Continuity and Change in Grammar
Linguistik Aktuell/Linguistics Today (LA) Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.
General Editors Werner Abraham University of Vienna / Rijksuniversiteit Groningen
Elly van Gelderen Arizona State University
Advisory Editorial Board Josef Bayer
Christer Platzack
Cedric Boeckx
Ian Roberts
Guglielmo Cinque
Lisa deMena Travis
Liliane Haegeman
Sten Vikner
Hubert Haider
C. Jan-Wouter Zwart
University of Konstanz ICREA/Universitat Autònoma de Barcelona University of Venice University of Ghent University of Salzburg
University of Lund University of Cambridge McGill University
University of Aarhus University of Groningen
Terje Lohndal
University of Maryland
Volume 159 Continuity and Change in Grammar Edited by Anne Breitbarth, Christopher Lucas, Sheila Watts and David Willis
Continuity and Change in Grammar
Edited by
Anne Breitbarth Ghent University
Christopher Lucas Sheila Watts David Willis University of Cambridge
John Benjamins Publishing Company Amsterdam / Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data Continuity and change in grammar / edited by Anne Breitbarth...[et al.]. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 159) Includes bibliographical references and index. 1. Language and languages--Variation. 2. Grammar, Comparative and general. 3. Sociolinguistics. I. Breitbarth, Anne, 1976P120.V37C667 2010 417’.2--dc22 2010014516 isbn 978 90 272 5542 6 (Hb ; alk. paper) isbn 978 90 272 8807 3 (Eb)
© 2010 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
Table of contents List of contributors Introduction: Continuity and change in grammar Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis
vii 1
part i. Continuity What changed where? A plea for the re-evaluation of dialectal evidence Katrin Axel & Helmut Weiß I mpossible changes and impossible borrowings: The Final-over-Final Constraint Theresa Biberauer, Michelle Sheehan & Glenda Newton Continuity is change: The long tail of Jespersen’s cycle in Flemish Anne Breitbarth & Liliane Haegeman
13
35 61
sing the Matrix Language Frame model to measure the extent of U word-order convergence in Welsh–English bilingual speech Peredur Davies & Margaret Deuchar
77
n language contact as an inhibitor of language change: O The Spanish of Catalan bilinguals in Majorca Andrés Enrique-Arias
97
Towards notions of comparative continuity in English and French Remus Gergel
ariation, continuity and contact in Middle Norwegian V and Middle Low German John D. Sundquist
119
145
part ii. Change Directionality in word-order change in Austronesian languages Edith Aldridge
169
Negative co-ordination in the history of English Richard Ingham
181
Continuity and Change in Grammar
Formal features and the development of the Spanish D-system Masataka Ishikawa
201
The rise of OV word order in Irish verbal-noun clauses Elliott Lash
225
The great siSwati locative shift Lutz Marten
249
The impact of failed changes Gertjan Postma
269
A case of degrammaticalization in northern Swedish Henrik Rosenkvist
303
J espersen’s Cycle in German from the phonological perspective of syllable and word languages Renata Szczepaniak
321
n article on the rise: Contact-induced change and the rise A and fall of N-to-D movement Mila Dimitrova-Vulchanova & Valentin Vulchanov
335
Language index
355
Subject index
357
List of contributors Edith Aldridge Department of Linguistics University of Washington A210 Padelford Hall Box 354340 Seattle, WA 98195-4340 USA
[email protected] Katrin Axel Universiteit Göttingen Seminar für Deutsche Philologie Jacob-Grimm-Haus Käte-Hamburger-Weg 3 D-37073 Göttingen Germany
[email protected] Theresa Biberauer Faculty of Modern and Medieval Languages University of Cambridge Sidgwick Avenue Cambridge CB3 9DA UK
[email protected] Anne Breitbarth Universiteit Gent Vakgroep Engels Rozier 44 9000 Ghent Belgium
[email protected] Peredur Davies The Department of Linguistics & English Language University of Wales Bangor Gwynedd LL57 2DG UK
[email protected]
Margaret Deuchar The Department of Linguistics & English Language University of Wales Bangor Gwynedd LL57 2DG UK
[email protected] Andrés Enrique-Arias Universitat de les Illes Balears Departament de Filologia Espanyola, Moderna i Llatina Edifici Ramon Llull Cra de Valldemossa Km 7,5 E-07122 Palma (Balears) Spain
[email protected] Remus Gergel Descriptive & Theoretical Linguistics English Department University of Tübingen Wilhelmstr. 50, R. 407 D-72074 Tübingen Germany
[email protected] Liliane Haegeman Universiteit Gent Vakgroep Engels Rozier 44 9000 Ghent Belgium
[email protected] Richard Ingham The School of English Birmingham City University City North Campus Perry Barr Birmingham B42 2SU
[email protected]
Continuity and Change in Grammar Masataka Ishikawa Sogokagakubu Hiroshima University Kagamiyama 1-7-1 Japan 739-8521
[email protected] Elliott Lash Faculty of Modern and Medieval Languages University of Cambridge Sidgwick Avenue Cambridge CB3 9DA UK
[email protected] Lutz Marten Department of Africa School of Oriental and African Studies Thornhaugh Street Russell Square London WC1H 0XG UK
[email protected] Glenda Newton Faculty of Modern and Medieval Languages University of Cambridge Sidgwick Avenue Cambridge CB3 9DA UK
[email protected] Gertjan Postma Meertens Instituut Postbus 94264 1090 GG Amsterdam
[email protected] Henrik Rosenkvist Centre for Language and Literature Lund University Box 201 221 00 Lund Sweden
[email protected]
Michelle Sheehan School of English Literature, Language and Linguistics Percy Building University of Newcastle upon Tyne Newcastle upon Tyne NE1 7RU UK
[email protected] John Sunquist Department of Foreign Languages 640 Oval Drive Stanley Coulter 170 West Lafayette, IN 47907 USA
[email protected] Renata Szczepaniak Institut für Germanistik I Von-Melle-Park 6 20146 Hamburg Germany
[email protected] Valentin Vulchanov Institutt for moderne fremmedspråk NTNU 7491 Trondheim Norway
[email protected] Mila Vulchanova Institutt for moderne fremmedspråk NTNU 7491 Trondheim Norway
[email protected] Helmut Weiß Institut für Kognitive Linguistik Johann Wolfgang Goethe-Universität Grüneburgplatz 1 (Fach 161) 60629 Frankfurt a. M. (Briefe) 60323 Frankfurt a. M. (Pakete) Germany
[email protected]
Introduction Continuity and change in grammar Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis Ghent University/University of Cambridge
The present volume brings together 16 contributions selected from papers presented at the conference on Continuity and Change in Grammar that took place at the University of Cambridge 18–20 March 2008. The aim of the conference was to foster an exchange of ideas on various aspects of linguistic transmission in different frameworks (e.g. generative vs. functionalist) in order to enhance our understanding of what makes languages change and what in turn prevents them from changing. This volume is particularly timely, as there has been much interest recently in the topic of continuity, beginning with the publication of Longobardi’s (2001) and Keenan’s (2002) articles on ‘inertia’ as a principle of, or rather, a constraint on language change.
1. Approaches to continuity and change In addressing change and continuity, different views can be taken as to what constitutes a ‘change’. From the point of view of the ‘Inertia Principle’, syntactic change should, in principle, always be reducible to the categorial or structural reanalysis of lexical items following changes in phonology or lexical semantics (plus, presumably, language-external factors, such as contact): Syntactic change should not arise, unless it can be shown to be caused – that is, to be a well-motivated consequence of other types of change (phonological changes and semantic changes, including the appearance/disappearance of whole lexical items) or, recursively, of other syntactic changes, given a plausible theory of Universal Grammar (UG) and language acquisition. (Longobardi 2001: 278)
This kind of perspective sees the very existence of syntactic change as conflicting with an ideal model of a ‘perfect’ syntactic component which is replicated identically by each successive generation. Thus, in applications of Longobardi’s theory, it is not the functional structure of the clause that changes, but the filling of the terminal nodes – by reanalyses of material subject to phonological and semantic change.
Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis
Other approaches highlight the ways in which such reanalyses occur and therefore where the lexical material filling terminal nodes comes from historically; for instance, an element that used to realize a lower functional head (or its specifier) and move to a higher position may later come to realize the higher position by loss of movement as part of reanalysis (Roberts & Roussou 2003). Then there are approaches which seek to model the properties of linguistic change mathematically. Typical changes (not merely syntactic ones) follow an S-shaped curve, as the replacement of an old form by a new one proceeds slowly at first, then gathers speed, and slows down again as it nears completion (Bailey 1973; Kroch 1989). Since the work of Kroch, it has become standard to model this property of (linguistic) change by using the logistic function. In the present volume, Postma proposes refinements to this that allow him to capture the fact that the peak of a failed change coincides with the turning point of a related S-curve. Also relevant to this volume are approaches which see grammaticalization as an important driving force for change, and which see unidirectionality as a definitional property of grammaticalization (e.g. Haspelmath 1999). While it is clear that changes that involve more lexical items becoming more functional are extremely common, there is growing evidence to suggest that changes from more functional to more lexical are not as rare as has been thought. Two further instances of this latter type of change are discussed in the contributions by Marten and Rosenkvist. Many of these different approaches have in common that they see language acquirers as the agents of syntactic change, whether these are infants acquiring their first language or older children and adults acquiring a second language. Children acquiring their first language have to abduce the grammar of that language on the basis of indeterminate evidence from their linguistic environment, and may therefore arrive at a different analysis to previous generations of a given underlying structure. This can be explained as coming about due to surface ambiguity in the realization of that structure, arising for instance through morphological levelling (Andersen 1973; Timberlake 1977; Kroch 1989; Lightfoot 1991, 1999). However, reanalyses of this type can be hard to identify until their effects become evident by spreading beyond the original context of the reanalysis itself (cf. Timberlake 1977; Harris & Campbell 1995, among others). Aldridge’s chapter in the present volume describes such a case, where surface ambiguity leads to structural reanalysis, as a former topic in SpecCP is reanalysed as a subject in SpecTP. The ‘spreading’, or actualization, of this change is witnessed by an increase in the A-properties and a decrease in the A′-properties of the elements involved. On the other hand, the trigger experience of first language acquirers may also be altered by previous ‘imperfect’ language acquisition on the part of adult second language learners. The articles collected in this volume that address contact as a source of
Introduction
linguistic change make it clear that the degree of bilingualism in a given contact situation plays an important role in determining the type and quality of the change. First of all, it is clear that contact tends not to lead to significant structural change unless it is intensive and long-term, whereas lexical borrowing seems to be possible even in cases of minimal contact; cf. Thomason and Kaufman’s (1988: 50) ‘scale of interference’. Illustrating this point, Sundquist (this volume) argues that Middle Low German had no influence on the syntax of Middle Norwegian during the time of the Hansa, where the contacts were restricted to the domain of business and trade, and where the languages were similar enough for speakers to accommodate to each other’s word order by using variants available within their own language if necessary. However, it seems that situations of intensive and extensive contact, where many children simultaneously acquire two first languages, can also be environments for (syntactic) continuity. Paradoxical as this may seem, this follows from the fact that children have little difficulty differentiating and acquiring multiple languages in their environment, and can arrive at grammars of those languages which do not differ systematically from those of their monolingual peers (cf. Meisel 1995, 2001). This is illustrated by Davies and Deuchar’s contribution, which shows that there is very little evidence that the extensive Welsh–English bilingualism found in much of Wales has led the syntax of Welsh to converge with that of English. Vulchanov & Vulchanova suggest that Old Bulgarian (spoken in the ninth and tenth centuries) was influenced by contact with New Testament Greek. Although they do not discuss this within the confines of their chapter, this raises the intriguing question of whether transfer of basic syntactic properties can proceed entirely through the written medium or whether vernacular (rather than elite) bilingualism is required for transfer of syntactic features. In this case, they prefer to suggest contact via the New Testament Greek Gospel translations, rather than making primary reference to the extensive bilingualism between Slavonic varieties and Byzantine Greek in the Balkans. Such a view implicitly rejects child language acquisition as the locus of syntactic change. The absence of or delay in syntactic change in cases where it would be expected, for example on the basis of other languages having undergone a parallel development under otherwise similar conditions, is perhaps expected under an ideal model of first language acquisition, in which the language system of the parent generation is perfectly replicated in the acquiring generation. But ideal models are empirically inadequate. For instance, most languages are in reality exposed to some form of language contact. Already the existence of dialects is enough to create minimal but potentially significant changes in the trigger experience for language acquirers. Given the ubiquity of syntactic change, it is rather surprising to find the absence or at least delay of a change. One of the questions behind the present collection was therefore which factors
Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis
might prevent or impede a change that might be expected on the basis of otherwise similar input conditions. Biberauer, Newton and Sheehan in their contribution, for instance, argue for a universal constraint on certain types of structures, which implies restrictions on possible changes. Sundquist and Enrique-Arias in turn explore the possibility of language contact as a factor in delaying syntactic change. In keeping with the two major themes of the volume, the contributions are divided into two broad sections, the first addressing continuity, and the second addressing change.
2. Continuity The contributions to this part of the volume address problems of (apparent) syntactic continuity. In What changed where? A plea for the re-evaluation of dialectal evidence, Katrin Axel and Helmut Weiß make a case for studying linguistic change in dialects rather than standard languages. They present a number of case studies, focussing mainly on the development of pro-drop in the history of German (dialects). They show that the alleged loss of pro-drop during the Old High German period has only affected the standard language; many dialects show (partial) pro-drop to this day, licensed by c-commanding Agr in C. The only change is the innovation of the present-day restriction to pronominal Agr. Similar continuity at the dialectal level, Axel & Weiß argue, is found in the domain of negation (negative concord is preserved in several dialects) or (with qualifications) in the order of verbs within the verb cluster. Continuity at the level of the standard language, on the other hand, is found in possessive constructions, where the standard language preserves post-nominal genitive possessors. In Impossible changes and impossible borrowings: the Final-over-Final Constraint, Theresa Biberauer, Michelle Sheehan and Glenda Newton address the implications for syntactic change of a principle argued to restrict word–order possibilities across languages, the Final-Over-Final-Constraint (FOFC) (Biberauer, Holmberg & Roberts 2007). As the FOFC predicts certain word orders to be impossible, ruling out in particular head-final higher functional projections when lower projections are head-initial, change and borrowing are predicted to be constrained. The authors demonstrate in a number of case studies that changes from head-final to head-initial do respect the FOFC in that higher heads change first and that the presence of initial heads blocks the borrowing of final heads. Anne Breitbarth and Liliane Haegeman’s contribution, Continuity is change: The long tail of Jespersen’s cycle in Flemish, discusses an apparent case of syntactic continuity, the (at least optional) maintenance of the old preverbal negation marker e(n) in the Flemish dialects of Dutch. They argue that its survival is not due to these dialects
Introduction
going through Jespersen’s Cycle at a much slower pace than the other West Germanic languages, which have long lost the cognates of e(n), but rather to its reanalysis as a marker of (affective) polarity which has left the surface word order unchanged. Evidence for this comes from the pragmatic restrictions on the use of e(n) in the Flemish dialects and its availability in non-negative affective contexts. In their chapter, Using the Matrix Language Frame model to measure the extent of word-order convergence in Welsh–English bilingual speech, Peredur Davies and Margaret Deuchar aim to test the hypothesis that ongoing ‘convergence’ (contactinduced transfer) can be identified in bilingual code-switching speech, in this case the speech of Welsh–English bilinguals. In a corpus of Welsh with extensive codeswitching with English, they look for instances where the matrix language is Welsh, but the grammatical system used is unambiguously English. Any such instances are viewed as evidence of ongoing convergence of Welsh towards English. They investigate this with respect to word order within the nominal domain. Using their methodology, they identify only one possible instance of convergence (modifier–head order within noun phrases in this case), and conclude that Welsh noun phrases manifest continuity rather than change in the face of extensive bilingualism and contact with English. Andrés Enrique-Arias’ contribution, Contact-induced conservativism in the Spanish of Catalan bilinguals in Majorca, addresses the topic of grammatical continuity from the perspective of contact between Spanish and Catalan in Majorca. The principal claim is that, while contact is predominantly seen as a force for grammatical change, under the right circumstances it can, in fact, serve to prevent otherwise expected innovations in a given language. Enrique-Arias argues that the presence of parallels in Catalan for a number of structures which are obsolescent in general Spanish results in their retention in the Spanish spoken and written by Catalan native speakers in Majorca. In his chapter Comparative continuity, Remus Gergel investigates comparative clauses in English and French with primary focus on subject–verb inversion. He explores the derivation and diachronic evolution of these clauses and argues that they instantiate syntactic continuity in that they feature a ‘low subject’, that is, one that is not raised out of the verb phrase – an option that used to be much more widespread in both languages. On this analysis surface subject-verb inversion is achieved merely by the expected raising of the verb to the inflectional domain and not on to the complementizer domain, as in the standard generative analysis of comparative inversion. John Sundquist looks at Variation, continuity, and contact in Middle Norwegian and Middle Low German and concludes that the patterns of OV/VO variation in Middle Norwegian were not influenced by contact with Middle Low German. He attributes
Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis
this to the particular sociolinguistic character of this contact – restricted to business contacts in the context of Hanseatic trade.
3. Change Edith Aldridge’s chapter Directionality in word-order change in Austronesian languages reconstructs how SVO order in Austronesian languages has evolved from the older VOS order via topicalization of the subject. Absolutive arguments in Austronesian VOS languages are argued to end up in final position by moving the absolutive to a topic position and then fronting the remnant TP around it. Diachronically, SVO develops when only the first step of deriving VOS in that fashion is taken and the fronted topic is reanalysed as a subject in SpecTP. By comparing several Austronesian SVO languages, Aldridge shows that this is a gradual process as subjects across these languages show combinations of Aand A′-properties. In his contribution, Negative co-ordination in the history of English, Richard Ingham looks at the syntax of nor in the history of English, arguing that changes can be derived from more general changes in the syntax of negation. While the pattern affirmative clause + nor + negative clause is widely attested in Middle English, English today allows nor only where both conjoined clauses are negative. Conversely, the clause introduced by nor once required additional negative elements (negative concord), while, today, nor may act as the sole negator. Ingham accounts for this by proposing two changes in the grammar of English: the replacement of a null negative operator (NegOP) in SpecNegP by not and subsequently the loss of NegP in Early Modern English as n-words such as no become inherently negative. Initially nor has an uninterpretable [uNeg] feature checked by NegOP (giving rise to negative concord); with the loss of NegOP, it checks its [uNeg] feature from a c-commanding negative clause (hence the preceding clause must be negative); finally, nor itself becomes iNeg and cannot co-occur with other negative elements (loss of negative concord). Masataka Ishikawa’s chapter, A feature-driven analysis of syntactic change: A case study in the history of articles in Spanish, offers a minimalist analysis of the rise of the definite article from Old to Modern Spanish, arguing for a two-stage development of the structure and phasehood of the Spanish DP through a combination of grammarinternal and grammar-external factors. The first stage consists of the creation of a ‘double d-shell’ structure, which, while computationally more complex, is motivated by the (diachronic) pressure for a one-to-one assignment of formal features to functional heads. By contrast, in the second stage, in which the inner dP acquires phasehood, Ishikawa suggests ease of processing in acquisition as a possible motivation. Elliott Lash, looking at The rise of OV word order in Irish verbal-noun clauses, proposes a diachronic account of the emergence of the exceptional object–verb (OV)
Introduction
word order found in Irish nonfinite clauses, an otherwise regularly VSO language. He posits a series of reanalyses involving the preposition do ‘to, for’ plus a nonfinite verb, with noun phrases originally in a main clause being reanalysed as subjects or objects of an embedded clause. He associates the final emergence of OV clauses with a change in information stucture: the preverbal noun phrase in the order O do V was once fronted for reasons of focus, but that motivation was lost with reanalysis and the fronting became purely syntactic, as it is today. The preposition do in these structures became first a focus-marking particle, and then a nonfinite marker, with various dialectal variations. In his chapter The great siSwati locative shift, Lutz Marten presents a historicalcomparative analysis of the locative morphosyntax of the southern Bantu language siSwati. He argues that the Proto-Bantu locative noun class prefix ku- has degrammaticalized to become a preposition in siSwati. This position is supported with a range of syntactic evidence, such as the fact that noun-class agreement with a ku-marked noun is with the class of the noun itself, and demonstratives can intervene between ku- and a noun, both of which are unexpected if ku- is a noun class marker rather than a preposition. Thus we have a rather robust case of degrammaticalization, and the fact that this change is shown to be natural within the context of an overall reanalysis of the siSwati locative system arguably reinforces its status as a counterexample to the claimed unidirectionality of grammaticalization. In The impact of failed changes, Gertjan Postma establishes a mathematical relation between successful and failing linguistic changes. Based on two case studies, he argues that the latter can be understood as extra-grammatical variants introduced by L2 learners of a language which are not retained in subsequent generations, but which trigger successful changes that are. He argues that, under the logistic model commonly used to model successful changes which describe an S-curve, such unsuccessful changes can be shown to be mathematically correlated, as they can be modelled as the first derivative of the logistic function. Henrik Rosenkvist considers A case of degrammaticalization in northern Swedish dialects. These dialects have a modal verb bö ‘need, must, ought to’ which Rosenkvist argues has developed from the prefix of the verb behöva ‘need’. It is linked to the fact that these dialects also allow the prefix be- (bö- in these dialects) of the verb behöva ‘need’ to detach itself from the verbal root to give discontinuous word orders be X höva. He regards this as a case of the reanalysis of a prefix as an auxiliary which involves several changes in status (prefix > full word and prefix > auxiliary) that run counter to the developments expected in grammaticalization. As such it would fit into an existing known category of degrammaticalizations of affixes as independent words (cf. Irish agreement suffix -muid > pronoun ‘we’, Doyle 2002). Renata Szczepaniak looks at Jespersen’s Cycle in German from the phonological perspective of syllable and word languages. She argues that the replacement of the
Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis
old preverbal negator ne/en by a new postverbal one (niht) during the Middle High German period was triggered, or at least encouraged, by the simultaneous prosodic restructuring of the language from a syllable language to a word language, resulting in the loss of unstressed pretonic syllables such as the old preverbal negator ne/en. In An article on the rise: Contact-induced change and the rise and fall of N-to-D movement, Mila Vulchanova and Valentin Vulchanov examine the evolution of the postposed definite article in Bulgarian, a well-known feature of the Balkan linguistic area. They argue that, at one time, definiteness was marked either by head movement of N to D or by insertion of a determiner clitic in D. Historically, the determiner changed from a phrase (a specifier of DP) to a head merged directly in D, a shift in the expression of a functional category from Move > Merge in the sense of Roberts & Roussou (2003). They suggest that, while the article evolves syntactically using preexisting material, the development is assisted by contact with Greek, which had a longestablished article.
4. Conclusion One of the central tasks of historical linguistics has generally been seen as solving the actuation problem: ‘Why do changes in a structural feature take place in a particular language at a given time, but not in other languages with the same feature, or in the same language at other times?’ (Weinreich, Labov & Herzog 1968: 102). We hope that the individual contributions to this volume will help to elucidate both parts of this problem. They address the first part by investigating various aspects of why grammars change: such familiar issues as what processes need to be identified as the drivers of change, the role of children in change, and the role of syntax-external factors in triggering syntactic change (both phonological or lexical, and in the form of language contact). We also hope that they offer in part some new perspectives on the second, often neglected aspect of the actuation problem: why changes do not occur in other parallel languages. Progress here necessarily focuses more on the constraints on change, on factors which limit change, and on the role of contact in supporting conservative linguistic features.
Acknowledgements Our thanks go the various reviewers of the volume chapters, as well as to the series editor Werner Abraham and to Kees Vaes at Benjamins. We are indebted to various people and organizations for help with the Conference on Continuity and Change
Introduction
in Grammar: the scientific committee, Ian Roberts, Wim van der Wurff and Hedde Zeijlstra; the various other people who helped with the day-to-day organization of the conference: Deborah Anderson, Alastair Appleton, Yi-An Lin and George Walkden; to Andrew Winnard at Cambridge University Press, and to the Faculty of English and Selwyn College for providing a venue. The conference formed an integral part of the research project The Development of Negation in the Languages of Europe, which was possible due to the generosity of the Arts and Humanities Research Council (Research grant award no. AR119272), who provided funding to bring the invited speakers to Cambridge.
References Andersen, Henning. 1973. Abductive and deductive change. Language 49: 765–93. Bailey, Charles-James. 1973. Variation and Linguistic Theory. Washington DC: Center for Applied Linguistics. Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2007. Disharmonic word-order systems and the Final-over-Final-Constraint (FOFC). In Proceedings of XXXIII Incontro di Grammatica Generativa, Antonietta Bisetto & Francesco Barbieri (eds), 86–105. Bologna: Università di Bologna. Doyle, Aidan. 2002. Yesterday’s affixes as today’s clitics. In New Reflections on Grammaticalization [Typological Studies in Language 49], Ilse Wischer & Gabriela Diewald (eds), 67–81. Amsterdam: Benjamins. Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: CUP. Haspelmath, Martin. 1999. Why is grammaticalization irreversible? Linguistics 37(6): 1043–68. Keenan, Edward L. 2002. Explaining the creation of reflexive pronouns in English. In Studies in the History of English: A Millennial Perspective, Donka Minkova & Robert Stockwell (eds), 325–355. Berlin: Mouton de Gruyter. Kroch, Anthony S. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Lightfoot, David W. 1991. How to Set Parameters: Arguments from Language Change. Cambridge MA: The MIT Press. Lightfoot, David W. 1999. The Development of Language: Acquisition, Change and Evolution. Oxford: Blackwell. Longobardi, Giuseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The history of French chez. Linguistic Inquiry 32: 275–302. Meisel, Jürgen. 1995. Parameters in acquisition. In A Handbook of Child Language, Paul Fletcher & Brian MacWhinney (eds), 10–35. Oxford: Blackwell. Meisel, Jürgen. 2001. The simultaneous acquisition of two first languages. Early differentiation and subsequent development of grammars. In Trends in Bilingual Acquisition [Trends in Language Acquisition Research 1], Jasone Cenoz & Fred Genesee (eds), 11–41. Amsterdam: John Benjamins.
Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Thomason, Sarah & Kaufman, Terrence. 1988. Language Contact, Creolization, and Genetic Linguisitcs. Berkeley CA: University of California Press. Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Mechanisms of Syntactic Change, Charles N. Li (ed.), 141–177. Austin TX: University of Texas Press. Weinreich, Uriel, Labov, William & Herzog, Marvin I. 1968. Empirical foundations for a theory of language change. In Directions for Historical Linguistics, Winifred P. Lehman & Yakov Malkiel (eds), 95–195. Austin TX: University of Texas Press.
part i
Continuity
What changed where? A plea for the re-evaluation of dialectal evidence* Katrin Axel & Helmut Weiß
University of Göttingen/Goethe-Universität, Frankfurt-am-Main In the field of generative diachronic syntax, it has often been disregarded at which level of the language (dialect or Standard) syntactic change has occurred. However, just as in the case of phonological developments, the syntax often (though not always) turns out to be more conservative at the dialectal level. In this chapter we will present four cases studies on the syntax of German: the diachrony of pro-drop (null subjects), of negative concord, of possessive constructions and of word-order changes in the verbal cluster. Our plea for taking into account dialect data in historical linguistics converges with the growing significance dialects and dialectal data have gained within theoretical linguistics.
1. Introduction It is commonly assumed in generative linguistics that language change – and thus continuity (i.e. the non-occurrence of change) as well – is largely due to first language acquisition (L1 acquisition).1 Therefore, it would be quite natural to investigate language change only in languages (or dialects, varieties) which were transmitted from generation to generation by L1 acquisition alone.2 Contrary to expectation, this is not really the case. Research on language change within the generative framework is still mostly restricted to developments in languages which have undergone standardization. There are two misconceptions which can presumably be made responsible for this curious state of affairs. Firstly, based on the – no doubt – correct assumption that I-language is the primary object of investigation of linguistic research, generative linguists refuse to discuss what kind of objects languages like English or German are
*We are grateful to the audience at the conference on ‘Continuity and Change in Grammar’ for their comments. Thanks also to two anonymous reviewers and to the editors for their helpful suggestions. Special thanks to Sheila Watts for her assistance and many insightful comments. . Cf. also already Paul (1880). . There are other types of language change, e.g. contact-induced language change (Kroch 2000; Weiß 2005b).
Katrin Axel & Helmut Weiß
(though they use the terms as convenient abbreviations, for example, speaking of the loss of negative concord in English). Secondly, it is often assumed, that “standard languages are simply dialects which have gained a different [i.e. higher] social prestige” (Cornips & Poletto 2005: 939). Both these assumptions lead to the curious fact just mentioned that generative historical research is usually concerned with developments in standard languages.3 Not to deal with the concept of language may be a pardonable neglect (though it seems to be a little strange to leave such an essential term undefined, cf. Weiß 2009). However, to hold standard languages as nothing more than dialects is simply false. Standard languages are not simply codified forms of spoken and L1-acquired dialects, they almost always evolved as secondarily learned, written languages. Standard German may be an extreme case because it evolved out of several dialects and has never corresponded to any individual one, but, at least initially, all standard languages develop relatively independently of L1-acquisition (Weiß 2001, 2004a, 2004b, 2005b, 2005c). In Weiß (2001) it is therefore proposed that there are two kinds of natural languages based on the criterion of language acquisition. According to this theory, first order natural languages (N1 languages) are those which are acquired in the course of L1 acquisition, whereas second order natural languages (N2 languages) are learned, that is acquired mostly by instruction (e.g. in school). Dialects (and languages) as defined in Weiß (2001) are N1 languages, and standard languages are N2 languages, or at least used to be in the past.4 Given the fundamental role which L1-acquisition plays in modern historical linguistics, it follows naturally that dialects should be the objects of investigation. A second consequence for historical linguistics is that whenever there is a different development in dialects and standard languages, it is the development in the dialects
. There are a couple of notable exceptions: Poletto (1995) and Ledgeway (2000) diachronically investigate various aspects of Italian dialect syntax. Kroch & Taylor (1997) deal with the verbsecond property in different historical dialects of English. Sprouse & Vance (1999) discuss the null-subject properties of various Romance and Germanic languages from a diachronic perspective. Moreover, the work on creole languages should be mentioned here (cf. deGraff 1999 for an overview). Thanks to an anonymous reviewer for pointing this out. However, as far as German is concerned, there is hardly any generative work on historical syntax so far (but see Lenerz 1984; Axel 2007, Jäger 2008). Non-generative diachronic studies have been primarily concerned with syntactic developments leading to the standard; this in contrast to work in phonology, which has always taken into account the situation in the dialects. . Nowadays, standard languages are subject to L1 acquisition to a much greater extent than was the case in the past (Weiß 2004a,b, 2005c). These ‘re-naturalized’ languages then can develop new dialects as well as colloquial variants (e.g. social registers). This is, for example, the case with Standard English which has even developed several national varieties in Great Britain, America, Australia and so on (Durrell 1999: 299).
What changed where?
which should gain our attention as historical linguists,5 because these changes can be largely ascribed to language-internal factors alone, whereas standard languages have also been formed by extra-linguistic forces. In what follows we will present a case study on pro-drop in German and several other case studies in which changes in dialects differ from those in the standard. All these case studies show that taking into account dialectal evidence in historical syntax can lead to different and more appropriate results. These results constitute a more solid basis for judgments about what is stable and what is more open to change in the history of a language. This plea for taking into account dialect data in historical linguistics converges with the growing significance dialects and dialectal data have gained within theoretical linguistics. Besides their greater naturalness (see above), it is above all the issue of microvariation that makes the comparative study of dialects an invaluable source for linguistic research. In the line of Kayne’s (1996) concept of microvariation, dialects provide a unique data source for investigating what is possible or impossible in natural languages (Benincà & Poletto 2007).
2. Case study 1: Pro-drop in German Our major case study is on pro-drop in German. It is well-known that in Modern Standard German (referential) subject pronouns have to be overtly realized: (1) Kaffee magst *(du) nicht. coffee like you not ‘You don’t like coffee.’
However, modern German dialects do license pro-drop under certain conditions and the same is true for all the major dialect groups within the Continental WestGermanic group. In the previous literature, it has often been claimed that German lost the pro-drop property at a very early stage, even though there is a large consensus that ProtoGermanic and Proto-Indo-European were pro-drop languages. Paul (1919: 22) states that referential null subjects died out before the earliest attested stages of Old High German (OHG). Hopper (1975: 31), following Meillet (1908/1909: 89), speculates that the use of overt subject pronouns is the rule already in early Germanic. In the Mittelhochdeutsche Grammatik (Paul 2007: 315) it is claimed that in more recent stages of the language inflectional endings are not capable of ‘expressing’ the subject anymore, thereby implying that this was possible in older stages. What has been almost totally ignored is
. To the extent that there is sufficient documentation of the dialects. There is, of course, the problem that there are many dialects which have barely received syntactic description.
Katrin Axel & Helmut Weiß
that there are many dialects that still allow subject omission in certain morphosyntactic environments even in contemporary German.
2.1 Pro-drop in the modern dialects Though the Continental West Germanic dialects are only partial pro-drop languages in the sense defined in Weiß (2001), they nonetheless exhibit some indisputable instances of pro-drop (e.g. Bayer 1984; Weiß 1998: 116ff. on Bavarian, Haag-Merz 1996: 153ff. on Swabian, Cooper 1995 on Zurich German). In Bavarian, for instance, pro-drop is possible in the 2nd sg., (2), and 2nd pl. (cf. also Bayer 1993; Fuß 2005): (2) a.
moang bisd pro wieda gsund tomorrow are-2.sg again healthy ‘Tomorrow, you will be well again.’
(Bavarian)
b. wennsd pro moang wieda gsund bisd if-2.sg tomorrow again healthy are-2.sg ‘if you will be well again tomorrow.’
As can be seen in (2b), Bavarian shows so-called complementizer agreement: the complementizer wenn carries the inflectional morpheme -sd and thus agrees with the finite verb at the end of the clause. In subordinate clauses pro-drop is parasitic on complementizer agreement. In the 1st sg. there is no complementizer agreement in Bavarian. The subject pronoun has to be realized overtly in this case: (3) wenn-*(e) moang wieda gsund bin if-I tomorrow again healthy am ‘if I will be well again tomorrow.’
(Bavarian)
So we can conclude that the presence of overt Agr-in-C is necessary for pro-drop. Whether Agr-in-C is the result of verb raising to C or of inflected complementizers seems to be irrelevant. However, Agr-in-C alone does not suffice to license pro-drop. Consider the following contrast, which, according to Weise (1900: 56), shows up in the Thuringian dialect spoken in Altenburg:6 (4) a. schreib mir einmal den Brief, write-imp-2.sg me once the letter,
kriegst pro auch einen Groschen get-2.sg too a penny
‘Write me a letter and you will get a penny.’
(Thuringian)
. Example (4a) is taken from Weise (1900: 56). Example (4b) is constructed in accordance with Weise’s remark that ‘In the plural, this ellipsis of the subject pronoun is not attested’ (“Im Plural ist diese Ellipse [des Subjektspronomens] nicht nachweisbar” (Weise 1900: 56).
What changed where?
b. schreibt mir einmal den Brief, dann kriegt write-imp-2.pl me once the letter, then get-2.pl
*pro/ihr auch einen Groschen pro/you-2.pl too a penny ‘Write me a letter and you will get a penny.’
In this dialect pro-drop is possible with the 2nd sg., and not with the 2nd pl., as the contrast between (4a) and (4b) reveals. The difference between the inflectional marker -st in the 2nd sg. and the -t in the 2nd pl. is that only the former contains a pronominal element, namely /t/. As the OHG data show, the ending of the 2nd sg. (= -st), was originally only -s. The form with an additional -t originated from cases where the verb was followed by the pronoun thu, du in OHG times. In this configuration the word boundary was then reanalysed in such a way that the dental plosive became part of the verbal ending. As Weise (1907: 204) already notes, subject pronouns can be absent in the 2nd sg. and pl. in exactly those Bavarian and East Franconian dialects where complementizer agreement is pronominal (5).7 (5) a.
wennsd pro mogsd if-2.sg like-2.sg ‘if you like’
b. wennds8 pro megds if-2.pl like-2.pl ‘if you like’ (Central Bavarian, likewise North Bavarian and East Franconian)
In East Franconian, the 1st and 3rd pl exhibit non-pronominal complementizer agreement (6), and thus the subject pronoun cannot be dropped. (6) a. waaln *(mer) graad besamn senn because-1.pl we at-the-moment together are-1.pl ‘since we are together now.’ b. waaln *(se) graad besamn senn because-3.pl they at-the-moment together are-3.pl ‘since they are together now’ (East Franconian, Weise 1907: 204)
. When we speak of (non-)pronominal agreement, we do not imply that speakers can distinguish between the two because that would require information about the historical development which speakers normally do not have. Speakers are only capable of detecting that the one variant of agreement allows for pro-drop and the other one does not. . This inflectional suffix developed from the enclitic pronoun es parallel to the 2nd sg. suffix -sd.
Katrin Axel & Helmut Weiß
On the basis of the data discussed here (and comparable ones, cf. Weiß 1998, 2005a) we can postulate the following licensing principle for pro-drop in modern German dialects: (7) Licensing principle for pro-drop: pro must be c-commanded by pronominal agreement.
It should be noted that pro-drop and inflected complementizers are by no means restricted to Upper German dialects like Bavarian and East Franconian. It is attested in all major groups of modern German dialects, especially in Eastern Middle German dialects, but for the 2nd sg. even in Low German dialects like Low Saxon and Low Franconian (cf. Weiß 2005a).
2.2 Pro-drop in early German It is generally assumed that the pro-drop property inherited from Proto-Germanic was already largely lost in the OHG period, that is, in the oldest attested stage of the German language. The introduction of overt subject pronouns in pre-/early OHG has traditionally been linked to the weakening of verbal endings due a phonological weakening process which in turn has been argued to have been a consequence of the introduction of word-initial accent (see Kögel 1882: 127; Held 1903: XIII; Behaghel 1928: 442). However, as Grimm (1967 [1898]: 235) and later Eggenberger (1961) already objected, these phonological and morphological changes did not lead to a substantial leveling of inflectional distinctions during the OHG period. It should furthermore be noted that from a cross-linguistic perspective even the Standard Present-Day German paradigm is usually considered sufficiently strong to license pro-drop (Jaeggli & Safir 1989).9,10 The absence of pro-drop has thus been attributed to an incompatibility of the pro-drop and the verb-second (V2) parameters (Jaeggli & Safir 1989; Rohrbacher 1999). And what is even more problematic: we have seen that at a dialectal level the alleged loss of pro-drop has not occurred even though here the verbal paradigms have the same amount of syncretism or even more than in the standard language.
. This despite the fact that in the Present-Day German paradigm there is more syncretism than in OHG. For example, the OHG verbal paradigm attested in the OHG Tatian (c. 830 AD, East Franconian) has six different endings in the present indicative (-u, -is(t), -it, -emēs, (-ēn), -et, -ent. There is thus no syncretism at all. By contrast, in the modern Standard German paradigm (-e, -st, -t, -en, -t, -en) there is some syncretism. In the plural, the first and third person have identical endings. Also, both the 3rd sg. and the 2nd pl. end in -t. So indeed the verbal paradigm has been weakened. . However, more recently the opposite claim has been propagated (e.g. Müller 2006; Koeneman 2007).
What changed where?
As mentioned above, it is generally assumed that in OHG (referential) pro-drop has already been largely lost. However, there are numerous instances of pro-drop in the eighth and ninth century texts: (8) a.
Druhtin ist auh Lord is also ‘He is also the Lord.’
b. Sume hahet in cruci some-acc hang-2.pl to cross ‘Some of them you will crucify.’
et ex illis … crucifigetis,
c.
quidis zi uns thesa parabola say-2.sg to us this parable ‘Are you telling this parable to us?’
ad nos dicis hanc parabolam
(I 426; OHG)
(MF XVIII,17)
(T 529,2)
In the traditional literature, the standard approach towards the omission of referential subject pronouns in earlier OHG is to trace it back to a special type of ‘loan syntax’ (see Eggenberger 1961; Hopper 1975). The central idea is that the subject omissions have been ‘imposed’ on these OHG translated texts through the process of producing narrow – or even ‘slavish’ – translation from the Latin originals and should not be considered a native part of the OHG grammar. As is argued in Axel (2005, 2007: ch. 6) the loan-syntax hypothesis is not convincing. It cannot account for the fact that null-subject distribution in the Old High German is clearly governed by (morpho-)syntactic factors that do not play a role in the Latin originals. For example, as has already been observed by Eggenberger (1961), there is a clear main/subordinate asymmetry. While in main clauses subject pronouns are often omitted parallel to the Latin, they are almost always inserted contrary to the Latin in subordinate clauses. Axel argues that the main/subordinate asymmetry can be derived without further assumptions if we assume that the OHG null subjects were only licensed in post-finite position, see (9).11 (9) a. [CP Druhtini [C istj] pro auh ti tj ] b. [CP Sumei [C hahetj] pro ti in cruci tj ] c. [CP [C quidisj] pro zi uns thesa parabola tj ]
. It should be noted that – like modern German – OHG already largely exhibited the properties of an asymmetric V2 language: in subordinate clauses introduced by complementizers or interrogative/relative phrases the finite verb stayed in its base position at the end of the clause (the basic word order in the VP was OV). However, in (unintroduced) main clauses it moved to C0, and in declaratives there was usually also XP movement to SpecC (cf.Axel 2007).
Katrin Axel & Helmut Weiß
Thus, just like in the modern dialects, OHG null subjects were only licensed in a configuration where they were c-commanded by Agr in C. In OHG the only way to obtain the configuration in (10) was via verb movement to C0: (10) [CP [C V+AGRi ]k …. proi tk]
In OHG the phenomenon of complementizer agreement is not attested. Therefore, the licensing condition of c-commanding agreement was not fulfilled in subordinate clauses. As a consequence, subject pronouns were regularly inserted contrary to the Latin originals:12 (11) Dhar ir auh quhad …, dhar meinida pro where he also said there meant
(I 274)
leohtsamo zi archennenne dhen heilegan gheist easily to recognize the holy spirit ‘Where he said …, he clearly meant there the Holy Spirit’. Item dicendo … sanctum spiritum euidenter aperuit
In contrast to the modern dialects, pro-drop is not restricted to sentences with verb forms that show pronominal agreement. None of the finite verbs in the sentences in (8) above exhibits a suffix of pronominal origin, yet the subject pronoun is dropped in each case. Nevertheless, morphological factors do have a quantitative effect on null subject usage in OHG: even though referential null subjects are attested in all persons and numbers, it is only in the 3rd sg. and pl. that the null variant is used more
. As is outlined in Axel (2005, 2007: ch. 5), the main/subordinate asymmetry, of course, also emerges if one looks at the OHG data from a quantitative perspective (the figures she gives are based on Eggenberger’s (1961) quantitative analysis of null-subject usage in the major OHG texts). In the OHG Isidor, the subject pronoun is dropped in 44% of a total of 109 main clauses with pronoun subjects. By contrast only 9% of subordinate clauses (#93) do not have overt subject pronouns. In the Monsee Fragments even 64% of the main clauses with pronoun subjects (#132) do not exhibit overt subject pronouns, as opposed to only 15% of subordinate clauses (#86). In Tatian, 40% of main clauses with pronoun subjects (# 2394) show pro-drop, but only 8% of subordinate clauses (#1275). Nevertheless, in all three texts there is some amount of subject omission in subordinate clauses. Note, however that Eggenberger’s figures for subordinate clauses also include dependent clauses with V2 order and examples which are ambiguous between V2 and V-end order. We would like to claim that the remaining really problematic cases (i.e. null subjects in the context of verb-end/late order) do not really falsify the post-finiteness restriction, as they can still be due to translational errors, oversights etc. Also, we do concede that the Latin may have exerted some minor impact on the OHG translation. Nevertheless null-subject usage in general is not an instance of systematic ‘loan syntax’ since there is a clear overall tendency that overt subject pronouns are inserted in verb-end/late environments.
What changed where?
frequently than the overt one (Axel 2005, 2007: ch. 6). The reason for this person split is still unclear. In late OHG texts (i.e. in Notker’s works or in Williram von Ebersberg’s paraphrase of the Song of Songs), pro-drop is attested only sporadically. Interestingly, null-subject usage was quite frequent again in the Early New High German period.13 However, the surface distribution of omitted subject pronouns then was somewhat different from that of the OHG period. Instead, null-subject usage then already showed a similar distribution to that found in the modern dialects. The syntax and morphology of prodrop in the dialects is discussed in the following section.
2.3 From Old High German to the modern dialects Now that we have discussed the synchronic distribution of pro in both OHG and some modern dialects, the question arises as to what the diachronic development looks like. From a diachronic viewpoint, at the dialectal level, the phenomenon of pro-drop seems to be an example for continuity: both OHG and the modern dialects allow for partial pro-drop.14 But the grammatical continuity is even more far-reaching. Interestingly, the grammatical licensing conditions have remained the same: both in OHG and in some modern dialects, pro is licensed by c-commanding Agr-in-C. What must have changed is that the licensing conditions have become more specific: only pronominal Agr-in-C has come to license pro-drop. What has also changed is that there was a grammatical innovation – the development of complementizer inflection – which cancelled the restriction to root sentences so that pro-drop can now occur in embedded sentences also in the modern dialects. The restriction to pronominal Agr is related to a further phenomenon, namely to double agreement.
2.3.1 The emergence of double agreement Double agreement means that there exist two ways of inflecting the verb, and that the choice of inflectional marker depends on the position of the verb. In modern Central
. Held (1903: 111) argues that the difference between Middle High German and Early New High German is partly due to the fact that in the latter case, text genres had developed that were closer to oral communication. Middle High German court language, by contrast, was very remote from popular speech and strongly shaped by literary conventions. . When speaking of continuity, it would, of course, be necessary to provide data from Middle High German and Early Modern High German as well. Unfortunately, there are no systematic studies on null subject phenomena at these stages. See, however, Held (1903) for examples.
Katrin Axel & Helmut Weiß
Bavarian, for example, the ending in the 1st pl. is -n in verb-final sentences, but in verb-second and verb-first sentences the suffix is -ma: (12) a. dasma mia aaf Minga fahrn/*ma that-1.pl we to Munich go ‘that we go to Munich’
(Central Bavarian)
b. mia fahrma/*n aaf Minga we go-1.pl to Munich ‘We go to Munich’ c.
fahrma/*n mia aaf Minga? go-1.pl we to Munich ‘Do we go to Munich?’
As can be seen in (12a), -ma is also the ending that appears on inflected complementizers. The suffix -ma is derived from the pronominal subject mia in its clitic form ma: fahrn’ma > fahrm’ma > fahrma. In modern Central Bavarian, -ma is obligatory even when C0 is filled with a complementizer instead of a verb, and it can combine with the full pronoun mia. This shows that -ma is an inflectional suffix rather than a subject clitic. In German double agreement has a long history. A precursor can already be found in OHG. In the Isidor translation, for example, the 1st pl. has two alternative endings: a short ending -m, which was later replaced by -n, and the older long ending (-mēs): (13) a.
Endi dhes selben christes … chichundemes auh and of-the same Christ demonstrate-1.pl also
nu dhes ędhili endi odhil now of-that age and home
‘we now demonstrate the lineage and homeland of that same Christ.’
et cuius … demonstretur et genus et patria
b. Chiuuisso chioffanodom uuir nu hear dhazs … surely revealed-1.pl we now here that ‘surely we have now revealed here that …’
(I 520)
(I 484)
Probauimus …
In OHG, the occurrence of -mēs correlates with pro-drop in verb-second contexts where the subject pronoun would show up postfinitely if it was overtly realized (i.e. in ‘inversion’ contexts). But, in contrast to the modern dialects, the inflectional markers are not dependent on verb position in OHG: both the long and the short ending occur in second and in final position. This phenomenon does not yet constitute double agreement in the sense defined at the beginning of this section. However, soon after the replacement of -mēs, a new type of double agreement for the 1st pl. began to arise, which was then characteristic for the Middle High German
What changed where?
period (cf. Paul 2007: §E21, 2). In Middle High German the 1st pl. ending was often dropped completely. At the time only the -n-marker was used, as the long ending had already died out. Interestingly, this -n-drop was now dependent on verb position/subject–verb inversion: when the pronominal subject directly followed the finite verb in second position, the -n was dropped. A very early example can already be found in the OHG Tatian: (14) thesan niuuizuuuir uuanan her ist. this-acc neg.know.we whence he is ‘but as to this man we do not know where he is from.’
(T 457,10)
hunc autem nescimus unde sit.
A further interesting case is the 2nd sg. As mentioned above, this ending arose out of a reanalysis of the word boundary. In ‘Christ and the Samaritan Woman’ (late 9th century) we find two endings, the short ending -s, and the long ending -st whose additional dental plosive -t comes from the pronoun thu. Interestingly, in this text the old short form occurs only when the subject pronoun precedes the verb: (15) a. kerōst thu ‘chooses you’, gābist du ‘gave you’ b. thu wissīs ‘you know’, tu bātis ‘you begged’ (Christ and the Samaritan Woman; cited from Braune 2004: 261)
To sum up, the development of a second ending/of double agreement seems to have depended on two issues: on V-to-C raising and on subject-verb-inversion (i.e. a pronominal subject cliticizing onto the verb in C0).
2.3.2 Double agreement and the rise of pro-drop What does double agreement have to do with pronominal agreement? After all, we need an explanation for the innovation of pronominal agreement in order to describe the diachrony of pro-drop in the dialects. The above-mentioned Isidor examples may provide evidence for a putative diachronic relationship between the two phenomena: (16) a.
Endi dhes selben christes … chichundemes pro auh nu (= 13a)
b. Chiuuisso chioffanodom uuir nu hear dhazs … (= 13b)
Such minimal pairs, which must have occurred in spoken OHG as well, may have conveyed the impression to the child learner that a verbal ending must contain something special which enables it to license pro-drop and that this special property is its pronominal origin. Already Paul (1877) analyses -mēs as deriving from a personal pronoun. This assumption has turned out to be untenable, as later research has shown. Nevertheless, it is conceivable that the pronominal nature of -mēs was the result of a reanalysis of V-mēs as V-n+uuir in such minimal pairs: -mēs may have
Katrin Axel & Helmut Weiß
been eventually analysed as pronominal exactly in such contexts where it contrasts with overt pronouns. Since only the long ending, but not the short one, was able to license prodrop, it may well have been the case that its pronominal nature was made responsible for this.15 Additional evidence for this speculative scenario comes from two further pieces of data: first, the hybrid ending -nmes which looks like a combination of the short and long form (n + mēs), is also attested in Tatian (Franck 1971): (17) uuanne gisahun uuir thih gast uuesentan when saw we you guest being
(T 545,7)
Inti gihalotunmes thih. and took you ‘when saw we you a stranger, and took you in’ te uidimus hospitem & collegimus te.
Such forms have been analyzed as combinations of a verb regularly carrying the short ending -n to which -mēs was added. If this was indeed the case, it would provide strong evidence for the pronominal nature of -mēs in the eyes of speakers of OHG. As expected (and necessary for our explanation), all seven cases of hybrid forms appear in verb-second position in Tatian and all other texts where they are attested. Second, we know from several Upper German dialects that the w-onset of the pronoun wir ‘we’ has indeed developed into an m in exactly those contexts where the subject pronoun follows the verb in second position. This assimiliation may have occurred already in spoken OHG (cf. Schatz 1907: 163). To conclude: the reanalysis of the inflectional marker -mēs as pronominal may have then had the consequence that the ability to license pro-drop became restricted to pronominal agreement in C.
2.3.3 Conclusion The phenomenon of pro-drop in German is characterized by continuity at the dialectal level. The alleged loss of (referential) pro-drop has occurred only in the standard language. At first sight, the surface distribution of (referential) pro seems to be entirely different in OHG as compared to the modern dialects. However, the underlying syntactic licensing condition is the same: pro is only licensed if c-commanded by verbal Agr. What has changed is that the requirement on . As one of the reviewers rightly objects, this explanation still leaves open the question, why -mēs licensed pro-drop before it has been reanalysed as [+pronominal]. However, this problem may be only an artifact of the sparse text documentation: it may be the case that the reanalysis of -mēs had already begun at the time from which the oldest texts of OHG (like MF and I) are attested.
What changed where?
Agr has become more specific: in the modern dialects only pronominal agreement has come to be able to identify pro. We have argued that this change is probably related to a further phenomenon, namely to the rise of double agreement. Furthermore, the innovation of inflected complementizers, an independent phenomenon, led to a drastic change in the surface distribution of pro, as the requirement of a c-commanding Agr could then also be fulfilled in subordinate clause. It would, however, be beyond the scope of this chapter to discuss the history of complementizer inflection (but cf. Fuß 2005, 2008; Axel & Weiß to appear).
3. Further case studies We have only been able to present one case study in more detail. But there are many more cases where the question of dialect vs. standard has not received sufficient attention in diachronic studies. We will briefly discuss three further cases where the development at the dialectal level considerably differs from the development of the standard.
3.1 Negation A further case of more continuity at the dialectal level is negation. In Old and Middle High German there was negative concord (NC); consider (18a) from the Middle High German Prose Lancelot. Modern Standard German is not a negative concord language any more, (18b), but at the dialectal level negative concord has survived as can be seen in (18c) from Modern Bavarian. (18) a.
Da enwart nymand konig there neg-became nobody king ‘Nobody became king.’
b. Ich habe kein Bier (*nicht) zuhause I have no beer neg at-home c.
I hon koa Bia ned dahoam I have no beer neg at-home ‘I don’t have a beer at home.’
(Middle High German) (Lanc 10,9) (Modern Standard German) (Modern Bavarian)
The difference between (18a) and (18c) indicates that there was a change from Middle High German to the modern dialects as well: the negative particle ne/en vanished and nicht (Bavarian ned), originally a negation strengthener, became the particle expressing negation alone (Jäger 2008). But this change is only a superficial one in two respects: (i) it has not altered the fundamental property of being a NC language, and (ii) there was no accompanying change on the structural level. As Jäger (2008) has shown, the syntax of negation in German can be analysed as involving a NegP structure
Katrin Axel & Helmut Weiß
throughout its history. The only change that has happened was that the lexeme originally occupying the Neg0 position was lost in Middle High German times (19): (19) a. Da enwart
b.
nymand
konig
NegP Spec nicht VP
Neg′ Neg0 ni/ne/ø
Note that continuity on the structural level also holds for Standard German (Jäger 2008), but this is somewhat concealed through the apparent loss of the NC property. So the development of the dialects gives a much more adequate picture with respect to continuity or change.16
3.2 Possessive constructions A different case would be the development of possessive constructions. In this case there is more continuity in the standard language: here, the post-nominal genitive has survived, (20a). But in the dialects there emerged a new structure, namely a DP-internal pre-nominal possessor construction.17 (20) a.
das Haus des Vaters the house the-gen father-gen ‘the house of the father’
b. am Voda sei Haus the-dat father-dat his house ‘the house of the father’
(Modern Standard German)
(Modern Bavarian)
The case system of German developed in such a way that the genitive was lost first as verbal case and eventually adnominally as well (Behaghel 1923). This development
. The case of negation in German is much more complicated than described in the main text, since there is dispute in the literature whether German has lost NC (Jäger 2008) or not (Weiß 2004b). A further point of discussion is the question as to whether or not nicht was reanalysed as head or NegP (cf. Jäger 2008). However, independently of these issues, it is undisputable that all the dialects did not lose the NC property and they thus exhibit obvious continuity in a fundamental area of grammar. . Some dialects have, however, retained the adverbial genitive in fossilized expressions such as des Mittags ‘at noon’, des Abends ‘in the evening’ etc.
What changed where?
took place more than 500 years ago, but only in the dialects, whereas in the written language the adnominal genitive survived (Weiß 2001). The case of the Standard German genitive is an impressive example of the conserving ability of written varieties (even in the absence of prescriptive grammars) (Weiß 2001, 2004a,b). In the dialects there evolved a new possessive construction as in (20b) where the possessor appears pre-nominally and is marked as dative instead of genitive (Weiß 2008). This development conforms with the overall development of the case system in German with its loss of the genitive – though the emergence of the structure as such was presumably independent of this overall development (cf. Weiß 2008). Therefore, the possessive syntax of the standard with its continuity in the use of the genitive conveys a rather inadequate impression.
3.3 The consolidation of verb-final order It is a widely-held assumption that ‘German’ underwent a development towards more rigid final order of the finite verb, which amounts to the strengthening of the so-called sentence bracket formed by the finite verb or complementizer in C (left sentence bracket) and by the verbal cluster at the end of the clause (right sentence bracket). It has been argued that there was a diachronic decrease of variation in the verbal cluster with the effect that the finite verb became increasingly restricted to absolute final position.18 Again, if we take into consideration the situation in the modern dialects, a somewhat different picture emerges. In Modern Standard German, the word order in the verbal complex is very rigid in that the subordinate verb has to precede the superordinate verb (e.g. V2 > V1, cf. (21); V3 > V2 > V1; V4 > V3 >V2 > V1):19 (21) a.
dass Maria ein Lied gesungen hat that Mary a song sung2 has1 ‘…that Mary has sung a song.’
b. … *dass Maria ein Lied hat gesungen that Mary a song has1 sung2 ‘…that Mary has sung a song.’
. Besides the change in verbal clusters, the sentence bracket was also consolidated by a quantitative and qualitative decrease of extraposition patterns. We will not deal with this phenomenon here, but note that there are suggestive findings reported in the secondary literature that the contemporary dialects are more liberal towards what types of XPs can be extraposed and exhibit a higher frequency of extraposition than modern Standard German (e.g. Patocka 1997: 320–358; Weiß 1998: 55–58 on Bavarian). . The maximal superordinate verb, i.e. the finite verb, receives the index ‘1’ (= V1), the verb governed by it receives the index ‘2’ (= V2) and so forth.
Katrin Axel & Helmut Weiß
In general, deviations do not occur in bipartite verbal complexes as in (21). They are only possible in verbal clusters with three or more verbs, but even here there are many restrictions regarding the question as to which types of verbs may invert from the canonical order, which exact orders are allowed and so on. It is a well-known fact that in earlier stages of the language there was much more variation in verbal clusters and that in bipartite clusters the inverted order (V1 > V2) was also possible (see Robinson 1997; Axel 2007: ch. 2; Weiß to appear on Old High German; Paul 2007: 453–456 on Middle High German; Ebert et al. 1993: 438–440 on Early New High German). The decisive period for verb-order change in the verbal cluster was the Early New High German period (1350–1650). During this period there was a general decline of the inverted order (Härd 1981; Maurer 1926: 151; Ebert 1981, 1998; Bies 1996; Reifsnyder 2003; Sapp 2006). In more recent studies there has been growing consensus that this is a ‘change from above’, passed down from the chancery style (Ebert 1981, 1998; Bies 1996). Reifsnyder (2003), in her study of the Early New High German dialect of Augsburg, comes to the conclusion that there was a prevalence of V2 > V1 orders in the official texts, as well as an increase in frequency over time, which she argues results from the adoption of a standard language ideology. To our knowledge there have been no studies to date looking at when the inverted order became completely ungrammatical in bipartite clusters in the standard language. However, in most modern dialects bipartite verbal clusters may still be realized with inverted order (V1 > V2) and tripartite clusters show orders that are not possible in Standard German, as is illustrated by the Austrian German examples in (22) (cf. also Lötscher 1978; Wurmbrand 2004 on Swiss German; Weiß 1998 on Bavarian; Wurmbrand 2004 on Austrian dialects): (22) a. dea den Gaia håt åhagschossn rel the hawk has1 down-shot2 ‘who shot down the hawk.’
(Styria)
b. wånn a Kind is taaft gwoadn when a child is baptized-ppp aux-ppp2 ‘when a child was baptized.’
(Styria) (Patocka 1997: 290)
Sapp (2006) carried out controlled experiments (magnitude estimation) on verbal clusters in some contemporary dialects (mostly on Austrian and Swabian German) and compared these findings and those from an extensive survey of the secondary literature on various dialects with the results of his quantitative investigation of the texts from the online version of the Bonner Frühneuhochdeutschkorpus. He comes to the conclusion (p. 170) that although “many dialects appear to have the same kind of word-order variation as ENHG [Early New High German], there have also been some important historical developments”. With bipartite clusters, the Early New High German state of affairs is best preserved in Swiss German, which still
What changed where?
allows the V1 > V2 and the V2 > V1 order, while in Swabian and Austrian and most other dialects the former order is disprefered. In Early New High German, the V1 > V2 order is better with the modal–infinitive syntagm than with the perfect tense. Interestingly, however, other dialects, namely Swabian and Eastern Austrian, have developed the reverse preference. With tripartite clusters, most dialects have come to agree with Standard German in allowing only V3 > V2 > V1, however, as in Early New High German, in the IPP (= infintivus pro participio) and werden–modal–infinitive constructions (cf. Schmid & Vogel 2004; Schmid 2005) there is more variation than in Standard German. In both bi- and tripartite clusters, focus has some effect on verb order in Early New High German. More precisely, Sapp (2006: 168) has found that focus on the object has a favouring effect on the V1 > V2 order and probably on the V1 > V3 > V2 order. Interestingly, the same effect shows up in two contemporary dialects which he studied experimentally (i.e. Swabian and Austrian German) and in Modern Standard German. To sum up, Sapp’s comparison of the contemporary dialects with the Early New High German dialects has revealed that the diachrony of verbal clusters is characterized by both syntactic continuity and discontinuity. Compared to the developments in the standard languages, the dialects seem to be more conservative and reluctant to change at first. However, Sapp’s study suggests that the microvariation we still find in the contemporary dialects is no direct reflex of the situation in the Early New High German dialects, but it must be the product of some small-scale changes (‘micro-changes’).
4. Conclusion All these case studies show that taking into account dialectal evidence in historical syntax can lead to different and more appropriate results. These results constitute a more solid basis for judgements about what is stable and what is more open to change in the history of a language. There are especially three aspects speaking in favour of dialectal data in historical syntax:20 First: if there were different historical developments in dialects and the standard, the development in the dialects should be more relevant, because L1 acquisition can be
. There is another aspect speaking in favour of dialects in historical linguistics: the written (variety of a) language is not always based on the same variety/dialect in the historical development. English exemplifies this lack of continuity: whereas the corpus of Old English is written in the Wessex dialect, modern Standard English developed from the Mercian dialect (Crystal 1995: 29). Thanks to one of the reviewers for bringing this aspect to our attention.
Katrin Axel & Helmut Weiß
made responsible for them, whereas standard languages were shaped by extra-linguistic factors as well; see, for example, the loss of negative concord. Second: some of the innovations occurring in dialects have never reached the standard and these are mostly the interesting cases for linguistics. We mentioned one such case: complementizer agreement (cf. Corbett 2006 for the typological uniqueness of this feature). Other cases from German dialects, among many others, would be verb doubling in Swiss German (Schönenberger & Penner 1995) or matrix verb stranding in Swabian (Hiller 1999).21 Third: Just as dialects exhibit syntactic variation at the synchronic level, they do so at the diachronic level as well. The German verbal clusters show considerable wordorder variation both in the contemporary and in the Early New High German dialects, but there is no direct correspondence between the situation in the modern dialects and their respective Early New High German precursors. Though syntax is surely the most inert linguistic level with respect to change (Longobardi 2001), the syntactic systems of the dialects must be the product of many ‘micro-changes’ that have shaped the microvariation that we see today.
Primary sources [I] Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten. Hans Eggers (ed.). Tübingen: Niemeyer. 1964. [Lanc] Lancelot. Nach der Heidelberger Pergamenthandschrift Pal. Germ. 147, Reinhold Kluge (ed.). Berlin: Akademie-Verlag. Vol. I 1948. [MF] The Monsee Fragments. Newly collated text with introduction, notes, grammatical treatise and exhaustive glossary and a photo-lithographic fac-simile. George Allison Hench (ed.). Straßburg: Karl J. Trübner. 1890. [T] Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Achim Masser (ed.) (in cooperation with Elisabeth De Felip-Jaud). Vandenhoeck & Ruprecht. 1994.
References Axel, Katrin. 2005. Null subjects and verb placement in Old High German. In Linguistic Evidence. Empirical, Theoretical and Computational Perspectives, Stephan Kepser & Marga Reis (eds), 27–48. Berlin: Mouton de Gruyter.
. Verbs like to help/let/start and a few others behave curiously in Swabian in that it is the embedded verb they select that is inflected, whereas they themselves appear as infinitives, cf. (i). This is called Matrix Verb Stranding by Hiller (1999). (i)
Se šreibt des ets ãafangә uf she write-3.sg this-acc now begin-inf down ‘She is now starting writing this down’
What changed where?
Axel, Katrin. 2007. Studies on Old High German Syntax: Left Sentence Periphery, Verb Placement and Verb-Second [Linguistik Aktuell/Linguistics Today 112]. Amsterdam: John Benjamins. Axel, Katrin & Weiß, Helmut. To appear. Pro-drop in the history of German: From Old High German to the modern dialects. In Empty Pronouns, Peter Gallmann & Melanie Wratil (eds). Berlin: Mouton de Gruyter. Bayer, Josef. 1984. COMP in Bavarian Syntax. The Linguistic Review 3: 209–274. Bayer, Josef. 1993. ‘Zum’ in Bavarian and scrambling. In Dialektsyntax, Werner Abraham & Josef Bayer (eds), 50–70. Opladen: Westdeutscher Verlag. (Linguistische Berichte, Sonderheft 5/1993). Bies, Ann. 1996. Syntax and Discourse Factors in Early New High German: Evidence for Verbfinal Word Order. MA thesis, University of Pennsylvania, Philadelphia. Benincà, Paula & Poletto, Cecilia. 2007. The ASIS enterprise: A view on the construction of a syntactic atlas for the Northern Italian Dialects. In Nordlyd 34: Scandinavian Dialect Syntax 2005, Kristine Bentzen & Øystein Alexander Vangsnes (eds), 35–52. Tromsø: CASTL. Behaghel, Otto. 1923. Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 2: Die Wortklassen und Wortformen. Heidelberg: Winter. Behaghel, Otto. 1928. Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 3: Die Satzgebilde. Heidelberg: Winter. Braune, Wilhelm. 2004. Althochdeutsche Grammatik I, 15th edn, by Ingo Reiffenstein. Tübingen: Niemeyer. Corbett, Greville G. 2006. Agreement. Cambridge: CUP. Cornips, Leonie & Poletto, Cecilia. 2005. On standardising syntactic elicitation techniques, Part 1. Lingua 115: 939–957. Cooper, Kathrin. 1995. Null subjects and clitics in Zurich German. In Topics in Swiss German Syntax, Zvi Penner (ed.), 59–71. Bern: Peter Lang. Crystal, David. 1995. The Cambridge Encyclopedia of the English Language. Cambridge: CUP. DeGraff, Michel. 1999. Language Creation and Language Change: Creolization, Diachrony, and Development. Cambridge MA: The MIT Press. Durrell, Martin. 1999. Standardsprache in England und Deutschland. Zeitschrift für Germanistische Linguistik 27: 285–308. Ebert, Robert Peter. 1981. Social and stylistic variation in the order of auxiliary and non-finite verb in dependent clauses in Early New High German. Beiträge zur Geschichte der deutschen Sprache und Literatur 103: 204–237. Ebert, Robert Peter. 1998. Verbstellungswandel bei Jugendlichen, Frauen und Männern im 16. Jahrhundert. Tübingen: Niemeyer. Ebert, Robert Peter, Reichmann, Oskar, Solms, Hans-Joachim & Wegera, Klaus-Peter. 1993. Frühneuhochdeutsche Grammatik. Tübingen: Niemeyer. Eggenberger, Jakob. 1961. Das Subjektspronomen im Althochdeutschen. Chur: Selbstverlag. Franck, Johannes. 1971. Altfränkische Grammatik. Laut- und Flexionslehre, 2nd edn, by Rudolf Schützeichel. Göttingen: Vandenhoeck & Ruprecht. Fuß, Eric. 2005. The Rise of Agreement. A Formal Approach to the Syntax and Grammaticalization of Verbal Inflection [Linguistik Aktuell/Linguistics Today]. Amsterdam: John Benjamins. Fuß, Eric. 2008. Multiple agreement and the representation of inflection in the C-domain. Linguistische Berichte 213: 77–106. Grimm, Jacob. 1967. Deutsche Grammatik, Vol. 4. (4th Reprographic reprint of the edition Gütersloh 1898). Hildesheim: Georg Olms. Haag-Merz, Christine. 1996. Pronomen im Schwäbischen: Syntax und Erwerb. Marburg: Tectum-Verlag.
Katrin Axel & Helmut Weiß Härd, John Evert. 1981. Studien zur Struktur mehrgliedriger deutscher Nebensatzprädikate Diachronie und Synchronie [Göteborger germanistische Forschungen 21]. Göteborg: Acta Universitatis Gothoburgensis. Held, Karl. 1903. Das Verbum ohne pronominales Subjekt in der älteren deutschen Sprache. Berlin: Mayer & Müller. Hiller, Markus. 1999. Violable relativized minimality. Ms, Rutgers University. Hopper, Paul J. 1975. The Syntax of the Simple Sentence in Proto-Germanic. The Hague: Mouton. Jäger, Agnes 2008. History of German Negation [Linguistik Aktuell/Linguistics Today 118]. Amsterdam: John Benjamins. Jaeggli, Osvaldo & Safir, Kenneth J. 1989. The null-subject parameter and parametric theory. In The Null Subject Parameter, Osvaldo Jaeggli & Kenneth J. Safir (eds), 1–44. Dordrecht: Kluwer. Kayne, Richard. 1996. Microparametric syntax: Some introductory remarks. In Microparametric Syntax and Dialect Variation [Current Issues in Linguistic Theory 139], James R. Black & Virginia Motapanyane (eds), ix-xviii, Amsterdam: John Benjamins. Koeneman, Olaf. 2007. Deriving the difference between full and partial pro-drop. In Arguments and Agreement, Peter Ackema, Patrick Brandt, Maaike Schoorlemmer & Fred Weerman (eds), 76–100. Oxford: OUP. Kögel, Rudolf. 1882. Zum deutschen Verbum. Beiträge zur Geschichte der deutschen Sprache und Literatur 8: 126–139. Kroch, Anthony & Taylor, Ann. 1997. Verb movement in Old and Middle English: Dialect variation and language contact. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds), 297–325. Cambridge: CUP. Kroch, Anthony. 2000. Syntactic change. In The Handbook of Contemporary Syntactic Theory, Mark Baltin & Chris Collins (eds.), 629–739. Malden MA: Blackwells. Ledgeway, Adam. 2000. A Comparative Syntax of the Dialects of Southern Italy: A Minimalist Approach. Oxford: Blackwell. Lenerz, Jürgen. 1984. Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Lötscher, Andreas. 1978. Zur Verbstellung im Zürichdeutschen und in anderen Varianten des Deutschen. Zeitschrift für Dialektologie und Linguistik 45: 1–29. Longobardi, Guiseppe. 2001. Formal syntax, diachronic minimalism, and etymology. The History of French chez. Linguistic Inquiry 32: 275–302. Maurer, Friedrich. 1926. Untersuchungen über die deutsche Verbstellung in ihrer geschichtlichen Entwicklung. Heidelberg: Winter. Meillet, Antoine. 1908/09. Notes sur quelques faits gotiques. Mémoires de la Société de Linguistique de Paris 15: 73–103. Müller, Gereon. 2006. Pro-drop and impoverishment. In Form, Structure, and Grammar. A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, Patrick Brandt & Eric Fuß (eds), 93–115. Berlin: Akademie Verlag. Patocka, Franz. 1997. Satzgliedstellung in den bairischen Dialekten Österreichs. Frankfurt: Peter Lang. Paul, Hermann. 1877. Die Vocale der Flexions- und Ableitungssilben in den ältesten germanischen Dialecten. Beiträge zur Gechichte der deutschen Sprache und Literatur 4: 315–475. Paul, Hermann. 1880. Prinzipien der Sprachgeschichte. Tübingen: Niemeyer. Paul, Hermann. 1919. Deutsche Grammatik, Vol. III. Tübingen: Niemeyer.
What changed where?
Paul, Hermann. 2007. Mittelhochdeutsche Grammatik, 25th edn by Thomas Klein, Hans-Joachim Solms, Klaus-Peter Wegera, Ingeborg Schröbler and Heinz-Peter Prell. Tübingen: Niemeyer. Poletto, Cecilia. 1995. The diachronic development of subject clitics in North-Eastern Italian dialects. In Clause Structure and Language Change, Adrian Battye & Ian G. Roberts (eds), 295–324. Oxford: OUP. Reifsnyder, Kristen. 2003. Vernacular versus Emerging Standard: An Examination of Dialect Usage in Early Modern Augsburg (1500–1650). Ph.D. dissertation, University of Wisconsin, Madison. Robinson, Orrin W. 1997. Clause Subordination and Verb Placement in the Old High German Isidor Translation. Heidelberg: Winter. Rohrbacher, Bernhard Wolfgang. 1999. Morphology-Driven Syntax: A Theory of V to I Raising and Pro-Drop [Linguistik Aktuell/Linguistics Today 15]. Amsterdam: John Benjamins. Sapp, Christopher. 2006. Verb Order in Subordinate Clauses: From Early New High German to Modern German. Ph.D. dissertation, Indiana University. Schatz, Josef. 1907. Altbairische Grammatik. Laut- und Flexionslehre. Göttingen: Vandenhoeck & Ruprecht. Schmid, Tanja & Vogel, Ralf. 2004. Dialectal variation in German 3-verb-clusters. A surfaceoriented Optimality Theoretic account. Journal of Comparative Germanic Linguistics 7: 235–274. Schmid, Tanja. 2005. Infinitival Syntax: Infinitivus Pro Participio as a Repair Strategy [Linguistik Aktuell/Linguistics Today 79]. Amsterdam: John Benjamins. Schönenberger, Manuela & Penner, Zvi. 1995. Cross-dialectal variation in Swiss German: Doubling verbs, verb projection raising, barrierhood, and LF-Movement. In Studies in Comparative Germanic Syntax, Hubert Haider, Susan Olsen & Sten Vikner (eds), 285–305. Dordrecht: Kluwer Sprouse, Rex A. & Vance, Barbara. 1999. An explanation for the decline of null pronouns in certain Germanic and Romance languages. In Language Creation and Language Change: Creolization, Diachrony, and Development, Michel DeGraff (ed.), 257–284. Cambridge MA: The MIT Press. Weise, Oskar. 1900. Syntax der Altenburger Mundart. Leipzig: Breitkopf und Härtel. Weise, Oskar 1907. Die sogenannte Flexion der Konjunktionen. Zeitschrift für deutsche Mundarten 2: 199–205. Weiß, Helmut. 1998. Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: Niemeyer. Weiß, Helmut. 2001. On two types of natural languages. Some consequences for linguistics. Theoretical Linguistics 27: 87–103. Weiß, Helmut. 2004a. Zum linguistischen Status von Standardsprachen. In Indogermanistik, Germanistik, Linguistik, Akten der Arbeitsstagung der Indogermanischen Gesellschaft, Jena 18.–20.9.2002, Maria Kozianka, Rosemarie Lühr & Susanne Zeilfelder (eds), 591–643. Hamburg: Verlag Dr. Kovač. Weiß, Helmut. 2004b. A question of relevance: Some remarks on standard languages (Author’s response). In What Counts as Evidence in Linguistics? The Case of Innateness [Studies in Language 28(3)], Martina Penke & Anette Rosenbach (eds), 648–674. Amsterdam: John Benjamins. (Also published as Benjamins Current Topics 7, 2007). Weiß, Helmut. 2005a. Inflected complementizers in Continental West Germanic Dialects. Zeitschrift für Dialektologie und Linguistik 72: 148–166.
Katrin Axel & Helmut Weiß Weiß, Helmut. 2005b. The double competence hypothesis. On diachronic evidence. In Linguistic Evidence: Empirical, Theoretical, and Computational Perspectives, Stephan Kepser & Marga Reis (eds), 557–575. Berlin: Mouton de Gruyter. Weiß, Helmut. 2005c. Von den vier Lebensaltern einer Standardsprache. Zur Rolle von Spracherwerb und Medialität. Deutsche Sprache 33: 289–307. Weiß, Helmut. 2008. The possessor that appears twice? Variation, structure and function of possessive doubling in German. In Microvariation in Syntactic Doubling in German, Sjef Barbiers, Olaf Koeneman, Marika Lekakou, Margreet van der Ham (eds), 381–401. Bingley: Emerald. Weiß, Helmut. 2009. How to define dialect and language. A proposal for further discussion. Linguistische Berichte 219: 251–270. Weiß, Helmut. To appear. Die rechte Peripherie im Althochdeutschen. Zur Verbstellung in dass-Sätzen. In Tagungsakten der Arbeitstagung der Indogermanischen Gesellschaft. Wurmbrand, Susi. 2004. West Germanic verb clusters: The empirical domain. In Verb Clusters: A Study of Hungarian, German, and Dutch [Linguistik Aktuell/Linguistics Today 69], Katalin É. Kiss & Henk van Riemsdijk (eds), 43–85. Amsterdam: John Benjamins.
Impossible changes and impossible borrowings The Final-over-Final Constraint* Theresa Biberauer1,3, Michelle Sheehan2 & Glenda Newton1 1University
of Cambridge/2University of Newcastle/3Stellenbosch University
This chapter examines the predictions of Biberauer, Holmberg & Roberts’s (2007, 2008) Final-over-Final Constraint/FOFC for grammatical change and borrowing. As a putatively invariant syntactic principle, FOFC excludes the synchronic possibility of head-final phrases dominating categorially alike head-initial phrases. For diachrony, FOFC predicts certain word-order changes to be impossible, whether contact is involved or not: specifically, head-final to head-initial change is predicted to proceed top-down, whereas the reverse change should proceed bottom-up. Case studies from the history of English, Afrikaans and French seem to support the first of these predictions. Furthermore, we show on the basis of data from South Asian languages, that the presence of a phrase-initial head blocks the borrowing of a higher phrase-final head, thereby avoiding a FOFC-violation.
1. Introduction In word-order terms, the languages of the world can be classified as being either harmonic or disharmonic. The term “harmonic”, originating with Greenberg (1963) (cf. also Hawkins 1983), refers to a language that is either consistently head-initial or consistently head-final. Within a generative framework, patterns of this type have
*Our thanks to the audiences at the Continuity and Change Conference (Cambridge – March 2008), the “Past Meets the Present: A Dialogue Between Historical Linguistics and Theoretical Linguistics” Conference (Taipei – July 2008) and DiGS X (Cornell – August 2008), and to the members of the FOFC seminar (October-December 2007). Particular thanks to Edith Aldridge, Alastair Appleton, Özgür Aydın, Silvio Cruschina, Matthew Dryer, Jim Huang, Tony Kroch, Elliott Lash, Adam Ledgeway, Pino Longobardi, Iain Mobbs, Waltraud Paul, Chris Reintges, Sarah Thomason, Nigel Vincent, John Whitman and two very helpful anonymous reviewers. This work is supported by AHRC Grant No. AH/E009239/1 (“Structure and Linearization in Disharmonic Word Orders”).
Theresa Biberauer, Michelle Sheehan & Glenda Newton
traditionally been related to the Head Parameter, which may be informally stated as in (1): (1) The Head Parameter: X > YP (head-initial languages) YP > X (head-final languages)
However, a significant number of languages are not fully harmonic; instead they contain a mixture of head-initial and head-final phrases. A well-known example of a disharmonic language is German, which has a head-final VP and AuxP,1 but a headinitial CP, DP and, for the most part, PP, as in (2): (2) … dass das Mädchen in München gewohnt hat [CP that [AuxP[VP[DP the girl] [PP in Munich] lived] has]] ‘… that the girl lived in Munich.’
The existence of disharmonic word orders indicates that the Head Parameter must be formulated in such a way that it can be set not just in an across-the-board fashion, but also, where necessary, in a category-specific way (cf. Hawkins 1983 for discussion). This leads us to expect that disharmonic orders will be “equal” in the sense that all combinations of mixed headedness should, in principle, be equally available. In reality, however, the empirical record exhibits a striking skewing regarding the attestation of disharmonic word orders. The following section introduces a generalization capturing this skewing and then presents data supporting it, along with some apparent counterexamples. Section 3 considers the diachronic predictions made by the generalization in terms of word-order change, Section 4 focuses on the implications for borrowing and Section 5 concludes.
2. The Final-over-Final Constraint Holmberg (2000: 124) observes that the following configuration seems to be banned in many languages:2
. We employ the label AuxP here despite the fact that it has often been argued that German auxiliaries are in fact Vs (cf. Haider 1993, 2000; Müller & Reis 2001). This labeling should not be interpreted as signifying a commitment to an opposing analysis; here and elsewhere, we simply employ AuxP as a convenient, theory-neutral descriptive label designating the position(s) above the lexical VP that may be occupied by auxiliaries, not only in German, but in languages more generally. . The tree in (3) shows head-final orders as base-generated by a head parameter. Under a Kaynean view, head-final orders are derived via roll-up movement of a head’s complement to
Impossible changes and impossible borrowings
(3)
*
β′
αP α
β γP
where αP is the complement of β and γP is the complement of α
He accounts for this gap by postulating the following constraint:
(4) The Final-over-Final Constraint (FOFC) [First Version]:
If α is a head-initial phrase and β is a phrase immediately dominating α, then β must be head-initial. If α is a head-final phrase, and β is a phrase immediately dominating α, then β can be head-initial or head-final.
Evidence of the existence of a constraint of the type in (4) comes from a range of clausal and non-clausal contexts in unrelated languages. Thus, for example, Holmberg observes that in Finnish all permutations of the verb, object and auxiliary occur, except for FOFC-violating V-O-AUX (V underlined; O in smallcaps; AUX in bold): Milloin Jussi olisi kirjoittanut romaanin? when Jussi would-have written novel-def
[Aux-V-O]
b. Milloin Jussi olisi romaanin kirjoittanut? when Jussi would-have novel-def written
[Aux-O-V]
(5) a.
Milloin Jussi romaanin kirjoittanut olisi? when Jussi novel-def written would-have ‘When would Jussi have written a novel?’
[O-V-Aux]
d. *Milloin Jussi kirjoittanut romaanin olisi? when Jussi written novel-def would-have
*[V-O-Aux]
c.
The same pattern emerges if we consider both present-day and historical varieties of Germanic. We illustrate for Old English (formatting as above; cf. Biberauer,
its specifier position (cf. Kayne 1994). In this chapter, we abstract away from such theoretical details. These are, however, addressed in Biberauer, Newton & Sheehan (2009), where it is also made clear that a head parameter-based approach is untenable. For the purposes of this chapter, however, it should be assumed that FOFC violations surface under two structural conditions: (i) wherever a head-initial phrase is dominated by a head-final one that has been merged on top of it, and (ii) wherever a head-initial phrase is dominated by a head-final one as the result of a linearization-driven movement operation (see also Note 3). In Kaynean terms, all FOFC violations will, of course, have the structure in (ii).
Theresa Biberauer, Michelle Sheehan & Glenda Newton
Holmberg & Roberts 2007 for more detailed discussion and illustration; and cf. Pintzuk 1991/1999 on the absence of SVOAux throughout the history of English):
(6) O V AUX (“head-final” order in VP and AuxP):
Đa se Wisdom þa is fitte asungen hæfde … when the Wisdom then this poem sung had ‘When Wisdom had sung this poem …’ (Boethius 30.68.6; cf. Fischer et al. 2000: 143)
(7) O AUX V (“verb-raising”; cf. Evers 1975 and subsequent work):
… þe æfre on gefeohte his handa wolde afylan who ever in battle his hands would defile-inf ‘… whoever would defile his hands in battle’ (Ælfric’s Lives of Saints 25.858; cf. Pintzuk 1991: 102)
(8) AUX O V (“verb projection raising”; cf. Haegeman & van Riemsdijk 1986):
… þæt hie mihton swa bealdlice Godes geleafan bodian that they could so boldly God’s faith preach-inf ‘…that they could preach God’s faith so boldly’ (The Homilies of the Anglo-Saxon Church I 232; cf. van Kemenade 1987: 179)
(9) V AUX O (“object extraposition”):
… þæt ænig mon atellan mæge ealne one demm that any man relate can all the misery ‘… that any man can relate all the misery’ (Orosius 52.6–7; cf. Pintzuk 2002: 283) (10) AUX V O (“verb raising” combined with “object extraposition”): … þæt he mot ehtan godra manna that he might persecute good men ‘… that he might persecute good men’ (Wulfstan’s Homilies 130.37–38; cf. Pintzuk 2002: 282)
Crucially, every permutation of Aux, V and O is attested, except for the FOFC-violating V-O-AUX order (see i.a. den Besten 1986; Travis 1984: 157–8; Kiparsky 1996: 168–171; Pintzuk 1991/1999; Hróarsdóttir 2000; Fuß & Trips 2002 for discussion). Haddican (2004) observes that the same gap occurs in Basque: (11) a.
Jon-ek ez dio Miren-i egia esan Jon-erg not aux Miren-dat truth say-perf ‘Jon has not told Miren the truth.’
[Aux-O-V]
b. Jon-ek ez dio esan Miren-i egia [Aux-V-O] Jon-erg not aux say-perf Miren-dat truth ‘Jon has not told Miren the truth.’
Impossible changes and impossible borrowings
(12) a.
Jon-ek Miren-i egia esan dio Jon-erg Miren-dat truth say-perf aux ‘Jon has told Miren the truth.’
b. *Jon-ek esan Miren-i egia dio Jon- erg say-perf Miren-dat truth aux
[O-V-Aux]
[*V-O-Aux]
Furthermore, it is not only in this domain that we find such word-order gaps. It has been observed that VO languages cross-linguistically do not have clause-final complementizers (cf. Hawkins 1990: 256–7; Dryer 1992: 102). Although at first sight this does not seem to be ruled out by FOFC as stated above, because C does not directly dominate VP, on closer inspection it emerges that a final C in a VO language necessarily violates FOFC at some point in structures in which it occurs. Consider the following: (13) a. *[CP [AuxP [VP V O ] Aux ] C ] – violates FOFC (α = V, β = Aux) *
C′ AuxP
C
b. *[CP [AuxP Aux [VP V O ]] C ] – violates FOFC (α = Aux, β = C) *
C′
V
O
In (13a), a structure with a head-final CP and a head-final AuxP, the head-final AuxP dominates a head-initial VP, leading to a FOFC violation of the type already discussed above. In (13b) a structure with a head-final CP dominates a head-initial AuxP, leading to a FOFC violation at the CP/AuxP level. Biberauer, Holmberg & Roberts/BHR (2007, 2008) discuss a range of further structures excluded by FOFC, not only in the clausal domain, but also in the context of nominals, underlining the general nature of the gap characterized by (4). Furthermore, Cecchetto (to appear) shows that FOFC also holds in the domain of sign languages, thereby confirming the non-modality-specific nature of this grammatical constraint.
Theresa Biberauer, Michelle Sheehan & Glenda Newton
Although there is thus significant empirical support for FOFC, it is important to note that there are also apparent exceptions. For example in German, a head-final VP can dominate a head-initial DP (14a) or head-initial PP (14b): (14) a.
Er hat [VP[DP ein Buch] gelesen]3 he has a book read ‘He read the book.’
b. Sie ist [VP[PP nach Berlin] gefahren] she is to Berlin driven ‘She went to Berlin.’
BHR argue that such exceptions can be accounted for if FOFC is modified, such that it only holds over XPs that are categorially non-distinct. Although the notion of ‘categorial distinctness’ is by no means straightforward, in the German case it is clear that the VP is a verbal category, whereas the DP, and perhaps the PP, should be considered nominal.4 By contrast, both (inflected) auxiliaries and verbs are verbal, while Cs are thought to encode typically verbal features like [finiteness] and are thus at least in part verbal too (cf. Biberauer, Newton & Sheehan 2008, 2009 for further discussion); consequently, we would expect FOFC to hold between V and C, but not V and D. A further class of exceptions comes from clause-final particles. Many VO languages have clause-final force particles (e.g. Mandarin Chinese, Sre, Mon, Copala Trique) (15) a.
Hufei chi-le sheme ma? Hufei eat-asp thing QYES/NO ‘Did Hufei eat anything?’
(Mandarin Chinese)
. An anonymous reviewer points out that DPs in German may not all be VP-internal, as it is clear that some DPs surface to the left of middle-field elements like VP-adverbs. This does not affect our central point since both base- and movement-generated FOFC-violating structures are predicted to be out (cf. Note 2). It is, however, worth noting that BHR do distinguish between linearization-driven movements and A- and A¢-driven movements, with the latter not being subject to FOFC. The core idea here is that movement that takes place as a result of its being part of an item’s selection information must be FOFC-conforming: it is not possible for a given head to select a head-initial phrase and require this selected phrase to move to its specifier (cf. BHR, Biberauer, Newton & Sheehan 2009 for more detail on this). By contrast, it is possible for an initial functional head to bear an A- or A′-movement trigger – i.e. a movement trigger that is not associated with its selection profile – which might then trigger a FOFCviolating structure. This is arguably the case wherever VP-fronting occurs in head-initial languages (cf. English and, more strikingly, the Slavic example discussed in BHR). In connection with DP-placement, it should, however, be remembered that both A- and A′-movement of DPs cannot trigger FOFC violations since FOFC only holds between categorially alike phrases. . There are many predecessors to this idea in the literature, from Kayne’s (1984) Connectedness proposals to Grimshaw’s (1991) ‘extended projection’ to van Riemsdijk’s (1998) ‘Law of Categorial Feature Magnetism’ and Brody’s (1995, 1998) ideas on projection.
Impossible changes and impossible borrowings
b. drùsĭ mâ zâ \aa rá tomorrow I meat eat affirmative ‘Tomorrow I will eat meat.’
(Lugbara, Nilo-Saharan) (cf. Heine & Nurse 2000: 208)
Systematic positional differences between “full” and particle forms of a given type of element (e.g. auxiliary, complementizer, etc.) in some of the languages with clause-final particles suggest that the Greenbergian approach to particles, in terms of which these are systematically excluded from word-order placement generalizations on account of their peculiar properties (cf. Greenberg 1963), may be well-founded. Consider (16), which shows that inflected auxiliaries are barred from the clause-final position in which (uninflected) auxiliary particles obligatorily appear: (16) a.
yә- ca dεyo lf 1sg-see picture asp ‘I am looking at a picture.’
(Bwe, Karen)
b. ce-1f mι jә-khf´ phι má nf´ (*jә-khf´) 3- say comp 3-fut take what ‘What did he say that he would take?’
We leave aside here (apparently) FOFC-violating structures containing particles, whose properties are still poorly understood. To summarize, then, empirical evidence from a wide range of languages suggests that FOFC, as stated in (17), is an absolute principle which acts as a universal constraint on synchronic grammars (modulo particles). (17) The Final-over-Final Constraint (FOFC) [Final Version]: If α is a head-initial phrase and β is a categorially non-distinct phrase immediately dominating α, then β must be head-initial. If α is a head-final phrase, and β is a phrase immediately dominating α, then β can be head-initial or head-final.
It presently remains unresolved whether FOFC follows from a principle of UG (i.e. Chomsky (2005)’s “first factor”), a processing preference, or extra-linguistic, economyoriented “third factor” pressures (cf. once again, Chomsky 2005). Biberauer, Holmberg & Roberts initially (2007, 2008) take it to be an effect of ‘phase harmony’ within a Linear Correspondence Axiom (LCA)-based system, and, later (Biberauer, Holmberg & Roberts 2010), of Relativized Minimality, once again within an LCA-based system, while Sheehan (2009a, b) proposes that it stems directly from Kayne’s (1994) LCA and a copy theory of labeling. These accounts, then, take FOFC to be at least in part due to first factor properties of the faculty of language, notably the LCA. Alternatively, it might be that FOFC is an effect of a very strong parsing/processing preference (cf. Hawkins 1994, 2004 and Cecchetto 2007, to appear), though see Sheehan (to appear) for a critical appraisal of Hawkins’ Performance Grammar Correspondence
Theresa Biberauer, Michelle Sheehan & Glenda Newton
Hypothesis and its relation to FOFC.5 Given the uncertain nature of this question, we leave it aside here, focusing instead on the central claim that FOFC, whatever its precise origins, is a universal constraint on synchronic grammars. As a universal constraint, we expect FOFC not only to constrain synchronic grammars, but also to play a role in acquisition, restricting the types of grammars that can be acquired and, thereby, also the types of diachronic change which are possible. The remainder of this chapter will consider FOFC’s role in diachronic change, both where contact is and where it is not involved.
3. FOFC and diachronic change If, as proposed in the previous section, FOFC is a universal constraint on synchronic grammars, it has important implications for syntactic change. In terms of an influential view in the field of diachronic syntax (see in particular Lightfoot 1999 and also Roberts 2007 for discussion), syntactic change should be seen as a random walk through parametric space. On this view, any syntactic change is possible if the primary linguistic data (PLD) to which the language learner is exposed provides the appropriate evidence or trigger for it. It therefore follows from this view that it is impossible to make predictions regarding the pathways that word-order change will follow (see also Newmeyer 1998). However, we propose that this is not the case. Universal constraints on grammar, such as FOFC, which categorically rule out certain structures, suggest that there may in fact be UG-imposed structural constraints that determine specific possible and impossible pathways of syntactic change.6 If FOFC is a universal constraint, then it is operative not only in today’s languages, but also in all languages of the past. In other words, FOFC must apply at every stage of a language’s history. This predicts that a FOFC-violating order should never be able to develop, even transitionally as part of a larger series of changes. Bearing this in mind, FOFC allows us to make predictions about pathways of syntactic change. For example, when a language changes from being predominantly head-final (“OV”) to predominantly head-initial (“VO”), this change must proceed top-down, as illustrated in (18): (18) [[[O V] Aux] C] → [C [[O V] Aux]] → [C [Aux [O V]]] → [C [Aux [V O]]]
As shown above, FOFC requires that CP must change first, giving C-AuxP order in place of AuxP-C order. The AuxP can then follow, giving Aux-VP in place of VP-Aux . Taken a certain way, Hawkins’ principles might actually be considered ‘third factor’ general economy principles. This idea is explicitly developed in Mobbs (2008). . As noted in the main text, we are not committed to the view that FOFC is the exclusive consequence of a hard-wired invariant principle of UG. Our use of UG here should thus be interpreted as leaving open the various possibilities raised in the main text.
Impossible changes and impossible borrowings
order. Only once these two changes are at the very least underway (see below) can the VP start to exhibit variation, possibly ultimately leading to change from OV to VO. If the change proceeded in the opposite direction, with the VP undergoing the change first, this would give rise to V-O-Aux and V-O-C, orders that are ruled out by FOFC, as discussed above. Similarly, if we consider the opposite change, that is, from head-initial (VO) to head-final (OV), FOFC predicts that this must proceed bottom-up. The VP must change first, followed by the AuxP, then the CP: (19) [C [Aux [V O]]] → [C [ Aux [O V]]] → [C [[O V] Aux]] → [[[O V] Aux] C]
A change in the opposite direction, beginning with the CP or the AuxP would result in the FOFC-violating V-O-C or V-O-Aux orders from the outset.7 The first of these predictions is borne out within the history of Indo-European. The change from OV to VO is well attested in both the Germanic and Romance branches and in both cases the change seems to follow the FOFC-determined pathway.8
3.1 OV to VO in the history of English Although Modern English is predominantly a head-initial language, this was not always the case. Like Modern German, Old English was head-final in the VP and the AuxP; therefore, English has clearly undergone a change from head-final to head-initial within its recorded history. FOFC predicts that the first stage of this change should affect the CP. From the earliest attested evidence of Germanic, we only find head-initial CPs.9 Crucially, however, initial complementizers may appear with both head-initial
. Crucially it should be noted that the pathways we propose are in no sense deterministic. If a consistently head-final language becomes head-initial in the CP, there is no necessity for it to undergo further change in the AuxP and the VP. Further syntactic change will only take place as a result of a change in the PLD available to children acquiring the language. See Section 3.3 below for further discussion with reference to synchronic word-order variation in Afrikaans. . Evidence of the second pathway (VO to OV) can be found in the Ethiopian Semitic languages, which have seemingly changed from head-initial to head-final due to contact with Cushitic. Biberauer, Newton & Sheehan (2008) discuss this case in some detail. . It is not clear that Germanic or, indeed, Proto-Indo-European ever had clause-final complementisers. Complementisers in the attested Indo-European languages seem to have developed independently, and, as such, we cannot reconstruct complementisers for ProtoIndo-European (see Kiparsky 1995). Just as there is no need for a system to progress all the way down a given FOFC-determined pathway once it has “started” at one of the extremes, so there is no need for a system which undergoes FOFC-sanctioned changes to have started at one of these extremes: our proposal is simply that changes will follow the available pathways, independently of their precise starting point.
Theresa Biberauer, Michelle Sheehan & Glenda Newton
and head-final AuxPs and VPs (although, as we saw above, not in the FOFC violating V-O-Aux order). This is shown in the examples below (formatting as above). (20) a. O-V-AUX: … þæt Darius hie mid gefeohte secan wolde that Darius them for battle seek wanted ‘… that Darius wanted to seek them out for a battle.’ (Orosius 45.31; cf. van Kemenade 1987: 16) b. V-O-AUX unattested
(cf. Pintzuk 1991, 1999)
c. AUX-O-V: … þæt hie mihton swa bealdlice Godes geleafan bodian that they could so boldly God’s faith preach ‘…that they could preach God’s faith so boldly.’ (The Homilies of the Anglo-Saxon Church I 232; cf. van Kemenade 1987: 179) d. AUX-V-O: … þæt he mot ehtan godra manna that he might persecute good men ‘… that he might persecute good men.’ (Wulfstan’s Homilies 130.37–38; Pintzuk 2002: 282)
Here we see that a final AuxP may only combine with a head-final VP (cf. (20a)), whereas an initial AuxP has two combination possibilities, being able to combine both with a head-final VP as in (20c) and with a head-initial VP as in (20d). The availability of (20d) at a stage at which (20c)-type structures are still attested indicates that variation in VP order becomes possible as soon as initial AuxP becomes available. Crucially, however, head-initial VPs are strictly limited to initial AuxP structures: despite the independent availability of both head-final AuxP and head-initial VP, structures of this type are completely unattested. Turning next to the AuxP, Pintzuk (1991/1999) proposes that the transition from head-final AuxP to head-initial AuxP was a gradual process, progressing throughout the Old English period until early Middle English, when it reached completion. Variation within the VP, between OV and VO order, however, continues until the Late Middle English period (formatting as before): (21) AUX-O-V: a.
Þat ne haue noht here sinnes forleten who neg have not their sins forsaken ‘…who have not forsaken their sins.’ (Trinity Homilies 67.934; cited in Kroch & Taylor 2000: 154)
b.
AUX-V-O oðet he habbe iŠetted ou al et Še wulle until he has granted you all that you desire ‘Until he has granted you all that you desire.’ (Ancrene Riwle; cited in Kroch & Taylor 2000: 145)
Impossible changes and impossible borrowings
It seems then, that the shift from OV to VO order in the history of English proceeded top-down, exactly as FOFC predicts. Biberauer, Newton & Sheehan (2008) show how the same is true of word-order change in Yiddish and Icelandic.
3.2 OV to VO in the history of French The development of VO order in French seems to have followed a similar pattern. Like the early Germanic languages, Latin exhibits only head-initial CPs: (22) accidit perincommode quod eum nusquam vidisti happened-3sg unfortunately that him nowhere saw-2sg ‘It is unfortunate that you didn’t see him anywhere.’ (Cicero, Ad Attticum 1, 17, 2; cited in Roberts 2007: 169)
The development of the AuxP in Romance languages is somewhat more complicated than in Germanic, as most “auxiliary” meanings in Latin were expressed morphologically rather than by means of auxiliaries (cf. Benveniste 1968 and Ledgeway to appear for overview discussion). Classical Latin does, however, feature one compound tense, namely the perfect passive. As (23) illustrates, the auxiliary element in this structure, esse (‘to be’), could either precede or follow the participle: (23) a.
illa quae cum rege est pugnata that-nom which-nom with king-abl is fought ‘that (battle) which was fought with the king.’ (Cicero, Pro Murena; cited in Ledgeway to appear: 8)
b. diu et acriter pugnatum est long-time and fierce fought is ‘There was a long and hard battle.’ (Caesar, De Bello Gallico; cf. Ledgeway to appear: 8)
The auxiliary habere develops in later Latin, and, like esse, its position is variable but the unmarked option seems to be for it to appear clause-finally (Bauer 1995: 104–107). This is to be expected as the preferred position of the lexical verb, from which the auxiliary habere grammaticalized, is clause-final (cf. i.a. Bauer 1995: 89–92):10 (24) a.
haec omnia probatum habemus these all.nom.pl tried.nom.sg have.1pl ‘We have tried all these things.’ (Oribasius – 6th century C.E.; cited in Ledgeway to appear: 62)
. The grammaticalization of the Romance future tense (see Roberts & Roussou 2003: 48–57), formed from infinitive+habere, also suggests that, in this construction at least, the auxiliary habere appeared clause-finally.
Theresa Biberauer, Michelle Sheehan & Glenda Newton
b. sicut parabolatum habuistis thus spoken had.2pl ‘Thus you had spoken.’ (Formulae Salicae Merkelianae; cited in Ledgeway to appear: 62)
As part of the transition from Vulgar Latin to French, both the AuxP and the VP became fixed as head-initial. Bauer (1995: 106) suggests that this change progressed in exactly the order predicted by FOFC, with the AuxP becoming head-initial before the VP: (25) “Whereas the ordering habeo/sum + participle prevailed in twelfth century Old French, the development towards the modern structure, where the direct object follows the compound verb [[habeo/j’ai [participle] [direct object]] is a development of Middle French. Accordingly, the structure (a) epistulam scriptam habeo/je possède une lettre écrite ‘I possess a written letter’ changed first into (b) habeo epistulam scriptam/j’ai une lettre écrite and then into (c) j’ai écrit une lettre ‘I wrote a letter’, where the direct object follows the auxiliary and the participle”
Although there is clearly a great deal more work to be done in determining the exact details of the change from OV to VO in French, with the relevant data possibly being absent from the attested record (Adam Ledgeway, p.c.), the basic evidence presented above is suggestive of a change following the pathway predicted by FOFC.
3.3 Word-order variation in Afrikaans FOFC pathways are not only demonstrated by diachronic change, but also by synchronic variation. In Modern Spoken Afrikaans (MSA), we find both verb-final (26a) and “verb-early” (26b) embedded clauses, in which a finite verb surfaces in a position clearly higher than VP: (26) a.
Ek weet dat sy [VP dikwels Chopin gespeel] het I know that she often Chopin played have ‘I know that she has often played Chopin.’
(verb-final)
b. Ek weet dat sy het [VP dikwels Chopin gespeel] (verb early) I know that she have often Chopin played ‘I know that she has often played Chopin.’
(26a) represents the older pattern, that is, the “correct” one prescribed by grammars; the “verb-early” construction is an innovation (the equivalent structure is barred in all varieties of Dutch). However, both structures are common in MSA and they are interpretively identical. In contrast, “verb-early” constructions with main verbs are far less common and necessarily have a “main-clause” interpretation (cf. i.a. Holmberg & Platzack 1995 on so-called embedded root phenomena in V2 languages).
Impossible changes and impossible borrowings
(27) Hy dink dat sy [VP speel altyd Chopin] he think that she play always Chopin ‘He thinks that she always plays Chopin’
Since (27)-type structures systematically behave differently to the “verb-early” ones in (26), Biberauer (2003) proposes that only the (27)-type structures involve V2; the alternations in (26)-type structures, by contrast, feature auxiliaries located in Aux. As such, they constitute evidence that MSA permits both final ((26a)) and initial ((26b)) AuxP structures. Crucially, however, MSA does not permit head-initial VPs, as earlier English, for example, did (cf. (20c, d) above). This seems to be directly attributable to the salience of particle verbs in MSA (cf. Ponelis 1993). As argued by Lightfoot (1979 and following), particles serve as “signposts” signalling the location of the verb with respect to the object: where a particle precedes the object, as in VO languages, the acquirer can conclude that VP is initial; where it follows, as in OV languages, VP must be final. Against this background, robust attestation of particle verbs in the MSA input would be expected to contribute a clear signal to the acquirer that the system being acquired involves a final VP. That this reasoning is correct is strongly suggested by the fact that Kaaps, a variety of Afrikaans spoken in the Cape in which English borrowings/substitutions have drastically reduced the number of particle verbs, permits initial VPs in the presence of an initial TP. As in the earlier English case, initial VP is not compatible with final TP. The relevant facts are illustrated in (28): (28) a. Ek het [VP gekry R1400 van die Revenue] [AUX-V-O] I have received R1400 from the Revenue ‘I have received R1400 from the Receiver of Revenue.’ [Standard Afrikaans: Ek het R1400 van die Department van Inkomste gekry, i.e. [O-V-AUX]] b. *dat ek [VP gekry R1400 van die Revenue] het [*V-O-AUX] that I received R1400 from the Revenue have ‘I have received R1400 from the Receiver of the Revenue.’
Kaaps, then, appears to be one step further along the FOFC-predicted pathway than MSA. Crucially, this state of affairs reflects the fact that FOFC is not itself a driver of syntactic change, but merely a constraint which defines possible and impossible diachronic pathways. For change to occur, it needs, as Longobardi (2001: 278) points out, to be the “well-motivated consequence of other types of change (phonological changes and semantic changes, including the appearance/disappearance of whole lexical items) or, recursively, of other syntactic changes” (cf. also Keenan 2002 on this so-called Inertia Principle). Syntax therefore does not change unless there is specific input (which we might think of as a ‘cue’ – cf. i.a. Lightfoot 1991, 1999; Dresher 1999) signaling to the
Theresa Biberauer, Michelle Sheehan & Glenda Newton
acquirer that a grammar different to that initially acquired by the previous generation should be postulated (cf. Roberts 2007 for recent overview discussion of this general approach to syntactic change). In the following section we will consider a further example of word-orderrelated change that clearly involved a prior change, this time from typologically very different languages.
4. FOFC and borrowing Contact-induced change is of particular interest in the present context as it has sometimes been suggested that this type of change, in contrast to the non-contact-induced variety, may result in typologically unusual linguistic systems. Harris & Campbell (1995: 239), for example, claim that typologically rare word orders are often the result of contact-induced change, implying that the latter might fall beyond the remit of typological universals. Nonetheless, while it might be the case that ‘exotic’ word orders often arise as a result of borrowing, there is no reason to believe that borrowings will be immune to grammatical constraints (cf. also Bowern 2008 for a discussion). As far as FOFC is concerned, initial empirical research suggests that borrowing with wordorder implications is constrained in exactly the same way as change which may be less directly contact-induced (cf. the discussion in the previous section). We will consider just one case study here. The South Asian linguistic area provides an excellent testing ground for FOFC’s effect on borrowing as the more rigidly head-final Dravidian languages have a long history of contact with the more disharmonic Indo-Aryan languages. As noted by Hock & Joseph (1996: 61), there is controversy over when Indo-Aryan and Dravidian first came into contact, but it is uncontroversial that they came to “structurally converge after multilingual contact extending over several millennia”. According to Thomason & Kaufman (1988: 43), there is evidence of an early, shift-induced, Dravidian influence on Indic from the spread of retroflex consonants, despite the virtual lack of lexical borrowing (cf. also Kuiper 1974). Moreover, there is little dispute that many Indo-Aryan languages have borrowed many lexical and grammatical features from Dravidian in the more recent past, through continued contact. Interestingly, there are clear patterns across the area with respect to certain grammatical features, which, we argue, cannot be satisfactorily explained by sociolinguistic or genetic factors alone. In at least one such case, patterns of borrowing/development can be explained by grammatical factors, more specifically FOFC. It is well known that Indo-Aryan languages show variation in the placement of complementizers and polarity question particles (cf. Masica 1991; Marlow 1997; Bayer 1999, 2001; Davison 2007). It is generally assumed that Indo-Aryan borrowed final
Impossible changes and impossible borrowings
quotative complementizers from Dravidian, either very early on (Kuiper 1974) or at a late stage of contact (Meenakshi 1986), or perhaps both (but cf. Hock 1982 for an alternative account).11 In fact, while Sanskrit did have the (largely) final complementizer iti (ti/tti in Middle Indo-Aryan), this was later lost and no cognate remains in any of the modern mainland Indo-Aryan languages (Meenakshi 1986; Marlow 1997; Southworth 2005).11Instead, Modern Indo-Aryan languages either lack a final complementizer altogether or display final complementizers of the quotative type, typically derived from the verb of saying (cf. Bengali/Bangla bole ‘saying’), or from the demonstrative root (cf. Marathi asa ‘such’) or both in the case of Marathi (Bayer 1999, 2001; Masica 1991; Davison 2007). Figure 1 shows the distribution of these complementizers (Cs),12 based on appendix A of Davison (2007).13 No final C
Final C from ‘saying’
Hindi/Urdu, Panjabi, Sindhi, Sinhala, Dhivehi, Marathi, Kashmiri, Maithili, Kurmali Nepali, Dakkhini Hindi Assamese, Bengali/Bangla, Oriya
Final C from demonstrative Marathi, Gujarati
Figure 1. Distribution of final complementizers in Indo-Aryan
Interestingly, all modern Indic languages (with the exception of Sinhala, spoken in Sri Lanka) also have an initial C, the origins of which vary, and are not to be discussed at
. Southworth (2005) also notes that the Munda languages, which constitute a branch of the Austro-Asiatic language family, exhibit a similar pattern, so the borrowing might not necessarily have been from Dravidian. This is not crucial for our purposes as the Munda languages would also represent a typologically distinct borrowing source. . Although we will abbreviate complementizer as C, as is commonly done in Chomskyan work, it should be noted that we do not intend this abbreviation to be read as an indication of our interpreting the C-elements in the languages under discussion as Cs equivalent to Englishstyle complementizers. As seems fairly clear from the examples that follow, the elements in question would seem to lexicalize a rather different sub-portion of Rizzi’s (1997) articulated CP to the finite Cs familiar from Germanic and Romance. For example, while English that seems to encode both Force and Finiteness, with the result that we might think of it as a syncretised Force-Finiteness element (cf. Giorgi & Pianesi 1997), the Marathi and Hindi-Urdu Cs illustrated in (32–33) only seem to encode subordination, which may be a (sub-)species of Force (cf. Biberauer, Newton & Sheehan 2009 for further discussion). . We call these final ‘quotatives’ complementizers as they are not limited to contexts of direct/indirect speech, but rather occur with epistemic and perception verbs such as know, think and hear (cf. Bayer 2001).
Theresa Biberauer, Michelle Sheehan & Glenda Newton
length here (cf. Meenakshi 1986; Marlow 1997 for discussion).14 For our purposes, the interest lies in the varying availability of a final C in the languages under consideration. All Indic languages exhibit a degree of typological similarity with Dravidian. Typological similarity is known to aid borrowing (cf. Thomason & Kaufman 1988: 97–98), and in this case, it did so by making available the potential structure [[O V] say] in the native language. While it is true that many of the languages with a final C have been in heavy contact with Dravidian (e.g. Marathi, Gujarati), others have not (e.g. Nepali). This means that in at least this language, the structure must either have developed internally (cf. Klamer 2001) or been borrowed from another source (perhaps a neighbouring Tibeto-Burman language). See Marlow (1997) for a full discussion of these facts and the conclusion that the distribution of final quotatives in the area cannot be explained in purely historical/geographical terms. The question arises, then, as to whether this borrowing also follows the path of diachronic change predicted by FOFC. In fact, there is good evidence that it does. Davison’s (2007) observation is that those Indo-Aryan languages which lack a final complementizer are exactly those which have an initial Polarity question head (Pol) of the type illustrated here for Hindi-Urdu and Panjabi: (29) kyaa aap wahaaN aaeeNgii? pol you there come.fut.2pl ‘Are you coming?’
(Hindi-Urdu, Davison 2007: 180)
(30) (kii) tuslïï ajj káánii suNaavoge? pol you today story tell.fut.2mp ‘Will you tell a story today?’ (Panjabi, Davison 2007: 180, citing Bhatia 1993: 5)
Other languages in the area have non-initial Pol heads, which occur in final position or in final/medial position, but crucially never initially: (31) a.
to kaal parat aalaa kaa(y)? he yesterday back come.pst.3s pol ‘Did he come back yesterday?’
. The two main types of initial complementizers are the (Sanskrit-like) j-series, derived from the relative pronoun, and the k-series. Although it is commonly assumed that kicomplementizers in Hindi-Urdu and other Indo-Aryan languages (minus the Eastern branch) are a borrowing from Persian (Kellogg 1893), the initial complementizer kimti/kiti is also attested in the Inscriptional Prakrits, and might be the actual source of the modern form (Meenakshi 1986: 212). Moreover, as Bayer (2001) points out, ki is similar in form to the wh-paradigm, which has served as a source for complementizers in Romance, Slavic and Greek branches of Indo-European. Interestingly, Persian ke has been passed on to one Indo-European language as well as at least three other language families through contact: Indo-European (Asia Minor Greek), Altaic (several Turkic languages, Kartvelian (Laz), Nakho-Daghestanian (Lezgian)), and (Northern) Dravidian (Brahui) (cf. Haig 2001 for discussion).
Impossible changes and impossible borrowings
b. *kaa(y) to kaal parat aalaa pol he yesterday back come.pst.3ms (Marathi, based on Davison 2007: 181, citing Pandharipande 1997: 8) (32) a.
modhu aS-be ki (na)? Madhu come-fut pol not ‘Will/Won’t Madhu come?’
b. modhu ki aS -be? Madhu pol come-fut ‘Will Madhu come?’15 c. *ki modhu aS-be? pol Madhu come-fut (Bengali/Bangla, Davison 2007: 189, attributed to P. Dasgupta)
Previous research suggests that C is uniformly higher than the Pol head involved in yes/no questions (Laka 1994; Rizzi 2001; Holmberg 2003, but cf. Biberauer 2008 for the positing of a higher polarity head involved in negation). In fact, there is good evidence from Indic that C dominates Pol. As Davison shows, in many, but not quite all of the modern languages (the exceptions being Bengali/Bangla and other Eastern Indic languages), Pol can co-occur with C in embedded clauses.16 Where this happens, final Pol precedes final C, whereas initial Pol follows initial C. Moreover, the reverse orders are ungrammatical, as shown for Hindi-Urdu: (33) [[to kal parat aalaa kaa(y)] mhaaNun/asa] he yesterday back come.pst.3ms pol quot/such raam malaa witSaarat hotaa Ram I.dat ask.prog be.pst.3ms
‘Ram was asking me whether/if he came back yesterday.’ (Marathi, Davison 2007: 182, attributed to R. Pandharipande)
(34) a.
us-nee puuc-aa [ki [kyaa tum aa-oogee]] 3s-erg ask-perf that pol you come-fut ‘He asked whether you will come.’
b. *us-nee puuc-aa [kyaa ki tum aa-oogee] 3s-erg ask-perf pol that you come-fut ‘He asked whether you will come.’ (Hindi-Urdu, Davison 2007: 181)
. The positioning of the polarity head appears to be connected to focus, according to Davison (2007). See Nayudu (2008) for a similar effect in the domain of negation in Marathi. . The pattern in Bangla is reminiscent of that seen in Sanskrit and is probably connected to the fact that initial complementizers in Bangla and the other Eastern languages are derived from relative pronouns, whereas Western languages have borrowed Persian ki (cf. Bayer 1999, 2001; Davison 2005)
Theresa Biberauer, Michelle Sheehan & Glenda Newton
Where C is initial and Pol is final, the two can still co-occur: (35) raam maalaa witSaarat hotaa [ki to kal parat Ram I.dat ask.prog be.pst.3ms that he yesterday back aalaa kaa(y)] come.pst.3ms pol ‘Ram was asking me whether/if he came back yesterday.’ (Marathi, Davison 2007: 182, attributed to R. Pandharipande)
This is not problematic from the perspective of FOFC as, assuming that C is higher than Pol, this represents an inverse FOFC order, which is permitted by UG: (36) [CPC [PolP [AuxP] Pol ]] CP C
PolP AuxP
Pol
Under our assumptions, though, an initial Pol should block the borrowing/development of a final C as this would lead to the FOFC-violating structure in (37): (37)
* [CP [PolP Pol [AuxP]] C ] CP
* PolP Pol
C AuxP
This prediction appears to hold quite robustly in the area. As Davison shows at length, no language in the South Asian linguistic area has both initial Pol and final C, although all three of the other possible combinations readily occur. Hindi-Urdu, for example has not developed a final C of any kind, from either a verbal or demonstrative source. Thus both potential types of final C lead to ungrammaticality as shown in (38):17
. Bhatt & Takahashi (to appear) note that Hindi-Urdu postpositions are also unable to take clausal complements and this is potentially another instance of the same effect: the Pol head, whether overt or not, blocks the possibility of a final selecting (i.e. higher) head, regardless of whether this selector is verbal or adpositional. On the assumption that selectors share the property of being [-N] (cf. Chomsky 1981), this can be related to the categorial nondistinctness desideratum entailed by the FOFC constraint (cf. (17)). It is also worth noting that if a null Pol head can indeed “count” for FOFC purposes – in barring the postulation of
Impossible changes and impossible borrowings
(38) *usee [[ vee aa rahee haiN] yah/ kah-kar ] maaluum hai 3s.dat 3pl come prog are this/ say-prt known is ‘He/she knows that they are coming.’ (Hindi-Urdu, Davison 2007: 177)
Figure 2, adapted from Appendix B of Davison (2007), shows the pattern in the relevant area. In type A languages, the direct borrowing of a final C has been blocked by the presence of an initial Pol, with the result that type D languages, which would be the FOFC-violating type, are unattested. Type
Position of Pol
Position of C
Languages
A
Initial
Initial Only
Hindi-Urdu, Panjabi, Kashmiri, Sindhi, Maithili, Kurmali (plus some North Dravidian languages, i.e. Brahui)
B
Final/Medial
Initial and Final Marathi, Gujarati, Assamese, Bengali/ Bangla, Dakhini Hindi, Oriya, Nepali
C
Final/Medial
Final Only
Sinhala (plus some Dravidian languages)
D
Initial
Final
Unattested in the area (adapted from Davison 2007 Appendix B)
Figure 2. Polarity heads and complementizers in South Asia
One apparent problem with this proposal is that Sanskrit had both an initial question particle kim and an optional final quotative iti (cf. Davison 2005). It appears, though, that the syntax of kim was very different from that of modern Indo-Aryan Polarity heads, as kim failed to occur in embedded clauses and so would not have given rise to FOFC-violating structures (cf. Davison 2005 for a discussion of the syntax of iti). Assuming that the polarity heads in Modern Indo-Aryan are derived from initial kim, the prediction is that iti/tti must have been lost before kim was reanalysed as a Polarity head so that it could appear in embedded clauses. Further work is required to verify this claim. Of course, it remains true that most of the type A languages in Figure 2 have had little contact with Dravidian, and this in turn might explain why they have failed to develop either a final Polarity head or a final C. To verify that Figure 2 is really evidence for the FOFC-predicted pathway, it would have to be shown that development/borrowing of final Pol preceded borrowing/development of final C. Once again, further diachronic work is needed to confirm this. Further support for the FOFC-based prediction regarding the distribution of initial question particles and final complementizers comes from evidence from WALS a FOFC-violating structure – this constraint cannot be ascribed to processing considerations (cf. Cecchetto 2007, to appear for one proposal along these lines).
Theresa Biberauer, Michelle Sheehan & Glenda Newton
indicating that this pattern is not limited to South Asia, but is replicated throughout the languages of the world (cf. Dryer 2005a, 2005b). From a sample of 195 languages, the numbers of languages roughly equivalent to type A-D languages are as follows:18 Type Position of Pol
Position of C
Number of Languages (genera: families)
A
Initial
Initial Only
72:35:13
B
Final
Initial
74:40:16
C
Final
Final Only
45:33:20
D
Initial
Final
4:3:3
Figure 3. Typological positioning of Polarity heads and complementizers
Clearly there is a skewing in the data here, with large numbers of genetically diverse languages of types A-C. FOFC-violating type D is not, however, completely unattested. Four languages from three language families, all found in South America, are of this type: Tacana and Ese Ejja (Tacanan), Gavião (Tupi) and Resígaro (Arawakan). In fact, all of these languages appear to nominalize their embedded clauses (cf. Moore 1989 on Gavião, Allin 1976 on Resígaro and Ottaviano 1980 on Tacana) and so this might be the reason why FOFC does not appear to hold. Further research is required to establish exactly what the relationship between clausal nominalization and FOFC really is (cf. Biberauer & Sheehan to appear for some discussion). For the moment, it would, however, seem that there is good evidence that FOFC has constrained the borrowing of final complementizers in Indo-Aryan. Only those languages which lack an initial Polarity head have developed/borrowed a final C. Further evidence that this is what is at stake in Indic comes from typological trends, which mirror the skewing in Indo-Aryan.
5. Conclusion The aim of this chapter was to argue that FOFC, a universally valid syntactic principle in the domain of word order, constrains synchronic grammars and therefore . The data presented here is actually for “position of polar question particles” and “order of adverbial subordinator and clause”. These have, however, been cleansed to make them more comparable with the Indic facts, with other values such as second-position question particles removed. Dryer uses the term adverbial subordinators to refer to “because, although, when, while, and if”. While these subordinators often pattern with the positioning of more unequivocal C-elements such as that, this is not always the case. For this reason, these data are taken to be suggestive only (cf. also the discussion in Note 12).
Impossible changes and impossible borrowings
also diachronic change, regardless of whether this is triggered by contact or not. Evidence suggests that FOFC-based predictions hold in the history of English, Afrikaans and, insofar as relevant data are available, French. Moreover, patterns of final-complementizer borrowing in Indic languages also seem to be attributable to FOFC, the former being blocked by the presence of an initial Pol head lower in the clause. Taken together, we argue that these facts suggest that the notions ‘impossible change’ and ‘impossible borrowing’ may after all be contentful, thus justifying further investigation.
References Allin, Trevor R. 1976. A Grammar of Resígaro. Horsleys Green: Summer Institute of Linguistics. Bauer, Brigitte. 1995. The Emergence and Development of SVO Patterning in Latin and French: Diachronic and Psycholinguistic Perspectives. Oxford: OUP. Bayer, Josef. 1999. Final complementizers in hybrid languages. Journal of Linguistics 35: 233–271. Bayer, Josef. 2001. Two grammars in one: Sentential complements and complementizers in Bengali and other South Asian languages. In The Yearbook of South Asian Languages: Tokyo Symposium on South Asian Languages – Contact Convergence and Typology, Peri Bhaskararao & Karumuru Venkata Subbarao (eds), 11–36. New Delhi: Sage. Benveniste, Emile. 1968. Mutations of linguistic categories. In Directions for Historical Linguistics, Winifred P. Lehmann & Yakov Malkiel (eds), 85–94. Austin TX: University of Texas Press. Besten, Hans den. 1986. Decidability in the syntax of verbs of (not necessarily) West Germanic languages. Groninger Arbeiten zur Germanistischen Linguistik 28: 232–256. Reprinted as: den Besten, Hans. 1989. Studies in West Germanic Syntax, 111–135. Amsterdam: Rodopi. Bhatia, Tej K. 1993. Punjabi: A Cognitive-Descriptive Grammar. London: Routledge. Bhatt, Rajesh & Takahashi, Shoichi. To appear. Reduced and unreduced phrasal comparatives. In Natural Language and Linguistic Theory. Biberauer, Theresa. 2003. Verb Second (V2) in Afrikaans: A Minimalist Investigation of WordOrder Variation. Ph.D. dissertation, University of Cambridge. Biberauer, Theresa. 2008. Doubling and omission: Insights from Afrikaans negation. In Microvariations in Syntactic Doubling, Sjef Barbiers, Olaf Koeneman, Marika Lekakou & Margreet van der Ham, (eds), 103–140. Bingley: Emerald. Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2007. Disharmonic word-order systems and the Final-over-Final-Constraint (FOFC). In Proceedings of XXXIII Incontro di Grammatica Generativa, Antonietta Bisetto & Francesco Barbieri (eds), 86–105. Bologna: Università de Bologna. Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2008. Structure and Linearization in Disharmonic Word Orders. In Proceedings of the 26th West Coast Conference on Formal Linguistics, Charles B. Chang & Hannah J. Haynie (eds), 96–104. Somerville MA: Cascadilla Proceedings Project. Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2010. A syntactic universal and its consequences. Ms, Universities of Cambridge and Newcastle. Biberauer, Theresa, Newton, Glenda & Sheehan, Michelle. 2008. Defining diachronic pathways: The Final-over-Final Constraint. Paper presented at DiGS X (Cornell).
Theresa Biberauer, Michelle Sheehan & Glenda Newton Biberauer, Theresa, Newton, Glenda & Sheehan, Michelle. 2009. Limiting synchronic and diachronic variation and change: The Final-Over-Final Constraint. Language and Linguistics 10(4):701–743. Biberauer, Theresa & Sheehan, Michelle. To appear. Disharmony, antisymmetry, and the Final-over-Final Constraint. In Ways of Structure Building, Myriam Uribe-Etxebarria & Vidal Valmala (eds), Cambridge: CUP. Bowern, Claire. 2008. Syntactic change and syntactic borrowing in generative grammar. In Principles of Syntactic Reconstruction, Gisella Ferarresi & Maria Goldbach (eds), 187–216. Amsterdam: John Benjamins. Brody, Michael. 1995. Lexico-Logical Form: A Radically Minimalist Theory. Cambridge MA: The MIT Press. Brody, Michael. 1998. Projection and phrase structure. Linguistic Inquiry 29: 367–398. Cecchetto, Carlo. 2007. Some preliminary remarks on a ‘weak’ theory of linearization. Annali Online di Ferrara – Lettere 1: 1–13. Cecchetto, Carlo. To appear. Backwards dependencies must be short. A unified account of the Final-over-Final and the Right Roof Constraints and its consequences for the syntax/morphology interface. In Challenges to Linearization, Theresa Biberauer & Ian Roberts (eds). Berlin: Mouton de Gruyter. Chomsky, Noam. 1981. Lectures on Government and Binding Theory. Dordrecht: Foris. Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36: 1–22. Davison, Alice. 2005. Syntactic adjunction: Consequences for correlatives in Sanskrit and Hindi/ Urdu. Paper presented at Generative Linguistics in the Old World, GLOW in Asia V. (Delhi) Davison, Alice. 2007. Word order, parameters, and the Extended COMP projection. In Linguistic Theory and South Asian Languages, Essays in Honour of K.A. Jayaseelan [Linguistik Aktuell/ Linguistics Today 102], Josef Bayer, Tanmoy Bhattacharya & M.T. Hany Babu (eds), 175–198. Amsterdam: John Benjamins. Dresher, Elan. 1999. Charting the learning path: Cues to parameter setting. Linguistic Inquiry 30: 27–67. Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 81–138. Dryer, Matthew. 2005a. The position of polar question particles. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds), 374–377. Oxford: OUP. Dryer, Matthew. 2005b. The order of adverbial subordinator and clause. In The World Atlas of Language Structures, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds), 382–385. Oxford: OUP. Evers, Arnold. 1975. The Transformational Cycle in Dutch and German. Ph.D. dissertation, University of Utrecht. Fischer, Olga, van Kemenade, Ans, Koopman, Willem & van der Wurff, Wim. 2000. The Syntax of Early English. Cambridge: CUP. Fuß, Eric & Trips, Carola. 2002. Variation and change in Old and Middle English: On the validity of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics 4: 171–224. Giorgi, Alessandra & Pianesi, Fabio. 1997. Tense and Aspect. From Semantics to Morphosyntax. Oxford: OUP. Greenberg, Joseph. 1963. Some universals of grammar with particular reference to the order of meaningful elements. In Universals of Grammar, Joseph Greenberg (ed.), 73–113. Cambridge MA: The MIT Press.
Impossible changes and impossible borrowings
Grimshaw, Jane. 1991. Extended projections. Ms, Brandeis University. Published as: Grimshaw, Jane. 2000. Extended projection and locality. In Lexical Specification and Insertion [Current Issues in Linguistic Theory 197], Peter Coopmans, Martin Everaert & Jane Grimshaw (eds), 115–134. Amsterdam: John Benjamins. Haddican, William. 2004. Sentence polarity and word order in Basque. The Linguistic Review 21: 87–124. Haegeman, Liliane & Riemsdijk, Henk van. 1986. Verb projection raising, scope and the typology of rules affecting verbs. Linguistic Inquiry 7: 417–466. Haider, Hubert. 1993. Deutsche Syntax, generativ. Tübingen: Narr. Haider, Hubert. 2000. OV is more basic than VO. In The Derivation of VO and OV [Linguistik Aktuell/Linguistics Today 31], Peter Svenonius, (ed.) 45–67. Amsterdam: John Benjamins. Haig, Geoffrey. 2001. Linguistic diffusion in present-day East Anatolia: From top to bottom. In Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, Alexandra Y. Aikhenvald & Robert M.W. Dixon, (eds) 195–224. Oxford: OUP. Harris, Alice & Campbell, Lyle. 1995. Historical Syntax in Crosslinguistic Perspective. Cambridge: CUP. Hawkins, John. 1983. Word Order Universals. New York NY: Academic Press. Hawkins, John. 1990. Explaining Language Universals. Oxford: Blackwell. Hawkins, John. 1994. A Performance Theory of Order and Consituency. Cambridge: CUP. Heine, Bernd & Nurse, Derek. 2000. African Languages: An Introduction. Cambridge: CUP. Hock, Hans Henrich. 1982. The Sanskrit quotative: A historical and comparative study. Studies in the Linguistic Sciences 12(2): 39–85. Hock, Hans Henrich & Joseph, Brian. 1996. Language History, Language Change and Language Relationship: An Introduction to Historical and Comparative Linguistics. Berlin: Mouton de Gruyter. Holmberg, Anders. 2000. Deriving OV order in Finnish. In The Derivation of VO and OV [Linguistik Aktuell/Linguistics Today 31], Peter Svenonius (ed.), 123–152. Amsterdam: John Benjamins. Holmberg, Anders. 2003. Yes/no questions and the relation between tense and polarity in English and Finnish. Linguistic Variation Yearbook 3: 45–70. Holmberg, Anders & Platzack, Christer. 1995. The Role of Inflection in Scandinavian Syntax. Oxford: OUP. Hróarsdóttir, Thorbjörg. 2000. Verb Phrase Syntax in the History of Icelandic. Ph.D. dissertation, University of Tromsø. (Also published as: Hróarsdóttir, Thorbjörg. 2000. Word Order Change in Icelandic: From OV to VO [Linguistik Aktuell/Linguistics Today 35]. Amsterdam: John Benjamins. Kayne, Richard. 1984. Connectedness and Binary Branching. Dordrecht: Foris. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Keenan, Edward. 2002. Explaining the creation of reflexive pronouns in English. In Studies in the History of English: A Millenial Perspective, Donka Minkova & Robert Stockwell (eds), 325–355. Berlin: Mouton de Gruyter. Kellogg, Samuel. 1893. A Grammar of the Hindi Language. London: Kegan Paul & Co. Kemenade, Ans van. 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kiparsky, Paul. 1995. Indo-European origins of Germanic syntax. In Clause Structure and Language Change, Adrian Battye & Ian Roberts (eds), 140–170. Oxford: OUP.
Theresa Biberauer, Michelle Sheehan & Glenda Newton Kiparsky, Paul. 1996. The shift to head-initial VP in Germanic. In Studies in Comparative Germanic Syntax II, Höskuldur Thráinsson, Samuel Epstein & Steve Peter (eds), 140–179. Dordrecht: Kluwer. Klamer, Marian. 2001. Expressives and iconicity in the lexicon. In Ideophones [Typological Studies in Language 44], Erhard Voeltz & Christa Kilian-Hatz (eds), 165–181. Amsterdam: John Benjamins. Kroch, Anthony & Taylor, Ann. 2000. Verb-object order in Middle English. In Diachronic Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 132–163. Oxford: OUP. Kuiper, Franciscus Bernardus Jacobus. 1974. The genesis of a linguistic area. International Journal of Dravidian Languages 3(1): 135–153. Laka, Itziar. 1994. On the Syntax of Negation. New York NY: Garland. Ledgeway, Adam. To appear. Syntactic and morphosyntactic typology and change in Latin and Romance. In The Cambridge History of the Romance Languages, Martin Maiden, John Charles Smith & Adam Ledgeway (eds). Cambridge: CUP. Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: CUP. Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge MA: The MIT Press. Lightfoot, David. 1999. The Development of Language. Oxford: Blackwell. Longobardi, Giuseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The history of French chez. Linguistic Inquiry 32: 275–302. Marlow, Patrick Edward. 1997. Origin and Development of the Indo-Aryan Quotatives and Complementizers: An Areal Approach. Ph.D. dissertation, University of Illinois. Masica, Colin. 1991 The Indo-Aryan languages. Cambridge: CUP. Meenakshi, K. 1986. The quotative in Indo-Aryan. In South Asian Languages: Structure, Convergence and Diglossia, Bhadriraju Krishnamurti (ed.), 209–218. Dehli: Motilal Banarsidass. Mobbs, Iain. 2008. ‘Functionalism’, the Design of the Language Faculty, and (Disharmonic) Typology. Mphil thesis, University of Cambridge. Moore, Denny. 1989. Gaviao nominalisations as relative clause and sentential complement equivalents. IJAL 55: 309–25. Müller, Reinmar & Reis, Marga. 2001. Modalität und Modalverben im Deutschen. Hamburg: Buske Verlag. Nayudu, Aarti. 2008. Issues in the Syntax of Marathi: A Minimalist Approach. Ph.D. dissertation, Durham University. Newmeyer, Frederick. 1998. Language Form and Language Function. Cambridge MA: The MIT Press. Ottaviano, Ida de. 1980. Textos tacana. Riberalta, Bolivia: Instituto Linguístico de Verano en colaboración con el Ministerio de Educación y Cultura. Pandharipande, Rajeshwari. 1997. Marathi. London: Routledge. Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and Change in Old English Word Order. Ph.D. dissertation, University of Pennsylvania. (Also published as: Pintzuk, Susan. 1999. Phrase Structures in Competition. Variation and Change in Old English Word Order. New York NY: Garland). Pintzuk, Susan. 2002. Verb order in Old English: Variation as grammatical competition. In Diachronic Syntax: Models and Mechanisms, David Lightfoot (ed.), 132–163. Oxford: OUP. Ponelis, Frederick. 1993. The Development of Afrikaans. Duisburg: Peter Lang.
Impossible changes and impossible borrowings
Riemsdijk, Henk van. 1998. Categorial feature magnetism: The endocentricity and distribution of projections. Journal of Comparative Germanic Linguistics 2: 1–48. Rizzi, Luigi. 1997. On the fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281–338. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the position ‘Int(errogative)’ in the left periphery of the clause. In Current Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque & Giampaolo Salvi (eds), 287–296. Amsterdam: Elsevier. Roberts, Ian. 2007. Diachronic Syntax. Oxford: OUP. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Sheehan, Michelle. 2009a. The Final-over-Final Constraint as an effect of Complement Stranding. Newcastle Working Papers in Linguistics 15: 104–125. Sheehan, Michelle. 2009b. Labelling, multiple Spell-Out and the Final-over-Final Constraint. In Proceedings of XXXV Incontro di Grammatica Generativa, Vincezo Moscati & Emilio Servidio (eds), 231–243. Siena: CISCL. Sheehan, Michelle. To appear. Formal and functional approaches to disharmonic word orders. To appear in Newcastle Working Papers in Linguistics 16. Southworth, Franklin C. 2005. Linguistic Archaeology of South Asia. London: Routledge-Curzon. Thomason, Sarah. G. & Kaufman, Terrence. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley CA: University of California Press. Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT.
Continuity is change The long tail of Jespersen’s cycle in Flemish* Anne Breitbarth & Liliane Haegeman Ghent University
All West Germanic languages have completed Jespersen’s Cycle, replacing a preverbal negation marker (stage I) with a post-verbal one (stage III) after a period of transition during which the two could co-occur (stage II). Only the Flemish dialects have maintained the old preverbal marker to some extent. The present chapter addresses this apparent historical continuity, establishing that Flemish en was in fact saved from the fate of its West Germanic cognates by a morphosyntactic change. Based on an analysis of the contexts in which preverbal en is found in present-day Flemish dialects, it is argued that en has evolved into a marker of emphasis of polarity, and that contrary to appearances, the Flemish dialects have in fact reached stage III.
1. Introduction All West Germanic languages have undergone a development in the expression of negation known as Jespersen’s Cycle (term by Dahl 1979, after observations by Jespersen 1917), by which an original preverbal negator is first joined by a newly grammaticalized adverbial marker and eventually supplanted by it. The rough stages of the cycle are demonstrated for English in (1).1
*We gratefully acknowledge funding by the Arts and Humanities Research Council, award no. AR119272 (Anne Breitbarth) and the Vlaamse Fonds voor Wetenschappelijk Onderzoek, 2009-Odysseus grant no. G091409 (Liliane Haegeman). . More fine-grained subdivisions are possible and have been proposed in great numbers and variety, cf. e.g. Posner (1985), Schwegler (1988), Ladusaw (1993), Rowlett (1998), Van Kemenade (2000), Zeijlstra (2004), Schwenter (2006), Barbiers et al. (2008) and Willis (to appear). The rough division into three stages given in (1) is, however, sufficient for the purposes of this chapter.
Anne Breitbarth & Liliane Haegeman
(1) Stage I ic ne secge Old English
Stage II
Stage III
Stage I’
I seye not
I say not
Middle English
Early Modern English
I do not say > I don’t say Modern English
Jespersen (1917: 9–11)
All West Germanic languages have, furthermore, like English, completed this development, that is, they have all reached at least stage III. As is well known, the only exception to this is the Flemish dialects, which, as widely discussed in the literature (Haegeman 1995 and much subsequent work; Haegeman & Zanuttini 1991, 1996; Barbiers et al. 2008), have preserved, apparently optionally, the use of the preverbal marker. They therefore appear to be still in transition from stage II to stage III. (2) Valère (en) klaapt nie. Valère ‘neg’ talks ‘neg’ ‘Valère doesn’t talk.’
(West Flemish (Lapscheure); Haegeman 1995)
The question the present chapter wishes to address is why these Flemish dialects show this apparent historical continuity. By looking in detail at the contexts in which preverbal en is found in present-day Flemish dialects we will establish that it is in fact a morphosyntactic change that saved Flemish en from the fate of its other West Germanic cognates, and we will argue that, contrary to appearances, the Flemish dialects are not in transition between stage II and stage III of Jespersen’s Cycle, but have reached stage III.
2. Present-day Flemish en As just indicated, the Flemish dialects are known for having preserved the old preverbal marker en in the expression of negation, at least optionally. We will show in the following that this optionality is only apparent and that an argument can be made for en having acquired properties that suggest it is no longer a negation marker. This has already been observed for West Flemish by Haegeman (2000, 2001, 2002), but is, as we will see, a wider Flemish phenomenon.
2.1 En is a polarity marker in Flemish In both East and West Flemish, purely formal, that is, non-negative, uses of en in nonnegative affective contexts can be found. The examples below from East Flemish Ghent demonstrate the availability of non-negative en in conditional clauses (3), comparative clauses (4) and the context of maar ‘only’ (5).
Continuity is change
(3) en aa’t slecht weer en is and if it bad weather en is ‘and if the weather is bad’
(Ghent; Leemans 1966: 191)
(4) Ge moet ‘tzegge gelijk of ‘t en is. you must it say like if it en is ‘You must tell it the way it is.’
(Ghent; Leemans 1966: 191)
(5) ten is maar een kleintsje it.en is only a little-one ‘it’s only a little one’
(Ghent; Tavernier 1959: 246)
In West Flemish, non-negative en can be used in before-clauses (6), as well as the context of maar ‘only’ (7), as in East Flemish. (6) Je moet niet komen voordat ik geschreven en heb. you should not come before I written en have ‘You should not come before I have written.’ (Kortrijk, Barbiers et al. 2008: 60) (7) K’en een mo drie marbels. I en have but three marbles ‘I only have three marbles.’
(Lapscheure; LH’s intuitions)
Besides such non-negative uses of en, there is of course the well-known preservation of en in negative clauses. But here as well we will see that en is no longer simply a negation marker.
2.2 En marks emphasis on polarity In the literature it has occasionally been observed (cf. e.g. Overdiep 1937: 457), that en in the Flemish dialects lays a certain additional emphasis on the polarity of the clause it occurs in, in particular where it occurs with sentential negation. There are two types of context which show that Flemish en has somehow moved on to become a marker of emphasis. First, it is used in elliptical replies with doen ‘do’ functioning as emphatic contradictions. Second, it is used in negative clauses where its function is to emphatically negate or contradict an expectation either explicit or implicit in the context. We will provide empirical evidence for this proposal in the following two subsections.
2.2.1 Emphatic contradictions Emphatic contradictions with en are found in most Flemish dialects. These are elliptical replies emphatically contradicting a preceding statement (8a, b) or question (8c). They have a rather fossilized structure, consisting of a finite form of vicarious doen ‘do’ in the present tense and a weak pronominal subject. For many speakers this can also
Anne Breitbarth & Liliane Haegeman
be impersonal ‘t ‘it’, which then does not agree with the subject given in the preceding context.2 (8) a. A: Hij slaapt. he sleeps ‘He is asleep.’ B: Hij/’t en doet. he en does ‘No, he isn’t.’ b. A: Hij zal niet komen. he will not come ‘He won’t come.’ B: Hij/’t en doet. he en does ‘No, he WILL.’ c. A: Slaapt hij? sleeps he ‘Is he asleep?’ B: Hij/’t en doet. he en does ‘No he isn’t.’
(after Barbiers et al. 2008: 60)
Although (8a) and (8c) may give the impression that this is a fossilized use of archaic stage I negation, (8b) shows that this cannot be true: when a negative statement is contradicted by means of the ‘t en doet construction, this amounts to an emphatic affirmation. The same is found in the following West Flemish example from Haegeman (1995): (9) A: Valère verkuopt da nie Valère sells that not ‘Valère doesn’t sell that.’ B: J’en doet. Je verkoopt da wel he en does He sells that indeed
K’een der gisteren gekocht. I-have some yesterday bought
‘He does. He does sell that. I bought some yesterday.’ (Lapscheure; Haegeman 1995: 160)
Clearly, en in the ‘t en doet construction is not equivalent to the old stage I negation; rather, it expresses an emphatic contradiction, whether the contradicted statement or question is positive or negative.
. Cf. Ryckeboer (1986), Haegeman (2001, 2002), Barbiers et al. (2008).
Continuity is change
2.2.2 Contradicting an expectation in negative sentences Besides these emphatic contradictions with doen, where en occurs on its own, en is also used together with the sentential negator nie(t) or n-words,3 which is the basis of the commonly held view that the Flemish dialects are still in transition between stage II and stage III of Jespersen’s Cycle. The reason why this use of en is not likely to be ‘just’ an illustration of stage II negation, that is, negation expressed by a discontinuous marker, is that en has a low, though stable, frequency in the Flemish dialects, ranging around 10% of all negative sentences with nie(t) or n-words, as the data in studies such as Leemans (1966) or De Pauw (1973) show.4 Comparing the cases with and without en, it appears that en is used in contradictions of an (implicit) expectation. This is confirmed for both West and East Flemish. In (10), from West Flemish, the request uttered by speaker A presupposes that speaker B is able to comply with the request. B, in order to deny this ability emphatically by denying the presupposition, can use en. Examples (10a) and (10b) show that this is possible in both main and embedded clauses. (10) A: Geef me nen.keer Valère zenen telefoon. give me once Valère his phone ‘Can you give me Valère’s phone number?’ a. B: K’(en) een-k ik Valère zenen telefon nie. I (en) have-I I Valère his phone neg ‘I don’t have Valère’s number.’ b. B: K-zeggen jen toch dan-k em nie (en)-een. I say you prt that-I him neg (en) have ‘I am telling you I don’t have it.’ (Lapscheure, Haegeman 2002: 11)
The same holds in East Flemish. As already observed by Overdiep (1937: 457), negation with en…niet instead of niet alone, as found for example in the dialect of East Flemish Ghent, is more emphatic. In (11a), the common knowledge that Ghent is famous for its university is given in the background. By using en, the speaker expresses his/her surprise at the interlocutor not knowing this university despite having lived in the city for fifteen years. In (11b), the context (not given) seems to be an implication or direct accusation of the speaker’s having made a mistake, which is emphatically denied in the example.
. An n-word (Laka 1990) is an indefinite pronoun, determiner or adverb, such as, for instance, nadie ‘no one’, ningún ‘no’ or nunca ‘never’ in Spanish, which potentially participates in negative concord and is able to express negation on its own in a fragment answer (Giannakidou 2006). . Unfortunately, there are no more recent corpus studies of the use of en in Flemish dialects. We are aiming to address this desideratum in future work.
Anne Breitbarth & Liliane Haegeman
(11) a.
Ge weunt al vijftien jaor in Gent, in g’en ken nog you live already fifteen years in Ghent and you-en know still
d’Universiteit nie? the university not
‘What? You have lived in Ghent for fifteen years, and you are telling me you still don’t know the university?!’ b. Dat en is toch mijn schuld nie! that en is prt my fault not ‘You can’t say that this is MY fault!’
(Ghent, Overdiep 1937: 456–457)
Summarizing, what en in combination with sentential negation expresses is an emphasis on this negation, contradicting a presupposition or assumption that the contrary state of affairs holds present in the context.
3. The historical development of Flemish en Both purely formal uses as well as the emphatic use of en are already found in the history of Dutch. Purely formal, or expletive, uses are rare in the complement of adversative predicates (like doubt, deny, forbid etc.) in Middle Dutch (Burridge 1993: 184–185), but regular for example in the standard of comparison (Beheydt 1998): (12) Ghien moget niet vorder rechten dan u manne en wijsen. you-ne can neg more judge than your men ne tell ‘You cannot judge more than your men tell you.’ (from Beheydt 1998: 16)
Such non-negative uses spread further in the seventeenth century to clauses with maer ‘only’ and nauw ‘nearly, hardly’, besides the standard of comparison: (13) hoe wel ter nauwer noodt verhaalens waart en is how well to near need telling worth ne is ‘although it is hardly worth telling’ (Weijnen 1956: 73)
Furthermore, elliptical replies with vicarious doen ‘do’ are already attested in Middle Dutch, the (still only negative) predecessor of the emphatic contradictions in (8): (14) Ghi hout u spot. In (=ic+en) doe you hold your mockery. I=ne do ‘You are mocking. I do not!’
(from Beheydt 1998: 15)
As discussed by Burridge (1993), the preverbal marker en in combination with niet was lost during the seventeenth century in the northern Dutch provinces. It has been argued that this is the consequence of a ban by prescriptive grammarians and influential writers. Burridge shows how it disappears from the letters of P.C. Hooft (1581–1647) during his lifetime; similarly, Van der Wouden (1995: 23) shows how playwright Joost
Continuity is change
van den Vondel “after the model of […] writers such as P. C. Hooft” deliberately reduces the use of en in his plays over the years. However, the reduction of the use of en took place to a much lesser extent in the southern provinces. The reasons for this have hardly been addressed in the literature, which for the period since the Gouden Eeuw (‘Golden Age’, roughly, the seventeenth century) has almost exclusively focused on the north. Interestingly, Beheydt (1998) hints at a possible meaning difference arising in the early modern period between the older bipartite and the newer single postverbal expression of negation: (…) in vraagzinnen (…) blijkt de negatie steeds postverbaal te zijn als de betekenis positief is. Het lijkt heel aannemelijk dat de taalgebruikers vonden dat de tweeledige ontkenning de negatieve betekenis te zeer benadrukte, wat minder het geval was met de postverbale.5 (Beheydt 1998: 93)
We conclude that the emphatic use of en in conjunction with niet or n-words may thus have developed in the southern dialects already during the seventeenth century.
4. An asymmetric view of stage II The standard approach to stage II of Jespersen’s cycle, that is, the stage at which the original and the new negation marker coexist, is that both elements are negation markers. We can call this the symmetric approach to stage II of Jespersen’s cycle. Approaches differ with respect to the direction of the change, that is, whether it is the weakening of the preverbal marker that drives the development, or whether the grammaticalization of a new phrasal negator causes the weakening of the old preverbal marker and ultimately makes it superfluous. Both types of approach agree that it is the redundancy or lack of iconicity in the presence of two negation markers that causes the loss of the preverbal marker and the transition to stage III. Abraham (2003: 360), a proponent of the former approach writes, “Under criteria of economy, the attrition and total erosion of the weak verbal neg-prefix is entirely expectable”,6 and according to Detges & Waltereit
. “In questions, the negation seems to be invariably postverbal in case the meaning is positive. It is very likely that language users thought that the bipartite negation put too much emphasis on the negative meaning, which was less the case with the postverbal negator” (our emphasis). . However, it is not the case that systematically a ‘weakened’ clitic marker of negation is replaced by a ‘stronger’ phrasal one. Lindstad (2007) shows that in the development of Old Norse a preverbal clitic marker of negation is replaced by a postverbal clitic. Thus the “replacing marker is not an adverb, perceived as ‘stronger’ or ‘heavier’ than the replaced one, but a suffix on the verb” (Lindstad 2007: 35).
Anne Breitbarth & Liliane Haegeman
(2002), a functional principle of ‘constructional iconicity’ is behind the demise of the original negator, since a “formally discontinuous functional unit” (Detges & Waltereit 2002: 186) is less iconic than the expression of a single function by a single element, and therefore dispreferred and prone to simplification. Generative accounts of stage II of Jespersen’s Cycle operate with different variations on the theme of feature valuation or agreement, whereby one of the two elements in a bipartite negation licenses the other. Rowlett (1998), and following him, Wallage (2005, 2008), for example, propose that the preverbal marker loses the value [neg] of its polarity feature [pol: ] and therefore requires a licenser that is able to provide this value. We have seen that the old preverbal marker en in Flemish bipartite negation has developed a function separate from expressing negation, and that this development goes back to Middle Dutch. We therefore propose that the prolonged period of coexistence of the two elements, the old and the new negation marker, entails their functional differentiation. That is, we assume an asymmetric view on stage II of Jespersen’s Cycle in (historical) Flemish, under which the two elements present are not both at the same time related to the expression of negation. This provides us with an elegant account of the actual distribution of en in the Flemish dialects. Breitbarth (2009) proposes analysing the preverbal marker at stage II of Jespersen’s cycle in the West Germanic languages as a polarity marker rather than a negator. This leads to an “asymmetric” interpretation of stage II ‘bipartite’ ‘negation’, as the two markers do not conspire in the expression of negation. Rather, the two elements come to be associated with different functional heads. We will use this analysis here, finetuning it to the development of Flemish en.7 Like Moscati (2006), we propose that polarity is a clause-typing feature located on a functional head in the left periphery. Although a more fine-grained representation of the CP-system may be necessary, let us for simplicity assume that this head is C and that the polarity clause typing feature is [±affective], building on Klima’s (1964) characterization of weak NPI contexts as affective. Now, in Middle English, Middle Dutch, Middle High German and Middle Low German, the polarity element ne/en is a bound morpheme. It therefore needs a host, which in the West Germanic languages is the finite verb. As the finite verb is not always in a position in the left periphery, Breitbarth (2009) proposes that ne/en is spelt out on the finite verb by the mechanism of Alternative Realization (Emonds 1987, 2000):
. A similar analysis has been developed by Haegeman (2000, 2001, 2002) for West Flemish from a synchronic point of view. Haegeman proposes to analyse the preverbal marker en as a marker of polarity, rather than negation. Her analysis differs from the one presented here in that she places PolP between TP and a projection called FP situated between TP and AgrOP.
Continuity is change
(15) Alternative Realization A syntactic feature F canonically associated in UG with category B can be alternatively realized in a closed class grammatical morpheme under X0, provided X0 is the lexical head of a sister of B. (Emonds 2000: definition (4.20))
That is, the feature [+affective] is realized on the lexical head of a sister of C[+affective]. Intervening functional heads not lexicalized in a structure are ignored because of the definition of Extended Sisterhood, e.g. heads in the TP and CP-domains in our case, if the verb stays low as in embedded clauses in (continental) West Germanic. (16) Extended Sisterhood If Z0 and XP are sisters and if Z1 is the smallest phrase (besides structural projections of Y) whose lexical head is Y0, then Y0 and XP are extended sisters. (Emonds 2000: definition (4.28))
The reanalysis of the former negative clitic on the finite verb as an exponent of affective polarity was possible exactly because of the Extended Sisterhood relation between T and Pol. This mechanism is similar to Roberts & Roussou’s (2003) ‘upward reanalysis’ by which the expression of a lower functional head becomes associated with the features of a higher functional head due to a derivational link between the two heads.8 This reanalysis became possible once sentential negation had found a new exponent, namely when an erstwhile adverbial emphasizer is reanalysed as the expression of sentential negation. After its reanalysis as the Alternative Realization of the feature [+affective] of C in negative clauses, ne/en could be generalized to non-veridical contexts beyond negation. An argument for the initial reanalysis of en as the Alternative Realization of C’s [+affective] on the finite verb in Flemish comes from comparative historical evidence. In early Middle Dutch as well as in Middle High German (Jäger 2008: 289–294) and Middle Low German, n-marked indefinites, where they do not mark sentential negation on their own, only seem to co-occur with the old preverbal marker, but not with the new adverbial one. Table 1 illustrates this for early Middle Dutch, and Table 2 for Middle Low German.9
. Roberts & Roussou originally assume that the result of such a reanalysis is the loss of movement, which previously connected the two heads. We see no reason why agreement alone cannot already be a sufficient connection between two positions. . These figures are based on corpora of chancery documents (oorkonden/Urkunden). In the case of early Middle Dutch these are the first 300 documents from the Gysseling corpus (Gysseling 1977); the Middle Low German corpus is the one discussed in Breitbarth (2008). It should be noted that co-occurrence of the adverbial negator and n-words does become available in later Middle Dutch, late Middle Low German and Early New High German, but
Anne Breitbarth & Liliane Haegeman
Table 1. Preverbal and adverbial markers with n-words in early Middle Dutch 1200–1280 en..n-word 49 (75.4%)
n-word alone 16 (24.6%)
n-word…niet 0 (0.0%)
Total 65
Table 2. Preverbal and adverbial markers with n-words in Middle Low German 1325–1575 en..n-word 487 (37.4%)
n-word alone 812 (62.4%)
n-word…nicht 2 (0.2%)
total 1301
This evidence shows that already early on, n-words are able to mark sentential negation on their own. The fact that they do not or prefer not to co-occur with the new adverbial negator, which is taking over the expression of negation,10 suggests that the preverbal marker is no longer a marker of sentential negation, at least not when accompanied by an n-word. Instead, the expression of negation elsewhere in the clause must have made the reanalysis of the preverbal marker as the alternative realization of [+affective] on C seem likely to language acquirers. We have seen above already (Section 3) how preverbal en came to be used in non-negative contexts in Middle Dutch, and spread to further such contexts in the present-day Flemish dialects. We can understand this as the actualization (Timberlake 1977, Harris & Campbell 1995) of the reanalysis of en, making the reanalysis visible in the language. Summing up, there are developments that suggest to the language acquirer that the old preverbal marker may be a polarity marker, namely other negative markers such as n-words becoming able to express negation without it, and this reanalysis is actualized when non-negative affective uses appear.
5. Flemish preverbal en: Continuity is change We have seen that, instead of being lost like its cognates in the other West Germanic dialects, Flemish en has (a) spread to more non-negative affective contexts than are not initially when the assumed reanalysis of the preverbal element occurred. The later developments are therefore likely to have a different explanation which goes beyond the scope of this chapter. . Niet in Middle Dutch, nicht in Middle Low German and Middle High German and not Middle English are clearly no longer emphatic and are increasingly used on their own (without ne/en) as sentential negators. They are furthermore used for negating non-finite verbs and for constituent negation, which the inherited preverbal markers are not capable of.
Continuity is change
found in Middle Dutch, thereby stabilizing the position of en in the language, and (b) that it has acquired the ability to realize an additional feature, namely emphasis on polarity. We propose that en developed historically as follows. In a first step that all West Germanic languages underwent, the negation particle ne/en on the finite verb is reanalysed as the Alternative Realization of C’s [+affective] feature. This, as argued in Section 4, qualifies as a grammaticalization by ‘upward reanalysis’ in the sense of Roberts & Roussou (2003) as a lower head becomes associated with the features of a higher head:11 (17) stage I → stage II in Middle Dutch [CP C[+aff] [ … [VP … ] [T en/ne[+neg]-Vfin]] → [CP C[+aff] [ … niet[+neg] [VP …] [T en/ne[+aff]-Vfin]]
The first line in (17) represents the structure of a negative clause in Old Dutch. By default, C in negative clauses carries the feature [+affective], representing the polarity of the clause at its interface. Sentential negation is expressed by the clitic ne, which invariably attaches to the finite verb. When n-words become able to express sentential negation on their own (that is, when they become n-indefinites), without overt realization of ne, ne becomes ambiguous between being a sentential negator in negative clauses without n-words, and a potential spell-out of [+affective] in clauses with n-words. With the grammaticalization of the new phrasal negator niet, ne comes to be associated with the realization of [+affective] – as the Alternative Realization of C’s feature on the finite verb in T, as seen in the second line in (17). The spread of en to further non-negative affective contexts in the Flemish dialects beyond those found already in Middle Dutch can be taken to have prevented en from suffering the fate of its West Germanic cognates: it has survived to this day. As we have seen, however, en is still used only very infrequently today, in only around 10% of negative sentences. Its continued survival, we have argued, is due to the fact that it has acquired the ability to realize emphasis on polarity. Emphasis (on polarity) is recognized to be a feature of the left periphery as well, related to assertion and focus.12 This is of course not relevant or available in all negative clauses, hence its low, but stable frequency. We can visualize this second change as follows: (18) Middle Dutch stage II → Present-day Flemish emphatic ‘stage II’ [CP C[+aff] [ … niet[+neg] [VP …] [T en[+aff]-Vfin]] → [CP C[+aff+emph] [ … niet[+neg] [VP …] [T en[+aff+emph]-Vfin]]
. For simplicity, we represent the clause structure with a head-final TP in (15) and (16). As shown by Haegeman (2001), it is possible to give an analysis of Continental West Germanic OV orders that conforms with Kayne’s (1994) universal base hypothesis. . On polarity emphasis/polarity focus in various languages (mainly on emphatic affirmation, however), cf. e.g. Lipták (2003), Hernanz (2007), Holmberg (2007) and Martins (2007).
Anne Breitbarth & Liliane Haegeman
Just as the [+affective] feature on C came to be Alternatively Realized by the particle en on the finite verb in T in Middle Dutch, C’s (optional) polarity emphasis feature comes to be associated with the same particle on the finite verb in the development from Middle Dutch to the present-day Flemish dialects as another instance of Alternative Realization.
6. Conclusion In this chapter, we have argued that the continued availability of the old preverbal ‘negation’ particle en in the Flemish dialects is not a case of the language having stopped in the transition from stage II to stage III of Jespersen’s Cycle. Rather, we have provided evidence showing that the present-day situation is the result of further morphosyntactic change. We have argued that after an initial reanalysis of the preverbal marker as a polarity marker, which affects other West Germanic languages as well and leads to an asymmetric analysis of stage II in these languages, en in the Flemish dialects has undergone a further change by acquiring the ability to realize an additional feature, emphasis on polarity. That is, it is now used to emphatically contradict states of affairs, explicitly or implicitly given in the context. We have argued that it is (a) the extension of the use of en to further non-negative affective contexts, and (b) en’s acquisition of the ability to express emphasis on polarity that saves it from suffering the fate of its West Germanic cognates and allows it to survive in its functional niche until today. This development is very similar to the one seen in (dialectal) French. It has long been acknowledged that old preverbal ne in French has ceased to be a negation marker proper but, as it were, ‘switches off ’ the affirmative concept (Tesnière 1959: 224–225), sentential negation being expressed by a forclusif (pas or n-words), and thus ne acts as a polarity marker. It can also used in various non-negative contexts in Standard French. However, ne is rapidly being lost from the spoken language in most dialects.13 As Fonseca-Greber (2007) shows, however, its use has stabilized at a low percentage in spoken Swiss French, because it has been reanalysed as emphatic in this variety. Our conclusion is that, in spite of the availability of en, Flemish dialects have actually reached stage III of Jespersen’s cycle and that what appears to be continuity is really a case of change.
. Cf. the references cited and discussed in Fonseca-Greber (2007).
Continuity is change
References Abraham, Werner. 2003. Autonomous and non-autonomous components of ‘grammatic(al) ization’: Economy criteria in the emergence of German negation. Sprachtypologie und Universalienforschung 56: 325–365. Barbiers, Sjef, van der Auwera, Johan, Bennis, Hans, Boef, Eefje, De Vogelaer, Gunther & van der Ham, Margreet. 2008. Syntactic Atlas of the Dutch Dialects, vol. II. Amsterdam: University of Amsterdam Press. Beheydt, Griet. 1998. Het gebruik en de vorm van de negatie in het zuidelijke Nederlands in een diachronisch perspectief (15e–20e eeuw). Licentiate dissertation, Catholic University of Leuven. Breitbarth, Anne. 2008. The development of negation in Middle Low German. Paper presented at the Annual Meeting of the LAGB, University of Essex, September 2008. Breitbarth, Anne. 2009. A hybrid approach to Jespersen’s cycle in West Germanic. Journal of Comparative Germanic Linguistics 12: 81–114. Burridge, Kate. 1993. Syntactic Change in Germanic: Aspects of Language Change in Germanic with Particular Reference to Middle Dutch [Current Issues in Linguistic Theory 89]. Amsterdam: John Benjamins. Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79–106. Detges, Ulrich & Waltereit, Richard. 2002. Grammaticalization vs. reanalysis: A semanticpragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft 21(2): 151–195. Emonds, Joseph E. 1987. The invisible category principle. Linguistic Inquiry 18: 613–632. Emonds, Joseph E. 2000. Lexicon and Grammar: The English Syntacticon. Berlin: Mouton de Guyter. Fonseca-Greber, Bonnibeth Beale. 2007. The emergence of emphatic ‘ne’ in conversational Swiss French. French Language Studies 17: 249–275. Giannakidou, Anastasia. 2006. N-words and negative concord. In The Blackwell Companion to Syntax, Martin Everaert, Henk van Riemsdijk, Rob Goedemans & Bart Hollebrandse (eds), iii. 327–391. Oxford: Blackwell. Gysseling, Maurits 1977. Corpus van Middelnederlandse teksten (tot en met het jaar 1300). m.m.v. en van woordindices voorzien door Willy Pijnenburg. The Hague: Nijhoff. Haegeman, Liliane. 1995. The Syntax of Negation. Cambridge: CUP. Haegeman, Liliane. 2000. Remnant movement and OV order. In The Derivation of OV and VO [Linguistik Aktuell/Linguistics Today 31], Peter Svenonius (ed.), 69–96. Amsterdam: John Benjamins. Haegeman, Liliane. 2001 Antisymmetry and verb-final order in West Flemish. Journal of Comparative Germanic Linguistics 3: 207–232. Haegeman, Liliane. 2002. West Flemish negation and the derivation of SOV order in West Germanic. Nordic Journal of Linguistics 25 (2): 154–189. Haegeman, Liliane & Zannuttini, Raffaella. 1991. Negative heads and the NEG-criterion. Linguistic Review 8: 233–251. Haegeman, Liliane & Zannuttini, Raffaella. 1996. Negative concord in West Flemish. In Parameters and Functional Heads, Adriana Belletti & Luigi Rizzi (eds), 117–179. Oxford: OUP. Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: CUP.
Anne Breitbarth & Liliane Haegeman Hernanz, M. Lluïsa 2007. Emphatic polarity and C in Spanish. In Studies in Spanish Syntax, Laura Brugè (ed.), 104–150. Venice: Libreria Editrice Cafoscarina. Holmberg, Anders. 2007. Null subjects and polarity focus. Studia Linguistica 61(3): 212–236. Jäger, Agnes. 2008. History of German Negation. Amsterdam: John Benjamins. Jespersen, Otto. 1917. Negation in English and Other Languages [Historisk-filologiske Meddelelser I,5]. Copenhagen: A.F. Høst. Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kemenade, Ans van. 2000. Jespersen’s Cycle revisited. In Diachronic Syntax, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 51–74. Oxford: OUP. Klima, Edward S. 1964. Negation in English. In The Structure of Language: Readings in the Philosophy of Language, Jerry A. Fodor & Jerrold J. Katz (eds), 246–323. Englewood Cliffs NJ: Prentice-Hall. Ladusaw, William A. 1993. Negation, indefinites, and the Jespersen Cycle. Berkeley Linguistics Society 19: 437–446. Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections. Ph.D. dissertation, Massachusetts Institute of Technology. Leemans, Emilienne. 1966. Syntactische Kenmerken van het Gents Dialect. Licentiate dissertation, University of Ghent. Lindstad, Arne M. 2007. Analyses of Negation. Structure and Interpretation. Ph.D. dissertation, University of Oslo. Lipták, Anikó. 2003. The expression of sentential emphasis in Hungarian. Ms, Leiden University. Martins, Ana M. 2007. Double realization of verbal copies in European Portuguese emphatic affirmation. In The Copy Theory of Movement on the PF Side [Linguistik Aktuell/Linguistics Today 107] Norbert Corver & Jairo Nunes (eds), 77–118. Amsterdam: John Benjamins. Moscati, Vincenzo. 2006. The Scope of Negation. Ph.D. dissertation, University of Siena. Overdiep, Gerrit Siebe. 1937. Stilistische Grammatica van het Moderne Nederlandsch. Zwolle: Tjeenk Willink. Pauw, G. De. 1973. Syntaktische kenmerken van het dialect van Buggenhout-Opstal. Licentiate dissertation, University of Ghent. Posner, Rebecca. 1985. Post-verbal negation in non-standard French: A historical and comparative view. Romance Philology 7: 44–82. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Rowlett, Paul. 1998. Sentential Negation in French. Oxford: OUP. Ryckeboer, Hugo. 1986. Het hulpwerkwoord doen in replieken. In Vruchten van z’n akker. Opstellen van (oud-)medewerkers en oud-studenten voor Prof. V.F. Vanacker hem aangeboden bij zijn afscheid van de Rijksuniversiteit Gent, Magda Devos & Johan Taeldeman (eds), 321–337. Ghent: Seminarie voor Nederlandse Taalkunde en Vlaamse Dialectologie, Rijksuniversiteit Gent. Schwegler, Armin. 1988. Word-order changes in predicate negation strategies in Romance languages. Diachronica 5: 21–58. Schwenter, Scott A. 2006. Fine-tuning Jespersen’s Cycle. In Drawing the Boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn [Studies in Language Companion Series 80], Betty J. Birner & Gregory Ward (eds), 327–344. Amsterdam: John Benjamins. Tavernier, C. 1959. Over negatie en expletief en in het Gents dialect. Taal en Tongval 11: 245–252. Tesnière, Lucien. 1959. Eléments de syntaxe structurale. Paris: Klincksieck.
Continuity is change
Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Mechanisms of Syntactic Change, Charles N. Li (ed.), 141–77. Austin TX: University of Texas Press. Wallage, Philipp. 2005. Negation in Early English: Parametric Variation and Grammatical Competition. Ph.D. dissertation, University of York. Wallage, Phillip. 2008. Jespersen’s Cycle in Middle English: Parametric variation and grammatical competition. Lingua 118 (5): 643–674. Weijnen, Antonius Angelus. 1956. Zeventiende-eeuwse taal. 2nd edn. Zutphen: Thieme. Willis, David. To appear. A minimalist approach to Jespersen’s Cycle in Welsh. In Grammatical Change: Origins, Nature, Outcomes, Dianne Jonas, John Whitman and Andrew Garrett (eds). Oxford: OUP. Wouden, Ton van der. 1995. On the development of marked negation systems. Paper presented at the 12th International Conference on Historical Linguistics, University of Manchester, August 1995. Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. Ph.D. dissertation, University of Amsterdam.
Using the Matrix Language Frame model to measure the extent of word-order convergence in Welsh-English bilingual speech Peredur Davies & Margaret Deuchar Bangor University
Word-order convergence is argued to be a processual mechanism in which bilinguals use morphemes from language A in an order which is more prominent in language B than in language A. In this chapter we examine whether there is word-order convergence in the speech of Welsh–English bilinguals using Myers-Scotton’s (2002) Matrix Language Frame model. We use the term dichotomous matrix language to capture the phenomenon found where structural material within a clause does not point to either language as being the matrix language (ML), and suggest that this reflects convergence in such clauses. Applying this model to recorded spoken data from six Welsh–English bilinguals, we find that dichotomous ML clauses in the data are very rare, suggesting that word-order convergence in Welsh–English bilinguals is not commonly found when measured in this way.
1. Introduction In this chapter we address the question of whether word-order convergence is found in the speech of Welsh–English bilinguals. After outlining our view of convergence, we discuss a method of identifying convergence using Myers-Scotton’s (2002) Matrix Language Frame (MLF) model. Our method involves identifying clauses in bilingual speech which have structure from both participating languages as potentially showing convergence in progress. We then apply this method to spoken data from six Welsh– English bilinguals to identify the extent of the convergence found therein.
2. Defining convergence We define convergence, following Backus (2004: 179), as a contact-induced processual mechanism of language change which “results in greater similarity [than before] between two languages that are in contact with each other”. We interpret his definition to refer to convergence in individual constructions, rather than in all constructions in the language. Bullock & Toribio (2004) point out that convergence has traditionally
Peredur Davies & Margaret Deuchar
been seen mainly as an outcome rather than as a process, but we follow them in defining convergence as a process. Bullock & Toribio further state that the result of the diachronic process of convergence can be observed synchronically in the output of bilinguals (2004: 91), an approach also taken by, for instance, Myers-Scotton (2002) and Schmitt (2000). Convergence can occur in most parts of a language’s grammar, but in this chapter we will focus on word-order convergence. Myers-Scotton (2002: 164) writes that a clause in which word-order convergence has occurred will be one where the morphemes come from one language but the grammar includes structure from both participating languages. Schmitt’s (2000: 15) description of covert code-switching (see Section 3) is interpretable as a description of convergence at the utterance level. These definitions allow for the results of convergence to be observable in both monolingual and bilingual utterances. We define monolingual clauses as clauses containing morphemes from only one language and bilingual clauses as clauses containing morphemes from more than one language. Backus (2004) suggests that the processual mechanism which results in a change in frequency of a construction is the “use of one of two or more L1 patterns, congruent with [an] unmarked L2 pattern” (2004: 180); we take ‘congruent’ here to mean ‘similar’. Therefore a bilingual will choose a construction from one language (LA) which has a similar counterpart in the other language (LB) in preference to an LA construction which does not. This point is made by Thomason (2001: 93) when she suggests that constructions which are similar in two languages, but not actually identically distributed (e.g. constructions that are marked in one language but unmarked in the other), are most prone to convergence. A similar approach leads Bullock & Toribio (2004: 91) to view convergence as “the enhancement of inherent structural similarities found between two linguistic systems”. They distinguish convergence from the notions of interference and transfer on the grounds that each of these latter notions implies unidirectionality whereas convergence may be either unidirectional or bidirectional. The focus of our investigation into convergence is constructions which have one or more morphemes from Welsh but word order from English, where that word order is available in Welsh but only in restricted contexts. Such constructions will reflect convergence as a process. Diachronic changes which might result from frequent occurrences of this convergence are not the focus of this chapter. Having outlined our definition of convergence, in the next section we will briefly review some of the more recent studies of this phenomenon, interpreting the research in terms of convergence as being a processual mechanism.
3. Previous work on convergence We focus here on previous work supporting the notion of convergence (whether in word order or elsewhere) as a process rather than a result, on the assumption that the
Using the Matrix Language Frame model to measure the extent of word-order convergence
mechanism of convergence is reflected in individual instances which may or may not lead to change, depending of the frequency of those instances.1 Toribio’s (2004) analysis of Spanish–English bilinguals in the USA demonstrates that contact with English affects these speakers’ use of null pronouns, a typical feature of Spanish. Her study involved two tasks: first, participants were required to narrate a story “in Spanish” in monolingual mode based on a series of pictures representing Little Red Riding Hood. Then the participants were required to reproduce orally another story, The Beggar Prince, which they had previously read aloud in a mixed-language version; this was expected to lead to the production of both Spanish and English in bilingual mode. In the first task, where they were in monolingual mode, speakers used overt pronouns more often than was expected in monolingual Spanish discourse. Furthermore, in the second task, when in bilingual mode, the use of overt pronouns by these speakers was even more frequent than expected. So, in any given context where there was a choice whether or not to use an overt pronoun, speakers frequently converged to English usage in their Spanish. In a study with a similar focus, Montrul (2004) compared the speech of Spanish– English heritage bilinguals2 living in the USA with the speech of monolingually-raised Spanish speakers from different Spanish-speaking countries, examining, among other features, their use of null subjects. The heritage speakers in Montrul’s data used overt subjects (68.6%) more than null subjects (31.4%), compared to the reverse pattern in monolinguals’ use (42.8% overt and 57.2% null). As Spanish allows such subjects to be null whereas English does not, Montrul argues that convergence is occurring in the Spanish morphosyntax of these speakers and that the dominance of English has led the speakers to select overt subjects instead of null subjects. Schmitt (2000), analysing data from Russian–English bilingual immigrant children in the USA within the MLF framework, proposes the notion of covert codeswitching, which is what “takes place when abstract lexical structure from one language is influenced by abstract structure from another language, resulting in a composite ML [matrix language]” (2000: 19). A composite ML is found in a construction in which morphemes from one language have (some) structure from another language (2000: 15). Examples Schmitt gives of covert code-switching are bare forms, which are inserted LB morphemes (in a clause where the structure is otherwise from LA) appearing without the appropriate LA morphology (2000: 23). Children in her data produce
1. That is, what some authors would call ‘convergence’, but which we call ‘the result of convergence’. . Montrul uses the term ‘heritage speakers’ to describe second- or third-generation bilinguals who come from a home where a non-English language (Spanish here) was spoken, and who have some level of bilingualism in both English and the heritage language.
Peredur Davies & Margaret Deuchar
such forms, e.g. (1) below (2000: 23), where inserted English morphemes in what is otherwise a Russian ML clause lack the expected Russian morphological endings. (1) nu yest’ baseball v heaven? well is baseball in heaven ‘Well, is there baseball in heaven?’ [expected: nu yest’ baseball v heaven-e? well is baseball in heaven-prep.masc.sg]
In (1) the switched noun heaven does not have the expected Russian prepositional case marker -e. In inserting the switched item (heaven) from English, the speaker has incorporated with it the grammar of English (which does not have an inflectional affix of this kind on nouns). Schmitt views this as convergence, with the ML as a composite of Russian structure (e.g. verbal morphology) and English structure (e.g. no noun inflection on heaven), occurring because of an increase in the level of activation of the EL so that it competes with the ML in the providing the clause’s structure (2000: 24).3 These studies therefore show convergence as a contact phenomenon that is best viewed as a process rather than a result.
4. Convergence and code-switching Backus (2004: 179) suggests that evidence of convergence may be found particularly in the speech of bilinguals who code-switch extensively, and Toribio (2004) provides evidence that the same speakers may show more evidence of convergence when using both languages than just one. Bilingual clauses are therefore more likely loci for convergence than monolingual clauses in the speech of bilinguals. Nevertheless, Myers-Scotton (2002) indicates that convergence can occur with or without code-switching. Code-switching showing convergence is labelled “composite code-switching” (2002: 165), whereas evidence of convergence without codeswitching is a structure where “all surface morphemes come from one language” and “the abstract lexical structure projecting these morphemes no longer comes from one language, but includes some abstract structure from another language” (2002: 164), as noted above. Whether or not code-switching is involved, clauses showing evidence of convergence will have a composite morphosyntactic frame, which we shall label here
3. We are grateful to a reviewer for noting that apparently the choice of the preposition v ‘in’ here is an unusual one; na ‘on’ is more expected. We suggest that this is another example of the influence of English on the clause.
Using the Matrix Language Frame model to measure the extent of word-order convergence
a dichotomous matrix language (see below). Since Myers-Scotton’s MLF model applies to individual clauses in the speech of individual bilinguals rather than to a language as a whole, and as we define convergence as a synchronic process that can occur within a single construction, we argue below that the MLF model can be used to identify evidence of convergence in specific, quantitative terms.
5. Convergence in Welsh We now review studies that have been conducted on convergence in the Welsh language. While our data is from Welsh–English bilingual speech, our focus is on the influence of the majority language, English, on the minority language, Welsh.4 Deuchar (2005) examines whether paradigmatic and syntagmatic congruence between the grammars of Welsh and English facilitates code-switching. She concludes that where the grammatical categories of the two languages are similar, code-switching is facilitated. The results suggest that phenomena such as code-switching are more likely to occur in aspects of the grammar where the participating languages are similar than where they differ. Deuchar suggests that code-switching is enhanced or restricted depending on the extent of this similarity. Deuchar (2006) goes on to examine Welsh–English bilingual utterances which show possible convergence. An example is given in (2), taken from Deuchar (2006: 1996). (2) fi ‘di bod i ‘r bus lle 1s prt.past be.nonfin to det bus place5 ‘I have been to the bus place.’
. The UK census for 2001 shows that 20.8% of the population of Wales can speak Welsh; this was an increase from the 1991 census, which indicated that 18.7% of the people of Wales spoke Welsh. Figures taken from the Welsh Language Board website (http://www.byig-wlb. org.uk/English/faq/Pages/Howdothe2001resultscomparewiththe1991results.aspx) or (http:// tinyurl.com/2001cf1991). . We use the following abbreviations in glossing examples: 1S 1PL 2S 2PL 3SM 3SF DET IMP
1st person singular pronoun 1st person plural pronoun 2nd person singular pronoun 2nd person plural pronoun 3rd person singular masculine pronoun 3rd person singular feminine pronoun determiner imperfect tense
NEG NONFIN PAST PRES PRT PRT.PAST VBZ
negative particle or prefix non-finite verb past (perfect) tense present tense a pre-verbal particle perfect particle verbalizer suffix
Peredur Davies & Margaret Deuchar
This is a bilingual clause where all but one morpheme is Welsh, but the order found in the noun–noun NP bus lle is English. The expected Welsh order here would be lle bus, where the modifier bus follows the head noun lle, but instead the modifier is found preceding the head. Deuchar notes that the morphemes in this clause appear to have a one-to-one morpheme correspondence with English surface word order, where fi ‘di bod i is equivalent to I have been to. She therefore suggests that this clause might show word-order convergence from Welsh towards English. Deuchar & Davies (2009) evaluate two similar models of language change proposed by Thomas (1982) and Myers-Scotton (1998). Both these models postulate processes of contact-induced change which can ultimately lead to speakers in a community abandoning their original language for another. At a stage prior to this, composite code-switching occurs, which can lead to what Myers-Scotton calls matrix-language turnover. Deuchar & Davies use Myers-Scotton’s MLF model to analyse a dataset of speech from Welsh–English bilinguals (which is also among the data examined in this study) for evidence of English structure in bilingual clauses. The findings show that Welsh was the matrix language (ML) in the great majority (93.94%) of finite clauses overall, and that all bilingual clauses had Welsh as their ML. Clauses which do not demonstrate a clear ML were considered to show possible convergence, but only one such clause was identified (examined again as example (34) in the present study; see Section 12 below). The results point to the absence of ML turnover from Welsh to English, with little evidence of convergence. For further information regarding the influence of English on Welsh, see ParryWilliams (1923), Fynes-Clinton (1995 [1913]) and Lindsay (1993).
6. The differences and similarities in Welsh and English word order Convergence may occur in parts of the grammar which are similar in both languages in contact. Following Thomason’s (2001: 93) suggestion that an essential step in identifying contact-induced structural change in a language is to identify structural features shared by both languages, in this section we describe the similarities found in Welsh and English word order, specifically the relative positions of the subject and the finite verb in a finite clause and the head and the modifier within a NP. English is an SV-order language, with the finite verb occupying post-subject position in declarative clauses (The cat caught the mouse); see Quirk et al. (1989: 724). VS order is available only in yes-no interrogative constructions (Did the cat catch the mouse?), questions involving modal verbs (Can the mouse run?), questions involving to be or to have as the main verb (Is that a mouse?), WH-questions (Where is the cat?), and various other inversion contexts, in all of which the finite verb is raised
Using the Matrix Language Frame model to measure the extent of word-order convergence
to pre-subject position.6 English is therefore primarily SV order with VS in some constructions.7 Welsh, conversely, uses VS order (the inflected verb occupying the first main position, before the subject) in main clauses, as in (3), or periphrastic constructions using an auxiliary and a nonfinite verbal noun, as in (4); see King (1996: 21). (3) Daliodd y gath y llygoden. Caught the cat the mouse V
S
O
‘The cat caught the mouse.’ (4) Wnaeth y gath ddal y llygoden. Did the cat catch the mouse AuxV
S
V
O
‘The cat caught the mouse.’
Question clauses have rising intonation and/or the pre-verbal particle a (usually omitted). The position of the subject and verb is not changed, so Welsh interrogatives, like Welsh declaratives, have VS order (Ddaliodd y gath y llygoden? Wnaeth y gath ddal y llygoden?). SV order, however, is found in subject WH-question constructions, as in (5), and also where the subject constituent is fronted for emphasis, as in (6). (5) Pwy ddaliodd y llygoden? Who caught the mouse WH V
O
‘Who caught the mouse?’8 (6) Y gath ddaliodd y llygoden. The cat caught the mouse S V O ‘It was the cat which caught the mouse.’9
Therefore Welsh has primarily VS order, but SV order in certain constructions.
. An exception to this is a construction where the WH-word is the subject of the sentence, e.g. Who caught the mouse?, which has SV order. . There are other, rarer constructions in English that are VS order, such as those with fronted negatives, e.g. Never has the cat caught anything. . This could also mean ‘Whom did the mouse catch?’. . This could also mean ‘It was the cat that the mouse caught’ (i.e. OVS order), demonstrating object-fronting.
Peredur Davies & Margaret Deuchar
We now turn to English and Welsh head/modifier constructions in NPs. We use the following abbreviations: HM denotes the modifier following the head; MH denotes the modifier preceding the head. Note that when we use the term ‘head/modifier order’, we mean the relative order of head and modifier (without specifying what this is). In English head/modifier constructions (e.g. noun/adjective or noun/noun NPs), the order is MH in the great majority of cases, e.g. red wine, where red modifies wine. There are some English collocations where the order is HM, for instance, court martial, queen consort, Attorney General, and sergeant-major, but these are very rare.10 Welsh head/modifier NPs usually have HM order, for instance, gwin coch ‘red wine’, where the adjective coch precedes the head noun gwin, or gyrrwr bws ‘bus driver’, where the modifying noun bws follows the head noun gyrrwr. Some adjectives, however, precede the head noun they modify, such as hen ‘old’ and prif ‘main, prime’; for instance, hen ddyn ‘old man’, prif weinidog ‘prime minister’.11 Furthermore, any Welsh modifier can generally precede its head, causing soft mutation of the initial consonant of the head, e.g. cryno-ddisg ‘compact disc’, where the adjective cryno precedes the noun disg (the first consonant of which is mutated [d] > [ð]), and heddwas (< hedd ‘peace’ + gwas ‘servant’) ‘policeman’, where the modifying noun hedd precedes the head gwas, which again undergoes soft mutation of [g] > [ø]. Whilst this is primarily a literary style, it also occurs less frequently in colloquial use, and often results in the construction being interpreted as a compound; in fact, compounds in Welsh are usually MH order constructions where the initial consonant of the head is mutated. Therefore, Welsh has more than one option for NP word order, primarily HM but also MH. In this section we have demonstrated the overlap between possible word orders in English and Welsh. In main clauses, English has SV whilst Welsh has VS, but English also has VS and Welsh SV in limited contexts. In head/modifier NPs, English has MH word order whilst Welsh has both MH and (more frequently) HM. Where there are similarities such as these, it is possible that convergence will be found. We now proceed to discussing the Matrix Language Frame model, which we suggest may be adapted to become a framework for identifying sites of convergence.
. Moreover, some of these are borrowings from Romance languages (e.g. sergeant-major < French sergent-major, first attested in English in 1574 according to the Oxford English Dictionary), and others are anomalous, such as court martial, which is first-attested as martial court (normal English MH order) but apparently obtained the new HM order in the mid-17th century, perhaps due to conflation with the concurrent term court marshal. . Note that some of these adjectives, such as hen, however, are also found following the head (e.g. dyn hen ‘old man’), meaning that certain Welsh adjectives can occur in either HM or MH position.
Using the Matrix Language Frame model to measure the extent of word-order convergence
7. Considering the matrix language Central to Myers-Scotton’s theory of code-switching (1993, 2002 etc.) is the Uniform Structure Principle (USP), which claims that a given type of constituent in any language has a uniform abstract structure and must be well-formed. Related to this is the Asymmetry Principle (AP), which states that in bilingual speech the two (or more) languages used do not participate equally (2002: 9). The Matrix Language (ML) provides structure, whilst the Embedded Language (EL) is the language of switches, which can only be content morphemes and certain types of system morphemes. Content morphemes may come from either the ML or EL. This asymmetry between ML and EL is a key concept in Myers-Scotton’s MLF model (2002: 15). In this model, “classic codeswitching” is used to refer to the asymmetry between the languages as just detailed. The MLF model may be used to identify clauses in bilingual speech which show the appropriate asymmetry between ML and EL. It also follows that the model could also be used to identify those clauses where there is no asymmetry between the two participating languages. If structure in a clause is supplied by both languages, then the ML may not be uniquely sourced from either of the two participating languages. This latter case is what Myers-Scotton (2002: 105), referring to bilingual utterances, terms “composite code-switching”, where the clause’s morphosyntactic frame is derived from more than one language, and which she considers to demonstrate convergence. Clauses exhibiting composite code-switching have what we call a dichotomous matrix language, as explained above. We expect that clauses with a dichotomous matrix language will exhibit convergence. In the following section we outline the manner in which we apply the MLF model to our data to identify clauses, whether monolingual or bilingual, that show convergence.
8. Applying the MLF model to the data The MLF model specifies that a clause will adhere to two principles, which MyersScotton (2002: 59) terms the Morpheme Order Principle (MOP) and the System Morpheme Principle (SMP). The analytical process of applying the MLF model to data, therefore, takes the form of verifying whether or not both MOP and SMP point to the same language as being the ML for a given clause. If they indicate the same language, then the language indicated is identified as the ML of that clause. The principles are outlined below. The Morpheme Order Principle (MOP) states that the ML supplies the word order for a clause, with the exception of well-formed EL islands, which follow the word order of the EL. The primary word-order distinction between Welsh and English that we have selected is the position of the verb in relation to the subject, and this is the first
Peredur Davies & Margaret Deuchar
word-order criterion tested on a clause. The secondary word-order distinction we have selected is the position of the modifier in relation to the head in an NP. If a head/modifier NP in a clause was either (1) a bilingual NP or (2) a monolingual NP with morpheme order that was not expected for morphemes in that language; then the head/ modifier word order in that NP was used to identify the ML. Therefore, in a clause where the subject/verb positions and/or the head/modifier positions are what would be expected in Welsh, we identify ‘Welsh’ as the ML for that clause; where the order is what would be expected of English, we identify ‘English’ as the ML. Some clauses have the same basic word order in Welsh and English, as noted in Section 6, and in such instances the MOP alone was not sufficient to indicate the clause’s ML. The System Morpheme Principle (SMP) dictates that the ML supplies outside late system morphemes; following Myers-Scotton (2006: 244), we take verbal agreement markers to be outside late system morphemes.12 Welsh verbal morphology indicates a Welsh ML and English verbal morphology indicates an English ML. The SMP was only applied here to finite clauses: we considered clauses lacking a finite verb not to contain enough information to satisfy the SMP, except, as in (7), where we assume there is a null auxiliary. (7) ti ‘di siarad ybyty fo 2s prt.past talk.nonfin about 3sm ‘You’ve talked about it.’
[FUS27-LIS328]
[Alternative form: wyt ti di siarad ybyty fo be.2s.pres 2s prt.past talk.nonfin about 3sm]
In such cases as this the clause was assumed to be finite, with the null element consisting of outside late system morphemes from Welsh. However, the relative position of this null verb in relation to the subject was not ascertainable, and therefore these types of clauses could have the SMP successfully applied to them but not the subject/verb order criterion of the MOP. We will now demonstrate the application of the MLF to an example clause from the data, (8). English morphemes are underlined in examples. (8) dw i ‘n love-o ‘r gwlad though be.1s.pres 1s prt love-VBZ det countryside though ‘I love the countryside, though.’
[FUS27-LIS38]
First the MOP is tested. As the finite verb dw precedes the subject i, it appears that Welsh provides the word order. If there was a head/modifier construction available, it
. Outside late system morphemes are described by Myers-Scotton as morphemes that are “coindexed with forms outside the head of their maximal projections” (2002: 75–6).
Using the Matrix Language Frame model to measure the extent of word-order convergence
would be expected to follow the same order (Welsh), unless it was an EL island. Second, the SMP is tested. The inflected verb dw is in the first person singular, and this inflection matches the first person singular pronoun i. There is therefore agreement. The inflection on dw is Welsh. Therefore the SMP identifies Welsh as ML. Finally, the two languages indicated by the SMP and the MOP are compared. According to the MLF model, both principles must indicate either Welsh or English as supplying structure.13 Returning to Example (8), the SMP and MOP both indicate Welsh as ML, so Welsh can be positively identified as the clause’s ML, providing outside late system morphemes, word order and some content morphemes. English is the EL, providing some content morphemes (love and though). Assuming an ML could be positively identified, therefore, each clause in the data is marked as ‘Welsh’ or ‘English’ ML.14 Clauses with sufficient material to test the principles, but which cannot show one language as ML, are discussed in the next section.
9. Cases where the source of the ML is not positively identifiable Whilst it is expected that, in most clauses the testing of the MOP and SMP will result in the same language being indicated as ML, it is possible that the two principles will not agree; for instance, if the verb agreement morphology is from one language but the word order is from the other, or if one principle provides conflicting information about which language is ML. An example of the latter type of clause is (9), taken from Deuchar (2006: 1994). (9) Ddaru ni gyfweld […] am ddeg awr happen.past 1pl interview.nonfin for ten hour assistant i D.S. assistant for D.S. ‘We interviewed […] for a ten-hour assistant for D.S.’
It is not possible to test the SMP on this clause, as subject–verb agreement does not apply: the verb ddaru is not inflected for person or number. When the MOP is tested, the position of verb (ddaru) before subject (ni) indicates the word order as being Welsh. However, the NP ddeg awr assistant ‘ten-hour assistant’ appears to have English word order, as the modifier ddeg awr precedes the head assistant. This order is not . If the word order construction in a clause is available in either Welsh or English, the SMP has the ‘casting vote’, taken to specify which language is the source of the ML. That is, the subject-verb agreement positively identifies the ML which the MOP could only ambiguously identify. . Some clauses in the data contained insufficient material to test the principles, such as verbless clauses or non-finite clauses. These are not examined in this chapter.
Peredur Davies & Margaret Deuchar
usual for Welsh. As the material within the NP is not all English, it cannot be labelled an EL island. The English word assistant would therefore be expected to be positioned in a locus appropriate to the subject/verb order (Welsh), preceding the modifier, to produce assistant deg awr. As we find that English seems to provide the word order in this NP, however, the result is that the verb/subject order and head/modifier order indicate different languages as being the ML. Thus the asymmetry specified by the MLF model is not found in such clauses as (9) above. Here, both languages are supplying structural information. As the evidence for the ML points in different directions, we shall refer to any clauses where this occurs as having a dichotomous matrix language.15 Such clauses may show convergence. A benefit of adding this nuance to the MLF model, and to methods of identifying convergence in general, is that it provides a means to identify where a given structure (e.g. an NP) has word order which is not usual for the morphemes projected by that structure. We therefore posit that the MLF model can be used to identify loci of convergence in clauses which yield a dichotomous ML result. The reasons for using the MLF model as a means of identifying word-order convergence are twofold. First, where the MOP and the SMP both clearly indicate one language as being the source of the ML, we assume there is no word-order convergence. However, clauses where the two principles do not clearly indicate one language as being the source of the ML, as in the case of dichotomous ML clauses, seemed to us to be likely sites for convergence and therefore worth investigation. Second, the MLF model can be applied quantitatively to data on a clause-by-clause basis (see below), and therefore this method can generate quantitative information indicating the extent of wordorder convergence taking place in a given dataset. We believe that this is one of the first implementations of a model that identifies convergence quantitatively in this way. It will be recalled that convergence may be found in structures which contain morphemes from just one language (cf. the definitions given by Schmitt (2000) and Myers-Scotton (2002) cited in Section 3). We suggest that the MLF model may also be applied to monolingual clauses, following Jake, Myers-Scotton & Gross (2002: 88), who imply that monolingual speech will also have an ML providing the grammatical frame. Even if a clause’s morphemes are all from one language, the other language may be providing structure. Therefore monolingual constructions from bilinguals may show convergence, and the principles of the MLF model can be tested on such clauses.
. Example (9) could also be described as an example of composite code-switching, but note that this differs to a dichotomous ML clause in that the latter label can be applied to either monolingual or bilingual clauses, whereas we interpret composite code-switching only to be applicable to bilingual clauses.
Using the Matrix Language Frame model to measure the extent of word-order convergence
In this study we have included all clauses from the data, whether monolingual or bilingual, in our analysis, though this chapter concentrates on results from finite clauses.
10. The data The data analysed in this study were collected as part of an AHRC-funded project located at Bangor University, for which 40 hours of recorded spoken data were collected, transcribed according to the CHAT system (MacWhinney 2000), and submitted to Talkbank (http://talkbank.org/data/bilingbank/Bangor). Participants were found via social networks and advertising, and were selected on the basis of being bilingual in Welsh and English. The recordings were made in informal conditions, in a location chosen by the participants, who knew one another well either as friends, colleagues or family members. The participants had an informal conversation with no researcher nearby. In the majority of cases, conversations were between two people. For the purposes of the current study, a subset of six speakers in three recordings was chosen for analysis. The transcriptions are coded Davies6, Fusser6 and Fusser27. We will describe the six participants briefly next.16 Davies6 is a conversation between two male friends, Hector (23, a teacher) and Daniel (25, a student). Both had lived in the same part of north-west Wales since birth. For both, Welsh was the language spoken to them by their parents whilst growing up and the medium of their schooling, and both reported learning English at or just before primary school age. Fusser6 is a conversation between two female work colleagues, Antonia (52, a lecturer) and Amranwen (36, a secretary). Antonia was born in south Wales but moved to live permanently in the north-west at 21. Her father spoke Welsh and English to her but her mother was a monolingual speaker of English. She was educated through the medium of Welsh and English. Amranwen had always lived in north Wales apart from a period at an English university. Her parents and school teachers spoke Welsh to her. Both women reported having acquired both Welsh and English simultaneously. Fusser27 is a conversation between two female student friends, Mabli (19) and Lisa (20). Mabli had lived in north-west Wales since birth. Though she had received all her schooling through the medium of Welsh, she came from a home where her father spoke only English and her mother Welsh and English. Lisa had lived in south-west Wales from birth until she moved to north-west Wales to attend university. Her father spoke English to her, whilst her mother used Welsh, which was also the language
. Pseudonyms have been provided for the participants in order to hide their identity.
Peredur Davies & Margaret Deuchar
through which she received her schooling. Mabli and Lisa both reported that they acquired Welsh and English simultaneously.
11. Results The MLF model was applied to all transcribed clauses, monolingual and bilingual, totalling 3275 clauses. Of these, 1816 or 55.45% were finite. There was enough evidence to assign an ML to all finite clauses17 and the results are displayed in Figure 1. Of the finite clauses, 1733 (95.43%) had a Welsh ML; English was the ML for 82 (4.52%) clauses. Clearly Welsh is the predominant ML in these speakers’ finite clauses. 2000 1800
Number of clauses
1600
Bilingual 335
1400 1200 1000 800 600
Monolingual 1398
400 200 0
82 Welsh
English Matrix Language
1 Dichotomous
Figure 1. Distribution of finite clauses according to matrix language
18.50% of finite clauses (336 out of 1816) were bilingual. Of those bilingual clauses, 99.70% (335 out of 336) had a Welsh ML. English was never the ML in bilingual finite clauses. The one remaining bilingual finite clause (0.30%) had a dichotomous ML (see Section 12 below). Thus Welsh is clearly the most usual ML in bilingual clauses in these data. Example (10) is a bilingual clause with Welsh ML, whilst (11) is a monolingual clause with English ML. As above, English morphemes are underlined in examples.
. All clauses analysed were labelled as having either a Welsh, English or dichotomous ML.
Using the Matrix Language Frame model to measure the extent of word-order convergence
(10) ches i ‘m chance i ddarllen o get-imp.1s 1s neg chance to read.nonfin 3sm ‘I didn’t get a chance to read it.’
[DAV6-DAN40]
(11) oh there’s nothing wrong with you!
[FUS6-ANT283]
Proportion of finite clauses
Example (10) has Welsh as ML since the finite verb and subject (ches and i) have Welsh outside late system morphemes, and the subject/verb order is VS, also Welsh. Example (11) has English as ML since is has English outside late system morphemes and the subject/verb order is SV, also English. 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
MAB
LIS
AMR ANT Speaker
Dichotomous
English
DAN
HEC
Welsh
Figure 2. Matrix language distribution of clauses according to speaker.
There is little variation in the data between speakers, as shown in Figure 2. Hector produced the highest proportion of clauses with Welsh ML (291 out of 297 or 97.98%), whilst Daniel produced the lowest (232 out of 262 or 88.55%). The younger speakers (Mabli, Lisa, Daniel and Hector) used Welsh ML on average in 94.99% of their finite clauses; the older speakers (Amranwen and Antonia) did so in 96.82% of their finite clauses on average, therefore only slightly more frequently. Regarding where speakers grew up, northern speakers (Daniel, Hector and Amranwen) used Welsh ML in 95.66% of their finite clauses on average, similar to the 95.07% for southern speakers (Mabli, Lisa and Antonia). Females used Welsh ML slightly more frequently than males (96.26% to 93.56%). These differences, however, are very small. To summarize, the Welsh–English bilinguals that were recorded almost always produce clause structure reflecting Welsh as ML and therefore do not display convergence.
Peredur Davies & Margaret Deuchar
12. Clauses with a dichotomous Matrix Language Dichotomous ML clauses were defined in Section 9 as clauses where the evidence for the ML points in different directions. There was only one dichotomous ML clause in the data. It is shown in (12) below. (12) roedd drws-nesa # pobl yn wneud be.3s.imp next-door people prt make.nonfin sloe gin sloe gin ‘[The] next-door people made sloe gin.’
#
[FUS6-AMR371]
The SMP identifies Welsh as ML (roedd has Welsh outside late system morphemes). The verb/subject order of this clause is Welsh VS (roedd precedes the subject NP drws nesa pobl). However, the word order in the NP drws nesa pobl, where the head noun pobl ‘people’ follows the modifying lexical unit drws nesa ‘next door’, is Englishpredominant MH, rather than the expected Welsh order, HM (i.e. pobl drws nesa). The material in the rest of the clause points to Welsh as providing structure, but the information in the NP drws nesa pobl points to English as providing structure. Since there is disagreement between these two indicators of ML, the clause has a dichotomous ML. We view (12) as showing signs of word-order convergence in the NP drws nesa pobl, where the structure found is more prominent in one language, English, but where the morphemes are from the other language, Welsh. Other interpretations of (12) are possible, however. First, the speaker inserts a brief pause (<0.5 seconds) between drws nesaand pobl, which may suggest that she is aware that she is producing an unusual word order, but chooses not to correct herself. Second, this pause could alternatively be viewed as clause division with ellipsis, along the lines of “there were, next-door, people who made sloe gin”, in which case it is not an anomalous order and the clause ML is not dichotomous. Third, the English phrase people next door could be considered to be equally as expected as the phrase next-door people, as relative clause truncation resulting in this kind of post-modifier are quite common in English. If true, it is likely that Welsh speakers would use people next-door rather than next-door people as a pattern for calquing. However, it could be that it is the generic English order MH which influences the order here. Overall, we suggest convergence in this case is one valid interpretation. The proportion of dichotomous ML clauses in this data compared to non-dichotomous ML clauses is very small. As we suggest that the dichotomous ML clause found shows convergence within NPs, it appears that structural convergence is not prominent in the dataset. Previous studies (e.g. Thomason 2001: 89) suggest widespread convergence of a similar type can result in language change. If there were more dichotomous ML clauses in our data, we could suggest a change in progress in the word order of NPs
Using the Matrix Language Frame model to measure the extent of word-order convergence
produced by Welsh–English bilinguals, but with only one instance in the data, we do not predict a word-order shift. The issue remains as to why there should be so few apparent examples of convergence in these Welsh–English data. It may be that we have conducted too narrow a search for possible sites of convergence in restricting our attention to clauses with dichotomous ML. In future research we intend to broaden the search to other clauses, including those where the ML is decided by only one of the two principles. For example, in (7) we saw that the MOP criterion could not be applied because the relative order of the (null) auxiliary and subject could not be ascertained. However, if we look at the relative order of the subject and non-finite verb in (7) we see that it is SV, and that the result of the deletion of the presumed auxiliary is that the clause has word order which arguably converges towards English. We will investigate such clauses more thoroughly in the future. For the time being we suggest that there is almost no word-order convergence in these data. If widespread convergence is taken as a sign of language change or shift in progress, then conversely limited convergence may be taken to indicate language stability, which is what the results of this study suggests is the current state of Welsh. The predomination of Welsh rather than English as the ML in the clauses of these speakers supports the notion that Welsh structure is preferred by these speakers, and possibly by Welsh–English bilinguals in general. If there actually is more widespread convergence in the output of Welsh speakers, then it has not been identified using our current methodology.
13. Conclusions In this chapter we have made a theoretical claim and tested this claim empirically on a set of data. Our theoretical claim is that convergence can be found in clauses where the morpheme order is provided by both participating languages, and, more specifically, and in reference to the MLF model, where the testing of the model’s principles identifies more than one language as supplying clause structure. We called such clauses dichotomous ML clauses. We then tested this claim on a sample of data from six Welsh–English bilinguals, where we applied the Matrix Language Frame model to 1816 finite clauses. We found that only one clause displayed convergence using this method; the remainder of the clauses adhered to the asymmetry between ML and EL predicted by the Asymmetry Principle. So the evidence supplied by applying the model in this manner shows that the phenomenon where English word order influences structure in clauses which are otherwise Welsh ML is only present in the speech of Welsh–English bilinguals to a very limited degree. Word-order shift in NPs by such speakers does not seem, therefore, to be in progress. Our theoretical claim that the
Peredur Davies & Margaret Deuchar
MLF model can identify convergence, nevertheless, appears to be upheld, given that the dichotomous ML clause identified probably does show signs of convergence. However, the overall indication according to these data is that there is practically no wordorder convergence in the speech of Welsh–English bilinguals, and that what we find is continuity, not change, in their grammar.
References Backus, Ad. 2004. Convergence as a mechanism of language change. Bilingualism: Language and Cognition 7: 179–181. Bullock, Barbara E. & Toribio, Almeida J. 2004. Convergence as an emergent property in bilingual speech. Bilingualism: Language and Cognition 7: 91–93. Deuchar, Margaret. 2005. Congruence and Welsh–English code-switching. Bilingualism: Language and Cognition 8: 255–269. Deuchar, Margaret. 2006. Welsh–English code-switching and the Matrix Language Frame model. Lingua 116: 1986–2011. Deuchar, Margaret & Davies, Peredur. 2009. Code switching and the future of Welsh. International Journal of the Sociology of Language 195: 15–38. Fynes-Clinton, Osbert Henry. 1995 (reprint from 1913). The Welsh Vocabulary of the Bangor District. Felinfach: Llanerch (originally by OUP). Jake, Janice L., Myers-Scotton, Carol & Gross, Steven. 2002. Making a Minimalist approach to codeswitching work: Adding the Matrix Language. Bilingualism: Language and Cognition 5: 69–91. King, Gareth. 1996. Modern Welsh: A Comprehensive Grammar. London: Routledge. Lindsay, Cora F. 1993. Welsh and English in the city of Bangor: A study in functional differentiation. Language in Society 22: 1–17. MacWhinney, Brian. 2000. The CHILDES Project: Tools for Analyzing Talk, 3rd edn. Mahwah NJ: Lawrence Erlbaum Associates. Montrul, Silvina. 2004. Subject and object expression in Spanish heritage speakers: A case of morphosyntactic convergence. Bilingualism: Language and Cognition 7: 125–142. Myers-Scotton, Carol. 1993. Duelling Languages. Oxford: Clarendon Press. Myers-Scotton, Carol. 1998. A way to dusty death: the Matrix Language turnover hypothesis. In Endangered Languages: Language Loss and Community Response, Lenore A. Grenoble & Lindsay J. Whaley (eds.), 289–316. Cambridge: CUP. Myers-Scotton, Carol. 2002. Contact Linguistics: Bilingual Encounters and Grammatical Outcomes. Oxford: OUP. Myers-Scotton, Carol. 2006. Multiple Voices: An Introduction to Bilingualism. Malden MA: Blackwell. Parry-Williams, Thomas H. 1923. The English Element in Welsh: A Study of English Loan-words in Welsh. London: Honourable Society of Cymmrodorion. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1989. A Comprehensive Grammar of the English Language. London: Longman. Schmitt, Elena. 2000. Overt and covert codeswitching in immigrant children from Russia. International Journal of Bilingualism 4: 9–28.
Using the Matrix Language Frame model to measure the extent of word-order convergence
Thomas, Alan. 1982. Change and decay in language. In Linguistic Controversies: Essays in Linguistic Theory and Practice in Honour of F. R. Palmer, David Crystal (ed.), 209–219. London: Edward Arnold. Thomason, Sarah G. 2001. Language Contact: An Introduction. Edinburgh: EUP. Toribio, Almeida J. 2004. Convergence as an optimization strategy in bilingual speech: Evidence from code-switching. Bilingualism: Language and Cognition 7: 165–173.
On language contact as an inhibitor of language change The Spanish of Catalan bilinguals in Majorca* Andrés Enrique-Arias
Universitat de les Illes Balears This study explores the role of language contact as an inhibitor of language change through the historical analysis of several features that are characteristic of the Spanish in contact with Catalan in the island of Majorca. The analysis reveals that (i) these features are attested in Majorcan Spanish at least since the 1700s and (ii) they had some sort of existence in monolingual varieties of Spanish at the time when the language was introduced to the island. The historical data suggest that these features of Majorcan Spanish are not, as traditionally believed, contact-driven innovations, but the result of the preservation of structures that are recessive in monolingual Spanish but in Majorca have been reinforced by the existence of a parallel Catalan structure.
1. Introduction A widespread assumption in the linguistic literature is that language change is an expected, or even unavoidable, result of language contact.1 In fact, numerous studies of language variation and change in contact languages claim that contact accelerates change. In contrast, very little attention has been paid to the opposite scenario, that is, the possibility that the presence of two languages in the same speech community may constitute a factor promoting the retention of variants that are losing ground in noncontact varieties. In particular, a preliminary historical analysis of the Spanish spoken in Majorca, an island off the eastern coast of Spain where Spanish has been spoken
*I wish to thank Miriam Bouzouita, David Heap, Carmen Silva-Corvalán, and two anonymous reviewers for their comments on this paper. Any errors that remain are of course my responsibility. . As Thomason (2001: 10) puts it “The most common result of language contact is change in some or all of the languages: typically, though not always, at least one of the languages will exert at least some influence on at least one of the other languages”.
Andrés Enrique-Arias
alongside Catalan for centuries, has shown a remarkable historical continuity for some morphosyntactic and lexical features that are receding in other Spanish-speaking areas (Enrique-Arias 2006, 2008). In the same vein, research carried out using the variationist sociolinguistic methodology (Blas Arroyo 2007) has called attention to the considerable vitality in Castellón, another Spanish–Catalan bilingual area, of traits that are receding in non contact areas. The main objective of this study is to explore further the role of language contact as an inhibitor of language change through an analysis of a range of phonetic, morphosyntactic and semantic phenomena in the varieties of Spanish in contact with Catalan, mainly in Majorca but also in other Catalan speaking territories. The specific aims are: a. to evaluate the possible influence of language contact on the retention of variants that are subject to change in other dialect areas; b. to determine what linguistic and extralinguistic factors favor the retention of linguistic variables in language contact situations; c. to contribute new data from a historical corpus of Spanish in contact with Catalan in Majorca to shed new light on these issues. In order to achieve these objectives, in the present study I analyse the historical development of a number of structures that are characteristic of the variety of Spanish spoken in Majorca. For each phenomenon I look at the current description according to the available linguistic monographs, as well as its distribution in historical corpora that represent, on the one hand, Spanish produced by Majorcan bilinguals in the late 1700s and early 1800s and, on the other, comparable data from general Spanish. The historical analysis reveals that many of the features that characterize the variety of Spanish spoken in Majorca today were already present in the Spanish texts produced by Majorcans in the 1700s. The phenomena in question had in turn some sort of existence in monolingual varieties of Spanish at the time when the language was introduced to the island. This distribution of forms suggests that, for a number of linguistic structures, Catalan contact has resulted not so much in the introduction of innovative structures into Majorcan Spanish, but in the preservation of language usages that are recessive or have disappeared in general Spanish, reinforced by the existence of a parallel structure in Catalan.
2. Language contact as a promoter of linguistic change The effect of language contact as a promoter of language change is closely related to the fact that bilingualism sets the stage for a variety of linguistic phenomena that
On language contact as an inhibitor of language change
result in change of one kind or another: code-switching, borrowing, simplification of grammatical and lexical categories, development of periphrastic constructions or overgeneralization of forms following a regularizing pattern, among other kinds of structural interference. Contact-induced change appears to be particularly intense in situations of social bilingualism which are unbalanced in favour of one language. In these situations members of the subordinate group are more likely to become bilingual and adopt features from the socially dominant group’s language (Thomason 2001: 66). This scenario makes extensive interference possible, often resulting in the development of innovative language uses and accelerating changes that are less developed in monolingual varieties. By acceleration it is understood that the innovative variant occurs in more linguistic environments and at a higher frequency than it does in non-contact varieties of the same language. In the Spanish-speaking world, acceleration of language change is well illustrated by a number of phenomena from Spanish in contact with English in the United States. In the three examples presented below, Spanish–English bilinguals living in the Southwest United States exhibit a more advanced stage of linguistic change than their Spanish monolingual counterparts in other areas of the Spanish-speaking world. a. Advance of the periphrastic future at the expense of morphological future Spanish exhibits variable use of two future forms: a periphrastic future (voy a cantar ‘I am going to sing’) and a morphological future (cantaré ‘I will sing’). The advance of the former at the expense of the latter is a historical development documented in several studies (see, for instance, Aaron 2007). In regards to Spanish in contact with English in the US, Gutiérrez (1995, 2002) has shown that this change has been accelerated among the communities of Mexican origin of the United States when compared to monolingual speakers from Mexico. b. Increase in the use of preverbal clitics in verbal periphrases In a number of Spanish verbal periphrases composed of a finite semi-auxiliary plus an infinitive or a present participle, object clitics referring to an argument of the non-finite form may variably occur before the finite verb as shown in (1): (1) Vino a verme/me vino a ver. he.came to see.me/me he.came to see ‘He came to see me.’
The trend of placing clitics in preverbal position in periphrases, to a large extent as a consequence of a process of grammaticalization of a number of matrix verbs, reflects an ongoing change that is more advanced in language contact than it is in monolingual varieties, as demonstrated in Silva-Corvalán (1994: 130).
Andrés Enrique-Arias
c. The extension of estar at the expense of ser in the context of predicate adjectives Spanish has two copular verbs, ser and estar, both translated as ‘to be’. The extension of the copula estar to contexts traditionally limited to ser is another phenomenon that appears to be more advanced amongst bilingual speakers (Silva-Corvalán 1994: 119). Innovative uses such as mi marido está alto ‘my husband is tall’, mi casa está chiquita ‘my house is small’ (as opposed to traditional use of ser in these structures: mi marido es alto, mi casa es chiquita) among Los Angeles bilingual speakers (including Mexico-born speakers living in the U.S.) are higher, an overall 34% compared to 16% in a group of monolingual speakers in Morelia, Mexico (Gutiérrez 1992). A number of reasons have been adduced to explain the acceleration of change affecting subordinate varieties in contact situations. In regards to Spanish in Los Angeles, Silva-Corvalán (1994: 7) points out that the contact-induced changes that she studies are further “favored and accelerated” by three factors: low normative pressure in the subordinate language, restriction in the range of communicative uses of the subordinate language, and speakers’ positive attitudes toward the superordinate language. This is indeed the case in the US, where Spanish does not enjoy official status, most speakers have incomplete schooling, if any, in Spanish, and there is no academic and literary class serving as a normative model. As a result, speakers are to a great extent isolated from the Spanish-speaking norm (Enrique-Arias 2010). In summary, acceleration of language change in contact situations would work in the following manner: bilingualism promotes innovative linguistic uses, and the sociolinguistic situations that typically characterize subordinate varieties, such as lack of normative pressure, or the restriction in contexts of use, contribute to intensifying and spreading these innovations even further.
3. Language contact as an inhibitor of linguistic change In contrast, very little attention has been paid to the opposite scenario, that is, the possibility that the presence of two languages in the same speech community may constitute a factor promoting the retention of variants that are receding in non-contact varieties. Processes of language change are often characterized by the emergence of a new form or construction that becomes interchangeable in at least one linguistic environment with a pre-existing one. Once this new form increases its frequency to the point of replacing entirely the older one, the change is complete. However, in a language-contact situation use of the traditional variant may be reinforced by (a) the existence of a parallel structure in the contact language and/or (b) the absence of a structural equivalent for the innovative variant in the contact language. In such cases change could be delayed, and thus the spread of the innovative variant would progress more slowly than in non-contact varieties of the same language. Moreover, when the language in question is the socially dominant variety, that is, it is typically acquired and
On language contact as an inhibitor of language change
used in formal contexts, the sociolinguistic situation may further encourage the preservation of the traditional variant. This is especially true when the innovative form is characteristic of informal registers. As such, speakers that acquire and use the socially dominant language in formal settings have limited exposure to the innovative variant and are thus less likely to use it in their speech. It will be argued in the following pages that this is in fact the situation that can be found in the variety of Spanish spoken in the island of Majorca where there is intensive societal Spanish–Catalan bilingualism.2 The evolution of a number of phonetic phenomena illustrates how language contact has been a factor in the retention of traditional variants in the Spanish spoken in Majorca. 1. The phenomenon known as yeísmo (the loss of the phoneme /ˆ/ and its merger into the phoneme /N/) is a trend that originated in Western Andalusia, possibly as early as the sixteenth century and has spread considerably ever since (Guitarte 1983). In European Spanish yeísmo “has become part of the dominant urban speech pattern throughout the Peninsula” (Penny 2000: 121). Although some Catalan speakers are showing loss of the palatal lateral approximant, the distinction of the two phonemes is far more robust in Catalan than it is in standard urban varieties of Spanish (Julià i Muné 2002: 79); as a result bilingual speakers tend to retain it more. In Romera’s study (2003: 371–72) in Palma, Majorca’s main city and administrative capital, speakers with Catalan as L1 exhibited 52.3% of conservation of the distinction compared to 12.9% in the speech of those classified as Spanish L1. Romera’s findings seem to be supported by the historical data collected by EnriqueArias (2006, 2008), in which eighteenth-century Spanish documents produced by Majorcan writers show no traces of yeísmo (i.e. the graphemes
and exhibit a distribution consistent with a phonemic distinction of the sounds they represent) during a period when other comparable texts written by Spanish monolinguals already showed signs of the phonemic opposition being lost. 2. Weakening and loss of [d] in words ending in the sequence [ado] is another widespread phenomenon in the Spanish-speaking world that has been attested profusely since the 1500s (Moreno Fernández 2004: 999). In Romera’s (2003: 373) study it was found that amongst Palma speakers its loss is more likely to happen in the speech of older individuals, with low education levels and Spanish as L1.
. Currently, competence in Spanish in the Balearic Islands is almost universal while competence in Catalan depends on the speaker’s place of origin. Of those residents born in the Balearic Islands, 95.9% declare they speak Catalan “well” or “not too badly”. This percentage goes down to 47.2% among the immigrants born outside the Islands. Overall, 76.7% of the population in the Balearic Islands declared that they spoke Catalan in 2002 (Melià 2002: 61).
Andrés Enrique-Arias
These findings are consonant with the general trend, registered in several studies, that retention of [d] in words ending in [ado] is higher overall in areas where Spanish is in contact with Catalan.3 In Blas Arroyo’s (2007) study of the realizations of -ado participles in Castellón the influence of Catalan on this outcome is evident. His informants with Catalan as L1 exhibited 74% of retention of [d] as opposed to 32% in those with Spanish as L1 (cf. Blas Arroyo 2007: 269–270). This may be explained, first, because, as opposed to Spanish, in which /d/ in -ado participles is often fricativized or even deleted, the equivalent Catalan consonant in the participle ending is produced with greater articulatory tension, that is, as a voiceless occlusive. Another factor to take into account is that loss of intervocalic [d] is more common in casual speech (Moreno Fernández 2004: 1002). As most Catalan-dominant bilinguals learn Spanish in formal contexts, their exposure to weakening and loss of [d] is necessarily limited. 3. Weakening of /s/ in syllable-final position is another widespread expanding phenomenon in the Spanish-speaking world (Moreno Fernández 2004: 993–99). According to Romera’s (2003: 374) data from Palma, speakers classified as L1 Spanish show 24.6% maintenance of -/s/ while bilingual speakers that are Catalan-dominant exhibit a higher level of maintenance (92.7%). Again, the reason is that syllable-final /s/ in Catalan is rather robust, which explains its higher levels of maintenance amongst Catalan-dominant bilinguals; in addition, this is the formal, prestigious variant taught in school and used in the mass media. From the previous data and analyses it follows that a possible effect of language contact is the retention of conservative variants. All three examples are variable phenomena that, in monolingual Spanish varieties, exhibit a historical trend of increase in frequency and territorial expansion. However, we find that (a) in Spanish in contact with Catalan the innovative variant occurs in fewer linguistic environments and/or with lower frequency when compared to non-contact varieties of the same language; and (b) of those Catalan–Spanish bilinguals surveyed in the different studies, those classified as Catalan L1 exhibit lower frequency of the innovative variants than those classified as Spanish L1. There are two conditions that seem to favor this outcome: first, the conservative variant has a parallel equivalent in the contact language; second, the traditional variant is either closer to standard, prestigious uses ([ado] pronunciation, preservation of final [s]) or is neutral in this respect (maintenance of the palatal lateral
. As can be seen in Blas Arroyo’s (2007) comparative study of [ado] pronunciations in ten Spanish cities, the two localities where Spanish happens to exist side by side with Catalan (Castellón and Barcelona) are the ones that exhibit higher percentages of preservations of final [ado].
On language contact as an inhibitor of language change
approximant������������������������������������������������������������������������ ); as Catalan L1 speakers typically learn and use Spanish in formal contexts, they have limited exposure to innovative variants that are characteristic of less formal registers.
4. Historical clues for contact-induced continuity in grammar The data and analyses of phonetic and phonological phenomena discussed so far illustrate how language contact can be an effective inhibitor of language change. In order to explore further this kind of effect in a contact situation, I shall now turn to analyse a number of morphosyntactic phenomena that are characteristic of the Spanish spoken by Catalan bilinguals. The main question to be addressed is whether these morphosyntactic features are contact-induced innovations or, on the contrary, they result from the preservation of language uses that were already present in general Spanish at the time when it came into contact with Catalan. In order to answer this question it is pertinent to incorporate an analysis of historical data of the Spanish produced by Catalan bilinguals. In the case at hand, a diachronic analysis will show whether there has been a historical continuity in the features that characterize Majorcan Spanish, and, if that is the case, provide critical evidence in support of the view that a substantial number of the morphosyntactic traits that are characteristic of the Spanish spoken in Majorca are the result of the preservation of linguistic structures that were already present in general Spanish at the time when the language was introduced to the island. Majorca, along with the rest of the Balearic Islands, was incorporated into the Kingdom of Aragon and populated with Catalan speakers in the thirteenth century in the context of the Christian Reconquest of Spain from the Muslims. Following the dynastic union of Castile and Aragon in the fifteenth century, and the political, economic and cultural hegemony of Castile among the Peninsular kingdoms that characterized the ensuing centuries, Catalan-speaking territories fell under increasing cultural pressure and underwent a gradual process of linguistic castilianization. By the sixteenth century, educated, urban Majorcans had at the very least passive competence in Spanish as evinced by abundant records of books printed in Spanish, and sermons, poetry competitions, celebrations and theatre performances in Spanish. The gradual castilianization of Majorca intensified in the eighteenth century as the Bourbon monarchs implemented a number of measures intended to spread the use of Spanish in the administrative and educational domains (Martínez i Taberner 2000: 355–359.). As a result, from the late eighteenth century on, the languages of Majorca entered a period of diglossia, with Spanish being the only language used in formal domains. In the early 1980s, with the advent of democracy things took a drastic turn with Catalan finally achieving official status and use in domains such as education, administration and mass media.
Andrés Enrique-Arias
The historical corpus used in the study is composed of a collection of letters written in Majorca in the late eighteenth century representing a variety of writer demographics and registers (Enrique-Arias 2006, 2008, in press).4 These data are complemented by judicial statements from the late eighteenth and early nineteenth centuries that reproduce direct discourse taken from the excerpts published in Vibot (2004, 2005). The corpus thus represents the early stages in the formation of the variety of Spanish spoken by Catalan speakers in Majorca.
4.1 Morphological and periphrastic future Since the seventeenth century, future temporal reference in Spanish has undergone a significant change, whereby the territory of the morphological future has been progressively taken over by the periphrastic future (Aaron 2007). In the areas where Spanish coexists with Catalan, however, the substitution of the morphological future by the periphrastic one is less advanced (Blas Arroyo 2004: 1068). The effect of Catalan language contact on the retention of the morphological future has been illustrated by Blas Arroyo’s (2007) detailed study of the distribution of future forms in the Spanish of the Catalan-speaking region of Castellón: speakers that are Catalan-dominant exhibit systematically higher percentages of use of the morphological future (cf. Table 1, adapted from Blas Arroyo 2007). Table 1. Percentage of retention of morphological future in Castellón Uses of future Immediate distance e.g. ‘We’ll go to the movies at five.’ Intermediate distance e.g. ‘We’ll do it this week.’ Indefinite distance e.g. ‘I don’t know when we’ll do it.’ Maximal distance e.g. ‘When I graduate I’ll do a Ph.D.’
L1 Catalan
L1 Spanish
41
30
58
42
45
39
74
73
Both languages share the widespread Romance morphological future derived from the Latin periphrasis infinitive + habere ‘to have’ (amare habeo > amar-é ‘I will love’). Nevertheless, unlike Spanish, which has developed a new periphrastic future with ir ‘go’, the literal Catalan equivalent, that is, the periphrasis with anar ‘go’,
. The letters belong to file 6 of the epistolary archive of the Zaforteza family (henceforth ZA) which contains nearly 250 letters (roughly half of them in Spanish and the other half in Catalan) dated between 1739 and 1788. The examples are identified by letter number, line number and year of composition.
On language contact as an inhibitor of language change
is used for the expression of preterit actions and states. The result of this asymmetry is that, while all varieties of Spanish have shown a historical tendency to replace the morphological future with the periphrasis with ir, Catalan bilingual speakers tend to retain it more. In addition, the socially dominant status of Spanish in Majorca may have contributed to the retention of morphological future. Idealized models of grammar tend to ignore the existence of the periphrastic future, even though it has become by far the most common structure to express future in all Spanish varieties; most textbooks and Spanish grammars only include the inflectional forms, that is, the so-called future simple (amaré ‘I will love’) and the future perfect (habré amado ‘I will have loved’) as examples of the future tenses. This may have reinforced the morphological future amongst those speakers that have acquired Spanish in formal contexts.5 The data from the historical corpus suggest that this distribution of forms is a long-established trend in Majorcan Spanish. In the Spanish letters in the ZA we find 146 examples of morphological future as opposed to only two examples of periphrastic future with ir. In other words, the innovative use accounts for little over 1%. This figure seems to be rather low if we consider the available data for mainland Castilian texts: Aaron (2007) registers 4% of periphrastic future for the seventeenth and eighteenth centuries combined, and 12% for the nineteenth century. It should be noted that Aaron’s study is based on literary works, which typically reflect more formal language. Personal letters, on the other hand, tend to be closer to oral registers (Biber 1995: 283–300). Taking this into account, the actual difference in use between the Majorcan and Castilian varieties is likely to be even greater.
4.2 Obligation periphrases In Spanish there are three obligation periphrases involving the semi-auxiliaries tener and haber – both translated as ‘have’– plus infinitive: tener de, haber de and tener que. The first one, tener de, is the least common; haber de sounds somewhat more formal and archaic in general monolingual Spanish while tener que is by far the most common one. The spread of tener que in general Spanish is a relatively recent phenomenon, as can be seen in Figure 1 (data collected from Corpus del español).
. Lipski (1990: 53) gives a similar explanation for the quantitative preponderance of morphological future in Malabo (Equatorial Guinea), another place where Spanish is a socially dominant variety that, for many speakers, is learned as a second language in formal contexts.
Andrés Enrique-Arias 100 90 80 70 60 50 40 30 20 10 0
tener que tener de haber de
16C
17C
18C
19C
20C
Figure 1. Obligation periphrases in general Spanish (sixteenth–twentieth centuries)
The chart shows that, in the last three centuries, tener que has experienced a spectacular progression, from 7.4% in the eighteenth century, to 23.7% in the nineteenth century, to 81.3% in the last century. However, in areas of Spanish in contact with Catalan, including Majorca, several studies point out how the traditional forms tener de and haber de maintain their vitality (Blas Arroyo 2004: 1071). Once again this seems to result from the maintenance of structures that have receded in general Spanish but are maintained in Majorca due to the existence of a parallel structure in Catalan, whose most common obligation periphrases are tenir de and haver de.6 The data from the historical corpus suggest that this feature of the Spanish spoken in Majorca is a long-established trend. Out of twenty-two occurrences of obligation periphrases (sixteen haber de, six tener de) none corresponds to the innovative form tener que.
4.3 Preverbal negative concord General standard Spanish allows negative concord in postverbal position but not preverbally, as exemplified in (2). In contrast, Catalan allows negative concord pre- and post-verbally, while single preverbal negation is ungrammatical (3): (2) nadie ha venido / *nadie no ha venido / no ha venido nadie nobody has come / nobody not has come / not has come nobody (3) *ningú ha vingut / ningú no ha vingut / no ha vingut ningú nobody has come / nobody not has come / not has come nobody
In contrast to the standard variety, in Majorcan Spanish preverbal negative concord occurs quite frequently, as in the following example taken from a local newspaper:
. In Majorca some speakers use tenir que, a Spanish calque condemned by the norm.
On language contact as an inhibitor of language change
(4) La ex-alcaldesa afirmó que en ningún momento no se ha the ex-mayor stated that in no moment not self has planteado dimitir. considered resign
(DM March 14 2008: 19).
‘The former mayor said that she has at no point considered resigning.’
It is possible that the presence of this structure in the Spanish spoken in Majorca is the result of the retention of a structure that was common in earlier stages of the language. In Old Spanish negative concord was rather common both in preverbal and postverbal position, as illustrated in the following early thirteenth-century examples extracted from Camus (2006: 1175): (5) a. Ascondense de mio Cid, ca no-l osan dezir nada. they.hide of my Cid because not-him dare say nothing ‘They hide from my Cid for they dare not say anything to him.’ b. que a mio Cid Ruy Díaz, que nadi no-l diessen posada that to my Cid Ruy Diaz, that nobody not-him give lodging ‘That no one give lodging to my Cid.’
Although this structure declined by the end of the Middle Ages, examples are still found in the ensuing centuries (Vanrell (2008) records 37.5% of preverbal negative concord with ninguno in a corpus of sixteenth-century notary documents). In fact, preverbal negative concord has not disappeared completely as it is still common in rural western Spanish dialects (Borrego Nieto 1996: 157, Aguado Candanedo 1984: 221). Considering all this, we can conclude that preverbal negative concord was likely to have been a valid option in the Spanish introduced in Majorca in the 1600s and 1700s. Quite possibly, the existence of a parallel Catalan structure encouraged the retention of this structure. The historical corpus provides evidence of these uses in the Spanish produced by Majorcans in the 1700s as in the following example:7 (6) Nunca no he dejado de rogar a Nuestro Señor. never not I.have stopped of begging to our lord ‘I have never stopped praying to Our Lord.’ (ZA 116, 24, 1772)
. As Silva-Corvalán (1998: 237) points out, a similar situation and explanation can be found in studies of Spanish–Guaraní contact. Preverbal negative concord in Paraguayan Spanish (nada no dije, literally “nothing not said”–‘I said nothing’) has also been explained (Granda 1991) as the result of the retention of the structure which, as has been already indicated, existed in general Spanish at least until the sixteenth century. The retention has been favored by contact with Guaraní, where it exists as a standard construction (cf. Silva-Corvalán 1998: 237): (i)
mba’eve nda‑’ei nothing not-said ‘I said nothing.’
Andrés Enrique-Arias
4.4 Lexical-semantic transfer with syntactic consequences A common outcome of language contact is lexical borrowing with syntactic consequences as in the following case. General standard Spanish has two distinct verbs to convey the idea of ‘ask’: pedir ‘to ask someone to do or give something’ and preguntar ‘to ask a question’. Besides this lexical contrast, the two verbs have different argument structure, which in the case of pedir entails a syntactic constraint. In general standard Spanish, only preguntar but not pedir can subcategorize an interrogative clause (7a). In contrast, Catalan has one verb, demanar, which can be used in either sense, and which can therefore subcategorize interrogative clauses (7b). (7) a.
Le *pidió/preguntó cómo estaba. him asked how was
b. Li va.demanar com estava. him asked how was ‘He asked him how he was doing.’
Apparently as a result of contact with Catalan, pedir in the Spanish spoken in Majorca has the semantic and syntactic properties of Catalan demanar. Therefore structures like (7a) are grammatical with either verb. The studies carried out from a purely synchronic perspective (cf. Moll 1961: 471; Serrano Vázquez 1996: 384–85) attribute these uses of pedir to direct transfer from Catalan. However, they miss the fact that the use of pedir in the sense of ‘to ask a question’ was not a rarity in eighteenth century general standard Spanish; in fact, in the Royal Academy’s 1737 dictionary, and all its subsequent editions through 1852, pedir is defined in its second meaning as “to ask or obtain information from someone on something”. Furthermore, the very example provided by the Royal Academy to illustrate the correct use of the word features pedir subcategorizing an interrogative clause (Real Academia: s.v. pedir):8 (8) Pedían cuál era el camino. they.asked which was the road ‘They asked which road it was.’
. The example, though, is taken from Criticón, a literary work written one century earlier by Baltasar Gracián (1601–1658). With the aim of justifying the inclusion of words in their Dictionary and to illustrate their “correct use” the academics picked excerpts from prestigious literary works. The fact that this example was included in the dictionary suggests that this use of pedir was approved of and endorsed by the educated classes.
On language contact as an inhibitor of language change
Again, the historical corpus shows that this structure is not new to Majorcan Spanish. The eighteenth-century letters and court testimonies contain examples in which pedir subcategorizes an interrogative clause: (9) Pidióle si contava la verdad. he.asked.him whether he.told the truth ‘He asked him whether he was telling the truth.’
(Vibot 2005: 66, 1833)
Although lexical borrowing may be a mechanism leading to eventual syntactic change, the uses of pedir in the Spanish spoken in Majorca do not seem to be the outcome of direct borrowing of syntactic structures or rules from Catalan into Spanish. Catalan contact has resulted not so much in the introduction of an innovative structure, but in the preservation of a language usage that existed in general Spanish, reinforced by the existence of a parallel structure in Catalan.
4.5 Deixis in demonstratives Another common phenomenon in the Spanish of Catalan bilinguals is the tendency to adopt a two-member deixis in the demonstratives, in contrast with monolingual Spanish, which has a tripartite distinction (see Table 2, adapted from Vann 1998). Table 2. Catalan and Spanish deictic systems Catalan aquest
Spanish ‘this’ ‘that’ object near S or A
este ese
aquell aquí
‘yonder’ object not near S or A ‘here’ ‘there’ object near S or A
aquel aquí/acá ahí
allà
‘yonder’ object not near S or A
allí/allá
‘this’ object near S ‘that’ object not too far from S ‘yonder’ object very far from S ‘here’ object near S ‘there’ object not too far from S ‘yonder’ object very far from S
S = speaker, A = addressee.
In the Spanish spoken in Catalan-speaking territories there is a tendency to adopt the deictic and pragmatic properties of Catalan deictic terms, using este ‘this’ and aquí ‘here’ where monolingual Spanish speakers generally use ese ‘that’ and ahí ‘there’ respectively. For instance, in Vann’s (1998) data from 58 interviews of Barcelona speakers, all the informants demonstrated some kind of influence from Catalan deictic systems, in a number of individuals 80% of the time. The same phenomenon is widely found in Majorca, as pointed out by Moll (1961: 471). Vann (1998) repeatedly characterizes these uses as contact-driven innovations. But again a historical perspective suggests that the tendency towards a bipartite distinction
Andrés Enrique-Arias
in the deixis of the demonstratives may have its roots in Spanish itself. Classical Latin had a three-place system (hic, iste and ille). As hic was eliminated from speech in Late Latin, the three-member system became reduced to the two-member system that prevails in French and Italian among other Romance varieties. In Spanish (and Portuguese), however, the three-place system was subsequently restored by transfer of ipse from its emphatic role to second person demonstrative role. The restoration of the three-member system was a slow process only consummated in early modern times. Badia Margarit (1952: 5) explains that in The Song of el Cid (ca. 1200) esse and aquel do not have their respective uses clearly delimited. Likewise, in a study of demonstratives in fifteenth century Spanish, Eberenz (2000: 263) points out that confusion of este and esse was rather common. This loss of distinction occurs in spatial deixis and is even more common in anaphoric co-reference, where the opposition between the two forms is completely neutralized: for example, it is impossible to distinguish differences in use between expressions like por esto and por esso ‘that’s why’ (Eberenz 2000: 264). In a similar fashion, the series of Castilian adverbs of place related to the demonstratives maintained a two-member system throughout the Middle Ages (aquí/acá ‘here’ allí/allá ‘there’). The emergence of the third member ahí occurs rather late: the form is first documented ca. 1490 (Badía Margarit 1952: 6). According to Eberenz the persistence of the tendency to maintain a two-member system in Castilian is due to a combination of different historical factors: (a) in origin, Latin iste (> este) had the referential properties of Modern Spanish ese, that is, historically este referred to the hearer before esse assumed that function; (b) in many communicative situations speaker and hearer share common space and topics, which makes it unnecessary to establish clear-cut differences between este and esse; (c) in the anaphoric co-reference there has been a process of convergence of the deictic value of este and the identifying one of esse (< ipse). As can be seen in the historical data from the Majorcan corpus, the tendency to converge with the Catalan two-member deictic system is a long-established trend in the Spanish of Majorca (cf. Examples 10–11): (10) Llegamos en esse Puerto en donde me mantengo. we.arrived in this port in where me I.keep ‘We arrived at this port where I am staying.’
(ZA 70, 12, 1759)
(11) Suplico a V.S. diga [al Sr Tomas] reconozca essa por suya. I.beg to you tell [Mr. Tomas] recognize this for his ‘I beg you to ask Mr Tomas to recognize this letter as his own.’ (ZA 70, 20, 1759)
In both examples the writer employs esse in cases in which most monolingual speakers would use este. It is possible that the tendency to maintain a two member system in Majorcan Spanish is the result of the continuation of language uses that were present in the speech of at least some monolingual Spanish speakers at the time when the language was introduced to the island. The influence of the Catalan two-member system
On language contact as an inhibitor of language change
would have resulted in the reinforcement of a structure that already had some sort of existence in monolingual Spanish.
4.6 Collocations Certain collocations that employ dar ‘give’, tomar ‘take’, poner ‘put’ or other verbs in standard general Spanish are used with hacer ‘do/make’ as a support verb in the Spanish of Majorca. This has been explained as the result of influence from Catalan where all the equivalent collocations employ fer ‘do/make’. Table 3. Some collocations in Catalan and Spanish ‘have a coffee’ ‘make a face’ ‘be disgusting’ ‘smell’ ‘scare’
Catalan
Majorcan Spanish
General Spanish
fer un cafè fer mala cara fer fàstic fer olor fer por
hacer un café hacer mala cara hacer asco hacer olor hacer miedo
tomar un café poner mala cara dar asco dar olor dar miedo
Again this feature of the Spanish from Majorca represents the continuation of uses that were normal in general Spanish in earlier stages of the language. In Old Spanish fazer ‘make’ and dar ‘give’ competed in numerous collocations: tristura ‘grief ’, enojo ‘anger’, gozo ‘joy’, plazer ‘pleasure’ (Dubský 1962). In some cases (asco ‘disgust’, lástima ‘pity’ and rabia ‘rage’) occurrences with hacer outnumbered those with dar well into the 1600s (Alba-Salas 2007). Moreover, the overuse of hacer in collocations is also present in general Spanish in colloquial registers (Briz 2004: 126). There is evidence that the use of collocations with hacer in Spanish in contact with Catalan is not a recent innovation; the structure is condemned vigorously in grammars of the eighteenth and nineteenth centuries addressed to teach Catalans how to speak proper Spanish (see Solà 1980 for a detailed study of such grammars).
4.7 Prepositional uses In the Spanish spoken by Catalan–Spanish bilinguals in Majorca, the preposition en is often used to express direction, something already observed by Moll (1961: 472), who provides examples like the following: (12) Se fue en Barcelona. ref he.left in Barcelona ‘He left for Barcelona.’ (13) A las nueve llegamos en la ciudad. at the nine we.arrived in the city ‘At nine we arrived in the city.’
Andrés Enrique-Arias
I have collected similar examples from the local press in the Balearic Islands: (14) La pareja se trasladó en esta urbanización. the couple ref moved in this housing-complex ‘The couple moved to this housing complex.’ (DM June 9, 2006: 28)
The structure in examples (12–14) is ungrammatical in standard Spanish, which uses en for location and a for direction as shown in the examples under (15): (15) a.
Estoy *a/en Palma. ‘I am in Palma.’
b. Voy *en/a Palma. ‘I go to Palma.’
In contrast, the Catalan spoken in Majorca uses a to express both location and direction as exemplified in (16):9 (16) a.
Som *en/a Palma. ‘I am in Palma.’
b. Vaig *en/a Palma. ‘I go to Palma.’
Moll (1961: 472) attributes the non-standard preposition use in (12–13) to the different distribution of a and en in Catalan and Spanish, which purportedly causes confusion for bilingual speakers. But this explanation is at best incomplete: if Majorcan Catalan uses a for situation and direction, one would expect that in their Spanish utterances Majorcans would overuse the preposition a, not en, to express spatial relations; that is, from a comparison of the related Spanish and Catalan structures it follows that the directional uses of en in the Spanish from Majorca are not the result of simple direct transfer from present-day Catalan.
. In the Catalan of Majorca the preposition a develops an epenthetic -[n] segment when it occurs in front of articles and demonstratives that begin with a vowel. (i)
Vaig an es poble. ‘I go to the town.’
This phonetically conditioned variant of the preposition a is homophonous with the preposition en (they are both pronounced [6n]) and in fact is usually spelled en, which gives the impression that there is a generalized confusion of the prepositions a and en in Balearic Catalan. However, if we look at those cases in which there is no epenthetic -[n] it is clear that there is not such confusion: (ii)
Vaig *en/a sa plaça. ‘I go to the square.’
On language contact as an inhibitor of language change
As I have explained in greater detail in Enrique-Arias (2005), there are a number of synchronic and diachronic factors that help explain the existence of this structure in the grammar of bilingual Majorcans. A relevant piece of data is that the directional uses of en are common in the varieties of Spanish that come into contact with other languages, such as Galician in Galicia (Rojo 2004: 1097), Guaraní in Paraguay (Choi 2001), English in Texas (García 1982) or Fang in Equatorial Guinea (Granda 1988; Lipski 1990). This suggests that we are dealing with a phenomenon motivated by the tendency toward simplification and restructuring of grammatical rules that is common in language contact situations. It should also be noted that this area of grammar is particularly vulnerable to change, since the uses of a and en, or their equivalents, have fuzzy limits within monolingual varieties of all the Romance languages (cf. examples in Choi 2001: 188–193). For instance, in standard Spanish en may be used to express direction (entrar en la casa ‘enter the house’) while a in certain cases may express static location (esperar a la puerta ‘wait by the door’). Thus, the contact situation has the effect of extending a structure that already existed in general Spanish to more contexts. Finally, one must take into account the inherent difficulty in learning and using prepositions: in a study of the acquisition of ten Spanish morphemes by monolingual Spanish children ages 2:0 to 4:8 Kvaal et al. (1988) found that the preposition en was one of the latest features to be acquired, only second to the irregular preterit indicative. The difficulty of acquiring the standard use of prepositions may be even greater for unbalanced bilinguals especially when, as in the case of Catalan-dominant bilinguals, there is no clear frame of reference in the speaker’s first language. A diachronic perspective complements the synchronic factors outlined above, unveiling other circumstances that may have contributed to the introduction of the directional uses of en in Majorcan Spanish. Unlike present-day Catalan, in the language of the eighteenth century en was frequently used to express direction, as documented in the Catalan letters in the ZA corpus: (17) Aurà de passar en Vic. will.have of pass in Vic ‘He will have to get to Vic.’ (18) Arribà lo dia de Pasqua en Madrit. he.arrived the day of Easter in Madrid ‘He arrived in Madrid at Easter.’
(ZA 20, 21, 1743)
(ZA 52, 6, 1756)
Spanish letters written by Majorcans in the corpus also use en to express direction: (19) Llegamos en esse puerto. we.arrived in that port ‘We arrived in that port.’ (20) En el viernes pasado pasamos en Palma. in the Friday past we.passed in Palma ‘Last Friday we went to Palma.’
(ZA 70, 11, 1759)
(ZA 121, 24, 1773)
Andrés Enrique-Arias
The historical data suggest that the current directional uses of en in Majorcan Spanish are the continuation of a structure that was already firmly established in the Spanish used in Majorca in the eighteenth century. In turn, this feature of Majorcan Spanish would have been encouraged by the existence of a parallel construction in the Catalan of the eighteenth and nineteenth centuries.
5. Summary and conclusions In this chapter I have analysed the historical development of a number of structures found in the variety of Spanish spoken in Majorca in order to argue that language contact can be an effective inhibitor of language change. The phenomena studied involve linguistic variants that have decreased in frequency or even disappeared in general Spanish but exhibit considerable vitality in the Majorcan variety. While previous studies simply attribute these features of Majorcan Spanish to direct transfer from Catalan, the explanation offered here is that the Spanish spoken in Majorca has not so much borrowed or adopted Catalan morphosyntactic structures, but rather has retained existing variants that paralleled Catalan structures. Some of these, like the morphological future or the expression of obligation with haber de are, while receding in most varieties, valid options in standard Spanish. However, some other structures, like preverbal negative concord, have disappeared in general Spanish and survive only in marginal rural dialects. Their maintenance in the Spanish of Majorca can be seen as evidence of how parallel structures favour retention of archaic features. The historical analysis yields two findings that support the view that Catalan language contact has favored the retention of archaic features in Majorcan Spanish. To begin with, all the phenomena examined had some sort of existence in monolingual Spanish varieties at the time when the language was introduced to the island of Majorca. This poses a challenge to the view that these structures are due to direct transfer from Catalan, as one of the necessary conditions to establish the existence of contact-induced change is to prove that the proposed interference features were not present in the pre-contact variety (see for instance Thomason 2001: 93–94). Furthermore, the data from the historical corpus of Majorcan Spanish reveal that there has been a historical continuity for some of the features that characterize the variety of Spanish spoken in Majorca today. Additionally, it has been shown how for some phenomena the retention of traditional features may have been further encouraged by the socially dominant status of Spanish in the last three centuries. As Catalan-dominant bilinguals tended to acquire and use Spanish in formal settings, they used less those innovative variants that were characteristic of casual speech. Evidently, the whole topic of mechanisms of cross-linguistic interference in the history of Spanish–Catalan contact is far from being completely understood. In any
On language contact as an inhibitor of language change
case, I hope to have demonstrated the relevance of a diachronic perspective in obtaining a more complete understanding of the different factors involved in the formation of contact varieties.
Primary sources Davies, Mark. 2008. Corpus del español. [Consulted on line at http://www.corpusdelespanol.org]. Diario de Mallorca (1955-) [=DM]. Palma de Mallorca: Editorial Balear Real Academia Española. Nuevo tesoro lexicográfico de la lengua española [Consulted on line at http://www.rae.es]. Vibot, Tomàs. 2004. Crims, infàmies i immoralitats a Esporles i Banyalbufar (Segles XVII-XIX). Palma: Associació Cultural Index, Edicions el Moixet. Vibot, Tomàs. 2005. Crims, infàmies i immoralitats a Bunyola i Orient (Segles XVI-XIX). Bunyola: Collectiu Cultural Sitja. Zaforteza family epistolary archive [=ZA]. File 6. Vinagrella estate (Llubí), Majorca.
References Aaron, Jessi E. 2007. El futuro epistémico y la variación: Gramaticalización y expresión de la futuridad desde 1600. Moenia 13: 253–274. Aguado Candanedo, David. 1984. El habla en Bercianos del Real Camino (León): Estudio socio lingüístico. León: Diputación de León. Alba-Salas, Josep. 2007. On the life and death of a collocation: A corpus-based diachronic study of dar miedo/hacer miedo-type structures in Spanish. Diachronica 24(2): 207–252. Badia i Margarit, Antoni M. 1952. Los demostrativos y los verbos de movimiento en iberorrománico. In Estudios dedicados a Menéndez Pidal, Vol. 3, 3–31. Biber, Douglas. 1995. Dimensions of Register Variation: A Cross-Linguistic Comparison. Cambridge: CUP. Blas Arroyo, José Luis. 2004. El español actual en las comunidades del ámbito lingüístico catalán. In Historia de la lengua española, Rafael Cano Aguilar (ed.), 1065–1086. Barcelona: Ariel. Blas Arroyo, José Luis. 2007. El contacto de lenguas como factor de retención en procesos de variación y cambio lingüístico. Datos sobre el español en una comunidad bilingüe peninsular. Spanish in Context 4(2): 263–29. Borrego Nieto, Julio. 1983. Norma y dialecto en el sayagués actual. Salamanca: Acta Salmaticensia. Briz, Antonio. 2004. El castellano en la Comunidad Valenciana. Revista Internacional de Lingüística Iberoamericana 2(2): 119–130. Camus Bergareche, Bruno. 2006. La expresión de la negación. In Sintaxis histórica de la lengua española. Primera parte: La frase verbal, Concepcion Company (ed.), Vol. 2, 1165–1252. Mexico: Fondo de Cultura Económica and Universidad Nacional Autónoma de México. Choi, Jenny K. 2001. The genesis of voy en el mercado: The preposition en with directional meaning in Paraguayan Spanish. Word 52: 181–196. Dubský, Josef. 1962. Las formas descompuestas en el español antiguo. Revista de Filología Española 46: 31–48.
Andrés Enrique-Arias Eberenz, Rolf. 2000. El español en el otoño de la Edad Media. Madrid: Gredos. Enrique-Arias, Andrés. 2005. On the origin of the preposition en to express ‘direction to’ in the Spanish spoken in Mallorca. Paper read at the 17th International Conference on Historical Linguistics, Madison (Wisconsin). Enrique-Arias, Andrés. 2006. Spanish/Catalan contact in historical perspective: 18th century documents from Majorca. In New Perspectives in Iberian Dialectology/Nouvelles perspectives en dialectologie ibérienne, David Heap, Enrique Pato & Claire Gurski (eds). London: The University of Western Ontario. (online edition
On language contact as an inhibitor of language change
Moll, Francisco de B. 1961. El castellano en Mallorca. In Studia Philologica. Homenaje ofrecido a Dámaso Alonso por sus amigos y discípulos con ocasión de su 600 aniversario, Vol. 2, 469–474. Madrid: Gredos. Moreno Fernández, Francisco. 2004. Cambios vivos en el plano fónico del español: variación dialectal y sociolingüística. In Historia de la lengua española, Rafael Cano Aguilar (ed.), 973–1009. Barcelona: Ariel. Penny, Ralph. 2000. Variation and Change in Spanish. Cambridge: CUP. Rojo, Guillermo. 2004. El español de Galicia. In Historia de la lengua española, Rafael Cano Aguilar (ed.), 1087–1101. Barcelona: Ariel. Romera, Magdalena. 2003. La variedad del castellano actual en Baleares. Moenia 9: 359–381. Serrano Vázquez, María del Carmen. 1996. Interferencias léxicas y semánticas en una situación de contacto entre dos lenguas, catalán y castellano. In Las lenguas en la Europa Comunitaria II, Mercè Pujol Berché & Fermin Sierra Martínez (eds), 375–394. Amsterdam: Diálogos Hispánicos. Silva-Corvalán, Carmen. 1994. Language Contact and Change. Spanish in Los Angeles. Oxford: OUP. Silva-Corvalán, Carmen. 1998. On borrowing as a mechanism of syntactic change. In Romance Linguistics: Theoretical Perspectives [Current Issues in Linguistic Theory 160], Armin Schwegler, Bernard Tranel & Myriam Uribe-Etxebarria (eds), 225–246. Amsterdam: John Benjamins. Solà, Joan. 1980. Tractats de catalanismes. In Miscellània Aramon i Serra: estudis de llengua i literatura catalanes oferts a Ramon Aramon i Serra en el seu setantè aniversari, Emilio Alarcos Llorach et al. (eds), Vol. 2, 559–582. Barcelona: Curial. Thomason, Sarah. 2001. Language Contact. An Introduction. Edinburgh: EUP. Vann, Robert E. 1998. Pragmatic transfer from less developed to more developed systems: Spanish deictic terms in Barcelona. In Romance Linguistics: Theoretical Perspectives [Current Issues in Linguistic Theory 160], Armin Schwegler, Bernard Tranel & Myriam Uribe-Etxebarria (eds), 307–318. Amsterdam: John Benjamins. Vanrell, Bruno. 2008. La evolución de la doble negación preverbal en el castellano de la Edad Media. Ms, Universitat de les Illes Balears.
Towards notions of comparative continuity in English and French* Remus Gergel
Universität Tübingen The present chapter investigates comparative clauses in English and French with a special focus on subject–verb inversion. It explores the derivation and diachronic evolution of such clauses and makes a case for continuity by defending a simple derivation that requires a so-called low subject, that is one that is not forced to the edge of the clause. Based on well-known options from the diachrony of English and French, I argue that the non-movement syntax proposed is the key continuous aspect against the background of several changes given in the histories of the two languages.
1. Introduction The present chapter investigates comparative clauses in English and French with a special focus on subject–verb inversion. Initial cases of comparative inversion (CI) are illustrated in (1) below.1
*The present enterprise owes special thanks to John Vanderelst for discussions on French. I am very grateful to the organizers of the Cambridge workshop for the opportunity to explore comparative clauses in a continuous way, to Chris Lucas also for numerous helpful comments concerning both form and content, to Jacqueline Guéron for a very instructive discussion on inversion, to Polina Berezovskaya for hints on materials and proofreading, and to two reviewers for their exceptionally constructive comments. Only I am responsible for any remaining errors. . CI is register-based and optional. This chapter focuses on the continuity of the inverted clauses. Potentially interesting aspects in the relationship between inverted and non-inverted derivations from the point of view of grammatical theory are thus not within present scope (cf. Culicover & Winkler 2008; Gergel 2008; Gergel, Gengel & Winkler 2007; Merchant 2003; Winkler 2005, among others, for some suggestions). At the same time, nothing special is assumed for non-inverted comparatives here.
Remus Gergel
(1) a. [T]he Scots have spent less on their system [than has the prison service of England and Wales]. (http://news.bbc.co.uk/2/hi/uk_news/magazine/3932653.stm)
b.
And they were no more persuaded by democracy [than was the Pharaoh]. (http://www.nytimes.com/2008/09/11/opinion/11Cohen.html)
c. Optogenetic stimulation can potentially target problem cells much more precisely [than can the electrodes used in DBS]. (Scientific American 08/2008, p. 59)
We will argue that the key continuous element in the diachronic trajectory is a low subject. Standardly, the initial line-up of the subject is within the VP (and below the auxiliary position). But the subject is then usually assumed to move to the edge of the clause. A low subject on the derivation proposed here means in particular that the subject is not moved to the edge of the clause (technically, the inflectional domain), as would be required on standard generative approaches. Two observations worth making from the start are: (i) from a diachronic point of view, this type of syntactic behavior is well-known independently of comparatives from early Old French/English (OF/OE) onwards; (cf. Adams 1987; Haeberli 2002; Kroch 2008); (ii) several syntactic properties which could potentially affect comparatives have changed in both languages, though without always impacting on the derivation of CI (we will restrict attention to the most direct interferers). We will argue that the now relic possibility of a lack of subject not moved to the Spec,TP position is the factor that remained constant in comparatives in both languages and which derives the core similarities. The chapter is structured as follows. Section 2 gives some details on how the syntax of comparatives in English shows various signs of an archaism, which consists in the possibility of not filling the Spec,TP position overtly. This seems to go against the grain of the requirement that every finite clause have an overt subject in that very position (the so-called EPP feature). But we show that such structures have been possible in comparatives at all major stages of the language (via null expletives; see Haeberli 2002; Hulk & van Kemenade 1995; Williams 2000, among others). Section 3 discusses a similar phenomenon in French, for which we present evidence that it has also been persistent from early Old to Modern French and for which we develop a similar analysis. In Section 4, additional aspects relating to (lack of) change in comparatives as well as potential interferences with the simplest continuity hypothesis are discussed. We thus investigate the potential contact of English with French and other features that have been claimed in the literature to have an influence on V2 (e.g. Northern features). Section 5 concludes the chapter and formulates certain open questions including the (non)interaction with verb movement and the possibility of also having semantic continuity in comparatives.
Towards notions of comparative continuity in English and French
2. Comparative inversion in English In this section, we propose that CI is simpler than on the standard generative view and that the analysis proposed is historically consistent. It represents an archaic possibility of the language. Earlier English could produce apparent V2 (and with subject pronouns clearly V3) structures by moving the finite element to T (and not exclusively higher up to C); see Fischer et al. (2000); Haeberli (2002); Pintzuk (1991); Speyer (2008), among others. It is crucial that the subject is not displaced to the edge of TP in such structures. (Or if it were, there would be no inversion). We will argue that CI patterns with this type of simple inversion and show how its derivation fits the current grammars first (Section 2.1), before addressing its diachronic evolution (in Section 2.2).
2.1 English CI involves no movement to C At first sight, inversion in clausal comparatives, as in (1) above, seems amenable to a raising of the finite element to C. This corresponds to the analysis of matrix questions in English or run-of-the mill V2 structures in Germanic. (See Merchant 2003; Gergel, Gengel & Winkler 2007, among others.) But there is an alternative and diachronically more cogent analysis of CI. To illustrate the difference: on a standard analysis, the subject is moved to the edge of the clause (Spec,TP) and the finite element is moved to C; (2a). Under the alternative to be defended, however, the subject is in a sub-T position (Spec,VP here) and the finite element is in T; (2b).2
(2) a.
Structure for the C-based analysis CP wh (Comp Op)
C′ TP
C Finite
T′
DP Subject
T
VP DP
V′
tsubject
…
. This proposal in and of itself is independent of how a finite element can (or cannot) get to T. Unlike modals, do and have/be (cf. (1)), lexical verbs cannot appear in T in current English, as a reviewer notes. Clearly, French and earlier English have the additional possibility of V-to-T; see Section 5.
Remus Gergel
b. T-based analysis CP C′
wh (Comp Op)
TP
C than/as
T′
T Finite
VP DP
V′
Subject
…
Thus, following Culicover & Winkler (2008) and Gergel (2008), we argue that neither T nor the subject undergoes movement to C and Spec,TP, respectively. An argument for the alternative that does without the two movement steps is that, in sequences of auxiliaries, the subject is not forced to follow the first auxiliary, (3); see Huddleston & Pullum (2002), Potts (2002), among others. This contrasts with questions, which do not allow auxiliaries to invert together around the subject, as shown in (4).3
. The formulation “do not allow” for questions becomes crucial to capture the appropriate contrast. Comparatives do allow auxiliary clusters to appear inverted around the subject, as in (3). They do not force the sequences across all contexts, however, cf. (i). The interaction of the conditions in CI may thus be even more complex than has been thought (e.g. in Culicover & Winkler 2008; Gergel 2008; Huddleston & Pullum 2002). An important independent factor is that in limited cases under the scope of negation, English allows intercalated subjects in one type of CI after all, as in (i).
(i)
a. Abouten it is gras spryngyng,/For moiste so thikke and wel liking/That it ne may in wynter dye/No more than may the see be drye. (Romaunt of the Rose, Frag.A:1563) b. Of what he spoke I could tell you no more than could the children of Hamelin have told the tune the Pied Piper played. (J. Jerome, Paul Kelver, p. 34. London: Hutchinson)
Given the well-known inversion-triggering potential of negation in English over various environments including different types of degree constructions (cf. Borroff 2006; Green 1976, among many others) this is not entirely surprising. What is crucial for immediate purposes is that in particular in base examples such as (3) subject intervention is impossible, while in interrogatives as in (4) it is obligatory.
Towards notions of comparative continuity in English and French
(3) The notes […] are therefore more helpful [than could be the opinions of many other critics, including myself]. (http://www.amazon.com/review/RSKNPKI3HDRUB, 06/2007)
(4) a. Could that grumpy man be the next prime minister? b. *Could be that grumpy man the next minister?
Another possible argument for the proposal obtains under the assumption that the introducer of clausal comparatives than is inserted under C (Hankamer 1973; cf. (2b)). C-based V2 is, then, implausible since such movements targeting C do not take place in conjunction with overt complementizers (e.g. den Besten 1983). This derives the simple analysis of CI. But there is a valid objection, as pointed out by a reviewer.4 Current English grammar has the (restricted) capacity to generate certain inversion structures under a complementizer such as that with the follow-up possibility of CP-recursion (cf. Davison 1979; Green 1976; Iatridou & Kroch 1992, among others). The variety of such inversion structures would certainly deserve more space by itself. We discuss two types that are often invoked and then their immediate relationship to CI: negative preposing and locative inversion. While traditionally called main clause phenomena, such structures can surface within embedded contexts. We illustrate this in (5), for locative inversion, and (6), for negative inversion (cf. Green 1976: 384, 385).
(5) I realized that in would come the milkman, with me there, and my hair in curlers.
(6) I knew that never before had prices been so high.
. As the reviewer also points out, a further issue is colloquial English “than what” constructions. An illustration is (i), on a variation of (1a) in Section 1.
(i)
The Scots have spent less on their system than what the prison service of England has.
Such constructions have been invoked in particular since Chomsky (1977) as an argument for wh-movement. However, in a series of more thorough studies the observation has been refuted in its quality as an argument for genuine clausal comparatives (cf. especially den Besten 1978). What den Besten shows is that such structures are best treated as free relatives. That is, as an equivalent to Hankamer’s non-clausal “than XP”. Notice at the same time that if Chomsky’s remark hadn’t turned out to need refinement, the impossibility of inversion could be explained on the account we argue against. For instance, as follows, capitalizing on the reviewer in question’s remarks: assume a featural difference between C in comparatives and interrogatives, from which the inviolability of the Doubly-filled Comp filter in comparatives follows. Finally, there is one more caveat to the issue (independently of the analysis of CI one favors). The impossibility of inversion in “than what” structures does perhaps not have to be enforced within narrow syntax (Lasnik & Sobin 2000; Merchant 2003). Instead, the colloquialism of such structures and the high-register status of CI place them in disjoint grammars.
Remus Gergel
The structures contain a complementizer. Hence the possibility of recursion. (Simply put: to accommodate that in one CP, and the preposed constituent in the specifier position of the recursively subordinated one). However, a first incongruity obtains between locative inversion and than (see Aissen 1975: 8; Hartmann 2005: 95):
(7) *On the ground lie more apples than on the tree grow.
The recursion possibilities in comparative contexts with negative preposing can also be shown (empirically) to be more restricted than they appear with the standard complementizer that. Thus, even though negative preposing can surface under that, as we saw in (6), it is ungrammatical under the comparative complementizer than in (8).5
(8) *Prices got higher than never before had wages been.
Finally, from a general point of view, the literature on CP-recursion contains a fair amount of diverging suggestions as to the exact restrictions responsible for its licensing. However, one common feature is that the complementizers in question are explicitly selected by verbs (cf. Iatridou & Kroch 1992). This criterion is not fulfilled in the case of than either (selected by -er/more). We hence preliminarily conclude that recursion is unavailable in comparatives. Next, we turn to the diachronic aspects of CI.
2.2 Diachronic continuity in the syntax of CI In order to show that CI was T-based at earlier stages, we can draw on independent facts about the syntax of OE. First, OE had the canonical V2 order of Germanic (produced through movement to C, e.g. in matrix questions). Second, however, it had a mechanism that only displaced the finite verb to T (e.g. with topicalization; cf. Fischer et al. 2000; Kroch, Taylor & Ringe 2000; Pintzuk 1991).6 Importantly, the latter strategy would produce V3 orders, namely with subject pronouns between topicalized phrases
. It is possible to rule out such structures by using the semantics of comparison (von Stechow 1984). Playing devil’s advocate it could be counter-argued that recursion is still possible, then, but that it is ruled out for independent reasons. But a grammar in which such limited options of recursion are additionally restricted by important factors including semantic ones only has vanishingly small possibilities of keeping the hypothetical option as viable in comparatives. Chris Lucas (p.c.) notes that a non-inverted version of (8) is less ungrammatical, but it is extremely hard to parse or to assign meaning to. This observation is important since it is consistent both with the semantic background and the simple present argument. Modern negative inversion gains ground in the seventeenth century (Nevalainen 2006) and it is unlikely a factor of historical continuity. . Both the VP and the TP in OE could be head-final (as well as head-initial; cf. Pintzuk 1991). We discuss the significance of TP headedness in some detail from the perspective of CI in Section 4 below.
Towards notions of comparative continuity in English and French
and verbs. The asymmetry is sketched in (9) vs. (10) with a wh-question and topicalization, respectively (cf. Fischer et al. 2000; Kroch et al. 2000; the OE material in (9) and (10) is taken here from Fischer et al. 2000: 118). (9) For hwam noldest þu ðe sylfe me gecyðan þæt… for what not.wanted you you self me make.known that ‘Wherefore would you not want to make known to me yourself that…’ (10) Forðon we sceolan mid eally mod & mægene to Gode gecyrran therefore we must with all mind and power to God turn ‘Therefore we must turn to God with all our mind and power.’
The test gives the possibility of a clear prediction for OE, which could use both mechanisms in general, depending on the context (e.g. questions vs. topicalization as above). The main diagnostic to test the two types of apparent V2 effects are pronouns, which are taken to reside in the specifier of the highest functional head in the Infl-domain, here for concreteness: Spec,TP. If the finite verb surfaces to the left of pronouns, then it will be under C. (In which case the canonical type of Germanic V2 analysis would be confirmed). If the finite element is invariably on the right of the subject pronoun in OE, it must be under T. Applying the test to comparatives in OE shows that they show no inversion with pronouns, but appear only as shown in (11). (11) for.ðan.þe se Fæder is mare [þonne ic sy]. because the father is greater than I am ‘For the Father is greater than I.’ (YCOE, coaelhom, ÆHom 10:19.1417)
This means that the finite verb in OE comparatives does not go beyond the TP (or the Infl-domain, more generally).7 Another diachronically rooted argument is that if a representation that does not require an overt subject in Spec, TP is on the right track, then we expect null expletives in comparative clauses. An (EModE) example of this sort is given in (12). (12) [H]e left soch a companie of fellowes and scholers in S. Iohnes Colledge, [as _ can scarse be found now in some whole vniuersitie]. (PPCEME-ASCH-E1-H,55R.164)
There are some contexts in current English that naturally have null (and, as Chris Lucas, p.c., points out, in standard British English only null) subjects in comparatives; see the passives in (13a,b) and the impersonal in (13c).
. A point of change is that the diagnostic does not have the same force in Middle English (ME), where pronoun subjects start inverting (even though even here examples in comparatives are rare; cf. also the ModE situation, where contrast is needed for the speakers who accept CI with pronouns.)
Remus Gergel
(13) a. [T]he tropics play a more active role [than _ was thought] in controlling the Earth’s climate. (www.sciencedaily.com/releases/2005/10/051012084249.htm) b. Davis Cup exit poses more questions [than _ could be answered in a year]. (www.timesonline.co.uk/tol/sport/tennis/article4800983.ece) c. [P]ast experiences need to be reviewed in a rather more circumspect manner [than _ has been the tendency to date]. (www.bilaterals.org/article.php3?id_article=11194)
The motivation for this move lies in the historical links between null expletives and V2. For correlations between the loss of the English type of V2 and null expletives, we draw on Haeberli (2002) and transfer the basic logic to comparative clauses. (See also van Kemenade 1997 and especially Hulk & van Kemenade 1995: 246 with partly equivalent implementations, formulated in terms of expletive pro-drop.) We can thus suggest the following. The overt subject in CI can stay low because, structurally speaking, the higher subject position does not need to be overtly filled (or equivalently: it is filled by a null-expletive; this technicality avoids conflicts with the EPP, i.e. the principle that the clause must have a subject in Spec,TP) A further argument can be culled from the continuous presence of CI in the records. While CI has never seemed to be a high-frequency phenomenon, it has been attested throughout the history of English; (14) comes from the early OE text of Beowulf. (14) Næfre ic maran geseah eorla ofer eorþan ðonne is eower sum… never I greater seen of warriors on earth than is of.you one ‘I have never seen a greater warrior on earth than is one of you.’ (Beowulf, III.247)
To summarize, we have so far argued that CI involves a non-standard derivation with low subjects and that the phenomenon has been historically persistent in the history of English. In the next section, we will explore a similar phenomenon in French in which comparatives are the locus of conservativism. We will thus extend both the observation about continuity and the analysis proposed.
3. On the syntax of comparative inversion in French In this section we first clarify that French has inversion in comparatives too. We subsequently argue that French CI also involves low subjects and show that the phenomenon is also historically persistent from some of the earliest records onwards.
3.1 CI in Modern French Modern French allows subject–auxiliary inversion in comparative clauses:
Towards notions of comparative continuity in English and French
(15) a.
Pierre est plus grand que n’ était son père. Pierre is more tall than en-was his father ‘Pierre is taller than his father was.’
b.
[L]es délibérations […] sont ouvertes au public dans un degré the deliberations are open to.the public to a degree
plus large que ne le sont les activités d’autres institutions… more large than en it are the activities of-other institutions.
‘The deliberations are more open to the public than the activities of other institutions.’
(http://www.infocom.gc.ca/reports/section_display-f.asp?intSectionId=301)
The French examples additionally feature an expletive negative (EN) element, the larger consensus on which is that it does not have negative meaning (cf. Price 1990; von Stechow 1984). EN is not a characteristic of inversion but of comparatives; more specifically, of clausal comparatives (of inequality). Non-inverted clauses display it too and equatives need not contain the EN, whether inverted or not; see (16) and (17).8 (16) [L]a télévision est plus différente du cinéma the television is more different from.the cinema que le cinéma ne l’est de la photographie. than the cinema en it.is from the photography ‘The difference between television and cinema is greater than the difference between cinema and photography.’ (http://unesdoc.unesco.org/images/0014/001437/143768fb.pdf) (17) a.
Marie est aussi grande que l’ était sa mère. M. is as tall as it-was her mother ‘Marie is as tall as her mother (is).’
b.
… que sa mère l’ était.
c.
… que sa mère ne l’ était.
d. … que ne l’ était sa mère.
By testing sequences of verbal heads in French, we can observe that the subject also follows the entire sequence rather than just the structurally highest finite member
. The negative expletive intruding into the equative is a later development of modern French. OF had the negative morpheme in comparatives of inequality but not in equatives (cf. also Section 5 below).
Remus Gergel
of the sequence. Crucially, then, this option is possible (and virtually obligatory) in inverted comparatives in French; see (18).9 (18) a.
Elle est plus jolie [que n’a été sa mère]. she is more beautiful than en-has been her mother ‘She’s more beautiful than her mother was.’ (FM, p. 191)
b.
Ceux-ci ne seront pas touchés par la nouvelle carte militaire, those not will.be not touched by the new law military
pas plus que ne devrait l’être la base aéronavale de L.-B.]. not more than en should it-be the base aero-naval of L.-B.
‘They will not be more affected by the new military law, not more than the aero-naval base Lann-Bihoué should be.’ (www.ouest-france.fr/actu/actu_BN_-Carte-militaire-le-couperet-tombeaujourd-hui_8619-674581_actu.htm) c.
[U]ne langue qui change plus vite que ne sont rédigés les a language that changes more fast than en are redirected the
articles du future dictionnaire. articles of.the future dictionary
‘A language that changes faster than the articles of the future dictionary are updated.’
(www.chass.utoronto.ca/~wulfric/siehlda/actesmen/ilt1_2.htm)
d. [U]ne famille est pauvre si l’ essentiel, comme la nourriture, a family is poor if the-essential like the nourishment
le logement et l’habillement, absorbe un pourcentage the lodging and the-clothing absorbs a percentage
plus élevé de son revenu more high of its income
. The participial and the perfect auxiliary appear separated on the surface in the single (verse-based) example I was able to find in (i). John Vanderlest (p.c.) points out that any examples of this sort (i.e. separated) are marked at the very least in current French and I have no quantitative figures for early French. Thus while such examples merit further investigation, notice, again, that it is the existence of the clusters that argues against a standard raising to C. (Conversely, it is technically possible to have the subject still lower than the specifier of the highest inflectional projection in such examples.) (i) Et plus vermeille que n’est rose arousee. and more red than en.is rose made.rosy ‘And (she was) redder than a rose.’ (CLM corpus: Adenet le Roi, 8153, Enfances Ogier – Œuvres d’Adenet le Roi, (1269–1285))
Towards notions of comparative continuity in English and French
[que ne devrait dépenser la famille canadienne moyenne]. than en should spend the family Canadian average
‘A family is poor if the essentials such as food, lodging and clothing absorb a higher percentage of its income than the average Canadian family should spend.’
(www.unicef.org/french/sowc06/pdfs/repcard6f.pdf)
The existence of CI is then reminiscent to some degree of the situation in English. But the question arises whether more appropriate tests for the syntax of French can be applied. Let me therefore next introduce some background to the additional intricacies of subject–verb inversion in French. I capitalize here on ideas pertaining to inversion that have been suggested in Lahousse (2006) and Rowlett (2007), independently of comparatives, who consider a range of inversion possibilities in the language and classify them syntactically. We will test French CI on the basis of such tests and argue that it offers evidence that the subject is in a structurally low position. More specifically, French CI parallels to a large extent one particular type of stylistic inversion of French.10 A first distinction needs to be drawn between the so-called ‘pronominal’ inversion of the standard variety, as well as the related ‘complex’ inversion, which are illustrated in the two examples in (19), on the one hand, and so-called ‘stylistic’ inversion on the other (stylistic inversion will be further subdivided momentarily). (19) a.
Quand est-elle partie? b. when is-she left ‘When did she leave?’
Quand Claire est-elle partie? When Claire is-she left ‘When did Claire leave?’
By definition, pronominal inversion, such as the type witnessed in (19a), involves a pronoun subject, as does in fact the type of complex inversion seen in (19b) in addition to the full-DP subject it displays. Stylistic inversion, which will be of interest for current purposes, does not share the restriction to subject pronoun clitics of complex inversions in general (whether with or without accompanying full DPs) and shows distinct properties throughout.11 Since both types in (19) are clearly distinct from (the
. The terms Rowlett uses are ‘genuine SI’ for the low-subject inversion and ‘focus SI’ for the remnant-movement inversion, where focus is taken as wide-focus. I avoid these terms, as CI, though syntactically more on a par with genuine SI may not be independent of informationstructure (cf., e.g. Culicover & Winkler 2008; Gergel et al. 2007; Winkler 2005). . The CI of French is unacceptable with pronouns in current French and I could not find appropriate examples from earlier stages. The restriction cannot be blamed entirely on the clitic status of the subject pronouns as the tonic versions of the pronoun paradigm are degraded too.
Remus Gergel
various types of) inversion that include CI, we next focus on stylistic inversion. (See e.g. Kayne & Pollock 2008, Rizzi & Roberts 1989; Rowlett 2007 for discussions of complex inversion.) More specifically, we will next inspect a two-fold subdivision within the syntactic mechanisms in the class of stylistic inversion. According to Rowlett (2007: 217–223) (cf. also especially Lahousse 2006), there are two major types of stylistic inversion in French. The terms used in my two main sources are not fully identical (much less in the rest of the literature), but importantly the mechanisms are, and they can be summarized as follows. Crucially, one type of inversion is, simply put, a low-subject type of inversion. (The other type of inversion finds a remnant-movement-type implementation in Rowlett’s analysis. Since CI will turn out to be closely related to the first type, I leave the more technical details in the implementation of the type based on remnant evacuation aside.) A first characteristic of the low-subject inversion observed is that adverbs can be found intercalated between the finite verb and the subject: (20) Ici se cachent souvent des lapins. here self hide often some rabbits ‘Rabbits often hide here.’
Second, verbal dependants can follow a low subject (cf. (21)). But, third, the subject must be final, if the mechanism that brings about inversion is remnant movement (the latter case is illustrated with a stage topic in (22) below; cf. Rowlett 2007). (21) Quand passeront Jean et Pierre sous la fenêtre? when will.walk J. and P. under the window ‘When will Jean and Pierre walk under the window?’ (22) a.
Seuls passeront sous la fenêtre Jean et Pierre. alone will.walk under the window J. and P. ‘Jean and Pierre alone will walk under the window.’
b. *Seuls passeront Jean et Pierre sous la fenêtre.
Fourth, quantifier float is impossible in stylistic inversion, as the contrast below (following Rowlett) between low-subject and remnant-movement inversion shows: (23) *Quand ont tous fini les enfants? when have all finished the children ‘When did all the children finish?’ (24) Ont tous réussi ceux qui ont assez travaillé. have all succeeded those who have enough worked ‘Those who did enough work were all successful.’
The reason invoked by Lahousse (2006) for effects such as (23–24) is that the floating quantifier needs to be in a binding relationship with the trace of a raised subject. (This argument in turn is based on Doetjes’ 1992 analysis of quantifier float.)
Towards notions of comparative continuity in English and French
The immediate question is whether, within the class of stylistic inversions, CI patterns with low-subject inversion or with remnant-movement inversion in the grammar of French. We will argue that the former analysis is more appropriate for CI. First, we find that adverbs such as souvent, ‘often’, can be intercalated between the finite verb and the subject, as in (25). (25) La vraie vie est plus théâtrale [que ne l’est souvent le théâtre]. the true life is more theatrical than en it-is often the theatre ‘Real life is more theatrical than theatre often is.’
The standard logic of the argument is as follows: since the adverb delimits the edge of the VP (assume simply via adjunction), the subject can be observed to be within the VP. This offers, then, some indication for the low subject position.12 Second, various dependants can be found in linearly post-subject positions on the surface of CI structures. This confirms in particular the second test discussed above for stylistic inversion more generally. (In simplified terms: a remnant movement analysis, as opposed to the low-subject inversion, would require subjects to be final.) (26) a.
Je me défonce plus pour mon chef que ne le fait Martine I self hit more for my boss than en it does Martine
pour le sien. for the hers
‘I get more involved for my boss than does Martine for hers.’
b.
[O]n fait plus que ne légitime la déduction jusqu’ici. one does more than en justifies the deduction up.to.here ‘One does more than deduction justifies up to here.’ (D. Lories in Kant’s Ästhetik/L’ esthétique de Kant, Berlin: de Gruyter 1997, p. 575)
(John Vanderelst, p.c.)
. The adverb test is one standard heuristic (Pollock 1989). Rowlett goes further for the inversion types he analyses and assumes that the EPP feature of T is never satisfied. We may disagree with this for CI. CI derivations have non-inverted counterparts. A second note is also in order regarding adverbs. Other adverbs are also quite possible, as illustrated in (i) below. But, at least on certain syntactic accounts in the wake of Cinque (1999), they may be situated higher up and thus do not necessarily delimit the low boundary. Such adverbs would, then, be less clear as evidence for low subjects. Finally, their precise syntax is at the same time anything but clear (the invoked high position may, e.g. be a matter of LF). (i)
[L]e même Jacques Peyrat avait fait 2% de plus the same Jacques Peyrat had made 2% of more
[que ne fait aujourd’hui Christian Estrosi]. than en makes today Christian Estrosi
‘The same Jacques Peyrat has made 2% more than C.E. makes today.’ (http://boymottard.wordpress.com/page/19/)
Remus Gergel
Third, quantifier float of the type discussed becomes entirely incomprehensible in CI (as expected if CI patterns with low-subject inversion): (27) *Pierre a fait ses devoirs plus vite [que tous ont fini Pierre has done his homeworks more fast than all have finished les enfants]. the children Intended: ‘P. did his homework faster than all the children have finished theirs.’
Fourth, we can note that subject-less impersonals are in some cases (we assume restrictedly) possible (i.e., e.g., without the pronoun il): (28) un système qui a progressé vers un état d’ équilibre a system that has progressed toward a state of.balance durant une période de temps beaucoup plus longue during a period of time much more long que ne serait possible en système synthétique. than en would.be possible in system synthetic ‘a system that equilibrated for a much longer period of time than would be possible in synthetic systems.’ (http://pubs.nrc-cnrc.gc.ca/mineral/tcm-88939-3.html) (29) Il sera plus difficile d’ accroître ce pourcentage it will.be more difficult to.increase this percentage que n’ est le cas pour Genève… than en.is the case for Geneva ‘It will be more difficult to increase this percentage than is the case for Geneva.’ (http://www.litra.ch/Trains_trams_et_bus_offrent-ils_une_capacite_suffisante.html)
Overall then, these tests show that CI patterns with what Rowlett calls genuine stylistic inversion, that is, the type specifically based on low subjects. In the next section, we will focus on the main characteristics of CI in earlier French.
3.2 CI in earlier French From a diachronic vantage point, we can make two crucial supporting observations for the current analysis of CI right away. First, earlier stages of French had the possibility of a subject appearing in a low position (see especially Vance 1997 and Kroch 2008 for recent discussions and the references cited there). Second, we can empirically add that OF French had CI as well. This is shown in (30), with a more comparative, and in (31) with a comparative introduced by the adverb isnels, ‘nimbly, agilely’.
Towards notions of comparative continuity in English and French
(30) Plus curt a piet que ne fait un cheval. more runs by foot than en does a horse ‘He runs more than a horse does.’ (31) Plus est isnels que nen est uns falcuns. more is agile than en is a falcon ‘He is more agile than a falcon is.’
(CdR, 71.890)
(CdR, 118, 1572)
Third, examples from Middle and Classical French support the idea of a continuous phenomenon; compare the inversions in the comparatives in (32) and (33), respectively. (32) … plus rouge que n’est feu gregois. more red than en-is fire Greek ‘…redder than is Greek fire.’ (CLM corpus; Mystère de la Passion (ca. 1450)) (33) J’ay plus failly que ne peut un mortel. I-have more erred than en can a mortal ‘I erred more than a mortal can.’ (Frantext: Q757/VIAU Th. de/OEuvres poétiques: 1/1621, p. 113)
A fourth supporting argument is that subjectless impersonals are a time-honoured presence in the language, including some comparatives, as the OF (34) illustrates. (34) Il le dublat plus que ne solt. he it doubled more than en uses ‘He doubled it more than is usual.’
(BR, 1. 765)
Fifth, verb sequences containing more than just the finite verb can occasionally be observed in the records of earlier French as well. We note that they can display the same basic surface word order as do the modern examples. I assume that the subject is in the specifier position of the lowest (and theta-assigning) predicate phrase. (35) Guillaumes fu joianz et liez,[…] Guillaumes was happy and content Plus que ne puet dire nus hom. more than en can tell us man ‘Guillaumes was happier and more content than anyone could tell.’ (CLM corpus; De Guillaume au faucon)
Finally, the test based on verbal dependents after the subject (cf. 3.1 above) can also be applied to some extent to the diachronic record, see (36). The result of the test suggests once more continuity, namely in the availability of the low-subject position.
Remus Gergel
(36) Li palefrois sor quoi ele est montee the palfrey on which she is climbed Estoit plus blans que n’ est nois sor gelee; is more white than en-is snow on ice ‘The palfrey on which she climbed is whiter than snow on ice.’ (CLM corpus; 8145, Enfances Ogier – Œuvres d’Adenet le Roi)
Overall there are reasonable grounds for a low-subject derivation of CI. While I have no quantitative results on the developments in French, it is possible that with the advent of annotated corpora such as the upcoming Les voies du français (see Kroch 2008 for discussion), more will be learnt about the trajectory of the construction including numerical developments and what they may reveal about grammar.
4. Continuity in the syntax of CI We have so far illustrated the continuous presence of CI in the historical record. This section is concerned with a selection of facts which potentially complicate the picture. We concentrate on English, due to the data situation at present (in particular the wider availability of parsed historical corpora), but also raise certain questions in connection with French. On the basis of the Penn-Helsinki-York corpora of historical English, estimates as in (37) can be given for the ratio of inverted tokens within structures that qualified as clausal comparatives from the corpus annotation (cf. Gergel 2008). (37) OE: 223/5114 = 4.36%; ME: 135/1639 = 8.23%; EModE: 31/2497 = 1.24%
The estimates almost double in ME (compared to OE) and then drop dramatically towards EModE. The fall in EModE is straightforwardly explained by the erosion of the low subject position (which remains an archaic possibility in comparatives, as we have seen). However, the surge in ME needs an explanation. There is more than one potential source why the incidence of CI might have increased in ME regarding potential intervening factors. The main possibilities to be discussed are listed in (38). (38) a. b. c.
contact with Northern syntactic features in ME the influence of French independent factors relating to change/continuity in the diachrony of CI
We first investigate whether the incidence of CI in the records was influenced by factors that have been claimed to have a major influence on the loss of V2 in English. One such factor is given by Northern features possibly stemming from contact with Germanic languages that had a more regular C-based V2 than OE (see Fischer et al. 2000; van Kemenade 1997; Kroch et al. 2000; Roberts 2007, among others for
Towards notions of comparative continuity in English and French
discussions). But importantly, for our immediate purposes there are several points on which the development of CI does not go together with that of the progress of such Northern dialectal features and, as we will argue, cannot be explained on the basis of this type of contact. To begin, subject pronouns do start to invert during ME in contexts in which they did not invert in OE. But this increase cannot be plausibly explained by a putative increase in inverting pronouns; that is, neither in terms of its timing nor the quantity of change. For one, other major changes in the patterns of inversion between finite verb and subject become visible in the estimates from the period M2 onwards. But the current increase is from M1, the first Helsinki subdivision of ME. In addition, even if we abstract away from timing, the number of actual inverting pronouns in ME still remains particularly limited and cannot explain the surge in the CI estimates. Furthermore, we have another related negative result from an independent diagnostic, namely from a consideration of the incidence of CI in texts of Northern provenience in the PPCME2 corpus in particular. Out of the 6 Northern texts in the corpus just one contains CI. The thoroughly Northern-based Rule of St. Benet, for example, does not contain inversions in comparatives. The Northern version of the Mirror of St. Edmund is the sole text from this sample to contain (two) tokens of CI. The Northernbased ratio of texts thus stands in a stark contrast with the overall ratio of texts containing CI constructions in ME in general and in particular with the texts that are not classified as being from the North. For example, the average within the remaining 50 texts in the PPCME2 is as follows: 35 texts have CI, that is, as much as 70% (compared to the 16.6% above). Moreover, the overall total rate of CI per relevant tokens in the Northern texts is, at 1.35%, far below the overall average of 8.23 % for ME. The situation, then, does not yield support for a rise of CI due to Northern features. (If anything, a slowing down factor might be given). Clearly, this does not mean that interesting comparisons cannot be drawn between features of, say, Old Norse (which arguably influenced the Northern V2) and of ME including comparatives13 (or much less that more intricate scenarios than the ones presented are not possible). It only means that
. The Example (i) from the Laxdœla Saga illustrates a structure resembling CI with a dual pronoun in Old Norse (from Faarlund 2004: 267; (54c)): (i)
allt er mér sl’kt it sama nú í hug, all.neu.n is me.d such.neu.n the same.def.neu.n now in mind.d
sem þá rœddum vit as then talked.1p we.du
‘My thoughts are still the same regarding everything as we talked about then.’ [(Laxd 203.21)]
Remus Gergel
the Northern features cannot be blamed directly for the increase in CI that we witness in ME. We next turn to the influence of French. Parallels between the syntactic histories of English and French can be, and certainly have been, drawn (Adams 1987; Hulk & van Kemenade 1995; Kroch 2008; Roberts 1993, among others). An additional possibility arises, however, when we consider the potential of an indirect grammatical influence on the grammar reflected in the ME texts. Considering such texts in respect to CI, there is an observable contrast to the influence of Northern texts that we addressed above. The influence of French cannot be refuted the way that of Scandinavian has been. First, we have already said that this language had a productive process of CI at all times. Second, and perhaps more importantly, the corpus texts that are translations from French show the construction as well. We next focus on the texts of the PPCME2 that are translations from French to check their incidence of CI. The summary in (39) below indicates the ratios of CI per relevant clausal comparative tokens for the six texts that were French translations (using the standard corpus tags for the texts): (39) Aelr4 2/19; Ayenbi 10/85; Meli 2/64; Pars 6/67; Mandev 7/71; Vices 1/16.
The average frequency of CI in the six texts thus lies at 28/322 or 8.69%. This is notably more than six times the average in the Northern texts. A third asymmetry between the French-based and the Northern-based texts is that each individual text that is a French translation has occurrences of CI structures. (Compare the only 16.6%, or one in six of the texts, in the group of Northern texts). But, alas, all of this does not show that there was a French influence on the development of CI in English. All it does is show that it is not possible to dismiss (indirect) influence on the language. There are in fact some arguments that make a positive argument for a French import implausible. First, the construction was already available in the OE of Beowulf (cf. (14)), thus preceding French influence. Second, even though the frequency of CI in the French-based texts is much higher compared to the Northern texts, it is still around the ME average calculated over all texts given above. That means that a possible reinforcement through French might have taken place at most. Third, as was the case with the potential Northern influence, the surge in the estimates occurs too early, namely right from the very first period of ME. Fourth, even in a text such as the Ayenbite of Inwyt dated from 1340, which is a fairly literal translation from the French La somme le roi, we do not get a one-to-one mapping import of French in the translation. To illustrate this point, (40) preserves the structure from the French original, as expected, but (41) does not. In (41), the English (close) translation has inversion without inversion in the French original. (40) a.
Middle French:
Il i a son droit et sa soustenance, et tout comme il en he there has his right and his sustainment and all as he it of.it
Towards notions of comparative continuity in English and French
veut avoir, plus soufisemment [que n’a li rois]. wants have more sufficiently than en.has the king
‘He has his right and support and everything as he wants it, in more abundance than the king does.’ (LsR, p. 189, Ch. 47, §123) b.
Middle English:
Þer is his wone and his sustinonce and al þet he wyle there is his habit and his sustainment and all that he wants
habbe more ynogh have more sufficiently
[þanne habbe þe kyng]. than has the king
(41) a.
Middle French:
Vertu fet l’omme plus a.droit seigneur du monde virtue makes the man more rightfully lord of.the world
[que li rois n’est de son roiaume]. than the king en-is of his kingdom
(CMAYENBI, 85.1650)
‘Virtue makes man more legitimately lord of the world than the king is of his kingdom.’ (LsR, p. 189, Ch. 47-§120) b.
Middle English:
Virtue makeþ þane man more aright lhord of þe wordle: virtue makes the man more rightly lord of the world
þanne by þe kyng of his regne. than is the king of his kingdom
(CMAYENBI, 85.1648)
The fact that even in a literal translation we can detect home-grown instances of English inversion makes French, even as a source of re-import, then, unlikely. A more likely possibility (consistent with our proposal) is that the construction was due to a structural possibility given in the grammar of both languages, and which found propitious ground in the ME period due to a variety of factors in the output. But the question, then, of course remains what might have chiefly caused the surge in ME. An issue that has had a direct impact on the appearance of CI on the surface strings of the texts is an independent syntactic development, namely the headedness of TP. Since a head-final TP obscures CI under the low-subject proposal defended here, the doubling of the estimates in ME as compared to OE can be explained as follows. First, notice that at the descriptive level OE had both T-final and T-initial word orders. (Cf. Fischer et al. 2000; van Kemenade 1997; Kroch et al. 2000, among others. For an interesting theoretical account of the transition in headedness, see Biberauer, Sheehan & Newton this volume). Second, we use the independent assumption that roughly half of the relevant clauses of OE are head-final (cf. Pintzuk 1991 for comprehensive discussion of such findings). Next, we can note that for a T-based inversion to be
Remus Gergel
noticeable in the first place, the TP must have been T-initial. (In a head-final TP, the finite element may move to T, but still stays to the right of the subject. Hence no inversion is detectable.) That means that half of the structures that had the mechanics for CI in place in OE were obscured in the output. But by early ME T-final structures are already dead or only archaic (Kroch et al. 2000). This, then, yields a possible explanation for the doubling of the estimates in the CI output seen in the corpora. To summarize, we have observed that a major point of change was the directionality of the TP by the end of OE; pace, for example, possible textual influence from French. A syntactic suggestion has been made as to why this change may have affected, as a key factor, the ratio of CI in the output. Notably for the case in point, the low subject became more prominent at the stage at which English had only head-initial TPs. This strengthens the main proposal that the main effect is best explained by continuity in the sense that the low subject position was heavily used in conjunction with T, both before and after the change in directionality of headedness.
5. Open issues and conclusions We next discuss remaining questions, grouped around two main issues, before concluding. First, we address V-to-T. We then re-address continuity in comparatives. As noted by a reviewer, a non-trivial factor in the evolution of any process that makes reference to the inflectional domain is the trajectory of V-to-T. While it is beyond the scope of this chapter to elucidate the notorious controversies surrounding this issue in its general aspects (Kroch 1989; Pollock 1989; Roberts 1993; Warner 1997, among others), we offer a note from the perspective of comparative structures. In this connection, it is crucial to distinguish between two layers of potential interference. First, notice that whether French or English has V-to-T does not affect the underlying skeleton of the proposal. But putting the classical parametric difference (both between English and French and between the developmental stages of English; cf. Kroch 1989; Pollock 1989; Roberts 1993, among others) together with the current proposal makes a valid prediction. Namely (just) that CI involves a finite element under T. Relating to it, however, the proposal itself, as stated, does not (need to) make any reference as to which elements are under T. The important second point is that the parametric difference entirely takes care of this aspect. The parametric difference, then, does not change the phrase-structure of CI as such but it interacts with it in the output produced. There is a clear generative difference dependent upon whether CI interacts with a grammar with or without V-to-T. The predicted output difference is that, all things being equal, the former grammar will have the means to feed lexical verbs to be (re-)merged under the node T. (Simply put: lexical verbs will be able to participate in CI in conjunction with the low subject). The latter grammar, however, will in general lack the means to do so. This seems to be the case for Present Day English (PDE), which is forced
Towards notions of comparative continuity in English and French
to apply do-support with lexical verbs. CI in PDE thus only occurs with specialized T-elements (modals/do) and have/be, as a reviewer observes. French, expectedly, also allows lexical verbs in CI.14 We now turn to our second open question, namely why a construction such as CI might have remained in place so persistently while (almost) everything else around it changed. We tentatively suggest two types of reasons as main necessary conditions. The first major reason we suggest (and hope to have convinced the reader of through the evidence presented) is that CI has been naturally riding on a syntactic option (namely low subjects) that was historically widely available in both English and French. Today, this earlier phrase-structural option remains available in areas of the grammar in the two languages which are typically stylistically marked. While the specific syntactic reasons why such representations should be plausible have been explained in some detail, there is a second type of motivation with the potential to affect comparative clauses: the interface with the component of meaning. We just mention three points in this connection. First, one additional heuristic reason for stability may be the parallelism relationship given through the very nature of comparative constructions generally (cf. Gergel et al. 2007). A second, perhaps more crucial reason is that the core semantic representation of comparatives may not have changed significantly in either English or French. Following tests discussed in synchronic terms in Beck et al. (to appear) for a series of largely unrelated languages, it becomes possible to test for parameters at the syntax– semantics interface in the history of English, too. In this connection, OE and PDE seem to have the same semantics of the comparative in place.15 The final point that becomes
. As pointed out by the reviewer, there are several other (important) syntactic changes in both languages. But they did not seem to affect comparatives. One reason why the changes in the status of pronoun subjects in French might have not affected the output is that pronouns are particularly hard to find even with non-clitic forms (both in the records and with the new full forms in current French). E.g., *Pierre est plus grand que n’était lui (on a pronominalized (15a)) is not acceptable. A possible connection is that information structure interacts with syntax (cf., e.g., van Kemenade 2008 for OE). . The question of whether the same semantic parameters are operative over time is certainly not trivial. Since motivating a theoretical framework for this falls beyond present scope, I refer the interested reader to Beck et al. (to appear) for the background on possible syntactic and semantic options. Empirically, parametric (in)variation can then be tested through various degree constructions. To mention but one, let us take measure phrases. OE had measure phrases (e.g. Gergel 2008), largely like PDE. The case of French is more intricate and deserves special attention in future research. But we can illustrate the essentials. While the ancestor language Latin had measure phrases (ducentos pedes altus, ‘two hundred feet high’), such constructions were lost after the Old French period. However, what French seems to have exploited is a rescue strategy. Measure phrases are thus still available, introduced by the functional word de as in haut de 2m, ‘2m tall’.
Remus Gergel
conspicuous in the same connection of semantic variation in French is the need to study the role of expletive negation even more closely than a rich tradition has already done. While the negation morpheme ne does not have negative meaning in current French comparatives (cf. Price 1990; von Stechow 1984), it is a relic from a time when ne alone could potentially be the negative marker by itself. One major task for future research, then, remains to explain how the negative form came to essentially mark the clausal status of comparatives. (Phrasal comparatives lack ne). This fact may perhaps be interpreted as one option that effectively helps keep the semantic settings constant. As is well known, it is not a universal for a language to have clausal comparatives. To conclude, the chapter has raised several questions and purports to have provided a few answers relating to comparative clauses. It has primarily illustrated a case of simple, but thus far not systematically addressed, continuity in English and French by discussing comparative inversion. We have drawn on the history of English and largely extended the analysis to French, a language that in its diachronic evolution is independently also well known for having low subjects. By using knowledge from diachronic research on English and French, current syntactic diagnostics, and by having partly investigated additional factors such as language contact, we have reasoned that there is a case for a simple (and as described, identical) phrase-structural continuity based on low subjects. This may have been co-supported, though from the current evidence not caused, by language contact. Several questions have been left open here, in particular so with regard to the interface with the component of meaning.
Major corpora and primary textual sources used in this study Beowulf: A student edition. Jack, George (ed.). 1994. Oxford: Oxford University Press. BR: Le voyage de Saint Brendan. Short, Ian & Merrilees, Brian (eds.). 2006. Paris: Ed. Honoré Champion. CdR: Chanson de Roland. Steinsieck, Wolfgang (ed.). 1999. Das altfranzösische Rolandslied. Stuttgart: Reclam. CLM: Corpus de la littérature médiévale. Éditions Champion Électronique. Available online at: http://www.champion-electronique.net/bases. FM: G. de Maupassant. 1923. Fort comme la mort, Éditions du Boucher. FRANTEXT: Base textuelle Frantext. Nancy: Centre national de la recherche scientifique. Available online at: http://www. frantext.fr. LsR: La somme le roi. Édith Brayer and Anne-Françoise Labie-Leurquin (eds.). 2008. Paris – Abbeville, Société des anciens textes français – Paillart. PPCEME: Kroch, A., Santorini, B. & Delfs, L. 2004. The Penn-Helsinki Parsed Corpus of Early Modern English. PPCME2: Kroch, A. and Taylor, A. 2000. The Penn-Helsinki Parsed Corpus of Middle English. YCOE Taylor, A., Warner, A., Pintzuk, S. and Beths, F. 2003. The York-Toronto-Helsinki Parsed Corpus of Old English Prose.
Towards notions of comparative continuity in English and French
References Adams, Marianne P. 1987. From Old French to the theory of pro-drop. Natural Language and Linguistic Theory 5: 1–32. Aissen, Judith. 1975. Presentational there-insertion. CLS 11: 1–14. Beck, Sigrid, Krasikova, Sveta, Fleischer Daniel, Gergel, Remus, Hofstetter, Stefan, Savelsberg, Christiane, Vanderelst, John & Villalta, Elisabeth. To appear. Cross-linguistic variation in comparison constructions. In The Linguistic Variation Yearbook. Besten, Hans den. 1978. On the presence and absence of wh-elements in Dutch comparatives. Linguistic Inquiry 9: 641–671. Besten, Hans den. 1983. On the interaction of root transformations and lexical deletive rules. In On the Formal Syntax of the Westgermania. Papers from the 3 Groningen Grammar Talks Groningen, January 1981. Linguistik Aktuell/Linguistics Today 3], Werner Abraham (ed.), 47–131. Amsterdam: John Benjamins. Borroff, Marianne L. 2006. Degree phrase inversion in the scope of negation. Linguistic Inquiry 37: 514–521. Chomsky, Noam. 1977. On wh-movement. In Formal Syntax, Peter Culicover, Thomas Wasow & Adrian Akmajian (eds), 71–132. New York NY: Academic Press. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford: OUP. Culicover, Peter W. & Winkler, Susanne. 2008. English focus inversion. Journal of Linguistics 44: 625–658. Davison Alice. 1979. Some mysteries of subordination. Studies in the Linguistic Sciences 9: 105–128. Doetjes, Jenny. 1992. Rightward floating quantifiers float to the left. The Linguistic Review 9: 313–332. Faarlund, Jan T. 2004. The Syntax of Old Norse. Oxford: OUP. Fischer, Olga, van Kemenade, Ans, Koopman, Willem & van der Wurff. 2000. The Syntax of Early English. Cambridge: CUP. Gergel, Remus. 2008. Comparative inversion: A diachronic study. Journal of Comparative Germanic Linguistics 11: 191–211. Gergel, Remus, Gengel, Kirsten & Winkler, Susanne. 2007. Ellipsis and inversion: A featurebased focus account. In On Information Structure, Meaning and Form. Generalizations across Languages [Linguistik Aktuell/Linguistics Today 100] Kerstin Schwabe & Susanne Winkler (eds), 301–322. Amsterdam: John Benjamins. Green, Georgia. 1976. Main clause phenomena in subordinate clauses. Language 52: 382–397. Haeberli, Eric. 2002. Inflectional morphology and the loss of verb second in English. In Syntactic Effects of Morphological Change, David Lightfoot (ed.), 88–106. Oxford: OUP. Hankamer, Jorge. 1973. Why there are two ‘than’s in English. CLS 9: 179–191. Hartmann, Jutta. M. 2005. Why there is(n’t) wh-movement in there constructions. In Linguistics in the Netherlands 2005, Jenny Doetjes & Jeroen van de Weijer (eds), 87–99. Amsterdam: John Benjamins. Huddleston, Rodney & Pullum, Geoffrey. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Hulk, Aafke & van Kemenade, Ans. 1995. Verb second, pro-drop, functional projections and language change. In Clause Structure and Language Change, Adrian Battye & Ian Roberts (eds), 227–256. Oxford: OUP.
Remus Gergel Iatridou, Sabine & Kroch, Anthony. 1992. The licensing of CP-recursion and its relevance to the Germanic verb-second phenomenon. Working Papers in Scandinavian Syntax 50: 1–24. Kayne, Richard & Pollock, Jean-Yves. 2008. Toward an analysis of French hyper-complex inversion. Ms, New York University/Université Paris-Est Marne l.V. van Kemenade, Ans. 1997. V2 and embedded topicalisation in Old and Middle English. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds.), 326–352. Cambridge: CUP. van Kemenade, Ans. 2008. The interaction of syntax and information structure in Old and Middle English. Paper presented at the 10th Diachronic Generative Syntax Conference, Cornell University. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Kroch, Anthony. 2008. V2 loss revisited: A quantitative comparison of English and French. Paper presented at the 10th Diachronic Generative Syntax Conference, Cornell University. Kroch, Anthony, Taylor, Ann & Ringe, Donald. 2000. The Middle English verb-second constraint: A case study in language contact and language change. In Textual Parameters in Older Language [Current Issues in Linguistic Theory 195], Susan C. Herring, Pieter van Reenen & Lene Schøsler (eds), 353–391. Amsterdam: John Benjamins. Lahousse, Karen. 2006. NP-subject inversion in French: Two types, two configurations. Lingua 16: 424–461. Lasnik, Howard & Sobin, Nicholas. 2000. The who/whom puzzle: On the preservation of an archaic feature. Natural Language and Linguistic Theory 18: 343–371. Merchant, Jason. 2003. Subject-auxiliary inversion in comparatives and PF output constraints. In The Interfaces: Deriving and Interpreting Omitted Structures [Linguistik Aktuell/ Linguistics Today 61] Kerstin Schwabe & Susanne Winkler (eds), 55–77. Amsterdam: John Benjamins. Nevalainen, Terttu. 2006. An Introduction to Early Modern English. Edinburgh: EUP. Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and Change in Old English Word Order. Ph.D. dissertation, University of Pennsylvania. Pollock, Jean-Yves. 1989. Verb movement, universal grammar and the structure of IP. Linguistic Inquiry 20: 365–424. Potts, Christopher. 2002. The syntax and semantics of as-parentheticals. Natural Language and Linguistic Theory 20: 623–689. Price, Susan. 1990. Comparative constructions in Spanish and French Syntax. London: Routledge. Rizzi, Luigi & Roberts Ian G. 1989. Complex inversion in French. Probus 1: 1–30. Roberts, Ian G. 1993. Verbs and Diachronic Syntax. A Comparative History of English and French. Dordrecht: Kluwer. Roberts, Ian G. 2007. Diachronic Syntax. Oxford: OUP. Rowlett, Paul. 2007. The Syntax of French. Cambridge: CUP. Speyer, Augustin. 2008. On the Interaction of Prosody and Syntax in the History of English, with a Few Remarks on German. Ph.D. dissertation, University of Pennsylvania. Stechow, Arnim von. 1984. Comparing semantic theories of comparison. Journal of Semantics 3: 1–79. Vance, Barbara S. 1997. Syntactic Change in Medieval French. Dordrecht: Kluwer.
Towards notions of comparative continuity in English and French
Warner, Anthony. 1997. The structure of parametric change, and V-movement in the history of English. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds), 380–393. Cambridge: CUP. Williams, Alexander. 2000. Null subjects in Middle English existentials. In Diachronic Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 164–190. Oxford: OUP. Winkler, Susanne. 2005. Ellipsis and Focus in Generative Grammar. Berlin: de Gruyter.
Variation, continuity and contact in Middle Norwegian and Middle Low German John D. Sundquist Purdue University
This study focuses on the period of language contact between Middle Norwegian (MNw) and Middle Low German (MLG) in Norway between 1300 and 1525. Using corpora of MLG and MNw diplomatic letters written during this time, we examine the possibility that syntactic continuity in the form of retention of archaic forms may take place when two typologically similar languages are in prolonged contact with each other. Data on the variation between Object–Verb (OV) and Verb–Object (VO) order in both MNw and MLG indicate that the frequency of OV word order in MNw is not affected by the influx of variational patterns from MLG. We explore possible reasons for this lack of change and focus on the role of syntactic variation in both continuity and change. Drawing our conclusions from evidence on OV/VO variation, we argue that the syntactic variation in the MNw data disrupts continuity rather than helping to maintain it.
1. Introduction 1.1 Domain of inquiry In recent research on diachronic syntax, there has been increasing interest in examining the apparent absence of syntactic change – or continuity – in certain contexts in order to come closer to understanding the cause of change in other contexts. Interesting data emerge when we compare instances of change in one language, such as contactinduced borrowings or cyclical grammatical changes like Jespersen’s cycle, with stages of continuity in a typologically similar language. Since Keenan’s (2002) description of inertia in language change and Longobardi’s (2001) modifications of an earlier version of this concept, the notion of continuity in historical syntax has been cast in a new light. According to Keenan, “things stay the way they are unless acted on by an outside force or decay” (2002: 327). Following this line of argumentation, Longobardi focuses more specifically on syntactic change and proposes a reformulation of Keenan’s more general point: “syntactic change should not arise, unless it can be shown to be caused” (2001: 278). In other words, syntactic change is the exception to the rule: syntax
John D. Sundquist
remains stable unless such forces as language contact or morphological or phonological decay exert significant pressure and cause it to change. To examine this other side of the coin, we might ask what conditions are appropriate for continuity to be maintained. Following the logic of Longobardi’s principle of inertia, we can assume that syntactic continuity exists in the absence of an outside force. However, as this chapter will discuss, the equation is not so simple when we factor in syntactic variation. In the following case study on word order in Middle Norwegian in the fourteenth and fifteenth centuries, we ask whether variation acts to bring about change in a period of language contact with Middle Low German or whether it helps to maintain continuity. The latter possibility has been discussed in terms of ‘retention’, or as Braunmüller (1998: 329) calls it, ‘Driftumkehrung’ – a type of gradual reversion, in which an outgoing, archaic variant is retained or reactivated during periods of language contact with a typologically similar language. Contact in such situations would have the opposite effect to what Longobardi (2001) predicts: it might introduce more variation and strengthen certain outgoing patterns of usage.1 The main focus of this study is on the possibility of syntactic retention during a period of language contact between Middle Low German (MLG) and Middle Norwegian (MNw) in Norway. After assembling a sampling of texts from the fourteenth and fifteenth centuries, I first ask to what extent Norwegian word order remained unchanged when these two typologically similar languages were in contact with each other over several hundred years. Secondly, I ask what role variation plays in bringing about change or contributing to continuity in Norwegian. I focus on the word order patterns [Object–Verbmain] (OV) and [Verbmain–Object] (VO) in Norwegian and German corpora of diplomatic letters written between 1300 and 1500.2 The crucial question in the study is the extent to which the influx of variational patterns of OV word order from Middle Low German speakers affects the OV order that already existed but was dying out in Norwegian. The empirical portion of the chapter will present the general frequencies and patterns of variation in OV and VO orders in both languages, as represented in the two corpora. The discussion section will examine views on the role of syntactic variation in periods of contact and provide support for the argument that variation appears to disrupt continuity rather than help to maintain it.
. Thomason and Kaufman (1988: 58) also describe this phenomenon, referring to it as retention: “…interference through shift in particular may even be responsible sometimes for lack of change. This means that, since retention as well as innovation may be externally motivated, the presence of inherited features is not always adequately explained once one determines their genetic origin”. . For simplicity, I follow the conventional term “OV” to refer to various kinds of PP- and NPcomplements to the non-finite main verb.
Variation, continuity and contact in Middle Norwegian and Middle Low German
1.2 Corpus The Middle Norwegian (MNw) and Middle Low German (MLG) corpora that I used in this study are both collections of diplomatic letters. Although this type of text exhibits written language that is less representative of the spoken vernacular, I found it important to have a homogenous text type throughout each corpus as well as across the two collections.3 Moreover, the letters are useful when examining issues of language contact and syntactic change over time, since the provenance and date of composition are usually identifiable. For the Middle Norwegian corpus, I selected 300 charters (approximately 90,000 total words) from Diplomatarium Norvegicum written between 1300 and 1525. Because of the conservative, formulaic nature of these letters, I follow precautions suggested by Mørck (1980) and exclude archaisms or copied phrases that might skew the data.4 I selected 1300 as the starting point for the corpus, since many of the diplomatic letters prior to that date are in Latin, and I chose 1525 as the end point, since many of the letters written at that time are almost exclusively in Danish. Although Bergen is the city with the highest number of charters included in the corpus, letters from other major scribal centers like Trondheim and Stavanger were also included.5 The Middle Low German corpus consists of 200 diplomatic letters (c. 75,000 words) from Codex Diplomaticus Lubecensis (or Lübeckisches Urkundenbuch) written between 1325 and 1500. Because there are fewer letters written before 1325 in Lübeck in the vernacular and most of the significant contact between Germans and Norwegians began after this point, I used 1325 as the starting point for the Middle Low German corpus.6 I did not include texts after 1500 because of the influence of High
. Naturally, when analysing linguistic elements in these texts with a more conservative style, we must assume that there is a larger gap between the spoken and written language than would be the case in more informal texts with a colloquial style. However, it is the aim of this study to focus on general patterns in the data that would indicate larger trends of linguistic variation and change. In addition, the diplomatic letters constitute a less-than-ideal corpus for the study of unmarked word order, since their conservative style and structure exhibit strong influence from Latin. However, as will be discussed in this section, the consistency of text type both throughout and between the two corpora was important for this study. . See Sundquist (2002, 2006) for further discussion of the MNw corpus and the precautions that are necessary when analysing these texts. . Unlike other studies on MNw–MLG contact like Brattegard (1963), Jahr (1999) or Nesse (2002), which focus solely on the Bergen dialect, the current study is concerned more with MLG influence on Norwegian in general. Although it would be preferable to limit the study to letters written in Bergen, where the number of German merchants during this time was highest, there would be too few examples in this more limited corpus for a detailed quantitative analysis. . For a helpful overview of the most important dates and general historical information on Hanseatic merchants in Norway with particular emphasis on Bergen, see Gade (1951), Brattegard (1963), Helle (1994) and Nesse (2002).
John D. Sundquist
German that took hold in early part of the sixteenth century in Northern Germany and Scandinavia (Braunmüller 2004). I limited the MLG corpus to texts written in Lübeck by authors of that city in order to maintain consistency with respect to geographical region and to focus on the most dominant Low German variety that would have been used in Scandinavia in the period of the most intense contact in the fourteenth and fifteenth centuries.7
2. Quantitative analysis: OV and VO word order in Middle Norwegian and Middle Low German In the analysis of variation between OV and VO word order in Middle Norwegian and Middle Low German, I included any main or embedded clause in the corpora that contains both a finite and a non-finite verb followed by two or more non-verbal constituents. I also limited examples to those that contain a verbal complement – either an NP- or a PP-complement to the non-finite verb. In Norwegian, the unmarked order in such examples is VO, as in (1) and (2):8 (1) at hann hafðe suaret þænna sama vithnisburð aðr that he had sworn that same testimony before ‘that he had sworn that same testimony before.’ (DN 1:122; 1309 Bergen) (2) stod sa Haluard vp och vildæ gangæ at dørnen stood so Haluard up and wanted go to door-the ‘Haluard stood up and wanted to go to the door.’ (DN 1:947; 1486 Skien)
However, as we will see in the discussion of the results below, Middle Norwegian exhibits variation between the pattern in (1) and (2) and the archaic OV order in examples like (3) and (4): (3) skall hon þetta goz hafua shall he those wares have ‘He shall have those wares.’
(DN 2:120; 1314 Bergen)
. Note that I included letters written by a wide variety of authors from Lübeck, but I did not limit myself to those written by merchants or secretaries who had been to Norway. The purpose of using this larger, less exclusive corpus for this study is to obtain an approximate sampling of the language of this time and observe general norms, trends, and patterns of variation. Future research on contact might focus on only those MLG letters written by individuals who spent time in Norway and came into contact with Norwegians, as in Simensen’s (1989) study of some of the MLG letters in Diplomatarium Norvegicum. . MNw examples from Diplomatarium Norvegicum (DN) are cited with the volume number, letter number, year of composition, and provenance.
Variation, continuity and contact in Middle Norwegian and Middle Low German
(4) at han vildhe siælfuer til erkibiskupenn koma that he wanted himself to archbishopric-the come ‘that he himself wanted to come to the archbishopric.’ (DN 1:713; 1426 Bergen)
Following Sundquist (2002, 2006), we can assume that OV word order in MNw is inherited from Old Norse, as evidenced in earlier Norwegian manuscripts from the thirteenth century.9 I follow the same general procedures in selecting examples for analysis from the MLG corpus. MLG exhibits the opposite trend of MNw: OV order, as in (5) and (6), is the norm, while the VO order in examples like (7) and (8) is marked.10 (5) dar schal he sin teken vp setten there shall he his sign up set ‘There he shall set up his sign.’
(LU 1.8:DCCXXX; 1450 Lübeck)
(6) dat sodane vorscreuene schip vnd gudere in vnse hauene that that aforementioned ship and goods into our harbours komen mogen come can ‘so that the aforementioned ship and goods can come into our harbours.’ (LU 1.7:XLV; 1427 Lübeck) (7) de rat schal beboden alle de schuldemere the council shall summon all the guilty ‘The council shall summon all the guilty ones.’ (LU 1.2:CDII; 1320 Lübeck) (8) dat wi willen scriuen an den copman van Bergen that we want write to the merchants from Bergen ‘that we want to write to the merchants from Bergen.’ (LU 1.5:CCCXVIII; 1410 Lübeck)
This state of affairs in MLG and MNw is not unexpected: MNw exhibits primarily VO word order with some instances of the older OV order, while MLG exhibits the opposite pattern.
. Sundquist (2002, 2006) provides evidence that Norwegian texts such as the Konungs skuggsjá (‘The King’s Mirror’) and Den eldre Gulatingslova (‘The Older Gulathing Law’), whose manuscripts are from the middle of the 13th century, exhibit a higher rate of OV order than the diplomatic letters from the 14th century. . Examples from Lübeckisches Urkundenbuch are listed according to section, volume, letter number, date and the location of composition.
John D. Sundquist
2.1 Variables that affect OV/VO variation In order to track the variation between OV and VO word order in both the German and Norwegian corpora, I analysed a number of variables. First of all, I examined the role of syntactic weight, or the ‘heaviness’ of the NP- or PP-complement. According to the extensive literature on post-posing of PPs and NPs outside the sentence frame in earlier stages of German, the frequency of VO order is common in coordinated or modified complements.11 Consider the following examples: (9) dat de heren van Holzsten vnde wy scholden vnse that the men from Holstein and we should our boden zenden messenger send ‘that we and the men from Holstein should send our messenger.’ (LU 1.4:CDXXXIII; 1387: Lübeck) (10) dat sie ere brieue …senden scholen deme rade vnde der stat that they their letter …send should the council and the city tho Lubeke of Lübeck ‘that they should send their letter to the council and city and Lübeck.’ (LU 1.4:XCIII; 1364: Lübeck)
In order to observe the effects of syntactic length, I differentiated between those sentences with ‘simple’ PP/NP complements like (9) from ‘complex’ PPs and NPs that are modified or coordinated, as in (10).12 Other variables that were tracked in the quantitative analysis included the type of clause (main or embedded), type of non-finite verb (infinitive or past participle), type of complement (PP, non-pronominal NP, pronominal NP), syntactic length (simple vs. modified/coordinated PP/NP) and date of composition (see Pintzuk 1999 or Sunquist 2006 for information on the variables “type of clause” and “type of complement”). Using the statistical program Goldvarb to analyse the patterns of interaction and . For an overview of research on the sentence frame and Ausklammerung in early stages of German, see Schildt (1977), who includes analysis of Low German. Ebert (1980, 1986), Betten (1980), Ebert et. al. (1993) and Bies (1996) also discuss OV order in Early New High German. For a more specific analysis that focuses on differences and similarities between MLG and early Scandinavian with respect to the sentence frame, see Braunmüller (1993b, 1998). . I follow Bies (1996) in making this distinction between simple vs. complex complements as a measure of syntactic length. In her study of Early New High German word order, she found that neither the number of syllables nor the number of words was a significant factor in determining the position of heavy elements outside the sentence frame. Instead, she found modified NPs and PPs or coordinated elements to be much more likely to occur in postposition than simple complements of the verb.
Variation, continuity and contact in Middle Norwegian and Middle Low German
variation, I used these variables as independent variables and OV word order as the dependent variable. In both the Norwegian and German corpora, the period under investigation was divided into 25-year periods. I analysed enough letters to yield a minimum of 100 tokens of OV/VO word order for each short period. I included as many examples as the chosen letters yielded. Thus, for some of the 25-year periods, I analysed fewer letters because of the large number of examples available in those letters, while for other 25-year periods, I had to include more letters in order to collect the minimum number of tokens.
2.2 Results 2.2.1 OV word order in Middle Norwegian The general results of the MNw data analysis are listed in Table 1.13 Table 1. Goldvarb results for frequency of pre-verbal complements in Middle Norwegian Variables Type of Complement Type of Clause Type of Non-finite Verb Syntactic Length Date of Composition
Variants pronominals PPs non-pronominals main embedded past participle infinitive simple NPs/PPs coordinated/ modified NPs/PPs 1300–1324 1325–1349 1350–1374 1375–1399 1400–1424 1425–1449 1450–1474 1475–1499 1500–1525
N (OV)
%OV
Probabilistic Weight
148/315 33/184 106/639 126/489 161/649 132/519 155/619 341/853
46% 17% 16% 25% 24% 25% 25% 40%
.63 .55 .24 .52 .49 .49 .52 .61
50/285 56/146 33/104 55/140 33/120 35/123 33/123 22/129 10/119 10/134
18% 38% 31% 39% 28% 28% 26% 17% 8% 7%
.38 .69 .66 .68 .59 .56 .57 .42 .23 .17
There are some important trends in the Norwegian data that are worth noting. Goldvarb selected type of complement, syntactic length and date of composition as
. The data in Table 1 include a measurement called probabilistic weight. This figure is Goldvarb’s measurement of the effect of a variable on a scale of 0.00 to 1.00. A factor weight above .50 has a favouring effect on the occurrence of the dependent variable, while a weight below this threshold has a disfavouring effect.
John D. Sundquist
variables with a significant effect on the occurrence of OV word order. Pronominal NP objects favour OV word order, PPs as a general group favour this order less, and NPs significantly disfavour OV word order. As is to be expected, coordinated and modified verb complements occur less often in OV position than the simple NPs. In terms of the general trend in OV-frequency over time, there is a period of fluctuation between 1300 and 1450 before a steady decline in frequency among all three types of complements between 1450 and 1525. Goldvarb selected type of clause and type of non-finite verb as insignificant variables that have no effect on the frequency of OV word order.
2.2.2 OV word order in Middle Low German The results for the MLG data are presented in Table 2. Table 2. Goldvarb results for frequency of pre-verbal complements in Middle Low German Variables Type of Complement Type of Clause Type of Non-finite Verb Syntactic Length Date of Composition
Variants pronominals non-pronominals PPs Embedded main past participle infinitive simple NPs/PPs coordinated/ modified NPs/PPs 1325–1349 1350–1374 1375–1399 1400–1424 1425–1449 1450–1474 1475–1499
N (OV)
%OV
Probabilistic Weight
191/196 361/457 140/222 366/434 326/441 303/368 389/507 595/707
97% 78% 63% 84% 73% 82% 76% 84%
.89 .41 .25 .57 .42 .53 .48 .71
98/168 100/154 100/125 94/132 102/128 92/105 109/127 95/104
58% 64% 80% 71% 79% 87% 85% 91%
.45 .32 .48 .39 .49 .60 .62 .71
Much like in the analysis of the Norwegian data, Goldvarb selected type of complement, syntactic length and date of composition as variables that significantly affect the occurrence of OV word order in MLG, but it also found an additional factor group, type of clause, to be significant. Pronominal NP-complements highly favour OV word order, and according to the measurement of probabilistic weight, non-pronominal NPs and PPs disfavour OV word order at a rate of .41 and .25 respectively. We see here that there is a statistically significant difference between coordinated and modified NPs and PPs in comparison to simple NPs and PPs: the coordinated or modified NPs/PPs, as expected, occur more frequently in post-position (.45) while simple complements significantly favour OV word order (.71). Unlike the Norwegian dataset, the German examples indicate that OV word order is significantly higher in embedded clauses than
Variation, continuity and contact in Middle Norwegian and Middle Low German
in main clauses, with probabilistic weights of .57 vs. .41. In terms of the general trends in frequency data, the percentage of OV word order is very high throughout the period under investigation: the rate is 64% in the first stage and increases steadily throughout the fourteenth and fifteenth centuries before it reaches 91% by 1500. Goldvarb found there to be no significant difference between sentences with a past participle and an infinitive with respect to OV word order, disregarding type of non-finite verb as a significant factor group.
3. Discussion 3.1 Retention as a result of contact The data from the two corpora in Section 2 indicate that there was no increase in the frequency of OV word order in MNw during the period under investigation. The figures in Table 1 show that the percentage frequency of OV order between 1300 and 1525 does not increase for any extended period of time – in fact, it decreases from 38% to 7% over the MNw period. During the same period, the frequency of OV order in MLG increases steadily from 64% to 91%. These findings pose problems for any proposal that suggests that contact between MLG and MNw in Norway may have contributed to continuity and the retention of archaic syntactic forms in Norwegian. In Christoffersen’s (2000: 166) discussion of the effects of MLG contact on embedded clause word order in the history of Norwegian, she argues that Low German had a type of “strengthening effect” on native Scandinavian word-order patterns that MNw inherited from Old Norse. She cites her (1993) study of the law code of Magnus Lagabøter from c. 1350 and points out the high level of OV word order in embedded clauses.14 During the time of contact with MLG in Norway, Christoffersen maintains, the similar structures in the two languages made it possible that certain word order patterns did not diminish in frequency in subsequent stages of Norwegian. In other words, MLG contact helped preserve an older, native Scandinavian tendency like OV word order that was otherwise a marked syntactic variant on its way out of use (Christoffersen 2000: 165). Braunmüller (1998: 329) also addresses this issue when he discusses the lack of syntactic borrowings from MLG into Scandinavian. He warns that what may appear to be a borrowing may actually . Christoffersen notes that 43% of embedded clauses in her (1993) study exhibit what she calls “OV” word order (Christoffersen 2000: 161). She uses this term to describe the relative order of non-finite verbs and their NP- or PP-complements, as I do in this study. However, she also includes examples of “OV” order in which an NP- or PP-complement precedes the finite verb as well. Thus, my data on OV order in this study and her data are not directly comparable. For this reason, I am suggesting merely a general comparison of her findings and conclusions with mine in this study.
John D. Sundquist
be the result of ‘Driftumkehrung’, or a gradual reversion, whereby the frequency of a remnant, archaic word order pattern might remain the same rather than decrease during contact between two similar linguistic varieties (1998: 331). He adds that such reactivation, or strengthening, of older word order patterns, may only be temporary during a period of dialect contact (1998: 332). As marked patterns become less tolerated in spoken language, syntactic norms begin to be established in the written and spoken language and the marginal patterns gradually drop out of use. The MNw data here provide no evidence of retention. If we were to assume that contact with Low German had even a slightly strengthening effect on the use of OV word order in MNw, the percentage frequency of this pattern would flatten out over much of the period under investigation or increase temporarily over the course of the fourteenth and fifteenth centuries. The figures in Table 1 provide no evidence of such a trend. Secondly, this decrease in OV frequency in MNw takes place at the same time as OV word order becomes very common in MLG. Recall that the OV-rate in MLG texts rises steadily from 64% to 91% between 1325 and 1500 with only slight decreases in some 25-year periods of the fourteenth century. We must assume that, although the written language of the German letters differs from the spoken language in many ways, it is still a reflection of the spoken language to a certain extent. Thus, there is a problem with a hypothesis which assumes that this pattern was becoming commoner in MLG and that exposure to MLG might have aided the reactivation of OV word order in Norwegian. In fact, the statistical trends for the two languages run in the opposite direction of each other: the frequency of OV word order in Norwegian is at its lowest in the fifteenth century, just when it reaches its highest level in MLG. The data on syntactic length provide further counter-evidence. Recall from Tables 1 and 2 that syntactic length was a significant variable in both the Norwegian and German data. It is clear that coordinated or modified NPs/PPs occur more frequently in post-position (VO order) than simple NPs/PPs in both languages. This result is to be expected if we consider the interaction of word order and pragmatic constraints: heavier constituents are often post-posed near the end of the sentence as a type of afterthought when such syntactic options are available. However, the data on syntactic length raise the question why Norwegian does not behave even more like MLG with respect to the placement of simple NPs. As Table 2 indicates, 84% (probabilistic weight .71) of the simple NPs/PPs in German occur in OV position. As Table 1 demonstrates, however, only 40% (.61) of the same category of verbal complements occurs in OV word order in MNw. If there were a natural, universal tendency for simple complements to occur in pre-verbal position when such options are available, and if contact with MLG had any significant effects on Norwegian word order, we would expect a much higher OV-rate among simple NPs. In other words, with all else being equal, there are relatively few departures from the unmarked VO word order in MNw – even in situations where the language naturally allows this option and where the tendency is so robust in MLG.
Variation, continuity and contact in Middle Norwegian and Middle Low German
3.2 Continuity as a result of syntactic variation The previous section demonstrated that there is little evidence to support the hypothesis that contact with MLG had an effect on the retention of OV word order structures in Norwegian. In the corpora for this study, the frequency of OV structures in MNw declines during the same period in which contact with MLG was at its strongest in Bergen and in other parts of the country.15 The commonest explanation for the lack of MLG influence on Norwegian syntax during the fourteenth and fifteenth centuries is that the sociolinguistic conditions were not appropriate for syntactic borrowing to take place. It is beyond the scope of the current study to discuss the sociolinguistic factors that may have played a role in lexical or syntactic borrowing in Norwegian. Suffice it to say that, in terms of Thomason and Kaufman’s (1988: 74) description of social and linguistic factors that lead to borrowing, the Norwegian–MLG contact situation was not ripe for much change outside of lexical borrowing.16 Although MLG merchants were in contact with Norwegians for several centuries in Norway, especially in Bergen, there was little interaction between Norwegians and Germans other than in business transactions. Germans did not intermarry with Norwegians, they lived in isolated communities, and both groups had separate laws, churches, social groups, and living quarters. As Ersland (2003: 45) points out, there is no evidence to indicate that Germans in Bergen, for instance, mingled with the local population or assimilated with locals. Thus, one can assume that the lack of syntactic influence from German is attributable, in part, to a lack of intense contact between the languages. However, another hypothesis that has been suggested is that syntactic variation played a decisive role in limiting the amount of syntactic borrowing. Following Trudgill (1994), proponents of this view focus on the unique characteristics of contact between MLG and MNw (or Mainland Scandinavian in general) as typical for a type of dialect contact rather than language contact. As Braunmüller (1998) and Christoffersen (2000) suggest, the extensive amount of syntactic variation in these
. For more background information on the sociolinguistic conditions for contact in Norway, see Jahr (1999). For specific discussion of sociolinguistic and cultural aspects of German– Norwegian contact in Bergen, see Nesse (2002) or Ersland (2003). . This conclusion is supported by the findings in Zeevaert’s (1995) corpus-based analysis of lexical borrowing in Scandinavian. Although he focuses on Swedish, the findings are relevant to any analysis which attempts to determine the level of intensity of contact between MLG and Scandinavian in general. Using Thomason and Kaufman’s (1988) classification system for intensity, he concludes that MLG–Scandinavian contact was between Levels 1 (light contact) and 2 (somewhat intense contact) (1995: 175).
John D. Sundquist
similar languages allowed speakers to use accommodation strategies in their speech.17 Braunmüller (1998) speculates that both spoken Scandinavian and MLG had relatively few strict word order norms in the fourteenth and fifteenth century. Because of the deeply rooted structural similarities between the languages, speakers on both sides of business transactions could speak their native language and understand each other. One reason for the success of this passive multidialectalism, according to Braunmüller (1998), is that speakers took advantage of the wide variety of word order patterns in their own language and used those patterns more frequently which could be easily processed by their interlocutors. As an example, he suggests the possibility that MLG speakers could accommodate Scandinavians and ease processing constraints by using VO word order when pausing in speech to post-pose elements to positions to the right of the non-finite verb (Braunmüller 1998: 325). Similarly, Scandinavians may have taken advantage of the grammatical OV option in their language more frequently in order to accommodate their MLG trading partners. More marked syntactic patterns may have been used more often as part of functionally motivated accommodation strategies. As a result of this variation, MLG exerted little influence on the word order of MNw: the languages were too similar and they exhibited too much variation for any type of long-term or even temporary impact to take place. The data in the current study, however, provide little support for this view. Quantitative analysis of MLG in Section 2 reveals that there is not as much OV/VO variation in MLG as has been previously assumed. From the beginning of the period under investigation, OV word order occurs in 61% of the examples in the Low German letters as early as 1325. This rate increases steadily before it is reaches 91% in the fifteenth century. We can assume that the lack of OV/VO variation in the diplomatic letters may be, in part, a result of these external factors and the conservative style of the text type.18 However, the general trend in OV word order in letters from Lübeck is consistent throughout the period under investigation, and the steady increase in frequency of OV word order in the written language begins before the influence of the High German standard in the sixteenth century. Although there are no strict norms in word order from this period, there are clear preferences that follow definite trends in the language as early as the fourteenth century.19
. For a more detailed discussion of this argument, see Braunmüller (1993a, 1996, 1998, 2005), Christoffersen (2000) or Mæhlum (2005). . As has been noted by Schildt (1977) for Low and High German and by Ebert (1980) for Early New High German, various language-external factors, including educational level and gender of the author, style, and geographical region, contribute to the frequency with which authors have post-posed PPs and NPs in their written language. . As in any quantitative analysis of historical syntax, we are forced to draw conclusions about the spoken language of an earlier period from our analysis of the written language.
Variation, continuity and contact in Middle Norwegian and Middle Low German
Furthermore, the statistical patterns of use of OV and VO word orders reveal a form of systematic variation rather than random fluctuation between syntactic variants. Recall that both type of complement and syntactic length were variables that had a favouring effect on the occurrence of OV word order in MLG. For instance, there is a highly significant difference between pronominal objects and prepositional phrases with respect to their tendency to occur in OV position (.89 vs. .25 probabilistic weight). These findings indicate that there is a clear underlying system in place that regulates the positioning of various complement types. Although there are various syntactic options available to MLG speakers, the choice between these options is not random and fluctuation between OV and VO order is not as free as previously assumed. In the same way, the distinction between complex and simple complements also shapes the variable use of word order patterns: coordinated and modified NPs/PPs favour OV word order at the level of .71 while simple complements disfavour it at a rate of .45. Once again, specific language-internal factors shape the patterns of OV/VO usage and demonstrate the systematic characteristics of these patterns. Thus, the Low German data here cause problems for any analysis that assumes that MLG was a language with a high degree of variation with few word-order norms. Empirical analysis of texts from this period reveals that there were clear preferences in OV/VO usage throughout the fourteenth and fifteenth centuries and that there were systematic rules that governed this usage. In terms of contact between MLG and Norwegian, therefore, it is less easy to assume from analysis of the written language that the MLG spoken language exhibited as much variation as others have previously claimed.20
3.3 Change as a result of syntactic variation Previous sections of this chapter focused on the role of syntactic variation in a period of contact between MLG and MNw. It was shown that, contrary to previous claims, new patterns of variation introduced during contact with MLG had no stabilizing effects on OV frequency in MNw. In fact, there is little retention of OV order during Furthermore, we can assume that the MLG letters do not reflect the exact type of language to which the Norwegians would have been exposed. However, as Pintzuk (1999: 14) concludes, the written language is our only source for drawing general conclusions about the spoken language. I assume that the precautions taken in text selection discussed in Section 1.2 have made the MLG corpus here as reliable as possible for drawing such conclusions. . Note that I agree with the conclusion in Braunmüller (1996, 1998, 2004, 2005), Christoffersen (2000) and Mæhlum (2005) that the MLG–Scandinavian contact was a type of dialect contact and that speakers of both languages could communicate with each other in their respective native languages by using accommodation strategies. However, I wish merely to point out that the evidence here from the written languages suggests that the extent of variability in the two languages is less than previously assumed.
John D. Sundquist
the period under investigation, as the archaic forms begin to decline more rapidly in frequency and VO order becomes dominant. These findings support a hypothesis proposed recently by Reintges (2009), who suggests that syntactic variation is an important source for change. Reintges states this as follows: If we are right in claiming that syntactic variation can arise spontaneously and that syntactic change follows from syntactic variation whatever the source of the variation is, we are in fact making the controversial claim that syntactic change can happen without outside factors playing any role. Syntactic variation initiates syntactic change, as novel patterns emerge or existing patterns compete with and eventually replace one another. (Reintges 2009: 46)
In other words, this view is an alternative to Longobardi’s (2001) principle of inertia: syntactic variation can be a cause of change and the source of this variation need not necessarily be some sort of outside force. The data on OV/VO variation in MNw lend support to this argument. In addition to the sociolinguistic factors addressed above, I propose that internal linguistic variables also contributed to the lack of retention during the period of MLG contact. Recall from the empirical analysis that MNw is primarily a VO language with some occasional OV patterns and that type of complement is one of the factors that significantly affects this marked word order. As discussed in Sundquist (2006), if we look more closely at types of complements in general and the non-pronominal objects in particular, it becomes clear that the remnant OV word order in MNw is regulated by information structure constraints on certain grammatical elements. Following Diesing’s (1997) analysis of Yiddish ‘scrambling’ and OV word order, Sundquist (2006) points out that there is a significant difference between the three main semantic types of NPs — quantificational, referential, and predicational — with respect to their position in the sentence. These three NP-types in MNw are exemplified in (11) with a referential NP, in (12) with a predicational NP and in (13) with a quantificational NP (from Sundquist 2006): (11) skall hon þetta goz hafua shall he those wares have ‘He shall have those wares.’
(DN 2:120; 1314 Bergen)
(12) j huat matta han hafde vpboret landskylder in what way he had paid land taxes ‘in what way he had paid the country taxes.’
(DN 1:630; 1411 Bergen)
(13) at hann hafðe huærn pænning tekit that he had each piece of property taken ‘that he had taken each piece of property.’
(DN 1:100; 1303 Bergen)
Variation, continuity and contact in Middle Norwegian and Middle Low German
In (11), the demonstrative refers to something stated previously in the discourse, while in (12), the indefinite plural NP provides new information that is not stated in the previous discourse. Example (13) contains an NP that is preceded by a quantifier. In Table 3, I propose the tripartite distinction between the types of NPs that occur in MNw.21 Table 3. Non-pronominal objects in MNw according to semantic NP-type Referential NPs
Predicational NPs
Quantificational NPs
Bare NPs with an antecedent in discourse Bare NPs with light verbs Definite NPs Possessive NPs Proper Names Demonstrative NPs
Bare NPs without an antecedent in discourse Indefinite NPs with article Indefinite plurals Cardinal Numbers
Quantificational NPs
These distinctions are based on general differences between “old” vs. “new information.” Referential NPs refer to specific entities previously mentioned in the discourse while predicational NPs provide new information and refer to items not mentioned in the preceding discourse. Quantificational NPs are NPs introduced by quantifiers like the English forms ‘some’, ‘none’, ‘many’ or ‘more’, among others. In a set of 677 examples with the NP-types in Table 3 taken from MNw texts written between 1250 and 1450, the data on the three semantic NP-types reveal significant differences with respect to the frequency of OV word order. The results are listed in Table 4 (from Sundquist 2006): Table 4. Frequency of preverbal non-pronominal object types in MNw (1250–1450) N (preverbal)
% preverbal
Quantificational NPs Referential NPs Predicational NPs
39/71 101/312 45/294
55% 32% 15%
TOTALS
185/677
27%
. For more extensive discussion of these NP-types and examples, see Sundquist (2006). Bare NPs include those non-pronominal objects which are singular NPs not preceded by a definite article or indefinite article. Following Delsing (2000), Sundquist (2006) also provides a separate category for the object–verb pairings with so-called ‘light verbs’ like ha ‘have’ or fa ‘get/receive’ which often occur with singular NPs like behof ‘need’ or lof ‘praise’. Because these object–verb combinations generally occur more frequently in OV order and may also be paraphrased by a single verb on their own, it is necessary to create a separate category for them to avoid skewing the data.
John D. Sundquist
The higher frequency of quantificational NPs in preverbal position is to be expected, given the status of quantified NPs that occur frequently in OV order in many Germanic languages such as Modern Icelandic, Old English, and Yiddish (cf. Pintzuk 2002 and Diesing 1997). However, the difference between referential and predicational NPs in MNw is more surprising and more relevant to our discussion here. Referential NPs occur at a significantly higher rate in OV position than predicational NPs (p=.0000001, χ2=24.10519623, df=1). As in Diesing’s (1997) analysis of “semantically-driven movement” in Yiddish, the position of non-pronominal objects in MNw is dependent on the inherent semantic properties of NP-types. In simple terms, old information tends to occur earlier in the sentence than new information, which generally occurs more frequently near the end of the sentence. OV word order in MNw is, in part, regulated by semantic constraints on the position of certain grammatical elements in the sentence. In other words, OV word order serves a definite function in MNw: it allows certain types of NPs to occur outside their usual position for interpretative reasons. As the analyses in Sundquist (2006) and Haugan (2000) show, information structure plays a similar role in regulating the positioning of elements in the sentence already in earlier stages of Scandinavian, including thirteenth-century Old Norwegian texts and Old Norse sagas. This native Scandinavian pattern of usage continues in MNw, even though it gradually disappears in the latter/second half of the fifteenth century. When we compare the Norwegian and Low German data on semantic NP-type, several differences become apparent. Recall that OV word order in MLG is unmarked and the number of sentences with VO order gradually diminishes throughout the fourteenth and fifteenth centuries. The frequency of all the specific types of nonpronominal objects in MLG is provided in Table 5. Table 5. Frequency of preverbal non-pronominal object types in Middle Low German (1325–1500)
Definite NPs Indefinite NPs with article Possessive NPs Bare NPs (not in previous discourse) Demonstrative NPs Indefinite NP plurals Quantified NPs Bare NPs with light verbs Proper Names Bare NPs w/ antecedent in previous discourse Cardinal Numbers
N (pre-verbal)
% pre-verbal
123/142 26/30 85/104 14/17 33/41 23/29 32/42 8/11 5/7 5/8 7/26
87% 87% 82% 82% 80% 79% 76% 73% 71% 63% 27%
Variation, continuity and contact in Middle Norwegian and Middle Low German
Following the classification of these types of NP-objects into the three groupings above, we see, in Table 6, that MLG exhibits no significant differences with respect to OV word order. Table 6. Frequency of pre-verbal non-pronominal object-types in Middle Low German (1325–1500)
Referential NPs Quantificational NPs Predicational NPs
N (pre-verbal)
% pre-verbal
254/313 32/42 75/102
81% 76% 74%
Unlike in Norwegian, all three NP-types behave similarly in Low German. Since this word order is the unmarked option, we expect the OV-rate among all three semantic types to be high. However, the fact that there is no difference between them in MLG reveals a fundamental difference in the function of OV word order in the two languages. In Low German, OV order is a default, unmarked option in which semantic distinctions between NP-types have little effect on word order. In MNw, however, OV order performs a specialized function inherited from Old Norse as a means by which interpretative contrasts in NP-types may be accentuated. Thus, the source of variation in MNw that leads to changes in the frequency of OV word order is tightened constraints on positioning of various semantic NP-types and not the influx of MLG through contact. In terms of Reintges’ (2009) alternative view on Longobardi’s principle of inertia, the MNw data provide evidence of syntactic change that came about without an outside force: syntactic change was the result of syntactic variation that arose from the interaction of internal linguistic variables in Norwegian and not language contact. It appears that OV word order in MNw was not susceptible to any external influence and ran its own course of development in early Norwegian. It is a native Scandinavian word order pattern in MNw whose frequency and distribution were dependent on internal linguistic variables and underlying rules of variation inherited from Old Norse. As we saw above, the specific variables like semantic NP-type that affect the occurrence of OV word order in MNw differ from those which affect this same pattern in MLG. Thus, despite the fact that both languages exhibit OV order on the surface, the contexts in which it occurs and the internal variables that underlie its frequency are fundamentally different. These same internal factors that shape OV usage are also responsible for its gradual drift toward markedness in Norwegian. In other words, it appears that the native pattern of variable use of syntactic options and the underlying system of internal rules that affect this pattern have a more significant effect than outside influences and external linguistic forces.
John D. Sundquist
4. Summary and conclusions This chapter has examined three main issues. First, it explored the possibility that contact between two typologically similar languages might lead to a period of continuity characterized by the retention of syntactic forms. Secondly, the focus shifted to the issue of syntactic variation and whether or not continuity and stability might possibly arise when two or more syntactic variants exist during a period of contact. Thirdly, the discussion focused on syntactic variation as a cause of change rather than continuity. The empirical portion of the study is based on a Middle Low German and Middle Norwegian corpus of diplomatic letters written in Norway and in northern Germany between 1300 and 1525. Data on the alternation between OV and VO order and on the variables that shape this variation are used to explore these three main issues. Data from this particular corpus provide no evidence of retention of archaic OV word-order patterns in Norwegian during the period of contact between German and Norwegian merchants in Norway. In fact, empirical evidence indicates the opposite trend: OV order declines during the same period when it is on the rise in MLG. There is, in relevant respects, a general lack of syntactic change in MNw, although this lack of change is not characterized by retention of older, outgoing syntactic forms. This syntactic continuity in MNw, despite the prolonged contact with MLG, is most likely a result of the sociolinguistic conditions of contact in Norway. Germans and Norwegians in Bergen, for instance, had little interaction with each other beyond business transactions. However, an alternative possibility was discussed, as proposed by Braunmüller (1998) and Christoffersen (2000), who suggest that the lack of syntactic borrowing in Norwegian, or any of the Mainland Scandinavian languages, is the result of the high degree of syntactic variation in two very similar languages. Both Norwegians and Germans could speak their own language and use accommodation strategies in a type of multi-dialectal communication. Thus, according to this view, MLG contact with Norwegian had little impact beyond the lexical level. Evidence from the written language as represented in this corpus, however, shows relatively little OV/ VO variation in MLG, especially in the fifteenth century when OV is almost categorical, and MNw exhibits its own underlying system of variation that has little random fluctuation of OV and VO forms. The data point to a situation in which each language exhibits a unique pattern of usage and suggest that these patterns do not influence each other. A final point is made in the last section of the chapter regarding variation as a catalyst for change rather than continuity. The evidence here support the proposal by Reintges (2009) that syntactic variation may bring about change despite the absence of any outside forces like language contact. This alternative to Longobardi’s (2001) principle of inertia claims that variation may arise spontaneously or through the interaction of internal linguistic variables rather than outside forces and that this variation
Variation, continuity and contact in Middle Norwegian and Middle Low German
may result in syntactic change. The data on non-pronominal objects in MNw indicate that the decline of OV word order may have come about in a similar way. Information structure constraints and tightened restrictions on the movement of certain semantic NP-types slowed down the rate of OV word order and limited syntactic options in MNw. Thus, variational patterns that were inherited from earlier stages of Norwegian, and not the influx of variational patterns through language contact, are responsible for disruptions in syntactic continuity along the way.
References Primary sources Codex Diplomaticus Lubecensis: Lübeckisches Urkundenbuch. Part 1, vols. 4, 5, 7, 8, 11. Lübeck: Ferdinand Grautoff, 1877–1905. Diplomatarium Norvegicum: Oldbreve til Kundskab om Norges indre og ydre Forhold, Sprog Slægter, sæder, Lovgivning og Rettergang i Middelalderen, vols. 1–22, Kommisjonen for Diplomatarium Norvegicum (eds). Christiania, 1847–1995.
Secondary sources Betten, Anne. 1980. Zu Satzbau und Satzkomplexität im mittelhochdeutschen Prosa-Lancelot. Sprachwissenschaft 5: 15–42. Bies, Ann. 1996. Syntax and Discourse Factors in Early New High German: Evidence for Verbfinal Word Order. MA thesis, University of Pennsylvania. Brattegard, Olav. 1963. Niederdeutsch und Norwegisch am hansischen Kontor zu Bergen in Norwegen. Niederdeutsches Jahrbuch 86: 7–16. Braunmüller, Kurt. 1993a. Voraussetzungen für die Übernahme mittelniederdeutscher Sprachstrukturen in die skandinavischen Sprachen. In Niederdeutsch und die skandinavischen Sprachen I, Kurt Braunmüller & Willy Diercks (eds), 137–160. Heidelberg: Carl Winter. Braunmüller, Kurt. 1993b. Syntaktische Divergenzen und Transferenzen. In Niederdeutsch und die skandinavischen Sprachen I, Kurt Braunmüller & Willy Diercks (eds), 231–270. Heidelberg: Carl Winter. Braunmüller, Kurt. 1996. Forms of language contact in the area of the Hanseatic League: Dialect contact phenomena and semicommunication. Nordic Journal of Linguistics 19: 141–154. Braunmüller, Kurt. 1998. Wortstellungstypologische Untersuchungen zu den Kontaktsprachen der Hansezeit. In Historische germanische und deutsche Syntax, John Ole Askedal (ed.), 315–334. Frankfurt: Peter Lang. Braunmüller, Kurt. 2004. Niederdeutsch und Hochdeutsch im Kontakt mit den skandinavischen Sprachen: Eine Übersicht. In Deutsch im Kontakt mit germanischen Sprachen, Horst Haider Munske (ed.), 1–30. Tübingen: Max Niemeyer. Braunmüller, Kurt. 2005. Language contacts in the late Middle Ages and in early modern times. In The Nordic Languages, Vol. II, Oskar Bandle, Kurt Braunmüller, Ernst Håkon Jahr, Allan Karker, Hans-Peter Naumann & Ulf Teleman (eds), 1222–1233. Berlin: Walter de Gruyter. Christoffersen, Marit. 1993. Setning og sammenheng: Syntaktiske studier i Magnus Lagabøters Landslov. ADH-serien 65. Agder: Agder Distriktshøgskole.
John D. Sundquist Christoffersen, Marit. 2000. Leddsetninga i norsk – er her spor av mellomnedertysk påvirkning? In Språkkontakt – Innverknaden frå nedertysk på andre nordeuropeiska språk, Ernst Håkon Jahr (ed.), 153–169. Copenhagen: Nordisk Ministerråd. Delsing, Lars-Olof. 2000. From OV to VO in Swedish. In Diachronic Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 255–274. Oxford: OUP. Diesing, Molly. 1997. Yiddish VP order and the typology of object movement in Germanic. Natural Language and Linguistic Theory 15: 369–427. Ebert, Roger Peter. 1980. Social and stylistic variation in Early New High German word order: The sentence frame (‘Satzrahmen’). Beiträge zur Geschichte der deutschen Sprache 102: 357–398. Ebert, Roger Peter. 1986. Historische Syntax des Deutschen II: 1300–1750. Bern: Peter Lang. Ebert, Roger Peter, Reichmann, Oskar, Solms, Hans-Joachim & Wegera, Klaus-Peter. 1993. Frühneuhochdeutsche Grammatik. Tübingen: Max Niemeyer. Ersland, Geir Atle. 2003. Was the kontor in Bergen a topographically closed entity? In Das Hansische Kontor zu Bergen und die Lübecker Bergenfahrer, Antjekathrin Graßmann (ed.), 41–57. Lübeck: Archiv der Hansestadt Lübeck. Gade, John Allyne. 1951. The Hanseatic Control of Norwegian Commerce During the Late Middle Ages. Leiden: E.J. Brill. Haugan, Jens. 2000. Old Norse Word Order and Information Structure. Ph.D. dissertation, University of Trondheim. Helle, Knut. 1994. The Hanseatic merchants in Bergen in the Middle Ages. In Bergen and the German Hansa, Ingvild Øye (ed.), 11–28. Bergen: Bryggens Museum. Jahr, Ernst H. 1999. Sociolinguistics in historical language contact: The Scandinavian languages and Low German during the Hanseatic period. In Language Change, Ernst Håkon Jahr (ed.), 119–140. Berlin: Mouton de Gruyter. Keenan, Edward L. 2002. Explaining the creation of reflexive pronouns in English. In Studies in the History of English: A Millennial Perspective, Donka Minkova & Robert Stockwell (eds), 325–355. Berlin: Mouton de Gruyter. Longobardi, Giuseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The history of French chez. Linguistic Inquiry 32: 275–302. Mæhlum, Brit. 2005. Sociolinguistic structures chronologically III: Norwegian. In The Nordic Languages, Vol. II, Oskar Bandle, Kurt Braunmüller, Ernst Håkon Jahr, Allan Karker, Hans-Peter Naumann & Ulf Teleman (eds), 1907–1922. Berlin: Walter de Gruyter. Mørck, Endre. 1980. Passiv i mellomnorske diplom. MA thesis, University of Oslo. Nesse, Agnete. 2002. Språkkontakt mellom norsk og tysk i hansatidens Bergen. Oslo: Novus. Pintzuk, Susan. 1999. Phrase Structures in Competition: Variation and Change in Old English Word Order. New York NY: Garland. Pintzuk, Susan. 2002. Verb-object order in Old English: Variation as grammatical competition. In Syntactic Effects of Morphological Change, David Lightfoot (ed.), 276–299. Oxford: OUP. Reintges, Chris. 2009. Spontaneous syntactic change. In Historical Syntax and Linguistic Theory, Paola Crisma & Giuseppe Longobardi (eds), 41–57. Oxford: OUP. Schildt, Joachim. 1977. Zur Ausbildung des Satzrahmens. Zur Ausbildung der Norm der deutschen Literatursprache auf der syntaktischen Ebene, Gerhard Kettmann & Joachim Schildt (eds), 235–284. Berlin: Akademie Verlag. Simensen, Erik. 1989. Zum Kontakt zwischen Niederdeutsch und Nordisch im Mittelalter unter besonderer Berücksichtigung der Frage nach einer niederdeutsch-nordischen
Variation, continuity and contact in Middle Norwegian and Middle Low German
Mischsprache. In Niederdeutsch in Skandinavien II, Karl Hyldgaard-Jensen, Vibeke Winge & Birgit Christiansen (eds), 65–77. Berlin: Erich Schmidt. Sundquist, John D. 2002. Morphosyntactic Change in History of the Mainland Scandinavian Languages. Ph.D. dissertation, Indiana University. Sundquist, John D. 2006. Syntactic variation in the history of Norwegian and the decline of XV word order. Diachronica 23: 105–145. Thomason, Sarah Grey & Kaufman, Terrence. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley CA: University of California Press. Trudgill, Peter. 1994. Language contact and dialect contact in linguistic change. In Dialektkontakt, språkkontakt och språkförändring i Norden, Ulla-Britt Kotsinas & John Helgander (eds), 13–22. Stockholm: Stockholm University Press. Zeevaert, Ludger. 1995. Wie intensiv war der mittelniederdeutsch-skandinavische Sprachkontakt wirklich? Untersuchungen von Transferenzen im Grundwortschatz. In Niederdeutsch und die skandinavischen Sprachen II, Kurt Braunmüller (ed.), 145–179. Heidelberg: Carl Winter.
part ii
Change
Directionality in word-order change in Austronesian languages Edith Aldridge
University of Washington This chapter proposes a diachronic account of VOS and SVO word-order variation in Austronesian languages. I argue that the ‘S’ argument in VOS languages is a topic which moves to an A′-position in the left periphery of the clause. Movement of the topic is followed by fronting of the remnant TP to a higher position in the CP layer. The change to SVO word order results from the loss of TP fronting. Subsequent to this, the clause-initial topic is reanalyzed as an A-position subject. I show that the ‘S’ arguments in SVO Austronesian languages have characteristics of both A-position subjects and A′-position topics, providing evidence for the direction of change from topic to subject.
1. Introduction This chapter proposes a diachronic scenario to account for the synchronic variation between SVO and verb-initial word-order types in Austronesian languages. It is uncontroversial that verb-initial word order is more conservative than SVO (Starosta et al. 1982; Cumming 1991; Donohue 2007, 2008, among others). What I propose in this chapter is that SVO has evolved from verb-initial order via topicalization of the subject. The verb-initial language which I primarily use for illustration in this chapter is the VOS language Seediq, an Atayalic language spoken in Taiwan. Seediq is an ergative language; the absolutive argument always appears in clause-final position, as in (1a). The SVO word order type is illustrated by the Indonesian example in (1b). (1) a.
Wada biq-un hulama na Ape ka laqi. past give-tr treat erg Ape abs child ‘Ape gave the child a treat.’
b.
Ali mem-beli buku pada Nuri. Ali act-buy book for Nuri ‘Ali bought a book for Nuri.’
What I focus on in this chapter is showing that the absolutive argument in the verbinitial languages has characteristics of a A′-position topic, rather than a subject residing
Edith Aldridge
in an A-position like [Spec, TP]. In SVO languages, the clause-initial DP has properties of both A-position subjects and A′-position topics. Crucially, what this suggests is that this DP in SVO languages is currently in the process of being reanalyzed from a topic to an A-position subject.
2. Verb-initial languages In this section I discuss the position and interpretive properties of absolutives in verbinitial languages. I focus primarily on VOS Seediq, since the clause-final absolutive in this type of language can clearly be demonstrated to reside in a TP-external topic position and not in an A-position like [Spec, TP]. The absolutive in VSO languages like Tagalog, however, also has some interpretative characteristics of a topic. Therefore, absolutives in verb-initial languages in general should not be analyzed as subjects residing in an A-position. There are two possible derivations which result in absolutive-final word order. One is simply rightward movement of the absolutive, similar to the analysis of VOS order in Malagasy by Guilfoyle, Hung, and Travis (1992). (2)
CP C′ C
DPAbs TP
…tAbs…
The other is leftward movement of the absolutive, followed by subsequent fronting the remnant TP to the left of this nominal. (3)
CP TP …tAbs…
C′ DPAbs C[EPP] > C
C′ tTP
Recent work on VOS word order in Austronesian languages (Massam 2000, 2001, 2003 for Niuean; Rackowski & Travis 2000; Pearson 2001 for Malagasy; Cole & Hermon 2008 for Toba Batak) has converged on a derivation involving fronting of the clause or predicate, along the lines of (3). Aldridge (2002, 2004) and Cole and Hermon (2008) argue specifically in favour of the predicate-fronting approach, based on the empirical observation that the fronted TP forms an island to extraction. As is well-known, only
Directionality in word-order change in Austronesian languages
the absolutive nominal in VOS Austronesian languages is capable of being A′-extracted. All other clause-internal constituents must remain in TP, which is accounted for by the derivation in (3), the fronted TP forming an island to extraction due to the ‘freezing effect’ on fronted constituents (Wexler & Culicover 1980) or the Condition on Extraction Domain (Huang 1982).1 A theoretical reason to favour (3) for the purposes of the present chapter is that it provides a direct input to the change to SVO word order, which involves a leftward position for the topic. I will assume the derivation in (3) in what follows. The reader is referred to cited works for detailed discussion. The main focus here is showing that the absolutive nominal in VOS Seediq has a fixed position in the clause and that this position is an A′ topic position. First, the absolutive DP must be in clause-final position. No other clause-internal constituent can follow it.
(4) Seediq
a.
Wada biq-un hulama na Ape ka laqi. past give-tr treat erg Ape abs child ‘Ape gave the child a treat.’
b. *Wada biq-un hulama ka laki na Ape. past give-tr treat abs child erg Ape
‘Ape gave the child a treat.’
c. *Wada biq-un na Ape ka laki hulama. past give-tr erg Ape abs child treat ‘Ape gave the child a treat.’
Secondly, the absolutive argument can be shown to occupy a structurally prominent position. Clauses can be coordinated, excluding the absolutive.
(5) Seediq
a.
[tp S-bari=na
app-buy=3.sg.erg treat
hulama] ma
[tp s-smalu=na lukus dungan] app-make=3.sg.erg clothes also
ka laqi=na. abs child=3.sg.gen
‘She buys a treat for and also makes clothes for her child.’
b.
[tp M-usa Purishia] [tp m-ari sama] ka Ape. intr-go Puli intr.buy vegetable abs Ape ‘Ape went to Puli and bought vegetables.’
and
. See Nunes and Uriagereka (2000), Stepanovic (2007), and others for more recent approaches to CED effects.
Edith Aldridge
NPIs cannot be licensed in absolutive position, as in (6a), indicating that the absolutive DP resides in a position not c-commanded by negation. An NPI can, however, appear in direct object position in an antipassive, as in (6b). In an antipassive construction, the verb is marked as intransitive, the external argument has absolutive case, and the theme or patient is treated syntactically as an oblique. In Seediq, oblique objects remain in their base positions in VP and therefore can be c-commanded by negation.
(6) Seediq
a. *[Ini burig-i na Ape] ka animumaanu. neg buy-tr.irr erg Ape abs anything ‘Ape didn’t buy anything.’ b.
[Wada ini bari animumaanu] ka Ape. past neg buy anything abs Ape ‘Ape didn’t buy anything.’
The topic status of the absolutive is demonstrated by the next set of examples. Pearson (2001, 2005) has argued extensively that the clause-final DP in VOS Malagasy resides in an A′-position. One argument he employs is to show that the clause-final DP must be definite. This is also true for Seediq. For example, Chang (1997) shows that whwords, which are inherently focused and not presupposed, are prohibited from occupying this position. The wh-word undergoes fronting to clause-initial position in (7a), while it remains in situ in (7b).
(7) Seediq
a.
Ima s<m>ebut laqi. who hit child ‘Who hits a child?’
b.
*S<m>ebut laqi ka ima? hit child abs who ‘Who hits a child?’
(Chang 1997: 146)
The asymmetry between (7a) and (7b) should not be understood merely as a requirement that wh-words move to [Spec, CP], since adjunct wh-words appear in post-verbal position, as in (8). (8)
Seediq Mari inu patis Ape intrbuy where book Ape ‘Where did Ape buy books?’
If an object in Seediq is indefinite, it must be packaged as the oblique in an antipassive. In (9a), the verb has intransitive morphology. The object remains in VP, immediately following the verb. The external argument has absolutive case. (9b) is a transitive clause, in which the object has absolutive case. The object in the transitive clause is definite.
Directionality in word-order change in Austronesian languages
(9) Seediq
a.
Moda m-ari qushia mutaso Hori ka Ape. intrgo intr-buy water clean Puli abs Ape ‘Ape went to Puli to buy clean water.’
b.
Wada=na s-pahu lukus ka qushia mutaso. past=3.sg.erg app-wash clothes abs water clean ‘She washed clothes with the clean water.’
Finally, Aldridge (2004) shows that absolutives in Seediq are resumed by clitic pronouns in the clause, while DPs remaining in argument position are not doubled by clitics. (10) Seediq a.
Gaga=ku=daha ngal-un ka yaku duri. pres=1.sg.abs=3.pl choose-tr abs 1.sg also ‘They have also chosen me.’
b.
M-usa=ku mu-huma kyuuri=na intr-go=1.sg.abs intr-plant cucumber=3.sg.gen
ka yaku. abs 1.sg
‘I went to plant his cucumbers.’
No clitic appears with the ergative DP, however, since it remains in situ in vP. Ergative clitics are indeed possible, but only when there is no full DP in [Spec, vP], as in (11c). (11) Seediq a.
Wada burig-un na Ape ka patis-ni. past buy-tr erg Ape abs book-dem ‘Ape bought this book.’
b.
*Wada=na burig-un na Ape ka patis-ni. past=3.sg.erg buy-tr erg Ape abs book-dem ‘Ape bought this book.’
c.
Wada=na biq-un hulama laqi past=3.sg.erg give-tr treat child.abs ‘He/she gave the child a treat.’
Another important indication that absolutives reside in an A′-topic position and not [Spec, TP] comes from reconstruction effects. Pearson (2001, 2005) shows for Malagasy that a variable in the clause-final DP can be bound by the clause-internal ergative argument. He proposes that this is possible because the surface position of the clausefinal argument is a TP-external topic position and not [Spec, TP] subject position. This allows reconstruction for the purposes of binding, with the result that the possessors in the clause-final topics in (12) can be interpreted as variables bound by the quantified DPs in external argument position within TP.
Edith Aldridge
(12) Malagasy
(Pearson 2005: 424)
a.
Novangian’ny mpianatra tsirairayi ny rainyi past.tt.visit.det student each det father.3
omaly. yesterday
‘Hisi father, eachi student visited yesterday.’
b.
Norohan’
past.tt.kiss det woman all det spouse.3 ‘Theiri spouses, all the womeni kissed.’
ny
vehivavy rehetrai ny
vadinyi
Absolutives in the VSO2 language Tagalog must also be interpreted as definite, as in (13a). If an object is indefinite (or nonspecific), then the clause must be antipassive and the object given oblique status, as in (13b). Richards (2000), Rackowski (2002), Aldridge (2004, 2005, 2008), and Rackowski and Richards (2005) all assume that basic word order is derived by head-movement of the verb to a position external to vP. The ergative and absolutive DPs remain in their base positions in vP in overt syntax. The presuppositional interpretation of the absolutive is obtained by covert movement out of the VP, along the lines of Diesing (1992). (13) Tagalog a.
Bili ng babae ang isda. buy erg woman abs fish ‘The woman bought the/*a fish.’
b.
Bili ang babae ng isda. buy abs woman obl fish ‘The woman bought a/*the fish.’
This section has shown that absolutives in verb-initial languages have characteristics of topics and not A-position subjects.
3. Subject in SVO languages This section discusses SVO, or Topic-initial, word order in two languages in Indonesia. The diachronic scenario is shown in (14) and (15). First, the absolutive DP undergoes topicalization to clause-initial position. This is the first step in the VOS word order
. Strictly speaking, Tagalog word order is quite free. VSO is the basic order; permutations are permitted via scrambling. The key difference between VSO and VOS order is whether the absolutive argument has a fixed position. This is true only for VOS languages, which do not generally allow scrambling.
Directionality in word-order change in Austronesian languages
erivation assumed in (3). However, the second step does not take place, leaving TP in d its base position following the topic. CP
(14)
DPTop
C′ C
TP …tTop…
The clause-initial topic is subsequently reanalyzed as an A-position subject located in [Spec, TP]. TP
(15)
DPSubj
T′ T
nP …tSubj…
However, this is a gradual process. What I show in this section is that clause-initial DPs in Indonesian and Balinese have a combination of A and A′ properties. What this suggests is that the reanalysis from topic to subject is not yet complete. In structural terms, this means that both [Spec, TP] and [Spec, CP] are available to clause-initial DP arguments in these languages. I begin with the A-properties. (16) shows first that clause-initial position can host the controlled gap in a nonfinite clause. (16) Indonesian a.
Saya ingin [pro mem-ilih dia sebagai presiden]. 1.sg want act-choose 3.sg as president ‘I want to choose him/her as president.’
b.
Saya ingin [pro di-pilih sebagai presiden]. 1.sg want pass-choose as president ‘I want to be chosen as president.’
Secondly, this position is the target of raising-to-object or exceptional case-marking. This is shown for Balinese. The fact that the embedded subject is accessible to caselicensing from the matrix clause indicates that it must be located in an A-position. Similarly, the asymmetry between the reflexive in (17a) and the pronoun in (17b) shows that the embedded subject position counts as an A-position for the purposes of binding.
Edith Aldridge
In other words, if the pronoun were in an A′-position, we would expect a reconstruction effect, which would yield opposite predictions to the binding relations exhibited in (17). For example, if reconstruction took place, then we would expect binding to be possible in (17b), since the pronoun would be interpreted in its base position in the embedded clause and therefore not in the local domain of the antecedent. (17) Balinese
(Wechsler & Arka 1998: 437)
a.
Iai nawang awaknei lakar tangkep polisi. 3 av.know self fut ov.arrest police ‘He knew that the police would arrest self.’
b.
*Iai nawang iai lakar tangkep polisi. 3 av.know 3 fut ov.arrest police ‘He knew that the police would arrest him.’
This is not to say that reconstruction of clause-initial DPs is never observed in Indonesian and Balinese. In Balinese, internal argument reflexive pronouns can be bound by the agent, regardless of whether they remain in VP or move to clause-initial position. In the actor voice construction in (18a), the agent is the subject and can bind the object reflexive in VP. In the object voice construction in (18b), however, the internal argument has been moved to clause-initial position, while the agent remains in situ to the right of the verb. Note that binding still obtains, indicating that movement of the object cannot be A-movement which would alter binding relations. Therefore, examples like (18) provide evidence that clause-initial position sometimes has A′-properties. (18) Balinese a.
Ia ningalin awakne. 3 av.see self ‘(S)he saw herself/himself.’
b.
Awakne tingalin=a. self ov.see=3 ‘(S)he saw herself/himself.’
(Wechsler & Arka 1998: 406)
Similar examples can also be found in Indonesian. A fronted reflexive can be bound by a pronominal agent which follows it in surface order. (19) Indonesian
(Arka & Manning 1998: 8)
a.
Diri-saya saya serah-kan ke polisi. self-1.sg 1.sg surrender-app to police ‘I surrendered myself to the police.’
b.
Diri-nya mesti dia serah-kan ke polisi. self-3.sg.gen must 3.sg surrender-app to police ‘(S)he must surrender herself/himself to the police.’
Reconstruction effects as shown in (18) for Balinese and (19) for Indonesian are parallel to the reconstruction effects seen for VOS Malagasy in (12), indicating that the
Directionality in word-order change in Austronesian languages
clause-initial DP in the SVO languages retains some residual characteristics of topics found in the verb-initial languages. This combination of the presence and absence of reconstruction effects is reminiscent of clause-internal scrambling in languages like Japanese and Hindi. Movement of an internal argument to clause-initial position sometimes creates a new binding relation in these languages. (20a) and (20b) show that the reciprocal otagai requires a c-commanding antecedent. In (20c), the intended antecedent is scrambled to clauseinitial position, where it c-commands and binds the reciprocal. (20) Japanese
(Saito 2003: 485)
a.
Karera-ga [otagai-no sensei]-o hihansita. they-nom each.other-gen teacher-acc criticized ‘They criticized each other’s teachers.’
b.
?*[otagai-no
each.other-gen teacher-nom they-acc criticized ‘Each other’s teachers criticized them.’
c.
?Karera-o [otagai-no sensei]-ga t hihansita. they-acc each.other-gen teacher-nom criticized ‘Them, each other’s teachers criticized.’
sensei]-ga
karera-o hihansita.
On the other hand, clause-internal scrambling sometimes appears to require reconstruction for the purposes of binding. In (21), a reflexive anaphor has been scrambled to a position where it c-commands its intended antecedent. If reconstruction did not take place, both Condition C and Condition A would be violated, since the anaphor is not c-commanded by the antecedent and in fact itself c-commands the antecedent. Since binding is possible, we must conclude that reconstruction takes place. (21) Zibunzisin-o Taroo-ga t semeta. self-acc Taro-nom blamed ‘Himself, Taro blamed.’
This seemingly contradictory behaviour of Japanese scrambling is traditionally treated as evidence that the landing site of this type of movement is sometimes an A-position and sometimes an A′-position (Mahajan 1990; Tada 1993; Saito 1992, 2003; and others). The mixture of A and A′ properties in Indonesian and Balinese subjects can be accounted for on the analysis presented in (14) and (15). Specifically, the reanalysis from topic to subject is not yet complete, meaning that clause-initial DP arguments are able to occupy either [Spec, TP] or [Spec, CP].
4. Conclusion In this chapter I have argued that synchronic variation between verb-initial and SVO word order in Austronesian languages can be accounted for diachronically by positing
Edith Aldridge
topicalization of the absolutive argument, followed by subsequent reanalysis of this DP as an A-position subject. I have shown that the absolutive in verb-initial languages has characteristics of a topic and not an A-position subject. Subjects in the SVO languages, on the other hand, show a mixture of A and A′ properties, indicating that the reanalysis is currently in progress in these languages. I might point out here that Donohue (2007, 2008) offers an alternative view of the change from verb-initial to SVO word order in some Austronesian languages. He shows that there are large concentrations of SVO Austronesian languages in geographic proximity to SVO and SOV languages in mainland Southeast Asia and western New Guinea and proposes that the change to SVO in Austronesian languages was the result of contact with subject-initial neighbours. I offer in passing an informal opinion that this proposal does not contradict the claims made in the current chapter. In fact, the two proposals can perhaps be seen to complement each other. Areal contact could very well be the external catalyst for the change in basic word order. What I have endeavoured to argue for in this chapter is a structural account of the change itself. My claim that the post-verbal subject had access to a clause-peripheral topic position facilitated the change to subject (or topic) initial order, since all that was required syntactically was the loss of TP-fronting, which left the topic in a clause-initial [Spec, CP], as shown in (14). The proposal put forth in this chapter has interesting consequences for the study of diachronic syntax. First, the identification of absolutives with topics (rather than subjects) in the ergative verb-initial languages is not surprising, given the well-known fact that syntactically ergative languages do not have a well-defined notion of grammatical subject, the cross-linguistic characteristics of subjects being distributed between the ergative and absolutive DPs (Bok-Bennema 1991; Murasugi 1992; Campana 1992; Bittner 1994; Bittner & Hale 1996a, b; Manning 1996; Ura 2000). Therefore, it should not be surprising that the change to SVO word order is accompanied by the reanalysis of absolutive to subject, concomitant with the change from ergative to accusative alignment, as argued for Austronesian languages by Aldridge (2008). Another consequence of the analysis is the clear evidence of continuity of older characteristics during the process of change, rather than an abrupt shift from ergative verb-initial syntax to accusative SVO. The relative stability of the mixture of A and A′ properties of subjects in Indonesian languages is clear evidence for the gradual nature of syntactic change.
References Aldridge, Edith. 2002. Wh-movement in Seediq and Tagalog. In Proceedings of AFLA 8: The 8th Meeting of the Austronesian Formal Linguistics Association [MIT Working Papers in Linguistics 44], Andrea Rackowski & Norvin Richards (eds), 1–28. Cambridge MA: MITWPL. Aldridge, Edith. 2004. Ergativity and Word Order in Austronesian Languages. Ph.D. dissertation, Cornell University.
Directionality in word-order change in Austronesian languages
Aldridge, Edith. 2005. Phase theory account of absolutive extraction in Tagalog. In Perspectives on Phases [MIT Working Papers in Linguistics 49], Martha McGinnis & Norvin Richards (eds), 1–27. Cambridge MA: MITWPL. Aldridge, Edith. 2008. Phase-based account of extraction in Indonesian. Lingua 118(10): 1440–1469. Arka, Wayan & Manning, Christopher. 1998. Voice and grammatical relations in Indonesian: A new perspective, In Proceedings of the LFG98 Conferences, Miriam Butt & Tracy Holloway King (eds). Stanford CA: CSLI. Bittner, Maria. 1994. Case, Scope, and Binding. Dordrecht: Kluwer. Bittner, Maria & Hale, Ken. 1996a. The structural determination of case and agreement. Linguistic Inquiry 27: 1–68. Bittner, Maria & Hale, Ken. 1996b. Ergativity: Toward a theory of a heterogeneous class. Linguistic Inquiry 27: 531–604. Bok-Bennema, Reineke. 1991. Case and Agreement in Inuit. Berlin: Foris. Campana, Mark. 1992. A Movement Theory of Ergativity. Ph.D. dissertation, McGill University. Chang, Yung-li. 1997. Voice, Case and Agreement in Seediq and Kavalan. Ph.D. dissertation, National Tsing Hua University. Cole, Peter & Hermon, Gabriella. 2008. VP raising in a VOS language. Syntax 11(2): 144–197. Cumming, Susanna. 1991. Functional Change: The Case of Malay Constituent Order [Discourse Perspectives on Grammar 2], Paul Hopper & Sandra A. Thompson (eds). Berlin: Mouton de Gruyter. Diesing, Molly. 1992. Indefinites [Linguistic Inquiry Monograph 20]. Cambridge MA: The MIT Press. Donohue, Mark. 2007. Word order in New Guinea: Dispelling a myth. Oceanic Linguistics 44(2): 527–536. Donohue, Mark. 2008. Word order in Austronesian from north to south and west to east. Linguistic Typology 11: 349–391. Guilfoyle, Eithne, Hung, Henrietta & Travis, Lisa. 1992. Spec of IP and Spec of VP: Two subjects in Austronesian languages. Natural Language and Linguistic Theory 10: 375–414. Huang, James C.-T. 1982. Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, MIT. Mahajan, Anoop. 1990. The A/A-bar Distinction and Movement Theory. Ph.D. dissertation, MIT. Manning, Christopher. 1996. Ergativity: Argument Structure and Grammatical Relations. Stanford CA: Center for the Study of Language and Information. Massam, Diane. 2000. VSO and VOS: Aspects of Niuean word order. In The Syntax of Verb Initial Languages, Andrew Carnie & Eithne Guilfoyle (eds), 97–116. Oxford: OUP. Massam, Diane. 2001. Pseudo noun incorporation in Niuean. Natural Language and Linguistic Theory 19: 153–197. Massam, Diane. 2003. Questions and the left periphery in Niuean. In Proceedings of the Ninth Annual Meeting of the Austronesian Formal Linguistics Association (AFLA 9) [Cornell Working Papers in Linguistics 19], Anastasia Riehl & Thess Savella (eds), 94–106. Ithaca NY: CLC Publications. Murasugi, Kumiko G. 1992. Crossing and Nested Paths: NP Movement in Accusative and Ergative Languages. Ph.D. dissertation, MIT. Nunes, Jairo & Uriagereka, Juan. 2000. Cyclicity and extraction domains. Syntax 3(1): 20–43. Pearson, Matthew. 2001. The Clause Structure of Malagasy: A Minimalist Approach. Ph.D. dissertation, UCLA.
Edith Aldridge Pearson, Matthew. 2005. The Malagasy subject/topic as an A′-element. Natural Language and Linguistic Theory 23(2): 381–457. Rackowski, Andrea. 2002. The Structure of Tagalog: Specificity, Voice, and the Distribution of Arguments. Ph.D. dissertation, MIT. Rackowski, Andrea & Richards, Norvin. 2005. Phase edge and extraction: A Tagalog case study. Linguistic Inquiry 36(4): 565–599. Rackowski, Andrea & Travis, Lisa. 2000. V-initial languages: X or XP movement and adverbial placement. In The Syntax of Verb Initial Languages, Andrew Carnie & Eithne Guilfoyle (eds), 117–142. Oxford: OUP. Richards, Norvin. 2000. Another look at Tagalog subjects. In Formal Issues in Austronesian Linguistics, Ilena Paul, Vivianne Phillips & Lisa Travis (eds), 105–116. Dordrecht: Kluwer. Saito, Mamoru. 1992. Long distance scrambling in Japanese. Journal of East Asian Linguistics 1: 96–118. Saito, Mamoru. 2003. A derivational approach to the interpretation of scrambling chains. Lingua 113: 481–518. Starosta, Stanley, Pawley, Andrew & Reid, Lawrence. 1982. The evolution of focus in Austronesian. In Papers from the Third International Conference on Austronesian Linguistics, Vol. 2, Amran Halim, Lois Carrington & Stephen Adolphe Wurm (eds), 145–170. Canberra: Pacific Linguistics. Stepanov, Arthur. 2007. The end of CED? Minimalism and extraction domains. Syntax 10(1): 80–126. Tada, Hiroaki. 1993. A/A-bar Partition in Derivation. Ph.D. dissertation, MIT. Ura, Hiroyuki. 2000. Checking Theory and Grammatical Functions in Universal Grammar. Oxford: OUP. Wechsler, Stephen & Arka, Wayan. 1998. Syntactic ergativity in Balinese: An argument structure based theory. Natural Language and Linguistic Theory 16: 387–441. Wexler, Kenneth & Culicover, Peter W. 1980. Formal Principles of Language Acquisition. Cambridge MA: The MIT Press.
Negative co-ordination in the history of English Richard Ingham
Birmingham City University This chapter examines the syntax of negative coordination in the history of English. Two major developments can be observed: (a) in Late Middle English it becomes impossible for a negative conjunct clause introduced by the negative conjunction ne to follow an initial affirmative conjunct clause – since the fourteenth century the initial conjunct must also be negative; (b) nor replaces ne in the second conjunct, which is interpreted as negative but contains no other negative element. These developments follow from two proposed changes in the structure of English clausal negation. First, not replaces a null negative operator in SpecNegP. This operator had checked the uninterpretable feature of ne/nor. Its loss thus requires a c-commanding negative clause to check this feature. Second, nor itself gains an interpretable negative feature, making the following clause negative in interpretation without any further negative element.
1. Introduction Despite a large and growing number of studies devoted to the structure of negation in the history of English, there remain areas of this topic that have seen little or no systematic investigation in terms of diachronic changes and their relevance to syntactic theory. Among them is the case of co-ordinated clauses linked by the morphologically negative conjunctions ne in Old and Early Middle English, and nor in the more recent history of the language. The main changes to sentence negation discussed by a large number of commentators since Jespersen (1917) are widely known, and feature in most textbook treatments of English diachronic linguistics, but no attempt has to our knowledge been made to consider coordinate sentential negation within this broader picture. Yet here too there were significant developments which should be of interest to those concerned to elaborate a theoretical account of the evolution of negation in English. As will be shown in this chapter, it was regularly possible until the fourteenth century for a negative conjunct clause to be introduced by ne, even though the first conjunct clause was affirmative, for instance:
Richard Ingham
(1) Ah we schule sechen efter wrake on alle þeo þt te biwiteð, ‘But we shall seek vengeance on all those that protect you, ne ne schulen ha neauer beo sker of ure weorre. and they they shall never be quit of our war.’ (Helsinki Corpus, St. Juliana Bodl. 34 51,11)
We shall refer to this as Affirmative + Negative coordination (abbreviated as Aff.+Neg. coordination). By the end of the medieval period such constructions are no longer found, leaving only cases where the clause introduced by ne or nor (variably spelt ner) is conjoined with a preceding negative clause:
(2) The seid Paston hath no certeyn informacion therof, ner non cam to his handes. (Paston Vol I, 109)
Here, as in (1), the coordinator ner does not suffice to negate the second conjunct, which contains a negative element (non), producing negative concord. Negative concord (henceforth NC) lasts into the seventeenth century in coordinate constructions, such as (2), in educated written English (Nevalainen & Raumolin-Brunberg 2003: 91). In the meanwhile, during the sixteenth century, it became possible, and then normal in Modern Standard English, for the negative coordinator nor to introduce negation to a clause having no other negative element:1
(3) Ye have not hearde his voyce at eny tyme, nor ye have sene his shape. (Helsinki Corpus, Tyndale Bible V 39)
These changes in sentential coordination can be summarized descriptively as an evolution from the stages presented in (4) to (6):
(4) Early English
In clausal coordination, a negative coordinator introduces, but need not follow, a clause containing a negative element
(5) Late Middle English and Early Modern English
In clausal coordination, a negative coordinator follows, and also introduces, a clause containing a negative element (6) Modern (Standard) English In clausal coordination, a negative coordinator follows a clause continuing a negative element, and introduces a clause containing no other negative element
The purpose of this chapter is to examine these changes in the light of current theoretical analyses of negative syntax. It will be proposed that the structure of clausal negation changed in two distinct ways between the thirteenth and seventeenth centuries, giving . As shown by Jacobsson (1951), subject–verb inversion after nor normally precluded the appearance of other n-items in the clause introduced by nor. In the 16th century, however, nor could negate a clause by itself without inversion.
Negative co-ordination in the history of English
rise to the successive stages (4) to (6). Both changes involved the licensing of n-items, the theoretical characterization of which is set out in the next section. In the process of upholding this claim, we shall focus on two structural issues in particular. First, what was the relationship between the negative coordinator and the presence or absence of negative concord? The recorded history of English shows n-marked coordinators continuing to co-occur with other formally negative items. Yet their distribution has clearly changed, as witness the differences between (1), (2), and (3). Secondly, what formal element licensed the negative co-ordinators ne and nor at different periods? In the course of seeking answers to these questions, we will uncover further evidence that structural change in negation took place around the fourteenth century with the grammaticalisation of not, as argued by numerous researchers, and in addition around the sixteenth century, a period that has likewise received some attention as regards the syntax of negation. We shall also suggest certain theoretical modifications that seem appropriate to account for the data studied.
2. Theoretical background The structure of negative clauses in generative grammar has offered fertile ground for theoretical controversy in recent years. Following a proposal by Pollock (1989) that, at least in some languages, a NegP is projected, Haegeman (1995) posited that all syntactically negative clauses have a NegP constituent, containing a negative operator in Spec NegP. This applied in English to clauses negated only by a member of the no-series of pronouns/determiners (no, no-one, nothing etc.) conventionally referred to as n-items. Thus the structure of a negative clause such as Mary can have no complaints was taken to be as follows:
(7) [IP Mary can [[NegP NegOP [VP have no complaints ]]]
Lacking the sentence negator not, the NegP constituent is here occupied only by a null element, the negative operator (henceforth NegOP). Identified by the n-item no within the VP, it heads an operator chain with no. In other studies, on the other hand, such as Haegeman & Zanuttini (1991) and Zeijlstra (2004, 2008), NegP was envisaged as absent from the structure of clauses negated only by an n-item: it was present in clause structure only in languages where an n-item is obligatorily accompanied by a negator, as in Slavic languages such as Czech (Zeijlstra 2004: 249ff.): (8) Nikdo ne-vola nikoho. nobody-subj neg-call-3psg-pres nobody-obj ‘Nobody calls anybody.’
Here, the head negator ne as well as the n-items nikdo and nikoho participate in a NC relation. In his analysis of negation in a strict NC language, Zeijlstra recasts Haegeman’s
Richard Ingham
approach in the framework of Minimalist Program feature-checking (Chomsky 2001). A negative operator (NegOP) stands in Spec NegP, bearing an interpretable feature [iNEG]. The n-items are not semantically negative and each bears an uninterpretable feature, requiring to be checked and deleted against the [iNEG] feature on NegOP. The null NegOP is identified by a sentential negator, ne in (8). The relevant features are shown in (8):
(8′) Nikdo[uNEG]
ne-vola[uNEG]
Op[iNEG] nikoho [uNEG]
The [iNEG] NegOP checks/deletes the [uNEG] features in a process referred to as Multiple Agree. In this analysis, following the phase-based derivation of the clause (Chomsky 2001), checking of [uNEG] features takes place under a c-command relationship: the [iNEG] constituent c-commands all [uNEG] constituents. A NegP containing a null operator is added to vP while the latter still contains all the n-items: (8″) [NegP Op[iNEG] Neg0 [vP Nikdo [uNEG] ne-vola[uNEG]
nikoho [uNEG] ]]
At this point in the derivation, the [iNEG]-bearing Neg operator checks/deletes not only the [uNEG] features of the negated items nikdo, nikoho and ne-vola but also that of the negative particle preceding the verb (ne-vola), which raises to Nego to identify the Neg Operator, as indicated below:2 (8″′) [TP Nikdo[uNEG] ne-vola[uNEG] [NegP Op[iNEG] Neg0 ne-vola[uNEG] [vP nikdo[uNEG] ne-vola[uNEG] nikoho[uNEG] ]]]
As in Chomsky (2001), the [uNEG] items are taken to be probes seeking an [iNEG] element as a goal. Usually, as for example in subject–verb agreement, a probe c-commands its goal. In this analysis of NC, however, n-items show an ‘Inverse Agree’ relationship (Roberts 2007), that is, they are c-commanded by their goal, the null NegOP. As Roberts (2007: 72) points out, this could be seen as anomalous, yet, when analysed in this way, n-items show an interesting commonality with polarity items, such as the any series in English, which indisputably need to be c-commanded by their licensing element. Roberts assumes that polarity items have an uninterpretable feature needing to be checked against a c-commanding goal. In feature-checking terms, polarity items and n-items can therefore be given a uniform treatment, in which uninterpretable features always probe upwards. Zeijlstra (2008) explicitly adopts the Inverse Agree formulation. A key aspect of his approach is that, unlike Haegeman (1995), Zeijlstra postulates NegP and a null NegOP only for NC languages. Thus in his analysis the sentence shown in (7) is negated by the inherently negative element no, and there is no NegP projection.
. Movement to Spec TP of the subject nikdo for case checking is also shown.
Negative co-ordination in the history of English
Modern Standard English is a non-NC language, so the absence of a null NegOP correctly predicts that it is unable to check multiple n-items under a single negative reading, as happens in Czech, for example. In the present study, this approach to the representation of negative syntax will be adapted to negative coordination in the successive phases of English set out in (4) to (6). These will be captured in terms of differing featural properties of n-items, and the presence or absence of a null NegOP in NegP. The n-items under discussion are the negative coordinator ne in medieval English, replaced by nor from the fifteenth century onwards; we begin with a brief survey of the featural properties of nor (those of ne are further discussed in Section 4 below). Nor in Modern Standard English has phrasal and clausal coordination uses. It introduces a conjoined phrase as the second member of the neither… nor construction, and also following another negative item such as not or no: (9) a. b. c.
The change won’t be as abrupt as in 1958 nor as severe as in 1959. No state shall have a share less than 50% nor more than 70%. Serious art is not for the lazy, nor for the untrained. (Huddleston & Pullum 2002: 1309)
Accordingly, phrasal coordinator nor may be designated as a [uNEG] item, unlike polarity items, such as any, which have no negative feature of any kind and can be used in affective contexts more generally: (10)
a. b. c. d.
He left without any warning or explanation. *He left without any warning nor explanation. It seems unlikely there’ll be any rain or snow. *It seems unlikely there’ll be any rain nor snow.
Matters are different when nor introduces a clause, following one that is syntactically or semantically negative: (11) a. b. c.
He gave us no idea he was leaving, nor did he explain his decision later. He left without any warning, nor did he explain his decision later. It seems unlikely to rain, nor is it expected to snow.
Here, nor is acceptable not only after an overt negative element no, in (11a), but also in the affective contexts (11b–c), where the first conjunct lacks syntactic negation, as can be tested by its inability to take affirmative tag questions (Haegeman 1995).3 So if clausal nor were taken to bear a [uNEG] feature, the latter would fail to be checked in (11b–c). . It should be added that the conjunction status of non-correlative nor is somewhat debatable. Quirk et al. (1985: 939) point out that nor can alternatively be seen as an adverb co-occurring with asyndetic (null) coordination. In what follows we shall accordingly refer to non-correlative nor as a clause coordinator, without seeking to finalize its status as a conjunction or as an adverb.
Richard Ingham
Clausal nor clearly introduces negation to the second conjunct clause in each case, so we assume it bears a semantically interpretable NEG feature, that is [iNEG]. Two values of nor should therefore be posited, one with a [uNEG] and one with an [iNEG] feature. As will be seen below in Section 3, this would not be an unexpected state of affairs, given the variable behaviour of n-items in earlier stages of English. Before definitively adopting this conclusion, however, another possibility consistent with Zeijlstra’s (2004) framework must be addressed. Conceivably, in the examples in (11) above, nor could be acting as an identifier for an [iNEG] operator in the second conjunct clause. If so, nor would not itself be [iNEG], or else a double negation reading would be produced. By hypothesis, it would bear a [uNEG] feature serving to identify a null [iNEG] operator responsible for checking the [uNEG] feature posited for nor. Now this combination of a [uNEG] overt identifier and a null NegOP is, as we saw, used in the present framework to give an account of Negative Concord. But nor in Modern English comes to disallow NC in the clause it introduces, so to propose that a [uNEG] nor identifies a null NegOP here would make entirely the wrong prediction. What is needed is an analysis which predicts the absence of NC, and in which nor itself negates the second conjunct; in order to do so, clausal nor must bear an [iNEG] feature. Ahn (2003: 174) likewise considers that nor in negative clause coordination is inherently negative, rejecting the notion of a null NegOP distinct from clausal nor on the grounds that, when nor introduces a nominal conjunct, “it does not lose its negative force”. We shall adopt Ahn’s characterisation of nor, specified in the formal analysis presented here as [iNEG]. It should be added that, although clausal nor is not in a syntactic feature-checking relationship, it still has to be licensed by a c-commanding semantically negative element, as in (12b–c). In this respect, its behaviour is that of a polarity item such as any. To summarize this part of the discussion, we have argued that Present-day English clausal coordinator nor is an [iNEG] element, whereas phrasal nor is [uNEG]. In what follows we shall have nothing further to say about phrasal coordination, and will be concerned only with clausal nor, seeking to explicate diachronically its syntactic behaviour and that of its predecessor ne from the Old English period onwards. We wish to consider how far the analysis of changes in ordinary clausal negation can handle changes in coordinate negation, or whether a special analysis of the latter is required.
3. Negation in pre-Modern English The main developments in the syntax of negation in pre-Modern English have been related to the role of NegP from Pollock (1989) onwards (see also Beukema & Tomić 1996; Rowlett 1998; Willis 2004; and Giannakidou 2006). Originally an optional reinforcing negator, not comes to be a necessary and sufficient element in the expression of English negation, and can thus be assigned an [iNEG] feature. This change was
Negative co-ordination in the history of English
essentially complete by the fifteenth century. Old and Early Middle English (henceforth OE and EME respectively) had a head negator ne which we assume served to identify an [iNEG] null operator, as in Zeijlstra’s analysis of Slavic languages such as Czech. As regards negative clauses with indefinites, there were two Negative Concord phases in medieval English: in OE and EME, n-items such as na ‘no’ and nan ‘no/none’ normally co-occurred with the negator ne: (12) Þe feont ne mei neden nan mon to na sunne. ‘The devil may not compel any man to any sin.’ Helsinki Corpus, Ancrene Riwle II.224.3246)
After the demise of the head negator ne, which occurred in the fourteenth century (Frisch 1997; Ingham 2005), the null NegOP could no longer be identified and was lost. N-words such as no now came to bear [iNEG] features, appearing alone in negative clauses: (13) She kowth haf no money as for þe ffyndyng of þe Offyces
(Stonor 58 (1462))
NC, however, did not disappear: n-items optionally co-occurred with the negator not, and could co-occur with each other. (14) a. b.
It can not come better at no tyme ….that he talkyd neuer wyth no man in no mater
(Stonor 168,17) (Paston D 89,27 (a.1450))
Evidently, as is made particularly clear by comparing (13) and (14a), n-items had dual status as regards their inherent negative properties. In (13), no introduces a negative reading, whereas in (14a) it must not. To suppose that in this clearly transitional period there were two lexical entries for an n-item, as does Martins (2000) for indefinites in Romance undergoing change in the later medieval period and beyond, appears to be a valid approach to the Late Middle English (LME) data illustrated here. A related development concerns negative concord across a clause boundary. In EME, as shown by Ingham (2003), NC is strictly avoided between a matrix clause and an embedded clause, for instance: (15) Nule ich naut þat ani seo ow bute leaue habbe ‘I do not wish anyone to see you unless he has leave of ouwer special meister of your special master.’
(Helsinki Corpus, Ancr. Riw, II.47.438)
In the fifteenth century, cases appear which violate the EME clause-boundedness constraint on NC:4
. It may be noted that the same construction is found in modern NC varieties of nonstandard English, for example (Coles 1998):
(i)
That boy don’t think he done nothing wrong.
Richard Ingham
(16) a. b.
I trowe no man can thynk that this treté is to no good purpose (Paston D. 39, 51) Ther wos none that bad me to no porpos (Helsinki Corpus, Cely 55, 23)
It was in this period that not took over the function of clausal negator, and clausal negation by ne alone became moribund (Frisch 1997). A further development at this time was that object–verb order in auxiliated clauses, which until the fourteenth century had been common with objects of any type, became limited essentially to negated objects, for instance: (17) I may no leysour haue
(Paston Vol. I, 48)
The negated object followed the finite auxiliary and preceded the non-finite main verb, in complementary distribution with not. For this reason, Ingham (1998, 2000) analysed such constructions in terms of Neg Movement, that is the movement of negated objects into Spec NegP, a position barred to ordinary objects.5 If a shifted object occupied Spec NegP, this position was not available to not. The changing ways in which these properties cluster can be economically accounted for by a structural evolution involving NegP, from a structure featuring a null NegOP in NegP, via a structure where NegP could be present but filled with lexical material only, either not or a shifted negative XP as in (17). Subsequently a further change took place such that NegP in negative clauses with n-items (Ingham 2007a, b) was eliminated. These successive stages are shown below, taking the structure at a point in the derivation of a clause comparable to Zeijlstra’s analysis in (8) above, for ease of comparison. Note that we have used examples featuring auxiliated clauses, which we have assumed involve recursive vP structures: the precise nature of the analysis of auxiliaries is not directly relevant to the point being illustrated here, however. (18) a. b. c.
[NegP NegOP [vP ne mei [vP þe feont neden nan mon…]]] [NegP not [vP it can [vP come better at no time ]]] [vP it can [vP get no better…]]
Examples (18a) and (18b) show the presence of NegP, occupied by a null NegOP in EME, a lexical negator in LME; (18c) shows the absence of NegP in Early Modern English, in which we assume following Zeijlstra that the n-item has become an inherently negative element.
. In OE and EME all types of objects could be scrambled into a linear position embraciated by a finite auxiliary and the main verb (Pintzuk 2005). No special position for negated objects is proposed in that study.
Negative co-ordination in the history of English
In the first phase, the analysis accords with Zeijlstra (2004, 2008): in OE/EME, a NegOP occupied Spec NegP, and cross-clausal NC was not possible, since checking by a NegOP is local. Furthermore, Neg Movement could not take place because a null NegOP occupied Spec NegP. The same can be said of the third phase: in Modern Standard English, where NegP is absent in clauses with n-items, NC is ungrammatical, and negated objects cannot be displaced to an embraciated position. However, in the second phase, spanning roughly the later fourteenth to the midsixteenth century, conflicting properties existed side by side, in terms of Zeijlstra (2004, 2008). Although the clausal negator ne, which identified a null NegOP in EME, was lost, NC continued, involving [uNEG] items. We must therefore depart here from Zeijlstra’s approach in which [uNEG] must be checked by a NegOP. Instead, in the absence of a null NegOp, we posit that a [uNEG] feature may be checked under a c-commanding relation by a preceding [iNEG] item. In effect, [uNEG] items are handled as polarity items, which is why cross-clausal NC as in (16) became possible in this period. The analysis presented above has distinguished three successive stages in the syntax of ordinary clausal negation in the history of English. Up to the fourteenth century [uNEG] items required to be licensed in a local relationship; though not one of c-command. Subsequently, for a period of about 200 years, they needed to be c-commanded by a negative element, and could now be non-locally licensed. Thereafter, with the demise of NC in Standard English, [uNEG] items generally disappeared. These three stages will in the remainder of the study be shown to be relevant to the evolution of coordinate negative clauses as well. In the first stage, lasting until about the fourteenth century, the negative coordinator’s [uNEG] feature is checked by NegOP, and so it does not require the preceding conjunct to be negative; in the second, lasting into the seventeenth century, it continues to require the clause it introduces to contain an [iNEG] element, but its [uNEG] feature is now checked by a c-commanding [iNEG] element; in the final stage, Modern Standard English from the later seventeenth century onwards, the negative coordinator is no longer a [uNEG] but an [iNEG] element. The chronology suggested above should be seen as only an approximate idealization, since at certain periods in particular both the older and the newer grammars would have been in use at the same time, with the usual stylistic and sociolinguistic variation observable in data sources. In the next section the main developments in the clause types associated with these changing grammars are presented.
4. Diachronic trends in English coordinate negation No published study of clausal coordinator ne using diachronic corpus resources has to our knowledge been carried out, and it was not the aim of the present chapter to
Richard Ingham
conduct a corpus study.6 In order to get a rough indication of change in progress, nevertheless, some kind of assessment of empirical developments was felt necessary. Accordingly, frequencies of sentential negative coordination with ne and nor were identified visually in prose texts found in four electronic resources, the Helsinki Corpus (Kytö 1991) for OE, EME and Early Modern English as far as 1640, supplemented for the thirteenth century by the online Ancrene Riwle, for the fifteenth century by the online Paston Letters, Volume 1 (Paston), and for the sixteenth century by the relevant files of the Corpus of Early English Correspondence Sampler (CEECS). Only finite clauses with an overt subject, as in examples (1)–(3), were selected for this purpose. The results, totalling 145 instances, are displayed in the table in the Appendix. As observed by Andrew (1940) and Mitchell (1985), the affirmative-negative coordination pattern shown in (1) existed already in Old English (our examples): (19) a.
Ealle þas þinc synd fram eagan gewitene, ne hit nu nan all these things are from eyes witnessed, nor it now no
þinc þære sawle ne helpeð. thing to-the soul neg helps.
‘All these things are eyewitnessed, and it will not now help the soul at all.’ (Helsinki Corpus, Homily Epiphany, p. 165)
b.
Hwylc bið he þonne buton swylce stan, oþþe treow? Ne he what is he thon except such.as stone or tree nor he
hine na ne onstyreþ, syþþan seo ungesynelice sawl it never neg rouses, since the invisible soul
him of biþ. him away is
‘What will he be then but as a stone or a tree? And he will not rouse it [the soul] because the invisible soul has left him.’ (Helsinki Corpus, Blickling Homilies, p. 21)
Andrew claimed that this construction had a particular discourse value, of enhancing a contrast between the affirmative first and the negative second conjunct. Mitchell disagreed, stating that coordinator ne here meant no more than ‘and… not’. Although he did not produce supporting examples, cases which clearly favour Mitchell’s position may readily be found, for instance:
. It should be noted that the search apparatus of the available historical parsed corpora such as PPCME2 does not currently make it feasible to identify coordinator ne in the context of a preceding affirmative conjunct, a key issue in this study.
Negative co-ordination in the history of English
(20) His metes gereord he ana underfo, (þam gemete and on þære tide, þe se abbod understande, þe him gebyrige.) Ne he ne sy gebletsod fram nanum þara, þe… ‘Let him take his food alone… And he may not be blessed by any of them, who…’ (Helsinki Corpus, Benedictine Rule, p. 50)
Here, it is clearly stated that an offender will suffer a number of penalties, of which the last, expressed as a negative, adds to rather than contrasting with the previous provisions. This upholds Mitchell’s view of such cases as an ordinary co-ordinate construction in which the second conjunct is introduced by ne, rather than and. The presence of ne conveys nothing more than and would in its place; indeed negative conjuncts could also be introduced by and, in Old and Early Middle English, for example (21a), as well as in LME, for example (21b): (21) a.
Nu ic soðlice þe to sprece and for.þi ne mæg ic now I truly thee to speak and so neg may I
na læng beon. neg long be
‘Now I speak truly to you and so I shall not be long.’ (Helsinki Corpus, OE St Margaret, p. 177)
b.
Let meknesse be ioyned to chastete, and noþyng ‘Let meekness be joined to chastity and nothing
schal be bryŠtere. shall be brighter.’ (Helsinki Corpus, AELR 33)
However, as can be seen in the Appendix, the EME period is when Aff. + Neg. coordination becomes very common. Inversion in negative main clauses was optional at this time (Ingham 2005), and the Aff. + Neg. construction is indifferent to the internal structure of the second conjunct, being found with both inverted and uninverted negative clauses: (22) a. Ant he haueð iþolet us þe þolemode lauerd ne we nusten hwet we duden. ‘And the merciful Lord has been merciful to us, and we did not know what we did.’ (Helsinki Corpus, St Katherine, p. 42) b. Hwenne & hwuch time þe þeof walde cume to his hus; he walde wakien. ne nalde he nawt þolien þe þeof forte breoken hire. ‘At the time when the thief would come to his house, he would wake and would not allow the thief to break into it.’ (Helsinki Corpus, Sawles Warde, p. 166)
Ne therefore clearly functioned as a conjunction linking CP clauses, and had no adverbial properties, as was suggested for nor in Modern English by Quirk et al. (1985), see footnote 3.
Richard Ingham
In the EME data we obtained, such cases as (22) actually outnumbered those of Neg. + Neg. coordination, possibly a rather surprising outcome, and certainly the reverse of what was found for earlier and later periods. Let us recognize that the EME textual record is heavily skewed towards particular types of prose text, chiefly religious prose from the West Midlands, and may not be representative of the language more generally. There is also the issue of potential contact influence from Old French, in which the coordinator ne was commonly used in Aff. + Neg. contexts, for instance: (23) Il fu mout angwyssous e juna troys jurs e troys nous, he was much afraid and fasted three days and three nights, ne ne manja, ne ne boust ne ne dormy. nor neg ate nor neg drank nor neg slept ‘He was very afraid and fasted 3 days and 3 nights, and did not eat, drink or sleep.’ (Pseudo-Turpin, 72, (c. 1215))
In this example, taken from an Anglo-Norman text version written in England, all the negative clauses beginning ne manja are conjoined to the clause e juna troys jurs e troys nous by the conjunction ne. Old French was not a negative concord language (Roberts 2000), so the licensing of coordinator ne in that language was not done via a null NegOP. Coordinator ne was also licensed in affective contexts such as interrogatives and conditional sentences. The grammars of Old French and Middle English thus differed in terms of their formal analysis, but the surface patterns were close enough to make contact influence highly plausible. Since examples can be found in Old English, as we have seen, there is no suggestion that Aff.+ Neg. coordination was of French origin. However, as is common in contact situations such as that of England after 1066, the presence of extensive bilingualism amongst the educated classes may have had the effect of increasing the frequency of the phenomenon.7 Although numbers in our sample data were inevitably very limited, the overall results show a clear trend from the fourteenth century onwards towards eliminating the Aff. + Neg. coordination construction. It seems to have disappeared in the sixteenth century, so a satisfactory account of it should invoke a factor that remained constant from the tenth to the fourteenth centuries, but then changed. The conjunction used in all the pre-fourteenth century data in Appendix Table 1 was ne. It might be thought that the disappearance of Aff. + Neg. coordination is thus a product of the loss of coordinator ne, but ne is not the only coordinator form found here, for instance:
. It may in particular have influenced Chaucer’s translations of the French texts Boece and Melibee which appear in electronic sources as among the main prose works of the late 14th century.
Negative co-ordination in the history of English
(24) a. For I sawe trulye that god dothe alle thynge, be itt nevere so litille; nor nathynge es done be happe ne be eventure. (Helsinki Corpus, Julian of Norwich p.49) b.
The best of alle and the fayrest ys cleymyd, ner yt ys not in hys jnventory. (Paston Vol. I, 373)
In the next section a proposal is put forward in terms of the changing structure of English sentential negation as analysed above.
5. Analysis of affirmative + negative coordination Let us now consider the formal structure of the negated second conjunct in OE/EME, and how it changed in LME and ENE. We begin by taking the example of an OE Aff. + Neg. coordination clause such as Ne he ne-sy gebletsod fram nanum þara in (20) above. Following the approach of Zeijlstra (2004, 2008), the constituents of the clause are first merged, and appropriate movement of constituents to functional positions within that clause takes place, viz.: (25) [TP he T ne-sy [NegP Op[iNEG] ne-sy [vP he ne-sy gebletsod fram nanum[uNEG] þara ]]]
Now the negative coordinator is merged (checked elements within the conjunct clause removed for ease of presentation): (25′) [ConjP ne[uNEG] [TP he T ne-sy [NegP Op[iNEG] [vP gebletsod fram nanum þara]]]]
According to Zeijlstra’s (2008) account of NC, a [uNEG] feature is checked by the [iNEG] null Operator, by Inverse Agree, since [uNEG] features on constituents merged in vP probe upwards for a Goal, the NegOP which c-commands them. But in a negative coordinate clause, as in (25´), the [uNEG] feature on the coordinator ne is not commanded by the NegOP. Unlike other [uNEG] items, the coordinator ne does not originate in the vP at a point where NegOP can c-command it. On almost any structural account of clausal coordination, coordinator ne c-commands the NegOP, since the maximal projection TP of the coordinate clause containing the NegOP does not dominate the coordinator itself. For concreteness, we illustrate this using the Johannessen (1998) approach to clausal coordination, in which the second conjunct is the complement of the conjunction head, while the first conjunct stands in its specifier: (26)
ConjP (first conjunct)
Conj′ Conj
CP
ne[uNEG]
…NegOp[iNEG]…
Richard Ingham
Consequently, the licensing of the negative coordinator, though actually conforming to the standard Probe-Goal relationship of probing downwards, is out of line with the checking of [uNEG] features adopted above, following Zeijlstra and Roberts. Naturally, where a negative first conjunct c-commands ne introducing the second conjunct, an [iNEG] element in the latter would in principle be able to act as the Goal checking the [uNEG] feature of coordinator ne, by Inverse Agree. But this will not handle the Aff. + Neg. coordination construction, where coordinator ne’s [uNEG] Probe will find no c-commanding [iNEG] element to act as a Goal. Thus to account for the licensing of the coordinator ne in an Aff. + Neg. coordination construction, only the downwards Probe-Goal operation can work. As clarified by Roberts (2007), Zeijlstra’s analysis has shown a requirement for NC to operate on an Inverse Agree basis; the present analysis suggests that [uNEG] features as a whole need to be handled in terms of both Agree and Inverse Agree. For approaches based on Chomsky (2001), Agree operates within a local domain: the checking relationship is between a functional head and its complement. The same will be true of the bi-directional Agree operation proposed here. In ordinary NC, involving Inverse Agree, as analysed in (8), the n-items stand within vP, the complement of Neg. In Agree with coordinate negation, as in (26) the conjunction and the CP stand in the same local relationship. It can be seen as a functional complex equivalent to the T + vP complex that figures as the local domain in the Minimalist paradigm case of Agree, subject–verb agreement. Thus with Aff. + Neg. coordination, the [uNEG] feature on the Conj functional head acts a Probe, and the [iNEG] Operator within the checking domain (CP) is the Goal. From this perspective, the intuitive sense that the negative coordinator is agreeing with the polarity of the second conjunct receives a natural explanation. Given the assumption in Chomsky (2001) that CP forms a completed phase within which checking takes place, however, a further issue needs to be considered. Once uninterpretable features within CP have been checked, CP is no longer accessible to syntactic computation, so an [iNEG] feature within CP could not form a Goal allowing checking of the [uNEG] conjunction ne in (26). Second conjuncts in Aff. + Neg. conjunction in Early Middle English could involve verb-second clauses, as in (22b) above, in which the finite verb moved to C, so the construction indisputably involved a CP as a second conjunct. Does this mean that the coordinator ne could not be within the CP phase? Such a strict formulation of the phase notion may be over-strong. The fact that ne in pre-Modern English introduces a negative clause but never an affirmative clause seems to require that CP remain open to syntactic computation. That is, Conj and its CP complement form a functional complex within which checking may be effected. Though descriptively adequate for present purposes, this analysis would be strengthened if comparable cases of conjunction CP checking can be identified, which we leave to future research.
Negative co-ordination in the history of English
It is proposed, then, that in OE/EME a [uNEG] conjunction was checked by a null NegOP under Agree, and that this mechanism was lost in LME, with the demise of the null operator. Licensing of [uNEG] items now shifted to one of licensing under c-command by a structurally higher [iNEG] element. This eliminated Aff. + Neg. coordination, since negative second conjuncts in the new grammatical system had to be c-commanded by a negative element in a higher clause, viz.: ConjP
(27)
Conj′
TP[iNEG]
Conj
TP[iNEG]
ne/nor[uNEG]
…
Were the first conjunct to be affirmative, there would be no [iNEG] element in the first conjunct to c-command the [uNEG] conjunction introducing the second conjunct. However, nothing prevents a [uNEG] conjunction from introducing an overtly negative clause, since a [uNEG] element does not introduce negation. Hence we continue to find examples like (2) long after the null NEG operator was lost. The final stage in the process occurred in Early Modern English, when [iNEG] nor appeared, as in cases such as (3). Since all other English n-items, such as no, as discussed above, had by then acquired [iNEG] entries, this development can be seen as a very natural analogical extension within the repertoire of negative elements in English. In the fifteenth century none, never, etc. were [uNEG], as we have seen, yet can be found without another negative element in the clause, while nor was [uNEG] but required the presence of another negative element. In the sixteenth century, the behaviour of none and never was extended to nor when the latter became able to appear in surface patterns lacking another negator, giving the alternation observed in Early Modern English between [iNEG] nor, which prohibited other [iNEG] elements in the conjunct clause it introduced, and [uNEG] nor, which permitted them. Because nor is now [iNEG], it should in principle be able to appear after an affirmative first conjunct, and this is indeed occasionally observed, though in the data this study examined it does not show up until the seventeenth century: (28) Ther is upon the way towards your Lordship a whole kennall of houndes; five cople of them are for me, nor was I ever maister of soe many before in all my life. (CEECS, WENTWORTH, p. 2 (1632))
Why this construction has not achieved the productivity that Aff. + Neg. coordination had in medieval English is not clear, and is an issue that needs greater attention on the basis of a large corpus of Late Modern English. For present purposes, it seems that for about two centuries before the date of example (28) the combination of single negation
Richard Ingham
and a preceding affirmative conjunct, such as five cople of them are for me in (28), was not found. During this phase of development of the grammar of negation, the negative coordinator was obligatorily licensed by a preceding negative element. In Old and Middle English, a [uNEG] clause coordinator was licensed iff it stood in a functional complex including a Null NegOP, that is: (29) [ConjP [Conj[uNEG]] [CP … NegOp… ne …]].
When [uNEG] could no longer be checked by a NegOP, it then needed to be licensed by a c-commanding n-item, so the first conjunct was now required to be negative. Our account of the change in negative coordination in terms of the loss of the null negative operator thus seems consistent with the diachronic evidence.8
6. Summary Affirmative + Negative Coordination of clauses, found in earlier stages of English, died out in the sixteenth century, around the time when clausal coordinator nor first began to negate a second conjunct by itself. We have proposed that the process took place in three stages rather than two. Until Late Middle English, the negative coordinator was a [uNEG] item licensed by being within a functional complex including NegOP. In a transitional stage including the fifteenth century, a [uNEG] coordinator continued to be licensed, but now by a c-commanding [iNEG] Goal. At that point, with the loss of [uNEG] licensing by null NegOP, the ability of a [uNEG] coordinator to follow an affirmative first conjunct disappeared. Finally, in Modern English, at least in the standard variety, the [uNEG] clausal coordinator was eliminated from negative clause-introducing contexts, possibly under normative pressure, and [iNEG] nor became regular. These developments in the syntax of negative coordinate clauses have been analysed in terms of changes in the structure of negative clauses, and of the entries of n-items, that are already motivated for non-coordinate negation. The loss of the null NegOP meant
. The correlative structure neither…ne or neither…nor appears in Middle English; though neither is found first as an adverb accompanying the first ne of the EME correlative pair ne… ne, according to the Middle English Dictionary (Kurath et al. 1954–2001): (i)
Do hyne into þan huse, þe beo nærþer ne to hæt ne to ceald. ‘Put it into the house that is neither too hot nor too cold.’ (PDidax 31/20 (c.1150))
(ii)
Ne sweriŠeð, naiðer ne be heuene ne be ierðe, ne bie nan oðer ðing. ‘Do not swear, neither by heaven nor earth nor any other thing.’ (Vices & V. 9/12 (c.1225))
Negative co-ordination in the history of English
that Aff. + Neg. coordination was no longer a possible structure. The later reanalysis of nor as an [iNEG] item conformed to the lexical reanalysis of other n-items as [iNEG], so that NC in coordinate clauses was no longer licensed. Our proposal has brought negative coordination within an explanatory framework for the structure of negation as a whole, using the theoretical concepts developed especially by Zeijlstra (2004, 2008) and Roberts (2007), though with certain specific modifications of detail. Numerous issues are raised that must be addressed in detail in further work, such as the structure of coordinate clauses, the possible existence of counterparts to bi-directional [uNEG] licensing with other elements, the development of negative coordination in non-Standard English, and the structure of, and motivation for, Negative Inversion after nor in Modern English. What we hope to have made clear, however, are broad outlines of the diachronic developments regarding coordinate clausal negation in English, and of how the formal analysis employed here can shed light on their evolution. Appendix. Full clause coordination with a negative coordinator (ne or nor) Neg. + Neg.
Aff. + Neg
Total
OE (9th–11th centuries) EME (13th century) LME (14th century) LME (15th century) ENE (1500–1640)* + NC – NC
23 14 10 25 27 13 14
6 27 8 2 3 0 3
29 41 18 27 30
Total
99
46
145
*In this period, unlike previously, the negative coordinator may appear with no other negative element in the second conjunct clause (-NC). At all other periods, in addition to the negative coordinator, the second conjunct clause must contain a negative element (+NC).
Primary sources Ancrene Riwle: Ancrene Wisse, ed. R. Hasenfratz. Kalamazoo, Michigan: Medieval Institute Publications, 2000, http://www.lib.rochester.edu/camelot/teams/awintro.htm. CEECS: Corpus of Early English Correspondence Sampler, ed. A. Nurmi. Helsinki: Department of English, University of Helsinki, 1998. The ICAME Corpus Collection on CD-ROM, version 2, 1999. Helsinki Corpus: The Helsinki Corpus of English Texts (Diachronic Part), compiled by M. Kytö. Helsinki: Department of English, University of Helsinki, 1991. The ICAME Corpus Collection on CD-ROM, version 2, 1999. Paston: Letters and Papers of the Paston family, ed. N. Davis. Oxford: Oxford University Press, 1971. http://etext.lib.virginia.edu/toc/modeng/public/PasLett.html. Pseudo-Turpin: The Anglo-Norman Pseudo-Turpin Chronicle of Williame de Briane, ed. I. Short. Oxford: Anglo-Norman Text Society, 1973.
Richard Ingham
References Ahn, Sung-ho. 2003. Coordination: A case for multiple categorial projection. Studies in Generative Grammar 13(1): 171–177. Andrew, Samuel Ogden. 1940. Syntax and Style in Old English. Cambridge: CUP. Beukema, Frits & Tomić, Olga Mišeska. 1996. Negation in English: A diachronic view. Folia Linguistica Historica 16(1–2): 123–135. Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, Michael Kenstowicz (ed.), 1–52. Cambridge MA: The MIT Press. Coles, D’Jaris Renee. 1998. Barrier Constraints on Negative Concord in African-American English. Ph.D. dissertation, University of Massachusetts, Amherst. Frisch, Stefan. 1997. The change in negation in Middle English: A NEGP licensing account. Lingua 101: 21–64. Giannakidou, Anastasia. 2006. N-words and negative concord. In The Blackwell Companion to Syntax, Martin Everaert & Henk van Riemsdijk (eds), 327–391. Oxford: Blackwell. Haegeman, Liliane. 1995. The Syntax of Negation. Cambridge: CUP. Haegeman, Liliane & Zanuttini, Raffaella. 1991. Negative heads and the Neg criterion. Linguistic Review 8: 233–251. Huddleston, Rodney & Pullum, Geoffrey. 2002. The Cambridge Grammar of the English Language. Cambridge: CUP. Ingham, Richard. 1998. Negation and OV order in Late Middle English. Paper presented at the Symposium on Negation, Syntax and Semantics, University of Salford, November 1998. Ingham, Richard. 2000. Negation and OV order in Late Middle English. Journal of Linguistics 36: 1–38. Ingham, Richard. 2003. Negative concord in Middle English: A canonical agreement relation. Paper presented at the Linguistics Association of Great Britain, University of Oxford, September 2003. Ingham, Richard. 2005. The loss of Neg V → C in Middle English. Linguistische Berichte 202: 171–206. Ingham, Richard. 2007a. A structural constraint on multiple negation in Late Middle and Early Modern English. Medieval English Mirror 3: 55–67. Ingham, Richard. 2007b. NegP and negated constituent movement in the history of English. Transactions of the Philological Society 105 (3): 1–33. Jacobsson, Bengt. 1951. Inversion in English, with special reference to the Early Modern English Period. Uppsala: Almqvist & Wiksell. Jespersen, Otto. 1917. Negation in English and other Languages. Copenhagen: Hoest & Son. Johannessen, Janne Bondi. 1998. Coordination [Oxford Studies in Comparative Syntax]. Oxford: OUP. Kurath, Hans, Kuhn, Sherman M. & Lewis, Robert E. (eds.). 1954–2001. Middle English Dictionary. Ann Arbor MI: University of Michigan Press. Kytö, M. 1991. Manual to the Diachronic Part of the Helsinki Corpus of English Texts: Coding Conventions and Lists of Source Texts. Helsinki: Department of English, University of Helsinki. Martins, Anna-Maria. 2000. Polarity items in Romance: Underspecification and lexical change. In Diachronic Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 191–210. Oxford: OUP. Mitchell, Bruce. 1985. Old English Syntax, 2 Vols. Oxford: Clarendon Press.
Negative co-ordination in the history of English
Nevalainen, Terttu & Raumolin-Brunberg, Helena. 2003. Historical Sociolinguistics: Language Change in Tudor and Stuart England. London: Longman. Pintzuk, Susan. 2005. Arguments against a Universal Base: Evidence from Old English. English Language and Linguistics 9: 115–138. Pollock, Jean-Yves. 1989. Verb movement, UG, and the structure of IP. Linguistic Inquiry 20: 365–424. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A Comprehensive Grammar of the English Language. London: Longman. Roberts, Ian. 2000. Some remarks on the diachrony of French negation. Revista de Documentação de Estudos em Lingüística Teórica e Aplicada (D.E.L.T.A.) 16: 201–219. Roberts, Ian. 2007. Diachronic Syntax [Oxford Textbooks in Linguistics]. Oxford: OUP. Rowlett, Paul. 1998. Sentential Negation in French. Oxford: OUP. Willis, David. 2004. Formal approaches to Jespersen’s cycle. Ms, Dept of Linguistics, University of Cambridge. Zeijlstra, Hedzer Hugo. 2004. Sentential negation and negative concord. Unpublished Ph.D. dissertation, University of Amsterdam. Zeijlstra, Hedzer Hugo. 2008. Negative Concord is syntactic Agreement. Ms, lingBuzz/000645.
Formal features and the development of the Spanish D-system* Masataka Ishikawa Hiroshima University
This chapter studies the evolution of the article in Spanish within the framework of the parametric approach to language (Chomsky 1999, 2000). Based on the chronology of the extension of the definite article in different syntactic contexts in its history, the chapter extends the diachronic analysis of the structure of the Spanish DP (Ishikawa 2001) attributing the (crucial) difference among the older and modern D-systems to the difference in the phase level. The chapter concludes by discussing implications of the pattern of syntactic change suggested here for the feature-driven approach to language history.
1. Introduction According to the parametric approach to language, diachronic differences among grammars are located in the difference in the value of parameters permitted by Universal Grammar (UG). Following this idea, the present chapter deals with the question of how particular (patterns of) syntactic change may be characterized in terms of parameterized options by examining the history of the Spanish D(eterminer)-system in relation to the (definite) article. The chapter is organized as follows. Section 2 summarizes basic differences between Old Spanish (OS) and Modern Spanish (MS) with respect to the use of (definite) articles. Within the scope of this chapter, I will focus on the use of articles mainly with unmodified nominals. Section 3 identifies two major structural changes that took place in the history of the Spanish D-system. In doing so, I discuss how the notion of
*Due to space limitations, only a limited number of representative examples are included here. I thank two anonymous reviewers and Chris Lucas for their comments and suggestions on an earlier version of this chapter. All remaining shortcomings are mine.
Masataka Ishikawa
‘phase’ may be applied to the analysis of language change.1 Section 4 presents concluding remarks.
1.1 Assumptions As a general framework, I adopt the view that lexical and functional elements in the lexicon are made up of semantic, phonological, and formal features (FFs). Furthermore, FFs are either interpretable or uninterpretable at the two interface leves, PF and LF. According to the standard minimalist assumption, (linguistic) expressions generated by the narrow syntax need to be ‘legible’ to PF and LF (Chomsky 1998b, 1999, Giusti 2002). Uninterpretable FFs need to be eliminated (by checking) before the derivation reaches the interface levels for convergence (Chomsky 1995, 1998a, 2000; Johnson & Lappin 1997; Uriagereka 1999 and others). The narrow syntax contains two basic operations, Merge and Move (Chomsky 1998b).2 Spell-Out (at which portions of the derivation are sent off to the PF side) can apply recursively during the syntactic derivation (Chomsky 1999: 9).
2. Some characteristics of the article in Old and Modern Spanish As is well-known, Latin had no articles, while Romance languages do. In Spanish (and in Romance in general) the definite article (and third person pronouns) developed out of Latin demonstratives ILLE, ILLA, ILLUD through a series of functional (semantic weakening) and lexical (morphological reduction) processes, which in one sense represents a switch from the synthetic system to the analytic one (e.g. Álvarez Martínez 1986 and references cited therein).3 The development in question may be viewed as a change from the lexical element to the functional element. In contrast . In this study, the following “criteria” (among others) for phasehood are adopted (Chomsky 2001, cited by Den Dikken 2007: 27): (i)
a. b.
semantic and phonological independence; ability to host uninterpretable formal features (i.e. act as probes)
. I assume that each step of syntactic movement needs to be motivated by the checking requirement on features obeying the minimality condition (Rizzi 1990; cf. Chomsky 1993). . The morpho(phono)logical development of the definite articles from Latin to Spanish may be summarized (in a schematic fashion) as follows (see Hanssen 1913; Lapesa 1981, 2000, Menéndez Pidal 1982; Lathrop 1984; Cano Aguilar 1988 and Urrutia Cárdenas & Álvarez 1988, among many others): (i)
a. b. c.
ille/illu(m)/illos > el/los (M, SG/PL) ‘(that/those >) the’ illa(m)/illas > la/las/(el) (F, SG/PL) illud > lo (N, the so-called neuter lo (as in lo bueno ‘the good thing/point’))
Formal features and the development of the Spanish D-system
to the (rather complex) lexico-functional adjustments suffered by the demonstratives (in their development from Latin to Spanish), the morphological (gender/number) agreement properties of the definite article are stable and virtually identical in OS and MS (apart from some minor morphemic alternations).4 Yet one finds quite different syntactic behaviors of the (definite) article between the two periods. OS articles often exhibit a non-obligatory character in several syntactic contexts in which their presence is expected in MS. In this section, I will present some of the basic properties of the (definite) article in OS and MS.
2.1 Introduction: Scope of the study This subsection gives a brief introductory sketch of the use of the (definite) article in MS. In Spanish the use of the definite article expresses such ideas as the identified/ accessible “reality” (e.g. locating an entity in time and space), individuation/delimitation or (generic) class, while the lack of an article generally conveys such notions as the “essence”, “virtual state”, lack of specificity or indefinite amount (e.g. Keniston 1937: 221; Alarcos Llorach 1984: 233; Gili Gaya 1985: 242–243; Cano Aguilar 1988: 145–146 and Marcos Marín, Satorre Grau & Viejo Sánchez 1998: 139, among others; cf. Leonetti 1996: 6, 8–10; Laca 1999: 895, 899). In the simplest case (i.e. with single unmodified nouns involved), the following basic (and schematic) distributional patterns of the (definite) article with respect to subject and object may be observed in MS: (1) a. *(La) camioneta ha llegado. the van has arrived ‘The van has arrived.’ b. *(Las) camionetas han llegado. the vans have arrived ‘The vans have arrived.’ c.
Llegó *(la) carta. arrived the letter ‘The letter arrived.’
d. Llegarán (las) cartas. will.arrive the letters ‘(The) letters will arrive.’
(preverbal singular subject)
(preverbal plural subject)
(postverbal singular subject)
(postverbal plural subject)
For the development of Romance definite articles (from ille/ipse), see, e.g. Lausberg 1966; Menéndez Pidal 1982; Harris 1980; Renzi 1993; Posner 1996; Vincent 1997; Lyons 1999; Giusti 2001; Roberts & Roussou 2003 and the references cited. . In Spanish the definite article agrees in gender and number with the head noun; e.g., el perro ‘the.M.SG dog.M.SG’, las casas ‘the.F.PL houses.F.PL’ (MS); el fijo ‘the.M.SG son.M.SG’, las mançanas ‘the.F.PL apples.F.PL’ (OS).
Masataka Ishikawa
e.
Escribirá *(la) carta. will.write the letter ‘He/She will write the letter.’
f.
Recibirá (las) cartas. will.receive the letters ‘He/She will receive (the) letters.’
(singular object of V)
(plural object of V)
Definite/specific subjects and objects take definite articles, while plural objects with an existential interpretation (here intened to convey an indefinite/unspecified amount of objects or material) such as cartas without the definite article in (1d, f)) may lack an article. An articleless simple plural subject may not appear in preverbal position. In other words, (unmodified) preverbal subjects generally require an article. Articleless plural (or mass) subjects may be acceptable in postverbal position with certain (e.g. existential) verbs in some (generally stylistically or pragmatically marked) contexts (Alarcos Llorach 1984: 226; Marcos Marín et al. 1998: 130; Laca 1999: 906–907; Contreras 1996 for an overview of the issues involved; see also Butt & Benjamin 2004: 27–30). In this construction, factors such as contrastive focus in some cases and formation of a [V–(postverbal)subject] semantic unit (rather than that of a predicational relation) in others are involved (see, e.g. De Mello 1980: 558; Laca 1999: 907–908).5 The above illustration shows that the appearance of an articleless singular noun in argument position is most restricted. Articleless singular arguments (without modification) are only allowed under specific (mostly) non-syntactic or extra-linguistic conditions. For example, a singular count noun without an article may be found as a complement of some verbs conveying ‘possession’ or ‘acquisition’ (of a certain class of objects), such as, tener ‘to have’, comprar ‘to buy’ (De Mello 1980: 558, Laca 1999: 919); e.g. ¿Tiene jardín? ‘Does it (e.g. your house) have a garden’; Compró coche (vs. *castillo) ‘He/She bought a car/(castle)’; Se puso corbata ‘He put on a tie’ (see Laca 1999). It has been noted that this usage is related to certain ‘socio-cultural expectation’ (De Mello 1980 and Laca 1999). That is, it is normally expected in a(n ordinary) society that an average individual, for example, owns, buys, or wears one unit of the object in question (e.g. one car, one house, or a tie (at a time), etc.) and the use of the (indefinite) article un(a) ‘a/one’ may be avoided.6 . Coordination (or enumeration) and modification also play a role in this respect: Hombres y mujeres son iguales ante la ley en España (TVE (Spanish TV news), March 8, 2009) ‘Men and women are equal before the law in Spain’ (see, e.g. Company 1991: 404, Note 3; Marcos Marín et al. 1998: 140; Laca 1999: 908; Bosque 1996 for a good overview of the issues involved; cf. Keniston 1937: 239 for the sixteenth-century situation). . Note also such examples as the following involving a telic context; “Consiguió piso en pocos días” ‘He/She found an apartment in a few days’ (with a verb of acquisition (Laca 1999: 904–905)). Determinerless objects are generally not acceptable in telic contexts (as pointed out by Laca 1999: 904). Although the issue of the acceptability of articleless (existential) objects in general
Formal features and the development of the Spanish D-system
In another context, namely in advertisement language, one finds examples such as “Necesita asistenta” ‘Assistant needed’ and Se alquila piso ‘Apartment for rent’. The singular nouns asistenta and piso (and others in this usage) do not refer to an individual(ilzed member), but rather designate (or represent) a(n) (abstract) category (Marcos Marín et al. 1998: 141). Negative contexts with verbs expressing ‘existence’ may also accept bare singular nouns (Bosque 1980: 132–135; Laca 1999: 920); e.g. No tiene novio ‘She doesn’t have a boyfriend’. Negation of the existence of some object/ entity would exclude the identification/individuation or quantification of the object in question, inviting a singular noun (Section 3.1) without an article. Another relevant issue in the presence vs. absence of the article with singular nouns seems to be the following. A casual examination of languages (with articles) reveals that the use of articles varies both within a single language and across languages. Along both axes one finds some uncanny or peculiar (phraseological) variability in the article usage. Observe: (2) a. b. c. d. e. f.
A mediodía/al mediodía (al = a + el) ‘at noon’ (Spanish) En 1998 vs. en el 2000 ‘in (the year) 1998/2000’ (Laca 1999: 921) Fare colazione ‘to have breakfast’ vs. fare il bagno ‘to take a bath’ (Italian) Have breakfast vs. have/take a bath (English) Por primera vez (Spanish) vs. for the first time (English) Ministro de Asuntos Exteriores (Spanish) vs. Ministro degli Esteri (Italian) ‘Foreign Minister’
In examples like the following the nouns in question do not function as (core) arguments: (3)
Reloj de bolsillo ‘pocket watch’ (adjectival) (see also Laca 1999: 913–914) Campo de juego ‘playing ground’ (adjectival) Tener miedo ‘to fear’ (lexical, equivalent to temer ‘to fear’) (Marcos Marín et al. 1998: 141) d. En casa ‘at home’ (adverbial)
a. b. c.
vs. their unacceptability in telic contexts seems to be essentially semantic (rather than syntactic) in nature, the acceptability of the determinerless singular noun piso in the above example may suggest that it (and possibly determinerless singular nouns in the above mentioned usage in general) can function as a kind of “unique” nominal expression (rather than an existential expression). Keniston (1937: 237) observes, commenting on the articleless usage of nouns like casa ‘house’ (and tierra ‘earth/world’) in such expressions as en casa ‘at home’ (in the sixteenth century), that in the usage in question the (determinerless) noun (e.g. casa) “was originally felt as unique and in effect a proper noun” (which might have been a socio-culturally motivated solution to this particular grammatical structure).
Masataka Ishikawa
In adjectival (and probably adverbial) expressions, each lexical word does not have a syntactic autonomy of its own, but the whole phrase functions as a (conceptual) unit.7 These uses do not seem to faithfully reflect the basic syntactic properties of the (singular) noun. An adequate analysis of the above cases would call for a non-syntactic treatment. It is clear that various discourse-semantic factors play an important role (and could override the basic grammatical rules) in the use of articles (given their functional properties). The fact that some additional (linguistic or extralinguistic) factors are involved when sentences or phrases override the general constraints on (e.g. (bare) singular) nouns suggests in my view that they do not directly bear on the most fundamental structural properties of the D-system. Although analyses of both structural and discourse-pragmatic aspects of article usage are essential to an understanding of its overall communicative-linguistic functions, I will deal with those grammatical aspects of article usage which seem to bear on the structural aspects of the change under investigation. Within the scope of the present study, I will focus on article usage with respect to the (required) argumental nominals, which are more directly subject to the (core) syntactic constraints, dealing with nominals of the first of the following types, rather than those of the second; poner la llave a la puerta ‘to lock the door (lit. to put the key to the door)’ (argumental) vs. cerrar la puerta con llave ‘to lock the door (lit. to close the door with (a) key’ (non-argumental/adverbial); Eligieron presidente a Carlos ‘They elected Carlos as president,’ in which presidente has a predicative function (vs They met the president yesterday). In the following subsections, some of the main differences between OS and MS in the use of (definite) articles will be summarized.
2.2 Definite nominals OS differs rather significantly from MS in marking definite/specific nominal expressions. In MS, the definite article normally signals a definite/specific interpretation of a nominal argument involved, while the lack of it indicates the non-referential character of an argument: (4) a.
¿Dónde están *(las) revistas (que estabas leyendo)? where are the magazines which were reading ‘Where are the magazines (you were reading)?’
. Compare curiosidad de niño (de Ana) (= curiosidad infantil (de Ana)) ‘(Ana’s) childlike curiosity’ (adjectival) vs. curiosidad del niño (= curiosidad de él) ‘the child’s curiosity’ (Laca 1999: 913, cf. Lapesa 1996). For cases such as (ir) a casa ‘go home’, (cerrar) con llave ‘to lock (= to close with a key)’ and proverbs, see, e.g. Laca (1999: 922–924 and references cited therein, cf. Note 6).
Formal features and the development of the Spanish D-system
b.
¿Tú también vas a cortarte *(el) pelo? you also go to cut.inf.yourself the hair ‘Are you going to get your hair cut too?
c.
Brillaban muchísimas estrellas en *(el) cielo. shined very.many stars in the sky ‘So many stars were shining in the sky.’
In OS, on the other hand, definite or referential nominals were not necessarily marked by determiner-type elements: (5) a.
Fezist estrelas e luna. made stars and moon ‘You created (the) stars and (the) moon.’
b.
Pastores te glorifficaron, ouieron te a laudare. shepherds you glorified had you to praise ‘The shepherds glorified you (and) praised you.’ (Cid 335, 12th c.; MP 1964: 1037)
c.
El vno es en parayso, ca el otro non entro ala. the one is in Paradise, because the other not entered there ‘One (of them) is in Paradise and (= but) the other did not enter there.’ (Cid 350; MP1964:920)
(Cid 332; MP 1964: 919)
d. Faze aues e bestias en çelo entrar. makes birds and beasts in sky enter.inf ‘It makes birds and beasts enter the (celestial) sky (i.e. Heaven).’ (Alexandre 2393d; MP 1982: 164) e.
Sueles lavar pies e manos. are.in.the.habit.of wash.inf feet and hands ‘You usually wash your feet and hands.’
f.
Ora es venjda. time is come ‘The time has come.’
g.
EN mundo tal cabdal non a como el saber. in world such asset not there.is as the knowledge ‘In the world there is no (other) asset like (i.e. as important as) knowledge.’ (Proverbios Morales 326; F: 125)
(DAV 74; K: 34)
(Planto por Jerusalem 14a; MP 1982: 288)
h. Çiertamente natura ha grant poder. certainly nature has great power ‘Certainly, nature has a great power.’
(GS, lines 14–15; MP 1976: 558)
That is, the absence of a definite article did not necessarily convey a non-referential interpretation of the nominal involved (nor sometimes did its presence signal a specific interpretation of the nominal involved) (see, e.g. Cano Auilar 1988: 146). The
Masataka Ishikawa
non-obligatory use of the definite article with definite (or identifiable) nominals in OS suggests that it was not yet an entirely grammaticalized element, while the obligatory marking of such nominals with the definite article in MS would indicate that the definite article had lost its etymological (deictic) content and become a grammatical element.
2.3 Generics In MS generic nominals (in either subject or object position) require (‘expletive’ definite) articles (Vergnaud & Zubizarreta 1992; Butt & Benjamin 2004; Marcos Marín et al. 1998; Laca 1999, among others; cf. Longobardi 1994): (6) a.
En aquella época, *(la) gente no se preocupaba mucho in that period the people not self worried much de esas cosas. of those things ‘In those days, people did not worry a lot about those things.’
b.
*(El) fino se sirve muy fresco. the dry.sherry self serves very cold ‘Dry sherry is to be served very cold.’
c.
En esta zona abundan *(los) gatos monteses. in this area be.abundant the cats wild ‘This area is abundant in wildcats.’ (adapted from Laca 1999: 896)
In OS, on the other hand, generics could appear without an article (Menéndez Pidal 1954, 1964 and Company 1991, among others): (7) a. Venido es a moros, exido es de cristianos. come is to Moors gone.out is from Christians ‘He has come to (the land of) the Moors, he has left (the land of) the Christians.’ (Cid 566; MP 1964: 1047) b.
Non le llorassen cristianos. (Cid 1295; MP 1964: 1075) not him cry Christians ‘Christians should not cry for him (i.e. over his death).’
c.
Nin da conseio padre a fijo nin fijo a padre. neither gives advice father to son neither son to father ‘A father could not help his son, nor a son his father.’ (Cid 1176; Cano Aguilar 1988: 145)
d.
Et si todas mis rique zas partiere en dar a and if all my wealth divide in give.inf to comer a pobres. eat.inf to poor (people) ‘And if I divide all my wealth to feed the poor’ (San Pablo 17–18; F: 99)
Formal features and the development of the Spanish D-system
In this respect, OS is (partly) similar to present-day English (Cell phones are not always reliable/Life is full of surprises). Generics and referentially unique nouns (e.g. luna ‘moon’; see above) may be considered similar in (semantic) character.
2.4 Singular vs. non-singular nouns MS generally does not allow bare singular arguments in either subject or object position (8) (and (1a, c, e) (Section 2.1)), while OS did allow them in either position (with or without a modifier) (9): (8) *(El) perro está jugando con la pelota. the dog is playing with the ball ‘The dog is playing with the ball.’ (9) a.
Caridat nunca se pierde. charity never self loses ‘Virtue is never lost.’
(San Pablo XIII; F: 99)
b.
Pero que omne non coma nin comiença la mançana. but that man not eat nor begin the apple ‘Although a man (= one) may not eat (up) an apple or (even) take a bite.’ (Lba 678a; B: 169)
c.
De pequeña cosa nasçe fama en la vezindat. from small thing is.born fame in the neighborhood ‘Fame is born from a small thing in the neighborhood.’ (Lba 707a; B: 175)
d. Mas ell olor… A omne muerto rressuçetarya. but the odor to man dead would.resuscitate ‘But the odor (that was coming out from there) would resuscitate a dead man.’ (Razon feita d’amor 49–50; F: 68) e.
¿Quando fallaras muger quet ame…? when will.find woman who.you would.love ‘When will you find a woman who would love you (as much…)?’ (EE, line 19; MP 1982: 227)
f.
et demandoles conseio. and asked.for.them advice ‘and he asked them for a piece of advice.’ (San Fernando, line 200; MP 1982: 324)
g.
Nunca vi cauallero que lo a su cuello echasse. never saw knight who it to his neck threw ‘I never saw a knight who threw it around his neck.’ (Sancto Grial 179, I, 43; Riiho 1988: 102)
Masataka Ishikawa
In this respect, MS is similar to English; *(The) driver didn’t see *(the) hole (but cf. Love conquers everything). In general, (non-coordinated/unmodified) bare plural (and mass) nouns are not acceptable in (stylistically unmarked preverbal) subject position in MS (10) (and (1b); see Section 2.1 and Note 5): (10) a.
Finalmente *(los) gobiernos han llegado a un acuerdo. finally the governments have arrived to a agreement ‘Finally the governments have come to an agreement.’
b. *(Unos) jugadores no quisieron participar en la ceremonia. some players not wanted participate in the ceremony ‘Some players did not want (i.e. refused) to participate in the ceremony.’ c. *(La) nieve no se ha derretido todavía. the snow not self has thawed yet ‘It has not thawed yet.’
The article is basically required in the above infelicitous cases. On the other hand, (non-coordinated/unmodified) bare plural (and mass) subjects were acceptable preverbally (with either a specific or non-specific interpretation) in OS: (11) a.
Pastores te glorifficaron, ouieron te a laudare shepherds you glorified had you to praise ‘The shepherds glorified you (and) praised you.’ (Cid 335; MP 1964: 1037)
b.
Moros lo çercaron e lo derribaron. Moors it looked.for and it knocked.down ‘Moors searched for it and knocked it down.’ (Planto por Jerusalem 4c-d; MP 1982: 287)
c.
Tienpo es ya pasado. time is already passed ‘(A lot of) time has already passed.’
d. Cartas le fueron venidas. letters him were come ‘Letters were sent (lit. had come) to him.’
(Lba 661b; B: 166)
(Romances, line 4; F: 184)
With respect to the acceptability of bare plural and mass subjects with a non-specific interpretation, OS is again similar to English; Airlines are offering free upgrades.
2.5 Summary Although the presence or absence of the article in the cases discussed above was not categorical in OS, in my view the differences illustrated in the preceding subsections point to the conclusion that from a syntactic point of view, the OS D-system was fundamentally different from its MS counterpart. From a cognitive-semantic point of view, nouns denoting unique referents (which are, by their nature, always identifiable in discourse),
Formal features and the development of the Spanish D-system
generics (which are anaphoric in the sense that they are stored in the permanent registry of discourse (Kuno 1973: 41)), and possibly inalienable body-part nouns, are similar in their semantic character. The possible absence of the definite article with these nouns in the early OS period would point to the lexical (e.g. emphatic (with some deictic force), rather than expletive) character of the (early) article. The obligatory presence of the definite article with the same types of nouns in (late OS and) MS would, on the other hand, indicate that the definite article became a functional element.
3. A diachronic path in the development of the Spanish D-system 3.1 Nominal features As a point of departure, I adopt the view that nouns have the following (morphosyntactic) FFs (among others, see Longobardi 1994 and the references cited therein; Chomsky 1995): (12) a. b. c.
Proper name = [+R(eferential)] (Longobardi 1994) Plural and mass nouns = [+P(artitive)] (with no [R] specification) Singular (and abstract) nouns = [–R(eferential)] (with no [P] specification)
[R] and [P] are taken to be FFs with semantic relevance. Singular nouns refer to kinds (‘essence’), but not to individual(ized) objects, while plural and mass nouns represent unspecified amounts of entities (‘existence’) (Alonso 1982, Alarcos Llorach 1984, Gili Gaya 1985, Vergnaud & Zubizarreta 1992; Longobardi 1994; Laca 1999, among many others). Singular (and abstract) nouns (SANs) are specified (negatively) for the feature [R] only, while plural and mass nouns (PMNs) are marked (positively) for the feature [P] only. The referential status of a nominal expression is a property of the whole (nominal) phrase.8 [R] needs to be checked in order for the nominal expression to be interpreted referentially (definite or indefinite). In this study, the following (abbreviated) nominal structure is assumed:9 (13) [dP d0 … (Dem) (Pos) … Nmax …] (d = D; Nmax = [NumP … Num(ber)0 … N0 …])
. [–R] (e.g. on perro/dog), unlike [+R] (e.g. on Elena), does not necessarily entail a nonreferential/non-specific interpretation. The referential status of (common) nouns depends (partially) on a particular instantiation of the determiner (position) (see Alarcos Llorach 1984, Gili Gaya 1985; Longobardi 1994; Lyons 1999; Giusti 2002, among others). . I abstract away from the intermediate projections irrelevant to the issues dealt with in this study, such as Pos(sessive Phrase) and Dem(onstrative Phrase). For different analyses of the internal structure of DP (in Romance languages), see, e.g. Abney 1987; Mallén 1989, 1990; Delfitto & Schroten 1991; Harris 1991; Bianchi 1992; Bernstein 1993, 2001; Cinque 1994; Giusti 1993, 1997, 2002; Brugè 2002; Déprez 2006, and the references cited therein.
Masataka Ishikawa
The (highest) d-head contains [αR]. The head noun moves to Num0 to license a particular (morpholexical) instantiation of a noun (e.g. libro ‘book’). The morphological (singular/plural) number marking on the noun (in Spanish) is distinguished from the [αP] distinction since [+P] does not necessarily entail the morphologically plural marking (e.g. (las) chicas ‘(the.PL) girls.PL’ (chicas = [+P], morphologically plural marking) vs. (la) gente ‘(the.SG) people.SG’ (gente = [+P], morphologically singular marking).
3.2 OS and MS nominal structures: A basic difference The analysis given in (14) is a first approximation (based on the assumption that [R] and [P] are checked differently in OS and MS (due to different grammatical properties (cf. Ishikawa 2000)): (14) Development of the Spanish D-system … Nmax]
i. (early) OS: [dP [R/P] ii. MS: [dP [R] [dP [P] … Nmax]]
(first approximation)
(single d-shell DP structure; dP = phase) (double d-shell DP structure; dP = phase)10
Let us first consider bare SANs, which were acceptable in OS, but not in MS ((8) vs. (9), Section 2.4). In the framework adopted here, [–R] attracts SANs, while [+P] attracts PMNs. In the single d-shell structure of OS (14i = [dP [R]/[P] … [Nmax … N …]]), (bare) SANs check [–R] and may also check [–P] if one assumes (basically following Longobardi 1994) that a feature F with the negative value (i.e. [-F]) may be checked in a chain not containing [+F] within the minimal phase as a last resort. This would satisfy the checking requirement on [R]/[P] for bare singular arguments in OS. In MS (bare) SANs can check neither [–R] nor [–P] in the double d-shell structure of (14ii = [dP [R] [dP [P] … [Nmax … N …]]]) (minimality condition). The historical scenario in (14) would also appear to be compatible with the situation with bare PMNs with an existential interpretation (in subject position) in OS. In OS (e.g. (11b, c, d), Section 2.4), [+P] may be checked by the head noun. If one assumes that [–R] may be checked by the assignment of the (default) existential interpretation (for bare PMNs (Longobardi 1994)) applying once within a phase (before the checking of [P] by the head noun), bare plural and mass subjects in OS would satisfy the checking
. In some languages, more than one (definite) article (which may take a free or bound form) can appear within the single noun phrase as noted by Giusti (2002: 61): (i)
ha-bxina ha-tedira shel ha-mismaxim the-examination the-frequent of the-documents ‘the frequent examination of the documents’
(Hebrew = Giusti’s (17))
This possibility in some languages would support the nominal structure with two D-type functional categories.
Formal features and the development of the Spanish D-system
requirement on [R]/[P]. According to (14), then, MS (parallel to OS) would also allow (unmarked) bare (existential) PMNs in preverbal subject position, which is not borne out (10b). It must then be the case that the checking of [–R] is somehow blocked with a bare PMN in preverbal subject (but not object) position in MS. Note that certain nominals (e.g. generics and referentially unique nouns (Company 1991)) seem to have come to take definite articles more consistently than others in the second half of the OS period, which would be problematic to (14i) (in the present form). One finds what may be interpreted as a contrastive situation with respect to generics and existential/non-specific plurals in the period between the mid-fourteenth century and the mid-fifteenth century (according to the preliminary survey based on (selected parts of) the texts in Menéndez Pidal 1976, 1982). Namely, bare generic/ definite (plural) nominals seem to have been becoming rare (16 cases (1320 ~ c. 1416) vs. 6 (1416 ~1462)), while bare existential subjects seem to have been somewhat more resistant (12 (1320 ~ c. 1416) vs. 22 (1416 ~1462)). This state of affairs in the later OS period would be unexpected if (14) represents the whole story. My modifications to (14) are the following; (i) the OS stage consisted of two subperiods OS1 and OS2 (cf. Ishikawa 2000); (ii) the Spanish DP underwent the change in the phasehood (of dP): (15) Development of the Spanish D-system
(revised)
i. OS1: [dP-phase [R/P] … Nmax] ii. OS2: [dP-phase [R] [dP [P] … Nmax]] (emergence of the double d-shell DP) iii. MS: [dP-phase [R] [dP-phase [P] … Nmax]] (emergence of the double phase DP)
The OS nominal phrase consisted of one phase, while its MS counterpart consists of two phases. In what follows, I adopt the model of grammar in (16): (16) a. The assignment of the (default) existential interpretation (as a checking mechanism) to the lexically empty D (with a FF) is only possible as a last resort. b. FFs (of the functional head) need be checked in order for a representation to be legible (‘visible’) at the interface levels (Section 1.1). c. Structural Case assignment requires that the highest D (with [R]) be made ‘visible’ (by checking). An ([R]-less) ‘invisible’ nominal can only be assigned a default (lexical) ‘partitive’ Case by the verb (Belletti 1988 and subsequent work).11
. Overt case marking on demonstratives occupying “the outermost specifier position of the DP” in Hungarian (e.g. ez-t az asztal-t ‘this-ACC the table-ACC’ (Den Dikken 2007: 36)) may be seen as a realization of this condition on D.
Masataka Ishikawa
With these assumptions, I will re-examine the history of the Spanish D-system below.
3.3 From OS1 to OS2 In OS1 the (single d-shell) structure [dP-phase [R/P] … Nmax] is always generated. Recall that in (early) OS definite/generic nominals could appear without articles (e.g. (5) and (7) (Sections 2.2–2.3)). In the present approach [–R] (and [–P]) may be checked by the singular noun (attracted by [–R]) or by a plural (or mass) noun (attracted by [+P]) for a definite or generic interpretation to be assigned (as discussed above (Section 3.2)).12 In the resulting structure, both [–R] and [+P] would be checked (adopting Longobardi’s 1994 suggestion as before).13 In OS2 the (double d-shell) structure [dP-phase [R] [dP [P] … Nmax]] is always generated. This is because, I assume, the phasehood is essential to Spell-Out and representations without a phase projection are not ‘processable’ at Spell-Out. In this period, bare PMNs may receive an existential (or non-specific) interpretation (in either subject or object position) if one assumes as before that the lexically empty D can (only) be assigned the default (existential) interpretation (checking [–R] in this case) as a last resort. In OS2 (and MS (see below)), due to the newly created double d-shell structure, checking of [–R] with PMNs with a definite/generic reading involves the insertion of a determiner since [–R] needs to be checked for structural Case assignment. For bare singular nouns (in OS2), it may be suggested that [–R] may be checked (for a specific reading) in an Agree relation between the higher d and the closest (singular) N with a matching feature (i.e. [–R]), assuming that the lexically empty lower d-position does not induce the minimality violation for Agree to hold (since the lower dP is not a phase in this period), and that [–P] can be checked by Form Chain (a last resort option) linking the intervening head(s) (of the same category) in the process of Agree (Move).14
. An existential interpretation may be assigned to PMNs if the relevant assignment operation takes place before the checking of [P]. . As for the definite-vs.-generic distinction of the nominal expression, I assume that in a given context the appropriateness of one or the other interpretation is measured (ultimately) at the discourse level (as in MS (Alarcos Llorach 1984; see also Déprez 2006)). . I assume the following characterization of Agree:
(i)
Agree
(adapted from Chomsky 1998a, 1999, and Roberts 2007)
A Agrees with B iff: a. A asymmetrically c-commands B in its (local) domain; b. A and B are compatible in terms of the formal feature specification.
In Chomsky’s (1998a: 37) terminology, given a matching pair a probe P (= A) and a goal G (= B), “G must (at least) be in the domain D(P) of P and satisfy locality conditions.” The (local) domain is
Formal features and the development of the Spanish D-system
In the evolution of the Spanish D-system, the (gradual) semantic weakening of the definite article (in the Middle Ages) must have favored the use of articles with non-existential nominal expressions. This change in primary data (with an increased number of tokens of noun phrases in which the d-position was filled by the article) must have functioned as trigger and would have induced the two d-shell structure.
3.4 MS stage In MS, I assume that since the lower dP has acquired phasehood, a nominal expression with only one d-shell may be generated, i.e. [dP [P] … Nmax] (by selecting one D element by Numeration), which, being in a sense ‘complete’, can be sent to PF at Spell-Out. If this option is chosen with PMNs, the resulting nominal receives an existential/non-specific interpretation (in object position with (the default) partitive Case due to the visibility condition on the structural Case assignment).15 If, on the other hand, two functional (D) categories are selected with either a SAN or a PMN as a head, the resulting (double d-shell) nominal only has a possibility of having a non-existential interpretation with a determiner. This is because in accordance with the (general) economy guidelines, if the (structurally simpler) single d-shell structure expresses an existential meaning (by default), also allowing a double d-shell configuration to express the same meaning (with a last resort option) would be uneconomical. Namely, one kernel configuration would suffice for one (kernel) semantic meaning.16 For a non-existential interpretation to be assigned, [–R] must be checked (for structural Case). The direct linking of [–R] and the head noun (Agree) (without an article) would not be licensed in MS (with either SANs or PMNs) since the lower d-projection is now a cyclic node (the phase level), which would block Agree between [R] and the head N (Chomsky 1999: 9–10). Thus, an overt determiner is
understood here in terms of c-command, i.e. the sister of A (= P) (Chomsky 1998a: 38). ‘Closeness’ is defined as follows:
(ii) Given B and B’, both matching A and in D(A), B is closer to A than B’ if B’ is in D(B) (following Chomsky 1998a: 38).
. As for postverbal subjects, I assume tentatively that the (inflected) verb may assign (the default) partitive Case to the (agreeing) plural/mass subject (under c-command) similar to the case of existential objects. This (or similar) Case-assignment mechanism may also be assumed for the postverbal subject in such constructions as the following; Al llegar Pedro… ‘When Pedro arrived…’ and Llegado el paquete… ‘After the package had arrived…’. . When the quantificational meaning is emphasized, (double d-shell) plural nouns may accommodate some lexical elements (such as unos (libros)/unas (casas) some (books/houses)’).
Masataka Ishikawa
required for the checking of [R] (and [P]) with definite/generic nominals in MS (Sections 2.2–2.3). I assume that the article is merged (Section 3.6) in the specifier position of the lower dP and enters into an Agree relation with the head noun (at the same time checking [P] in the process) in the cycle of the lower phase. Then it cliticizes onto the higher d (checking [R]) (Bernstein 1993; Parodi 1994 and references cited therein). In this connection, note in passing the following contrast with respect to unstressed object pronouns (etymological siblings of definite articles) in MS. When the object of a verb is semantically definite, the corresponding clitic pronoun (referring to the specific object) is obligatory; *(la) comí hace un rato (e. g., la for la manzana ‘the apple’) ‘I ate it a while ago’. On the other hand, when the object is interpreted as non-specific/ existential, the corresponding clitic (with strict identity) may not appear; sí, ø tengo (e.g. ø for dinero ‘money’) ‘Yes, I have (some)’ (see also Laca 1999: 897–898, 901–902, 910). In the present analysis, the presence in one case and the absence in the other of the object clitic may be accounted for in terms of structural Case assignment. That is, if the pronominal clitic is viewed as a realization of [R] under the higher d-head (see also Uriagereka 1995), its appearance needs to be Case-licensed.17 Only the definite nominal meets the structural condition for Case (through the checking of [R]).18 In sum, the new requirement on the use of the definite article in MS was brought about by the change in the phase status of the lower dP.
3.5 Boundaries between the sub-periods The definite article was extended to increasingly larger syntactic domains in the fourteenth and fifteenth centuries (based on the OS texts in Marden 1904; Menéndez Pidal 1964, 1976, 1982; Bolaño e Isla 1966; Fotitch 1969; Kohler 1970; Daly & Zahareas 1978; Blecua 1992 and Whinnom 1971). In this section I will touch on the question of when the two grammatical changes in question may possibly have taken place in the history of Spanish (without intending to provide quantitatively comprehensive results within the scope of this chapter). Important evidence for detecting the chronological boundaries in the development of the Spanish D-system appears to come from (bare) generics and (bare) subjects. According to the present analysis, the obsolescence (and eventual loss) of bare generic (and definite) plurals must have preceded that of bare
. Spanish pronouns exhibit the morphological case distinction, e.g. ella (NOM) ‘she’ vs. la (ACC) ‘her’ vs. le (DAT) ‘her’, etc. . In the present analysis, the lack (in MS) of a ‘partitive’ (clitic) pronoun is case-theoretic; cf. presence of partitive clitics in Italian (ne) and French (en), which suggests that Italian and French would have different DP structures.
Formal features and the development of the Spanish D-system
singular and (existential) plural subjects in OS (as suggested above). (That is, the latter must have been available in both OS1 and OS2, while the former only in OS1.) Previous studies provide the following general picture for generics. Bare generic subjects were becoming rare in the early fourteenth century and generic expressions (in general) appear to have come to take definite articles roughly by the mid-to-late fourteenth century (Menéndez Pidal 1954: 7, 1964: 943; Company 1991: 405, 417).19 The following are the summarized results (from the survey mentioned above (Section 3.2)) regarding plural generics (which must be taken to be a preliminary approximation):20 (17)
a. b. c. d.
Bare (plural) generics: 13 Article + (plural) generics: 14 Bare (plural) generics: 6 Article + (plural) generics: 95
(1320 ~ c. 1400) (1320 ~ c. 1400) (1401 ~ 1462) (1401 ~ 1462)
It may be suggested that the more and more frequent appearance of the (semantically weakened) article with generic PMNs would have induced the independent d-projection for [R] (in the double d-shell structure), which must have eventually ousted bare generics. It may then be speculated that the boundary between OS1 and OS2 (or the onset of OS2) could (tentatively) be situated roughly in the late-fourteenth century (considering the conservative nature of written texts). Singular nouns denoting (uniquely) identifiable referents (e.g. (el) mundo ‘(the) world’) came to take articles more consistently in (or towards the end of) the second half of the fourteenth century (Company 1991: 419; Marden 1904: xliii; Note 19). The preliminary survey results concerning referentially unique nouns are the following; between 1250 and 1350, 18 cases without an article vs. 9 cases with an article; from 1350 to c.1416, 12 cases (without an article) vs. 19 cases (with an article). (Between 1416 and 1462, only sporadic cases of bare unique nouns (four) were found in the texts surveyed.) Bare SA subjects seem to have become obsolete about a century later (and have come to take an article by the early sixteenth century), 21 which (due to their semantic
. The generalization of the (definite) article to generics and uniquely identifiable nouns (e.g. (el) ç(i)elo ‘(the) sky’) roughly coincided (in the late fourteenth to the early fifteenth century), which might suggest that the semantic notion of ‘class’ or ‘totality’ was extended from the former type to the latter since each unique referent represents a whole (i.e. singlemember) class of its own (Company 1991: 406). In other words, this extension may be considered semantically motivated. . A quantitative survey of bare definite PMNs (which seem to have become rare by the end of the fourteenth century) will be left for future work. . Keniston (1937: 221) notes that “[t]he use of the article with abstract nouns is firmly established by the sixteenth century”.
Masataka Ishikawa
affinity) might first have been influenced by the (increasingly more consistent) use of the article with referentially unique nouns. A preliminary summary of the use of the definite article with singular subjects (from the same survey) is as follows. Bare SA subjects: 15; vs. article + SA subjects: 35 (1416 ~ 1462). The Cárcel de amor (1492) provides the following figures (in affirmative tensed contexts (in the first half of the book)): (18) a. Bare (unmodified) singular subjects: 17 (of these 17 cases, 9 occur in one single occasion in an enumeration) b. (Unmodified) singular subjects with an article: 55
About half a century later, the Lazarillo de Tormes (1554) gives the following figures (in affirmative tensed contexts (in the entire text)): (19) a. b.
Bare (unmodified) singular subjects: 1 (Unmodified) singular subjects with an article: 80
Bare existential/non-specific PM subjects seem to have been available until the mid(-to-late)-fifteenth century (15 cases in the period from the late 12th century and c. 1416, and 22 cases in the period from 1416 to 1462) based on the preliminary survey (with an unfortunately small number of tokens) (cf. Keniston 1937: 345). When limited to unmodified subjects in preverbal position, the first half of the Cárcel de amor gives no examples of bare existential (or non-specific) plural subjects (in affirmative tensed contexts) (as opposed to 7 with an article). The Lazarillo de Tormes provides one unmodified existential plural subject without an article in preverbal position (as opposed to 4 with an article) (in affirmative tensed contexts in the entire text). Once the double d-shell DP with two phases emerged in the grammar, bare singular subjects and bare existential plural subjects (in preverbal position) should be unavailable since the representation with a lexically unfilled D (i.e. unchecked [R]) for the intended readings would be rejected at the interface. The loss of bare existential plural subjects and that of bare singular subjects roughly coincided in the late fifteenth century to the early sixteenth century, which would lead one to suspect that the (completion of the) switch from the OS2 to the MS stage could likely be located (tentatively) in the late second half of the fifteenth century (or possibly in the early sixteenth century).
3.6 Some comments on untouched issues It has often been suggested that the creation of articles (in Spanish/Romance) is related to the loss of overt case morphology of Latin (e.g. Álvarez Martínez 1986: 19 and references cited therein). When Latin cases were reduced to two, “Recto” (for subject) and “Oblicuo” (for non-subject nominals) in Old Romance, the former came to take an article (more) consistently prior to the latter (though not as consistently in HispanoRomance as in Gallo-Romance (Álvarez Martínez 1986: 19)). Originally the demonstrative-article (in Vulgar Latin/Old Romance) might still have had some case function,
Formal features and the development of the Spanish D-system
which was later lost in the vernacular Romance of Hispania. It then came to acquire the function of definiteness/specificity marker (i.e. became the (morphosyntactic) article) in (late) OS or early MS. It may be considered (with Álvarez Martínez 1986) that in the early stages of its creation, the article may have played a functional role of indicating Nominative case (subject), analogous to the function of the so-called personal a in MS, which accompanies a (definite) [human/animate] accusative object (e.g. Vi a Ana ‘I saw Ana’). In other words, the lexical semantic meaning and the grammatical function expressed by a lexical noun in Latin (e.g. amicus = ‘friend’ and nominative (in addition to number irrelevantly here)) came to be expressed (initially) by the article (= grammatical function) and the noun (= lexical meaning), i.e. el amigo (vs. ø amigo) (Alonso 1982: 127, Note 2). In structural terms, it may be suggested that the case ending in Latin (e.g. -um/-am) could make the highest nominal head visible (see, e.g., Giusti 2001). Here I will simply sketch the crudest picture of the evolution in question (further investigation of which will be left for future research). Latin had determiners (e.g. demonstratives). Thus its lexicon had the functional category D and its nominal structure consisted of a D projection. One may assume that the highest d-head (be it d or K, technical aspects of which will be left open here) hosted a case feature in Latin (and possibly in Vulgar Latin/Old Romance). Overt case morphology of the head noun would need to be licensed in the (checking) domain of the functional head with an appropriate case feature. For example, this may be achieved if the complement of the highest d-head moves to SpecdP (licensing overt case morphology on the lexical items involved in the whole nominal expression (i.e. N and modifying As)). This raising would make the highest head visible for structural Case. As the nominal case ending was (partially) lost in Old Romance/early OS, the newly emerging grammatical element ‘(demonstrative-transforming) article’ came to fulfill this visibility condition. It may be supposed (with Giusti 2001, 2002) that movement of the demonstrative to a position within DP (in Old Romance), which might be viewed as a remnant XP movement of Latin, was reanalysed and replaced by merge of the article in D (in OS/MS). In (OS and) MS, the visibility requirement for Case is satisfied by the raising of N or merge of a determiner.
4. Summary and concluding remarks The focus of this chapter has been to characterize the fundamental syntactic differences between the OS and MS D-systems which reflect the change in the article usage (rather than to pinpoint the exact chronological boundaries (or the waves of the article extension) among the subperiods in the development of the D-system). By way of conclusion, I will briefly consider some aspects of the relation between the language acquirer’s interpretive computation and language change. Considering that acquisition takes place based, at least partly, on the decoding of available linguistic input, it
Masataka Ishikawa
seems possible that the (interpretive/inferential) parsing strategy plays an important role in the process of grammar formation. If at early stages of language acquisition the language parser prefers structural analyses that would allow quick on-line processing in decoding (Johnson & Lappin 1997: 275–276; Crain & Thornton 1998; cf. Johnson 1991; Roeper 1996), then the notion of economy in processing may well be different in adult grammar (synchrony) and early child grammar (micro-diachrony).22 During early language acquisition (i.e. by extension, in micro-diachrony), ‘maximize local optimality’ could be the preferred strategy in parsing (cf. Roeper 1996: 434). If the presence of a lexical element should (ideally) be the reflection of a necessary Merge/ Move, rather than the reflection of a last resort operation (i.e. the lexical override (for a default rule); see, e.g., Hale 2007 on this subject in the phonological domain), then linguistic change may be guided by the acquirer’s inferential prejudice which would lead him/her to assume whenever possible that incoming strings encode necessary (rather than optional or last resort) grammatical operations (for convergence) in the most ‘economical’ manner possible. In the historical development proposed here, in early OS the noun phrase had a single d-shell structure. In late OS and MS the noun phrase had/has a double d-shell structure with one phase and two phase levels, respectively. The first change, by which [R] came to be merged under the higher d, may be regarded as a (supposed diachronic) tendency to make the feature makeup of a functional category as simple as possible. This change may be viewed as an instance of structural simplification defined in terms of the ‘avoidance of syncretism’ (Roberts & Roussou 2003), rather than in terms of the number of projections.23 The second change may be related (or due) to the preference to ‘hand over’ substrings (phases) as quickly as possible for efficient on-line processing. Then the first type is a change induced by a factor internal to the core grammar/syntax, while the second type is a change motivated by a factor external to the core grammar/ syntax. In sum, the development of the Spanish D-system (with respect to the use of the definite article) can be considered the compound results of the two changes in the direction of reducing the application of last resort options.
. Cf. Guasti (1996) for discussion of certain asymmetries between children’s interpretive and productive preferences. . If the ideas discussed here prove to be on the right track, the development of the double d-shell structure would represent the selection of a (supposedly) less elegant grammar over a (supposedly) more elegant one (with a more compact representation and fewer computational steps, cf. Clark & Roberts 1993). From a processing perspective, the creation of the double phase structure may be viewed as a way to ‘compensate’ this structural complication.
Formal features and the development of the Spanish D-system
To conclude, the present study seems to show that the outcome of syntactic change may be (partially) influenced by structure-internal factors. From the perspective of the stipulated (inferential) preference (seen as a type of decoding strategy) in (early) language acquisition, the type of linguistic change proposed here for the Spanish D-system may be considered an instance of a possible type of (economy-driven) ‘syntactic drift’ in language change.
Abbreviations Cid DAV EE Guillelme GS Lba San Fernando San Pablo
Cantar de Mio Cid Disputa del agua y el vino Estoria de España Estoria del Rey Guillelme Generaciones y semblanzas Libro de buen amor Crónica particular de San Fernando La Primera epístola de San Pablo a los Corintios
B = Blecua (1992) F = Fotitch (1969) K = Kohler (1970) MP = Menéndez Pidal (1964, 1976, 1982)
Primary sources Blecua, Alberto (ed.). 1975. La vida de Lazarillo de Tormes. Madrid: Editorial Castalia. Blecua, Alberto (ed.). 1992. Libro de buen amor. (Juan Ruiz Arcipreste de Hita) Madrid: Cátedra, S.A. Bolaño e Isla, Amancio (ed.). 1966. El Conde Lucanor. México DF: Editorial Porrua. Daly, Saralyn R. & Zahareas, Anthony N. (eds). 1978. The Book of True Love (A Bilingual Edition), with an introduction by Saralyn R. Daly. University Park PA: The Pennsylvania State University Press. Fotitch, Tatiana (ed.). 1969. An Anthology of Old Spanish. Washington DC: The Catholic University of America Press. Kohler, Eugène. 1970. Antología de la literatura española de la Edad Media. París: Éditions Klincksieck. Marden, C. Carroll (ed.). 1904. Poema de Fernan Gonçalez. Baltimore: The Johns Hopkins Press. Menéndez Pidal, Ramón. 1964. Cantar de Mio Cid, Vol. 3, 4th edn. Madrid: Espasa-Calpe. Menéndez Pidal, Ramón. 1976. Crestomatía del español medieval, Vol. 1. Madrid: Gredos. Menéndez Pidal, Ramón. 1982. Crestomatía del español medieval, Vol. 2. Madrid: Gredos. Whinnom, Keith (ed.). 1971. Cárcel de amor. Madrid: Editorial Castalia.
Masataka Ishikawa
References Abney, Steve. 1987. The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, MIT. Alarcos Llorach, Emilio. 1984. Estudios de gramática funcional del español, 3rd edn. Madrid: Gredos. Alonso, Amado. 1982. Estudios lingüísticos: Temas españoles. 3rd edn. Madrid: Gredos. Álvarez Martínez, M. Ángel. 1986. El artículo como entidad functional en el español de hoy. Madrid: Gredos. Belletti, Adriana. 1988. The Case of unaccusative. Linguistic Inquiry 19: 1–34. Bernstein, Judy. 1993. The syntactic role of word markers in null nominal constructions. Probus 5: 5–38. Bernstein, Judy. 2001. The DP hypothesis: Identifying clausal properties in the nominal domain. In The Handbook of Contemporary Syntactic Theory, Mark Baltin & Chris Collins (eds), 536–561. Oxford: Blackwell. Bianchi, Valentina. 1992. Sulla struttura funzionale del sintagma nominale italiano. Rivista di Grammatica Generativa 17: 39–64. Bosque, Ignacio. 1980. Sobre la negación. Madrid: Catedra. Bosque, Ignacio. 1996. Por qué determinados sustantivos no son sustantivos determinados. Repaso y balance. In El sustantivo sin determinación. La ausencia de determinante en la lengua española, Ignacio Bosque (ed.), 13–119. Madrid: Visor Libros. Brugè, Laura. 2002. The positions of demonstratives in the extended nominal projection. In Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1, Guglielmo Cinque (ed.), 15–53. Oxford:OUP. Butt, John & Benjamin, Carmen. 2004. A New Reference Grammar of Modern Spanish, 4th edn. London: Edward Arnold. Cano Aguilar, Rafael. 1988. El español a través de los tiempos. Madrid: Arco/Libros, S.A. Chomsky, Noam. 1993. A minimalist program for linguistic theory. In The View from Building 20, Ken Hale & Samuel Jay Keyser (eds), 1–52. Cambridge MA: The MIT Press. Chomsky, Noam. 1995. The Minimalist Program. Cambridge MA: The MIT Press. Chomsky, Noam. 1998a. Minimalist inquiries: The framework [MIT Occasional Papers in Linguistics 15]. Cambridge MA: Department of Linguistics, MIT. Chomsky, Noam. 1998b. Some observations on economy in generative grammar. In Is the Best Good Enough?, Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis & David Pesetsky (eds), 115–127. Cambridge MA: The MIT Press. Chomsky, Noam. 1999. Derivation by phase [MIT Occasional Papers in Linguistics 18]. Cambridge MA: Department of Linguistics, MIT. Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays on Minimalism in Honor of Howard Lasnik, Roger Martin, David Michaels & Juan Uriagereka (eds), 89–155. Cambridge MA: The MIT Press. Chomsky, Noam. 2001. Beyond explanatory adequacy [MIT Occasional Papers in Linguistics 20]. Cambridge MA: Department of Linguistics, MIT. Cinque, Guglielmo. 1994. On the evidence for partial N-movement in the Romance DP. In Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne, Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaella Zanuttini (eds), 85–110. Washington D.C.: Georgetown University Press. Clark, Robin & Roberts, Ian. 1993. A computational model of language learnability and language change. Linguistic Inquiry 24: 299–345.
Formal features and the development of the Spanish D-system
Company, Concepción. 1991. La extensión del artículo en el español medieval. Romance Philology 44: 402–424. Contreras, Heles. 1996. Sobre la distribución de los sintagmas nominales no predicativos sin determinante. In El sustantivo sin determinación. La ausencia de determinante en la lengua española, Ignacio Bosque (ed.), 141–168. Madrid: Visor Libros. Crain, Stephen & Thornton, Rosalid. 1998. Investigations in Universal Grammar: A Guide to Experiments on the Acquisition of Syntax and Semantics. Cambridge MA: The MIT Press. De Mello, George. 1980. On the use of the article in Spanish. Hispania 63: 557–60. Delfitto, Denis & Schroten, Jan. 1991. Bare plurals and the number affix in DP. Probus 3: 155–185. Déprez, Viviane. 2006. On the conceptual role of number. In New Perspectives on Romance Linguistics, Vol.1: Morphology, Syntax, Semantics and Pragmatics [Current Issues in Linguistic Theory 275], Chiyo Nishida & Jean-Pierre Y. Montreuil (eds), 67–81. Amsterdam: John Benjamins. Dikken, Marcel den. 2007. Phase extension: Contours of a theory of the role of head movement in phrasal extraction. Theoretical Linguistics 33: 1–41. Gili Gaya, Samuel. 1985. Curso superior de sintaxis española, 15th edn. Barcelona: Biblograf. Giusti, Giuliana. 1993. La sintassi dei determinanti. Padova: Unipress. Giusti, Giuliana. 1997. The categorial status of determiners. In The New Comparative Syntax, Liliane Haegeman (ed.), 95–123. London: Longman. Giusti, Giuliana. 2001. The birth of a functional category: From Latin ILLE to the Romance article and personal pronoun. In Current Studies in Italian Syntax: Essays Offered to Lorenzo Renzi, Guglielmo Cinque & Giampaolo Salvi (eds), 157–171. Amsterdam: North Holland. Giusti, Giuliana. 2002. The functional structure of noun phrases: A bare phrase structure approach. In Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1, Guglielmo Cinque (ed.), 54–90. Oxford: OUP. Guasti, Maria Teresa. 1996. Acquisition of Italian interrogatives. In Generative Perspectives on Language Acquisition. Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons [Language Acquisition & Language Disorders 14], Harald Clahsen (ed.), 241–269. Amsterdam: John Benjamins. Hale, Mark. 2007. Historical Linguistics. Oxford: Blackwell. Hanssen, Federico. 1913. Gramática histórica de la lengua castellana. New York NY: Hafner Publishing. Harris, James. 1991. The exponence of gender in Spanish. Linguistic Inquiry 22: 27–62. Harris, Martin. 1980. The marking of definiteness in Romance. In Historical Morphology, Jacek Fisiak (ed.), 141–156. The Hague: Mouton. Ishikawa, Masataka. 2000. Sobre la cronología del desarrollo del sistema determinante en castellano: Dos cambios sintácticos en los siglos XIII–XIV. Paper presented at the V Congreso Internacional de Historia de la Lengua Española, January 31-February 4, 2000, Valencia, Spain. Johnson, David and Lappin, Shalom. 1997. A critique of the minimalist program. Linguistics and Philosophy 20: 273–333. Johnson, Mark. 1991. Deductive parsing: The use of knowledge of language. In Principle-Based Parsing: Computation and Psycholinguistics, Robert C. Berwick, Steven P. Abney & Carol L. Tenny (eds), 39–64. Dordrecht: Kluwer. Keniston, Hayward. 1937. The Syntax of Castilian Prose. The Sixteenth Century. Chicago IL: The University of Chicago Press. Kuno, Susumu. 1973. The Structure of the Japanese Language. Cambridge MA: The MIT Press. Laca, Brenda. 1999. Presencia y ausencia de determinante. In Gramática descriptiva de la lengua española, Vol. 1, Ignacio Bosque & Violeta Demonte (eds), 891–928. Madrid: Espasa-Calpe. Lapesa, Rafael. 1981. Historia de la lengua española, 9th edn. Madrid: Gredos.
Masataka Ishikawa Lapesa, Rafael. 1996. El sustantivo sin actualizador en espanñol. In El sustantivo sin determinación. La ausencia de determinante en la lengua española, Ignacio Bosque (ed.), 121–137. Madrid: Visor Libros. Lapesa, Rafael. 2000. Del demostrativo al artículo. In Estudios de morfosintaxis histórica del español. Vol. I, Rafael Cano Aguilar and Maria Teresa Echenique Elizondo (eds), 360–387. Madrid: Gredos. Lathrop, Thomas A. 1984. Curso de gramática histórica española. Barcelona: Ariel. Lausberg, Heinrich. 1966. Lingüística románica. Madrid: Gredos. Leonetti, Manuel. 1996. Determinantes y contenido descriptivo. Español Actual 66: 5–23. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Mallén, Enrique. 1989. The Internal Structure of Determiner Phrases. Ph.D. dissertation, Cornell University. Mallén, Enrique. 1990. Clitic movement inside noun phrases. Studia Linguistica 44: 1–29. Marcos Marín, Francisco, Satorre Grau, F. Javier and Viejo Sánchez, María Luisa. 1998. Gramática española. Madrid: Editorial Síntesis. Marden, C. Carroll (ed.). 1904. Introduction to Poema de Fernan Gonçalez. Baltimore MD: The Johns Hopkins Press. Menéndez Pidal, Ramón. 1954. Cantar de Mio Cid, Vol. 1, 3rd edn. Madrid: Espasa-Calpe. Menéndez Pidal, Ramón. 1982. Manual de gramática histórica española. Madrid: Espasa-Calpe. Parodi, Claudia. 1994. On case and agreement in Spanish and English DPs. In Issues and Theory in Romance Linguistics, Michael L. Mazzola (ed.), 403–416. Washington DC: Georgetown University Press. Posner, Rebecca. 1996. The Romance Languages. Cambridge: CUP. Renzi, Lorenzo. 1993. Da dove viene l’articolo il. In Verbum Romanicum, Johannes Kramer & Guntram A. Plangg (eds), 215–230. Hamburg: Helmut Buske. Riiho, Timo. 1988. La redundancia pronominal en el iberorromance medieval. Tübingen: Niemeyer. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge MA: The MIT Press. Roberts, Ian. 2007. Diachronic Syntax. Oxford: OUP. Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: CUP. Roeper, Thomas. 1996. The role of merger theory and formal features in acquisition. In Generative Perspectives on Language Acquisition. Empirical Findings, Theoretical Considerations and Crosslinguistic Comparisons [Language Acquisition & Language Disorders 14], Harald Clahsen (ed.), 415–449. Amsterdam: John Benjamins. Uriagereka, Juan. 1995. Syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79–123. Uriagereka, Juan. 1999. Multiple spell-out. In Working Minimalism, Samuel David Epstein & Norbert Hornstein (eds), 251–282. Cambridge MA: The MIT Press. Urrutia Cárdenas, Hernán & Álvarez Álvarez, Manuela. 1988. Esquema de morfosintaxis histórica del español. Bilbao: Publicaciones de la Universidad de Deusto. Vergnaud, Jean-Roger & Zubizarreta, Maria Luisa. 1992. The definite determiner and the inalienable constructions in French and in English. Linguistic Inquiry 23: 595–652. Vincent, Nigel. 1997. The emergence of the D-system in Romance. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds), 149–169. Cambridge: CUP.
The rise of OV word order in Irish verbal-noun clauses* Elliott Lash
University of Cambridge In this chapter I propose a new diachronic analysis of Old Irish OV word order in verbal noun phrases, which replaces an earlier synchronic approach to this OV word order found in Disterheft (1982). The analysis proposed here relies on bracketing ambiguity and morphological underspecification as motivations for change. I argue that these prerequisites contributed to the increased verbal nature of verbal nouns in Old Irish, starting from the 9th century. This is formalized as the introduction of verbal functional material projected above the verbal noun. Specifically, I claim that this verbal functional material licensed OV word order as a focalization strategy. This claim is based on a corpus of texts spanning the 8th to the 12th centuries.
1. Introduction In this chapter a diachronic approach is adopted in attempting to account for the OV order found in non-finite clauses in Old and Middle Irish (along with some speculation about certain modern dialects). I will argue that this word order arose through a series of reanalyses affecting sentences containing a verbal noun selected by the preposition do ‘to/for’. The reanalyses argued for are summarized directly. First, ambiguity between two possible analyses induced a shift in bracketing, bringing new nominal material into the verbal-noun clause, where the nominal material could be plausibly understoood as part of that clause. Additionally, morphological underspecification contributed to the collapse of distinctions between various types of verbal-noun clause. Moreover, the pre-verbal-noun order of the nominal within the new OV verbal-noun clause became associated in the Old Irish period with focus interpretation of the nominal. Subsequent loss of this focus interpretation in the (late) Middle Irish period must therefore be seen as a conventionalization of the word order and a grammaticalization of the focus movement as a kind of object shift. * I would like to thank David Willis, Theresa Biberauer, Glenda Newton, Alastair Appleton, the participants in the Continuity and Change Conference and an anonymous reviewer for helpful comments on this chapter.
Elliott Lash
These reanalyses forced various grammaticalizations of the preposition do. First, during the Middle Irish period the categorial status of this element changed to a movement-inducing focus-marking particle found in the extended projection of VP (in the spirit of Belletti’s (2004) work on the left periphery of the VP). Second, after focus interpretation was lost, do was likely grammaticalized again, but differently, in two dialect areas; however, in both, OV word order was preserved as a historical remnant (on historical remnants see, for example, Simpson 2004). Third, the Modern Irish verbal noun, at least in some circumstances, has become much more infinitival, which is to say that it is embedded beneath verbal functional structure instead of nominal functional structure. This chapter is organized in the following way. In Section 2 I will provide a brief overview of the type of verbal-noun clauses with which we are concerned, as well as a critique of two previous accounts which propose a raising analysis of the phenomenon to be discussed (Disterheft 1982 and Disterheft 1984). In Section 3 data from copular clauses in the Old Irish text known as the Würzburg Glosses (8th c.) will be used to further argue against the raising analysis. Instead, I will claim that, at some point prior to the eighth century, there may have been a bracketing reanalysis in copular sentences leading to the introduction of a nominal element before the verbal noun. In Section 4 data from non-copular clauses mostly from the Milan Glosses (9th c.) will show that a later reanalysis involving verbal nouns in non-copular clauses occurred, which allowed further nominal elements to occur before the verbal noun. In Section 5 various Middle Irish developments leading to the situation in modern dialects will be discussed, in particular the loss of focus interpretation. Section 6 will conclude this chapter with a short discussion of grammaticalization pathways and clause cartography.
2. Types of non-finite clauses In Old Irish, non-finite clauses were primarily constructed in two ways. In the first, an accusative-marked verbal noun was used together with an object marked with genitive case as in (1). The second method utilized a preposition, such as do ‘to/for’, which selected the verbal noun, marking it with dative case. The object of the verbal noun was still marked with genitive as in (2). A further possible order is shown in (3). It should be noted here that the preposed object is in the accusative case and the logical subject of the verbal noun is within a prepositional phrase.1 (1) Ní foil-s-itis [deics-in a gnús-a.] not endure-sbj-impf.3s [watching-acc his face-gen] ‘They could not bear to watch his face.’
(Wb. 15b20)
. When Irish morphology involves processes other than affixing (i.e. internal modification), I have separated the morpheme from the stem by a dot in the gloss. When the morpheme is easily separable (i.e it is an affix), I have indicated this by a hyphen.
The rise of OV word order in Irish verbal-noun clauses
(2) Com mi-mmis angraib dú-ib si [do gabaal desimrecht-a so be-sbj.1p model for-2p you [to taking.dat example-gen dí-in ni.] from.1p us] ‘So that we may be a model for you to take example from us.’ (3) Huare nad n-ac-at hi frecndairc [gnim-u since not NAS-see-3p in present.dat [deed-acc.pl
(Wb. 26b18) cosmail-i same-acc.pl
du denum du Christ.] to doing.dat by Christ.dat] ‘Since, at present, they don’t see that similar deeds are done by Christ.’ (Ml. 93d14)
This chapter concentrates on [do + Verbal-Noun] constructions such as (3), in which the object (or subject) of the verbal noun is linearly before the verbal noun and marked with the nominative or accusative case. Such a construction is found in all periods of the language; however, it seems to be fairly rare in the Old Irish period and very common in Modern Irish (to the practical exclusion of the construction found in (2)). In the following sections I will argue that this surface word order was the result of a series of reanalyses that occurred affecting the category of both the preposition and the verbal noun itself, as well as the placement of the arguments of that verbal noun. However, before presenting my proposal, I will first review two prior studies of this phenomenon. Disterheft (1982) argues at length that the OV word order seen in (3) is due to various raising rules that are present in the synchronic grammar of Old Irish. She posits two main rules: Raising-to-Subject and Subject-to-Object Raising. In her account, Raising-to-Subject was innovated first and is found from the time of the Würzburg Glosses on the Pauline Epistles (Wb., mid-8th c.) onward. Subsequently, the second rule was added on the analogy of this first rule, sometime between Wb. and the Milan Glosses on the Psalms (Ml., mid. 9th c.). I will call Disterheft’s analysis ‘the raising analysis’. We shall see that the raising analysis contrasts with my analysis (to be proposed below) in that, on the raising analysis, the argument of the verbal noun is moved from the lower verbal-noun clause into the matrix clause synchronically. In my analysis, however, the argument of the verbal noun is synchronically merged within the lower verbal-noun clause, entering this clause diachronically through a bracketing reanalysis. To see why this is a preferable analysis, let us consider in some detail the raising analysis. First, considering examples (4) and (5), Disterheft notes that the logical subjects of the verbal nouns do not seem to be behaving like most verbal-noun subjects in that they are not marked either by the preposition do or by a postposed genitive (with unaccusative verbs). Either way, the case marking of the subject seems to be anomalous since this subject is marked with nominative case.
Elliott Lash
(4) Bi-d héet li-b si [geint-i do bith cop-ft.3s jealousy.nom with-2p you [gentile-nom.pl to being.dat in hiris.] in faith.dat] ‘You will be jealous that the gentiles have faith.’
(Wb. 5a13)
(5) Ni fiu le-u [bás n-aill du Not fitting.nom with-3p [death.nom nas-another to techt forr-u.] coming.dat on-3p] ‘They don’t find it fitting for another death to come upon them.’
(Ml. 139c3)
In order to account for this, Disterheft proposes the Raising-to-Subject rule, which is an Extended Standard Theory transformational rule targeting a noun in the VP. The rule is given below in (6). What this rule states is that in the context of the copula and a nominal or adjectival predicate marked with nominative case, the NP subject of a verb in an embedded clause can be moved into the matrix clause and marked with nominative case. The verb of the embedded clause is simultaneously changed to a [do + Verbal-Noun] structure. (6) [be {NP/ADJ [nom]} [V NP X]] becomes [be {NP/ADJ [nom]} NP[nom] [do VN X]]
Of course, this type of transformation is completely out in the context of a more minimalist theory in which the only linguistically relevant levels are the interfaces; hence, there is no difference between a deep structure, the transformational component and surface structure. Furthermore, this type of transformation is out in a minimalist theory in which the Inclusiveness Condition holds (Chomsky 1995: 228). This is because the condition is violated by the insertion of a new element by the transformation in (6), namely the preposition do. Even updating this rule to conform to the current theory is a challenge. Such a rule seems to presuppose that, underlyingly (i.e. at Deep Structure), a verbal noun is in fact just a normal non-finite verb. This assumption is made clearer by further examining the types of data used to argue for the raising analysis. Consider (7): (7) Is festae in trocaire mór [do todlugud.] cop.3s wise.nom the.nom mercy.nom great.nom [to craving.dat] ‘It is wise to crave great mercy.’ (Ml. 71a2)
Here, the Raising-to-Subject rule appears to take an object from the verbal-noun phrase and raise it to the ‘subject’ position of the main clause. In order for this to make sense, one might suppose that the object of the verbal noun has first been promoted to subject within its own clause, by some sort of passivization operation. Otherwise, the object would surely be inaccessible to any probe within the matrix clause. Consider a
The rise of OV word order in Irish verbal-noun clauses
similar structure, (8) in English, assuming ‘be wise’ to be a raising predicate. Here the first nominal element accessible to the higher clause in a non-passivized embedded clause would be John, while the object would be too far away (given something like Relativized Minimality), as we see in (8a) compared to (8b). (8) — is wise [John to crave mercy] a. John is wise [John to crave mercy] b. *Mercy is wise [John to crave mercy].
So, in order for this raising rule to be updated and accommodated within a minimalist conception of syntax, we would need to show that the Old Irish verbal noun was no more than a non-finite verbal category, that is, it wasn’t nominal at all, and thus it was capable of undergoing a passivization rule. However, it is not clear that this is the case. Verbal nouns have case and gender, may appear with adjectives and articles and may be plural (in certain constructions). Furthermore, the genitive case-marking of their complement in normal clauses (i.e. sentences like (1)) parallels nominal phrase structure. Hence, it does not seem to be necessarily the case that Old Irish verbal nouns were non-finite verbs, at least in the Würzburg Glosses. Turning now to constructions described by Disterheft as Subject-to-Object raising, we will see that many problems arise if we follow the raising analysis. Subjectto-Object raising is illustrated by examples (9) and (10). (9) Huare nad ro-gaid huisce do thinnacul. since not prf-ask.3s water.acc/nom to bringing.dat ‘Since he didn’t ask for water to be brought.’
(Ml. 124c10)2
(10) Huare nad n-ac-at hi frecndairc gnim-u cosmail-i since not NAS-see-3p in present.dat deed-acc.pl same-acc.pl du denum du Christ. to doing.dat by Christ.dat ‘Since, at present, they don’t see that similar deeds are done by Christ.’ (Ml. 93d14)
Disterheft claims, in support of the raising analysis, that nouns in sentences like the above behave as if they were arguments of the main clause with regard to passivization and case marking. As examples of passivization, she gives the sentences (11) and (12). (11) Do-archet do Israhel húathad ndóine do pv-foretold.pss.3s to Israel.dat few.nom/acc.sg people.gen.pl to
. The noun huisce is ambiguously nominative or accusative in form.
Elliott Lash
chretim di-ib. believing.dat from.3p ‘It was foretold to Israel that few of them would believe.’
(Wb. 4d14)3
(12) Is ed in-choisecht tri sodin coic bliadn-i cop.3s this PV-(LEN)indicated.PSS.3s through this.acc five year-acc.pl dëac do thormuch for a saigul som. ten to adding.dat on his life.dat his ‘It’s this that was indicated thereby, that fifteen years would be added to his life.’ (Ml. 16c10)
The crucial noun phrase in (11) is húathad ndoine ‘few people’ and in (12), coic bliadni dëac ‘fifteen years’, which are claimed to be the subject of the passive matrix verb in each respective sentence. However, this is not obvious. Given the general pattern of VSO word order in Old Irish, it is unclear, without some more detailed theory, why the subjects in these sentences would occur to the right of the PPs do Israhel ‘to Israel’ and tri sodin ‘thereby’. Furthermore, in at least (12), the expected agreement between coic bliadni dëac (plural) and the passive verb in-choisecht ‘is indicated’ (singular) is not present. Also, note that the morphophonology of the passive verb in this sentence is such that it indicates that an argument has been extracted by wh-movement. Here, I am referring to the lenition4 of the verb in-choisecht indicating that a neuter argument has been extracted. This argument is ed, which is in the cleft clause. It seems as though this is the true subject of the passive verb and hence, under the raising analysis there would be two elements competing for the same position. Thus, Subject-to-Object raising with subsequent passivization doesn’t seem to be the appropriate analysis. Finally, the following sentence (13) is given as an example of raising to the object of a verbal noun, with appropriate genitive-case marking: (13) Ar na follaig-dis gud-i nDae di a tabairt for not neglect-impf.3p asking-acc God.gen to their bringing.dat a sin doiri sin. from the.dat captivity.dat that ‘For they didn’t neglect to ask God to bring them from that captivity.’ (Ml. 113d5)
. Pace Disterheft, húathad is ambiguous between nominative and accusative case. . Lenition in Irish turns stops into the corresponding fricatives. In this case, the base of the verb is in-coisig, with no lenition of the initial stop consonant of the verb. Extraction by whmovement is commonly indicated by various morphophonogical mutations, such as lenition. See Thurneysen (1946), Breatnach (1980), McCone (1980), Ahlqvist (1983), Ahlqvist (1985), and Ó hUiginn (1986) for details.
The rise of OV word order in Irish verbal-noun clauses
The logical subject of the verbal noun tabairt, that is Dae ‘God’, is marked as genitive because it seems to be acting as the object of the higher verbal noun. However, one may view this in at least two other ways. First, it could be an ECM construction and, secondly, it could be the case that the argument Dae is merged as the object of the higher verbal noun with the lower verbal-noun clause an adjunct. This second possibility fits in with my analysis. There thus seem to be some grounds for rejecting the raising analysis. In the following sections I will develop a new analysis of these constructions that is fundamentally diachronic in nature. I will continue to refer back to the raising analysis and show how my new analysis accounts better for the data.
3. Verbal nouns in copular clauses In the earliest Old Irish, the Würzburg Glosses, the [do + Verbal-Noun] construction is frequent only in predicative/copular sentences, as demonstrated in (14). This is the environment that Disterheft labels ‘Raising-to-Subject’. (14) Ní date le-u in coimdiu do chrochad. not pleasing.nom to-3p the.nom lord.nom to crucifying.dat ‘The Lord is not pleasing to crucify, in their opinion.’ (Wb. 8a6)
According to the raising analysis, the phrase in coimdiu has raised to the subject of the matrix predicate. The negative particle ní is functioning as the third person singular of the copula; the plural form being nítat. Given the singular nature of the copula in this instance, we may posit Agree relation between it and the subject of the predication. A superficially similar type is found in (15). Here again, according to the raising analysis, the argument NP of the lower verbal-noun clause is raised to the subject of matrix predicate. However, here there is a lack of agreement between the copula is (3rd sg.) and in daim (3rd pl.),5 the element which seems to be in the same position as in coimdiu occupies in (14). (15) Is bés le-o som in daim do cop.3s custom.nom with-3p them the.nom oxen.nom.pl to thúarcuin ind arb-e. threshing.dat the.gen corn-gen ‘It is a custom with them that the oxen thresh the corn.’
(Wb. 10d6)
. A reviewer points out that the copula agrees with the noun bés. This is superficially the case, as both are singular in form. However, it is unclear to me how a predicate nominal, which undoubtedly bés is in this context, can control subject–verb agreement. No other agreement in the language works like this; verbs and verb-like elements such as the copula agree with only subjects, not predicates.
Elliott Lash
Usually, however, there is complete agreement between the copula and the subject of predication. This is seen in (16). Here, the copula is negative, but it still agrees with the plural subject (plural nítat instead of singular ní). (16) Ni-ta-t aecn-i huil-i. not-cop-3p wise-nom.pl all-nom.pl ‘All are not wise.’
(Wb. 8a7)
Not only does the raising rule fail to account for the lack of agreement, it also fails to account for the semantic incongruity that may sometimes result. In (15), raising in daim ‘the oxen’ from the position of the subject of the verbal noun to subject of the main-clause predication would result in the incongruous situation of predicating the noun ‘custom’ of the subject ‘oxen’. However, as a predicate, the noun ‘custom’ most naturally takes an activity as its subject. This is unlike standard raising predicates, such as ‘appear’ or ‘seem’ which do not require any semantic subjects. Even non-verbal raising predicates like ‘certain’ in a clause like ‘John is certain to win’ do not have a semantic subject. The predicate in (15), however, requires some element denoting an activity; in other words, if the DP in daim were raised from the verbal-noun phrase to the matrix subject position, the derivation would fail at LF. The sentence would have an exceedingly odd interpretation, something like ‘*?The activity denoted by oxen is a custom amongst them and oxen thresh the corn’, instead of the required ‘The activity denoted by oxen threshing corn is a custom amongst them’. Therefore, instead of a Raising-to-Subject account in sentences with the copula, I would like to propose that the unusual agreement patterns and the placement of the nominal in relation to the verbal noun should be viewed as evidence for the extension of syntactic patterns that result from a bracketing reanalysis. Given input like example (14), repeated here as (17) and (18), an acquirer is faced with a somewhat ambiguous structure. This is indicated below by the two different translations, glosses and bracketing patterns. (17) Ní date le-u in coimdiu [do chrochad.] not pleasing.nom to-3p the.nom lord.nom [to crucifying.dat] ‘The Lord is not pleasing [for PROarb crucifixion] in their opinion.’ (18) Ní date le-u [in coimdiu do chrochad.] not pleasing.nom to-3p the.nom lord.nom to crucifying.dat ‘[To crucify the Lord] is not pleasing in their opinion.’
In translation (17), in coimdiu ‘the lord’ is viewed as the subject of the main-clause predicate date ‘pleasant’, the verbal-noun clause is an adjunct. In translation (18), it is viewed as syntactically part of the verbal-noun clause. Because the noun is
The rise of OV word order in Irish verbal-noun clauses
ominative rather than genitive (the normal case for objects of verbal nouns), it n seems that the acquirer would be forced to analyse it as a subject. Therefore, if the acquirer of Old Irish analysed the sentence such that in coimdiu were part of the verbal-noun clause the sequence in coimdiu do chrochad could be viewed as a predicative small clause meaning something like ‘(that) the Lord (is) for crucifying’, with nominative marking on the subject of predication as in other predicative situations in Old Irish. This entire small clause is then analysed as the subject of the main predicate. Thus, instead of synchronic raising we have a diachronic change in bracketing and the introduction of a predicative head (Pr0) mediating the relationship between the two elements in the small clause. The singular number of the copula in the main clause arises by agreement with a null determiner that selects the small clause (corresponding to ‘(the fact) that’ in the English translation). Since the null determiner is likely underspecified for phi-features, the copula is given third singular agreement by default. This is shown in (19). (19) Is bés le-o som [DP [D] [PRP [DP cop.3s custom.nom with-3p them [phi[default]] daim] [ Pr¢ Pr0 [ PP oxen.nom.pl
in the
do thúarcuin ind arb-e.] ] ] ] for threshing.dat the corn-gen
‘[(The fact) that oxen (are) for threshing the corn] is a custom amongst them.’
The analysis in this section accounts for all of the following issues. First, the appearance of agreement mismatch is explained because third singular arises through an Agree relation between the copula and the null D0 that is the head of the subject free-relative. Second, the semantic incongruity is erased: the predicative phrase headed by Pred/Pr0 denotes an activity, thus being compatible with predicates such as bés ‘custom’ in (19). There is no need to posit a raising rule that targets subject and object indiscriminately, or to assume that the verbal noun is in this case a non-finite predicate capable of passivization. Rather, the facts are accounted for by viewing the preposed nominal as a subject of a predicative small clause. Thus the issue is resolved by idea that the operation Merge may take any two syntactic elements and create a third. In this case the two elements would be a DP and the complex object created by merging together the PP with the head of PredP. Finally, I assume that the nominative case of the element in SpecPredP is assigned by Pred/Pr0. Below, this diachronic change will be called Change A and the clause type that was created by the change and its extension, Type A. This clause type is essentially the same as the type subsumed under Disterheft’s Raising-to-Subject rule. However, as I have said, this is a diachornic change. The synchronic facts would be that the prepositional phrase [do + Verbal Noun] could appear in the complement of a predicative
Elliott Lash
head (Pred/Pr0) while a subject could appear in the specifier of this head.6 Once innovated as the result of the reanalysis, this structure could be spread to other contexts or could interact with other structures created by similar reanalyses. We shall see that this is exactly what occurred when we turn to Disterheft’s Subject-to-Object raising constructions in the next section.
4. Do + verbal noun in non-copular clauses With the exception of its use in the copula clauses discussed above, the structure [do + Verbal Noun] seems to be not as common in other types of clauses, where the accusative verbal noun is more frequently found, at least in the Würzburg Glosses. Gagnepain (1963) notes only two examples of the [do + Verbal-Noun] construction as the complement of transitive verbs in the Würzburg Glosses. However, as an adjunct, [do + Verbal Noun] was found very frequently indeed in the glosses. In these cases, the adjunct clause follows some nominal element, which often is an object of a verb or the complement of a preposition (although subjects can occur as well). Occasionally, due to its adjunct status, the verbal-noun phrase does not directly follow the noun to which it is semantically ‘connected’. Some examples of this adjunct use of the verbal-noun phrase are given below in (20), (21) and (22). (20) It inn-a tairngere du-r-airnger-t Dia do cop.3p the-nom.pl promise.nom.pl pv-prf-promise-pst.3s God.nom to Duaid [du soirad in popuil.] David.dat [to delivering.dat the.gen people.gen] ‘They are the promises that God promised to David to deliver the people.’ (Ml. 108b8) (21) briathar inso [do fuilled i ssin salm.] verb.nom this [to adding.dat in the.dat psalm.dat] ‘This is a verb to add to the psalm.’
(Ml. 69b6)
. It is also interesting that, in Wb., one occasionally finds glosses that consist solely of this structure (i.e. without an overt matrix verb to trigger any type of raising). An example of this is gloss 23a7:
suánem-uin do dénum in aidch-i rope-pl.nom to making.dat in night.dat ‘ropes (are) for making at night’
The rise of OV word order in Irish verbal-noun clauses
(22) Is Dia as tosach [du thuistin cop.3s God.nom rel.cop.3s first.nom [ to creating.dat cech dul-o.] each element-gen] ‘It’s God who is first to create each element.’
(Ml. 85b16)
Many examples of this sort are cases that have been analysed as involving Subjectto-Object raising. Under the raising analysis, the nominal element has reached its ‘surface’ position via raising from the subject position of the verbal-noun clause. We have already seen examples of this type. I repeat them here for clarity as examples (23) and (24) as well as adding examples (25) and (26). We must note first that example (26) is said to fall under the Subject-to-Object raising rule, in that remdéicsin ‘providence’ is the logical subject of the verbal noun but has accusative case. However, this example in fact seems to contradict the raising rule since there already seems to be an object in the matrix clause, namely the pronoun són. Presumably, a derivation in which two items compete for the same position should crash if they both wind up in that position (cf. example (12)). I will use example (26) below to argue for extension of the proposed reanalyses. (23) Huare nad ro-gaid huisce do thinnacul. since not prf-ask.3s water.acc to bringing.dat ‘Since he didn’t ask for water to be brought.’
(Ml. 124c10)
(24) Huare nad n-ac-at hi frecndairc gnim-u cosmail-i du since not NAS-see-3p in present.dat deed-acc.pl same-acc.pl to denum du Christ. doing.dat by Christ.dat
‘Since, at present, they don’t see that similar deeds are done by Christ.’ (Ml. 93d14)
(25) Rocúala in n-eress cetna do forbairt. heard.3s the.acc NAS-heresy.acc same to increasing.dat ‘He heard that that very heresy was increasing.’
(LH 311.34)
(26) Do-ru-sluind-s-et són remdéics-in Dae du buith pv-prf-deny-pst-3p that.acc providence-acc God.gen to being.dat dí-ib di an ícc. for-3p to their saving.dat ‘They have denied the aforementioned that God’s providence is among them.’ (Ml. 90b17)
These examples differ in two ways from the structure that provided the input to Change A. First, the matrix predicate is not a non-verbal copular predicate; rather, it is verbal. The second difference is that the logical object (or subject) of the verb is accusative. In three of the examples, (24), (25) and (26) accusative case is marked explicitly
Elliott Lash
by the morphology. Like the structure underlying Change A, these examples seem to have structures that are somewhat ambiguous, and hence they may have been open to reanalysis by the acquirer. While it is unclear why an analogous change did not occur at the stage that Type A was introduced (i.e. pre-Würzburg), it is clear that by the time of the Milan Glosses (early 9th c.), a change had occurred. In the Milan Glosses we find an increase of the frequency of examples of a [do + Verbal-Noun] clause used as the complement of verbs. It is likely that such an increase is the result of reanalysis very similar to that which produced Type A. An acquirer hearing sentences like (27a) may have analysed the sequence either as (27b) with an adjunction structure or as (27c) with shift in the placement of brackets. (27) a.
Huare nad ro-gaid huisce do thinnacul. since not prf-demand.3s water.acc for bringing.dat ‘Since he has not asked for water to be brought.’ (Wb. 124c10)
b. c.
PP-adjunct: [do thinnacul.] ‘[to be brought/for bringing]’ COMPLEMENT: [huisce do thinnacul.] ‘[that water be brought.]’
An essential difference between this reanalysis and Change A is that the nominal element is marked with accusative case because it was originally the object of the mainclause verb. This case difference did not allow the acquirer to posit the same structure as in Type A for this new sequence. More specifically, the acquirer could not posit a predicational relationship between the noun and the PP containing the verbal noun, since subjects of predication in Old Irish had nominative case. Rather, the reanalysis represented by (27c) created a new structure,7 which I will label Type B. Because Type B has an accusative marked noun, we must assume that some element has been introduced into the phrase that can license accusative case. This means that the verbal noun, in this case tinnacul ‘bringing’, is no longer a nominal, but rather has become verbal in nature. This may be represented as (28) becoming (29). (28)
nP n
NP verbal noun
. A reviewer asks if this step is strictly necessary because the the two constructions, Type A and Type B, are superficially so similar. It seems preferable that the two types should be kept separate, if only to account for the chronology of events. First, Type A was innovated prior to the Würzburg Glosses and then type B was innovated between Würzburg and Milan. Of course, the argument I am making is that Type A and B were so similar (especially after the distinction between accusative and nominative became less robust) that, once Type B was innovated, it likely didn’t last very long as a type separate from A.
The rise of OV word order in Irish verbal-noun clauses
(29)
vP v*
VP verbal noun
This v*-head licenses accusative case on the object. Furthermore, because the verbal noun no longer had nominal functional structure above it, the preposition do anomalously selected a vP, rather than the usual nominal categories. This led to a concomitant shift of PP to some functional element. This shift of prepositions to functional elements is well documented in the literature (see for example Roberts and Roussou (2003), Hopper and Traugott (1993) and Heine and Kuteva (2002)) and was helped along in Old Irish, due to the morphological impoverishment of the verbal noun. The verbal noun was inflected in four cases: nominative, accusative, genitive and dative. These cases were only distinguished in a few of the stem types and there was almost never a distinction between the accusative and the dative. Table 1 gives the declension of verbal nouns from the major stem types. Because the genitive is not used in complementation, I have omitted it from this table. It is significant to note that the final two stem types are particularly common. Table 1. Old Irish verbal-noun declensions: data from Thurneysen (1946: 446–454) TYPE
NOM
ACC
DAT
TRANSLATION
o-stem jo-stem ā-stem jā-stem i/a-stem u-stem n-stem (f.) n-stem (n.) -ad -iud
accomol fuine caithem claide brith (breith) ammus airitiu réim mórad léiciud
accomol fuine caithim claidi brith (breith) ammus airitin réim mórad léiciud
accomul fuiniu caithim claidi brith (breith) ammus airitin réim(imm) mórad léiciud
adding cooking consuming digging bringing attempting accepting riding praising allowing
The two cases used when the verb functioned as a complement of a main clause are accusative (on its own or with a preposition) and dative (with a preposition). That they were not distinct for the majority of the verbal-noun types led to a situation in which it appeared to acquirers that the PP do failed to case mark the following verbal noun. This is the same situation discussed in Los (1999) for the development of infinitival to in English, which developed from the preposition to when the case of a following verbal noun became ambiguous, thus allowing for the reanalysis of PP as some other functional category dissociated with case marking.
Elliott Lash
Change B, then, is a three-fold change that can be described as follows. First, bracketing ambiguity caused reanalysis from [DP-object] [PP-adjunct do+VN] to [FP object do VN]. Then, case marking of the object likely caused acquirers to posit a v* dominating the VN, in order to license accusative case. This shift in functional category caused a reanalysis of PP to some type of functional category (which we will discuss in detail below); a shift that was facilitated by the lack of overt case marking in many of the verbal-noun types. After Change B created Type B at some point prior to the mid-ninth century, we see a spread of this type of phrase in the Milan Glosses, where complementation or adjunction with [object do VN] is introduced into sentences where accusative case cannot be licensed in the main clause. In the following examples of Type B, I have glossed the element do as PTC, however, I will discuss the category of this element further below. Some examples of Type B are: (30) Ad-fiad-ar [ airechas du thabairt do macc.] pv-show-pss.3s [ leadership.acc ptc giving.dat to son.dat] ‘It is shown that leadership is given to the son.’ (Ml.128a15) (31) Do-ru-sluind-s-et són [remdeéics-in Dae du pv-prf-denied.pt.3p that [ providence-acc god.gen ptc buith dí-ib.] being.dat among.3p] ‘They have denied the aforementioned that God’s providence is among them.’ (Ml. 90b17) (32) Con e-r-bair-t fri-a [na lin-a do chor i so pv-prf-said-pst.3s to.3sf [the net-pl.acc ptc putting.dat in sin usce.] the.dat water.dat] ‘So that he said to her to cast the nets into the water.’
(LH 345.21–22)
Having established that Type B represents the result of a possible reanalysis, we should now examine the motivation for the OV word order in the post-reanalysis grammar. One hypothesis that we may immediately reject is that the object is fronted in the presence of v* in non-finite clauses. However, in Old Irish, it is unclear whether or not the object is fronted in the presence of v* in finite transitive sentences, but as the null hypothesis we should posit that it does not. Why then would the presence of v* in non-finite transitive clauses cause fronting? Instead, I argue that the object is moved to a preverbal position due to the features of do in its new role as a functional category. Given that it is no longer a preposition (it cannot robustly case-mark and it selects for a verbal element), it may be a type of C0, although we shall see that there are arguments against this as well. Accepting for the moment that it is a type of complementizer, we may draw a parallel between this complementizer and the (non-overt) C0 in cleft sentences and other sentences
The rise of OV word order in Irish verbal-noun clauses
having focus interpretation. The focus bearing C0 in such sentences is the only type of complementizer that induces fronting in Old Irish. Example (33) gives an instance of a cleft sentence, in which the predicate pronoun éside is fronted to pre-subject position. Note that the subject here is represented as a free-relative clause (CP). (33) [ CP is é-side Cfoc dano [ CP as éola [ CP cop.3s he-emph C then [ CP cop.3s.rel knowing.nom and] éside] therein] he-EMPH] ‘Its he, then, who is knowledgeable about it.’
(Wb. 6b25)
It seems then that one possibility for the category of do in Type B constructions is as a focalizing element. One of the major findings of Genee’s (1997) work is that, in most cases where a verbal-noun clause has a preposed object nominal, the preposed NP/DP has a focalized interpretation: either representing new/presentational information or having a contrastive force. Genee’s data for the early period (8th and 9th cc.) concentrate on the more common Type A construction. An example of focus in this construction is shown below in (34). Here, in fid ‘the tree/wood’ is being contrasted with fid aile ‘another tree/wood’. (34) Ar is bésad le-u som [ in fid do for cop.3s custom.nom with-3p them the.nom tree.nom ptc thób-u] ET [fid ail-e do esnid and.] cutting-dat and tree.nom other-nom ptc inserting.dat therein ‘For they have a custom to cut the tree and insert another tree therein.’ (Wb 5b42)
For Type B, an example of contrastive focus is found in the Tripartite Life of St. Patrick, which is from the late ninth century or early tenth century. This is shown in (35). Here, ‘the druid’ is being contrasted with Binén (proper name). (35) Fri-th arol-i comarl-i ann … [in draí do find-pss.3s another-nom counsel-nom therein … [the druid.nom ptc chur i sin leth úr] … [Binén do putting.dat into the.acc side.acc new.acc] … [Binén.acc ptc thabairt i sin leth crín.] giving.dat into the.acc side.acc old.acc] ‘Another counsel was taken therein to put THE DRUID into the new side (and) to put BINÉN into the old side.’ (Trip. 26:13)
We might say that do in some way marks this focus interpretation in this example. Notice that the first part, in draí do chur ‘to put the druid’, is Type A as the preposed nominal is nominative, while the structure of the second part, Binén do thabairt
Elliott Lash
‘to place Binén’, is unclear as Binén shows no distinction between accusative and nominative. Further examples of a do + VN phrase as a focus-marking strategy can be found in many different texts after the reanalyses argued for above occurred. A particularly striking example (37) is found in the poem Saltair na Rann (10th c.). The first example (36) gives the context and the second gives the phrase in the focus structure. (36) At-ber-t Día d’ Ádam co glan, do chin húair pv-say-pst.3s God.nom to Adam.dat adv clear your sin.nom since na-ch a-tam-ar not-it pv-confess-sbj.2s ‘God said to Adam clearly, since you do not confess YOUR SIN … ’ (SnR: 1405–06) (37) Do-gén-at do chlann-a iartain dogrés a cin pv-do.fut-3p your descendant-pl afterwards always their sin.nom d’ imresain. ptc contesting-dat ‘Your children will ever afterward contest THEIR SIN.’
(SnR: 1407–08)
Sentence (37) is particularly interesting because of the use of the verb do-gní (here future dogénat) ‘do’ as a sort of auxiliary. This usage is found repeatedly in combination with a do + VN phrase. This may indicate that the introduction of a focus structure in the left-periphery of VP for some reason prevents V-to-T movement, so that a dummy element is inserted in T, i.e. an Irish version of do-support. Whatever the analysis here, it is clear that (36) introduces a unit do chin ‘your sin’, which contrasts with the displaced argument in the verbal-noun phrase, a cin ‘their sin’. Other examples of focus are examples (38), (39) and (40). (38) Ro-indis do Dáre a ech do marbad do n prf-tell.3s to Dáire.dat his horse.nom ptc killing.dat to the.dat chlerech. Dixit Dáre in clerech fen do cleric.dat said Dáire the.nom cleric.nom self ptc marbad índ. killing.dat due.to.3s ‘He told Dáire that HIS HORSE was killed by the cleric. Dáire said that he would kill the cleric himself because of that.’ (Trip. 34:8–9) (39) do seinm suantraige do nt sluag innund & to singing.dat sleep-songs.gen to the.dat host.dat inside & in sluag ammuig do thabairt an aigth-e fri the.nom host.nom outside ptc giving.dat their face-gen to
The rise of OV word order in Irish verbal-noun clauses
lár ammuig do thabairt an aigth-e fri lár ground.acc outside ptc giving.dat their face-gen to ground.acc ‘to sing a sleep-song to THE HOST INSIDE and THE HOST OUTSIDE would put their faces to the ground.’ (ODR. 399) (40) Fé amai mo mac sa do bith immaig a oenur Woe alas my son.nom my ptc being.dat outside his lonesome ocus a lín di a ta-bair mathius. and his multitude.nom to whom pv-give.3s wealth.acc (here) ‘Woe, alas that MY SON is out alone, while his multitude to whom he gives wealth (is here).’ (FR 105)
In each of these examples one constituent is being contrasted or placed in opposition with another element. This seems to be among the major functions of the [do + VN] structure, whereas non-focus interpretation is usually found with postposed genitive objects, i.e. in verbal-noun constructions that have remained nominal in nature and that have not and could not have undergone the same series of reanalyses. There are two choices with regard to the category of do: it might be a C0 element, as we have been assuming thus far. If this were the case, we could identify it with FocP, following Rizzi (1997), i.e. a functional category that bears a focus feature and an edge feature driving movement to its specifier. This would be directly parallel to main-clause focus-driven movement in cleft constructions as well as other sentences with focus interpretation in Old Irish. The second possibility would be for do to be some light-verb head with a focus feature. In this case it would be a focus-bearing head in the ‘left-periphery’ of VP, following Belletti (2004). If it were a light verb, one would not need to posit a non-overt category in this structure: do would both license accusative case and drive focus movement. In both scenarios it has a focus feature and an edge feature to drive movement. We shall see below that there may be a theoretical argument against the C0 analysis of do and in support of a left-periphery of VP analysis.
5. Middle and Modern Irish developments It was noted above that both Type A and Type B seem to have had a focus interpretation. This, in addition to the fact that the case alternations in Old Irish rapidly broke down in the Middle Irish period according to Breatnach (1994), contributed to a situation in which the distinction between Type A and Type B was no longer robustly present in the input. In Old Irish, nominative cases and accusative case were distinguished in various ways depending on the stem class of the noun. See Table 2, which gives an
Elliott Lash
example of both cases from each of the main stem classes. The cases in which the accusative is different than the nominative are in bold. Table 2. Noun declensions: data from Thurneysen (1946: 176–217) TYPE
NOM.SG
ACC.SG
NOM.PL
ACC.PL
TRANSLATION
o-masc. o-neut. jo-masc. jo-neut. ā-stem jā -stem i-stem i-stem u-masc. u-neut. g-stem th-stem t-stem n-stem nn-stem r-stem s-stem
fer scél cél-e crid-e túath soils-e sétig súil mug dorus rí cin carae noídiu ainm athair glenn
fer scél cél-e crid-e túaith soils-i sétch-i súil mug dorus rí-g cin-aid car-ait noíd-in ainm athair glenn
fir scél(-a) cél-i crid-e túath-a soils-i sétch-i súil-i mog-ae dorus (doirs-ea) rí-g cin-aid car-ait noíd-in anm-ann aithir glinn-e
fir-u scél(-a) cél-iu crid-e túath-a soils-i sétch-i súil-i mug-u dorus (doirs-ea) rí-ga cin-ta cair-tea noíd-nea anm-ann aithr-ea glinn-e
man story spouse heart people light wife eye slave door king sin friend child name father valley
In Table 2 we see that there is quite a high degree of overlap between the two cases. According to Breatnach (1994), this overlap increased as the language became Middle Irish. The distinctions in the consonant stems were, for the most part, erased. Nominative and accusative singular became indistinguishable and to a large extent this occurred in the plural as well. In ā-stems, the distinction between accusative and nominative was lost in favour of the accusative. In the jā- and jo-stems the distinction between nominative and accusative was lost due to the neutralization of final vowels to schwa. In consonant stems, as Breatnach shows, there was wide spread use of the former nominative for accusative and vice versa, both in the plural and the singular. For example, he notes cathracha ‘cities’ as both nominative and accusative plural, although originally it is accusative. Similarly, he notes that cathraig ‘cities’ may be used for both nominative and accusative plural, though it is originally nominative. This wide-spread neutralization almost completely obliterated the distinction between Type A and Type B. This case neutralization affected not only the nominal in Type A and Type B, but the rest of the phrase as well. As case distinctions were erased, mostly in favour of the accusative, the evidence for the distinction between two was lost and the acquirer may have posited a verbal functional structure for all clauses, creating what I will call Type C, which is essentially the same as Type B, but is the product of a reanalysis of Type A.
The rise of OV word order in Irish verbal-noun clauses
This reanalysis, Change C, was also helped by the fact that Type A and Type B seem to have shared focus interpretation. Acquirers may have been led to create a unified treatment of focus in their new grammars. Type C only arose from Type A in contexts where there was no postposed genitive object of the verbal noun. Hence (15), repeated here as (41), would not have been reanalysed as Type C, since the evidence that this was the same as Type B is clearly not present. Constructions in which the object of the verbal noun was marked as genitive as in (41) would have likely been replaced as the frequency of Type C rose (which according to Genee (1997) seems to have happened sometime in the eleventh or twelfth centuries).8 In Modern Irish it is rare and limited to certain dialect areas. (41) Is bés le-o som in daim do cop.3s custom.nom with-3p them the.nom oxen.nom.pl for thúarcuin ind arb-e threshing.dat the.gen corn-gen ‘It is a custom with them that the oxen thresh the corn.’
(Wb. 10d6)
The next step in the development from Middle Irish to Modern Irish verbal-noun clauses is the loss of focus interpretation. Genee (1997) argues that focus interpretation was a short-term stage and was lost rapidly in Middle Irish. A corpus study that was conducted for texts between 850 to 1050, in Lash (2007),9 showed that focus interpretation is in fact rather stable throughout this time period. It is thus somewhat uncertain around what point focus interpretation was lost for the moved object/subject of intransitive. That being said, Genee’s last text, Betha Colaim Chille (1532) shows no sign of focus interpretation and it seems to be on the wane from the early twelfth century at least. In any event, it is certain that at some point focus interpretation was lost, as Modern Irish OV order is not pragmatically driven. Syntactically, loss of focus interpretation has to be related to a reanalysis of do. We have seen that, in Old and early Middle Irish, this element has become the realization of a head bearing a focus feature and an movement-inducing edge feature found in Type C constructions. Having discussed the Old and Modern Irish changes, we may speculate, to some extent, about how the early Type C construction became the construction found in the modern language. In Modern Irish, two systems are found, according to Guilfoyle (1997) and Ó Siadhail (1989). The northern system . A reviewer points out that this type is found in seventeenth-century texts. I have not checked this; however, it is worth noting that seventeenth-century texts, such as Keating’s Foras Feasa ar Éirinn (1634) and Ó Maolchonaire’s Desiderius (1616), are in general fairly conservative in their language. . This is an unpublished MPhil thesis. The dates 850 to 1050 were chosen in order to fill in a gap in Genee’s data for early Middle Irish.
Elliott Lash
(Donegal and Connaught) is shown in (42), while the southern system (Munster) is shown in (43). (42) INTRANSITIVE [S Intrns-VN] — TRANSITIVE [O a Trns-VN] (43) INTRANSITIVE [S a Intrns-VN] — TRANSITIVE [O a Trns-VN]
Note that, in the north, the particle a (equivalent to Old and Middle Irish do), appears only with transitive verbal nouns when the object is fronted. In the south, on the other hand, the particle appears before both intransitive and transitive verbal nouns, whenever an argument precedes the VN. One can take this distribution to mean that in the north the particle a is a marker for transitivity and non-finiteness together, whereas in the south it is only a marker for non-finiteness. Following Bowers (2002), the difference between a transitive and an intransitive is the presence or absence of a transitive light verb, Tr (or v*). Thus, the northern transitive verb facts in (42) may be derived from the structure in (44) with do/a as the exponent of the light verb head and the intransitive verb facts may be derived from the structure (45) without a light verb. The southern facts, on the other hand, may be derived from locating the particle do/a in the head of CP (or FinP) as shown in (46). CP
(44)
PredP
C[non−finite] subject
Pred′ Pred
vP object
v′ do[EPP]
verbal noun
CP
(45)
C[non−finite]
PredP subject
VP
Pred′ Pred verbal-noun
The rise of OV word order in Irish verbal-noun clauses
(46)
CP subject/object
C′ do[non−finite]verbal noun
Both of these Modern Irish structures seem to suggest that, for Old and Middle Irish, a focus position in the left periphery of C is too high. The reanalysis of do to a transitive light verb in the north would only be possible if do was already low in the structure, as Roberts and Roussou (2003) have shown that, when elements grammaticalize, they tend to be found higher than, rather than lower than their original position in the tree. Similarly, if the particle do/a is associated with finiteness in the south, it seems likely that it originated further down, perhaps in a focus position in the left periphery of VP.10 With this in mind, we may picture the Early Modern Irish acquisition process as follows. As the focus meaning became conventionalized, the placement of the particle do/a became ambiguous between two positions: a low vP position and a high position associated with finiteness. Thus, the acquirer could have mapped the particle either to a lower position in the tree (as it presumably was in their parents’ grammar, as a focus marking v0) or they could map it to a higher position. If it was mapped to a low position, it became associated with non-finite transitive verbs alone, whereas if it was mapped to a high position, it became the exponent of non-finiteness generally.11
6. Conclusion In this chapter I have shown that the raising analysis to OV word order in verbal noun constructions must be abandoned. Rather, we have seen that OV word order in Irish verbal-noun clauses arises from a variety of factors. First a series of bracketing ambiguities forced acquirers to reanalyse both the verbal noun itself and subsequently the preposition do. These shifts caused a nominal constituent from the main clause to be reinterpreted as a part of the verbal-noun clause. In Change A, a predicative small clause structure was created, shown in (47), whereas in Change B the structure became
. In terms of the proposal in Roberts and Roussou (2003: 198–199), the change in the southern variety could be seen as roughly equivalent to the loss of Agree between a higher head (the non-finite C) and a lower head (the focus-marking v* in the left-periphery of vP), such that the features once associated with v* become associated with C0. This reanalysis is proposed for complementizers such as mu (Calabrian), na (Greek) and to (English). . Obviously, these remarks are preliminary and more work is necessary in order for the full picture of the transition to Modern Irish to become clear.
Elliott Lash
fully verbal (due to the accusative case of the new nominal argument). This is shown in (48). (47) [DP D0[phi] [CP C0 [PredP [DP] [Pred′ Pred0 [PP do + Verbal Noun]]]]] (48) [vP [DP-object] [v′ v0 [VP verbal noun]]]
At least in the second construction, the verbal noun was reanalysed as fully verbal (but non-finite). In this construction, the preposition do was reanalysed as a functional element: a focus particle in a low vP focus position. After these reanalyses took place, these structures spread to contexts in which the nominal could only be licensed in the lower clause. Because of this extension in usage, we see a gradual decrease in the frequency of the accusative verbal noun with a genitive-marked object construction, such that in Modern Irish OV is the norm. Then, due to the breakdown of the case system, the difference between Type A and Type B was no longer robust and hence Type C arose, as a generalized focus-marking non-finite construction. Finally, with a late Middle/early Modern Irish loss of focus interpretation, the OV word order becomes fully conventionalized. Another reanalysis of do/a results in the difference between the southern and northern varieties of Irish. Before concluding this chapter, I would like to look at two issues that arise from the various reanalyses of the preposition do that have been suggested here. First, in addition to the commonly observed P-to-C grammaticalization, we now see that P-to-v grammaticalization is a possible grammaticalization pathway as well and an option that has been taken in the course of the history of Irish. In fact, this is grammaticalization pathway is particularly common in Irish (and Celtic languages generally) resulting in a number of aspectual markers, such as Irish ag ‘progressive’ from the preposition oc ‘at’.12 If we assume Roberts and Roussou’s (2003) grammaticalization proposal, i.e. grammaticalization proceeds upwards along the clausal spine, not downwards, then we may make a prediction regarding the possibilities open to P if it is grammaticalized as either a C or a v. The main prediction is that, if P-to-C grammaticalization occurs, then further grammaticalization of the same element to a lower position in the same clause will not be possible. However, if P-to-v grammaticalization occurs, then the element has the possibility of either staying low, in the space occupied by various v-heads, or moving higher in the clause as happened in the southern variety of Irish. Finally, if we assume grammaticalization is in an ‘upward’ direction, we may need to rethink some aspects of the cartographic approach. If we think that each feature has
. A reviewer queries on what grounds an aspectual marker can be compared to little v0. Here I am following the idea that aspectual heads are within the extended projection of VP under TP (see for example Travis (2000), and Ritter and Rosen (1998)). If this is the case then we can at least compare v0 heads with aspectual heads in that they both appear in the extended projection of VP.
The rise of OV word order in Irish verbal-noun clauses
a dedicated head position, we would need to posit a representation like (49) for the Old/Middle Irish situation: (49) [C-nonfinite [v-Foc = DO [v*… ]]]
Here, do is merged in vFoc, the left-periphery of vP. Acquirers of the northern variety would then grammaticalize downward (to the normal v* position), while acquirers of the southern variety would grammaticalize upward (to the non-finite C0 position). If, on the other hand, we think that a strict cartographic approach should be replaced by a view of functional heads in which they may be filled by multiple features, we may maintain the insight of the Roberts and Roussou approach to grammaticalization and account for the reanalyses posited in this chapter.
Primary Sources Greene, David, (ed.). 1975. Fingal Rónáin and other stories. Dublin: Dublin Insitute for Advanced Studies, 2nd edition. Bernard, J.H., and R. Atkinson, (eds.). 1898. The Liber Hymorum: Edited from the MSS with Translations, notes and glossary, volume 1: Texts and Introduction. Henry Bradshaw Society. Stokes, Whitley and Strachan, John (eds.). 1901–3. Thesauraus Palaeohibernicus, volume 1: Biblical glosses and scholia. Cambridge: Cambridge University Press. Stokes, Whitley (ed.). 1883. The Saltair na Rann, volume 1 of Anecdota Oxoniensia, intro. vi–155. Oxford: Clarendon Press. Mulchrone, Kathleen, (ed.). 1939. The Tripartite Life of Patrick: Edited with Translation and Notes. Dublin: Royal Irish Academy.
References Ahlqvist, Anders. 1983. Two notes on relative marking in Old Irish. Celtica 15: 10–12. Ahlqvist, Anders. 1985. The syntax of relative marking in Old Irish. Folia Linguistica Historica 6: 323–346. Belletti, Adriana. 2004. Aspects of the low IP area. In The Structures of CP and IP: The Cartography of Syntactic Structures, Luigi Rizzi (ed.), 16–51. Oxford: OUP. Bowers, John. 2002. Transitivity. Linguistic Inquiry 33: 183–224. Breatnach, Liam. 1980. Some remarks on the relative in Old Irish. Ériu 31: 1–9. Breatnach, Liam. 1994. An Mheán-Ghaelige. In Stair na Gaelge in ómós do: Pádraig Ó fiannachta, Liam Breatnach, Kim McCone, Damian McManus, Nicholas Williams, and Cathal Ó hÁinle (eds), 221–333. Maigh Ruad: Roinn na Sean-Ghaelige, Coláiste Phádraig. Chomsky, Noam. 1995. A Minimalist Program for Linguistic Theory. Cambridge MA: The MIT Press. Disterheft, Dorothy. 1982. Subject raising in Old Irish. In 5th International Congress of Historical Linguistics [Current Issues in Linguistic Theory 21], Anders Ahlqvist (ed.), 183–224. Amsterdam: John Benjamins.
Elliott Lash Disterheft, Dorothy. 1984. Irish complementation: A case study in two types of syntactic change. In Historical Syntax [Trends in Linguistics, Studies and Monographs 23], Jacek Fisiak (ed.), 89–107. Berlin: Mouton. Gagnepain, Jean. 1963. Le syntaxe du nom verbal dans les langue celtiques, 1: Irlandais. Paris: Librairie C. Klincksieck. Genee, Inge. 1997. Sentential Complementation in a Functional Grammar of Irish [LOT International Series 7].The Hague: HAG. Guilfoyle, Eithne. 1997. The verbal noun in Irish nonfinite clauses. In Dán do oide: Essays in Memory of Conn R. Ó Cléirigh, Anders Ahlqvist & Véra Čapková (eds), 187–200. Dublin: Institiúid Teangeolaíchta Éireann. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Hopper, Paul & Traugott, Elizabeth Closs. 1993. Gramaticalization. Cambridge: CUP. Lash, Elliott. 2007. Syntacic Developments in Irish Complementation. MPhil dissertation, University of Cambridge. Los, Bettelou. 1999. Infinitival Complementation in Old and Middle English [LOT Dissertation Series 31]. The Hague: Holland Institute of Generative Linguistics (Theseus). McCone, Kim. 1980. The nasalizing relative clause with object antecedent in the glosses. Ériu 31: 10–27. Ó hUiginn, Ruairí. 1986. Old Irish nasalizing relative clauses. Ériu 37: 33–86. Ó Siadhail, Mícheál. 1989. Modern Irish: Grammatical Structure and Dialectal Variation. Cambridge: CUP. Ritter, Elizabeth & Rosen, Sarah. 1998. Delimiting events in syntax. In Projection of Arguments: Lexical and Syntactic Constraints, Miriam Butt & Willy Geuder (eds.), 135–164. Stanford CA: CSLI. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar, Lilliane Haegeman (ed.), 281–338. Dordrecht: Kluwer. Roberts, Ian & Roussou, Anna. 2003. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Simpson, Andrew. 2004. The EPP, fossilized movement and reanalysis. In Diachronic Clues to Synchronic Grammar [Linguistik Aktuell/Linguistics Today 72], Eric Fuß & Carola Tripps (eds), 161–189. Amsterdam: John Benjamins. Thurneysen, Rudolf. 1946. A Grammar of Old Irish. Dublin: Dublin Insitute for Advanced Studies. Translated by D.A. Binchy and Osborn Bergin.
The great siSwati locative shift* Lutz Marten
School of Oriental and African Studies In siSwati the accumulation of a number of changes in the morphology and syntax of locative phrases has led to a more fundamental shift of restructuring of the underlying grammatical system – the great siSwati locative shift – so that locatives in siSwati are no longer, as in Proto-Bantu and most other present-day Bantu languages, part of the noun class system, but are prepositional. This shift explains aspects of changes in the siSwati locative system which are not otherwise independently motivated, including the degrammaticalization of a historic noun class marker into a preposition and distinct relative clause marking of locatives, and provides an example of a complex, systematic historical change of a subsystem of the grammar.
1. Introduction Bantu languages provide a rich empirical base for the study of language change. Against a background of fairly high typological similarity among the family overall, especially south-eastern Bantu languages exhibit a high degree of morphosyntactic microvaration (Marten, Kula & Thwala 2007), which provides the basis for detailed case studies of different aspects of language change. In this chapter I present one such case study, namely the development of the locative system of siSwati, a Bantu language of the Nguni group spoken in Swaziland and South Africa. I will show that the system is historically related to Proto-Bantu (PB) locative noun classes, but that in presentday siSwati the locative system is prepositional, and no longer part of the noun class
*I am grateful to Nomsile Gamedze, Thandekile Mndzebele, Mbongiseni Msibi, Lindi Nkambule,
to Margaret Kumbuka and to Bupe Kula, Fred Kula and Nancy Kula for discussion of relevant siSwati, Swahili and Bemba data, as well as to Leston Buell, Denis Creissels, Peter Nichols and to audiences in Cambridge, Tervuren and Windhoek, where parts of this material were presented, for helpful comments and suggestions. Parts of the research reported here were supported by AHRC grant B/RG/AN8675/APN16312 and by a British Academy UK-Africa network grant on Languages and Linguistic Studies of Southern African Languages, which are hereby gratefully acknowledged. All errors remain, of course, mine.
Lutz Marten
system. Most of the changes which characterize the development in siSwati are also found in other Bantu languages, and can often be individually motivated, for example as loss of morphological markers, instances of grammaticalization or language contact. However, it is only in siSwati and closely related languages that the accumulation of a number of individual changes in the morphosyntax of locative phrases has led to the more fundamental shift of restructuring of the underlying grammatical system – the great siSwati locative shift, during which locative morphology was lost and innovated, locative noun class markers were reanalysed as prepositions, and locative phrases, historically noun phrases, were reanalysed as prepositional phrases. This shift explains aspects of changes in the siSwati locative system which are not independently motivated, such as the unusual case of the degrammaticalization of a historic noun class marker into a preposition. The siSwati locative shift, as an example of a complex, systematic change of a sub-system of the grammar, is thus of interest from a comparative Bantu perspective as well as for the study of language change more generally.1 In the following sections I will provide a brief outline of the PB locative system and then show in detail how the siSwati system differs from it. The perspective adopted is historical–comparative, and I will provide a number of examples from other Bantu languages to show how the siSwati system relates to the wider Bantu situation. After describing the system in some detail, I will summarize the changes and show how they can best be explained by assuming a change in the underlying morphosyntax of siSwati locatives, which has not occurred in most other Bantu languages. Finally I will briefly address the question whether the locative shift can be explained as resulting from the accumulation of a number of independent changes, or whether in addition language-external factors, such as language contact, could have played a role as well.
2. The Bantu locative system Before discussing the siSwati locative system in more detail, I will provide in this section a brief outline of a ‘typical’ Bantu locative system, that is, aspects of the grammar of locatives which are highly frequent in comparative Bantu and/or have been reconstructed for Proto-Bantu. Locative marking in Bantu is typically part of the noun class system. Three different locative noun classes can be reconstructed confidently and are found very frequently in
. I take siSwati here as a representative of a wider group of southern Bantu languages. Changes similar, or in some cases identical, to the changes found in siSwati are also found in Zulu and other Nguni languages, and in other southern Bantu languages outside of Nguni, and possibly also in northwestern Bantu languages; cf. Gregoire (1975) for a general overview, and Creissels (2009) for Tswana. However, more empirical, especially syntactic data are needed to ascertain how wide-spread the locative shift is.
The great siSwati locative shift
present-day Bantu languages. These are conventionally labelled classes 16, 17, and 18. The relevant noun class prefixes and associated semantics are class 16 *pà-, referring to proximate or specific location, class 17 *kù-, referring to distal or non-specific location, class 18 *mù-, referring to interiority. Locative morphology is found both in the nominal and the verbal domain. All three locative classes can be used to mark nouns as locative, and there are three locative subject and object markers (e.g. Meinhof 1910, 1948; Meeussen 1967; Maho 1999). A typical system is found in Bemba, illustrated below:2 (1) Pà-ngándá pà-lì àbà-nà. 16-9.house sm16-be 2-children ‘There are children at home.’
(Bemba)
(2) Kú-ngándá kwà-lí-ìs-à áb-ènì. 17-9.house sm17-recpast-come-fv 2-guests ‘Visitors have come to the house.’ (3) Mù-ngándá mù-lé-ímb-á ábà-nà. 18-9.house sm18-progr-sing-fv 2-children ‘The children are singing in the house.’
The examples show that three locative noun class prefixes pa-, ku- and mu- are distinguished, and that the verb agrees with the relevant noun class of the locative subject. Similarly, class 16–18 locative agreement affixes typically exist to express other agreement relations usually expressed by noun class morphology, such as adjective, demonstrative or possessive agreement. (4) mù-ngándá mw-à-bá-kàfúndíshà ù-mú-súmà 18-9.house 18-poss-2-teacher aug-18-beautiful ‘in the beautiful house of the teachers’
Syntactically, locative nouns behave in most respects like other nouns, and can function as subjects and objects of verbs. For example, in (1), the locative noun pàngándá ‘at home’ functions as the subject of the copula -lì ‘be’. The locative nouns in (1) to (3) are grammatical subjects in so-called locative inversion constructions, in which a locative phrase is coded as grammatical subject and the logical or thematic subject is placed in immediately post-verbal position (see Bresnan & Kanerva 1989; Marten
. The following less widely used abbreviations are used in the glosses: 1, 2, 3 … = Noun class number 1, 2, 3 when directly followed by SG/PL = Person AUG = Augment or pre-prefix FV = Final vowel OM = Object marker SM = subject marker
Lutz Marten
2006). The availability of locative inversion constructions is related to the functionality of the locative system, a point I will return to in relation to siSwati below. In addition to classes 16–18, a fourth locative prefix is often assumed to have existed in PB, class 24 *ì-. Reflexes of this prefix are limited, and only a few Bantu languages show class 24 agreement patterns in addition to the nominal noun class marker. However, the siSwati locative marker e- might be related to this PB form (Gregoire 1975: 94, Maho 1999: 205–6). I will briefly return to this point below, but in the main will concentrate on the development of classes 16–18 for which better comparative material is available. Although typical Bantu locative systems as illustrated from Bemba above are part of the noun class system, there are nevertheless some differences between the locative classes 16, 17 and 18 and other, non-locative noun classes. First, while typical noun classes have basic members, that is words which are found only in that particular class or class-pairing, locative classes typically have no basic members at all, or only one member such as the word for place. Thus, for example, in Bemba, ùmùntù ‘person’, úmwàná ‘child’ and ùmwènì ‘visitor’, among many others, are class 1 nouns (with corresponding plurals in class 2). In contrast, the locative classes are only used with one nominal root: úkùntù ‘to a place’, ápàntù ‘at a place’ and úmùntù ‘in a place’, all based on the nominal root -ntù, ‘entity’ which is also found in ùmùntù ‘person’ with the human class 1 prefix. The vast majority of locative nouns are derived from nouns from other classes, so that locative noun class prefixes are prefixed not to nominal stems as normal class prefixes are, but to forms which already contain a prefix. Thus, ímì-tí ‘trees’ consists of the class 4 prefix ímì- and the nominal stem -tí ‘tree’, and can be used as a locative noun as for example in kú-mì-tí ‘on the tree’ or mú-mì-tí ‘in the tree’. In contrast to the basic form ímìtí, the locative forms contain two class prefixes, the locative class prefix and the original class 4 prefix. In this respect the locative classes can be seen as part of the derivational use of the noun class system, where nouns are assigned to secondary classes, often with a clear semantic effect (cf. Schadeberg 2003). Bantu locative marking is thus part of the Bantu noun class system, and locative nouns are essentially like other nouns in terms of morphology, agreement and syntactic behaviour. However, the locative noun classes are peripheral members of the noun class system as they have no or a very restricted set of basic members, and are mainly used derivationally, a fact which is reflected in the common presence of two noun class prefixes in locative nouns. In many present-day Bantu languages, the locative system has changed in different ways. For example, Swahili has lost the original nominal locative morphology appearing on nouns, and locative nouns are marked with a newly developed locative suffix. However, the three-way distinction between the three locative noun classes in the agreement system has been retained, so that, for example, three different locative
The great siSwati locative shift
subject markers are still part of the grammar. In contrast, Lozi has maintained the three locative noun class prefixes for nouns, but has lost two of the three agreement prefixes, so that class 16, 17 and 18 nouns in Lozi all have a class 17 subject agreement marker (cf. Marten et al. 2007). However, in most Bantu languages the locative system is still part of the noun class system, even if some morphological markers have been lost. In contrast, I will show below that the locative system of siSwati – and probably also that of related southern Bantu languages – has been restructured to such an extent that locatives are no longer coded as nouns, and thus not as part of the noun class system, but as prepositional phrases. The results of the restructuring process can be seen in both the morphology and the syntax of locatives in siSwati, as I will show in more detail below. What is of particular interest in this restructuring process is that many aspects of the new system are also found in other Bantu languages, and that it is rather the totality of the changes which has led to the underlying more fundamental change of the complete system.
3. The restructuring of the siSwati locative system Locative marking in siSwati shows clear traces of the PB locative system. In both nominal and verbal morphology, the class 17 locative marker ku- is still found, even though, as I will show below, with innovated morphology, syntax and semantics. Traces of class 16 and 18 are also found, although only in lexicalized forms. Furthermore, the locative prefix e- is probably a reflex of PB class 24. However, the innovations outnumber the retentions. For example, as in Swahili, a new suffix is used for locative marking. Furthermore, syntactically, a number of typical Bantu locative characteristics are not found in siSwati. In fact, the number of innovations amounts to a systematic change whereby locative marking has ceased to be part of the noun class system, and is now better analysed as a prepositional system. The specific aspects of the locative system relevant to the discussion include both morphological and syntactic facts. In the following sections I will discuss in more detail the development of new nominal locative morphology, the loss of the locative subject markers pa- and mu- and the reanalysis of the locative subject marker ku- as expletive subject marker, the reanalysis of historical locative noun class prefixes as locative prepositions, the loss of locative subject and object marking, and finally the loss of relative agreement except in locative relatives.
3.1 Development of new locative morphology Although traces of the three typical members of PB locative class morphology are found in siSwati, locatives are formed productively in siSwati only with class 17 ku-, or by
Lutz Marten
using different morphemes, namely the prefix e- and the suffix -ini, sometimes in combination with each other (examples from Taljaard, Khumalo & Bosch 1991: 43–46).3
(5) bafana ‘boys’ > ku-bafana ‘to/at the boys’ (6) sitolo ‘shop’ > e-sitolo ‘at the shop’ (7) indlu ‘house’ > e-ndl-ini ‘at the house’
The use of e- … ini as in (7) is the most productive locative strategy, while e- is used with place names and localities, as well as with some lexical exceptions (6), and ku- is most commonly found with nouns of class 1 and 2, which include animate nouns, personifications and some loan-words (5), as well as with demonstratives, further discussed below (cf. e.g. Ziervogel & Mabuza 1976: 34–37, Taljaard et al. 1991: 43–46). The locative suffix -ini has been argued to result from a grammaticalization process with the lexical form *-ini ‘liver’ as starting point, involving semantic, metaphorical extension to interiority (comparable to metaphorical uses of English heart) and then further extension to denote other locative relations (Samsom & Schadeberg 1994). According to Samsom & Schadeberg (1994), the grammaticalization process originated in East African Bantu languages, and the formative then spread to other languages in the east and south of the Bantu area, replacing to a greater or lesser extent the historically older locative morphology involving all three locative noun classes (see Gregoire 1975: 185–204; Güldemann 1999 for further discussion of -ini). Further evidence for the previous existence of class 16 and 18 in siSwati comes from fact that class 16 and class 18 locative noun class morphology is found in lexicalized expressions such as adverbials and demonstratives (examples from Ziervogel & Mabuza 1976: 27–28): (8) phandle (9) phansi (10) lapha (11) mshiya loyi (12) mshiya lowa
‘outside’ ‘underneath’ ‘here’ ‘this side’ ‘that side’
The forms in (8) to (10) involve the class 16 formative pha-, while (11) and (12) appear to include grammaticalized forms of the class 18 prefix m(u)-. The existence of these forms shows that PB locative class morphology was used in an older stage of the language, but that productive locative marking became associated with only class 17 or the affixes e- and -ini. This development in nominal morphology is mirrored in verbal agreement morphology. As in the nominal domain, class 16 and class 18 agreement morphology has
. SiSwati data are given without tone marking, following usage in most sources, and due to the absence of a satisfactory analysis of siSwati tone.
The great siSwati locative shift
been lost, and there is no agreement morphology associated with siSwati e-. While class 17 ku- is still used as a subject marker, its function and syntax have changed, and it is now used as a marker for expletive, impersonal or default subjects: (13) Ku-ya-bandza. sm17-pres-be.cold ‘It is cold.’ (14) Ku-mnandzi. sm17-be.nice ‘It is nice.’ (15) Ku-ne-kudla e-dladla-ini. sm17-poss.cop-food loc-kitchen-loc ‘There is food in the kitchen.’ (16) Ku-ya-dl-iwa. sm17-pres-eat-pass ‘There is being eaten.’ (idiomatic, ‘There is a feast.’) (17) Kw-ent-e njani? Ku-buta unina. sm17-happen-perf what? sm17-ask mother.his ‘What happened?, asked his mother.’ (18) Ku-ya-bonakala kutsi u-ya-gula. sm17-pres-evident that sm3sg-pres-be.ill ‘It is evident that he is ill.’ (19) Tim-buti, li-hhashi ne-ma-tfole ku-ya-baleka. 10-goat 5-horse conj-6-calf sm17-pres-run ‘The goats, the horse and the calves run away.’
Examples (13) and (14) show expletive ku- used as a subject marker with the weather verb -bandza ‘be cold’ and the stative verb -mnadzi ‘be nice’. In (15) and (16) ku- serves as subject marker in existential constructions: in (15) with the possessive copula -ne, and in (16) with an intransitive passive. In (17) ku- is found both as an expletive subject marker with -enta, ‘happen’ and as subject marker in a quotative construction with inverted subject, while in (18), ku- functions as the expletive subject marker of a propositional attitude verb. Finally, in (19) ku- is used as the subject marker of a conjoined subject with conjuncts of different noun classes. In these examples ku- is used in contexts where the grammatical subject does not correspond to the logical or thematic subject, where a thematic subject is not encoded at all, or where agreement between the noun class of the grammatical subject and the subject marker is not possible. The function of ku- in these examples is as a default, indefinite or expletive subject marker, showing that ku- has lost its locative semantics and can be used to encode more abstract semantic functions. For simplicity, I will refer to all these functions as ‘expletive’ in the following discussion.
Lutz Marten
Locative marked nouns can be used with the expletive subject maker, but show no class agreement (20): (20) E-ndl-ini ku-ya-shisa. loc-house-loc sm17-pres-be.hot ‘In the house it is hot.’ (21) Phandle ku-ya-bandza. 16.outside sm17-pres-be.cold ‘Outside it is cold.’
Based on similar evidence, Buell (2007) proposes for Zulu that when locative NPs precede the expletive subject markers, the locative NP is in topic position, rather than in subject position, a point to which I return in Section 3.3 below. The changes in class morphology in siSwati are also found in a number of other Bantu languages. As mentioned above, both Swahili and Lozi have lost part of the three-way distinction of the historical locative morphology. Furthermore, the new locative suffix -ini is also found in Swahili and a number of other southern and eastern Bantu languages. The use of the class 17 locative marker ku- as an expletive marker is similarly not restricted to siSwati, and is found, for example, in Swahili and Bemba: (22) Ku-na m-vua. sm17-cop 9-rain ‘It is raining.’
(Swahili)
(23) Ku-na nini? sm17-cop what ‘What is happening?’ (24) Nàà-kù-tálálá. pres-sm17-be.cold ‘It is cold.’
(Bemba)
Neither the change in morphological marking nor the change in function of ku- is thus only found in siSwati, and it is not particularly exceptional from a cross-Bantu perspective. However, the siSwati situation is unusual in that expletive use seems to be the only, or at least predominant function associated with the class 17 subject marker. If the development of expletive function from previous locative semantics is seen as a grammaticalization path (cf. Heine & Kuteva 2002: 203), then in siSwati this development appears to have proceeded further than in most other Bantu languages. A similar situation can also be found in the internal structure of locative phrases discussed below.
3.2 R eanalysis of historical locative noun class prefixes as locative prepositions A second type of evidence relevant for the understanding of the siSwati locative system comes from modified locative phrases. Both in terms of agreement within complex
The great siSwati locative shift
locative phrases, and in terms of morpheme order, modified locatives show that locative phrases in siSwati are not DPs, but behave like prepositional phrases. Agreement of dependent nominals with locative phrases provides useful evidence about the internal structure of the locative phrase. As was mentioned above, locative morphology is typically added onto original noun class morphology, and so the question arises of whether dependent nominals such as adjectives or possessives can show agreement with the original noun class, the locative noun class, or both. In siSwati, only the first possibility is realized: dependent nominals agree with the original class, not with the locative: (25) ba-fana ba-mi 2-boys 2-my ‘my boys’ (26) ku-ba-fana ba-mi loc-2-boys 2-my ‘at my boys’
The agreement facts show that, irrespective of locative marking, the possessive pronoun shows class agreement with the class 2 noun bafana ‘boys’. In other words, the locative morphology in siSwati cannot function as the head of the locative phrase for agreement purposes. In terms of internal structure of the locative phrase, the agreement facts thus indicate that locatives in siSwati are not locative noun phrases, but rather locative prepositional phrases. Under the assumption that dependent nominals can agree with nominal heads within DPs, but not with prepositional heads, which take DP complements, this would explain the absence of agreement with the locative. Further support for this analysis comes from locatives with pre-nominal demonstratives. Demonstratives can be placed before the noun in siSwati, and both full forms (27) and short forms (28) are possible. Full demonstratives are phonologically independent forms, while the short forms are phonological clitics, which can function like a definite article rather than a proximate demonstrative. However, both forms can also follow the noun they modify. Given the relative positional freedom of demonstratives, it is not expected that they could intervene between two prefixes, or a prefix and a stem. However, this is what occurs in locative phrases. When nouns modified by a demonstrative are combined with the locative marker ku-, the locative marker is attached before the demonstrative (29, 30) (cf. Ziervogel & Mabuza 1976: 199–200): (27) leyo n-dvodza / n-dvodza leyo dem9 9-man 9-man dem9 ‘this man’ (28) le-n-dvodza / n-dvodza-le dem-9-man 9-man-dem ‘this man/the man’
Lutz Marten
(29) ku-leyo n-dvodza / ku-n-dvodza leyo 17-dem9 9-man 17-9-man dem9 ‘to this man’ (30) ku-le-n-dvodza / ku-n-dvodza-le 17-dem-9-man 17-9-man-dem ‘to this man’
The locative marker ku- behaves in this respect like other preposition-like elements in siSwati, for example nga- ‘with’ which also precedes pre-nominal demonstratives (cf. Thwala 2006: 213–14): (31) nga-le-moto with-dem-9.car ‘with this car’ (32) ku-le-si-kolwa 17-dem-7-school ‘at the school’
These examples show that the locative marker ku- does not behave morphologically like a noun class marker, but behaves like a preposition taking a complex nominal complement. In descriptive grammars of siSwati (e.g. Ziervogel & Mabuza 1976; Taljaard et al. 1991) both ku- and nga- are analysed as prefixes. However, unlike noun class prefixes, they precede a range of hosts, and crucially precede pre-verbal demonstrative clitics, and might thus better analysed as (phonological) clitics.4 The relevant difference is that ku- patterns with nga- in preceding le-, unlike other noun class prefixes, which follow le- as the example of the class 7 noun class prefix si- in (32) shows. Evidence from the interaction of locative marking with modified nouns, both with respect to agreement and with respect to morpheme order, thus shows that the locative marker ku- behaves like a preposition rather than as a noun class marker, and that a locative phrase such as (33) can be analysed as a PP as shown in (34):5 (33) ku-le-si-kolwa loc-dem-7-school ‘at the school’
. Further prosodic evidence may shed more light on the phonological status of morphemes like nga- and ku-. For further discussion of potential mismatches between the phonological and morphosyntactic status of Bantu noun-class prefixes, see e.g. Bresnan & Mchombo (1995); Carstens (1991), (1997); Myers (1987). . The internal structure of the siSwati NP is likely to be more complex than shown here; the point of (34) is to show the function of ku- as prepositional head.
The great siSwati locative shift
PP
(34) P
DP
ku Dem le
NP NCl
Nstem
si
kolwa
Cl 7
school
From a comparative perspective, the situation in siSwati is unusual. To my knowledge no other (at least non-southern) Bantu language allows a demonstrative or other modifier to intervene between the locative marker and the noun stem. While it is possible that further empirical study will show other cases like this, it still appears to be very rare. From the perspective of the analysis proposed here, it is this fact which provides the strongest morphological indication that a systematic reanalysis of the locative system has taken place, and that locative markers have been reanalysed as prepositions. Cross-linguistically, this reanalysis is remarkable as it constitutes an instance of degrammaticalization, that is, an exception to the unidirectionality of grammaticalization paths, often claimed to be universal, from less morphologically bound forms – like prepositions – to more morphologically bound forms, like noun class prefixes (e.g. Heine, Claudi & Hünnemeyer 1991: 212–15; Hopper & Traugott 1993: 94–129). The context in which this process takes place in siSwati may explain it. The accumulation of the individually innocuous changes discussed so far has led in siSwati to a more far-reaching change of the underlying system of locative marking – from being part of the noun class system to becoming a prepositional system. Rather than occurring in isolation, the locative marker ku- has lost the morphosyntax associated with noun class markers because of its part in this overall reanalysis of the locative system. While a shift from noun class marker to preposition is in itself unusual, it is clearly motivated in the context in which the relevant (locative) system is no longer seen as part of the noun class system. I will return below to the question how this larger, systematic shift can be explained. With respect to agreement with dependent nominals, the situation in siSwati is comparable to the situation of the interpretation of the expletive marker ku- discussed above. That is, from a comparative perspective, the possibility of agreeing with the original noun, rather than with the locative is, although probably not typical, not unusual either. For example, while in Swahili dependent nominals can only agree with the locative, in Luganda agreement with either the locative or the original noun is possible (Ashton et al. 1954; Gregoire 1975: 82).
Lutz Marten
(35) m-oyo w-angu 3-heart 3-my ‘my heart’
(Swahili)
(36) m-oyo-ni mw-angu/*w-angu 3-heart-loc 18-my/3-my ‘in my heart’ (37) ku-bbalaza kw-ange 17-9.terrace 17-my ‘on my terrace’
(Luganda)
(38) ku-ky-alo ky-ange 17-7-village 7-my ‘in my village’
In Swahili, even though locative nouns are not marked with a noun class prefix, but with the locative suffix -ni, dependent nominals still show class 18 locative agreement. In contrast, Luganda allows both agreement with the locative class and agreement with the original class, where the two options are exploited for expressing pragmatic information by placing different emphasis on the locative relation (37) and the possessive relation (38) (see Ashton et al. 1954: 253). However, as shown above, in siSwati only agreement with the original class is possible. This is expected under an analysis in which ku- in siSwati functions as a preposition, taking a DP complement, as the agreement relation obtains between different constituents of the DP. The following section presents further evidence for this view from the syntax of locative phrases and verbal agreement in siSwati.
3.3 Loss of locative verbal agreement marking In addition to morphological criteria, the syntactic behaviour of locatives in siSwati shows that locatives are prepositional adjuncts rather than nominal complements, as has been argued by Thwala (2006), and as often seems to be assumed in descriptive grammars which class locatives as ‘adverbials’ (e.g. Ziervogel & Mabuza 1976: 129). Relevant evidence comes from the absence of locative object markers as well as from the restriction of the historically locative subject marker ku- to expletive contexts, already noted above. Subject and object agreement are often taken as tests for the argument status of DPs in Bantu (though see e.g. Schadeberg 1995, Marten & Kula 2008 for some problems with this). In siSwati, verbs show agreement with their subjects in most syntactic contexts, while objects are marked typically when they are dislocated. In addition, both subject and object DPs can remain unexpressed, and only be indexed by subject and object markers. However, agreement differs between locative and other phrases, and both object marking and subject marking in siSwati show that locative constituents are not arguments like DPs of other noun classes.
The great siSwati locative shift
With respect to object marking, (39) and (40) illustrate the difference between a DP object such as inja ‘dog’, which can be expressed by an object marker, and a locative constituent for which no object marker exists, and which therefore cannot be expressed by an object marker irrespective of whether the lexical locative is present or omitted.6 (39) Ng-a-yi-bon-a kahle (in-ja). sm1sg-past-om10-see-fv well 10-dog ‘I saw it well, the dog.’ (40) *Ngi-ku-bon-e (e-Thekw-ini). sm1sg-om17-see-past loc-Durban-loc Intended: ‘I saw it/there (Durban).’ (41) Ku-Lilongwe n-a-ku-ziw-a. 17-Lilongwe sm1sg-pres-om17-know-fv ‘Lilongwe, I know it/there.’
(Nsenga)
As (40) shows, locative object marking is not possible in siSwati. This contrasts with typical Bantu systems of object agreement, in which locative object markers are included, as illustrated by the Nsenga example in (41). However, there are a number of Bantu languages which like siSwati do not have locative object markers, including Lozi and the Tanzanian Bantu languages Ciruri and Chasu (cf. Marten et al. 2007). The role of locative subject agreement markers has already been noted above, when discussing the development of the historically locative subject marker ku- as an expletive subject marker. Further relevant evidence comes from locative inversion constructions. In Bantu locative inversion constructions a locative phrase is coded as grammatical subject and the logical subject follows the verb, as seen in the Bemba examples (1–3) above, and illustrated in (42) from Herero (Marten 2006). (42) M-òn-djúwó mw-á hìtí é-rùngà. 18-9-house sm18-past enter 5-thief ‘The thief entered the house.’ (‘Into the house entered a/the thief.’)
(Herero)
In contrast, in siSwati locatives do not appear to function as subjects. A locative phrase preceding the verb behaves like an adjunct, since there is a strong tendency to use a resumptive locative demonstrative like khona in (43) with pre-verbal locatives, unlike with true subjects in siSwati, or locative subjects in locative inversion constructions, where resumptive pronouns are marginal (cf. Buell 2007). The subject marker in (43) thus appears to be the invariable expletive subject marker, discussed above, rather than an agreeing subject marker as it is in Herero.
. Example (40) is perfectly acceptable if the object marker is interpreted as the homophonous 2nd person singular object marker ku-, resulting in the reading ‘I saw you in Durban’. However, the point at hand is that it cannot be interpreted as locative.
Lutz Marten
(43) Ku-le-ti-ndlu ku-hlala (khona) ba-ntfu la-ba-dzala. loc-dem-10-houses 17-stay there 2-people rel-2-old ‘In these houses (there) live elderly people.’
Alternatively, locative inversion is expressed with a non-locative-marked subject functioning as subject, where the subject marker agrees with the (non-locative) class of the subject (Thwala 2006: 213), a difference termed formal as opposed to semantic locative inversion by Buell (2007): (44) Le-si-kolwe si-to-fundza ba-ntfa-ba-in-khosi. dem-7-school sm7-fut-read 2-chidren-poss-9-king ‘The king’s children will study at this school.’ (45) Le-ti tin-dlu ti-hlala ba-ntfu la-ba-dzala. dem-10 10-houses 10-stay 2-people rel-2-old ‘Elderly people live in these houses.’
Thus, while siSwati has constructions similar to canonical locative inversion constructions, syntactically these either involve a locative adjunct or a non-locative subject. The same facts are true for Zulu (Buell 2007), but it is not clear how wide-spread they are beyond southern Bantu. In so far as data are available, locative inversion is common in Bantu and has been reconstructed for PB (Meeussen 1967: 120), while inversion constructions such as in (44) and (45) have been documented less frequently. It is interesting to note in passing that the siSwati facts show that the availability of (locative) inversion constructions appears to be independent of the existence of locative noun classes. While in siSwati the locative system has been restructured and is no longer part of the noun class system, inversion constructions are still found, even though without explicit locative marking. In summary, then, syntactic evidence thus shows that locative phrases in siSwati do not function as either subject or object, which further supports an analysis of locative phrases in siSwati as prepositional phrases rather than as noun (or determiner) phrases.
3.4 Loss of relative agreement except in locative relatives A final set of relevant data comes from relative clause marking. Relativized arguments in siSwati are marked by a relative marker la- prefixed to the verb and a co-referential subject or object maker (46, 47). However, this strategy is not available for relativized locative phrases (48). These require the use of a special locative relative agreement marker -pho, historically related to the noun class 16 marker pa-, to which the relative marker la- is prefixed. Furthermore, instead of a co-referential agreement marker in the verb form, locative relatives show a resumptive demonstrative in the relative clause, such as khona in (48).7
. The use of khona in (48) further supports the analysis of the locative in (43) as an adjunct rather than as a subject: in both cases, the locative is outside of the core clause.
The great siSwati locative shift
(46) Nge-ba-fanai la-bai-to-natsa tjwala. foc-2-boys rel-sm2-fut-drink 14.alcohol ‘It is boys who will drink alcohol.’ (47) Tjwalai, ba-fana la-ba-to-bui-natsa. foc.14.acohol 2-boys rel-sm2-fut-om14-drink ‘It is alcohol which the boys will drink.’ (48) Ku-se-khayai la-phoi ba-fana ba-to-natsa tjwala khonai. foc.17-7-home rel-loc 2-boys sm2-fut-drink 14.alcohol there ‘It is at home where the boys will drink alcohol.’
In contrast to siSwati, Bantu languages with more typical locative systems allow the same relative strategy to be used for both non-locative and locative phrases, as illustrated below from Swahili. (49) Ni wa-vulanai amba-oi wai-ta-nunua pombe hapa. cop 2-boys rel-2 sm2-fut-buy 9.beer loc16 ‘It is boys who will buy beer here.’
(Swahili)
(50) Ni pombei amba-yoi wa-vulana wa-ta-(ii)-nunua hapa. cop 9.beer rel-9 2-boys sm2-fut-(om9)-buy loc16 ‘It is beer which the boys will buy here.’ (51) Ni ha-pai amba-poi wa-vulana wa-ta-(pai)-nunua pombe. cop dem-16 rel-16 2-boys sm2-fut-(om16)-buy 9.beer ‘It is here that the boys will buy beer.’
As the examples in (49) to (51) show, in Swahili relative clauses can be formed with a relative marker amba-, to which an anaphoric element agreeing in class with the head noun is suffixed. This strategy is used for both locative and non-locative relatives, and in both a co-referential agreement marker is found in the verb form, obligatorily for subjects and optionally for non-locative objects and locatives. Historically, siSwati relative marking is likely to be an innovation, through generalization of the relative marker la-, based on a demonstrative form, to be used in all relative contexts, without specific agreement morphology. In contrast, relativized locatives retain a more conservative pattern in which the relative marker agrees with the head noun. The change of relative clause marking is thus likely to have occurred after the restructuring of the locative system, as at the time of the change in relative marking locatives had already been reanalysed as prepositional phrases.
4. Accumulation of changes and the restructuring of the locative system The preceding sections have provided a detailed description of locative phrases in siSwati. Taken together, the particular morphosyntactic structures of the siSwati locative system show many differences to the original historical starting point of the
Lutz Marten
PB locative system. The changes which distinguish the modern system are summarized below: –– –– –– –– –– –– –– ––
loss of class 16 and 18 in both nominal and verbal morphology use of new nominal locative morphology development of locative subject marker ku- as expletive subject marker demonstratives intervene between the locative marker ku- and the nominal stem in modified locative phrases, agreement is only possible with the original noun no locative object marker no (formal) locative inversion: locatives are not used as subject special relative marking strategy for locatives
The majority of these changes can be found in other Bantu languages, although this does not seem to be true for the possibility of a demonstrative intervening between the locative marker ku- and the nominal stem, or possibly either the absence of formal locative inversion. However, it appears that no other Bantu language exhibits all these changes. Furthermore, as briefly discussed above, the effect of all the changes taken together amounts to a systematic change in the underlying system of locative morphosyntax in siSwati. The locative system started out as part of the noun class system, so that locative phrases were syntactically noun phrases, while the siSwati locative system is independent of the noun class system, and locative phrases are syntactically prepositional phrases. The majority of the changes in this shift from nominal to prepositional morphosyntax are widely attested and well motivated in themselves, and may have resulted from both external and internal factors, for example resulting from a combination of loss of morphological distinctions, semantic changes of individual locative markers, and the increased use of new locative markers in addition to contact-related factors, such as the adoption of the new locative suffix -ini from Eastern Bantu. The exception to this is the degrammaticalization of ku- from a noun class marker to a preposition. This indicates an analysis in which a number of independent changes led to a restructuring of the system, and where subsequently in a second step, as a consequence of this restructuring, the remaining noun class marker ku- was reanalysed as a preposition – an unmotivated reanalysis in itself, but highly motivated in the context of the wider structural changes. Since the noun class system was no longer associated with locative morphosyntax, a locative noun class marker would be inconsistent within the system overall, so that the reanalysis as preposition resolved this inconsistency.8
. Three partly alternative analyses might be worth mentioning at this stage. Gregoire (1975) proposes that Southern Bantu languages, including siSwati, have lost reflexes of the class 17 noun class prefix ku- almost entirely, except in lexicalized forms, and that forms like the ones
The great siSwati locative shift
In this more wide-ranging restructuring, language contact may well have played a role as well. More comparative and historical work is needed to investigate this question fully, however evidence for influence from Khoisan languages is found on the phonological level by click consonants found in Nguni languages, as well as in the lexicon and in aspects of morphosyntax. Although most South African Khoisan languages, that is, Khoisan languages with which early speakers of Bantu languages in the area are likely to have had extended contact, are extinct, it appears that several relevant languages had prepositions, even though in most other respects structures are head-final (Güldemann 1999: 67; Güldemann & Vossen 2000). The prepositional marking may have served as syntactic blueprint for the development of the prepositional system of siSwati, although more detailed study is needed develop this idea further.
5. Conclusions This chapter has presented an analysis of a locative morphosyntax in siSwati. Historically, the siSwati locative system has developed from nominal morphosyntax, where the locative system was part of the Bantu noun class system, to prepositional morphosyntax, where the locative system is no longer part of the noun class system, but constitutes a separate subsystem of the grammar. The siSwati locative shift presents relevant case studies for further understanding of grammaticalization – and degrammaticalization – processes, and constitutes
discussed in this chapter result historically from a construction involving *kúdí ‘où est’ (‘where is’), found in several other Bantu languages, and which developed into *ku- through use with an empty copula, with the meaning ‘chez’ (‘at’), in the sense of ‘at someone’s place’. This explains well the use of ku- with animate nouns, but not so well the use with non-animate class 1/2 nouns, or with demonstratives. Creissels (2009), on the other hand, develops an analysis of locative marking in Tswana, where he points out that what appears to be a locative preposition is in fact historically derived from a locative agreement prefix, not from a noun class prefix. If translated into siSwati, this would explain the use ku- with demonstratives, but less successfully the use with class 1/2 nouns. In fact, neither analysis is entirely incompatible with the analysis developed here, although either would imply a more complex historical derivation of present day siSwati ku-. However, both analyses are compatible with the claim that the locative system of siSwati (and related southern Bantu languages) has been restructured. On the other hand, Bresnan & Mchombo (1995) and Carstens (1997) analyse Chichewa locative phrases syntactically as nominal – although differing in formal detail from each other – rather than prepositional, and it is interesting to note that this is partly motivated by properties of Chichewa locatives which differ from siSwati (e.g. presence of subject and object marking, no intervening demonstratives between locative markers and a following DP, locative agreement of dependent nominals, formal locative inversion), thus indirectly supporting the present analysis.
Lutz Marten
an example of the interaction between isolated, individually motivated changes and systematic changes of underlying complex sub-systems of the grammar. While most of the changes found in siSwati locative morphosyntax are also found in other Bantu languages, the specific combination of all of them is not. It is thus the accumulation of individual changes which has led to the restructuring of the system, with further attendant changes following this restructuring. The question of whether in addition language contact has played a role in the restructuring process has been raised, but a detailed discussion remains for future research. Similarly, more comparative work is needed to show to what extent the analysis presented here can be extended to other Bantu languages, which share the changes outlined here to a greater or lesser extent, and which may be analysed as presenting intermediate stages of development from the PB locative system to a reanalysed system such as found in siSwati. The restructuring of the siSwati locative system thus also provides relevant evidence for the study of noun class systems in general, and in particular for the question of how noun classes change, and can be lost. Given the central role of the noun class system in Bantu grammar, the study of aspects of the system, and their diachronic developments, is important for further understanding of Bantu grammar and historical-comparative Bantu studies more widely.
References Ashton, Ethel O., Mulira, Enoch M.K, Ndawula, E.G.M. & Tucker, Archibald Norman. 1954. A Luganda Grammar. London: Longmans, Green and Co. Bresnan, Joan & Kanerva, Jonni M. 1989. Locative inversion in Chichewa. Linguistic Inquiry 20: 1–50. Bresnan, Joan & Mchombo, Sam A. 1995. The Lexical Integrity Principle: Evidence from Bantu. Natural Language and Linguistic Theory 13: 181–254. Buell, Leston. 2007. Semantic and formal locatives: Implications for the Bantu locative inversion typology. SOAS Working Papers in Linguistics 15: 105–120. Carstens, Vicki. 1991. The Morphology and Syntax of Determiner Phrases in Kiswahili. Ph.D. dissertation, University of California, Los Angeles. Carstens, Vicki. 1997. Empty nouns in Bantu locatives. The Linguistic Review 14: 361–410. Creissels, Denis. 2009. The presentational focus construction in Tswana: A case against ‘locative subjects’. Paper presented at the Bantu Syntax Workshop in Inversion Constructions in Bantu, Tervuren, 28 March 2009. Gregoire, Claire. 1975. Les locatifs en bantou. Ph.D. dissertation, University of Leiden. Published in Annales Sciences Humaines, Vol 83. Tervuren: Musée royal de l’Afrique Centrale. Güldemann, Tom. 1999. Head-initial meets head-final: Nominal suffixes in Eastern and Southern Bantu from a historical perspective. Studies in African Linguistics 28: 49–91. Güldemann, Tom & Vossen, Rainer. 2000. Khoisan. In African Languages: An Introduction, Bernd Heine & Derek Nurse (eds), 99–122. Cambridge: CUP. Heine, Bernd & Kuteva, Tania. 2002. World Lexicon of Grammaticalization. Cambridge: CUP.
The great siSwati locative shift
Heine, Bernd, Claudi, Ulrike & Hünnemeyer, Friederike. 1991. Grammaticalization: A Conceptual Framework. Chicago IL: The University of Chicago Press. Hopper, Paul J. & Traugott, Elizabeth Closs. 1993. Grammaticalization. Cambridge: CUP. Maho, Jouni. 1999. A Comparative Study of Bantu Noun Classes. Göteborg: Acta Universitatis Gothoburgensis. Marten, Lutz. 2006. Locative inversion in Herero: More on morphosyntactic variation in Bantu. Papers in Bantu Grammar and Description, ZAS Papers in Linguistics 43: 97–122. Marten, Lutz & Kula, Nancy C. 2008. Object marking and morphosyntactic variation in Bantu. Ms, SOAS & University of Essex. Marten, Lutz, Kula, Nancy C. & Thwala, Nhlanhla. 2007. Parameters of morphosyntactic variation in Bantu. Transactions of the Philological Society 105: 253–338. Meeussen, Achille E. 1967, Bantu grammatical reconstructions. Africana Linguistica 3: 79–121. Meinhof, Carl. 1910. Grundriss einer Lautlehre der Bantusprachen, 2nd edn. Berlin: Reimer. Meinhof, Carl. 1948. Grundzüge einer vergleichenden Grammatik der Bantusprachen, 2nd edn. Berlin: Reimer. Myers, Scott P. 1987. Tone and the Structure of Words in Shona. Ph.D. dissertation, University of Massachusetts, Amherst. Samsom, Ridder & Schadeberg, Thilo C. 1994. Kiinimacho cha mahali: Kiambishi tamati cha mahali -ni. Afrikanistische Arbeitspapiere 37: 127–138. Schadeberg, Thilo C. 1995. Object diagnostics in Bantu. In Issues in African Languages and Linguistics: Essays in Honour of Kay Williamson, E. Nolue Emenanjo & Ozo-Mekuri Ndimele (eds), 173–180. Aba: National Institute for Nigerian Languages. Schadeberg, Thilo C. 2003. Derivation. In The Bantu Languages, Derek Nurse & Gerard Philippson (eds), 71–89. London: Routledge. Taljaard, Petrus C., Khumalo, James S.N. & Bosch, Sonja E. 1991. Handbook of Siswati. Pretoria: J.L. van Schaik. Thwala, Nhlanhla. 2006. Parameters of variation and complement licensing in Bantu. Papers in Bantu Grammar and Description, ZAS Papers in Linguistics 43: 209–232. Ziervogel, Dirk & Mabuza, Enos John. 1976. A Grammar of the Swati Language. Pretoria: J.L. van Schaik.
The impact of failed changes Gertjan Postma
Meertens Institute, Academy of Sciences* We study linguistic changes that failed, and their relation to successful changes. The inclusion of failed changes into Kroch’s logistic model of linguistic change is possible and, in fact, necessary. The logistic functions are solutions of a differential equation that also describes how failed changes are temporaldynamically related to successful changes: the failed change is the first derivative of the successful change. We interpret this relation in Weerman’s Peripheral Rules Model: failed changes are L2 innovations supported by peripheral rules rather than by core grammar, which are subsequently accommodated and modified by L1 adopters and turned into a successful change. Two case studies support the claims: the rise of the reflexive in Dutch and the rise of do-support in English.
1. Introduction Language changes. This fact of everyday life is a problem for generative linguistic theories, which, in the structuralist tradition, take language to be a system, composed of elements whose only status in reality lies in their relations with other elements. As such, change takes place as a replacement of one system by another, and is caused by factors outside the system itself. It can neither be stated nor explained within the linguistic system, that is within I-language. To the extent that the linguistic system is stable, all changes and all small deviations from the optimal realization return to the stable state, and any change would necessarily fail. Consequently, failures are to be expected, successes are the exception. Probably as a consequence of this reasoning, virtually no attention has been paid to the dynamics of failed changes and all attention has been focused on the dynamics of successful changes. As structuralist linguistics does not provide a theory or model for a successful change, change is considered to be the replacement of one system by the other, be this a jump in the individual or group of *I am grateful to Charlotte Galves, Tony Kroch, Marc van Oostendorp, Ans van Kemenade and the audiences of the Workshop on Language Change, Amsterdam, 1 Nov 2007, the Theelezing, Meertens Institute Amsterdam, 20 Dec 2007, the Workshop on Continuity and Change, Cambridge, 18–20 March 2008, and the Diachronic Generative Syntax Workshop, Campinas, 22–24 July 2009 for discussions and comments.
Gertjan Postma
individuals as an adult innovation (L2 model of linguistic change), or a jump between individuals as incomplete or imperfect intergenerational transmission of language (L1 model). Therefore, theories that describe the dynamics of the change necessarily involve the realization and behaviour of the E-language. In this study we explore the properties of failed changes in linguistics and propose a model for them. It is our aim to study diffusion that goes hand in hand with a change of the system, and in order not to complicate things beyond reason, we study the simplest case of diffusion in a community as a whole, ignoring spatial or sociological effects.1 We do so using a case study from Middle Dutch syntax and compare it with the scarce data on failures in the literature. It turns out that failed changes are related to successful changes in a systematic way. This relation can be expressed algebraically. Failed changes are the first derivative of the successful change. We interpret failed changes as what we propose to call ‘off-grammatical innovations’, that is, ‘tricks’, peripheral rules or stylistic rules (cf. Weerman 1993; Hinterhölzl 2004) made by adults or L2 learners.2 The successful change is a response to the off-grammatical input of their predecessors and establishes a newly accommodated I-language. First we review the existing models of successful change as proposed in the linguistic literature, especially Kroch’s (1989) logistic implementation and the Constant Rate Hypothesis (CRH) in Section 2. Then we show the shortcomings of this model, especially its non-application to failed changes. A study of the phenomena that have led Kroch to his logistic model leads us to the proposal that failed and successful changes are related, contradicting the CRH. Related changes do not have (seemingly) equal slopes. In order to confirm the relation, we construct a case where a failed and a successful change are clearly connected. This is our case study of the replacement of the reflexive use of the pronoun hem ‘him’ in a Middle Dutch border dialect by a lexical reflexive: sick, borrowed from Low German (the failed change), and sich, borrowed from High German (which finally succeeds) (Section 3). In Section 4 we then restore Kroch’s CRH by slightly modifying the model so that it includes failed changes. In Section 4.1 we will model the relation between failed and successful changes, derive the relation (Section 4.2) from other principles, and explain it in pre-theoretical terms (Section 4.4). As a nice consequence, we do not need to go beyond the logistic model to express the relation. In fact, only the logistic model is able to capture the relation between failure and success in a conceptually elegant way. The CRH, which links the
. For diffusion over social networks, see Van der Wurff (1990). . We take the maturation period as the only watershed in acquisition and capture, therefore, all types of acquisition of linguistic features after the maturation period as L2, be it adult innovations in their proper mother tongue or true second language acquisition. Weerman (1993) assumes these to be distinct, which is meant only sociologically not linguistically (personal communication).
The impact of failed changes
dynamics of the change to switches in the Principles and Parameter Model will be critically evaluated and replaced by a stronger version. In Section 5 we evaluate the results thus far. We come to the conclusion that the model can best be interpreted in terms of an L2-theory of linguistic change as sketched out in Weerman (1993). In Section 6 we reanalyse the data in Ellegård (1953) and Kroch (1989) in terms of the failed change model. The chapter closes with a pre-theoretical reflection on the model, its results and its consequences.
2. Successful changes Gradual linguistic changes that can be observed in a non-homogeneous language community were studied by Weinreich, Labov & Herzog (1968). Not contradicting the Chomskyan idealization of linguistic competence as displayed in a homogeneous community, but rather complementing it, Weinreich, Labov & Herzog study the gradual shift in a non-homogeneous environment from a linguistic configuration producing A to an alternative configuration producing B through all intermediate stages. Bailey (1973) proposes a quantificational model in which change starts slowly, picks up speed in the middle and then decelerates as it proceeds to completion (see the review of Naro 1984). This describes a typical S-shaped curve as often displayed in empirical studies (Piotrovskaja & Piotrovskij 1974) and illustrated in Figure 1. 1.0
logistic function
0.8 0.6 0.4 0.2 0.0
–5
0 x
5
Figure 1. The logistic function as an example of an S-shaped curve
There are various mathematical functions to model such an S-curve, e.g. the Gauss cumulative function (Aldrich & Nelson 1984, Piotrovskaja & Piotrovskij 1974), the Lorentz cumulative function, and the logistic function. Altmann et al. (1983) propose a model in which the change starts out with an increase which is proportional
Gertjan Postma
to the fraction of new forms, multiplied by the fraction of old forms. This establishes Verhulst’s differential equation:
(1) dS/dt = S․(1–S) 1 (2) S(t ) = (1+e −t )
(Verhulst differential equation) (logistic function)
The left-hand side of the equation in (1) represents the growth in a population S per time unit (dS represents the infinitesimal change in S, dt the infinitesimal change in t). The equation claims that the growth is proportional to two factors, S and (1–S), to be discussed below. The differential equation in (1) has the logistic function of (2) as its solution. This means that populations that satisfy equation (1) grow with a rate represented by the logistic function. This differential equation was first proposed by Verhulst in 1838 as an alternative to the differential equation of Malthusian growth, dS/dt = S, in which the growth in S is proportional to S. The latter captures the concept of unrestricted procreation. It gives an exponential growth curve (S = et ). The Verhulst equation in (3), on the other hand, has an additional saturation factor (1–S), which can be taken as a counterforce to unlimited growth. Instances of this counterforce are the decrease in food or space or some other factor with limited availability. In diffusion processes of linguistic innovations, the two terms may receive a probabilistic interpretation: the change for a switch in usage from A to B is proportional to the number of speakers who use B (S), multiplied by the chance of them meeting speakers that do not use B yet (1–S). This latter factor, of course, decreases and forms the saturating factor. In this way, Altmann et al. (1983) provide a specific S-curve, the logistic curve, as a model of linguistic change. This model is basically a twoparameter model. The actuation time (t0) can be modified leading to the term (t–t0) and the steepness of the change (a) can be modified, leading to (t–t0)/a.3
(3)
S(t ) =
1 t −t 0 − a 1+ e
(logistic model)
In the rest of this chapter we use the logistic model.
2.1 Kroch’s model Kroch (1989) adopts the logistic model without much discussion. For Kroch, any twoparameter S-shaped curve suffices (such as the logistic function, the Gauss cumulative . The actuation parameter in a model is that parameter that defines the process’s orientation on the time scale. The saturation rate is a third parameter but it is normalized to a dimensionless 1 (100%).
The impact of failed changes
function, or the Lorentz cumulative function). Kroch remains agnostic as to which model fits best: “these generally differ so little from the logistic that they can provide no improvement in fit to empirical data.” This is correct. Kroch adopts the logistic because of its widespread use in the life sciences and because of its graphical properties: when the occurrence rate is displayed on a semi-logarithmic graph against the time axis, straight lines are obtained.4 In Section 4.2 we propose theoretical reasons for exclusively adopting the logistic model derived from theoretical considerations of failed changes. The innovative part of Kroch’s study is that it makes a connection between the gradual diffusion of a change in the E-language, as described by the logistic curve, and more abstract properties of the I-language such as the parameter switch in the grammatical system. This relation is established in the so-called Constant Rate Hypothesis (CRH), which states that related changes have equal slope parameters. The CRH states that an underlying change in the parameter setting may manifest itself in various constructions with different actuation times but must proceed in these constructions with equal speed. In a two-parameter model, this means that the slope parameter of the different related S-curves must be equal, but the actuation parameter may be distinct. For instance, the curves in Figure 2a with equal slope but distinct actuation times may be tied to one parameter switch (they do not need to be!), but the curves in Figure 2b, with distinct slopes and equal actuation cannot to be reduced to one parameter reset. t0=1450, a=30 t0=1500, a=30 t0=1550, a=30
100
rate
80 60 40 20 0 1300
1400
period
1500
1600
Figure 2a. Kroch’s (1989) model – Logistic curves with distinct actuation times and equal slope
. Straight lines are only obtained provided that on the logarithmic scale the Fisher-Pry transform of the occurrence rate (‘odds’) is used.
Gertjan Postma t0=1450, a=30 t0=1450, a=20 t0=1450, a=10
1.0
rate
0.8 0.6 0.4 0.2 0.0
1400
1440 x
1480
Figure 2b. Logistic curves with equal actuation times, distinct slopes
Kroch does not derive the CRH from basic principles but motivates it empirically using four case studies taken from the literature: (i) the replacement of have by have got in British English (studied in Noble 1985) in four different environments; (ii) the rise of the definite article in Portuguese possessive noun phrases (studied in Oliveira e Silva 1982) in 6 different environments; (iii) loss of verb-second word order in French in two different environments (studied in Fontaine 1985 and Priestley 1955 respectively); and (iv) the case of periphrastic do (“do-support”) in six different types of constructions, as studied by Ellegård (1953). Kroch argues that in all these cases the slopes of the various S-curves are equal, while their actuation times may be different (as in Kroch’s case studies 3 and 4) but these do not need to be (as in Kroch’s case studies 1 and 2, where both activation times and rates are equal).5 On the assumption that these changes are tied to one and the same parameter, Kroch derives the CRH.6 Its contraposition says that whenever two changes have distinct slopes, they must be tied to distinct parameter resets.
2.2 Evaluation of Kroch (1989) Kroch’s model has been a major step in analysing linguistic change on the basis of E-language (corpora), without losing the connection to changes at the level of I-language,
. In Kroch’s case studies 1 and 2, the stability factors (i.e. the ratio between the rates of the constructions) remain equal during the change, which implies that both their slope parameters and their actuation times must be equal. . Kroch uses the unfortunate term ‘constant rate hypothesis’ for the more correct ‘equal rate hypothesis’. In the logistic model any change that fits to the logistic function has a constant rate during the change. We stick to the existing term, though.
The impact of failed changes
that is, tying the gradual change in the linguistic community to the discrete change at speaker level and on the level of the (individual) grammars. We take Kroch’s model as a starting point. However, there are data that are difficult to evaluate within Kroch’s logistic model, simply because of the fact that their diffusion rates do not have the shape of S-curves: some have bell-shaped curves. Nevertheless, Kroch simply treats them as having a logistic pattern. For instance, within the study of d o-support in English, the diffusion of do-support in positive declarative clauses is first forced into the logistic model, two parameters are extracted, one of which, the slope, turns out to be significantly distinct. Subsequently, under application of the CRH, it is concluded that this process is parametrically disconnected from the other instances of the diffusion of do-support. The logistic fits are given in Figure 3 as straight lines.7 At first glance, positive do-support (the black line) isolates itself as a process separated from the other lines.
2
rate
0 –2 –4 1400
1500
year
1600
1700
neg decl neg decl neg quest neg quest trans y/n trans y/n intr y/n intr y/n aff WH aff WH poss decl
Figure 3. The rise of do-support in English on a logarithmic scale (data from Ellegård 1953)
Some provisos must be made here, however. In the first place, the separation of the two sets of lines is not established on the basis of rates obtained by significant fits. The black line in particular is a bad logistic fit to the underlying data. Moreover, it is unattractive to disconnect the failed process in time from the first actuation of do-support, as they obviously occur more or less in the same timespan, as Kroch notices himself
. The vertical axis is logarithmic in addition to a Fisher-Pry transformation of the data. The regression calculations throughout this chapter are performed using Pro-Fit 6.1.11 for Mac. The algorithm is Levenberg-Marquardt.
Gertjan Postma
(1989: 29). This can be more clearly seen from Figure 4 without a logarithmic scale, where the successful changes are approximated by S-curves and the failed change by a significant fit to a bell-curve. 1.0 0.8
rate
0.6 0.4 0.2 0.0 1400
1500
year
1600
1700
neg decl neg decl neg quest neg quest aff trans yes/no aff trans yes/no aff int yes/no aff int yes/no aff WH aff WH poss decl poss decl
Figure 4. The rise of do-support in English on a linear scale (data from Ellegård 1953)
With respect to the actuation parameter, the first successful change(s) and the failed change seem to be related somehow: the peak of the bell curve coincides with the inflection point of the initial S-curves. Their activation times are equal.8 This cannot be expressed in Kroch’s theory, as the CRH does not connect processes on the basis of an equal actuation parameter, but disconnects processes with unequal slope. If one assigns a slope to the black line as if it were an S-curve (as Kroch does, cf. his Table 9), a slope comes out that is unequal to the slope(s) of the successful lines. What can be done if no slope can be assigned, because it is not a good fit? Kroch suggests that some “more complex model should be applied” (1989: 29). An important feature of this “more complex model” should include his observation that “the rate of use of do in the affirmative declarative context rises along with the other contexts … which suggests that the context is tied to the others”. It is this programme that we intend to design in this study. In the second case study used by Kroch, two processes are considered with respect to the increase of the determiner in Portuguese (cf. Oliveira e Silva 1982). Portuguese changes from the use of bare possessives as in meu livro to the use of possessive
. Since the inflection point of the S-curve and the maximum of the peak are the only independently motivated orientation points on the time axis, we can only take these as the activation time.
The impact of failed changes
constructions with definite determiners: o meu livro ‘the my book’. A similar increase occurs in colonial Brazil, which does not persist.9 The Portuguese data on the increase of the use of the pronoun in the motherland and in colonial Brazil are compared.10 The data and the logistic fits are graphically displayed in Figure 5. 11 Portugal Fit for Portugal Brazil Fit for Brazil
ln(odds)
2 0 –2 –4 1400
1600
century
1800
2000
Figure 5. Logarithmic plot of the article in possessive constructions in Portuguese with S-shaped fits (data from Oliveira e Silva 1982: 314–315)
Can we conclude from these data that these changes are not connected? One should keep in mind that one of these data sets (stars/dashed line) does not have the shape of an S-curve, and is not appropriate to be fit to an S-curve. Consequently the extracted parameters do not have statistical significance. One cannot evaluate two processes on the basis of these parameters and decide on their non-relatedness. If, on the other hand, we fit the failed change to a bell curve the following picture arises.
. The data are taken from the Rio dialect. In modern times, probably under influence of the southern dialects, a new increase of the article can be observed. We leave this later period out of consideration. . Oliveira e Silva includes data for the period long after the independence of Brazil. We only compare Portugal and Brazil as long as they formed a linguistic community of some kind. So we ignored the data after 1811. . Oliveira e Silva makes the observation that the fit is not significant.
Gertjan Postma 1.0
article in Portugal logostic fit article in Brazil logistic peak
0.8
rate
0.6 0.4 0.2 0.0 1400
1600
1800
2000
period Figure 6. Linear plot of the article in possessive constructions in Portuguese in S- and bellshaped fits (data from Oliveira e Silva 1982: 314–315)
Once again, the top of the failed change (dashed curve) seems to be related in time to the inflection point of the S-curve (drawn curve). Apparently, an attempt for change in colonial Portuguese and the success in local Portuguese are related, at least in time (i.e. in actuation). In sum, it is suggestive that the failed change and the successful change are connected in time and possibly in shape. The top of the failed change seems, in the case of do-support, to be connected to the first actuation of the successful change. So the following questions arise:
(4) – Is there a relation of some kind between a failed change and a successful change? – Can we construct conclusive evidence for such a relation? And if so… – Can we construe a model of failed and successful changes? – Can we connect them on the basis of (a modified form of) the CRH?
If the CRH is to be retained, we must attach some form of “slope parameter” to the bell-shaped curve of the failed change.
3. Failed changes In the previous section we discussed two successful linguistic changes that seem to be flanked with a failed change: the maximum of the failed innovation appears to be connected in time with the inflection point of the successful innovation. In the standard logistic model enriched with the CRH, there is a problem to connect the
The impact of failed changes
two innovations, since the framework does not make a connection on the basis of shared actuation times. Moreover, their slope parameters, assigned upon a fit to a logistic curve, are distinct (if assigned at all). It is plausible that the two developments are connected some way (as Kroch observes himself), although the framework is not able to express this. So, the logistic model seems to have a defect or limitation here, since application of the CRH leads to the undesirable conclusion that two processes, the successful and the failed process, are not connected. Moreover, the fit of one data set to a logistic function is not significant. Therefore, it would be good to have a conclusive case for which it can be argued on independent grounds that the failed and the successful change are connected. This can be done if the failed and the successful change show up in exactly the same context and they are in nature intertwined. This case is outlined in the next section and elaborated in the subsequent section.
3.1 Constructing a compelling case In this section we construct a case where a failed and a successful change are connected on independent grounds. In that way, Kroch’s model of logistic diffusion + the CRH will be proved to be incomplete: if the slope parameters of the fits are distinct, we have construed a CRH-refuting case where two processes with distinct slopes are connected, in conflict with the CRH. It must be kept in mind that one of these slopes is assigned by brute force and is not significant. Subsequently, therefore, we show that in a slightly modified logistic model, a more meaningful slope parameter can be assigned to a failed change. An additional advantage will be that in this modified logistic model the CRH can be maintained. The construction of a refuting case involves a failed and a successful change in one and the same context. For instance, where the failed and the successful innovations share a feature, call it [+s] (which is successful) but differ in a feature where one is successful and the other a failure, call it [+f]. (5) old situation innovation [–s] [+s, –f] (successful) [+s, +f] (failure)
Under the assumption that the f-feature is irrelevant in the old situation and that [+s] is orthogonal to the [f] feature, it follows that [+s, +f] and [+s,–f] are connected because of [+s]. As the change from [–s] to [+s, ±f] and from [–s] to [+s,–f] are related to one parameter, one would on the basis of the CRH predict that the changes have equal rates. If we then show that the empirical data provide distinct slopes, it follows that the CRH cannot be retained in its present form, or in its present interpretation. Furthermore, the failed and successful changes have equal actuation times. So the CRH is both too weak and too strong. It is too strong as two changes with distinct slopes are related, which is excluded by the CRH. On the other hand, it is too weak: related changes (the failed and
Gertjan Postma
the successful) have equal actuation times, which cannot be expressed in the theory. In the next section we work out this case. In Section 4 we develop an extension of the logistic framework that includes both successful and failed changes. This framework provides an interpretation of the CRH that makes it possible to retain it.
3.2 The refuting case: The rise of the reflexive in Middle Dutch The case that refutes the CRH, as designed in the previous section, can be found in the rise of the reflexive in Middle Dutch, as studied in Postma (2004). Middle Dutch did not have a separate reflexive pronoun comparable to German sich and Latin se/sibi, but simply used the pronoun hem ‘him’ in reflexive contexts, just as Modern Frisian does and Old English did. Later Dutch developed a separate reflexive, which is obligatorily used in locally bound contexts. The change happened at different times for the different dialects. Earliest in the northeastern dialects, latest in Holland. In Drenthe the change happened during the fifteenth century. Using a corpus of 5000 verdicts (AD 1400–1500, 215,000 tokens) from Drenthe, a northeastern region in the Netherlands, Postma (2004) argues that, during the course of one century, new reflexive pronouns sich and sick ‘himself ’, and their variants, collectively referred to as ‘s-reflexives’, arise at the cost of the reflexive use of hem. The numerical data are in Table 1.12 Table 1. Reflexive contexts in 15th century Drenthe Period
sick
sich
ø
hem
s-refl
total
%hem
% s-refl
%sick
1403 1420 1435 1450 1465 1481 1497 1516
2 0 9 4 11 21 0 3
1 0 6 6 4 42 24 11
0 0 0 0 0 0 3 0
33 17 18 10 21 11 12 3
3 0 15 10 15 63 24 14
36 17 33 20 36 74 36 17
0.895 1 0.568 0.450 0.566 0.135 0.325 0.176
0.104 0.0 0.431 0.549 0.433 0.864 0.675 0.823
0.08 0.0 0.25 0.18 0.33 0.28 0.0 0.17
After an initial increase of sick, this form gradually disappears in favour of sich halfway through the fifteenth century. Although sich is successful and the sick form is not, it is theoretically better to say that the s-reflexive forms gradually replace the hem-form in reflexive contexts and that the form sick fails. In Figure 7 we have represented the occurrence of the form sick and the total rate of the s-reflexive (sich+sick) in the corpus.
. The division into 8 periods is obtained from the 15 period division in Postma (2004) by merging the rough data of two periods (leaving the last one untouched).
The impact of failed changes 1.0
% s- reflexive % sick logistic fit to s- reflexive logistic peak fit to sick
0.8
rate
0.6 0.4 0.2 0.0 1400
1440
period
1480
1520
Figure 7. Rise of the s-reflexive and sick in 15th-century Drenthe
To connect this change to a parameter switch, it is essential to discuss the change that occurred in the I-language. In Postma (2004) I show that the change occurred as a consequence of the specialization of the pronoun hem going from being underspecified with respect to number (‘him’, ‘them’) to being exclusively singular (‘him’). Middle Dutch hem had, therefore, the same underspecification as German sich and Latin sibi/se. In the course of the fifteenth century, there is a decline of the underspecification of hem. The requirement of referential underspecification is a formal requirement of reflexivity (Reinhart & Reuland 1993). In other words, during the fifteenth century, the pronoun hem loses its capability to be used in reflexive contexts. The problem was resolved by borrowing reflexive pronouns from neighbouring German dialects, which had sick (Low German) and sich (Central German), respectively. So, hem-to-sick and hem-to-sich are both related to hem’s loss of number neutralization. These changes are therefore related to a switch in the parameter in the I-language. These changes are related with respect to E-language as well: we see that the peak of the failed change coincides with the inflection point of the successful change. Their actuation times are related. This refutes the CRH, which is too strong in this case. Moreover, with the naked eye, it can be seen from Figure 7 that the spreading around this value, that is the speed of the switch from 0 to 1 and the speed of rise and decline, are related as well. While the steepness of the solid curve increases, the dashed curve increases, and while the solid curve starts to flatten (inflection point), the dashed curve starts to decrease (it has its maximum). At this stage, we cannot make this precise, as we lack a model for the failed change.13 We may, however, take these points as evidence that failed and successful
. Anticipating the model to be developed, we give the data fitted with a logistic curve S and a logistic peak, respectively. The fitted parameters (actuation and slope) are 1452 (± 0.1) and 21 (± 0.1) for the S-curve, and 1457 (± 1.3) and 20.1 (± 0.9) for the peak, which are identical within the error.
Gertjan Postma
relations are connected. The CRH cannot express this connection. It is too weak. In the next section we will propose a model for it that minimally departs from the logistic model that solves these two problems. What still has to be cleared up is the reason that sick failed and the reason that sich succeeded. That will be discussed in Section 4.5.
4. Towards a model of failed changes In this section we model the relation between successful and failed changes. The aims are to relate successful and failed changes to each other in a systematic way. Moreover, we would like to assign parameters to failed changes that can be compared to Kroch’s parameters of successful changes, actuation and the slope parameter.
4.1 The model Intuitively, successful changes are S-curves and failed changes are bell-shaped curves. The question is how to model these concepts algebraically. Interestingly, S-curves and bell curves are mathematically related according to differential calculus: a bell shape is the first derivative of a corresponding S-curve. Alternatively, S-curves are the integral function of a corresponding bell curve. These relations have geometrical counterparts to which a linguistic interpretation can be given. The geometrical relations can be easily seen by inspecting Figure 8. 1.0
C
failed change succesful change
0.8
rate
0.6 B
0.4
D
0.2 0.0
A
1400x
1600 period
1800
Figure 8. Arithmetic relations between failed and successful changes
Consider the subsequent slope values of the S-curve, rather than the values themselves. The slope starts with a zero value (at A), gradually increases until its maximal slope at the inflection point (at B). The slope then declines until zero again (at C). If we put the slope values into a diagram of its own, the slope function has a bell shape.
The impact of failed changes
Alternatively, consider the area under the bell-curve until a certain value of the x-axis (the shaded area in Figure 8.). If the value x gradually increases, the area under the bell-curve that is cut off increases as well. The area grows most rapidly at the point where the bell-curve has its top (at D). Subsequently, the area continues growing but at a lower and lower rate. The S-curve is the cumulative function of the bell curve. This is true for all S-curves and corresponding bell curves. If we now apply this to successful changes and failed changes we get the hypotheses in (6), in mathematical form in (7).
(6) a. The failed innovation is proportional to the slope function of the successful innovation b. The successful innovation is proportional to the cumulative failed innovation
(7) Let S be an occurrence rate in a successful change and F an occurrence rate in a corresponding failed change, then
a.
dS(t )/dt = F (t )
b.
S(t ) = ∫−∞ F (t)d t
t
Although (7a) and (7b) are mathematically equivalent, they receive different linguistic interpretations. These interpretations are in fact reversed in cause and effect. We may interpret (7a) as expressing that the successful change S is basic and the failed change F a side effect: when the successful change has its highest impetus (at B), it causes resonance outside its proper domain, that is, in contexts that are less susceptible or appropriate for the change. The failed change is a spurious side effect. The formula in (7b), on the other hand, has a reversed interpretation as to cause and effect, which is slightly more difficult to grasp: it may be interpreted as a (failing) innovation proceeding autonomously for a certain period of time (because of fashion, language contact, or whatever reason). The successful change, then, is proportional to the cumulative usage of this (failing) innovation: a language community is susceptible to the number of innovative occurrences that they have heard and start to use systematically. We work out this interpretation in Section 4.4. Thus far, the theory has had nothing to say on what type of S-curve is used in the model. The most widely used S-curves are the logistic function, the Gauss cumulative function and the Lorentz cumulative function. They are related to their respective bell curves: the logistic peak, the Gauss peak, or the Lorentz function according to the mathematical relations in (7). In the next section, we see that the logistic model is the only model that is compatible with pre-theoretical intuitions of what a failed change is.
4.2 Deriving the model: Accidental and inherent failures In this section we derive the algebraic relations from basic intuitions of what a failed change is. What do we mean by a failed change? Now, nothing is eternal in this world, so even the most successful innovation will decline after a certain period and will be replaced by something else. We do not want to call the type of curve rendered in
Gertjan Postma
Figure 9a a failure, simply because the innovation disappears after some time. Changes that, after a certain period of reasonable success, turn out to be unsuccessful and decline, simply do exist. They should not be considered failures. Such an accidentally failing change, AF, we can model as AF = Sa · (1–Sb), with a and b representing distinct actuation times of the rise and the decline. On the other hand we have changes whose decline is intimately related to their rise, that is, whose decline is inherent and unavoidable. They are inherent failures or nonstarters. We rather mean the latter type when we speak in this chapter of ‘failed changes’. For accidental failures, the actuation of the decline is (much) later than the actuation of its rise, as displayed in Figure 9a. 1.0
actuation at 1450 actuation at 1800
rate
0.8 0.6 0.4 0.2 0.0
1400
1600 year
1800
Figure 9a. Representation of an accidentally failed change: composed of two S-curves with remote actuation times
The actuation times are indicated with dashed lines. The basic intuition now is that the more the actuation times of the rise and fall come closer to each other, that is, the more the dashed lines come closer, the more ‘failed’ the change is, as in Figure 9b. 1.0
logist with activation time 1500 logist with activation time 1650
rate
0.8 0.6 0.4 0.2 0.0
1400
1600 year
1800
Figure 9b. Representation of an accidentally failed change: composed of two S-curves with close actuation times
The impact of failed changes
If the two actuation times coincide, that is, the innovation starts declining from the moment of its rise, the change is an inherent failure. From this perspective, the logistic model allows a change that includes from its very beginning the ingredient of failure, but nevertheless obtains some non-zero value. Put differently, although the failed change does not exist in the I-language, it nevertheless shows up in the E-language, albeit ephemerally. This situation is displayed in Figure 9c, where the two actuation points coincide. 1.0
S(t).(1-S(t)) with S the logistic function (double activation time at 1500)
rate
0.8 0.6 0.4 0.2 0.0
1400
1600 year
1800
Figure 9c. Representation of an inherently failed change: composed of two S-curves with equal actuation times
Let us, therefore, define a failed change as a change that is composed of two factors, a rising S-curve multiplied by a falling S-curve, whose decline has an activation parameter equal to its rise. We make a similar assumption concerning the slope parameter, since this assumption constitutes the simplest model with the fewest parameters possible. We come back to this choice in the next section. So let us assume that a failed change is a change whose actuation and slope parameters are equal to those of its decline, that is, Fa = Sa · (1–Sa) with two S-curves with equal parameter sets. We call Sa the constituting function of the failed change Fa. Now that we have defined a failed change, we can see what predictions it gives in (and only in) the logistic model. In the logistic model, the S-curve is a solution of Verhulst’s differential equation given in (1), repeated in (8).
(8) dS/dt = S ⋅ (1 − S )
(Verhulst 1838)
Now, the right-hand member of (8) is identical to a failed change, as we just have defined it: F = S(1–S). Consequently the failed change is proportional to the first derivative of a successful change that constitutes it. The hypothesis is now that the constituting S of the failed change de facto is realized in the E-language as the flanking successful change. In other words, Verhulst’s equation in (8) establishes an algebraic connection of the failed and the successful change. The successful change is the solution of the equation (S, the logistic function). The failed change is shown by the equation. Both the left and the right
Gertjan Postma
member show the algebraic expression of the failed change: it is the first derivative of the successful change, as well as a superposition of two S-curves, one going up (S) and one going down (1–S). In (9) we give the algebraic shape of this failed change. It is the first derivative of the logistic curve. It is also known as the Hubbert curve.14 (9)
F (t ) =
e
−
(t − t0 ) a
(t − t0 ) − a e 1+
2
In Section 4.4 we give a conceptual justification for the model. For the time being, we are only concerned with the empirical justification. Let us, therefore, provide best fits to the failed and successful trajectories of the various reflexive pronouns in Drenthe, and use the logistic function as given in (2) and the logistic peak as given in (9). We then we arrive at the following values for the parameters, t0 = a[1] and a = a[2]. Table 2. Parameter fits by the logistic function and the logistic peak to the rates of the s-reflexive and sick, respectively sick (Chi squared = 5.8851e–2)
s-reflexive (Chi squared = 0.1369) Parameters
Standard deviations
Parameters
Standard deviations
a[1] = 1458.25
Δa[1] = 7.9
a[1] = 1463.6411
Δa[1] = 7.9816
a[2] = 28.1549
Δa[2] = 8.4
a[2] = 16.9110 a[3] = 1.2251
Δa[2] = 5.9411 Δa[3] = 0.3
The two fits are made in separate fitting procedures. As we see, the a[1] parameters which represent t0, that is, the curving point, and the peak value of the failed change (1458 and 1463 respectively), are equal within the error. Similarly, the a[2] parameters, which represent a, that is, the steepness of the successful change, and the failed change, are also equal within the error (28±8 and 17±6 respectively). The latter identity confirms the CRH applied to failed changes. The former is the additional identity of the actuation times of the failed and the successful changes, which identity is not predicted by the CRH but has its natural place in our model.
4.3 The symmetry of the failed change In the previous section, we modelled the failed change as the superposition of two successful changes with equal actuation parameters (by virtue of being an inherent failure) and equal slope parameters (for the sake of simplicity), algebraically: Fa = Sa · (1–Sa). This assumption causes the peak curve of the failed change to be symmetric. A similar
. An alternative mathematical representation is:
1 2 + 2cosh ((t – t 0 )/a)
The impact of failed changes
assumption of (point) symmetry is commonly made about the shape of S-curves (the initial rise is in shape similar to its final saturation behaviour). Not only does this assumption realize the simplest model, it is essential for Verhulst’s equation as the algebraic expression of the failed change model. It is, of course, an empirical issue whe ther this presupposed symmetry of the peak is legitimate or not. The data suggest it is tenable, though a 4-parameter model, obviously, gives better fits than a 2-parameter model. The symmetry assumption is natural in a theory that models failed changes in a principled way. Tony Kroch (pers. comm.) and an anonymous reviewer suggest giving an independent interpretation to the two S-curves, S1 and S2, that constitute the failed change according to the formula F = S1 . (1–S2). The factor S1, describing the rising flank of the failed peak, might then be connected to some (successful) change in one direction, and S2, describing the declining flank of the failed peak, might be connected to another (successful) change in the other direction. This is in fact the accidental failure model. Under such a model, there is no principled reason to assume both slope parameters that jointly constitute the peak to be equal. Moreover, there would be no reason that the two actuation parameters should be equal either, the only restriction being that the actuation of S2 is after S1. Let us apply this suggestion to do-support, where the failed change (positive declarative do, the lowest solid curve in Figure 3) is tied to successful changes in negative and interrogative contexts (especially the initial dashed curve in Figure 3). In terms of the features [s] and [f] of Section 4.2, this proposal can be formulated as follows. The rise and decline of do-support can be analysed as [–s] = V-to-I and [+f] = positive declarative do-support (PAD). The PAD-peak can then be analysed as in (10), where PAD is split into two periods, one in which [+s, –f] competes with verb raising [–s] (this is period 2) and one in which [+s, +f] competes with affix hopping [+s, –f] (this is period 3), both following an S-curve. The assumption is now that these two S-curves, multiplied with each other, form a peak. (10)
period 1 [–s]
period 2 →
[–s] [+s,+f]
period 3 →
[+s, –f] [+s, +f]
period 4 →
[+s, –f]
Now, as one reviewer argues, since PAD (i.e. [+s, +f]) competes with different alternatives in period 2 and 3, (with V-to-I and affix hopping, respectively), there is no reason why the CRH would predict an equal slope parameter to the rising logist S1 and the declining logist S2. This approach, therefore, allows for an asymmetrical peak for the failed change, its increase tied to the slope parameters of the other do-support processes (the S-curves of negative do-support), its decline to some other process. Now, the only slope parameter that governs the change is the slope parameter of the S-curves, which are, according to the CRH, all connected to each other. So there is no possible slope parameter within the overall process. So we either state that the slope
Gertjan Postma
parameter of the peak’s right flank is equal to the slope parameter of its left flank, or we find a second process to which it is tied. The onus of finding this process is on this asymmetric model approach, which is probably at odds with the CRH itself. For, if such process could be found at all, it must be triggered by a parameter switch that is different from the one we are investigating, and do-support as a unitary process would be lost. In this study we simply adopt the null hypothesis, that the slope parameter of the decline is equal to all the other S-curves, in full agreement within the CRH. This is, of course, an empirical question. So, if these slopes turn out to be systematically and significantly unequal (with opposite sign), our model would be incorrect. Unfortunately, the test must be left for further research as none of the data are sufficiently detailed to perform this discriminating test.15
4.4 Explaining the model In this section we discuss the two ingredients of the mathematical model sketched in the previous section. The first ingredient is the assumption that failed changes exist in natural language. This raises the question: how can a change be an inherent failure? The second ingredient is the hypothesis of a flanking successful change that is identical to the constituting logistic function within the failed change. This raises the question: how can a failed change be dynamically connected to a successful change? In order to legitimate these assumptions, we must return to the nature of linguistic innovations. The older assumption in the principles and parameters model of generative grammar is that innovations occur as an imperfect transmission in L1 learning. Language change is a change of the grammar in L1 learners (Lightfoot 1999, Andersen & Lightfoot 2002). The reanalysis model is the standard implementation of this idea. This L1scenario has faced various objections: why would language change? Are children such bad learners? It conflicts with the idea that the language acquisition device is extremely robust (Pinker 1984). And there is, of course, the conflicting data problem: if children have a grammar distinct from their parents, how do they deal with the conflicting input data that will certainly be heard? To avoid some of these problems, it has been proposed that L2 acquisition be the locus of language change (Bybee & Slobin 1982; Aitchison 1991). These authors locate the change in peer group innovations, which subsequently spread through the language through social networks. Now, the status of L2 learning has been an issue of much research and debate (cf. Epstein, Flynn & Martohardjono 1996). Despite the differences in approach, it is commonly felt that the essential difference between L1
. The rather technical reason is that the fitting procedure becomes unstable in this asymmetric model. Being de facto a 5-parameter model, it is extremely flexible and allows for too many, often bizarre, best fits.
The impact of failed changes
and L2 learning is that L2 learning involves persons after the maturation period, that is, persons who do not have full access to Universal Grammar (UG) or the Language Acquisition Device (LAD) and cannot acquire new parameter settings after the maturation. It may be clear that the idea of adult innovations as the trigger of language change poses a problem for the parameter theory, as it must assumed that adults are able to set (new) parameters. Consequently, the L2-hypothesis cannot be applied to the parameter model of grammar. Some enriched model is necessary. An interesting model of language change where both L1 and L2 factors are active is found in Weerman (1993). Basing himself on Clahsen & Muysken (1986), Weerman formulates two principles of acquisition: (11) a. b.
Children cannot deviate from positive evidence Parameter change can only take place under L1 acquisition
The first principle is taken from a deterministic hypothesis of acquisition. The second is the standard maturation assumption in acquisition studies. The two principles exclude change through parameter resetting in the Principles & Parameters model, because, firstly, children cannot deviate from what is offered, and secondly, only children can set parameters. To escape the dilemma of a deterministic connection between E-language exposure and acquired settings in the I-language, Weerman suggest a ‘peripheral rules solution’. This model adds a secondary ingredient to the model that determines the interaction between E and I-language. Apart from standard L1 acquisition by setting parameters, Weerman assumes adult innovations which are stored in the I-language in the form of peripheral rules. Peripheral rules are rules that do not belong to the core grammar but are applied to the output strings. Application of peripheral rules is not produced by parameterized UG, but by later sociological accommodation. These peripheral rules are active in later stages of first language acquisition (‘late L1’) as well as in the process of second language acquisition. Reformulated in modern terms, peripheral rules are not produced by the language faculty proper but by the human problem-solving faculty and, hence, subject to variation. In a system enriched with peripheral rules we can model language change under retention of the principles in (11). Weerman’s peripheral rules solution to language change boils down to the following sequence of events: parameter settings are transmitted flawlessly to the next generation, but adolescents and adults change their output by adding peripheral rules (they cannot change parameter settings). The next generation, now, taking all data as primary linguistic data, construct a new parameter setting based on the complete input. Speaking on the change from OV to VO in English, Weerman writes about the increased number of VO-patterns: “My suggestion is that these (VO) leakages were so to speak exaggerated via L2-M acquisition [adult innovations; GJP], both quantitatively and qualitatively. The
Gertjan Postma
relevant speakers do not change their internalised setting of the head parameter. From the perspective of their L1-grammar these over-generalisations are ungrammatical [my italics; GJP]. What they do is add a peripheral rule. A next generation, however, could set the head parameter”. (Weerman 1993).
We will speak of ‘off-grammatical’ constructions whenever innovations are not produced by core grammar but by peripheral rules. A similar, though less explicit, proposal is found in Van der Wurff (1990: 30). Van der Wurff considers a language produced by grammar G2 to which isolated new constructions (which he calls Cc, be they borrowings or innovations from G1) are attached. If these innovations are compatible with G2, basically nothing happens. Suppose, however, the more interesting case that these additional constructions are incompatible with the grammar G2. This creates an unstable complex (G2, Cc) which is then transformed to a stable (G3, Cc’). In general, language change proceeds through such unstable intermediate complexes. A more interface-type approach of the adaptive nature of L2 acquirers can be found in Vainikka & Young-Scholten (1994). They show that L2 learners fall short on morphology or, rather, they fall short on the fine tuning between syntax and morphology, that is, the alignment of morphological forms and the accompanying syntax. An example of such a mismatch between morphology and syntax is an interlanguage that has lost rich inflection but retains V-to-I or vice versa. The idea of failed changes makes it possible to operationalize these ideas. Combining the unstable intermediate complexes with the morphology-syntax misalignment proposed by Vainikka and Young-Scholten with Weerman’s off-grammatical adult innovation, we may think of a failed change as an off-grammatical adult innovation that serves some linguistic or sociological goal but does not have the perfect alignment of morphology and the parameter settings. A successful change, on the other hand, is a change that is compatible with all principles of UG and has a perfect alignment. Successful changes are introduced by L1 learners, for instance as accommodations of the failed change initiated by adults. Let us, therefore, represent language as a pair consisting of a parameterized Grammar G1 and a lexicon D1. Suppose adults start using a new lexeme (be it a deflected verb or a pronoun with a different specification, etc.), that is, their lexicon changes into D2 that is not fully compatible with the old grammar. They fail to align the lexeme and the grammar properly: *[G1, D2]. Alternatively, they can change the grammar but fail to align it properly with the morphology: *[G2, D1]. Both situations are unstable and resolve to a fully aligned [G2, D2]. Only non adult speakers are able to make this alignment before the maturation time. We call it an L1-accommodation of the instability represented by [G1, D2] or [G2, D1].
The impact of failed changes
(12) Old situation → Intermediate situation grammatical [G1, D1]
→ New situation
off-grammatical mismatch
grammatical
adult innovations (L2)
L1-accommodations
*[G1, D2] or *[G2, D1]
[G2, D2]
failure
success
This scenario, essentially taken from Weerman (1993) and Van der Wurff (1990), resolves various problems stated above, most notably the relation between the failed and the successful change. The problem of the conflicting data in the reanalysis scenario disappears as well: these conflicting data are the very trigger of the change. The essential ingredient of the model is that it allows for off-grammatical data, that is, data without full alignment between syntax and morphology for a relatively short time: the innovation inherently fails, but may have an impact on the system.
4.5 Applying off-grammaticality to the case study In order to illustrate the proposed L2 mechanism of language change and the role of failed changes in it, let us apply it to our case study of the rise of the reflexive in Dutch. First we must adopt the assumption that sick was first introduced into Dutch to solve the growing unsuitability of hem as a reflexive. The unsuitability of hem was felt with the L1 speakers of Drentish. (For the reason why hem became unsuitable, cf. Section 3.2 and Postma 2004). So, the adult speaker borrowed sick or sich into their language. Both sick and sich were suitable as reflexives since they display the desirable number neutrali zation. Obviously, this suitability is not accessible to the innovative adults, they simply borrowed these upon the unsuitability of hem, but the suitability of sick/sich is accessible to the next generation. Though feasible as reflexives, both sick and sich were as such incompatible with the structure of the Dutch morphological system (to be elaborated below), and should fail unless something happened. The idea is that this something was possible with sich but not with sick. Let us start with the incompatibility. The reason why sick/sich are incompatible can be extracted from the scheme in (13). (13) Accusative reflexive pronouns 1p 2p 3p
Drenthe mi di hem
new mi di *syk/sych
Low German mik dik sik
Central German mich dich sich
In (13) we give the (reflexive) pronominal system (‘me/myself, you/yourself, him(self)’) in Drenthe of 1400 and of two types of German. The German pronouns have a transparent accusative marker -k/ch. After specialisation for number, hem was replaced by
Gertjan Postma
syk/sych. Now, while sich and sik have a compositional status in the German system (person+case), this cannot be retained upon borrowing into Dutch, since these accusative morphemes are absent. This means that syk/sych must enter the system in a morphologically non-analysed way. If it were analysed bi-morphemically, some new feature should be tied to the -k/-ch ending. Now, as we will see, it could not be monomorphemic for phonological reasons (see Section 5). In other words, syk/sych had a problem: they had to be reanalysed. Why then could sych could do this but syck could not? A potential answer comes from a study on the structure of the reflexive by Barbiers & Bennis (2003). Transposing modern dialectal Dutch ze eigen ( = ‘ze+own’) to other reflexive structures in Dutch, they come to the conclusion that it is “reasonable to consider zich as a morphologically complex lexical item in which the possessive pronominal element ze is combined with the possessive affix -ig”. So, sych (just as its modernly spelled counterpart zich), could be reanalysed as composed of a pronominal part ze plus a possessive morpheme -ig. The older borrowing sik could not be so reanalysed, as no (possessive) morpheme -k existed in the Drenthe dialect).
4.6 Evidence for the off-grammatical outset In this section we provide supporting evidence for the off-grammatical outset of the borrowings of reflexive pronouns in Drenthe. This evidence can be drawn from the vocalism of the borrowings. In the beginning of the fifteenth century the reflexive pronouns sik and sich, borrowed from Low and Central German, show up in the texts as syck and sych respectively, that is written with /y/. In the course of this century, the y-spelling decreases and is replaced by /i/: sick and sich. The rise of /i/ is graphically represented in Figure 10 by the diamonds, together with a logistic fit (solid line). For comparison, we also put the curve of the rise of the reflexive itself (the dashed line). The curves are remarkably similar. 100
S-refl/S+hem S-refl/S+hem lax vowel i/i+y lax vowel i/i+y
80
rate
60 40 20 0
1400
1440 year
1480
1520
Figure 10. Increase of s-reflexive and its lax vocalisation /i/ at the expense of /y/
The impact of failed changes
Within the tradition of older Dutch spelling (the Drenthe corpus is written in an Utrecht-based spelling),16 the spelling /y/ is used for the high and tense sound [i(:)], be it long or half long, while the spelling /i/ is used for the high-mid vowel [I], probably lax. Now, this is an argument for the L2 innovation status of this innovation. The reason is that German short [I] in [sIk] or [sIx] is perceived by Dutch speakers as long, tense /ie/, that is, German [mIk] is perceived by the Dutch as a long [mi:k], probably because of the quality cueing of the Dutch short-long opposition. Now, according to phonological analysis, a long [si:k] is to be analysed phonologically as bisyllabic and, therefore, bi-morphemic in the morphological domain. Now, [si:x] could be reanalysed as bi-morphematic, namely as possessive, but [si:k] could not. The following development sketches how the -k variant succumbed. (14) stage 1 hem
→
stage 2 stage 3 *syk → *si+ik → *sych si+ich
stage 4 – sich
Notice that [sIk] could have survived, in principle, but this step came too late for this variant. It had already disappeared. In fact, many present-day Low German dialects show stable mi/di/sik, with short lax /i/ in sik. It is the contact variant mi/di/s[i:]k that was instable and selected s[i:]ch. So, the grammaticalization process of the off-grammatical innovation [si:k] not only comprises a consonantal change of /k/ → /x/ but, simultaneously, a vocalic change (shortening the vowel). The change in the vowel is evidence for its borrowed nature by adult innovators, and for its initial marked status in the receiving language.
5. Discussion Let us evaluate the results thus far. Kroch (1989) has proposed a logistic model of linguistic change on the basis of the qualities listed in (15). (15)
– The logistic model establishes a simple two-parameter model – The model is well known from life sciences – The model can be easily displayed and graphically manipulated – The CRH ties its dynamics to the P&P model (connection between I-language and E-language)
In the previous section we argued that the existence of failed changes constitutes a problem for the logistic model, but only at first glance. On second thought, the
. The verdicts are written by the clerk of the bailiff of Coevorden, on behalf of the bishop of Utrecht.
Gertjan Postma
inclusion of failed changes confirms the superiority of the logistic model on the basis of the additional qualities listed in (16). (16)
– The logistic is a solution of a simple differential equation – The model is applicable to failed changes – The model relates failed and successful changes algebraically – The failed change is the first derivative of the successful change – The successful change is the cumulative function of the failed change – The model allows ephemeral virtual changes (in the E-language but outside the I-language)
Furthermore, the failed change model has the theoretical interpretation given in (17). (17) – Failed changes are the initiators of successful changes – Failed changes are off-grammatical adult/L2 innovations – Successful changes are L1-fixings of off-grammatical adult/L2 innovations
Despite the advantages, there are some problems with this model. These show up when we try to apply the model to the rise of do-support in the history of English.
6. Do-support in positive declaratives in English revisited In the previous sections we designed an algebraic model that included failed changes along with successful changes. The failed change is proportional to the first derivative of the successful change. The logistic model constitutes a two-parameter model with the actuation time and the slope as its parameters. These two parameters define when and how quickly a change proceeds from a penetration of 0 (or 0%) to a penetration of 1 (or 100%). Notice that in this model the occurrence rate of the innovation is normalized with respect to the absolute occurrence of the construction. It is therefore strictly speaking a 3-parameter model. If the failed and the successful innovations occur in exactly the same pre-defined set of constructions, that is, when they target the same ‘market’, the model predicts that the failed change will reach a maximal penetration of 1/4 (25%) before it declines. This is a reasonably correct prediction for the rise of sick in Dutch. In most of the cases, however, there is no predefined target. In brief, there is a normalization problem in the modelling of successful changes, which becomes even more urgent in the modelling of failed changes (we cannot simply choose some relatively late t). Normalization is especially difficult in its relation to the successful change if the failed change targets another set of constructions. The normalization problem is one of the reasons that we find deviations from the predicted 25%: for example, it is much too low in the failed do-support (8%), displayed in Figure 4, repeated here as Figure 11.
The impact of failed changes
1.0 0.8
rate
0.6 0.4 0.2 0.0 1400
1500
year
1600
1700
neg decl neg decl neg quest neg quest aff trans yes/no aff trans yes/no aff int yes/no aff int yes/no aff WH aff WH poss decl poss decl
Figure 11. The rise of do-support in English on a linear scale (data from Ellegård 1953)
Why is the failed change rate so low in some cases? In this section we address the normalization problem for the case of the failed do-support, as it has been discussed from the very beginning and was one of the reasons for Ellegård to start a quantitative analysis. Therefore, do-support is illustrative for what problems the analysis may encounter with respect to normalization.
6.2 The normalization problem The earlier scholars who studied do-support noticed that the occurrence of periphrastic ‘do’ was initially (i.e. during the 15th century) predominantly observed in positive declarative clauses (Dietze 1895; Trnka 1930: 53; Engblom 1938), that is, precisely where it no longer occurs in modern English. These scholars came to this impression because they considered the set of do-support clauses. In this set, the majority of clauses are positive declarative. However, as Ellegård noticed, to draw a conclusion of dominancy in general from this fact is a methodological flaw as positive assertions exceed by far the number of interrogative and negative sentences (by roughly a factor 17). Ellegård’s quantitative work made it possible to shift from considering absolute frequencies to relative frequencies in the respective constructions. Ellegård argues that there is “absolutely no justification for supposing that the frequency was at any time higher in affirmative sentences than in others”. Ever since, the rate numbers have been calculated proportional to the occurrence rate of the construction in which they show up. All data in Kroch (1989) are normalized in this way. From the perspective of I-language this seems justifiable. Things change, however, if we calculate the impact of a change in a construction (for instance a failure) on other constructions (where it
Gertjan Postma
might become successful). This might be an E-language phenomenon and hence the absolute rates might be important. The nature of the too small peak of the failed change in the case of do-support (8%) can be understood from the perspective of the overall increase of do-support. The relevant data are not in Kroch (1989) but can be found in Ellegård’s Table 7. We have plotted these in Figure 12 (the dots).
rate of clauses with ‘do’
0.12
rate
0.08
0.04
0.00 1400
1500
1600
1700
period Figure 12. Overall rate of the do-construction (data from Ellegård 1953: table 7)
We see that do-support starts out in the early fifteenth century, increases rapidly in the sixteenth century, declines slightly and goes to a saturation of 7% during the seventeenth century. Interestingly, the increase displays an overshoot up to 11% around 1550, the nature of which is not discussed in Ellegård, nor in the literature afterwards. From the ‘failed change’ perspective, it is natural to identify the peak with the failed change, whose peak coincides with the inflection point of the successful change. The question is of course whether the failed change perspective can appropriately model these data. The modelling of the empirical data given by Ellegård needs requires careful scrutiny. As said, the increase in do-support has a peak around 1550 and then goes asymptotically down to the level of 7%. This value is the relative number of negative and interrogative contexts in corpora in general. This reflects the fact that the rate of dosupport in Neg and Q contexts goes to 100%. In contrast to the case of sick/sich, the failed change in the case of do-support is not part of the successful change but realizes itself outside the domain of success. The failures are not included in the success, but cause a peak that is superposed to it. This peak is rather strong because of the fact that the contexts of affirmative declaratives are much more numerous than the contexts of successful changes (in negative and interrogative clauses). In this case we need two
The impact of failed changes
normalization factors, a normalization factor of the overall process (a3) and a relative weight factor of the failed and the successful change (a4). So we arrive at the formula in (18) (For a derivation, see the appendix). (18) Rate = a3 ․ [S+ a4 •F], with S a logistic function and F a logistic peak
As to the internal parameters (a1,a2) of S and F, these are as follows. According to our hypothesis, the activation and the slope of the successful and failed changes are equal. So we predict that the measured data should fit to a function of the form in (19). (19) Rate = a3 ․ [S(a1,a2) + a4F(a1,a2)]
where a1, a2 are the activation time and the slope parameter. In this fit, the slope and the activation parameter of the failed change (the peak) are taken to be identical. The third parameter, a3, is the saturation level, that is, the proportion of negative and interrogative constructions that are feasible for do-support (7%±2). This parameter is needed in any growth process (already for dimensional purposes). The only additional parameter in this model is a4, the relative strength of the peak and the logistic function. In Figure 13 we show a best fit of this function to the data given by Ellegård. It shows that this model is quite faithful to the data.
fraction of clauses with ‘do’
0.12
rate fit to logistic+ logist peak
0.08
0.04
0.00 1400
1500
1600
1700
period Chi squared
=
Parameters: a[1] = 1557.2296 a[2] = 15.0347 a[3] = 7.0679e-2 a[4] = 4.0024
1.1473e-3 Standard deviations: ∆a[1]= 5.9884 ∆a[2]= 3.2376 ∆a[3]= 8.5145e-3 ∆a[4]= 1.1146
Figure 13. Overall rate of do-support with algebraic fit in the failed change model
Gertjan Postma
We conclude that the failed change model can be adequately applied not only to cases where the failed change is part of the successful change but also to cases where the failed and the successful change apply in separate syntactic contexts.
6.3 Calculation of the 8% level of the failed change Now that we have disentangled the failed and the successful change algebraically, we are in a position to normalize the failed change peak. Consider the decomposition of the curve in Figure 14 into a successful part and a failed part. 0.12 0.10 0.08 rate
observed rate success + failure successful change
B
0.075
0.06 A
0.04
0.035
0.02 0.00 1400
1500
1600
1700
period Figure 14. Data analysis of do-support in the failed change model
The peak of the failed changes corresponds with the inflection point of the successful change. Now, the failed contribution in the overall rate of do-support is the segment A–B, which is 0.075, the level at the peak value A (0.11) minus the level at inflection point B of the successful change, 0.035 ( = 0.07/2). In order to get the rate of do-support in positive declarative clauses, we must not normalize it to the total number of clauses but to the positive declarative clauses only, that is, 100%–7%, the ratio of WH and Neg clauses. Using the subtraction 1–0.07 = 0.93, we arrive at the value of 8% for the peak of the failed change, according to the calculation in (20). (20) 0.11–0.035/0.93 = 0.075/0.93 = 0.08
This corresponds to the measured value of 8% in Figure 11. We conclude that the low peak of the failed change in the relative frequency diagram in Figure 11 is a mere consequence of the numerical dominance of positive declarative clauses. If we use absolute levels of occurrence, both the successful and failed changes have the same normalization factor. There is an important theoretical implication of this result. It means that the failed and successful change interact at absolute occurrence levels, not
The impact of failed changes
at relative occurrence levels. Put differently, the interaction of failed and successful change occurs in the domain of E-language, not at the level of I-language. The relative occurrence level, on the other hand, is a consequence of the high occurrence of positive declarative clauses (where it fails) and the low occurrence of negative and interrogative (where it succeeds).
7. Should all linguistic data be considered grammatical? While all linguistic data are to be taken with the same scrutiny, it is far from clear whether all data are to be considered grammatical. Especially in periods of linguistic change, the general assumption that all data are grammatical cannot be maintained, for, if this were so, no change would be possible. One way to go is to assume a twogrammar system where the data that are ungrammatical in one system are taken to be grammatical in the other and vice versa. This approach is generally applied in generative studies and captures the change as a gradual diffusing replacement of one grammar by the other in a community by direct grammar competition (Kroch 1989; Santorini 1992; Henry 1995; Pintzuk 2002; Postma 2004). This grammatical approach considers the extent to which the two grammars overlap with respect to their respective outputs to be the possible locus of (the) change and has functioned as a successful heuristic tool in diachronic research. Within this overlap, reanalysis is possible: a string produced by one grammar can be parsed within the other. However, the reanalysis hypothesis, with Lightfoot as the major advocate, has struggled with the conflicting data problem: on what basis does an L1 learner ignore data that conflict with his new parameter setting? As yet, this problem has not been overcome. Furthermore, the L1 approach does not give consideration to the influence of small sociolinguistic changes under influence of peer groups who deliberately change their E-language (Labov 1972; Bybee & Slobin 1981; Sankoff & Blondeau 2007). Another way to go is to assume that linguistic change proceeds through a linguistic stage that includes fundamentally off-grammatical data. Under the assumption that L1 acquisition fully complies with the rules of UG and produces configurations that are generated by an instance of UG, the off-grammatical approach utilizes L2 changes imposed by adult speakers or adult L2 learners. Such changes imposed by L2 learners, then, (may) cause a reshuffle in the core grammatical system. By its very assumptions, the second approach deals with the relation between E-language and I-language, and is fundamentally sociolinguistic. The obvious problem here is how to distinguish the grammatical and the off-grammatical data. In this study, we have developed a simple model that is a variant of Kroch’s model and which includes the sociolinguistic factors that feed the transition as claimed by Labov. We have argued that to any change that follows an S-curve (the so-called ‘successful change’), there is a ‘failed change’ of sociolinguistic nature that
Gertjan Postma
fuels the successful change. The unsuccessful change consists of off-grammatical variants that die out after a while. In those cases where we can empirically distinguish the off-grammatical data from the grammatical data, the peak-type curve of the unsuccessful change maintains an algebraic relation with the successful change. In this way, we take the transitional data, despite their off-grammaticality, seriously as produced by peripheral rules. More precisely, we have shown that these transitional data are offgrammatical in the sense that they are ungrammatical in the system both before and after the linguistic change. Nevertheless, these off-grammatical changes do not escape from Kroch’s (1989) Constant Rate Hypothesis.
Appendix If we ignore the number of focussed positive declarative do sentences (just as Ellegård does), we have N (do) = Suc(do) + Fail(do)
Hence, dividing both members by TOT, the total number of clauses considered, we get: N (do) Suc(do) Fail(do) = + TOT TOT TOT
Furthermore, if we divide and multiply by NPI (the total number of WH and NEG clauses) and divide and multiply by POS (the total number of positive declarative clauses), we obtain: N (do) Suc(do) Fail(do) Suc(do) NPI Fail(do) POS NPI = + = ⋅ + ⋅ ⋅ TOT TOT TOT NPI TOT POS NPI TOT NPI Suc(do) Fail(do) POS = NPI + POS ⋅ NPI TOT = a3[S + F ⋅a4 ]=
In sum, rate(do) = a3[S + F ⋅ a4 ]
where, a3 is the overall fraction of interrogative and negative clauses (NPI/TOT) and a4 is the overall fraction of positive declarative and interrogative + negative clauses (POS/ NPI). These ratios are not known, but are supposed to be constant in time and can be treated as fitting parameters to the graph. Results are a3 = 0.07 (7%) and a4 = 4.0. Note that we have taken the rate of do-sentences with respect to all clauses, modal clauses included.
The impact of failed changes
References Aitchison, Jean. 1991. Language Change: Progress or Decay? Cambridge: CUP. Aldrich, John Herbert & Nelson, Forrest D. 1984. Linear Probability, Logit and Profit Models. Beverly Hills CA: Sage. Altmann, Gabriel, Buttlar, Haro von, Rott, Walter & Strauß, Udo. 1983. A law of change in language. In Historical linguistics [Quantitative Linguistics 18], Barron Brainerd (ed.), 104–115. Bochum: Brockmeyer. Andersen, Stephen & Lightfoot, David. 2002. The Language Organ: Linguistics as Cognitive Physiology. Cambridge: CUP. Bailey, Charles-James. 1973. Variation and Linguistic Theory. Washington DC: Center for Applied Linguistics. Barbiers, Sjef & Bennis, Hans. 2003. Reflexives in dialects of Dutch. In Germania et Alia: A Linguistic Webschrift for Hans den Besten, Jan Koster & Henk van Riemsdijk (eds). 〈http:// www.let.rug.nl/~koster/DenBesten/contents.htm〉. Bybee, Joan & Slobin, Dan. 1982. Why small children cannot change language on their own – evidence from the past tense. In Papers from the Fifth International Conference on Historical Linguistics, Galway, April 6–10 1981, [Current Issues in Linguistic Theory 21], Anders Ahlqvist (ed.), 29–37. Amsterdam: John Benjamins. Clahsen, Harald & Muysken, Pieter. 1989. The availability of universal grammar in adult and child learners: A study of the acquisition of German word order. Second Language Research 2: 93–119. Dietze, H. 1895. Das umschreibende Do in der neuenglischen Prosa. Ph.D. dissertation, Universität Jena. Ellegård, Alvar. 1953. The Auxiliary do: The Establishment and Regulation of its Use in English. [Gothenburg Studies in English]. Stockholm: Almqvist and Wiksell. Engblom, V. 1938. On the Origin and Early Development of the Auxiliary do [Lund studies in English 6]. Lund: C.W.K. Gleerup. Epstein, Samuel, Flynn, Suzanne & Martohardjono, Gita. 1996. Second language acquisition: Theoretical and experimental issues in contemporary research. Brain and Behavioural Sciences 19: 677–714. Fontaine, Carmen. 1985. Application de methodes quantitatives en diachronie: L’inversion du sujet en français. MA thesis, Université du Québec à Montréal. Henry, Allison. 1995. Belfast English and Standard English: Dialect Variation and Parameter Setting. Oxford: OUP. Hinterhölzl, R. 2004. Language change versus grammar change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, Eric Fuss & Carola Trips (eds), 131–160. Amsterdam: John Benjamins. Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199–244. Labov, William. 1972. Language in the Inner City: Studies in Black English Vernacular. Philadelphia PA: University of Pennsylvania Press. Lightfoot, David W. 1999. The Development of Language: Acquisition, Change, and Evolution. Oxford: Blackwell. Naro, Anthony J. 1984. Review of ‘Historical Linguistics’ by Barron Brainerd. Language 60: 929–932.
Gertjan Postma Noble, Shawn. 1985. To have and have got. Paper presented at NWAVE 14, Georgetown University. Oliveira e Silva, Giselle. 1982. Estudo da regularidade na variação dos possessivos no Português do Rio de Janeiro. Ph.D. dissertation, Universidade Federal do Rio de Janeiro. Pinker, Steven. 1984. Language Learnability and Language Development. Cambridge MA: Harvard University Press. Pintzuk, Susan. 2002. Verb–object order in Old English: Variation as grammatical competition. In Syntactic Effects of Morphological Change, David Lightfoot (ed.), 276–299. Oxford: OUP. Piotrovskaja, A.A., & Piotrovskij, R.G. 1974. Matematiceskie modeli diachronii i tekstoobrazovanija. In Statistika reci i avtomaticeskij analiz teksta, 361–400. Leningrad: Nauka. Postma, Gertjan. 2004. Structurele tendensen in de opkomst van het reflexief pronomen ‘zich’ in het 15de-eeuwse Drenthe en de Theorie van Reflexiviteit. Nederlandse Taalkunde 9: 144–168. Priestley, L. 1955. Reprise constructions in French. Archivum Linguisticum 7: 1–28. Reinhart, Tanya & Reuland, Eric J. 1993. Reflexivity. Linguistic Inquiry 24: 657–720. Sankoff, Gillian & Blondeau, Helene. 2007. Language change across the lifespan: /r/ in Montreal French. Language 83(3): 560–88. Santorini, Beatice. 1992. Variation and change in Yiddish subordinate clause word order. Natural Language and Linguistic Theory 10: 595–640. Trnka, B. 1930. On the Syntax of the English Verb from Caxton to Dryden [Traveaux du Cercle Linguistique de Prague 3]. Prague. Vainikka, Anna & Young-Scholten, Martha. 1994. Direct access to X′ structure: Evidence from Korean and Turkish adults learning German. In Language Acquisition Studies in Generative Grammar [Language Acquisition & Language Disorders 8], Teun Hoekstra & Bonnie Schwartz (eds), 265–316. Amsterdam: John Benjamins. Verhulst, Pierre-François. 1838. Notice sur la loi que la population poursuit dans son accroissement. Correspondance mathématique et physique 10: 113–121. Weerman, Fred. 1993. The diachronic consequences of first and second language acquisition: The change from OV to VO. Linguistics 31: 903–931. Weinreich, Uriel, Labov, William & Herzog, Marvin I. 1968. Empirical foundations for a theory of language change. In Directions for Historical Linguistics: A Symposium, Winifred P. Lehmann (ed.), 95–195. Austin TX: University of Texas Press. van der Wurff, Wim. 1990. Diffusion and Reanalysis in Syntax. Ph.D. dissertation, Leiden University.
A case of degrammaticalization in northern Swedish* Henrik Rosenkvist Lund University
1. Introduction In northern dialects of Sweden there are two verbs that are etymologically related to the Swedish verb behöva ‘need’, see (1). (1) Det behöver han inte göra. it need he not do ‘He doesn’t need to do that.’
(Swedish)
In the Piteå-dialect,1 it is, for example, possible to use all of the variants below ((2a–c); Brännström 1933: 122 presents similar examples from Nederluleå): (2) a. b. c.
Hä böhöv’n it djära. it need.hecl not do ‘He doesn’t need to do that.’ Hä bö’n it höv djära. it bö.hecl not höv do ‘He doesn’t need to do that.’ Hä bö’n it djära. it bö.hecl not do ‘He doesn’t need to do that.’
(Piteå)
(Piteå)
(Piteå)
In northern Swedish, the Swedish infinitive ending -a is in general lost after bisyllabic or “heavy” stems, and the present tense suffix -(e)r is absent in most northern Swedish dialects. The bare verb form böhöv, without suffixes, is thus not unexpected, and it may function as an infinitive as well as a verb in the present tense. As can be seen, the Swedish
*Two anonymous reviewers have commented on an earlier version of this chapter, for which I thank them. I also wish to express my gratitude to Ulf Teleman, who has suggested several improvements of the same earlier version and to David Willis, a very patient and careful editor. I alone am responsible for remaining faults. . I will mostly use data from Piteå throughout this chapter; I thank Mattias Johansson and his family for providing data, as well as a number of informants from other areas in northern Sweden.
Henrik Rosenkvist
behöva is in the Piteå dialect matched by böhöv, bö höv or just by bö. Below, I will use the term bö X höv-word order for those cases where items separate bö and höv as in (2b). This variation was mentioned by Noreen (1906: 7–8), but it was first discussed by Brännström (1933), who states that this phenomenon is present in all areas north of Ångermanland, with the exception for the northernmost districts (Kalix and Överkalix). Due to the general decline of genuine dialects2 in Sweden in recent times (Pettersson 2005: 201ff., Hallberg 2005: 1691ff.), there is at present a marked difference between generations (younger speakers do not use höv at all; see further below), but the speakers that do use the bö X höv-word order are still to be found in the same area as when Brännström performed his study. The approximate distribution of the bö X höv-word order (the northern and southern boundaries are better established than the inland spread) is depicted in Figure 1.
NORTHERN SWEDEN
Luleå Piteå
Kalix
Skellefteå
Umeå Ömsköldsvik Sundsvall
Figure 1. The approximate distribution of the bö X höv-word order
. A genuine dialect is used in a more restricted area than a regional dialect, and deviates more from the standard language (in this case, Swedish) (Pettersson 2005: 202). Hallberg (2005: 1692) uses the term rural local dialect.
A case of degrammaticalization in northern Swedish
The modal verb bö is, I argue, derived from böhöv, with bö X höv as an intermediate stage, as suggested by Brännström (1933). This language change thus appears to be a clear case of degrammaticalization, namely, prefix > auxiliary. In the present chapter I will first briefly introduce the notions of grammaticalization and degrammaticalization (Section 2). Then I discuss the origin and the detailed development of the dialectal varieties (in Section 3), describing the phonetic, semantic and syntactic aspects of the proposed change. It is concluded that bö is derived from böhöv, and that the change böhöv > bö indeed is an indisputable case of degrammaticalization.3 Section 4 sums up the discussion, and in Section 5 I make some final comments.
2. Grammaticalization and degrammaticalization – A brief overview 2.1 Two different notions of grammaticalization Due to a number of foundational works in the 1980s (e.g. Traugott 1980; Heine & Reh 1984; Lehmann 1985), the process of language change that gives rise to different kinds of grammatical markers has gained a prominent position on the linguistic agenda.4 Following Meillet (1912), this type of language change has been named grammaticalization. Since the 1980s, grammaticalization has been a hot topic within the field of historical linguistics: “During the past two decades, the revived interest in grammaticalization phenomena has firmly established itself as one of the most flourishing branches of historical linguistics.” (Fischer, Norde & Perridon 2004: 1). Several definitions of the notion have been proposed (Campbell & Janda 2001 present a compilation of various definitions). There are, however, two main approaches: the functional, which may be represented, for example, by Bybee, Perkins & Pagliuca (1994) and Heine & Kuteva (2002), and the formal, represented, for example, by Roberts & Roussou (2003) as well as van Gelderen (2004). Indeed, in his book about formal and functional perspectives in linguistics, Newmeyer (1998) spends an entire chapter discussing grammaticalization.5 The main differences between the functional and the formal camps with respect to the study of grammaticalization may be summarized as follows. First, the functionalists
. From a standard Swedish perspective, the bö X höv-construction actually appears so exotic that, when Westroth and Holm compiled Swedish dialectal data as a preparation for the Swedish Academy Grammar (Teleman, Hellberg & Anderson 1999), they chose not to include it in their work (Westroth & Holm 1987) (Lisa Christensen (née Holm), p.c.). . A thorough introduction to the history of grammaticalization is provided by Hopper & Traugott (2003: chapter 2). . It is reprinted in Language Sciences 23 (2001), an issue of the journal containing only articles criticizing grammaticalization theory.
Henrik Rosenkvist
concentrate on the semantic and pragmatic aspects of grammaticalization, whereas the formalists’ focus is syntactic change. Second, functionalists do not assume that there are clear-cut borders between grammatical categories, whereas formalists do; and finally, a majority of the functionalists claim that grammaticalization is a distinct process of language change that is unidirectional, gradual and partly determined by the source construction, and thus allows for reconstruction of earlier stages of language based on just synchronic data – that is, they are proponents of grammaticalization theory. The notion of grammaticalization theory and its applications is firmly rejected by linguists with a formal inclination (cf. e.g. Campbell 2001). However, in spite of these differences, some researchers, for instance, van Kemenade (1999), point out that the two approaches “complement rather than contradict each other in important ways” (van Kemenade 1999: 998; cf. also Rosenkvist 2005), and in this chapter no possible explanations will be excluded merely on theoretical grounds.
2.2 Two similar notions of degrammaticalization One often discussed point of grammaticalization theory is the assumed unidirectionality; why do there seem to be a number of language changes that proceed in the opposite direction and which hence seem to refute unidirectionality? A great number of such changes, i.e. degrammaticalizations, are presented by e.g. Campbell (2001: 127–8, 132).6 In a reply to this question, Haspelmath (2004) sets out to “examine a substantial number of alleged exceptions to unidriectionality and show that only a very few of them can be accepted as real exceptions” (2004: 17). In the end, Haspelmath recognizes eight cases of genuine degrammaticalizations (2004: 29; I return to these cases in Section 4), claiming that all other cases that have been proposed to constitute examples of degrammaticalization actually are not.,7,8 A genuine case of degrammaticalization is, according to Haspelmath, “a change that leads from the endpoint to the starting point of a potential grammaticalization
. As for the development of the Swedish modal verb må ‘may’, one of the most cited examples of purported degrammaticalizations (cf. van der Auwera & Plungian 1998: 105; Lehmann 2004: 177), it has recently been demonstrated that this verb actually did not change from grammatical to lexical meaning in the history of Swedish (Andersson 2007: Chapter 4). Må appears with lexical meaning already in the earliest texts. . Heine (2003), on the other hand, deconstructs the notion of degrammaticalization and concludes that “none of the processes underlying degrammaticalization […] clearly exhibits any directionality” (2003: 175) and that neither do these processes share any common denominator. Hence, Heine suggests, degrammaticalization has no explanatory or descriptive value. Cf. also Lehmann (2004: 170ff.). . Haspelmath (2004: 27) introduces the term antigrammaticalization for such changes that are true opposites of grammaticalization.
A case of degrammaticalization in northern Swedish
and also shows the same intermediate stages” (2004: 27–8). Obviously, Haspelmath’s definition depends on which notion of grammaticalization one is committed to; Haspelmath seems to take a more or less standard functionalist view of grammaticalization, as defined, for instance, by Heine & Reh (1984: 15): “With the term grammaticalization we refer essentially to an evolution whereby lingustic units lose in semantic complexity, pragmatic significance, syntactic freedom, and phonetic substance.” From a morphosyntactic perspective, which is particularly relevant where the emergence of bö is concerned, Haspelmath also appears to encompass the view of grammaticalization as encoded in the cline of grammaticality (Hopper & Traugott 2003: 7):9
(3) content item > grammatical word > clitic > affix
Also Norde (2006) starts out from this cline, in a recent contribution to the discussion of degrammaticalization. She assumes a more stringent position when defining degrammaticalization, however, suggesting that degrammaticalization involves “a SINGLE SHIFT to the left on the cline of grammaticality” (Norde 2006: 202ff.), and that it also must include more than one type of change; that is, for instance, a morphosyntactic change must coincide with a semantic change. (In Norde 2009 the discussion of degrammaticalization is elaborated further, but the main view is unaltered.) Both Haspelmath (2004) and Norde (2006) stress that the identity of a construction must be preserved in a grammaticalization (and, accordingly, also in a degrammaticalization): “grammaticalization changes modify the constructions they affect but preserve their identity” (Haspelmath 2004: 28). That is, grammaticalization never results in new constructions – this is, however, quite a vague concept, and Haspelmath clarifies his definition by stating that “a change […] is said to preserve its identity because it consists of numerous imperceptible changes with no radical break at any point” (2004: 28). So, although there are some minor differences between the definitions of degrammaticalization suggested by Haspelmath (2004) and Norde (2006), they agree on most points. Degrammaticalization is a construction-preserving change which operates at more than one domain of language, the direction of which is opposite to that of grammaticalization.
3. The emergence of the verb bö 3.1 The verb böhöv According to Brännström (1933), bö is derived from böhöv, which in turn is related to the Swedish verb behöva ‘need’. This verb was borrowed from Low German (behoven) . The cline itself encompasses different types of change, and its descriptive and explanatory values can be questioned. Nevertheless, Haspelmath (2004) and Norde (2006) use the cline as a point of departure in their discussions.
Henrik Rosenkvist
in the Middle Ages, and the prefix appears to have had the same phonetic form in Swedish since then (Hellquist 1980: 61). In northern Swedish dialects the verb böhöv is unique in three respects. First, it is the only verb where the prefix be- has changed into bö-. The cause of the change be- > bö- is probably assimilation to the vowel [ø:] in the root. Lindgren (1890–1918: 69) proposes that “be-höva has probably become bö- through assimilation of e with the following ö, hence a kind of ö-umlaut, so to speak” (my translation). Apart from böhöv, there are also a few other instances of unstressed [e] before [ø:] changing into [ø] in northern Swedish dialects,10 but, crucially, no other verbs with the be-prefix have changed in a similar way, as was stated above. The distribution of this sound change coincides with the distribution of the bö X höv-word order; north of the area where böhöv is found (in Kalix and Överkalix), be- has changed into bi- (Rutberg 1924–31, Pihl 1948: 26, 66) and south of this area, in Ångermanland, be- is preserved. This is an indication that the two linguistic changes are related to each other. Second, böhöv is also the only one of the verbs with this prefix (be-/bö-) that may express a modal meaning. It may, for instance, have the meanings ‘must’ or ‘ought to’ (Lidström et al. 1991: 26). (4) I böhöv djär’e. I böhöv do.itcl ‘I need/must/ought to do it.’
(Piteå)
interpretation 1: ‘I need to do it.’ interpretation 2: ‘I must do it.’ interpretation 3: ‘I ought to do it.’
The meanings of böhöv are discussed further in Section 3.3. The third factor is that the root -höv is a meaningless morpheme (a so-called cranberry morpheme) – it is not related to any other item in the language – whereas most other roots in the group of be-verbs have clear lexical meanings.11 One example is the verb be-söka (‘visit’; the meaning of söka is ‘search’). In the remainder of this section I will first discuss and criticize Brännström’s (1933) hypothesis concerning the origin of the bö, and then I will present the relevant semantic and syntactic circumstances in greater detail.
. The matter may be more complex; Lindgren (1890–1918) makes a point out of the fact that be- in the noun behov ‘need’ has not changed into bö-, which seems to support his umlauthypothesis. However, in Old Swedish the same word actually appears as böhoff in at least one manuscript (Söderwall 1884–1973 i.10). The crucial point for the present discussion is, however, that no other verb than böhöv had a bö-prefix in 19th-century northern Swedish. . I thank the CCG-audience for pointing this out.
A case of degrammaticalization in northern Swedish
3.2 Brännström’s (1933) hypothesis The first explanation of the emergence of the verb bö (as in (2c)), was issued by Brännström (1933: 121–2).12 He proposes that bö originated from the verb böhöv. When böhöv was used as a modal verb it was always unstressed, he claims, and, through sound change, a weak form (bö) eventually emerged, in the following pattern:13
(5) böhöv > böhö > böö > bö
Accordingly, we would expect a meaning difference separating böhöv from bö, and Lidström et al. (1991: 26) actually state that the verb böhöv only has a root meaning (‘need’), while bö has modal meanings (‘must’, ‘ought to’). Clearly, the short form seems to function solely as a modal verb in the Piteå dialect at present. However, in the Nederluleå dialect, spoken in an area adjacent to Piteå, Brännström (1933: 115) noted that bö and the longer form böhöv both appear to be used as modal verbs: (6) a.
No böhö jö gä häim. now böhö I go home ‘Now I need to/must go home.’
b. No böhövd’n fa smisk. surely böhöv-past.hecl get smack ‘Surely he ought to be smacked.’ (Nederluleå; Brännström 1933: 115)
At the time of Brännström’s investigations, both böhöv and bö were accordingly used as modal verbs in Nederluleå, and there was no functional division of the type suggested by Brännström. This is an argument against Brännström’s hypothesis (but not a very strong one, admittedly).
. The bö X höv-word order is also mentioned by Noreen (1906: 8–9), who suggests that the element intervening between bö and höv may be an infix. This hypothesis is clearly untenable. . An anonymous reviewer claims that the Swedish verb behöva is regularly contracted into [bø:] in colloquial speech, and that this contraction is involved in the bö X hövconstruction. However, there are two pieces of evidence that indicate that this is on the wrong track: first, it cannot explain the presence of höv in northern Swedish, and, second, it seems to be empirically incorrect. During the period 2005–8, Swedish dialect recordings from all over Sweden were transcribed in the project Swedish Dialect Syntax (under my supervision). The transcriptions now constitute a corpus with 222 informants and approximately 500 000 words. The transcriptions are linked to the recordings, and a random sample of the verb behöver (85 tokens from all over Sweden) seems to confirm that it is only in the northern Swedish region that behöver may be pronounced as [bø:] by some informants. A preliminary version of the corpus can be found via the ScanDiaSyn webpage: http://uit. no/scandiasyn/3517/.
Henrik Rosenkvist
Another argument against Brännström’s hypothesis is the fact that bö and höv may appear in separate positions in the clause, with one or more element between them. Bö and höv may, for instance, be separated by one or more adverbial (7a) or the subject and an adverbial (7b) – the word order actually indicates that bö is a finite verb, situated in the V2-position, while höv in these samples seems to appear in the slot for infinitival main verbs (Teleman, Hellberg & Andersson 1999: iv.14ff.). (7) a.
Han bö so jäddransh väl höv e he bö so darned well höv it ‘He really needs it.’
b. Bö dö it höv’e? bö you-sg not höv.itcl ‘Don’t you need it?’
(Norsjö; Brännström 1933: 120) (Piteå)
Brännström did of course observe this word order pattern, and explains it as a contamination between the two forms of the verb (böhöv and bö): the clause with separable compounding [i.e, bö X höv] may have emerged from the other in the following way: when the pronunciation had reached bö, this was associated and identified with the prefix in the more complete form böhöv and then this was completed with höv, inserted after the negation. (Brännström 1933: 122; my translation)
There are a number of problems with Brännström’s explanation for the emergence of the bö X höv-word order, however. First, it is supposed that a speaker first selects bö, and then regrets this choice and inserts a semantically and syntactically undefined höv. This element is furthermore inserted further down in the clause and not adjacent to bö, which would have been natural if höv had been interpreted as a missing verb root. Furthermore, other verbs beginning with bö- (such as böre ‘begin’) have never been subject to hövinsertion, which perhaps would have been expected given Brännström’s explanation. To conclude, Brännström’s explanation for the emergence of the verb bö is based on a Neogrammarian approach to language change. Since there appears to be more involved than mere sound change in this particular case, his explanation is unsatisfactory in several respects; the data contradict it, and he has to postulate höv-insertion, an otherwise completely unheard-of mechanism in order to explain the bö X höv-word order. Below, I therefore present a more coherent hypothesis concerning the emergence of bö, utilizing current theories of semantic and morphosyntactic change.14
. An anonymous reviewer suggests that the Swedish verb bör ‘should’ may have been involved in the emergence of bö. However, Brännström (1933: 123) explicitly rejects this hypothesis, showing that in the genuine northern Swedish dialect, Swedish bör corresponds
A case of degrammaticalization in northern Swedish
3.3 Semantic aspects of böhöv, bö X höv and bö As discussed above, the source for bö is the verb böhöv. Böhöv may express a lexical meaning (8a) as well as modal meanings (following the definition of modality suggested by van der Auwera & Plungian 1998: 80): (8) a.
I böhövd it öksa. I need-past not axe ‘I didn’t need the axe.’
‘need’
b. Kleda böhöv no twettes väsamma. ‘need’, root modal clothes-def need probably wash-pass immediately ‘The clothes must probably be washed immediately.’ c.
No böhövd’n fa smisk. surely böhöv-past.hecl get smack ‘Surely he ought to be smacked.’
d. Nöckeln böhöv it ha köme bort. key-def need not have come away ‘The key needn’t have been lost.’
‘ought to’, root modal. subject autonomous ‘needn’t’, epistemic modal
In (8b), böhöv is used as a modal auxiliary, expressing a possible obligation relating to the subject – this is therefore a root modal meaning. Also böhöv in (8c) is a root modal auxiliary, although in this case it is subject autonomous (cf. Abraham 2002: 36ff.); the subject may have another opinion than the speaker (and in this particular case, this is highly probable). Interestingly, behöva may also be used as an epistemic modal auxiliary (8d), although only in negative contexts – epistemic böhöv is hence a negative polarity item, just as Swedish behöva (Teleman et al. 1999: iv.290ff; cf. also van der Auwera 2001, who discusses Danish behøve, Dutch hoeven and German brauchen). In example (8d), böhöv is used to express the speaker’s belief that it is not probable that the key has been lost, and it is hence epistemic: “Epistemic expressions qualify the truth of the proposition” (Traugott & Dasher 2002: 16). So, böhöv had a number of meanings, and it was used both as a main verb (with lexical meaning) and as a modal verb. These functions and meanings are also available for the bö X höv-variant, but, crucially, the verb bö can only be used as a modal verb with root meaning; see Table 1.
to the verb borda or bola, and that possible occurrences of bör are late loans from standard Swedish. Furthermore, this hypothesis also fails to explain the bö X höv-construction.
Henrik Rosenkvist
Table 1. Meanings of böhöv, bö X höv and bö
böhöv bö X höv bö
Main verb (lexical)
Modal (root)
Modal (epistemic)
+ + –
+ + +
+ + –
Hence, a sentence such as (9a) (cf. (8a)), in which bö is used as a lexical verb is judged to be ungrammatical by speakers of northern Swedish, and epistemic interpretations of bö are also ungrammatical (9b) (cf. 8d): (9) a. *I bö it öksa. I bö not axe ‘I don’t need the axe.’ b. *Nöckeln bö it ha köme bort. key-def need not have come away ‘The key needn’t have been lost.’
To conclude, the meaning of bö indicates that this verb has lost the possibility of expressing lexical as well as epistemic meaning – it is clearly not more grammatical than the source construction böhöv or bö X höv. Rather, it has been semantically degrammaticalized.
3.4 The syntactic reanalyses of böhöv, bö X höv and bö Reanalysis is a well-known mechanism of language change (cf. Langacker 1977; Harris & Campbell 1995: Chapter 4 etc.) and it is plausible to assume that bö is the result of a segmental reanalysis (Rosenkvist 2004: 35–6). If so, the starting point must have been subject-initial clauses without clause adverbials. In such clauses, there is no material that indicates whether böhöv is a simple word or two separate words, and hence these clauses are syntactically ambiguous: böhöv can be analysed as a verb, or as an auxiliary bö plus the main verb höv (the analytical options are discussed further below). Furthermore, in some cases, at least three semantic interpretations of böhöv are possible in northern Swedish (as shown above). Thus, the sound change be- > bö- separated the verb böhöv from other verbs with the be-prefix, making it possible to dissociate bö- from the be-prefixes. Böhöv also developed a modal meaning (which in general is a feature of auxiliaries)15 and it is probable that böhöv appeared quite frequently in ambiguous clauses. Given these
. Eide (2002: chapter 5) discusses modal verbs in the Scandinavian languages, and their interaction with other categories.
A case of degrammaticalization in northern Swedish
phonetic, semantic and syntactic prerequisites, I propose that clauses such as I böhöv’e (‘I need it’) were reanalysed. The suggested reanalysis is illustrated in Figure 2. CP
CP
SpecCP I C0 böhöv′ei
IP… ti
SpecCP I C0
böi
IP… ti höv′e
Figure 2. Reanalysis from verb to free bö and remnant höv
A reanalysis such as the one in Figure 2 would yield a new syntactic option for the language user: bö is analysed as an auxiliary verb (the form of which fits well among auxiliaries such as ha ‘have’, få ‘get’, må ‘may’ and ska ‘shall’) and höv as a completely meaningless main verb which, considering that bö expresses the salient semantic features (cf. (2c)), must either be assigned some other function or replaced with another main verb such as, for instance, djära ‘do’. Overt elements in positions between C0 and höv, such as the subject in SpecIP or clause adverbials, will indicate that bö and höv are separate items – the bö X höv-word order is thus a result as well as an indicator of the reanalysis. The exact meaning of höv is actually hard to pin down. Brännström (1933: 123) notes that höv in clause-final position may give a clause “a strong emphatic, rebuffing quality” (my translation), but he does not mention what höv might have meant in other positions. Interestingly, it seems that older contemporary speakers are quite uncertain regarding the meaning of höv – some say that is has no meaning at all, and some that it “intensifies” the clause (as in (10b)). However, all speakers agree that höv cannot appear with any verb other than bö (11a–b): (10) a.
I bö it fåra no. I bö not leave now ‘I don’t need to leave now.’
b. I bö it höv fåra no. I bö not höv leave now ‘I REALLY don’t need to leave now.’ (11) a.
I nöges it fåra no. I must not go now ‘I don’t have to leave now.’
(Piteå)
(Piteå: some older speakers)
(Piteå)
b. *I nöges it höv fåra no.
So, we find some confusion regarding the exact meaning of höv, and the contrast in (11) indicates that höv is neither an independent verb, nor an independent adverbial.
Henrik Rosenkvist
Furthermore, younger speakers of northern Swedish16 are not even aware of the existence of höv – an indication that höv is disappearing from northern Swedish dialects rather rapidly. It is also notable that höv is not included in Lidström et al. (1991), which is an attempt to document the dialect of Piteå. This is what we would expect if bö emerged through a reanalysis, stranding höv. There is no reason why speakers should have stopped inserting höv due to contamination with böhöv (Brännström’s hypothesis), given that bö and böhöv are both still in use. This is another reason why it is difficult to maintain Brännström’s explanation for the bö X höv-word order. Hence, I find it reasonable to assume that bö actually emerged through reanalysis as illustrated in Figure 2, and that höv, once a verb root, accordingly became superfluous and eventually dropped out of use. Had höv regularly occurred in certain specific pragmatic contexts, it is plausible that, by pragmatic inferencing (Hopper & Traugott 2003: 7ff.), speakers might have assigned some specific meaning to höv, but this was apparently not the case. The former prefix bö has on the other hand taken over some of the meanings of böhöv and developed into a regular verb; it is as such inflected for tense in the same pattern as, for instance, the verbs ro ‘row’ and bo ‘live in’:17 (12) I bödd it fåra åt stan. I bö-past not go to city-def ‘I didn’t need/have to go to the city.’
(Piteå)
By abandoning Brännströms’s explanation – as I have shown, there are several arguments against it – it becomes necessary to assume that the prefix bö- has developed into a verb, leaving höv behind. In the following section I argue that this change constitutes an instance of degrammaticalization.
4. Is the emergence of bö a true case of degrammaticalization? In this section the changes of bö and höv will be discussed further, and I aim to show that the emergence of the verb bö is a genuine case of degrammaticalization, given the definitions quoted in Section 2.2. Schematically, the change from böhöv to bö and höv may be as illustrated in Figure 3. [böhöv] prefix+root
[bö] auxiliary
[höv] verb (?)
Figure 3. The split of böhöv
. I have interviewed four younger speakers (aged 22–26) from Skellefteå. . Swedish modal auxiliaries have preterite forms, unlike English auxiliaries.
A case of degrammaticalization in northern Swedish
A prefix (bö-) has changed into a regular free verb (bö), while höv has an unclear status at first, and at a later stage becomes obsolete. Given Haspelmath’s definition (cf. Section 2.2), the rise of bö was construction-preserving – the semantic changes occurred before the reanalysis, and the reanalysis itself occurred in syntactic contexts which were structurally ambiguous. Hence the reanalysis was imperceptible until it was revealed by the bö X höv-word order. This is of course a regular feature of reanalyses; as noted by Harris & Campbell (1995: 61), a syntactic reanalysis “does not involve any immediate or intrinsic modification of its surface manifestation”. In his list of eight undisputed attested degrammaticalizations, Haspelmath (2004: 29) includes no less than three changes of the type affix > independent word. In Irish, the suffix muid has change into a personal pronoun (Doyle 2002), in Saame the reconstructed suffix ptaken has become a postposition (Nevis 1986) and in Modern Greek the prefix ksana has turned into an adverb (Méndez Dozuna 1997). Bö- fits nicely in this group, being the fourth example of this kind of change. An important difference is however that bö- was a unique prefix, only occurring in one single verb, and thus the change did not affect an entire paradigm of prefixes (unless, of course, one claims that bö- constituted a paradigm in itself, albeit a very small one). Also höv emerged through this change, but it is not clear whether the emergence of höv is a case of degrammaticalization too. Consider the changes in the linguistic properties of bö and höv, respectively, with Heine & Reh’s definition of grammaticalization (1984: 5), quoted above, and Hopper & Traugott’s (2003: 7) cline of grammaticality as theoretical backgrounds (cf. Table 2). Table 2. Bö and höv degrammaticalization scores
bö höv
Semantic complexity
Pragmatic significance
+ – (?)
+ – (?)
Syntactic freedom Leftwards on cline + + (?)
+ ?
The former prefix bö- has subsumed some of the meanings of the verb böhöv, and has thus clearly gained in semantic complexity. It is also more significant from a pragmatic perspective (e.g. it can now be stressed), and it has furthermore become an independent word and is thus liberated from its former host, the verb root. Naturally, this implies that it also has moved leftwards on the cline of grammaticality – even more than one step (cf. Norde’s 2006 view of degrammaticalization). The change of höv is on the other hand considerably harder to pin down – it seems at present to have been abandoned by the community of speakers, left without proper meaning or function. Nothing of this kind happened in the similar degrammaticalizations mentioned by Haspelmath (2004: 29) – the reason is of course that böhöv had a unique prefix, and the split of bö and höv thus was a unique change, only affecting a single item in the language. For some older speakers, höv now appears to function
Henrik Rosenkvist
as an intensifying adverb, while others just reject it and find it meaningless. One may claim that it has gained in syntactic freedom, since the root has become a free word, but as has been shown above (11), höv is bound to bö and is not accepted in any other context. Since höv seems to be quite impossible to define syntactically and semantically, it follows that the change that gave rise to höv cannot be properly analysed in terms of grammaticalization or degrammaticalization. Interestingly, this change is merely the flip side of the change that brought about the new verb bö. If the explanation for the change of bö that has been presented here is accepted, it cannot really be disputed that the emergence of the verb bö is a true instance of degrammaticalization. But this change also resulted in a stranded höv, and from the perspective of höv it is not at all clear what kind of change we are dealing with. It can be concluded that this particular change assumes different guises depending on the stance chosen by the analyst.
5. Final comments In northern Swedish dialects the former prefix bö- has developed into a regular verb, which is inflected for tense just as other verbs. The starting point for the process of change was the verb böhöv, and I have argued that there were four essential prerequisites that facilitated the change. The first of these is the sound change that changed behöv into böhöv, thereby separating bö- from the group of commmon be-prefixes. This dissociation can be considered to be a case of what could be called de-analogy – in an analogous change an item is included in a new category/paradigm due to salient similarites with the elements that are understood as prototypical members of this category/ paradigm, whereas, in this case, bö- was recategorized since it did not share phonetic properties with any class of prefixes in the language. The second prerequisite was the change of meaning, whereby böhöv ‘need’ developed the modal meanings ‘ought to’ and ‘must’. A consequence of this change was that speakers could analyse bö as an auxiliary, and the former verb root höv as a main verb (or, possibly, as an adverbial). The third factor is that the root höv was a meaningless morpheme, unlike most other roots occurring in verbs with the prefix be-. The final factor is that böhöv probably appeared in structurally ambiguous contexts quite frequently, meaning that there were numerous opportunities for language users to perform the structural reanalysis described in Section 3.4. These phonetic, phonotactic, semantic and syntactic prerequisites thus formed a linguistic conspiracy which eventually led to the degrammaticalization of bö. Actually, such conspiracies seem to participate in degrammaticalizations quite often (cf. the cases discussed by Haspelmath 2004), and one possible solution to the problem of why degrammaticalizations are so infrequent may be that degrammaticalizations require not only one but several triggers for change, all pointing in the same direction.
A case of degrammaticalization in northern Swedish
The changes that led to the emergence of bö highlight the fact that language change may proceed in idiosyncratic and unpredictable patterns, and although a majority of grammatical language changes follow certain principles, it is never possible to state which grams have developed in a regular fashion and which have not, taking only synchronic facts about a certain language into account. For instance, in Rosenkvist (2004), the development of five Swedish conditional subordinators is studied in detail, and it is shown that the changes of at least two (om and ifall) contradict the predictions of grammaticalization theory (Heine & Kuteva 2002). Hence, it appears that reconstruction along the lines of baisc grammaticalization theory may yield a true picture of the diachronic facts, but need not necessarily do so. There is thus no way to determine actual paths of change in language history other than through detailed diachronic studies (cf. Hopper & Traugott 2003: 138–9). Accordingly, it must be pointed out that any discussion about the relation between böhöv and bö will be dependent on reconstruction, since there is no historical dialectal material in which the proposed change can be studied in detail, to the best of my knowledge. All caveats inherent in reconstruction as a method for historical linguistics must thus be considered (cf. Newmeyer 1998: 279ff). The validity of this specific reconstruction thus rests on factors such as general analytical coherence and plausibility in conjunction with detailed empirical data concerning the current and early-twentiethcentury usage of böhöv, bö X höv and bö. In comparison with, for instance, the results presented by Heine & Kuteva (2002: 6), whose work mainly rests upon information from grammar books, and who point out that “most of the over 400 grammaticalization paths discussed in this book are based on fairly reliable [sic!] reconstruction work”, the reconstruction of the emergence of the verb bö must be considered to be quite credible.
References Abraham, Werner, 2002. Modal verbs. Epistemics in German and English. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Fritz Beukema & Wim van der Wurff (eds), 19–50. Amsterdam: John Benjamins. Andersson, Peter. 2007. Modalitet och förändring. En studie av må och kunna i fornsvenska. Ph.D. dissertation, Dept. of Swedish, Gothenburg University. Auwera, Johan van der. 2001. On the typology of negative modals. In Perspectives on Negation and Polarity Items [Linguistik Aktuell/Linguistics Today 40], Jack Hoeksema, Hotze Rullmann, Víctor Sánchez-Valencia & Ton van der Wouden (eds), 23–48. Amsterdam: John Benjamins. Auwera, Johan van der & Plungian, Vladimir. 1998. Modality’s semantic map. Linguistic Typology 2: 79–124.
Henrik Rosenkvist Brännström, Edvin. 1933. En syntaktisk egendomlighet i norrländska dialekter. Nysvenska Studier 13: 112–127. Bybee, Joan, Perkins, Revere and Pagliuca, William. 1994. The Evolution of grammar. Chicago and London: University of Chicago Press. Campbell, Lyle. 2001. What’s wrong with grammaticalization? Language Sciences 23: 113–161. Campbell, Lyle and Janda, Richard. 2001. Introduction: Conceptions of grammaticalization and their problems. Language Sciences 23: 93–112. Doyle, Aidan. 2002. Yesterday’s affixes as today’s clitics. A case-study in degrammaticalization. In New Reflections on Grammaticalization [Typological Studies in Language 49], Ilse Wischer & Gabriela Diewald (eds), 67–81. Amsterdam: John Benjamins. Eide, Kristin. 2002. Norwegian Modals. Ph.D. dissertation, Norwegian University of Science and Technology, Trondheim. Fischer, Olga, Norde, Muriel & Perridon, Harry (eds). 2004. Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59]. Amsterdam: John Benjamins. Gelderen, Elly van. 2004. Grammaticalization as Economy [Linguistik Aktuell/Linguistics Today 71]. Amsterdam: John Benjamins. Hallberg, Göran 2005. Dialects and regional linguistic varieties in the 20th century I: Sweden and Finland. In The Nordic Languages 2, Oskar Bandle (main ed.), 1692–1706. Berlin: Walter de Gruyter. Harris, Alice C. and Campbell, Lyle. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin. 2004. On directionality in language change with particular reference to grammaticalization. In Up and Down the Cline: The Nature of Grammaticalization [Typological Studies in Language 59], Olga Fischer, Muriel Norde & Harry Peridon (eds), 17–44. Amsterdam: John Benjamins. Heine, Bernd. 2003. On degrammaticalization. Historical Linguistics 2001 [Current Issues in Linguistic Theory 237], Barry J. Blake & Kate Burridge (eds), 163–179. Amsterdam: John Benjamins. Heine, Bernd & Reh, Mechtild. 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Heine, Bernd & Kuteva, Tanya. 2002. World Lexicon of Grammaticalization. Cambridge: CUP. Hellquist, Elof. 1980. Svensk etymologisk ordbok. Lund: LiberLäromedel. Hopper, Paul J. & Traugott, Elizabeth. 2003. Grammaticalization, 2nd edn. Cambridge: CUP. Kemenade, Ans van. 1999. Functional categories, morphosyntactic change, grammaticalization. Linguistics 37: 997–1010. Langacker, Ronald W. 1977. Syntactic reanalysis. Mechanisms of Syntactic Change, Charles Li (ed.), 57–139. Austin: University of Texas Press. Lehmann, Christian. 1985. Grammaticalization: Synchronic variation and diachronic change. Lingua e stile 20: 303–318. Lehmann, Christian. 2004. Theory and method in grammaticalization. Zeitschrift für Germanistische Linguistik 32: 152–187. Lidström, Gun. 1991. Pitemålet – ållt mila aagg å ööx. ABF Piteåbygden. Lindgren, Johan Valfrid. 1890–1918. Burträskmålets grammatik. Svenska Landsmål 12. Stockholm. Meillet, A. 1912. L’ évolution des formes grammaticales. Scientia 12: 384–400. Méndez Dozuna, J. 1997. Fusion, fission and relevance in language change. De-univerbation in Greek morphology. Studies in Language 21: 577–612. Nevis, J. 1986. Decliticization in Old Estonian. In Studies on Language Change, Brian Joseph (ed.), 10–27. Columbus OH: Department of Linguistics, Ohio State University.
A case of degrammaticalization in northern Swedish
Newmeyer, Frederick. 1998. Language Form and Language Function. Cambridge MA: The MIT Press. Norde, Muriel. 2006. Demarcating degrammaticalization: The Swedish s-genitive revisited. Nordic Journal of Linguistics 29: 201–238. Norde, Muriel. 2009. Degrammaticalization. Oxford: OUP. Noreen, Adolf. 1906. Vårt språk VII. Lund: C.W.K. Gleerups Förlag. Pettersson, Gertrud. 2005. Svenska språket under sjuhundra år. Lund: Studentlitteratur. Pihl, Carin. 1948. Verben i Överkalixmålet. Skrifter utgivna genom Landsmåls- och Folkminnesarkivet i Uppsala A:5. Uppsala: Landsmåls- och Folkminnesarkivet. Roberts, Ian and Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Rosenkvist, Henrik. 2004. The Emergence of Conditional Subordinators in Swedish. Ph.D. dissertation, Lund University. Rosenkvist, Henrik. 2005. Modern Swedish bara – from adjective to conditional subordinator. In Historical Linguistics 2003, Michael Fortescue, Eva Skafte Jensen, Jens Erik Mogensen & Lene Schøsler (eds), 225–239. Amsterdam: John Benjamins. Rutberg, Hulda. 1924–1931. Folkmålet i Nederkalix ock Töre socknar. Svenska landsmål och svenskt folkliv B 28. Stockholm: Landsmåls- och folkminnesarkivet i Uppsala. Söderwall, Knut F. 1884–1973. Ordbok över svenska medeltidsspråket I–IV. Lund. Teleman, Ulf, Staffan Hellberg & Erik Andersson. 1999. Svenska Akademiens Grammatik I–IV. Stockholm: Svenska Akademien. Traugott, Elizabeth 1980. Meaning-change in the development of grammatical markers. Language Sciences 2: 44–61. Traugott, Elizabeth & Dasher, Richard B. 2002. Regularity in Semantic Change. Cambridge: CUP. Westroth, Helena & Lisa Holm. 1987. Förteckning över syntaktiska drag i regionalt talspråk. Ms, Lund University.
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages Renata Szczepaniak University of Hamburg
This chapter deals with the phonological aspect of Jespersen’s Cycle in German. This is the process whereby the older, Germanic negator ni was replaced by the more recent particle niht during the Middle High German period. At the same time, German changed from a syllable language to a word language. This typological shift was primarily a result of the replacement of the phonological syllable by the phonological word as the central prosodic domain. The tendency towards optimization of the phonological word involved many processes, including vowel reduction and deletion. Unstressed pretonic syllables such as the Middle High German preverbal negation marker ne/en were affected by this prosodic restructuring. I argue that this typological change was an additional factor that set Jespersen’s Cycle in motion in German.
1. Introduction Over the course of its history, German has clearly undergone Jespersen’s Cycle, as shown by Behaghel (1918), Donhauser (1996, 1998), Lenz (1996) and recently Jäger (2005, 2008), among others. This chapter deals with the phonological aspect of the renovation of the verbal negator. It will be shown that the typological shift towards a word language, initiated in Middle High German (MHG), was an additional factor that set the Jespersen’s Cycle in motion. The old, Germanic verbal negator ni was grad‑ ually replaced by the former intensifier ni(eo)wiht, lit. ‘not any thing’. This renovation took place in the MHG period, after the double negation phase: MHG ne/en + niht. In this period, constructions with niht as a single verbal negator began to appear. The replacement process was completed in the Early New High German (ENHG) period, in the southern German dialects earlier than in the northern ones (Pensel 1981). It has been observed by Behaghel (1918: 244–245) that the loss of the older, Germanic verbal negator in MHG occurred in certain morphological environments earlier than in others. It first disappeared before verbs with unstressed prefixes. Behaghel assumes that the stress structure was the crucial factor in this develop‑
Renata Szczepaniak
ment, because “[d]ie größte Masse der Vorsilben ist unbetont; es traten also durch die Verneinung zwei unbetonte Silben nebeneinander”.1 Yet it is not clear why this stress structure might have caused the loss of the negation particle in MHG. The adjacency was not bothersome in Old High German (OHG). In the following OHG example (1), two unstressed syllables, the negation particle and the unstressed verbal prefix be‑, occur before the stressed stem syllable: (1) mer ne-bedarf er more ne-need he ‘He doesn’t need more.’
(Notker Psalm 9 33, 12)
In New High German (NHG), there are still cases of two or more adjacent unstressed syllables, but only in the posttonic position, for example bitterere [‘bFt6r6r6] ‘bitter (fem.sg.comp)’. This chapter will answer the question of why this particular prosodic structure could have caused the loss of the negator. It takes into account the phonologicaltypological change of German from a syllable language (OHG) to a word language (MHG > ENHG > NHG). As will be shown, this shift involved a modification of German’s phonological word structure. The crucial moment in the evolution of the German Jespersen’s Cycle was the period of transition from OHG to MHG. At the same time, the typological shift of German began, bringing about a series of phono‑ logical changes optimizing the phonological word at the expense of the phonological syllable. The next section includes an outline of this typological change, with special attention to the phonological processes that played an important role in the renovation of the verbal negator: the MHG vowel reduction and subsequent vowel deletion. Both processes are highly relevant to the phonological structure of negated verbs. Section 3 discusses the phonological behaviour of the OHG and MHG forms of the negation particle, ni and ne/en, respectively. It will be shown that the phonological behaviour of this particle changed in MHG. In Section 4 the MHG negation is analysed from a phonological point of view, with particular attention to the morphological and phono‑ logical status of negated verbs.
2. Th e phonological change of German from a syllable language to a word language Over time, German has changed typologically from a syllable language to a word language (see Szczepaniak 2007a; Auer 2001). This shift, which began in Late OHG/
. “Most prefixes are unstressed. Thus, negation led to two adjacent unstressed syllables.”
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages
Early MHG, manifested itself in a series of phonological processes that made (pho‑ nological) words much more easily perceivable. It brought advantages for the listener, because it helped him decode the morphological structure of an utterance. Generally, the base of a phonological word is a morpheme, which can be phono‑ logically emphasized by means of word stress, vowel and consonant quality, number of syllables, among other strategies. The actual size of a phonological word is languagespecific (see Nespor & Vogel 1986). In NHG, phonological words (ω) can consist of simplexes or inflected words each containing a monomorphemic stem and subsequent grammatical morphemes, see Table 1 (a) and (b): Table 1. Phonological word structure in NHG (a) morph. simplexes (b) inflected words (c) derivatives with unstressed prefix (e.g. ver-, ge-) (d) derivatives with stressed affix (e.g. an-, -tum) (e) compound nouns
Blume Baum segel+te frisch+er+es verbinden Verband geduldig
['blu:m6]ω [ba~Ûm]ω ['ze:g6lt6]ω ['frFw6r6s]ω [f!'bFnd6n]ω [f!'bant]ω [g6'd~ldFç]ω
‘flower’ ‘tree’ ‘(s/he) sailed’ ‘fresher, n.sg.neutr.’ ‘to bind’ ‘bandage’ ‘patient’
an+rufen Reich+tum Ségel+bòot
[an]ω[rufen]ω [Reich]ω[tum]ω [Segel]ω[boot]ω
‘to call’ ‘wealth’ ‘sailboat’
Derivatives with unstressed prefixes (for example ver-, be-, ge-) constitute more complex and marked phonological words, in which the reduced syllable of the prefix precedes the stressed stem syllable, see Table 1 (c). Stressed prefixes and suffixes such as an-, auf-, ‑tum constitute separate phonological words, such that derivations like the following contain two phonological words: an+rufen [an]ω[rufen]ω ‘to call’ (see (d) in Table 1). Each morphological component bears either the main stress (′) or the secondary stress (`), for example ánrùfen, Réichtùm. The same structure is exhibited by compounds, for example Ségel+bòot [Segel]ω[boot]ω ‘sailboat’ (see Raffelsiefen 2000; Hall 1999; Nübling & Szczepaniak 2008, 2009). An important word-optimizing strategy introduced in MHG was emphasis of stressed (stem) syllables through reduction of unstressed vowels to schwa, for example OHG geban > MHG geben > NHG geb[6]n ‘give’. Another crucial strategy was regula‑ tion of the number of syllables within a phonological word. Short phonological words of equal length and stress structure support the decoding of morpheme boundaries. The MHG period saw the beginnings of the development of the trochaic (disyllabic, initially stressed) word as a prototype. Trochee is the predominant structure in the main word classes in contemporary German: nouns, verbs, and adjectives (see Eisenberg 1991). Verbs (and other derivatives) with unstressed prefixes like ver‑ in verbinden ‘to bind’ deviate from this prototype, as they have a reduced syllable at the left word edge (see Section 4).
Renata Szczepaniak
The earlier OHG period is characterized by phonological rules and processes typi‑ cal of syllable languages. Unlike in word languages, the phonological processes that occur in syllable languages are related to optimizing (phonological) syllables, thus ensuring easy articulation and bringing benefits to the speaker. The optimal, most easily pronounceable syllable contains one consonant (C) followed by one vowel (V) (Ven‑ nemann 1988). Most of the phonological processes in OHG clearly express a strong tendency to equalize syllable sequences where the ideal is CVCV… For example, h iatus positions were eliminated in OHG by means of vowel elision – see (2a). The resulting sequence was optimized, as it contained only syllables beginning with a consonant: zellu ih ‘I tell’ [CVC.CV.VC] > zelluh [CVC.CVC], ni ist [CV.VCC] > nist [CVCC]. (The point marks the syllable boundaries.) However, vowel elision and the subsequent resyllabi‑ fication blurred the word boundaries, which were no longer aligned with the syllable boundaries: [zellu]ω [ih]ω > [zel.luh]. Thus, phonological words were made less “visible”. A similar effect was brought about by vowel epenthesis, which broke up consonantal clusters – see (2b). Consonantal clusters are generally more difficult to articulate than single intervocalic consonants. This syllable optimization, which benefits the speaker, was achieved at the expense of the phonological word, which took on an unstable size and whose form varied between disyllabic (farwa) and trisyllabic (farawa). (2) a.
OHG vowel elision in hiatus position OHG zellu ih > zelluh ‘I tell’ OHG santa inan > santa-nan ‘sent him’ OHG ni ist > nist ‘is not’
b.
OHG vowel epenthesis OHG wurm > wurum ‘worm’ OHG farwa > farawa ‘color’
In OHG there were many other phonological processes that benefited the speaker by making word strings (or syllable sequences) more easily pronounceable. The same effect resulted from the OHG umlaut, a regressive palatalization of the stem vowel, for example OHG farit > ferit > NHG fährt ‘(s/he) drives’. Just like the vowel harmony of the middle vowel (OHG wuntar ‘wonder’, wuntoro ‘gen.pl.’) or epenthetic vowel (OHG wurm > wurum ‘wurm’ but berg > bereg ‘mountain’) – largely untreated in the literature – umlaut made adjacent syllables more similar to each other (Szczepaniak 2007a, b). As early as Late OHG/MHG, the typological tendency changed to the listener’s advantage. The above mentioned vowel reduction and deletion, as well as a series of other processes emphasizing the phonological word, led to the deterioration of syllable structure, which now differs depending on its position within a word, for example by means of reducing unstressed syllables to schwa syllables in MHG. Other processes include word-final consonantal epenthesis in ENHG, which marked the word end, for
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages
example MHG mane > (E)NHG Mond ‘moon’ or MHG saf > (E)NHG saft ‘juice’ (cf. engl. sap which does not have a final epenthetic consonant), and word-licensed conso‑ nantal changes, for example the word medial h deletion in MHG sehen > NHG sehen ['ze:.6n], which produced a new hiatus (for more details see Szczepaniak 2007a). To summarize, the crucial moment in the typological shift from OHG as a syllable language to NHG as a word language was the MHG period. It was exactly at this point in time that the replacement of the original verbal negator with the new one took place (see Table 2). The phonological status and development of the original negator ni will be analysed in the next section. Table 2. Phonological and syntactic (negation) change in MHG OHG
MHG
phonological type
syllable language
word language
tendency to
easy articulation (to the speaker’s favour)
easy decoding of morphological structure (to the listener’s favour)
phonological processes
– avoiding hiatus through vowel elision – avoiding consonantal clusters through vowel epenthesis – keeping syllable quality equal (through umlaut, vowel harmony, vowel quality etc.) – resyllabification
– marking word structure through vowel reduction and deletion, – marking word boundaries through initial stressed syllable, glottal stop, final consonant epenthesis – blocking resyllabification beyond word boundaries
syntactic change (negation)
preverbal particle ni (single negation)
double negation and subsequent formal renovation of the negation marker niht
3. Jespersen’s Cycle in German from the phonological perspective In this section, the phonological behaviour of the negation particle in OHG and in MHG will be discussed. A corpus analysis of OHG and MHG texts was conducted for this purpose. Special attention was paid to the OHG work of Otfrid (Evangelienharmonie, Ms D) and Notker as well as that of Iwein (Ms A) and the Nibelungenlied (Ms C) in MHG. The online version of Titus was used to find negation contexts. However, proper analysis of MHG negation was performed on the basis of the respective manuscripts.
3.1 Old High German ni The OHG negation element ni is usually said to be a verbal clitic (Jäger 2005, 2008). Braune & Reiffenstein (2004: 73) discuss ni in the chapter dealing with vowels in
Renata Szczepaniak
unstressed (verbal) prefixes such as gi‑/ga‑, bi‑, ant- etc. This also suggests that the OHG negation particle had a clitic, unstressed status. However, it is worth mention‑ ing that unlike other prefixes (Valentin 1978; Becker 2000), ni maintained a stable full i‑vowel during the OHG period. Only in the final stages of OHG can a graphematic reflex of a gradual vowel reduction be observed, for example daz du commen ne hebist ‘that you husband no have’ (Christus und die Samariterin, mid‑10th c., 25), mainly in the writings of Notker and Williram (10th/11th c.). In the OHG records negation with a full vowel was practically never written as a free graphical word, but usually as part of one word, attached to a verb, see (3a), or even a member of a different word class as in (3b). ni mostly occurred at the begin‑ ning of one of these graphical units and less frequently within it, (3c). All examples in (3) are taken from Otfrid’s Evangelienharmonie (Ms D, Wolfenbüttel Cod. Guelf. 131.1 Extrav.). (3) a.
nigangent ni.go.3pl ‘they don’t go’
(Otfrid Evangelienharmonie II 22, 10; fol. 68r; 00011)2
b.
nisiesih ginerien ni.they.refl subsist ‘not in order to subsist’
(Otfrid Evangelienharmonie II 22, 12; fol. 68r; 00011)
c.
auuarsoniduat but that way ni.do.3sg ‘but (who) doesn’t act this way’
(Otfrid Evangelienharmonie II 21, 43; fol. 67v; 00010)
The graphical behaviour of ni corresponds to the (so far unsystematic) findings about the OHG writing system. The OHG graphical word (or unit) could contain a whole prepositional phrase, for example inhimile ‘in heaven’ (Paternoster, 8th c.), or even larger syntactic units as in (3c). At the same time, words could be graphically divided – sometimes within a line – at syllable boundaries, which did not necessarily correspond to morpheme boundaries, for example 〈ent fan gan〉 ‘received’ (see Frey 1988; Voetz 2006). This tendency to write as one word anything that comes between two pauses or to divide a written word into syllables is quite strong in the OHG documents and correlates with the status of OHG as a syllable language. In OHG, syllables were of comparable quality. Their structure did not depend on their position within the word. Thus, the word was not phonologically emphasized and hence not easily perceivable. Much more perceivable were the edges of larger phonological constituents, such as
. The last number refers to the pagination of the digitized manuscript (Wolfenbüttel Cod. Guelf. 131.1 Extrav.).
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages
phonological or, even, intonational phrases. They were signalled by means of pauses (see Nespor & Vogel 1986). A phonological phrase roughly corresponds to a syntactic phrase of the type noun/verb/adjective or a prepositional phrase (see Wiese 1996). This prosodic unit plays an important role in syllable languages, serving as the domain of phonological processes such as resyllabification beyond word boundaries (see also Kleinhenz 1996). It appears that OHG scribes, although well versed in the Latin sys‑ tem, which is based on the concept of graphical words, often wrote whole phonologi‑ cal phrases in one word, presumably because of the problems they faced in “correctly” detecting word boundaries. If that was the case, the negation particle can be analysed as having had the status of a free (phonological) word occurring at the beginning of or within a phonological phrase (φ), for example [[ni]ω[gangent]ω]φ. However, OHG can be considered a period in which the originally free negation word was on its way to losing this status and gradually turning into a clitic (the 1st stage of Jespersen’s Cycle). At this time, two variants of ni coexisted: a stressed and an unstressed form (Nübling 1992: 133). The evidence for the stressed variant comes from meter, for example nísiesih ginerien ‘not in order to subsist’ (O II 22, 12; fol. 68r). The formal reduction of the unstressed clitic ni was limited to a hiatus position, in which ni preceded a vowel-initial word (usually a verb). Only in such contexts could the vowel be dropped. The vowel elision took place only before the (un)stressed i of verbal prefixes ir‑, in‑, and int‑ (see de Boor 1928: 86). The vowel i was also lost before members of other word classes (conjunctions, adverbs, and indefinites), for example ni oba > noba ‘but, unless’, *ni alles > nalles ‘not at all’. Generally, the deletion led to optimization of the syllable structure by resolving the hiatus, as is typical of syllable languages:
(4) [ni]ω[intfahan]ω > (vowel elision and resyllabification) [nintfahan]ω
As early as Late OHG, the preverbal negator could be assumed to have an unstressed, reduced syllable status. However, vowel deletion was still limited to the hiatus position: (5) a.
with vowel deletion: nersterbint (< *ne ersterbint) ne.kill.3pl ‘they don’t kill’
b.
without vowel deletion: nefindest ne.find.2sg ‘you don’t find’
The OHG ni underwent the following phonological stages:
(Notker Psalm 29 88, 22)
(Notker Psalm 36 119, 24)
Renata Szczepaniak
Table 3. The phonological development of OHG ni 1st stage (Pre‑)OHG
2nd stage OHG
3rd stage Late OHG
free (phonological) word
unstressed clitic with full vowel (vowel elision in hiatus possible) [nintfahan]ω
unstressed clitic with reduced vowel
[ni]ω[gangan]ω
[ne'findest]ω [ner'sterbint]ω < *ne ersterbint
3.2 Middle High German ne/en In MHG, the situation gradually changed. The process of vowel reduction, which began in the Late OHG period, was completed, as can be deduced from graphical forms. The vowel grapheme 〈e〉, which has also been used for posttonic reduced vow‑ els since MHG, appears very regularly in the manuscripts. Vowel reduction provides a reliable proof of the unstressed status of ne. In this section, the occurrence of the old negator in Iwein (Ms A, CPg 397, 2nd half of 13th c.) and Nibelungenlied (Ms C, Cod. Don. 63, 1st half of 13th c.) will be analysed. The manuscripts were examined in addi‑ tion to the printed text editions and electronic versions of both texts, available in the Titus online text corpus. This way, parts of the normalized editions that deviate from the MHG original could be detected and eliminated.
3.2.1 Iwein (Low Franconian) In Iwein the preverbal negator ne is generally maintained and spelled ne and written as part of a word as in (6a). Unlike in the seventh edition by Benecke, Lachmann & Wolff (1986), ne sometimes also occurs as a free graphical word as in (6b). Generally, the graphical word boundaries are not always easy to determine. (6) a.
donebestont there.ne.stayed.3sg ‘no one has stayed there’
(Iwein 7733, fol. 36v)
b.
wān man ne geloubete me niht because one ne believed.3sg me niht ‘because no one would have believed him’
(Iwein 1730, fol. 21r)
For the purpose of this chapter, special attention was paid to prefixed verbs. In this manuscript ne is mostly written as part of one word including the following prefixed verb (37 occurrences) or even as a free word (29 occurrences). Only in four cases can a clear enclisis be observed. Table 4. Graphical status of the verbal negator ne in Iwein (Ms A) Proclitic
Free graphical word
Enclitic
37 cases
29 cases
4 cases
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages
In 22 cases, the negation marker ne is absent before a prefixed verb in the manu‑ script, although it appears in the normalized edition, for example daz sie sin niht uir drieze ‘that she won’t be weary of it’ (Iwein 2875, fol. 50r). In all but two cases, an unstressed prefix (here: uir‑, NHG ver-) is involved, mostly in subordinate clauses. Additionally, the ne-negated verb occurs 25 times without prefix (unlike in the nor‑ malized edition).
3.2.2 The Nibelungenlied (Bavarian) The tendency to avoid ne or en as a preverbal negator before a verb with an unstressed prefix is much stronger in the Nibelungenlied (C). There are only six cases in which a verb with an unstressed prefix is negated with ne as in (7). (7) sine gesach in she.ne saw him ‘she did not see him’
(Nibelungenlied C 933, 4; fol. 35v)
The preceding word, which in all but one case is the host of the enclitic negator, is always monosyllabic, for example si in (7). In the following section a phonological analysis of this phenomenon will be presented.
4. Th e gradual loss of the preverbal negation particle and the phonological structure of negated words The phonological restructuring of the word that took place in MHG also included foot structure. Within a (phonological) foot (F), syllables are arranged according to their relative prominence (stress structure), with the prominent syllable forming the foot head, for example NHG Segel ‘sail’ [[ze:.g6l]F]ω. As a result of the MHG vowel reduction (and deletion), the number of feet within a word decreased to only one (for details see Szczepaniak 2007a). This is still a constitutive characteristic of the native vocabulary of Modern German. As in MHG, the monopedal structure of simplexes as well as of inflected words is additionally expressed through vowel quality (full stressed and reduced unstressed vowels). Thus, the prototypical phonological word in German starts with a stressed full vowel, contains only one phonological foot with an unstressed reduced vowel, and is disyllabic. This prototypical structure is very regu‑ larly found in native German nouns which have monosyllabic or (at most) disyllabic stems and always create disyllabic (trochaic) plural forms, for example Frau – Frau+en, Blume – Blume+n, Wurzel – Wurzel+n. Verbs and adjectives can also exhibit trisyllabic and (seldom) quadrisyllabic forms, for example segelte ‘(s/he) sailed’, muntererer ‘more animated (masc., sg.)’ (see Eisenberg 1991). The more syllables a monopedal foot con‑ tains, the further away it is from the trochaic prototype.
Renata Szczepaniak
Verbs with unstressed prefixes deviate from this prototype very strongly. The first, reduced syllable (σr) cannot form its own foot (or even a phonological word), because it is unstressed, and is thus only incorporated in the phonological word of the verb: [〈σr〉[σ(σ)]F]ω, for example MHG gediente [〈g6〉[dient6]F]ω ‘deserved’ (see (a) in Figure 1). The unstressed MHG negation particle, when attached as a proclitic to such a verb, increased the number of unstressed syllables and deteriorated the phonological word structure: [〈σr〉〈σr〉[σσ]F]ω, for example ne gediente [〈n6〉〈g6〉[dient6]F]ω ‘did not deserve’, see Figure 1 (b). (a)
ω
(b)
ω
(c)
ω
F
F
F
σr σr σw
σr σr σs
σw
g6 dien t6
n6 g6 dien t6
σr σr σs
σw
n6 g6 dien t6
Figure 1. Prosodic structure of negated verbs with unstressed prefixes
With the increasing relevance of the word as the phonological category, ill-formed pho‑ nological words (such as this in Figure 1(b)) were eliminated through the dropping of the negation particle in double negation contexts – see (c) in Figure 1. The corpus analysis (see Section 3.2 above) shows that this tendency is weak in Iwein, whereas in the Nibelungenlied, ne is lacking almost completely. Hence, the phonological tendency to optimize the phonological word structure led to the loss of the verbal negator ne before verbs with unstressed prefix. In the intermediate state, forms without a vowel (n-) and forms that are vowelinitial (en‑) arose. They demonstrate the strong tendency to suppress initial unstressed vowels. Unlike in OHG, in MHG clisis and vowel reduction was also possible before consonants and led to complex and less optimal syllable structures, for example nebot > nbot > enbot ‘don’t allow’ (Nibelungenlied (Ms C) 1183, 3, fol. 45r; cf. manuscripts A and B; see Nübling 1992: 136; Paul 1998: 37). In MHG phonological word structure increased in importance. Here, the tendency to begin the phonological word with a stressed syllable led to the vowel reduction ne > n and deteriorated the syllable struc‑ ture. The negator consisted only of the syllabic n [nŸ ]. The subsequent epenthesis of a weak e (Paul 1998: 37) was the “last effort” to maintain the negator and to make the consonant cluster pronounceable: [nŸ .bot] > [6n.bot]. Interestingly, the preverbal negator also underwent enclisis in MHG. It attached mostly to monosyllabic, less stressed words such as pronouns, adverbs and conjunc‑ tions, for example dune, jane, sone, producing trochaic structures. In this position its form was unstable because of vowel reduction, for example dune > dun ‘you not’. Also, the morphological fact that in such cases the negator ne was attached as an enclitic to diverse word classes may have been responsible for its successive loss. Conjunctions
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages
or personal pronouns are not “ideal” hosts for clitic or inflectional sentence negation (see Dahl 1979).
5. Summary The phonological shift of German from a syllable to a word language can be seen as an additional factor that set Jespersen’s Cycle in motion. The loss of the original OHG negator ni (MHG ne/en) continued during the phonological-typological drift of German towards a word language. The tendency to optimize the phonological word led to a loss of the preverbal negation clitic, which occupied the initial position within a phonologi‑ cal word. The negator was first lost before verbs with unstressed prefixes, whose struc‑ ture strongly deviates from the new prototype of syllabic trochee (xʹ x). In such cases, the unstressed negation syllable further deteriorated the structure of such verbs as verbinden ‘to bind’ (xxʹ x) by producing xxxʹ x structures. Instead, the enclitic negator created a tro‑ chaic structure with a monosyllabic host. Its loss was supported by morphological fac‑ tors, because its hosts belonged to diverse word classes such as conjunctions, adverbs, or personal pronouns, which were not directly related to sentence negation.
References Primary sources Das Nibelungenlied. Paralleldruck der Handschriften A, B und C nebst Lesarten der übrigen Handschriften. Michael S. Batts (ed.). 1971. Tübingen: Niemeyer. Titus entry by B. Karner and H. Reichert. Titus version by Jost Gippert, Frankfurt/M. 1998–2003. Das Nibelungenlied. MS C. [=N]. Hartmann von Aue: Iwein. G.F. Benecke, K. Lachmann and L. Wolff (eds). 1986. Berlin: de Gruyter. Titus entry by T. Parchert. Titus version by Jost Gippert, Bayreuth, Frankfurt/M. 1999–2003. Hartmann von Aue: Iwein. MS A. Cpg 397. Notker der Deutsche. Werke. Boethius. De consolatione Philosopiæ. Petrus W. Tax (ed.). 1988. Tübingen: Niemeyer. Titus entry electronically prepared under the direction of R. Lühr. Titus version by Jost Gippert, Frankfurt/M. 1997–2003. Boethius. Aristotelis Categoriae. Titus entry electronically prepared under the direction of R. Lühr. Titus version by Jost Gippert, Frankfurt/M. 1997–2003. Boethius. Aristotelis De interpretatione. James King (ed.). 1975. Tübingen: Niemeyer. Titus entry electronically prepared under the direction of R. Lühr. Titus version by Jost Gippert, Frankfurt/M. 1997–2003.
Renata Szczepaniak Martianus Capella. De nuptiis Philologiae et Mercurii. James King (ed.). 1979. Tübingen: Niemeyer. Titus entry electronically prepared under the direction of R. Lühr. Titus version by Jost Gippert, Frankfurt/M. 1997–2003. Psalterium. Psalm 1–50. Petrus W. Tax (ed.). Tübingen: Niemeyer. 1979. Titus entry electroni‑ cally prepared under the direction of R. Lühr. Titus version by Jost Gippert, Frankfurt/M. 1997–2003. Otfrid von Weissenburg: Evangelienbuch. Oscar Erdmann (ed.). 1973. Tübingen: Niemeyer. Titus text entry by R. Schuhmann, M. Bayer et al. Titus version by Jost Gippert, Frankfurt/M. 1997–2003. Otfrid von Weissenburg: Evangelienbuch. Wolfgang Kleiber (ed.). Vol. 1. 2004. Tübingen: Niemeyer. [=O]. Otfrid von Weissenburg: Evangelienbuch. Wolfenbüttel Cod. Guelf. 131.1 Extrav. Paternoster. Cod. Sang. 911, 320.
Secondary references Auer, Peter. 2001. Silben- und akzentzählende Sprachen. In Language Typology and Language Universals. An International Handbook, Vol. 2.2 [Handbücher zur Sprach- und Kommu‑ nikationswissenschaft 20] Martin Haspelmath, Ekkehard König, Wulf Oesterreicher and Wolfgang Raible (eds), 1391–1399. Berlin: de Gruyter. Becker, Thomas. 2000. Zur Vokalreduktion im Althochdeutschen. In Angemessene Strukturen. Systemorganisation in Phonologie, Morphologie und Syntax, Andreas Bittner, Dagmar Bittner & Klaus-Michael Köpcke (eds), 31–46. Hildesheim: Olms. Behaghel, Otto. 1918. Die Verneinung in den deutschen Sprachen. In Wissenschaftliche Beihefte zur Zeitschrift des allgemeinen deutschen Sprachvereins 5(38/40), 225–252. Boor, Helmut de. 1928. Untersuchungen zur Sprachbehandlung Otfrids. Hiatus und Synalöphe [Germanistische Abhandlungen 60]. Breslau: Marcus. Braune, Wilhelm & Reiffenstein, Ingo. 2004. Althochdeutsche Grammatik I: Laut- und Formenlehre, 15th edn, ed. by Ingo Reiffenstein [Sammlung kurzer Grammatiken germanischer Dialekte A, Hauptreihe 5.1]. Tübingen: Niemeyer. Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79–106. Donhauser, Karin. 1996. Negationssyntax in der deutschen Sprachgeschichte. Grammatika‑ lisierung oder Degrammatikalisierung? In Deutsch – typologisch [Jahrbuch/Institut für Deutsche Sprache 1995], Ewald Lang & Gisella Zifonun (eds), 201–217. Berlin: de Gruyter. Donhauser, Karin. 1998. Negationssyntax im Althochdeutschen. Ein sprachhistorisches Rätsel und der Weg zu seiner Lösung. In Deutsche Grammatik. Thema in Variationen. Festschrift für Hans-Werner Eroms zum 60. Geburtstag, Karin Donhauser & Ludwig M. Eichinger (eds), 283–298. Heidelberg: Universitätsverlag Winter. Eisenberg, Peter. 1991. Syllabische Struktur und Wortakzent. Prinzipien der Prosodik deutscher Wörter. Zeitschrift für Sprachwissenschaft 10(1): 37–84. Frey, Evelyn. 1988. Wortteilung und Silbenstruktur im Althochdeutschen. München: CD Copy & Druck. Hall, Tracy Alan. 1999. The phonological word. A review. In Studies on the Phonological Word, [Amsterdam Studies in the Theory and History of Linguistic Science. Series 4. Current
Jespersen’s Cycle in German from the phonological perspective of syllable and word languages Issues in Linguistic Theory 174], Tracy Alan Hall & Ursula Kleinhenz (eds), 1–22. Amsterdam: John Benjamins. Jäger, Agnes. 2005. Negation in Old High German. Zeitschrift für Sprachwissenschaft 24: 227–262. Jäger, Agnes. 2008. History of German Negation [Linguistik Aktuell/Linguistics Today 118]. Amsterdam: John Benjamins. Kleinhenz, Ursula. 1996. Zur Typologie phonologischer Domänen. In Deutsch – typologisch [Jahrbuch/Institut für Deutsche Sprache 1995], Ewald Lang & Gisella Zifonun (eds), 569–584. Berlin: de Gruyter. Lenz, Barbara. 1996. Negationsverstärkung und Jespersens Zyklus im Deutschen und in anderen europäischen Sprachen. In Deutsch – typologisch [Jahrbuch/Institut für Deutsche Sprache 1995], Ewald Lang & Gisella Zifonun (eds), 183–200. Berlin: de Gruyter. Nespor, Marina and Irene Vogel. 1986. Prosodic Phonology [Studies in Generative Grammar 28]. Dordrecht: Foris. Nübling, Damaris. 1992. Klitika im Deutschen. Schriftsprache, Umgangssprache, alemannische Dialekte [ScriptOralia 42]. Tübingen: Narr. Nübling, Damaris & Szczepaniak, Renata. 2008. On the way from morphology to phonology. German linking elements and the role of the phonological word. Morphology 18: 1–25. Nübling, Damaris & Szczepaniak, Renata. 2009. Religion+s+freiheit, Stabilität+s+pakt und Subjekt(+s+)pronomen: Fugenelemente als Marker phonologischer Wortgrenzen. In Studien zur Fremdwortbildung [Germanistische Linguistik 197–198], Peter O. Müller (ed.), 195–222. Hildesheim: Olms. Paul, Hermann. 1998. Mittelhochdeutsche Grammatik [Sammlung kurzer Grammatiken ger‑ manischer Dialekte A, Hauptreihe 2], 24th edn, ed. by Peter Wiehl & Siegfried Grosse. Tübingen: Niemeyer. Pensel, Franzjosef. 1981. Die Satznegation. In Zur Ausbildung der Norm der deutschen Literatursprache auf der syntaktischen Ebene (1470–1730). I. Der Einfachsatz [Bausteine zur Sprachgeschichte des Neuhochdeutschen 56], Gerhard Kettmann & Joachim Schildt (eds), 285–326. Berlin: Akademie-Verlag. Raffelsiefen, Renate. 2000. Evidence for word-internal phonological words in German. In Deutsche Grammatik in Theorie und Praxis, Rolf Thieroff, Mathias Tamrat, Nanna Fuhrhop & Oliver Teuber (eds), 43–56. Tübingen: Niemeyer. Szczepaniak, Renata. 2007a. Der phonologisch-typologische Wandel des Deutschen von einer Silben- zu einer Wortsprache [Studia Linguistica Germanica 85]. Berlin: de Gruyter. Szczepaniak, Renata. 2007b. Vokalharmonie im Althochdeutschen und im Walserdeutschen – ein Vergleich. Zeitschrift für Dialektologie und Linguistik 74(1): 38–60. Valentin, Paul. 1978. The simplification of the unstressed vowel systems in Old High German. In Recent Developments in Historical Phonology [Trends in Linguistics. Studies and Mono‑ graphs 4], Jacek Fisiak (ed.), 373–389. The Hague: Mouton. Vennemann, Theo. 1988. Preference Laws for Syllable Structure and the Explanation of Sound Change. With Special Reference to German, Germanic, Italian, and Latin. Berlin: de Gruyter. Voetz, Lothar. 2006. Einige Beobachtungen zur Getrennt- und Zusammenschreibung im Althochdeutschen. In Neue Perspektiven der Sprachgeschichte [Germanistische Bibliothek 26], Ursula Götz & Stefanie Stricker (eds), 51–64. Heidelberg: Winter. Wiese, Richard. 1996. The Phonology of German [The Phonology of the World’s Languages]. Oxford: Clarendon Press.
An article on the rise Contact-induced change and the rise and fall of N-to-D movement Mila Dimitrova-Vulchanova & Valentin Vulchanov
Norwegian University of Science and Technology, Trondheim/KVAB-VLAC In this chapter we take a somewhat unconventional perspective and address the evolution of the article as attested in early South Slavic (10th–11th c.) in the light of contact with New Testament Greek, which was the source language for the early translations in a situation of active language interaction (bilingualism). Our analysis builds on comprehensive corpus data which suggest, counter to traditional views, that, at this stage, the article is already a functional category inside the nominal expression. We propose that this is the result of a reanalysis (Harris & Campbell 1995; Roberts 2007) of the item at hand of the type maximal (phrasal) constituent > head (SpecDP > D0), or in terms of grammatical mechanism move > merge (Roberts & Roussou 2003), and show that in the structure of the nominal phrase at this stage the article and the demonstrative are already distinct syntactic categories, despite the complete phonological/ morphological overlap. Our data suggest further that the process of the evolution of the article is in direct correlation with the loss of N-to-D0 movement as a way of signaling definiteness in the DP.
1. Introduction In this chapter we address the evolution of the definite article in Old Bulgarian1 using extensive corpus data and discuss its consequences for the structure of the DP. On the basis of our data we conclude that, at this stage of the language, definiteness is marked by two alternative means, full movement of the N-head all the way to D0 or insertion of
. Several terms are used by scholars to refer to the language of the earliest Slavic texts. They reflect different aspects of this early literary language. Old Church Slavonic (OCS) reflects the primarily ecclesiastical nature of these texts, while Old Slavic focuses on the functional aspect of the language as a common literary language of the Orthodox Slavic community. Old Bulgarian reflects the language-proper basis of these early texts. In this chapter we use the latter term due to the linguistic nature of our analysis.
Mila Dimitrova-Vulchanova & Valentin Vulchanov
a determiner clitic in D0. We suggest that there are two diachronic processes at work, a parameter shift and a reanalysis of the type Move > Merge in the sense of Roberts & Roussou (2003) involving the determiner clitic, which changes its status from an XP constituent found in SpecDP to a head directly inserted in the highest functional head of the nominal projection, D0. In Section 2 we provide the general background and hypotheses for our analysis. In Section 3 we present our sources and the data. In Section 4 we address the nature of the Old Bulgarian (OB) determiner and its morphosyntactic status. Section 5 introduces an analysis of OB nominal structure and the parameter shift, while in Section 6 we address the issue of whether this change can be viewed as contact-induced or not. In Section 7 we provide a summary of our results and some conclusions.
2. Background and hypotheses The Balkan linguistic area is a standard example of contact-driven change (Hock & Joseph 1996; Thomason 2001; Asenova 2000, among others). From the point of view of diachrony, as well as linguistic typology, what is interesting about the core languages of the area, Albanian, Romanian, Bulgarian, Macedonian and Greek, is that they not only display mutual convergence, but also “individual and collective divergence from their historical antecedents” (Hock & Joseph 1996). The development of a post-posed definite article is among the most salient structural features mentioned in the literature. While Bulgarian, Macedonian, Romanian and Albanian display this feature, Greek clearly stands apart in featuring a pre-posed article which has led linguists to ignore Greek and its role in the process of evolving a common article typology. In this chapter we take a somewhat unconventional perspective and address the evolution of the article as attested in some of the earliest South Slavic texts (11th c.) in the light of contact with New Testament (NT) Greek, which was the source language for the early translations in a situation of active language interaction (bilingualism). Even though the enclitic article is grammaticalized in the core languages that display it and heads the top functional layer of the nominal projection, its distribution and placement inside the DP is subject to parametric variation. The variation is displayed in which categories are attracted to the article in D0, or possibly SpecDP, and what constraints obtain. Thus, AP movement to SpecDP gives rise to overt orders in which the definite article is enclitic to the AP, that is, ‘the post-posed definite article’. In a similar fashion, the article may encliticize onto the nominal head, again giving rise to postposition. In contrast, in Greek the article precedes both adjectives and/or the head noun; however, the Greek situation can be accounted for in terms of the specific requirements on well-formedness and licensing inside the DP.
An article on the rise
The statistical observations and tendencies displayed by our data suggest, counter to traditional views (Gaˇlaˇbov 1950 being the only exception), that at this stage (10–11th cc.) the article is already a functional category inside the nominal expression with all structural and semantic consequences of this status. We propose that this situation is the result of a reanalysis (in the sense of Harris & Campbell 1995, and the development in Roberts 2007) of the item at hand of the type maximal (phrasal) constituent > head (SpecDP > D0), and show that in the structure of the nominal phrase at this stage the article and the demonstrative are already distinct syntactic categories, despite the complete phonological/morphological overlap. Structural and formal criteria are used in support of the analysis proposed, and in addition we suggest that the two items at hand can be distinguished on the basis of a discourse function provisionally labelled the short-term memory recall function. In the absence of parallel developments in the other Slavic languages, the Balkan Slavic article is clearly an innovation (Hock & Joseph 1996), while the corresponding Greek item had a long history behind it already at the NT Greek stage (Manolessou 2000; Manolessou & Horrocks 2007). We suggest that, while, syntactically, the early Slavic article evolves in a language-internal way consistent with syntactic tendencies already present in the system, such as the shift from a basic SOV syntax to an SVO pattern and the development of functional categories at clausal and phrasal level (cf. Dimitrova-Vulchanova & Vulchanov 2006, 2008), its rise at this time may have been sped up and influenced by the existence of a comparable functional category in NT Greek. Most probably, the NT Greek functional category acted as some kind of catalyst to a parameter shift already well under way that forced the reanalysis at hand (Roberts 2007). We speculate that the NT Greek influence applies to the necessity of evolving a top functional layer and the parameters that apply to the well-formedness of DPs, such as the Spec-Head Visibility Parameter proposed in Dimitrova-Vulchanova & Giusti (1998), and do not involve any direct ‘imitation’ or structure-copying, as may be the case in pidgins, and as was held to be the case by earlier scholars investigating the influence of Greek on OB syntax. Our hypothesis, however, is in line with traditional accounts of the mechanism underlying the Balkan post-posed articles in terms of a reanalysis in contact position (cf. Asenova 2000 and references therein).
3. Sources and data Our analysis builds on comprehensive data from the specialized electronic corpus of Old Bulgarian nominal phrases (available at http://www.hf.ntnu.no/hf/adm/forskning/ prosjekter/balkansim/index.html) gathered from one of the earliest Slavic manuscripts, the Codex Suprasliensis (CS), from the early eleventh century (based on the standard edition by Zaimov & Capaldo 1982). The corpus contains over 10,000 nominal expressions
Mila Dimitrova-Vulchanova & Valentin Vulchanov
and, as such, provides a solid basis for statistical generalizations and reliable observations concerning the earliest documented period.
4. Article or demonstrative 4.1 The Old Bulgarian determiner system The OB determiner system is characterized by the co-existence of two categories, a demonstrative and a clitic article on the rise, which, at this stage of the language, are homophonous. To the extent that our data attest the earliest possible stage of the language on record (10–11th cc.), it is impossible to state with certainty whether this stage was possibly preceded by a state featuring a non-homophonous, only demonstrative item (cf. also Svane 1962). The ambiguities arise due to: (a) the exact morphological overlap of the two categories at hand; and (b) the partial overlap in discourse function, in that both display an anaphoric function in referring to entities previously mentioned in the discourse. Lyons (1999) defines this common feature of demonstratives and definite articles as ‘identifiability’, in that the hearer can identify the referents of such expressions on the basis of the information provided by the discourse/speaker. A situation similar to the OB one is also attested diachronically for Greek (cf. Manolessou 2000). Moreover, both the article and demonstrative occur only in definite expressions due to their referential properties (e.g. cf. Longobardi 1994). The two items in question diverge only in the deictic proper function, typical only of the demonstrative (Renzi 1997; Giusti 2001). Lyons (1999) provides a clearer description of this distinction in terms of the demonstrative signalling that the referent is immediately accessible to the hearer, while, when the article is used, there is inferencing involved. Observe the examples in (1)–(2) below. (1) a.
ñ¸¸ æå ãëàñú sii že glasъ this.nom.m discourse cl voice.nom.m
(OB, CS 35, 4)
a′. ταύτας бε τάς φωνάς this.pl.acc.f discourse cl art.pl.acc.f voice.pl.acc.f (2) a.
ì©æ¸ ñ¸ì¸ ñâťòû¸ì¸ mo˛ži simi svętymi man.pl.inst.m art.pl.inst.m holy.pl.inst.m
(NTGreek, CS 35, 4)
(OB, CS 78, 28)
ν ˜ν a′. τω ἀγίων τoύτων ἀνбĕω art.pl.gen holy.pl.gen these.pl.gen men.pl.gen (NTGreek, CS 78, 28)
The pair in (1) most likely attest the demonstrative evidenced by the phrase-initial position and the reinformcement of the deictic feature by the immediately following discourse clitic æå že and бε respectively, with the possible interpretation ‘this very voice’.
An article on the rise
In contrast, the pair in (2) instantiate what we claim to be an instance of the evolving article, again evidenced by its syntactic position in the phrase immediately after the first XP constituent ì©æ¸ mo˛ži ‘men’, which is in the process of getting established as the standard site for inserting the article. It is worth noting that the OB example in (2a) corresponds to a NT Greek phrase in (2a′), which, in addition to the article, features the demonstrative τoύτων ‘these’. The latter, however, does not weaken our hypothesis about the emerging OB article, since the Greek demonstrative actually occurs in a weak position, as established by Horrocks & Stavrou (1987) for Modern Greek, and Manolessou (2000) for Greek historically. We adress those data in our analysis in 4.2 below. Thus, despite the partial overlap, there are clear tendencies in the syntactic placement of the two items that suggest a split in function already at this stage. In earlier work (Dimitrova-Vulchanova & Vulchanov in press) we claimed that the article can be distinguished on the following grounds: a. With regard to discourse function the article does not have the deictic proper function, and is underspecified for the features definite and anaphoric, while lacking the feature deictic (cf. Renzi 1997 & Giusti 2001); and b. The article is a structural marker of the phrase as a whole and provides its functional layer (the DP part of the projection). Below we elaborate on these criteria.
4.2 Criteria We rely on the following criteria to distinguish between the two categories. The first criterion has to do with the nature of functional categories. As already observed, functional categories exhibit a low selection for their host (Zwicky 1977, Zwicky & Pullum 1983; the grammaticalization tradition, and for Old Bulgarian, Gaˇlaˇbov 1950). The emerging OB article is no exception to this rule. Already at this stage it can cliticize to a vast variety of categories inside the nominal expression – adjectival modifiers, the head noun and quantifiers. (3) a.
ïðýñëàâúíú¸ òú äåíü prěslavъnъi tъ denь most glorious.acc.m art.acc.m day.acc.m
a′. τē ἐνбóξῳ ἡμέĕᾳ art.dat.f honoured.dat.f day.dat.f b.
äąõà ñåãî íå÷¸ñòààãî duha sego nečistaago spirit.gen.m art.gen.m impure.gen.m
(OB, CS 35, 5–6) (NTGreek, CS 35,5–6)
(OB, CS 173, 11)
b′. τoèυ πνεύματoς τoèυ ἀϰαϑάĕτoυ art.gen.m spirit.gen.m art.gen.m impure.gen.m (NTGreek, CS, 173, 11)
Mila Dimitrova-Vulchanova & Valentin Vulchanov
c.
âüñå òî ñåëî vьse to selo all.acc.n art.acc.n place (village).acc.n
c′. τoèυ παντòς τóπoυ art.gen.m all.gen.m place.gen.m (4) a.
áëΰζενû¸ òú γλΰñú blaženyi tъ glasъ blessed.acc.m art.acc.m voice.acc.m
(OB, CS 207, 4) (NTGreek, 207, 4)
(OB, CS 123, 28–9)
a′. τèης μαϰαĕίας ϕωνηèς ταύτης art.gen.f blessed.gen.f voice.gen.f dem.gen.f (NTGreek, CS 123, 28–9)
The example in (3a) above illustrates encliticization on an adjective phrase ïðýñëàâúíú¸ prěslavъnъi ‘most glorious’, while (3b) provides an example of attachment to the nominal head, and (3c) shows the article-to-be immediately following the universal quantifier âüñü vьsь.2 For an analysis and discussion of the syntax of the latter item, see Dimitrova-Vulchanova, Giusti & Vulchanov (to appear). Parallel OB–NT Greek data of the type given in (4) are crucial for our claim, since they attest the innovative character of the Bulgarian item at hand. As can be seen, in those examples the NT Greek expression containing a demonstrative in its weak anaphoric function and an article is rendered by an expression in OB featuring only one item, namely the determiner. The latter, we hypothesize, functionally corresponds to the Greek article by occurring in the top DP Spec–Head configuration that is being established at that stage. For Greek it has been claimed independently that the phrase-internal/final position of the demonstrative is a weak position and corresponds to an anaphoric interpretation in the sense of earlier mention of the referent as opposed to the pre-article phrase-initial position, which is strong and clearly deictic (cf. Horrocks & Stavrou 1987; Manolessou 2000). Clearly then, as far as the Greek text in (4a′) is concerned, we are not dealing with the “immediate accessibility” (deictic) demonstrative feature of Lyons (1999). Thus, in terms of discourse function and interpretation the OB and the NT Greek example are exactly parallel, despite the differences in structure and overt linearization. Such data mutually support each other, and provide clear evidence of the new category emerging in OB and its functional nature. Furthermore, these data attest that the emerging OB article corresponds only functionally to the NT Greek article, while not imitating literally the Greek structure or linearization.
. That these are instances of encliticization is corroborated by the fact that in other texts from the same period, the article is actually spelled together with the preceding word, usually the noun, while during the 13th century this becomes a very common pattern (Svane 1962).
An article on the rise
Further criteria that point in the direction of a functional category include the semantic nature of the item in question. Functional categories tend to be underspecified for semantic features, and the OB article is no exception. It is in sharp contrast with the demonstrative, which in many contexts is semantically marked (e.g. in the deictic proper function). This is demonstrated in the examples in (5) below, where the OB demonstrative corresponds to the strong distal demonstrative in Greek, and serves to mark deixis by way of direct pointing, as judged by the discourse context. Observe also that the Greek example in (5a′) contains the strong deictic demonstrative occurring in the phrase-initial pre-article position. (5) a.
îíû ν¸âû против© ony nivy protivo˛ that.pl.nom.f field.pl.nom.n opposite
(OB, CS 37, 22)
a′. ἐϰειναι αί ἀπεναντι χω˜ĕαι that.pl.nom.f art.pl.nom.f opposite estate.pl.nom.f ‘those fields (on) the opposite (side)’ (NTGreek, CS 37, 22)
As observed by Gaˇlaˇbov (1950), the semantic distinction between the demonstrative and the article is that the demonstrative is “immediately anchored in the situation of utterance”, while the article is bleached/underspecified and instantiates removed “intellectualized” reference. This observation is surprisingly similar to Lyons’ (1999) “immediate accessibility” (demonstrative) vs. “inferencing” (article) distinction, though phrased in terms of a different paradigm. Furthermore, in parallel with other functional/phonologically weak categories in OB at this stage, which exhibit classical Wackernagel (2P) effects (cf. Anderson 1993 for a parallel with V2 phenomena), the emerging article occurs in the second position of the phrase. In (6a) below we find one of the first, still rare, occurrences of the possessive dative clitic in the nominal-internal Wackernagel position. Observe that the second position may also host discourse clitics that serve to mark focus or topic phrase-internally in parallel with the domain of the clause (for the DP-clause parallel cf. Giusti 2006, and for information-structure phenomena in the Balkan DP see Dimitrova-Vulchanova & Giusti 1998). (6) a.
ïðúâî¬ ò¸ âúïðàøàí¸¬ prъvoje ti vъprašanije first.acc.n you.cl.dat questioning.acc.n
(OB, CS 177, 26–27)
˜ν a′. τω πĕoτέĕων σoυ ἐπεĕωτήσεων art.pl.gen prior.pl.gen you.cl questioning.pl.gen ‘your first/prior questioning’ (NTGreek, CS 177, 26–27)
The existence of items displaying Wackernagel effects bears witness to the emergence of an articulated left periphery in the nominal expression, for instance, a process of grammaticalization and the rise of syntactic structure related to the expression of functional
Mila Dimitrova-Vulchanova & Valentin Vulchanov
categories. To the extent that, at this point, we do not have examples in our data featuring simultaneous co-occurrence of these functional categories, there is no evidence of the fine structure of the left periphery and the number of projections there. Concerning the rising article, we propose to analyse this process as a reanalysis in terms of structural position inside the nominal expression of the type SpecDP > D0 (cf. Giusti 2001 for Italian), whereby the item at hand no longer moves to SpecDP and is instead inserted directly in D0. Diachronically this type of reanalysis can be described as a parameter change in the syntactic mechanism involved of the type Move > Merge (cf. Roberts & Roussou 2003). Crucially, the ambiguous category splits into two distinct items distinguished in syntactic behaviour. One we label dem, which retains its XP nature and moves to SpecDP, the other det (the article on the rise) which is directly merged in D0 attracting other categories to raise either to SpecDP (modifiers) or adjoin to it in D0 (the nominal head). As demonstrated by our data, at this stage (10th–11th cc.), the article and the demonstrative are already distinct categories found in complementary distribution. We elaborate on the diachronic processes involved in the following sections.
4.3 Other modifiers inside the nominal expression For the purposes of our discussion we introduce the general patterns of modification inside the OB nominal expression. Of particular interest for the placement of the demonstrative and the article is the ordering of adjective phrases, possessive adjectives and denominal adjectives relative to each other, and relative to the nominal head. Possessive adjectives (PAs) have been described in the literature as a category typical of Slavic (cf. Trubetzkoy 1937; Corbett 1987). These are modifiers derived from a pronominal base with adjectival morphology, and, as such, represent a hybrid category, simultaneously agreeing with the nominal head and participating in binding relations. In our data they are overwhelmingly post-nominal at a ratio 2 (N > PA): 1 (PA > N). The OB DP is characterized by a number of permutations, both in the preand post-nominal position, thus the extended pattern is either N > PA > AP or N > AP > PA. Observe that PAs are not in complementary distribution with the emerging article or demonstratives (in contrast to e.g. Modern English, but not Modern Bulgarian), as shown in (7) below. (7) a.
áîãà íàøåãî ¸ñò¸íüíààãî boga našego istinьnaago God.acc/gen.m our.pa.acc/gen.m true.acc/gen.m
a′. τòν ἀληϑινòν art.acc.m true.acc.m b.
˜ν ϑεòν ἡμω God.acc.m your.pl.gen
îáðàґîìú ñЃòû¸ìú òâо¸ìú obrazomъ svętyimъ tvoimъ image.inst.m holy.inst.m your.pa.inst.m
(OB, CS, 68, 14)
(NTGreek, CS 68, 14)
(OB, CS 22, 3)
An article on the rise
b′. εἰϰόνι ἁγίᾳ σoυ image.dat.n holy.dat.n your.gen c.
м©ками сими твоими mo˛kami simi tvoimi torment.pl.inst.f art.pl.inst.f your.pl.inst.f
˜ν c′. τω βασάνων σoυ τoύτων art.pl.gen torment.pl.gen you.gen this.pl.gen
(NTGreek, CS, 22, 3)
(OB, CS 113, 4) (NTGreek, CS 113, 4)
Another inherently Slavic category consists of denominal adjectives (DAs), again hybrid in nature, formed by a nominal stem and adjectival morphology (Trubetzkoy 1937; Corbett 1987). Typically, they realize a variety of thematic roles, and as a rule translate NT Greek genitive DPs. As shown in the examples below, they occur maximally close to the head noun, either immediately preceding it or immediately following it (cf. the discussion in Dimitrova-Vulchanova & Vulchanov 2003, in press). Such a situation is not surprising on the assumption that they are thematically related to the head noun and originate in a low specifier position (e.g. the projection immediately dominating N). (8) a.
íàêà´àí¸¸ ãîñïîäüíΈ (Agent) nakazanii gospodьni admonition.loc.n (of) Lord.da.loc.n
a′. b.
τē πĕoτĕoπē τoυ Xĕιστoυ art.dat.f persuasion.dat.f art.gen.m Christ.gen.m (NTGreek, CS 75, 30) îñòàíüö¸ îãíüí¶¸ (Source) ostanьci ognьnii remnant.pl.acc.m (of) fire.da.pl.acc.m (OB, CS 94, 13)
(OB, CS 75, 30)
b′. τὰ τoυ πυĕòϚ λείψανα art.pl.nom.n art.gen.m fire.gen.m remnant.pl.nom.n (NTGreek, CS 94, 13)
Now when we look at the patterns involving the demonstrative and the article-to-be, two positions appear to be crucial, the prenominal, and the postnominal one. The prenominal position, which is also the phrase-initial position, is typical only of the demonstrative. Some examples are given in (9c–d) below. Clearly, in these examples, the demonstrative occupies the left edge of the phrase (SpecDP), since both phrases are governed by a preposition. However, like other modifiers, including dem-like adjectives, the demonstrative occurs also post-nominally, as shown in the examples in (9a–b). As we will see, in the absence of other nominal elements, such examples are ambiguous. It is only in the phrase-final position, as in (9e), that there is no ambiguity concerning the XP status of the item at hand.
Mila Dimitrova-Vulchanova & Valentin Vulchanov
(9) a.
въ исповýдани· семъ vъ ispovědanii semъ in confession.dat.n this.dat.n
a′. ἐν τē ὁμoλoγίᾳ ταύτῃ in art.dat.f confession.dat.f this.dat.f b.
γπΰδΰ ςξγξ grada togo town.gen.m that.gen.m
b′. τηϚ πόλεωϚ art. gen.f town.gen.f c.
βό ςξμü ăđаäý vь tomь gradě in that.loc.m town.loc.m
c′. τηϚ πόλεωϚ art.gen.f town.gen.f d. âü ñü ÷àñú vь sь časъ at this.acc.m hour.acc.m d′. ἐν τē ὥĕᾳ ταύτῃ at art.dat.f hour.dat.f this.dat.f e.
(OB, CS 16, 27–28) (NTGreek, CS, 16, 27–28)
(OB, CS 81, 2) (NTGreek, CS, 81, 2)
(OB, CS 26, 25) (NTGreek, CS, 81, 2)
(OB, CS 22, 10) (NTGreek, CS 22, 10)
пръвûими сЃтûими м©жи онýми prъvyimi svętyimi mo˛ži oněmi first.pl.inst.m holy.pl.inst.m man.pl.inst.m that.pl.inst.m ‘these primary holy men’ (OB, CS 371, 8–9)
The most intriguing position for our analysis is the immediately post-nominal position. This position is ambiguous when the nominal expression contains only the nominal head and the item under analysis. On the current analysis, this may be a default instance of the rising article (e.g. in terms of the postnominal position coinciding with the second position). Alternatively, this could be a case of a phrase-final demonstrative, just like the one in (9e) above, the only difference being that there is no other material occurring between the noun and the demonstrative. This structure is illustrated in examples in (9a–b) above and in (10) below. (10) a.
ґвэри¬ òи· zvěrije tii beast.pl.nom.m the/these.pl.nom.m
b.
ìýñòý òîìü městě tomь place.loc.n the/that.loc.n
(OB, CS 15, 3)
(OB, CS 14, 27)
Structurally, this position can correspond to either N – dem or N – det, both coherent with the predominant pattern of marking definiteness in OB, namely full N-to-D0
An article on the rise
movement, the only difference being the position and status of the determiner that follows, either in an intermediate functional specifier between N and D0 (if dem) or in D0 and adjoined to N, if det. We consider this pattern crucial in triggering the parameter shift that followed, firstly because it was the source of structural ambiguity, and secondly, due to its high frequency of occurrence in the data. It is worth mentioning that out of a total of 244 examples in our corpus featuring a post-nominal determiner item, 215 involve the ambiguous pattern at hand with only a noun followed by a determiner. We address these data and the diachronic process in detail in Section 5 below. Interestingly, our analysis is in contrast with the OB tradition otherwise (Zaimov & Capaldo 1982, Ivanova-Mirčeva & Haralampiev 1999: 125–129), where such contexts are taken as clear evidence of the emerging article by analogy with the modern post-posed article. Only extended modification patterns provide clear evidence and disambiguate between the two homophonous categories. Thus, patterns of the type AP > det > N or N > det > *(AP), illustrated in (11), provide crucial evidence for our claim that already at this stage we are dealing with two distinct categories: a head (det) and a demonstrative (dem), the latter of a phrasal nature. Roberts & Roussou (2003) propose that grammaticalization of D0 involves a change of the type Move > Merge, and in our case the category det is directly merged in D0. (11) a.
ìàëîâðýìåíüíààãî ñåãî æ¸ò¸ÿ malovrěmenьnaago sego žitija transitory.gen.n art.gen.n life.gen.n
(OB, CS 72, 8)
a′. τῂ πĕoσϰαίĕῳ ταύτῃ στĕατείᾳ art.dat.f transitory.dat.f this.dat.f military.service.dat.f (NTGreek, CS, 72, 8) b.
íå÷¸ñòąąěą òîìą äЃõą nečistuumu tomu duhu impure (evil).dat.m art.dat.m spirit.dat.m
(OB, CS 36, 12–13)
b′. τῳ ἀϰαϑάĕτῳ ἐϰεινῳ πνεύματι art.dat.m impure.dat.m that.dat.m spirit.dat.m (NTGreek, CS, 36, 12–13) c.
áðΰò¸ť ñåť μξť υσδϋť bratię seę moę hudyę brother.coll.gen.f art.gen.f my.pa.gen.f poor.gen.f (OB, CS 123, 25)
˜ν ˜ν c′. τω ἀбελφω˜ν μoυ τω ἐλαχίστων art.pl.gen brother.pl.gen my.gen art.pl.gen poor.pl.gen (NTGreek, CS 123, 25) d. υąäϋť ñåť íàøåť жи´ни hudyę seę našeę žizni short.gen.f art.gen.f our.gen.f life.gen.f
(OB, CS 67, 29)
d′. τò βĕαχὺ της ζωης ἡμω˜ν art.nom.n shortness.nom.n art.gen.f life.gen.f our.pl.gen (NTGreek, CS 67, 29)
Mila Dimitrova-Vulchanova & Valentin Vulchanov
The examples in (11) demonstrate the emerging article as found in the extended pattern of modification in the context of an AP in SpecDP, to which it encliticizes, as in (11a–b, d) or with N adjoined to it in D, as in (11c). Observe that the syntactic status of the item at hand is not related to phonological/morphological heaviness, as seen in the item in (11b),3 but rather to a specialized syntactic position (cf. Zwicky 1977, Zwicky & Pullum 1983, van Riemsdijk 1999) and, more importantly, to the complementary distribution between the light/clitic item (what we label det) and the full constituent (dem in our case). In (12) below we provide further evidence of the functional nature of det in adjective coordination data. (12) a.
ìðüòâààãî òîãî ¸ íýìààãî êàï¸øòà mrьtvaago togo i němaago kapišta dead.gen.n art.gen.n and dumb.gen.n idol.gen.n (OB, CS 226, 25–26)
a′. ϰoυφoυ ϰαὶ ἀλάλoυ ξoάνoυ idle.gen.n and dumb.gen.n idol.gen.n
(NTGreek, CS 226, 25–26)
b. b′.
áå´áîæüíî¬ îíî ¸ íå÷üñò¸âî¬ ïîâýäàí¸¬ bezbožьnoje ono i nečьstivoje povědanije godless.acc.n art.acc.n and evil.acc.n proclamation.acc.n (OB, CS 85, 5) τò ἄϑεoν ἐϰεινo ϰαὶ art.nom.n godless.nom.n that.nom.n and
ἀσεβὲς ϰήĕυγμα ungodly.nom.n proclamation.nom.n
c.
ρии [ëüñòиâûи и íåąêðîøòåíûи] ïüñú sii [lьstivyi i neukroštenyi] pьsъ this.nom.m alluring.nom.m and untamable.nom.m dog.nom.m (OB, CS 60, 7)
(NTGreek, CS 85, 5)
c′. αὐτὸς бὲ ὁ δóλιoϚ καὶ ἀνήμεĕoς this.nom.m discourse cl art.nom.m deceitful.nom.n and fierce.nom.m
κύων dog.nom.n
(NTGreek, CS 60, 7)
The examples in (12) above further demonstrate the functional status of the emerging article and provide evidence that the article, or rather D and the top DP layer, is in the process of being grammaticalized. In a fashion similar to the modern language, the article needs to encliticize only on the first conjunct in APs in coordination. This is to
. Special clitics are not necessarily phonologically light, e.g. loro in Italian.
An article on the rise
be expected on the assumption that the first conjunct is higher and only one XP constituent needs to raise to SpecDP (cf. the analysis of Modern Bulgarian in DimitrovaVulchanova & Giusti 1998). The Modern Bulgarian data are given in (13a). In contrast, when SpecDP is already filled by a demonstrative, as in (12c), raising of the first AP in the conjunct is pre-empted, and both APs are preceded by the demonstrative, ρии ‘this’ in this case. (13) a.
[DP [SpecDP [FP1 новаma] …[ SpecFP/ConjP [Conj и [FP2 удобна] … къща]]] novata i udobna kăšta new-the and comfortable house ‘the new and comfortable house’ (one referent)
b.
[DP/ConjP [DP [SpecDPумнаmа___ ]] и [DP [SpecDP хубаваmа] секретарка]]] umnata i xubavata sekretarka clever-the and beautiful-the secretary ‘the clever (secretary) and the beautiful secretary’(two referents) (Modern Bulgarian)
The minimal pair in (13) illustrates an important property of Modern Bulgarian concerning the role of the article as a referential index. As proposed by Longobardi (1994), the definite article functions as a referential index, thus keeping track of individuals in the discourse. This property is well-attested in Modern Bulgarian (DimitrovaVulchanova 2002; Penchev 1998), in that the presence of the article can disambiguate between one-referent interpretations and two-referent interpretations in the case of adjectives in coordination. As already mentioned above, for the purposes of licensing definite DPs only one of the adjectives in coordination needs to raise to SpecDP to a spec-head configuration with the D head. In contrast, in the case of DPs in coordination both APs need to raise in their respective phrases. Thus, as shown by the contrast in the pair in (13) above, the simple rule is one article per phrase (DP), and if both APs bear the article, the level of coordination is that of the DP, respectively corresponding to an interpretation in terms of two referents (cf. also Bouchard 1998 on the semantic realization of the category of number and its relation to the article).
5. Parameter shift and the loss of N-to-D movement Despite the clear grammaticalization tendencies already visible at this stage, the system is unstable and is characterized by the co-existence of different parameter settings, as well as ambiguous linearization patterns that give rise to two potential underlying grammars. Our data attest both partial and full N-to-D movement as a way of signalling definiteness. Out of 10,000 nominal expressions present in the corpus, 4,820 (48%) are N-initial, while 640 expressions (6%) display partial N-movement.
Mila Dimitrova-Vulchanova & Valentin Vulchanov
Thus, full N-to-D movement is on the way to being established as the mechanism that ensures DP licensing in terms of the visibility of the top DP layer. This parameter has been proposed in work by Dimitrova-Vulchanova & Giusti (1998) to capture the cross-linguistic variation and choices languages make in terms of making the top Spec–Head configuration visible, the obvious options including overt movement to SpecDP (as in the case of APs in Modern Bulgarian, and demonstratives cross-linguistically), or merging a lexical head in D (an article cross-linguistically), or through overt N-to-D movement, as in Old Bulgarian. Now, obviously, languages tend to be parsimonious in the mechanisms at play at any given time (cf. the parallel with the doubly-filled Comp filter of Chomsky & Lasnik 1977), and this is reflected in diachronic processes that tend to reduce optionality and make specific choices illicit because they are redundant. The reanalysis witnessed in our data and when compared to Modern Bulgarian involves a number of parameters: the Spec–Head Visibility Parameter, and the availability of N-movement (partial N-movement > N-to-D) as one setting of this parameter. From the point of view of a more general mechanism employed by a specific grammar at a specific synchronic stage, Move is being replaced by Merge (Move > Merge) seen in the evolution of the lexical head that is overtly inserted in D (det), and the split with the item that moves to SpecDP (dem). When we look at settings obtaining for the Spec–Head Visibility Parameter (Dimitrova-Vulchanova & Giusti 1998), a valid observation is that dem raising is still optional, and we find a lot of N-movement (partial, but predominantly N-to-D). At the same time, AP raising to SpecDP, even though optional, also becomes increasingly common. Observe that, in definite expressions both AP raising and N-to-D movement are found in the context of an overt head (det) merged in D. Diachronically, N-to-D movement ceased to be an option,4 while the Spec–Head Visibility Parameter is satisfied exclusively through overtly filling SpecDP, e.g. through the highest AP or dem that raise to that position. Observe that, in both cases, the D head is not overt in the modern language; it only contains the D feature. DP licensing is ensured either through agreement with an AP whose head bears the article or, in the absence of modifiers and when N is inflected with the article, at LF in a chain with the head noun in situ. In (14) below we give the structure of the OB nominal expression featuring the options of satisfying the Spec–Head Visibility Parameter available at that point.
. As attested in Modern Bulgarian (cf. the analysis in Dimitrova-Vulchanova & Giusti 1998). Exactly what stage was crucial in this process is subject to future detailed research.
An article on the rise
DP
(14) Spec
D′
APj/demk D0
FP
Ni+det Spec
FP
tj/tk Spec
NP Spec
demlow N0 ti
N′ (XP)
The phrase marker in (14) above sketches all available possibilities, which, however, may obtain only one at a time. If an article (det) is merged in D, this would serve as trigger for movement to the DP area. There are two options: either N has to move to det in D, successively passing through head positions on the way, or the highest AP modifier has to move to SpecDP, thus occurring in strict adjacency to det in D, respectively illustrated in (15a) and (15b) below. Of the two mechanisms, full N-to-D movement was the prevailing one at that stage, as demonstrated by the corpus data: out of 10,000 nominal expressions in the corpus, 4,820 (48%) are N-initial. The AP movement alternative was just emerging in that period, but already gaining ground and is amply represented in our data.5 A third option would be for a demonstrative to move to the highest specifier position, SpecDP. There appear to have been two positions where the demonstrative could merge, a low specifier position immediately above N,6 and the highest functional specifier position, the latter gradually becoming established as the only available merge site in Modern Bulgarian. Quite likely, the consolidation of the latter option is related to the modern setting of the Spec–Head Visibility Parameter and the loss of N-to-D movement. Examples are given in (15c) and (15d), respectively. The data in (15c) provide an excellent illustration of the pervasiveness of N-movement as a mechanism of marking definiteness in the nominal expression at that stage. While (15c) shows short N-movement just across the low demonstrative, (15c¢–c″) demonstrate N-movement all the way to the phrase-initial position, thus giving rise to
. Our data already witness 29 instances of this pattern (12% of post-nominal modification occurrences featuring the determiner). . The low demonstrative position was originally proposed by Brugè (1996) for Spanish and has been attested in many languages (cf. Brugè 2002), including Classical and New Testament Greek (cf. Manolessou 2000, 2001).
Mila Dimitrova-Vulchanova & Valentin Vulchanov
post-nominal ordering of modifiers, which is very typical of the earliest recorded data (Dimitrova-Vulchanova & Vulchanov 2003). Dem could only move to SpecDP from the highest functional specifier in the DP in any given structure.7 The independent application of any of these mechanisms ensures that no minimality constraints will be violated. (15) a.
ì©æà òîãî ρЃςΰΰγξ mo˛ža togo svętaago man.gen.m that.art.gen.m holy.gen.m
b.
íå÷üñòиâ©« ò© ´ΰοξβýδü nečьstivo˛j to˛ zapovědь evil.acc.f this.art.acc.f command.acc.f
c.
ëþòúи ñëąãà òú ljutъi sluga tъ ferocious.nom.m servant.nom.m this.nom.m
(CS 58, 25)
c¢. đàáú òâîиõú ñиυϊ rabъ tvoihъ sihъ slave.pl.gen.m your.pa.gen.m these.pl.gen.m
(CS 109, 29–30)
c″. ребро мо¬ се rebro moje se rib.acc.n my.pa.acc.n this.acc.n
(CS 504, 11–12)
d. ñиõú äîáðûиõú ì©÷åíиêú sihъ dobryihъ mo˛čenikъ this.pl.gen.m good.pl.gen.m martyr.pl.gen.m
(CS 216, 9–10)
(CS 174, 21–22)
(CS 66, 23–24)
The subsequent steps in that process are: ––
––
The emerging article, being an item that is syntactically weak (Zwicky 1997; van Riemsdijk 1999), evolves into one that is also phonologically weak, a change of the type syntactic clitic > phonological clitic (special clitic) > morpheme. Overt N-to-D raising ceased to be an option when det became phonologically weak (e.g. an affix, cf. Modern Bulgarian), and it may be speculated that overt N-movement became too costly (cf. Dimitrova-Vulchanova & Giusti’s 1998 claim that there is no N-to-D movement, not even at LF), compared to satisfying the Spec–Head Visibility Parameter by raising the highest AP or dem to SpecDP.
Another over-arching parallel process is the reanalysis underlying the complete split between the article-to-be and the demonstrative, which we believe is triggered by
. This would depend on whether other modifiers were present or not, and whether functional specifiers on the way were filled. In the absence of other modifiers, one could assume that dem may move from the lower dem position.
An article on the rise
parameter ambiguity (in the sense of Roberts 2007) involving the ambiguous N-dem/ det strings. Quite likely subsequent generations of speakers8 ended up parsing the N-dem string into an N-det string, thus (a) resolving grammar ambiguity, and (b) making the complementarity of the two items complete, leaving pre-nominal dem as the only option for the strong demonstrative (e.g. in line with the re-setting of the Spec–Head Visibility Parameter). Svane (1962) convincingly claims that the article is already fully evolved during the thirteenth century, while Mirčev (1953, 1955) dates this to an even earlier stage, as documented in manuscripts from the twelfth century. Alongside the functional change, our data also witness an interesting development in the direction of a semantic distinction. Already in our data we see a discourse function that we provisionally label the short-term memory recall function, which is typical of dem and serves to distinguish the latter from the emerging article. Generally, articles have a cohesive function in discourse and serve to keep track of referents mentioned previously. Lyons (1999) also suggests that articles require establishing the identity of the referent by searching the spatio-temporal or textual context, and involve a process of inferencing on part of the hearer, while demonstratives serve to direct the hearer’s attention by ostension (pointing). The examples in (16) below illustrate this difference. (16) a.
ñ¸¸ äóõîâå sii duhove this.pl.nom.m spirit.pl.nom.m
(OB, CS 53, 16–17)
b.
ìýñòî òî město to place.nom.n art.nom.n
(OB, CS 26, 19–20)
In (16a) we find a demonstrative judged by its prenominal (strong) position. It is used in a context that reads ‘these spirits, the ones we have been talking about’ and corresponds to quite a number of references to the spirits in previous discourse, but is separated from the instances of earlier mention by a number of lines. Thus, this use can be described as one that picks up a referent as topic by pointing, as indicated by the use of the proximal demonstrative ñ¸¸ ‘these’, and serves as a kind of reminder to the speaker to recall those earlier mentions, some kind of ‘memory index’. Hence the label we suggest. In contrast, in (16b) the article simply serves as a cohesion device (cf. Halliday & Hasan 1976) in referring to a specific definite individual (the place in question) and its use presupposes using knowledge of the situation (e.g. inferencing) on part of the hearer. . This is a well-attested process in language diachrony, with the only reservation that “subsequent generations” should be interpreted broadly. In actual time this process may take up to several hundred years before the change/reanalysis is complete.
Mila Dimitrova-Vulchanova & Valentin Vulchanov
6. Contact-induced change? An interesting question concerning the parameter shifts is whether contact with NT Greek played a role at all. Here we refer to contact in the psycholinguistic sense applying to situations of bilingualism, that is, two (or more) languages co-existing in the same individual, which was most likely the case of the first translators of NT texts into OB. Clearly, the Spec–Head Visibility Parameter is set differently for NT Greek; however, the article is already a fully-fledged functional category with all the consequences of this status. We speculate that this very situation may have speeded up/catalyzed the evolution of the article in Bulgarian due to language contact (and bilingualism). Thus, the rising article in OB does not imitate blindly the NT Greek linearization pattern or structure. Instead, it follows the language-specific parameters for the licensing of DPs in the language, and the constraints on the just emerging top functional layer of nominal structure. It is exactly in this functional aspect that the Greek system might have influenced the parallel evolution observed in OB: the very existence of a functional category (D) in Greek may have accelerated the grammaticalization of OB nominal structure seen in the evolution of an overt head inserted in D. Our hypothesis is in line with the discussion in Gaˇlaˇbov (1950), who underlines the fact that despite appearances (e.g. the suffixal nature of the article in Modern Bulgarian), the evolution of the article is of an exclusively syntactic nature, clearly related to the development of a well-articulated DP structure.
7. Conclusions In this chapter we have addressed data from one of the earliest OB manuscripts which demonstrate some clear tendencies in the evolution of OB nominal structure. Our data attest that, at this stage, despite the complete morphological overlap with the demonstrative pronoun, the (enclitic) article is already a distinct category. Furthermore, as demonstrated in our data, ambiguities arise due to the (written) nature of the data (e.g. the absence of extended modification patterns), but it is exactly those same ambiguities that drive diachronic change and shifts in parameter settings to resolve parameter ambiguity (in the sense of Roberts 2007). In the case of Bulgarian, the ambiguous N-dem/N-det orders are ultimately resolved in just one structure, one involving enclitizing the article on the N head. Our data attest a reanalysis underway of the type Move > Merge, resulting in the ultimate split between the article and the homophonous demonstrative, and a change of the type partial N-movement > N-to-D movement related to the newly evolved functional (D) layer of the DP, a parameter setting which will, over time, eventually give way to XP raising to SpecDP as a more economical way of satisfying the Spec– Head Visibility Parameter.
An article on the rise
Our data show convincingly that the parameter change at hand is language-internal; however, quite likely, NT Greek acted as a catalyst in featuring a well-developed functional DP layer.
References Anderson, Stephen R. 1993. Wackernagel’s revenge: Clitics, morphology and the syntax of second position. Language 69: 68–99. Asenova, P. 1989/2000. Balkansko ezikoznanie. Osnovni problemi na balkanskija ezikov saˇjuz. Sofia: Nauka i izkustvo. Bouchard, Denis. 1998. The distribution and interpretation of adjectives in French. Probus 10: 139–183. Brugè, Laura. 1996. Demonstrative movement in Spanish: A comparative approach. University of Venice Working Papers in Linguistics 6(1): 1–53. Brugè, Laura. 2002. The positions of demonstrative in the extended nominal projection. In Functional Structure in DP and IP: The Cartography of Syntactic Structures, Vol. 1, Guglielmo Cinque (ed.), 54–90. Oxford: OUP. Chomsky, Noam & Lasnik, Howard. 1977. Filters and control. Linguistic Inquiry 8: 425–504. Corbett, Greville. 1987. The morphology/syntax interface: Evidence from possessive adjectives in Slavonic. Language 63(2): 299–345. Dimitrova-Vulchanova, Mila & Giusti, Giuliana. 1998. Fragments of Balkan nominal phrase structure. In Possessors, Predicates and Movement in the Determiner Phrase [Linguistik Aktuel/Linguistics Today 22], Artemis Alexiadou & Chris Wilder (eds), 333–360. Amsterdam: John Benjamins. Dimitrova-Vulchanova, Mila & Vulchanov, Valentin. 2003. Observations on the ordering of modifiers in nominal expressions in some Old Bulgarian texts with reference to Greek. In Slavia Orthodoxa. Ezik i kultura, Siebka Bogdanova (ed.), 75–84. Sofia: Universitetsko izdatelstvo Sv. Kliment Ohridski. Dimitrova-Vulchanova, Mila & Vulchanov, Valentin. 2006. Relative clauses in Old Bulgarian. In Ezikovedski izsledvanija v čest na 75-godišninata ot roždenieto na Jordan Penčev, Ivan Kutsarov et al. (eds), 77–88. Sofia: Artgraf Publishing House. Dimitrova-Vulchanova, Mila & Vulchanov, Valentin. 2008. Old Bulgarian syntax: The basics. In Problemi na Kirilo-Metodievoto delo i na baˇlgarskata kultura prez IX–X vek, Svetlina Nikolova (ed.), 243–259. Sofia: Marin Drinov Academic Publishing House. Dimitrova-Vulchanova, Mila & Vulchanov, Valentin. In press. An article evolving. In Grammatical Change: Origins, Nature, Outcomes, Dianne Jonas, John Whitman & Andrew Garrett (eds). Oxford: OUP. Gaˇlaˇbov, Ivan. 1950. Za člena v starobaˇlgarskija ezik. Izvestija na narodnija muzej v Burgas 1: 171–227 (reprinted in Gaˇlaˇbov, Ivan. 1986 Izbrani trudove po ezikoznanie, 282–329. Sofia: Nauka i izkustvo). Giusti, Giuliana. 2001. The birth of a functional category: From Latin ille to the Romance article and personal pronoun. In Essays in Honour of Lorenzo Renzi, Guglielmo Cinque & Giampaolo P. Salvi (eds), 157–171. Amsterdam: North Holland Publishers. Giusti, Giuliana. 2006. Parallels in clausal and nominal periphery. In Phases of Interpretation, Mara Frascarelli (ed.), 151–172. Berlin: Mouton de Gruyter. Halliday, Michael A. K. & Hasan, Ruqaiya. 1976. Cohesion in English. London: Longman.
Mila Dimitrova-Vulchanova & Valentin Vulchanov Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: CUP. Hock, Hans Henrich & Joseph, Brian D. 1996. Language History, Language Change and Language Relationship: An Introduction to Historical and Comparative Linguistics. Berlin: Mouton de Gruyter. Horrocks, Geoffrey & Stavrou, Melita. 1987. Bounding theory and Greek syntax: Evidence for wh-movement in NP. Journal of Linguistics 23: 79–123. Ivanova-Mirčeva, Dora & Haralampiev, Ivan. 1999. Istorija na baˇlgarskija ezik. Sofia: Faber. Lyons, Christopher. 1999. Definiteness. Cambridge: CUP. Longobardi, Giuseppe. 1994. Reference and proper names: A theory of N-movement in syntax and logical form. Linguistic Inquiry 25: 609–665. Manolessou, Io. 2000. Greek Noun Phrase Structure: A Study in Syntactic Evolution. Ph.D. dissertation, University of Cambridge. Manolessou, Io. 2001. The evolution of the demonstrative system in Greek. Journal of Greek Linguistics 2: 119–148. Manolessou, Io & Horrocks, Geoffrey. 2007. The development of the definite article in Greek. Studies in Greek Linguistics 27: 224–236. Mirčev, K. 1953. Koga vaˇznikva člennata forma v bulgarskija ezik. Baˇlgarski ezik 3 (1): 48–57. Mirčev, K. 1955. Istoričeska gramatika na baˇlgarskija ezik. Sofia: University Press. Penchev, Iordan. 1998. Intonationally determined movement. In Papers from the 2nd Conference on FASSL. University of Trondheim Working Papers in Linguistics 31, Mila DimitrovaVulchanova, Lars Hellan, Ivan Kassabov and Ilyana Krapova (eds.), 180–185. Trondheim: University of Trondheim. Renzi, Lorenzo. 1997. Fissione di lat. ILLE nelle lingue romanze. In Italica et Romanica. Festschrift für Max Pfister zum 65. Geburtstag, Gunter Holtus, Johannes Kramer & Wolfgang Schweickard (eds). Tübingen: Niemeyer. van Riemsdijk, Henk. (ed.) 1999. Empirical Approaches to Language Typology. Clitics in the Languages of Europe. Berlin: Mouton de Gruyter. Roberts, Ian. 2007. Diachronic Syntax. Oxford: OUP. Roberts, Ian. & Roussou, Anna. 2003. Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: CUP. Sproat, Richard & Shih, Chilin. 1991. The cross-linguistic distribution of adjective ordering restrictions. In Interdisciplinary Approaches to Language. Essays in honor of S.-Y. Kuroda, Carol J. Georgopoulos & Roberta Ishihara (eds), 565–593. Dodrecht: Kluwer. Svane, Gunnar. 1962. O sintaksičeskom primenenii bolgarskogo člena v XIII veke. B. Dobrejšovo evangelie. Scando-Slavica 8: 224–238. Thomason, Sarah G. 2001. Language Contact: An Introduction. Edinburgh: EUP. Trubetzkoy, Nikolai. 1937. O pritjažatel’nyh (possessiva) starocerkovno-slavjanskogo jazyka. In Zbornik lingvističkih i filoloških rasprava A. Beliću o četrdesetogodišnjici njegova naučnog rada posvećuju njegovi prijatelji i učenici, 15–20. Belgrade: Mlada Srbija. Vulchanov, Valentin & Dimitrova-Vulchanova, Mila. In progress. An Electronic Corpus of Old Bulgarian Nominal Phrases. Trondheim: Norwegian University of Science and Technology. Zaimov, Iordan & Capaldo, Mario. (eds). 1982. Suprasaˇlski ili Retkov Sbornik. Sofia: Izdatelstvo na Baˇlgarskata Akademija na Naukite. Zwicky, Arnold. 1977. On Clitics. Bloomington IN: Indiana University Linguistics Club. Zwicky, Arnold & Pullum, Geoffrey. 1983. Cliticization versus inflection: English n’t. Language 59: 502–513. Wackernagel, Jacob. 1892. Über ein Gesetz der indogermanischen Wortstellung. Indogermanischen Forschungen 1: 333–436.
Language index
A Afrikaans 46–47 Albanian 336 b Balinese 175–176 Bantu languages 249–266 Basque 38–39 Bemba 251–252, 256 Bengali 49, 51 Bulgarian 336, 352, 342, 347–349 Old Bulgarian (OB; = Old Church Slavonic) 335–353 Bwe 41 c Calabrian 245 Catalan 101–115 Chasu 261 Chichewa 265 Chinese (Mandarin) 40 Ciruri 261 Cushitic 43 Czech 183–184, 187 d Dravidian 48–50 Dutch 291–293, 311 Middle Dutch 66, 68–72, 280–281 e English 13–14, 29, 47, 62, 99, 113, 195, 237, 245, 254, 274–276, 289, 294–299 Old English 29, 37–38, 44, 62, 124–126, 134–136, 187, 190–191, 280 Middle English 134–137, 182, 187–188, 191–193, 196 Early Modern English 187–193
f Fang 113 Finnish 37 French 45–46, 72, 84, 110, 120, 126–132, 138 Old French 192, 132–134 Middle French 46, 132–134, 136–137 g Galician 113 German 13–30, 36, 40, 270, 281, 291–293, 311 Old High German (OHG) 17, 19–24, 322–328 Middle High German (MHG) 25–26, 68–70, 321–325, 328–331 Early New High German (ENHG) 21, 28–29, 69, 150, 321–322, 325 New High German (NHG) 321–325, 329 German dialects Austrian German 28 Bavarian 16–17, 22, 25–26 Franconian 17, 328 Swiss German 28 Thuringian 16 Low German Middle Low German 68–70, 145–163 Germanic 37–38, 43, 160, 321 Proto-Germanic 15, 18 West Germanic 61–62, 68–70 Greek 245, 315 Asia Minor Greek 50 New Testament Greek 335–353 Guaraní 107, 113
h Herero 261 Hindi–Urdu 49–53, 177 i Indo-Aryan 48–53 Indo-European 43 Indonesian 169, 175–176 Italian 14, 110, 205, 216, 346 Irish Old Irish 225–241 Middle Irish 241–245 j Japanese 177 k Khoisan languages 265 l Latin 19–20, 45–46, 104, 110, 139, 202, 218–219, 280–281 Lozi 253, 256, 261 Luganda 260 Lugbara 41 m Macedonian 336 Malagasy 170, 173–174 Marathi 49–52 n Niuean 170 Norse Old Norse 67, 135, 149, 153, 160–161 Middle Norwegian 145–163 Scandinavian 153, 155–157, 160 Nsenga 261 p Portuguese 276–278
Language index r Romance languages 45, 104, 110, 113, 202–203 Romanian 336 Russian 79–80 s Seediq 169–173 Semitic (Ethiopian) 43 siSwati 249–266 Slavonic Old Church Slavonic, see Bulgarian, Old
Spanish 65, 97–115, 203–210, 212–214, 349 Old Spanish 107, 111, 207–210, 212–220 Majorcan Spanish 101–115 Paraguayan Spanish 107 Spanish in the USA 79, 99–100 Swahili 252, 256, 259–260, 263 Swedish 303–317
t Tagalog 174 Toba Batak 170 w Welsh 81–93 y Yiddish 158 z Zulu 262
Subject index
A absolutive case 169–172, 174 acceleration 99–100 accusative case 226–247 acquisition 42, 113, 219–220, 245 L1 2, 13–14, 29, 113, 270, 288–291 L2 2–3, 14, 78, 269–271, 288–291, 293, 299 adverbs, position of 130–131 affective contexts 62, 68–72, 185, 192 see also negative polarity item agreement 86–87, 203, 230–233, 251–265, 342 complementizer agreement 16–17 double agreement 21–23 pronominal agreement 20, 23–25, 64 of syntactic features 68, 184, 193–195, 214–216, 231, 233, 348 Alternative Realization 68–72 antipassive 172, 174 auxiliaries 37–39, 42–48, 83, 86, 122–123, 126–127, 188, 240, 311–314 semi-auxiliaries 99, 105–106 Ayenbite of Inwyt 136 B before-clauses 63 Beowulf 126, 136 Behaghel, Otto 26, 321–2 bilingualism 3, 77–94, 97–115, 192, 352 borrowing 48–54, 108–109, 153–154, 281 Brännström, Edvin 309–310 Braunmüller, Kurt 146, 153–156
C c-command 18, 20–21, 24–25, 172, 177, 184, 186, 189, 193–196, 214–215 Chomsky, Noam 41 Christoffersen, Marit 153, 155–156 clitics 22–23, 67, 71, 99, 173, 216, 257–259, 307, 325–331, 336, 338–341, 346, 350 code-switching 78–82, 85 collocations 84, 111 comparative clause 62–63, 119–140 complementizers 39, 48–54, 123–124, 238–239, 245 see also agreement compounds 84, 323 conditionals 62–63 Constant Rate Hypothesis (CRH) 269–300 see also Kroch contradiction 63–66 control 175 coordination 171, 181–197 copula 99–100, 231–235, 251, 255 convergence 77–94, 336 D definiteness 160–161, 172–174, 335, 344, 347–349 definite article 202–221, 257–260, 277, 335–353 degrammaticalization 259, 264–265, 303–317 deixis 109–110, 341 demonstratives 49, 52, 109–111, 159–160, 202–203, 219, 254, 257–259, 261, 262–264, 338–351 Disterheft, Dorothy 227–234
do-support 139, 240, 274–278, 287–288, 294–298 Driftumkehrung 146, 154 E ellipsis 63–64, 66 emphasis 63–66, 323 ergative case 173–174 Exceptional Case Marking 175–176 expletive articles 208, 211 negation 66–67, 127, 140 subjects 120, 125–126, 140, 255–256 Extended Sisterhood 69 F Final-over-Final Constraint (FOFC) 36–55 focus 29, 51, 71, 129, 172, 204, 225–226, 239–241, 243, 245–246, 341 foot structure 329–330 future tense infinitive + habere (Romance) 45, 99, 104–105 G generic nouns 208–209 genitive case 26–27, 226–227, 229–230, 243, 246, 343 see also possession Goldvarb 150–153 grammaticalization 2, 45, 67, 71, 99, 225–226, 246–247, 250, 254, 256, 303–317, 345 see also degrammaticalization Greenberg, Joseph 35, 41 H Haspelmath, Martin 2, 306–307, 315–316 Hawkins, John 41–42
Subject index head movement 20, 120–124, 174, 240, 344–345, 347–351 Head Parameter 35–37, 289–290 Holmberg, Anders 36–37 I iconicity 67–68 impersonals 64, 125–126, 132–133, 255 Inclusiveness Condition 228 indefinites 65, 69–71, 187–188 inertia 1, 47, 145–146, 158, 161–162 see also Longobardi, Giuseppe infinitives 99, 105, 303 information structure 158–161 inversion 22–23, 82, 119–140, 182, 191, 197 locative inversion 123–124, 251–252, 261–262, 264 stylistic inversion 129–132 Isidor 20, 22–23 Iwein 328–329 J Jäger, Agnes 25–26 Jespersen, Otto 61–62, 181 Jespersen’s Cycle 61–72, 145, 321–331 K Kayne, Richard 15, 36–37, 40–41, 71, 130 Keenan, Edward 145 Kroch, Tony 2, 272–279 L La Somme le Roi 136 language contact 1–4, 8, 13, 42–43, 47–54, 77–94, 97–115, 120, 134–135, 145–163, 178, 192, 264–265, 291–293, 307–308, 335–353 lexical-semantic transfer 108–109 Linear Correspondence Axiom (LCA) 41 linguistic area 48–54, 178 Balkan linguistic area 336–337 locatives 250–265
logistic function 2, 271–300 Longobardi, Giuseppe 1, 30, 47, 145–146, 158, 161–162, 347 see also inertia M Matrix Language Frame model 85, 88, 90, 93 modal verbs 82, 139, 305, 308–314 Monsee Fragments 20 Morpheme Order Principle 85–86, 87, 93 Myers-Scotton, Carol 78, 80–81, 85–86 N negation 25–26, 61–72, 122, 181–197, 205, 287, 295, 321–331 expletive negation 66–67, 127, 140 negative preposing 123–124 negative concord (NC) 25–26, 106–107, 114, 181–197 negative polarity item (NPI) 171–172, 184–186, 189, 311 see also affective contexts Nibelungenlied 325, 328–330 nominal declension 237, 242 nonfinite clauses 175, 225–247 subjects of 226–231 noun classes 249–266 noun phrases 81–82, 84, 92, 158–161, 201–221 null elements 86, 93, 233 negative operator 183–189, 192–193, 195–196 subjects 14–25, 79, 120, 125–126 numerals 160 n-word 65–66, 70, 182–189, 194–197 O obligation 105–6, 311 P palatal lateral approximant 101 pathways of change 42–43 particles 40–41, 53, 83, 226, 244–245
particle verbs 47 question particles 48–49, 50–51 passivization 228–231, 255 Paul, Hermann 15 periphrasis 83, 99, 104–106, phases 41, 184, 194, 202 possession 26–27, 173, 204, 251, 255, 257, 260, 276–278, 292–293, 342 prepositions 111–114, 225–241, 256–260 presupposition 65–66, 174 primary linguistic data (PLD) 42–43, 289 pro-drop 15–25, 29, 126 prosodic structure 322–331 Q quantifiers 159–161 quantifier float 130, 132 questions 51, 63–64, 67, 82–83, 122–123 R reanalysis 1–2, 23–24, 69–72, 197, 226–227, 232–234, 236–238, 242–245, 256–260, 288, 291, 299, 312–316, 342, 348 reconstruction effects 173–174 register variation 100–101, 103–104, 119, 123 reflexive pronoun 175–177, 280–282, 286, 291–293 relative clauses 239, 262–263 Relativized Minimality 41, 229 Roberts, Ian 2, 69, 71, 184, 192, 194, 197, 220, 245, 246–7, 336, 342, 345 Roussou, Anna 2, 69, 71, 220, 245–247, 336, 342, 345 S S-curve 2, 271–300 standard languages 13–15, 25–30, 102, 106, 108, 111–114, 189 subject position 126–129 subordinate clauses 51–54, 123–124, 150–53, 175–176, 187, 228
asymmetry with main clauses 19–20, 46–47 System Morpheme Principle 85, 86–87 T Tatian 18, 20, 23–24 topic 124–125, 130, 169–178, 256, 341, 351 translation influence from French 136–137, 192 influence from Greek 352 influence from Latin 19–20
Subject index V verb clusters 27–29, 37–38 vowels reduction of 323, 325, 328–329 elision of 324–325, 327–328 vowel harmony 324–325 W Williram von Ebersberg 21, 326 word order change 27–29, 43–54, 157–161, 169–178, 347–349, 352–353 disharmonic 35–36, 48 OV 43–48, 148–163, 225–247
SVO 82–83, 169–178 verb-initial 169–178 verb-final 27–28 verb-second 18, 47, 120–126, 134–135, 194, 310 VO 43–48, 146, 148–163 VOS 169–178 VSO 82–83, 92 weakening of preverbal negator 67 of verbal endings 18 of /d/ 101–102 of /s/ 102 Würzburg Glosses 226–227, 229, 231, 234, 236
Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 167 MERCADO, Raphael, Eric POTSDAM and Lisa deMena TRAVIS (eds.): Austronesian and Theoretical Linguistics. vii, 374 pp. + index. Expected October 2010 166 BRANDT, Patrick and Marco GARCÍA GARCÍA (eds.): Transitivity. Form, Meaning, Acquisition, and Processing. vi, 300 pp. + index. Expected September 2010 165 BREUL, Carsten and Edward GÖBBEL (eds.): Comparative and Contrastive Studies of Information Structure. xii, 303 pp. + index. Expected September 2010 164 ZWART, Jan-Wouter and Mark de VRIES (eds.): Structure Preserved. Studies in syntax for Jan Koster. xxiii, 383 pp. + index. Expected August 2010 163 KIZIAK, Tanja: Extraction Asymmetries. Experimental evidence from German. xvi, 271 pp. + index. Expected July 2010 162 BOTT, Oliver: The Processing of Events. xix, 379 pp. + index. Expected September 2010 161 HAAN, Germen J. de: Studies in West Frisian Grammar. Edited by Jarich Hoekstra, Willem Visser and Goffe Jensma. x, 380 pp. + index. Expected August 2010 160 MAVROGIORGOS, Marios: Clitics in Greek. A minimalist account of proclisis and enclisis. x, 286 pp. + index. Expected July 2010 159 BREITBARTH, Anne, Christopher LUCAS, Sheila WATTS and David WILLIS (eds.): Continuity and Change in Grammar. 2010. viii, 359 pp. 158 DUGUINE, Maia, Susana HUIDOBRO and Nerea MADARIAGA (eds.): Argument Structure and Syntactic Relations. A cross-linguistic perspective. 2010. vi, 348 pp. 157 FISCHER, Susann: Word-Order Change as a Source of Grammaticalisation. 2010. ix, 200 pp. 156 DI SCIULLO, Anna Maria and Virginia HILL (eds.): Edges, Heads, and Projections. Interface properties. 2010. vii, 265 pp. 155 SATO, Yosuke: Minimalist Interfaces. Evidence from Indonesian and Javanese. 2010. xiii, 159 pp. 154 HORNSTEIN, Norbert and Maria POLINSKY (eds.): Movement Theory of Control. 2010. vii, 330 pp. 153 CABREDO HOFHERR, Patricia and Ora MATUSHANSKY (eds.): Adjectives. Formal analyses in syntax and semantics. 2010. vii, 335 pp. 152 GALLEGO, Ángel J.: Phase Theory. 2010. xii, 365 pp. 151 SUDHOFF, Stefan: Focus Particles in German. Syntax, prosody, and information structure. 2010. xiii, 335 pp. 150 EVERAERT, Martin, Tom LENTZ, Hannah de MULDER, Øystein NILSEN and Arjen ZONDERVAN (eds.): The Linguistics Enterprise. From knowledge of language to knowledge in linguistics. 2010. ix, 379 pp. 149 AELBRECHT, Lobke: The Syntactic Licensing of Ellipsis. 2010. xii, 230 pp. 148 HOGEWEG, Lotte, Helen de HOOP and Andrej MALCHUKOV (eds.): Cross-linguistic Semantics of Tense, Aspect, and Modality. 2009. vii, 406 pp. 147 GHOMESHI, Jila, Ileana PAUL and Martina WILTSCHKO (eds.): Determiners. Universals and variation. 2009. vii, 247 pp. 146 GELDEREN, Elly van (ed.): Cyclical Change. 2009. viii, 329 pp. 145 WESTERGAARD, Marit: The Acquisition of Word Order. Micro-cues, information structure, and economy. 2009. xii, 245 pp. 144 PUTNAM, Michael T. (ed.): Towards a Derivational Syntax. Survive-minimalism. 2009. x, 269 pp. 143 ROTHMAYR, Antonia: The Structure of Stative Verbs. 2009. xv, 216 pp. 142 NUNES, Jairo (ed.): Minimalist Essays on Brazilian Portuguese Syntax. 2009. vi, 243 pp. 141 ALEXIADOU, Artemis, Jorge HANKAMER, Thomas McFADDEN, Justin NUGER and Florian SCHÄFER (eds.): Advances in Comparative Germanic Syntax. 2009. xv, 395 pp. 140 ROEHRS, Dorian: Demonstratives and Definite Articles as Nominal Auxiliaries. 2009. xii, 196 pp. 139 HICKS, Glyn: The Derivation of Anaphoric Relations. 2009. xii, 309 pp. 138 SIDDIQI, Daniel: Syntax within the Word. Economy, allomorphy, and argument selection in Distributed Morphology. 2009. xii, 138 pp. 137 PFAU, Roland: Grammar as Processor. A Distributed Morphology account of spontaneous speech errors. 2009. xiii, 372 pp.
