epdf pub continuity-and-change-in-grammar-linguistik-aktuel - Elementos de Máquinas (2024)

Prévia do material em texto

<p>Continuity and Change in Grammar</p><p>Volume 159</p><p>Continuity and Change in Grammar</p><p>Edited by Anne Breitbarth, Christopher Lucas, Sheila Watts and David Willis</p><p>General Editors</p><p>Werner Abraham</p><p>University of Vienna /</p><p>Rijksuniversiteit Groningen</p><p>Elly van Gelderen</p><p>Arizona State University</p><p>Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph</p><p>studies into synchronic and diachronic linguistics. Studies in LA confront empirical</p><p>and theoretical problems as these are currently discussed in syntax, semantics,</p><p>morphology, phonology, and systematic pragmatics with the aim to establish robust</p><p>empirical generalizations within a universalistic perspective.</p><p>Linguistik Aktuell/Linguistics Today (LA)</p><p>Advisory Editorial Board</p><p>Josef Bayer</p><p>University of Konstanz</p><p>Cedric Boeckx</p><p>ICREA/Universitat Autònoma de Barcelona</p><p>Guglielmo Cinque</p><p>University of Venice</p><p>Liliane Haegeman</p><p>University of Ghent</p><p>Hubert Haider</p><p>University of Salzburg</p><p>Terje Lohndal</p><p>University of Maryland</p><p>Christer Platzack</p><p>University of Lund</p><p>Ian Roberts</p><p>University of Cambridge</p><p>Lisa deMena Travis</p><p>McGill University</p><p>Sten Vikner</p><p>University of Aarhus</p><p>C. Jan-Wouter Zwart</p><p>University of Groningen</p><p>Continuity and Change</p><p>in Grammar</p><p>Edited by</p><p>Anne Breitbarth</p><p>Ghent University</p><p>Christopher Lucas</p><p>Sheila Watts</p><p>David Willis</p><p>University of Cambridge</p><p>John Benjamins Publishing Company</p><p>Amsterdam / Philadelphia</p><p>Library of Congress Cataloging-in-Publication Data</p><p>Continuity and change in grammar / edited by Anne Breitbarth...[et al.].</p><p>p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166-0829 ; v. 159)</p><p>Includes bibliographical references and index.</p><p>1. Language and languages--Variation. 2. Grammar, Comparative and general. 3. Socio-</p><p>linguistics. I. Breitbarth, Anne, 1976-</p><p>P120.V37C667 2010</p><p>417’.2--dc22 2010014516</p><p>isbn 978 90 272 5542 6 (Hb ; alk. paper)</p><p>isbn 978 90 272 8807 3 (Eb)</p><p>© 2010 – John Benjamins B.V.</p><p>No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any</p><p>other means, without written permission from the publisher.</p><p>John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands</p><p>John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa</p><p>The paper used in this publication meets the minimum requirements of</p><p>American National Standard for Information Sciences – Permanence of</p><p>Paper for Printed Library Materials, ansi z39.48-1984.</p><p>8 TM</p><p>Table of contents</p><p>List of contributors vii</p><p>Introduction: Continuity and change in grammar 1</p><p>Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>part i. Continuity</p><p>What changed where? A plea for the re-evaluation of dialectal evidence 13</p><p>Katrin Axel & Helmut Weiß</p><p>Impossible changes and impossible borrowings:</p><p>The Final-over-Final Constraint 35</p><p>Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Continuity is change: The long tail of Jespersen’s cycle in Flemish 61</p><p>Anne Breitbarth & Liliane Haegeman</p><p>Using the Matrix Language Frame model to measure the extent of</p><p>word-order convergence in Welsh–English bilingual speech 77</p><p>Peredur Davies & Margaret Deuchar</p><p>On language contact as an inhibitor of language change:</p><p>The Spanish of Catalan bilinguals in Majorca 97</p><p>Andrés Enrique-Arias</p><p>Towards notions of comparative continuity in English and French 119</p><p>Remus Gergel</p><p>Variation, continuity and contact in Middle Norwegian</p><p>and Middle Low German 145</p><p>John D. Sundquist</p><p>part ii. Change</p><p>Directionality in word-order change in Austronesian languages 169</p><p>Edith Aldridge</p><p>Negative co-ordination in the history of English 181</p><p>Richard Ingham</p><p>Formal features and the development of the Spanish D-system 201</p><p>Masataka Ishikawa</p><p>The rise of OV word order in Irish verbal-noun clauses 225</p><p>Elliott Lash</p><p>The great siSwati locative shift 249</p><p>Lutz Marten</p><p>The impact of failed changes 269</p><p>Gertjan Postma</p><p>A case of degrammaticalization in northern Swedish 303</p><p>Henrik Rosenkvist</p><p>Jespersen’s Cycle in German from the phonological perspective</p><p>of syllable and word languages 321</p><p>Renata Szczepaniak</p><p>An article on the rise: Contact-induced change and the rise</p><p>and fall of N-to-D movement 335</p><p>Mila Dimitrova-Vulchanova & Valentin Vulchanov</p><p>Language index 355</p><p>Subject index 357</p><p>vi Continuity and Change in Grammar</p><p>Edith Aldridge</p><p>Department of Linguistics</p><p>University of Washington</p><p>A210 Padelford Hall</p><p>Box 354340</p><p>Seattle, WA 98195-4340</p><p>USA</p><p>eca1@u.washington.edu</p><p>Katrin Axel</p><p>Universiteit Göttingen</p><p>Seminar für Deutsche Philologie</p><p>Jacob-Grimm-Haus</p><p>Käte-Hamburger-Weg 3</p><p>D-37073 Göttingen</p><p>Germany</p><p>katrin.axel@phil.uni-goettingen.de</p><p>Theresa Biberauer</p><p>Faculty of Modern and Medieval Languages</p><p>University of Cambridge</p><p>Sidgwick Avenue</p><p>Cambridge</p><p>CB3 9DA</p><p>UK</p><p>mtb23@cam.ac.uk</p><p>Anne Breitbarth</p><p>Universiteit Gent</p><p>Vakgroep Engels</p><p>Rozier 44</p><p>9000 Ghent</p><p>Belgium</p><p>anne.breitbarth@ugent.be</p><p>Peredur Davies</p><p>The Department of Linguistics & English</p><p>Language</p><p>University of Wales</p><p>Bangor</p><p>Gwynedd</p><p>LL57 2DG</p><p>UK</p><p>p.davies@bangor.ac.uk</p><p>Margaret Deuchar</p><p>The Department of Linguistics &</p><p>English Language</p><p>University of Wales</p><p>Bangor</p><p>Gwynedd</p><p>LL57 2DG</p><p>UK</p><p>m.deuchar@bangor.ac.uk</p><p>Andrés Enrique-Arias</p><p>Universitat de les Illes Balears</p><p>Departament de Filologia Espanyola,</p><p>Moderna i Llatina</p><p>Edifici Ramon Llull</p><p>Cra de Valldemossa Km 7,5</p><p>E-07122 Palma (Balears)</p><p>Spain</p><p>andres.enrique@uib.es</p><p>Remus Gergel</p><p>Descriptive & Theoretical Linguistics</p><p>English Department</p><p>University of Tübingen</p><p>Wilhelmstr. 50, R. 407</p><p>D-72074 Tübingen</p><p>Germany</p><p>remus.gergel@uni-tuebingen.de</p><p>Liliane Haegeman</p><p>Universiteit Gent</p><p>Vakgroep Engels</p><p>Rozier 44</p><p>9000 Ghent</p><p>Belgium</p><p>Liliane.Haegeman@UGent.be</p><p>Richard Ingham</p><p>The School of English</p><p>Birmingham City University</p><p>City North Campus</p><p>Perry Barr</p><p>Birmingham</p><p>B42 2SU</p><p>richard.ingham@bcu.ac.uk</p><p>List of contributors</p><p>Masataka Ishikawa</p><p>Sogokagakubu</p><p>Hiroshima University</p><p>Kagamiyama 1-7-1</p><p>Japan 739-8521</p><p>ishkwammgb@hiroshima-u.ac.jp</p><p>Elliott Lash</p><p>Faculty of Modern and Medieval</p><p>Languages</p><p>University of Cambridge</p><p>Sidgwick Avenue</p><p>Cambridge</p><p>CB3 9DA</p><p>UK</p><p>ejfl2@cam.ac.uk</p><p>Lutz Marten</p><p>Department of Africa</p><p>School of Oriental and African Studies</p><p>Thornhaugh Street</p><p>Russell Square</p><p>London WC1H 0XG</p><p>UK</p><p>lm5@soas.ac.uk</p><p>Glenda Newton</p><p>Faculty of Modern and Medieval</p><p>Languages</p><p>University of Cambridge</p><p>Sidgwick Avenue</p><p>Cambridge</p><p>CB3 9DA</p><p>UK</p><p>gen21@cam.ac.uk</p><p>Gertjan Postma</p><p>Meertens Instituut</p><p>Postbus 94264</p><p>1090 GG Amsterdam</p><p>Gertjan.Postma@meertens.knaw.nl</p><p>Henrik Rosenkvist</p><p>Centre for Language and Literature</p><p>Lund University</p><p>Box 201</p><p>221 00 Lund</p><p>Sweden</p><p>Henrik.Rosenkvist@nordlund.lu.se</p><p>Michelle Sheehan</p><p>School of English Literature, Language and</p><p>Linguistics</p><p>Percy Building</p><p>University of Newcastle upon Tyne</p><p>Newcastle upon Tyne</p><p>NE1 7RU</p><p>UK</p><p>michelle.sheehan@ncl.ac.uk</p><p>John Sunquist</p><p>Department of Foreign Languages</p><p>640 Oval Drive</p><p>Stanley Coulter 170</p><p>West Lafayette, IN 47907</p><p>USA</p><p>sundquist@purdue.edu</p><p>Renata Szczepaniak</p><p>Institut für Germanistik I</p><p>Von-Melle-Park 6</p><p>20146 Hamburg</p><p>Germany</p><p>renata.szczepaniak@uni-hamburg.de</p><p>Valentin Vulchanov</p><p>Institutt for moderne fremmedspråk</p><p>NTNU</p><p>7491 Trondheim</p><p>Norway</p><p>valentin.vulchanov@hf.ntnu.no</p><p>Mila Vulchanova</p><p>Institutt for moderne fremmedspråk</p><p>NTNU</p><p>7491 Trondheim</p><p>Norway</p><p>mila.vulchanova@hf.ntnu.no</p><p>Helmut Weiß</p><p>Institut für Kognitive Linguistik</p><p>Johann Wolfgang Goethe-Universität</p><p>Grüneburgplatz 1 (Fach 161)</p><p>60629 Frankfurt a. M. (Briefe)</p><p>60323 Frankfurt a. M. (Pakete)</p><p>Germany</p><p>weiss@lingua.uni-frankfurt.de</p><p>viii Continuity and Change in Grammar</p><p>Introduction</p><p>Continuity and change in grammar</p><p>Anne Breitbarth, Christopher Lucas,</p><p>Sheila Watts & David Willis</p><p>Ghent University/University of Cambridge</p><p>The present volume brings together 16 contributions selected from papers presented</p><p>at the conference on Continuity and Change in Grammar that took place at the Uni-</p><p>versity of Cambridge 18–20 March 2008. The aim of the conference was to foster an</p><p>exchange of ideas on various aspects of linguistic transmission in different frameworks</p><p>(e.g. generative vs. functionalist)</p><p>a higher frequency of extraposition than modern Standard German</p><p>(e.g. Patocka 1997: 320–358; Weiß 1998: 55–58 on Bavarian).</p><p>1.  The maximal superordinate verb, i.e. the finite verb, receives the index ‘1’ (= V1), the verb</p><p>governed by it receives the index ‘2’ (= V2) and so forth.</p><p> Katrin Axel & Helmut Weiß</p><p>In general, deviations do not occur in bipartite verbal complexes as in (21). They are</p><p>only possible in verbal clusters with three or more verbs, but even here there are many</p><p>restrictions regarding the question as to which types of verbs may invert from the</p><p>canonical order, which exact orders are allowed and so on. It is a well-known fact that</p><p>in earlier stages of the language there was much more variation in verbal clusters and</p><p>that in bipartite clusters the inverted order (V1 > V2) was also possible (see Robinson</p><p>1997; Axel 2007: ch. 2; Weiß to appear on Old High German; Paul 2007: 453–456 on</p><p>Middle High German; Ebert et al. 1993: 438–440 on Early New High German). The</p><p>decisive period for verb-order change in the verbal cluster was the Early New High</p><p>German period (1350–1650). During this period there was a general decline of the</p><p>inverted order (Härd 1981; Maurer 1926: 151; Ebert 1981, 1998; Bies 1996; Reifsnyder</p><p>2003; Sapp 2006). In more recent studies there has been growing consensus that this</p><p>is a ‘change from above’, passed down from the chancery style (Ebert 1981, 1998; Bies</p><p>1996). Reifsnyder (2003), in her study of the Early New High German dialect of</p><p>Augsburg, comes to the conclusion that there was a prevalence of V2 > V1 orders in</p><p>the official texts, as well as an increase in frequency over time, which she argues results</p><p>from the adoption of a standard language ideology. To our knowledge there have been</p><p>no studies to date looking at when the inverted order became completely ungrammati-</p><p>cal in bipartite clusters in the standard language.</p><p>However, in most modern dialects bipartite verbal clusters may still be realized</p><p>with inverted order (V1 > V2) and tripartite clusters show orders that are not possible</p><p>in Standard German, as is illustrated by the Austrian German examples in (22) (cf. also</p><p>Lötscher 1978; Wurmbrand 2004 on Swiss German; Weiß 1998 on Bavarian; Wurmbrand</p><p>2004 on Austrian dialects):</p><p>(22) a. dea den Gaia håt åhagschossn (Styria)</p><p>rel the hawk has1 down-shot2</p><p>‘who shot down the hawk.’</p><p>b. wånn a Kind is taaft gwoadn (Styria)</p><p>when a child is baptized-ppp aux-ppp2</p><p>‘when a child was baptized.’ (Patocka 1997: 290)</p><p>Sapp (2006) carried out controlled experiments (magnitude estimation) on verbal</p><p>clusters in some contemporary dialects (mostly on Austrian and Swabian German)</p><p>and compared these findings and those from an extensive survey of the secondary</p><p>literature on various dialects with the results of his quantitative investigation of the</p><p>texts from the online version of the Bonner Frühneuhochdeutschkorpus. He comes</p><p>to the conclusion (p. 170) that although “many dialects appear to have the same</p><p>kind of word-order variation as ENHG [Early New High German], there have also</p><p>been some important historical developments”. With bipartite clusters, the Early</p><p>New High German state of affairs is best preserved in Swiss German, which still</p><p>What changed where? </p><p>allows the V1 > V2 and the V2 > V1 order, while in Swabian and Austrian and most</p><p>other dialects the former order is disprefered. In Early New High German, the</p><p>V1 > V2 order is better with the modal–infinitive syntagm than with the perfect tense.</p><p>Interestingly, however, other dialects, namely Swabian and Eastern Austrian, have</p><p>developed the reverse preference. With tripartite clusters, most dialects have come to</p><p>agree with Standard German in allowing only V3 > V2 > V1, however, as in Early New</p><p>High German, in the IPP (= infintivus pro participio) and werden–modal–infinitive</p><p>constructions (cf. Schmid & Vogel 2004; Schmid 2005) there is more variation than</p><p>in Standard German. In both bi- and tripartite clusters, focus has some effect on verb</p><p>order in Early New High German. More precisely, Sapp (2006: 168) has found that</p><p>focus on the object has a favouring effect on the V1 > V2 order and probably on the V1</p><p>> V3 > V2 order. Interestingly, the same effect shows up in two contemporary dialects</p><p>which he studied experimentally (i.e. Swabian and Austrian German) and in Modern</p><p>Standard German. To sum up, Sapp’s comparison of the contemporary dialects with</p><p>the Early New High German dialects has revealed that the diachrony of verbal clus-</p><p>ters is characterized by both syntactic continuity and discontinuity. Compared to the</p><p>developments in the standard languages, the dialects seem to be more conservative</p><p>and reluctant to change at first. However, Sapp’s study suggests that the microvariation</p><p>we still find in the contemporary dialects is no direct reflex of the situation in the Early</p><p>New High German dialects, but it must be the product of some small-scale changes</p><p>(‘micro-changes’).</p><p>.  Conclusion</p><p>All these case studies show that taking into account dialectal evidence in historical</p><p>syntax can lead to different and more appropriate results. These results constitute a</p><p>more solid basis for judgements about what is stable and what is more open to change</p><p>in the history of a language.</p><p>There are especially three aspects speaking in favour of dialectal data in historical</p><p>syntax:20</p><p>First: if there were different historical developments in dialects and the standard, the</p><p>development in the dialects should be more relevant, because L1 acquisition can be</p><p>.  There is another aspect speaking in favour of dialects in historical linguistics: the written</p><p>(variety of a) language is not always based on the same variety/dialect in the historical develop-</p><p>ment. English exemplifies this lack of continuity: whereas the corpus of Old English is written</p><p>in the Wessex dialect, modern Standard English developed from the Mercian dialect (Crystal</p><p>1995: 29). Thanks to one of the reviewers for bringing this aspect to our attention.</p><p> Katrin Axel & Helmut Weiß</p><p>made responsible for them, whereas standard languages were shaped by extra-linguistic</p><p>factors as well; see, for example, the loss of negative concord.</p><p>Second: some of the innovations occurring in dialects have never reached the stan-</p><p>dard and these are mostly the interesting cases for linguistics. We mentioned one such</p><p>case: complementizer agreement (cf. Corbett 2006 for the typological uniqueness of</p><p>this feature). Other cases from German dialects, among many others, would be verb</p><p>doubling in Swiss German (Schönenberger & Penner 1995) or matrix verb stranding</p><p>in Swabian (Hiller 1999).21</p><p>Third: Just as dialects exhibit syntactic variation at the synchronic level, they do so</p><p>at the diachronic level as well. The German verbal clusters show considerable word-</p><p>order variation both in the contemporary and in the Early New High German dialects,</p><p>but there is no direct correspondence between the situation in the modern dialects</p><p>and their respective Early New High German precursors. Though syntax is surely the</p><p>most inert linguistic level with respect to change (Longobardi 2001), the syntactic sys-</p><p>tems of the dialects must be the product of many ‘micro-changes’ that have shaped the</p><p>microvariation that we see today.</p><p>Primary sources</p><p>[I] Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten.</p><p>Hans Eggers (ed.). Tübingen: Niemeyer. 1964.</p><p>[Lanc] Lancelot. Nach der Heidelberger Pergamenthandschrift Pal. Germ. 147, Reinhold Kluge</p><p>(ed.). Berlin: Akademie-Verlag. Vol. I 1948.</p><p>[MF] The Monsee Fragments. Newly collated text with introduction, notes, grammatical treatise</p><p>and exhaustive glossary and a photo-lithographic fac-simile. George Allison Hench (ed.).</p><p>Straßburg: Karl J. Trübner. 1890.</p><p>[T] Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Achim Masser</p><p>(ed.) (in cooperation with Elisabeth De Felip-Jaud). Vandenhoeck & Ruprecht. 1994.</p><p>References</p><p>Axel,</p><p>Katrin. 2005. Null subjects and verb placement in Old High German. In Linguistic Evidence.</p><p>Empirical, Theoretical and Computational Perspectives, Stephan Kepser & Marga Reis (eds),</p><p>27–48. Berlin: Mouton de Gruyter.</p><p>1.  Verbs like to help/let/start and a few others behave curiously in Swabian in that it is the</p><p>embedded verb they select that is inflected, whereas they themselves appear as infinitives,</p><p>cf. (i). This is called Matrix Verb Stranding by Hiller (1999).</p><p>(i) Se šreibt des ets ãafangә uf</p><p>she write-3.sg this-acc now begin-inf down</p><p>‘She is now starting writing this down’</p><p>What changed where? 1</p><p>Axel, Katrin. 2007. Studies on Old High German Syntax: Left Sentence Periphery, Verb Placement</p><p>and Verb-Second [Linguistik Aktuell/Linguistics Today 112]. Amsterdam: John Benjamins.</p><p>Axel, Katrin & Weiß, Helmut. To appear. Pro-drop in the history of German: From Old High</p><p>German to the modern dialects. In Empty Pronouns, Peter Gallmann & Melanie Wratil (eds).</p><p>Berlin: Mouton de Gruyter.</p><p>Bayer, Josef. 1984. COMP in Bavarian Syntax. The Linguistic Review 3: 209–274.</p><p>Bayer, Josef. 1993. ‘Zum’ in Bavarian and scrambling. In Dialektsyntax, Werner Abraham &</p><p>Josef Bayer (eds), 50–70. Opladen: Westdeutscher Verlag. (Linguistische Berichte, Sonderheft</p><p>5/1993).</p><p>Bies, Ann. 1996. Syntax and Discourse Factors in Early New High German: Evidence for Verb-</p><p>final Word Order. MA thesis, University of Pennsylvania, Philadelphia.</p><p>Benincà, Paula & Poletto, Cecilia. 2007. The ASIS enterprise: A view on the construction of a</p><p>syntactic atlas for the Northern Italian Dialects. In Nordlyd 34: Scandinavian Dialect Syntax</p><p>2005, Kristine Bentzen & Øystein Alexander Vangsnes (eds), 35–52. Tromsø: CASTL.</p><p>Behaghel, Otto. 1923. Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 2: Die Wortklassen</p><p>und Wortformen. Heidelberg: Winter.</p><p>Behaghel, Otto. 1928. Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 3: Die Satzgebilde.</p><p>Heidelberg: Winter.</p><p>Braune, Wilhelm. 2004. Althochdeutsche Grammatik I, 15th edn, by Ingo Reiffenstein. Tübingen:</p><p>Niemeyer.</p><p>Corbett, Greville G. 2006. Agreement. Cambridge: CUP.</p><p>Cornips, Leonie & Poletto, Cecilia. 2005. On standardising syntactic elicitation techniques,</p><p>Part 1. Lingua 115: 939–957.</p><p>Cooper, Kathrin. 1995. Null subjects and clitics in Zurich German. In Topics in Swiss German</p><p>Syntax, Zvi Penner (ed.), 59–71. Bern: Peter Lang.</p><p>Crystal, David. 1995. The Cambridge Encyclopedia of the English Language. Cambridge: CUP.</p><p>DeGraff, Michel. 1999. Language Creation and Language Change: Creolization, Diachrony, and</p><p>Development. Cambridge MA: The MIT Press.</p><p>Durrell, Martin. 1999. Standardsprache in England und Deutschland. Zeitschrift für Germanis-</p><p>tische Linguistik 27: 285–308.</p><p>Ebert, Robert Peter. 1981. Social and stylistic variation in the order of auxiliary and non-finite</p><p>verb in dependent clauses in Early New High German. Beiträge zur Geschichte der deutschen</p><p>Sprache und Literatur 103: 204–237.</p><p>Ebert, Robert Peter. 1998. Verbstellungswandel bei Jugendlichen, Frauen und Männern im 16.</p><p>Jahrhundert. Tübingen: Niemeyer.</p><p>Ebert, Robert Peter, Reichmann, Oskar, Solms, Hans-Joachim & Wegera, Klaus-Peter. 1993.</p><p>Frühneuhochdeutsche Grammatik. Tübingen: Niemeyer.</p><p>Eggenberger, Jakob. 1961. Das Subjektspronomen im Althochdeutschen. Chur: Selbstverlag.</p><p>Franck, Johannes. 1971. Altfränkische Grammatik. Laut- und Flexionslehre, 2nd edn, by Rudolf</p><p>Schützeichel. Göttingen: Vandenhoeck & Ruprecht.</p><p>Fuß, Eric. 2005. The Rise of Agreement. A Formal Approach to the Syntax and Grammaticalization</p><p>of Verbal Inflection [Linguistik Aktuell/Linguistics Today]. Amsterdam: John Benjamins.</p><p>Fuß, Eric. 2008. Multiple agreement and the representation of inflection in the C-domain.</p><p>Linguistische Berichte 213: 77–106.</p><p>Grimm, Jacob. 1967. Deutsche Grammatik, Vol. 4. (4th Reprographic reprint of the edition</p><p>Gütersloh 1898). Hildesheim: Georg Olms.</p><p>Haag-Merz, Christine. 1996. Pronomen im Schwäbischen: Syntax und Erwerb. Marburg:</p><p>Tectum-Verlag.</p><p> Katrin Axel & Helmut Weiß</p><p>Härd, John Evert. 1981. Studien zur Struktur mehrgliedriger deutscher Nebensatzprädikate</p><p>Diachronie und Synchronie [Göteborger germanistische Forschungen 21]. Göteborg: Acta</p><p>Universitatis Gothoburgensis.</p><p>Held, Karl. 1903. Das Verbum ohne pronominales Subjekt in der älteren deutschen Sprache.</p><p>Berlin: Mayer & Müller.</p><p>Hiller, Markus. 1999. Violable relativized minimality. Ms, Rutgers University.</p><p>Hopper, Paul J. 1975. The Syntax of the Simple Sentence in Proto-Germanic. The Hague: Mouton.</p><p>Jäger, Agnes 2008. History of German Negation [Linguistik Aktuell/Linguistics Today 118].</p><p>Amsterdam: John Benjamins.</p><p>Jaeggli, Osvaldo & Safir, Kenneth J. 1989. The null-subject parameter and parametric theory.</p><p>In The Null Subject Parameter, Osvaldo Jaeggli & Kenneth J. Safir (eds), 1–44. Dordrecht:</p><p>Kluwer.</p><p>Kayne, Richard. 1996. Microparametric syntax: Some introductory remarks. In Microparametric</p><p>Syntax and Dialect Variation [Current Issues in Linguistic Theory 139], James R. Black &</p><p>Virginia Motapanyane (eds), ix-xviii, Amsterdam: John Benjamins.</p><p>Koeneman, Olaf. 2007. Deriving the difference between full and partial pro-drop. In Arguments</p><p>and Agreement, Peter Ackema, Patrick Brandt, Maaike Schoorlemmer & Fred Weerman</p><p>(eds), 76–100. Oxford: OUP.</p><p>Kögel, Rudolf. 1882. Zum deutschen Verbum. Beiträge zur Geschichte der deutschen Sprache und</p><p>Literatur 8: 126–139.</p><p>Kroch, Anthony & Taylor, Ann. 1997. Verb movement in Old and Middle English: Dialect varia-</p><p>tion and language contact. In Parameters of Morphosyntactic Change, Ans van Kemenade &</p><p>Nigel Vincent (eds), 297–325. Cambridge: CUP.</p><p>Kroch, Anthony. 2000. Syntactic change. In The Handbook of Contemporary Syntactic Theory,</p><p>Mark Baltin & Chris Collins (eds.), 629–739. Malden MA: Blackwells.</p><p>Ledgeway, Adam. 2000. A Comparative Syntax of the Dialects of Southern Italy: A Minimalist</p><p>Approach. Oxford: Blackwell.</p><p>Lenerz, Jürgen. 1984. Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an</p><p>Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer.</p><p>Lötscher, Andreas. 1978. Zur Verbstellung im Zürichdeutschen und in anderen Varianten des</p><p>Deutschen. Zeitschrift für Dialektologie und Linguistik 45: 1–29.</p><p>Longobardi, Guiseppe. 2001. Formal syntax, diachronic minimalism, and etymology. The History</p><p>of French chez. Linguistic Inquiry 32: 275–302.</p><p>Maurer, Friedrich. 1926. Untersuchungen über die deutsche Verbstellung in ihrer geschichtlichen</p><p>Entwicklung. Heidelberg: Winter.</p><p>Meillet, Antoine. 1908/09. Notes sur quelques faits gotiques. Mémoires de la Société de Linguistique</p><p>de Paris 15: 73–103.</p><p>Müller, Gereon. 2006. Pro-drop and impoverishment. In Form, Structure, and Grammar. A Fest-</p><p>schrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, Patrick Brandt &</p><p>Eric Fuß (eds), 93–115. Berlin: Akademie Verlag.</p><p>Patocka, Franz. 1997. Satzgliedstellung in den bairischen Dialekten Österreichs. Frankfurt:</p><p>Peter Lang.</p><p>Paul, Hermann. 1877. Die Vocale der Flexions- und Ableitungssilben in den ältesten germanischen</p><p>Dialecten. Beiträge zur Gechichte der deutschen Sprache und Literatur 4: 315–475.</p><p>Paul, Hermann. 1880. Prinzipien der Sprachgeschichte. Tübingen: Niemeyer.</p><p>Paul, Hermann. 1919. Deutsche Grammatik, Vol. III. Tübingen: Niemeyer.</p><p>What changed where? </p><p>Paul, Hermann. 2007. Mittelhochdeutsche Grammatik, 25th edn by Thomas Klein, Hans-Joachim</p><p>Solms, Klaus-Peter Wegera, Ingeborg Schröbler and Heinz-Peter Prell. Tübingen: Niemeyer.</p><p>Poletto, Cecilia. 1995. The diachronic development of subject clitics in North-Eastern Italian</p><p>dialects. In Clause Structure and Language Change, Adrian Battye & Ian G. Roberts (eds),</p><p>295–324. Oxford: OUP.</p><p>Reifsnyder, Kristen. 2003. Vernacular versus Emerging Standard: An Examination of Dialect</p><p>Usage in Early Modern Augsburg (1500–1650). Ph.D. dissertation, University of Wisconsin,</p><p>Madison.</p><p>Robinson, Orrin W. 1997. Clause Subordination and Verb Placement in the Old High German</p><p>Isidor Translation.</p><p>Heidelberg: Winter.</p><p>Rohrbacher, Bernhard Wolfgang. 1999. Morphology-Driven Syntax: A Theory of V to I Raising</p><p>and Pro-Drop [Linguistik Aktuell/Linguistics Today 15]. Amsterdam: John Benjamins.</p><p>Sapp, Christopher. 2006. Verb Order in Subordinate Clauses: From Early New High German to</p><p>Modern German. Ph.D. dissertation, Indiana University.</p><p>Schatz, Josef. 1907. Altbairische Grammatik. Laut- und Flexionslehre. Göttingen: Vandenhoeck &</p><p>Ruprecht.</p><p>Schmid, Tanja & Vogel, Ralf. 2004. Dialectal variation in German 3-verb-clusters. A surface-</p><p>oriented Optimality Theoretic account. Journal of Comparative Germanic Linguistics 7:</p><p>235–274.</p><p>Schmid, Tanja. 2005. Infinitival Syntax: Infinitivus Pro Participio as a Repair Strategy [Linguistik</p><p>Aktuell/Linguistics Today 79]. Amsterdam: John Benjamins.</p><p>Schönenberger, Manuela & Penner, Zvi. 1995. Cross-dialectal variation in Swiss German:</p><p>Doubling verbs, verb projection raising, barrierhood, and LF-Movement. In Studies in</p><p>Comparative Germanic Syntax, Hubert Haider, Susan Olsen & Sten Vikner (eds), 285–305.</p><p>Dordrecht: Kluwer</p><p>Sprouse, Rex A. & Vance, Barbara. 1999. An explanation for the decline of null pronouns in</p><p>certain Germanic and Romance languages. In Language Creation and Language Change:</p><p>Creolization, Diachrony, and Development, Michel DeGraff (ed.), 257–284. Cambridge MA:</p><p>The MIT Press.</p><p>Weise, Oskar. 1900. Syntax der Altenburger Mundart. Leipzig: Breitkopf und Härtel.</p><p>Weise, Oskar 1907. Die sogenannte Flexion der Konjunktionen. Zeitschrift für deutsche Mundarten</p><p>2: 199–205.</p><p>Weiß, Helmut. 1998. Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache.</p><p>Tübingen: Niemeyer.</p><p>Weiß, Helmut. 2001. On two types of natural languages. Some consequences for linguistics.</p><p>Theoretical Linguistics 27: 87–103.</p><p>Weiß, Helmut. 2004a. Zum linguistischen Status von Standardsprachen. In Indogermanistik,</p><p>Germanistik, Linguistik, Akten der Arbeitsstagung der Indogermanischen Gesellschaft, Jena</p><p>18.–20.9.2002, Maria Kozianka, Rosemarie Lühr & Susanne Zeilfelder (eds), 591–643.</p><p>Hamburg: Verlag Dr. Kovač.</p><p>Weiß, Helmut. 2004b. A question of relevance: Some remarks on standard languages (Author’s</p><p>response). In What Counts as Evidence in Linguistics? The Case of Innateness [Studies in</p><p>Language 28(3)], Martina Penke & Anette Rosenbach (eds), 648–674. Amsterdam: John</p><p>Benjamins. (Also published as Benjamins Current Topics 7, 2007).</p><p>Weiß, Helmut. 2005a. Inflected complementizers in Continental West Germanic Dialects.</p><p>Zeitschrift für Dialektologie und Linguistik 72: 148–166.</p><p> Katrin Axel & Helmut Weiß</p><p>Weiß, Helmut. 2005b. The double competence hypothesis. On diachronic evidence. In Linguistic</p><p>Evidence: Empirical, Theoretical, and Computational Perspectives, Stephan Kepser & Marga</p><p>Reis (eds), 557–575. Berlin: Mouton de Gruyter.</p><p>Weiß, Helmut. 2005c. Von den vier Lebensaltern einer Standardsprache. Zur Rolle von</p><p>Spracherwerb und Medialität. Deutsche Sprache 33: 289–307.</p><p>Weiß, Helmut. 2008. The possessor that appears twice? Variation, structure and function of</p><p>possessive doubling in German. In Microvariation in Syntactic Doubling in German, Sjef</p><p>Barbiers, Olaf Koeneman, Marika Lekakou, Margreet van der Ham (eds), 381–401. Bingley:</p><p>Emerald.</p><p>Weiß, Helmut. 2009. How to define dialect and language. A proposal for further discussion.</p><p>Linguistische Berichte 219: 251–270.</p><p>Weiß, Helmut. To appear. Die rechte Peripherie im Althochdeutschen. Zur Verbstellung in</p><p>dass-Sätzen. In Tagungsakten der Arbeitstagung der Indogermanischen Gesellschaft.</p><p>Wurmbrand, Susi. 2004. West Germanic verb clusters: The empirical domain. In Verb Clusters:</p><p>A Study of Hungarian, German, and Dutch [Linguistik Aktuell/Linguistics Today 69],</p><p>Katalin É. Kiss & Henk van Riemsdijk (eds), 43–85. Amsterdam: John Benjamins.</p><p>Impossible changes and impossible borrowings</p><p>The Final-over-Final Constraint*</p><p>Theresa Biberauer1,3, Michelle Sheehan2 & Glenda Newton1</p><p>1University of Cambridge/2University of Newcastle/3Stellenbosch University</p><p>This chapter examines the predictions of Biberauer, Holmberg & Roberts’s</p><p>(2007, 2008) Final-over-Final Constraint/FOFC for grammatical change and</p><p>borrowing. As a putatively invariant syntactic principle, FOFC excludes the</p><p>synchronic possibility of head-final phrases dominating categorially alike</p><p>head-initial phrases. For diachrony, FOFC predicts certain word-order changes</p><p>to be impossible, whether contact is involved or not: specifically, head-final to</p><p>head-initial change is predicted to proceed top-down, whereas the reverse</p><p>change should proceed bottom-up. Case studies from the history of English,</p><p>Afrikaans and French seem to support the first of these predictions. Furthermore,</p><p>we show on the basis of data from South Asian languages, that the presence of a</p><p>phrase-initial head blocks the borrowing of a higher phrase-final head, thereby</p><p>avoiding a FOFC-violation.</p><p>1.  Introduction</p><p>In word-order terms, the languages of the world can be classified as being either</p><p>harmonic or disharmonic. The term “harmonic”, originating with Greenberg (1963)</p><p>(cf. also Hawkins 1983), refers to a language that is either consistently head-initial</p><p>or consistently head-final. Within a generative framework, patterns of this type have</p><p>*Our thanks to the audiences at the Continuity and Change Conference (Cambridge – March</p><p>2008), the “Past Meets the Present: A Dialogue Between Historical Linguistics and Theoretical</p><p>Linguistics” Conference (Taipei – July 2008) and DiGS X (Cornell – August 2008), and to the</p><p>members of the FOFC seminar (October-December 2007). Particular thanks to Edith Aldridge,</p><p>Alastair Appleton, Özgür Aydın, Silvio Cruschina, Matthew Dryer, Jim Huang, Tony Kroch,</p><p>Elliott Lash, Adam Ledgeway, Pino Longobardi, Iain Mobbs, Waltraud Paul, Chris Reintges,</p><p>Sarah Thomason, Nigel Vincent, John Whitman and two very helpful anonymous reviewers.</p><p>This work is supported by AHRC Grant No. AH/E009239/1 (“Structure and Linearization in</p><p>Disharmonic Word Orders”).</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>traditionally been related to the Head Parameter, which may be informally stated as</p><p>in (1):</p><p>(1) The Head Parameter: X > YP (head-initial languages)</p><p>YP > X (head-final languages)</p><p>However, a significant number of languages are not fully harmonic; instead they con-</p><p>tain a mixture of head-initial and head-final phrases. A well-known example of a</p><p>disharmonic language is German, which has a head-final VP and AuxP,1 but a head-</p><p>initial CP, DP and, for the most part, PP, as in (2):</p><p>(2) … dass das Mädchen in München gewohnt hat</p><p>[CP that [AuxP[VP[DP the girl] [PP in Munich] lived] has]]</p><p>‘… that the girl lived in Munich.’</p><p>The existence of disharmonic word orders indicates that the Head Parameter must be</p><p>formulated in such a way that it can be set not just in an across-the-board fashion, but</p><p>also, where necessary, in a category-specific way (cf. Hawkins 1983 for discussion).</p><p>This leads us to expect that disharmonic orders will be “equal” in the sense that all</p><p>combinations of mixed headedness should, in principle, be equally available. In reality,</p><p>however, the empirical record exhibits a striking skewing regarding the attestation of</p><p>disharmonic word orders. The following section introduces a generalization capturing</p><p>this skewing and then presents data supporting it, along with some apparent counter-</p><p>examples. Section 3 considers the diachronic predictions made by the generalization</p><p>in terms of word-order change, Section 4 focuses on the implications for borrowing</p><p>and Section 5 concludes.</p><p>2.  The Final-over-Final Constraint</p><p>Holmberg (2000: 124) observes that the following configuration seems to be banned</p><p>in many languages:2</p><p>1.  We employ the label AuxP here despite the fact that it has often been argued that German</p><p>auxiliaries are in fact Vs (cf. Haider 1993, 2000; Müller & Reis 2001). This labeling should</p><p>not be interpreted as signifying a commitment to an opposing analysis; here and elsewhere,</p><p>we simply</p><p>employ AuxP as a convenient, theory-neutral descriptive label designating the</p><p>position(s) above the lexical VP that may be occupied by auxiliaries, not only in German, but</p><p>in languages more generally.</p><p>2.  The tree in (3) shows head-final orders as base-generated by a head parameter. Under a</p><p>Kaynean view, head-final orders are derived via roll-up movement of a head’s complement to</p><p>Impossible changes and impossible borrowings </p><p>(3) * β′</p><p>αP β</p><p>α γP</p><p>where αP is the complement of β and γP is the complement of α</p><p>He accounts for this gap by postulating the following constraint:</p><p>(4) The Final-over-Final Constraint (FOFC) [First Version]:</p><p>If α is a head-initial phrase and β is a phrase immediately dominating α, then β</p><p>must be head-initial. If α is a head-final phrase, and β is a phrase immediately</p><p>dominating α, then β can be head-initial or head-final.</p><p>Evidence of the existence of a constraint of the type in (4) comes from a range of</p><p>clausal and non-clausal contexts in unrelated languages. Thus, for example, Holmberg</p><p>observes that in Finnish all permutations of the verb, object and auxiliary occur, except</p><p>for FOFC-violating V-O-AUX (V underlined; O in smallcaps; AUX in bold):</p><p>(5) a. Milloin Jussi olisi kirjoittanut romaanin? [Aux-V-O]</p><p>when Jussi would-have written novel-def</p><p>b. Milloin Jussi olisi romaanin kirjoittanut? [Aux-O-V]</p><p>when Jussi would-have novel-def written</p><p>c. Milloin Jussi romaanin kirjoittanut olisi? [O-V-Aux]</p><p>when Jussi novel-def written would-have</p><p>‘When would Jussi have written a novel?’</p><p>d. *Milloin Jussi kirjoittanut romaanin olisi? *[V-O-Aux]</p><p>when Jussi written novel-def would-have</p><p>The same pattern emerges if we consider both present-day and historical variet-</p><p>ies of Germanic. We illustrate for Old English (formatting as above; cf. Biberauer,</p><p>its specifier position (cf. Kayne 1994). In this chapter, we abstract away from such theoretical</p><p>details. These are, however, addressed in Biberauer, Newton & Sheehan (2009), where it is</p><p>also made clear that a head parameter-based approach is untenable. For the purposes of this</p><p>chapter, however, it should be assumed that FOFC violations surface under two structural</p><p>conditions: (i) wherever a head-initial phrase is dominated by a head-final one that has been</p><p>merged on top of it, and (ii) wherever a head-initial phrase is dominated by a head-final one as</p><p>the result of a linearization-driven movement operation (see also Note 3). In Kaynean terms,</p><p>all FOFC violations will, of course, have the structure in (ii).</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Holmberg & Roberts 2007 for more detailed discussion and illustration; and cf. Pintzuk</p><p>1991/1999 on the absence of SVOAux throughout the history of English):</p><p>(6) O V AUX (“head-final” order in VP and AuxP):</p><p>Đa se Wisdom þa is fitte asungen hæfde …</p><p>when the Wisdom then this poem sung had</p><p>‘When Wisdom had sung this poem …’</p><p>(Boethius 30.68.6; cf. Fischer et al. 2000: 143)</p><p>(7) O AUX V (“verb-raising”; cf. Evers 1975 and subsequent work):</p><p>… þe æfre on gefeohte his handa wolde afylan</p><p>who ever in battle his hands would defile-inf</p><p>‘… whoever would defile his hands in battle’</p><p>(Ælfric’s Lives of Saints 25.858; cf. Pintzuk 1991: 102)</p><p>(8) AUX O V (“verb projection raising”; cf. Haegeman & van Riemsdijk 1986):</p><p>… þæt hie mihton swa bealdlice Godes geleafan bodian</p><p>that they could so boldly God’s faith preach-inf</p><p>‘…that they could preach God’s faith so boldly’</p><p>(The Homilies of the Anglo-Saxon Church I 232; cf. van Kemenade 1987: 179)</p><p>(9) V AUX O (“object extraposition”):</p><p>… þæt ænig mon atellan mæge ealne one demm</p><p>that any man relate can all the misery</p><p>‘… that any man can relate all the misery’</p><p>(Orosius 52.6–7; cf. Pintzuk 2002: 283)</p><p>(10) AUX V O (“verb raising” combined with “object extraposition”):</p><p>… þæt he mot ehtan godra manna</p><p>that he might persecute good men</p><p>‘… that he might persecute good men’</p><p>(Wulfstan’s Homilies 130.37–38; cf. Pintzuk 2002: 282)</p><p>Crucially, every permutation of Aux, V and O is attested, except for the FOFC-violating</p><p>V-O-AUX order (see i.a. den Besten 1986; Travis 1984: 157–8; Kiparsky 1996: 168–171;</p><p>Pintzuk 1991/1999; Hróarsdóttir 2000; Fuß & Trips 2002 for discussion).</p><p>Haddican (2004) observes that the same gap occurs in Basque:</p><p>(11) a. Jon-ek ez dio Miren-i egia esan [Aux-O-V]</p><p>Jon-erg not aux Miren-dat truth say-perf</p><p>‘Jon has not told Miren the truth.’</p><p>b. Jon-ek ez dio esan Miren-i egia [Aux-V-O]</p><p>Jon-erg not aux say-perf Miren-dat truth</p><p>‘Jon has not told Miren the truth.’</p><p>Impossible changes and impossible borrowings </p><p>(12) a. Jon-ek Miren-i egia esan dio [O-V-Aux]</p><p>Jon-erg Miren-dat truth say-perf aux</p><p>‘Jon has told Miren the truth.’</p><p>b. *Jon-ek esan Miren-i egia dio [*V-O-Aux]</p><p>Jon-erg say-perf Miren-dat truth aux</p><p>Furthermore, it is not only in this domain that we find such word-order gaps. It</p><p>has been observed that VO languages cross-linguistically do not have clause-final</p><p>complementizers (cf. Hawkins 1990: 256–7; Dryer 1992: 102). Although at first sight</p><p>this does not seem to be ruled out by FOFC as stated above, because C does not</p><p>directly dominate VP, on closer inspection it emerges that a final C in a VO language</p><p>necessarily violates FOFC at some point in structures in which it occurs. Consider</p><p>the following:</p><p>(13) a. *[CP [AuxP [VP V O ] Aux ] C ] – violates FOFC (α = V, β = Aux)</p><p>*[CP [AuxP Aux [VP V O ]] C ] – violates FOFC (α = Aux, β = C)</p><p>* C′</p><p>AuxP C</p><p>b.</p><p>* C′</p><p>V O</p><p>In (13a), a structure with a head-final CP and a head-final AuxP, the head-final AuxP</p><p>dominates a head-initial VP, leading to a FOFC violation of the type already discussed</p><p>above. In (13b) a structure with a head-final CP dominates a head-initial AuxP, leading</p><p>to a FOFC violation at the CP/AuxP level. Biberauer, Holmberg & Roberts/BHR (2007,</p><p>2008) discuss a range of further structures excluded by FOFC, not only in the clausal</p><p>domain, but also in the context of nominals, underlining the general nature of the gap</p><p>characterized by (4). Furthermore, Cecchetto (to appear) shows that FOFC also holds</p><p>in the domain of sign languages, thereby confirming the non-modality-specific nature</p><p>of this grammatical constraint.</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Although there is thus significant empirical support for FOFC, it is important to</p><p>note that there are also apparent exceptions. For example in German, a head-final VP</p><p>can dominate a head-initial DP (14a) or head-initial PP (14b):</p><p>(14) a. Er hat [VP[DP ein Buch] gelesen]3</p><p>he has a book read</p><p>‘He read the book.’</p><p>b. Sie ist [VP[PP nach Berlin] gefahren]</p><p>she is to Berlin driven</p><p>‘She went to Berlin.’</p><p>BHR argue that such exceptions can be accounted for if FOFC is modified, such that it</p><p>only holds over XPs that are categorially non-distinct. Although the notion of ‘categorial</p><p>distinctness’ is by no means straightforward, in the German case it is clear that the</p><p>VP is a verbal category, whereas the DP, and perhaps the PP, should be considered</p><p>nominal.4 By contrast, both (inflected) auxiliaries and verbs are verbal, while Cs are</p><p>thought to encode typically verbal features like [finiteness] and are thus at least in part</p><p>verbal too (cf. Biberauer, Newton & Sheehan 2008, 2009 for further discussion); con-</p><p>sequently, we would expect FOFC to hold between V and C, but not V and D.</p><p>A further class of exceptions comes from clause-final particles. Many VO languages</p><p>have clause-final force particles (e.g. Mandarin Chinese, Sre, Mon, Copala Trique)</p><p>(15) a. Hufei chi-le sheme ma? (Mandarin Chinese)</p><p>Hufei eat-asp thing QYES/NO</p><p>‘Did Hufei eat anything?’</p><p>.  An anonymous reviewer points out that DPs in German may not all be VP-internal, as it</p><p>is clear that some DPs surface to the left of middle-field elements like VP-adverbs. This does</p><p>not affect our central point since both base- and movement-generated FOFC-violating struc-</p><p>tures are predicted to</p><p>be out (cf. Note 2). It is, however, worth noting that BHR do distinguish</p><p>between linearization-driven movements and A- and A¢-driven movements, with the latter not</p><p>being subject to FOFC. The core idea here is that movement that takes place as a result of its</p><p>being part of an item’s selection information must be FOFC-conforming: it is not possible for</p><p>a given head to select a head-initial phrase and require this selected phrase to move to its spec-</p><p>ifier (cf. BHR, Biberauer, Newton & Sheehan 2009 for more detail on this). By contrast, it is</p><p>possible for an initial functional head to bear an A- or A′-movement trigger – i.e. a movement</p><p>trigger that is not associated with its selection profile – which might then trigger a FOFC-</p><p>violating structure. This is arguably the case wherever VP-fronting occurs in head-initial lan-</p><p>guages (cf. English and, more strikingly, the Slavic example discussed in BHR). In connection</p><p>with DP-placement, it should, however, be remembered that both A- and A′-movement of DPs</p><p>cannot trigger FOFC violations since FOFC only holds between categorially alike phrases.</p><p>.  There are many predecessors to this idea in the literature, from Kayne’s (1984) Connected-</p><p>ness proposals to Grimshaw’s (1991) ‘extended projection’ to van Riemsdijk’s (1998) ‘Law of</p><p>Categorial Feature Magnetism’ and Brody’s (1995, 1998) ideas on projection.</p><p>Impossible changes and impossible borrowings 1</p><p>b. drùsĭ mâ zâ \aa rá (Lugbara, Nilo-Saharan)</p><p>tomorrow I meat eat affirmative</p><p>‘Tomorrow I will eat meat.’ (cf. Heine & Nurse 2000: 208)</p><p>Systematic positional differences between “full” and particle forms of a given type</p><p>of element (e.g. auxiliary, complementizer, etc.) in some of the languages with</p><p>clause-final particles suggest that the Greenbergian approach to particles, in terms</p><p>of which these are systematically excluded from word-order placement gener-</p><p>alizations on account of their peculiar properties (cf. Greenberg 1963), may be</p><p>well-founded. Consider (16), which shows that inflected auxiliaries are barred</p><p>from the clause-final position in which (uninflected) auxiliary particles obligato-</p><p>rily appear:</p><p>(16) a. yә- ca dεyo lf (Bwe, Karen)</p><p>1sg-see picture asp</p><p>‘I am looking at a picture.’</p><p>b. ce-1f mι jә-khf́ phι má nf́ (*jә-khf́)</p><p>3- say comp 3-fut take what</p><p>‘What did he say that he would take?’</p><p>We leave aside here (apparently) FOFC-violating structures containing particles,</p><p>whose properties are still poorly understood.</p><p>To summarize, then, empirical evidence from a wide range of languages suggests</p><p>that FOFC, as stated in (17), is an absolute principle which acts as a universal con-</p><p>straint on synchronic grammars (modulo particles).</p><p>(17) The Final-over-Final Constraint (FOFC) [Final Version]:</p><p>If α is a head-initial phrase and β is a categorially non-distinct phrase immediately</p><p>dominating α, then β must be head-initial. If α is a head-final phrase, and β is a</p><p>phrase immediately dominating α, then β can be head-initial or head-final.</p><p>It presently remains unresolved whether FOFC follows from a principle of UG (i.e.</p><p>Chomsky (2005)’s “first factor”), a processing preference, or extra-linguistic, economy-</p><p>oriented “third factor” pressures (cf. once again, Chomsky 2005). Biberauer, Holmberg</p><p>& Roberts initially (2007, 2008) take it to be an effect of ‘phase harmony’ within a</p><p>Linear Correspondence Axiom (LCA)-based system, and, later (Biberauer, Holmberg &</p><p>Roberts 2010), of Relativized Minimality, once again within an LCA-based system,</p><p>while Sheehan (2009a, b) proposes that it stems directly from Kayne’s (1994) LCA</p><p>and a copy theory of labeling. These accounts, then, take FOFC to be at least in part</p><p>due to first factor properties of the faculty of language, notably the LCA. Alterna-</p><p>tively, it might be that FOFC is an effect of a very strong parsing/processing prefer-</p><p>ence (cf. Hawkins 1994, 2004 and Cecchetto 2007, to appear), though see Sheehan (to</p><p>appear) for a critical appraisal of Hawkins’ Performance Grammar Correspondence</p><p>2 Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Hypothesis and its relation to FOFC.5 Given the uncertain nature of this question, we</p><p>leave it aside here, focusing instead on the central claim that FOFC, whatever its pre-</p><p>cise origins, is a universal constraint on synchronic grammars.</p><p>As a universal constraint, we expect FOFC not only to constrain synchronic gram-</p><p>mars, but also to play a role in acquisition, restricting the types of grammars that can</p><p>be acquired and, thereby, also the types of diachronic change which are possible. The</p><p>remainder of this chapter will consider FOFC’s role in diachronic change, both where</p><p>contact is and where it is not involved.</p><p>.  FOFC and diachronic change</p><p>If, as proposed in the previous section, FOFC is a universal constraint on synchronic</p><p>grammars, it has important implications for syntactic change. In terms of an influential</p><p>view in the field of diachronic syntax (see in particular Lightfoot 1999 and also Roberts</p><p>2007 for discussion), syntactic change should be seen as a random walk through para-</p><p>metric space. On this view, any syntactic change is possible if the primary linguistic</p><p>data (PLD) to which the language learner is exposed provides the appropriate evidence</p><p>or trigger for it. It therefore follows from this view that it is impossible to make predic-</p><p>tions regarding the pathways that word-order change will follow (see also Newmeyer</p><p>1998). However, we propose that this is not the case. Universal constraints on gram-</p><p>mar, such as FOFC, which categorically rule out certain structures, suggest that there</p><p>may in fact be UG-imposed structural constraints that determine specific possible and</p><p>impossible pathways of syntactic change.6</p><p>If FOFC is a universal constraint, then it is operative not only in today’s languages,</p><p>but also in all languages of the past. In other words, FOFC must apply at every stage of</p><p>a language’s history. This predicts that a FOFC-violating order should never be able to</p><p>develop, even transitionally as part of a larger series of changes. Bearing this in mind,</p><p>FOFC allows us to make predictions about pathways of syntactic change. For example,</p><p>when a language changes from being predominantly head-final (“OV”) to predomi-</p><p>nantly head-initial (“VO”), this change must proceed top-down, as illustrated in (18):</p><p>(18) [[[O V] Aux] C] → [C [[O V] Aux]] → [C [Aux [O V]]] → [C [Aux [V O]]]</p><p>As shown above, FOFC requires that CP must change first, giving C-AuxP order in</p><p>place of AuxP-C order. The AuxP can then follow, giving Aux-VP in place of VP-Aux</p><p>.  Taken a certain way, Hawkins’ principles might actually be considered ‘third factor’ general</p><p>economy principles. This idea is explicitly developed in Mobbs (2008).</p><p>.  As noted in the main text, we are not committed to the view that FOFC is the exclusive</p><p>consequence of a hard-wired invariant principle of UG. Our use of UG here should thus be</p><p>interpreted as leaving open the various possibilities raised in the main text.</p><p>Impossible changes and impossible borrowings </p><p>order. Only once these two changes are at the very least underway (see below) can the</p><p>VP start to exhibit variation, possibly ultimately leading to change from OV to VO. If</p><p>the change proceeded in the opposite direction, with the VP undergoing the change</p><p>first, this would give rise to V-O-Aux and V-O-C, orders that are ruled out by FOFC,</p><p>as discussed above.</p><p>Similarly, if we consider the opposite change, that is, from head-initial (VO) to</p><p>head-final (OV), FOFC predicts that this must proceed bottom-up. The VP must</p><p>change first, followed by the AuxP, then the CP:</p><p>(19) [C [Aux [V O]]] → [C [ Aux [O V]]] → [C [[O V] Aux]] → [[[O V] Aux] C]</p><p>A change in the opposite direction, beginning with the CP or the AuxP would result in</p><p>the FOFC-violating V-O-C or V-O-Aux orders from the outset.7</p><p>The first of these predictions is borne out within the history of Indo-European. The</p><p>change from OV to VO is well attested in</p><p>both the Germanic and Romance branches</p><p>and in both cases the change seems to follow the FOFC-determined pathway.8</p><p>.1  OV to VO in the history of English</p><p>Although Modern English is predominantly a head-initial language, this was not</p><p>always the case. Like Modern German, Old English was head-final in the VP and the</p><p>AuxP; therefore, English has clearly undergone a change from head-final to head-initial</p><p>within its recorded history. FOFC predicts that the first stage of this change should</p><p>affect the CP. From the earliest attested evidence of Germanic, we only find head-initial</p><p>CPs.9 Crucially, however, initial complementizers may appear with both head-initial</p><p>.  Crucially it should be noted that the pathways we propose are in no sense deterministic. If</p><p>a consistently head-final language becomes head-initial in the CP, there is no necessity for it to</p><p>undergo further change in the AuxP and the VP. Further syntactic change will only take place</p><p>as a result of a change in the PLD available to children acquiring the language. See Section 3.3</p><p>below for further discussion with reference to synchronic word-order variation in Afrikaans.</p><p>.  Evidence of the second pathway (VO to OV) can be found in the Ethiopian Semitic lan-</p><p>guages, which have seemingly changed from head-initial to head-final due to contact with</p><p>Cushitic. Biberauer, Newton & Sheehan (2008) discuss this case in some detail.</p><p>.  It is not clear that Germanic or, indeed, Proto-Indo-European ever had clause-final</p><p>complementisers. Complementisers in the attested Indo-European languages seem to have</p><p>developed independently, and, as such, we cannot reconstruct complementisers for Proto-</p><p>Indo-European (see Kiparsky 1995). Just as there is no need for a system to progress all the</p><p>way down a given FOFC-determined pathway once it has “started” at one of the extremes, so</p><p>there is no need for a system which undergoes FOFC-sanctioned changes to have started at</p><p>one of these extremes: our proposal is simply that changes will follow the available pathways,</p><p>independently of their precise starting point.</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>and head-final AuxPs and VPs (although, as we saw above, not in the FOFC violating</p><p>V-O-Aux order). This is shown in the examples below (formatting as above).</p><p>(20) a. O-V-AUX:</p><p>… þæt Darius hie mid gefeohte secan wolde</p><p>that Darius them for battle seek wanted</p><p>‘… that Darius wanted to seek them out for a battle.’</p><p>(Orosius 45.31; cf. van Kemenade 1987: 16)</p><p>b. V-O-AUX unattested (cf. Pintzuk 1991, 1999)</p><p>c. AUX-O-V:</p><p>… þæt hie mihton swa bealdlice Godes geleafan bodian</p><p>that they could so boldly God’s faith preach</p><p>‘…that they could preach God’s faith so boldly.’</p><p>(The Homilies of the Anglo-Saxon Church I 232; cf. van Kemenade 1987: 179)</p><p>d. AUX-V-O:</p><p>… þæt he mot ehtan godra manna</p><p>that he might persecute good men</p><p>‘… that he might persecute good men.’</p><p>(Wulfstan’s Homilies 130.37–38; Pintzuk 2002: 282)</p><p>Here we see that a final AuxP may only combine with a head-final VP (cf. (20a)),</p><p>whereas an initial AuxP has two combination possibilities, being able to combine both</p><p>with a head-final VP as in (20c) and with a head-initial VP as in (20d). The availability</p><p>of (20d) at a stage at which (20c)-type structures are still attested indicates that varia-</p><p>tion in VP order becomes possible as soon as initial AuxP becomes available. Crucially,</p><p>however, head-initial VPs are strictly limited to initial AuxP structures: despite the</p><p>independent availability of both head-final AuxP and head-initial VP, structures of</p><p>this type are completely unattested.</p><p>Turning next to the AuxP, Pintzuk (1991/1999) proposes that the transition from</p><p>head-final AuxP to head-initial AuxP was a gradual process, progressing throughout</p><p>the Old English period until early Middle English, when it reached completion. Varia-</p><p>tion within the VP, between OV and VO order, however, continues until the Late</p><p>Middle English period (formatting as before):</p><p>(21) AUX-O-V:</p><p>a. Þat ne haue noht here sinnes forleten</p><p>who neg have not their sins forsaken</p><p>‘…who have not forsaken their sins.’</p><p>(Trinity Homilies 67.934; cited in Kroch & Taylor 2000: 154)</p><p>b. AUX-V-O</p><p>oðet he habbe iŠetted ou al et Še wulle</p><p>until he has granted you all that you desire</p><p>‘Until he has granted you all that you desire.’</p><p>(Ancrene Riwle; cited in Kroch & Taylor 2000: 145)</p><p>Impossible changes and impossible borrowings </p><p>It seems then, that the shift from OV to VO order in the history of English proceeded</p><p>top-down, exactly as FOFC predicts. Biberauer, Newton & Sheehan (2008) show how</p><p>the same is true of word-order change in Yiddish and Icelandic.</p><p>.2  OV to VO in the history of French</p><p>The development of VO order in French seems to have followed a similar pattern. Like</p><p>the early Germanic languages, Latin exhibits only head-initial CPs:</p><p>(22) accidit perincommode quod eum nusquam vidisti</p><p>happened-3sg unfortunately that him nowhere saw-2sg</p><p>‘It is unfortunate that you didn’t see him anywhere.’</p><p>(Cicero, Ad Attticum 1, 17, 2; cited in Roberts 2007: 169)</p><p>The development of the AuxP in Romance languages is somewhat more complicated</p><p>than in Germanic, as most “auxiliary” meanings in Latin were expressed morphologi-</p><p>cally rather than by means of auxiliaries (cf. Benveniste 1968 and Ledgeway to appear</p><p>for overview discussion). Classical Latin does, however, feature one compound tense,</p><p>namely the perfect passive. As (23) illustrates, the auxiliary element in this structure,</p><p>esse (‘to be’), could either precede or follow the participle:</p><p>(23) a. illa quae cum rege est pugnata</p><p>that-nom which-nom with king-abl is fought</p><p>‘that (battle) which was fought with the king.’</p><p>(Cicero, Pro Murena; cited in Ledgeway to appear: 8)</p><p>b. diu et acriter pugnatum est</p><p>long-time and fierce fought is</p><p>‘There was a long and hard battle.’</p><p>(Caesar, De Bello Gallico; cf. Ledgeway to appear: 8)</p><p>The auxiliary habere develops in later Latin, and, like esse, its position is variable but</p><p>the unmarked option seems to be for it to appear clause-finally (Bauer 1995: 104–107).</p><p>This is to be expected as the preferred position of the lexical verb, from which the</p><p>auxiliary habere grammaticalized, is clause-final (cf. i.a. Bauer 1995: 89–92):10</p><p>(24) a. haec omnia probatum habemus</p><p>these all.nom.pl tried.nom.sg have.1pl</p><p>‘We have tried all these things.’</p><p>(Oribasius – 6th century C.E.; cited in Ledgeway to appear: 62)</p><p>1.  The grammaticalization of the Romance future tense (see Roberts & Roussou 2003:</p><p>48–57), formed from infinitive+habere, also suggests that, in this construction at least, the</p><p>auxiliary habere appeared clause-finally.</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>b. sicut parabolatum habuistis</p><p>thus spoken had.2pl</p><p>‘Thus you had spoken.’</p><p>(Formulae Salicae Merkelianae; cited in Ledgeway to appear: 62)</p><p>As part of the transition from Vulgar Latin to French, both the AuxP and the VP</p><p>became fixed as head-initial. Bauer (1995: 106) suggests that this change progressed</p><p>in exactly the order predicted by FOFC, with the AuxP becoming head-initial before</p><p>the VP:</p><p>(25) “Whereas the ordering habeo/sum + participle prevailed in twelfth century</p><p>Old French, the development towards the modern structure, where the direct</p><p>object follows the compound verb [[habeo/j’ai [participle] [direct object]] is</p><p>a development of Middle French. Accordingly, the structure (a) epistulam</p><p>scriptam habeo/je possède une lettre écrite ‘I possess a written letter’ changed</p><p>first into (b) habeo epistulam scriptam/j’ai une lettre écrite and then into (c) j’ai</p><p>écrit une lettre ‘I wrote a letter’, where the direct object follows the auxiliary</p><p>and the participle”</p><p>Although there is clearly a great deal more work to be done in determining the exact</p><p>details of the change from OV to VO in French, with the relevant data possibly being</p><p>absent from the attested record (Adam Ledgeway, p.c.), the basic evidence</p><p>presented</p><p>above is suggestive of a change following the pathway predicted by FOFC.</p><p>.  Word-order variation in Afrikaans</p><p>FOFC pathways are not only demonstrated by diachronic change, but also by syn-</p><p>chronic variation. In Modern Spoken Afrikaans (MSA), we find both verb-final (26a)</p><p>and “verb-early” (26b) embedded clauses, in which a finite verb surfaces in a position</p><p>clearly higher than VP:</p><p>(26) a. Ek weet dat sy [VP dikwels Chopin gespeel] het (verb-final)</p><p>I know that she often Chopin played have</p><p>‘I know that she has often played Chopin.’</p><p>b. Ek weet dat sy het [VP dikwels Chopin gespeel] (verb early)</p><p>I know that she have often Chopin played</p><p>‘I know that she has often played Chopin.’</p><p>(26a) represents the older pattern, that is, the “correct” one prescribed by grammars;</p><p>the “verb-early” construction is an innovation (the equivalent structure is barred in</p><p>all varieties of Dutch). However, both structures are common in MSA and they are</p><p>interpretively identical. In contrast, “verb-early” constructions with main verbs are far</p><p>less common and necessarily have a “main-clause” interpretation (cf. i.a. Holmberg &</p><p>Platzack 1995 on so-called embedded root phenomena in V2 languages).</p><p>Impossible changes and impossible borrowings </p><p>(27) Hy dink dat sy [VP speel altyd Chopin]</p><p>he think that she play always Chopin</p><p>‘He thinks that she always plays Chopin’</p><p>Since (27)-type structures systematically behave differently to the “verb-early” ones</p><p>in (26), Biberauer (2003) proposes that only the (27)-type structures involve V2; the</p><p>alternations in (26)-type structures, by contrast, feature auxiliaries located in Aux. As</p><p>such, they constitute evidence that MSA permits both final ((26a)) and initial ((26b))</p><p>AuxP structures.</p><p>Crucially, however, MSA does not permit head-initial VPs, as earlier English, for</p><p>example, did (cf. (20c, d) above). This seems to be directly attributable to the salience</p><p>of particle verbs in MSA (cf. Ponelis 1993). As argued by Lightfoot (1979 and follow-</p><p>ing), particles serve as “signposts” signalling the location of the verb with respect to</p><p>the object: where a particle precedes the object, as in VO languages, the acquirer can</p><p>conclude that VP is initial; where it follows, as in OV languages, VP must be final.</p><p>Against this background, robust attestation of particle verbs in the MSA input would</p><p>be expected to contribute a clear signal to the acquirer that the system being acquired</p><p>involves a final VP.</p><p>That this reasoning is correct is strongly suggested by the fact that Kaaps, a variety</p><p>of Afrikaans spoken in the Cape in which English borrowings/substitutions have dras-</p><p>tically reduced the number of particle verbs, permits initial VPs in the presence of an</p><p>initial TP. As in the earlier English case, initial VP is not compatible with final TP. The</p><p>relevant facts are illustrated in (28):</p><p>(28) a. Ek het [VP gekry R1400 van die Revenue] [AUX-V-O]</p><p>I have received R1400 from the Revenue</p><p>‘I have received R1400 from the Receiver of Revenue.’</p><p>[Standard Afrikaans: Ek het R1400 van die Department van Inkomste gekry,</p><p>i.e. [O-V-AUX]]</p><p>b. *dat ek [VP gekry R1400 van die Revenue] het [*V-O-AUX]</p><p>that I received R1400 from the Revenue have</p><p>‘I have received R1400 from the Receiver of the Revenue.’</p><p>Kaaps, then, appears to be one step further along the FOFC-predicted pathway than</p><p>MSA. Crucially, this state of affairs reflects the fact that FOFC is not itself a driver of</p><p>syntactic change, but merely a constraint which defines possible and impossible dia-</p><p>chronic pathways. For change to occur, it needs, as Longobardi (2001: 278) points out,</p><p>to be the “well-motivated consequence of other types of change (phonological changes</p><p>and semantic changes, including the appearance/disappearance of whole lexical items)</p><p>or, recursively, of other syntactic changes” (cf. also Keenan 2002 on this so-called Inertia</p><p>Principle). Syntax therefore does not change unless there is specific input (which we</p><p>might think of as a ‘cue’ – cf. i.a. Lightfoot 1991, 1999; Dresher 1999) signaling to the</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>acquirer that a grammar different to that initially acquired by the previous generation</p><p>should be postulated (cf. Roberts 2007 for recent overview discussion of this general</p><p>approach to syntactic change).</p><p>In the following section we will consider a further example of word-order-</p><p>related change that clearly involved a prior change, this time from typologically very</p><p>different languages.</p><p>.  FOFC and borrowing</p><p>Contact-induced change is of particular interest in the present context as it has some-</p><p>times been suggested that this type of change, in contrast to the non-contact-induced</p><p>variety, may result in typologically unusual linguistic systems. Harris & Campbell</p><p>(1995: 239), for example, claim that typologically rare word orders are often the result</p><p>of contact-induced change, implying that the latter might fall beyond the remit of</p><p>typological universals. Nonetheless, while it might be the case that ‘exotic’ word orders</p><p>often arise as a result of borrowing, there is no reason to believe that borrowings will</p><p>be immune to grammatical constraints (cf. also Bowern 2008 for a discussion). As far</p><p>as FOFC is concerned, initial empirical research suggests that borrowing with word-</p><p>order implications is constrained in exactly the same way as change which may be less</p><p>directly contact-induced (cf. the discussion in the previous section). We will consider</p><p>just one case study here.</p><p>The South Asian linguistic area provides an excellent testing ground for FOFC’s</p><p>effect on borrowing as the more rigidly head-final Dravidian languages have a long</p><p>history of contact with the more disharmonic Indo-Aryan languages. As noted by</p><p>Hock & Joseph (1996: 61), there is controversy over when Indo-Aryan and Dravidian</p><p>first came into contact, but it is uncontroversial that they came to “structurally converge</p><p>after multilingual contact extending over several millennia”. According to Thomason &</p><p>Kaufman (1988: 43), there is evidence of an early, shift-induced, Dravidian influence</p><p>on Indic from the spread of retroflex consonants, despite the virtual lack of lexical</p><p>borrowing (cf. also Kuiper 1974). Moreover, there is little dispute that many Indo-Aryan</p><p>languages have borrowed many lexical and grammatical features from Dravidian in</p><p>the more recent past, through continued contact. Interestingly, there are clear patterns</p><p>across the area with respect to certain grammatical features, which, we argue, cannot</p><p>be satisfactorily explained by sociolinguistic or genetic factors alone. In at least one</p><p>such case, patterns of borrowing/development can be explained by grammatical factors,</p><p>more specifically FOFC.</p><p>It is well known that Indo-Aryan languages show variation in the placement of</p><p>complementizers and polarity question particles (cf. Masica 1991; Marlow 1997; Bayer</p><p>1999, 2001; Davison 2007). It is generally assumed that Indo-Aryan borrowed final</p><p>Impossible changes and impossible borrowings </p><p>quotative complementizers from Dravidian, either very early on (Kuiper 1974) or at</p><p>a late stage of contact (Meenakshi 1986), or perhaps both (but cf. Hock 1982 for an</p><p>alternative account).11 In fact, while Sanskrit did have the (largely) final complemen-</p><p>tizer iti (ti/tti in Middle Indo-Aryan), this was later lost and no cognate remains in</p><p>any of the modern mainland Indo-Aryan languages (Meenakshi 1986; Marlow 1997;</p><p>Southworth 2005).11Instead, Modern Indo-Aryan languages either lack a final comple-</p><p>mentizer altogether or display final complementizers of the quotative type, typically</p><p>derived from the verb of saying (cf. Bengali/Bangla bole ‘saying’), or from the demon-</p><p>strative root (cf. Marathi asa ‘such’) or both in the case of Marathi (Bayer 1999, 2001;</p><p>Masica 1991; Davison 2007). Figure 1 shows the distribution of these complementizers</p><p>(Cs),12 based on appendix A of Davison (2007).13</p><p>No final C Final C from ‘saying’</p><p>Final C from demonstrative</p><p>Hindi/Urdu, Panjabi, Sindhi,</p><p>Kashmiri, Maithili, Kurmali</p><p>Sinhala, Dhivehi, Marathi,</p><p>Nepali, Dakkhini Hindi</p><p>Assamese, Bengali/Bangla,</p><p>Oriya</p><p>Marathi, Gujarati</p><p>Figure 1. Distribution of final complementizers in Indo-Aryan</p><p>Interestingly, all modern Indic languages (with the exception of Sinhala, spoken in</p><p>Sri Lanka) also have an initial C, the origins of which vary, and are not to be discussed at</p><p>11.  Southworth (2005) also notes that the Munda languages, which constitute a branch of</p><p>the Austro-Asiatic language family, exhibit a similar pattern, so the borrowing might not</p><p>necessarily have been from Dravidian. This is not crucial for our purposes as the Munda</p><p>languages would also represent a typologically distinct borrowing source.</p><p>12.  Although we will abbreviate complementizer as C, as is commonly done in Chomskyan</p><p>work, it should be noted that we do not intend this abbreviation to be read as an indication of</p><p>our interpreting the C-elements in the languages under discussion as Cs equivalent to English-</p><p>style complementizers. As seems fairly clear from the examples that follow, the elements in</p><p>question would seem to lexicalize a rather different sub-portion of Rizzi’s (1997) articulated</p><p>CP to the finite Cs familiar from Germanic and Romance. For example, while English that</p><p>seems to encode both Force and Finiteness, with the result that we might think of it as a syn-</p><p>cretised Force-Finiteness element (cf. Giorgi & Pianesi 1997), the Marathi and Hindi-Urdu</p><p>Cs illustrated in (32–33) only seem to encode subordination, which may be a (sub-)species of</p><p>Force (cf. Biberauer, Newton & Sheehan 2009 for further discussion).</p><p>1.  We call these final ‘quotatives’ complementizers as they are not limited to contexts of</p><p>direct/indirect speech, but rather occur with epistemic and perception verbs such as know,</p><p>think and hear (cf. Bayer 2001).</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>length here (cf. Meenakshi 1986; Marlow 1997 for discussion).14 For our purposes, the</p><p>interest lies in the varying availability of a final C in the languages under consideration.</p><p>All Indic languages exhibit a degree of typological similarity with Dravidian. Typo-</p><p>logical similarity is known to aid borrowing (cf. Thomason & Kaufman 1988: 97–98),</p><p>and in this case, it did so by making available the potential structure [[O V] say] in the</p><p>native language. While it is true that many of the languages with a final C have been</p><p>in heavy contact with Dravidian (e.g. Marathi, Gujarati), others have not (e.g. Nepali).</p><p>This means that in at least this language, the structure must either have developed</p><p>internally (cf. Klamer 2001) or been borrowed from another source (perhaps a neigh-</p><p>bouring Tibeto-Burman language). See Marlow (1997) for a full discussion of these</p><p>facts and the conclusion that the distribution of final quotatives in the area cannot be</p><p>explained in purely historical/geographical terms.</p><p>The question arises, then, as to whether this borrowing also follows the path of</p><p>diachronic change predicted by FOFC. In fact, there is good evidence that it does.</p><p>Davison’s (2007) observation is that those Indo-Aryan languages which lack a final</p><p>complementizer are exactly those which have an initial Polarity question head (Pol) of</p><p>the type illustrated here for Hindi-Urdu and Panjabi:</p><p>(29) kyaa aap wahaaN aaeeNgii?</p><p>pol you there come.fut.2pl</p><p>‘Are you coming?’ (Hindi-Urdu, Davison 2007: 180)</p><p>(30) (kii) tuslïï ajj káánii suNaavoge?</p><p>pol you today story tell.fut.2mp</p><p>‘Will you tell a story today?’ (Panjabi, Davison 2007: 180, citing Bhatia 1993: 5)</p><p>Other languages in the area have non-initial Pol heads, which occur in final position</p><p>or in final/medial position, but crucially never initially:</p><p>(31) a. to kaal parat aalaa kaa(y)?</p><p>he yesterday back come.pst.3s pol</p><p>‘Did he come back yesterday?’</p><p>1.  The two main types of initial complementizers are the (Sanskrit-like) j-series, derived</p><p>from the relative pronoun, and the k-series. Although it is commonly assumed that ki-</p><p>complementizers in Hindi-Urdu and other Indo-Aryan languages (minus the Eastern branch)</p><p>are a borrowing from Persian (Kellogg 1893), the initial complementizer kimti/kiti is also attested</p><p>in the Inscriptional Prakrits, and might be the actual source of the modern form (Meenakshi</p><p>1986: 212). Moreover, as Bayer (2001) points out, ki is similar in form to the wh-paradigm,</p><p>which has served as a source for complementizers in Romance, Slavic and Greek branches of</p><p>Indo-European. Interestingly, Persian ke has been passed on to one Indo-European language</p><p>as well as at least three other language families through contact: Indo-European (Asia Minor</p><p>Greek), Altaic (several Turkic languages, Kartvelian (Laz), Nakho-Daghestanian (Lezgian)),</p><p>and (Northern) Dravidian (Brahui) (cf. Haig 2001 for discussion).</p><p>Impossible changes and impossible borrowings 1</p><p>b. *kaa(y) to kaal parat aalaa</p><p>pol he yesterday back come.pst.3ms</p><p>(Marathi, based on Davison 2007: 181, citing Pandharipande 1997: 8)</p><p>(32) a. modhu aS-be ki (na)?</p><p>Madhu come-fut pol not</p><p>‘Will/Won’t Madhu come?’</p><p>b. modhu ki aS -be?</p><p>Madhu pol come-fut</p><p>‘Will Madhu come?’15</p><p>c. *ki modhu aS-be?</p><p>pol Madhu come-fut</p><p>(Bengali/Bangla, Davison 2007: 189, attributed to P. Dasgupta)</p><p>Previous research suggests that C is uniformly higher than the Pol head involved in</p><p>yes/no questions (Laka 1994; Rizzi 2001; Holmberg 2003, but cf. Biberauer 2008 for</p><p>the positing of a higher polarity head involved in negation). In fact, there is good evi-</p><p>dence from Indic that C dominates Pol. As Davison shows, in many, but not quite all of</p><p>the modern languages (the exceptions being Bengali/Bangla and other Eastern Indic</p><p>languages), Pol can co-occur with C in embedded clauses.16 Where this happens, final</p><p>Pol precedes final C, whereas initial Pol follows initial C. Moreover, the reverse orders</p><p>are ungrammatical, as shown for Hindi-Urdu:</p><p>(33) [[to kal parat aalaa kaa(y)] mhaaNun/asa]</p><p>he yesterday back come.pst.3ms pol quot/such</p><p>raam malaa witSaarat hotaa</p><p>Ram I.dat ask.prog be.pst.3ms</p><p>‘Ram was asking me whether/if he came back yesterday.’</p><p>(Marathi, Davison 2007: 182, attributed to R. Pandharipande)</p><p>(34) a. us-nee puuc-aa [ki [kyaa tum aa-oogee]]</p><p>3s-erg ask-perf that pol you come-fut</p><p>‘He asked whether you will come.’</p><p>b. *us-nee puuc-aa [kyaa ki tum aa-oogee]</p><p>3s-erg ask-perf pol that you come-fut</p><p>‘He asked whether you will come.’ (Hindi-Urdu, Davison 2007: 181)</p><p>1.  The positioning of the polarity head appears to be connected to focus, according to</p><p>Davison (2007). See Nayudu (2008) for a similar effect in the domain of negation in Marathi.</p><p>1.  The pattern in Bangla is reminiscent of that seen in Sanskrit and is probably connected</p><p>to the fact that initial complementizers in Bangla and the other Eastern languages are derived</p><p>from relative pronouns, whereas Western languages have borrowed Persian ki (cf. Bayer 1999,</p><p>2001; Davison 2005)</p><p>2 Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Where C is initial and Pol is final, the two can still co-occur:</p><p>(35) raam maalaa witSaarat hotaa [ki to kal parat</p><p>Ram I.dat ask.prog be.pst.3ms that he yesterday back</p><p>aalaa kaa(y)]</p><p>come.pst.3ms pol</p><p>‘Ram was asking me whether/if he came back yesterday.’</p><p>(Marathi, Davison 2007: 182, attributed to R. Pandharipande)</p><p>This is not problematic from the perspective of FOFC as, assuming that C is higher</p><p>than Pol, this represents an inverse FOFC order, which is permitted by UG:</p><p>(36) [CPC [PolP [AuxP] Pol ]]</p><p>CP</p><p>C PolP</p><p>AuxP Pol</p><p>Under our assumptions, though, an initial Pol should block the borrowing/develop-</p><p>ment of a final C as this would lead to the FOFC-violating structure in (37):</p><p>(37) * [CP [PolP Pol [AuxP]] C ]</p><p>* CP</p><p>PolP C</p><p>Pol AuxP</p><p>This prediction appears to hold quite robustly in the area. As Davison shows at length,</p><p>no language in the South Asian linguistic area has both initial Pol and final C, although</p><p>all three of the other possible combinations readily occur. Hindi-Urdu, for example</p><p>has not developed a final C of any kind, from either a verbal or demonstrative source.</p><p>Thus both potential types of final C lead to ungrammaticality as shown in (38):17</p><p>1.  Bhatt & Takahashi (to appear) note that Hindi-Urdu postpositions are also unable to</p><p>take clausal complements and this is potentially another instance of the same effect: the Pol</p><p>head, whether overt or not, blocks the possibility of a final selecting (i.e. higher) head, regard-</p><p>less of whether this selector is verbal or adpositional. On the assumption that selectors share</p><p>the property of being [-N] (cf. Chomsky 1981), this can be related to the categorial non-</p><p>distinctness desideratum entailed by the FOFC constraint (cf. (17)). It is also worth noting</p><p>that if a null Pol head can indeed “count” for FOFC purposes – in barring the postulation of</p><p>Impossible changes and impossible borrowings </p><p>(38) *usee [[ vee aa rahee haiN] yah/ kah-kar ] maaluum hai</p><p>3s.dat 3pl come prog are this/ say-prt known is</p><p>‘He/she knows that they are coming.’ (Hindi-Urdu, Davison 2007: 177)</p><p>Figure 2, adapted from Appendix B of Davison (2007), shows the pattern in the relevant</p><p>area. In type A languages, the direct borrowing of a final C has been blocked by the</p><p>presence of an initial Pol, with the result that type D languages, which would be the</p><p>FOFC-violating type, are unattested.</p><p>Type Position of Pol Position of C Languages</p><p>A Initial Initial Only Hindi-Urdu, Panjabi, Kashmiri, Sindhi,</p><p>Maithili, Kurmali (plus some North</p><p>Dravidian languages, i.e. Brahui)</p><p>B Final/Medial Initial and Final Marathi, Gujarati, Assamese, Bengali/</p><p>Bangla, Dakhini Hindi, Oriya, Nepali</p><p>C Final/Medial Final Only Sinhala (plus some Dravidian languages)</p><p>D Initial Final Unattested in the area</p><p>(adapted from Davison 2007 Appendix B)</p><p>Figure 2. Polarity heads and complementizers in South Asia</p><p>One apparent problem with this proposal is that Sanskrit had both an initial ques-</p><p>tion particle kim and an optional final quotative iti (cf. Davison 2005). It appears,</p><p>though, that the syntax of kim was very different from that of modern Indo-Aryan</p><p>Polarity heads, as kim failed to occur in embedded clauses and so would not have given</p><p>rise to FOFC-violating structures (cf. Davison 2005 for a discussion of the syntax of</p><p>iti). Assuming that the polarity heads in Modern Indo-Aryan are derived from initial</p><p>kim, the prediction is that iti/tti must have been lost before kim was reanalysed as a</p><p>Polarity head so that it could appear in embedded clauses. Further work is required</p><p>to verify this claim. Of course, it remains true that most of the type A languages in</p><p>Figure 2 have had little contact with Dravidian, and this in turn might explain why</p><p>they have failed to develop either a final Polarity head or a final C. To verify that</p><p>Figure 2 is really evidence for the FOFC-predicted pathway, it would have to be shown</p><p>that development/borrowing of final Pol preceded borrowing/development of final C.</p><p>Once again, further diachronic work is needed to confirm this.</p><p>Further support for the FOFC-based prediction regarding the distribution of ini-</p><p>tial question particles and final complementizers comes from evidence from WALS</p><p>a FOFC-violating structure – this constraint cannot be ascribed to processing considerations</p><p>(cf. Cecchetto 2007, to appear for one proposal along these lines).</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>indicating that this pattern is not limited to South Asia, but is replicated throughout</p><p>the languages of the world (cf. Dryer 2005a, 2005b). From a sample of 195 languages,</p><p>the numbers of languages roughly equivalent to type A-D languages are as follows:18</p><p>Type Position of Pol Position of C Number of Languages</p><p>(genera: families)</p><p>A</p><p>B</p><p>C</p><p>D</p><p>Initial</p><p>Final</p><p>Final</p><p>Initial</p><p>Initial Only</p><p>Initial</p><p>Final Only</p><p>Final</p><p>72:35:13</p><p>74:40:16</p><p>45:33:20</p><p>4:3:3</p><p>Figure 3. Typological positioning of Polarity heads and complementizers</p><p>Clearly there is a skewing in the data here, with large numbers of genetically</p><p>diverse languages of types A-C. FOFC-violating type D is not, however, completely</p><p>unattested. Four languages from three language families, all found in South America,</p><p>are of this type: Tacana and Ese Ejja (Tacanan), Gavião (Tupi) and Resígaro (Arawakan).</p><p>In fact, all of these languages appear to nominalize their embedded clauses (cf. Moore</p><p>1989 on Gavião, Allin 1976 on Resígaro and Ottaviano 1980 on Tacana) and so this</p><p>might be the reason why FOFC does not appear to hold. Further research is required</p><p>to establish exactly what the relationship between clausal nominalization and FOFC</p><p>really is (cf. Biberauer & Sheehan to appear for some discussion).</p><p>For the moment, it would, however, seem that there is good evidence that FOFC</p><p>has constrained the borrowing of final complementizers in Indo-Aryan. Only those</p><p>languages which lack an initial Polarity head have developed/borrowed a final C. Further</p><p>evidence that this is what is at stake in Indic comes from typological trends, which</p><p>mirror the skewing in Indo-Aryan.</p><p>.  Conclusion</p><p>The aim of this chapter was to argue that FOFC, a universally valid syntactic prin-</p><p>ciple in the domain of word order, constrains synchronic grammars and therefore</p><p>1.  The data presented here is actually for “position of polar question particles” and “order</p><p>of adverbial subordinator and clause”. These have, however, been cleansed to make them</p><p>more comparable with the Indic facts, with other values such as second-position question</p><p>particles removed. Dryer uses the term adverbial subordinators to refer to “because, although,</p><p>when, while, and if”. While these subordinators often pattern with the positioning of more</p><p>unequivocal C-elements such as that, this is not always the case. For this reason, these data</p><p>are taken to be suggestive only (cf. also the discussion in Note 12).</p><p>Impossible changes and impossible borrowings </p><p>also diachronic change, regardless of whether this is triggered by contact or not.</p><p>Evidence suggests that FOFC-based predictions hold in the history of English,</p><p>Afrikaans and, insofar as relevant data are available, French. Moreover, patterns of</p><p>final-complementizer borrowing in Indic languages also seem to be attributable to</p><p>FOFC, the former being blocked by the presence of an initial Pol head lower in the</p><p>clause. Taken together, we argue that these facts suggest that the notions ‘impossible</p><p>change’ and ‘impossible borrowing’ may after all be contentful, thus justifying further</p><p>investigation.</p><p>References</p><p>Allin, Trevor R. 1976. A Grammar of Resígaro. Horsleys Green: Summer Institute of Linguistics.</p><p>Bauer, Brigitte. 1995. The Emergence and Development of SVO Patterning in Latin and French:</p><p>Diachronic and Psycholinguistic Perspectives. Oxford: OUP.</p><p>Bayer, Josef. 1999. Final complementizers in hybrid languages. Journal of Linguistics 35: 233–271.</p><p>Bayer, Josef. 2001. Two grammars in one: Sentential complements and complementizers in Bengali</p><p>and other South Asian languages. In The Yearbook of South Asian Languages: Tokyo Sympo-</p><p>sium on South Asian Languages – Contact Convergence and Typology, Peri Bhaskararao &</p><p>Karumuru Venkata Subbarao (eds), 11–36. New Delhi: Sage.</p><p>Benveniste, Emile. 1968. Mutations of linguistic categories. In Directions for Historical Linguistics,</p><p>Winifred P. Lehmann & Yakov Malkiel (eds), 85–94. Austin TX: University of Texas Press.</p><p>Besten, Hans den. 1986. Decidability in the syntax of verbs of (not necessarily) West Germanic</p><p>languages. Groninger Arbeiten zur Germanistischen Linguistik 28: 232–256. Reprinted as:</p><p>den Besten, Hans. 1989. Studies in West Germanic Syntax, 111–135. Amsterdam: Rodopi.</p><p>Bhatia, Tej K. 1993. Punjabi: A Cognitive-Descriptive Grammar. London: Routledge.</p><p>Bhatt, Rajesh & Takahashi, Shoichi. To appear. Reduced and unreduced phrasal comparatives.</p><p>In Natural Language and Linguistic Theory.</p><p>Biberauer, Theresa. 2003. Verb Second</p><p>(V2) in Afrikaans: A Minimalist Investigation of Word-</p><p>Order Variation. Ph.D. dissertation, University of Cambridge.</p><p>Biberauer, Theresa. 2008. Doubling and omission: Insights from Afrikaans negation. In</p><p>Microvariations in Syntactic Doubling, Sjef Barbiers, Olaf Koeneman, Marika Lekakou &</p><p>Margreet van der Ham, (eds), 103–140. Bingley: Emerald.</p><p>Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2007. Disharmonic word-order systems</p><p>and the Final-over-Final-Constraint (FOFC). In Proceedings of XXXIII Incontro di</p><p>Grammatica Generativa, Antonietta Bisetto & Francesco Barbieri (eds), 86–105. Bologna:</p><p>Università de Bologna.</p><p>Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2008. Structure and Linearization in</p><p>Disharmonic Word Orders. In Proceedings of the 26th West Coast Conference on Formal</p><p>Linguistics, Charles B. Chang & Hannah J. Haynie (eds), 96–104. Somerville MA: Cascadilla</p><p>Proceedings Project.</p><p>Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2010. A syntactic universal and its</p><p>consequences. Ms, Universities of Cambridge and Newcastle.</p><p>Biberauer, Theresa, Newton, Glenda & Sheehan, Michelle. 2008. Defining diachronic pathways:</p><p>The Final-over-Final Constraint. Paper presented at DiGS X (Cornell).</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Biberauer, Theresa, Newton, Glenda & Sheehan, Michelle. 2009. Limiting synchronic and dia-</p><p>chronic variation and change: The Final-Over-Final Constraint. Language and Linguistics</p><p>10(4):701–743.</p><p>Biberauer, Theresa & Sheehan, Michelle. To appear. Disharmony, antisymmetry, and the</p><p>Final-over-Final Constraint. In Ways of Structure Building, Myriam Uribe-Etxebarria &</p><p>Vidal Valmala (eds), Cambridge: CUP.</p><p>Bowern, Claire. 2008. Syntactic change and syntactic borrowing in generative grammar. In</p><p>Principles of Syntactic Reconstruction, Gisella Ferarresi & Maria Goldbach (eds), 187–216.</p><p>Amsterdam: John Benjamins.</p><p>Brody, Michael. 1995. Lexico-Logical Form: A Radically Minimalist Theory. Cambridge MA: The</p><p>MIT Press.</p><p>Brody, Michael. 1998. Projection and phrase structure. Linguistic Inquiry 29: 367–398.</p><p>Cecchetto, Carlo. 2007. Some preliminary remarks on a ‘weak’ theory of linearization. Annali</p><p>Online di Ferrara – Lettere 1: 1–13.</p><p>Cecchetto, Carlo. To appear. Backwards dependencies must be short. A unified account of the</p><p>Final-over-Final and the Right Roof Constraints and its consequences for the syntax/mor-</p><p>phology interface. In Challenges to Linearization, Theresa Biberauer & Ian Roberts (eds).</p><p>Berlin: Mouton de Gruyter.</p><p>Chomsky, Noam. 1981. Lectures on Government and Binding Theory. Dordrecht: Foris.</p><p>Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36: 1–22.</p><p>Davison, Alice. 2005. Syntactic adjunction: Consequences for correlatives in Sanskrit and Hindi/</p><p>Urdu. Paper presented at Generative Linguistics in the Old World, GLOW in Asia V. (Delhi)</p><p>Davison, Alice. 2007. Word order, parameters, and the Extended COMP projection. In Linguistic</p><p>Theory and South Asian Languages, Essays in Honour of K.A. Jayaseelan [Linguistik Aktuell/</p><p>Linguistics Today 102], Josef Bayer, Tanmoy Bhattacharya & M.T. Hany Babu (eds), 175–198.</p><p>Amsterdam: John Benjamins.</p><p>Dresher, Elan. 1999. Charting the learning path: Cues to parameter setting. Linguistic Inquiry</p><p>30: 27–67.</p><p>Dryer, Matthew. 1992. The Greenbergian word order correlations. Language 68: 81–138.</p><p>Dryer, Matthew. 2005a. The position of polar question particles. In The World Atlas of Lan-</p><p>guage Structures, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds),</p><p>374–377. Oxford: OUP.</p><p>Dryer, Matthew. 2005b. The order of adverbial subordinator and clause. In The World Atlas of</p><p>Language Structures, Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie</p><p>(eds), 382–385. Oxford: OUP.</p><p>Evers, Arnold. 1975. The Transformational Cycle in Dutch and German. Ph.D. dissertation,</p><p>University of Utrecht.</p><p>Fischer, Olga, van Kemenade, Ans, Koopman, Willem & van der Wurff, Wim. 2000. The Syntax</p><p>of Early English. Cambridge: CUP.</p><p>Fuß, Eric & Trips, Carola. 2002. Variation and change in Old and Middle English: On the validity</p><p>of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics 4: 171–224.</p><p>Giorgi, Alessandra & Pianesi, Fabio. 1997. Tense and Aspect. From Semantics to Morphosyntax.</p><p>Oxford: OUP.</p><p>Greenberg, Joseph. 1963. Some universals of grammar with particular reference to the order</p><p>of meaningful elements. In Universals of Grammar, Joseph Greenberg (ed.), 73–113.</p><p>Cambridge MA: The MIT Press.</p><p>Impossible changes and impossible borrowings </p><p>Grimshaw, Jane. 1991. Extended projections. Ms, Brandeis University. Published as: Grimshaw,</p><p>Jane. 2000. Extended projection and locality. In Lexical Specification and Insertion [Current</p><p>Issues in Linguistic Theory 197], Peter Coopmans, Martin Everaert & Jane Grimshaw (eds),</p><p>115–134. Amsterdam: John Benjamins.</p><p>Haddican, William. 2004. Sentence polarity and word order in Basque. The Linguistic Review</p><p>21: 87–124.</p><p>Haegeman, Liliane & Riemsdijk, Henk van. 1986. Verb projection raising, scope and the typology</p><p>of rules affecting verbs. Linguistic Inquiry 7: 417–466.</p><p>Haider, Hubert. 1993. Deutsche Syntax, generativ. Tübingen: Narr.</p><p>Haider, Hubert. 2000. OV is more basic than VO. In The Derivation of VO and OV [Linguistik</p><p>Aktuell/Linguistics Today 31], Peter Svenonius, (ed.) 45–67. Amsterdam: John Benjamins.</p><p>Haig, Geoffrey. 2001. Linguistic diffusion in present-day East Anatolia: From top to bottom. In</p><p>Areal Diffusion and Genetic Inheritance: Problems in Comparative Linguistics, Alexandra Y.</p><p>Aikhenvald & Robert M.W. Dixon, (eds) 195–224. Oxford: OUP.</p><p>Harris, Alice & Campbell, Lyle. 1995. Historical Syntax in Crosslinguistic Perspective. Cambridge:</p><p>CUP.</p><p>Hawkins, John. 1983. Word Order Universals. New York NY: Academic Press.</p><p>Hawkins, John. 1990. Explaining Language Universals. Oxford: Blackwell.</p><p>Hawkins, John. 1994. A Performance Theory of Order and Consituency. Cambridge: CUP.</p><p>Heine, Bernd & Nurse, Derek. 2000. African Languages: An Introduction. Cambridge: CUP.</p><p>Hock, Hans Henrich. 1982. The Sanskrit quotative: A historical and comparative study. Studies</p><p>in the Linguistic Sciences 12(2): 39–85.</p><p>Hock, Hans Henrich & Joseph, Brian. 1996. Language History, Language Change and Language</p><p>Relationship: An Introduction to Historical and Comparative Linguistics. Berlin: Mouton</p><p>de Gruyter.</p><p>Holmberg, Anders. 2000. Deriving OV order in Finnish. In The Derivation of VO and OV</p><p>[Linguistik Aktuell/Linguistics Today 31], Peter Svenonius (ed.), 123–152. Amsterdam:</p><p>John Benjamins.</p><p>Holmberg, Anders. 2003. Yes/no questions and the relation between tense and polarity in English</p><p>and Finnish. Linguistic Variation Yearbook 3: 45–70.</p><p>Holmberg, Anders & Platzack, Christer. 1995. The Role of Inflection in Scandinavian Syntax.</p><p>Oxford: OUP.</p><p>Hróarsdóttir, Thorbjörg. 2000. Verb Phrase Syntax in the History of Icelandic. Ph.D. disserta-</p><p>tion, University of Tromsø. (Also published as: Hróarsdóttir, Thorbjörg. 2000. Word Order</p><p>Change in Icelandic: From OV to VO [Linguistik Aktuell/Linguistics Today 35]. Amsterdam:</p><p>John Benjamins.</p><p>Kayne, Richard. 1984. Connectedness and Binary Branching. Dordrecht: Foris.</p><p>Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press.</p><p>Keenan, Edward. 2002. Explaining the creation of reflexive pronouns in English. In Studies in</p><p>the History of English: A Millenial Perspective, Donka Minkova & Robert Stockwell (eds),</p><p>325–355. Berlin: Mouton de Gruyter.</p><p>Kellogg, Samuel. 1893. A Grammar of the Hindi Language. London: Kegan Paul & Co.</p><p>Kemenade, Ans van. 1987. Syntactic Case and Morphological Case in the History of English.</p><p>Dordrecht: Foris.</p><p>Kiparsky, Paul. 1995. Indo-European origins of Germanic syntax. In Clause Structure and</p><p>Language Change, Adrian Battye & Ian Roberts (eds), 140–170. Oxford: OUP.</p><p> Theresa Biberauer, Michelle Sheehan & Glenda Newton</p><p>Kiparsky, Paul. 1996. The shift to head-initial VP in Germanic. In Studies in Comparative</p><p>Germanic</p><p>in order to enhance our understanding of what makes</p><p>languages change and what in turn prevents them from changing. This volume is par-</p><p>ticularly timely, as there has been much interest recently in the topic of continuity,</p><p>beginning with the publication of Longobardi’s (2001) and Keenan’s (2002) articles on</p><p>‘inertia’ as a principle of, or rather, a constraint on language change.</p><p>1. Approaches to continuity and change</p><p>In addressing change and continuity, different views can be taken as to what constitutes</p><p>a ‘change’. From the point of view of the ‘Inertia Principle’, syntactic change should, in</p><p>principle, always be reducible to the categorial or structural reanalysis of lexical items</p><p>following changes in phonology or lexical semantics (plus, presumably, language-ex-</p><p>ternal factors, such as contact):</p><p>Syntactic change should not arise, unless it can be shown to be caused – that</p><p>is, to be a well-motivated consequence of other types of change (phonological</p><p>changes and semantic changes, including the appearance/disappearance of</p><p>whole lexical items) or, recursively, of other syntactic changes, given a plausible</p><p>theory of Universal Grammar (UG) and language acquisition.</p><p>(Longobardi 2001: 278)</p><p>This kind of perspective sees the very existence of syntactic change as conflicting with</p><p>an ideal model of a ‘perfect’ syntactic component which is replicated identically by</p><p>each successive generation. Thus, in applications of Longobardi’s theory, it is not the</p><p>functional structure of the clause that changes, but the filling of the terminal nodes –</p><p>by reanalyses of material subject to phonological and semantic change.</p><p>2 Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>Other approaches highlight the ways in which such reanalyses occur and therefore</p><p>where the lexical material filling terminal nodes comes from historically; for instance,</p><p>an element that used to realize a lower functional head (or its specifier) and move to a</p><p>higher position may later come to realize the higher position by loss of movement as</p><p>part of reanalysis (Roberts & Roussou 2003).</p><p>Then there are approaches which seek to model the properties of linguistic change</p><p>mathematically. Typical changes (not merely syntactic ones) follow an S-shaped curve,</p><p>as the replacement of an old form by a new one proceeds slowly at first, then gathers</p><p>speed, and slows down again as it nears completion (Bailey 1973; Kroch 1989). Since</p><p>the work of Kroch, it has become standard to model this property of (linguistic) change</p><p>by using the logistic function. In the present volume, Postma proposes refinements to</p><p>this that allow him to capture the fact that the peak of a failed change coincides with</p><p>the turning point of a related S-curve.</p><p>Also relevant to this volume are approaches which see grammaticalization as an</p><p>important driving force for change, and which see unidirectionality as a definitional</p><p>property of grammaticalization (e.g. Haspelmath 1999). While it is clear that changes</p><p>that involve more lexical items becoming more functional are extremely common,</p><p>there is growing evidence to suggest that changes from more functional to more lexical</p><p>are not as rare as has been thought. Two further instances of this latter type of change</p><p>are discussed in the contributions by Marten and Rosenkvist.</p><p>Many of these different approaches have in common that they see language</p><p>acquirers as the agents of syntactic change, whether these are infants acquiring their</p><p>first language or older children and adults acquiring a second language.</p><p>Children acquiring their first language have to abduce the grammar of that lan-</p><p>guage on the basis of indeterminate evidence from their linguistic environment, and</p><p>may therefore arrive at a different analysis to previous generations of a given underlying</p><p>structure. This can be explained as coming about due to surface ambiguity in the real-</p><p>ization of that structure, arising for instance through morphological levelling (Ander-</p><p>sen 1973; Timberlake 1977; Kroch 1989; Lightfoot 1991, 1999). However, reanalyses of</p><p>this type can be hard to identify until their effects become evident by spreading beyond</p><p>the original context of the reanalysis itself (cf. Timberlake 1977; Harris & Campbell</p><p>1995, among others). Aldridge’s chapter in the present volume describes such a case,</p><p>where surface ambiguity leads to structural reanalysis, as a former topic in SpecCP is</p><p>reanalysed as a subject in SpecTP. The ‘spreading’, or actualization, of this change is</p><p>witnessed by an increase in the A-properties and a decrease in the A′-properties of the</p><p>elements involved.</p><p>On the other hand, the trigger experience of first language acquirers may also be</p><p>altered by previous ‘imperfect’ language acquisition on the part of adult second lan-</p><p>guage learners. The articles collected in this volume that address contact as a source of</p><p>Introduction 3</p><p>linguistic change make it clear that the degree of bilingualism in a given contact situa-</p><p>tion plays an important role in determining the type and quality of the change.</p><p>First of all, it is clear that contact tends not to lead to significant structural change</p><p>unless it is intensive and long-term, whereas lexical borrowing seems to be possible</p><p>even in cases of minimal contact; cf. Thomason and Kaufman’s (1988: 50) ‘scale of</p><p>interference’. Illustrating this point, Sundquist (this volume) argues that Middle Low</p><p>German had no influence on the syntax of Middle Norwegian during the time of the</p><p>Hansa, where the contacts were restricted to the domain of business and trade, and</p><p>where the languages were similar enough for speakers to accommodate to each oth-</p><p>er’s word order by using variants available within their own language if necessary.</p><p>However, it seems that situations of intensive and extensive contact, where many</p><p>children simultaneously acquire two first languages, can also be environments for (syn-</p><p>tactic) continuity. Paradoxical as this may seem, this follows from the fact that children</p><p>have little difficulty differentiating and acquiring multiple languages in their environ-</p><p>ment, and can arrive at grammars of those languages which do not differ systemati-</p><p>cally from those of their monolingual peers (cf. Meisel 1995, 2001). This is illustrated</p><p>by Davies and Deuchar’s contribution, which shows that there is very little evidence</p><p>that the extensive Welsh–English bilingualism found in much of Wales has led the</p><p>syntax of Welsh to converge with that of English.</p><p>Vulchanov & Vulchanova suggest that Old Bulgarian (spoken in the ninth and</p><p>tenth centuries) was influenced by contact with New Testament Greek. Although they</p><p>do not discuss this within the confines of their chapter, this raises the intriguing ques-</p><p>tion of whether transfer of basic syntactic properties can proceed entirely through</p><p>the written medium or whether vernacular (rather than elite) bilingualism is required</p><p>for transfer of syntactic features. In this case, they prefer to suggest contact via the</p><p>New Testament Greek Gospel translations, rather than making primary reference</p><p>to the extensive bilingualism between Slavonic varieties and Byzantine Greek in the</p><p>Balkans. Such a view implicitly rejects child language acquisition as the locus of syn-</p><p>tactic change.</p><p>The absence of or delay in syntactic change in cases where it would be expected,</p><p>for example on the basis of other languages having undergone a parallel development</p><p>under otherwise similar conditions, is perhaps expected under an ideal model of first</p><p>language acquisition, in which the language system of the parent generation is per-</p><p>fectly replicated in the acquiring generation. But ideal models are empirically inad-</p><p>equate. For instance, most languages are in reality exposed to some form of language</p><p>contact. Already the existence of dialects is enough to create minimal but potentially</p><p>significant changes in the trigger experience for language acquirers. Given the ubiq-</p><p>uity of syntactic change, it is rather surprising to find the absence</p><p>Syntax II, Höskuldur Thráinsson, Samuel Epstein & Steve Peter (eds), 140–179.</p><p>Dordrecht: Kluwer.</p><p>Klamer, Marian. 2001. Expressives and iconicity in the lexicon. In Ideophones [Typological</p><p>Studies in Language 44], Erhard Voeltz & Christa Kilian-Hatz (eds), 165–181. Amsterdam:</p><p>John Benjamins.</p><p>Kroch, Anthony & Taylor, Ann. 2000. Verb-object order in Middle English. In Diachronic</p><p>Syntax: Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds),</p><p>132–163. Oxford: OUP.</p><p>Kuiper, Franciscus Bernardus Jacobus. 1974. The genesis of a linguistic area. International Journal</p><p>of Dravidian Languages 3(1): 135–153.</p><p>Laka, Itziar. 1994. On the Syntax of Negation. New York NY: Garland.</p><p>Ledgeway, Adam. To appear. Syntactic and morphosyntactic typology and change in Latin</p><p>and Romance. In The Cambridge History of the Romance Languages, Martin Maiden,</p><p>John Charles Smith & Adam Ledgeway (eds). Cambridge: CUP.</p><p>Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: CUP.</p><p>Lightfoot, David. 1991. How to Set Parameters: Arguments from Language Change. Cambridge</p><p>MA: The MIT Press.</p><p>Lightfoot, David. 1999. The Development of Language. Oxford: Blackwell.</p><p>Longobardi, Giuseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The history</p><p>of French chez. Linguistic Inquiry 32: 275–302.</p><p>Marlow, Patrick Edward. 1997. Origin and Development of the Indo-Aryan Quotatives and</p><p>Complementizers: An Areal Approach. Ph.D. dissertation, University of Illinois.</p><p>Masica, Colin. 1991 The Indo-Aryan languages. Cambridge: CUP.</p><p>Meenakshi, K. 1986. The quotative in Indo-Aryan. In South Asian Languages: Structure, Conver-</p><p>gence and Diglossia, Bhadriraju Krishnamurti (ed.), 209–218. Dehli: Motilal Banarsidass.</p><p>Mobbs, Iain. 2008. ‘Functionalism’, the Design of the Language Faculty, and (Disharmonic)</p><p>Typology. Mphil thesis, University of Cambridge.</p><p>Moore, Denny. 1989. Gaviao nominalisations as relative clause and sentential complement</p><p>equivalents. IJAL 55: 309–25.</p><p>Müller, Reinmar & Reis, Marga. 2001. Modalität und Modalverben im Deutschen. Hamburg:</p><p>Buske Verlag.</p><p>Nayudu, Aarti. 2008. Issues in the Syntax of Marathi: A Minimalist Approach. Ph.D. dissertation,</p><p>Durham University.</p><p>Newmeyer, Frederick. 1998. Language Form and Language Function. Cambridge MA: The</p><p>MIT Press.</p><p>Ottaviano, Ida de. 1980. Textos tacana. Riberalta, Bolivia: Instituto Linguístico de Verano en</p><p>colaboración con el Ministerio de Educación y Cultura.</p><p>Pandharipande, Rajeshwari. 1997. Marathi. London: Routledge.</p><p>Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and Change in Old English</p><p>Word Order. Ph.D. dissertation, University of Pennsylvania. (Also published as: Pintzuk,</p><p>Susan. 1999. Phrase Structures in Competition. Variation and Change in Old English Word</p><p>Order. New York NY: Garland).</p><p>Pintzuk, Susan. 2002. Verb order in Old English: Variation as grammatical competition. In</p><p>Diachronic Syntax: Models and Mechanisms, David Lightfoot (ed.), 132–163. Oxford: OUP.</p><p>Ponelis, Frederick. 1993. The Development of Afrikaans. Duisburg: Peter Lang.</p><p>Impossible changes and impossible borrowings </p><p>Riemsdijk, Henk van. 1998. Categorial feature magnetism: The endocentricity and distribution</p><p>of projections. Journal of Comparative Germanic Linguistics 2: 1–48.</p><p>Rizzi, Luigi. 1997. On the fine structure of the left periphery. In Elements of Grammar, Liliane</p><p>Haegeman (ed.), 281–338. Dordrecht: Kluwer.</p><p>Rizzi, Luigi. 2001. On the position ‘Int(errogative)’ in the left periphery of the clause. In Current</p><p>Studies in Italian Syntax. Essays Offered to Lorenzo Renzi, Guglielmo Cinque & Giampaolo</p><p>Salvi (eds), 287–296. Amsterdam: Elsevier.</p><p>Roberts, Ian. 2007. Diachronic Syntax. Oxford: OUP.</p><p>Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammaticaliza-</p><p>tion. Cambridge: CUP.</p><p>Sheehan, Michelle. 2009a. The Final-over-Final Constraint as an effect of Complement Stranding.</p><p>Newcastle Working Papers in Linguistics 15: 104–125.</p><p>Sheehan, Michelle. 2009b. Labelling, multiple Spell-Out and the Final-over-Final Constraint. In</p><p>Proceedings of XXXV Incontro di Grammatica Generativa, Vincezo Moscati & Emilio Servidio</p><p>(eds), 231–243. Siena: CISCL.</p><p>Sheehan, Michelle. To appear. Formal and functional approaches to disharmonic word orders. To</p><p>appear in Newcastle Working Papers in Linguistics 16.</p><p>Southworth, Franklin C. 2005. Linguistic Archaeology of South Asia. London: Routledge-Curzon.</p><p>Thomason, Sarah. G. & Kaufman, Terrence. 1988. Language Contact, Creolization, and Genetic</p><p>Linguistics. Berkeley CA: University of California Press.</p><p>Travis, Lisa. 1984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MIT.</p><p>Continuity is change</p><p>The long tail of Jespersen’s cycle in Flemish*</p><p>Anne Breitbarth & Liliane Haegeman</p><p>Ghent University</p><p>All West Germanic languages have completed Jespersen’s Cycle, replacing a</p><p>preverbal negation marker (stage I) with a post-verbal one (stage III) after a</p><p>period of transition during which the two could co-occur (stage II). Only the</p><p>Flemish dialects have maintained the old preverbal marker to some extent. The</p><p>present chapter addresses this apparent historical continuity, establishing that</p><p>Flemish en was in fact saved from the fate of its West Germanic cognates by a</p><p>morphosyntactic change. Based on an analysis of the contexts in which preverbal</p><p>en is found in present-day Flemish dialects, it is argued that en has evolved into</p><p>a marker of emphasis of polarity, and that contrary to appearances, the Flemish</p><p>dialects have in fact reached stage III.</p><p>1. Introduction</p><p>All West Germanic languages have undergone a development in the expression of nega-</p><p>tion known as Jespersen’s Cycle (term by Dahl 1979, after observations by Jespersen</p><p>1917), by which an original preverbal negator is first joined by a newly grammatical-</p><p>ized adverbial marker and eventually supplanted by it. The rough stages of the cycle are</p><p>demonstrated for English in (1).1</p><p>*We gratefully acknowledge funding by the Arts and Humanities Research Council, award</p><p>no. AR119272 (Anne Breitbarth) and the Vlaamse Fonds voor Wetenschappelijk Onderzoek,</p><p>2009-Odysseus grant no. G091409 (Liliane Haegeman).</p><p>1.  More fine-grained subdivisions are possible and have been proposed in great numbers</p><p>and variety, cf. e.g. Posner (1985), Schwegler (1988), Ladusaw (1993), Rowlett (1998), Van</p><p>Kemenade (2000), Zeijlstra (2004), Schwenter (2006), Barbiers et al. (2008) and Willis (to</p><p>appear). The rough division into three stages given in (1) is, however, sufficient for the pur-</p><p>poses of this chapter.</p><p> Anne Breitbarth & Liliane Haegeman</p><p>(1)</p><p>Jespersen (1917: 9–11)</p><p>All West Germanic languages have, furthermore, like English, completed this develop-</p><p>ment, that is, they have all reached at least stage III. As is well known, the only exception</p><p>to this is the Flemish dialects, which, as widely discussed in the literature (Haegeman</p><p>1995 and much subsequent work; Haegeman & Zanuttini 1991, 1996; Barbiers et al. 2008),</p><p>have preserved, apparently optionally, the use of the preverbal marker. They therefore</p><p>appear to be still in transition from stage II to stage III.</p><p>(2) Valère (en) klaapt nie.</p><p>Valère ‘neg’ talks ‘neg’</p><p>‘Valère doesn’t talk.’ (West Flemish (Lapscheure); Haegeman 1995)</p><p>The question the present chapter wishes to address is why these Flemish dialects show</p><p>this apparent historical continuity. By looking in detail at the contexts in which pre-</p><p>verbal en is found in present-day Flemish dialects we will establish that it is in fact a</p><p>morphosyntactic change that saved Flemish en from the fate of its other West Germanic</p><p>cognates, and we will argue that, contrary to appearances, the Flemish dialects are not in</p><p>transition between stage II and stage III of Jespersen’s Cycle, but have reached stage III.</p><p>. Present-day Flemish en</p><p>As just indicated, the Flemish dialects are known for having preserved the old prever-</p><p>bal marker en in the expression of negation, at least optionally.</p><p>We will show in the</p><p>following that this optionality is only apparent and that an argument can be made for</p><p>en having acquired properties that suggest it is no longer a negation marker. This has</p><p>already been observed for West Flemish by Haegeman (2000, 2001, 2002), but is, as we</p><p>will see, a wider Flemish phenomenon.</p><p>.1 En is a polarity marker in Flemish</p><p>In both East and West Flemish, purely formal, that is, non-negative, uses of en in non-</p><p>negative affective contexts can be found. The examples below from East Flemish Ghent</p><p>demonstrate the availability of non-negative en in conditional clauses (3), comparative</p><p>clauses (4) and the context of maar ‘only’ (5).</p><p>Stage I Stage II Stage III Stage I’</p><p>ic ne secge I seye not I say not I do not say ></p><p>I don’t say</p><p>Old English Middle</p><p>English</p><p>Early Modern</p><p>English</p><p>Modern</p><p>English</p><p>Continuity is change </p><p>(3) en aa’t slecht weer en is</p><p>and if it bad weather en is</p><p>‘and if the weather is bad’ (Ghent; Leemans 1966: 191)</p><p>(4) Ge moet ‘tzegge gelijk of ‘t en is.</p><p>you must it say like if it en is</p><p>‘You must tell it the way it is.’ (Ghent; Leemans 1966: 191)</p><p>(5) ten is maar een kleintsje</p><p>it.en is only a little-one</p><p>‘it’s only a little one’ (Ghent; Tavernier 1959: 246)</p><p>In West Flemish, non-negative en can be used in before-clauses (6), as well as the con-</p><p>text of maar ‘only’ (7), as in East Flemish.</p><p>(6) Je moet niet komen voordat ik geschreven en heb.</p><p>you should not come before I written en have</p><p>‘You should not come before I have written.’</p><p>(Kortrijk, Barbiers et al. 2008: 60)</p><p>(7) K’en een mo drie marbels.</p><p>I en have but three marbles</p><p>‘I only have three marbles.’ (Lapscheure; LH’s intuitions)</p><p>Besides such non-negative uses of en, there is of course the well-known preservation</p><p>of en in negative clauses. But here as well we will see that en is no longer simply a</p><p>negation marker.</p><p>. En marks emphasis on polarity</p><p>In the literature it has occasionally been observed (cf. e.g. Overdiep 1937: 457), that</p><p>en in the Flemish dialects lays a certain additional emphasis on the polarity of the</p><p>clause it occurs in, in particular where it occurs with sentential negation. There are</p><p>two types of context which show that Flemish en has somehow moved on to become</p><p>a marker of emphasis. First, it is used in elliptical replies with doen ‘do’ functioning</p><p>as emphatic contradictions. Second, it is used in negative clauses where its function</p><p>is to emphatically negate or contradict an expectation either explicit or implicit in</p><p>the context. We will provide empirical evidence for this proposal in the following</p><p>two subsections.</p><p>..1 Emphatic contradictions</p><p>Emphatic contradictions with en are found in most Flemish dialects. These are ellipti-</p><p>cal replies emphatically contradicting a preceding statement (8a, b) or question (8c).</p><p>They have a rather fossilized structure, consisting of a finite form of vicarious doen ‘do’</p><p>in the present tense and a weak pronominal subject. For many speakers this can also</p><p> Anne Breitbarth & Liliane Haegeman</p><p>be impersonal ‘t ‘it’, which then does not agree with the subject given in the preceding</p><p>context.2</p><p>(8) a. A: Hij slaapt.</p><p>he sleeps</p><p>‘He is asleep.’</p><p>B: Hij/’t en doet.</p><p>he en does</p><p>‘No, he isn’t.’</p><p>b. A: Hij zal niet komen.</p><p>he will not come</p><p>‘He won’t come.’</p><p>B: Hij/’t en doet.</p><p>he en does</p><p>‘No, he WILL.’</p><p>c. A: Slaapt hij?</p><p>sleeps he</p><p>‘Is he asleep?’</p><p>B: Hij/’t en doet.</p><p>he en does</p><p>‘No he isn’t.’ (after Barbiers et al. 2008: 60)</p><p>Although (8a) and (8c) may give the impression that this is a fossilized use of archaic</p><p>stage I negation, (8b) shows that this cannot be true: when a negative statement is contra-</p><p>dicted by means of the ‘t en doet construction, this amounts to an emphatic affirmation.</p><p>The same is found in the following West Flemish example from Haegeman (1995):</p><p>(9) A: Valère verkuopt da nie</p><p>Valère sells that not</p><p>‘Valère doesn’t sell that.’</p><p>B: J’en doet. Je verkoopt da wel</p><p>he en does He sells that indeed</p><p>K’een der gisteren gekocht.</p><p>I-have some yesterday bought</p><p>‘He does. He does sell that. I bought some yesterday.’</p><p>(Lapscheure; Haegeman 1995: 160)</p><p>Clearly, en in the ‘t en doet construction is not equivalent to the old stage I negation;</p><p>rather, it expresses an emphatic contradiction, whether the contradicted statement or</p><p>question is positive or negative.</p><p>.  Cf. Ryckeboer (1986), Haegeman (2001, 2002), Barbiers et al. (2008).</p><p>Continuity is change </p><p>.. Contradicting an expectation in negative sentences</p><p>Besides these emphatic contradictions with doen, where en occurs on its own, en is</p><p>also used together with the sentential negator nie(t) or n-words,3 which is the basis of</p><p>the commonly held view that the Flemish dialects are still in transition between stage</p><p>II and stage III of Jespersen’s Cycle. The reason why this use of en is not likely to be</p><p>‘just’ an illustration of stage II negation, that is, negation expressed by a discontinuous</p><p>marker, is that en has a low, though stable, frequency in the Flemish dialects, ranging</p><p>around 10% of all negative sentences with nie(t) or n-words, as the data in studies such</p><p>as Leemans (1966) or De Pauw (1973) show.4</p><p>Comparing the cases with and without en, it appears that en is used in contradic-</p><p>tions of an (implicit) expectation. This is confirmed for both West and East Flemish. In</p><p>(10), from West Flemish, the request uttered by speaker A presupposes that speaker B</p><p>is able to comply with the request. B, in order to deny this ability emphatically by deny-</p><p>ing the presupposition, can use en. Examples (10a) and (10b) show that this is possible</p><p>in both main and embedded clauses.</p><p>(10) A: Geef me nen.keer Valère zenen telefoon.</p><p>give me once Valère his phone</p><p>‘Can you give me Valère’s phone number?’</p><p>a. B: K’(en) een-k ik Valère zenen telefon nie.</p><p>I (en) have-I I Valère his phone neg</p><p>‘I don’t have Valère’s number.’</p><p>b. B: K-zeggen jen toch dan-k em nie (en)-een.</p><p>I say you prt that-I him neg (en) have</p><p>‘I am telling you I don’t have it.’ (Lapscheure, Haegeman 2002: 11)</p><p>The same holds in East Flemish. As already observed by Overdiep (1937: 457), nega-</p><p>tion with en…niet instead of niet alone, as found for example in the dialect of East</p><p>Flemish Ghent, is more emphatic. In (11a), the common knowledge that Ghent is</p><p>famous for its university is given in the background. By using en, the speaker expresses</p><p>his/her surprise at the interlocutor not knowing this university despite having lived in</p><p>the city for fifteen years. In (11b), the context (not given) seems to be an implication or</p><p>direct accusation of the speaker’s having made a mistake, which is emphatically denied</p><p>in the example.</p><p>.  An n-word (Laka 1990) is an indefinite pronoun, determiner or adverb, such as, for</p><p>instance, nadie ‘no one’, ningún ‘no’ or nunca ‘never’ in Spanish, which potentially partici-</p><p>pates in negative concord and is able to express negation on its own in a fragment answer</p><p>(Giannakidou 2006).</p><p>.  Unfortunately, there are no more recent corpus studies of the use of en in Flemish dialects.</p><p>We are aiming to address this desideratum in future work.</p><p> Anne Breitbarth & Liliane Haegeman</p><p>(11) a. Ge weunt al vijftien jaor in Gent, in g’en ken nog</p><p>you live already fifteen years in Ghent and you-en know still</p><p>d’Universiteit nie?</p><p>the university not</p><p>‘What? You have lived in Ghent for fifteen years, and you are telling me you</p><p>still don’t know the university?!’</p><p>b. Dat en is toch mijn schuld nie!</p><p>that en is prt my fault not</p><p>‘You can’t say that this is MY fault!’ (Ghent, Overdiep 1937: 456–457)</p><p>Summarizing, what en in combination with sentential negation expresses is an empha-</p><p>sis on this negation, contradicting a presupposition or assumption that the contrary</p><p>state of affairs holds present in the context.</p><p>.  The historical development of Flemish en</p><p>Both purely formal uses as well as the emphatic use of en are already found in the his-</p><p>tory</p><p>of Dutch. Purely formal, or expletive, uses are rare in the complement of adversa-</p><p>tive predicates (like doubt, deny, forbid etc.) in Middle Dutch (Burridge 1993: 184 –185),</p><p>but regular for example in the standard of comparison (Beheydt 1998):</p><p>(12) Ghien moget niet vorder rechten dan u manne en wijsen.</p><p>you-ne can neg more judge than your men ne tell</p><p>‘You cannot judge more than your men tell you.’ (from Beheydt 1998: 16)</p><p>Such non-negative uses spread further in the seventeenth century to clauses with maer</p><p>‘only’ and nauw ‘nearly, hardly’, besides the standard of comparison:</p><p>(13) hoe wel ter nauwer noodt verhaalens waart en is</p><p>how well to near need telling worth ne is</p><p>‘although it is hardly worth telling’ (Weijnen 1956: 73)</p><p>Furthermore, elliptical replies with vicarious doen ‘do’ are already attested in Middle</p><p>Dutch, the (still only negative) predecessor of the emphatic contradictions in (8):</p><p>(14) Ghi hout u spot. In (=ic+en) doe</p><p>you hold your mockery. I=ne do</p><p>‘You are mocking. I do not!’ (from Beheydt 1998: 15)</p><p>As discussed by Burridge (1993), the preverbal marker en in combination with niet</p><p>was lost during the seventeenth century in the northern Dutch provinces. It has been</p><p>argued that this is the consequence of a ban by prescriptive grammarians and influen-</p><p>tial writers. Burridge shows how it disappears from the letters of P.C. Hooft (1581–1647)</p><p>during his lifetime; similarly, Van der Wouden (1995: 23) shows how playwright Joost</p><p>Continuity is change </p><p>van den Vondel “after the model of […] writers such as P. C. Hooft” deliberately reduces</p><p>the use of en in his plays over the years.</p><p>However, the reduction of the use of en took place to a much lesser extent in the</p><p>southern provinces. The reasons for this have hardly been addressed in the literature,</p><p>which for the period since the Gouden Eeuw (‘Golden Age’, roughly, the seventeenth</p><p>century) has almost exclusively focused on the north. Interestingly, Beheydt (1998)</p><p>hints at a possible meaning difference arising in the early modern period between the</p><p>older bipartite and the newer single postverbal expression of negation:</p><p>(…) in vraagzinnen (…) blijkt de negatie steeds postverbaal te zijn als de</p><p>betekenis positief is. Het lijkt heel aannemelijk dat de taalgebruikers vonden dat</p><p>de tweeledige ontkenning de negatieve betekenis te zeer benadrukte, wat minder het</p><p>geval was met de postverbale.5 (Beheydt 1998: 93)</p><p>We conclude that the emphatic use of en in conjunction with niet or n-words may thus</p><p>have developed in the southern dialects already during the seventeenth century.</p><p>. An asymmetric view of stage II</p><p>The standard approach to stage II of Jespersen’s cycle, that is, the stage at which the</p><p>original and the new negation marker coexist, is that both elements are negation mark-</p><p>ers. We can call this the symmetric approach to stage II of Jespersen’s cycle. Approaches</p><p>differ with respect to the direction of the change, that is, whether it is the weakening of</p><p>the preverbal marker that drives the development, or whether the grammaticalization</p><p>of a new phrasal negator causes the weakening of the old preverbal marker and ulti-</p><p>mately makes it superfluous. Both types of approach agree that it is the redundancy or</p><p>lack of iconicity in the presence of two negation markers that causes the loss of the pre-</p><p>verbal marker and the transition to stage III. Abraham (2003: 360), a proponent of the</p><p>former approach writes, “Under criteria of economy, the attrition and total erosion of</p><p>the weak verbal neg-prefix is entirely expectable”,6 and according to Detges & Waltereit</p><p>.  “In questions, the negation seems to be invariably postverbal in case the meaning is posi-</p><p>tive. It is very likely that language users thought that the bipartite negation put too much</p><p>emphasis on the negative meaning, which was less the case with the postverbal negator” (our</p><p>emphasis).</p><p>.  However, it is not the case that systematically a ‘weakened’ clitic marker of negation</p><p>is replaced by a ‘stronger’ phrasal one. Lindstad (2007) shows that in the development of</p><p>Old Norse a preverbal clitic marker of negation is replaced by a postverbal clitic. Thus the</p><p>“replacing marker is not an adverb, perceived as ‘stronger’ or ‘heavier’ than the replaced one,</p><p>but a suffix on the verb” (Lindstad 2007: 35).</p><p> Anne Breitbarth & Liliane Haegeman</p><p>(2002), a functional principle of ‘constructional iconicity’ is behind the demise of the</p><p>original negator, since a “formally discontinuous functional unit” (Detges & Waltereit</p><p>2002: 186) is less iconic than the expression of a single function by a single element,</p><p>and therefore dispreferred and prone to simplification.</p><p>Generative accounts of stage II of Jespersen’s Cycle operate with different varia-</p><p>tions on the theme of feature valuation or agreement, whereby one of the two ele-</p><p>ments in a bipartite negation licenses the other. Rowlett (1998), and following him,</p><p>Wallage (2005, 2008), for example, propose that the preverbal marker loses the value</p><p>[neg] of its polarity feature [pol: ] and therefore requires a licenser that is able to pro-</p><p>vide this value.</p><p>We have seen that the old preverbal marker en in Flemish bipartite negation has</p><p>developed a function separate from expressing negation, and that this development</p><p>goes back to Middle Dutch. We therefore propose that the prolonged period of coex-</p><p>istence of the two elements, the old and the new negation marker, entails their func-</p><p>tional differentiation. That is, we assume an asymmetric view on stage II of Jespersen’s</p><p>Cycle in (historical) Flemish, under which the two elements present are not both at</p><p>the same time related to the expression of negation. This provides us with an elegant</p><p>account of the actual distribution of en in the Flemish dialects.</p><p>Breitbarth (2009) proposes analysing the preverbal marker at stage II of Jespers-</p><p>en’s cycle in the West Germanic languages as a polarity marker rather than a negator.</p><p>This leads to an “asymmetric” interpretation of stage II ‘bipartite’ ‘negation’, as the two</p><p>markers do not conspire in the expression of negation. Rather, the two elements come</p><p>to be associated with different functional heads. We will use this analysis here, fine-</p><p>tuning it to the development of Flemish en.7</p><p>Like Moscati (2006), we propose that polarity is a clause-typing feature located on</p><p>a functional head in the left periphery. Although a more fine-grained representation</p><p>of the CP-system may be necessary, let us for simplicity assume that this head is C and</p><p>that the polarity clause typing feature is [±affective], building on Klima’s (1964) char-</p><p>acterization of weak NPI contexts as affective. Now, in Middle English, Middle Dutch,</p><p>Middle High German and Middle Low German, the polarity element ne/en is a bound</p><p>morpheme. It therefore needs a host, which in the West Germanic languages is the</p><p>finite verb. As the finite verb is not always in a position in the left periphery, Breitbarth</p><p>(2009) proposes that ne/en is spelt out on the finite verb by the mechanism of Alterna-</p><p>tive Realization (Emonds 1987, 2000):</p><p>.  A similar analysis has been developed by Haegeman (2000, 2001, 2002) for West Flemish</p><p>from a synchronic point of view. Haegeman proposes to analyse the preverbal marker en as a</p><p>marker of polarity, rather than negation. Her analysis differs from the one presented here in</p><p>that she places PolP between TP and a projection called FP situated between TP and AgrOP.</p><p>Continuity is change </p><p>(15) Alternative Realization</p><p>A syntactic feature F canonically associated in UG with category B can be alter-</p><p>natively realized in a closed class grammatical morpheme under X0, provided</p><p>X0 is the lexical head of a sister of B. (Emonds 2000: definition (4.20))</p><p>That is, the feature [+affective] is realized on the lexical head of a sister of C[+affective].</p><p>Intervening functional heads not lexicalized in a structure are ignored because of the</p><p>definition of Extended Sisterhood, e.g. heads in the TP and CP-domains in our case,</p><p>if</p><p>the verb stays low as in embedded clauses in (continental) West Germanic.</p><p>(16) Extended Sisterhood</p><p>If Z0 and XP are sisters and if Z1 is the smallest phrase (besides structural</p><p>projections of Y) whose lexical head is Y0, then Y0 and XP are extended sisters.</p><p>(Emonds 2000: definition (4.28))</p><p>The reanalysis of the former negative clitic on the finite verb as an exponent of affective</p><p>polarity was possible exactly because of the Extended Sisterhood relation between T</p><p>and Pol. This mechanism is similar to Roberts & Roussou’s (2003) ‘upward reanalysis’</p><p>by which the expression of a lower functional head becomes associated with the fea-</p><p>tures of a higher functional head due to a derivational link between the two heads.8</p><p>This reanalysis became possible once sentential negation had found a new exponent,</p><p>namely when an erstwhile adverbial emphasizer is reanalysed as the expression of</p><p>sentential negation. After its reanalysis as the Alternative Realization of the feature</p><p>[+affective] of C in negative clauses, ne/en could be generalized to non-veridical con-</p><p>texts beyond negation.</p><p>An argument for the initial reanalysis of en as the Alternative Realization of C’s</p><p>[+affective] on the finite verb in Flemish comes from comparative historical evidence.</p><p>In early Middle Dutch as well as in Middle High German (Jäger 2008: 289–294) and</p><p>Middle Low German, n-marked indefinites, where they do not mark sentential nega-</p><p>tion on their own, only seem to co-occur with the old preverbal marker, but not with</p><p>the new adverbial one. Table 1 illustrates this for early Middle Dutch, and Table 2 for</p><p>Middle Low German.9</p><p>.  Roberts & Roussou originally assume that the result of such a reanalysis is the loss of</p><p>movement, which previously connected the two heads. We see no reason why agreement</p><p>alone cannot already be a sufficient connection between two positions.</p><p>.  These figures are based on corpora of chancery documents (oorkonden/Urkunden). In</p><p>the case of early Middle Dutch these are the first 300 documents from the Gysseling corpus</p><p>(Gysseling 1977); the Middle Low German corpus is the one discussed in Breitbarth (2008).</p><p>It should be noted that co-occurrence of the adverbial negator and n-words does become</p><p>available in later Middle Dutch, late Middle Low German and Early New High German, but</p><p> Anne Breitbarth & Liliane Haegeman</p><p>Table 1. Preverbal and adverbial markers with n-words in early Middle Dutch</p><p>1200–1280</p><p>en..n-word n-word alone n-word…niet Total</p><p>49 (75.4%) 16 (24.6%) 0 (0.0%) 65</p><p>Table 2. Preverbal and adverbial markers with n-words in Middle Low German</p><p>1325–1575</p><p>en..n-word n-word alone n-word…nicht total</p><p>487 (37.4%) 812 (62.4%) 2 (0.2%) 1301</p><p>This evidence shows that already early on, n-words are able to mark sentential</p><p>negation on their own. The fact that they do not or prefer not to co-occur with the</p><p>new adverbial negator, which is taking over the expression of negation,10 suggests that</p><p>the preverbal marker is no longer a marker of sentential negation, at least not when</p><p>accompanied by an n-word. Instead, the expression of negation elsewhere in the clause</p><p>must have made the reanalysis of the preverbal marker as the alternative realization of</p><p>[+affective] on C seem likely to language acquirers.</p><p>We have seen above already (Section 3) how preverbal en came to be used in</p><p>non-negative contexts in Middle Dutch, and spread to further such contexts in the</p><p>present-day Flemish dialects. We can understand this as the actualization (Timberlake</p><p>1977, Harris & Campbell 1995) of the reanalysis of en, making the reanalysis visible</p><p>in the language.</p><p>Summing up, there are developments that suggest to the language acquirer that</p><p>the old preverbal marker may be a polarity marker, namely other negative markers</p><p>such as n-words becoming able to express negation without it, and this reanalysis is</p><p>actualized when non-negative affective uses appear.</p><p>. Flemish preverbal en: Continuity is change</p><p>We have seen that, instead of being lost like its cognates in the other West Germanic</p><p>dialects, Flemish en has (a) spread to more non-negative affective contexts than are</p><p>not initially when the assumed reanalysis of the preverbal element occurred. The later devel-</p><p>opments are therefore likely to have a different explanation which goes beyond the scope of</p><p>this chapter.</p><p>1.  Niet in Middle Dutch, nicht in Middle Low German and Middle High German and not</p><p>Middle English are clearly no longer emphatic and are increasingly used on their own (without</p><p>ne/en) as sentential negators. They are furthermore used for negating non-finite verbs and for</p><p>constituent negation, which the inherited preverbal markers are not capable of.</p><p>Continuity is change 1</p><p>found in Middle Dutch, thereby stabilizing the position of en in the language, and (b)</p><p>that it has acquired the ability to realize an additional feature, namely emphasis on</p><p>polarity. We propose that en developed historically as follows.</p><p>In a first step that all West Germanic languages underwent, the negation particle</p><p>ne/en on the finite verb is reanalysed as the Alternative Realization of C’s [+affective]</p><p>feature. This, as argued in Section 4, qualifies as a grammaticalization by ‘upward</p><p>reanalysis’ in the sense of Roberts & Roussou (2003) as a lower head becomes associ-</p><p>ated with the features of a higher head:11</p><p>(17) stage I → stage II in Middle Dutch</p><p>[CP C[+aff] [ … [VP … ] [T en/ne[+neg]-Vfin]] →</p><p>[CP C[+aff] [ … niet[+neg] [VP …] [T en/ne[+aff]-Vfin]]</p><p>The first line in (17) represents the structure of a negative clause in Old Dutch. By</p><p>default, C in negative clauses carries the feature [+affective], representing the polarity</p><p>of the clause at its interface. Sentential negation is expressed by the clitic ne, which</p><p>invariably attaches to the finite verb. When n-words become able to express senten-</p><p>tial negation on their own (that is, when they become n-indefinites), without overt</p><p>realization of ne, ne becomes ambiguous between being a sentential negator in nega-</p><p>tive clauses without n-words, and a potential spell-out of [+affective] in clauses with</p><p>n-words. With the grammaticalization of the new phrasal negator niet, ne comes to</p><p>be associated with the realization of [+affective] – as the Alternative Realization of C’s</p><p>feature on the finite verb in T, as seen in the second line in (17).</p><p>The spread of en to further non-negative affective contexts in the Flemish dialects</p><p>beyond those found already in Middle Dutch can be taken to have prevented en from</p><p>suffering the fate of its West Germanic cognates: it has survived to this day. As we have</p><p>seen, however, en is still used only very infrequently today, in only around 10% of</p><p>negative sentences. Its continued survival, we have argued, is due to the fact that it has</p><p>acquired the ability to realize emphasis on polarity. Emphasis (on polarity) is recog-</p><p>nized to be a feature of the left periphery as well, related to assertion and focus.12 This</p><p>is of course not relevant or available in all negative clauses, hence its low, but stable</p><p>frequency. We can visualize this second change as follows:</p><p>(18) Middle Dutch stage II → Present-day Flemish emphatic ‘stage II’</p><p>[CP C[+aff] [ … niet[+neg] [VP …] [T en[+aff]-Vfin]] →</p><p>[CP C[+aff+emph] [ … niet[+neg] [VP …] [T en[+aff+emph]-Vfin]]</p><p>11.  For simplicity, we represent the clause structure with a head-final TP in (15) and (16). As</p><p>shown by Haegeman (2001), it is possible to give an analysis of Continental West Germanic</p><p>OV orders that conforms with Kayne’s (1994) universal base hypothesis.</p><p>1.  On polarity emphasis/polarity focus in various languages (mainly on emphatic affirma-</p><p>tion, however), cf. e.g. Lipták (2003), Hernanz (2007), Holmberg (2007) and Martins (2007).</p><p> Anne Breitbarth & Liliane Haegeman</p><p>Just as the [+affective] feature on C came to be Alternatively Realized by the particle</p><p>en on the finite verb in T in Middle Dutch, C’s (optional) polarity emphasis feature</p><p>comes to be associated with the same particle on the</p><p>finite verb in the development</p><p>from Middle Dutch to the present-day Flemish dialects as another instance of Alterna-</p><p>tive Realization.</p><p>. Conclusion</p><p>In this chapter, we have argued that the continued availability of the old preverbal</p><p>‘negation’ particle en in the Flemish dialects is not a case of the language having</p><p>stopped in the transition from stage II to stage III of Jespersen’s Cycle. Rather, we have</p><p>provided evidence showing that the present-day situation is the result of further mor-</p><p>phosyntactic change. We have argued that after an initial reanalysis of the preverbal</p><p>marker as a polarity marker, which affects other West Germanic languages as well and</p><p>leads to an asymmetric analysis of stage II in these languages, en in the Flemish dia-</p><p>lects has undergone a further change by acquiring the ability to realize an additional</p><p>feature, emphasis on polarity. That is, it is now used to emphatically contradict states</p><p>of affairs, explicitly or implicitly given in the context.</p><p>We have argued that it is (a) the extension of the use of en to further non-negative</p><p>affective contexts, and (b) en’s acquisition of the ability to express emphasis on polar-</p><p>ity that saves it from suffering the fate of its West Germanic cognates and allows it</p><p>to survive in its functional niche until today. This development is very similar to the</p><p>one seen in (dialectal) French. It has long been acknowledged that old preverbal ne</p><p>in French has ceased to be a negation marker proper but, as it were, ‘switches off ’ the</p><p>affirmative concept (Tesnière 1959: 224–225), sentential negation being expressed by a</p><p>forclusif (pas or n-words), and thus ne acts as a polarity marker. It can also used in vari-</p><p>ous non-negative contexts in Standard French. However, ne is rapidly being lost from</p><p>the spoken language in most dialects.13 As Fonseca-Greber (2007) shows, however,</p><p>its use has stabilized at a low percentage in spoken Swiss French, because it has been</p><p>reanalysed as emphatic in this variety.</p><p>Our conclusion is that, in spite of the availability of en, Flemish dialects have actu-</p><p>ally reached stage III of Jespersen’s cycle and that what appears to be continuity is really</p><p>a case of change.</p><p>1.  Cf. the references cited and discussed in Fonseca-Greber (2007).</p><p>Continuity is change </p><p>References</p><p>Abraham, Werner. 2003. Autonomous and non-autonomous components of ‘grammatic(al)</p><p>ization’: Economy criteria in the emergence of German negation. Sprachtypologie und</p><p>Universalienforschung 56: 325–365.</p><p>Barbiers, Sjef, van der Auwera, Johan, Bennis, Hans, Boef, Eefje, De Vogelaer, Gunther & van der</p><p>Ham, Margreet. 2008. Syntactic Atlas of the Dutch Dialects, vol. II. Amsterdam: University</p><p>of Amsterdam Press.</p><p>Beheydt, Griet. 1998. Het gebruik en de vorm van de negatie in het zuidelijke Nederlands in</p><p>een diachronisch perspectief (15e–20e eeuw). Licentiate dissertation, Catholic University</p><p>of Leuven.</p><p>Breitbarth, Anne. 2008. The development of negation in Middle Low German. Paper presented</p><p>at the Annual Meeting of the LAGB, University of Essex, September 2008.</p><p>Breitbarth, Anne. 2009. A hybrid approach to Jespersen’s cycle in West Germanic. Journal of</p><p>Comparative Germanic Linguistics 12: 81–114.</p><p>Burridge, Kate. 1993. Syntactic Change in Germanic: Aspects of Language Change in Germanic</p><p>with Particular Reference to Middle Dutch [Current Issues in Linguistic Theory 89]. Amster-</p><p>dam: John Benjamins.</p><p>Dahl, Östen. 1979. Typology of sentence negation. Linguistics 17: 79–106.</p><p>Detges, Ulrich & Waltereit, Richard. 2002. Grammaticalization vs. reanalysis: A semantic-</p><p>pragmatic account of functional change in grammar. Zeitschrift für Sprachwissenschaft</p><p>21(2): 151–195.</p><p>Emonds, Joseph E. 1987. The invisible category principle. Linguistic Inquiry 18: 613–632.</p><p>Emonds, Joseph E. 2000. Lexicon and Grammar: The English Syntacticon. Berlin: Mouton de</p><p>Guyter.</p><p>Fonseca-Greber, Bonnibeth Beale. 2007. The emergence of emphatic ‘ne’ in conversational Swiss</p><p>French. French Language Studies 17: 249–275.</p><p>Giannakidou, Anastasia. 2006. N-words and negative concord. In The Blackwell Companion to</p><p>Syntax, Martin Everaert, Henk van Riemsdijk, Rob Goedemans & Bart Hollebrandse (eds),</p><p>iii. 327–391. Oxford: Blackwell.</p><p>Gysseling, Maurits 1977. Corpus van Middelnederlandse teksten (tot en met het jaar 1300). m.m.v.</p><p>en van woordindices voorzien door Willy Pijnenburg. The Hague: Nijhoff.</p><p>Haegeman, Liliane. 1995. The Syntax of Negation. Cambridge: CUP.</p><p>Haegeman, Liliane. 2000. Remnant movement and OV order. In The Derivation of OV and VO</p><p>[Linguistik Aktuell/Linguistics Today 31], Peter Svenonius (ed.), 69–96. Amsterdam: John</p><p>Benjamins.</p><p>Haegeman, Liliane. 2001 Antisymmetry and verb-final order in West Flemish. Journal of Com-</p><p>parative Germanic Linguistics 3: 207–232.</p><p>Haegeman, Liliane. 2002. West Flemish negation and the derivation of SOV order in West</p><p>Germanic. Nordic Journal of Linguistics 25 (2): 154–189.</p><p>Haegeman, Liliane & Zannuttini, Raffaella. 1991. Negative heads and the NEG-criterion. Lin-</p><p>guistic Review 8: 233–251.</p><p>Haegeman, Liliane & Zannuttini, Raffaella. 1996. Negative concord in West Flemish. In Param-</p><p>eters and Functional Heads, Adriana Belletti & Luigi Rizzi (eds), 117–179. Oxford: OUP.</p><p>Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cam-</p><p>bridge: CUP.</p><p> Anne Breitbarth & Liliane Haegeman</p><p>Hernanz, M. Lluïsa 2007. Emphatic polarity and C in Spanish. In Studies in Spanish Syntax,</p><p>Laura Brugè (ed.), 104–150. Venice: Libreria Editrice Cafoscarina.</p><p>Holmberg, Anders. 2007. Null subjects and polarity focus. Studia Linguistica 61(3): 212–236.</p><p>Jäger, Agnes. 2008. History of German Negation. Amsterdam: John Benjamins.</p><p>Jespersen, Otto. 1917. Negation in English and Other Languages [Historisk-filologiske Med-</p><p>delelser I,5]. Copenhagen: A.F. Høst.</p><p>Kayne, Richard S. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press.</p><p>Kemenade, Ans van. 2000. Jespersen’s Cycle revisited. In Diachronic Syntax, Susan Pintzuk,</p><p>George Tsoulas & Anthony Warner (eds), 51–74. Oxford: OUP.</p><p>Klima, Edward S. 1964. Negation in English. In The Structure of Language: Readings in the Phi-</p><p>losophy of Language, Jerry A. Fodor & Jerrold J. Katz (eds), 246–323. Englewood Cliffs NJ:</p><p>Prentice-Hall.</p><p>Ladusaw, William A. 1993. Negation, indefinites, and the Jespersen Cycle. Berkeley Linguistics</p><p>Society 19: 437–446.</p><p>Laka, Itziar. 1990. Negation in Syntax: On the Nature of Functional Categories and Projections.</p><p>Ph.D. dissertation, Massachusetts Institute of Technology.</p><p>Leemans, Emilienne. 1966. Syntactische Kenmerken van het Gents Dialect. Licentiate disserta-</p><p>tion, University of Ghent.</p><p>Lindstad, Arne M. 2007. Analyses of Negation. Structure and Interpretation. Ph.D. dissertation,</p><p>University of Oslo.</p><p>Lipták, Anikó. 2003. The expression of sentential emphasis in Hungarian. Ms, Leiden University.</p><p>Martins, Ana M. 2007. Double realization of verbal copies in European Portuguese emphatic</p><p>affirmation. In The Copy Theory of Movement on the PF Side [Linguistik Aktuell/Linguistics</p><p>Today 107] Norbert Corver & Jairo Nunes (eds), 77–118. Amsterdam: John Benjamins.</p><p>Moscati, Vincenzo. 2006. The Scope of Negation. Ph.D. dissertation, University of Siena.</p><p>Overdiep, Gerrit Siebe. 1937. Stilistische Grammatica van het Moderne Nederlandsch. Zwolle:</p><p>Tjeenk Willink.</p><p>Pauw, G. De. 1973. Syntaktische kenmerken van het dialect van Buggenhout-Opstal. Licentiate</p><p>dissertation, University of Ghent.</p><p>Posner, Rebecca. 1985. Post-verbal negation in non-standard French: A historical and compara-</p><p>tive view. Romance Philology 7: 44–82.</p><p>Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammatical-</p><p>ization. Cambridge: CUP.</p><p>Rowlett, Paul. 1998. Sentential Negation in French. Oxford: OUP.</p><p>Ryckeboer, Hugo. 1986. Het hulpwerkwoord doen in replieken. In Vruchten van z’n akker.</p><p>Opstellen van (oud-)medewerkers en oud-studenten voor Prof. V.F. Vanacker hem aange-</p><p>boden bij zijn afscheid van de Rijksuniversiteit</p><p>Gent, Magda Devos & Johan Taeldeman</p><p>(eds), 321–337. Ghent: Seminarie voor Nederlandse Taalkunde en Vlaamse Dialectologie,</p><p>Rijksuniversiteit Gent.</p><p>Schwegler, Armin. 1988. Word-order changes in predicate negation strategies in Romance lan-</p><p>guages. Diachronica 5: 21–58.</p><p>Schwenter, Scott A. 2006. Fine-tuning Jespersen’s Cycle. In Drawing the Boundaries of Meaning:</p><p>Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn [Studies in</p><p>Language Companion Series 80], Betty J. Birner & Gregory Ward (eds), 327–344. Amster-</p><p>dam: John Benjamins.</p><p>Tavernier, C. 1959. Over negatie en expletief en in het Gents dialect. Taal en Tongval 11: 245–252.</p><p>Tesnière, Lucien. 1959. Eléments de syntaxe structurale. Paris: Klincksieck.</p><p>Continuity is change </p><p>Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Mechanisms of</p><p>Syntactic Change, Charles N. Li (ed.), 141–77. Austin TX: University of Texas Press.</p><p>Wallage, Philipp. 2005. Negation in Early English: Parametric Variation and Grammatical Com-</p><p>petition. Ph.D. dissertation, University of York.</p><p>Wallage, Phillip. 2008. Jespersen’s Cycle in Middle English: Parametric variation and grammati-</p><p>cal competition. Lingua 118 (5): 643–674.</p><p>Weijnen, Antonius Angelus. 1956. Zeventiende-eeuwse taal. 2nd edn. Zutphen: Thieme.</p><p>Willis, David. To appear. A minimalist approach to Jespersen’s Cycle in Welsh. In Grammatical</p><p>Change: Origins, Nature, Outcomes, Dianne Jonas, John Whitman and Andrew Garrett (eds).</p><p>Oxford: OUP.</p><p>Wouden, Ton van der. 1995. On the development of marked negation systems. Paper presented</p><p>at the 12th International Conference on Historical Linguistics, University of Manchester,</p><p>August 1995.</p><p>Zeijlstra, Hedde. 2004. Sentential Negation and Negative Concord. Ph.D. dissertation, Univer-</p><p>sity of Amsterdam.</p><p>Using the Matrix Language Frame model</p><p>to measure the extent of word-order</p><p>convergence in Welsh-English bilingual speech</p><p>Peredur Davies & Margaret Deuchar</p><p>Bangor University</p><p>Word-order convergence is argued to be a processual mechanism in which</p><p>bilinguals use morphemes from language A in an order which is more prominent</p><p>in language B than in language A. In this chapter we examine whether there is</p><p>word-order convergence in the speech of Welsh–English bilinguals using</p><p>Myers-Scotton’s (2002) Matrix Language Frame model. We use the term dichotomous</p><p>matrix language to capture the phenomenon found where structural material</p><p>within a clause does not point to either language as being the matrix language (ML),</p><p>and suggest that this reflects convergence in such clauses. Applying this model to</p><p>recorded spoken data from six Welsh–English bilinguals, we find that dichotomous</p><p>ML clauses in the data are very rare, suggesting that word-order convergence in</p><p>Welsh–English bilinguals is not commonly found when measured in this way.</p><p>1.  Introduction</p><p>In this chapter we address the question of whether word-order convergence is found</p><p>in the speech of Welsh–English bilinguals. After outlining our view of convergence,</p><p>we discuss a method of identifying convergence using Myers-Scotton’s (2002) Matrix</p><p>Language Frame (MLF) model. Our method involves identifying clauses in bilingual</p><p>speech which have structure from both participating languages as potentially showing</p><p>convergence in progress. We then apply this method to spoken data from six Welsh–</p><p>English bilinguals to identify the extent of the convergence found therein.</p><p>2. Defining convergence</p><p>We define convergence, following Backus (2004: 179), as a contact-induced proces-</p><p>sual mechanism of language change which “results in greater similarity [than before]</p><p>between two languages that are in contact with each other”. We interpret his definition</p><p>to refer to convergence in individual constructions, rather than in all constructions</p><p>in the language. Bullock & Toribio (2004) point out that convergence has traditionally</p><p> Peredur Davies & Margaret Deuchar</p><p>been seen mainly as an outcome rather than as a process, but we follow them in defin-</p><p>ing convergence as a process. Bullock & Toribio further state that the result of the</p><p>diachronic process of convergence can be observed synchronically in the output of</p><p>bilinguals (2004: 91), an approach also taken by, for instance, Myers-Scotton (2002)</p><p>and Schmitt (2000). Convergence can occur in most parts of a language’s grammar, but</p><p>in this chapter we will focus on word-order convergence.</p><p>Myers-Scotton (2002: 164) writes that a clause in which word-order convergence</p><p>has occurred will be one where the morphemes come from one language but the</p><p>grammar includes structure from both participating languages. Schmitt’s (2000: 15)</p><p>description of covert code-switching (see Section 3) is interpretable as a description</p><p>of convergence at the utterance level. These definitions allow for the results of con-</p><p>vergence to be observable in both monolingual and bilingual utterances. We define</p><p>monolingual clauses as clauses containing morphemes from only one language and</p><p>bilingual clauses as clauses containing morphemes from more than one language.</p><p>Backus (2004) suggests that the processual mechanism which results in a change</p><p>in frequency of a construction is the “use of one of two or more L1 patterns, congruent</p><p>with [an] unmarked L2 pattern” (2004: 180); we take ‘congruent’ here to mean ‘similar’.</p><p>Therefore a bilingual will choose a construction from one language (LA) which has a</p><p>similar counterpart in the other language (LB) in preference to an LA construction</p><p>which does not. This point is made by Thomason (2001: 93) when she suggests that</p><p>constructions which are similar in two languages, but not actually identically distrib-</p><p>uted (e.g. constructions that are marked in one language but unmarked in the other),</p><p>are most prone to convergence. A similar approach leads Bullock & Toribio (2004: 91)</p><p>to view convergence as “the enhancement of inherent structural similarities found</p><p>between two linguistic systems”. They distinguish convergence from the notions of</p><p>interference and transfer on the grounds that each of these latter notions implies uni-</p><p>directionality whereas convergence may be either unidirectional or bidirectional.</p><p>The focus of our investigation into convergence is constructions which have one</p><p>or more morphemes from Welsh but word order from English, where that word order</p><p>is available in Welsh but only in restricted contexts. Such constructions will reflect</p><p>convergence as a process. Diachronic changes which might result from frequent occur-</p><p>rences of this convergence are not the focus of this chapter.</p><p>Having outlined our definition of convergence, in the next section we will briefly</p><p>review some of the more recent studies of this phenomenon, interpreting the research</p><p>in terms of convergence as being a processual mechanism.</p><p>3. Previous work on convergence</p><p>We focus here on previous work supporting the notion of convergence (whether in</p><p>word order or elsewhere) as a process rather than a result, on the assumption that the</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence </p><p>mechanism of convergence is reflected in individual instances which may or may not</p><p>lead to change, depending of the frequency of those instances.1</p><p>Toribio’s (2004) analysis of Spanish–English bilinguals in the USA demonstrates</p><p>that contact with English affects these speakers’ use of null pronouns, a typical feature</p><p>of Spanish. Her study involved two tasks: first, participants were required to narrate a</p><p>story “in Spanish” in monolingual mode based on a series of pictures representing Lit-</p><p>tle Red Riding Hood. Then the participants were required to reproduce orally another</p><p>story, The Beggar Prince, which they had previously read aloud in a mixed-language</p><p>version; this was expected to lead to the production of both Spanish and English in</p><p>bilingual mode. In the first task, where they were in monolingual mode, speakers used</p><p>overt pronouns more often than was expected in monolingual Spanish</p><p>discourse. Fur-</p><p>thermore, in the second task, when in bilingual mode, the use of overt pronouns by</p><p>these speakers was even more frequent than expected. So, in any given context where</p><p>there was a choice whether or not to use an overt pronoun, speakers frequently con-</p><p>verged to English usage in their Spanish.</p><p>In a study with a similar focus, Montrul (2004) compared the speech of Spanish–</p><p>English heritage bilinguals2 living in the USA with the speech of monolingually-raised</p><p>Spanish speakers from different Spanish-speaking countries, examining, among other</p><p>features, their use of null subjects. The heritage speakers in Montrul’s data used overt</p><p>subjects (68.6%) more than null subjects (31.4%), compared to the reverse pattern in</p><p>monolinguals’ use (42.8% overt and 57.2% null). As Spanish allows such subjects to</p><p>be null whereas English does not, Montrul argues that convergence is occurring in the</p><p>Spanish morphosyntax of these speakers and that the dominance of English has led the</p><p>speakers to select overt subjects instead of null subjects.</p><p>Schmitt (2000), analysing data from Russian–English bilingual immigrant chil-</p><p>dren in the USA within the MLF framework, proposes the notion of covert code-</p><p>switching, which is what “takes place when abstract lexical structure from one language</p><p>is influenced by abstract structure from another language, resulting in a composite</p><p>ML [matrix language]” (2000: 19). A composite ML is found in a construction in</p><p>which morphemes from one language have (some) structure from another language</p><p>(2000: 15). Examples Schmitt gives of covert code-switching are bare forms, which are</p><p>inserted LB morphemes (in a clause where the structure is otherwise from LA) appear-</p><p>ing without the appropriate LA morphology (2000: 23). Children in her data produce</p><p>1.  That is, what some authors would call ‘convergence’, but which we call ‘the result of</p><p>convergence’.</p><p>2.  Montrul uses the term ‘heritage speakers’ to describe second- or third-generation bilin-</p><p>guals who come from a home where a non-English language (Spanish here) was spoken, and</p><p>who have some level of bilingualism in both English and the heritage language.</p><p> Peredur Davies & Margaret Deuchar</p><p>such forms, e.g. (1) below (2000: 23), where inserted English morphemes in what is</p><p>otherwise a Russian ML clause lack the expected Russian morphological endings.</p><p>(1) nu yest’ baseball v heaven?</p><p>well is baseball in heaven</p><p>‘Well, is there baseball in heaven?’</p><p>[expected: nu yest’ baseball v heaven-e?</p><p>well is baseball in heaven-prep.masc.sg]</p><p>In (1) the switched noun heaven does not have the expected Russian prepositional case</p><p>marker -e. In inserting the switched item (heaven) from English, the speaker has incor-</p><p>porated with it the grammar of English (which does not have an inflectional affix of</p><p>this kind on nouns). Schmitt views this as convergence, with the ML as a composite of</p><p>Russian structure (e.g. verbal morphology) and English structure (e.g. no noun inflec-</p><p>tion on heaven), occurring because of an increase in the level of activation of the EL so</p><p>that it competes with the ML in the providing the clause’s structure (2000: 24).3</p><p>These studies therefore show convergence as a contact phenomenon that is best</p><p>viewed as a process rather than a result.</p><p>4. Convergence and code-switching</p><p>Backus (2004: 179) suggests that evidence of convergence may be found particularly</p><p>in the speech of bilinguals who code-switch extensively, and Toribio (2004) provides</p><p>evidence that the same speakers may show more evidence of convergence when using</p><p>both languages than just one. Bilingual clauses are therefore more likely loci for con-</p><p>vergence than monolingual clauses in the speech of bilinguals.</p><p>Nevertheless, Myers-Scotton (2002) indicates that convergence can occur with</p><p>or without code-switching. Code-switching showing convergence is labelled “com-</p><p>posite code-switching” (2002: 165), whereas evidence of convergence without code-</p><p>switching is a structure where “all surface morphemes come from one language” and</p><p>“the abstract lexical structure projecting these morphemes no longer comes from one</p><p>language, but includes some abstract structure from another language” (2002: 164), as</p><p>noted above. Whether or not code-switching is involved, clauses showing evidence of</p><p>convergence will have a composite morphosyntactic frame, which we shall label here</p><p>3.  We are grateful to a reviewer for noting that apparently the choice of the preposition v ‘in’</p><p>here is an unusual one; na ‘on’ is more expected. We suggest that this is another example of the</p><p>influence of English on the clause.</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence 1</p><p>a dichotomous matrix language (see below). Since Myers-Scotton’s MLF model applies</p><p>to individual clauses in the speech of individual bilinguals rather than to a language as</p><p>a whole, and as we define convergence as a synchronic process that can occur within</p><p>a single construction, we argue below that the MLF model can be used to identify evi-</p><p>dence of convergence in specific, quantitative terms.</p><p>5. Convergence in Welsh</p><p>We now review studies that have been conducted on convergence in the Welsh language.</p><p>While our data is from Welsh–English bilingual speech, our focus is on the influence of</p><p>the majority language, English, on the minority language, Welsh.4 Deuchar (2005) exam-</p><p>ines whether paradigmatic and syntagmatic congruence between the grammars of Welsh</p><p>and English facilitates code-switching. She concludes that where the grammatical catego-</p><p>ries of the two languages are similar, code-switching is facilitated. The results suggest that</p><p>phenomena such as code-switching are more likely to occur in aspects of the grammar</p><p>where the participating languages are similar than where they differ. Deuchar suggests</p><p>that code-switching is enhanced or restricted depending on the extent of this similarity.</p><p>Deuchar (2006) goes on to examine Welsh–English bilingual utterances which show</p><p>possible convergence. An example is given in (2), taken from Deuchar (2006: 1996).</p><p>(2) fi ‘di bod i ‘r bus lle</p><p>1s prt.past be.nonfin to det bus place5</p><p>‘I have been to the bus place.’</p><p>4.  The UK census for 2001 shows that 20.8% of the population of Wales can speak Welsh;</p><p>this was an increase from the 1991 census, which indicated that 18.7% of the people of Wales</p><p>spoke Welsh. Figures taken from the Welsh Language Board website (http://www.byig-wlb.</p><p>org.uk/English/faq/Pages/Howdothe2001resultscomparewiththe1991results.aspx) or (http://</p><p>tinyurl.com/2001cf1991).</p><p>5.  We use the following abbreviations in glossing examples:</p><p>1S 1st person singular pronoun NEG negative particle or prefix</p><p>1PL 1st person plural pronoun NONFIN non-finite verb</p><p>2S 2nd person singular pronoun PAST past (perfect) tense</p><p>2PL 2nd person plural pronoun PRES present tense</p><p>3SM 3rd person singular masculine pronoun PRT a pre-verbal particle</p><p>3SF 3rd person singular feminine pronoun PRT.PAST perfect particle</p><p>DET determiner VBZ verbalizer suffix</p><p>IMP imperfect tense</p><p>2 Peredur Davies & Margaret Deuchar</p><p>This is a bilingual clause where all but one morpheme is Welsh, but the order found in</p><p>the noun–noun NP bus lle is English. The expected Welsh order here would be lle bus,</p><p>where the modifier bus follows the head noun lle, but instead the modifier is found</p><p>preceding the head. Deuchar notes that the morphemes in this clause appear to have</p><p>a one-to-one morpheme correspondence with English surface word order, where fi ‘di</p><p>bod i is equivalent to I have been to. She therefore suggests that this clause might show</p><p>word-order convergence from Welsh towards English.</p><p>Deuchar & Davies (2009) evaluate two similar models of language change pro-</p><p>posed by Thomas (1982) and Myers-Scotton (1998). Both these models postulate pro-</p><p>cesses of contact-induced change which can ultimately lead to speakers in a community</p><p>abandoning their original language for another. At a stage prior to this, composite</p><p>code-switching occurs,</p><p>which can lead to what Myers-Scotton calls matrix-language</p><p>turnover. Deuchar & Davies use Myers-Scotton’s MLF model to analyse a dataset of</p><p>speech from Welsh–English bilinguals (which is also among the data examined in this</p><p>study) for evidence of English structure in bilingual clauses. The findings show that</p><p>Welsh was the matrix language (ML) in the great majority (93.94%) of finite clauses</p><p>overall, and that all bilingual clauses had Welsh as their ML. Clauses which do not</p><p>demonstrate a clear ML were considered to show possible convergence, but only one</p><p>such clause was identified (examined again as example (34) in the present study; see</p><p>Section 12 below). The results point to the absence of ML turnover from Welsh to</p><p>English, with little evidence of convergence.</p><p>For further information regarding the influence of English on Welsh, see Parry-</p><p>Williams (1923), Fynes-Clinton (1995 [1913]) and Lindsay (1993).</p><p>6. The differences and similarities in Welsh and English word order</p><p>Convergence may occur in parts of the grammar which are similar in both languages</p><p>in contact. Following Thomason’s (2001: 93) suggestion that an essential step in identi-</p><p>fying contact-induced structural change in a language is to identify structural features</p><p>shared by both languages, in this section we describe the similarities found in Welsh</p><p>and English word order, specifically the relative positions of the subject and the finite</p><p>verb in a finite clause and the head and the modifier within a NP.</p><p>English is an SV-order language, with the finite verb occupying post-subject</p><p>position in declarative clauses (The cat caught the mouse); see Quirk et al. (1989: 724).</p><p>VS order is available only in yes-no interrogative constructions (Did the cat catch the</p><p>mouse?), questions involving modal verbs (Can the mouse run?), questions involv-</p><p>ing to be or to have as the main verb (Is that a mouse?), WH-questions (Where is the</p><p>cat?), and various other inversion contexts, in all of which the finite verb is raised</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence 3</p><p>to pre-subject position.6 English is therefore primarily SV order with VS in some</p><p>constructions.7</p><p>Welsh, conversely, uses VS order (the inflected verb occupying the first main posi-</p><p>tion, before the subject) in main clauses, as in (3), or periphrastic constructions using</p><p>an auxiliary and a nonfinite verbal noun, as in (4); see King (1996: 21).</p><p>(3) Daliodd y gath y llygoden.</p><p>Caught the cat the mouse</p><p>V S O</p><p>‘The cat caught the mouse.’</p><p>(4) Wnaeth y gath ddal y llygoden.</p><p>Did the cat catch the mouse</p><p>AuxV S V O</p><p>‘The cat caught the mouse.’</p><p>Question clauses have rising intonation and/or the pre-verbal particle a (usually omit-</p><p>ted). The position of the subject and verb is not changed, so Welsh interrogatives, like</p><p>Welsh declaratives, have VS order (Ddaliodd y gath y llygoden? Wnaeth y gath ddal y</p><p>llygoden?). SV order, however, is found in subject WH-question constructions, as in</p><p>(5), and also where the subject constituent is fronted for emphasis, as in (6).</p><p>(5) Pwy ddaliodd y llygoden?</p><p>Who caught the mouse</p><p>WH V O</p><p>‘Who caught the mouse?’8</p><p>(6) Y gath ddaliodd y llygoden.</p><p>The cat caught the mouse</p><p>S V O</p><p>‘It was the cat which caught the mouse.’9</p><p>Therefore Welsh has primarily VS order, but SV order in certain constructions.</p><p>6.  An exception to this is a construction where the WH-word is the subject of the sentence,</p><p>e.g. Who caught the mouse?, which has SV order.</p><p>.  There are other, rarer constructions in English that are VS order, such as those with fronted</p><p>negatives, e.g. Never has the cat caught anything.</p><p>.  This could also mean ‘Whom did the mouse catch?’.</p><p>.  This could also mean ‘It was the cat that the mouse caught’ (i.e. OVS order), demon-</p><p>strating object-fronting.</p><p>4 Peredur Davies & Margaret Deuchar</p><p>We now turn to English and Welsh head/modifier constructions in NPs. We use</p><p>the following abbreviations: HM denotes the modifier following the head; MH denotes</p><p>the modifier preceding the head. Note that when we use the term ‘head/modifier order’,</p><p>we mean the relative order of head and modifier (without specifying what this is).</p><p>In English head/modifier constructions (e.g. noun/adjective or noun/noun NPs),</p><p>the order is MH in the great majority of cases, e.g. red wine, where red modifies wine.</p><p>There are some English collocations where the order is HM, for instance, court mar-</p><p>tial, queen consort, Attorney General, and sergeant-major, but these are very rare.10</p><p>Welsh head/modifier NPs usually have HM order, for instance, gwin coch ‘red</p><p>wine’, where the adjective coch precedes the head noun gwin, or gyrrwr bws ‘bus driver’,</p><p>where the modifying noun bws follows the head noun gyrrwr. Some adjectives, how-</p><p>ever, precede the head noun they modify, such as hen ‘old’ and prif ‘main, prime’; for</p><p>instance, hen ddyn ‘old man’, prif weinidog ‘prime minister’.11 Furthermore, any Welsh</p><p>modifier can generally precede its head, causing soft mutation of the initial consonant</p><p>of the head, e.g. cryno-ddisg ‘compact disc’, where the adjective cryno precedes the noun</p><p>disg (the first consonant of which is mutated [d] > [ð]), and heddwas ( [ø]. Whilst this is primarily a literary</p><p>style, it also occurs less frequently in colloquial use, and often results in the construc-</p><p>tion being interpreted as a compound; in fact, compounds in Welsh are usually MH</p><p>order constructions where the initial consonant of the head is mutated. Therefore,</p><p>Welsh has more than one option for NP word order, primarily HM but also MH.</p><p>In this section we have demonstrated the overlap between possible word orders</p><p>in English and Welsh. In main clauses, English has SV whilst Welsh has VS, but Eng-</p><p>lish also has VS and Welsh SV in limited contexts. In head/modifier NPs, English has</p><p>MH word order whilst Welsh has both MH and (more frequently) HM. Where there</p><p>are similarities such as these, it is possible that convergence will be found. We now</p><p>proceed to discussing the Matrix Language Frame model, which we suggest may be</p><p>adapted to become a framework for identifying sites of convergence.</p><p>1.  Moreover, some of these are borrowings from Romance languages (e.g. sergeant-major</p><p>between ML and EL. It also follows that the model could</p><p>also be used to identify those clauses where there is no asymmetry between the two</p><p>participating languages. If structure in a clause is supplied by both languages, then the</p><p>ML may not be uniquely sourced from either of the two participating languages. This</p><p>latter case is what Myers-Scotton (2002: 105), referring to bilingual utterances, terms</p><p>“composite code-switching”, where the clause’s morphosyntactic frame is derived</p><p>from more than one language, and which she considers to demonstrate convergence.</p><p>Clauses exhibiting composite code-switching have what we call a dichotomous matrix</p><p>language, as explained above. We expect that clauses with a dichotomous matrix lan-</p><p>guage will exhibit convergence.</p><p>In the following section we outline the manner in which we apply the MLF model to</p><p>our data to identify clauses, whether monolingual or bilingual, that show convergence.</p><p>. Applying the MLF model to the data</p><p>The MLF model specifies that a clause will adhere to two principles, which Myers-</p><p>Scotton (2002: 59) terms the Morpheme Order Principle (MOP) and the System Mor-</p><p>pheme Principle (SMP). The analytical process of applying the MLF model to data,</p><p>therefore, takes the form of verifying whether or not both MOP and SMP point to the</p><p>same language as being the ML for a given clause. If they indicate the same language,</p><p>then the language indicated is identified as the ML of that clause. The principles are</p><p>outlined below.</p><p>The Morpheme Order Principle (MOP) states that the ML supplies the word order</p><p>for a clause, with the exception of well-formed EL islands, which follow the word order</p><p>of the EL. The primary word-order distinction between Welsh and English that we</p><p>have selected is the position of the verb in relation to the subject, and this is the first</p><p>6 Peredur Davies & Margaret Deuchar</p><p>word-order criterion tested on a clause. The secondary word-order distinction we have</p><p>selected is the position of the modifier in relation to the head in an NP. If a head/modi-</p><p>fier NP in a clause was either (1) a bilingual NP or (2) a monolingual NP with mor-</p><p>pheme order that was not expected for morphemes in that language; then the head/</p><p>modifier word order in that NP was used to identify the ML. Therefore, in a clause</p><p>where the subject/verb positions and/or the head/modifier positions are what would</p><p>be expected in Welsh, we identify ‘Welsh’ as the ML for that clause; where the order is</p><p>what would be expected of English, we identify ‘English’ as the ML. Some clauses have</p><p>the same basic word order in Welsh and English, as noted in Section 6, and in such</p><p>instances the MOP alone was not sufficient to indicate the clause’s ML.</p><p>The System Morpheme Principle (SMP) dictates that the ML supplies outside late</p><p>system morphemes; following Myers-Scotton (2006: 244), we take verbal agreement</p><p>markers to be outside late system morphemes.12 Welsh verbal morphology indicates a</p><p>Welsh ML and English verbal morphology indicates an English ML. The SMP was only</p><p>applied here to finite clauses: we considered clauses lacking a finite verb not to contain</p><p>enough information to satisfy the SMP, except, as in (7), where we assume there is a</p><p>null auxiliary.</p><p>(7) ti ‘di siarad ybyty fo</p><p>2s prt.past talk.nonfin about 3sm</p><p>‘You’ve talked about it.’ [FUS27-LIS328]</p><p>[Alternative form:</p><p>wyt ti di siarad ybyty fo</p><p>be.2s.pres 2s prt.past talk.nonfin about 3sm]</p><p>In such cases as this the clause was assumed to be finite, with the null element consist-</p><p>ing of outside late system morphemes from Welsh. However, the relative position of</p><p>this null verb in relation to the subject was not ascertainable, and therefore these types</p><p>of clauses could have the SMP successfully applied to them but not the subject/verb</p><p>order criterion of the MOP.</p><p>We will now demonstrate the application of the MLF to an example clause from</p><p>the data, (8). English morphemes are underlined in examples.</p><p>(8) dw i ‘n love-o ‘r gwlad though</p><p>be.1s.pres 1s prt love-VBZ det countryside though</p><p>‘I love the countryside, though.’ [FUS27-LIS38]</p><p>First the MOP is tested. As the finite verb dw precedes the subject i, it appears that</p><p>Welsh provides the word order. If there was a head/modifier construction available, it</p><p>12.  Outside late system morphemes are described by Myers-Scotton as morphemes that are</p><p>“coindexed with forms outside the head of their maximal projections” (2002: 75–6).</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence </p><p>would be expected to follow the same order (Welsh), unless it was an EL island. Sec-</p><p>ond, the SMP is tested. The inflected verb dw is in the first person singular, and this</p><p>inflection matches the first person singular pronoun i. There is therefore agreement.</p><p>The inflection on dw is Welsh. Therefore the SMP identifies Welsh as ML.</p><p>Finally, the two languages indicated by the SMP and the MOP are compared.</p><p>According to the MLF model, both principles must indicate either Welsh or English</p><p>as supplying structure.13 Returning to Example (8), the SMP and MOP both indicate</p><p>Welsh as ML, so Welsh can be positively identified as the clause’s ML, providing out-</p><p>side late system morphemes, word order and some content morphemes. English is the</p><p>EL, providing some content morphemes (love and though).</p><p>Assuming an ML could be positively identified, therefore, each clause in the data</p><p>is marked as ‘Welsh’ or ‘English’ ML.14 Clauses with sufficient material to test the prin-</p><p>ciples, but which cannot show one language as ML, are discussed in the next section.</p><p>. Cases where the source of the ML is not positively identifiable</p><p>Whilst it is expected that, in most clauses the testing of the MOP and SMP will result</p><p>in the same language being indicated as ML, it is possible that the two principles will</p><p>not agree; for instance, if the verb agreement morphology is from one language but</p><p>the word order is from the other, or if one principle provides conflicting information</p><p>about which language is ML. An example of the latter type of clause is (9), taken from</p><p>Deuchar (2006: 1994).</p><p>(9) Ddaru ni gyfweld […] am ddeg awr</p><p>happen.past 1pl interview.nonfin for ten hour</p><p>assistant i D.S.</p><p>assistant for D.S.</p><p>‘We interviewed […] for a ten-hour assistant for D.S.’</p><p>It is not possible to test the SMP on this clause, as subject–verb agreement does not</p><p>apply: the verb ddaru is not inflected for person or number. When the MOP is tested,</p><p>the position of verb (ddaru) before subject (ni) indicates the word order as being</p><p>Welsh. However, the NP ddeg awr assistant ‘ten-hour assistant’ appears to have English</p><p>word order, as the modifier ddeg awr precedes the head assistant. This order is not</p><p>13.  If the word order construction in a clause is available in either Welsh or English, the</p><p>SMP has the ‘casting vote’, taken to specify which language is the source of the ML. That is, the</p><p>subject-verb agreement positively identifies the ML which the MOP could only ambiguously</p><p>identify.</p><p>14.  Some clauses in the data contained insufficient material to test the principles, such as</p><p>verbless clauses or non-finite clauses. These are not examined in this chapter.</p><p> Peredur Davies & Margaret Deuchar</p><p>usual for Welsh. As the material within the NP is not all English, it cannot be labelled</p><p>an EL island. The English word assistant would therefore be expected to be positioned</p><p>in a locus appropriate to the subject/verb order (Welsh), preceding the modifier, to</p><p>produce assistant deg awr. As we find that English seems to provide the word order</p><p>in this NP, however, the result is that the verb/subject order and head/modifier order</p><p>indicate different languages as being the ML.</p><p>Thus the asymmetry specified by the MLF model is not found in such clauses as</p><p>(9) above. Here, both languages are supplying structural information. As the evidence</p><p>for the ML points in different directions, we shall refer to any clauses where this occurs</p><p>as having a dichotomous matrix</p><p>language.15 Such clauses may show convergence. A</p><p>benefit of adding this nuance to the MLF model, and to methods of identifying conver-</p><p>gence in general, is that it provides a means to identify where a given structure (e.g. an</p><p>NP) has word order which is not usual for the morphemes projected by that structure.</p><p>We therefore posit that the MLF model can be used to identify loci of convergence in</p><p>clauses which yield a dichotomous ML result.</p><p>The reasons for using the MLF model as a means of identifying word-order con-</p><p>vergence are twofold. First, where the MOP and the SMP both clearly indicate one</p><p>language as being the source of the ML, we assume there is no word-order convergence.</p><p>However, clauses where the two principles do not clearly indicate one language as being</p><p>the source of the ML, as in the case of dichotomous ML clauses, seemed to us to be</p><p>likely sites for convergence and therefore worth investigation. Second, the MLF model</p><p>can be applied quantitatively to data on a clause-by-clause basis (see below), and there-</p><p>fore this method can generate quantitative information indicating the extent of word-</p><p>order convergence taking place in a given dataset. We believe that this is one of the first</p><p>implementations of a model that identifies convergence quantitatively in this way.</p><p>It will be recalled that convergence may be found in structures which contain</p><p>morphemes from just one language (cf. the definitions given by Schmitt (2000) and</p><p>Myers-Scotton (2002) cited in Section 3). We suggest that the MLF model may also</p><p>be applied to monolingual clauses, following Jake, Myers-Scotton & Gross (2002: 88),</p><p>who imply that monolingual speech will also have an ML providing the grammatical</p><p>frame. Even if a clause’s morphemes are all from one language, the other language</p><p>may be providing structure. Therefore monolingual constructions from bilinguals may</p><p>show convergence, and the principles of the MLF model can be tested on such clauses.</p><p>15.  Example (9) could also be described as an example of composite code-switching, but</p><p>note that this differs to a dichotomous ML clause in that the latter label can be applied to</p><p>either monolingual or bilingual clauses, whereas we interpret composite code-switching only</p><p>to be applicable to bilingual clauses.</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence </p><p>In this study we have included all clauses from the data, whether monolingual or bilin-</p><p>gual, in our analysis, though this chapter concentrates on results from finite clauses.</p><p>1. The data</p><p>The data analysed in this study were collected as part of an AHRC-funded project</p><p>located at Bangor University, for which 40 hours of recorded spoken data were col-</p><p>lected, transcribed according to the CHAT system (MacWhinney 2000), and sub-</p><p>mitted to Talkbank (http://talkbank.org/data/bilingbank/Bangor). Participants were</p><p>found via social networks and advertising, and were selected on the basis of being</p><p>bilingual in Welsh and English. The recordings were made in informal conditions, in</p><p>a location chosen by the participants, who knew one another well either as friends,</p><p>colleagues or family members. The participants had an informal conversation with no</p><p>researcher nearby. In the majority of cases, conversations were between two people.</p><p>For the purposes of the current study, a subset of six speakers in three recordings</p><p>was chosen for analysis. The transcriptions are coded Davies6, Fusser6 and Fusser27.</p><p>We will describe the six participants briefly next.16</p><p>Davies6 is a conversation between two male friends, Hector (23, a teacher) and</p><p>Daniel (25, a student). Both had lived in the same part of north-west Wales since birth.</p><p>For both, Welsh was the language spoken to them by their parents whilst growing</p><p>up and the medium of their schooling, and both reported learning English at or just</p><p>before primary school age.</p><p>Fusser6 is a conversation between two female work colleagues, Antonia (52, a lec-</p><p>turer) and Amranwen (36, a secretary). Antonia was born in south Wales but moved</p><p>to live permanently in the north-west at 21. Her father spoke Welsh and English to her</p><p>but her mother was a monolingual speaker of English. She was educated through the</p><p>medium of Welsh and English. Amranwen had always lived in north Wales apart from</p><p>a period at an English university. Her parents and school teachers spoke Welsh to her.</p><p>Both women reported having acquired both Welsh and English simultaneously.</p><p>Fusser27 is a conversation between two female student friends, Mabli (19) and</p><p>Lisa (20). Mabli had lived in north-west Wales since birth. Though she had received all</p><p>her schooling through the medium of Welsh, she came from a home where her father</p><p>spoke only English and her mother Welsh and English. Lisa had lived in south-west</p><p>Wales from birth until she moved to north-west Wales to attend university. Her father</p><p>spoke English to her, whilst her mother used Welsh, which was also the language</p><p>16.  Pseudonyms have been provided for the participants in order to hide their identity.</p><p> Peredur Davies & Margaret Deuchar</p><p>through which she received her schooling. Mabli and Lisa both reported that they</p><p>acquired Welsh and English simultaneously.</p><p>11. Results</p><p>The MLF model was applied to all transcribed clauses, monolingual and bilingual,</p><p>totalling 3275 clauses. Of these, 1816 or 55.45% were finite. There was enough evi-</p><p>dence to assign an ML to all finite clauses17 and the results are displayed in Figure 1. Of</p><p>the finite clauses, 1733 (95.43%) had a Welsh ML; English was the ML for 82 (4.52%)</p><p>clauses. Clearly Welsh is the predominant ML in these speakers’ finite clauses.</p><p>2000</p><p>1800</p><p>1600</p><p>1400</p><p>1200</p><p>N</p><p>um</p><p>be</p><p>r o</p><p>f c</p><p>la</p><p>us</p><p>es</p><p>1000</p><p>800</p><p>600</p><p>400</p><p>200</p><p>0</p><p>Welsh English</p><p>Matrix Language</p><p>Dichotomous</p><p>182</p><p>Bilingual</p><p>335</p><p>Mono-</p><p>lingual</p><p>1398</p><p>Figure 1. Distribution of finite clauses according to matrix language</p><p>18.50% of finite clauses (336 out of 1816) were bilingual. Of those bilingual clauses,</p><p>99.70% (335 out of 336) had a Welsh ML. English was never the ML in bilingual finite</p><p>clauses. The one remaining bilingual finite clause (0.30%) had a dichotomous ML (see Sec-</p><p>tion 12 below). Thus Welsh is clearly the most usual ML in bilingual clauses in these data.</p><p>Example (10) is a bilingual clause with Welsh ML, whilst (11) is a monolingual</p><p>clause with English ML. As above, English morphemes are underlined in examples.</p><p>1.  All clauses analysed were labelled as having either a Welsh, English or dichotomous ML.</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence 1</p><p>(10) ches i ‘m chance i ddarllen o</p><p>get-imp.1s 1s neg chance to read.nonfin 3sm</p><p>‘I didn’t get a chance to read it.’ [DAV6-DAN40]</p><p>(11) oh there’s nothing wrong with you! [FUS6-ANT283]</p><p>Example (10) has Welsh as ML since the finite verb and subject (ches and i) have Welsh</p><p>outside late system morphemes, and the subject/verb order is VS, also Welsh. Example</p><p>(11) has English as ML since is has English outside late system morphemes and the</p><p>subject/verb order is SV, also English.</p><p>100%</p><p>90%</p><p>80%</p><p>70%</p><p>60%</p><p>50%</p><p>40%</p><p>30%</p><p>20%</p><p>10%</p><p>0%</p><p>MAB LIS AMR</p><p>Speaker</p><p>Pr</p><p>op</p><p>or</p><p>tio</p><p>n</p><p>of</p><p>�</p><p>ni</p><p>te</p><p>c</p><p>la</p><p>us</p><p>es</p><p>ANT DAN HEC</p><p>Dichotomous English Welsh</p><p>Figure 2. Matrix language distribution of clauses according to speaker.</p><p>There is little variation in the data between speakers, as shown in Figure 2. Hector</p><p>produced the highest proportion of clauses with Welsh ML (291 out of 297 or 97.98%),</p><p>whilst Daniel produced the lowest (232 out of 262 or 88.55%). The younger speakers</p><p>(Mabli, Lisa, Daniel and Hector) used Welsh ML on average in 94.99% of their finite</p><p>clauses; the older speakers (Amranwen and Antonia) did so in 96.82% of their finite</p><p>clauses on average, therefore only slightly more frequently. Regarding where speak-</p><p>ers grew up, northern speakers (Daniel, Hector and Amranwen) used Welsh ML in</p><p>95.66% of their finite clauses on average, similar to the 95.07% for southern speak-</p><p>ers (Mabli, Lisa and</p><p>or at least delay of a</p><p>change. One of the questions behind the present collection was therefore which factors</p><p>4 Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>might prevent or impede a change that might be expected on the basis of otherwise</p><p>similar input conditions. Biberauer, Newton and Sheehan in their contribution, for</p><p>instance, argue for a universal constraint on certain types of structures, which implies</p><p>restrictions on possible changes. Sundquist and Enrique-Arias in turn explore the pos-</p><p>sibility of language contact as a factor in delaying syntactic change.</p><p>In keeping with the two major themes of the volume, the contributions are divided</p><p>into two broad sections, the first addressing continuity, and the second addressing</p><p>change.</p><p>2. Continuity</p><p>The contributions to this part of the volume address problems of (apparent) syntactic</p><p>continuity.</p><p>In What changed where? A plea for the re-evaluation of dialectal evidence, Katrin</p><p>Axel and Helmut Weiß make a case for studying linguistic change in dialects rather</p><p>than standard languages. They present a number of case studies, focussing mainly</p><p>on the development of pro-drop in the history of German (dialects). They show that</p><p>the alleged loss of pro-drop during the Old High German period has only affected</p><p>the standard language; many dialects show (partial) pro-drop to this day, licensed by</p><p>c-commanding Agr in C. The only change is the innovation of the present-day restric-</p><p>tion to pronominal Agr. Similar continuity at the dialectal level, Axel & Weiß argue, is</p><p>found in the domain of negation (negative concord is preserved in several dialects)</p><p>or (with qualifications) in the order of verbs within the verb cluster. Continuity at the</p><p>level of the standard language, on the other hand, is found in possessive constructions,</p><p>where the standard language preserves post-nominal genitive possessors.</p><p>In Impossible changes and impossible borrowings: the Final-over-Final Constraint,</p><p>Theresa Biberauer, Michelle Sheehan and Glenda Newton address the implications</p><p>for syntactic change of a principle argued to restrict word–order possibilities across</p><p>languages, the Final-Over-Final-Constraint (FOFC) (Biberauer, Holmberg & Roberts</p><p>2007). As the FOFC predicts certain word orders to be impossible, ruling out in partic-</p><p>ular head-final higher functional projections when lower projections are head-initial,</p><p>change and borrowing are predicted to be constrained. The authors demonstrate in</p><p>a number of case studies that changes from head-final to head-initial do respect the</p><p>FOFC in that higher heads change first and that the presence of initial heads blocks</p><p>the borrowing of final heads.</p><p>Anne Breitbarth and Liliane Haegeman’s contribution, Continuity is change: The</p><p>long tail of Jespersen’s cycle in Flemish, discusses an apparent case of syntactic continu-</p><p>ity, the (at least optional) maintenance of the old preverbal negation marker e(n) in</p><p>the Flemish dialects of Dutch. They argue that its survival is not due to these dialects</p><p>Introduction 5</p><p>going through Jespersen’s Cycle at a much slower pace than the other West Germanic</p><p>languages, which have long lost the cognates of e(n), but rather to its reanalysis as a</p><p>marker of (affective) polarity which has left the surface word order unchanged. Evi-</p><p>dence for this comes from the pragmatic restrictions on the use of e(n) in the Flemish</p><p>dialects and its availability in non-negative affective contexts.</p><p>In their chapter, Using the Matrix Language Frame model to measure the extent</p><p>of word-order convergence in Welsh–English bilingual speech, Peredur Davies and</p><p>Margaret Deuchar aim to test the hypothesis that ongoing ‘convergence’ (contact-</p><p>induced transfer) can be identified in bilingual code-switching speech, in this case</p><p>the speech of Welsh–English bilinguals. In a corpus of Welsh with extensive code-</p><p>switching with English, they look for instances where the matrix language is Welsh,</p><p>but the grammatical system used is unambiguously English. Any such instances are</p><p>viewed as evidence of ongoing convergence of Welsh towards English. They investi-</p><p>gate this with respect to word order within the nominal domain. Using their meth-</p><p>odology, they identify only one possible instance of convergence (modifier–head</p><p>order within noun phrases in this case), and conclude that Welsh noun phrases man-</p><p>ifest continuity rather than change in the face of extensive bilingualism and contact</p><p>with English.</p><p>Andrés Enrique-Arias’ contribution, Contact-induced conservativism in the</p><p>Spanish of Catalan bilinguals in Majorca, addresses the topic of grammatical conti-</p><p>nuity from the perspective of contact between Spanish and Catalan in Majorca. The</p><p>principal claim is that, while contact is predominantly seen as a force for grammati-</p><p>cal change, under the right circumstances it can, in fact, serve to prevent otherwise</p><p>expected innovations in a given language. Enrique-Arias argues that the presence</p><p>of parallels in Catalan for a number of structures which are obsolescent in gen-</p><p>eral Spanish results in their retention in the Spanish spoken and written by Catalan</p><p>native speakers in Majorca.</p><p>In his chapter Comparative continuity, Remus Gergel investigates comparative</p><p>clauses in English and French with primary focus on subject–verb inversion. He</p><p>explores the derivation and diachronic evolution of these clauses and argues that</p><p>they instantiate syntactic continuity in that they feature a ‘low subject’, that is, one</p><p>that is not raised out of the verb phrase – an option that used to be much more wide-</p><p>spread in both languages. On this analysis surface subject-verb inversion is achieved</p><p>merely by the expected raising of the verb to the inflectional domain and not on</p><p>to the complementizer domain, as in the standard generative analysis of compara-</p><p>tive inversion.</p><p>John Sundquist looks at Variation, continuity, and contact in Middle Norwegian</p><p>and Middle Low German and concludes that the patterns of OV/VO variation in Mid-</p><p>dle Norwegian were not influenced by contact with Middle Low German. He attributes</p><p>6 Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>this to the particular sociolinguistic character of this contact – restricted to business</p><p>contacts in the context of Hanseatic trade.</p><p>3. Change</p><p>Edith Aldridge’s chapter Directionality in word-order change in Austronesian languages</p><p>reconstructs how SVO order in Austronesian languages has evolved from the older VOS</p><p>order via topicalization of the subject. Absolutive arguments in Austronesian VOS lan-</p><p>guages are argued to end up in final position by moving the absolutive to a topic position</p><p>and then fronting the remnant TP around it. Diachronically, SVO develops when only the</p><p>first step of deriving VOS in that fashion is taken and the fronted topic is reanalysed as</p><p>a subject in SpecTP. By comparing several Austronesian SVO languages, Aldridge shows</p><p>that this is a gradual process as subjects across these languages show combinations of A-</p><p>and A′-properties.</p><p>In his contribution, Negative co-ordination in the history of English, Richard Ingham</p><p>looks at the syntax of nor in the history of English, arguing that changes can be derived</p><p>from more general changes in the syntax of negation. While the pattern affirmative</p><p>clause + nor + negative clause is widely attested in Middle English, English today allows</p><p>nor only where both conjoined clauses are negative. Conversely, the clause introduced</p><p>by nor once required additional negative elements (negative concord), while, today, nor</p><p>may act as the sole negator. Ingham accounts for this by proposing two changes in the</p><p>grammar of English: the replacement of a null negative operator (NegOP) in SpecNegP</p><p>by not and subsequently the loss of NegP in Early Modern English as n-words such</p><p>as no become inherently negative. Initially nor has an uninterpretable [uNeg] feature</p><p>checked by NegOP (giving rise to negative concord); with the loss of NegOP, it checks</p><p>its [uNeg] feature from a c-commanding</p><p>Antonia). Females used Welsh ML slightly more frequently than</p><p>males (96.26% to 93.56%). These differences, however, are very small.</p><p>To summarize, the Welsh–English bilinguals that were recorded almost</p><p>always produce clause structure reflecting Welsh as ML and therefore do not display</p><p>convergence.</p><p>2 Peredur Davies & Margaret Deuchar</p><p>12. Clauses with a dichotomous Matrix Language</p><p>Dichotomous ML clauses were defined in Section 9 as clauses where the evidence for</p><p>the ML points in different directions. There was only one dichotomous ML clause in</p><p>the data. It is shown in (12) below.</p><p>(12) roedd drws-nesa # pobl yn wneud #</p><p>be.3s.imp next-door people prt make.nonfin</p><p>sloe gin</p><p>sloe gin</p><p>‘[The] next-door people made sloe gin.’ [FUS6-AMR371]</p><p>The SMP identifies Welsh as ML (roedd has Welsh outside late system morphemes).</p><p>The verb/subject order of this clause is Welsh VS (roedd precedes the subject NP</p><p>drws nesa pobl). However, the word order in the NP drws nesa pobl, where the head</p><p>noun pobl ‘people’ follows the modifying lexical unit drws nesa ‘next door’, is English-</p><p>predominant MH, rather than the expected Welsh order, HM (i.e. pobl drws nesa). The</p><p>material in the rest of the clause points to Welsh as providing structure, but the infor-</p><p>mation in the NP drws nesa pobl points to English as providing structure. Since there is</p><p>disagreement between these two indicators of ML, the clause has a dichotomous ML.</p><p>We view (12) as showing signs of word-order convergence in the NP drws nesa pobl,</p><p>where the structure found is more prominent in one language, English, but where the</p><p>morphemes are from the other language, Welsh.</p><p>Other interpretations of (12) are possible, however. First, the speaker inserts a</p><p>brief pause (</p><p>of</p><p>morphosyntactic convergence. Bilingualism: Language and Cognition 7: 125–142.</p><p>Myers-Scotton, Carol. 1993. Duelling Languages. Oxford: Clarendon Press.</p><p>Myers-Scotton, Carol. 1998. A way to dusty death: the Matrix Language turnover hypothesis.</p><p>In Endangered Languages: Language Loss and Community Response, Lenore A. Grenoble &</p><p>Lindsay J. Whaley (eds.), 289–316. Cambridge: CUP.</p><p>Myers-Scotton, Carol. 2002. Contact Linguistics: Bilingual Encounters and Grammatical Out-</p><p>comes. Oxford: OUP.</p><p>Myers-Scotton, Carol. 2006. Multiple Voices: An Introduction to Bilingualism. Malden MA:</p><p>Blackwell.</p><p>Parry-Williams, Thomas H. 1923. The English Element in Welsh: A Study of English Loan-words</p><p>in Welsh. London: Honourable Society of Cymmrodorion.</p><p>Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1989. A Comprehensive</p><p>Grammar of the English Language. London: Longman.</p><p>Schmitt, Elena. 2000. Overt and covert codeswitching in immigrant children from Russia. Inter-</p><p>national Journal of Bilingualism 4: 9–28.</p><p>Using the Matrix Language Frame model to measure the extent of word-order convergence 5</p><p>Thomas, Alan. 1982. Change and decay in language. In Linguistic Controversies: Essays in Lin-</p><p>guistic Theory and Practice in Honour of F. R. Palmer, David Crystal (ed.), 209–219. London:</p><p>Edward Arnold.</p><p>Thomason, Sarah G. 2001. Language Contact: An Introduction. Edinburgh: EUP.</p><p>Toribio, Almeida J. 2004. Convergence as an optimization strategy in bilingual speech: Evidence</p><p>from code-switching. Bilingualism: Language and Cognition 7: 165–173.</p><p>On language contact as an inhibitor</p><p>of language change</p><p>The Spanish of Catalan bilinguals in Majorca*</p><p>Andrés Enrique-Arias</p><p>Universitat de les Illes Balears</p><p>This study explores the role of language contact as an inhibitor of language</p><p>change through the historical analysis of several features that are characteristic</p><p>of the Spanish in contact with Catalan in the island of Majorca. The analysis</p><p>reveals that (i) these features are attested in Majorcan Spanish at least since the</p><p>1700s and (ii) they had some sort of existence in monolingual varieties of Spanish</p><p>at the time when the language was introduced to the island. The historical data</p><p>suggest that these features of Majorcan Spanish are not, as traditionally believed,</p><p>contact-driven innovations, but the result of the preservation of structures that</p><p>are recessive in monolingual Spanish but in Majorca have been reinforced by the</p><p>existence of a parallel Catalan structure.</p><p>1. Introduction</p><p>A widespread assumption in the linguistic literature is that language change is an</p><p>expected, or even unavoidable, result of language contact.1 In fact, numerous studies</p><p>of language variation and change in contact languages claim that contact accelerates</p><p>change. In contrast, very little attention has been paid to the opposite scenario, that is,</p><p>the possibility that the presence of two languages in the same speech community may</p><p>constitute a factor promoting the retention of variants that are losing ground in non-</p><p>contact varieties. In particular, a preliminary historical analysis of the Spanish spoken</p><p>in Majorca, an island off the eastern coast of Spain where Spanish has been spoken</p><p>*I wish to thank Miriam Bouzouita, David Heap, Carmen Silva-Corvalán, and two anony-</p><p>mous reviewers for their comments on this paper. Any errors that remain are of course my</p><p>responsibility.</p><p>1. As Thomason (2001: 10) puts it “The most common result of language contact is change</p><p>in some or all of the languages: typically, though not always, at least one of the languages will</p><p>exert at least some influence on at least one of the other languages”.</p><p> Andrés Enrique-Arias</p><p>alongside Catalan for centuries, has shown a remarkable historical continuity for some</p><p>morphosyntactic and lexical features that are receding in other Spanish-speaking areas</p><p>(Enrique-Arias 2006, 2008). In the same vein, research carried out using the variation-</p><p>ist sociolinguistic methodology (Blas Arroyo 2007) has called attention to the consid-</p><p>erable vitality in Castellón, another Spanish–Catalan bilingual area, of traits that are</p><p>receding in non contact areas.</p><p>The main objective of this study is to explore further the role of language con-</p><p>tact as an inhibitor of language change through an analysis of a range of phonetic,</p><p>morphosyntactic and semantic phenomena in the varieties of Spanish in contact with</p><p>Catalan, mainly in Majorca but also in other Catalan speaking territories. The specific</p><p>aims are:</p><p>a. to evaluate the possible influence of language contact on the retention of variants</p><p>that are subject to change in other dialect areas;</p><p>b. to determine what linguistic and extralinguistic factors favor the retention of lin-</p><p>guistic variables in language contact situations;</p><p>c. to contribute new data from a historical corpus of Spanish in contact with Catalan</p><p>in Majorca to shed new light on these issues.</p><p>In order to achieve these objectives, in the present study I analyse the historical devel-</p><p>opment of a number of structures that are characteristic of the variety of Spanish spo-</p><p>ken in Majorca. For each phenomenon I look at the current description according to</p><p>the available linguistic monographs, as well as its distribution in historical corpora that</p><p>represent, on the one hand, Spanish produced by Majorcan bilinguals in the late 1700s</p><p>and early 1800s and, on the other, comparable data from general Spanish. The historical</p><p>analysis reveals that many of the features that characterize the variety of Spanish spoken</p><p>in Majorca today were already present in the Spanish texts produced by Majorcans in the</p><p>1700s. The phenomena in question had in turn some sort of existence in monolingual</p><p>varieties of Spanish at the time when the language was introduced to the island. This</p><p>distribution of forms suggests that, for a number of linguistic structures, Catalan contact</p><p>has resulted not so much in the introduction of innovative structures into Majorcan</p><p>Spanish, but in the preservation of language usages that are recessive or have disappeared</p><p>in general Spanish, reinforced by the existence of a parallel structure in Catalan.</p><p>2. Language contact as a promoter of linguistic change</p><p>The effect of language contact as a promoter of language change is closely related</p><p>to the fact that bilingualism sets the stage for a variety of linguistic phenomena that</p><p>On language contact as an inhibitor of language change </p><p>result in change of one kind or another: code-switching, borrowing, simplification</p><p>of grammatical and lexical categories, development of periphrastic constructions or</p><p>overgeneralization of forms following a regularizing pattern, among other kinds of</p><p>structural interference.</p><p>Contact-induced change appears to be particularly intense in situations of social</p><p>bilingualism which are unbalanced in favour of one language. In these situations</p><p>members of the subordinate group are more likely to become bilingual and adopt fea-</p><p>tures from the socially dominant group’s language (Thomason 2001: 66). This scenario</p><p>makes extensive interference possible, often resulting in the development of innova-</p><p>tive language uses and accelerating changes that are less developed in monolingual</p><p>varieties. By acceleration it is understood that the innovative variant occurs in more</p><p>linguistic environments and at a higher frequency than it does in non-contact varieties</p><p>of the same language.</p><p>In the Spanish-speaking world, acceleration of language change is well illustrated</p><p>by a number of phenomena from Spanish in contact with English in the United States.</p><p>In the three examples presented below, Spanish–English bilinguals living in the South-</p><p>west United States exhibit a more advanced stage of linguistic change than their Spanish</p><p>monolingual counterparts in other areas of the Spanish-speaking world.</p><p>a. Advance of the periphrastic future at the expense of morphological future</p><p>Spanish exhibits variable use of two future forms: a periphrastic future (voy a</p><p>cantar</p><p>‘I am going to sing’) and a morphological future (cantaré ‘I will sing’). The</p><p>advance of the former at the expense of the latter is a historical development doc-</p><p>umented in several studies (see, for instance, Aaron 2007). In regards to Spanish</p><p>in contact with English in the US, Gutiérrez (1995, 2002) has shown that this</p><p>change has been accelerated among the communities of Mexican origin of the</p><p>United States when compared to monolingual speakers from Mexico.</p><p>b. Increase in the use of preverbal clitics in verbal periphrases</p><p>In a number of Spanish verbal periphrases composed of a finite semi-auxiliary</p><p>plus an infinitive or a present participle, object clitics referring to an argument of</p><p>the non-finite form may variably occur before the finite verb as shown in (1):</p><p>(1) Vino a verme/me vino a ver.</p><p>he.came to see.me/me he.came to see</p><p>‘He came to see me.’</p><p>The trend of placing clitics in preverbal position in periphrases, to a large extent</p><p>as a consequence of a process of grammaticalization of a number of matrix verbs,</p><p>reflects an ongoing change that is more advanced in language contact than it is in</p><p>monolingual varieties, as demonstrated in Silva-Corvalán (1994: 130).</p><p>1 Andrés Enrique-Arias</p><p>c. The extension of estar at the expense of ser in the context of predicate adjectives</p><p>Spanish has two copular verbs, ser and estar, both translated as ‘to be’. The exten-</p><p>sion of the copula estar to contexts traditionally limited to ser is another phenomenon</p><p>that appears to be more advanced amongst bilingual speakers (Silva-Corvalán 1994:</p><p>119). Innovative uses such as mi marido está alto ‘my husband is tall’, mi casa está</p><p>chiquita ‘my house is small’ (as opposed to traditional use of ser in these structures: mi</p><p>marido es alto, mi casa es chiquita) among Los Angeles bilingual speakers (including</p><p>Mexico-born speakers living in the U.S.) are higher, an overall 34% compared to 16%</p><p>in a group of monolingual speakers in Morelia, Mexico (Gutiérrez 1992).</p><p>A number of reasons have been adduced to explain the acceleration of change affect-</p><p>ing subordinate varieties in contact situations. In regards to Spanish in Los Angeles,</p><p>Silva-Corvalán (1994: 7) points out that the contact-induced changes that she studies</p><p>are further “favored and accelerated” by three factors: low normative pressure in the</p><p>subordinate language, restriction in the range of communicative uses of the subordi-</p><p>nate language, and speakers’ positive attitudes toward the superordinate language. This</p><p>is indeed the case in the US, where Spanish does not enjoy official status, most speak-</p><p>ers have incomplete schooling, if any, in Spanish, and there is no academic and literary</p><p>class serving as a normative model. As a result, speakers are to a great extent isolated</p><p>from the Spanish-speaking norm (Enrique-Arias 2010).</p><p>In summary, acceleration of language change in contact situations would work</p><p>in the following manner: bilingualism promotes innovative linguistic uses, and the</p><p>sociolinguistic situations that typically characterize subordinate varieties, such as lack</p><p>of normative pressure, or the restriction in contexts of use, contribute to intensifying</p><p>and spreading these innovations even further.</p><p>3. Language contact as an inhibitor of linguistic change</p><p>In contrast, very little attention has been paid to the opposite scenario, that is, the</p><p>possibility that the presence of two languages in the same speech community may</p><p>constitute a factor promoting the retention of variants that are receding in non-contact</p><p>varieties. Processes of language change are often characterized by the emergence of</p><p>a new form or construction that becomes interchangeable in at least one linguistic</p><p>environment with a pre-existing one. Once this new form increases its frequency to</p><p>the point of replacing entirely the older one, the change is complete. However, in a</p><p>language-contact situation use of the traditional variant may be reinforced by (a) the</p><p>existence of a parallel structure in the contact language and/or (b) the absence of a</p><p>structural equivalent for the innovative variant in the contact language. In such cases</p><p>change could be delayed, and thus the spread of the innovative variant would progress</p><p>more slowly than in non-contact varieties of the same language. Moreover, when the</p><p>language in question is the socially dominant variety, that is, it is typically acquired and</p><p>On language contact as an inhibitor of language change 11</p><p>used in formal contexts, the sociolinguistic situation may further encourage the pres-</p><p>ervation of the traditional variant. This is especially true when the innovative form is</p><p>characteristic of informal registers. As such, speakers that acquire and use the socially</p><p>dominant language in formal settings have limited exposure to the innovative variant</p><p>and are thus less likely to use it in their speech.</p><p>It will be argued in the following pages that this is in fact the situation that can be</p><p>found in the variety of Spanish spoken in the island of Majorca where there is intensive</p><p>societal Spanish–Catalan bilingualism.2</p><p>The evolution of a number of phonetic phenomena illustrates how language con-</p><p>tact has been a factor in the retention of traditional variants in the Spanish spoken in</p><p>Majorca.</p><p>1. The phenomenon known as yeísmo (the loss of the phoneme /ˆ/ and its merger into</p><p>the phoneme /N/) is a trend that originated in Western Andalusia, possibly as early</p><p>as the sixteenth century and has spread considerably ever since (Guitarte 1983).</p><p>In European Spanish yeísmo “has become part of the dominant urban speech pat-</p><p>tern throughout the Peninsula” (Penny 2000: 121). Although some Catalan speak-</p><p>ers are showing loss of the palatal lateral approximant, the distinction of the two</p><p>phonemes is far more robust in Catalan than it is in standard urban varieties of</p><p>Spanish (Julià i Muné 2002: 79); as a result bilingual speakers tend to retain it more.</p><p>In Romera’s study (2003: 371–72) in Palma, Majorca’s main city and administra-</p><p>tive capital, speakers with Catalan as L1 exhibited 52.3% of conservation of the</p><p>distinction compared to 12.9% in the speech of those classified as Spanish L1.</p><p>Romera’s findings seem to be supported by the historical data collected by Enrique-</p><p>Arias (2006, 2008), in which eighteenth-century Spanish documents produced by</p><p>Majorcan writers show no traces of yeísmo (i.e. the graphemes and exhibit</p><p>a distribution consistent with a phonemic distinction of the sounds they represent)</p><p>during a period when other comparable texts written by Spanish monolinguals</p><p>already showed signs of the phonemic opposition being lost.</p><p>2. Weakening and loss of [d] in words ending in the sequence [ado] is another wide-</p><p>spread phenomenon in the Spanish-speaking world that has been attested pro-</p><p>fusely since the 1500s (Moreno Fernández 2004: 999). In Romera’s (2003: 373)</p><p>study it was found that amongst Palma speakers its loss is more likely to happen</p><p>in the speech of older individuals, with low education levels and Spanish as L1.</p><p>2.  Currently, competence in Spanish in the Balearic Islands is almost universal while com-</p><p>petence in Catalan depends on the speaker’s place of origin. Of those residents born in the</p><p>Balearic Islands, 95.9% declare they speak Catalan “well” or “not too badly”. This percentage</p><p>goes down to 47.2% among the immigrants born outside the Islands. Overall, 76.7% of the</p><p>population in the Balearic Islands declared that they spoke Catalan in 2002 (Melià 2002: 61).</p><p>12 Andrés Enrique-Arias</p><p>These findings are consonant with the general trend, registered in several stud-</p><p>ies, that retention of [d] in words ending in [ado] is higher overall in areas where</p><p>Spanish is in contact with Catalan.3 In Blas Arroyo’s (2007) study of the realiza-</p><p>tions of -ado participles in Castellón the influence of Catalan on this outcome</p><p>is evident. His informants with Catalan as L1 exhibited 74% of retention of [d]</p><p>as opposed to 32% in those with Spanish as L1 (cf. Blas Arroyo 2007: 269–270).</p><p>This may be explained, first, because, as opposed to Spanish, in which /d/ in -ado</p><p>participles is often fricativized or even deleted, the equivalent Catalan consonant</p><p>in the participle ending is produced with greater articulatory tension, that is, as a</p><p>voiceless occlusive. Another factor to take into account is that loss of intervocalic</p><p>[d] is more common in casual speech (Moreno Fernández 2004: 1002). As most</p><p>Catalan-dominant bilinguals learn Spanish in formal contexts, their exposure to</p><p>weakening and loss of [d] is necessarily limited.</p><p>3. Weakening of /s/ in syllable-final position is another widespread expanding phe-</p><p>nomenon in the Spanish-speaking world (Moreno Fernández 2004: 993–99).</p><p>According to Romera’s (2003: 374) data from Palma, speakers classified as L1</p><p>Spanish show 24.6% maintenance of -/s/ while bilingual speakers that are Cata-</p><p>lan-dominant exhibit a higher level of maintenance (92.7%). Again, the reason is</p><p>that syllable-final /s/ in Catalan is rather robust, which explains its higher levels</p><p>of maintenance amongst Catalan-dominant bilinguals; in addition, this is the for-</p><p>mal, prestigious variant taught in school and used in the mass media.</p><p>From the previous data and analyses it follows that a possible effect of language contact</p><p>is the retention of conservative variants. All three examples are variable phenomena</p><p>that, in monolingual Spanish varieties, exhibit a historical trend of increase in fre-</p><p>quency and territorial expansion. However, we find that (a) in Spanish in contact with</p><p>Catalan the innovative variant occurs in fewer linguistic environments and/or with</p><p>lower frequency when compared to non-contact varieties of the same language; and</p><p>(b) of those Catalan–Spanish bilinguals surveyed in the different studies, those classi-</p><p>fied as Catalan L1 exhibit lower frequency of the innovative variants than those clas-</p><p>sified as Spanish L1.</p><p>There are two conditions that seem to favor this outcome: first, the conserva-</p><p>tive variant has a parallel equivalent in the contact language; second, the traditional</p><p>variant is either closer to standard, prestigious uses ([ado] pronunciation, preser-</p><p>vation of final [s]) or is neutral in this respect (maintenance of the palatal lateral</p><p>3.  As can be seen in Blas Arroyo’s (2007) comparative study of [ado] pronunciations in ten</p><p>Spanish cities, the two localities where Spanish happens to exist side by side with Catalan</p><p>(Castellón and Barcelona) are the ones that exhibit higher percentages of preservations of</p><p>final [ado].</p><p>On language contact as an inhibitor of language change 13</p><p>approximant); as Catalan L1 speakers typically learn and use Spanish in formal con-); as Catalan L1 speakers typically learn and use Spanish in formal con-</p><p>texts, they have limited exposure to innovative variants that are characteristic of less</p><p>formal registers.</p><p>4. Historical clues for contact-induced continuity in grammar</p><p>The data and analyses of phonetic and phonological phenomena discussed so far</p><p>illustrate how language contact can be an effective inhibitor of language change. In</p><p>order to explore further this kind of effect in a contact situation, I shall now turn to</p><p>analyse a number of morphosyntactic phenomena that are characteristic of the Span-</p><p>ish spoken by Catalan bilinguals. The main question to be addressed is whether these</p><p>morphosyntactic features are contact-induced innovations or, on the contrary, they</p><p>result from the preservation of language uses that were already present in general</p><p>Spanish at the time when it came into contact with Catalan. In order to answer this</p><p>question it is pertinent to incorporate an analysis of historical data of the Spanish</p><p>produced by Catalan bilinguals. In the case at hand, a diachronic analysis will show</p><p>whether there has been a historical continuity in the features that characterize Major-</p><p>can Spanish, and, if that is the case, provide critical evidence in support of the view</p><p>that a substantial number of the morphosyntactic traits that are characteristic of the</p><p>Spanish spoken in Majorca are the result of the preservation of linguistic structures</p><p>that were already present in general Spanish at the time when the language was intro-</p><p>duced to the island.</p><p>Majorca, along with the rest of the Balearic Islands, was incorporated into the King-</p><p>dom of Aragon and populated with Catalan speakers in the thirteenth century in the</p><p>context of the Christian Reconquest of Spain from the Muslims. Following the dynastic</p><p>union of Castile and Aragon in the fifteenth century, and the political, economic and</p><p>cultural hegemony of Castile among the Peninsular kingdoms that characterized the</p><p>ensuing centuries, Catalan-speaking territories fell under increasing cultural pressure</p><p>and underwent a gradual process of linguistic castilianization. By the sixteenth cen-</p><p>tury, educated, urban Majorcans had at the very least passive competence in Spanish as</p><p>evinced by abundant records of books printed in Spanish, and sermons, poetry compe-</p><p>titions, celebrations and theatre performances in Spanish. The gradual castilianization</p><p>of Majorca intensified in the eighteenth century as the Bourbon monarchs implemented</p><p>a number of measures intended to spread the use of Spanish in the administrative and</p><p>educational domains (Martínez i Taberner 2000: 355–359.). As a result, from the late</p><p>eighteenth century on, the languages of Majorca entered a period of diglossia, with</p><p>Spanish being the only language used in formal domains. In the early 1980s, with the</p><p>advent of democracy things took a drastic turn with Catalan finally achieving official</p><p>status and use in domains such as education, administration and mass media.</p><p>14 Andrés Enrique-Arias</p><p>The historical corpus used in the study is composed of a collection of letters written</p><p>in Majorca in the late eighteenth century representing a variety of writer demographics</p><p>and registers (Enrique-Arias 2006, 2008, in press).4 These data are complemented by</p><p>judicial statements from the late eighteenth and early nineteenth centuries that repro-</p><p>duce direct discourse taken from the excerpts published in Vibot (2004, 2005). The</p><p>corpus thus represents the early stages in the formation of the variety of Spanish spo-</p><p>ken by Catalan speakers in Majorca.</p><p>4.1  Morphological and periphrastic future</p><p>Since the seventeenth century, future temporal reference in Spanish has undergone a</p><p>significant change, whereby the territory of the morphological future has been pro-</p><p>gressively taken over by the periphrastic future (Aaron 2007). In the areas where Span-</p><p>ish coexists with Catalan, however, the substitution of the morphological future by the</p><p>periphrastic one is less advanced (Blas Arroyo 2004: 1068). The effect of Catalan lan-</p><p>guage contact on the retention of the morphological future has been illustrated by Blas</p><p>Arroyo’s (2007) detailed study of the distribution of future forms in the Spanish of the</p><p>Catalan-speaking region of Castellón: speakers that are Catalan-dominant exhibit sys-</p><p>tematically higher percentages of use of the morphological future (cf. Table 1, adapted</p><p>from Blas Arroyo 2007).</p><p>Table 1. Percentage of retention of morphological future in Castellón</p><p>Uses of future L1 Catalan L1 Spanish</p><p>Immediate distance</p><p>e.g. ‘We’ll go to the movies at five.’</p><p>41 30</p><p>Intermediate distance</p><p>e.g. ‘We’ll do it this week.’</p><p>58 42</p><p>Indefinite distance</p><p>e.g. ‘I don’t know when we’ll do it.’</p><p>45 39</p><p>Maximal distance</p><p>e.g. ‘When I graduate I’ll do a Ph.D.’</p><p>74 73</p><p>Both languages share the widespread Romance morphological future derived</p><p>from the Latin periphrasis infinitive + habere ‘to have’ (amare habeo > amar-é</p><p>‘I will love’). Nevertheless, unlike Spanish, which has developed a new periphrastic</p><p>future with ir ‘go’, the literal Catalan equivalent, that is, the periphrasis with anar ‘go’,</p><p>4.  The letters belong to file 6 of the epistolary archive of the Zaforteza family (henceforth</p><p>ZA) which contains nearly 250 letters (roughly half of them in Spanish and the other half</p><p>in Catalan) dated between</p><p>1739 and 1788. The examples are identified by letter number, line</p><p>number and year of composition.</p><p>On language contact as an inhibitor of language change 1</p><p>is used for the expression of preterit actions and states. The result of this asymmetry</p><p>is that, while all varieties of Spanish have shown a historical tendency to replace the</p><p>morphological future with the periphrasis with ir, Catalan bilingual speakers tend to</p><p>retain it more.</p><p>In addition, the socially dominant status of Spanish in Majorca may have con-</p><p>tributed to the retention of morphological future. Idealized models of grammar tend</p><p>to ignore the existence of the periphrastic future, even though it has become by far</p><p>the most common structure to express future in all Spanish varieties; most textbooks</p><p>and Spanish grammars only include the inflectional forms, that is, the so-called future</p><p>simple (amaré ‘I will love’) and the future perfect (habré amado ‘I will have loved’)</p><p>as examples of the future tenses. This may have reinforced the morphological future</p><p>amongst those speakers that have acquired Spanish in formal contexts.5</p><p>The data from the historical corpus suggest that this distribution of forms is a</p><p>long-established trend in Majorcan Spanish. In the Spanish letters in the ZA we find</p><p>146 examples of morphological future as opposed to only two examples of periphras-</p><p>tic future with ir. In other words, the innovative use accounts for little over 1%. This</p><p>figure seems to be rather low if we consider the available data for mainland Castilian</p><p>texts: Aaron (2007) registers 4% of periphrastic future for the seventeenth and eigh-</p><p>teenth centuries combined, and 12% for the nineteenth century. It should be noted</p><p>that Aaron’s study is based on literary works, which typically reflect more formal lan-</p><p>guage. Personal letters, on the other hand, tend to be closer to oral registers (Biber</p><p>1995: 283–300). Taking this into account, the actual difference in use between the</p><p>Majorcan and Castilian varieties is likely to be even greater.</p><p>4.2  Obligation periphrases</p><p>In Spanish there are three obligation periphrases involving the semi-auxiliaries tener</p><p>and haber – both translated as ‘have’– plus infinitive: tener de, haber de and tener que.</p><p>The first one, tener de, is the least common; haber de sounds somewhat more formal</p><p>and archaic in general monolingual Spanish while tener que is by far the most common</p><p>one. The spread of tener que in general Spanish is a relatively recent phenomenon, as</p><p>can be seen in Figure 1 (data collected from Corpus del español).</p><p>.  Lipski (1990: 53) gives a similar explanation for the quantitative preponderance of mor-</p><p>phological future in Malabo (Equatorial Guinea), another place where Spanish is a socially</p><p>dominant variety that, for many speakers, is learned as a second language in formal contexts.</p><p>1 Andrés Enrique-Arias</p><p>0</p><p>10</p><p>20</p><p>30</p><p>40</p><p>50</p><p>60</p><p>70</p><p>80</p><p>90</p><p>100</p><p>16C 17C 18C 19C 20C</p><p>tener que</p><p>tener de</p><p>haber de</p><p>Figure 1. Obligation periphrases in general Spanish (sixteenth–twentieth centuries)</p><p>The chart shows that, in the last three centuries, tener que has experienced a spec-</p><p>tacular progression, from 7.4% in the eighteenth century, to 23.7% in the nineteenth</p><p>century, to 81.3% in the last century.</p><p>However, in areas of Spanish in contact with Catalan, including Majorca, several</p><p>studies point out how the traditional forms tener de and haber de maintain their vital-</p><p>ity (Blas Arroyo 2004: 1071). Once again this seems to result from the maintenance</p><p>of structures that have receded in general Spanish but are maintained in Majorca due</p><p>to the existence of a parallel structure in Catalan, whose most common obligation</p><p>periphrases are tenir de and haver de.6 The data from the historical corpus suggest</p><p>that this feature of the Spanish spoken in Majorca is a long-established trend. Out of</p><p>twenty-two occurrences of obligation periphrases (sixteen haber de, six tener de) none</p><p>corresponds to the innovative form tener que.</p><p>4.3  Preverbal negative concord</p><p>General standard Spanish allows negative concord in postverbal position but not pre-</p><p>verbally, as exemplified in (2). In contrast, Catalan allows negative concord pre- and</p><p>post-verbally, while single preverbal negation is ungrammatical (3):</p><p>(2) nadie ha venido / *nadie no ha venido / no ha venido nadie</p><p>nobody has come / nobody not has come / not has come nobody</p><p>(3) *ningú ha vingut / ningú no ha vingut / no ha vingut ningú</p><p>nobody has come / nobody not has come / not has come nobody</p><p>In contrast to the standard variety, in Majorcan Spanish preverbal negative concord</p><p>occurs quite frequently, as in the following example taken from a local newspaper:</p><p>.  In Majorca some speakers use tenir que, a Spanish calque condemned by the norm.</p><p>On language contact as an inhibitor of language change 1</p><p>(4) La ex-alcaldesa afirmó que en ningún momento no se ha</p><p>the ex-mayor stated that in no moment not self has</p><p>planteado dimitir.</p><p>considered resign (DM March 14 2008: 19).</p><p>‘The former mayor said that she has at no point considered resigning.’</p><p>It is possible that the presence of this structure in the Spanish spoken in Majorca is the</p><p>result of the retention of a structure that was common in earlier stages of the language.</p><p>In Old Spanish negative concord was rather common both in preverbal and postverbal</p><p>position, as illustrated in the following early thirteenth-century examples extracted</p><p>from Camus (2006: 1175):</p><p>(5) a. Ascondense de mio Cid, ca no-l osan dezir nada.</p><p>they.hide of my Cid because not-him dare say nothing</p><p>‘They hide from my Cid for they dare not say anything to him.’</p><p>b. que a mio Cid Ruy Díaz, que nadi no-l diessen posada</p><p>that to my Cid Ruy Diaz, that nobody not-him give lodging</p><p>‘That no one give lodging to my Cid.’</p><p>Although this structure declined by the end of the Middle Ages, examples are still</p><p>found in the ensuing centuries (Vanrell (2008) records 37.5% of preverbal negative</p><p>concord with ninguno in a corpus of sixteenth-century notary documents). In fact, pre-</p><p>verbal negative concord has not disappeared completely as it is still common in rural</p><p>western Spanish dialects (Borrego Nieto 1996: 157, Aguado Candanedo 1984: 221).</p><p>Considering all this, we can conclude that preverbal negative concord was likely to</p><p>have been a valid option in the Spanish introduced in Majorca in the 1600s and 1700s.</p><p>Quite possibly, the existence of a parallel Catalan structure encouraged the retention</p><p>of this structure. The historical corpus provides evidence of these uses in the Spanish</p><p>produced by Majorcans in the 1700s as in the following example:7</p><p>(6) Nunca no he dejado de rogar a Nuestro Señor.</p><p>never not I.have stopped of begging to our lord</p><p>‘I have never stopped praying to Our Lord.’ (ZA 116, 24, 1772)</p><p>.  As Silva-Corvalán (1998: 237) points out, a similar situation and explanation can be found</p><p>in studies of Spanish–Guaraní contact. Preverbal negative concord in Paraguayan Spanish (nada</p><p>no dije, literally “nothing not said”–‘I said nothing’) has also been explained (Granda 1991)</p><p>as the result of the retention of the structure which, as has been already indicated, existed in</p><p>general Spanish at least until the sixteenth century. The retention has been favored by contact</p><p>with Guaraní, where it exists as a standard construction (cf. Silva-Corvalán 1998: 237):</p><p>(i) mba’eve nda-’ei</p><p>nothing not-said</p><p>‘I said nothing.’</p><p>1 Andrés Enrique-Arias</p><p>4.4  Lexical-semantic transfer with syntactic consequences</p><p>A common outcome of language contact is lexical borrowing with syntactic con-</p><p>sequences as in the following case. General standard Spanish has two distinct</p><p>verbs to convey the idea of ‘ask’: pedir ‘to ask someone to do or give something’</p><p>and preguntar ‘to ask a question’. Besides this lexical contrast, the two verbs have</p><p>different argument structure, which in the case of pedir entails a syntactic con-</p><p>straint. In general standard Spanish, only preguntar but not pedir can subcatego-</p><p>rize an interrogative</p><p>clause (7a). In contrast, Catalan has one verb, demanar, which</p><p>can be used in either sense, and which can therefore subcategorize interrogative</p><p>clauses (7b).</p><p>(7) a. Le *pidió/preguntó cómo estaba.</p><p>him asked how was</p><p>b. Li va.demanar com estava.</p><p>him asked how was</p><p>‘He asked him how he was doing.’</p><p>Apparently as a result of contact with Catalan, pedir in the Spanish spoken in Majorca</p><p>has the semantic and syntactic properties of Catalan demanar. Therefore structures</p><p>like (7a) are grammatical with either verb.</p><p>The studies carried out from a purely synchronic perspective (cf. Moll 1961: 471;</p><p>Serrano Vázquez 1996: 384–85) attribute these uses of pedir to direct transfer from</p><p>Catalan. However, they miss the fact that the use of pedir in the sense of ‘to ask a ques-</p><p>tion’ was not a rarity in eighteenth century general standard Spanish; in fact, in the</p><p>Royal Academy’s 1737 dictionary, and all its subsequent editions through 1852, pedir</p><p>is defined in its second meaning as “to ask or obtain information from someone on</p><p>something”. Furthermore, the very example provided by the Royal Academy to illus-</p><p>trate the correct use of the word features pedir subcategorizing an interrogative clause</p><p>(Real Academia: s.v. pedir):8</p><p>(8) Pedían cuál era el camino.</p><p>they.asked which was the road</p><p>‘They asked which road it was.’</p><p>.  The example, though, is taken from Criticón, a literary work written one century earlier</p><p>by Baltasar Gracián (1601–1658). With the aim of justifying the inclusion of words in their</p><p>Dictionary and to illustrate their “correct use” the academics picked excerpts from prestigious</p><p>literary works. The fact that this example was included in the dictionary suggests that this use</p><p>of pedir was approved of and endorsed by the educated classes.</p><p>On language contact as an inhibitor of language change 1</p><p>Again, the historical corpus shows that this structure is not new to Majorcan Spanish.</p><p>The eighteenth-century letters and court testimonies contain examples in which pedir</p><p>subcategorizes an interrogative clause:</p><p>(9) Pidióle si contava la verdad.</p><p>he.asked.him whether he.told the truth</p><p>‘He asked him whether he was telling the truth.’ (Vibot 2005: 66, 1833)</p><p>Although lexical borrowing may be a mechanism leading to eventual syntactic change,</p><p>the uses of pedir in the Spanish spoken in Majorca do not seem to be the outcome of</p><p>direct borrowing of syntactic structures or rules from Catalan into Spanish. Catalan</p><p>contact has resulted not so much in the introduction of an innovative structure, but in</p><p>the preservation of a language usage that existed in general Spanish, reinforced by the</p><p>existence of a parallel structure in Catalan.</p><p>4.  Deixis in demonstratives</p><p>Another common phenomenon in the Spanish of Catalan bilinguals is the tendency to</p><p>adopt a two-member deixis in the demonstratives, in contrast with monolingual Span-</p><p>ish, which has a tripartite distinction (see Table 2, adapted from Vann 1998).</p><p>Table 2. Catalan and Spanish deictic systems</p><p>Catalan Spanish</p><p>aquest ‘this’ ‘that’</p><p>object near S or A</p><p>este ‘this’</p><p>object near S</p><p>ese ‘that’</p><p>object not too far from S</p><p>aquell ‘yonder’</p><p>object not near S or A</p><p>aquel ‘yonder’</p><p>object very far from S</p><p>aquí ‘here’ ‘there’</p><p>object near S or A</p><p>aquí/acá ‘here’</p><p>object near S</p><p>ahí ‘there’</p><p>object not too far from S</p><p>allà ‘yonder’</p><p>object not near S or A</p><p>allí/allá ‘yonder’</p><p>object very far from S</p><p>S = speaker, A = addressee.</p><p>In the Spanish spoken in Catalan-speaking territories there is a tendency to</p><p>adopt the deictic and pragmatic properties of Catalan deictic terms, using este ‘this’</p><p>and aquí ‘here’ where monolingual Spanish speakers generally use ese ‘that’ and ahí</p><p>‘there’ respectively. For instance, in Vann’s (1998) data from 58 interviews of Barcelona</p><p>speakers, all the informants demonstrated some kind of influence from Catalan deictic</p><p>systems, in a number of individuals 80% of the time. The same phenomenon is widely</p><p>found in Majorca, as pointed out by Moll (1961: 471).</p><p>Vann (1998) repeatedly characterizes these uses as contact-driven innovations. But</p><p>again a historical perspective suggests that the tendency towards a bipartite distinction</p><p>11 Andrés Enrique-Arias</p><p>in the deixis of the demonstratives may have its roots in Spanish itself. Classical Latin</p><p>had a three-place system (hic, iste and ille). As hic was eliminated from speech</p><p>in Late Latin, the three-member system became reduced to the two-member system</p><p>that prevails in French and Italian among other Romance varieties. In Spanish (and</p><p>Portuguese), however, the three-place system was subsequently restored by transfer of</p><p>ipse from its emphatic role to second person demonstrative role. The restoration of the</p><p>three-member system was a slow process only consummated in early modern times.</p><p>Badia Margarit (1952: 5) explains that in The Song of el Cid (ca. 1200) esse and aquel</p><p>do not have their respective uses clearly delimited. Likewise, in a study of demonstra-</p><p>tives in fifteenth century Spanish, Eberenz (2000: 263) points out that confusion of</p><p>este and esse was rather common. This loss of distinction occurs in spatial deixis and is</p><p>even more common in anaphoric co-reference, where the opposition between the two</p><p>forms is completely neutralized: for example, it is impossible to distinguish differences</p><p>in use between expressions like por esto and por esso ‘that’s why’ (Eberenz 2000: 264).</p><p>In a similar fashion, the series of Castilian adverbs of place related to the demonstra-</p><p>tives maintained a two-member system throughout the Middle Ages (aquí/acá ‘here’</p><p>allí/allá ‘there’). The emergence of the third member ahí occurs rather late: the form is</p><p>first documented ca. 1490 (Badía Margarit 1952: 6).</p><p>According to Eberenz the persistence of the tendency to maintain a two-member</p><p>system in Castilian is due to a combination of different historical factors: (a) in ori-</p><p>gin, Latin iste (> este) had the referential properties of Modern Spanish ese, that is,</p><p>historically este referred to the hearer before esse assumed that function; (b) in many</p><p>communicative situations speaker and hearer share common space and topics, which</p><p>makes it unnecessary to establish clear-cut differences between este and esse; (c) in the</p><p>anaphoric co-reference there has been a process of convergence of the deictic value of</p><p>este and the identifying one of esse (</p><p>un cafè hacer un café tomar un café</p><p>‘make a face’ fer mala cara hacer mala cara poner mala cara</p><p>‘be disgusting’ fer fàstic hacer asco dar asco</p><p>‘smell’ fer olor hacer olor dar olor</p><p>‘scare’ fer por hacer miedo dar miedo</p><p>Again this feature of the Spanish from Majorca represents the continuation of uses</p><p>that were normal in general Spanish in earlier stages of the language. In Old Spanish</p><p>fazer ‘make’ and dar ‘give’ competed in numerous collocations: tristura ‘grief ’, enojo</p><p>‘anger’, gozo ‘joy’, plazer ‘pleasure’ (Dubský 1962). In some cases (asco ‘disgust’, lástima</p><p>‘pity’ and rabia ‘rage’) occurrences with hacer outnumbered those with dar well into</p><p>the 1600s (Alba-Salas 2007). Moreover, the overuse of hacer in collocations is also</p><p>present in general Spanish in colloquial registers (Briz 2004: 126).</p><p>There is evidence that the use of collocations with hacer in Spanish in contact with</p><p>Catalan is not a recent innovation; the structure is condemned vigorously in gram-</p><p>mars of the eighteenth and nineteenth centuries addressed to teach Catalans how to</p><p>speak proper Spanish (see Solà 1980 for a detailed study of such grammars).</p><p>4.  Prepositional uses</p><p>In the Spanish spoken by Catalan–Spanish bilinguals in Majorca, the preposition en is</p><p>often used to express direction, something already observed by Moll (1961: 472), who</p><p>provides examples like the following:</p><p>(12) Se fue en Barcelona.</p><p>ref he.left in Barcelona</p><p>‘He left for Barcelona.’</p><p>(13) A las nueve llegamos en la ciudad.</p><p>at the nine we.arrived in the city</p><p>‘At nine we arrived in the city.’</p><p>112 Andrés Enrique-Arias</p><p>I have collected similar examples from the local press in the Balearic Islands:</p><p>(14) La pareja se trasladó en esta urbanización.</p><p>the couple ref moved in this housing-complex</p><p>‘The couple moved to this housing complex.’ (DM June 9, 2006: 28)</p><p>The structure in examples (12–14) is ungrammatical in standard Spanish, which uses</p><p>en for location and a for direction as shown in the examples under (15):</p><p>(15) a. Estoy *a/en Palma.</p><p>‘I am in Palma.’</p><p>b. Voy *en/a Palma.</p><p>‘I go to Palma.’</p><p>In contrast, the Catalan spoken in Majorca uses a to express both location and direc-</p><p>tion as exemplified in (16):9</p><p>(16) a. Som *en/a Palma.</p><p>‘I am in Palma.’</p><p>b. Vaig *en/a Palma.</p><p>‘I go to Palma.’</p><p>Moll (1961: 472) attributes the non-standard preposition use in (12–13) to the different</p><p>distribution of a and en in Catalan and Spanish, which purportedly causes confusion</p><p>for bilingual speakers. But this explanation is at best incomplete: if Majorcan Catalan</p><p>uses a for situation and direction, one would expect that in their Spanish utterances</p><p>Majorcans would overuse the preposition a, not en, to express spatial relations; that</p><p>is, from a comparison of the related Spanish and Catalan structures it follows that the</p><p>directional uses of en in the Spanish from Majorca are not the result of simple direct</p><p>transfer from present-day Catalan.</p><p>.  In the Catalan of Majorca the preposition a develops an epenthetic -[n] segment when it</p><p>occurs in front of articles and demonstratives that begin with a vowel.</p><p>(i) Vaig an es poble.</p><p>‘I go to the town.’</p><p>This phonetically conditioned variant of the preposition a is homophonous with the preposition</p><p>en (they are both pronounced [6n]) and in fact is usually spelled en, which gives the impression</p><p>that there is a generalized confusion of the prepositions a and en in Balearic Catalan.</p><p>However, if we look at those cases in which there is no epenthetic -[n] it is clear that there is</p><p>not such confusion:</p><p>(ii) Vaig *en/a sa plaça.</p><p>‘I go to the square.’</p><p>On language contact as an inhibitor of language change 113</p><p>As I have explained in greater detail in Enrique-Arias (2005), there are a number</p><p>of synchronic and diachronic factors that help explain the existence of this structure</p><p>in the grammar of bilingual Majorcans. A relevant piece of data is that the directional</p><p>uses of en are common in the varieties of Spanish that come into contact with other</p><p>languages, such as Galician in Galicia (Rojo 2004: 1097), Guaraní in Paraguay (Choi</p><p>2001), English in Texas (García 1982) or Fang in Equatorial Guinea (Granda 1988;</p><p>Lipski 1990). This suggests that we are dealing with a phenomenon motivated by the</p><p>tendency toward simplification and restructuring of grammatical rules that is com-</p><p>mon in language contact situations. It should also be noted that this area of grammar is</p><p>particularly vulnerable to change, since the uses of a and en, or their equivalents, have</p><p>fuzzy limits within monolingual varieties of all the Romance languages (cf. examples</p><p>in Choi 2001: 188–193). For instance, in standard Spanish en may be used to express</p><p>direction (entrar en la casa ‘enter the house’) while a in certain cases may express static</p><p>location (esperar a la puerta ‘wait by the door’). Thus, the contact situation has the</p><p>effect of extending a structure that already existed in general Spanish to more con-</p><p>texts. Finally, one must take into account the inherent difficulty in learning and using</p><p>prepositions: in a study of the acquisition of ten Spanish morphemes by monolingual</p><p>Spanish children ages 2:0 to 4:8 Kvaal et al. (1988) found that the preposition en was</p><p>one of the latest features to be acquired, only second to the irregular preterit indicative.</p><p>The difficulty of acquiring the standard use of prepositions may be even greater for</p><p>unbalanced bilinguals especially when, as in the case of Catalan-dominant bilinguals,</p><p>there is no clear frame of reference in the speaker’s first language.</p><p>A diachronic perspective complements the synchronic factors outlined above,</p><p>unveiling other circumstances that may have contributed to the introduction of the</p><p>directional uses of en in Majorcan Spanish. Unlike present-day Catalan, in the language</p><p>of the eighteenth century en was frequently used to express direction, as documented</p><p>in the Catalan letters in the ZA corpus:</p><p>(17) Aurà de passar en Vic. (ZA 20, 21, 1743)</p><p>will.have of pass in Vic</p><p>‘He will have to get to Vic.’</p><p>(18) Arribà lo dia de Pasqua en Madrit. (ZA 52, 6, 1756)</p><p>he.arrived the day of Easter in Madrid</p><p>‘He arrived in Madrid at Easter.’</p><p>Spanish letters written by Majorcans in the corpus also use en to express direction:</p><p>(19) Llegamos en esse puerto. (ZA 70, 11, 1759)</p><p>we.arrived in that port</p><p>‘We arrived in that port.’</p><p>(20) En el viernes pasado pasamos en Palma. (ZA 121, 24, 1773)</p><p>in the Friday past we.passed in Palma</p><p>‘Last Friday we went to Palma.’</p><p>114 Andrés Enrique-Arias</p><p>The historical data suggest that the current directional uses of en in Majorcan</p><p>Spanish are the continuation of a structure that was already firmly established in the</p><p>Spanish used in Majorca in the eighteenth century. In turn, this feature of Majorcan</p><p>Spanish would have been encouraged by the existence of a parallel construction in</p><p>the Catalan of the eighteenth and nineteenth centuries.</p><p>. Summary and conclusions</p><p>In this chapter I have analysed the historical development of a number of structures</p><p>found in the variety of Spanish spoken in Majorca in order to argue that language</p><p>contact can be an effective inhibitor of language change. The phenomena studied</p><p>involve linguistic variants that have decreased in frequency or even disappeared in</p><p>general Spanish but exhibit considerable vitality in the Majorcan variety. While pre-</p><p>vious studies simply attribute these features of Majorcan Spanish to direct transfer</p><p>from Catalan, the explanation offered here is that the Spanish spoken in Majorca has</p><p>not so much borrowed or adopted Catalan morphosyntactic structures, but rather has</p><p>retained existing variants that paralleled Catalan structures. Some of these, like the</p><p>morphological future or the expression of obligation with haber de are, while receding</p><p>in most varieties, valid options in standard Spanish. However, some other structures,</p><p>like preverbal negative concord, have disappeared in general Spanish and survive only</p><p>in marginal rural dialects.</p><p>Their maintenance in the Spanish of Majorca can be seen as</p><p>evidence of how parallel structures favour retention of archaic features.</p><p>The historical analysis yields two findings that support the view that Catalan</p><p>language contact has favored the retention of archaic features in Majorcan Spanish.</p><p>To begin with, all the phenomena examined had some sort of existence in monolin-</p><p>gual Spanish varieties at the time when the language was introduced to the island</p><p>of Majorca. This poses a challenge to the view that these structures are due to direct</p><p>transfer from Catalan, as one of the necessary conditions to establish the existence of</p><p>contact-induced change is to prove that the proposed interference features were not</p><p>present in the pre-contact variety (see for instance Thomason 2001: 93–94). Further-</p><p>more, the data from the historical corpus of Majorcan Spanish reveal that there has</p><p>been a historical continuity for some of the features that characterize the variety of</p><p>Spanish spoken in Majorca today. Additionally, it has been shown how for some phe-</p><p>nomena the retention of traditional features may have been further encouraged by the</p><p>socially dominant status of Spanish in the last three centuries. As Catalan-dominant</p><p>bilinguals tended to acquire and use Spanish in formal settings, they used less those</p><p>innovative variants that were characteristic of casual speech.</p><p>Evidently, the whole topic of mechanisms of cross-linguistic interference in the</p><p>history of Spanish–Catalan contact is far from being completely understood. In any</p><p>On language contact as an inhibitor of language change 11</p><p>case, I hope to have demonstrated the relevance of a diachronic perspective in obtain-</p><p>ing a more complete understanding of the different factors involved in the formation</p><p>of contact varieties.</p><p>Primary sources</p><p>Davies, Mark. 2008. Corpus del español. [Consulted on line at http://www.corpusdelespanol.org].</p><p>Diario de Mallorca (1955-) [=DM]. Palma de Mallorca: Editorial Balear</p><p>Real Academia Española. Nuevo tesoro lexicográfico de la lengua española [Consulted on line at</p><p>http://www.rae.es].</p><p>Vibot, Tomàs. 2004. Crims, infàmies i immoralitats a Esporles i Banyalbufar (Segles XVII-XIX).</p><p>Palma: Associació Cultural Index, Edicions el Moixet.</p><p>Vibot, Tomàs. 2005. Crims, infàmies i immoralitats a Bunyola i Orient (Segles XVI-XIX). Bunyola:</p><p>Collectiu Cultural Sitja.</p><p>Zaforteza family epistolary archive [=ZA]. File 6. Vinagrella estate (Llubí), Majorca.</p><p>References</p><p>Aaron, Jessi E. 2007. El futuro epistémico y la variación: Gramaticalización y expresión de la</p><p>futuridad desde 1600. Moenia 13: 253–274.</p><p>Aguado Candanedo, David. 1984. El habla en Bercianos del Real Camino (León): Estudio socio-</p><p>lingüístico. León: Diputación de León.</p><p>Alba-Salas, Josep. 2007. On the life and death of a collocation: A corpus-based diachronic study</p><p>of dar miedo/hacer miedo-type structures in Spanish. Diachronica 24(2): 207–252.</p><p>Badia i Margarit, Antoni M. 1952. Los demostrativos y los verbos de movimiento en iberor-</p><p>románico. In Estudios dedicados a Menéndez Pidal, Vol. 3, 3–31.</p><p>Biber, Douglas. 1995. Dimensions of Register Variation: A Cross-Linguistic Comparison. Cam-</p><p>bridge: CUP.</p><p>Blas Arroyo, José Luis. 2004. El español actual en las comunidades del ámbito lingüístico catalán.</p><p>In Historia de la lengua española, Rafael Cano Aguilar (ed.), 1065–1086. Barcelona: Ariel.</p><p>Blas Arroyo, José Luis. 2007. El contacto de lenguas como factor de retención en procesos de</p><p>variación y cambio lingüístico. Datos sobre el español en una comunidad bilingüe penin-</p><p>sular. Spanish in Context 4(2): 263–29.</p><p>Borrego Nieto, Julio. 1983. Norma y dialecto en el sayagués actual. Salamanca: Acta Salmaticensia.</p><p>Briz, Antonio. 2004. El castellano en la Comunidad Valenciana. Revista Internacional de Lingüís-</p><p>tica Iberoamericana 2(2): 119–130.</p><p>Camus Bergareche, Bruno. 2006. La expresión de la negación. In Sintaxis histórica de la lengua</p><p>española. Primera parte: La frase verbal, Concepcion Company (ed.), Vol. 2, 1165–1252.</p><p>Mexico: Fondo de Cultura Económica and Universidad Nacional Autónoma de México.</p><p>Choi, Jenny K. 2001. The genesis of voy en el mercado: The preposition en with directional mean-</p><p>ing in Paraguayan Spanish. Word 52: 181–196.</p><p>Dubský, Josef. 1962. Las formas descompuestas en el español antiguo. Revista de Filología Espa-</p><p>ñola 46: 31–48.</p><p>11 Andrés Enrique-Arias</p><p>Eberenz, Rolf. 2000. El español en el otoño de la Edad Media. Madrid: Gredos.</p><p>Enrique-Arias, Andrés. 2005. On the origin of the preposition en to express ‘direction to’ in the</p><p>Spanish spoken in Mallorca. Paper read at the 17th International Conference on Historical</p><p>Linguistics, Madison (Wisconsin).</p><p>Enrique-Arias, Andrés. 2006. Spanish/Catalan contact in historical perspective: 18th century</p><p>documents from Majorca. In New Perspectives in Iberian Dialectology/Nouvelles perspec-</p><p>tives en dialectologie ibérienne, David Heap, Enrique Pato & Claire Gurski (eds). London:</p><p>The University of Western Ontario. (online edition</p><p>Joan. 2002. La situació lingüística a les Illes Balears: Comentaris al voltant d’una enquesta.</p><p>Llengua i Ús: Revista Tècnica de Política Lingüística 24: 61–64.</p><p>On language contact as an inhibitor of language change 11</p><p>Moll, Francisco de B. 1961. El castellano en Mallorca. In Studia Philologica. Homenaje ofrecido a</p><p>Dámaso Alonso por sus amigos y discípulos con ocasión de su 600 aniversario, Vol. 2, 469–474.</p><p>Madrid: Gredos.</p><p>Moreno Fernández, Francisco. 2004. Cambios vivos en el plano fónico del español: variación</p><p>dialectal y sociolingüística. In Historia de la lengua española, Rafael Cano Aguilar (ed.),</p><p>973–1009. Barcelona: Ariel.</p><p>Penny, Ralph. 2000. Variation and Change in Spanish. Cambridge: CUP.</p><p>Rojo, Guillermo. 2004. El español de Galicia. In Historia de la lengua española, Rafael Cano</p><p>Aguilar (ed.), 1087–1101. Barcelona: Ariel.</p><p>Romera, Magdalena. 2003. La variedad del castellano actual en Baleares. Moenia 9: 359–381.</p><p>Serrano Vázquez, María del Carmen. 1996. Interferencias léxicas y semánticas en una situación</p><p>de contacto entre dos lenguas, catalán y castellano. In Las lenguas en la Europa Comuni-</p><p>taria II, Mercè Pujol Berché & Fermin Sierra Martínez (eds), 375–394. Amsterdam: Diálo-</p><p>gos Hispánicos.</p><p>Silva-Corvalán, Carmen. 1994. Language Contact and Change. Spanish in Los Angeles. Oxford: OUP.</p><p>Silva-Corvalán, Carmen. 1998. On borrowing as a mechanism of syntactic change. In Romance</p><p>Linguistics: Theoretical Perspectives [Current Issues in Linguistic Theory 160], Armin</p><p>Schwegler, Bernard Tranel & Myriam Uribe-Etxebarria (eds), 225–246. Amsterdam: John</p><p>Benjamins.</p><p>Solà, Joan. 1980. Tractats de catalanismes. In Miscellània Aramon i Serra: estudis de llengua i lit-</p><p>eratura catalanes oferts a Ramon Aramon i Serra en el seu setantè aniversari, Emilio Alarcos</p><p>Llorach et al. (eds), Vol. 2, 559–582. Barcelona: Curial.</p><p>Thomason, Sarah. 2001. Language Contact. An Introduction. Edinburgh: EUP.</p><p>Vann, Robert E. 1998. Pragmatic transfer from less developed to more developed systems: Spanish</p><p>deictic terms in Barcelona. In Romance Linguistics: Theoretical Perspectives [Current Issues</p><p>in Linguistic Theory 160], Armin Schwegler, Bernard Tranel & Myriam Uribe-Etxebarria</p><p>(eds), 307–318. Amsterdam: John Benjamins.</p><p>Vanrell, Bruno. 2008. La evolución de la doble negación preverbal en el castellano de la Edad</p><p>Media. Ms, Universitat de les Illes Balears.</p><p>Towards notions of comparative continuity</p><p>in English and French*</p><p>Remus Gergel</p><p>Universität Tübingen</p><p>The present chapter investigates comparative clauses in English and French</p><p>with a special focus on subject–verb inversion. It explores the derivation and</p><p>diachronic evolution of such clauses and makes a case for continuity by defending</p><p>a simple derivation that requires a so-called low subject, that is one that is not</p><p>forced to the edge of the clause. Based on well-known options from the diachrony</p><p>of English and French, I argue that the non-movement syntax proposed is the</p><p>key continuous aspect against the background of several changes given in the</p><p>histories of the two languages.</p><p>1.  Introduction</p><p>The present chapter investigates comparative clauses in English and French with a</p><p>special focus on subject–verb inversion. Initial cases of comparative inversion (CI) are</p><p>illustrated in (1) below.1</p><p>*The present enterprise owes special thanks to John Vanderelst for discussions on French.</p><p>I am very grateful to the organizers of the Cambridge workshop for the opportunity to</p><p>explore comparative clauses in a continuous way, to Chris Lucas also for numerous helpful</p><p>comments concerning both form and content, to Jacqueline Guéron for a very instructive</p><p>discussion on inversion, to Polina Berezovskaya for hints on materials and proofreading, and</p><p>to two reviewers for their exceptionally constructive comments. Only I am responsible for</p><p>any remaining errors.</p><p>1.  CI is register-based and optional. This chapter focuses on the continuity of the inverted</p><p>clauses. Potentially interesting aspects in the relationship between inverted and non-inverted</p><p>derivations from the point of view of grammatical theory are thus not within present scope</p><p>(cf. Culicover & Winkler 2008; Gergel 2008; Gergel, Gengel & Winkler 2007; Merchant 2003;</p><p>Winkler 2005, among others, for some suggestions). At the same time, nothing special is</p><p>assumed for non-inverted comparatives here.</p><p>1 Remus Gergel</p><p>(1) a. [T]he Scots have spent less on their system [than has the prison service of</p><p>England and Wales].</p><p>(http://news.bbc.co.uk/2/hi/uk_news/magazine/3932653.stm)</p><p>b. And they were no more persuaded by democracy [than was the Pharaoh].</p><p>(http://www.nytimes.com/2008/09/11/opinion/11Cohen.html)</p><p>c. Optogenetic stimulation can potentially target problem cells much more</p><p>precisely [than can the electrodes used in DBS].</p><p>(Scientific American 08/2008, p. 59)</p><p>We will argue that the key continuous element in the diachronic trajectory is a low</p><p>subject. Standardly, the initial line-up of the subject is within the VP (and below the</p><p>auxiliary position). But the subject is then usually assumed to move to the edge of the</p><p>clause. A low subject on the derivation proposed here means in particular that the</p><p>subject is not moved to the edge of the clause (technically, the inflectional domain),</p><p>as would be required on standard generative approaches. Two observations worth</p><p>making from the start are: (i) from a diachronic point of view, this type of syntactic</p><p>behavior is well-known independently of comparatives from early Old French/Eng-</p><p>lish (OF/OE) onwards; (cf. Adams 1987; Haeberli 2002; Kroch 2008); (ii) several syn-</p><p>tactic properties which could potentially affect comparatives have changed in both</p><p>languages, though without always impacting on the derivation of CI (we will restrict</p><p>attention to the most direct interferers). We will argue that the now relic possibility of a</p><p>lack of subject not moved to the Spec,TP position is the factor that remained constant</p><p>in comparatives in both languages and which derives the core similarities.</p><p>The chapter is structured as follows. Section 2 gives some details on how the syntax</p><p>of comparatives in English shows various signs of an archaism, which consists in the</p><p>possibility of not filling the Spec,TP position overtly. This seems to go against the grain</p><p>of the requirement that every finite clause have an overt subject in that very position</p><p>(the so-called EPP feature). But we show that such structures have been possible in</p><p>comparatives at all major stages of the language (via null expletives; see Haeberli 2002;</p><p>Hulk & van Kemenade 1995; Williams 2000, among others). Section 3 discusses a</p><p>similar phenomenon in French, for which we present evidence that it has also been</p><p>persistent from early Old to Modern French and for which we develop a similar analy-</p><p>sis. In Section 4, additional aspects relating to (lack of) change in comparatives as well</p><p>as potential interferences with the simplest continuity hypothesis are discussed. We</p><p>thus investigate the potential contact of English with French and other features that</p><p>have been claimed in the literature to have an influence on V2 (e.g. Northern features).</p><p>Section 5 concludes the chapter and formulates certain open questions including the</p><p>(non)interaction with verb movement and the possibility of also having semantic con-</p><p>tinuity in comparatives.</p><p>Towards notions of comparative continuity in English and French 11</p><p>.  Comparative inversion in English</p><p>In this section, we propose that CI is simpler than on the standard generative view and</p><p>that the analysis proposed is historically consistent. It represents an archaic possibility</p><p>of the language. Earlier English could produce apparent V2 (and with subject pronouns</p><p>clearly V3) structures by moving the finite element to T (and not exclusively higher up</p><p>to C); see Fischer et al. (2000); Haeberli (2002); Pintzuk (1991); Speyer (2008), among</p><p>others. It is crucial that the subject is not displaced to the edge of TP in</p><p>such structures.</p><p>(Or if it were, there would be no inversion). We will argue that CI patterns with this</p><p>type of simple inversion and show how its derivation fits the current grammars first</p><p>(Section 2.1), before addressing its diachronic evolution (in Section 2.2).</p><p>.1  English CI involves no movement to C</p><p>At first sight, inversion in clausal comparatives, as in (1) above, seems amenable to a</p><p>raising of the finite element to C. This corresponds to the analysis of matrix questions</p><p>in English or run-of-the mill V2 structures in Germanic. (See Merchant 2003; Gergel,</p><p>Gengel & Winkler 2007, among others.) But there is an alternative and diachronically</p><p>more cogent analysis of CI. To illustrate the difference: on a standard analysis, the sub-</p><p>ject is moved to the edge of the clause (Spec,TP) and the finite element is moved to C;</p><p>(2a). Under the alternative to be defended, however, the subject is in a sub-T position</p><p>(Spec,VP here) and the finite element is in T; (2b).2</p><p>(2) a. Structure for the C-based analysis</p><p>CP</p><p>C</p><p>Finite</p><p>(Comp Op) TP</p><p>C′</p><p>DP</p><p>Subject T VP</p><p>T′</p><p>V′DP</p><p>tsubject …</p><p>wh</p><p>.  This proposal in and of itself is independent of how a finite element can (or cannot)</p><p>get to T. Unlike modals, do and have/be (cf. (1)), lexical verbs cannot appear in T in current</p><p>English, as a reviewer notes. Clearly, French and earlier English have the additional possibility</p><p>of V-to-T; see Section 5.</p><p>1 Remus Gergel</p><p>b. T-based analysis</p><p>CP</p><p>C</p><p>than/as</p><p>(Comp Op) TP</p><p>C′</p><p>-</p><p>T VP</p><p>T′</p><p>V′DP</p><p>Subject …</p><p>wh</p><p>Finite</p><p>Thus, following Culicover & Winkler (2008) and Gergel (2008), we argue that neither</p><p>T nor the subject undergoes movement to C and Spec,TP, respectively.</p><p>An argument for the alternative that does without the two movement steps is that,</p><p>in sequences of auxiliaries, the subject is not forced to follow the first auxiliary, (3); see</p><p>Huddleston & Pullum (2002), Potts (2002), among others. This contrasts with ques-</p><p>tions, which do not allow auxiliaries to invert together around the subject, as shown</p><p>in (4).3</p><p>.  The formulation “do not allow” for questions becomes crucial to capture the appropriate</p><p>contrast. Comparatives do allow auxiliary clusters to appear inverted around the subject, as in</p><p>(3). They do not force the sequences across all contexts, however, cf. (i). The interaction of the</p><p>conditions in CI may thus be even more complex than has been thought (e.g. in Culicover &</p><p>Winkler 2008; Gergel 2008; Huddleston & Pullum 2002). An important independent factor is</p><p>that in limited cases under the scope of negation, English allows intercalated subjects in one</p><p>type of CI after all, as in (i).</p><p>(i) a. Abouten it is gras spryngyng,/For moiste so thikke and wel liking/That it ne</p><p>may in wynter dye/No more than may the see be drye.</p><p>(Romaunt of the Rose, Frag.A:1563)</p><p>b. Of what he spoke I could tell you no more than could the children of Hamelin</p><p>have told the tune the Pied Piper played.</p><p>(J. Jerome, Paul Kelver, p. 34. London: Hutchinson)</p><p>Given the well-known inversion-triggering potential of negation in English over various</p><p>environments including different types of degree constructions (cf. Borroff 2006; Green 1976,</p><p>among many others) this is not entirely surprising. What is crucial for immediate purposes</p><p>is that in particular in base examples such as (3) subject intervention is impossible, while in</p><p>interrogatives as in (4) it is obligatory.</p><p>Towards notions of comparative continuity in English and French 1</p><p>(3) The notes […] are therefore more helpful [than could be the opinions of many</p><p>other critics, including myself].</p><p>(http://www.amazon.com/review/RSKNPKI3HDRUB, 06/2007)</p><p>(4) a. Could that grumpy man be the next prime minister?</p><p>b. *Could be that grumpy man the next minister?</p><p>Another possible argument for the proposal obtains under the assumption that the</p><p>introducer of clausal comparatives than is inserted under C (Hankamer 1973; cf. (2b)).</p><p>C-based V2 is, then, implausible since such movements targeting C do not take place</p><p>in conjunction with overt complementizers (e.g. den Besten 1983). This derives the sim-</p><p>ple analysis of CI. But there is a valid objection, as pointed out by a reviewer.4 Current</p><p>English grammar has the (restricted) capacity to generate certain inversion structures</p><p>under a complementizer such as that with the follow-up possibility of CP-recursion</p><p>(cf. Davison 1979; Green 1976; Iatridou & Kroch 1992, among others). The variety</p><p>of such inversion structures would certainly deserve more space by itself. We discuss</p><p>two types that are often invoked and then their immediate relationship to CI: negative</p><p>preposing and locative inversion. While traditionally called main clause phenomena,</p><p>such structures can surface within embedded contexts. We illustrate this in (5), for</p><p>locative inversion, and (6), for negative inversion (cf. Green 1976: 384, 385).</p><p>(5) I realized that in would come the milkman, with me there, and my hair in curlers.</p><p>(6) I knew that never before had prices been so high.</p><p>.  As the reviewer also points out, a further issue is colloquial English “than what” construc-</p><p>tions. An illustration is (i), on a variation of (1a) in Section 1.</p><p>(i) The Scots have spent less on their system than what the prison service of England has.</p><p>Such constructions have been invoked in particular since Chomsky (1977) as an argument</p><p>for wh-movement. However, in a series of more thorough studies the observation has been</p><p>refuted in its quality as an argument for genuine clausal comparatives (cf. especially den</p><p>Besten 1978). What den Besten shows is that such structures are best treated as free relatives.</p><p>That is, as an equivalent to Hankamer’s non-clausal “than XP”. Notice at the same time that if</p><p>Chomsky’s remark hadn’t turned out to need refinement, the impossibility of inversion could</p><p>be explained on the account we argue against. For instance, as follows, capitalizing on the</p><p>reviewer in question’s remarks: assume a featural difference between C in comparatives and</p><p>interrogatives, from which the inviolability of the Doubly-filled Comp filter in comparatives</p><p>follows. Finally, there is one more caveat to the issue (independently of the analysis of CI one</p><p>favors). The impossibility of inversion in “than what” structures does perhaps not have to be</p><p>enforced within narrow syntax (Lasnik & Sobin 2000; Merchant 2003). Instead, the colloquialism</p><p>of such structures and the high-register status of CI place them in disjoint grammars.</p><p>1 Remus Gergel</p><p>The structures contain a complementizer. Hence the possibility of recursion. (Simply</p><p>put: to accommodate that in one CP, and the preposed constituent in the specifier</p><p>position of the recursively subordinated one). However, a first incongruity obtains</p><p>between locative inversion and than (see Aissen 1975: 8; Hartmann 2005: 95):</p><p>(7) *On the ground lie more apples than on the tree grow.</p><p>The recursion possibilities in comparative contexts with negative preposing can also</p><p>be shown (empirically) to be more restricted than they appear with the standard com-</p><p>plementizer that. Thus, even though negative preposing can surface under that, as we</p><p>saw in (6), it is ungrammatical under the comparative complementizer than in (8).5</p><p>(8) *Prices got higher than never before had wages been.</p><p>Finally, from a general point of view, the literature on CP-recursion contains a fair</p><p>amount of diverging suggestions as to the exact restrictions responsible for its licensing.</p><p>However, one common feature is that the complementizers in question are explicitly</p><p>selected by verbs (cf. Iatridou & Kroch 1992). This criterion is not fulfilled in the case</p><p>of than either (selected by -er/more). We hence preliminarily conclude that recursion</p><p>is unavailable in comparatives. Next, we turn to the diachronic aspects of CI.</p><p>.  Diachronic continuity in the syntax of CI</p><p>In order to show that CI was T-based at earlier stages, we can draw on independent</p><p>facts about the syntax of OE. First, OE had the canonical V2 order</p><p>negative clause (hence the preceding clause</p><p>must be negative); finally, nor itself becomes iNeg and cannot co-occur with other nega-</p><p>tive elements (loss of negative concord).</p><p>Masataka Ishikawa’s chapter, A feature-driven analysis of syntactic change: A case</p><p>study in the history of articles in Spanish, offers a minimalist analysis of the rise of the</p><p>definite article from Old to Modern Spanish, arguing for a two-stage development of</p><p>the structure and phasehood of the Spanish DP through a combination of grammar-</p><p>internal and grammar-external factors. The first stage consists of the creation of a</p><p>‘double d-shell’ structure, which, while computationally more complex, is motivated by</p><p>the (diachronic) pressure for a one-to-one assignment of formal features to functional</p><p>heads. By contrast, in the second stage, in which the inner dP acquires phasehood,</p><p>Ishikawa suggests ease of processing in acquisition as a possible motivation.</p><p>Elliott Lash, looking at The rise of OV word order in Irish verbal-noun clauses,</p><p>proposes a diachronic account of the emergence of the exceptional object–verb (OV)</p><p>Introduction 7</p><p>word order found in Irish nonfinite clauses, an otherwise regularly VSO language.</p><p>He posits a series of reanalyses involving the preposition do ‘to, for’ plus a nonfinite</p><p>verb, with noun phrases originally in a main clause being reanalysed as subjects or</p><p>objects of an embedded clause. He associates the final emergence of OV clauses with</p><p>a change in information stucture: the preverbal noun phrase in the order O do V was</p><p>once fronted for reasons of focus, but that motivation was lost with reanalysis and the</p><p>fronting became purely syntactic, as it is today. The preposition do in these structures</p><p>became first a focus-marking particle, and then a nonfinite marker, with various dia-</p><p>lectal variations.</p><p>In his chapter The great siSwati locative shift, Lutz Marten presents a historical-</p><p>comparative analysis of the locative morphosyntax of the southern Bantu language</p><p>siSwati. He argues that the Proto-Bantu locative noun class prefix ku- has degrammati-</p><p>calized to become a preposition in siSwati. This position is supported with a range of</p><p>syntactic evidence, such as the fact that noun-class agreement with a ku-marked noun</p><p>is with the class of the noun itself, and demonstratives can intervene between ku- and a</p><p>noun, both of which are unexpected if ku- is a noun class marker rather than a preposi-</p><p>tion. Thus we have a rather robust case of degrammaticalization, and the fact that this</p><p>change is shown to be natural within the context of an overall reanalysis of the siSwati</p><p>locative system arguably reinforces its status as a counterexample to the claimed unidi-</p><p>rectionality of grammaticalization.</p><p>In The impact of failed changes, Gertjan Postma establishes a mathematical relation</p><p>between successful and failing linguistic changes. Based on two case studies, he argues</p><p>that the latter can be understood as extra-grammatical variants introduced by L2 learn-</p><p>ers of a language which are not retained in subsequent generations, but which trigger</p><p>successful changes that are. He argues that, under the logistic model commonly used to</p><p>model successful changes which describe an S-curve, such unsuccessful changes can be</p><p>shown to be mathematically correlated, as they can be modelled as the first derivative of</p><p>the logistic function.</p><p>Henrik Rosenkvist considers A case of degrammaticalization in northern Swed-</p><p>ish dialects. These dialects have a modal verb bö ‘need, must, ought to’ which Rosen-</p><p>kvist argues has developed from the prefix of the verb behöva ‘need’. It is linked to</p><p>the fact that these dialects also allow the prefix be- (bö- in these dialects) of the verb</p><p>behöva ‘need’ to detach itself from the verbal root to give discontinuous word orders</p><p>be X höva. He regards this as a case of the reanalysis of a prefix as an auxiliary which</p><p>involves several changes in status (prefix > full word and prefix > auxiliary) that run</p><p>counter to the developments expected in grammaticalization. As such it would fit into</p><p>an existing known category of degrammaticalizations of affixes as independent words</p><p>(cf. Irish agreement suffix -muid > pronoun ‘we’, Doyle 2002).</p><p>Renata Szczepaniak looks at Jespersen’s Cycle in German from the phonologi-</p><p>cal perspective of syllable and word languages. She argues that the replacement of the</p><p>8 Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>old preverbal negator ne/en by a new postverbal one (niht) during the Middle High</p><p>German period was triggered, or at least encouraged, by the simultaneous prosodic</p><p>restructuring of the language from a syllable language to a word language, resulting in</p><p>the loss of unstressed pretonic syllables such as the old preverbal negator ne/en.</p><p>In An article on the rise: Contact-induced change and the rise and fall of N-to-D</p><p>movement, Mila Vulchanova and Valentin Vulchanov examine the evolution of the</p><p>postposed definite article in Bulgarian, a well-known feature of the Balkan linguis-</p><p>tic area. They argue that, at one time, definiteness was marked either by head move-</p><p>ment of N to D or by insertion of a determiner clitic in D. Historically, the determiner</p><p>changed from a phrase (a specifier of DP) to a head merged directly in D, a shift in</p><p>the expression of a functional category from Move > Merge in the sense of Roberts &</p><p>Roussou (2003). They suggest that, while the article evolves syntactically using pre-</p><p>existing material, the development is assisted by contact with Greek, which had a long-</p><p>established article.</p><p>4. Conclusion</p><p>One of the central tasks of historical linguistics has generally been seen as solving the</p><p>actuation problem: ‘Why do changes in a structural feature take place in a particular</p><p>language at a given time, but not in other languages with the same feature, or in the</p><p>same language at other times?’ (Weinreich, Labov & Herzog 1968: 102). We hope that</p><p>the individual contributions to this volume will help to elucidate both parts of this</p><p>problem. They address the first part by investigating various aspects of why grammars</p><p>change: such familiar issues as what processes need to be identified as the drivers of</p><p>change, the role of children in change, and the role of syntax-external factors in trig-</p><p>gering syntactic change (both phonological or lexical, and in the form of language con-</p><p>tact). We also hope that they offer in part some new perspectives on the second, often</p><p>neglected aspect of the actuation problem: why changes do not occur in other parallel</p><p>languages. Progress here necessarily focuses more on the constraints on change, on</p><p>factors which limit change, and on the role of contact in supporting conservative lin-</p><p>guistic features.</p><p>Acknowledgements</p><p>Our thanks go the various reviewers of the volume chapters, as well as to the series</p><p>editor Werner Abraham and to Kees Vaes at Benjamins. We are indebted to various</p><p>people and organizations for help with the Conference on Continuity and Change</p><p>Introduction 9</p><p>in Grammar: the scientific committee, Ian Roberts, Wim van der Wurff and Hedde</p><p>Zeijlstra; the various other people who helped with the day-to-day organization of the</p><p>conference: Deborah Anderson, Alastair Appleton, Yi-An Lin and George Walkden;</p><p>to Andrew Winnard at Cambridge University Press, and to the Faculty of English</p><p>and Selwyn College for providing a venue. The conference formed an integral part of</p><p>the research project The Development of Negation in the Languages of Europe, which</p><p>was possible due to the generosity of the Arts and Humanities Research Council</p><p>(Research grant award no. AR119272), who provided funding to bring the invited</p><p>speakers to Cambridge.</p><p>References</p><p>Andersen, Henning. 1973. Abductive and deductive change. Language 49: 765–93.</p><p>Bailey, Charles-James. 1973. Variation and Linguistic Theory. Washington DC: Center for Applied</p><p>Linguistics.</p><p>Biberauer, Theresa, Holmberg, Anders & Roberts, Ian. 2007. Disharmonic word-order systems</p><p>and the Final-over-Final-Constraint</p><p>of Germanic (pro-</p><p>duced through movement to C, e.g. in matrix questions). Second, however, it had a</p><p>mechanism that only displaced the finite verb to T (e.g. with topicalization; cf. Fischer</p><p>et al. 2000; Kroch, Taylor & Ringe 2000; Pintzuk 1991).6 Importantly, the latter strategy</p><p>would produce V3 orders, namely with subject pronouns between topicalized phrases</p><p>.  It is possible to rule out such structures by using the semantics of comparison (von Stechow</p><p>1984). Playing devil’s advocate it could be counter-argued that recursion is still possible, then,</p><p>but that it is ruled out for independent reasons. But a grammar in which such limited options</p><p>of recursion are additionally restricted by important factors including semantic ones only has</p><p>vanishingly small possibilities of keeping the hypothetical option as viable in comparatives.</p><p>Chris Lucas (p.c.) notes that a non-inverted version of (8) is less ungrammatical, but it is</p><p>extremely hard to parse or to assign meaning to. This observation is important since it is con-</p><p>sistent both with the semantic background and the simple present argument. Modern nega-</p><p>tive inversion gains ground in the seventeenth century (Nevalainen 2006) and it is unlikely a</p><p>factor of historical continuity.</p><p>.  Both the VP and the TP in OE could be head-final (as well as head-initial; cf. Pintzuk</p><p>1991). We discuss the significance of TP headedness in some detail from the perspective of</p><p>CI in Section 4 below.</p><p>Towards notions of comparative continuity in English and French 1</p><p>and verbs. The asymmetry is sketched in (9) vs. (10) with a wh-question and topicaliza-</p><p>tion, respectively (cf. Fischer et al. 2000; Kroch et al. 2000; the OE material in (9) and</p><p>(10) is taken here from Fischer et al. 2000: 118).</p><p>(9) For hwam noldest þu ðe sylfe me gecyðan þæt…</p><p>for what not.wanted you you self me make.known that</p><p>‘Wherefore would you not want to make known to me yourself that…’</p><p>(10) Forðon we sceolan mid eally mod & mægene to Gode gecyrran</p><p>therefore we must with all mind and power to God turn</p><p>‘Therefore we must turn to God with all our mind and power.’</p><p>The test gives the possibility of a clear prediction for OE, which could use both mecha-</p><p>nisms in general, depending on the context (e.g. questions vs. topicalization as above).</p><p>The main diagnostic to test the two types of apparent V2 effects are pronouns, which</p><p>are taken to reside in the specifier of the highest functional head in the Infl-domain,</p><p>here for concreteness: Spec,TP. If the finite verb surfaces to the left of pronouns, then</p><p>it will be under C. (In which case the canonical type of Germanic V2 analysis would</p><p>be confirmed). If the finite element is invariably on the right of the subject pronoun in</p><p>OE, it must be under T. Applying the test to comparatives in OE shows that they show</p><p>no inversion with pronouns, but appear only as shown in (11).</p><p>(11) for.ðan.þe se Fæder is mare [þonne ic sy].</p><p>because the father is greater than I am</p><p>‘For the Father is greater than I.’ (YCOE, coaelhom, ÆHom 10:19.1417)</p><p>This means that the finite verb in OE comparatives does not go beyond the TP (or the</p><p>Infl-domain, more generally).7</p><p>Another diachronically rooted argument is that if a representation that does not</p><p>require an overt subject in Spec, TP is on the right track, then we expect null expletives</p><p>in comparative clauses. An (EModE) example of this sort is given in (12).</p><p>(12) [H]e left soch a companie of fellowes and scholers in S. Iohnes Colledge,</p><p>[as _ can scarse be found now in some whole vniuersitie].</p><p>(PPCEME-ASCH-E1-H,55R.164)</p><p>There are some contexts in current English that naturally have null (and, as Chris</p><p>Lucas, p.c., points out, in standard British English only null) subjects in comparatives;</p><p>see the passives in (13a,b) and the impersonal in (13c).</p><p>.  A point of change is that the diagnostic does not have the same force in Middle English</p><p>(ME), where pronoun subjects start inverting (even though even here examples in compara-</p><p>tives are rare; cf. also the ModE situation, where contrast is needed for the speakers who</p><p>accept CI with pronouns.)</p><p>1 Remus Gergel</p><p>(13) a. [T]he tropics play a more active role [than _ was thought] in controlling the</p><p>Earth’s climate.</p><p>(www.sciencedaily.com/releases/2005/10/051012084249.htm)</p><p>b. Davis Cup exit poses more questions [than _ could be answered in a year].</p><p>(www.timesonline.co.uk/tol/sport/tennis/article4800983.ece)</p><p>c. [P]ast experiences need to be reviewed in a rather more circumspect manner</p><p>[than _ has been the tendency to date].</p><p>(www.bilaterals.org/article.php3?id_article=11194)</p><p>The motivation for this move lies in the historical links between null expletives and</p><p>V2. For correlations between the loss of the English type of V2 and null expletives,</p><p>we draw on Haeberli (2002) and transfer the basic logic to comparative clauses. (See</p><p>also van Kemenade 1997 and especially Hulk & van Kemenade 1995: 246 with partly</p><p>equivalent implementations, formulated in terms of expletive pro-drop.) We can thus</p><p>suggest the following. The overt subject in CI can stay low because, structurally speak-</p><p>ing, the higher subject position does not need to be overtly filled (or equivalently: it is</p><p>filled by a null-expletive; this technicality avoids conflicts with the EPP, i.e. the prin-</p><p>ciple that the clause must have a subject in Spec,TP)</p><p>A further argument can be culled from the continuous presence of CI in the records.</p><p>While CI has never seemed to be a high-frequency phenomenon, it has been attested</p><p>throughout the history of English; (14) comes from the early OE text of Beowulf.</p><p>(14) Næfre ic maran geseah eorla ofer eorþan ðonne is eower sum…</p><p>never I greater seen of warriors on earth than is of.you one</p><p>‘I have never seen a greater warrior on earth than is one of you.’</p><p>(Beowulf, III.247)</p><p>To summarize, we have so far argued that CI involves a non-standard derivation with</p><p>low subjects and that the phenomenon has been historically persistent in the history of</p><p>English. In the next section, we will explore a similar phenomenon in French in which</p><p>comparatives are the locus of conservativism. We will thus extend both the observa-</p><p>tion about continuity and the analysis proposed.</p><p>.  On the syntax of comparative inversion in French</p><p>In this section we first clarify that French has inversion in comparatives too. We subse-</p><p>quently argue that French CI also involves low subjects and show that the phenomenon</p><p>is also historically persistent from some of the earliest records onwards.</p><p>.1  CI in Modern French</p><p>Modern French allows subject–auxiliary inversion in comparative clauses:</p><p>Towards notions of comparative continuity in English and French 1</p><p>(15) a. Pierre est plus grand que n’ était son père.</p><p>Pierre is more tall than en-was his father</p><p>‘Pierre is taller than his father was.’</p><p>b. [L]es délibérations […] sont ouvertes au public dans un degré</p><p>the deliberations are open to.the public to a degree</p><p>plus large que ne le sont les activités d’autres institutions…</p><p>more large than en it are the activities of-other institutions.</p><p>‘The deliberations are more open to the public than the activities</p><p>of other institutions.’</p><p>(http://www.infocom.gc.ca/reports/section_display-f.asp?intSectionId=301)</p><p>The French examples additionally feature an expletive negative (EN) element, the</p><p>larger consensus on which is that it does not have negative meaning (cf. Price 1990;</p><p>von Stechow 1984). EN is not a characteristic of inversion but of comparatives; more</p><p>specifically, of clausal comparatives (of inequality). Non-inverted clauses display it too</p><p>and equatives need not contain the EN, whether inverted or not; see (16) and (17).8</p><p>(16) [L]a télévision est plus différente du cinéma</p><p>the television is more different from.the cinema</p><p>que le cinéma ne l’est de la photographie.</p><p>than the cinema en it.is from the photography</p><p>‘The difference between television and cinema is greater than the difference</p><p>between cinema and photography.’</p><p>(http://unesdoc.unesco.org/images/0014/001437/143768fb.pdf)</p><p>(17) a. Marie est aussi grande que l’ était sa mère.</p><p>M. is as tall as it-was her mother</p><p>‘Marie is as tall as her mother (is).’</p><p>b. … que sa mère l’ était.</p><p>c. … que sa mère ne l’ était.</p><p>d. … que ne l’ était sa mère.</p><p>By testing sequences of verbal heads in French, we can observe that the subject also</p><p>follows the entire sequence rather than just the structurally highest finite member</p><p>.  The negative expletive intruding into the equative is a later development of modern</p><p>French. OF had the negative morpheme in comparatives of inequality but not in equatives</p><p>(cf. also Section 5 below).</p><p>1 Remus Gergel</p><p>of the sequence. Crucially, then, this option is possible (and virtually obligatory) in</p><p>inverted comparatives in French; see (18).9</p><p>(18) a. Elle est plus jolie [que n’a été sa mère].</p><p>she is more beautiful than en-has been her mother</p><p>‘She’s more beautiful than her mother was.’ (FM, p. 191)</p><p>b. Ceux-ci ne seront pas touchés par la nouvelle carte militaire,</p><p>those not will.be not touched by the new law military</p><p>pas plus que ne devrait l’être la base aéronavale de L.-B.].</p><p>not more than en should it-be the base aero-naval of L.-B.</p><p>‘They will not be more affected by the new military law, not more than the</p><p>aero-naval base Lann-Bihoué should be.’</p><p>(www.ouest-france.fr/actu/actu_BN_-Carte-militaire-le-couperet-tombe-</p><p>aujourd-hui_8619-674581_actu.htm)</p><p>c. [U]ne langue qui change plus vite que ne sont rédigés les</p><p>a language that changes more fast than en are redirected the</p><p>articles du future dictionnaire.</p><p>articles of.the future dictionary</p><p>‘A language that changes faster than the articles of the future dictionary are</p><p>updated.’</p><p>(www.chass.utoronto.ca/~wulfric/siehlda/actesmen/ilt1_2.htm)</p><p>d. [U]ne famille est pauvre si l’ essentiel, comme la nourriture,</p><p>a family is poor if the-essential like the nourishment</p><p>le logement et l’habillement, absorbe un pourcentage</p><p>the lodging and the-clothing absorbs a percentage</p><p>plus élevé de son revenu</p><p>more high of its income</p><p>.  The participial and the perfect auxiliary appear separated on the surface in the single</p><p>(verse-based) example I was able to find in (i). John Vanderlest (p.c.) points out that any</p><p>examples of this sort (i.e. separated) are marked at the very least in current French and I have</p><p>no quantitative figures for early French. Thus while such examples merit further investigation,</p><p>notice, again, that it is the existence of the clusters that argues against a standard raising to C.</p><p>(Conversely, it is technically possible to have the subject still lower than the specifier of the</p><p>highest inflectional projection in such examples.)</p><p>(i) Et plus vermeille que n’est rose arousee.</p><p>and more red than en.is rose made.rosy</p><p>‘And (she was) redder than a rose.’</p><p>(CLM corpus: Adenet le Roi, 8153, Enfances Ogier – Œuvres d’Adenet le Roi, (1269–1285))</p><p>Towards notions of comparative continuity in English and French 1</p><p>[que ne devrait dépenser la famille canadienne moyenne].</p><p>than en should spend the family Canadian average</p><p>‘A family is poor if the essentials such as food, lodging and clothing</p><p>absorb a higher percentage of its income than the average Canadian family</p><p>should spend.’</p><p>(www.unicef.org/french/sowc06/pdfs/repcard6f.pdf)</p><p>The existence of CI is then reminiscent to some degree of the situation in English. But</p><p>the question arises whether more appropriate tests for the syntax of French can be</p><p>applied. Let me therefore next introduce some background to the additional intrica-</p><p>cies of subject–verb inversion in French. I capitalize here on ideas pertaining to inver-</p><p>sion that have been suggested in Lahousse (2006) and Rowlett (2007), independently</p><p>of comparatives, who consider a range of inversion possibilities in the language and</p><p>classify them syntactically. We will test French CI on the basis of such tests and argue</p><p>that it offers evidence that the subject is in a structurally low position. More specifi-</p><p>cally, French CI parallels to a large extent one particular type of stylistic inversion</p><p>of French.10</p><p>A first distinction needs to be drawn between the so-called ‘pronominal’ inversion</p><p>of the standard variety, as well as the related ‘complex’ inversion, which are illustrated</p><p>in the two examples in (19), on the one hand, and so-called ‘stylistic’ inversion on the</p><p>other (stylistic inversion will be further subdivided momentarily).</p><p>(19) a. Quand est-elle partie? b. Quand Claire est-elle partie?</p><p>when is-she left When Claire is-she left</p><p>‘When did she leave?’ ‘When did Claire leave?’</p><p>By definition, pronominal inversion, such as the type witnessed in (19a), involves a</p><p>pronoun subject, as does in fact the type of complex inversion seen in (19b) in addi-</p><p>tion to the full-DP subject it displays. Stylistic inversion, which will be of interest for</p><p>current purposes, does not share the restriction to subject pronoun clitics of complex</p><p>inversions in general (whether with or without accompanying full DPs) and shows</p><p>distinct properties throughout.11 Since both types in (19) are clearly distinct from (the</p><p>1.  The terms Rowlett uses are ‘genuine SI’ for the low-subject inversion and ‘focus SI’ for the</p><p>remnant-movement inversion, where focus is taken as wide-focus. I avoid these terms, as CI,</p><p>though syntactically more on a par with genuine SI may not be independent of information-</p><p>structure (cf., e.g. Culicover & Winkler 2008; Gergel et al. 2007; Winkler 2005).</p><p>11.  The CI of French is unacceptable with pronouns in current French and I could not</p><p>find appropriate examples from earlier stages. The restriction cannot be blamed entirely on</p><p>the clitic status of the subject pronouns as the tonic versions of the pronoun paradigm are</p><p>degraded too.</p><p>1 Remus Gergel</p><p>various types of) inversion that include CI, we next focus on stylistic inversion. (See</p><p>e.g. Kayne & Pollock 2008, Rizzi & Roberts 1989; Rowlett 2007 for discussions of</p><p>complex inversion.) More specifically, we will next inspect a two-fold subdivision</p><p>within the syntactic mechanisms in the class of stylistic inversion.</p><p>According to Rowlett (2007: 217–223) (cf. also especially Lahousse 2006), there</p><p>are two major types of stylistic inversion in French. The terms used in my two main</p><p>sources are not fully identical (much less in the rest of the literature), but importantly</p><p>the mechanisms are, and they can be summarized as follows. Crucially, one type of</p><p>inversion is, simply put, a low-subject type of inversion. (The other type of inversion</p><p>finds a remnant-movement-type implementation in Rowlett’s analysis. Since CI will</p><p>turn out to be closely related to the first type, I leave the more technical details in the</p><p>implementation of the type based on remnant evacuation aside.)</p><p>A first characteristic of the low-subject inversion observed is that adverbs can be</p><p>found intercalated between the finite verb and the subject:</p><p>(20) Ici se cachent souvent des lapins.</p><p>here self hide often some rabbits</p><p>‘Rabbits often hide here.’</p><p>Second, verbal dependants can follow a low subject (cf. (21)). But, third, the subject</p><p>must be final, if the mechanism that brings about inversion is remnant movement (the</p><p>latter case is illustrated with a stage topic in (22) below; cf. Rowlett 2007).</p><p>(21) Quand passeront Jean et Pierre sous la fenêtre?</p><p>when will.walk J. and P. under the window</p><p>‘When will Jean and Pierre walk under the window?’</p><p>(22) a. Seuls passeront sous la fenêtre Jean et Pierre.</p><p>alone will.walk under the window J. and P.</p><p>‘Jean and Pierre alone will walk under the window.’</p><p>b. *Seuls passeront Jean et Pierre sous la fenêtre.</p><p>Fourth, quantifier float is impossible in stylistic inversion, as the contrast below (follow-</p><p>ing Rowlett) between low-subject and remnant-movement inversion shows:</p><p>(23) *Quand ont tous fini les enfants?</p><p>when have all finished the children</p><p>‘When did all the children finish?’</p><p>(24) Ont tous</p><p>réussi ceux qui ont assez travaillé.</p><p>have all succeeded those who have enough worked</p><p>‘Those who did enough work were all successful.’</p><p>The reason invoked by Lahousse (2006) for effects such as (23–24) is that the floating</p><p>quantifier needs to be in a binding relationship with the trace of a raised subject. (This</p><p>argument in turn is based on Doetjes’ 1992 analysis of quantifier float.)</p><p>Towards notions of comparative continuity in English and French 11</p><p>The immediate question is whether, within the class of stylistic inversions, CI pat-</p><p>terns with low-subject inversion or with remnant-movement inversion in the grammar</p><p>of French. We will argue that the former analysis is more appropriate for CI.</p><p>First, we find that adverbs such as souvent, ‘often’, can be intercalated between the</p><p>finite verb and the subject, as in (25).</p><p>(25) La vraie vie est plus théâtrale [que ne l’est souvent le théâtre].</p><p>the true life is more theatrical than en it-is often the theatre</p><p>‘Real life is more theatrical than theatre often is.’</p><p>The standard logic of the argument is as follows: since the adverb delimits the edge of</p><p>the VP (assume simply via adjunction), the subject can be observed to be within the</p><p>VP. This offers, then, some indication for the low subject position.12</p><p>Second, various dependants can be found in linearly post-subject positions on the</p><p>surface of CI structures. This confirms in particular the second test discussed above for</p><p>stylistic inversion more generally. (In simplified terms: a remnant movement analysis,</p><p>as opposed to the low-subject inversion, would require subjects to be final.)</p><p>(26) a. Je me défonce plus pour mon chef que ne le fait Martine</p><p>I self hit more for my boss than en it does Martine</p><p>pour le sien. (John Vanderelst, p.c.)</p><p>for the hers</p><p>‘I get more involved for my boss than does Martine for hers.’</p><p>b. [O]n fait plus que ne légitime la déduction jusqu’ici.</p><p>one does more than en justifies the deduction up.to.here</p><p>‘One does more than deduction justifies up to here.’</p><p>(D. Lories in Kant’s Ästhetik/L’ esthétique de Kant,</p><p>Berlin: de Gruyter 1997, p. 575)</p><p>1.  The adverb test is one standard heuristic (Pollock 1989). Rowlett goes further for the</p><p>inversion types he analyses and assumes that the EPP feature of T is never satisfied. We may</p><p>disagree with this for CI. CI derivations have non-inverted counterparts. A second note is also</p><p>in order regarding adverbs. Other adverbs are also quite possible, as illustrated in (i) below.</p><p>But, at least on certain syntactic accounts in the wake of Cinque (1999), they may be situated</p><p>higher up and thus do not necessarily delimit the low boundary. Such adverbs would, then, be</p><p>less clear as evidence for low subjects. Finally, their precise syntax is at the same time anything</p><p>but clear (the invoked high position may, e.g. be a matter of LF).</p><p>(i) [L]e même Jacques Peyrat avait fait 2% de plus</p><p>the same Jacques Peyrat had made 2% of more</p><p>[que ne fait aujourd’hui Christian Estrosi].</p><p>than en makes today Christian Estrosi</p><p>‘The same Jacques Peyrat has made 2% more than C.E. makes today.’</p><p>(http://boymottard.wordpress.com/page/19/)</p><p>1 Remus Gergel</p><p>Third, quantifier float of the type discussed becomes entirely incomprehensible in CI</p><p>(as expected if CI patterns with low-subject inversion):</p><p>(27) *Pierre a fait ses devoirs plus vite [que tous ont fini</p><p>Pierre has done his homeworks more fast than all have finished</p><p>les enfants].</p><p>the children</p><p>Intended: ‘P. did his homework faster than all the children have finished theirs.’</p><p>Fourth, we can note that subject-less impersonals are in some cases (we assume</p><p>restrictedly) possible (i.e., e.g., without the pronoun il):</p><p>(28) un système qui a progressé vers un état d’ équilibre</p><p>a system that has progressed toward a state of.balance</p><p>durant une période de temps beaucoup plus longue</p><p>during a period of time much more long</p><p>que ne serait possible en système synthétique.</p><p>than en would.be possible in system synthetic</p><p>‘a system that equilibrated for a much longer period of time than would be</p><p>possible in synthetic systems.’</p><p>(http://pubs.nrc-cnrc.gc.ca/mineral/tcm-88939-3.html)</p><p>(29) Il sera plus difficile d’ accroître ce pourcentage</p><p>it will.be more difficult to.increase this percentage</p><p>que n’ est le cas pour Genève…</p><p>than en.is the case for Geneva</p><p>‘It will be more difficult to increase this percentage than is the case for Geneva.’</p><p>(http://www.litra.ch/Trains_trams_et_bus_offrent-ils_une_capacite_suffisante.html)</p><p>Overall then, these tests show that CI patterns with what Rowlett calls genuine stylistic</p><p>inversion, that is, the type specifically based on low subjects. In the next section, we</p><p>will focus on the main characteristics of CI in earlier French.</p><p>.  CI in earlier French</p><p>From a diachronic vantage point, we can make two crucial supporting observations</p><p>for the current analysis of CI right away. First, earlier stages of French had the possibil-</p><p>ity of a subject appearing in a low position (see especially Vance 1997 and Kroch 2008</p><p>for recent discussions and the references cited there). Second, we can empirically add</p><p>that OF French had CI as well. This is shown in (30), with a more comparative, and in</p><p>(31) with a comparative introduced by the adverb isnels, ‘nimbly, agilely’.</p><p>Towards notions of comparative continuity in English and French 1</p><p>(30) Plus curt a piet que ne fait un cheval.</p><p>more runs by foot than en does a horse</p><p>‘He runs more than a horse does.’ (CdR, 71.890)</p><p>(31) Plus est isnels que nen est uns falcuns.</p><p>more is agile than en is a falcon</p><p>‘He is more agile than a falcon is.’ (CdR, 118, 1572)</p><p>Third, examples from Middle and Classical French support the idea of a continuous</p><p>phenomenon; compare the inversions in the comparatives in (32) and (33), respectively.</p><p>(32) … plus rouge que n’est feu gregois.</p><p>more red than en-is fire Greek</p><p>‘…redder than is Greek fire.’ (CLM corpus; Mystère de la Passion (ca. 1450))</p><p>(33) J’ay plus failly que ne peut un mortel.</p><p>I-have more erred than en can a mortal</p><p>‘I erred more than a mortal can.’</p><p>(Frantext: Q757/VIAU Th. de/OEuvres poétiques: 1/1621, p. 113)</p><p>A fourth supporting argument is that subjectless impersonals are a time-honoured</p><p>presence in the language, including some comparatives, as the OF (34) illustrates.</p><p>(34) Il le dublat plus que ne solt.</p><p>he it doubled more than en uses</p><p>‘He doubled it more than is usual.’ (BR, 1. 765)</p><p>Fifth, verb sequences containing more than just the finite verb can occasionally be</p><p>observed in the records of earlier French as well. We note that they can display the</p><p>same basic surface word order as do the modern examples. I assume that the subject is</p><p>in the specifier position of the lowest (and theta-assigning) predicate phrase.</p><p>(35) Guillaumes fu joianz et liez,[…]</p><p>Guillaumes was happy and content</p><p>Plus que ne puet dire nus hom.</p><p>more than en can tell us man</p><p>‘Guillaumes was happier and more content than anyone could tell.’</p><p>(CLM corpus; De Guillaume au faucon)</p><p>Finally, the test based on verbal dependents after the subject (cf. 3.1 above) can also be</p><p>applied to some extent to the diachronic record, see (36). The result of the test suggests</p><p>once more continuity, namely in the availability of the low-subject position.</p><p>1 Remus Gergel</p><p>(36) Li palefrois sor quoi ele est montee</p><p>the palfrey on which she is climbed</p><p>Estoit plus blans que n’ est nois sor gelee;</p><p>is more white than en-is snow on ice</p><p>‘The palfrey on which she climbed is whiter than snow on ice.’</p><p>(CLM corpus; 8145, Enfances Ogier – Œuvres d’Adenet le Roi)</p><p>Overall there are reasonable grounds for a low-subject derivation of CI. While I have no</p><p>quantitative results on the developments in French, it is possible that with the advent</p><p>of annotated corpora such as the upcoming Les voies du français (see Kroch 2008 for</p><p>discussion), more will be learnt about the trajectory of the construction including</p><p>numerical developments</p><p>and what they may reveal about grammar.</p><p>.  Continuity in the syntax of CI</p><p>We have so far illustrated the continuous presence of CI in the historical record. This</p><p>section is concerned with a selection of facts which potentially complicate the picture.</p><p>We concentrate on English, due to the data situation at present (in particular the wider</p><p>availability of parsed historical corpora), but also raise certain questions in connection</p><p>with French.</p><p>On the basis of the Penn-Helsinki-York corpora of historical English, estimates as</p><p>in (37) can be given for the ratio of inverted tokens within structures that qualified as</p><p>clausal comparatives from the corpus annotation (cf. Gergel 2008).</p><p>(37) OE: 223/5114 = 4.36%; ME: 135/1639 = 8.23%; EModE: 31/2497 = 1.24%</p><p>The estimates almost double in ME (compared to OE) and then drop dramatically</p><p>towards EModE. The fall in EModE is straightforwardly explained by the erosion of</p><p>the low subject position (which remains an archaic possibility in comparatives, as we</p><p>have seen). However, the surge in ME needs an explanation. There is more than one</p><p>potential source why the incidence of CI might have increased in ME regarding poten-</p><p>tial intervening factors. The main possibilities to be discussed are listed in (38).</p><p>(38) a. contact with Northern syntactic features in ME</p><p>b. the influence of French</p><p>c. independent factors relating to change/continuity in the diachrony of CI</p><p>We first investigate whether the incidence of CI in the records was influenced by</p><p>factors that have been claimed to have a major influence on the loss of V2 in English.</p><p>One such factor is given by Northern features possibly stemming from contact with</p><p>Germanic languages that had a more regular C-based V2 than OE (see Fischer</p><p>et al. 2000; van Kemenade 1997; Kroch et al. 2000; Roberts 2007, among others for</p><p>Towards notions of comparative continuity in English and French 1</p><p>discussions). But importantly, for our immediate purposes there are several points</p><p>on which the development of CI does not go together with that of the progress of</p><p>such Northern dialectal features and, as we will argue, cannot be explained on the</p><p>basis of this type of contact. To begin, subject pronouns do start to invert during ME</p><p>in contexts in which they did not invert in OE. But this increase cannot be plausibly</p><p>explained by a putative increase in inverting pronouns; that is, neither in terms of its</p><p>timing nor the quantity of change. For one, other major changes in the patterns of</p><p>inversion between finite verb and subject become visible in the estimates from the</p><p>period M2 onwards. But the current increase is from M1, the first Helsinki subdivision</p><p>of ME. In addition, even if we abstract away from timing, the number of actual inverting</p><p>pronouns in ME still remains particularly limited and cannot explain the surge in the</p><p>CI estimates.</p><p>Furthermore, we have another related negative result from an independent diag-</p><p>nostic, namely from a consideration of the incidence of CI in texts of Northern prove-</p><p>nience in the PPCME2 corpus in particular. Out of the 6 Northern texts in the corpus</p><p>just one contains CI. The thoroughly Northern-based Rule of St. Benet, for example,</p><p>does not contain inversions in comparatives. The Northern version of the Mirror of St.</p><p>Edmund is the sole text from this sample to contain (two) tokens of CI. The Northern-</p><p>based ratio of texts thus stands in a stark contrast with the overall ratio of texts con-</p><p>taining CI constructions in ME in general and in particular with the texts that are not</p><p>classified as being from the North. For example, the average within the remaining 50</p><p>texts in the PPCME2 is as follows: 35 texts have CI, that is, as much as 70% (compared</p><p>to the 16.6% above). Moreover, the overall total rate of CI per relevant tokens in the</p><p>Northern texts is, at 1.35%, far below the overall average of 8.23 % for ME. The situa-</p><p>tion, then, does not yield support for a rise of CI due to Northern features. (If anything,</p><p>a slowing down factor might be given). Clearly, this does not mean that interesting</p><p>comparisons cannot be drawn between features of, say, Old Norse (which arguably</p><p>influenced the Northern V2) and of ME including comparatives13 (or much less that</p><p>more intricate scenarios than the ones presented are not possible). It only means that</p><p>1.  The Example (i) from the Laxdœla Saga illustrates a structure resembling CI with a dual</p><p>pronoun in Old Norse (from Faarlund 2004: 267; (54c)):</p><p>(i) allt er mér sl’kt it sama nú í hug,</p><p>all.neu.n is me.d such.neu.n the same.def.neu.n now in mind.d</p><p>sem þá rœddum vit</p><p>as then talked.1p we.du</p><p>‘My thoughts are still the same regarding everything as we talked about then.’</p><p>[(Laxd 203.21)]</p><p>1 Remus Gergel</p><p>the Northern features cannot be blamed directly for the increase in CI that we witness</p><p>in ME. We next turn to the influence of French.</p><p>Parallels between the syntactic histories of English and French can be, and cer-</p><p>tainly have been, drawn (Adams 1987; Hulk & van Kemenade 1995; Kroch 2008;</p><p>Roberts 1993, among others). An additional possibility arises, however, when we con-</p><p>sider the potential of an indirect grammatical influence on the grammar reflected in</p><p>the ME texts. Considering such texts in respect to CI, there is an observable contrast to</p><p>the influence of Northern texts that we addressed above. The influence of French can-</p><p>not be refuted the way that of Scandinavian has been. First, we have already said that</p><p>this language had a productive process of CI at all times. Second, and perhaps more</p><p>importantly, the corpus texts that are translations from French show the construction</p><p>as well. We next focus on the texts of the PPCME2 that are translations from French</p><p>to check their incidence of CI. The summary in (39) below indicates the ratios of CI</p><p>per relevant clausal comparative tokens for the six texts that were French translations</p><p>(using the standard corpus tags for the texts):</p><p>(39) Aelr4 2/19; Ayenbi 10/85; Meli 2/64; Pars 6/67; Mandev 7/71; Vices 1/16.</p><p>The average frequency of CI in the six texts thus lies at 28/322 or 8.69%. This is notably</p><p>more than six times the average in the Northern texts.</p><p>A third asymmetry between the French-based and the Northern-based texts is</p><p>that each individual text that is a French translation has occurrences of CI structures.</p><p>(Compare the only 16.6%, or one in six of the texts, in the group of Northern texts).</p><p>But, alas, all of this does not show that there was a French influence on the develop-</p><p>ment of CI in English. All it does is show that it is not possible to dismiss (indirect)</p><p>influence on the language. There are in fact some arguments that make a positive argu-</p><p>ment for a French import implausible. First, the construction was already available in</p><p>the OE of Beowulf (cf. (14)), thus preceding French influence. Second, even though</p><p>the frequency of CI in the French-based texts is much higher compared to the Northern</p><p>texts, it is still around the ME average calculated over all texts given above. That means</p><p>that a possible reinforcement through French might have taken place at most. Third,</p><p>as was the case with the potential Northern influence, the surge in the estimates occurs</p><p>too early, namely right from the very first period of ME. Fourth, even in a text such</p><p>as the Ayenbite of Inwyt dated from 1340, which is a fairly literal translation from the</p><p>French La somme le roi, we do not get a one-to-one mapping import of French in the</p><p>translation. To illustrate this point, (40) preserves the structure from the French origi-</p><p>nal, as expected, but (41) does not. In (41), the English (close) translation has inversion</p><p>without inversion in the French original.</p><p>(40) a. Middle French:</p><p>Il i a son droit et sa soustenance, et tout comme il en</p><p>he there has his right and his sustainment and all as he it of.it</p><p>Towards notions of comparative continuity in English and French 1</p><p>veut avoir, plus soufisemment [que n’a li rois].</p><p>wants have more sufficiently</p><p>than en.has the king</p><p>‘He has his right and support and everything as he wants it, in more</p><p>abundance than the king does.’ (LsR, p. 189, Ch. 47, §123)</p><p>b. Middle English:</p><p>Þer is his wone and his sustinonce and al þet he wyle</p><p>there is his habit and his sustainment and all that he wants</p><p>habbe more ynogh</p><p>have more sufficiently</p><p>[þanne habbe þe kyng]. (CMAYENBI, 85.1650)</p><p>than has the king</p><p>(41) a. Middle French:</p><p>Vertu fet l’omme plus a.droit seigneur du monde</p><p>virtue makes the man more rightfully lord of.the world</p><p>[que li rois n’est de son roiaume].</p><p>than the king en-is of his kingdom</p><p>‘Virtue makes man more legitimately lord of the world than the king is of</p><p>his kingdom.’ (LsR, p. 189, Ch. 47-§120)</p><p>b. Middle English:</p><p>Virtue makeþ þane man more aright lhord of þe wordle:</p><p>virtue makes the man more rightly lord of the world</p><p>þanne by þe kyng of his regne. (CMAYENBI, 85.1648)</p><p>than is the king of his kingdom</p><p>The fact that even in a literal translation we can detect home-grown instances of English</p><p>inversion makes French, even as a source of re-import, then, unlikely. A more likely</p><p>possibility (consistent with our proposal) is that the construction was due to a struc-</p><p>tural possibility given in the grammar of both languages, and which found propitious</p><p>ground in the ME period due to a variety of factors in the output. But the question,</p><p>then, of course remains what might have chiefly caused the surge in ME.</p><p>An issue that has had a direct impact on the appearance of CI on the surface strings</p><p>of the texts is an independent syntactic development, namely the headedness of TP.</p><p>Since a head-final TP obscures CI under the low-subject proposal defended here, the</p><p>doubling of the estimates in ME as compared to OE can be explained as follows. First,</p><p>notice that at the descriptive level OE had both T-final and T-initial word orders. (Cf.</p><p>Fischer et al. 2000; van Kemenade 1997; Kroch et al. 2000, among others. For an inter-</p><p>esting theoretical account of the transition in headedness, see Biberauer, Sheehan &</p><p>Newton this volume). Second, we use the independent assumption that roughly half</p><p>of the relevant clauses of OE are head-final (cf. Pintzuk 1991 for comprehensive</p><p>discussion of such findings). Next, we can note that for a T-based inversion to be</p><p>1 Remus Gergel</p><p>noticeable in the first place, the TP must have been T-initial. (In a head-final TP, the</p><p>finite element may move to T, but still stays to the right of the subject. Hence no inver-</p><p>sion is detectable.) That means that half of the structures that had the mechanics for</p><p>CI in place in OE were obscured in the output. But by early ME T-final structures are</p><p>already dead or only archaic (Kroch et al. 2000). This, then, yields a possible explana-</p><p>tion for the doubling of the estimates in the CI output seen in the corpora.</p><p>To summarize, we have observed that a major point of change was the directionality</p><p>of the TP by the end of OE; pace, for example, possible textual influence from French.</p><p>A syntactic suggestion has been made as to why this change may have affected, as a</p><p>key factor, the ratio of CI in the output. Notably for the case in point, the low subject</p><p>became more prominent at the stage at which English had only head-initial TPs. This</p><p>strengthens the main proposal that the main effect is best explained by continuity in</p><p>the sense that the low subject position was heavily used in conjunction with T, both</p><p>before and after the change in directionality of headedness.</p><p>.  Open issues and conclusions</p><p>We next discuss remaining questions, grouped around two main issues, before con-</p><p>cluding. First, we address V-to-T. We then re-address continuity in comparatives.</p><p>As noted by a reviewer, a non-trivial factor in the evolution of any process that</p><p>makes reference to the inflectional domain is the trajectory of V-to-T. While it is</p><p>beyond the scope of this chapter to elucidate the notorious controversies surrounding</p><p>this issue in its general aspects (Kroch 1989; Pollock 1989; Roberts 1993; Warner 1997,</p><p>among others), we offer a note from the perspective of comparative structures. In this</p><p>connection, it is crucial to distinguish between two layers of potential interference.</p><p>First, notice that whether French or English has V-to-T does not affect the underlying</p><p>skeleton of the proposal. But putting the classical parametric difference (both between</p><p>English and French and between the developmental stages of English; cf. Kroch 1989;</p><p>Pollock 1989; Roberts 1993, among others) together with the current proposal makes</p><p>a valid prediction. Namely (just) that CI involves a finite element under T. Relating to</p><p>it, however, the proposal itself, as stated, does not (need to) make any reference as to</p><p>which elements are under T. The important second point is that the parametric differ-</p><p>ence entirely takes care of this aspect. The parametric difference, then, does not change</p><p>the phrase-structure of CI as such but it interacts with it in the output produced. There</p><p>is a clear generative difference dependent upon whether CI interacts with a grammar</p><p>with or without V-to-T. The predicted output difference is that, all things being equal,</p><p>the former grammar will have the means to feed lexical verbs to be (re-)merged under</p><p>the node T. (Simply put: lexical verbs will be able to participate in CI in conjunction</p><p>with the low subject). The latter grammar, however, will in general lack the means</p><p>to do so. This seems to be the case for Present Day English (PDE), which is forced</p><p>Towards notions of comparative continuity in English and French 1</p><p>to apply do-support with lexical verbs. CI in PDE thus only occurs with specialized</p><p>T-elements (modals/do) and have/be, as a reviewer observes. French, expectedly, also</p><p>allows lexical verbs in CI.14</p><p>We now turn to our second open question, namely why a construction such as CI</p><p>might have remained in place so persistently while (almost) everything else around it</p><p>changed. We tentatively suggest two types of reasons as main necessary conditions. The</p><p>first major reason we suggest (and hope to have convinced the reader of through the</p><p>evidence presented) is that CI has been naturally riding on a syntactic option (namely</p><p>low subjects) that was historically widely available in both English and French. Today,</p><p>this earlier phrase-structural option remains available in areas of the grammar in the</p><p>two languages which are typically stylistically marked. While the specific syntactic</p><p>reasons why such representations should be plausible have been explained in some</p><p>detail, there is a second type of motivation with the potential to affect comparative</p><p>clauses: the interface with the component of meaning. We just mention three points</p><p>in this connection.</p><p>First, one additional heuristic reason for stability may be the parallelism relation-</p><p>ship given through the very nature of comparative constructions generally (cf. Gergel</p><p>et al. 2007). A second, perhaps more crucial reason is that the core semantic represen-</p><p>tation of comparatives may not have changed significantly in either English or French.</p><p>Following tests discussed in synchronic terms in Beck et al. (to appear) for a series of</p><p>largely unrelated languages, it becomes possible to test for parameters at the syntax–</p><p>semantics interface in the history of English, too. In this connection, OE and PDE seem</p><p>to have the same semantics of the comparative in place.15 The final point that becomes</p><p>1.  As pointed out by the reviewer, there are several other (important) syntactic changes in</p><p>both languages. But they did not seem to affect comparatives. One reason why the changes in</p><p>the status of pronoun subjects in French might have not affected the output is that pronouns</p><p>are particularly hard to find even with non-clitic forms (both in the records and with the new</p><p>full forms in current French). E.g., *Pierre est plus grand que n’était lui (on a pronominalized</p><p>(15a)) is not acceptable. A possible connection is</p><p>that information structure interacts with</p><p>syntax (cf., e.g., van Kemenade 2008 for OE).</p><p>1.  The question of whether the same semantic parameters are operative over time is cer-</p><p>tainly not trivial. Since motivating a theoretical framework for this falls beyond present scope,</p><p>I refer the interested reader to Beck et al. (to appear) for the background on possible syn-</p><p>tactic and semantic options. Empirically, parametric (in)variation can then be tested through</p><p>various degree constructions. To mention but one, let us take measure phrases. OE had</p><p>measure phrases (e.g. Gergel 2008), largely like PDE. The case of French is more intricate and</p><p>deserves special attention in future research. But we can illustrate the essentials. While the</p><p>ancestor language Latin had measure phrases (ducentos pedes altus, ‘two hundred feet high’),</p><p>such constructions were lost after the Old French period. However, what French seems to</p><p>have exploited is a rescue strategy. Measure phrases are thus still available, introduced by the</p><p>functional word de as in haut de 2m, ‘2m tall’.</p><p>1 Remus Gergel</p><p>conspicuous in the same connection of semantic variation in French is the need to</p><p>study the role of expletive negation even more closely than a rich tradition has already</p><p>done. While the negation morpheme ne does not have negative meaning in current</p><p>French comparatives (cf. Price 1990; von Stechow 1984), it is a relic from a time when</p><p>ne alone could potentially be the negative marker by itself. One major task for future</p><p>research, then, remains to explain how the negative form came to essentially mark the</p><p>clausal status of comparatives. (Phrasal comparatives lack ne). This fact may perhaps</p><p>be interpreted as one option that effectively helps keep the semantic settings constant.</p><p>As is well known, it is not a universal for a language to have clausal comparatives.</p><p>To conclude, the chapter has raised several questions and purports to have pro-</p><p>vided a few answers relating to comparative clauses. It has primarily illustrated a case</p><p>of simple, but thus far not systematically addressed, continuity in English and French</p><p>by discussing comparative inversion. We have drawn on the history of English and</p><p>largely extended the analysis to French, a language that in its diachronic evolution is</p><p>independently also well known for having low subjects. By using knowledge from dia-</p><p>chronic research on English and French, current syntactic diagnostics, and by having</p><p>partly investigated additional factors such as language contact, we have reasoned that</p><p>there is a case for a simple (and as described, identical) phrase-structural continuity</p><p>based on low subjects. This may have been co-supported, though from the current</p><p>evidence not caused, by language contact. Several questions have been left open here,</p><p>in particular so with regard to the interface with the component of meaning.</p><p>Major corpora and primary textual sources used in this study</p><p>Beowulf: A student edition. Jack, George (ed.). 1994. Oxford: Oxford University Press.</p><p>BR: Le voyage de Saint Brendan. Short, Ian & Merrilees, Brian (eds.). 2006. Paris: Ed. Honoré</p><p>Champion.</p><p>CdR: Chanson de Roland. Steinsieck, Wolfgang (ed.). 1999. Das altfranzösische Rolandslied.</p><p>Stuttgart: Reclam.</p><p>CLM: Corpus de la littérature médiévale. Éditions Champion Électronique. Available online at:</p><p>http://www.champion-electronique.net/bases.</p><p>FM: G. de Maupassant. 1923. Fort comme la mort, Éditions du Boucher.</p><p>FRANTEXT: Base textuelle Frantext. Nancy: Centre national de la recherche scientifique.</p><p>Available online at: http://www. frantext.fr.</p><p>LsR: La somme le roi. Édith Brayer and Anne-Françoise Labie-Leurquin (eds.). 2008. Paris –</p><p>Abbeville, Société des anciens textes français – Paillart.</p><p>PPCEME: Kroch, A., Santorini, B. & Delfs, L. 2004. The Penn-Helsinki Parsed Corpus of Early</p><p>Modern English.</p><p>PPCME2: Kroch, A. and Taylor, A. 2000. The Penn-Helsinki Parsed Corpus of Middle English.</p><p>YCOE Taylor, A., Warner, A., Pintzuk, S. and Beths, F. 2003. The York-Toronto-Helsinki Parsed</p><p>Corpus of Old English Prose.</p><p>Towards notions of comparative continuity in English and French 11</p><p>References</p><p>Adams, Marianne P. 1987. From Old French to the theory of pro-drop. Natural Language and</p><p>Linguistic Theory 5: 1–32.</p><p>Aissen, Judith. 1975. Presentational there-insertion. CLS 11: 1–14.</p><p>Beck, Sigrid, Krasikova, Sveta, Fleischer Daniel, Gergel, Remus, Hofstetter, Stefan, Savelsberg,</p><p>Christiane, Vanderelst, John & Villalta, Elisabeth. To appear. Cross-linguistic variation in</p><p>comparison constructions. In The Linguistic Variation Yearbook.</p><p>Besten, Hans den. 1978. On the presence and absence of wh-elements in Dutch comparatives.</p><p>Linguistic Inquiry 9: 641–671.</p><p>Besten, Hans den. 1983. On the interaction of root transformations and lexical deletive rules.</p><p>In On the Formal Syntax of the Westgermania. Papers from the 3 Groningen Grammar Talks</p><p>Groningen, January 1981. Linguistik Aktuell/Linguistics Today 3], Werner Abraham (ed.),</p><p>47–131. Amsterdam: John Benjamins.</p><p>Borroff, Marianne L. 2006. Degree phrase inversion in the scope of negation. Linguistic Inquiry</p><p>37: 514–521.</p><p>Chomsky, Noam. 1977. On wh-movement. In Formal Syntax, Peter Culicover, Thomas Wasow &</p><p>Adrian Akmajian (eds), 71–132. New York NY: Academic Press.</p><p>Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford: OUP.</p><p>Culicover, Peter W. & Winkler, Susanne. 2008. English focus inversion. Journal of Linguistics 44:</p><p>625–658.</p><p>Davison Alice. 1979. Some mysteries of subordination. Studies in the Linguistic Sciences 9:</p><p>105–128.</p><p>Doetjes, Jenny. 1992. Rightward floating quantifiers float to the left. The Linguistic Review 9:</p><p>313–332.</p><p>Faarlund, Jan T. 2004. The Syntax of Old Norse. Oxford: OUP.</p><p>Fischer, Olga, van Kemenade, Ans, Koopman, Willem & van der Wurff. 2000. The Syntax of</p><p>Early English. Cambridge: CUP.</p><p>Gergel, Remus. 2008. Comparative inversion: A diachronic study. Journal of Comparative</p><p>Germanic Linguistics 11: 191–211.</p><p>Gergel, Remus, Gengel, Kirsten & Winkler, Susanne. 2007. Ellipsis and inversion: A feature-</p><p>based focus account. In On Information Structure, Meaning and Form. Generalizations</p><p>across Languages [Linguistik Aktuell/Linguistics Today 100] Kerstin Schwabe & Susanne</p><p>Winkler (eds), 301–322. Amsterdam: John Benjamins.</p><p>Green, Georgia. 1976. Main clause phenomena in subordinate clauses. Language 52: 382–397.</p><p>Haeberli, Eric. 2002. Inflectional morphology and the loss of verb second in English. In Syntactic</p><p>Effects of Morphological Change, David Lightfoot (ed.), 88–106. Oxford: OUP.</p><p>Hankamer, Jorge. 1973. Why there are two ‘than’s in English. CLS 9: 179–191.</p><p>Hartmann, Jutta. M. 2005. Why there is(n’t) wh-movement in there constructions. In Linguistics</p><p>in the Netherlands 2005, Jenny Doetjes & Jeroen van de Weijer (eds), 87–99. Amsterdam:</p><p>John Benjamins.</p><p>Huddleston, Rodney & Pullum, Geoffrey. 2002. The Cambridge Grammar of the English Language.</p><p>Cambridge: CUP.</p><p>Hulk, Aafke & van Kemenade, Ans. 1995. Verb second, pro-drop, functional projections and</p><p>language change. In Clause Structure and Language Change, Adrian Battye & Ian Roberts</p><p>(eds), 227–256. Oxford: OUP.</p><p>1 Remus Gergel</p><p>Iatridou, Sabine & Kroch, Anthony. 1992. The licensing of CP-recursion and its relevance to the</p><p>Germanic verb-second phenomenon. Working Papers in Scandinavian Syntax 50: 1–24.</p><p>Kayne, Richard & Pollock, Jean-Yves. 2008. Toward an analysis of French hyper-complex</p><p>inversion. Ms, New York University/Université Paris-Est Marne l.V.</p><p>van Kemenade, Ans. 1997. V2 and embedded topicalisation in Old and Middle English. In</p><p>Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent (eds.), 326–352.</p><p>Cambridge: CUP.</p><p>van Kemenade, Ans. 2008. The interaction of syntax and information structure in Old and</p><p>Middle English. Paper presented at the 10th Diachronic Generative Syntax Conference,</p><p>Cornell University.</p><p>Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation</p><p>and Change 1: 199–244.</p><p>Kroch,</p><p>Anthony. 2008. V2 loss revisited: A quantitative comparison of English and French. Paper</p><p>presented at the 10th Diachronic Generative Syntax Conference, Cornell University.</p><p>Kroch, Anthony, Taylor, Ann & Ringe, Donald. 2000. The Middle English verb-second con-</p><p>straint: A case study in language contact and language change. In Textual Parameters in</p><p>Older Language [Current Issues in Linguistic Theory 195], Susan C. Herring, Pieter van</p><p>Reenen & Lene Schøsler (eds), 353–391. Amsterdam: John Benjamins.</p><p>Lahousse, Karen. 2006. NP-subject inversion in French: Two types, two configurations. Lingua</p><p>16: 424–461.</p><p>Lasnik, Howard & Sobin, Nicholas. 2000. The who/whom puzzle: On the preservation of an</p><p>archaic feature. Natural Language and Linguistic Theory 18: 343–371.</p><p>Merchant, Jason. 2003. Subject-auxiliary inversion in comparatives and PF output constraints.</p><p>In The Interfaces: Deriving and Interpreting Omitted Structures [Linguistik Aktuell/</p><p>Linguistics Today 61] Kerstin Schwabe & Susanne Winkler (eds), 55–77. Amsterdam:</p><p>John Benjamins.</p><p>Nevalainen, Terttu. 2006. An Introduction to Early Modern English. Edinburgh: EUP.</p><p>Pintzuk, Susan. 1991. Phrase Structures in Competition: Variation and Change in Old English</p><p>Word Order. Ph.D. dissertation, University of Pennsylvania.</p><p>Pollock, Jean-Yves. 1989. Verb movement, universal grammar and the structure of IP. Linguistic</p><p>Inquiry 20: 365–424.</p><p>Potts, Christopher. 2002. The syntax and semantics of as-parentheticals. Natural Language and</p><p>Linguistic Theory 20: 623–689.</p><p>Price, Susan. 1990. Comparative constructions in Spanish and French Syntax. London:</p><p>Routledge.</p><p>Rizzi, Luigi & Roberts Ian G. 1989. Complex inversion in French. Probus 1: 1–30.</p><p>Roberts, Ian G. 1993. Verbs and Diachronic Syntax. A Comparative History of English and French.</p><p>Dordrecht: Kluwer.</p><p>Roberts, Ian G. 2007. Diachronic Syntax. Oxford: OUP.</p><p>Rowlett, Paul. 2007. The Syntax of French. Cambridge: CUP.</p><p>Speyer, Augustin. 2008. On the Interaction of Prosody and Syntax in the History of English, with</p><p>a Few Remarks on German. Ph.D. dissertation, University of Pennsylvania.</p><p>Stechow, Arnim von. 1984. Comparing semantic theories of comparison. Journal of Semantics</p><p>3: 1–79.</p><p>Vance, Barbara S. 1997. Syntactic Change in Medieval French. Dordrecht: Kluwer.</p><p>Towards notions of comparative continuity in English and French 1</p><p>Warner, Anthony. 1997. The structure of parametric change, and V-movement in the history of</p><p>English. In Parameters of Morphosyntactic Change, Ans van Kemenade & Nigel Vincent</p><p>(eds), 380–393. Cambridge: CUP.</p><p>Williams, Alexander. 2000. Null subjects in Middle English existentials. In Diachronic Syntax:</p><p>Models and Mechanisms, Susan Pintzuk, George Tsoulas & Anthony Warner (eds), 164–190.</p><p>Oxford: OUP.</p><p>Winkler, Susanne. 2005. Ellipsis and Focus in Generative Grammar. Berlin: de Gruyter.</p><p>Variation, continuity and contact in Middle</p><p>Norwegian and Middle Low German</p><p>John D. Sundquist</p><p>Purdue University</p><p>This study focuses on the period of language contact between Middle Norwegian</p><p>(MNw) and Middle Low German (MLG) in Norway between 1300 and 1525.</p><p>Using corpora of MLG and MNw diplomatic letters written during this time,</p><p>we examine the possibility that syntactic continuity in the form of retention of</p><p>archaic forms may take place when two typologically similar languages are in</p><p>prolonged contact with each other. Data on the variation between Object–Verb</p><p>(OV) and Verb–Object (VO) order in both MNw and MLG indicate that the</p><p>frequency of OV word order in MNw is not affected by the influx of variational</p><p>patterns from MLG. We explore possible reasons for this lack of change and</p><p>focus on the role of syntactic variation in both continuity and change. Drawing</p><p>our conclusions from evidence on OV/VO variation, we argue that the syntactic</p><p>variation in the MNw data disrupts continuity rather than helping to maintain it.</p><p>1.  Introduction</p><p>1.1  Domain of inquiry</p><p>In recent research on diachronic syntax, there has been increasing interest in examin-</p><p>ing the apparent absence of syntactic change – or continuity – in certain contexts in</p><p>order to come closer to understanding the cause of change in other contexts. Interesting</p><p>data emerge when we compare instances of change in one language, such as contact-</p><p>induced borrowings or cyclical grammatical changes like Jespersen’s cycle, with stages</p><p>of continuity in a typologically similar language. Since Keenan’s (2002) description of</p><p>inertia in language change and Longobardi’s (2001) modifications of an earlier version</p><p>of this concept, the notion of continuity in historical syntax has been cast in a new light.</p><p>According to Keenan, “things stay the way they are unless acted on by an outside force</p><p>or decay” (2002: 327). Following this line of argumentation, Longobardi focuses more</p><p>specifically on syntactic change and proposes a reformulation of Keenan’s more gen-</p><p>eral point: “syntactic change should not arise, unless it can be shown to be caused”</p><p>(2001: 278). In other words, syntactic change is the exception to the rule: syntax</p><p>1 John D. Sundquist</p><p>remains stable unless such forces as language contact or morphological or phonologi-</p><p>cal decay exert significant pressure and cause it to change.</p><p>To examine this other side of the coin, we might ask what conditions are appro-</p><p>priate for continuity to be maintained. Following the logic of Longobardi’s principle</p><p>of inertia, we can assume that syntactic continuity exists in the absence of an out-</p><p>side force. However, as this chapter will discuss, the equation is not so simple when</p><p>we factor in syntactic variation. In the following case study on word order in Middle</p><p>Norwegian in the fourteenth and fifteenth centuries, we ask whether variation acts</p><p>to bring about change in a period of language contact with Middle Low German or</p><p>whether it helps to maintain continuity. The latter possibility has been discussed in</p><p>terms of ‘retention’, or as Braunmüller (1998: 329) calls it, ‘Driftumkehrung’ – a type of</p><p>gradual reversion, in which an outgoing, archaic variant is retained or reactivated dur-</p><p>ing periods of language contact with a typologically similar language. Contact in such</p><p>situations would have the opposite effect to what Longobardi (2001) predicts: it might</p><p>introduce more variation and strengthen certain outgoing patterns of usage.1</p><p>The main focus of this study is on the possibility of syntactic retention during a</p><p>period of language contact between Middle Low German (MLG) and Middle Norwegian</p><p>(MNw) in Norway. After assembling a sampling of texts from the fourteenth and</p><p>fifteenth centuries, I first ask to what extent Norwegian word order remained unchanged</p><p>when these two typologically similar languages were in contact with each other over</p><p>several hundred years. Secondly, I ask what role variation plays in bringing about</p><p>change or contributing to continuity in Norwegian. I focus on the word order pat-</p><p>terns [Object–Verbmain] (OV) and [Verbmain–Object] (VO) in Norwegian and German</p><p>corpora of diplomatic letters written between 1300 and 1500.2 The crucial question</p><p>in the study is the extent to which the influx of variational patterns of OV word order</p><p>from Middle Low German speakers affects the OV order that already existed but was</p><p>dying out in Norwegian. The empirical portion of the chapter will present the general</p><p>frequencies and patterns of variation in OV and VO orders in both languages, as rep-</p><p>resented in the two corpora. The discussion section will examine views on the role of</p><p>syntactic variation in periods of contact and provide support for the argument that</p><p>variation appears to disrupt continuity rather than help to maintain it.</p><p>1.  Thomason and Kaufman (1988: 58) also describe this phenomenon, referring to it as</p><p>retention: “…interference through shift in particular may even be responsible sometimes for</p><p>lack of change. This means that, since retention as well as innovation may be externally moti-</p><p>vated, the presence of inherited features is not always adequately</p><p>explained once one determines</p><p>their genetic origin”.</p><p>.  For simplicity, I follow the conventional term “OV” to refer to various kinds of PP- and NP-</p><p>complements to the non-finite main verb.</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 1</p><p>1.  Corpus</p><p>The Middle Norwegian (MNw) and Middle Low German (MLG) corpora that I used in</p><p>this study are both collections of diplomatic letters. Although this type of text exhibits writ-</p><p>ten language that is less representative of the spoken vernacular, I found it important to</p><p>have a homogenous text type throughout each corpus as well as across the two collections.3</p><p>Moreover, the letters are useful when examining issues of language contact and syntactic</p><p>change over time, since the provenance and date of composition are usually identifiable.</p><p>For the Middle Norwegian corpus, I selected 300 charters (approximately 90,000</p><p>total words) from Diplomatarium Norvegicum written between 1300 and 1525. Because</p><p>of the conservative, formulaic nature of these letters, I follow precautions suggested</p><p>by Mørck (1980) and exclude archaisms or copied phrases that might skew the data.4</p><p>I selected 1300 as the starting point for the corpus, since many of the diplomatic letters</p><p>prior to that date are in Latin, and I chose 1525 as the end point, since many of the let-</p><p>ters written at that time are almost exclusively in Danish. Although Bergen is the city</p><p>with the highest number of charters included in the corpus, letters from other major</p><p>scribal centers like Trondheim and Stavanger were also included.5</p><p>The Middle Low German corpus consists of 200 diplomatic letters (c. 75,000</p><p>words) from Codex Diplomaticus Lubecensis (or Lübeckisches Urkundenbuch) writ-</p><p>ten between 1325 and 1500. Because there are fewer letters written before 1325 in</p><p>Lübeck in the vernacular and most of the significant contact between Germans and</p><p>Norwegians began after this point, I used 1325 as the starting point for the Middle Low</p><p>German corpus.6 I did not include texts after 1500 because of the influence of High</p><p>.  Naturally, when analysing linguistic elements in these texts with a more conservative</p><p>style, we must assume that there is a larger gap between the spoken and written language than</p><p>would be the case in more informal texts with a colloquial style. However, it is the aim of this</p><p>study to focus on general patterns in the data that would indicate larger trends of linguistic</p><p>variation and change. In addition, the diplomatic letters constitute a less-than-ideal corpus</p><p>for the study of unmarked word order, since their conservative style and structure exhibit</p><p>strong influence from Latin. However, as will be discussed in this section, the consistency of</p><p>text type both throughout and between the two corpora was important for this study.</p><p>.  See Sundquist (2002, 2006) for further discussion of the MNw corpus and the precautions</p><p>that are necessary when analysing these texts.</p><p>.  Unlike other studies on MNw–MLG contact like Brattegard (1963), Jahr (1999) or Nesse</p><p>(2002), which focus solely on the Bergen dialect, the current study is concerned more with MLG</p><p>influence on Norwegian in general. Although it would be preferable to limit the study to letters</p><p>written in Bergen, where the number of German merchants during this time was highest, there</p><p>would be too few examples in this more limited corpus for a detailed quantitative analysis.</p><p>.  For a helpful overview of the most important dates and general historical information</p><p>on Hanseatic merchants in Norway with particular emphasis on Bergen, see Gade (1951),</p><p>Brattegard (1963), Helle (1994) and Nesse (2002).</p><p>1 John D. Sundquist</p><p>German that took hold in early part of the sixteenth century in Northern Germany and</p><p>Scandinavia (Braunmüller 2004). I limited the MLG corpus to texts written in Lübeck</p><p>by authors of that city in order to maintain consistency with respect to geographical</p><p>region and to focus on the most dominant Low German variety that would have been</p><p>used in Scandinavia in the period of the most intense contact in the fourteenth and</p><p>fifteenth centuries.7</p><p>.  Quantitative analysis: OV and VO word order in Middle Norwegian</p><p>and Middle Low German</p><p>In the analysis of variation between OV and VO word order in Middle Norwegian and</p><p>Middle Low German, I included any main or embedded clause in the corpora that</p><p>contains both a finite and a non-finite verb followed by two or more non-verbal con-</p><p>stituents. I also limited examples to those that contain a verbal complement – either an</p><p>NP- or a PP-complement to the non-finite verb. In Norwegian, the unmarked order in</p><p>such examples is VO, as in (1) and (2):8</p><p>(1) at hann hafðe suaret þænna sama vithnisburð aðr</p><p>that he had sworn that same testimony before</p><p>‘that he had sworn that same testimony before.’ (DN 1:122; 1309 Bergen)</p><p>(2) stod sa Haluard vp och vildæ gangæ at dørnen</p><p>stood so Haluard up and wanted go to door-the</p><p>‘Haluard stood up and wanted to go to the door.’ (DN 1:947; 1486 Skien)</p><p>However, as we will see in the discussion of the results below, Middle Norwegian exhib-</p><p>its variation between the pattern in (1) and (2) and the archaic OV order in examples</p><p>like (3) and (4):</p><p>(3) skall hon þetta goz hafua</p><p>shall he those wares have</p><p>‘He shall have those wares.’ (DN 2:120; 1314 Bergen)</p><p>.  Note that I included letters written by a wide variety of authors from Lübeck, but I did</p><p>not limit myself to those written by merchants or secretaries who had been to Norway. The</p><p>purpose of using this larger, less exclusive corpus for this study is to obtain an approximate</p><p>sampling of the language of this time and observe general norms, trends, and patterns of</p><p>variation. Future research on contact might focus on only those MLG letters written by indi-</p><p>viduals who spent time in Norway and came into contact with Norwegians, as in Simensen’s</p><p>(1989) study of some of the MLG letters in Diplomatarium Norvegicum.</p><p>.  MNw examples from Diplomatarium Norvegicum (DN) are cited with the volume number,</p><p>letter number, year of composition, and provenance.</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 1</p><p>(4) at han vildhe siælfuer til erkibiskupenn koma</p><p>that he wanted himself to archbishopric-the come</p><p>‘that he himself wanted to come to the archbishopric.’ (DN 1:713; 1426 Bergen)</p><p>Following Sundquist (2002, 2006), we can assume that OV word order in MNw is</p><p>inherited from Old Norse, as evidenced in earlier Norwegian manuscripts from the</p><p>thirteenth century.9</p><p>I follow the same general procedures in selecting examples for analysis from the</p><p>MLG corpus. MLG exhibits the opposite trend of MNw: OV order, as in (5) and (6), is</p><p>the norm, while the VO order in examples like (7) and (8) is marked.10</p><p>(5) dar schal he sin teken vp setten</p><p>there shall he his sign up set</p><p>‘There he shall set up his sign.’ (LU 1.8:DCCXXX; 1450 Lübeck)</p><p>(6) dat sodane vorscreuene schip vnd gudere in vnse hauene</p><p>that that aforementioned ship and goods into our harbours</p><p>komen mogen</p><p>come can</p><p>‘so that the aforementioned ship and goods can come into our harbours.’</p><p>(LU 1.7:XLV; 1427 Lübeck)</p><p>(7) de rat schal beboden alle de schuldemere</p><p>the council shall summon all the guilty</p><p>‘The council shall summon all the guilty ones.’ (LU 1.2:CDII; 1320 Lübeck)</p><p>(8) dat wi willen scriuen an den copman van Bergen</p><p>that we want write to the merchants from Bergen</p><p>‘that we want to write to the merchants from Bergen.’</p><p>(LU 1.5:CCCXVIII; 1410 Lübeck)</p><p>This state of affairs in MLG and MNw is not unexpected: MNw exhibits primarily</p><p>VO word order with some instances of the older OV order, while MLG exhibits the</p><p>opposite pattern.</p><p>.  Sundquist (2002, 2006) provides evidence that Norwegian texts such as the Konungs</p><p>skuggsjá (‘The King’s Mirror’) and Den eldre Gulatingslova (‘The Older Gulathing Law’),</p><p>whose manuscripts are from the middle of the 13th century, exhibit a higher rate of OV order</p><p>than the diplomatic letters from the 14th</p><p>century.</p><p>1.  Examples from Lübeckisches Urkundenbuch are listed according to section, volume, letter</p><p>number, date and the location of composition.</p><p>1 John D. Sundquist</p><p>.1  Variables that affect OV/VO variation</p><p>In order to track the variation between OV and VO word order in both the German</p><p>and Norwegian corpora, I analysed a number of variables. First of all, I examined the</p><p>role of syntactic weight, or the ‘heaviness’ of the NP- or PP-complement. According</p><p>to the extensive literature on post-posing of PPs and NPs outside the sentence frame</p><p>in earlier stages of German, the frequency of VO order is common in coordinated or</p><p>modified complements.11 Consider the following examples:</p><p>(9) dat de heren van Holzsten vnde wy scholden vnse</p><p>that the men from Holstein and we should our</p><p>boden zenden</p><p>messenger send</p><p>‘that we and the men from Holstein should send our messenger.’</p><p>(LU 1.4:CDXXXIII; 1387: Lübeck)</p><p>(10) dat sie ere brieue …senden scholen deme rade vnde der stat</p><p>that they their letter …send should the council and the city</p><p>tho Lubeke</p><p>of Lübeck</p><p>‘that they should send their letter to the council and city and Lübeck.’</p><p>(LU 1.4:XCIII; 1364: Lübeck)</p><p>In order to observe the effects of syntactic length, I differentiated between those sen-</p><p>tences with ‘simple’ PP/NP complements like (9) from ‘complex’ PPs and NPs that are</p><p>modified or coordinated, as in (10).12</p><p>Other variables that were tracked in the quantitative analysis included the type of</p><p>clause (main or embedded), type of non-finite verb (infinitive or past participle), type</p><p>of complement (PP, non-pronominal NP, pronominal NP), syntactic length (simple vs.</p><p>modified/coordinated PP/NP) and date of composition (see Pintzuk 1999 or Sunquist</p><p>2006 for information on the variables “type of clause” and “type of complement”).</p><p>Using the statistical program Goldvarb to analyse the patterns of interaction and</p><p>11.  For an overview of research on the sentence frame and Ausklammerung in early stages of</p><p>German, see Schildt (1977), who includes analysis of Low German. Ebert (1980, 1986), Betten</p><p>(1980), Ebert et. al. (1993) and Bies (1996) also discuss OV order in Early New High German.</p><p>For a more specific analysis that focuses on differences and similarities between MLG and</p><p>early Scandinavian with respect to the sentence frame, see Braunmüller (1993b, 1998).</p><p>1.  I follow Bies (1996) in making this distinction between simple vs. complex complements</p><p>as a measure of syntactic length. In her study of Early New High German word order, she</p><p>found that neither the number of syllables nor the number of words was a significant factor</p><p>in determining the position of heavy elements outside the sentence frame. Instead, she found</p><p>modified NPs and PPs or coordinated elements to be much more likely to occur in post-</p><p>position than simple complements of the verb.</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 11</p><p>variation, I used these variables as independent variables and OV word order as the</p><p>dependent variable.</p><p>In both the Norwegian and German corpora, the period under investigation was</p><p>divided into 25-year periods. I analysed enough letters to yield a minimum of 100 tokens</p><p>of OV/VO word order for each short period. I included as many examples as the chosen</p><p>letters yielded. Thus, for some of the 25-year periods, I analysed fewer letters because of</p><p>the large number of examples available in those letters, while for other 25-year periods, I</p><p>had to include more letters in order to collect the minimum number of tokens.</p><p>.  Results</p><p>..1  OV word order in Middle Norwegian</p><p>The general results of the MNw data analysis are listed in Table 1.13</p><p>Table 1. Goldvarb results for frequency of pre-verbal complements in Middle Norwegian</p><p>Variables Variants N (OV) %OV Probabilistic Weight</p><p>Type of</p><p>Complement</p><p>pronominals 148/315 46% .63</p><p>PPs 33/184 17% .55</p><p>non-pronominals 106/639 16% .24</p><p>Type of Clause main 126/489 25% .52</p><p>embedded 161/649 24% .49</p><p>Type of</p><p>Non-finite Verb</p><p>past participle 132/519 25% .49</p><p>infinitive 155/619 25% .52</p><p>Syntactic Length simple NPs/PPs 341/853 40% .61</p><p>coordinated/</p><p>modified NPs/PPs 50/285 18% .38</p><p>Date of</p><p>Composition</p><p>1300–1324 56/146 38% .69</p><p>1325–1349 33/104 31% .66</p><p>1350–1374 55/140 39% .68</p><p>1375–1399 33/120 28% .59</p><p>1400–1424 35/123 28% .56</p><p>1425–1449 33/123 26% .57</p><p>1450–1474 22/129 17% .42</p><p>1475–1499 10/119 8% .23</p><p>1500–1525 10/134 7% .17</p><p>There are some important trends in the Norwegian data that are worth noting.</p><p>Goldvarb selected type of complement, syntactic length and date of composition as</p><p>1.  The data in Table 1 include a measurement called probabilistic weight. This figure is Gold-</p><p>varb’s measurement of the effect of a variable on a scale of 0.00 to 1.00. A factor weight above</p><p>.50 has a favouring effect on the occurrence of the dependent variable, while a weight below</p><p>this threshold has a disfavouring effect.</p><p>1 John D. Sundquist</p><p>variables with a significant effect on the occurrence of OV word order. Pronominal</p><p>NP objects favour OV word order, PPs as a general group favour this order less, and</p><p>NPs significantly disfavour OV word order. As is to be expected, coordinated and modi-</p><p>fied verb complements occur less often in OV position than the simple NPs. In terms of</p><p>the general trend in OV-frequency over time, there is a period of fluctuation between</p><p>1300 and 1450 before a steady decline in frequency among all three types of comple-</p><p>ments between 1450 and 1525. Goldvarb selected type of clause and type of non-finite</p><p>verb as insignificant variables that have no effect on the frequency of OV word order.</p><p>..  OV word order in Middle Low German</p><p>The results for the MLG data are presented in Table 2.</p><p>Table 2. Goldvarb results for frequency of pre-verbal complements in Middle Low German</p><p>Variables Variants N (OV) %OV Probabilistic Weight</p><p>Type of</p><p>Complement</p><p>pronominals 191/196 97% .89</p><p>non-pronominals 361/457 78% .41</p><p>PPs 140/222 63% .25</p><p>Type of Clause Embedded 366/434 84% .57</p><p>main 326/441 73% .42</p><p>Type of</p><p>Non-finite Verb</p><p>past participle 303/368 82% .53</p><p>infinitive 389/507 76% .48</p><p>Syntactic Length simple NPs/PPs 595/707 84% .71</p><p>coordinated/</p><p>modified NPs/PPs 98/168 58% .45</p><p>Date of</p><p>Composition</p><p>1325–1349 100/154 64% .32</p><p>1350–1374 100/125 80% .48</p><p>1375–1399 94/132 71% .39</p><p>1400–1424 102/128 79% .49</p><p>1425–1449 92/105 87% .60</p><p>1450–1474 109/127 85% .62</p><p>1475–1499 95/104 91% .71</p><p>Much like in the analysis of the Norwegian data, Goldvarb selected type of com-</p><p>plement, syntactic length and date of composition as variables that significantly affect</p><p>the occurrence of OV word order in MLG, but it also found an additional factor group,</p><p>type of clause, to be significant. Pronominal NP-complements highly favour OV word</p><p>order, and according to the measurement of probabilistic weight, non-pronominal NPs</p><p>and PPs disfavour OV word order at a rate of .41 and .25 respectively. We see here that</p><p>there is a statistically significant difference between coordinated and modified NPs</p><p>and PPs in comparison to simple NPs and PPs: the coordinated or modified NPs/PPs,</p><p>as expected, occur more frequently in post-position (.45) while simple complements</p><p>significantly favour OV word order (.71). Unlike the Norwegian dataset, the German</p><p>examples indicate that OV word order is significantly higher in embedded clauses than</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 1</p><p>in main clauses, with probabilistic weights of .57 vs. .41. In terms of the general trends</p><p>in frequency data, the percentage of OV word order is very high throughout the period</p><p>under investigation: the rate is 64% in the first stage and increases steadily throughout</p><p>the fourteenth and fifteenth centuries before it reaches 91% by 1500. Goldvarb found</p><p>there to be no significant difference between sentences with a past participle and an</p><p>infinitive with respect to OV word order, disregarding type of non-finite verb as a</p><p>significant factor group.</p><p>.  Discussion</p><p>.1  Retention</p><p>as a result of contact</p><p>The data from the two corpora in Section 2 indicate that there was no increase in</p><p>the frequency of OV word order in MNw during the period under investigation. The</p><p>figures in Table 1 show that the percentage frequency of OV order between 1300 and</p><p>1525 does not increase for any extended period of time – in fact, it decreases from 38%</p><p>to 7% over the MNw period. During the same period, the frequency of OV order in</p><p>MLG increases steadily from 64% to 91%.</p><p>These findings pose problems for any proposal that suggests that contact between</p><p>MLG and MNw in Norway may have contributed to continuity and the retention of</p><p>archaic syntactic forms in Norwegian. In Christoffersen’s (2000: 166) discussion of the</p><p>effects of MLG contact on embedded clause word order in the history of Norwegian,</p><p>she argues that Low German had a type of “strengthening effect” on native Scandinavian</p><p>word-order patterns that MNw inherited from Old Norse. She cites her (1993) study</p><p>of the law code of Magnus Lagabøter from c. 1350 and points out the high level of OV</p><p>word order in embedded clauses.14 During the time of contact with MLG in Norway,</p><p>Christoffersen maintains, the similar structures in the two languages made it pos-</p><p>sible that certain word order patterns did not diminish in frequency in subsequent</p><p>stages of Norwegian. In other words, MLG contact helped preserve an older, native</p><p>Scandinavian tendency like OV word order that was otherwise a marked syntactic</p><p>variant on its way out of use (Christoffersen 2000: 165). Braunmüller (1998: 329) also</p><p>addresses this issue when he discusses the lack of syntactic borrowings from MLG</p><p>into Scandinavian. He warns that what may appear to be a borrowing may actually</p><p>1.  Christoffersen notes that 43% of embedded clauses in her (1993) study exhibit what she</p><p>calls “OV” word order (Christoffersen 2000: 161). She uses this term to describe the relative</p><p>order of non-finite verbs and their NP- or PP-complements, as I do in this study. However, she</p><p>also includes examples of “OV” order in which an NP- or PP-complement precedes the finite</p><p>verb as well. Thus, my data on OV order in this study and her data are not directly comparable.</p><p>For this reason, I am suggesting merely a general comparison of her findings and conclusions</p><p>with mine in this study.</p><p>1 John D. Sundquist</p><p>be the result of ‘Driftumkehrung’, or a gradual reversion, whereby the frequency of</p><p>a remnant, archaic word order pattern might remain the same rather than decrease</p><p>during contact between two similar linguistic varieties (1998: 331). He adds that such</p><p>reactivation, or strengthening, of older word order patterns, may only be temporary</p><p>during a period of dialect contact (1998: 332). As marked patterns become less tolerated</p><p>in spoken language, syntactic norms begin to be established in the written and spoken</p><p>language and the marginal patterns gradually drop out of use.</p><p>The MNw data here provide no evidence of retention. If we were to assume that</p><p>contact with Low German had even a slightly strengthening effect on the use of OV word</p><p>order in MNw, the percentage frequency of this pattern would flatten out over much of</p><p>the period under investigation or increase temporarily over the course of the fourteenth</p><p>and fifteenth centuries. The figures in Table 1 provide no evidence of such a trend.</p><p>Secondly, this decrease in OV frequency in MNw takes place at the same time as</p><p>OV word order becomes very common in MLG. Recall that the OV-rate in MLG texts</p><p>rises steadily from 64% to 91% between 1325 and 1500 with only slight decreases in</p><p>some 25-year periods of the fourteenth century. We must assume that, although the</p><p>written language of the German letters differs from the spoken language in many ways,</p><p>it is still a reflection of the spoken language to a certain extent. Thus, there is a problem</p><p>with a hypothesis which assumes that this pattern was becoming commoner in MLG</p><p>and that exposure to MLG might have aided the reactivation of OV word order in</p><p>Norwegian. In fact, the statistical trends for the two languages run in the opposite</p><p>direction of each other: the frequency of OV word order in Norwegian is at its lowest</p><p>in the fifteenth century, just when it reaches its highest level in MLG.</p><p>The data on syntactic length provide further counter-evidence. Recall from</p><p>Tables 1 and 2 that syntactic length was a significant variable in both the Norwegian</p><p>and German data. It is clear that coordinated or modified NPs/PPs occur more fre-</p><p>quently in post-position (VO order) than simple NPs/PPs in both languages. This</p><p>result is to be expected if we consider the interaction of word order and pragmatic</p><p>constraints: heavier constituents are often post-posed near the end of the sentence as</p><p>a type of afterthought when such syntactic options are available. However, the data on</p><p>syntactic length raise the question why Norwegian does not behave even more like</p><p>MLG with respect to the placement of simple NPs. As Table 2 indicates, 84% (proba-</p><p>bilistic weight .71) of the simple NPs/PPs in German occur in OV position. As Table 1</p><p>demonstrates, however, only 40% (.61) of the same category of verbal complements</p><p>occurs in OV word order in MNw. If there were a natural, universal tendency for</p><p>simple complements to occur in pre-verbal position when such options are available,</p><p>and if contact with MLG had any significant effects on Norwegian word order, we</p><p>would expect a much higher OV-rate among simple NPs. In other words, with all else</p><p>being equal, there are relatively few departures from the unmarked VO word order in</p><p>MNw – even in situations where the language naturally allows this option and where</p><p>the tendency is so robust in MLG.</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 1</p><p>.  Continuity as a result of syntactic variation</p><p>The previous section demonstrated that there is little evidence to support the hypothesis</p><p>that contact with MLG had an effect on the retention of OV word order structures</p><p>in Norwegian. In the corpora for this study, the frequency of OV structures in MNw</p><p>declines during the same period in which contact with MLG was at its strongest in</p><p>Bergen and in other parts of the country.15</p><p>The commonest explanation for the lack of MLG influence on Norwegian syntax</p><p>during the fourteenth and fifteenth centuries is that the sociolinguistic conditions</p><p>were not appropriate for syntactic borrowing to take place. It is beyond the scope of the</p><p>current study to discuss the sociolinguistic factors that may have played a role in lexi-</p><p>cal or syntactic borrowing in Norwegian. Suffice it to say that, in terms of Thomason</p><p>and Kaufman’s (1988: 74) description of social and linguistic factors that lead to borrow-</p><p>ing, the Norwegian–MLG contact situation was not ripe for much change outside of</p><p>lexical borrowing.16 Although MLG merchants were in contact with Norwegians for</p><p>several centuries in Norway, especially in Bergen, there was little interaction between</p><p>Norwegians and Germans other than in business transactions. Germans did not inter-</p><p>marry with Norwegians, they lived in isolated communities, and both groups had sep-</p><p>arate laws, churches, social groups, and living quarters. As Ersland (2003: 45) points</p><p>out, there is no evidence to indicate that Germans in Bergen, for instance, mingled</p><p>with the local population or assimilated with locals. Thus, one can assume that the lack</p><p>of syntactic influence from German is attributable, in part, to a lack of intense contact</p><p>between the languages.</p><p>However, another hypothesis that has been suggested is that syntactic varia-</p><p>tion played a decisive role in limiting the amount of syntactic borrowing. Following</p><p>Trudgill (1994), proponents of this view focus on the unique characteristics of con-</p><p>tact between MLG and MNw (or Mainland Scandinavian in general) as typical for</p><p>a type of dialect contact rather than language contact. As Braunmüller (1998) and</p><p>Christoffersen (2000) suggest, the extensive amount of</p><p>(FOFC). In Proceedings of XXXIII Incontro di Gram-</p><p>matica Generativa, Antonietta Bisetto & Francesco Barbieri (eds), 86–105. Bologna: Uni-</p><p>versità di Bologna.</p><p>Doyle, Aidan. 2002. Yesterday’s affixes as today’s clitics. In New Reflections on Grammaticaliza-</p><p>tion [Typological Studies in Language 49], Ilse Wischer & Gabriela Diewald (eds), 67–81.</p><p>Amsterdam: Benjamins.</p><p>Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross-linguistic Perspective. Cam-</p><p>bridge: CUP.</p><p>Haspelmath, Martin. 1999. Why is grammaticalization irreversible? Linguistics 37(6): 1043–68.</p><p>Keenan, Edward L. 2002. Explaining the creation of reflexive pronouns in English. In Studies in</p><p>the History of English: A Millennial Perspective, Donka Minkova & Robert Stockwell (eds),</p><p>325–355. Berlin: Mouton de Gruyter.</p><p>Kroch, Anthony S. 1989. Reflexes of grammar in patterns of language change. Language Varia-</p><p>tion and Change 1: 199–244.</p><p>Lightfoot, David W. 1991. How to Set Parameters: Arguments from Language Change. Cambridge</p><p>MA: The MIT Press.</p><p>Lightfoot, David W. 1999. The Development of Language: Acquisition, Change and Evolution.</p><p>Oxford: Blackwell.</p><p>Longobardi, Giuseppe. 2001. Formal syntax, diachronic minimalism, and etymology: The his-</p><p>tory of French chez. Linguistic Inquiry 32: 275–302.</p><p>Meisel, Jürgen. 1995. Parameters in acquisition. In A Handbook of Child Language, Paul Fletcher</p><p>& Brian MacWhinney (eds), 10–35. Oxford: Blackwell.</p><p>Meisel, Jürgen. 2001. The simultaneous acquisition of two first languages. Early differentiation</p><p>and subsequent development of grammars. In Trends in Bilingual Acquisition [Trends in</p><p>Language Acquisition Research 1], Jasone Cenoz & Fred Genesee (eds), 11–41. Amsterdam:</p><p>John Benjamins.</p><p>10 Anne Breitbarth, Christopher Lucas, Sheila Watts & David Willis</p><p>Roberts, Ian & Roussou, Anna. 2003. Syntactic Change: A Minimalist Approach to Grammatical-</p><p>ization. Cambridge: CUP.</p><p>Thomason, Sarah & Kaufman, Terrence. 1988. Language Contact, Creolization, and Genetic Lin-</p><p>guisitcs. Berkeley CA: University of California Press.</p><p>Timberlake, Alan. 1977. Reanalysis and actualization in syntactic change. In Mechanisms of Syn-</p><p>tactic Change, Charles N. Li (ed.), 141–177. Austin TX: University of Texas Press.</p><p>Weinreich, Uriel, Labov, William & Herzog, Marvin I. 1968. Empirical foundations for a theory</p><p>of language change. In Directions for Historical Linguistics, Winifred P. Lehman & Yakov</p><p>Malkiel (eds), 95–195. Austin TX: University of Texas Press.</p><p>part i</p><p>Continuity</p><p>What changed where?</p><p>A plea for the re-evaluation of dialectal evidence*</p><p>Katrin Axel & Helmut Weiß</p><p>University of Göttingen/Goethe-Universität, Frankfurt-am-Main</p><p>In the field of generative diachronic syntax, it has often been disregarded at</p><p>which level of the language (dialect or Standard) syntactic change has occurred.</p><p>However, just as in the case of phonological developments, the syntax often</p><p>(though not always) turns out to be more conservative at the dialectal level.</p><p>In this chapter we will present four cases studies on the syntax of German:</p><p>the diachrony of pro-drop (null subjects), of negative concord, of possessive</p><p>constructions and of word-order changes in the verbal cluster. Our plea for taking</p><p>into account dialect data in historical linguistics converges with the growing</p><p>significance dialects and dialectal data have gained within theoretical linguistics.</p><p>1.  Introduction</p><p>It is commonly assumed in generative linguistics that language change – and thus conti-</p><p>nuity (i.e. the non-occurrence of change) as well – is largely due to first language acqui-</p><p>sition (L1 acquisition).1 Therefore, it would be quite natural to investigate language</p><p>change only in languages (or dialects, varieties) which were transmitted from genera-</p><p>tion to generation by L1 acquisition alone.2 Contrary to expectation, this is not really</p><p>the case. Research on language change within the generative framework is still mostly</p><p>restricted to developments in languages which have undergone standardization.</p><p>There are two misconceptions which can presumably be made responsible for this</p><p>curious state of affairs. Firstly, based on the – no doubt – correct assumption that</p><p>I-language is the primary object of investigation of linguistic research, generative lin-</p><p>guists refuse to discuss what kind of objects languages like English or German are</p><p>*We are grateful to the audience at the conference on ‘Continuity and Change in Grammar’ for</p><p>their comments. Thanks also to two anonymous reviewers and to the editors for their helpful</p><p>suggestions. Special thanks to Sheila Watts for her assistance and many insightful comments.</p><p>1.  Cf. also already Paul (1880).</p><p>.  There are other types of language change, e.g. contact-induced language change (Kroch</p><p>2000; Weiß 2005b).</p><p>1 Katrin Axel & Helmut Weiß</p><p>(though they use the terms as convenient abbreviations, for example, speaking of the</p><p>loss of negative concord in English). Secondly, it is often assumed, that “standard lan-</p><p>guages are simply dialects which have gained a different [i.e. higher] social prestige”</p><p>(Cornips & Poletto 2005: 939). Both these assumptions lead to the curious fact just</p><p>mentioned that generative historical research is usually concerned with developments</p><p>in standard languages.3</p><p>Not to deal with the concept of language may be a pardonable neglect (though it</p><p>seems to be a little strange to leave such an essential term undefined, cf. Weiß 2009).</p><p>However, to hold standard languages as nothing more than dialects is simply false.</p><p>Standard languages are not simply codified forms of spoken and L1-acquired dia-</p><p>lects, they almost always evolved as secondarily learned, written languages. Standard</p><p>German may be an extreme case because it evolved out of several dialects and has</p><p>never corresponded to any individual one, but, at least initially, all standard languages</p><p>develop relatively independently of L1-acquisition (Weiß 2001, 2004a, 2004b, 2005b,</p><p>2005c).</p><p>In Weiß (2001) it is therefore proposed that there are two kinds of natural lan-</p><p>guages based on the criterion of language acquisition. According to this theory, first</p><p>order natural languages (N1 languages) are those which are acquired in the course of</p><p>L1 acquisition, whereas second order natural languages (N2 languages) are learned,</p><p>that is acquired mostly by instruction (e.g. in school). Dialects (and languages) as</p><p>defined in Weiß (2001) are N1 languages, and standard languages are N2 languages, or</p><p>at least used to be in the past.4</p><p>Given the fundamental role which L1-acquisition plays in modern historical</p><p>linguistics, it follows naturally that dialects should be the objects of investigation.</p><p>A second consequence for historical linguistics is that whenever there is a different</p><p>development in dialects and standard languages, it is the development in the dialects</p><p>.  There are a couple of notable exceptions: Poletto (1995) and Ledgeway (2000) diachronically</p><p>investigate various aspects of Italian dialect syntax. Kroch & Taylor (1997) deal with the verb-</p><p>second property in different historical dialects of English. Sprouse & Vance (1999) discuss the</p><p>null-subject properties of various Romance and Germanic languages from a diachronic per-</p><p>spective. Moreover, the work on creole languages should be mentioned here (cf. deGraff 1999</p><p>for an overview). Thanks to an anonymous reviewer for pointing this out. However, as far as</p><p>German is concerned, there is hardly any generative work on historical syntax so far (but see</p><p>Lenerz 1984; Axel 2007, Jäger 2008). Non-generative diachronic studies have been primarily</p><p>concerned with syntactic developments leading to the standard; this in contrast to work in</p><p>phonology, which has always taken into account the situation in the dialects.</p><p>.  Nowadays, standard languages are subject to L1 acquisition to a much greater extent</p><p>than was the case in the past (Weiß 2004a,b, 2005c). These ‘re-naturalized’ languages then can</p><p>develop new dialects as well as colloquial variants (e.g. social registers).</p><p>syntactic variation in these</p><p>1.  For more background information on the sociolinguistic conditions for contact in Norway,</p><p>see Jahr (1999). For specific discussion of sociolinguistic and cultural aspects of German–</p><p>Norwegian contact in Bergen, see Nesse (2002) or Ersland (2003).</p><p>1.  This conclusion is supported by the findings in Zeevaert’s (1995) corpus-based analysis of</p><p>lexical borrowing in Scandinavian. Although he focuses on Swedish, the findings are relevant</p><p>to any analysis which attempts to determine the level of intensity of contact between MLG</p><p>and Scandinavian in general. Using Thomason and Kaufman’s (1988) classification system for</p><p>intensity, he concludes that MLG–Scandinavian contact was between Levels 1 (light contact)</p><p>and 2 (somewhat intense contact) (1995: 175).</p><p>1 John D. Sundquist</p><p>similar languages allowed speakers to use accommodation strategies in their speech.17</p><p>Braunmüller (1998) speculates that both spoken Scandinavian and MLG had relatively</p><p>few strict word order norms in the fourteenth and fifteenth century. Because of the</p><p>deeply rooted structural similarities between the languages, speakers on both sides of</p><p>business transactions could speak their native language and understand each other.</p><p>One reason for the success of this passive multidialectalism, according to Braunmül-</p><p>ler (1998), is that speakers took advantage of the wide variety of word order patterns</p><p>in their own language and used those patterns more frequently which could be easily</p><p>processed by their interlocutors. As an example, he suggests the possibility that MLG</p><p>speakers could accommodate Scandinavians and ease processing constraints by using</p><p>VO word order when pausing in speech to post-pose elements to positions to the right</p><p>of the non-finite verb (Braunmüller 1998: 325). Similarly, Scandinavians may have</p><p>taken advantage of the grammatical OV option in their language more frequently in</p><p>order to accommodate their MLG trading partners. More marked syntactic patterns</p><p>may have been used more often as part of functionally motivated accommodation</p><p>strategies. As a result of this variation, MLG exerted little influence on the word order</p><p>of MNw: the languages were too similar and they exhibited too much variation for any</p><p>type of long-term or even temporary impact to take place.</p><p>The data in the current study, however, provide little support for this view. Quanti-</p><p>tative analysis of MLG in Section 2 reveals that there is not as much OV/VO variation</p><p>in MLG as has been previously assumed. From the beginning of the period under</p><p>investigation, OV word order occurs in 61% of the examples in the Low German letters</p><p>as early as 1325. This rate increases steadily before it is reaches 91% in the fifteenth</p><p>century. We can assume that the lack of OV/VO variation in the diplomatic letters may</p><p>be, in part, a result of these external factors and the conservative style of the text type.18</p><p>However, the general trend in OV word order in letters from Lübeck is consistent</p><p>throughout the period under investigation, and the steady increase in frequency of OV</p><p>word order in the written language begins before the influence of the High German</p><p>standard in the sixteenth century. Although there are no strict norms in word order</p><p>from this period, there are clear preferences that follow definite trends in the language</p><p>as early as the fourteenth century.19</p><p>1.  For a more detailed discussion of this argument, see Braunmüller (1993a, 1996, 1998,</p><p>2005), Christoffersen (2000) or Mæhlum (2005).</p><p>1.  As has been noted by Schildt (1977) for Low and High German and by Ebert (1980) for</p><p>Early New High German, various language-external factors, including educational level and</p><p>gender of the author, style, and geographical region, contribute to the frequency with which</p><p>authors have post-posed PPs and NPs in their written language.</p><p>1.  As in any quantitative analysis of historical syntax, we are forced to draw conclusions</p><p>about the spoken language of an earlier period from our analysis of the written language.</p><p>Variation, continuity and contact in Middle Norwegian and Middle Low German 1</p><p>Furthermore, the statistical patterns of use of OV and VO word orders reveal a form</p><p>of systematic variation rather than random fluctuation between syntactic variants.</p><p>Recall that both type of complement and syntactic length were variables that had</p><p>a favouring effect on the occurrence of OV word order in MLG. For instance, there is</p><p>a highly significant difference between pronominal objects and prepositional phrases</p><p>with respect to their tendency to occur in OV position (.89 vs. .25 probabilistic weight).</p><p>These findings indicate that there is a clear underlying system in place that regulates</p><p>the positioning of various complement types. Although there are various syntactic</p><p>options available to MLG speakers, the choice between these options is not random</p><p>and fluctuation between OV and VO order is not as free as previously assumed. In the</p><p>same way, the distinction between complex and simple complements also shapes the</p><p>variable use of word order patterns: coordinated and modified NPs/PPs favour OV</p><p>word order at the level of .71 while simple complements disfavour it at a rate of .45.</p><p>Once again, specific language-internal factors shape the patterns of OV/VO usage and</p><p>demonstrate the systematic characteristics of these patterns.</p><p>Thus, the Low German data here cause problems for any analysis that assumes</p><p>that MLG was a language with a high degree of variation with few word-order norms.</p><p>Empirical analysis of texts from this period reveals that there were clear preferences</p><p>in OV/VO usage throughout the fourteenth and fifteenth centuries and that there</p><p>were systematic rules that governed this usage. In terms of contact between MLG</p><p>and Norwegian, therefore, it is less easy to assume from analysis of the written lan-</p><p>guage that the MLG spoken language exhibited as much variation as others have</p><p>previously claimed.20</p><p>.  Change as a result of syntactic variation</p><p>Previous sections of this chapter focused on the role of syntactic variation in a period</p><p>of contact between MLG and MNw. It was shown that, contrary to previous claims,</p><p>new patterns of variation introduced during contact with MLG had no stabilizing</p><p>effects on OV frequency in MNw. In fact, there is little retention of OV order during</p><p>Furthermore, we can assume that the MLG letters do not reflect the exact type of language to</p><p>which the Norwegians would have been exposed. However, as Pintzuk (1999: 14) concludes,</p><p>the written language is our only source for drawing general conclusions about the spoken</p><p>language. I assume that the precautions taken in text selection discussed in Section 1.2 have</p><p>made the MLG corpus here as reliable as possible for drawing such conclusions.</p><p>.  Note that I agree with the conclusion in Braunmüller (1996, 1998, 2004, 2005), Christof-</p><p>fersen (2000) and Mæhlum (2005) that the MLG–Scandinavian contact was a type of dialect</p><p>contact and that speakers of both languages could communicate with each other in their re-</p><p>spective native languages by using accommodation strategies. However, I wish merely to point</p><p>out that the evidence here from the written languages suggests that the extent of variability in</p><p>the two languages is less than previously assumed.</p><p>1 John D. Sundquist</p><p>the period under investigation, as the archaic forms begin to decline more rapidly in</p><p>frequency and VO order becomes dominant.</p><p>These findings support a hypothesis proposed recently by Reintges (2009), who</p><p>suggests that syntactic variation is an important source for change. Reintges states this</p><p>as follows:</p><p>If we are right in claiming that syntactic variation can arise spontaneously and</p><p>that syntactic change follows from syntactic variation whatever the source of the</p><p>variation is, we are in fact making the controversial claim that syntactic change</p><p>can happen without outside factors playing any role. Syntactic variation initiates</p><p>syntactic change,</p><p>This is, for example,</p><p>the case with Standard English which has even developed several national varieties in Great</p><p>Britain, America, Australia and so on (Durrell 1999: 299).</p><p>What changed where? 1</p><p>which should gain our attention as historical linguists,5 because these changes can be</p><p>largely ascribed to language-internal factors alone, whereas standard languages have</p><p>also been formed by extra-linguistic forces.</p><p>In what follows we will present a case study on pro-drop in German and several</p><p>other case studies in which changes in dialects differ from those in the standard. All</p><p>these case studies show that taking into account dialectal evidence in historical syntax</p><p>can lead to different and more appropriate results. These results constitute a more solid</p><p>basis for judgments about what is stable and what is more open to change in the history</p><p>of a language.</p><p>This plea for taking into account dialect data in historical linguistics converges</p><p>with the growing significance dialects and dialectal data have gained within theoretical</p><p>linguistics. Besides their greater naturalness (see above), it is above all the issue of</p><p>microvariation that makes the comparative study of dialects an invaluable source for</p><p>linguistic research. In the line of Kayne’s (1996) concept of microvariation, dialects</p><p>provide a unique data source for investigating what is possible or impossible in natural</p><p>languages (Benincà & Poletto 2007).</p><p>.  Case study 1: Pro-drop in German</p><p>Our major case study is on pro-drop in German. It is well-known that in Modern Stan-</p><p>dard German (referential) subject pronouns have to be overtly realized:</p><p>(1) Kaffee magst *(du) nicht.</p><p>coffee like you not</p><p>‘You don’t like coffee.’</p><p>However, modern German dialects do license pro-drop under certain conditions</p><p>and the same is true for all the major dialect groups within the Continental West-</p><p>Germanic group.</p><p>In the previous literature, it has often been claimed that German lost the pro-drop</p><p>property at a very early stage, even though there is a large consensus that Proto-</p><p>Germanic and Proto-Indo-European were pro-drop languages. Paul (1919: 22) states</p><p>that referential null subjects died out before the earliest attested stages of Old High</p><p>German (OHG). Hopper (1975: 31), following Meillet (1908/1909: 89), speculates that</p><p>the use of overt subject pronouns is the rule already in early Germanic. In the Mittelhoch-</p><p>deutsche Grammatik (Paul 2007: 315) it is claimed that in more recent stages of the lan-</p><p>guage inflectional endings are not capable of ‘expressing’ the subject anymore, thereby</p><p>implying that this was possible in older stages. What has been almost totally ignored is</p><p>.  To the extent that there is sufficient documentation of the dialects. There is, of course, the</p><p>problem that there are many dialects which have barely received syntactic description.</p><p>1 Katrin Axel & Helmut Weiß</p><p>that there are many dialects that still allow subject omission in certain morphosyntactic</p><p>environments even in contemporary German.</p><p>.1  Pro-drop in the modern dialects</p><p>Though the Continental West Germanic dialects are only partial pro-drop languages in</p><p>the sense defined in Weiß (2001), they nonetheless exhibit some indisputable instances</p><p>of pro-drop (e.g. Bayer 1984; Weiß 1998: 116ff. on Bavarian, Haag-Merz 1996: 153ff.</p><p>on Swabian, Cooper 1995 on Zurich German). In Bavarian, for instance, pro-drop is</p><p>possible in the 2nd sg., (2), and 2nd pl. (cf. also Bayer 1993; Fuß 2005):</p><p>(2) a. moang bisd pro wieda gsund (Bavarian)</p><p>tomorrow are-2.sg again healthy</p><p>‘Tomorrow, you will be well again.’</p><p>b. wennsd pro moang wieda gsund bisd</p><p>if-2.sg tomorrow again healthy are-2.sg</p><p>‘if you will be well again tomorrow.’</p><p>As can be seen in (2b), Bavarian shows so-called complementizer agreement: the</p><p>complementizer wenn carries the inflectional morpheme -sd and thus agrees with the</p><p>finite verb at the end of the clause. In subordinate clauses pro-drop is parasitic on</p><p>complementizer agreement. In the 1st sg. there is no complementizer agreement in</p><p>Bavarian. The subject pronoun has to be realized overtly in this case:</p><p>(3) wenn-*(e) moang wieda gsund bin (Bavarian)</p><p>if-I tomorrow again healthy am</p><p>‘if I will be well again tomorrow.’</p><p>So we can conclude that the presence of overt Agr-in-C is necessary for pro-drop.</p><p>Whether Agr-in-C is the result of verb raising to C or of inflected complementizers</p><p>seems to be irrelevant.</p><p>However, Agr-in-C alone does not suffice to license pro-drop. Consider the follow-</p><p>ing contrast, which, according to Weise (1900: 56), shows up in the Thuringian dialect</p><p>spoken in Altenburg:6</p><p>(4) a. schreib mir einmal den Brief, (Thuringian)</p><p>write-imp-2.sg me once the letter,</p><p>kriegst pro auch einen Groschen</p><p>get-2.sg too a penny</p><p>‘Write me a letter and you will get a penny.’</p><p>.  Example (4a) is taken from Weise (1900: 56). Example (4b) is constructed in accordance</p><p>with Weise’s remark that ‘In the plural, this ellipsis of the subject pronoun is not attested’</p><p>(“Im Plural ist diese Ellipse [des Subjektspronomens] nicht nachweisbar” (Weise 1900: 56).</p><p>What changed where? 1</p><p>b. schreibt mir einmal den Brief, dann kriegt</p><p>write-imp-2.pl me once the letter, then get-2.pl</p><p>*pro/ihr auch einen Groschen</p><p>pro/you-2.pl too a penny</p><p>‘Write me a letter and you will get a penny.’</p><p>In this dialect pro-drop is possible with the 2nd sg., and not with the 2nd pl., as the</p><p>contrast between (4a) and (4b) reveals. The difference between the inflectional marker</p><p>-st in the 2nd sg. and the -t in the 2nd pl. is that only the former contains a pronominal</p><p>element, namely /t/.</p><p>As the OHG data show, the ending of the 2nd sg. (= -st), was originally only -s.</p><p>The form with an additional -t originated from cases where the verb was followed</p><p>by the pronoun thu, du in OHG times. In this configuration the word boundary</p><p>was then reanalysed in such a way that the dental plosive became part of the verbal</p><p>ending.</p><p>As Weise (1907: 204) already notes, subject pronouns can be absent in the 2nd sg.</p><p>and pl. in exactly those Bavarian and East Franconian dialects where complementizer</p><p>agreement is pronominal (5).7</p><p>(5) a. wennsd pro mogsd</p><p>if-2.sg like-2.sg</p><p>‘if you like’</p><p>b. wennds8 pro megds</p><p>if-2.pl like-2.pl</p><p>‘if you like’</p><p>(Central Bavarian, likewise North Bavarian and East Franconian)</p><p>In East Franconian, the 1st and 3rd pl exhibit non-pronominal complementizer agree-</p><p>ment (6), and thus the subject pronoun cannot be dropped.</p><p>(6) a. waaln *(mer) graad besamn senn</p><p>because-1.pl we at-the-moment together are-1.pl</p><p>‘since we are together now.’</p><p>b. waaln *(se) graad besamn senn</p><p>because-3.pl they at-the-moment together are-3.pl</p><p>‘since they are together now’ (East Franconian, Weise 1907: 204)</p><p>.  When we speak of (non-)pronominal agreement, we do not imply that speakers can dis-</p><p>tinguish between the two because that would require information about the historical develop-</p><p>ment which speakers normally do not have. Speakers are only capable of detecting that the</p><p>one variant of agreement allows for pro-drop and the other one does not.</p><p>.  This inflectional suffix developed from the enclitic pronoun es parallel to the 2nd sg.</p><p>suffix -sd.</p><p>1 Katrin Axel & Helmut Weiß</p><p>On the basis of the data discussed here (and comparable ones, cf. Weiß 1998, 2005a) we</p><p>can postulate the following licensing principle for pro-drop in modern German dialects:</p><p>(7) Licensing principle for pro-drop:</p><p>pro must be c-commanded by pronominal agreement.</p><p>It should be noted that pro-drop and inflected complementizers are by no means</p><p>restricted to Upper German dialects like Bavarian and East Franconian. It is attested</p><p>in all major groups of modern German dialects, especially in Eastern Middle German</p><p>dialects, but for the 2nd sg. even in Low German dialects like Low Saxon and Low</p><p>Franconian (cf. Weiß 2005a).</p><p>.  Pro-drop in early German</p><p>It is generally assumed that the pro-drop property inherited from Proto-Germanic</p><p>was</p><p>already largely lost in the OHG period, that is, in the oldest attested stage of the German</p><p>language. The introduction of overt subject pronouns in pre-/early OHG has tradition-</p><p>ally been linked to the weakening of verbal endings due a phonological weakening</p><p>process which in turn has been argued to have been a consequence of the introduc-</p><p>tion of word-initial accent (see Kögel 1882: 127; Held 1903: XIII; Behaghel 1928: 442).</p><p>However, as Grimm (1967 [1898]: 235) and later Eggenberger (1961) already objected,</p><p>these phonological and morphological changes did not lead to a substantial leveling of</p><p>inflectional distinctions during the OHG period. It should furthermore be noted that</p><p>from a cross-linguistic perspective even the Standard Present-Day German paradigm</p><p>is usually considered sufficiently strong to license pro-drop (Jaeggli & Safir 1989).9,10</p><p>The absence of pro-drop has thus been attributed to an incompatibility of the pro-drop</p><p>and the verb-second (V2) parameters (Jaeggli & Safir 1989; Rohrbacher 1999). And</p><p>what is even more problematic: we have seen that at a dialectal level the alleged loss</p><p>of pro-drop has not occurred even though here the verbal paradigms have the same</p><p>amount of syncretism or even more than in the standard language.</p><p>.  This despite the fact that in the Present-Day German paradigm there is more syncretism</p><p>than in OHG. For example, the OHG verbal paradigm attested in the OHG Tatian (c. 830 AD,</p><p>East Franconian) has six different endings in the present indicative (-u, -is(t), -it, -emēs, (-ēn),</p><p>-et, -ent. There is thus no syncretism at all. By contrast, in the modern Standard German para-</p><p>digm (-e, -st, -t, -en, -t, -en) there is some syncretism. In the plural, the first and third person</p><p>have identical endings. Also, both the 3rd sg. and the 2nd pl. end in -t. So indeed the verbal</p><p>paradigm has been weakened.</p><p>1.  However, more recently the opposite claim has been propagated (e.g. Müller 2006;</p><p>Koeneman 2007).</p><p>What changed where? 1</p><p>As mentioned above, it is generally assumed that in OHG (referential) pro-drop</p><p>has already been largely lost. However, there are numerous instances of pro-drop in the</p><p>eighth and ninth century texts:</p><p>(8) a. Druhtin ist auh (I 426; OHG)</p><p>Lord is also</p><p>‘He is also the Lord.’</p><p>b. Sume hahet in cruci (MF XVIII,17)</p><p>some-acc hang-2.pl to cross</p><p>‘Some of them you will crucify.’</p><p>et ex illis … crucifigetis,</p><p>c. quidis zi uns thesa parabola (T 529,2)</p><p>say-2.sg to us this parable</p><p>‘Are you telling this parable to us?’</p><p>ad nos dicis hanc parabolam</p><p>In the traditional literature, the standard approach towards the omission of refer-</p><p>ential subject pronouns in earlier OHG is to trace it back to a special type of ‘loan</p><p>syntax’ (see Eggenberger 1961; Hopper 1975). The central idea is that the subject</p><p>omissions have been ‘imposed’ on these OHG translated texts through the process</p><p>of producing narrow – or even ‘slavish’ – translation from the Latin originals and</p><p>should not be considered a native part of the OHG grammar. As is argued in Axel</p><p>(2005, 2007: ch. 6) the loan-syntax hypothesis is not convincing. It cannot account for</p><p>the fact that null-subject distribution in the Old High German is clearly governed by</p><p>(morpho-)syntactic factors that do not play a role in the Latin originals. For example,</p><p>as has already been observed by Eggenberger (1961), there is a clear main/subordinate</p><p>asymmetry. While in main clauses subject pronouns are often omitted parallel to the</p><p>Latin, they are almost always inserted contrary to the Latin in subordinate clauses.</p><p>Axel argues that the main/subordinate asymmetry can be derived without further</p><p>assumptions if we assume that the OHG null subjects were only licensed in post-finite</p><p>position, see (9).11</p><p>(9) a. [CP Druhtini [C istj] pro auh ti tj ]</p><p>b. [CP Sumei [C hahetj] pro ti in cruci tj ]</p><p>c. [CP [C quidisj] pro zi uns thesa parabola tj ]</p><p>11.  It should be noted that – like modern German – OHG already largely exhibited the proper-</p><p>ties of an asymmetric V2 language: in subordinate clauses introduced by complementizers or</p><p>interrogative/relative phrases the finite verb stayed in its base position at the end of the clause</p><p>(the basic word order in the VP was OV). However, in (unintroduced) main clauses it moved</p><p>to C0, and in declaratives there was usually also XP movement to SpecC (cf.Axel 2007).</p><p> Katrin Axel & Helmut Weiß</p><p>Thus, just like in the modern dialects, OHG null subjects were only licensed in a con-</p><p>figuration where they were c-commanded by Agr in C. In OHG the only way to obtain</p><p>the configuration in (10) was via verb movement to C0:</p><p>(10) [CP [C V+AGRi ]k …. proi tk]</p><p>In OHG the phenomenon of complementizer agreement is not attested. Therefore,</p><p>the licensing condition of c-commanding agreement was not fulfilled in subordinate</p><p>clauses. As a consequence, subject pronouns were regularly inserted contrary to the</p><p>Latin originals:12</p><p>(11) Dhar ir auh quhad …, dhar meinida pro (I 274)</p><p>where he also said there meant</p><p>leohtsamo zi archennenne dhen heilegan gheist</p><p>easily to recognize the holy spirit</p><p>‘Where he said …, he clearly meant there the Holy Spirit’.</p><p>Item dicendo … sanctum spiritum euidenter aperuit</p><p>In contrast to the modern dialects, pro-drop is not restricted to sentences with verb</p><p>forms that show pronominal agreement. None of the finite verbs in the sentences in</p><p>(8) above exhibits a suffix of pronominal origin, yet the subject pronoun is dropped</p><p>in each case. Nevertheless, morphological factors do have a quantitative effect on null</p><p>subject usage in OHG: even though referential null subjects are attested in all per-</p><p>sons and numbers, it is only in the 3rd sg. and pl. that the null variant is used more</p><p>1.  As is outlined in Axel (2005, 2007: ch. 5), the main/subordinate asymmetry, of course,</p><p>also emerges if one looks at the OHG data from a quantitative perspective (the figures she</p><p>gives are based on Eggenberger’s (1961) quantitative analysis of null-subject usage in the</p><p>major OHG texts). In the OHG Isidor, the subject pronoun is dropped in 44% of a total of</p><p>109 main clauses with pronoun subjects. By contrast only 9% of subordinate clauses (#93)</p><p>do not have overt subject pronouns. In the Monsee Fragments even 64% of the main clauses</p><p>with pronoun subjects (#132) do not exhibit overt subject pronouns, as opposed to only 15%</p><p>of subordinate clauses (#86). In Tatian, 40% of main clauses with pronoun subjects (# 2394)</p><p>show pro-drop, but only 8% of subordinate clauses (#1275). Nevertheless, in all three texts there</p><p>is some amount of subject omission in subordinate clauses. Note, however that Eggenberger’s</p><p>figures for subordinate clauses also include dependent clauses with V2 order and examples</p><p>which are ambiguous between V2 and V-end order. We would like to claim that the remaining</p><p>really problematic cases (i.e. null subjects in the context of verb-end/late order) do not really</p><p>falsify the post-finiteness restriction, as they can still be due to translational errors, over-</p><p>sights etc. Also, we do concede that the Latin may have exerted some minor impact on the</p><p>OHG translation. Nevertheless null-subject usage in general is not an instance of systematic</p><p>‘loan syntax’ since there is a clear overall tendency that overt subject pronouns are inserted in</p><p>verb-end/late environments.</p><p>What changed where? 1</p><p>frequently than the overt one (Axel 2005, 2007: ch. 6). The reason for this person split</p><p>is still unclear.</p><p>In late OHG texts (i.e. in Notker’s works or in Williram von Ebersberg’s paraphrase</p><p>of the Song of Songs), pro-drop is attested only sporadically. Interestingly, null-subject</p><p>usage was quite frequent again in the Early New High German period.13 However, the</p><p>surface distribution of omitted subject pronouns then was somewhat different from</p><p>that of the OHG period. Instead, null-subject usage then already showed a similar</p><p>distribution to that found in the modern dialects. The syntax and morphology of pro-</p><p>drop in the dialects</p><p>is discussed in the following section.</p><p>.  From Old High German to the modern dialects</p><p>Now that we have discussed the synchronic distribution of pro in both OHG</p><p>and some modern dialects, the question arises as to what the diachronic develop-</p><p>ment looks like. From a diachronic viewpoint, at the dialectal level, the phenomenon</p><p>of pro-drop seems to be an example for continuity: both OHG and the modern dia-</p><p>lects allow for partial pro-drop.14</p><p>But the grammatical continuity is even more far-reaching. Interestingly, the</p><p>grammatical licensing conditions have remained the same: both in OHG and in some</p><p>modern dialects, pro is licensed by c-commanding Agr-in-C.</p><p>What must have changed is that the licensing conditions have become more</p><p>specific: only pronominal Agr-in-C has come to license pro-drop.</p><p>What has also changed is that there was a grammatical innovation – the develop-</p><p>ment of complementizer inflection – which cancelled the restriction to root sentences</p><p>so that pro-drop can now occur in embedded sentences also in the modern dialects.</p><p>The restriction to pronominal Agr is related to a further phenomenon, namely to</p><p>double agreement.</p><p>..1  The emergence of double agreement</p><p>Double agreement means that there exist two ways of inflecting the verb, and that the</p><p>choice of inflectional marker depends on the position of the verb. In modern Central</p><p>1.  Held (1903: 111) argues that the difference between Middle High German and Early New</p><p>High German is partly due to the fact that in the latter case, text genres had developed that</p><p>were closer to oral communication. Middle High German court language, by contrast, was</p><p>very remote from popular speech and strongly shaped by literary conventions.</p><p>1.  When speaking of continuity, it would, of course, be necessary to provide data from</p><p>Middle High German and Early Modern High German as well. Unfortunately, there are no</p><p>systematic studies on null subject phenomena at these stages. See, however, Held (1903) for</p><p>examples.</p><p> Katrin Axel & Helmut Weiß</p><p>Bavarian, for example, the ending in the 1st pl. is -n in verb-final sentences, but in</p><p>verb-second and verb-first sentences the suffix is -ma:</p><p>(12) a. dasma mia aaf Minga fahrn/*ma (Central Bavarian)</p><p>that-1.pl we to Munich go</p><p>‘that we go to Munich’</p><p>b. mia fahrma/*n aaf Minga</p><p>we go-1.pl to Munich</p><p>‘We go to Munich’</p><p>c. fahrma/*n mia aaf Minga?</p><p>go-1.pl we to Munich</p><p>‘Do we go to Munich?’</p><p>As can be seen in (12a), -ma is also the ending that appears on inflected complemen-</p><p>tizers. The suffix -ma is derived from the pronominal subject mia in its clitic form</p><p>ma: fahrn’ma > fahrm’ma > fahrma. In modern Central Bavarian, -ma is obligatory</p><p>even when C0 is filled with a complementizer instead of a verb, and it can combine</p><p>with the full pronoun mia. This shows that -ma is an inflectional suffix rather than a</p><p>subject clitic.</p><p>In German double agreement has a long history. A precursor can already be</p><p>found in OHG. In the Isidor translation, for example, the 1st pl. has two alterna-</p><p>tive endings: a short ending -m, which was later replaced by -n, and the older long</p><p>ending (-mēs):</p><p>(13) a. Endi dhes selben christes … chichundemes auh</p><p>and of-the same Christ demonstrate-1.pl also</p><p>nu dhes ędhili endi odhil (I 520)</p><p>now of-that age and home</p><p>‘we now demonstrate the lineage and homeland of that same Christ.’</p><p>et cuius … demonstretur et genus et patria</p><p>b. Chiuuisso chioffanodom uuir nu hear dhazs … (I 484)</p><p>surely revealed-1.pl we now here that</p><p>‘surely we have now revealed here that …’</p><p>Probauimus …</p><p>In OHG, the occurrence of -mēs correlates with pro-drop in verb-second contexts</p><p>where the subject pronoun would show up postfinitely if it was overtly realized (i.e. in</p><p>‘inversion’ contexts). But, in contrast to the modern dialects, the inflectional markers</p><p>are not dependent on verb position in OHG: both the long and the short ending occur</p><p>in second and in final position. This phenomenon does not yet constitute double</p><p>agreement in the sense defined at the beginning of this section.</p><p>However, soon after the replacement of -mēs, a new type of double agreement for</p><p>the 1st pl. began to arise, which was then characteristic for the Middle High German</p><p>What changed where? </p><p>period (cf. Paul 2007: §E21, 2). In Middle High German the 1st pl. ending was often</p><p>dropped completely. At the time only the -n-marker was used, as the long ending had</p><p>already died out. Interestingly, this -n-drop was now dependent on verb position/sub-</p><p>ject–verb inversion: when the pronominal subject directly followed the finite verb in</p><p>second position, the -n was dropped. A very early example can already be found in the</p><p>OHG Tatian:</p><p>(14) thesan niuuizuuuir uuanan her ist. (T 457,10)</p><p>this-acc neg.know.we whence he is</p><p>‘but as to this man we do not know where he is from.’</p><p>hunc autem nescimus unde sit.</p><p>A further interesting case is the 2nd sg. As mentioned above, this ending arose out</p><p>of a reanalysis of the word boundary. In ‘Christ and the Samaritan Woman’ (late 9th</p><p>century) we find two endings, the short ending -s, and the long ending -st whose addi-</p><p>tional dental plosive -t comes from the pronoun thu. Interestingly, in this text the old</p><p>short form occurs only when the subject pronoun precedes the verb:</p><p>(15) a. kerōst thu ‘chooses you’, gābist du ‘gave you’</p><p>b. thu wissīs ‘you know’, tu bātis ‘you begged’</p><p>(Christ and the Samaritan Woman; cited from Braune 2004: 261)</p><p>To sum up, the development of a second ending/of double agreement seems to have</p><p>depended on two issues: on V-to-C raising and on subject-verb-inversion (i.e. a pro-</p><p>nominal subject cliticizing onto the verb in C0).</p><p>..  Double agreement and the rise of pro-drop</p><p>What does double agreement have to do with pronominal agreement? After all, we</p><p>need an explanation for the innovation of pronominal agreement in order to describe</p><p>the diachrony of pro-drop in the dialects. The above-mentioned Isidor examples may</p><p>provide evidence for a putative diachronic relationship between the two phenomena:</p><p>(16) a. Endi dhes selben christes … chichundemes pro auh nu</p><p>(= 13a)</p><p>b. Chiuuisso chioffanodom uuir nu hear dhazs …</p><p>(= 13b)</p><p>Such minimal pairs, which must have occurred in spoken OHG as well, may have</p><p>conveyed the impression to the child learner that a verbal ending must contain</p><p>something special which enables it to license pro-drop and that this special prop-</p><p>erty is its pronominal origin. Already Paul (1877) analyses -mēs as deriving from a</p><p>personal pronoun. This assumption has turned out to be untenable, as later research</p><p>has shown. Nevertheless, it is conceivable that the pronominal nature of -mēs was</p><p>the result of a reanalysis of V-mēs as V-n+uuir in such minimal pairs: -mēs may have</p><p> Katrin Axel & Helmut Weiß</p><p>been eventually analysed as pronominal exactly in such contexts where it contrasts with</p><p>overt pronouns. Since only the long ending, but not the short one, was able to license pro-</p><p>drop, it may well have been the case that its pronominal nature was made responsible</p><p>for this.15</p><p>Additional evidence for this speculative scenario comes from two further pieces</p><p>of data: first, the hybrid ending -nmes which looks like a combination of the short and</p><p>long form (n + mēs), is also attested in Tatian (Franck 1971):</p><p>(17) uuanne gisahun uuir thih gast uuesentan (T 545,7)</p><p>when saw we you guest being</p><p>Inti gihalotunmes thih.</p><p>and took you</p><p>‘when saw we you a stranger, and took you in’</p><p>te uidimus hospitem & collegimus te.</p><p>Such forms have been analyzed as combinations of a verb regularly carrying the short</p><p>ending -n to which -mēs was added. If this was indeed the case, it would provide strong</p><p>evidence for the pronominal nature of -mēs in the eyes of speakers of OHG. As expected</p><p>(and necessary for our explanation), all seven cases of hybrid forms appear in verb-sec-</p><p>ond position in Tatian and all other texts where they are attested.</p><p>Second, we know from several Upper German dialects that the</p><p>w-onset of the</p><p>pronoun wir ‘we’ has indeed developed into an m in exactly those contexts where</p><p>the subject pronoun follows the verb in second position. This assimiliation may have</p><p>occurred already in spoken OHG (cf. Schatz 1907: 163).</p><p>To conclude: the reanalysis of the inflectional marker -mēs as pronominal may</p><p>have then had the consequence that the ability to license pro-drop became restricted</p><p>to pronominal agreement in C.</p><p>..  Conclusion</p><p>The phenomenon of pro-drop in German is characterized by continuity at the dialec-</p><p>tal level. The alleged loss of (referential) pro-drop has occurred only in the standard</p><p>language. At first sight, the surface distribution of (referential) pro seems to be entirely</p><p>different in OHG as compared to the modern dialects.</p><p>However, the underlying syntactic licensing condition is the same: pro is only</p><p>licensed if c-commanded by verbal Agr. What has changed is that the requirement on</p><p>1.  As one of the reviewers rightly objects, this explanation still leaves open the question,</p><p>why -mēs licensed pro-drop before it has been reanalysed as [+pronominal]. However, this</p><p>problem may be only an artifact of the sparse text documentation: it may be the case that the</p><p>reanalysis of -mēs had already begun at the time from which the oldest texts of OHG (like MF</p><p>and I) are attested.</p><p>What changed where? </p><p>Agr has become more specific: in the modern dialects only pronominal agreement has</p><p>come to be able to identify pro. We have argued that this change is probably related to</p><p>a further phenomenon, namely to the rise of double agreement.</p><p>Furthermore, the innovation of inflected complementizers, an independent phe-</p><p>nomenon, led to a drastic change in the surface distribution of pro, as the requirement</p><p>of a c-commanding Agr could then also be fulfilled in subordinate clause. It would,</p><p>however, be beyond the scope of this chapter to discuss the history of complementizer</p><p>inflection (but cf. Fuß 2005, 2008; Axel & Weiß to appear).</p><p>.  Further case studies</p><p>We have only been able to present one case study in more detail. But there are many</p><p>more cases where the question of dialect vs. standard has not received sufficient attention</p><p>in diachronic studies. We will briefly discuss three further cases where the develop-</p><p>ment at the dialectal level considerably differs from the development of the standard.</p><p>.1  Negation</p><p>A further case of more continuity at the dialectal level is negation. In Old and Middle</p><p>High German there was negative concord (NC); consider (18a) from the Middle High</p><p>German Prose Lancelot. Modern Standard German is not a negative concord language</p><p>any more, (18b), but at the dialectal level negative concord has survived as can be seen</p><p>in (18c) from Modern Bavarian.</p><p>(18) a. Da enwart nymand konig (Middle High German)</p><p>there neg-became nobody king</p><p>‘Nobody became king.’ (Lanc 10,9)</p><p>b. Ich habe kein Bier (*nicht) zuhause (Modern Standard German)</p><p>I have no beer neg at-home</p><p>c. I hon koa Bia ned dahoam (Modern Bavarian)</p><p>I have no beer neg at-home</p><p>‘I don’t have a beer at home.’</p><p>The difference between (18a) and (18c) indicates that there was a change from Middle</p><p>High German to the modern dialects as well: the negative particle ne/en vanished</p><p>and nicht (Bavarian ned), originally a negation strengthener, became the particle</p><p>expressing negation alone (Jäger 2008). But this change is only a superficial one in</p><p>two respects: (i) it has not altered the fundamental property of being a NC language,</p><p>and (ii) there was no accompanying change on the structural level. As Jäger (2008) has</p><p>shown, the syntax of negation in German can be analysed as involving a NegP structure</p><p> Katrin Axel & Helmut Weiß</p><p>throughout its history. The only change that has happened was that the lexeme origi-</p><p>nally occupying the Neg0 position was lost in Middle High German times (19):</p><p>(19) a. Da enwart nymand konig</p><p>b. NegP</p><p>Spec Neg′</p><p>nicht VP Neg0</p><p>ni/ne/ø</p><p>Note that continuity on the structural level also holds for Standard German (Jäger</p><p>2008), but this is somewhat concealed through the apparent loss of the NC property.</p><p>So the development of the dialects gives a much more adequate picture with respect to</p><p>continuity or change.16</p><p>.  Possessive constructions</p><p>A different case would be the development of possessive constructions. In this case</p><p>there is more continuity in the standard language: here, the post-nominal genitive has</p><p>survived, (20a). But in the dialects there emerged a new structure, namely a DP-internal</p><p>pre-nominal possessor construction.17</p><p>(20) a. das Haus des Vaters (Modern Standard German)</p><p>the house the-gen father-gen</p><p>‘the house of the father’</p><p>b. am Voda sei Haus (Modern Bavarian)</p><p>the-dat father-dat his house</p><p>‘the house of the father’</p><p>The case system of German developed in such a way that the genitive was lost first as</p><p>verbal case and eventually adnominally as well (Behaghel 1923). This development</p><p>1.  The case of negation in German is much more complicated than described in the main</p><p>text, since there is dispute in the literature whether German has lost NC (Jäger 2008) or not</p><p>(Weiß 2004b). A further point of discussion is the question as to whether or not nicht was</p><p>reanalysed as head or NegP (cf. Jäger 2008). However, independently of these issues, it is</p><p>undisputable that all the dialects did not lose the NC property and they thus exhibit obvious</p><p>continuity in a fundamental area of grammar.</p><p>1.  Some dialects have, however, retained the adverbial genitive in fossilized expressions</p><p>such as des Mittags ‘at noon’, des Abends ‘in the evening’ etc.</p><p>What changed where? </p><p>took place more than 500 years ago, but only in the dialects, whereas in the written lan-</p><p>guage the adnominal genitive survived (Weiß 2001). The case of the Standard German</p><p>genitive is an impressive example of the conserving ability of written varieties (even in</p><p>the absence of prescriptive grammars) (Weiß 2001, 2004a,b).</p><p>In the dialects there evolved a new possessive construction as in (20b) where the</p><p>possessor appears pre-nominally and is marked as dative instead of genitive (Weiß</p><p>2008). This development conforms with the overall development of the case system in</p><p>German with its loss of the genitive – though the emergence of the structure as such</p><p>was presumably independent of this overall development (cf. Weiß 2008). Therefore,</p><p>the possessive syntax of the standard with its continuity in the use of the genitive con-</p><p>veys a rather inadequate impression.</p><p>.  The consolidation of verb-final order</p><p>It is a widely-held assumption that ‘German’ underwent a development towards more</p><p>rigid final order of the finite verb, which amounts to the strengthening of the so-called</p><p>sentence bracket formed by the finite verb or complementizer in C (left sentence</p><p>bracket) and by the verbal cluster at the end of the clause (right sentence bracket). It has</p><p>been argued that there was a diachronic decrease of variation in the verbal cluster with</p><p>the effect that the finite verb became increasingly restricted to absolute final position.18</p><p>Again, if we take into consideration the situation in the modern dialects, a somewhat</p><p>different picture emerges.</p><p>In Modern Standard German, the word order in the verbal complex is very rigid in</p><p>that the subordinate verb has to precede the superordinate verb (e.g. V2 > V1, cf. (21);</p><p>V3 > V2 > V1; V4 > V3 >V2 > V1):19</p><p>(21) a. dass Maria ein Lied gesungen hat</p><p>that Mary a song sung2 has1</p><p>‘…that Mary has sung a song.’</p><p>b. … *dass Maria ein Lied hat gesungen</p><p>that Mary a song has1 sung2</p><p>‘…that Mary has sung a song.’</p><p>1.  Besides the change in verbal clusters, the sentence bracket was also consolidated by a</p><p>quantitative and qualitative decrease of extraposition patterns. We will not deal with this</p><p>phenomenon here, but note that there are suggestive findings reported in the secondary</p><p>literature that the contemporary dialects are more liberal towards what types of XPs can be</p><p>extraposed and exhibit</p>