [they.Nom] increased‑ST daily ‘The congregations became firmer in their faith and they increased daily.’ b.

af nokkrum vondum bókum svo sem margar ... fjölga daglega. (1541–1550) of some evil books such as many.Nom increase.3PL every.day ‘Of some evil books, which become more numerous every day.’

c. eftir því, sem fénaður manna fjölgaðist … (1930) as sheep.Nom people.Gen increased-ST ‘As people’s sheep increased’ d. eftir því sem fólki fjölgaði í landinu, … (1930) as people.Dat began to multiply in country.the ‘As people began to grow in number in the country.’

In a very late seventeenth century’s manuscript of the saga of Eyrbyggja, there is an example with fækkast with the subject in dative but the object in the nominative.22 (15) honum fækkuðust skotvopnin. him.Dat decreased-ST firearms.the.Nom ‘He was running out of firearms.’

Notice that the verb is in the plural and agrees with the plural object. Constructions with the same kind of plural agreement are well known in Modern Ice­ landic.23 In Icelandic, examples with fjölgast and fækkast mostly disappeared in the middle of the seventeenth century, cf. (12b), but after that time, some examples have been found, some in recent texts, cf. (13). The “disappearance” of the st-forms fjölgast and fækkast mostly happened at the same time as the oldest examples of nominative theme subjects were found: the middle of the sixteenth century; cf. (14). Furthermore, note that the function, with or without ‑st, is more or less the same, as can be seen in (14). It is difficult to say whether the sentences in (13) show a new construction, or whether the old one is reappearing, or, indeed, if these constructions have co-existed throughout the entire period. It is however not so important which

22 In this case, it is interesting to refer to Old Swedish. According to Falk (1997: 95–99), there are examples of anticausative constructions with oblique subject case; a so-called s-form, which is comparable to the –st; cf. (15). It is, however, not the rule in Old Swedish. 23 See Sigurðsson (1990–1991: 62) for a discussion of these constructions. Also, note that the sentence, which is from ONP, exists in two other variants, with fækka and eyðast ‘delete’, but in both instances with a dative subject.

From accusative to dative (via nominative) 

 193

answer is the correct one. More important is the fact that usage events such as (13) have not become conventionalized and generally accepted. More verbs, however, exhibit the same behaviour. Examples can be found in various types of texts, cf. (16).

4.2 Some other -k(k)a(st-)verbs At the end of 4.1, it was mentioned that other -k(k)a-verbs with -st exhibit the same behaviour as fjölgast and fækkast, and examples can be found in various types of text. Some of the verbs showing the same behaviour are given in Table 1. In A the causative/ anticausative form is given, the anticausative with -st in B. Table 1: Some -k(k)a-verbs and the same verbs with -st A

B

dýpka ‘deepen, dredge’ grynnka ‘make shallow’ hækka ‘raise’ lækka ‘lower, decrease, reduce’ minnka ‘make smaller, reduce’ seinka ‘delay, slow down’ smækka ‘diminish, reduce’ stækka ‘enlarge’ víkka ‘enlarge, expand, widen’

dýpkast grynnkast hækkast lækkast minnkast seinkast smækkast stækkast víkkast

In (16), there are sentences with some of the st-verbs in Table 1. The oldest is from around 1200 while the youngest is from this century. No difference can be seen in meaning of the verb depending on whether it occurs with or without -st. (16) a. því var höggvit höfuð af Jóhanni at minnkaðist maðurinn. (around 1200) for his was cut head.Acc off Jóhanni.Dat that reduced-ST man.the.Nom ‘Jóhanns’ head was cut off so that he became smaller.’ b. landsmanna góss og vara ... stendur í stað og hækkast ekki að sínu verði. (1615) population.Gen assets.Nom and goods.Nom … stands still and increases-ST not to its price.Dat ‘The population’s assets and goods are stationary and do not increase in price.’

194 

 Margrét Jónsdóttir

c.  Húsið þarf að stækkast og endurbætast meir og minna, svo að það verði aðgöngulegt til íbúðar. (1880–1882) house.the.Nom must be enlarged-ST and restored-ST more and less so that it will be acceptable for habitation.Gen. ‘The house must be enlarged and restored more and less so that it will be acceptable for habitation.’ d. Birgðaflutningar hafa seinkast vegna skorts á bátum og samgöngu­erfiðleika. (1978) transportprovisions.Nom have delayed-ST because lack.Gen on boats.Dat and difficult communications.Gen ‘Transport of provisions has become slower due to lack of boats and difficult communications.’ e. Svo víkkast ljósið, úr tóminu heyri ég tón. (2006) then expands-ST light.the.Nom, from void.the.Dat hear I tone ‘Then the light spreads wider, out of the void I hear a tone.’ f.

Ég er að minnkast, smækkast, hverfa, undursmækkast. (2007) I am to shrink-ST, diminish-ST, disappear, vanish-ST ‘I am shrinking, diminishing, disappearing, vanishing.’

There is no perceptible difference between the anticausatives with or without -st. Compare (17a) to (16f), with minnka(st) og smækka(st), and (17b) to (13b) fjölga(st): (17) a. Ég er að minnka, smækka. I am to shrink, diminish ‘I am shrinking, diminishing.’ b.

Gæsinni … er búið að fjölga mikið. goose.the.Dat has increased much ‘The geese have greatly increased in number.’

In all the sentences in (16), the subject is in the nominative. Among them is seinkast, cf. (16d). Note that, in addition to the dative subject example with fækkast, cf. (15), a few examples of seinkast with a dative subject have been found, too, as well as one example of víkkast, cf. (18c).24 (18) a. honum seinkaðist barneign meður henni. (1360–1370) him.Dat delayed-ST childbirth.Nom with her.Dat ‘Her bearing of a child to him with her drew out.’

24 Two other verbs on the A-list in Table 1 have been found with subject in an oblique case. These are lækka with a dative subject and smækka with subject in accusative, only one of each. One example with síkka has been found; also þurrka (one example) and aumka (many examples) but their syntactic behaviour is different. Also, in modern usage, the verb seinka still takes dative subject but without -st, unlike víkkast.

From accusative to dative (via nominative) 

 195

b.

Seinkaðist honum ferðin ... (1884) delayed-ST him.Dat trip.the.Nom ‘His journey was delayed.’

c.

og leiði sinn hug æ sem skjótast í brott frá þeim sem honum blíðkast. (1360–1370) and turn his.Acc mind.Acc always as soonest from those.Dat who him.Dat appeal-ST ‘And always turn his mind quickly away from those who appear appealing to him.’

d.  þá hefir bæði eldri kynslóðinni víkkast sjóndeildarhringur að aðdáanlega miklum mun. (1919) then has both older generation.the.Dat widened-ST horizon.the.Nom to an admirably high degree ‘Then the horizon of the older generation has widened admirably.’

In modern usage, the verb seinka still occurs with dative subjects, but without ‑st. On the other hand, the verb blíðkast takes nominative subjects.

4.3 Fjölgast, fækkast and ‑na-+st-verbs According to Sigurðsson (1989: 272), -k(k)a-verbs that can be used as anticausa­ tives can not be used with -st-formation. Sigurðsson’s claims are not compatible with the facts presented here. Furthermore, anticausative verbs with the inchoative suffix -na, verbs like brotna, hitna and stirðna and many others, were not supposed to be with -st. The verb sofnast ‘fall asleep, sleep’ is the only exception, according to Sigurðsson (1989: 272). However, many such examples exist. A few are very old, cf. ONP, but most of them are young, starting with the last century; cf. Jónsdóttir (2013): (19) a. Gormur rosknaðist … (1387–1395) Gormur aged-ST ‘Gormur came of age.’ b. En höndin stirðnaðist á miðri leið. (1947) but hand.the stiffened-ST up on middle.Dat way.Dat ‘But the hand stiffened up half way through.’ c. Amma, glasið brotnaðist. (2003) granny, glass.the broke-ST ‘Granny, the glass broke.’

Here before, there were examples of fjölgast (13b) and fjölga (17b), both with the same meaning. They will be repeated here as (20a.,b.) to compare with batna and batnast which show the same difference.

196 

 Margrét Jónsdóttir

(20) a. Gæsin … er búin að fjölgast mikið. goose.the.Nom is finished increased-ST much ‘The geese have greatly increased in number.’ b.

Gæsinni … er búið að fjölga mikið. goose.the.Dat has increased much ‘The geese have greatly increased in number.’

c.

Fjögurra ára tyrkneskur drengur batnaðist ... 4 years.Gen Turkish boy.Nom recovered-ST ‘A 4 year old Turkish boy recovered.’

d.

Fjögurra ára tyrkneskum dreng batnaði ... 4 years.Gen Turkish.Dat boy.Dat recovered ‘A 4 year old Turkish boy recovered.’

It is interesting to compare these examples to German, where middle verbs are anticausative and denote a process of becoming (inchoative meaning); among those are verbs like sich erhöhen ‘become higher’, sich verdunkeln ‘become dark’ and many others, cf. Haspelmath (1987: 18); also sich vermehren and sich redu­ zieren (or their synonyms), both anticausative. Comparing Icelandic and German syntactically and semantically, the verb sich vermehren corresponds to fjölgast, and sich reduzieren corresponds to fækkast. However, it is only semantically that sich vermehren corresponds to fjölga and sich reduzieren to fækka respectively. In all instances, the German verbs have a causative counterpart: erhöhen, ver­ dunkeln, vermehren, and reduzieren.

5 Fjölga(st) and fækka(st): Overview So far, we have traced the changes that occurred regarding the use of cases, illustrating them with the oldest examples attested. These have been collected in Table 2. On the basis of the examples presented, one may well argue that the period from 1600–1700 is the most important time window in this historical overview.25 At this point in time, causatives started taking dative subjects and examples with an accusative subject disappeared. The oldest examples with nominative subjects are attested, but the st-forms have all but disappeared. Finally, the oldest examples with a dative subject are from the first third of the eighteenth century. As repeatedly mentioned, many verbs with accusative theme subjects have switched to the nominative, cf. (8), or the dative, cf. (1a). The B-part of Table 2

25 With the proviso that these examples may not be numerous but they are unambiguous.

From accusative to dative (via nominative) 

 197

Table 2: The sequence of the changes A CAUSATIVES ~ fjölga/fækka Accusative object: Most recent example from 1584, cf. (3c). Dative object: Oldest example from 1702–1714, cf. (4a); still rule. ANTICAUSATIVES ~ fjölga/fækka Nominative subject: Oldest example from 1541–1550, cf. (5a); in frequent use until the 20th century. Accusative subject: Oldest secure example from 1584, cf. (6a), but without a doubt much older. Most recent example from 1738, cf. (6d). Dative subject: Oldest example from 1726, cf. (9a); still rule. ST-FORMS ~ fjölgast/fækkast Old Icelandic: fjölgast/fækkast in frequent use. Examples with st-forms nearly disappear in the middle of the 17th century, cf. (12b), after that some examples have been found but not many. In use, cf. (13). B ANTICAUSATIVES ~ fjölga/fækka SUBJECT IN SUBJECT IN ACCUSATIVE NOMINATIVE (?1584) –1738 1541/1550–20th century

SUBJECT IN DATIVE 1726–

shows this. The final section of this paper will now deal with the theoretical implications.

6 Conclusion The main topic of this article concerns change in case assignment. Now, I will deal with this theoretically and in the light of other changes occurring in Icelandic. In Icelandic, according to Jónsson (1997‒1998: 20–21), lexical case on subjects should be divided into two groups. On the one hand, there are the thematic cases, namely dative subjects on goals and experiencers; these cases are regular lexical cases. On the other hand, there are accusative subjects and dative themes/ patients which are idiosyncratic or quirky cases; in other words irregular lexical cases. These terms are given in Table 3 which is from Eythórsson (2000: 190), based on Jónsson (1997–1998). Table 3: Semantic (thematic) roles and lexical case of subjects SEMANTIC ROLE

ACCUSATIVE

DATIVE

theme goal experiencer

irregular lexical case

irregular lexical case regular lexical case regular lexical case

irregular lexical case

198 

 Margrét Jónsdóttir

Jónsson (1997–1998: 24) indicates that this is common but that idiosyncratic accusative marking on theme subjects gets replaced by the structural nominative case.26 Furthermore, referring to eralier studies, (1997–1998: 28), he points out that many verbs in Icelandic have changed their subject case from nominative or accusative to dative; in all instances, these subjects have been experiencer arguments. While Jónsson’s explanation is mostly semantic, Eythórsson’s view is not only semantically conditioned but structurally motivated as well. He (2000: 187– 190, 194–198, 2002: 196) points out the strong status of the nominative as the most common subject case and many verbs have changed their subject case to nominative; he claims that not only verbs with theme subject in accusative have changed but also verbs taking an experiencer subject in the dative. This is what Eythórsson (2000: 187, 2002: 196) calls Nominative Sickness/Substitution, cf. (1b) as opposed to Dative Sickness/Substitution, cf. (1a). Semantically conditioned are those subjects that are experiencers or goals. Barðdal’s view (2011a) is lexically based. Her main point is type frequency: Verbs with accusative and dative subjects could be divided into two main groups, namely happenstance verbs and experience-based verbs. The dative subjects could be divided into 13 semantic subgroups, but the accusative subjects only into four; verbs with experiencer subjects were found in both cases. In Modern Icelandic, experiencer subject verbs have a much higher type frequency or 76 % than those subject verbs belonging to the happenstance group or 24 %; in Old Icelandic they were equal or 50 %. Therefore, those experiencer subject accusative verbs had easy access to the dative case. At the same time, the accusative subject became increasingly associated with verbs denoting changes in the landscape and in nature as Barðdal (2011a) points out. But how does one explain the change of the theme subject cases of fjölga and fækka? Recall that they have not been used with the accusative subject for a long time; cf. (6d) and Table 2. Recall also the many examples of such verbs taking a dative subject. The unsolved problem concerns the nominative and dative cases: Why did the subject change its case? Why did it become nominative? Why dative? There are some possibilities. In the light of Eythórsson’s view about the strong status of the nominative it may be easy to understand why fjölga and fækka became personalized and changed their subject case from accusative to nominative. The dative case subject of fjölga and fækka, a theme, however, is still an unanswered question. From Jónsson’s explanation the dative case should not get any support either: As theme dative (and accusative) are irregular lexical cases,

26 The same happened in related languages, e. g. Faroese; cf. Jónsson and Eythórsson (2011: 233–35).

From accusative to dative (via nominative) 

 199

cf. Table 3. From Barðdal’s point of view, the accusative is rather unlikely, but the dative is not excluded. The relationship between the causatives fjölga and fækka and the corresponding anticausatives is interesting. The semantic properties of the verbs themselves are clear. Referring to Zaenen and Maling (1984), cf. footnote 5, anticausative verbs that do preserve case are related to corresponding anticausative verbs. Accordingly, the connection between the anticausative and causative fjölga and fækka respectively is clear. In the light of this, there could be more approaches to solving the question of the dative case of fjölga and fækka as causatives. Recall that the most recent example of the accusative subject is from 1738, cf. (6d), and the oldest example of the dative subject is from the beginning of the eighteenth century or 1726, cf. (9a). The most recent example of fjölga and fækka as causatives governing the accusative is from 1584, cf. (3c), but the oldest example of the dative object is from the beginning of the eighteenth century or 1702–1714, cf. (4a). This can all be seen in Table 2. This could imply a close connection between the causative verb and its anticausative counterpart. Maybe the causative verb has influenced the anticausative? There seems to be no forthcoming answer to that. Campbell (2013: 282–283) states that if middle forms change or disappear, they will typically be replaced with an impersonal form. As already mentioned, the so-called disappearance of fjölgast and fækkast is closely connected in time to the emergence nominative theme subjects which are not used any more. The use of dative subjects, which became common in the nineteenth century, is still a rule. The question then arises what would happen if the st-construction will become common? Will they coexist, develop with semantic differentiation in meaning or will one disappear? However, no attempt will be made to answer this question here.

References Allen, Cynthia L. (1995): Case Marking and Reanalysis: Grammatical Relations from Old to Early Modern English. Oxford: Oxford University Press. 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: John Benjamins. Barðdal, Jóhanna (2011a): The Rise of Dative Substitution in the History of Icelandic: A Diachronic Construction Grammar Approach. Lingua 121(1): 60–79. Barðdal, Jóhanna (2011b): Lexical vs. Structural Case: A False Dichotomy. Morphology 21(3–4): 619–659. Barðdal, Jóhanna (2013): Valency Classes in Icelandic: Oblique Subjects, Oblique Ambitransitives and the Actional Passive. To appear in Bernard Comrie and Andrej Malchukov (eds.), Valency Classes: A Comparative Handbook.

200 

 Margrét Jónsdóttir

Barnes, Michael (1986): Subject, Nominative and Oblique Case in Faroese. Scripta Islandica 37: 13–46. Campbell, Lyle (2013): Historical linguistics: An introduction. Third edition. Edinburgh: Edinburgh University Press. Cleasby, Richard and Guðbrandur Vigfússon (1874): Icelandic English Dictionary. Based on the ms. collections of the late Richard Cleasby. Enlarged and completed by Gudbrand Vigfusson. Oxford: Clarendon Press. Evans, Jonathan D. M. (1993): Hrólfs saga kraka. In: Phillip Pulsiano and Kirsten Wolf (eds.), Medieval Scandinavia: An Encyclopedia, 304–305. New York and London: Garland. Eythórsson, Thórhallur (2000): Þórhallur Eyþórsson: Fall á fallanda fæti? Um breytingar á frumlagsfalli í íslensku. Íslenskt mál og almenn málfræði 22: 185–204. Eythórsson, Thórhallur (2002): Changes in Subject Case-Marking in Icelandic. In: David W. Lightfoot (ed.), Syntactic Effects of Morphological Change, 196–212. Oxford: Oxford University Press. Falk, Cecilia (1997): Fornsvenska upplevarverb. Lund: Lund University Press. Fritzner, Johan (1954): Ordbog over Det gamle norske Sprog. Nytt uforandret opptrykk av 2. utgave (1883–1896). I-II. Oslo: Tryggve Juul Møller Forlag. Halldórsson, Halldór (1982): Um méranir: Drög að samtímalegri og sögulegri athugun. Íslenskt mál og almenn málfræði 4: 159–189. Haspelmath, Martin (1987): Transitivity Alternations of the Anticausative Type. Köln: Institut für Sprachwissenschaft der Universität zu Köln. Hermodsson, Lars (1952): Reflexive und intransitive Verba im älteren Westgermanischen. Uppsala: Almqvist & Wiksell. Heusler, Andreas (1932): Altisländisches Elementarbuch. Heidelberg: Winter. Holland, Gary (1993): Transitivity, Causativity, and Surface Case in Old Norse. Arkiv för nordisk filologi 108: 19–37. Íslensk orðabók (2002): Ritstjóri: Mörður Árnason. Þriðja útgáfa, aukin og endurbætt. Reykjavík: Edda. Jónsdóttir, Margrét (2013): Inchoative na-verbs with -st. Unpublished paper. Jónsson, Jóhannes Gísli (1996): Clausal Architecture and Case in Icelandic. Doctoral dissertation, University of Massachusetts, Amherst. Jónsson, Jóhannes Gísli (1997–1998): Sagnir með aukafallsfrumlagi. Íslenskt mál og almenn málfræði 19–20: 11–43. Jónsson, Jóhannes Gísli and Thórhallur Eythórsson (2005): Variation in Subject Case Marking in Insular Scandinavian. Nordic Journal of Linguistics 28: 223–245. Jónsson, Jóhannes Gísli and Thórhallur Eythórsson (2011): Structured exceptions and Case Selections in Insular Scandinavian. In: Horst Simon and Heike Wiese (eds.), Expecting the Unexpected. Exceptions in the Grammar, 213–241. Berlin and New York: Mouton de Gruyter. Maling, Joan (2002): Það rignir þágufalli á Íslandi. Verbs with Dative Objects in Icelandic. Íslenskt mál og almenn málfræði 24: 31–105. Nygaard, M[arius] (1906): Norrøn syntax. Kristiania: Aschehoug. ONP = Dictionary of Old Norse Prose: http://dataonp.hum.ku.dk/ (accessed June 4, 2014). Ottósson, Kjartan G. (2013): The Anticausative and Related Categories in the Old Germanic Languages. In: Folke Josephson and Ingmar Söhrman (eds.), Diachronic and Typological Perspectives on Verbs, 329–381. Amsterdam and Philadelphia: Benjamins.

From accusative to dative (via nominative) 

 201

Ottósson, Kjartan G. (1992): The Icelandic Middle Voice: The Morphological and Phonological Development. Doctoral Dissertation, Lund University. ROH = Ritmálssafn Orðabókar Háskólans, the Written Language Archive: http://www. arnastofnun.is/ (accessed June 4, 2014). Sandal, Catrine (2011): Akkusative subjekt og antikausativitet i norrønt. M. A. Thesis, University of Bergen. Sigurðsson, Halldór Ármann (1989): Verbal Syntax and Case in Icelandic. Doctoral dissertation, Lund University. Sigurðsson, Halldór Ármann (1990–1991): Beygingarsamræmi. Íslenskt mál og almenn málfræði 12–13: 31–77. Sigurðsson, Halldór Ármann (2006): The Nom/Acc alternation in Germanic. In: Jutta M. Hartmann and László Molnárfi (eds.), Comparative Studies in Germanic Syntax, 13–50. Philadelphia: John Benjamins. Smirnickaja, Olga (1972): The Impersonal Sentence Patterns in the Edda and in the Sagas. Arkiv för nordisk filologi 87: 56–88. Thráinsson, Höskuldur (2007): The Syntax of Icelandic. Cambridge: Cambridge University Press. Viðarsson, Heimir Freyr (2009): Tilbrigði í fallmörkun aukafallsfrumlaga. Þágufallshneigð í forníslensku? Íslenskt mál og almenn málfræði 31: 15–66. Wood, Jim (2012): Icelandic Morphosyntax and Argument Structure. Doctoral dissertation, New York University. Zaenen, Annie and Joan Maling (1984): Unaccusative, Passive and Quirky Case. Proceedings of WCCFL 3: 317–329. Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson (1985): Case and Grammatical Functions: The Icelandic Passive. Natural Language and Linguistic Theory 3: 441–483.

Section 4: Syntax

Karl Erland Gadelii

A generative interpretation of Diderichsen’s positional grammar1 Abstract: In this paper it is proposed that Diderichsen’s positional grammar (1946) has a generative power similar to that of Chomskyan grammar, mainly because its slots can (i) be inherently empty, (ii) be empty because of movement of their initial inhabitants, (iii) host more than one constituent, and/or (iv) host a constituent which is internally complex. Diderichsen’s positional grammar is tested against a large number of Scandinavian language structures, whereby certain problematic cases are encountered. Instead of making ad hoc adjustments to Diderichsen’s grammar by adding new slots, the forbinderfelt, needed for independent reasons, is exploited. Counterexamples involving the left periphery can then be accounted for, whereas four structures resist analysis: (a) Negative or quantified object preceding nonfinite verb, (b) Nominal object shift, (c)  Quantifier floating, (d) Sentence-final subject. It is observed that these phenomena mostly concern Icelandic, which has a richer mid- and endfield than Continental Scandinavian languages, which were Diderichsen’s main object of study.

1 Introduction This paper reinterprets Paul Diderichsen’s (1946) positional grammar from for Scandinavian languages (in particular Danish) in generative terms, trying to show that Diderichsen’s model has larger explanatory power than what he himself could have possibly foreseen. It also tries to account for various Scandinavian constructions that raise problems for Diderichsen’s model by exploiting a position independently needed in his grammar, the so called forbinderfelt, ‘linking field’. Topological or positional grammar boasts a long tradition in Germanic linguistics, going back to at least Herling’s (1821) topological scheme for German. Zezerijnen and Zwart (2008) describe the history of positional grammar in Ger-

1 This paper comprises a theoretical development of certain issues pertaining to Scandinavian syntax, taking as its point of departure the more general account of Diderichsen’s model in Gadelii (2011).

206 

 Karl Erland Gadelii

manic linguistics, where they point out that the Scandinavian topological tradition has preserved many properties of earlier models, notably Drach (1937), whereas current Dutch and German topological grammars have both diverged from Drach, but are today quite close to each other (see also Gadelii 2011). Different positional schemes for West Germanic and North Germanic are needed because the former languages, English excepted, are (superficially) OV, and allow for extensive scrambling.2 Although Dutch and German also have a highly underspecified fundament ‘foundation’ (see below) and are V2 languages, the topological tradition in these languages highlight the respective positions of the finite verb/complementizer and the nonfinite verb, the two so called “poles” or Satzklammer, indicated by boldface in footnote 2. By contrast, Diderichsen’s model and modernized versions thereof pay close attention to the fundament and the second position of the clause. In a West Germanic sentence model, the area hosting scrambled constituents (cf. footnote 2) needs to be quite rich, and at the same time an articulate endfield is required for constituents which do not scramble. The difference between mono- and multiverbal sentences is of course more notable in Dutch and German than in Scandinavian, as is the difference between main and embedded clauses. In Scandinavian, the intermediate field and the endfield display less complexity and are less varied. Diderichsen first presented his theory at the 8e Rencontre des philologues nordiques in Copenhagen 1935 (see Heltoft and Andersen 1986), published in Diderichsen (1936), whereafter he regularly published books and articles on positional grammar until 1986 (cf. in particular Diderichsen 1966, 1986). The formal traits of Diderichsen’s model have of course been analyzed in generative terms by various authors in the past, two outstanding examples being Teleman (1972) and Bjerre et al. (2008). However, the present paper intends to introduce some ideas not previously brought to the fore. As was mentioned above, in Diderichsen’s original model the two first positions in the main clause are of crucial importance. They are called the fundament ‘foundation’ and the finite verb position. The fundament can in principle host any type of constituent (including operators and other silent elements), whereas

2 I. e. insertion of material between AUX and V in main clauses and between COMP and V in embedded clauses: (i) Ich will morgen die Zeitung im Zug lesen. (German) I will tomorrow the journal in train read ‘I will read the journal on the train tomorrow.’ dass ich morgen die Zeitung im Zug lesen will. (German) (ii) … that I tomorrow the journal in train read will ‘… that I will read the journal on the train tomorrow.’

A generative interpretation of Diderichsen’s positional grammar 

 207

the second position must contain the finite verb (= the V2 parameter). Embedded clauses have a scheme of their own in Diderichsen’s model where there is no foundation, but the clause is introduced by the complementizer. In addition, in Diderichsen’s scheme for the embedded clause the sentential adverbial and the finite verb switch places. In an innovative paper, Christer Platzack (1986) collapses the schemes for the main and the embedded clause into a single scheme, where the finite verb and the complementizer are in complementary distribution in second position (see also den Besten 1981 and Koster 1975). Platzack calls this slot the ‘categorial’ position, since its content informs us about the category of the clause in question. There is no switching of places between the sentential adverbial and the finite verb in embedded clauses, but the verb simply remains in situ under V (see below) in subordinate clauses. Diderichsen’s positional grammar contains seven slots forming three hyperfields. These fields were originally labelled the fundamentfelt ‘foundation field’, neksusfelt ‘nexus field’ and inholdsfelt ‘content field’, but they have subsequently been renamed in various ways by different researchers. I will here stick to the terminology of the Grammar of the Swedish Academy (SAG 1999) and call them the initial field, the midfield and the endfield. The individual slots have also received different labels in the past. Here I will use the following notation, indicated by boldface in the example below: (1) Initial field

Midfield

“Foundation” (F) v/c

n

a

V

Kallei lästev ti tydligen tv Charles read obviously ‘Charles obviously read a book on the train.’

Endfield N

A

en bok på tåget. a book on the train

– … att  Kalle tydligen läste en bok på tåget. that  Charles obviously read a book on the train ‘… that Charles obviously read a book on the train.’

The attractiveness of Diderichsen’s positional scheme lies in its simplicity – it contains seven positions, no more no less, but is nevertheless supposed to be able to analyze any type of phrase in Danish, and by extension in continental Scandinavian. Because of its clarity, the positional scheme is very popular in coursebooks on Scandinavian syntax, but here I would like to show that Diderichsen’s model is not only pedagogically useful, but theoretically interesting to an extent that may not always have been fully appreciated. Many attempts have been made in order to challenge this super-simple model of Scandinavian language structure, an endeavour which, it would seem, has had surprisingly little success. In fact, several leading generative grammarians have

208 

 Karl Erland Gadelii

shown a continued interest in Diderichsen’s model, reinforced by a certain skepticism towards x-bar-theory (this concerns in particular Sigurdsson 2005, but see also Engdahl 2003, 2010; Platzack 1986, 1998, 2011; and Þráinsson 2007). This is all the more interesting since Diderichsen’s scheme is an old-fashioned holistic, top-down model of the phrase, very different from current generative theories. However, I would like to claim that Diderichsen’s positional grammar models Universal Grammar in a highly interesting way, distinct from the cartographic as well as the interface approach. Positional grammar is closer to the cartographic approach than to the interface one, but does not suffer from the abundance of structure which often characterizes the former. Diderichsen positional grammar presents a restricted skeleton of the phrase which can be seen as a hardwired template, against which sentences are matched in analysis and generation. However, it is not a simplistic representational model, but manages to combine a linear outline with derivational facilities. In my opinion, positional grammar presents a sound compromise between the opulent architecture of the cartographic view and the extreme bottom-up approach of interface theory. In the following I will reinterpret Diderichsen’s model in generative terms, thereby sometimes using terms that were unknown at the time of his writing. As pointed out above, I will try to show that Diderichsen’s model has a generative power that he himself could not have possibly foreseen. It would seem that Diderichsen’s model contains various facilities which enable it to handle structures which at first sight seem impossible to analyze within the seven-slot scheme. These facilities are listed below: (2) A slot may be intrinsically empty. (3) A slot may contain the trace of a moved constituent (cf. subject and verb movement in example (1) above). (4) A slot may host more than one constituent of the same type (this concerns a, V, N, A). (5) A slot may host a constituent which triggers indirect recursion.

Diderichsen himself did of course not use terms such as ‘trace’, ‘moved constituent’ or ‘recursion’, but his theory does include empty positions, constituents in non-canonical positions, etc., which can be reformulated in generative terms in the way I have done in (2) to (5) above. Taken together, it would seem that these four facilities endow Diderichsen’s positional grammar with an analytic capacity that goes far beyond simplistic topological models. Facility (2) just means that all seven slots are present, although all of them may not be used at all times. (3) implies that Diderichsen’s grammar is a deriva-

A generative interpretation of Diderichsen’s positional grammar 

 209

tional model allowing for two levels of analysis. (5) simply means that the system is indirectly recursive, which allows it to handle embedded clauses, for example. I will not dwell on these three facilities here, powerful though they may be, but will instead focus on number (4): a slot may host more than one constituent of the same type.

2 Multiply filled slots As already pointed out, a, V, N and A can host more than one constituent, whereas F, v/c or n cannot. We can express this as: (6) F v/c n a* V* N* A*

As to the three first slots, we can make the following observations: (7) The “foundation” (F) cannot host more than one constituent (= one of the two facets of the V2 constraint), but see below. (8) A clause cannot have more than one finite verb (v) or one complementizer (c), complementizers being considered as heads (possibly complex).

(9) The clause cannot have more than one subject (n).3

In connection with (7), it could be noted that doubly filled F has however been observed in Scandinavian languages, e. g. by Lundin Åkesson (2009) and Hultman (2003: 293): (10) F



v/c

n

a

V

N

A

[På puben]a [igår]b ville Mia inte dricka cider ta tb in the pub yesterday wanted Mia not drink cider ‘In the pub yesterday, Mia did not want to drink cider.’

It is possible that the two content adverbials in the “foundation” can be regarded as a single constituent in the sense that it sets a single spatiotemporal frame for the rest of the sentence. It would seem that fronting of the object with an adverbial is less grammatical:

3 Expletive constructions are easily analyzed in the Diderichsen model, but constructions that require more than two subject positions, for example certain structures involving Quantifier floating (see section 4) will present a problem for the positional scheme.

210  (11) F

 Karl Erland Gadelii

v/c

n

a

V

N A

??[Cider]n [igår]a ville Mia inte dricka tn på puben ta cider yesterday wanted Mia not drink in the pub “Cider yesterday, Mia did not want to drink in the pub”

This type of doubly filled “foundation” however seems to be grammatical in German, where it is called scheinbar mehrfache Vorfeldbesetzung (‘apparent multiple filling of the foundation’) (cf. Müller 2005, 2011). Stefan Müller gives examples such as (12) F

v/c

n a V

N A

[Die Kinder]n [nach Stuttgart]a sollst du – bringen tn ta the kids from Stuttgart should you bring “The kids from Stuttgart you should bring” (S Müller, cited by Engel 1970: 81)4

Presumably cases such as these could be analyzed as Remnant VP movement (though Müller does not do so since he works in the HPSG framework), but interestingly, Müller suggests, referring to work by Bjarne Ørsnes, that scrambling is a prerequisite for this construction, which could explain why it is ungrammatical in Scandinavian but grammatical in German. Whenever a certain construction proves incompatible with the positional scheme, one possible solution is to add a slot to the model. This could be done in various ways: Sigurðsson (2005) proposes an extension of the basic template to 13 slots (in order to account for all Germanic languages, including German and English), and Togeby (2003) arrives at the same number of slots in order to describe Danish. It would however seem that such rather dramatic extensions of Diderichsen’s model make it lose a lot of its initial appeal.5 The fact that slots a, V, N and A allow for multiple occupation means that more complex sentences can be fitted into Diderichsen’s original model without extending it. However, multiple occupants of a, V, N and A must be of the same category. We may for example note the presence of several light adverbs in position a: (13) F

v/c

n a*

V N

A

Kallei lästev ti [tydligen] [inte] tv någon bok på tåget. Charles read obviously not any book on the train ‘Charles obviously did not read any book on the train.’

4 According to an anonyous reviewer, this sentence is somewhat marginal and should be attributed a question mark. 5 The discussion reminds one of the cartographic vs. interface approach in minimalism, where the former has been criticized for its excessive adding of functional projections.

A generative interpretation of Diderichsen’s positional grammar 

 211

We expressed this above by using the Kleene star, indicating multiple occurrences of a. In x-bar theory, this configuration would equal stacking of ADVPs to the left of VP: (14) VP→ ADVP VP

Multiple occurrences of content adverbials in A work the same way: (15) F

v/c

n a

V N

A*

Kallei lästev ti tydligen tv en bok [på tåget] [i söndags] Charles read obviously a book on the train at Sunday ‘Charles obviously read a book on the train last Sunday.’

In x-bar theory this would be expressed as right-hand stacking of ADVPs to VP: (16) VP → VP ADVP

Left and right dislocation proper, as in the following examples, would seem to constitute a problem for Diderichsen’s model: (17) Kalle, han är snäll. Charles, he is nice. (18) Han är snäll, Kalle. He is nice, Charles.

Kalle ‘Charles’ in (17) cannot occur in the foundation since, as was just pointed out, this slot cannot be doubly filled, and in (18) Kalle ‘Charles’ cannot be part of a doubly filled A for categorial reasons. One solution to this problem would be to discard these examples on the grounds that they display clause-external extraposition, not relevant for the present purposes. Alternatively, one could adopt the line pursued by the Grammar of the Swedish Academy (1999) and place the dislocated constituents in two additional fields, called förfält and efterfält (prefield and postfield), though this would indeed constitute an extension of the model. Multiple inhabitation of the V slot simply means that the sentence contains several non-finite verbs: (19) Initial field

F

Midfield

v/c n a

Endfield V*

N

A

Kallei har ti tydligen [velat] [läsa] en bok på tåget Charles has obviously wanted read a book on the train ‘Charles has obviously wanted to read a book on the train.’

212 

 Karl Erland Gadelii

In x-bar terms this would be expressed as direct recursion of VP: (20) VP → V VP

Since this is not a case of stacking maximal projections onto another maximal projection, as in the case of ADVP stacking to the left or right of VP, one might instead want to analyze it in terms of clausal complementation, exploiting Diderichsen’s facility (5) above. That is, läsa en bok på tåget ‘read a book on the train’ could be regarded as a complex N, rewritten as follows: (21a) Initial field Midfield

Endfield



V

F

v/c n a

N

A

Kallei har ti tydligen [velat] [läsa en bok på tåget] – Charles has obviously wanted read a book on the train ‘Charles has obviously wanted to read a book on the train.’ (21b) Initial field Midfield

Endfield



V

N A

läsa read

en bok a book

F

v/c n a

– – – – ‘… read a book on the train.’

på tåget on the train

It may be objected that the clausal complement of velat ‘wanted to’ contains quite a few empty slots, but this is a problem common to all cartographic-type approaches to the sentence, rather than a particular flaw in Diderichsen’s model. The final case, multiply filled N, can be exemplified by the double object construction, a phenomenon that has been notoriously difficult to describe in binary x-bar terms. It has been analyzed in terms of small clause structure, scrambling, incorporation, or VP shell (see Larson 1988, 1990). Among those analyses, the scrambling one seems to be particularly easy to implement in Diderichsen’s model. Scrambling in x-bar theory would involve repetition of NP in the following manner: (22) VP → VP NP VP → V NP

Cf. doubly filled N in positional grammar: (23) Initial field Midfield

Endfield



V N*

F

v/c n a

A

Kallei gavv ti tydligen tv [Eva] [en bok] på tåget Charles gave obviously Eva a book on the train ‘Charles obviously gave Eva a book on the train.’

A generative interpretation of Diderichsen’s positional grammar 

 213

The small clause analysis also seems quite compatible with Diderichsen’s model, and would involve indirect recursion in the same way other types of small clause or complement clause do: (24a) Initial field





v/c n

F

Midfield a

Endfield V N

A

Kallei gavv ti tydligen tv [Eva en bok] på tåget Charles gave obviously Eva a book on the train ‘Charles obviously gave Eva a book on the train.’ (24b) Initial field





v/c n

F

– –

Midfield a

Endfield V N

[Eva]6 – – Eva

[en bok] a book

A –

Once again, it may be objected that the small clause layer contains quite a few empty slots, cf. the point made above. The analyses in terms of incorporation and VP-shell would seem to be more difficult to fit into Diderichsen’s positional grammar, in particular the latter, which demands a larger number of positions than provided for by the Diderichsen model. Summing up, the Scandinavian sentence can contain several sentence adverbials, non-finite verbs, objects, and content adverbials. All these four cases of multiple slot-filling in Diderichsen’s positional model correspond to well-known operations in x-bar syntax, involving direct or indirect recursion. Let us now turn to some more complicated cases, where it is less clear whether we are allowed to insert more than one constituent into the same slot.

3 Head adjunction As was noted in the previous section, doubly filled slots either involve direct or indirect recursion. We have already seen that indirect recursion does not raise a problem for the positional scheme, nor does direct recursion in the form of adverbial or object stacking to VP, or iteration of heads (VP → V VP). Since stacking and iteration require that stacked or iterated entities be of the same category, it follows that the resulting structures are equal to doubly filled slots, allowed on certain occasions in Diderichsen’s model.

6 The small clause subject is in n and not F, following Lundin (2009: 146) who actually considers small clauses to lack the F position.

214 

 Karl Erland Gadelii

However, in the following examples we find structures where two items are not of the same type but can still be argued to occur in the same slot, which accordingly cannot be “Kleene-starred”. Instead of introducing additional slots, thus making Diderichsen’s positional scheme less elegant and more ad hoc, we will use its derivational facility (cf. (3) above) and treat the following examples as instances of head adjunction to the verb (cf. also Jørgensen 2000, as well as later work by this author): (25) F

v

n a V

Jags sågv deno ts I saw it ‘I did not see it.’ (26) F

v

inte tv to not

n a V

Ändå sågv hans deno ts yet saw he it ‘Yet he did not see it.’ (27) F

v

N

inte tv not

n a V

A –

N

A

to

–

N

A

tp Då hade han inte tagit framp boken then had he not taken out the book ‘Then he had not taken out the book.’ (28) F

v

n a V

Pelle ta köpt Sedan hade noga then had probably Peter bought ‘Then Peter had probably bought a book in town.’

  (object shift)

  (clitic cluster)

  (particle   movement)

N

A

en bok a book

på stan   (adverbial in town   clitic)

These examples show that Diderichsen’s model can analyze sentences where a non-verbal head has adjoined to the verbal head by head adjunction, a well-established operation in x-bar syntax. The items that can be adjoined to the verb in this way are pronominal arguments, particles and light adverbials. We can thus make the following statement: (29) A verb-related non-verbal head is allowed to adjoin to the verb in the positional scheme, thereby creating a multiply filled slot.

Talking of adverbials, another problem is raised by examples such as the following: (30) F

v

n

a

?

V

N

A

Sen ska de väl kritiskt granska denna fråga i morgon. then will they probably critically examine this question tomorrow ‘Then they will probably critically examine this question tomorrow.’

A generative interpretation of Diderichsen’s positional grammar 

 215

The adverb kritiskt ‘critically’ is a content adverbial for which there is no immediate slot available, and we can neither adjoin it to V (since V is simple but content adverbials are complex), nor create a doubly filled a, since content adverbials and sentential adverbials are categorially incompatible in generative thinking. X-bar syntax would of course adjoin the content adverbial to the left of VP, but since VP as such does not exist in Diderichsen’s model, this option is not available to us.7 One solution would be to follow Hansen (1997: 51–52) who distinguishes between “free adverbials” (in a or A), “central adverbials” (sentential adverbials, in a), and final adverbials (normally temporal and locative adverbials, in A). Manner adverbials such as kritiskt ‘critically’ would then be underspecified as to their adverbial category, which means that they could be inserted under a as well as A.8

4 Various counterexamples Several other counterexamples have been raised over the years against Diderichsen’s positional scheme (cf. notably Jörgensen and Svensson 1987). Below is

7 Of course, the superposed Endfield could be considered equivalent to VP, but letting superposed field structure enter into syntactic operations would be a very radical deviation from Diderichsen’s original model. 8 An anonymous reviewer points out that the following position of the manner adverbial seems to be problematic (my emphasis): (i) F v/c n a* V ? ? A* Så hade judendomen möjligtvis kunnat fullständigt förenas med Jesus so had Jewish-faith possibly been-able-to fully combine with Jesus ‘In that way, Jewish faith could have possibly been fully combined with Jesus.’ As predicted by Hansen (1997), locative and temporal adverbials do not seem to be grammatical in this slot: (ii) F v/c n a* V ? ? A* *Så hade judendomen möjligtvis kunnat i västvärlden förenas med Jesus so had Jewish-faith possibly been-able-to in west-world combine with Jesus (iii) F v/c n a* V ? ? A* *Så hade judendomen möjligtvis kunnat förra seklet förenas med Jesus so had Jewish-faith possibly been-able-to last century combine with Jesus In order to account for (i), I suggest, following Hansen (1997), that manner adverbials, when specified as a, behave like sentential adverbs and can thus adjoin to verbs (cf. (28) in the text). In the example at hand, this means that the two verbs kunnat ‘been able to’ and förenas ‘combine’ multiply occupy V (as two non-finite verbs do in (21) in the text), whereafter the manner adverbial left-adjoins to förenas ‘combine’. This is not an extension of the model, since we are only applying two facilities present for independent reasons, namely the insertion of two non-finite verbs in V and adjunction of a to a verbal head.

216 

 Karl Erland Gadelii

a non-exhaustive list of additional problematic cases, which can not be handled by multiple slot-filling of any kind. The examples chosen represent classical phenomena in Scandinavian syntax which have attracted the attention of researchers in the field (cf. overviews such as Holmberg and Platzack 1995, 2005 and Þráinsson 2007). The list below is necessarily incomplete but includes constructions which seem to constitute serious counterexamples to Diderichsen’s positional scheme. (31) Verb raising in embedded clause (in particular Icelandic):

F v/c n

?

a

V

N

A

– að Jón hefurv ekki tv lesið bókina í gaer. that John has not read the book yesterday ‘... that John has not read the book yesterday.’

No solution is available in terms of multiple slot-filling: n cannot host more than one constituent, and the verb and a are incompatible categorially speaking. Head adjunction of verbs is not possible. (32) Topicalization in embedded clause:

F v/c ?

?

n

a

V

N A

– att kaffe hadev hon aldrig tv gillat t på kvällen that coffee had she never appreciated at night ‘… that coffee she had never appreciated at night.’

The finite verb is surrounded by items of categories which do not allow multiple slot-filling, or whose position is unclear, as in the case of the finite verb. As in the previous example, verbs cannot head-adjoin. The topicalized object is likewise in the vicinity of categorially incompatible items, and cannot head-adjoin to the verb since it is nominal, thus complex. (33) Narrative inversion in embedded clause (Icelandic):

F v/c ?

n

a

V

N A

– að hefðuv Gunnar þá tv komið – að stórum helli. that had Gunnar then arrived at a big cave ‘… that Gunnar had then arrived at a big cave.’ (after Þráinsson 2007: 173)

The verbal head cannot be inserted into v/c because this slot is inherently simple, and furthermore already occupied by the complementizer. It cannot be inserted into n for the same reason as above, nor can it head-adjoin since it is a verb.

A generative interpretation of Diderichsen’s positional grammar 

 217

(34) Stylistic fronting (Icelandic) (cf. Holmberg 2000; see also Ott 2009)



F v/c ?

?

n a V N

A

(Hvern heldur þú) – að stóliðw hafiv tn – tv tw hjólinu – who think you that stolen has the bike ‘Who do you think has stolen the bike ?’

This construction is unanalyzable partly for the same reason as topicalization in embedded clauses (cf. 32). It is supposed that the stylistically fronted element is complex, and therefore cannot adjoin to the finite verb, whose position is unclear anyway. (35) Negative or quantified object preceding nonfinite transitive verb:

F v/c n a

?

V

N A

Jagn skall tn verkligen inget kaffe ha tn idag I will definitely no coffee have today ‘I will definitely not have any coffee today.’

The object is categorially different from a and V, and it cannot adjoin to V since it is not a head. (36) Nominal object shift (Icelandic):

F v/c ?

n a V N A

Jónm lasv bókinan tm ekki tv tn John read the book not ‘John did not read the book.’

–.

Multiple slot-filling is ruled out for the reasons already outline above. In contrast to pronominal object shift, the nominal object cannot adjoin to the finite verb, since maximal projections cannot adjoin to heads. As in the case of (35), (36) illustrates the difficulty of integrating intermediate objects into the positional scheme. (37) Quantifier floating:

F v/c n

a

? V N

A

Då hade studenterna tydligen alla läst samma artikel igår. then had the students obviously all read the same article yesterday ‘Then the students had obviously all read the same article yesterday.’

The quantifier is located in a position in the scheme where no slot exists. It is categorially incompatible with a and V, and it does not seem natural to adjoin it to the verbal head, though this could in principle be possible. In x-bar syntax the quantifier is situated in the VP-internal subject position, a slot not available in Diderichsen’s positional scheme.

218 

 Karl Erland Gadelii

(38) Sentence-final subject (Icelandic, cf. Holmberg and Platzack 1995: chapter 5)

F

v/c

n

a

V

N

A

?

Það munu ? – kaupa þessa bók – margir stúdentar EXPL will but this book many students ‘Many students will buy this book .’ (“There will buy this book many students”)

No subject position is available in sentence-final position, unless the construction is analyzed as right dislocation to the clause, which seems questionable. Holmberg and Platzack (1995: 137) consider this kind of subject to be right-adjoined to VP, an alternative which for reasons developed earlier is not implementable in Diderichsen’s model.

5 A new proposal We have now run through eight constructions which seem incompatible with Diderichsen’s positional scheme in the sense that they can not be analyzed in terms of multiple slot-filling of any kind. One could of course argue that these counterexamples mostly come from Icelandic, and that Diderichsen only aimed at analyzing Mainland Scandinavian. But in fact, four of the eight constructions above also involve continental Scandinavian languages. Does this mean that the positional scheme is not sufficiently complex in order to correctly describe Scandinavian languages, and that the model needs to be extended after all? Instead of considerably increasing the number of slots, as has sometimes been proposed in the past (cf. Sigurðsson 2005 and Togeby 2003, mentioned above), I would like to suggest that the positional scheme be extended from seven to eight slots, no more, and that furthermore, the eighth slot is achieved by exploiting a position already present in Diderichsen (1946) for independent reasons. The position in question is the forbinderfelt (“linking field”), used by Diderichsen to account for conjunctions and prepositions: (39) Forbinderfelt F v/c n a V N

A

Og saa vilde Petra ikke følge Børnene hjem. and so wanted Petra not follow children home ‘And so Petra did not want to walk the children home.’ (Diderichsen (1986: 10)

(40) Forbinderfelt Bestemmerfelt Beskriverfelt Kernefelt Adverbfelt i to smaa Rum in two small rooms ‘in two small rooms at the back of the shop’

Tungtledsfelt

bag butikken – behind the shop

A generative interpretation of Diderichsen’s positional grammar 

 219

I would like to suggest that the preposition be assigned a category of its own, whereas conjunctions and complementizers (the latter called fog ‘linking element’, constituting the only member of the konjunktionalfelt ‘conjunctional field’ in Diderichsen 1946) be collapsed into a single category, C, positioned at the left periphery of the positional scheme. This means that we will have to abandon the economic collapsing of v and c into a common slot v/c. We are thus missing a generalization, but we are in keeping with the current idea in generative grammar that complementizers and verbs are merged in different positions. As will be shown later, this solution will admittedly not be able to deal with constituents in atypical positions in the midfield and endfield, but it will be quite successful in analyzing left periphery phenomena, typical of Scandinavian languages. The eight-slot positional scheme now has the following structure: (41) C F v n a V N A

One reason for collapsing conjunctions and complementizers is that they normally constitute two categories in complementary distribution, used for combining clauses. A caveat is however in order since although sequences COMP + CONJ seem marginal or ungrammatical (cf. *att men ‘that but’, *att och “that and”), the opposite structure is in fact found in what Julien (2009) calls “minimal matrix clauses “ (consider examples such as plus att ‘plus that’, men att ‘but that’). These latter constructions will have to be analyzed as recursive structures (C*, or in x-bar terms CP → C CP, parallel to the VP recursion we saw above (VP → V VP)). This would mean that v remains a single-item slot, whereas C can be complex.9 This slight modification of Diderichsen’s positional scheme, which in fact does not add any new slot to the model but merely exploits a position already present for independent purposes, allows us to handle all counterexamples above which comprise a complex left periphery:

9 It is not entirely satisfactory that C can only iterate in the order CONJ COMP. I have for the moment no solution to this problem, but one possibility, in the spirit of Julien, would be to propose a two-level analysis, where the first level consists of the conjunction only. Consider the example below: (i) Men att detta kaffe hade pojkarna väl druckit igår. but that this coffee had the boys probably drunk yesterday ‘But that this coffee, the boys had probably drunk yesterday.’ F v n a V N A (ii) C

[Men att detta kaffe hade pojkarna väl druckit igår] – – – C F v n a V N A

att detta kaffen hade pojkarna väl druckit tn igår But maybe this would be pushing the positional scheme too far.

–

– – ]

220 

 Karl Erland Gadelii

(42) Verb raising in embedded clause:

C F

v

n a V

N

A

að Jónn hefurv tn ekki tv lesið bókina í gaer. that John has not read the book yesterday ‘... that John has not read the book yesterday.’ (43) Topicalization in embedded clause:

C F

v

n a

V

N A

att kaffen hadev hon aldrig tv gillat tn that coffee had she never appreciated ‘… that coffee she had never appreciated at night.’

på kvällen at night

(44) Narrative inversion in embedded clause:

C

F v

n

a

V

N A

að – hefðuv Gunnar þá tv komið – að stórum helli. that had Gunnar then arrived at a big cave ‘… that Gunnar had then arrived at a big cave.’ (after Þráinsson 2007: 173) (45) Stylistic fronting: C F

v

n a V N

A

(Hvern heldur þú) að stóliðw hafiv tn – tv tw hjólinu – who think you that stolen has the bike ‘Who do you think has stolen the bike ?’

So in fact, these four examples only show that the introduction of an initial C slot creates a more fine-grained left periphery, presumably indispensable for any analysis of Scandinavian languages due to their complexity in this domain. This is in keeping with the cartographic approach in generative grammar, and also, as was just pointed out, with current analyses of Scandinavian languages as to the merging sites of complementizers vs. verbs. It comes as no surprise that the four cases above all involve embedded clauses, where collapsing v and c has more notable effects than in main clauses.

6 Remaining counterexamples and concluding remarks The last four examples, (a) Negative or quantified object preceding nonfinite verb, (b) Nominal object shift, (c) Quantifier floating, (d) Sentence-final subject, remain impossible to analyze in this generative version of Diderichsen’s model, even after introduction of the eight-slot scheme, whereas it is possible to treat them in an articulated version of x-bar-syntax. We may simply conclude that the only thing that this slightly extended model of Diderichsen’s cannot handle is items that appear in atypical positions in the

A generative interpretation of Diderichsen’s positional grammar 

 221

midfield and endfield, where they cannot be stacked or iterated. The objects mentioned in (a) and (b) occur to the left of V, and are thus unanalyzable in the Diderichsen model, where objects cannot occur in the midfield. The subjects in (c) and (d), which respectively occur between a and V and sentence-finally, are equally impossible to analyze. These four cases do not involve stacking or iterating, but adjoining of one maximal projection to another, category-incompatible maximal projection, in principle NP to VP. This is not possible in Diderichsen’s model. It is interesting to note that none of these constructions involve the left periphery, and it may further be observed that these constructions are rare or ungrammatical in Mainland Scandinavian,10 whereas they are all grammatical in Icelandic.11 So once again Diderichsen cannot be blamed for not being able to treat these expressions, since he never set out to describe the particulars of Icelandic syntax. We have to conclude that the midfield in Diderichsens positional grammar remains insufficient when it comes to dealing with phenomena such as intermediate objects and VP-internal subjects, and the endfield in his positional scheme has no place for sentence-final subjects. It would be interesting to test these four remaining exceptions in positional grammars developed for scrambling languages such as German and Dutch where, as we saw in the beginning of this paper, finite and nonfinite verbs constitute two “poles”, surrounded by richer mid- and endfields than what is the case in Diderichsen’s model.

References Bjerre, Tafs, Eva Engels, Henrik Jørgensen, and Sten Vikner (2008): Points of convergence between functional and formal approaches to syntactic analysis. Working Papers in Scandinavian Syntax 82: 131–166.

10 With the possible exception of Quantifier floating, a favourite topic in generative grammar, but whose overall centrality for the theory of grammar remains to be determined. 11 Þráinsson (2007:314) lists no less than five different subject positions in the Icelandic clause, whereas, as we have seen, Diderichsen only offers two, F and n (to which could be added N in the case of postverbal subjects). However, in the Icelandic sentence (38), repeated below for convenience, the direct object þessa bók ‘this book’ already occupies N, which means that margir stúdentar ’many students’ cannot simultaneously occupy N, and thus has to occur farther to the right: v/c n a V N A ? (i) F Það munu ? – kaupa þessa bók – margir stúdentar but this book many students EXPLETIVE will ‘Many students will buy this book .’ (“There will buy this book many students”)

222 

 Karl Erland Gadelii

den Besten, Hans (1981): On the interaction of root transformations and lexical deletive rules. Groninger Arbeiten zur Germanistischen Linguistik 20: 1–78. Diderichsen, Paul (1936): Prolegomena til en metodisk dansk Syntax. Forhandlinger paa det ottende nordisk Filologmøde i København, 1935, 41–46. Diderichsen, Paul (1946): Elementær dansk Grammatik, København: Gyldendal. Diderichsen, Paul (1966): Helhed og Struktur: Udvalgte sprogvidenskabelige afhandlinger, København: Gads Forlag. Diderichsen, Paul (1986): Dansk Saetningsanalyse. Dens Formaal og Metode. In: Lars Heltoft and John E. Andersen (eds.), Sætningsskemaet og dets stilling – 50 år efter, 7–17. København: Akademisk Forlag. Drach, Erich (1937): Grundgedanken der deutschen Satzlehre. Frankfurt a. M.: Diesterweg. Engdahl, Elisabet (2003): Med fokus på subjektet. In: Lars-Olof Delsing, Cecilia Falk, Gunlög Josefsson, and Halldór Á. Sigurðsson (eds.), Grammatik i fokus: Festskrift till Christer Platzack den 18 november 2003, Volume 1, 95–105. Department of Nordic languages, Lund University. Engdahl, Elisabet (2010): Vad händer med subjektstvånget? Om det-inledda satser utan subjekt. Språk och stil. Tidskrift för svensk språkforskning 20: 81–104. Engel, Ulrich (1970): Regeln zur Wortstellung. In: Ulrich Engel (ed.), Forschungsberichte des Instituts für deutsche Sprache Mannheim 5, 1–148. Tübingen: Narr. Gadelii, Karl Erland (2011): Le schéma positionnel dans l‘analyse des langues scandinaves. Habilitation thesis, Etudes nordiques, Université Paris–Sorbonne (Paris IV). Hansen, Erik (1997): Dæmonernes port. Støttemateriale til undervisningen i nydansk, København: H. Reitzel. Heltoft, Lars and John E. Andersen (eds.) (1986): Sætningsskemaet og dets stilling – 50 år efter. København: Akademisk forlag. Herling, Simon Heinrich Adolf (1821): Über die Topik der deutschen Sprache, 296–362. Frankfurt a. M.: Varrentrapp. Holmberg, Anders (2000): Scandinavian stylistic fronting: How any category can become an expletive. Linguistic Inquiry 31, 445–483. Holmberg, Anders and Christer Platzack (1995): The role of inflection in Scandinavian syntax. Oxford and New York: Oxford University Press. Holmberg, Anders and Christer Platzack (2005): The Scandinavian languages. In: Guglielmo Cinque (ed.), The comparative syntax handbook, 420–458. Oxford and New York: Oxford University Press. Hultman, Tor (2003): Svenska Akademiens språklära. Stockholm: Svenska Akademien. Julien, Marit (2009): Plus(s) at(t) i skandinaviska – en minimal matris. Språk och stil 19: 124–141. Jørgensen, Henrik (2000): Indføring i dansk syntax. Scandinavian Department, University of Aarhus. www.hum.au.dk/engelsk/engsv/papers/joer00b.pdf (accessed June 4, 2014). Jörgensen, Nils and Jan Svensson (1987): Nusvensk grammatik, Malmö: Liber. Koster, Jan (1975): Dutch as an SOV language. Linguistic Analysis 1: 111–136. Larson, Richard (1988): On the double object construction. Linguistic Inquiry 19: 39–91. Larson, Richard (1990): Double objects revisited: A reply to Jackendoff. Linguistic Inquiry 21: 589–632. Lundin Åkesson, Katarina (2009): Tala om språk: Grammatik för lärarstuderande. Lund: Student­litteratur.

A generative interpretation of Diderichsen’s positional grammar 

 223

Müller, Stefan (2005): Zur Analyse der scheinbar mehrfachen Vorfeldbesetzung. Linguistische Berichte 203: 297–330. Müller, Stefan (2011): Zur Syntax und Informationsstruktur der scheinbar mehrfachen Vorfeldbesetzung im Deutschen. Talk given at Colloque “Satzanfang”, Sorbonne University, Paris, January 27–29, 2011. Ott, Dennis (2009): Stylistic fronting as remnant movement. Working Papers in Scandinavian Syntax 37: 141–178. Platzack, Christer (1986): Diderichsens positionsschema och generativ transformationsgrammatik. In: Lars Heltoft and John E. Andersen (eds.), Sætningsskemaet og dets stilling – 50 år efter, 161–170. København: Akademisk Forlag. Platzack, Christer (1998): Svenskans inre grammatik – det minimalistiska programmet. En introduktion till modern generativ grammatik, Lund: Studentlitteratur. Platzack, Christer (2011): Den fantastiska grammatiken. En minimalistisk beskrivning av svenskan. Stockholm: Norstedts. Svenska Akademiens Grammatik (SAG). Volume 1–4 (1999). Ulf Teleman, Staffan Hellberg, and Erik Andersson (eds.). Stockholm: Svenska Akademien. Sigurðsson, Halldór Ármann (2005): Om ordföljd i germanska språk: en liten studie i språkvariation. Manuscript, Lund University. Teleman, Ulf (1972): Om Paul Diderichsens syntaktiska modell. In: Ulf Teleman (ed.), Tre uppsatser om grammatik, 33–57. Lund: Studentlitteratur. Togeby, Ole (2003): Fungerer denne saetning? København: Gads Forlag. Þráinsson, Höskuldur (2007): The syntax of Icelandic, Cambridge: Cambridge University Press. Zwezerijnen, Linda and Jan-Wouter Zwart (2008): Comparative topology: the fields analysis of sentence structure in the Danish and Dutch linguistic traditions. Talk given at TABU-dag, Groningen, December 6, 2008.

Henrik Rosenkvist

Evidence for a syntactic Parameter at work in Övdalian Abstract: Within the Scandinavian languages, there is a notable variation regarding verb agreement. Holmberg and Platzack (1995) suggested that this basic feature is linked to both verb raising and the presence of a handful of syntactic phenomena, such as stylistic fronting, oblique subjects, transitive expletives etc. In subsequent works, addressing dialect syntax as well as variation within the standard languages (Julien 2007; Bentzen 2009; Wiklund et al. 2009; Heycock et al. 2010, 2011 etc.), it has been shown that the correlation between agreement, verb raising and for instance oblique subjects is not as straightforward as was proposed by Holmberg and Platzack (1995) and others (cf. Roberts and Holmberg 2010; Holmberg 2010). In this paper, I argue that null referential subjects in Övdalian require both distinct verb agreement and verb raising, which indicates that these linguistic features are related to each other, possibly through the setting of a Parameter. Thereby the gist of the analyses presented by Holmberg and Platzack (1995) is supported.

1 Introduction1 In Chomsky (1981), it was suggested that universal Principles restrict human language (making all natural languages similar at a certain level of abstraction), while the specific setting of Parameters vary across languages. The Parameters were seen as language-internal settings which govern surface word order, and since the number of Parameters must be limited, it was predicted that the setting of one single Parameter would result in a cluster of syntactic features (a detailed discussion of parameters in generative syntax is provided in Roberts and Holmberg (2010: 1–57), while Angantýsson (2011) presents a theoretically oriented study of the Scandinavian languages from this perspective). In an influential work, Holmberg and Platzack (1995) proposed that the verb morphology of Icelandic, Faroese and Övdalian makes a number of syntactic con-

1 Previous versions of this paper have been commented by Piotr Garbacz, Christer Platzack, Lars Steensland and an anonymous reviewer. I am grateful for their constructive critique; only I can be held responsible for remaining errors and shortcomings.

Evidence for a syntactic Parameter at work in Övdalian 

 225

structions in these languages possible (verb raising, expletive null subjects, transitive expletives, etc.). In Swedish, Danish and Norwegian, on the other hand, verbs do not inflect for person and number, and accordingly the specific syntactic constructions are impossible in these languages. The underlying cause for these differences was assumed to be a parameter associated with the syntactic head I (i. e., T). The notion of Parameters has subsequently been seriously challenged on conceptual as well as theory-internal grounds (cf. Newmeyer 2004; Haspelmath 2008; Boeckx and Hornstein 2009 etc.), and the generalizations suggested by Holmberg and Platzack (1995) have also been found to be problematic, as new data from varieties of Scandinavian have become accessible (cf. Julien 2007; Bentzen 2009; Wiklund et al. 2009; Heycock et al. 2010, 2011; Garbacz 2011; Angantýsson 2011 etc.). In short, it has been shown that the variation within the Scandinavian standard languages is greater than was previously presumed, and the same is true for the dialectal variation within Scandinavia. In this study I claim that there is one specific syntactic phenomenon in Scandinavian which does indicate that a syntactic Parameter is at work: Övdalian referential null subjects. Verb inflection, word order and this specific syntactic phenomen all correlate in this particular case, as would be expected if a Para­ meter was involved. In section 2, background to the research context is provided, whereupon referential null subjects in Övdalian are briefly introduced in section 3. The high-NegP, which causes difficulties in observing verb raising in Övdalian, is discussed in section 4. Section 5 shows that there is a way to solve this problem, and section 6 contains a brief concluding discussion.

2 Background In the wake of seminal works such as Falk (1993), Holmberg and Platzack (1995), and Vikner (1995), the relation between the position of the finite verb and a handful of other syntactic phenomena has been a recurrent topic of research within Scandinavian syntax. Specific syntactic constructions, such as null expletive subjects, transitive expletives and stylistic fronting, were in these works assumed to be strongly linked to verb raising in embedded clauses, which in turn was caused by “strong” or “rich” verb morphology (cf. Angantýsson 2011 for an overview of this research). However, while these conjectures initially seemed to hold relatively well for the Scandinavian standard languages, it has been shown that Scandinavian dialects vary considerably in this respect. During the last decade, there has been a growing interest in Scandinavian dialect syntax, and ongoing research

226 

 Henrik Rosenkvist

suggests that the syntactic features of various dialects do not support strong generalizations regarding agreement, word order in embedded clauses and other phenomena. There is also considerable word order variation within the standard languages when one looks at different types of embedded clauses, different types of clause adverbials, and different types of subjects. Current research thus indicates that there is no straightforward connection between verb raising, verb morphology and the syntactic constructions that were mentioned above, contra for instance Holmberg and Platzack (1995). In turn, this might indicate that there are no underlying macro-parameters in syntax which, when turned on or off, simultaneously influence different parts of syntax that on the linguistic surface seem to be unrelated (cf. Newmeyer 2004; Haspelmath 2008). Acknowledging the problems with the hypotheses presented in for instance Holmberg and Platzack (1995), Holmberg (2010) incorporates recent theoretical developments and empirical findings in a new version of a parameter-based approach to the differences between Scandinavian languages, maintaining the idea that the syntactic differences can be attributed to “richness” of inflection and arguing, consequently, that Holmberg and Platzack (1995) were actually on the right track: What I will argue is that we were basically right, descriptively, in that most (though not all) of these differences are due to a parameter to do with the features of I. Later empirical findings and theoretical developments do not justify abandonment of that theory, only a refinement of it. The conclusion is that there are ‘deep parameters’, and furthermore, I will argue that this is perfectly consistent with minimalist theorizing. (Holmberg 2010: 3)

Holmberg (2010: 13) suggests that there are six syntactic features that are directly related to agreement differences in the Scandinavian languages. However, it is shown by Garbacz (2011) that when Holmberg’s predictions are tested in northern Dalecarlia, an area where several agreement-rich vernaculars are spoken (one of them is Övdalian), the predictions are not borne out. Garbacz (2011: 117) shows that there are no null expletives, no null impersonal subjects, no right-dislocated heavy subjects, no oblique subjects and no stylistic fronting in this region. Interestingly, in some places transitive expletives do occur, but, contrary to what would be expected, transitive expletives are possible also in the only variety which lacks “rich” agreement, i. e. the vernacular of Venjan, while it is missing in Övdalian. It can thus be concluded that in spite of Holmberg’s recent revision of the Parameter-related rich agreement hypothesis, new data from Dalecarlia present additional problems which cannot be ignored. In this study I will nevertheless argue that there is at least one Scandinavian syntactic phenomenon that without any exception is intertwined with “rich” verb morphology and verb raising in embedded clauses: referential null subjects in

Evidence for a syntactic Parameter at work in Övdalian 

 227

Övdalian (cf. Rosenkvist 1994, 2009, 2010; Garbacz 2010). This implies that in this language, verb raising yields particular syntactic effects (cf. the discussion about verb movement in the minimalist program in Roberts 2010). Among the Scandinavian languages, Övdalian and the adjacent Våmhus-variety are unique, since referential null subjects do not occur anywhere else, and for this reason it is only possible to attest and test this correlation in Älvdalen and Våmhus. In the following section, I briefly present null subjects in Övdalian.

3 Null wįð (‘we’) and ið (‘you’ plural) in Övdalian2 In Övdalian, the pronouns corresponding to we and you (plural) are in general omitted, just as in well-known null subject languages such as Spanish or Turkish. Examples of the phenomenon in Övdalian are given in (1) and (2); omitted pronouns are in bold in the English translations. (1) a. Byddjum i Övdalim. live.1PL in Älvdalen ‘We live in Älvdalen.’

b. Ulið fårå nų. shall.2PL leave now ‘You ought to leave now.’

(2) a. Witið at byddjum i Övdalim. know.2PL that live.1PL in Älvdalen ‘You know that we live in Älvdalen.’

b. Mienum ulið fårå nų. think.1PL shall.2PL leave now ‘We think that you ought to leave now.’

No other pronouns are regularly omitted – not even impersonal or expletive pronouns. As shown in (1) and (2), wi̢ð and ið are in general omitted, in main clauses as well as in (all types of) embedded clauses. The omission of wi̢ð and ið correlates with agreement on the finite verb; the verb forms for 1pl and 2pl are distinct, i. e., these forms may unambiguously serve as a basis for reconstruction of the missing subject – see table 1, where Övdalian, Icelandic and Faroese finite verb agreement and personal pronouns are shown. The Övdalian form for 3pl is fur-

2 In Rosenkvist (2010), Övdalian null subjects are discussed in more detail. For an introduction to Övdalian, see Garbacz (2010) or Garbacz and Johannessen (in progress).

228 

 Henrik Rosenkvist

thermore identical either with the infinitival form (as illustrated in table 1) or with the form for the singular – or both. In non-final position the ending -a is deleted due to apocope, a prominent feature of spoken Övdalian which also is manifested in writing. This phonetic process applies to both infinitival verbs and verbs in 3pl. The verb form baita will thus often appear as bait in spoken and written Övdalian, coinciding with the singular form. Table 1: Verb agreement and personal pronouns in Övdalian, Icelandic and Faroese

infinitive ‘to bite’ sg.

pl.

1. 2. 3. 1. 2. 3.

Övdalian

Icelandic

Faroese

baita ig bait du bait an bait (wįð) baitum (ið) baitið dier baita

bíta ég bít þú bítur hann bítur við bítum þið bítið þeir bíta

bíta eg bíti tú bítur hann bítur vit bíta tit bíta teir bíta

Table 1 also shows that there are at least three distinct forms in the Icelandic verb agreement paradigm, but null referential subjects are nevertheless not possible in Icelandic. The agreement patterns that are illustrated in table 1 constitute the basis for dividing the Scandinavian languages in Mainland Scandinavian (Swedish and other non-agreeing languages) and Insular Scandinavian (Icelandic, Faroese and Övdalian), as suggested by Holmberg and Platzack (1995) as well as by Holmberg (2010) – although Holmberg considers Icelandic to be the only Insular Scandinavian language.

4 High negations and obscure verb positions One of the most significant differences between the Mainland Scandinavian languages and the Insular Scandinavian languages is that in the former, finite verbs occur to the right of clause adverbials (such as the negation) in embedded clauses, while the reverse order is the default in e. g. Icelandic (see Heycock et al. 2010 for a detailed study of embedded word order in contemporary Faroese and Angantýsson 2011 for details about Icelandic). These differences are illustrated in (3). (3) a.

Detta är brevet som jag inte har läst. (Swedish) this is letter-def. that I not have read ‘This is the letter that I haven’t read’

Evidence for a syntactic Parameter at work in Övdalian 

 229

b. Þetta er bréfið sem ég hefur ekki lesið. (Icelandic) this is letter-def. that I have not read ‘This is the letter that I haven’t read’

The difference between Icelandic and Swedish has been attributed to verb raising in the embedded clause. The verb moves from a lower position (vP) to a higher position (TP) in Icelandic, across the negation, whereas the verb remains in vP in Swedish. It is thus generally assumed that the negation occupies a fixed position between the lower vP and the higher TP – indeed, the immobility of the negation is a prerequisite for establishing the contrast in (3).3 However, in standard Swedish it is possible to place the negation (inte) directly adjacent to the subordinator in virtually any embedded clause, as illustrated in (4). (4) a.

Jag vet att inte tomten finns. I know that not Santa exists ‘I know that Santa Clause doesn’t exist’

b. Detta är brevet som inte jag har läst. this is letter-def. that not I have read ‘This is the letter that I haven’t read’ c. Vi åker till havet om inte det regnar. we go to sea-def. if not it rains ‘We are going to the sea, if it isn’t raining’

Since the negation occurs between the subordinator and the subject in the sentences in (4), it is in principle impossible to tell whether the finite verb remains in vP or if it has raised to TP. As for Övdalian, Levander (1909: 123) points out that “The word not cannot as in Swedish occur between the subject and the finite verb in embedded clauses; if it is not situated in the beginning of the clause, it must be put after the verb” [my translation]. In the beginning of the 20th century, the Övdalian negation thus occurred either after the finite verb in embedded clauses (as in the Icelandic example in 3 b) or directly after the subordinator, in the high-NegP (Garbacz 2010). However, in a small study of Övdalian word order by Rosenkvist (1994), no less than 80 % of the negations in embedded clauses occurred in the high-NegP. No other adverbials were found in this position. Some examples are provided in (5). (5) a.

… fast int eð ir finwedreð olltiett. although not it is fine-weather always ‘although the weather isn’t always fine’

3 The idea that verbs occupy different positions in different languages and that clause adverbials reveal their position goes back to Emonds (1976) and Pollock (1989).

230 

 Henrik Rosenkvist

b. … um int du kumb. if not you come ‘if you’re not coming’ c. … um int eð war iðer fil if not it was your.2PL fault ‘if it wasn’t your fault’

The survey made by Rosenkvist thus suggests that the default option is to place the negation in high-NegP in contemporary Övdalian. In the longest Övdalian text ever published by a native speaker of Övdalian, Larsson (1985), this placement of the negation is also very frequent, and in the detailed study of Övdalian word order by Garbacz (2010), it is clear that speakers prefer to place the negation in the high-NegP in embedded clauses. The judgements of the sentences in (6) are taken from Garbacz (2010: 228); the acceptability scale goes from 1 to 5, with 5 as the highest grade. (6) a.

Eð ir bar i iss-jär buðn so int Marit andler jätå. (mean score: 4,66) it is only in this-here shop-def. that not Marit buys food-def. ‘It is only in this shop that Marit doesn’t buy food’

b. Eð ir bar i iss-jär buðn so Marit int andler jätå. (mean score: 3,83)

While both of the sentences in (6) are accepted, (6 a), with the negation in highNegP, receives a higher score and should therefore be seen as the unmarked alternative (cf. also Garbacz 2010: 132, 139). It is also shown by (6 b) that embedded clauses without verb raising are possible in modern Övdalian (as amply shown by Garbacz 2010), although embedded clauses with verb raising are equally frequent. Just as in Faroese (cf. Heycock et al. 2010, 2011), word order varies in modern Övdalian embedded clauses. The possibility to place the negation in high-NegP in Övdalian obscures verb raising. Furthermore, this circumstance becomes particularly irksome when there is a referential null subject in the embedded clause, since both of the salient overt constituents, the finite verb and the negation, may occur in different positions. The possible analyses of the sentence in (7), which is quoted from Rosenkvist (1994), are presented in table 2.4 (7) … um int windið brott o̢n. if not throw.2PL away her ‘if you don’t throw it away’

4 Garbacz (2010: 113) claims that there are two possible analyses of sentences such as (7), but Rosenkvist (1994: 22) shows that there are in fact three, as illustrated in table 2.

Evidence for a syntactic Parameter at work in Övdalian 

 231

Table 2: Three possible analyses of embedded clauses with negation and null subject

analysis 1 analysis 2 analysis 3

CP

High-NegP

TP

um um um

int int

windið

NegP

vP

int

brott o̢n windið brott o̢n windið brott o̢n

The analytical alternatives in table 2 are unique for Övdalian (and the Våmhus-variety), since these languages are the only Scandinavian language varieties in which null referential subjects occur. For this reason, negations and adverbials in high-NegP are irrelevant in studies of verb raising in other Scandinavian languages and dialects – the subject in SpecTP will always reveal the position of the negation or the adverbial (see for instance Heycock et al. 2011, where high-NegP is not an issue). In order to investigate whether there is a correlation between verb raising and referential null subjects in Övdalian, it is clear that embedded clauses with a negation do not constitute an operational testing ground. To get a clear view of the position of the finite verb, especially in combination with a null subject, an adverbial which cannot appear in high-NegP is necessary. Garbacz (2010: 113, fn. 123) reports that “I have not yet found an adverbial of this kind”, but in the following section I will demonstrate that there are such adverbials in Övdalian, and that differences between speakers (acceptance of null subjects with the verb (seemingly) in vP or not) should be attributed not to syntactic variation, but to lexical variation.

5 Non-high adverbs and speaker-related lexical differences In order to test the hypothesis that referential null subjects in Övdalian are only possible in an embedded clause if the verb has raised from vP to TP, as originally suggested by Rosenkvist (1994), it is thus necessary to find an Övdalian adverbial that always remains in the middle field and that accordingly can never appear in the high-NegP. When consulting Övdalian speakers, it appears that there are such adverbs, but that there is some variation between the informants as to which adverbs are possible in high-NegP.5 All informants accept the negation (inte), a

5 During the last years, I have had regular sessions with a handful of Övdalians in Lund, but I have also on several occasions done interviews on site, in Älvdalen. The regular contacts with my informants have led to the emergence of an elicitation methodology along the lines of Henry (2005).

232 

 Henrik Rosenkvist

majority accept aldri (‘never’) (or the older variant older) whereas very few accept other adverbials, such as fel (a highly polysemic adverb), sakta (‘actually’), naug (‘probably’), kringgt (‘often’) etc. Garbacz (2010) has also investigated the possibility to place adverbs in the high-NegP, inter alia. He tested the adverbs inte (‘not’), sakta (‘actually’), aldri (‘never’), kringgt (‘often’) and milumað (‘sometimes’) in high-NegP in relative clauses – one of the test sentences is quoted in (8): cf. Garbacz 2010: 125. (8) Eð ir iend buotje̢ so aldri ig har lesið. it is only book-def. that never I have read ‘It is the only book that I have never read’

Although Garbacz reports that he has not found any adverbs that cannot appear in high-NegP (see the quote above), data that contradict this statement are presented in his table 6.4. In this table, it is shown that the adverbs kringgt (‘often’) and milumað (‘sometimes’) cannot appear between the subordinator and a pronominal subject in an embedded clause, and neither can they appear in this position with a DP-subject, if there is an auxiliary in the embedded clause (Garbacz 2010: 124). However, the information presented in the table is an interpretation of the data compiled in the informant studies; the complete set of informant data provided by Garbacz in the appendix (2010: 225–227) gives a slightly different picture. There are 12 informants in the study, and it actually appears that an acceptability hierarchy can be attested in their judgements of various adverbials in high-NegP. According to my own results, all Övdalians accept inte (‘not’) in high-NegP, but one of Garbacz’s informants (informant 12) considers that to be ungrammatical. The second best adverbial in Garbacz’s study is aldri (‘never’), followed by sakta (‘actually’), kringgt (‘often’) and milumað (‘sometimes’) in that order. In Figure 1, the different adverbials are shown together with the informants (1–12) that accepted them. The figure illustrates that Garbacz’s informants actually can be ranked according to their acceptability scores; all informants that accept milumað in high-NegP also accept all other adverbials; all informants that accept kringgt also accept sakta, aldri and inte etc. It is thus possible to see the informants that accept for instance kringgt in high-NegP as a subset of a group of informants that accept inte, aldri, sakta and kringgt. In other words, it seems to be the case that the informants have different lexical categorizations of these adverbials. One informant (informant 12) has no high-NegP-adverbials at all (and for this reason he/she is absent from figure 1), one informant has one single high-NegP-adverbial, inte (informant 6), and so forth. There is only one exception to this pattern – informant 2 accepts kringgt but not sakta.

Evidence for a syntactic Parameter at work in Övdalian 

INTE

ALDRI

1 –11

1 –5, 7–11

SAKTA 3, 5, 7 –8, 10–11

KRINGGT 2–3, 7–8

MILUMAÐ

 233

3, 8

Fig. 1: Acceptability hierarchy in high-NegP in Övdalian

The results from the field work done by Garbacz as well as from my own studies show that there is individual variation among Övdalian informants as to which adverbials are possible in high-NegP, and that there is an implicational acceptability hierarchy (as illustrated in Figure 1) which separates the speakers from each other.6 Let us now recall the hypothesis (dubbed “Rosenkvist’s generalization” by Garbacz 2010:  113) that null subjects require verb raising in Övdalian, and the fact that adverbials may occur in a pre-verbal high-NegP. The informants’ grammaticality judgements in Garbacz (2010) indicate that informants 3 and 8 (see Figure 1) would consider a sentence such as (9) grammatical, whereas informant 6 and 12 would consider it ungrammatical – if the generalization is correct. Informants 3 and 8 would put kringgt in high-NegP and raise the verb to TP, but informants 6 and 12 would not be able to put kringgt in high-NegP and would therefore be forced to assume that the verb remains in vP, which would make (9) ungrammatical. (9) Ittað-jär ir ie̢ buok so kringgt wilum leså. this-here is a book that often will.1PL read ‘This is a book that we will read often’ expected scores for (9): informant 3+8: OK informant 6+12: *

The possible correlation between adverbials in high-NegP, verb raising and referential null subjects is not tested by Garbacz, but the informants that I have consulted fully comply with these conjectures. All of my informants accept inte in

6 It is not clear to me if or how this pattern ties in with cartographic approaches to adverbial hierarchies, such as Cinque (1999) and subsequent works.

234 

 Henrik Rosenkvist

high-NegP (10 a) and they also accept a referential null subject in an embedded clause with inte preceding the verb (10 b). (10) a. b.

Ittað-jär ir ie̢ buok so int ig wil leså. this-here is a book that not I will read ‘This is a book that I will not read’ Ittað-jär ir ie̢ buok so int wilum leså. this-here is a book that not will.1PL read ‘This is a book that we will not read’

Most informants also accept aldri (‘never’) in the same positions as inte in (10 a) and (10 b), but reject all other adverbials, whereas one informant, who is not from Älvdalen but from the neighbouring village of Våmhus, accepts all tested adverbials in both positions. Crucially, I have not encountered any informant who accepts a specific pre-verbal adverbial in an embedded clause with a null subject (as in 10 b) and at the same time rejects sentences with the very same adverbial in high-NegP (as in 10 a). My interpretation of the informants’ responses is that they have slightly different lexical categorizations – some informants have only one high-NegP-adverbial (inte ‘not’), but most informants have two such adverbials: inte and aldri (‘never). One of my informants and two of the informants (3 and 8) in Garbacz’s study consider all adverbials to belong to the high-NegPclass.7 My conclusion is that a positive judgement of a sentence such as (11) by an Övdalian informant cannot be considered to be a counter-argument against the generalization under discussion, unless it can be shown that this informant also rejects sentences with the same adverbial in high-NegP with an overt subject in an embedded clause. (11) Ittað-jär ir ie̢ buok so int/aldri/sakt/kringgt/milumað wilum leså. this-here is a book that not/never/actually/often/sometimes will.1PL read ‘This is a book that we will not/never/actually/often/sometimes read’

The attested acceptability correlation between sentences such as (10 a) and (10 b) underlines that while the Övdalian informants differ in their lexical categorizations, they all consistently apply one and the same syntactic principle: a referential null subject is only possible if the verb has raised to T. In other words: there

7 These informants are from Åsen, Loka and Våmhus, respectively, three villages which are not particularly close to each other. The liberal attitude towards putting anything in high-NegP is thus not a geographically determined feature.

Evidence for a syntactic Parameter at work in Övdalian 

 235

is no syntactic variation. In (12), this generalization is formulated in more formal terms. (12) Every speaker of Övdalian follows generalizations A and B. A. Iff B. Iff

OK[COMP AdvlA subject FV] → OK [COMP AdvlA pro FV] *[COMP AdvlB subject FV] → *[COMP AdvlB pro FV]

The generalization that was proposed by Rosenkvist (1994) is thus corroborated. It can also be concluded that two out of the three possible analyses that are illustrated in Table 2 are untenable. Only analysis 1 can be maintained.

6 Concluding discussion The correlation between word order in embedded clauses and “strong” or “rich” agreement morphology on the finite verb, as formulated by for instance Falk (1993), Holmberg and Platzack 1995, and Vikner (1995), is not as straightforward as was originally supposed. For Icelandic, a language with both person and number agreement (see Table 1), it has been shown that the word order in embedded clauses varies (see Angantýsson 2011), while verb raising seems to be possible in some cases in Faroese, a language with relatively poor agreement (see Heycock et al. 2010, 2011). The relation between verb agreement and verb raising is accordingly more complex than previously thought, and when also embedded V2, stylistic fronting, different types of subject, different types of adverbials and different types of embedded clauses are included in the syntactic puzzle, the picture that emerges is hard to explicate coherently. Paying heed to these problems, Holmberg (2010) suggests a theoretical revision that captures the empirical facts while also retaining the idea that there is an agreement-related parameter which is the underlying cause of several syntactic differences between Insular and Mainland Scandinavian. However, not even his version of the hypothesis can explain the data presented by Garbacz (2011). In this paper, I have argued that there is one Scandinavian syntactic construction that nevertheless requires robust verb raising: null referential subjects in Övdalian. It is argued that the apparent exceptions to this generalization are misleading, since the informants differ in their lexical categorizations. Some of them may for instance put all adverbials in the high-NegP, thereby creating a word order which seems to be a case of null subject with the verb in vP. But since these informants also accept all adverbials in high-NegP in embedded clauses with a pronounced subject, I conclude that all informants follow the same syntactic principle: referential null subjects require verb raising.

236 

 Henrik Rosenkvist

The notion of a syntactic parameter, the settings of which influence several aspects of the syntax simultaneously, has been criticized by Newmeyer (2004) and Haspelmath (2008), among others. Övdalian null subjects constitute an interesting example of how verb agreement seems to play a decisive role for syntax. The forms for 1pl and 2pl are distinct (see table 1), and it is only these forms that license null referential subjects. In this particular case, the agreement seems to be sufficiently rich for this syntactic option, although the other constructions that are predicted to occur by Holmberg (2010) are absent from Övdalian. Broadening the view and including other non-standard Germanic languages in the discussion, such as for instance Bavarian and Frisian (see Rosenkvist 2009, 2010), we find that distinct verb agreement is a prerequisite for referential null subjects in all modern Germanic language varieties in which referential null subjects are attested. This correlation offers a new possibility of restoring the link between verb agreement and a specific syntactic phenomenon within Germanic, which possibly is connected to a Parameter-setting. Verb raising is furthermore a form of head movement. Chomsky (2001: 37–38) claimed that head movement is not a part of narrow syntax, motivating this both with theory-internal technical difficulties and the observation that head movement rarely (if ever) affects the interpretation of the clause, i. e. the LF-interface (see the comprehensive discussion in Roberts 2010, chapters 1 and 4). Accordingly, head movement, including verb raising, is assumed to be a PF-phenomenon – see Platzack (2010) for a more recent version of this idea. I have shown that verb raising is a necessary condition for referential null subjects in Övdalian; if verb raising is interpreted as a pure PF-phenomenon, it follows that also referential null subjects should be a PF-phenomenon – we do not expect PF-conditions for LF-syntax. In recent theorizing about referential null subjects (see the articles in Biberauer et al. 2010 and Sigurðsson 2011), referential null subjects are on the contrary analyzed as a part of core syntax, relating the possibility of null subjects to pronominal features in T. The data that I have presented in this paper can accordingly also be seen as an argument for the hypothesis that verb movement is not (always) just a PF-feature.

References Agantýsson, Ásgrímur (2011): The Syntax of Embedded Clauses in Icelandic and Related Languages. Reykjavík: Hugvísindastofnun Háskóla Íslands. Biberauer, Theresa, Anders Holmberg, Ian Roberts, and Michelle Sheehan (2010): Parametric Variation: Null Subjects in Minimalist Theory. Cambridge: Cambridge University Press. Bentzen, Kristine (2009): Subject positions and their interaction with verb movement. Studia Linguistica 63(3): 1–31.

Evidence for a syntactic Parameter at work in Övdalian 

 237

Boeckx, Cedric and Norbert Hornstein (2009): Approaching universals from below: I-universals in light of a minimalist program for linguistic theory. In: M. Christiansen, C. Collins, and S. Edelman (eds.), Language Universals, 79–99. Oxford and New York: Oxford University Press. Cinque, Guglielmo (1999): Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Chomsky, Noam (1981): Lectures on Government and Binding: The Pisa Lectures. Dordrecht: Foris. Chomsky, Noam (2001): Derivation by Phase. In: Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–54. Cambridge, MA: MIT Press. Emonds, Joseph E. (1976): A Unified Theory of Syntactic Categories. Dordrecht: Foris. Falk, Cecilia (1993): Non-Referential Subjects in the History of Swedish. Lund: Dept. of Scandinavian Languages. Garbacz, Piotr (2010): Word Order in Övdalian. Lund: Centre for Languages and Literature. Garbacz, Piotr (2011): Sambandet mellan morfologi och syntax i de nordiska dialekterna. In: Lars-Erik Edlund, Lennart Elmevik and Maj Reinhammar (eds.), Studier i dialektologi och sociolingvistik, 113–122. Uppsala: Kungl. Gustav Adolfs Akademien för svensk folkkultur. Garbacz, Piotr and Janne Bondi Johannesson (In progress): Övdalian from 1909 to 2009. To appear in: Kristine Bentzen and Henrik Rosenkvist (eds.), Studies in Övdalian Syntax. Amsterdam: John Benjamins. Haspelmath, Martin (2008): Parametric versus functional explanations of syntactic universals. In: Theresa Biberauer (ed.), The limits of syntactic variation, 75–108. Amsterdam: John Benjamins. Henry, Alison (2005): Idiolectal variation and syntactic theory. In: Leonie Cornips and Karen P. Corrigan (eds.), Syntax and Variation, 109–122. Amsterdam: John Benjamins. Heycock, Caroline, Antonella Sorace, and Zakaris Svabo Hansen (2010): V–to–I and V2 in subordinate clauses: an investigation of Faroese in relation to Icelandic and Danish. The Journal of Comparative Germanic Linguistics 13(1): 61–97. Heycock, Caroline, Antonella Sorace, Zakaris Svabo Hansen, Sten Vikner, and Frances Wilson (2011): Residual V–to–I in Faroese and its lack in Danish: detecting the final stages of a syntactic change. WPSS 87: 137–165. Holmberg, Anders and Christer Platzack (1995): The Role of Inflection in Scandinavian Syntax. Oxford: Oxford University Press. Holmberg, Anders (2010): Parameters in minimalist theory: The case of Scandinavian. Theoretical Linguistics 36(1): 1–48. Julien, Marit (2007): Embedded V2 in Norwegian and Swedish. Working Papers in Scandinavian Syntax 80: 103–161. Larsson, Hjalmar (1985): Kunundsin kumb. Älvdalen. Levander, Lars (1909): Älvdalsmålet i Dalarna. Ordböjning ock syntax. Stockholm: Norstedt & Söner. Newmeyer, Frederick (2004.): Against a parameter-setting approach to typological variation. In: Pierre Pica (ed.), Linguistic Variation Yearbook 4, 181–234. Amsterdam: John Benjamins. Platzack, Christer (2010): Head Movement as a Phonological Operation. LingBuzz/001111. Pollock, Jean-Yves (1989): Verb Movement, Universal Grammar and the Structure of IP. Linguistic Inquiry 20: 365–424. Roberts, Ian (2010): Agreement and Head Movement. Cambridge, MA, and London: MIT Press.

238 

 Henrik Rosenkvist

Roberts, Ian and Anders Holmberg (2010): Introduction: parameters in minimalist theory. In: Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan (eds.), Parametric Variation: Null Subjects in Minimalist Theory, 1–57. Cambridge: Cambridge University Press. Rosenkvist, Henrik (1994): Tecken på syntaktisk utveckling i älvdalsmålet under senare tid. Unpublished B. A.-Diss. Lund: Dept. for Scandinavian Languages. Rosenkvist, Henrik (2009): Null referential subjects in Germanic – an overview. Working Papers in Scandinavian Syntax 84: 151–180. Rosenkvist, Henrik (2010): Null Referential Subjects in Övdalian. Nordic Journal of Linguistics 33(3): 231–267. Sigurðsson, Halldór Ármann (2011): Conditions on Argument Drop. Linguistic Inquiry 42(2): 289–308. Wiklund, Anna-Lena, Kristine Bentzen, Gunnar Hrafn Hrafnbjargarson, and Þorbjörg Hróarsdóttir (2009): On the distribution and illocution of V2 in Scandinavian that-clauses. Lingua 119(12): 1914–1938. Vikner, Sten (1995): Verb Movement and Expletive Subjects in the Germanic Languages. New York: Oxford University Press. Vikner, Sten (1997): V°-to-I° movement and inflection for person in all tenses. In: Liliane Haegeman (ed.), New Comparative Syntax, 189–213. London: Longman.

Ida Larsson and Janne Bondi Johannessen*

Embedded word order in Heritage Scandinavian Abstract: This paper investigates the word order in embedded clauses in Heritage Scandinavian (American Norwegian and American Swedish). It is shown that Heritage Scandinavian has a substantial amount of verb-raising across negation in embedded clauses. Verb-raising (i. e. the order verb–negation) is found in thatclauses, relative clauses and embedded questions; only the first allow the order verb–adverb in European Norwegian and Swedish. It is argued that the non-standard order (from a European perspective) should not be viewed as a consequence of direct transfer from English, and not as attrition. Instead, limited or incomplete acquisition of the heritage language leads to a grammar with V-to-I movement in addition to V-to-C movement. The acquisition of the heritage languages is assumed to follow the same path as the acquisition of the European Scandinavian languages, but it is delayed due to the limited input and interrupted when the heritage speakers start school.

1 Introduction1 Between 1820 and 1920 there was a massive emigration to North America. 1 300 000 Swedes and 800 000 Norwegians emigrated; most went to the Midwest and settled as farmers, often in communities where others from the same area of their homeland were residing. Norwegian and Swedish communities therefore kept the language from home for many years, and through several generations.

* Ida Larsson: University of Oslo, Janne Bondi Johannessen: University of Oslo 1 We would like to thank the audience at The 11th International Conference of Nordic and General Linguistics in Freiburg, April 2012, as well as local audiences at the universities of Oslo and Uppsala for good comments. We would also like to thank two reviewers who have helped making the paper much better. We are grateful to André Lynum for helping us to calculate the significance tests in Section 4. The Swedish work (by Ida Larsson) was partly funded by The Letter­stedt Association, The Helge Ax:son Johnson Foundation, and The Torsten Söderberg Foundation. The Norwegian work (by Janne Bondi Johannessen) was partly supported by the Research Council of Norway through its Centres of Excellence funding scheme, project number 223265, and through its funding of the project NorAmDiaSyn, project number 218878, under the BILATGRUNN/FRIHUM scheme. They are hereby gratefully acknowledged.

240 

 Ida Larsson and Janne Bondi Johannessen*

This language was used in public life, in schools and in church, and in many local newspapers. Still today, in 2013, there are descendants of these immigrants that continue to speak Norwegian and Swedish, and who have a Scandinavian language as their first language. We use the term heritage language to refer to the language of the American Scandinavians who have been born in America, following Rothman’s (2009: 156) definition:2 A language qualifies as a heritage language if it is a language spoken at home or otherwise readily available to young children, and crucially this language is not a dominant language of the larger (national) society. Like the acquisition of a primary language in monolingual situations and the acquisition of two or more languages in situations of societal bilingualism/multilingualism, the heritage language is acquired on the basis of an interaction with naturalistic input and whatever in-born linguistic mechanisms are at play in any instance of child language acquisition. Differently, however, there is the possibility that quantitative and qualitative differences in heritage language input and the introduction, influence of the societal majority language, and differences in literacy and formal education can result in what on the surface seems to be arrested development of the heritage language or attrition in adult bilingual knowledge.

Rothman’s definition includes the possibility that the heritage variety of a language may be different from the variety that is spoken as a majority language, given its special context of use. Heritage languages are worth studying in order to learn more about the nature of linguistic change. They represent a development of a language that often differs from that of the language of origin, and they have developed in the vicinity of a majority language, with ample possibilities of influence on the heritage language. The study of heritage languages therefore potentially reveals which factors are central in language change and which are not, which factors affect language acquisition and in what way, and they also present important data on possible individual language loss (attrition) in the situations where heritage speakers no longer use their first language (see e. g. Håkansson 1995; Montrul 2008; Johannessen and Salmons forthcoming). In this study, we investigate word order in embedded clauses in Heritage Scandinavian, which is different from that of European Scandinavian. In Heritage Scandinavian we find a substantial amount of verb–adverb word order in

2 We thus refer to Heritage Norwegian and Heritage Swedish, and to Heritage Scandinavian to cover both. Likewise, we call the languages of origin European Norwegian and European Swedish, as well as European Scandinavian, whenever we want to refer to both at the same time.

Embedded word order in Heritage Scandinavian 

 241

embedded clauses. In European Scandinavian this order is not possible, with the exception of one type of subordinate clause. The article is organized as follows. Section 2 gives an overview of the recordings available for Heritage Scandinavian. Section 3 describes the word order of main and subordinate clauses in European Scandinavian. Section 4 then presents embedded word order in Heritage Scandinavian, including the unexpected non-standard word order. This section contains both linguistic examples and frequency counts. Section 5 argues for why this non-standard order must be seen as a change in the grammatical system rather than as something that belonged to the language of origin, and it argues against an explanation based on the idea that the new word order is a loan from the majority language, English. We also look at an explanation based on attrition, which we also reject. Finally, we try to see this new language change as a consequence of incomplete acquisition. This analysis turns out to explain our data well, and is in accordance with established knowledge of acquisition of embedded word order in the European Scandinavian languages. Section 6 sums up our findings.

2 American Scandinavian recordings In this section, we give a short introduction to the American Norwegian and American Swedish recordings.

2.1 American Norwegian The Norwegian language in America has been studied off and on for more than a century, first by professors of linguistics, George T. Flom (U Iowa) and Nils Flaten (St. Olaf College), at the very beginning of the 20th century (see Flom 1900–1904, 1903, 1912, 1926, 1929, 1931, and Flaten 1900–1904). The first recordings were done in 1931, when the Norwegian professors Ernst W. Selmer and Didrik Arup Seip went to the American Midwest to study the language of the Norwegians there. They left behind a collection of their recordings. Unfortunately, these phonograph rolls were fragile, and many of them could not be recovered (Haugen 1992). The rest are now available at the Text Laboratory web site at the University of Oslo (UiO). So are the recordings of Einar Haugen, the legendary Harvard professor who wrote The Norwegian Language in America (1953). Haugen’s recordings were collected between 1936 and 1948 with a total of 207 informants and nearly 55 hours. The informants ranged from 1st to 3rd generation immigrants, and were from the Midwest. Their Norwegian background covers the major parts of southern Norway.

242 

 Ida Larsson and Janne Bondi Johannessen*

In the 1980–1990s, Arnstein Hjelde (1992, 1996a, 1996b) made recordings of people with a Trøndelag and Gudbrandsdalen dialect background in the Midwest. Students of Joseph Salmons at the University of Wisconsin made recordings in the 2010s. Since 2010, Janne Bondi Johannessen (with Signe Laake, Arnstein Hjelde and others) has conducted five field trips recording mainly 2nd–4th generation speakers across the northern part of the U. S. A., from the state of Washington to Wisconsin. Some studies based on these recordings are Johannessen and Laake (2011) and several articles in a special issue of Norsk Lingvistisk Tidsskrift on the Norwegian language in America (edited by Johannessen and Salmons in 2012), as well as in an anthology appearing soon at John Benjamins Publishing Company (edited by Johannessen and Salmons forthcoming). All recordings are available at the Text Laboratory, UiO.

2.2 American Swedish American Swedish has also been studied previously (see in particular Hasselmo 1974). In the 1960s, Folke Hedblom and Torsten Ordéus collected a large material of American Swedish, comprising over 300 hours of recorded speech, and based on this material a few studies were published (see Hedblom 1963, 1970, 1974 1978, 1981). These recordings are available at the Institute for Language and Folklore in Uppsala. They consist of interviews of varying length (and varying degree of formality) with 1st–4th generation speakers of Swedish, with family from almost all different parts of Sweden. Many of the informants are first generation immigrants who emigrated from Sweden as children or young adults. Other informants are descendants of Swedes that emigrated during the 19th or early 20th century, and many of them grew up with Swedish as the only first language.3 In the last couple of years, new recordings of American Swedish have been made, as part of the project Swedish in America (see Larsson et al. to appear and Andréasson et al. 2013 for an overview). In 2011–2012, 88 informants were interviewed and recorded in Minnesota and Illinois. Most of the informants were second or third generation immigrants with parents, grandparents or great grandparents from different parts of Sweden, but a few had themselves emigrated from Sweden. Some of the American-born informants had Swedish as their only

3 There are also other older collections of American Swedish varieties. For instance, a large material of Heritage Finno-Swedish was collected in the beginning of the 1970s (see Ahlbäck et al. 1976 and Ivars 2003). Those recordings have not been used in the present study.

Embedded word order in Heritage Scandinavian 

 243

first language, and had learned English in school; others had English as the dominant language when they grew up, or were simultaneous bilinguals. In this paper we focus on heritage speakers born in America. Unless otherwise noted, they have Norwegian or Swedish as their first language and are still fluent speakers. The data used are from recordings of (informal) interviews and dialogues between two or more heritage speakers.

3 Word order in Norway and Sweden In this section we give a brief overview of main and embedded word order in European Norwegian and Swedish. We exemplify with Swedish, but the pattern is the same in Norwegian. The Scandinavian languages in Europe and in America are all verb second (V2) languages. In declarative main clauses, the verb therefore immediately follows the first constituent, whether it is a subject, an adverbial or some other fronted phrase; cf. (1a) and (1b). Wh-questions also have the verb in second position (disregarding some Norwegian dialects); see (2a) and (2b).4 (1) a.

Den tröjan köpte han inte. that shirt.the bought he not ‘He didn’t buy that shirt.’

(Swedish)

b. * Den tröjan han köpte inte. that shirt.the he bought not (2) a. Vad köpte han? what bought he ‘What did he buy?’

(Swedish)

b. * Vad han köpte? what he bought

Following e. g. Rizzi (1997), Westergaard and Vangsnes (2005), and others, we will assume that the verb in main clauses spells out a head in the C-domain, which determines clause type. The composition and properties of the C-domain can therefore vary depending on clause type. However, for the present purposes the simplified structures in (3) will suffice, where C, I and V are used to cover a combination of heads (lower copies are marked with strikethrough). We will assume that in V2-contexts the verb is in a position in C, and that V2 involves V-to-C movement.

4 Forms like tröjan ’shirt + definite suffix’ are glossed as ’shirt.the’ throughout.

244 

 Ida Larsson and Janne Bondi Johannessen*

(3) a. [CP [Den tröjan] köpte [IP han köpte [NegP inte Neg° [VP han köpte ‘He didn’t buy that shirt.’ (= 1a) b. [CP vad köpte [IP han köpte [VP han köpte ‘What did he buy?’ (=2a)

In European Scandinavian, the finite verb follows both the subject and sentence adverbs in embedded clauses; see the relative clause in (4) and the embedded question in (5) (cf. among many others Holmberg and Platzack 1995; Vikner 1995): (4) a.

tröjan som han inte köpte shirt.the that he not bought ‘the shirt that he didn’t buy’

(Swedish)

b. * tröjan som han köpte inte shirt.the that he bought not (5) a.

… vad han inte köpte what he not bought ‘what he didn’t buy’

(Swedish)

b. * … vad han köpte inte what he bought not

Embedded clauses typically do not express illocutionary force, and they are often assumed to have a more restricted C-domain (which presumably still varies depending on clause type). Unlike direct questions, embedded questions do not, for instance, express interrogative force, and the interrogative head that triggers V2 in the main question can be assumed to be missing from the structure of the embedded question (see e. g. Westergaard 2006; Westergaard and Bentzen 2007 and references cited there). The finite verb will therefore remain in the verb phrase. Partial structures for (4a) and (5a) are given in (6). European Scandinavian does not have V-to-I movement. (6) a. [CP Oprel som [IP han I°  [NegP inte Neg° [VP han köpte ‘… that he didn’t buy’ (=4a) b.

[CP vad C° [IP han I°  [NegP inte Neg° [VP han köpte ‘... what he didn’t buy’ (=5a)

In an investigation of embedded word order, it is important to treat that-clauses (No. and Sw. at/att-clauses) separately. As discussed by Heycock (2006), Julien (2007, 2009), Wiklund et al. (2009) and many others, there is variation with respect to verb placement in certain types of that-clauses depending on the matrix predicate. In examples such as (7), the embedded verb can optionally be placed before negation. (The brackets in (7) mark that the complementizer is optional in Swedish.)

Embedded word order in Heritage Scandinavian  (7) a. b.

 245

Jag tycker (att) den passar inte bra. (Swedish) I think that it fits not well ‘I don’t think that it fits well.’ Jag tycker (att) den inte passar bra. I think that it not fits well ‘I don’t think that it fits well.’

Examples such as (7a) are often analyzed as involving embedded V2, and as having the verb in a position in the C-domain. In examples like these, it is also possible to have topicalisation and verb–subject order: (8)

Jag tycker verkligen (att) den boken ska du inte läsa. I think really that that book.the should you not read ‘I really don’t think that you should read that book.’

(Swedish)

The precise restrictions of embedded V2 are the subject of some debate (see e.g Julien 2007, 2009 and Wiklund et al. 2009), but it is typically restricted to asserted or semi-factive that-clauses in (European) Norwegian and Swedish. There is, however, considerable variation between speakers and dialects (see Bentzen 2014a for an overview). Importantly, V-to-C movement (i. e. V2) is never obligatory in that-clauses in Norway and Sweden, but it is not infrequent in the spoken language. On the contrary, in the Norwegian part of the Nordic Dialect Corpus (Johannessen et al. 2009), half of the that-clauses (474/937, 50.6 %) have the order verb–negation (Bentzen et al. 2013; cf. Julien 2008, and also Jensen and Christensen 2011, who argue that embedded V2 is more common in Danish than has generally been assumed). With respect to other kinds of embedded clauses, there is little variation in the placement of verbs relative to negation in Mainland Scandinavian (see Bentzen 2014b for an overview of the word order in relative clauses). A small number of dialects sometimes seem to have verb movement to I, independent of clause type.5 This is the case in Övdalian, for instance, where V-to-I movement appears to be required in certain very restricted contexts (in clauses with a null subject; see Rosenkvist 2011). Verb movement is, however, generally not obligatory in Övdalian.6 In spite of this word order variation in embedded clauses in

5 As shown by Bentzen (2007), Northern Norwegian has optional verb movement across adverbs such as ofte ‘often’ in e. g. embedded questions and relative clauses. The verb cannot move past negation, however. 6 Sandøy (2008: 189) claims that verb movement is possible in many types of embedded clauses in the dialect of Romsdal. With one exception with somewhat unclear origin, the examples he gives are all that-clauses. In part of the Nordic Dialect Corpus from locations in Møre og Romsdal,

246 

 Ida Larsson and Janne Bondi Johannessen*

European Scandinavian, what we find in Heritage Scandinavian is unexpected, as we shall see.

4 Embedded word order in Heritage Scandinavian We have good reason to assume that the emigrants brought with them to America a grammatical system with V2 in main clauses and certain that-clauses but without verb movement over negation in embedded questions and relative clauses. This is the pattern found in Swedish since the 17th century (see e. g. Falk 1993; Petzell 2012 and references cited there). As mentioned in Section 3, there are isolated dialects with embedded verb­–adverb order, but we have no reason to assume that the heritage speakers under discussion are influenced by rare dialects such as Övdalian through their input when growing up. In both Haugen’s and Hedblom’s recordings, first generation immigrants have the standard pattern (see section 5.2 below). In this section, we investigate the word order in Heritage Scandinavian embedded clauses.

4.1 Adverb–verb and verb–adverb order In the American Scandinavian recordings, we find examples of the standard patterns described in Section 3, but also a striking number of embedded clauses that do not follow the European Scandinavian pattern. While we would expect there to be some variation in subordinate declaratives (typically initiated by at/ att ‘that’), we would not expect anything other than the adverb–verb order in other subordinate clauses. However, we find many more examples with the order verb–adverb than expected with the former type, and a substantial number with the other subordinate types. This will be illustrated in this subsection. We start by showing examples that are in accordance with the European standard.7 (9a) shows a subordinate declarative with the standard adverb–verb order.

we do not find any examples of relative clauses where the finite verb precedes negation. (The search included sentences with the relative complementizer som followed by ikke ’not’, separated by up to five words.) 7 Most heritage language examples are represented in standard orthography. Haugen’s examples are represented in the way they are given in Haugen (1953) or in Oftedal’s transcriptions (available at the Text Laboratory, UiO). We have chosen to limit the glossing from Scandinavian to English by providing a word-by-word translation, and using punctuation when one Scandina-

Embedded word order in Heritage Scandinavian 

 247

(9b) shows an example of the same word order in an embedded polarity question, and (9c) in a relative clause. (9) a. vi er lykkelige på den måten at vi ikke bruker we are lucky on the way.the that we not use

stav eller noe slikt stick or something such ‘We are lucky that we don’t use staves or such things.’ (Norwegian, Elnor, coon_valley_WI_02gm)

b.

jeg kan gjøre det hvis om det ikke regner I can do it if whether it not rains ‘I can do it if it does not rain.’ (Norwegian, Tip, coon_valley_WI_06gm)

c.

det er mange som ikke har slutta òg there are many who not have stopped too ‘There are also many who have not stopped.’ (Norwegian, Elnor, coon_valley_WI_02gm)

As in European Scandinavian, we find the order verb–adverb in subordinate declaratives; a Heritage Norwegian example is given in (10) and a Heritage Swedish example in (11). (10) (11)

det er så lenge sia at jeg kommer messom ikke i hug it is so long ago that I come vaguely not in memory akkurat hvor vi var hen just where we were LOC ‘It’s so long ago that I hardly remember where exactly we were.’ (Norwegian, Tip, coon_valley_WI_06gm) jag visste att han skulle inte leva mycket längs I knew that he would not live much longer ‘I knew that he wouldn’t live much longer.’ (Swedish, Konrad, mn11_m013)

However, in Heritage Scandinavian we also find examples of the order verb– adverb in embedded polarity questions; see (12) and (13). Examples like these are not found in European Scandinavian.

vian word has to be translated to two English words. Grammatical morphemes (other than the definiteness suffix) are not represented.

248 

 Ida Larsson and Janne Bondi Johannessen*

(12)

om du finner ikke ut så if you find not out then ‘If you don’t find out, then.’ (Norwegian, Colleen, sunburg_MN_04gk)

(13)

Om ja sätter inte en sten på farfars grav if I put not a stone on granddad’s grave



så kommer det aldrig en sten där so comes there never a stone there ‘If I don’t put a stone on granddad’s grave there will never be a stone there.’ (Swedish, Konrad, mn11_m013)

Relative clauses can also have the unexpected verb–adverb order; see (14)–(15). Relative clauses in European Scandinavian, by contrast, are known to have a rigid adverb–verb word order, as noted in Section 3 above. (14)

det var en som arbeida med dem som forstår ikke there was one who worked with them who understands not

så mye norsk so much Norwegian ‘There was one who works with them who doesn’t understand much Norwegian.’ (Norwegian, Irene, zumbrota_MN_01gk) (15)

hon hade [… ] en äldre bror som arbeta också där oppe she had an older brother that worked also there up ‘She had an older brother that also worked up there.’ (Swedish, Konrad, mn11_m013)

Interestingly, there were examples of Heritage Scandinavian with the new word order as early as sixty years ago. The examples in (16) are from Haugen’s recordings. (16) a. Då di kåm ti detti landi då settla dæ på en when they came to this country.the then settled they on a homstedde som e no Taon åv Farmington homestead that is now town of Farmington ‘When they came to this country, they settled in a homestead that is now the town of Farmington.’ (Norwegian, Winfield Krostu, Waupacs co., born in Wis 1884, rec.1942) b.

Då me ha no blitt jipte, […] så kjæm hornaran when we have now become married so came musicians.the ‘When we had been married came the musicians.’ (Norwegian, Winfield Krostu, Waupacs co., born in Wis 1884, rec.1942)

Embedded word order in Heritage Scandinavian 

 249

Perhaps equally interesting is the fact that the verb type in the embedded clause appears to be irrelevant. One might have thought that the order verb–adverb would be restricted to (or more common with) auxiliary verbs, but in the present section we seen examples of the equivalents of content verbs, such as ’come’, ’find’, ’put’, ’understand’ and ’work’. This point is relevant for our discussion in 5.1, on whether it is an English pattern that is borrowed.

4.2 Corpus data In order to understand to what extent the cases of new order are more than just idiosyncratic findings, we need to count them. For this purpose it is convenient to make use of The Corpus of American Norwegian Speech. It is a searchable corpus of recordings of eleven speakers (at the time of writing) in the American Midwest; five from Minnesota (Rushford, Sunburg and Zumbrota) and six from Wisconsin (Coon Valley). Although it is in the initial development phase, and thus small (only 50,000 words in total in March 2013), it is already a good research tool. A search for a given complementizer followed by up to 3 words and then an appropriate adverb gave us a sizeable amount of results, see Table 1. Table 1: Results from a search in the Corpus of American Norwegian Speech Complementizer

Adverb

om ‘whether’ som ‘which/who/that’ at ‘that’

ikke ‘not’ ikke ‘not’ ikke ‘not’

Sum

Verb–Adverb

Adverb–Verb

Sum

1 6 15

2 7 1

3 13 16

22

10

32

The result for the complementizer om ‘whether’, which introduces embedded polarity questions, only gave three hits. Interestingly, even with this small number, we find the order verb–adverb, which is non-existent in European Norwegian. The relative clauses (here those introduced by som ‘which/who/that/ as’) are also worth noticing. Six out of 13, that is, approximately half, have verb– adverb order, an order that is not considered to be grammatical or existent in European Norwegian.8 This number is in agreement with the findings of Taranrød

8 We have not noted any difference between subject and object relatives (but the numbers are small).

250 

 Ida Larsson and Janne Bondi Johannessen*

(2011: 53).9 She found that out of twelve relative clauses containing an adverb, six, that is half, had verb–adverb order. She also looked at her findings in relation to the overall size of the text material. Her six non-standard word order relative clauses were found in a material of altogether 60,000 words. When searching the (European) Norwegian part of the Nordic Dialect Corpus, she found only three such clauses in a material of a total of 1.5 million words (Taranrød 2011: 64). With respect to that-clauses, which in European Scandinavian can have either verb–adverb or adverb–verb word order depending on the meaning of the superordinate predicate, we find that practically all have verb–adverb word order (15 against 1). This is much more than the 50.6 % found in European Norwegian (see Section 3). With respect to that-clauses, then, there is also a difference between American and European Scandinavian. We have made a closer comparison between Heritage and European Scandinavian by investigating the same clause types in the Nordic Dialect Corpus, choosing a subgroup that amounts to the same number of 49,000 words in total. This subgroup is limited to old informants from Oppland county (and from there, only from Brandbu, Brekkom, Gausdal, Jevnaker, Nordreland and Skreia), to make them maximally similar to the American informants with respect to age and place of linguistic origin. The results are presented in Table 2. Table 2: Results for European Norwegian from a search in the Nordic Dialect Corpus Complementizer

Adverb

om ‘whether’ som ‘which/who/that’ at ‘that’

ikke ‘not’ ikke ‘not’ ikke ‘not’

Sum

Verb–Adverb

Adverb–Verb

Sum

0 1 13

5 16 19

5 17 32

14

40

54

Table 2 shows a very different distribution of word order types, as we could have predicted based on the previous research mentioned above. In European Norwegian there is no example of verb–adverb order with the complementizer om, only one such order for the som-clauses, and the number of that-clauses with verb– adverb order constitutes less than half of the total number of that-clauses. We have calculated whether the differences between the tables for each subjunction type is significant. Using Fisher’s Exact Test, we find that the difference between Heritage and European Norwegian for that-clauses and som-clauses is highly sig-

9 It should be mentioned that there is a partial overlap between her informants and those in the Corpus of American Norwegian Speech.

Embedded word order in Heritage Scandinavian 

 251

nificant, with p-values of p  õue ‘to the outdoors’ [lative] välja [genitive] + poole ‘towards’ [lative] > väljapoole ‘to(wards) the out(er) side’

3 The topic of grammatical replication (cf. Heine and Kuteva 2006: 49) remains outside the scope of this paper, but it is certainly a possibility in the case of these languages, considering their long history of mutual contact. 4 All the Estonian grams have elative and inessive parallels as well, as in väljast ‘from the outside’ and väljas ‘outside’; so does Võro. In modern Estonian, ‘into the field’ and ‘into the court-

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 269

(2) Võro: väli > vällä ‘out’ [lative] vällä [genitive] + poolõ ‘towards’ [lative] > välläpoolõ ‘to the out(er) side’ uss ‘door’ > ussõ ‘to the outdoors’ [lative] (3) Latvian5 (‘bright place, clearing’ >) lauks ‘field’ > laukā ‘out’ [locative] ārs ‘arable land, clearing, field’ > ārā ‘out’ [locative] uz āru [preposition ‘to’ + accusative] (4) Lithuanian6 laukas ‘field’ > į lauką ‘out, to the outdoors’ [preposition ‘to’ + accusative] laukan ‘out’ [illative] lauk ‘out’ oras ‘air’ (cognate of ārs) > į išorę ‘out’ [preposition ‘to’ + accusative]

This study concentrates only on the usage of those lative-direction spatial grams with a sense of ‘out, to the outdoors’ – not the domain of expressions for movement from the interior to the exterior in general (for which a variety of other means can be employed, such as verbs with the meaning ‘to exit’, case-marking, and also verbal prefixes in the Baltic languages that often focus on the elative rather than the lative aspect of the movement). Only spatial usage is considered here, but it should be noted that some of the grams have also grammaticalized further: the Finnic välja/vällä also function as semi-productive aspect markers of completed action, of terminative movement with a clear goal, and have also lexicalized with numerous verbs producing constructions where the original interior-exterior opposition is barely present if at all. Note that many of the aspectual usages are likely to be grammatical replications of similar constructions in German (cf. Hasselblatt 1990). Below are some simple examples from Estonian (examples 5–9).7

yard’ would be välja-le and õue-le, respectively (with an allative case suffix), while välja and õue seems to have retained an old lative suffix that is no longer productive (the form is used by the modern illative). Here and below, I am using small caps to distinguish spatial concepts (e. g., elative) from cases (e. g., elative) that are traditionally used in the grammars of the languages involved. 5 laukā and ārā, etymologically bearing the locative suffix, function both in the lative ‘to the outside’ and locative ‘outside’ sense; ārā also has a parallel elative form, no ārās with the preposition ‘from’ governing genitive. The macron on the vowels in Latvian marks long length. 6 Both stems have parallel elative and locative forms as well, as in iš lauko ‘from the outside’ with the preposition governing genitive, and lauke ‘outside’. Not unlike the Latvian macron, the ogonek ( ̨ ) marks vowel length in Lithuanian. 7 The Leipzig Glossing Rules and abbreviations are used. Additionally, TRA signifies the translative and ELA the elative case.

270 

 Andres Karjus

(5) Spatial: mees sõit-is garaaži-st välja man drive-3SG.PST garage-ELA out The man drove out of the garage. (6) Completive (only partially productive, semi-lexicalized): poiss joonista-s=ki lõpuks välja kõik detaili-d boy draw-3SG.PST=indeed finally out all detail-PL Finally, the boy did draw all the details. (7) Terminative movement: mees sõit-is õhtu-ks Lätti välja man drive-3SG.PST evening-TRA Latvia.ILL out The man reached (drove all the way to) Latvia by the evening. (8) Semi-lexicalized: mees astu-s partei-st välja man step-3SG.PST party-ELA out The man left [lit.: stepped out of] the political party. (9) Lexicalized: mees ei tei-nud poisi-st välja ja läks edasi man NEG do-PST boy-ELA out and go.3SG.PST forward The man did not pay attention to [lit.: did not do out of] the boy and went on.

2 D  evelopment of the spatial grams in the light of grammaticalization theory Considering the vast multitude of definitions of and approaches to grammaticalization in the relevant literature, it should be established here that this study primarily bases itself on the model of grammaticalization (consisting of the interrelated mechanisms of desemanticization, extension, decategorialization and erosion/phonetic reduction) promoted by Heine and Kuteva (2004, 2007), and also on notions by Haspelmath (1999, 2004), who predicts that “[a]s an item expands to the right and forms a grammaticalization chain, some of its earlier manifestations on the left typically disappear, so that the chain loses on the left what it gains on the right”. (Haspelmath 2004: 33) ‘Left’ and ‘right’ refer to his models of grammaticalization (rightward), retraction (rightward with retraction) and antigrammaticalization (rightward, then leftward) that are visually depicted as movements beginning from the left, such as in Figure 1.

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

degree of grammaticalization 1 A1 2 A1 A2 3 A2 A3 4 A2 A3 A4 5 A4 A5 6 A5 A6

time

 271

Fig. 1: The scheme of the rightward expansion of grammaticalization, replicated from Haspelmath 2004: 33.

Following Traugott and Trousdale (2010: 26, 39‒40), gradience is understood as synchronic variation within a language community that results from small-step changes that are associated with the gradualness of language change (such as grammaticalization). They see gradience as “an ideal testing ground for hypotheses about the expressions that are most likely to undergo grammaticalization, and perhaps more critically, the way in which grammatical constructions emerge.” (Traugott and Trousdale 2010) Consider now the following hypothetical grammaticalization path for spatial grams with the sense of ‘out’, ‘to the outdoors’. Note that a gram may, of course, also stop in its grammaticalization path at any point along the cline. A lexical item that began its path of grammaticalization towards ‘out’ was probably used frequently (cf. Hopper and Traugott 2003: 127; Krug 2001) referring to a place one would exit into. The typical ‘inside’ where a human being could be found would be a house or dwelling place, so one would exit into a space that, depending on where the house is located, could bear the lexical designation of ‘street’, ‘field’ or ‘(court)yard’, or perhaps just ‘air’.8 As desemanticization takes place, the usage of the item extends to other contexts as well, for example, to describe exiting a building in general, regardless of what is actually outside of the door (the meaning of the noun bleaches with the exception of the spatial direction that remains intact (Heine and Kuteva 2007: 64)). If the lexical sense of the grammaticalizing word is that of ‘field’, it may, for example, be extended to be used for describing exiting into the street as well; consequently, the other way around if it is ‘street’, as in the Russian na ulicu ‘to the outdoors’ (ulicu is the singular accusative of ulica ‘street’). Decategorializa-

8 The grams in this study were already grammaticalized by the 16th/17th century – taking into account the agricultural society of the time, it makes sense – the space surrounding a house was indeed most likely a field (immediately surrounding the house could also be the yard –like the Estonian example of õu).

272 

 Andres Karjus

tion gradually washes away the morphosyntactic properties of the unit, while the stem and the marker of direction of movement (for example, a case suffix) are joined into one non-analyzable unit (or a small partial paradigm thereof) through lexicalization.9 Climbing the cline, the new gram goes through further desemanticization and extends to contexts other than exiting a house or confines on solid ground in general. The process can also be explained as the loosening of semantic selection restrictions (following the respective notion in Lehmann 1989) that an item imposes on its context: as an item makes its way from an open word class towards a closed one, its suitability to be used in more situations increases. If the gram grammaticalizes further (and semantic restrictions loosen), it would also become usable in situations quite different from exiting a house. For instance, the moving entity might be manipulated instead of moving itself, or the moving entity might be an inanimate object instead of an agent capable of self-propelled motion. Under the assumption that some common tendencies would exist in the development of the spatial grams under observation, and taking into account the semantic similarity of their etymological origins, the usage of the grams (described above in the first section) will be compared to see if gradience in their usage corresponds to what would be expected as the outcome of a gradual diachronic process. In order to adequately represent the variable usage of these forms, a relatively large data set is required. For the exploration of that data, we need a sufficiently robust method. Both of these issues will be addressed in the next section.

3 The data and method A total of 195 native speaker subjects filled in the online questionnaires that provide the data for this investigation. The questionnaire consisted of 10 handdrawn images, with a selection of grams below each image. Each image was separated horizontally into two parts, the initial state of affairs pictured above and the event of movement below it. The participants were asked to rate the grams according to whether or not they would use them in a description of the pictured movement when asked the question “where to?” The images depicted different

9 Here, lexicalization is not understood as the opposite process to grammaticalization, but rather as another type of reduction mechanism, potentially co-occurring with the latter (cf. Lehmann 2002: 13, 16).

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 273

kinds of exiting, a figure moving or being moved out of the confines of a building, settlement, room, etc. Since the original subject samples were not of equal size and the goal at hand is a modeling exercise rather than a representative overview, a reduced subsample of 20 random subjects per language is used. Furthermore, only the following grams are considered: the Estonian välja, õue, Võro vällä, Latvian ārā, laukā, and Lithuanian į lauką. In the following, the languages will be abbreviated as E, V, LA and LI respectively. The other grams are excluded because they were being used scarcely or just by a minority of the subjects, or, in the case of the remaining Lithuanian grams, there was so much variation that it was nearly impossible to draw conclusions (cf. Karjus 2012: 59‒65). It should be kept in mind though that all of the discussion is limited to the senses that would emerge in the context of the images used in the study; for example, the E väljapoole ‘to(wards) the outer side’ is a fairly common gram, yet it is apparently not used often to answer the question ‘where to?’ in the context of the situations depicted on the questionnaire images. A possible explanation is that its sense puts the focus on the path or the opposition between the inner and outer space, rather than the goal, which could be seen as the focal point in the images. Only data on four images is considered here (Figure 2), those depicting the exiting from a house to an empty field, to a street, from a plane in flight, and the pouring of spherical objects out of a container, while the rest of six images are left out (taking balls out of a box, a car exiting a town, a man from the water, from a room inside a house, a dog from a kennel and an abstract exiting scheme). The inclusion of two rather similar images is justified by the fact that they are considered different situations by some subjects, who also used (rated) the grams differently between the two. This will become apparent later on in the discussion of the statistical model. Originally, the subjects had to rate the grams on a scale of 0 to 5. To simplify the analysis and presentation here, these values have been factorized. Scores 0 to 3 are reinterpreted as ‘would not or would rather not use’ and scores 4 and 5 have been reinterpreted as ‘would use’, yielding a clearer binary distinction.10

10 One will notice that the split is not symmetric – in some other cases it might be hard to justify, but it serves a purpose here. First of all, in all the language samples, the minimal (0) and maximal (5) scores were used by far the most often, the intermediate scores forming a rather small minority (ibid., 40). Secondly, the split filters out scores that express doubt in the suitability to use this or that gram in a given context; only fairly certain opinions (scores 4, 5) are presented.

274 

 Andres Karjus

Fig. 2: The four images under observation in this study. The moving figures were colored red in the original questionnaire. The following shorthand names will be used to refer to them (from left to right) – “field”, “street”, “plane”, and “balls”.

The simplifications described so far are necessary to filter out noise resulting from micro-variation that would otherwise pollute the model to be constructed below and make it difficult to comprehend, especially in the static medium that is the printed page. This constrained approach allows the presentation to remain focused and avoid dealing with the considerable variation in the full data set (stemming from a number of potential factors discussed in Karjus 2012: 35‒39, 60‒66). To sum things up: the data consist of the descriptions by 80 subjects (20 per language) of 4 exiting images, represented by ratings given to each gram option in the context of each image; 6 grams (see above) in total are being observed. The collection of ratings given to one gram by one subject is operationalized as the unit of analysis. It might be useful to think of the data as a table of 4 columns and 20 * 6 = 120 rows, with 1’s (“would use”) and 0’s (“would not use”) populating the cells (Table 1). Table 1: An example fragment of the data, showing the usages of välja by three Estonian subjects subject_gram

“field”

“street”

“plane”

“balls”

E18_välja E19_välja E20_välja

0 0 1

0 1 1

0 1 1

1 1

1

A multidimensional data set such as this one can be visualized using multidimensional scaling (MDS), an explorative statistical technique for measuring and explicating similarity and dissimilarity between the entities being analyzed. When applied to linguistic phenomena, it produces a spatial representation of similarity between linguistic entities or functions (Croft and Poole 2008; the method has also been used for linguistic data by Levinson et al. 2003 and Ruette et al 2011, among others). Similarly to Principal Component Analysis, it is quite useful as a

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 275

dimension reduction technique (Baayen 2008: 136). While multiple dimensions would be difficult to grasp for the human eye, a scatterplot-like two-dimensional space like the one in the next section is easy to comprehend, while the interpretation of relative spatial distance as (dis)similarity should be fairly intuitive. Naturally, there are diagnostic measures (e. g., stress and dispersion accounted for) to determine how well the transformed visual representation matches the structure of the “real” data – they are within acceptable limits in the model below.

4 Mapping the gradience in a cross-linguistic sample Owing to the factorization and subsampling carried out in the data (discussed above), the MDS model below, which visualizes the gradience in the usage of the spatial grams under observation (Figure 3), is relatively clear and readable also on paper.11 The configuration of the objects in the plot is by no means coincidental. The Euclidean distances between the (centers of the groups of) labels are to be interpreted as dissimilarity measures – the absolute location of an object in the plot is not meaningful as such, but its relative position with regards to other objects is. The further apart any two objects are placed on the graph, the more different they are. The closer they are, the more similar they are, which in this context indicates how similarly or differently the grams were used (rated) by the subjects of the questionnaires. Some jitter is added to the label placement in order to maintain readability. Therefore, labels standing in a close group should be interpreted as essentially overlapping. Overlap in an MDS plot indicates an exact match (here: identical ratings), and binning (cf. Maciejewski 2011: 52) is used to visualize the number of overlapping gram usages. The larger the binned marker (here: label font size), the more usages are stacked “under” it; labels with the smallest font stand for usage by a single subject, while the largest size (that of õue in group 2) incidentally stands for 16 overlapping usages. The following is a short summary of what the map represents. At the top-left corner, marked by (0), is a small cluster of grams that those subjects actually did not use for any of the four images. (1) Moving anticlockwise, there are E õue and LI į lauką, used to describe the exiting situation in “field” (a man stepping out of the house and onto an empty patch of land, possibly a yard).

11 The PROXSCAL algorithm (with ordinal proximity transformations) in the software package IBM SPSS v20.0 was used for calculating the coordinates for the MDS model. The plot, utilizing the coordinates, was created using the functions available in R (v2.15.1; R Core Team 2013).

276 

 Andres Karjus

(7)

(0)

välja lauka

õue ara

lauka

lauka

(6) lauka ara välja

vällä

(5)

vällä

välja

ara

ilauka

(1) õue ilauka

(2) ara

õue vällä

ilauka lauka

vällä välja ara

vällä

ara lauka

vällä välja (4)

ilauka lauka välja (3) ara

Fig. 3: MDS model of spatial grams with the sense of ‘out, to the outdoors’ in Estonian (õue, välja), Võro (vällä), Latvian (ārā, laukā), and Lithuanian (į lauką), according to the usage by the respective subjects (20 per language), based on their gram ratings for the four images. Close clustering should be interpreted as overlap; font size illustrates the number of overlapping usages of that gram. Orthography has been simplified due to technical reasons.

(2) A number of grams that were used to describe both “field” and “street”; the usages of those grams could perhaps be translated as ‘(to the) outdoors’ in English. The font sizes speak of the fact that most of the usages of LI į lauką and E õue are found here. Note that õue is grammaticalized from ‘yard, courtyard’, apparently some speakers (in (1)) would not use it to describe exiting into the street. (3) The next group stands for grams that were used to describe the “plane” situation in addition to the aforementioned, but not the manipulated exiting depicted by the “balls” image. (4) This large stack of labels signifies those grams that were used by those subjects to describe all four images, including “balls”. This is the cardinal opposite point to (0) that signified the grams that were not used for any of the images. It is worth noticing that there appears to be another small cluster “leading” here, where the grams were used to describe “field”, “street”, and “balls”, but not “plane”. All in all, from this perspective, the map of synchronic usages corresponds quite well to the hypothetical cline of grammaticalization outlined above in

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 277

Section 2. The process of lexical items gradually losing their semantic contents and categorical properties, becoming grammaticalized enough to describe a variety of situations – is reflected by usage variation within the language communities. Taking into account multiple languages is the key – the full scale of gradience might not manifest itself in the usages of a single gram, but it may do so in a larger sample of different grams (with similar senses) from different languages. Naturally, including more languages could further support (or detract from) the hypothesis. Although the maximal number of contextual functions (that is four in this sample) has been attributed to some grams, as used by certain subjects, three more clusters occur on the map. (5) Clustered on the right side of the map are some grams that were used to describe three images, but “field”. (6) The next group consists of single usages of E välja, LA ārā and LA laukā, only used to describe “plane” and “balls”. (7) Finally, there are usages of E välja and of LA laukā that were used only to describe “balls”, but none of the other images. The adjacent two usages of laukā stand for descriptions of only “plane”. At first, the existence of the last three clusters appears to contradict the cline proposed above – but see Figure 4 below.

5 (Re)construction of a cline of grammaticalization The MDS model (Figure 3) and its description in the preceding section, except for a few outliers, is now summarized in Figure 4, which corresponds to the hypothesized grammaticalization cline in Section 2. The vertical axis represents a hypothetical time-line – not an absolute time-line, but rather the proposed location of a gram in relation to its own grammaticalization process. Figure 4 reflects the aforementioned model of grammaticalization (adopted from Haspelmath 2004: 33) (Figure 1, above) that predicts the likely disappearance of earlier functions. The difference is that the model in Figure 4 does not follow the grammaticalization of a single linguistic item but instead, combines cross-linguistic information on the grammaticalization of numerous items as used by numerous speakers – and nevertheless presents a similar conclusion. Of course, these inferences are based on fairly small samples and some of the usages in Figure 4 indeed represent the behavior of single speakers. There is also a chance that there could be errors or misunderstandings on the side of the subjects who filled in the questionnaires. Nevertheless, I find it reasonable to believe that the model does show some tendency in accord with grammaticalization theory, not just random variation and noise.

 Andres Karjus

balls

plane

< ---

field street

< --

location

278 

image [ █] = gram(s) used for this image [ █] = gram(s) not used for this image

E õue LI į lauką V vällä

(0)

õue

(1)

õue

į lauką

(2)

õue

į lauką į lauką

LA ārā LA laukā E välja ārā

laukā

vällä

ārā

laukā

vällä

ārā

laukā

välja

(4)

vällä

ārā

laukā

välja

(5)

vällä

ārā

(3)

(6)

ārā

(7) assumed degree of grammaticalization relative timeline

välja laukā

välja

laukā

välja

Fig. 4: The gradience in the usage of the spatial grams and the proposed grammaticalization path. The location refers to the enumeration in Figure 3 (above). Bold font indicates a dominant cluster.

The arrangement of the grams according to the hypothesized direction of change is supported by two considerations. For one, there is the fairly obvious etymological relation between the spatial grams and their respective lexical (i. e., non-grammaticalized) parallels. One might argue, though, against the direction – could it not be, for example, that the Estonian õu ‘yard, courtyard’ derives from the small paradigm of spatial grams with the sense of ‘to/from/in the outdoors’, or, that the Latvian laukā was first used in abstract contexts (requiring a highly desemanticized gram) and only later to describe situations like exiting a building into the field? While this is conceivable, this is where the grammaticalization framework supports the former interpretation with a strong tendency, with an abundance of cross-linguistic examples of the lexical becoming grammatical/ functional (cf. Heine and Kuteva 2004), with far fewer examples of the grammatical becoming less grammaticalized or lexical (cf. Haspelmath 2004: 27‒33; Lehmann 2002: 14‒15). The fact that laukā and ārā also occur at (0) might indicate that for some speakers, those grams have made the “full circle”, in other words, become grammaticalized to the point that they would not be used to describe any

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 279

of those situations anymore (but perhaps more abstract or schematic exiting, or a different aspect of the movement).12 One does not even have to look far to find similar examples of the kind of grammaticalization paths as presented above, where a lexical item denoting a space that is found outside one’s door develops into a spatial gram with a sense of ‘to the outdoors’ or ‘out’ in general. In fact, this appears to be almost an areal trait of the eastern Circum-Baltic region. For the same sense, Polish has the spatial gram na dwór (grammaticalized from the combination of the preposition na ‘to, at’ and dwór ‘courtyard’) and, in some southern varieties, na pole (with ‘field’), similar to of the Serbian napolje ‘to the outdoors’. Russian has na ulicu (with the noun for ‘street’) and na dvor in some varieties. Votic, an endangered Finnic language has the grams kujalõõ (derived from ‘street’), but also the Finnic cognate vällää (cf. VKS), while Karelian (also Finnic) has pihale (allative of piha ‘yard’ (KKVS)) (see respective entries in Torikka 2009). In the extinct Baltic language of Prussian, the spatial gram used for that function was winna, derived from the noun wins ‘air’ (Mažiulis 1997: 240), much like in the Lithuanian į išorę ‘to the outside’. Similar examples can be found beyond the Circum-Baltic language area as well: the Samoyedic (Uralic) language Selkup has pōnä ‘out’ that derives from po ‘tree, forest’ (Hell and Katz 1982: 175‒177) and Basque (isolate) has kanpora ‘out’, from the Spanish loanword for ‘field’ (de Rjik 2008: 75), while the Võro ussõ (cf. Section 1) by chance happens to mirror the Latin foris ‘out’, which is also grammaticalized from ‘door’ (Svorou 1994: 258; cf. also the English outdoors).

6 Conclusions This paper has outlined a simple model of grammaticalization of grams with the sense of ‘out, to the outdoors’, arguably reflected by the gradience in a synchronic sample containing comparable experimental data from four languages, Estonian, Võro, Latvian, and Lithuanian. It has also been demonstrated how multidimensional scaling, a mathematically complex yet easily interpretable method from the domain of multivariate statistics, can be used to explore variation in a large set of linguistic data and reveal its structure within. It was shown that the usage variation of the grams, when viewed from a cross-linguistic perspective, forms

12 Such as from the inside to(wards) the outer side, with emphasis on the path – this meaning was for example attributed to the otherwise low rated uz āru (‘out’, with the preposition ‘to’) by some subjects in the comments section of the questionnaire – all of the images had their visual emphasis on the goal, rather than the path of the movement.

280 

 Andres Karjus

a cline that corresponds fairly well to what grammaticalization theory would predict to be the gradual path of diachronic change of spatial grams with that sense. Further research that would include more languages and more data could possibly shed more light on the issue. The approach could also be extended to the study of any other element of language that has, or is expected to have changed in time. Approaching language change from a synchronic point of view does not devalue the diachronic study of language – rather, it could be used as a supplement to compare gradual changes in a system against the synchronic gradience either in the same language or cross-linguistically, serving as a testing ground for hypotheses on the tendencies, if not universalities, in grammaticalization. The matter becomes particularly interesting if one is to adopt the uniformitarian assumption that “[t]here is no intrinsic reason to doubt that language change and the functional motivations underlying it were of the same kind in early language as what we observe in modern languages. Accordingly, grammaticalization theory can be extended from modern languages to early language by extrapolating from the known to the unknown.” (Heine and Kuteva 2007: 24)

References Baayen, Harald R. (2008): Analyzing Linguistic Data. A Practical Introduction to Statistics Using R. Cambridge: Cambridge University Press. Croft, William and Keith T. Poole (2008): Inferring universals from grammatical variation: multidimensional scaling for typological analysis. Theoretical Linguistics 34: 1–37. Fraenkel, Ernst (1962): Litauisches Etymologisches Wörterbuch, Band I. Heidelberg and Göttingen: Winter. Häkkinen, Kaisa (2004): Nykysuomen etymologinen sanakirja. Helsinki: WSOY. Haspelmath, Martin (2004): On directionality in language change with particular reference to grammaticalization. In: Olga Fischer and Muriel Norde/Harry Perridon (eds.), Up and down the Cline – The Nature of Grammaticalization, 17–44. Amsterdam and Philadelphia: John Benjamins. Haspelmath, Martin (1999): Why is grammaticalization irreversible? Linguistics 37: 1043–1068. Hasselblatt, Cornelius (1990): Das Estnische Partikelverb als Lehnübersetzung aus dem Deutschen. Wiesbaden: Harrassowitz. Heine, Bernd and Tania Kuteva (2006): The Changing Languages of Europe. Oxford: Oxford University Press. Heine, Bernd and Tania Kuteva (2007): The Genesis of Grammar. A Reconstruction. New York: Oxford University Press. Heine, Bernd and Tania Kuteva (2004): World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hell, Karin and Harmut Katz (1982): Zu selkupisch pō. Finno-Ugrische Mitteilungen 6: 175–177.

Grammaticalization of the exterior space in the Eastern Circum-Baltic 

 281

Hilpert, Martin (2010): What can synchronic gradience tell us about reanalysis? Verb-first conditionals in written German and Swedish. In: Elizabeth Closs Traugott and Graeme Trousdale (eds.), Gradience, Gradualness and Grammaticalization, 181–201. Amsterdam: John Benjamins. Hopper, Paul J. and Elizabeth Closs Traugott (2003): Grammaticalization. Second Edition. Cambridge: Cambridge University Press. Judžentytė, Gintarė (2009): Latvian Adverbs of Place: Synchrony and Diachrony. Summary of Doctoral Thesis. http://vddb.laba.lt/obj/LT-eLABa-0001:E.02 ~2010~D_20100204_095742–66783 (accessed June 4, 2014). Karjus, Andres (2012): out in Estonian, Võro, Latvian and Lithuanian: A usage-based approach to the grammaticalization and gradience of spatial grams of the exterior-region. Master’s thesis, University of Tartu. http://www.academia.edu/2103624 (accessed June 4, 2014). Karulis, Konstantīns (2001): Latviešu etimologijas vārdnīca. Riga: Avots. Krug, Manfred G. (2001): Frequency, iconicity, categorization: Evidence from emerging rnodals. In: Joan Bybee and Paul J. Hopper (eds.), Frequency and the emergence of linguistic structure, 309–335. Amsterdam and Philadelphia: John Benjamins. Lehmann, Christian (1989): Markedness and grammaticalization. In: Olga Mišeska Tomic (ed.), Markedness in synchrony and diachrony, 175–190. Berlin and New York: Mouton de Gruyter. Lehmann, Christian (2002): New reflections on grammaticalization and lexicalization. In: Gabriele Diewald and Ilse Wischer (eds), New reflections on grammaticalization, 1–18. Amsterdam and Philadelphia: John Benjamins. Levinson, Stephen, Sérgio Meira, and The Language and Cognition Group (2003): “Natural concepts” in the Spatial Topological Domain – Adpositional Meanings in Crosslinguistic Perspective: an Exercise in Semantic Typology. Language 79(3): 485–515. Maciejewski, Ross (2011): Data representations, transformations, and statistics for visual reasoning. San Rafael, CA: Morgan & Claypool Publishers. Marja Torikka (ed.) (2009): Karjalan kielen verkkosanakirja. http://kaino.kotus.fi/kks (accessed 09. 04. 2013). Mažiulis, Vytautas (1997): Prūsų kalbos etimologijos žodynas 4. Vilnius: Mokslas. Metsmägi, Iris, Meeli Sedrik and Sven-Erik Soosaar (2012): Eesti keele etümoloogiline sõnaraamat. Tallinn: Eesti Keele Instituut. R Core Team (2013): R: A language and environment for statistical computing. http:// www.R-project.org (accessed June 4, 2014). de Rijk, Rudolf P. G. (2008): Standard Basque: A Progressive Grammar, Volume 1. Cambridge, MA: MIT Press. Ruette, Tom, Dirk Speelman, and Dirk Geeraerts (2011): Measuring the lexical distance between registers in national varieties of Dutch. In: Miguel Gonçalves, Amadeu Torres, and Augusto Soares da Silva (eds.), Línguas Pluricêntricas. Variação Linguística e Dimensóes Sociocognitivas, 541–554. Braga: Faculdade de Filosofia, Universidade Católica Portuguesa. Sullõv, Jüvä (2001): Võro-eesti synaraamat. http://www.folklore.ee/synaraamat (accessed 09. 04. 2013). Svorou, Soteria (1994): The Grammar of Space. Amsterdam and Philadelphia: John Benjamins. Traugott, Elizabeth Closs and Graeme Trousdale (2010): Gradience, Gradualness and Grammaticalization. How do they intersect? In: Elizabeth Closs Traugott and Graeme Trousdale (eds), Gradience, Gradualness and Grammaticalization, 45–73. Amsterdam: John Benjamins.

282 

 Andres Karjus

VAKK = Vana kirjakeele korpus. http://www.murre.ut.ee/vakkur/korpused (accessed 09. 04. 2013) VKS = Grünberg, Silja, Elna Adler and Merle Leppik (eds.) (1990–2011): Vadja keele sõnaraamat 1–7. Tallinn: Signalet.

Appendix 1. A Fragment from the Questionnaires.

Helle Metslang, Karl Pajusalu and Külli Habicht

Conjunctive markers of polar questions in Estonian1 Abstract: In Estonian, as is common in Circum-Baltic languages, polar questions are typically formed by means of particles. The article focuses on question markers in Estonian that originate from several subtypes of conjunctive coordination (additive, contrastive, adversative), and will discuss the genesis and usage dynamics of different question markers. The analysis is based on the corpora of the University of Tartu. The emergence of coordinative interrogative particles could be explained by the interplay of linking clauses in texts and presuppositions of polar questions. The use of coordination markers as polar question markers is based on reanalysis of a part of the presupposition of a potential question to the question itself.

1 Introduction The article focuses on question markers in Estonian that originate from conjunctive coordination, and will discuss the genesis and usage dynamics of different question markers. Our analysis will address several subtypes of conjunctive coordination (additive, contrastive, adversative) as a source of question markers in Standard Estonian. We will try to find an answer to the question of why conjunctive coordination is a constantly productive source of question markers and whether one can observe some trends in the grammaticalization of particles representing subtypes of conjunctive coordination when they have changed into question markers. The material for analysis comes from the following corpora of Standard Estonian at the University of Tartu: the Corpus of Standard Estonian from 1890 to 1990 (CSE), the Corpus of Old Written Estonian (COWE), and the Estonian Reference Corpus (ERC). The analysis is mainly based on prose texts from of the Corpus of Standard Estonian. We will analyse the occurrence and dynamics of the uses under study, which can be found in CSE texts from the 1890s, 1950s, and 1990s. We will also look at these forms in contemporary daily usage based on ERC mate-

1 The study was funded by the Estonian Research Council (projects PUT475 and IUT2–037).

284 

 Helle Metslang, Karl Pajusalu and Külli Habicht

rial from the popular magazine Kroonika (2001−2002), while new media is represented by the language of internet forums (2000−2008). In order to establish the diachronic dynamics of the usages under study, we analysed the usage of these speech forms in the old Estonian literary language (COWE), and also prose texts and media texts in CSE and in texts from other decades when the need arose. The first part of the article gives an overview of the primary ways to express polar questions in Estonian. The second part of the article discusses some general principles of the formation of polar questions and interrogative sentences; the third part focuses on the types of conjunctive coordination and the prerequisites for them to acquire new communicative additional meanings. The fourth part deals with the functioning of various types of conjunctive conjunctions as question markers on the basis of corpus examples. The concluding section sums up the results of the analysis.

2 The formation of polar questions in Estonian In Estonian, polar questions are typically formed by means of sentence-initial or sentence-final particles. The formation of questions by means of particles, especially sentence-initial particles, is typical of the Circum-Baltic area, with Estonian being one of the central languages among the languages with particle questions (cf. Koptjevskaja-Tamm and Wälchli 2001: 712−714). The principal method of expressing polar questions is the particle kas (1) (sometimes polar questions are called kas-questions). In negative interrogative sentences, the question can be expressed by means of the particle ega (2), while the sentence-final particle või~vä seems to be emerging (Lindström 2001). (1) Kas sul on arvutis viirus? q you:ade is computer:ine virus ‘Do you have a virus in your computer?’ (2) Ega sul arvutis viirust ei ole? q.neg you:ade computer:ine virus:prtv neg be.cng ‘Don’t you have a virus in your computer?’ (3) Sul on arvutis viirus või? you:ade is computer:ine virus q ‘You have a virus in your computer or?’

The other methods of forming questions (inversion and tag) are marginal, and their use is restricted. In Estonian, intonation is not a regular device for express-

Conjunctive markers of polar questions in Estonian 

 285

ing speech acts; a statement utterance can act in a certain context as a presumed question that requires affirmation (4). (4) See on sinu arvuti? this is you.gen computer ‘This is your computer?’

Estonian has lost the Finnic-Sami interrogative clitic -ko/-go; polar questions were formed by means of particles already in texts in the old written language (starting in the 16th century). The main markers of Estonian polar questions have undergone grammaticalization or are in the process of grammaticalization from particles and conjunctions with a coordinative meaning. The particles can be divided into conjunctive and disjunctive ones by their origin (Table 1). Table 1: The etymology and periods of use in the corpora of the more important Estonian interrogative particles (Metslang et al. 2011)2 Particle

Formal source

Semantic source

Period of use

eks (North Estonian)

ei-ko-s neg-q-conf1 ei-s neg-conf kaas ‚also‘ ei kaas neg also või ‚or‘

Disjunctive

17th−19th c.

Disjunctive

17th−18th c.

Conjunctive Conjunctive

17th c. − present 18th c. − present

Disjunctive

20th c. − present

es (South Estonian) kas ega või

The grammaticalization of the main question particle kas (original meaning ‘also’) in present-day Estonian and the question particle ega (original meaning ‘also not’) began in the 17th−18th centuries. The previous century saw the interrogative use of the sentence-final disjunctive või~vä (original meaning ‘or’). In addition, some sentence structures beginning with copulative (i. e. conjunctive) conjunctions act as polar questions.

2 neg − negation, q − question, conf − confirmative particle

286 

 Helle Metslang, Karl Pajusalu and Külli Habicht

3 Where do polar questions come from? Studies of the structure of interrogative sentences compare the structure of questions with that of the declarative sentence and proceed by default from the assumption that the structure of the declarative sentence serves as the basis for the structure of the interrogative sentence with some changes (e. g. Chisholm 1984; Dryer 2008). This assumption seems to be justified because the declarative sentence is unmarked; the interrogative sentence is marked with regard both to its communicative modal meaning and often also its form. The statement is the simplest and unmarked type of utterance, which is implicitly found in every other type of utterance, including questions. The other utterance types convey descriptive information, which is characteristic of the statement as a basis to which communicative (directive, emotional, etc.) information is added. By using a statement, a speaker simply informs others that some proposition is true or not; however, a question informs us about what he or she knows about the truthfulness of the proposition or what he or she does not know about it, wishing his/her conversation partner to fill in a blank in one’s knowledge. Usually, the form of the polar question is marked by comparison with the relevant declarative sentence. Thus, one might assume that questions have arisen cognitively on the basis of statements, and that the structures of questions develop on the basis of structures of the declarative sentence. Therefore, we suppose the following order in the genesis of speech acts: 1) asserting a proposition, and 2) asking something about the proposition; the utterance types are: 1) statement, 2) question; the sentence types: 1) declarative sentence, 2) interrogative sentence. Polar questions (yes-no questions) ask about the truth of a proposition. Modal logic (Åqvist 1965; Hintikka 1978) highlights two main components of the logical pragmatic structure of polar questions, which are the epistemic component (basis knowledge) and the pragmatic component (the directive to do away with uncertainty in basis knowledge). The epistemic component acts at the same time as a presupposition of the question (evaluation of the truth of the proposition and its negation). (5) Did Peter sell his house?

For example, we can see the presupposition of question (5) could be worded as follows: it is possible that he could have sold his house, and it is possible that he did not sell it. An equal wording is that one or the other is necessary – Peter either sold his house or he did not sell it. The epistemic component can be formalized, for example, in (6). (Metslang 1981: 6−8)

Conjunctive markers of polar questions in Estonian 

 287

(6) poss P & poss (neg P) ↔ nec P v nec (neg P)3

How then does a statement become a question, or under which circumstances do speakers begin to interpret utterances as polar questions? It seems to happen when it is apparent from the context that the truthfulness of the sentence content is uncertain, and the speaker has some reason to infer that one might expect the hearer to reduce this uncertainty. One such case is the linking of sentences in conversation. The linking of declarative sentences is usually coordinative; the following sentence adds the next statement to what was stated in the preceding sentence (conjunctive coordination); sometimes an alternative statement is added (disjunctive coordination). The coordinative link is often marked by such markers as and, too in the case of conjunction (7) and or in the case of disjunction (8). (7) This is Peter’s house. And he has a nice garden, too. (8) This is Peter’s house. Or he has sold it.

However if a sentence, as the second sentence in example (7), belongs to the turn of the other interlocutor, who is not fully convinced in the truthfulness of its content and assumes that the interlocutor might know the thing better, a question might emerge. Content like ‘He has a nice garden, too’ is interpreted not as a statement, but as part of the presupposition of the question, which may be true, but doesn’t need to be true. The recipient is expected to do away with the uncertainty of this knowledge (9). (9) This is Peter’s house. − And he has a nice garden, too?

The transition between the statement and the question is gradual; interpretations of utterances expressing one or the other depend on the speaker and the hearer’s knowledge about the world and each other (see Heritage 2012; for Estonian see Hennoste 2012). The emergence of interrogative particles with a coordinative meaning is based on the linking of sentences in the text, more precisely, on the possibility of reanalysis associated with the linking of sentences. If the following sentence that is associated with the preceding text by means of a copulative conjunction is interpreted as epistemically possible, it could be interpreted as a presupposition of a question, and the sentence-initial conjunction is reanalysed as an interrogative particle. This process could be illustrated as a sequence (10),

3 Thus, it is possible that P is true and it is possible that P is not true, or it is necessary that either P or the negation of P is true.

288 

 Helle Metslang, Karl Pajusalu and Külli Habicht

where the second sentence expresses an assumption that could further be interpreted as a question. (10) Peter will not be at home today. − But you will, of course?

In the 17th−18th centuries, Estonian witnessed the emergence of the interrogative particle kas (primary meaning ‘also’), the genesis of which could possibly be based on linking such a conjectural proposition with the preceding context (Alvre 1997; Campbell 1991; Metslang et al. 2011). The initial stage of the old written language reveals some examples of bridge contexts, where kas can be interpreted as a connective particle meaning ‘also’, associated with the preceding text, and as a question marker (see below, (19)). The lexicalization and grammaticalization of particles in the interrogative function is associated with their fixation in the periphery, beginning, and end of sentences, which are often the positions in languages where those particles emerge that indicate textual relations and pragmatic functions (Haselow 2012).

4 Conjunctive coordination and its markers in Estonian In descriptions of Estonian, coordination has been classified in a number of ways; the present study will follow the classification to be found in the book “Estonian Grammar II” (EKG II: 215−219). The first way it is divided up into categories is conjunctive coordination (traditionally called copulative coordination; it corresponds to logical conjunction) and disjunctive coordination (which corresponds to logical disjunction). Conjunctive coordination is further subdivided into unmarked conjunctive coordination (which corresponds to logical conjunction without any additional meanings; in Estonian it is mainly expressed by the conjunction ja ‘and’) and marked conjunctive coordination (the meaning of logical conjunction is supplemented with such semantic relations between coordinated constituents like contrast, inference, and explanation). The present study focuses on the functioning of markers of unmarked and adversative conjunctive coordination as interrogative markers. Adversative coordination has been divided into 1) contrastive, 2) concessive, and 3) corrective coordination (Izutsu 2008; Erelt 2010). 1) Contrast implies comparison, which shows some difference, but it does not imply denial of expectation, for example (11).

Conjunctive markers of polar questions in Estonian 

 289

(11) Minu kruus on sinine, aga tema oma on punane. I.gen mug is blue conj s/he.gen own is red ‘My mug is blue, but his is red.’

Contrast can be specified by means of connective particles, such as seevastu ‘in contrast’, jällegi ‘in fact’, omakorda ‘in return’, hoopis(ki) ‘actually’ (12). (12) Minu kruus on sinine, aga tema oma on jällegi punane. I.gen mug is blue conj s/he.gen own is ptcl red ‘My mug is blue, but his one is, in fact, red.’

2) The concessive opposition shows that the association of the post-member with the pre-member fails to match the speaker’s expectations. (13) Juhan on seagripis, aga käib tööl. Juhan is swine_flu:ine conj go:3sg work:ade ‘Juhan has got swine flu but goes to work.’

The concessive opposition is highlighted by concessive particles such as siiski ‘however’, ikkagi ‘anyway, nevertheless’, ometi ‘yet’, and sellest hoolimata ‘nevertheless’ (14). (14) Juhan on seagripis, aga käib ikkagi tööl. Juhan is swine_flu:ine conj go:3sg ptcl work:ade ‘Juhan has got swine flu, but nevertheless he goes to work.’

3) Corrective coordination links a positive and a negative member (15). (15) Juhan ei ole kodus voodis, vaid käib tööl. Juhan neg be.cng home:ine bed:ine conj go:3sg work:ade ‘Juhan is not at home in bed but goes to work.’

In Estonian aga ‘but, yet’ is a general adversative conjunction, which is used both in the case of contrastive and concessive opposition, however, ent ‘yet’ and kuid ‘but’ specify the denial of expectation. In addition, both types could be expressed by the general copulative conjunction ja/ning ‘and’ and the concessive particle (16). (Erelt 2010) (16) Juhan on seagripis, ja käib ikkagi tööl. Juhan is swine_flu:ine conj go:3sg ptcl work:ade ‘Juhan has got swine flu and nevertheless he goes to work.’

290 

 Helle Metslang, Karl Pajusalu and Külli Habicht

While in English the adversative conjunction but functions as a marker of all three types of adversative coordination, in German and Swedish, for example, these functions are divided between two conjunctions: aber (contrastive and concessive) and sondern (corrective) and men (contrastive and concessive) and utan (corrective). In Estonian, specialization is even more detailed – aga is a general adversative conjunction that functions both as a contrastive and concessive conjunction; the denial of expectation is expressed by kuid and ent (17), while vaid ‘but, only’ is corrective (18). (Erelt 2010; Izutsu 2008: 655) (17) Juhan on seagripis, ent käib tööl. Juhan is swine_flu:ine conj go:3sg work:ade ‘Juhan has got swine flu but goes to work.’ (18) Minu kruus ei ole mitte sinine, vaid punane. I.gen mug neg be.cng neg blue conj red ‘My mug is not blue, but red.’

In the present study, the term ‘additive coordination’ (see Malchukov 2004) stands for all unmarked conjunctive coordination. We will also analyse the functioning of two types of adversative coordination, corrective and concessive, as the basis for forming questions.

5 Markers of conjunctive coordination as question formers 5.1 Unmarked conjunctive coordination 5.1.1 Particle kas Two of the well-established question particles, kas and ega, are based on unmarked conjunctive coordination. In both cases, the additive meaning has already disappeared in Estonian. They will be discussed briefly as a background to the other markers in the grammaticalization process. Evidence of written Estonian texts from the 16th−18th centuries reflects the lexicalization and grammaticalization processes that proceeded from the word kansak ‘into the company’, which yielded the particle ka ‘also’, the interrogative particle kas, comitative marker -ga, and the adverb kaasa ‘together, along’ in contemporary Estonian. The source of the interrogative particle in this development was the connective particle in the form kaas ‘also’ (Alvre 1983, 1997; Metslang et al. 2011). Kas became an interrogative particle positioned at the beginning of the sentence in a position that connects utterances additively. Example (19) is a transitional case between the additive and interrogative functions of the particle.

Conjunctive markers of polar questions in Estonian 

 291

The particle kahs is used already at the beginning of the question but has also preserved its additive function (cf. auch in the German equivalent). (19) Kahs temma woip jelles omma Emma Jho sisse Q s/he can:3sg again own­_ mother.gen body.gen in minna/ ninck uhest sünditut sahma? proceed:inf and anew bear:ppptc shall:sup Kan er auch wiederumb in seiner Mutter Leib gehen/ vnd geboren werden? (COWE1649Stahl, 597) ‘Can he once again return to the body of his mother and be reborn?’

The first examples of its usage as an interrogative particle date back to the 17th century. In modern Estonian, kas is the principal and most widespread marker of polar questions, e. g. (20). (20) Kas teadus suudab selgitada armastust? Q science be_able:3sg explain:inf love:prtv ‘Can science explain love?’ (FORUM)

Interrogative particles with the source meaning ‘also’ can be found in other languages of the Circum-Baltic area, including in Livonian, Lithuanian, Latvian dialects, and Latgalian (cf. Koptjevskaja-Tamm and Wälchli 2001: 714).

5.1.2 Particle ega The sentence-initial particle ega in the negative question, which started to emerge in the 18th century, has a similar origin (ei/ep’neg + kaas ‘also’). (Metslang et al. 2011) (21) Eiga nemmad se külmaga kanna wilja kokko? neg.q they this frost:com gather.cng grain.prtv together ‘Don’t you think they might gather up grain during this frost?’ (COWE\1732-Helle, 391)

The emergence of the sentence-initial interrogative particle ega (21) was at first preceded by its development into a conjunction (22), which could be placed at the beginning of either a clause or a sentence. Though the particle still occurs in the form of two words eb kaas ~ ep kahs (22) in texts from the beginning of the 17th century, texts from the beginning of the 18th century have the shortened form eiga (23) alongside the currently used ega4 (24).

4 In the old spelling system, a double g indicated a short e in the first syllable.

292 

 Helle Metslang, Karl Pajusalu and Külli Habicht

(22) sina eb pidda sest mittekit errawothma, you neg must.cng this:ela anything:prtv take_away:sup sinna iure pannema. eb kaas mittekit neg ptcl anything:prtv there to put:sup ‘you don’t have to take away anything from there or add something there.’ (COWE\1601Müller, 36/10) üllemb Öppetajat/ eiga (23) Jünger ep olle disciple neg be.cng higher teacher:prtv neg.conj Sullane üllemb omma Jssandat servant higher own lord:prtv ‘A disciple is not above his teacher / nor a servant above his lord.’ (COWE\1715-NewTestament, Mt. 10:24) need Kaarnid/ et nemmad ei külwa/ (24) Pange tähhele pay_attention:imp:2pl art raven:pl conj they neg sow.cng leika; Neil ep olle Kamrid/ egga neg.conj reap.cng they:ade neg be.cng warehouse.prtv egga Aita / ning Jummal toitab neid. neg.conj barn.prtv conj God feed:3sg they:prtv (COWE\1715-New Testament, Lk. 12:24) ‘Consider the ravens: they do not sow, they do not reap, they have no warehouse or barn, and God feeds them.’

The interrogative function of ega developed in the sentence-initial position on the basis of its use as a conjunction. It is indicated by the function of linking with the preceding context or prior knowledge, which has survived in the interrogative word to this day. In modern Estonian, ega is used as marker of negative questions, which are often biased (25). (25) Ega teil seal midagi juhtunud ei ole? neg.q you:ade there anything happen.pptc neg be.cng ‘Is it all right with you over there?’ (1990FICT)

5.1.3 Particles ja, ning We will now proceed to the emerging question particles, which are multifunctional; the coordinative meaning has been preserved, too. The utterances formed with these particles serve as questions. It appears in the following sentences that are response utterances (26), (28), and (29) or continue in turn with the next question (27), (30). The functioning of these questions will be discussed from three different aspects. 1) the shades of meaning of a question depend on the meaning type of opposition, 2) the linking types of a sentence with the preceding comment or text, and 3) the main structural types of questions.

Conjunctive markers of polar questions in Estonian 

 293

Historically there have been two conjunctions – ja and ning – vying for the role of the principal general copulative conjunction. In texts from the 16th−17th centuries it was ning, which grammaticalized into a conjunction from the combination niin-kA of the adverb niin ‘so’ and the clitic -ka with a confirmative meaning. The word ja is a Germanic borrowing in Finnic languages. The conjunction ja occurs for the first time in old Estonian texts in 1600 as a copulative conjunction. In addition, there are some German-inspired uses as an emphatic particle (meanings ‘(and) even’, ‘surely’), where ja is positioned sentence-internally. It appears the conjunction ja became predominant in the written language at the beginning of the 18th century due to the fact that the translators and editors of the New Testament preferred it. In the language of the 20th−21st centuries, ja has been the most common copulative conjunction; ning has remained in the background and is a stylistic synonym for the conjunction ja. A sentence with an additive conjunction, which does not contain any other markers of interrogativity, acts as a question by mainly offering 1) a possible continuation to a list (26), 2) a parallel to the preceding sentence (27), or 3) a possible specifying follow-up to a narrative that developed in the preceding conversation (28), (29). The use of the ja-initial sentence as a question is manifested in dialogues in the following sentence which usually is a response to the presented question; also, it is manifested in the use and perception of the ja-initial sentence as a question in case there is no link with the preceding context. At the same time the marker ja is multifunctional; the additive meaning has also been preserved (unlike the established question particles kas and ega, where the additive meaning has disappeared). (26) [Küti lähedal oli üks väikemõis. ‘Close to Küti there was a small manor.’] Ja park? conj.q park ‘And a park?’ [Park oli ka. ‘There was a park, too.’] (1990FICT) (27) [Kas 10000. posti puhul tuleb samasugune hingeliigutus? ‘On the occasion of the 10,000th posting will there be the same kind of emotion?’] Ja vahepealsetel tuhandetel? conj.q intermediate:pl:ade thousand:pl:ade ‘And during the other many thousands in the meanwhile?’ [Soovime siis õnne, või mis? ‘Congratulations then, or what?’] (FORUM)

294 

 Helle Metslang, Karl Pajusalu and Külli Habicht

(28) [“Ta tõi mulle kallid pärlid ja ehtekiwa.” ‘“He brought me some expensive pearls and gems.”’] “Ja sa wõtsid kõik wastu ja pidasid salajas conj.q you.sg take:pst:2sg all on conj keep:pst.2sg secret oma mehe eest?” own husband.gen for ‘And you accepted everything and didn’t tell your husband about them?’ [“Ma wiskasin talle ta kingitused jalge ette ja katsusin temale wastumeelseks saada.” ‘“I threw the presents at his feet and tried to make myself disgusting for him.”’] (1890NEWS) (29) [Lapsed ootavad minu juures ema ja isa, nemad aga istuvad Pajanõmme soos ega tee lastest asjagi, nagu poleks neid maailmas olemaski. ‘The kids are waiting for their mum and dad at my place, but they are in Pajanõmme Swamp and ignore the kids, as if they didn’t even exist in the world.’] “Ja Armin?” conj.q Armin ‘“And how about Armin?”’ [“Keelas meid, et me välja ei tuleks.” ‘“He forbade us from coming out.”’] (1990FICT)

However, there are also some cases of concessive coordination, which is marked by the corresponding particle, e. g. ometi ‘nevertheless’. (30) [“Aga agressiivsed joodikud on alati ohtlikud, eriti siin turu ligiduses.” ‘“But aggressive drunks are always dangerous, especially here in the neighbourhood of the market.”’] “Ja ometi jalutate te igal öösel koeraga?” conj.q ptcl walk:2pl you every:ade night:ade dog:com poetas Paavo Ralm vahele. slip:pst:3sg Paavo Ralm between ‘“And nevertheless you walk your dog every night?” Paavo Ralm intervened.’ [“Te olete siis märganud?” uuris Anna õhinal. ‘“You have noticed it then?” Anna enquired enthusiastically.’] (1990FICT)

Sentences where ja is the only device indicating interrogativity are divided into two types of sentence structures. 1. One should mention the minor sentence that consists of the particle ja and word form or phrase, such as Ja park? ‘and a park’ Ja vahepealsetel tuhandetel? ‘and during the other many thousands in the meanwhile’, Ja Armin? ‘and Armin’ (26), (27), (29) as a special type. This structure does not unambiguously determine the type of the question it is; the question could be interpreted either as a polar question or a content question. For example, the question Ja Armin? ‘and Armin’ could be understood, depending on the context, as Ja kas Armin istub ja

Conjunctive markers of polar questions in Estonian 

 295

ootab ka? ‘And is Armin sitting and waiting too?’ or Ja mis Armin teeb? ‘And what is Armin doing?’. 2. The second type is the traditional complete sentence, which unambiguously expresses a polar question, for example, Ja sa võtsid kõik vastu? ‘And you accepted everything?’ (28), (30). The first type usually continues a list (26), but there can be some other methods of linking. The second type first and foremost offers a follow-up to the narrative. Old Estonian texts did not reveal any ja-questions, and ning-questions are also rare. Example (31) shows a ning-question of the second type. (31) [Romelt. Siit sadik ollete keik walleks ajanud mis ma räkisin, agga sedda ei woi teie ommeti mitte keelda, et Herra selle kässoga, et ta meid kässib ouna puid istutada ja potida meie ennese aedas, et ta sest kül omma kasso püab. ‘Romelt: From here onwards you have lied about everything that I said, but in fact you cannot prohibit that the master with this order that he orders us to plant apple trees in our own garden that he’s seeking profit from it.’] Hans. Ning puud on meie ennese aedas? Hans conj.q tree:pl be.3 we.gen own.gen garden:ine ‘Hans. And the trees will be in our own gardens?’ [Romelt. Noh! noh! Kül teie sate nähha, kui nemmad hakkawad kandma, siis peame kül kohto ounad maksma. ‘Romelt: Well, well. You will see that when the trees start to bear fruit, then we’ll have to pay our dues in apples.’] (COWE\1812-Luce, 204)

The ja-question does not occur in the old written Estonian; however, as ja became established as a conjunction, it also underwent a functional shift, turning into a question marker. Both types of the ja-question are represented in the corpus of standard Estonian since its first texts, i. e. since the 1890s (28); the ja-question is already rather common in the material of the 1990s and 2000s.

5.2 Adversative coordination 5.2.1 Particle aga In modern Estonian, the conjunction aga ‘but’ acts as a general adversative conjunction – it is used both for contrast and concessive coordination. In addition to aga, the denial of expectation can be expressed also by means of the words kuid ‘but’ and ent ‘but’. The origin of aga is associated with the word aeg