Studies on German-Language Islands

Studies on German-Language Islands

Studies in Language Companion Series (SLCS) This series has been established as a companion series to the periodical

Studies in Language.

Editors Werner Abraham

Elly van Gelderen

University ofVienna

Arizona State University

Editorial Board Bernard Comrie

Christian Lehmann

Max Planck Institute, Leipzig and University of California, Santa Barbara

University of Erfurt

William Croft

Carnegie-Mellon University

University of New Mexico

Marianne Mithun

Osten Dahl UnivtYsity of Stockholm

Gerrit J, Dimmendaal

Brian MacWhinney

University of California, Santa Barbara

Heiko Nanog Tohuku University

University of Cologne

Johanna L. Wood

Ekkehard Konig

University of Aarhus

Free University of Berlin

Volume 123 Studies on German--Language Islands

Edited by Michael T Putnam

Studies on German-Language Islands Edited by

Michael T. Putnam The Pennsylvania State University

John Benjarnins Publishing Company Amsterdam/ Philadelphia

The paper used in tllis publication meets the minimwn requirements of American National Standard for Information Scie.t1~-:es Permanence of Paper for Printed Library Materials, ANSI z39.48-198-1.

Library of Congress C'.ataloging-in-Publication Data Studies on German-language islands I edited by Michael T. Putnam. p. em. (Studies in Language Companion Series, ISSN 0165-7763; v.12.3) Includes bibliographical references and in~x. German language--Syntax.

2..

Grammar, Comparative and general--Syntax._. Ger-

man language--Foreign countries. 4· German language--Variation. I. Putnam, Michae!T. PF5861.S68

2011

435--dc22 ISBN 978

90 272 0590 2 (Hb; alk. paper)

ISBN 978

90:1.7:1.8740 3 (Eb)

2010045325

© 2011- John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm. or any other means, without 'A-Titten permission fmm the publisher.

John Benjamins Publishing Co. · P. 0. Box 36224 • 1020 ME Amsterdam · 1he Netherlands John Benjarnins North America· P.O. Box 27519 • Philadelphia PA 19118-0519 • usA

Table of contents

Acknowledgements

VII

List of abbreviations

IX

List of contributors

XI

Why study Spra.chinseln from generative or structural perspectives? Introductory remarks

1

Michael r Putnam SECTION 1.

Phonetics & Phonology

On final laryngeal distinctions in Wisconsin Standard German

13

ReneeRemy Past participles in Mocheno: Allomorphy, alignment and the distribution of obstruents

33

Birgit Alber SECTION 2.

Morphology & texical studies

Plautdietsch gender: Between Dutch and German

Annemarie Toebosch Anaphors in contact: 'Ihe distribution of intensifiers and reflexives in Amana German

111

Michael r Putnam Lexical developments in Texas German Hans C. Boas & Marc Pierce Gender assignment of English loanwords in Pennsylvania German: Is there a feminine tendency? B. Richard Page SECTION 3·

1;1

Syntax I - Verb clusters

Synchrony and dial~hrony of verb dusters in Pennsylvania Dutch Mark L Louden

165

VI

Studies on German-Language Islands

Looking for order in chaos: Standard convergence and divergence in Mennonite Low German Gaz Kaufmann SECTION

4- Syntax II

'lhe syntax of Cimbrian German

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy- and what they (do not) betray about language universals and change under areal contact with Italo-Romance

233

Werner Abraham Diachronic clues to granunaticalization phenomena in the Cimbrian CP

279

Andrea Padovan Hidden verb second: The case of Cimbrian Gunther Grewendorf & Cecilia Poletto Revisiting the Wackernagelposition: The evolution of the Cimbrian pronominal system

301

347

Ermenegildo Bidese SECTION 5·

Syntax. Ill- 'Ihe syntax of Pennsylvania German

Changes in frequency as a measure of language change: Extraposition in Pennsylvania German

Jil

Gesche Westphal Fitch From preposition to purposive to infinitival marker: The Pennsylvania Germanfer... zu construction Kersti Borjars & Kate Burridge SECTION 6.

Pragmatics & Conversation analysis

Word choice, turn construction, and topic management in German conversation: Adverbs that are sensitive to interactional positioning

415

EmmaM. Betz Texas German discourse pragmatics: A preliminary study of the English-origin discourse markers of course, see, and now

455

Hunter Weilbacher Index

475

Acknowledgements

I would like to thank the series editors, Werner Abraham and Elly van Gelderen, f()r their interest in this project and for the invitation to submit this volume for review. I would also be remiss not to profusely thank and laud the efforts of the contributors of this volume; I am thankful for their patient support of this project and for their insightful contributions. Thanks are also in order to Vanessa Pearson tor her assistance in proofreading some of the contributions to this volume. I am also indebted to my wife, Jill Putnam, for her last-minute help in compiling the index. Lastly, although my interest in German--language islands extends as far back as my childhood, it was not until my graduate studies at the University of Kansas that I gained a true appreciation for the formal, systematic study of these Sprachinsellanguages. It was under the tutelage of Dr. William Keel, "The Kaiser;' that the exciting world of Sprachinselforschung was opened up to me in full bloom. Thanks for the advice and mentorship over the years and the many trips throughout the Midwest to visit and investigate the communities who still speak these Sprachinsellanguages. University Park, PA December 2010

List of abbreviations

ACC AE AG AM CA CP CD DAT DM DO DOM DP EPG EPP(-feature) FM FPP IDS

IM 10

IPP IR

Ll L2 LF LPG

MF

Accusative Case American English Amana German Agglutinative Morphology Conversation Analysis Complementizer Phrase Clitic Doubling Dative Case Discourse Marker Direct Object Dative-· object Differentiation Determiner Phrase Early Pennsylvania German Extended Projection Principle Fusional Morphology First Pair Part Institut fUr Deutsche Sprache Inflectional Morphology Indirect Object Infinitivus pro Participio Italo·· Romance First language Second language Logical Form Lexical Functional Grammar Middle Field

MHG

MLG MPG NHG

NOM NP OD OPS

OT OV OVE PF PG PL pp

PPT SG SGG SLA

SOV svo TCU TGDA TGDP TP TxG USPG vP

Middle High German Mennonite Low German Modern Pennsylvania German New High German Nominative Case Noun Phrase Other-directed predicates Oberdeutscher Priiteritumsscwhundl Upper-German Preterite Demise Optimality 1heory Object-Verb ordering Ottawa Valley English Phonological Fonn Pennsylvania German Palatinate German Prepositional Phrase Past Participle Standard Genna.tl Strong Gra.rnmatical Gender Second language acquisition Subject--Object--Verb ordering Subject-Verb-Object ordering Turn Construction Unit Texas German Dialect Archive 1exas German Dialect Project Tense Phrase Texas German United States Pennsylvania German light/little-vP, i.e., lower phase (vP-phase)

x

Studies on German-Language Islands

v2 v3 Vlast VO VOT

VP WCPG WGG

Verb-second language Verb-third language/ordering Verb-last Verb-Object ordering Voice Onset Time Verb Phrase Waterloo County Pennsylvania German (Canada) Weak Grammatical Gender

List of contributors

Werner Abraham

UniversW:it Wien Lindwurmstr. 120c D-80337 Munich Germany Email: wemer.abraham@t--online .de Birgit Alber

Facolta di Lingue e Letterature Straniere Universita di Verona via S. Fnmcesco 22 I-37134 Verona Italy Email: [email protected] Emma Betz

Kansas State University Department of Modem Languages 104 Eisenhower Hall Manhattan, KS 66506 USA Email: [email protected]

Kersti Borjars 'Ihe University of Manchester School of Languages, Linguistics & Cultures Room N 1.11., Samuel Alexander Building Manchester, M13 9PL, UK Email: [email protected]. uk Katie Burridge Monash University Linguistics Program School of Languages, Cultures and Linguistics Building 11 Victoria 3800 Australia Email: [email protected]. edu.au Gesche Westphal Fitch

Ermenegildo Bidese

Universita degli Studi di Trento, Facolta di Lettere e Filosofia ViaS. Croce 65 Italy Email: e.bidese@lett. unitn.it Hans Boas

Department of Germanic Studies 1 University Station C 3300, Burdine 328 The University of Texas at Austin Austin, TX 78712-0304 USA Email: [email protected]

Universitiit Wien Department for Neurobiology and Cognition Research Faculty of Life Sciences Althanstrasse 1.4 A--1090 Vienna Austria Email: [email protected] GUnther Grewendorf

Universitiit Frankfurt Institut ftir Kognitive Linguistik Griineburgplatz 1 D-60629 Frankfurt (Germany) Email: [email protected]· turt.de

XII

Studies on German-Language Island~

Gi.1z Kaufmann Albert- Ludwigs- Universitat Freiburg Deutsches Seminar Germanistische Linguistik Platz der Universitat 3 79085 Freiburg (Germany) Email: goez.kaufinann@gennanistik. uni-- freiburg.de Mark Louden University of Wisconsin Department of German 812 Van Hise Hall Madison, WI 53706 USA Email: [email protected] Andrea Padovan Universita degli Studi di Trento, Facolta di Lettere e Filosofia Via S. Croce 65 Italy :Email: [email protected] Richard Page The Pennsylvania State University Department of Germanic & Slavic Languages & Literatures 427 Burrowes Building University Park, PA 1.6802 USA Email: [email protected] Marc Pierce Department of Germanic Studies l University Station C 3300, Burdine 328 The University of Texas at Austin Austin, TX 78712-0304 USA Email: [email protected]

Cecilia Poletto Dipartimento di Scienze del Linguaggio Universita Ca' Foscari Venezia Ca' Bembo Dorsoduro 1075 30123 Venezia Italy Email: [email protected] Michael T. Putnam The Pennsylvania State University Department of Germanic & Slavic Languages & Literatures 427 Burrowes Building University Park, PA 16802 USA Email: [email protected] Renee Remy Fremdsprc:1cheninstitut der Landeshauptstadt Munchen Amalienstr. 36 80799 Munich (Germany) Email: [email protected] Annemarie Toebosch University of Michigan-- Flint Department of English 326 French Hall Flint. MI 48502 USA Email: [email protected] u Hunter Weilbacher University of Texas at Austin Department of Germanic Studies 1 University Station C3300 Burdine 382 Austin, TX 78712-0304 USA Email: weilbacher@mail. utexas .ed u

Why study Sprachinseln from generative or structural perspectives? Introductory remarks* Michael 'I: Putnam The Pennsylvania State University

1.

Introduction

German-language islands, also known as Sprachinseln, have been the focus of inquiry for centuries (cf. see Abraham 2009; Armand 1867; Csaba 2005; Haldeman 1872; Schmeller 1855/1984; Schweizer 1939; Schweizer 1952). The onset of the 21st century bears witness to the last potential (semi- )functional speakers of many of these dialect communities throughout the world. Given the moribund status of the majority of these dialect communities, it would seem that the time to conduct meaningful research on these dialects is now. To date, most volumes dedicated to the study of Spmchinsel languages have focused primarily (if not exclusively) on socio- and ethnolinguistic aspects of these dialect communities (ct~ Berend & Mattheier 1994; Keel & Mattheier 2003). The focus of this volume, however, centers on generative and structural shtdies of Sprachinsellanguages world"'ide, thus serving as a much--needed supplement to the majority of sociolinguistic research that exists on these dialects and dialect communities. In fact, to the best of my knowledge, this volume represents the first of its kind with a generative/structural focus. Approaching complex data from these diverse dialect-enclave communities through a theoretical framework enables linguists to better understand and model the status of the grammar of the speakers of these individual Sprachinsellanguages. In my brief introductory remarks here, my goal is to elucidate the importance of the studies found in this volume based on recent research in the fields of contact linguistics and L1 attrition.

*Thanks are in order to Werner Abraham, Elly van Gelderen, and Joe Salmons for helpful suggestions on how to improve this introduction.

2

Michael T. Putnam

2.

Answering the 'why question'

Although certain Spra.chinsel languages continue to exist - and in some cases tlourish due to the sectarian religious beliefs of the communities of speakers (with Plain Pennsylvania Gemtan being the most famous example of such a situation), this is unfortunately the exception mther than the norm. Most Spra.chinsei languages today are on the verge of extinction, with the majority of speakers exhibiting very limited knowledge of the dialect. 1 As mentioned in the introduction above, most studies to date have shown a trend to focus almost exclusively on the sociolinguistic aspects of the language-attrition situations currently taking place in these languageenclave communities. 1he past two decades, however, have witnessed the development of detailed studies of Sprachinsellanguages in a multitude of generative and structural frameworks which have paved the way to further inquiry concerning the underlying grammatical competence of the Ll in these decaying dialects. An example of such an investigation is Silvia Dal Negro's (2004) detailed study of the language decay under way in the Pomattertisch dialect an Alemannic (Walser) dialect spoken in Forrnazza, a relatively small, isolated Alpine community in Italy. Clearly, the status of

Speakers of these dialects have been referred to in lhe literature as 'seffii..speakers' (Dorian 1981), 'in(Oinplete acquirers' (Montrul 2002; Polinsky 2006), \mbalan(ed, dominant, or pseudo ..bilinguals' (Baker & Tones 1998), and 'early bilinguals' (Kim et al 2006). The exact cognitive status of the internal grammars ofthese speakers, however, is still a matter of serious debate. As pointed out to me by Werner Abraham (p.c.), Bidese & Padovan (2010) point out some salient trnits that the grammar of'semi-speakers' share with aphasic, pigeon, and code· mixing speech patterns. They report: 1.

Auch bei der Pidginisierungwirdgrammatische Struktur abgebaut,gleichzeitigjedoch entsteht neue Struktur, vor allem indem lexikalische Elemente grammatikalisiert wurden. Auch beim Code-mixi11g im bili11gualen Spracherwerb weisen Kinder eine sttirkere bzw. vollstiindige Kompetenz in einer Sprache auf, withrend die Kompetenz in der anderen Sprache zeitlich mangelhaft ist. Nach der "I-ry Hypothesis" bildet die starke Sprache die funktionalen (also die oberen) Strukturen, an die sich die lexikalischen der schwacheren anlehnen. (cf Bernardini & Schlyter (2004)). [As found in (linguistic) pidi.gini.zation, grammatical structure is decomposed while simultaneously new structure is built, most importantly in environments where lexical elements are grammaticali:z.ed. Also as evidenced in code-switdting in bilingual language acquisition, children display a stronger/complete competence for one of the languages while the competence for the other 'weak.er' language is temporally stunted or defective. According to "The Ivy Hypothesis," the stronger/dominant language serves a~ the ba~is for the functional structures, onto which the lexical elements from the 'weaker' language a~similate to. (translation by M. Putnam)]

\Vhy study Sprachinsein from generative or structuraJ perspectives?

particular linguistic aspects ofthese grammars (e.g. phonology, morphology, syntax, semantics, etc.) illuminate exactly which components of the grammar are more susceptible to contact-induced change and which are not. 2 Detailed studies of Sprachinsel languages such as those in this volume stand to increase our working knowledge of the (in)stability of particular aspects of these grammars in flux. Not only do these descriptive/empirical studies enrich our knowledge of these Spmchinsel-language grammars, but recent psycholinguistic research also confirms that these grammars occupy a special status in the mind and, as a result, are unique in many aspects from 'regular' monolingual grammars. For example, as highlighted by Schmid and Kopke (2007: 3), "only recently has it been recognized that bilinguals may not have one 'normal' language (in which they are indistinguishable from monolinguals the Ll in the case ofL2-learners and the L2 in the case of Ll attrit-ters) and the one 'deviant' one (in which knowledge is less extensive than that of monolinguals, and also tainted by interference from L1 in SLA and from attrition):' Psycholinguistic studies have demonstrated that bilinguals have an 'in-between' way of processing sentences (Hernandez, Bates & Avila 1994), of structuring their phonetic space (Cutler, Mehler, Norris & Segui 1989; Flege 1987), and exhibiting an 'always on-line' status for lexical items regardless of the language in use (Van I-fell & Dijkstra 2002). As pointed out by Schmid & Kopke (2007: 3), based on these findings "if traffic is always bi-directional [... ] then L1 attrition may not be the special and rare 'condition' as which it is often perceived?' Findings such as these suggest that although certain sociolinguistic situations must arise that lead to the proper environment for L1 attrition to set in;~ a unified account of the structural properties of the L1 grammars that have undergone this attrition is sorely needed. 4 Formalized theories of grammar competence and performance provide us with an opportunity to gain fllrther insight into the nature of these L1 attrition grammars. As demonstrated by the psycholinguistic studies cited above, the research questions that

2. Of particular interest and relevance here is Abraham's (this volume) study that presents strong arguments against the Double Base Hypothesis, arguing rather than many changes undetway in Cimbrian German dialects should be understood as an increase in the frequency of already existing forms in both Italo-Romance and German(ic) rather than mere 'borrowing' from the former.

3· Some more commonly cited sociolinguistic situations generaily held to function as a catalyst for language attrition are (but definitely not limited to): emigration, a shift in the Ll speaking community to exclusively adopt an L2 as the dominant language, lack of usage of the Ll in daily life, a significant decline in the duration of actively speaking the Ll (sometime involving decade-long pauses in action), etc. 4· As noted by Cook, the most likely scenario is one in which the acquisition of any L2 in the lifetime of an individual can have irrevocable effects on the L1 of an individual.

3

4

Michael T. Putnam

need to be asked at this stage in the development of Sprachinselforschung (i.e. languageisland research) could be enhanced greatly through the implementation of generative and structural approaches to human cognition and grammar. Note that the inclusion of generative and structural approaches into the f(mnal study of the properties of these endangered dialects does not exclude or trump the on·· going sociolinguistic and experimentallpsycholinguistic research that is under way in bilingual and language attrition studies. On the contrary, the inclusion of generative/structural studies into this growing domain of linguistic research has the chance of improving our ever·· expanding knowledge of these grdlllmars that is underway. In this regard, we can answer the 'why question: i.e. 'VVhy should we concern ourselves with theoretical and structural studies of tl1ese dialects' as follows: Studies such as these have the potential of not only enhancing the research under way in related fields (e.g. generative linguistics, applied linguistics, and psycho linguistics to name a tew), but their addition to the ever-growing research in contact linguistics and L1 attrition also has the potential of establishing a working framework where insights into the behavior of these language and dialects can be formally analyzed. The study of German-language enclaves (i.e. Sprachinsellanguages) provides us with an even more unique opportunity to see how one particular language - one, I might add, that has been the extensive source of rigorous linguistic study for centuries- behaves in a multitude of diverse environments. For some, their Sprachinsel language exi.sts as an Ll. variant in a stable bilingual situation (as is commonly the case for speakers of Plain Pennsylvania German); for others, it behaves as an L1 no longer actively used by a c.ommunity of speakers. Regardless of the diverse environments these Spmchinsellanguages currently are situated in, generative and structural stud-· ies of these variants will undoubtedly strengthen and expand our current knowledge of the internal grammar of these languages. In this regard, this volume represents a 'call to arms' for continued, extensive research of these Spmchinsellanguages. This 'call to arms' serves as the raison dHre for the existence of this current volume, namely, to provide a forum tor the on-going discussion and research of German-.Janguage enclaves that exist in a language contact situation where the German-language variant is the minority and that also by in large exists in a moribund state. In most of these communities, the last generation of 'semi-speakers' of these dialects under study are all that remain.

3·

Scope and content of the contributions

The various foci of study of these individual contributions investigate topics from the phonology of Mt'lcheno to the morphos};·ntax of Plautdietsc.h in Henderson, Nebraska,

\\l'hy study Spradtinsein from generative or structural perspectives?

to the stalus of verb-second in Cimbrian German (Northern Italy). As noted above, not only are these papers rich in interesting typological data, but they also make use of many contemporary theoretical frameworks such as Optimality Theory, the Minimalist Program, and Conversation Analysis. Concerning the areas of phonetics and phonology, i.e. SECTION 1.: PHONETICS & PHONOLOGY, Renee Remy investigates the phonetic and phonological properties of German 'final devoicin.g' (i.e. Auslautverhiirtung) from data obtained from Wisconsin German speakers. Specifically, Remy's study takes a closer look at laryngeal neutralization patterns in order to determine the status of final /b, d, g, zi consonants in Wisconsin German. Her study reveals that, although these dialect speakers of Wisconsin German do exhibit patterns of final laryngeal neutralization, these patterns diverge in interesting ways and in varying degrees of consistence from similar segments and environments found in modern High German. Birgit Alber's study engages in an in depth investigation on the formation of past par-ticiples in the Afocheno dialect (a Southern Bavarian dialect spoken in the language island Bersntol, also referred to as Fersentalerisch in German). Alber contributes to the debate of what determines allomorph selection by demonstrating that it is possible to analyze the process of past participle fi.)rmation in Mocheno as a process through which unmarked structures arise. Furthermore. Alber argues that it is optimal to analyze this process as one of optimalization since the constraint hierarchy responsible for allomorph selection subsumes two partial constraint hierarchies which are active in the language as a whole. The second section of papers, i.e. SECTION 2: MoRPHOLOGY & LEXICAL STUDIES, explores lexical and morphosyntactic issues related to Sprachinsel languages. Annemarie 1oebosch investigates the morphosyntactic/semantic notion of gender in the Plautdietsch dialect spoken in Henderson, Nebraska. Toebosch's study shows that this dialect exhibits a mixed semantic-grammatical gender system similar to that found in modern standard Dutch ·with regard to its semantic gender assignment involving pronouns, however with stronger grc:lffimatical gender in attributive targets and with semantic gender having spread to a wider nmge of targets in some speakers. Toebosch argues that the loss of gender inflection in Henderson Plautdietsch is, in fact, a gain in formal animacy features. The morphosyntactic/semantic properties of intensifiers and reflexives in the moribund Amana (Iowa) German dialect are the focus of Michael T. Putnam's contribution. His research documents the apparent shift in the inventory of Amana Germans intensifier and reflexive inventory evidenced in the dialect's morphosyntactic/semantic realization of recognizing two distributional types of transitive predicates; namely, typically 'other-directed' (e.g. [+On)) and typically 'selfdirected' predicates (e.g. [-Oo]). Hans Boas and Marc Pierce explore one of the most prominent aspects of contact languages, namely, the borrowing of lexical material. Their paper investigates the lexical borrowings in Texas German (TxG) to establish

5

6

Michael T. Putnam

the degree of lexical change that has taken place in this dialect over the past century. Richard Page's study on the assignment of gender to English loan words in modem Pennsylvania German closes out this section. Page's research reveals a tendency for the assignment of masculine gender to English loan words in Pennsylvania German. Section 3 (i.e. Sn..'TAX I VERB CLUSTERS) consists of two related and corn·· prehensive studies of verb clusters in Sprachinsel languages, namely, Pennsylvania German and Mennonite Low German (as spoken in six Mennonite colonies in the Americas). Mark Louden's el!:ploration of verb clusters in Pennsylvania German covers not only synchronic aspects of these vexing. complex structures, but also incorporates diachronic data to illustrate historical developments that have taken place in the development of Pennsylvania German. 1his predominantly descriptive study highlights these changes as well as the relative uniformity in variants of Pennsylvania German with regard to this construction. Goz Kaufmann's research exposes some of the ongoing convergence under way shaping the differences in final verb cluster structures in these Mennonite Low German dialects. Kaufmann's contribution combines work in structural linguistics with sociolinguistic variables (such as age and gender) to deliver a more accurate study of exactly which extra linguistic factors appear to be functioning as a catalyst for some of these observed structural changes. The syntax of Cirnbrian German, i.e. SECTION 4: SYNTAX II - THE SYNTAX OF CIMBRIAN GERMAN, serves as tl1e topic of discussion for the next (fourth) section of papers. Werner Abraham exposes a common held fallacy in the study of Sprachinsel languages; namely, the so-called Double Base Hypothesis, which holds that the mixture of structural properties of German and Italian in Cimbrian German is due to the century--long isolation of this language enclave being surrounded by an Italian majority. Abraham's research focuses on the exclusive orality of Cirnbrian and the specific structural changes oral-only variants are subject to as opposed to written, standard vernaculars. As a result, Abraham establishes a methodological nullhypothesis that change occurs under the narrowest accompanying structural conditions accessible within one and the same language rather than by borrowing from the social majority. Andrea Padovan investigates grarnrnaticalization phenomena within the Cirnbrian cornplernentizer layer (i.e. CP-layer), with a particular emphasis on the expansion of the Italian cornplernentizer ke. The central claim of his work from a structural perspective is that ke has not infiltrated the Cirnbrian CP-layer randomly but along a precise hierarchy from top to bottom. The work of Gunther Grewendorf and Cecilia Poletto continues with the theme of the CP-layer in Cirnbrian German. Grewendorf and Poletto restrict their study of Cirnbrian German to the variety spoken in Lusema showing that the language still can be classified as a V2-language. This is demonstrated by data involving subject (clitic) inversion in declarative clauses, a first position expletive pronoun similar to German es, and the typical asymmetry between main and embedded clauses (although restricted to one subclass of embedded clauses).

\Vhy study Sprachinsein from generative or structuraJ perspectives?

Ermenegildo Bidese's study of the development of the pronominal system in Cimbrian German closes this section. Accordingly, the syntax of Cimbrian object pronouns moves closer to that of the clitics of North Italy's Romance dialects, but without ever totally converging with the status of Romance exceptionless proclitics, due to the fact that in any stage of historical Cimbrian proclitics also can be shown to appear (based on Abraham's study from this volume). Section 5 continues with the theme of syntactic studies of aspects of Sprachinsel languages, focusing on Pennsylvania German (i.e. SECTION 5: SYNTAX III 'THE SYNTAX OF PENNSYLVANIA GERMAN). Gesche Westphal Fitch surveys the increasing frequency of the extraposition of adverbials (in particular, event-related adverbials) and prepositional phrases in Pennsylvania German. Incorporating data from Standard German, Palatinate German, and Pennsylvania German, Fitch demonstrates that although Pennsylvania German dearly shows stronger tendencies toward the extraposition of these elements, the mother dialect (e.g. Palatinate German) itself displays more extraposed forms than St1ndard German. Fitch's research shows a high frequency of eventrelated adverb placement that corresponds to an English surface order and a lower frequency of the order corresponding to unmarked German in Pennsylvania German also is presented. Based on this study, Fitch argues that Pennsylvania German displays frequency changes in which previously marked variants rise in frequency, corresponding to English surface structure, which is possibly facilitated by pre-existing tendencies in Palatinate German. Kersti Borjars and Kate Burridge chronicle the development of fer in Pennsylvania German from a benefactive to a purposive to a general infinitival marker in the variant of Plain Pennsylvania German spoken in vVaterloo Country, zu" clauses from adjunct to subject and Canada. Although the spread of the "fer complement functions is common to a number of varieties of Germanic, Bc>rjars and Burridge argue that the unique aspect of this change in Pennsylvania German has taken place at the same time as a related and pervasive change in the language, namely the loss of the infinitival marker zu. Tt) model the phenomenon of grammaticalization and the role of reanalysis, Borjars and Burridge couch their study in Lexical Functional Grammar (LFG). The final section of this volume, SECTION 6: PRAGMATICS & CoNVERSATION ANALYSIS, contains studies involving the discourse particle markers of Texas German (TxG) and position-sensitive adverbs in Siebenbiirger Sachsisch. Emma Betz engages in a study of position-sensitive adverbs in Siebenbiirger Sachsisch, couched within the Conversation Analysis framework. Her study investigates the lexical items ientz{t) 'now' and ientzter 'now' and demonstrates that in interaction, speakers use these variants as resources to mark the status of a turn construction unit or turn as complete or incomplete. Moreover, the selection and use of these two lexical items is sensitive to their placement within a larger sequence of conversation: Turn-final ientzter is tied to the management of topics in interaction, specifically to promoting a new action

7

8

Michael T. Putnam

trajectory. Finally, Hunter Weilbacher's study explicates a predominant characteristic of TxG speech, namely the prevalent use of English-origin pragmatic elements such as discourse markers (DMs) including well or you know, as well as other discoursepragmatic phrases or 'particles' from Engli.sh. Weilbacher reveals that the primary function of these DMs in TxG is that they appear to "lessen the load" by pragmatically indicating that the speaker is processing the upcoming utterance. In sum, the purpose of this volume is to expose various related research commnnities to under-researched aspects of these Sprachinsellanguages as well as to provide new perspectives on long-standing debates. With the ocean of new and interesting data couched within modern structural and generative frameworks, this volume hopefully achieves this preliminary goal and beyond.

References Abraham, Werner. 2009. (re'.'i.ew). Bruno: Schweizer: Zimbrische Gesamtgrammatik: Vergleichende Darstellung der zimbrischen Dialekte. James R. Dow (ed.). Zeitschrift ftir Dlalektologie und Linguistlk. Belhefte. 132(1): 247·· 257. Armand, Friedrich. 1867. Friedrlchsburg: Die Colonie des deutschen Fe.sten--verreins in Texas. Leipzig: F. Fleischer. Baker, Colin & Jones, Sylvia P. 1998. Encydopedla of Bilingualism and Bilingual Education. Clevedon: Multilingual Matters. Berend, Nina & Mattheie1; Klaus (eels). 199·1. Sprachinselforschung: Bin Gedenksc.hrift ftir Hugo Jedig. Frankfi.trt: Peter Lang. Bernardini., Petra & Schlyte1; Suzanne. 2004. Growing syntactic structure and code--mixing in the weaker language: The I\')' Hypothesis. Bilingualism: Language and Cognition 7: 49··69. Cook, Vivian 2003. The changing Ll in the L2 user's mind. In Fffects of the Sec.ond Language otl the First, Vivian Cook (ed.), 1···18. Clevedon: Multilingual Matters. Csaba, Foldes. 2005. Kontaktdeutscil. Zur Theorie eines Varietatentyps unter transkuiturelletl Bedingungen von Mehrsprachigkeit. Tiibingen: Gunter Narr. Cutler, Anne, Mehler, Jacques, Norris, Dennis & Segui., Juan 1989. Limits on bilingualism. Nature 340: 229-230. Dal Negro, Silvia. 2004. The Decay of a Language: The Case of a German Dialect In the Italiatl Alps. Frankfurt: Peter Lang. Dorian, Nancy. 1981. Language Death: The Life Cycle of a Scottish Gaelic Dialect. Philadelphia PA: University of Pennsylvania Press. Flege, John. 1987. The production of 'new' and 'similar' phones in a foreign language: Evidence for the effect of equivalence classification. Journal ofPhonetics 15: 47···65. Haldeman, S.S. 1872. Pennsylvania Dutch: A diaiect ofSouth German with an lnfosion ofAmerican Bnglish. Philadelphia PA: Reformed Church Publication Board. Hernande7., Arturo, Bates, Eli.1.abeth & Avila, Luis. 1994. On-line sentence interpretation in Spanish-English bilinguals: What does it mean to be 'in-between'? Applied Psycholinguistics 15:417-466.

\Vhy study Sprachinsein from generative or structuraJ perspectives? Keel, William & Mattheier, Klaus (ed~). 2003. Germar1 Lar1guage Varieties Woridwide: lr1temal and R~:temal Perspectives. Frankfurt: Peter Lang. Kim, Ji-Hye, Montrul, Silvina & Yoon, James. 2006. On the logophoric long-distance binding interpretation of the Korean local anaphor 'Caki-casin' by early bilinguaJs. Proceedings of the Armuai Boston Universit-v Coriferen.:e on Language Development 30: 305-315. Kopke, Barbara, Schmid, Monika, Keijzer, Merel & Dostert, Susa!l (eds). 2007. Lan,'5uage Attrition: Theoretical Perspectives [Studies in Bilingualism 33]. Amsterdam: John Benjamins. Montrul, Silvina. 2002. Incomplete acquisition and attrition of Spanish tense/aspect distinctions in adult bilinguals. Bilingualism: Language and Cognition 5: 39--68. Padovan, Andrea & Bidese, Ermeneglldo. 2010. Erodierte Sprachstrukturen und Grammatiktheorie: Zur Morphosyntax der Halbsprecher in der zimbrischen Enklave Lusern. Handout. Conference: DSWI Deutsche Sprachwissenschaft in Italien. Rome, Italy. Feb mary 4--6. Polinsky, Monika. 2006. Incomplete acquisition: American Russian. Journal of Slavic Linguistics 14: 161--219. Schmeller, Johann Andreas. 1855/1984. Die Cimbern der VII rmd XIII Communen auf venedischen Aipen und ihre Sprache. 1855: Weiner Akademieverlag. 1984: Munich, Curatorium Cimbricum Bavarense. Facsimiledruck der Ausgabe von 1855. Schmid, Monika & Kopke, Barbara. 2007. Bilingualism and attritio!l In Language Attrition: Theoretical Perspectives [Studies in Bilingualism 33], Barbara Kopke, Monika Schmid, Merel Kei)zer & Susan Dostert, 1---7. Amsterdam: John Benjamins. Schweizer, Bnmo. 1939. 2imbrische Sprachreste. Teil1: Texte aus Giazza (Dreizehn Gemeinden

ob Veronaj. Nach dem Volksmunde aujgenommen und mit deutscher Oberssetzung heraus-gegeben. Halle/Saale: Max Niemeyer. Schweizer, Bruno. 1952. 2imbrische Gesamtgrammatik, Band V: Syntax der zimbrlschen Diaiekte in Oberitalien. Diesser1 am Ammersee. Tippskript. Marburg/Lahn: Institut ftir Forschung der deutschen Sprache. Van Hell, Janet & Dijkstra, Ton. 2002. Foreign language knowledge can influence native language performance in exclusively native contexts. Psycho nomic Bulietin & Review 9( 4): 780--789.

9

SECTION 1

Phonetics & Phonology

On final laryngeal distinctions in Wisconsin Standard German* ReneeRemy Fremdspracheninstitut der Landeshauptstadt Munchen

Final laryngeal distinctions in the speech oftwo Eastern Wisconsin Germans provide insight into the German dialects spoken in Eastern Wisconsin, and how Wisconsin English has apparently begun to develop patterns of final fortition (Auslautverhiirtung). The speech samples show that both speakers differ from standard German and each other, with regard to laryngeally neutral consonants. In Manitowoc county, some voicing is present on underlying /s/ consonants, while underlying /b,d,g/ consonants remain almost exclusively laryngeally neutral. The Ozaukee county speaker shows neutralization across ib,d,g,z/ consonants v.ith predictable exceptions. Looking at Wisconsin German also helps to \Ulderstand American English as spoken in Eastern Wisconsin. Thus, final laryngeal distinctions are important for obtaining a clear picture of speech patterns used by bilingual Germans in Eastern Wisconsin.

1.

Introduction

This paper treats the phonetics and phonology of German 'final devoicing' - also known as Ausla.utverhiirtung, tinal fortition, or 'final laryngeal neutralization'. A set of debates continues over the status of neutralizations like that between Rat 'advice' and Rad 'wheef. both pronounced with final [t] - these include what features are involved and whether it is or can be 'complete'. Still, remarkably few data have been presented from free speech on this topic. I present such data here from recordings of two Wisconsin German speakers, speaking a variety of American Standard German. I look at laryngeal neutralization patterns in order to determine the status of final /b,d,g,z/ consonants in Wisconsin German.

•-I would like to thank Joshua Bousquette, Ryan Carroll, 'Iodd Ehresmann, Richard Page, Rose Rittenhouse, Blake Rogers, Joseph Salmons, Charlie Webster, and Stephanie Yau for their comments and suggestions. Thanks also to Kevin Kurdylo of the Max Kade Institute for assistance with the recordings. 'Ihe usual disclaimers apply.

14

Renee Remy

I consider it possible that the laryngeal phonetics of final consonants in Wisconsin German will differ from what would be expected in Standard German. Moreover, even ifboth speakers neutralize final consonants laryngeally, the neutralization pattern may not be consistent between these two dialect speakers. These possibilities are supported by the current results. This may have several explanations, and I explore two. First, German is a pluricentric. language including mutually unintelligible dialects. Dialect speakers often learn High German in school, but do not necessarily rid themselves of dialectal characteristics, including the phonetics of final laryngeal distinctions. That is, the lack of complete neutralization may simply continue the earlier varieties imported from Germany. Second, in multilingual settings like those of German--speaking communities, language contad can lead to reshaping of the dialectal variety as well as the majority language in the area. That is, German L1 speakers who are bilingual in English and use English much more than German may show phonetic effects on their German. This second point is of particular interest, given evidence that some Wisconsin English speakers appear to be developing patterns of final laryngeal neutralization. In the rest of the paper, I will discuss previous research on the topic of final laryngeal neutralization, in §2, and introduce the present dataset in §3. I present the results of a small perceptual study in §4 and acoustic analysis in §5. §6 outlines a possible connection between German and Eastern Wisconsin English, while §7 concludes.

2.

Background and rationale

Before continuing, note that the traditional English term 'final devoicing' is problematic in itself for languages like German. 1hat term does not reflect the phonetic and phonological differences between twu types of languages. Some, like Romance, Slavic or Dutch, rely heavily on phonetic voicing that is, actual glottal pulsing to distinguish /dl from It/ and so on. Others, like most of Germanic, Somali, Kashmiri and many others, mark their 'voiceless' stops by aspiration in prosodically prominent positions, and do not use phonetic voicing as a central signal to their laryngeal distinctions. Under this view of 'laryngeal realism' (Honeybone 2005), the term final fortilion (Iverson & Salmons 2006) allows for a more precise and accurate characterization for final laryngeal neutralization in 'spread glottis' languages like German or English, while 'final devoic.ing' is reserved for 'voice' languages like Dutch. 'Ihe complete vs. partial status of phonetic and phonological neutralization has been discussed in a long string of previous studies, many focused specifically on German final obstruents (Inozuka 1991; Port & Crawford 1989; Houlihan & Iverson 1979; Kiparsky 1976). These have often compared words in carrier phrases (i.e. "Please say for me"), with the result that participants became aware of the nature

On final laryngeal distinctions in Wisconsin Standard German

of the experiment (Piroth & Janker 2004). 1 Another problem in determining what is complete vs. partial 'final laryngeal neutralization' is real world variation - it is dear that not all speakers of'Gennan' have neutralization, let alone complete neutralization, and such patterns need to be documented as part of the work in this arena. Pursuing the idea that sound change is a primary shaper of synchronic phono-logical patterning, Blevins (2004) shows that no single process leads to 'final devoicing' and that very few processes, when combined, consistently lead to 'final devoicing: For the case of German and its dialects, Mihm (2004) discusses the history starting with Middle High German (MHG) regional variation. He divides the German-speaking region into t'i.v-o sectors: the "northwesf' and the "southeasf'. Data in the current study come from a speaker with ancestry in the Kiel area (which lies in Mihn1s "northwesf' sector) and Pomerania (which does not fit directly in Mihm's sectors, although still exhibits typically northern characteristics). Mihm showed that the symmetry with p losives and fi·icatives in the sound system was not as clean in the "northwesf' as in the "southeast:' The data in the current study are consistent with Mihms analysis because some of the recording samples show final fortition across the board (with fricatives) even though other plosives are less consistent across regional recordings. Complete neutralization of [b,d,g] to [p,t,k] in New High German (NHG) is actually quite low (only 31% according to Mihm 2004: 185). Partial neutralization is found more often, showing up in some environments but not in others. Before moving on, note that fortition may be arising today in some varieties of American English, at least those with heavily German roots. According to Purnell et. al. (2005) speakers in a formerly German-speaking Wisconsin town produce final laryngeal distinctions differently in their spoken English than speakers of most other varieties of American English, showing partial laryngeal neutralization, especially of /s/ versus /z/. It is, in other words, possible that German speakers have contributed to the development of modern American English in the Wisconsin area. I will return to this in §6. As noted, many aspects of German final laryngeal neutralization remain unstudied and we lack information on the status of the process in many varieties of German, especially extraterritorial ones. A first goal, then, is simply to broaden the empirical foundation on the topic, drawing data from extraterritorial varieties. With that, let us turn now to the present study.

3·

The present study

The current study examines, as indicated, final tortition (.Auslautverhiirtung) among speakers of German living in Eastern \Visconsin. My data come from two heritage

Standard German speakers are clearly sensitive to spelling, and German-English bilinguals, 1. like the subjects in Port's work, can often produce such distinctions.

15

16

Renee Remy

speakers of what has been called 'American Standard German' (Salmons & Lucht 2005), one each from Manitowoc and Ozaukee counties. Both are from communities that spoke Low German natively but which used a standard-like variety in a full range of institutions, including schools and churches. 'Ihe Manitowoc and Ozaukee recordings were both recorded in 1968, and the language ofboth the interviewer and interviewee was (High) German. The Manitowoc sample is free speech by a female speaker whose German ancestors were ti·om a small area west of Kiel, Germany. The topics of the 27 minute recording include the occupations and lifestyle of her parents and grandparents, as well as her family's Christmas traditions. The Ozaukee sample is 29 minutes of free speech by a male speaker, with ancestry in Pomerania, about his grandfather and about various German societies in Milwaukee. There are, of course, advantages and disadvantages to using free speech samples. Although one cannot control for phonetic environment or number of tokens produced free speech recordings yield more natural production results than the randomized token words in carrier phrases used in previous studies, a problem already noted in Port's work above. In the present situation of advanced language loss in the community, studying free speech in older recordings is advantageous because newer recordings may show more attrition. The major disadvantage is naturally that we do not have directly parallel forms, minimal pairs like Rat vs. Radin the same phonetic context. But ultimately, this was the only approach possible given tlte data source. Finallary11.geal neutralization is a general trend in the data (especially with fricatives), but the samples show differences in the frequency of neutralized plosive consonants. The fact that German, including its northern spoken varieties, is a spread glottis language looks to be consistent with the Wisconsin German data. The data seem to shed light on specific dialectal differences, which are integrated into the standard variants. Some code-switching between Gennan and English can also be heard in the recordings. The tokens chosen were words ending in /b, d, g, z/. If these speakers were following standard High German patterns, one would expect all final obstruents to undergo tortition. I turn now to the results of a small perceptual study ainled at testing whether final obstruents in the speech of these two individuals are heard by American English speakers as voiced or voiceless.

4·

Perception of Wisconsin German final obstruents

Individual words from the recordings used in this study were played for six American respondents in order to determine their perceptions of the voiced/voiceless nature of the final obstruent~. I chose American respondents because a voicing contrast for final obstruents is found in most varieties of American English. It is important to know how these sounds are perceived and not just how they look acoustically. I assumed that German /p,t,k/ would be heard as [p,t,k], but underlyingly voiced consonants

On final laryngeal di.~tinction.~ in Wisconsin Standard German /b,d,g,:d would possibly be variable in terms of being perceived as voiced or voiceless. Therefore, I only chose to look at /b,d,g,v' and Is! consonants, where there were good examples in the data. Respondents heard each word twice and were asked to write down whether they heard the voiced or voiceless variant of the final obstruent. For example, subjects were asked: "is the final consonant a [b] or [p]?" The written words were not seen by the respondents. However, all respondents knew basic German and occasionally recognized certain words. Not all words analyzed in the section of phonetic data were used in perception testing. Words for perception testing were chosen based on their envi.-· ronment, so that (when possible) at least one final consonant followed by a voiced obstruent, a voiceless obstruent, and a pause were present for each consonant group (ip,bi, /t,di, /k,gi, /s,'ll respectively). Words with high and low glottal pulsing percent· ages were also chosen, in hopes of presenting a broad sample of tokens to respondents. The results were then combined according to the aforementioned consonant groups and compared across respondents. Each perception data table lists the word followed by the number of respondents who heard the final consonant as voiced or voiceless. The characterization of the first sound of the following word in the recording is listed in parentheses after the token: D = voiced obstruent; T = voiceless obstruent; N = nasal; pause = phrase final/pause in speech. Words that occurred more than once in the speech sample are numbered.

Perception of Manitowoc (speaker with ancestry near Kiel) final obstruents

4.1

The Manitowoc speaker produced final consonants that were fairly easily identified by respondents. Although each respondent heard each word twice, many respondents were able to identity the final consonant the first time they heard it. Table 1. Perceptions of Manitowoc German Underlying Final/sf Word

[s]

[z]

alies 1 (T) alies 2 (D) als (D)

6/6 3/6 5/6 2/6 4/6 5/6 5/6 2/6 5/6 5/6 6/6 2/6

0/6 3/6 1/6 4/6 2/6 1/6 1/6 4/6 1/6 1/6

hat~(D)

damals 1 (D) damals 3 (D) gab~ gab~

1 (D) 2 (D)

gab~

3 (D)

damals 2 (pause) ausgezeichnetes (N) Gans (1\1)

Oi6 4i6

17

18

Renee Remy

The vast majority of these underlying final/sf tokens were heard as /s/ (see Table 1). Generally speaking, tokens followed by 'D' were heard most often by five of the six respondents as I sf. Tokens followed by 'T were always heard as I sl, and tokens followed by 'pause' were heard by five of the six respondent~ as Is!. Tokens followed by 'N' were heard on average by four of six respondents as /s/ and two of six respondent~ as /'li. However, only three tokens 'gab~ 2: 'Cans: and 'hatr were heard more often as l7l than/sf by the respondents.

Table 2. Perceptions of Manitowoc German Underlying Final/z/ Word

[s]

[z]

logs(D)

6/6

0/6

This token was agreed upon by all six respondents as sounding like isi, even though it is an English word. The final consonant should be /z/ in American English, though this did not seem to lead respondents to call it/z/.

Table 3. Perceptions of Manitowoc German Underlying Final/g/ Word

[gl

[k]

logs (T)

6i6

0/6

All six respondents also agreed that the /g/ in 'l.ogs' was in fact a /g/ and not a /k!. 'lhis response conforms to American English norms.

Table 4. Perceptions of Manitowoc German Underlying Final/d./ Word

[d]

[t]

Hund (p) Pferd(D) Deutschland (T) secondhand (N)

0/6 1/6 6/6 6/6

6/6 5/6 0/6 0/6

Only two of the underl)'ing final idi tokens were unanimously deemed /d/. The other two were overw·helmingly heard as It/. Final /di tokens followed by 'D' were very often heard as /t/, as only one person reported hearing a /d/. Tokens followed by both 'T and 'N' were heard as /d/ by all respondents. 'Secondhand' is of course a loanword from English, which could explain why all respondent~ heard the final consonant as I d!. 'Ihere was more agreement here than there seems to be among the other final consonants.

On final laryngeal distinctions in Wisconsin Standard German

Table 5. Perceptions of Manitowoc German Underlying Final/b/

Word

[b]

[p]

gab (D) gab~ 1 (T)

6/6 5/6 2/6 5/6 1/6 5/6 5/6

0!6

gab~

2 (T) gab~ 3 (T) hab (N) ichhab~)

hab 1 (pause)

1!6 4/6 1!6 5!6

1!6 1!6

Five out of six respondents, in the majority of cases, he-ard tokens followed by 'T' as /b/. One token was perceived as /b/ and one as !pi in tokens followed by 'N: The token followed by 'pause' was heard by five out of six respondents as /b/, and the token followed by 'D' was heard by all respondents as /b/. Surprisingly, only two of the tokens in 'Iable 5 'gab's 2' and 'hab' were heard as !pl. This finding does not conform to Standard German norms, in which all final!b/ consonants should be heard as /p/. Table 6. Perceptions of Manitowoc German Underlying Final fbt/ Cluster

Word

[b]

[p]

erlaubt (I')

4/6 1!6 3/6

2/6 5/6 3/6

gehabt 1 (I') geiiebt(T)

Respondents exhibited widespread disagreement on whether they thought these final consonants clusters were /btl or /pt/. The 'erlaubt' token, however, had the most respondents hear the final/btl duster as /btl. 4.2

Perception of Ozaukee (speaker with ancestry in Pomerania) final obstruents

1he Ozaukee speaker produced final consonants that were difficult t{)r some respondents to identify. 1hus, respondents often did not agree with each other in their perceptions of these consonants. Table 7. Perceptions of Ozaukee German Underlying Final/sf

Word

[s]

[z]

alles (D) alles 2 (stress-D)

4/6 4/6 1/6

2/6 2/6 5/6

Dletze~(D)

19

20

Renee Remy

The final consonants in these tokens, all followed by 'IY, were primarily heard as Is/. The last token was heard as /z/ by all but one respondent, possibly because of the /z/ consonant that shows up towards the end of the word. Table 8. Perceptions of Ozaukee German Underlying Final lzfl Word

[s]

[z]

Klubs (D) Klubs2 (D)

·1/6 3/6

2/6 3/6

The respondents do not agree on the voiced/voiceless nature of the final consonants in the tokens with underlying final/zl (see Table 8).1he final consonant in the first token seems to have been heard as lsi by the majority of respondents, but there is no consensus for the second token. One would expect respondents to hear these tokens with final /z/ because of the 'D' in the following word, however many respondents report hearing Isl. Table 9. Perceptions of Ozaukee German Underlying Final/gl Word

[g]

[k]

fiinftig (D) genug(T)

4/6 0/6

2/6 6/6

Respondent~ marked the final consonant in the first token to sound primarily like a /gl, and the second token overwhelmingly like a /k/. This is possibly due to the sound that follows at the beginning of the next word in the recording (I will elaborate on this point in the section of glottal pulsing data).

Table 10. Perceptions of Ozaukee German Underlying Final/d/ Word

[d]

[t]

Deutschland 3 (pause) Kind (pause) Sc1zwarzwald (pause) Mitglled (T) Deutschland (D)

2/6 0/6 6/6 4/6 6/6

4/6 6/6 0/6

2/6 0/6

2. The word 'Klubs' may be d1aracterized as having an underlying final/z/ if it is a phonologically unintegrated loan word. It of course would have final [z] in English.

On final laryngeal distinction~ in Wisconsin Standard German

Underlying final/d/ consonants followed by 'pause' were heard on averc:tge by two of six respondents as/d/ and three of six respondents as/t/. Tokens followed by 'T and 'D' were heard as /d/ by four out of six and sil!: of six respondents respectively. Two final consonants were unanimously heard as /d/. Table ll. Perceptions of O:r.aukee German Underlying Final/b/ Word

[b]

[p]

Klub 1 (T) Klubs 1 (T) Klubs2 (T) giaub 1 (D) giaub 3 (D) Klub2 (D) Klub 5 (D) Klub 3 (pause) Klub 4 (pause)

1/6 4/6 3/6 5/6 1/6 5/6 6/6 4/6 4/6

5i6 2i6 3i6

1i6 5i6 1/6 0/6 2/6 2/6

Six out of nine tokens were heard by the respondents as having finalibi. 'Ibkens followed by 'D' were heard most often by five out of six respondents as /b/. Underlying final /b/ tokens followed by 'T' were heard more often as /p/, although many respondents heard them as /bi. When followed by 'pause' tokens were heard by four of six respondents as /b/. There was much more disagreement among the rest of the tokens than seen in the I d,g,z/ final consonants. 1he results of the small perception test show that respondents had more trouble deciphering whether or not certain consonants were voiceless or voiced in the Ozaukee tokens than in the Manitowoc tokens. 'Ihe Manitowoc speaker produced finallaryn·· geal obstruents that were interpreted by American listeners as definitely voiceless or definitely voiced. Respondents were not left guessing as to voiced/voiceless nature of the final consonant. With the Ozaukee speaker, respondents showed more instances of widespread disagreement. 'Ihese findings suggest that final laryngeal obstruents are not neutralized consistently within the same speaker and that the Ozaukee speaker may have been influenced by knowledge of and proximity to Wisconsin English. 1hus, the frequent use of Wisconsin English and infrequent use of German (as well as possible atrophy of German) may have led to such inconsistencies. Respondents had little trouble hearing when certain consonants were voiced or voiceless in the Manitowoc tokens. Here the majority of the respondents tended to agree ·with one another almost unanimously on the nature of the final consonants, especially with fricatives. Final /b/ consonants were heard as voiced by the majority of respondents, though the glottal pulsing data does not support this finding. Although the respondents' perceptions did not match the glottal pulsing data for final /b/, words

21

22

Renee Remy with high percentages of glottal pulsing (see phonetic data) still seemed to be heard by the majority of respondents as voiced, and words with low percentages of glottal pulsing as voiceless. Once again, the influence of German on Wisconsin English and Wisconsin English on German may be the reason for the agreement between what respondents perceived and the actual phonetic data

5·

A(:OUSti(:

analysis

After gathering perception data for these two German speakers, sound files from these interviews were analyzed using Praat Duration of final obstruents was measured and the amount of voicing during closure was then used to calculate the glottal pulsing percentages. This paper assumes neutralization to be presentin final laryngeal obstruents when the glottal pulsing percentage is below 50%. The duration of the final consonant (in seconds) was taken from the diagram (as shown below in the word 'hab') as well as the amount of voicing during closure. The top of the diagram shows the sound waveform and the bottom the spectrogram. The dark bands on the spectrogram indicate voicing. The voicing during closure divided by the duration of the final consonant yields the percentage of glottal pulsing. Higher percentages of glottal pulsing indicate a voiced consonant and lower percentages a voiceless consonant Example from Praat [0.029]

[ 0.097]

'hab' - Final/b! expected [p ]: 0.029 seconds with 0.097 seconds voicing during closure, thus 30.0% glottal pulsing.

On final laryngeal di.~tinction.~ in Wisconsin Standard German 5.1

Data from Manitowoc (speaker with ancestry in Kiel)

The speaker with ancestry in the Kiel area (the Manitowoc speaker) shows some voicing on underlying /z/ consonants, while underlying /b, d, gl consonants remain almost exclusively laryngeally neutr.U. The speaker with ancestry near Kiel produces many tokens with underlying final/sf consonants (see ·ntble 12). As noted above, the characterization of the first sound of the follmving word in the recording is listed in parentheses after the token: D = voiced obstruent; T = voiceless obstruent; N = nasal; pause == phrase final/pause in speech. Words that occurred more than once in the speech sample are numbered. The word is listed in the first column, followed by the percentage of glottal pulsing or voicing carryover (GP) in the second column, and the length (seconds) of the final consonant in the third column. Table 12. Manitowoc Tokens with Underlying Final/sf Length (sec.)

Word

GP%

alles 1 (T) alles 2 (D) alles 3 (T) als (D) ausgezeichnetes (N) bis..<;ehen antlers (T) damals 1 (D) damals 2 (pause) damals 3 (D)

Nachtisches (pause) namens (T)

0.070 0.181 0.185 0.811 0.605 0.312 0.427 0.274 0.600 0.220 0.177 0.400 0.571 0.714 0.635 0.263 0.250

0.272 0.158 0.075 0.053 0.212 0.170 0.175 0.187 0.151 0.081 0.122 0.105 0.137 0.066 0.164 0.192 0.120

Average (all) Average (T) Average (D) Average (N) Average (D + N) Average (pause)

0.392 0.205 0.439 0.597 0.482 0.268

0.144 0.159 0.114 0.171 0.129 0.190

gab~

1 (D) (D) gab~ 3 (D) gab~2

Gans(N) hat's (D) Mittags(N)

Tokens followed by a voiced obstruent have a much higher percentage of glottal pulsing and thus are less devoiced or fortis than tokens followed by a voiceless obstruent or a pause. Looking at the averages, the lowest percentage of glottal pulsing is

23

24

Renee Remy

found in the category final/ s/ consonants followed by 'T'. Those followed by a 'pause' have only a slightly higher percentage of glottal pulsing. Final/ s/ consonants followed by 'D' have more than double the percentage of glottal pulsing found in final i si followed by 'T' The only category producing an average percentage of glottal pulsing high enough to be considered voiced was the tokens followed by 'N'. 'Ihe speaker seems to be anticipating the voicing to come and thus exhibiting a shorter voice onset time (VOT) for the following word and a longer voicing carryover on the final consonant Some tokens occur several times with different glottal pulsing percentages each time. For example the word 'gabS' occurs three times, the first two times with a relatively low percentage of glottal pulsing and the third time with a slightly higher percentage. All three tokens, however, are still examples of final laryngeal neutralization. The only tokens where /s/ is produced in a voiced form are tokens which precede a voiced obstruent. 'This pattern is variable within a single speaker. Not all final consonants followed by 'D' are voiced (see Table 12). But, all final consonants followed by 'T' or a pause are laryngeally neutral.

Table 13. Manitowoc Token with Underlying Final/z/ Word

GP%

logs (D)

0.400

Length (sec.) 0.100

One token from the speaker ·with ancestry near Kiel has an underlying final/z/ consonant (see Table 13). This final consonant should exhibit a high percentage of voicing because of the voiced consonant that precedes it and the voiced consonant that follows, but it does in fact prove to be laryngeally neutralized with less than 50% glottal pulsing. The length of the final consonant does not seem to affect the amount of glottal pulsing produced. Given that 'logs' is an English token used in rapid German speech, one might expect the speaker to apply German phonological norms to this token. It is however possible that this speaker applied Wisconsin English norms to English words even when speaking German, due to the fact that this speaker was raised bilingually in Wisconsin. Purnell et. al. (2005) found that German speakers living in a formerly German speaking Wisconsin town show partial neutralization of final obstruents when speaking English. Thus, this speaker produced a percentage of glottal pulsing closer to what would be found in English (i.e. a higher percentage of glottal pulsing than what would be found in Standard German, but still laryngeally neutral). In Stan·· dard American English, the plural and possessive .. s, typically analyzed as /'1/, surfaces as [z] follo\\
On final laryngeal distinctions in Wisconsin Standard German

Wisconsin English speakers, however, tend to produce these as voiceless [s] regardless of the character of the preceding sound.

Table 14. Manitowoc Tokens with Underlying Final /g/ Word

GP%

Length (sec.)

gesagt(T) logs (T) sagt(T)

0.000 0.140 0.000

0.078 0.074 0.100

Average (all)

0.047

0.084

The speaker shows complete neutralization in the (gt) final consonant cluster (see 'Iable 14). The (g) in 'logs' does exhibit some voicing, once again due to the fact that it is an English token. All three tokens exhibit final laryngeal neutralization. Table 15. Manitowoc Tokens with Underlying Final /d/ Word

GP%

Length (sec.)

Huml (D) Pferd (D) secondhand (N) Deutschland (T)

0.333 0.400 0.500 0.220

0.060 0.046 0.043 0.047

Average (all) Average (D) Average (D + N)

0.363 0.367 0.411

0.049 0.053 0.050

The tokens with underlying /d/ consonants show higher percentages of glottal pulsing than would be expected due to the fact that almost every occurrence is followed by a voiced consonant (see Table 15). All of the averages for final /d/ are very close together and all show final laryngeal neutralization. The only token followed by a voiceless consonant does in fact show a lower percentage of glottal pulsing and would detl.nitely be considered laryngeally neutral. English words used by this speaker tend to have a higher than expected glottal pulsing percentage. In this subject's speech, English loanwords also tend to follow patterns of American English voicing. That being said, the word 'secondhand' has a higher than expected percentage of glottal pulsing for German /d/, but is still not as high as it would be in Standard American English. Three out of four tokens fall into the realm of being laryngeally neutralized.

25

26

Renee Remy Table 16. Manitowoc Tokens with Underlying Final /b/ Word

GP%

Length (sec.)

gab (D)

!chhab~)

0.524 0.469 0.333 0.300 0.286 0.300 0.550

0.021 0.032 0.037 0.100 0.068 0.029 0.011

Average (all) Average (T) Average (N) Average (D + N)

0.394 0.367 0.416 0.452

0.043 0.056 0.040 0.033

gab~ 1 (T) gab~ 2 (T) gab~ 3 (T)

hab (N) hab 1 (pause)

When looking at the glottal pulsing percentages for each 'gab's' token, the per·· centages fall under 50%, thus the 'gab's' tokens show at least partial final laryngeal neutralization (see Table 16). This may contribute to the mostly voiceless nature of the following I sf consonant. 1okens ending in /bi show considerable variation on glottal pulsing percentages. lhe averages for final/b/ consonants show the lowest percentage of glottal pulsing in final consonants followed by 'T' or a 'pause'. Tokens followed by 'D' or 'N' have on average 8.5% more glottal pulsing than consonants followed by 'T: The length of the consonant itself does not seem to be a predictor of final laryngeal neutrali1.ation, however the presence of'D' or 'N' following tends to yield higher percentages of glottal pulsing, sometimes even to the point of definite voicing. Five out of seven tokens show final laryngeal neutralization. Table 17. Manitowoc Tokens with Underlying Final !btl Cluster Word

GP%

Length (sec.)

eriaubt (T) gehabt 1 (T) gehabt2 (T) gehabt3 (T) gehabt4 (T) gehabt5 (T) gehabt6 (T) geliebt(T)

0.330 0.000 0.000 0.000 0.000 0.000 0.000 0.000

0.030 0.240 0.203 0.113 0.081 0.084 0.114 0.090

Average (all)

0.041

0.120

Similar to the (gt) cluster, the final consonant cluster (bt) shows complete neutralization. The speaker with Kiel ancestry produces only one exception to the (bt) neutralization trend with the token 'eru~ubt' (see Table 17). Recalling the perception

On final laryngeal distinctions in Wisconsin Standard German

data for this token, the majority of respondents heard the 'eriaubt' token as voiced, possibly due to the slight presence of glottal pulsing as compared to the complete absence of it in the rest of the final/btl cluster tokens. 5.2

Data from Ozaukee (speaker with ancestry in Pomerania)

The speaker with ancestry in Pomerania demonstrates final laryngeal neutralization across /b,d,g,z/ consonants and only produces voicing in cases of emphasis or when the next word starts with a voiced consonant. Nonetheless, neutralization does show up in some cases where the following word starts with a voiced consonant. These patterns look like early stages of neutralization in transition to the complete neutraliza-tion found in the Standard German variety. Table 18. Ozaukee Tokens with Underlying is/ Word

GP%

Length (sec.)

alles (D) alles 2 (stress-D) Dietze's (D) Average (all)

0.143 0.343 0.429

0.075 0.071 0.066

0.305

0.071

The two 'aUes' tokens are surprisingly close together in terms of glottal pulsing percentage (see Table 18). The second 'a lies' token is stressed as well as followed by a voiced obstruent and therefore would be expected to exhibit more voicing. But, both tokens remain laryngeally neutralized regardless of the voiced consonant following. 'Dietze's' is the name of a German 'Wursthaus' in Milwaukee. The final consonant shows partial neutralization due to the fact that it is a German last name and the possessive I sf would normally be voiceless in Standard German. Table 19. Ozaukee Tokens with Underlying /z/ Length (sec.)

Word

GP%

Klubs 1 (D) Klubs2 (D)

0.667 0.200

0.032 0.075

Average (all)

0.434

0.054

The two tokens with underlying final /z/ consonants are words borrowed from English (see Table 19). Here one would expect the plural/sf to sound more voiced due to the fact that English voices final [b] consonants. 'Ihe first 'Klubs' token can indeed be characterized as having the expected final voiced consonant. The second occurrence of this token suggests partial laryngeal neutralization with I sf consonants, as the glottal pulsing percentage is quite low.

27

28

Renee Remy Table 20. Ozaukee Tokens with Underlying Final/g/ l.ength (sec.)

Word

GP%

fiin.fzig (D) gettug(T)

0.320 0.600

0.053 0.052

Average (all)

0.460

0.053

There are very few instances of underl}ing final/ gl consonants in the speech sample (see Table 20). It is difficult to determine whether or not this speaker consistently exhibits final laryngeal neutralization with final/gl consonants because there are not enough tokens. Although many varieties of German do produce spirantized final/gl consonants, this speaker does not. These two tokens are opposite of what might be expected. A final consonant right next to a voiced consonant beginning the next word tends to lead to a higher percentage of glottal pulsing as shown with /b,d/ consonants from this speaker. A final consonant next to a voiceless consonant beginning the next word usually leads to a lower percentage of glottal pulsing, as seen in /b,d,s/ final consonants from this speaker. Table 21. 07..aukee Token with Underlying Final/'1)./ Cluster Word

GJ?%

gesagt(T)

0.350

l.ength (sec.) 0.020

The last token remains voiceless as part of the consonant cluster (gt) (see Table 21 ). 'Ihe speaker is anticipating the following voiceless consonant. However, one would expect the glottal pulsing percentage to have been lower because Standard German shows complete neutralization for this final consonant cluster. Table22. Ozaukee Tokens v.ith Underlying Final id/ l.ength (sec.)

Word

GP%

Deutschland (D) Deutschland 3 (pause) Feld (pause) Kind (pause) Mitglied (T) Rullestand (pause) Schwarzwllld (pause)

0.500 0.300 0.167 0.000 0.500 0.000 0.300

0.020 0.025 0.063 0.095 0.030 0.104 0.028

Average (all) Average (pause)

0.252 0.153

0.052 0.063

On final laryngeal distinctions in Wisconsin Standard German

Many of the tokens with underlying final /d/ consonants were phrase final (see Table 22). The overall average can definitely be considered laryngeally neutral, and the average for tokens followed by a 'pause' is even lower stilL The tokens followed by a pause (and/or phrase iinal) are almost exclusively laryngeally neutral given that they have very low percentages of glottal pulsing. One token is followed by a voiced obstruent, thus contributing to the increase of voicing on the final [d]. The remaining token 'Mitglied' is followed by a voiceless obstruent and shows 50% glottal pulsing. This may have something to do with the word 'Mitglied' specifically. However, five out of the seven tokens remain laryngeally neutral. Table 23. Ozaukee Tokens with Underlying Final /b/ Length (sec.)

Word

GP%

Klub 1 (T) Klub2 (D) Klub 3 (pause) Klub 4 (pause) Klub 5 (D) .f..."1ubs (T) .f..."1ubs 2 (T) giaub 1 (D) glaub2 (D) glaub 3 (D) hab 1 (D)

0.350 0.333 0.333 0.000 0.650 0.083 0.120 0.800 0.467 0.100 0.444

0.023 0.032 0.034 0.041 0.019 0.063 0.047 0.010 0.034 0.070 0.009

Average (all) Average (T) Average (D) Average (pause)

0.335 0.184 0.466 0.167

0.035 0.044 0.029 0.038

The tokens with an underlying final /b/ consonant show very close to complete final laryngeal neutralization (see Table 23). Nine of 11 tokens show final laryngeal neutralir..ation. The average glottal pulsing percentage for tokens followed by 'D' is more than twice that of tokens followed by 'T' or 'pause'. The percentages (on average) for tokens followed by 'T' and 'pause' fall fairly close together. Pre-pausal position is the classic place for devoicing (Blevins 2004; Iverson & Salmons forthcoming), and the current data support this finding. 1his speaker seems to show an intermediate step between the Standard German "no distinction'' and the Standard American English "consistent distinction:' All average glottal pulsing percentages show the presence of final laryngeal neutralization. The only place a high percentage of glottal pulsing occurs is before a voiced obstruent like in the tokens 'Klub 5' and 'glaub 1'. The rest of the tokens clearly tall below the margin of 50% glottal pulsing regardless of the nature of the consonant that follows in the next word. The English borrowing of the word

29

30

Renee Remy

'Klub' seems to be following Standard German norms, seeing as the final /b/ is laty'llgeally neutral and would not be in American English.

6.

Implications for Wisconsin English

In German, final/d/ consonants are almost exc.lusively devoiced, regardless of the fact that final /d/ consonants are always preceded by a nasal, liquid, or a vowel. Both the preceding and following sound seem to affect American English pronunciation, whereas German pronunciation seems to be affected much more by the following consonant. When the preceding sound is a vowel, it occasionally affects the nature of the final consonant, but not with any regularity across the vowel spectrum. In cases where an English loan is inserted in the midst of running German speech, the loan word is not fully phonetically and phonologically integrated into German. According to Pavlenko (2004), the use ofL2 words in the L1 amongst bilingual speakers comes from the need to name new objects and/or objects that do not exist in Germany. A L2 word may also be used to emphasize some particular conceptual distinction, which traditional L1 terms cannot adequately describe. It is not necessarily the case that language attrition is the reason for code-s"\<\itching. The phonetic realizations for English words in the current study reflect the combination of Ll and L2 norms, which exist~ due to long--time exposure to L2 (Engli.sh). Eastern Wisconsin English, however, seems to have taken its cues from the German speakers in the area. German-speaking communities in Wisconsin e..'Cperience language contact between German and English, which has led to the development of Wisconsin English according to German final laryngeal neutralization patterns, in addition to the German L1 being affected by Wisconsin English influence. Pavlenko & Jarvis (2002) looked at Ll speakers of Russian who had lived in the U.S. for 3-8 years. They found that language trc:Ulsfer can go both ways, namely the L1 (Russian) influenced the L2 (English) in addition to the L2 influencing the Ll. The German L1 speakers of the current study who are bilingual in English and use English much more than German may show English-like phonetic patterns when speaking German. These L1 German speakers show a higher occurrence of voicing on final laryngeal consonants than should be seen in Standard German, which could be explained by contact with Wisconsin English. By the same token Wisconsin English speakers show final laryngeal neutralization in places where speakers of Standard American English would produce a voic.ed variant. One place where this occurs in Eastern Wisconsin English with increasing regularity is with final/sf and /z/ consonants. Thus, some Eastern Wisconsin English speakers have adopted a more German--like pattern for final/sf and /z/ consonants.

On final laryngeal di.~tinction.~ in Wisconsin Standard German 7·

Summary and conclusion

1he Eastern Wisconsin German data provide an interesting perception of spoken German outside of Germany and insight into language transfer effects. 1he perception data support the actual glottal pulsing data in showing (or not) final laryngeal neu-tralization. The 171 tokens seem to devoice most often, but the possessive and plural morpheme-s is sometimes voiced following voiced obstruents, as is the case in Eastern Wisconsin American English. 1he voiced/voiceless nature of the final consonant also depends greatly on whether or not the first sound of the next word is voiced or voice-less. 'Ihe case with final/d! being neutralized remains consistent with what would be expected in Standard Gennan. The /b/ tokens exhibit final laryngeal neutralization everywhere in the speech of both Low German dialect speakers except when followed by a voiced consonant. As for the status of final/b,d,g,zi consonants in Wisconsin German, these dialect speakers both exhibit patterns of final laryngeal neutralization, but do not neutralize with exactly the same consistency as in modern High German. These German speakers seem to be converging on English norms at the same time as Eastern Wisconsin English speakers are converging on German norms. Both the preceding sound and the following sound of the next word impact the nature of the final consonant in a given word. In German, only the first sound in the following word impacts the character of a final consonant. English is affected more by what precedes the final consonant and sometimes also by the following consonant. Patterns of final laryngeal neutralization are not as strong as one would expect, and they seem to be (in certain words) taking cues from Wisconsin English patterns. However, speakers of Eastern Wisconsin English tend to devoice final/z/ almost exclusively, taking their cues from German. These two German speakers with roots in the north help shed light on how Eastern Wisconsin English has come to be spoken the way it is today.

References Blevins, Juliette. 2004. Evolutionary Phonology: 1he Emergence of Sound Patterns. Cambridge: CUP. Honeybone, Patrick. 2005. Sharing makes us stronger: Process inhibition and segmented stntc-ture. In Headhood, Elements. Spec!ftcation and Contrastivlty: Phonological Papers In Honour ofJohn Anderson [Current Issues in Linguistic Theory 259] Philip Carr, Jacques Durand & Colin J. Ewen (eds), 167-192. Amsterdam: John Benjamin~. Houlihan, Kathleen & Iverson, Gregory K. 1979. Functionally constrained phonology. In Current Approache-S to Phonological 1heory, Daniel A. Dinnsen (ed), 119-134. Bloomington IN: Indiana University Press. Inozuka, Emiko. 1991. 'Ihe realization of the German neutralized word-final plosives/g, k/: An acoustic analysis. Sophia L!ngulstica 30: 119-134.

31

32.

Renee Remy Iverson, Gregory K. & Salmons, Joseph. 2006. On the typology of final laryngeal neutralization: Evolutionary phonology and laryngeal realism. Theoretical Linguistics 32(2): 205-216. Iverson, Gregory K. & Salmons, Joseph. I1orthcoming. Hnal devoicing and final laryngeal neutralization. Companion to PI'IJ.mology, Marc van Oostendorp, Colin Ewen, Beth Hume & Keren Rice (eds). Oxford: BlackwelL Kiparsky, PauL 1976. Abstractness, opacity, and global rules. In 'l11e Appliwtio,l of and Ordering ofGrammatiwl Rules, .Andrea.~ Koutsouda.~ (ed.), 160-186. The Hague: Mouton. Mihm, Arend. 2004. Zur Geschichte der Auslautverhiirtung tmd ihrer Erforschung. Sprachwls-· senschaft 29(2): 133···206. Pavlenko, Aneta. 200·4. L2 influence and LI attrition in adult bilingualism. Studies In Bilingualism: First Language Attrltion-.Jnterdiscipli,lary Perspectives on Methodologic.al Issues [Studies in Bilingualism 28] Monika Schmid, Barbara K6pke, Merel Keijzer & Lina Wellemar (eds). 47--57. Amsterdam: John Benjamins. Pavlenko, Aneta &Janis, Scott. 2002. Bidirectional Transfer. Applied Linguistics 23(2): 190--214. Piroth, Hans Georg & Janker, Peter M. 2004. Speaker-dependent differences in voicing and devoicing of German obstruents. Journal of Phonetics 32(1): 81-109. Port, Robert & Crawford, Penny. 1989. Incomplete neutralization and pragmatics in German. Journal of Phonetics 17: 257-282. Purnell, Thomas, Salmons, Joseph. Tepeli. Dilara & Mercer, Jennifer. 2005. Structured heterogeneity and change in laryngeal phonetics: Upper midwestern final obstruents. Journal of English Linguistics 33(·1): 307-338. Salmons, Joseph & Lucht. Felecia A. 2005. Standard German in Texas. In Studies in Contact Linguistics: Essays in honor ofGlenn G. Gilbert, Linda Thornburg &Janet Fuller (eds), 165--186. Frankfurt: Peter Lang.

Past participles in Mocheno Allon1orphy, alignment and the distribution of obstruents* Birgit Alber Facolta di Lingue e Letterature Straniere, Universita di Verona

M.ocheno, a German variety spoken in Trentino (Italy), displays an interesting case of phonologically conditioned suppletive allomorphy in past participle formation. Past participle formation involves a variety of strategies, from absence of a prefix. to affrication, to prefuing a CV-prefixga-. I propose that two allomorphs are involved in the process, a subsegment [-cont. -voice] and a prefixga-- and that the distribution ofthe two allomorphs is regulated by a hierarchy of wellformedness constraints. This hierarchy in turn consists of two independent partial hierarchies, which are active in the grammar of the language in general, where they are responsible for morphoprosodic alignment at left stem edges and for the distribution of obstruents, respectively. This means that Mocheno past participles give us evidence in favor of the hypothesis that allomorph selection can, and sometimes must, be interpreted in terms of optimization. By adopting an alternative analysis in terms of subc.ategorization of the allomorphs for a certain phonological conte.xt, the relationship between the distribution of past participle allomorphs and other pieces of the Mocheno grammar would remain completely opaque.

1.

Preliminaries

Mbcheno (or, in German, Fersentalerisch) is a Southern Bavarian variety spoken in the speech island Bersntol (Italian: val dei Mocheni, German: Fersenta.l) in Trentino, Italy. The speech island is the result of the colonization of the val dei Mocheni by Tyrolean

*I want to thank Uirike Domahs, Shigeto Kawahara, Martin Krll.mer, Pranz Lanthaler,Andrew Nevins, Marc van Oostendorp, Ingo Plag, Anthony Rowley and an anonymous re"iewer, who have contributed considerably to my understanding of the data and to crucial turns in the analysis. Many thanks also to my informants Renzo Lenzi and Cristina Moltrer for their seemingly unlimited patience and to Lorenza Groff and Leo Toller for their double role as informants and providers of organisational help during my data collection.

34 Birgit Alber

and Bavarian farmers during the 13th century. It is unclear how many speakers the language currently has. 1 Mocheno, its history, sociolinguistic situation and its structure, has been described in much detail in numerous publications by Anthony Rowley (cf. especially Rowley 1982, 1.986, 2003). In this paper I will describe and analyze the fonnation of past participles in Mocheno. I will assume that past participles are fanned by prefixing one of two allomorphs to the verbal stem. Specifically, I propose that an allomorph [-cont, -voice] is prefixed to verbal stems beginning with a voiceless stop or a labial or coronal fricative while an allomorph ga- is prefixed in all other contexts: (I)

a.

Allomorph [····cont, ···voice]: viern -7 pfiert 'to c.onduct, conducted'

b.

Allomorph [ga]: rearn -7 garcart

'to cry, cried'

The principles of allomorph selection have been the topic of much debate in the recent phonologicallitemture (cf., among others, Lapointe 2001; Bye 2007; Paster 2006; Mascar6 2007; Wolf 2008 for discussion, Nevins, 2011, for a comprehensive overview). Specifically, it has been discussed to what degree phonologically conditioned allomorphs are selected by phonological markedness constraints. Analyses have been proposed which assume explicitly or implicitly that phonologically conditioned allomorphy is to a large e.xtent regulated by phonological markedness constraints favoring the optimization of structures (for this position see among others

1. 1he liiJb>ui!.tic census carried out by the Autonomous Province of Trento in 2001 reports that 2.278 people declared their affiliation to the Mocheno minority (Annuario Statistico 2006).lt is highly unlikely that this number reflects the number of speakers of Mocheno, since Mocheno is spoken - except for single individtLals - only in the three municipalities of Palai/ Palu del Fersina, Vlarotz/Fierozzo and Garait/Frassilongo (which incltLdes also Oachlait/ Roveda). The mLmber of people who declared their affiliation to the Mocheno comunity in these mtmicipalities amotmts to 947 and represents 95,4% of the population. Howeve1~ even this number cannot correspond to the actual number of speakers, since for the villages of Garait/Frassilongo and Vlarotz/Fierozzo Rowley (1986) describes a language situation where few families (Garait) or at most half of the families (Vlarotz) still speak Mocheno. The high percentages resulting in the census could reflect at most the number of speakers in the villages of Palai/Palu del Fersina and Oachlait/Roveda, in which, according to Rowley (1986), most families still speak Mocheno. If we ass-ume a percentage of 50% speakers in Garait -Oachlait (who figtLre as a single municipality in the statistics) and Vlarotz and 95,4% speakers for Palai, we arrive at the hypothetical mLmber of 583 Mocheno speakers (s. also the disctLssion in the local journal Lem, 27. Nov. 2002, 29. Sept. 2003, 30. Dec. 2003).

Past participles in Mocheno

McCarthy & Prince 1993b; Mester 1994; Drachman, Kager & Malikouti-Drachman 1996; Kager 1996; Mascar6 1996; Tranel1996; Plag 1999; Rubach & Booij 2001; Wolf 2008). Under this assumption, the best of two or more allomorphs will be selected, where phonological markedness conditions determine what is 'best' in a certain context and language. Other investigators have followed a more cautious approach, proposing that while optimization may play a prominent role in allomorph selection, at least in some cases of phonologically conditioned allomorphy the distribution of the allomorphs is due to subcategorization of the allomorphs for a specific phonological context (Lapointe 2001; Mascaro 2007; Bonet, Lloret & Mascar6 2007). More radically, Paster 2006 and Bye 2007 propose that all phonologically conditioned allomorphy should be described through subcategorization of the allomorph for a phonological context. For the phenomenon under scrutiny in this paper, under the optimization approach we expect the distribution of [---cont, ---voice] and ga- in Mocheno past participles to create structures which are phonologically unmarked, in some sense. Under the subcategorization approach, on the other hand, we will have to mark each allo-morph as being selected f{)r a specific phonological context. The contribution of the present paper to the debate about what determines allomorph selection is to show that (i) it is possible to analyze the process of past participle formation in Mocheno as a process through which unmarked structures arise (ii) it is desirable to analyze the process as a process of optimization sinc.e the constraint hier-archy responsible for allomorph selection subsumes two partial constraint hierarchies which are active in the language as a whole. I will therefore conclude that although there might be cases of phonologically conditioned allomorphy which (still) resist an interpretation in terms of optimization there are other cases where optimization is the most plausible analysis. In this paper, past participle formation in Mocheno will be analyzed as driven by markedness, since the choice between the two allomorphs [-cont, -voice] and ga-- is shown to be triggered to a large extent by the markedness constraint "[+vOicE], favoring the voiceless allomorph [ ---cont, ---voice], and by an alignment constraint, ALIGN (STEM, L, SYLL, L), favoring the morpho-prosodic alignment of left stem edges with syllable boundaries. In addition to these two constraints, the partial hierarchy determining the distribution of voiced and voiceless fricatives in Mocheno conditions the choice of the allmorph in the remaining contexts. 1he paper is structured as follows. In Section 2 an overview of the pattern of past participle formation in Mocheno is given. In Section 3 I argue for the existence of two past participle allomorphs and identify them as a floating subsegment [-cont, -voice] and a prefix ga-. In Section 4 I discuss alignment between left stem boundaries and

35

36

Birgit Alber syllable boundaries in Standard German and in Mocheno, and its significance for past participle formation. In Section 5 the distribution of obstruents in Mocheno is analyzed and a ranking deriving it is established. In Section 6, which contains the analysis proper, I analyze the single contexts of past participle formation and determine the constraint rankings that are relevant for the choice of the allomorphs in each context.

z. Past participle formation in Mocheno The regularities of past participle formation in Mocheno are exhaustively described in Rowley 1986, 2003. Andreolli 2004---05 has elicited an additional number of past participles with respect to those mentioned by Row ley, arriving at the same set of rebru-larities. I have double checked Rowley's and Andreolli's results by eliciting 68 past participle forms from two speakers, one from Palai (Palu del Fersina) and one from Vlarotz (Fierozzo). The regularities which emerge from these sources are consistent among each other and can be summarized as follows. 2 The past participle in Mocheno is formed by adding to the verbal root the suffixes -(a)t-3 (for so called 'weak: regular verbs, ct: a. and c. below) or -(a)n (for so called 'strong: irrebrular verbs, cf b. below). In strong verbs, the quality of the root vowel may change, in accordance to historical .Ablaut (b. below). In addition to the suffix and Ablaut. in certain phonologically conditioned contexts a prefix is added, which can be either realized as ga- (a. and b.) or take the form of an afiiicate by forming a complex segment with a root--initial fricative (c.): (2)

infinitive past participle a.

b. c.

bis--n nem-;m vier-n

ga--bis-t ga-nom-;m p-fier-t

'to know' 'to take' 'to conduct'

This paper focuses on the context in which past participle prefixes are realized as well as on the form they assume. The distribution of the past participle suffixes -(a)t and ·-(a)n is unpredictable and will be ignored from now on.

See Cognola, in press, for a discussion of the semantics of ga- prefixed verbs and de verbal nouns in Mocbeno. While the realization of the prefix is governed by phonological constraints in past participle formation, its realization in deverbal nouns seems to be sensitive to the semantics of the base nouns, as argued by Cognola. 2.

Tbe suffix can take the form of --et or --en in the variety of PalaL Tbe -(a)n suffix of strong verbs L'l not realized when the verbs end in a nasal (cf.jlo:IJ ~ t/lo:IJ•bea.t' and Sterm -7 (t)Stourm 'die' in (4)).

3·

Past participles in Mocheno 37 'lhe presence of a prefix depends, among other things, on the location of stress in the verbal root. As in Standard German, the prefix must be attached to a stressed syllable. It is never realized in verbal roots that do not bear initial stress, as in the following examples (cf. Rowley 1986: 239): (3)

infinitive past participle a. b.

konta:rn konta:rt vargtsn vargl:sn

'to tell' 'to forget'

Besides stress, the quality of the initial segment of the verbal root conditions the absence or presence of a prefix as well as its realization as ga- or as an affricate. As illustrated in the table below, the prefix ga-· is never reali1.ed when the root begins with a voiceless stop. When the first segment of the root is a labial [v), alveolar [z] or palatoalveolar [JI fricative, the past participle is formed by changing the initial segment into the corresponding voiceless affricates [pf], [ts] and [tJ].4 When roots begin with a sibilant-stop duster, the sibilant is always realized as [s] (described by Rowley 1986 as a postalveolar fricative) 5 before voiceless stops (e.g. spi:ln, 'to play'), and as [f) before sonorants (e.g. fnain, tfni:tn, 'to cut, cut past part:, Rowley 1982). Verbs with root-initial sibilant--stop clusters form their past participles either by creat-ing an initial affricate (similarly to roots where the first segment is a simple labial or a coronal fricative) or ·without a prefix. There is variation among speakers regarding these two strategies and even the same speaker might apply one strategy to one verb and the other strategy to another or might accept both possibilities for the same verb. In the analysis below I will treat sibilant-stop verbs in the same way as roots with a single initial fricative (hence forming the past participle through affrication) and will abstract away from the possibility of not realizing the prefix at all, a possibility I will attribute to cluster simplifkation. The prefix ga.-- is always reali1.ed when the root begins with a voiced stop, the voiceless glottal fricative [h] or a sonorant. 1he following table summarizes the regularities of past participles and gives examples for each phonological context. 'Ihe examples are transcriptions of verbs I elicited from one speaker of the variety ofVlarot:r., but except minor details, are identical to

4· Syllable-initial labial and alveolar fricatives are always reali:l.ed as voiced [v] and [z] and the palatoalveolar fricative [JJ is voiceless in tb.is conte.xt, in the native Mocheno lexicon (cf. Rowley 1986 and discussion in Section 4 of this paper). 5· Rowley distinguishes between postalveolar [s] and palatoalveolar [Jl Postalveolar [s] can be considered an allophone of [f] in the variety of Palai, but syllable finally still participates in phonemic contrast in tb.e variety ofVlarotz (cf. Rowley 1986 for detaiLc;).

38

Birgit Alber

those elicited for the variety of Palai.6 They confirm the regularities already described in Rowley 1986, 2003 and Andreolli 2004-2005. (4)

Mocheno past participle formation

past participle formation

first segment of verbal root

infinitive

past participle

gloss

0

voiceless stop

paizn ton.brn krotsn

pisn krotst

'to bite' 'to tlnmder' 'to scratch' 'to conduct' 'to look for' 'to beat'

labial and coronal fricatives

Jlo:l)

pfiert tsuaxt tJlo:l)

affricate or0

sibilant -stop cluster

spi:ln sterm

(t)spi:lt (t)stourm

'to play' 'to die'

ga

voiced stop

bisn grie:l;)n

gabist gadrukht gagriest

'to know' 'to press' 'to greet'

hupf:m

gahupft:

'to jump'

o:tn;~n

gao:tnt sa.iu:tst gareart galnxt ganutst gamnxt

'to breathe' 'to cl1eer' 'to c.ry' 'to laugh' 'to be of use' 'to make'

voiceless affricate

viern

tond;~rt

zuaX;~n

drukh;~n

h sonorant

ju:tsn rearn lnx;,n nutsn mnx.:m

The regularities exhibited by the verbs above don't seem to be relics of a historical process, but rather, are productive or, at least, have been productive at some point in time, since they are also active in a subpart of the loanword lexicon. Mocheno has integrated a certain number of verbs from the Italian dialect of Trentino (or, more recently, from Standard Italian) in its lexicon. The most common way of integration is to replace the inflectional ending of the Romance verbal root with the suffixes -a:rn, -e:rn, -i:rn (Rowley 2003: 247):

There is some variation e.g. in the realization of diphtongs (viern in Vlarotz, viarn in Palai) and in the length of the root vowel of one past participle (e.g. tjlo:l'} in Vlarotz, ~{lor,; in Palai). The rules that govern the patterns are the same.

6.

Past participles in Mocheno 39 (5)

'contar' (Trentino) 7 konta:rn 'to tell' 'rispondere' (Standard Italian) 7 respund:lrn 'to answer'

None of the loan verbs that have been integrated through this type of suffixation bear initial stress, since stress is preserved on the same syllable as in the Romance base verb usually the penultimate, or, in some cases, antepenultimate syllable of the verb. For this reason this class of loans will not tell us anything about the alternation of the ga- prefix with other strategies of past participle formation, since ga- is absent whenever stress is not root initiaL However, there is a small number of loan verbs which have not followed the usual path of integration and henc.e may bear initial stress. 'lhey are few in number and there are not examples for every single phonological context, but yet it is clear that the regularities of past participle formation hold for them as well. Below, I give a list of the ones I found, and the past participles produced by my informant in Vlarotz (to a large extent confirmed by the informant from Palai):7 (6)

Initially stressed loan verbs in Mocheno: past participle formation

first segment infinitive past part. of verbal root

gloss

Trentino base8 gloss

voiceless stop

pintsn

pintst

'to dent'

spintsar

'to dent, ruin'

tfu:dln

tfu:d;)lt

'to work badly'

tfondohir

'to work badly'

fricative

vri:gln

pfri:glt

'to grind'

sfregolar

'to grate, rub'

sibilant -stop

Skitsn

skitst

'to crush'

skit.far

'to crush'

Skaltsn

skaltst

'to kick'

skaltsar

'to kick'

cmbeJolarse

'to stutter'

voiced stop sonorant

7·

befin

gabcJlt

'to stutter'

dratJn

gadratJt

'to rain heavily' dratJ9

meiv~rn

gameiv;Jrt 'to move'

ritjln

garitflt

'to make cuds'

'sieve' (notm)

m5ver

'to move'

cnrioohh

'to make curls'

My informant in Palai did not know all the examples presented here.

8. Thanks to Lorenza Groff for providing the Trentino verbs from her native Trentino variety ofPine.

9.

See also the verb drazar 'separate' in Ricci ( 1904).

40

Birgit Alber

Loans form their past participles in the same way as verbs from the native lexicon: no prefix is inserted before voiceless stops, affricates are formed if the root begins with the voiced fricative [v] and the prefix ga- is present before voiced stops and sonorants. As for sibilant-stop clusters, in this case the speaker chose the strategy of not inserting any prefix in the verbs skitsn and skaltsn. Anthony Rowley has argued in two lectures (Trento, 1-3 April 2009; Verona, 17 December 2009) that the regularities of past participle formation in Mocheno cannot have arisen through historical change since they are unattested in this form among the German dialects. Historically, it would seem that syncope (i.e. deletion of the prefix vowel or of the entire prefix) targets first fricative-initial roots, then sonorantinitialroots, and evenutally stop-initial roots. Mocheno, however, would combine the first and the last stage of the development while leaving out the second. Rowley proposes that the pattern in Mocheno has developed through 'Dialektausgleich: which is assumed to have taken place early in the colonization, between 1220 and 1300 (Kranzmayer 1956). At least some of the mentioned Trentino loans can be assumed to have entered the language after that stage. The behavior of loans together with the fact that the Mocheno pattern of past participle formation is unattested as the result of diachronic change elsewhere among the German dialects point to the pattern as a productive process at least at some point in time.

3·

The allomorphs /ga-/ and [-cont, -voice]

The formation of past participles in Mocheno is interesting because of the variety of strategies used, which include prefixation, atlrication and absence of prefixation. In this section I will discuss the nature of the prefix and conclude that we are dealing here with a case of phonologically conditioned suppletive allomorphy (Paster 2006; Wolf 2008), i.e. a case of phonologically conditioned allomorphy where the two allomorphs are not linked to the same underlying form. In this sense, phonologically conditioned suppletive allomorphy is distinct from plain allomorphy as e.g. the [-z, -Iz, -s] allomorphs of English plural markers which can be seen as realizations (through epenthesis and voicing assimilation) of a single underlying morpheme /z/. To make clear the difference between the two types of allomorphy I will compare Mocheno past participle formation with past participle formation in some Tyrolean dialects of Sudtirol· Alto Adige, which, geographically speaking, are the closest Germanic dialect 'neighbors' of the Mocheno speech island. Only Mocheno displays phonologically conditioned suppletive allomorphy, while the

Past participles in Mocheno Tyrolean dialects can, to a large extent, be analyzed as cases of plain phonologically conditioned allomorphy. The Tyrolean dialects of Meran, Passeier and Upper Vinschgau display similar strategies as Mbcheno in past participle formation. The patterns are summarized in the table below, tollowing the description in Alber & Lanthaler 2005: (7)

Past participle formation in the Tyrolean dialects of Meran, Upper Vinschgau and Passeier

past participle formation

first segment of verbal root

infinitive

past participle

gloss

Dialect of Meran g;)

stop

pikxn del)k'-n

g;)-pikXt g;)-del)kxt

'to glue' 'to think'

g/k

fricative

fro:gl) visn

k-fmk g-vist

'to ask' , 'to know

g

sonorant

esn loxn

g-csn g-loxt

'to eat' 'to laugh'

Dialect of Upper Vinschgau

0

stop

pikxn del)kXn

pikXt deiJkxt

'to glue' 'to think'

g/k

fricative

fro:gl) visn

k-frok g-vist

'to ask' , 'to know

g

sonorant

f.Sn loxn

g-csn g-lnxt

'to eat' 'to laugh'

gr

stop

pikXn derJkxn

gr-pikxt gr -deiJkxt

'to glue' 'to think'

k

voiceless fricative

fro:gl)

k-frok

'to ask'

gr

voiced fricative

visn

gr-vist

'to know

g

vowel

esn

g-csn

'to eat'

gr

sonorant

loxn

gr-loxt

'to laugh'

Dialect ofPasseier

,

Mocheno shares with all three varieties the presence of a gV-- prefix. Similarly to the dialect of Upper Vinschgau it omits the prefix completely before voiceless stops, but differently from the Upper Vinschgau dialect, the prefix is present before voiced stops. Similarly to the dialect of Passeier, the gV- prefix is present also before sonorants. The only strategy exhibited by Mbcheno alone is the expression of the prefix through an affricate in the case oflabial and coronal fricatives.

41

42

Birgit Alber

At first glance, the comparison with the dialect of Meran and the dialect of Upper Vinschgau might suggest that we are dealing here with a process guaranteeing syllable wellformedness. For these two dialects it seems plausible to assume an underlying prefix /g-/, which is augmented by a epenthetic vowel (in the dialect of Meran) or dropped altogether (in the Upper Vinschgau variety) whenever illicit clusters would be created. However, there are several differences with respect to the Tyrolean varieties which suggest that an analysis in these terms is not feasible for Mocheno. First, in Mocheno the prefix never appears in the form [g] or [k]. This makes the postulation of an underlying fonn /gj for the prefix rather abstract. Furthermore, while the schwa in the ga- prefix of the dialect of Meran can be plausibly analyzed as an epenthetic element, the same is not true for the vowel present in the Mocheno prefix ga-·. Epenthetic vowels in Mocheno have a schwa-like quality and are clearly distinct from [a], at least in the variety ofvlarotz. 10 We find it for example in the context of the adjectival inflectional suffix-s when it is suffixed to a root ending in a sibilant: (8)

Epenthetic schwas between sibilants: (variety ofVlarotz) a kholt-s bns:.~r 'some cold water' (n.sg.nom.) a hoaz--:.~s boS:.lr 'some hot water' (n.sg.nom.) a vrif--:.~s boS:.lr 'some fresh water' (n.sg.nom.)

Since the vowel in the Mocheno past participle cannot be interpreted as epenthetic, I propose that the underlying form of the prefiX is /ga--/. However, /ga-/ cannot be the only underlying form playing a role in past palticiple formation in Mocheno. If this was the case, it would be very difficult to explain why /gal does not surface when the root begins with a voiceless stop or a labial or coronal fricative. After all, fi.)rms like *ga-pis-n or *ga-11ier-t would be perfectly wellformed, from the point of view of syllable structure. If we limit our attention to the contexts where the prefix ga- does not appear, we see that a second allomorph must be at play here. Consider again the environment of voiceless stops and labial and coronal fricatives:

to. In the variety of Palai, we do find a vowel similar to [-e) in contexts where epenthesis can be assumed (cf. Rowley 1986: 821f. for details of the variation between the local dialects of vowel quality in unstressed syllables.). Epenthetic schwa, as we find it in the variety ofVlarotz, has often a lower, more advanced quality than schwa in Standard German. Rowley (1986) in fact often transcribes it as [t ].

Past participles in Mocheno (9)

No /ga-/ prefu before voiceless stops and labial/coronal fricatives:

strategy of past participle formation

first segment of verbal root

infinitive

past participle

gloss

0

voiceless stop

paizn toncbrn krotsn

pisn krotst

'to bite' 'to thunder' 'to scratch'

pfiert tsuaxt tflo:I)

'to conduct' 'to look for' 'to beat'

voiceless affricate

labial and coronal fricatives

viern zuax~n

Jlo:I)

tond~rt

The past participle of verbs beginning with labial and coronal fricatives is formed by creating initial affricates which differ from root-initial fricatives in voicing. The prefix, in this case, can be assumed to be a subsegment specified for [-cont, -voice]. I will assume that it consists of a root node with the features [ -cont] and [-voice], but not specified for place: (10)

root

~

[---cont]

[---voice]

Adding these features to the initial fricatives, the complex segments lpf), Its] and [tJ] are created.l 1 Before voiceless stops, on the other hand, we can assume that the allomorph [-c.ont, -voice] undergoes coalescence with the root--initial stop, thus leading to the realization of a structure without a visible prefix. Note that it is not possible to reduce the allomorph to one of the two features, either [····cont] or I ---voice]. If we assumed that [····cont) was the relevant allomorph, we would expect coalescence to happen also with root--initial voiced stops. If, on the other hand, the allomorph consisted only of a [-voice] feature, we would not expect the par-ticiple to be realized by an affricate when roots begin with a voiceless [J]. To summarize, we can hypothesize that in Mocheno past participle formation two allomorphs are at play, one a more abstract prefix which takes the form of the features [-cont, -vok.e], the other a regular prefix ga.-: (ll)

Allomorphs in Mocheno past participle formation Allomorph I: [-cont, -voic.e] before voiceless stops and labial and coronal fricatives Allomorph II: /ga-/ before voiced stops, [h] and sonorants

n. For other analyses involving morphemes expressed as stLbsegments see Zoll 1996 and Rubach & Booij 2001.

4.1

44

Birgit Alber

If this approach to the data is correct, we are dealing here with a case of phonologically conditioned suppletive allomorphy, since the distribution of the two allomorphs depends on the quality of the first segment of the verbal root (and on the location of stress in the root, cf. above) but no common underlying form can be established for them. 12

4·

Alignment

It seems puzzling that in Mocheno past participle formation a CV allomorph should be chosen when roots begin with sonorants while a consonantal prefix is inserted before root-initial (voiceless) stops and fricatives. After all, from the point of view of syllable wellformedness we would rather expect the contrary. Indeed, we see that for example the dialect of Meran has more or less the opposite distribution, placing a gV- prefix before stops and a g- or k- prefix before fricatives and sonorants. At first glance thus it would seem that past participle formation in Mocheno does not easily yield to an analysis in terms of structure optimization. However, the distribution of past participle allomorphs in Mocheno does consistently respect a principle which is also at play in Standard German and many of its dialects, the principle that a stem should not be resyllabified with a prefix. 'Ihus, in Standard German consonant final prefixes do not readily resyllabify with vowel initial stems, instead a glottal stop is inserted at the stem boundary (s. Giegerich 1989, 1999; McCarthy & Prince 1993a; Alber 2001 for detailed descriptions of the phenomenon and analyses of it): 13 (12)

Vor.·· [?]ort 'suburb' ver.·· [?]antworten 'to take the responsibility' [?]auf.- [?]atmen 'to be relieved' [?]Er.·· [?]eignis 'event'

12. Things are yet more complicated in the dialect of Passeier, in which the [1] of the prefix could be epenthetic since it appears in clear epenthesis conte.xts such as adjectival inflection. Thus the correspondent form of Mocheno hoaziJs in this dialect would be hoas.rs (H Lanthaler, p.c.). Yet, an analysis in terms of syllable wellformedness is not as straightforward as for the dialects of Meran and Upper Vinschgau since the CV prefix appears also before sonorants and voiced fricatives. An analysis following Fleischhacker's (2001, 2005) findings that epenthesis is more readily obtainable before sonorants (and voiced fricatives?) rather than before obstruents might be possible but I have to leave a detailed investigation of this pattern for future research. 13. The relevant morphological boundary is indicated with a hyphen"-': the relevant syllable boundary with a dot ". ": Examples are given in orthographic form.

Past participles in Mocheno I will express this principle with an Alignment constraint (s. McCarthy & Prince 1993a for a similar proposal): (13)

ALIGN (STEM,

L,

SYLL,

L):

'if stem 3 syllable such that the left edge of the stem and the left edge of the

syllable coincide. 'align the left edge of a stem with the left edge of a syllable' If we want to analyze the lack of resyllabification across morpheme boundaries in Standard German in terms of alignment, we have to assume that the insertion of an epenthetic glottal stop violates alignment less than resyllabification of the prefix consonant. We can express this assumption informally by assigning less Alignment violations to misaligned structures containing an epenthetic consonant than to misalignment triggered by a prefix consonant: Tableau 1. Alignment ofleft stem edges in Standard German Vor-ort e1r

ONSET

a. Vor.Jl-ort b. Vor.-ort c. Vo.r-ort

ALIGN (STEM,

*

L, SYLL, L)

DEP

*

*! **!

The high-ranked constraint ONSET, requiring consonant initial syllables, is satisfied only by candidates a. and c. Between them, ALIGN chooses candidate a. which displays only an epenthetic consonant between the left morpheme edge and the syllable boundary. Epenthesis takes place at the cost of the low ranked constraint DEP, disfavoring the insertion of epenthetic elements. Ifwe now consider the different strategies of past participle formation in Mocheno, we see that the alignment constraint requiring coincidence of left edge and syllable boundaries is obeyed in all forms and that it determines the choice of the allomorph in some contexts (s. Kager 1996; Drachman, Kager & Malikouti-Drachman 1996; Klein 2003; Bonet, Lloret & Mascaro 2007 for other analyses of allomorphywhere alignment of a morphological and a prosodic category plays a role). Before roots beginning with a sonorant, a vowel, a voiced stop or [h], the presence of the prefix ga- allows the left edge of the verbal root to be perfectly aligned with a syllable edge. With sonorant-initial roots, alignment would be violated if the allomorph [-cont, -voice] was realized, e.g. through a voiceless stop with the same place features as the stem-initial sonorant. For instance, the participle for rearn could then be realized as *t-reart, in violation of the alignment constraint. In these cases, as will be discussed in the analysis below, the choice of the allomorph is conditioned by the alignment constraint.

45

46

Birgit Alber

Roots beginning with a labial or coronal fricative form their past participle by creating an initial affricate. If we assume that these affricates are complex segments, we can conclude that this strategy of past participle formation also leads to a well-aligned structure. 'Ihe leftmost segment of the root- which is the consonant associated to the affricate coincides "'ith the left edge of a syllable: 14 (14)

Alignment in verbs with root-initiallabial or coronal fricatives: ~·iern ~

pjiert

/[-cont, -voice]+

I

I

p

""

v/

f [0

C

/

Finally, if we assume that the past participle of verbs beginning with a voiceless stop are created through coalescence of the allomorph [-cont, -voice) and the first segment of the root, alignment of the root with a syllable edge is obeyed onc.e more, since the initial segment of the root which corresponds both to the root--initial segment and to the past participle allomorph- coincides with a syllable edge: (IS)

Alignment in verbs with root-initial voiceless stops: p11izr1 ~ pisr1 /[---cont. ---voiceh

"" Alignment explains why a gV- prefix is inserted before sonorants, and we see that this constraint is also obeyed when an affricate is formed before fricatives and when coalescence occurs with voiceless stops. Yet, it does not explain the distribution of the two allomorphs in every one of these context Why, for example, don't we insert ga-- before voiceless stops? After all, a torm like *ga.pisn would be syllabically well-formed, respect alignment of the stem with a syllable boundary and furthermore realize the prefix in a more transparent way than through coalescence. I propose that the answer to this question is that we have to consider [---·cont, -voice] to be the default allomorph of past participle f{)rmation in Mocheno. It will be

Note that this conclusion is independent of whether or not we assume that affricates have a two- or a one-root representation (see Clement'> & Hume 1995 for an overview), as long as we assume that alignment is calculated on the segment to which the single root node or, alternatively, the two root nodes are linked.

14.

Past participles in Mocheno

inserted whenever possible. Only when the insertion of [-cont, -voice] leads to the violation of high-ranked constraints in the language such as the constraint requiring a stem to be aligned with a syllable boundary, will the other allomorph, ga-, be chosen. The reason why [-cont, -voice] is the default allomorph is that it is less marked than the allomorph ga- with respect to voicing, since it bears the (unmarked) feature [---voice) while ga- contains a (marked) [+voice) consonant (s. the discussion in Section 6 ). Although alignment seems to play some role in the formation of past participles in Mbcheno, it is not the case that the constraint ALIGN (STEM, L, SYLL, L) is observed in the same contexts as in Standard German. Thus, differently from Standard German, in Mocheno no glottal stops are inserted in hiatus contexts at the prefix-stem boundary (examples a. below; cf. also Rowley 1986: 77) and resyllabification takes place between the prefix and the stem (examples b. below): 15 (16)

a.

Unresolved hiatus contexts: no.-nrbatn garo.-aus

b.

lo round off~ refine' 'straight ahead'

Resyllahification between prefix and stem: i.n-ntmn au.z-nstn va.r-nrbatn unt:l.r-oum u.r-nltn

'to breathe in:' 'to take off branches' 'to process, c.onvert' 'upside down' 'ancestors'

These data point to a language where onsetless syllables are not repaired in hiatus contexts, but misalignment of stems with syllable boundaries is tolerated in order to provide a syllable with an onset. Expressed in a constraint hierarchy, we will have a ranking where the faithfulness constraints MAx and DEP, which, respectively, militate against deletion of input material and insertion of epenthetic elements, will dominate the constraint ONSET, requiring syllables to have onsets. This constraint in turn will dominate ALIGN (STEM, L, SYLL, L). (17)

Constraint hierarchy for stem-prefix alignment in Mocheno: >>ONSET>> ALIGN (STEM, L, SYLL, L)

MAX, DEP

Thanks to Leo Toller for helping me find these examples and testing the syllabification for me; transcriptions are given in his native variety of Palai.

15.

47

48

Birgit Alber

The following tableau illustrates the constraint interaction: Tableau 2. Unresolved hiatus and misalignment of stems and syllables /no-orbat-n/

MAX

DI!P

ALIGN (STI!M,

L, SYLL, L)

...

..,. a no.orbatn b. nor.batn

ONSI!T

"'!

c. no. ?orbatn

,.

"'!

I in-otn-n/ ~

a.

i.n-otn~n

b.

in.-otn~n

c.

in.-tn~n

d

in.-?otn~n

...... "'! "'! "'!

...

The example no.orbatn illustrates that onsetless syllables are tolerated in hiatus contexts. They are neither repaired through deletion (candidate b.) nor through epenthesis (candidate c.). The evaluation of the example i.n-Dtnan shows us that onsetless syllables are repaired through resyllabification, as in candidate a., thus violating the alignment constraint twice. Not repairing the onsetless syllable, as in candidate b., or repairing it through deletion or epenthesis, as in candidates c. and d., is not an option. Alignment of a stem with a syllable boundary thus is a subordinate principle in the language as a whole. Yet it does play a certain role in the formation of past participles. The subordinate role of morphoprosodic alignment at the stem boundary is in line with the patterns that we find in Southern German varieties, where glottal stop epenthesis between stems and prefixes seems in general to be more restricted than in Standard German and resyllabification seems more readily available (Alber 2001). I interpret Mocheno as being most radically 'Southern' in the sense that the effects of morphoprosodic alignment are practically invisible in the language, except for the role that the alignment constraint plays in past participle formation. In the analysis in Section 6 I will show how the partial hierarchy established here for stem-prefix alignment in general interacts with other constraints in determining the distribution of allomorphs.

5·

Distribution of voiced and voiceless obstruents in Mocheno

Before turning to the analysis proper, it is necessary to take a closer look at the distribution of obstruents in Mocheno, since the partial hierarchy regulating it will play a role in allomorph selection in past participle formation. First of all, I will assume that obstruents in Mocheno, if at all, contrast for [±voice], since, as other Southern varieties of German, Mocheno does not exhibit any audible aspiration on stops (except, in some cases, on [k], see Footnote 24). In this sense it differs from Standard German, for which it has been proposed that stops contrast for the

Past participles in Mocheno 49

presence or absence of the feature [spread glottis] (for recent proposals in these terms see Jessen & Ringen 2002). Second, voiced and voiceless fricatives in Mocheno differ from voiced and voiceless stops in that they are in complementary distribution. 'Ihis means that the feature [±voice] is contrastive only in stops, not in fricatives. Thus, in the native lexicon, we find voiced [v, z] word-initially, when preceding a vowel or sonorant (ex. a. below) and word-medially either if they follow a so norant or if they follow a long vowel (ex. b below). We find voiceless [f, s] word--medially after a short vowel (ambisyllabic con-text, c.) and word--finally (ex. d.; examples are from Rowley 1982, 1986: 122ff.): (18)

a.

voiced [v, z]: word--initially, preceding a vowel or sonorant z:>k

vresn b.

'to say, 3.P.so: 'to eat (of animals)'

voiced [v, z]: word-medially, following a long vowel or a sonorant berven 'to throw' hdvrn 'to help' Jlo:vrn 'to sleep'

ri:vl c.

voiceless [f, s] :word-medially, after short vowels

Jafl pes:lr bas~rn

d.

'scab'

'little tub' 'better' 'to water'

voiceless [f, s]: word-finally birf i hilf Jlo:f

'to throw, imperative' 'to help, lP.SG.' 'sleep, noun'

tbf

'deep'

1he postalveolar fricatives [s, z] contrast with other sibilants only in the varieties of Vlarotz and Oachlait (and there only in few words) and are allophones of [JJ before stops in the variety of Palai. In the varieties where they are contrastive, their distribu-tion is similar to that of the fricatives [f/v, s/z] (Rowley 1986: 127). Since they are not contrastive in all varieties, I will ignore them in the following analysis of the distribution of voiced and voiceless obstruents. The palatoalveolar fricative [J] and the glottal fricative [h] are always voiceless .16 Summarizing, none of the fricatives exhibits a clear voicing contrast, at least when the native lexicon of Mocheno is considered.

16. Voiced [3] seems to appear sometimes when Iff undergoes assimilation to an adjacent voiced consonant (e.g. yne:r, 'grease: pe:ryz, 'Pergine, place name: Rowley 1986: 136), but Rowley does not note an assimilation of this fricative systematically. For example, verbs begin-ning with a sibilant--sonorant duster are transribed always \Vith voiceless [f] in Rowley 1982 (e.g. Jnai11, tfni:tn, 'to cut, cut past part.'). Pending clearer data on voicing assimilation of [fj I will ignore voiced realizations of this fricative.

50

Birgit Alber

The distribution of voiced and voiceless fricatives in Mocheno is reminiscent of that described for a variety of West Germanic dialects discussed by van Oostendorp 2003 (see also Bannert 1976 for the same phenomenon in Middle Bavarian). Van Oostendorp analyz~s Frisian, Thurgovian German (an Alemannic variety) and Roemtond Dutch, which tend to have an complementary distribution of fricatives in word-medial context: voiced fricatives follow short vowels and voiceless fricatives long vowels. Van Oostendorp concludes that fricatives in these varieties do not really differ in voice, but rather in length. Long fricatives appear after short vowels in order to dose the preceding syllable, in obeyance ofProkosch's law, which demands stressed syllables to be heavy. This leads to ambisyllabic structures as e.g. Mocheno (fajl] 'little tub: where the word-medial fricative is voiceless because in closing the preceding syllable containing a short vowel it undergoes final devoicing. On the other hand, syllables containing long vowels or closed by a sonorc:IIlt are heavy by themselves, hence obey Prokosch's law and therefore do not have to be closed by a following fricative. Thus, the word-medial fricatives in structures such as [hel.ven]'to help' and [flo:.ven]'to sleep' are not ambisyllabic and therefore do not undergo final devoicing. On the contrary, since the fricatives appear in a sonorant context, they will be voiced. Adapting the core of van Oostenorp's analysis, the complementary distribution of fricatives in Mocheno can be captured by the following constraint hierarchy: (19)

PROKOSCH's LAW, ~·GEM VOICE>> VOICE....SON >>"[VOICE)>> lDENT(VOICB)··fRIC

a. PRoKoscu's l..Aw: stressed syllables are heavy b. vmcB_SoN: an obstruent preceding a S<morant is voiced c. *[vorcn]: obstluents are not voic.ed d. lDBNT(vmcE)-FRIC: Correspondent fricatives have the same value for the feature [:!:voice] e. *GnMVmcE: no voiced geminate c.onsonants

The hierarchy characterizes a language where voicing cannot emerge as contrastive in fricatives since the markedness constraint " [VOICE) dominates the faithfulness constraint lDENT(vmcE)-FRIC, responsible for preserving the input value for [±voice] in fricatives. Word-finally, fricatives will therefore be voiceless (ex. d. above). On the other hand, voicing is mandatory before vowels and sonorants because of the constraint voiCE_SoN. Since VOICE_SoN dominates *[vOicE], we will have voiced fricatives in word-initial position preceding a vowel or sonorant (ex. a. above) and word-medially, when the fricative follows a heavy syllable which contains a long vowel or is closed in a consonant (ex. b. above). When the fricative follows a short vowel (ex. c. above), the fricative has to be long in order to close the preceding light syllable, thus satisfYing PRoKoscH's LAw. I will assume with van Oostendorp that the long fricative is attac.hed to two positions (two slots or two root nodes), having the structural representation

x.

Past participles in Mocheno of a geminate. As a geminate, it is targeted by the constraint *GEMVOICE, prohibiting the realization of voiced geminates, which are universally marked. Therefore the long consonants following short vowels emerge as voicelessP The following tableau summarizes the analysis of fricative distribution in Mocheno:

Tableau 3. Distribution of voiced and voiceless fricatives u

;

~

IJ.

j

·"'u

= lS

~ ~

""

u ....

~

rJl

C!l

6

~ *

il~

z

"''I

u

>

'iil' ~ 0

~

*

/hilv/ or !hilv/ ~a. hilf

vrtsn

b. frtsn

Q

(")

,.

(*) (*)

,.

hcl.wn

b. hcl.frn

"!

lfafll or 1/avl/ ~ a. Jaf.fl b. Javl c. Jav.vl

-

"!

"!

/hclven/ or !hclfrn/ ~a.

'~"'

(*)18

b. hilv /vrtsn/or /fr<-sn I ~a.

0 ~

,.

,.

*! *!

("") (*)

,.,.

("") ("") ("")

I do not follow van Oostendorp in attributing the voiceless realization oflong fricatives to a constraint FD (final devoicing). Final devoicing is often analyzed as the result of a partial hierarchy where a positional faithfulness constraint dominates a markedness constraint (Le. IDENT-VOICE (ONSET) » "[voiCE] » IDENT-VOICE; Lombardi 1999). If we combine this partial hierarchy with the ranking voiCE_SoN » "[voiCE], required to explain word-initial voicing, we obtain a hierarchy which, as it is, would predict long fricatives to be realized as VOiced, in satisfaction of IDENT-VOICE (ONSET) and VOICE_ SON. A constraint like *GEM VOICE is therefore required in any case.

17.

Since we cannot determine whether the underlying fricative in these examples is voiced or voiceless, we cannot determine whether the constraint IDENT is violated or not. What is crucial is that if a voiced fricative was present underlyingly, it could never emerge because of the dominant position of "[voicE].

18.

51

52

Birgit Alber

Summarizing, fricatives are voiceless by default, given the ranking of the markedness constraint *[vOicE] over the faithfulness constraint IDENT(VOICE)-Fruc. This explains the voicelessness of word-final fricatives as in hilf. They are voiced wordinitially, as in vrrsn, and word-medially, as in hel.w;n, when preceding a sonorant. because VOICE_SoN outranks *[vmcn]. However, word--medially after a short vowel, as in.faf.jl, PRo KoscH's LAw demands a long consonant. This long consonant can only emerge as voiceless bec.ause of*GEMVOicE. While voiced and voiceless fricatives are in complementary distribution, the situ·· ation is different for stops, who contrc:~st in voicing in all contexts (except for /d/, which is not attested word-finally, see Rowley lllff.).l9 According to Rowley, stops do not even undergo syllable final devoicing (Auslautverha1'tung) as the majority of German dialecb do. We can describe the different situation of fricatives and stops by assuming that they are targeted by two different faithfulness constraints, one requiring the preservation of the input specification of the feature [±voice] for stops. the other requiring it for fricatives (see Grijzenhout & Kriimer 2000 for a similar proposal): (20)

lnENT(vmcE)-SToP: Correspondent stops have the same value for the feature [±voice] IDENT(vorcB)··FRIC: Correspondent fricatives have the same value for the feature [±voice]

We can then assume that while IDENT(VOICB)-Fruc, as determined in the analysis above, is ranked below the markedness constraint ~[+vmcn], the constraint lDENT(VOICE)-ST(W is ranked above it: (21)

lDBNT(VOICE)-STOP >>*[+VOICE]>> IDBNT(VOICE)-FRIC

In fact, since stops do not exhibit the same word--medial and word--initial distribution as fricatives, we can assume that IDENT(VOICE)-STOP is ranked also above the other constraints leading to a complementary distribution of fricatives: (22)

IDBNT(VOICB)-STOP >> PRoKoscH's LAw, *GBMVoiCB >> VOICB_SON >>*[+VOICE]>> IDBNT(VOICB)-fRIC

The top position of IDBNT(VOICE)-STOP guarantees that stops are contrastive in all contexts. They do not participate in the length contrasts generated by PRoKoscH's LAw and they are not voiced before so no rants. Fricatives, on the other hand, cannot faithfully realize their input specification since IDENT(VOICE)-FRic is dominated by the markedness constraint *[+vOicE), which will favor a voiceless realization of

19. Rowley (1986: 111) mentions assimilation of voicing tor stops, but almost all the exam-· ples are cases where assimilation of voicing occurs across word boundaries, hence this might be a phenomenon of postlexical phonology.

Past participles in Mocheno

fricatives except in the contexts where the constraint VOIGE_SoN leads to voicing of fricatives. As we will see in the next section, the hierarchy established here for the distribution of voicing will play an important role also in allomorph selection. Specifically, the constraint*[ +voicE] will not only guarantee devoicing of fricatives, but it will also be the constraint that favors the allomorph [---cont, ---·voice) over the allomorphga-. Furthermore, the ranking of IDENT(VOICE)-STOP over IDE~'T(VOICE)-Fmc will allow us to explain the different allomorph selection tor root--initial voiced and voiceless stops.

6.

Analysis

In the analysis of phonologically conditioned suppletive allomorphy in Mbcheno I will follow the majority of the literature on the topic in assuming that suppletive allo-morphs are specified in the input and that the constraint hierarchy evaluates the collection of the individual candidate sets of each allomorph. 20 My proposal for the basic constraint hienrrchy responsible for allomorph selec-tion in Mocheno past participle formation is as follows: (23)

Constraint hierarchy ior allomorph selection: MAx,

DEP

ONSET

I ALIGN (STEM,

-

L, SYLL, L)

IDENT(VOICE)-STOP

------

-.............. *[+VOICE]

I

lDENT(VOICE) .. FRIC

Before going into the details of the analysis, let us take a step back and look at the geometry of the hierarchy. It is obvious that this hierc:U'chy is the simple result of pasting together the hierarchy proposed for alignment between stems and syllable boundaries and (part of) the hierarchy proposed for the distribution of obstruent voicing. 1his means that (i) the proposed hierarchy is compatible with pieces of the Mocheno grammar which have been established independently and, more importantly, (ii) the

20. See Mascar6 2007: 718 where this assumption is made explicit. Wolf 2008 proposes a somewhat different approach assuming that morphological feature structures are the only elements present in the input. Allomorphs are then morphs which are associated to iden-tical morphological feature structures. MAx constraints make sure that at least one morpho-logical feature structure is realized while markedness constraints decide which allomorph will be realized.

53

54

Birgit Alber hierarchy gives us an argument in favor of not interpreting allomorph selection as a simple process of subcategorization. If we were to propose that the two allmorphs of past participle formation are subcategorized for specific phonological contexts, we would discard an explanation that comes for free. In fact, two pieces of the grammar of the language which have been established independently from allomorph selection- when ranked among each other- can explain by themselves the selection of the allomorph, without any additional assumptions. The core of the hierarchy which is responsible for allomorph selection is the ranking of ALIGN (STEM, L, SYLL, L) and IDENT(VOICE)-SToP over *[+VOICE]. It tells us that, everything else being equal, we will prefer the voiceless allomorph [-cont. -voice] over the voiced allomorph ga-. However, if this choice leads to a violation of ALIGN (STEM, L, SYLL, L) or changes the input specification for voicing of a stop, we will select ga-, instead. In what follows I will discuss the single contexts of past participle formation beginning with those where [-cont. -voice] is selected. As outlined in Section 4, I will assume that the past participle of verbal roots beginning with voiceless stops is formed through coalescence of the allomorph [-cont. -voice] with the root initial segment (cf. structure (15) above). A structure of this type violates the anti-coalescence constraint UNIFORMITY, which will be ranked below the constraint*[ +vOicE]: (24)

UNIFORMITY: no output element has multiple correspondents in the input (no coalescence). (McCarthy & Prince 1995)

The following tableau illustrates the evaluation of the coalescence structure with respect to its competing candidates: Tableau 4. Coalescence before a voiceless consonant: paizn /{[-cont, -voice], ga}- paiz- n/

~

pisn

~

~ ~ 0

~

Q ~

<

~

..i

Ji

..;

:i I'll ~ z

>o

'gj' (.)

0

:>

..±. *

!:: ::!l

.......~

~

0 .... ....

< ~

a [-cont, -voiceh-

""'

/ ·Pt. 2

b. [-cont, -voice] 1 - p2

I ·Pt c. ga-pisn

>t

P2

*!

I

- p2 *!

Pa~t

participles in Mocheno

Following .Mascaro 2007 and Bonet, Lloret & Mascar6 2007, the input of the past participle is represented as consisting of the verbal root, the past participle suffix -n and the two past participle prefix allomorphs. The set of candidates evaluated by the hierarchy consists of possible realizations of both allomorphs. None of the three candidates violate the faithfulness constraints MAx and DEP, since all of them realize faithfully either one or the other allomorph without deleting input material or adding material that in the input was not present. With respect to candidate b., I am assuming that it agrees in its place features with the first segment of the root and therefore does not incur a DEP violation either. The first constraint making some decision among the candidates is ALIGN (STEM, L, SYLL, L). This constraint eliminates candidate b. from the list of competitors. 21 Candidate a. and c. would both guarantee alignment of the root edge with a syllable boundar}~ but a. is chosen over c. since a. obeys ~·[+VOICE] by realizing the voiceless allomorp h [-cont, -voice]. For this reason candidate a. wins although it violates the low ranked anti-coalescence constraint UNIFORMITY. The constraints lDENT(vorcE)-STOP and lDENT(vorcE)- Fmc have not been integrated into the tableau since they play no role here: none of the candidates exhibits any change of the input specifications of voicing. Note, however, that the ranking IDENT(vorcE)-STOP >> *[+VOICE] can eliminate candidates like ka-pisn, where the stop of the prefix ga- has been devoiced in order to obey* [+VOICE]. 'lhe second context where the allomorph [-cont, -voice] is chosen is when the verbal root begins with a fricative. ln this case, the participle is tormed by creating an initial complex segment, an affricate. Similarly to the coalescence structure, the affricate allows for alignment of the left root edge with a syllable boundary (cf the discussion of structure (1_4) ). The affricate structure itself will also violate certain constraints, such as a constraint against complex segments, which I will call ~·coMPLEXSEG. Furthermore, to realize the allmorph [-cont, -voice] in viern ~ pjiert, the first segment of the verbal root has to be devoiced. This means that the faithfulness constraint lDENT(VOICE)··FRIC, which favors the preservation of input voicing specifications of fricatives is violated in these cases. 22

Note that candidate b. could be eliminated also by a number of other constraints mili21. tating against initial geminates or stop clusters in onsets. However, ALIGN (STEM, L, SYI.J.., L), can eliminate this candidate and we will see that its position in the hierarchy will be crucial to decide between candidates in other phonological conte:x1:s. In principle, the underlying fricative of a verb like viern could be either voiced or voice22. less. Assuming that it am be voiced underlyingly, we have to account tor the fact that also this underlying structure is reali7.ed as voiceless in the affricate. We will not discuss the trivial case, where the underlying fricative is voiceless and is reali7.ed as such.

55

56

Birgit Alber

Tableau 5. Affricates before root-initial fricatives: viern /{[-cont, -voice], ga}- vier- t/

~

pjiert

~

~
z

...l'

""'

0

"' ..;

~

:i

Q

~

<

~

f::l

;..

..."" ~ z<:.!)

Iii'

~

:>

.±. *

a. [-cont, -voice] 1 -

Q,

:a 0 u *

~

"gl

~

~

...z ""

-.. ~

I

p

f

""'-

.c

/

b. [-cont, -voice] - v

I

.p c. ga.- vier-t

~

.

v

I

"' """"'

:::3

< ~

~

"!

I

- f

"!

None of the three candidates incurs violations of the top-ranked constraints MAx, DEP or ONSET. Candidate b., a structure where the allomorph does not form a complex segment with the first segment of the root, is excluded because of its violation of the alignment constraint Candidate c., which realizes the second allomorph, would be equally well-aligned as candidate a., but loses because it violates*[ +vOicE]. Candidate a. wins, even though it collects one violation of *CoMPLEXSEG and violates IDENT(VOICE)-FRIC. Note that the ranking that generates devoicing of the root-initial segment is not quite as simple as represented here. First of all, we must exclude the possibility of affricates in which the stop element and the fricative element disagree in voicing, as e.g. in the hypothetical affricate *pv. These are affricates (or, indeed, tautosyllabic obstruent clusters) which seem to be ruled out universally. Second, the ranking has to select a structure where the root-initial segment is devoiced and rule out a competitor where the past participle allomorph is instead voiced. Hypothetical affricates of this type, i.e. *bv-, *dz, *d3 can be excluded by the fact that voiced affricates are not licit structures in Mocheno and hence can be assumed to be ruled out by some high-ranked constraint.23

23. The affricate [d3] is attested in Mocheno, but is limited almost exclusively to loan words from Italian. Rowley (1986: 144f., 2003: 57) quotes only two native words containing it, indyr, 'our' and biald3ar, 'mole:

Past participles in Mocheno

There is one root-initial fricative which is different from the other fricatives in that it is prefixed by the allomorph ga- instead of forming an affricate with the allomorph [-cant, -voice]. This is the fricative [h], as in the example hupfan -7 gahupft. I will assume that the creation of an affricate is not an option in this case, since [kh] or [kx] is not a possible affricate in Mocheno. 24 Hence, the allomorph [-cant, -voice] cannot be selected before a root-initial [h] without creating an alignment violation and the allomorph ga- is chosen instead. Let us now consider the selection of the second allomorph, ga-, with root-initial sonorants, vowels and voiced stops. In the case of sonorants, ALIGN (STEM, L, SYLL, L) emerges as the decisive constraint, as illustrated in the following tableau:

Tableau 6. ga- before root-initial sonorants: rearn ~ garwrt /{[-cont, -voice], ga}- rear-t I

MAX DEP

ALIGN (STEM,

L, SYLL, L)

*[+VOICE]

ONSET

a. [-cont, -voice] - r

I

t - b. ga.- reart

*!

I r

*!

Candidate a. realizes faithfully the allomorph [-cant, -voice]. We can imagine it to be realized by a consonant which is assimilated in place to the following sonorant (as was the case with root-initial fricatives and voiceless stops), thus not violating any constraint disfavoring insertion of material not present in the input. Since we have assumed the allomorph to contain a root-node, its projection of a C-slot will not violate DEP either. However, since a faithful realization of the [-cant, -voice] features is not possible any more through an affricate or through coalescence and [-cant, -voice] has to be realized as a segment of its own, the structure necessarily incurs a violation of the alignment constraint requiring coincidence of the left edge of the verbal root with a syllable boundary. Since ALIGN (STEM, L, SYLL, L) cannot be fulfilled through faithful realization of [-cant, -voice], the other allomorph, ga- will be chosen. The presence of ga- will create a well-aligned structure, at the cost of violating the constraint *[+vOicE]. The behavior of sonorant-initial roots thus shows us that alignment of the left stem edge to a syllable boundary does play a role also in Mocheno: it drives the selection of ga- over the default allomorph [-cant, -voice].

24. There is a weakly contrastive aspirated velar stop /kh/ in Mocheno, which is sometimes pronounced as [kx] and often varies freely with [k]. Rowley (1986: 113f.) interprets it as a simple consonant, not as an affricate.

57

58

Birgit Alber

The role that alignment plays here in allomorph selection is a typical case of the Emergence of the Unmarked (TETU, McCarthy & Prince 1994) in the context of allomorph selection, a phenomenon observed in many analyses (see, among others, Kager 1996; Mascar6 1996, 2007; Rubach & Booij 2001; Bonet, Lloret & Mascar6 2007). TETU effects have been observed most clearly in reduplication or truncation, where, in the analysis of McCarthy & Prince 1994, 1995, the faithfulness constraint in a F-IO » M ranking is vacuously satisfied in reduplicants or truncation morphemes targeted by an M » F-R/TB ranking, since reduplicants or truncation morphemes are not subject to an input-output correspondence relation. 25 1he reason unmarked structures can emerge in allomorphy selection under an F >> M ranking is somewhat different: in the case of allomorphy, faithfulness constraints get two (or more) chances to be fulfilled, since they are satisfied by the realization of any of the underlying allomorphs. Since either of the allomorphs satisfies faithfulness, markedness can choose the most unmarked among them. For Mocheno this means that although in the overall language the faithfulness constraints MAx and DEP dominate the markedness constraint ALIGN (STEM, L, SYLL, L), which therefore does not seem to have any effect, the constraint emerges in a context where MAx and DEP can be satisfied in more than one way, i.e. through the faithful realization of more than one allomorph. Since MAx and DEP cannot decide between the two allomorphs, the decision can be passed to a lower markedness constraint, in this case ALIGN (STEM, L, SYLL, L). In the context of root-initial voiceless stops and fricatives, the alignment constraint will not make any decision, since both allomorphs would guarantee alignment of the stem to the syllable edge. The choice of the allomorph is decided by *[+voicE]. In the context of root-initial sonorants, though, only ga- fares well in terms of alignment, and hence is preferred over [-cont. -voice], even though it violates *[+vmcE]. Let us now consider the evaluation of the context of root-initial vowels: Tableau 7. ga- before root-initial vowels: o:tnan

I {[ -cont, -voice], ga}- o:tn-n I

a [-cont. -voice] - o:

I

.tn ~

b. ga.-o:tnt

I

o:

MAX

~

.. DBP . . .. I I

gao:tnt ONSET

..!

ALIGN

,.

..(+VOICE]

I I I I

.

,.

Candidate a. represents an attempt to realize the morpheme [-cont, -voice] in the most faithful way, by inserting a voiceless stop bearing an unmarked place feature,

25. F-10 = some faithfulness constraint relating input to output, F-R/TB = some faithfulness constraint relating a reduplicant or a truncation morpheme to its base, M = some markedness constraint.

Past participles in Mocheno i.e. a coronal [t] or a glottal [?]. 26 This insertion of a feature not present in the input leads to a violation of the constraint militating against insertion, DEP (or, probably more precisely, DEP-FEATURE, since only a place feature is inserted), excluding candidate (a) in favor of candidate (b) The constraint ALIGN (STEM, L, SYLL, L) is not the constraint choosing between the allomorphs, here, since at the prefix-root boundary, as we know from Section 4, misalignment is in principle allowed in order to satisfy ONSET. However, epenthesis is not, and since [-cont, -voice] is a subsegment, which requires the insertion of a place feature to be realized, it is excluded. 27 Note that if we were to assume that the underlying form of the allomorph was /k-/, we would expect this segment to show up in at least this context. The last context in which we find the allomorph ga- is before voiced stops, as in the example bisn -7 gabist. The challenge here is to explain why coalescence is not a possbility in this context as it is before voiceless stops. Why do we have the pattern bisn -7 gabist but paizn -7 pisn? My proposal is that a hypothetical coalescence candidate, which in the case of root-initial voiced stops would be something like bisn -7 pist is ruled out by IDENT(VOICE)-STOP. This constraint, as we know from the analysis of obstruent distribution in Section 5, is ranked above *[+vOicE] and hence blocks *[+vOicE] from selecting the allomorph [ -cont, -voice]: Tableau 8. ga- before voiced consonants: bisn /{[-cont, -voice], ga}- bis -t/

MAX

~

gabist

ALIGN

IDENT(VOICE)-STOP

*[+VOICE]

DEP ONSET a. [-cont, -voice ] 1 -

""'

/ ·P1.2

b. [-cont, -voice] 1

I

.pi

*!

P2

-

p2

*!

I

- b2

- c. ga.-bist

26.

*

I am assuming here with de Lacy (2006) that glottal stops are not placeless.

27. The fact that DEP and not ALIGN chooses between the allomorphs in this case, points to another possible analysis of Mocheno past participle formation, which does not make use of alignment. If in fact we were to assume that the default allomorph consisted oftwo independent floating features, [-cont] and [-voice], not connected by a root node, we could assume that the allomorph ga- was chosen in all those cases where realizing [-cont] and [-voice] required the insertion of a root node. Thus the burden of the choice of the allomorph gawould lie entirely on the constraint DEP. I have not followed this line of analysis since I find it difficult to conceive the allomorph [-cont, -voice] as not linked by anything and I do not know of cases where two floating subsegments combine to a single morpheme.

59

6o

Birgit Alber

For a coalescence structure as in a. to be realized, the voice value of the first segment of the root (or, in another possible coalescence candidate bisn -7 bist, the voice value of the defaultallomorph) has to be changed. This change, however, leads to a violation of the constraint IDENT(voicE)-SToP, which is tatal, since lDENT(W.liCE)-Sn)p, difterently from lDE~'T(VOICE)--FRrc, is ranked above the constraint ~[+vOicE]. For this reason, changing the voice value of a stop is not an option, even if this means that *[+vOicE] has to be violated by selecting the allomorphga-. Candidate c.., which realizes the 'second-best' allomorph but rates well on alignment and faithfulness, emerges as the winner. The non-coalesdng candidate b., as above, is eliminated because of its violation of aligmuent (and of numerous other syllable-structure constraints). As in the case of root-initial sonorants, where the insertion ofga- is due to ALIGN (STEM, L, Sur., I..), also in this case the choice between the two allomorphs is determined by a partial hierarchy which has been determined independently to hold in the language, i.e. IDE~'T(vOic)-SToP >>*[+vOicE].

7·

Conclusions

Mocheno displays a variety of strategies in forming the past participle of verbs, which can be described as prefixing of an allomorp h [ -cont, -voice] to roots beginning with a voiceless stop (resulting in coalescence) and labial and coronal fricatives (resulting in affrication) and prefixing an allomorph ga- elsewhere. The distribution of the two allomorphs would seem puzzling, since the consonantal allomorph [ -cont, -voice] is found before obstruent~ while the CV allomorph ga· is found before sonorants. From the point of syllable wellformedness we might expect the contrary. However, it has been shown that an analysis in terms of optimization is possible, since allomorph selection can be interpreted as being driven by a well-defined set of markedness and fathfulness constraints which play a role in other parts of the Mocheno grammar. Thus, the constraint *[+VOICE] will favor the allomorph [ -cont, -voice] over ga-, because of its unmarked voicing feature, while ALIGN (STEM, L, SnL, L) will select ga- before sonorants, where the realization of [-cont. -voice] would lead to misalignment between the left edge of the verbal root and the syllable boundary. In addition, the taithfulness constraint DEP favors ga- in the context of vowel-initial roots and IDENT(vorcE)-STOP favors it before root-initial voiced stops. The overall hierarchy determining the distribution of allomorph selection consists of two partial hierarchies which are relevant for pieces of the grammar of the Ianguage independent of past participle allomorph selection. They determine on the one hand the phenomenon of(mis)alignmentat the left edge of stems and on the other the distribution of voiced and voiceless obstruents. We can therefore claim that an analysis in terms of optimization is not only possible, but also desirable, since (i) it makes use

Past participles in Mocheno

of constraints which are either plausibly universal (*[ +vmcE], DEP, lDENT(VOICE)STOP) or active in other German varieties (ALIGN (STEM, L, SYLL, L); (ii) the grammar determining the distribution of allomorphs in Mocheno consists of constraint rankings which hold for the language in general. If on the other hand we were to subcategori7.e each allomorph for the phonological context it occurs in, the relationship between the distribution of past participle allomorphs and other pieces of the Mocheno grammar would remain completely opaque.

References Alber Birgit. 2001. Regional variation and edges: Glottal stop epenthesis and dissimilation in standard and southern varieties of German. Zeitsclzriftfiir Spraclzwissemchaft 20(1): 3-41. Alber, Birgit & Lanthale1~ Franz. 2005. Der Silbenonset in den Tiroler Dialekten. lnPer~pektiven Bin~~ Akten der 1. Tagtmg Deutsche Sprachwissenschaft in Italien (Rom, 2004), Claudio Di Meola, Antonie Hornung & Loren:l,a Rega (eds), 75-88, Roma: Istituto Italiano di Studi Germanici. Andreolli, Claudia. 2004-05. Sonoritat und Stimmhaftigkeit im Silbenanlaut. Bin Fa/lbebpiel aus dem Fersentalerischen, eir1er deutschen Minderheitenspmche des Trentirw, tesi di laurea triennale. Trento: Universita di Trento. Annuario Statistico, Provincia di Trento. . (5 August 2009). Bannert, Robert. 1976. Mittelbairische Phonologie auf akustisclzer tmd perzeptorisc/zer Grundlage. Lund: Libedi:iromedel. Bonet, Eulalia, Lloret, Maria-Rosa & Mascar6, Joan. 2007. Allomorph selecti.on and lexical preferences: Two case studies. Lingua 117: 903-927. Bye, Patrik. 2007. Allomorphy: Selection, not optimization. In Freedom (~Analysis, Sylvia Blaho, Patrik Bye & Martin Kraemer (eds), 63-91. Berlin: Mouton de Gruyter. Clements, Nick & Hmne, Elizabeth. 1995. Internal organization of speech sounds. In The Handbook of'Phonological Theory, John Goldsmith (ed.), 245-306. Oxf(n·d: Blackwell. Cognola, Federica. In press. Alcune considerazioni per tuianalisi unitaria del prefisso ge- del tedesco. In Quaderni patavini di linguistica, Elena Triantafillis (ed.). Drachman, Gaberell, Kager, Rene & Malikouti-D rachman, Angeliki. 1996. Greek allomorphy: An optimality theory account. In Papers }rom the First Conference on Formal Approaches to South Slavic Languages, Plovdiv, October 1995, [University ofTrondheim Working Papers in Linguistics 28], Mila Dimitrova-Vulchanova & Lars Hellan (eels), 345-361. Trondheim: University ofTrondheim. Fleischhacker, Heidi. 2001. Cluster-dependent epenlhesis asymmetries. In UCLA Working Papers in Linguistics 7, Papers in Phonology 5, Adam Albright & Taehong Cho (eds), 71-116. L<.--.s Angeles CA: UCLA. Fleischhacker, Heidi. 2005. Similarity in Phonology: Evidence from Reduplication and Loan AdaptatimL Ph.D. dissertation, UCLA. Giegerich Heinz. 1999. Lexical Strata in English. Aiorphological Causes, Photwlogical Effects. Cambridge: CUP.

61

62

Birgit Alber Giegerich, Heinz. 1989. Syllable Structure and Lexical Deri-vation in German. Bloomington IN: Indiana University Linguistics Club. Grijzenhout, Janet & Kramer, Martin. 2000. Hnal devotcing and voicing assimilation in Dutch derivation and clitl.cization. In Lexicon in f'ocu.~ [Studla Grammatica 45], Barbara Stiebels & Dieter Wunderlich (eds), 55-82. Berlin: Akademie Verlag. Jessen, Michael & Ringen, Catherine. 2002. Laryngeal Features in German. Phonology 19: 189-218.

Kage1; Rene. 1996. On affix allomorphy and syllable counting. In Interfaces in Phonology, Ursula Kleinhenz (ed.), 155---171. Berlin: Akademie Verlag. Klein, Thomas. 2003. Syllable structure and lexical markedness in creole morphophonology: Determiner allomorphy in Haitian and elsewhere. In Phonology and tr.orpiwlogy of Creole languages, Ingo Plag (ed.), 209--228. Thebingen: Niemeyer. Kranzmayer, Eberhard. 1956. Hlstorlsche Lautgeographie des gesamtbairischen Dialektraumes. Wien: Herrmann Bohlau. de Lacy, Paul. 2006. Markedness: Reduction and Preservation in Phonology [Cambridge Studies in Linguistics 112]. Cambridge: CUP. Lapointe, Steven. 2001. Stem selection and OT. In Yearbook of Morphology 1999, Geert Booij & Jaap van Marie (eds), 263-297. Dordrecht: Kluwer. Lem, Notiziario trimestrale del Kulturinstitut Bersntol Lusern, Istituto Culturale MochenoCimbro, I 38050 Palit del Fersina (Tn). N. 27 Nov. 2002, N. 29 Sept. 2003, N. 30 Dec. 2003. Lombardi, Linda 1999. Positional faithfulness and voicing assimilation in optimality theory. Natural Language and Linguistic 1heory 17: 267---302. Mascar6, JoaiL 1996. External allomorphy and contractions in Romance. Probus 8: 181--· 205. Mascar6, Jom 2007. External allomorphy and lexical representatioiL Linguistic Inquiry 38(4): 715--735.

McCarthy, John & Prince, Alan. 1993a Generali7.ed alignment In Yearbook ofMorphology .199.3, Geert Booij & Jaap van Marle (eds), 79-153. Dord.recht: Kluwer. McCarthy, john & Prince, Alan. 1993b. Prosodic Morphology, I: Constraint Interaction and Satisfaction [Technical Repo.rt 3]. New BrunswickNJ: Rutgers Center for Cognitive Science. McCarthy, John & Prince, Alan. 1994. The emergence of the unmarked. Optimality in prosodic morphology. In NBD) 24: Proceedings of the North .Hast Linguistic Society 24, Merce Gonzhlez (ed.), 333-379. Amherst MA: GLSA. McCarthy. John & Prince, Alan. 1995. Faithfulness and Reduplicative Identity. In Uni-versity of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, Jill Beckman, Laura Walsh Dickey & Suzan.ne Urbanczyk (eds.), 249-384. Amherst: GLSA. Mester, ArmiiL 1994. The quantitative trochee in Latin. Natural Language and Linguistic Theory 12: 1-61.

Nevins, Andrew. 2011. Phonologically-conditioned allomorph selection. In 111e Companion to Phonology, Colin Ewen, Elizabeth Hume, Marc van Oostendorp & Keren Rice (eds), 2357-2382. Oxford: Blackwell. van Oostendorp, Marc. 2003. Ambisyllabicity and fricative voicing in west Germanic dialects. In 1he Syllable in Optimality 1heory, Caroline l1ery & Ruben van de Vijver (eds), 304-337. Cambridge: CUP. Paster, Mary. 2006. Phonological Conditions on AffixatioiL Ph.D. dlssertatlon, University of California at Berkeley.

Past participles in Mocheno Plag, Ingo. 1999. _r..furphological Productivity. Structural Constraints in English Derivation. Berlin: Mouton de Gruyter. Ricci, Vittore. 1904. Vocabolario Trentino-Italiano compilato da alcune signorine di Trento col consiglio e con la revisione del prof Vittore Ricci. Zippel: Trento. (Reprinted by Forni, 1989). Rowley, Anthony. 1982. Fersentaier Wi."irterbud:. Fersentaler Dialekt - Deutsch - ltalienisch. Vocabolario del dialetto tedes.:o delia Viille del f'erslnti nel1're,lti,w. Dialetto- Tede.~co­ Italiano. Hamburg: Helmut Buske. Rowley, Anthony. 1986. Fer5{:~ntal (Val Fersina bei Trient/Oberitalien)- Untersuchung einer Sprachinselmundart. Tiibingen: Max Niemeyer. Rowley, Anthony. 2003. Liac.ht as de sproch Grammatica della lingua moc.hena/Grammatik des Deutsch-Fersentaleriscilen. Palu del Fersina: Istituto culturale mocheno--cimbro/Kulturinsitut Bersntol--Lusern Rubach, Jerzy & Booij, Geert. 200 I. Allomorphy in optimality theory: Polish iotation Language 77:26---60.

Tranel, Bernard 1996. French liaison and elision revisited: A unified account within optimality theory. In Aspects of Romance Linguistics. Selected Papers from the Linguistic Symposium on Romance Languages XXIV. Af.arch 10-13, 1994, Claudia Parodi, Carlos Quicoli, Mario Saltarelli & Maria Luisa Zubizarreta (eds), 433-455. Washington DC: Georgetown University Press. Wolf, Matthew. 2008. Optimal Interleaving: Serial Phonology-Morphology Interaction in a Constraint-Based Model. Ph.D. dissertation, University of Massachusetts at Amherst. Zoll, Cheryl 1996. Parsing Below the Segment in a Constraint Based Framework. Ph.D. dissertation, University of California at Berkeley.

63

SECTION 2

Morphology & Lexical studies

Plautdietsch gender Between Dutch and German* Annemarie Toebosch University of Michigan- Flint

In this paper, I desc.ribe the grammatical gender system of Plautdietsch as it relates to referent animacy. Plautdietsch (a settlement dialect of netherlandic Mennonites in the Americas), and more specifically Henderson Plautdietsch (the variety ofPlautdietsch spoken in Henderson, Nebraska), has a mixed syntactic-semantic gender system that is structurally positioned between that of Standard Dutch and High German. It has a three-way masculine-feminine-neuter grammatical gender contrast across gender agreement targets, and exhibits DP-internal hybrid gender assignment (where a noun assigns one gender to one and another gender to another type of DP-internal gender agreement target). I explain that a geometry of hierarchically organi7..ed gender and animacy features is uniquely suited to explain the hybrid gender assignment we find in Plautdietsch, as well as explains a number of other phenomena we observe in gender systems (such as default gender and the flexible gender assignment of inanimate nouns).

1.

Introduction

Gender, in the meaning of 'sex classification: is a characteristic of animate beings (humans, animals, and some plants). In a second meaning of the word, gender is a classification of referents of linguistic expressions. Thirdly, gender is 'noun classification: or as defined in Hockett (1958) a person's mental grouping of nouns "reflected in

*I would like to thank the audience at the Topics in Germanic Linguistics and Philology session of the convention of the Modern Language Association for their helpful comment'>, and the audience at the College of Arts and Sdences colloquium for their cross-disciplinary look at earlier versions of this work. I furthermore thank my anonymous reviewers for their invaluable input. Finally, I would like to thank my informants for their judgments on the data, their sharing of important stories about their linguistic identity, and their time and hospitality, and Lauren Friesen in particular for sparking my interest in Henderson Plautdietsch. 'Ihe usual disclaimers apply.

68

Annemarie Toebosch

the behavior of associated words" called targets. 1 Gender in this third meaning is said to be semantic when it is determined by the meaning of a noun, and grammatical when itis determined by a noun's phonological or morphologicalform. 2 As stated in Corbett (1991) citing Aksenov (1984), all gender systems have a semantic core.3 Grammars difter in the extent to which they have grammatical gender. For roughly the past five centuries, speakers of many varieties of Dutch have decreased their grammatical gender assignment. In the 16th and 17th centuries, when the shift. from grammatical to semantic gender was underway in Dutch, groups of Mennonites from The Netherlands, Belgium, and Germany moved to Prussia, and in later centuries to New Russia (now Ukraine) and the Americas. Their language4 Plautdietsch is a variety of Low Sa..xon5 with influences from Early Modern Dutch,

1. Gender in the second meaning can be seen as the bridge between gender in the first and third meanings. Referents are mental representations of entities in the real world, denoted by linguistic expressions. Referent gender is then a speaker's concept of the sex/animacy classification (male, female, inanimate) of an entity in the real world, expressed as a particular grantmatical gender (e.g. masculine, feminine, or neuter). The same real-world entity may be conceptualized differently by different !>peakers (i.e. have different referential properties), which may in turn lead to different lexical choices. For example, a particular boat may be con· ceptualized in the mind of its owner as having the quality FEMALE and be expressed v.1th the words boat; Lucy, and site. Another person may think of the same boat as INANIMATE and use the words boc1t and it to denote it. 2. I posit that both grammatical and semantic gender are encoded in the grammar, and are in that sense "grammaticaL" Following Chomsky (1995), I distinguish between semantic features (features of linguistic expressio11s that provide information about the referent of these expressions) and properties of referents oflinguistic expressions., which may or may not map onto such features. Examples of referential properties that are not grammatical in English are SMART, or WEARS »OW TIES, or LOVES MATH. There is no morpheme meaning 'smart' on for example the English noun genius or on its associated words, nor docs SMART affect the syntactic computation of genius or associated words. PEMALE on the other hand is grammatical in English (e.g. the morpheme -ess in lioness).

3· Interesting work has been done by Boroditsk.y et aL (2003) that provides evidence for the influence of grammatical gender and its semantic underpinnings on thought. 4· I define LANGUAGE as a set of similar grammars,. and GRAMMAR as the linguistic system in the mind of a speaker. In other words,. I adopt the Chomskian concept of I - language in that I do not consider languages or grammars to exist outside of the minds of speakers. As such, this study does not aim to give a unified account of Plautdictsch, or of the variety of Plautdietsch presented, as much as it aims to be another data point in the search for possible mental grammars. Any reference to language "change" (in terms of gender "loss" or "gain") should be interpreted in this light 5· 'Ihe term~ Low Saxon and Low German are often used interchangeably to refer to the combined West Germanic grammars used in the northeastem provinces of The Netherlands

Plautdietsch gender: Between Dutch and German 69

High German, Russian, and colonial American languages (English in North American communities; Spanish and Portuguese in South America). My study of one variety of Plautdietsch6 - that spoken in Henderson, Nebraska - shows a mixed semantic-grammatical gender system similar to that of Standard Dutch in its semantic gender of pronouns, yet with stronger gram-matical gender in attributive targets and with semantic gender having spread to a wider range of targets in some speakers. In this paper, I describe this system and analyze it in terms of a gender and animacy feature geometry. Crucially, I analyze Plautdietsch hybrid nouns (nouns with different gender in some tar-get classes than in others). as nouns with conflicting referential information reflected in their form and the form of their targets. I show that grammatical gender loss in settlement dialects like Plautdietsch provide us with a unique view of hybrid gender: the fact that in some languages, indefinite articles are more likely to be assigned semantic gender than other attributive targets. Finally, I will compare the Plautdietsch gender system to that of its close relatives (English, Dutch, and High German) and argue that the loss of gender inflection in Henderson Plautdietsch and in language in general is in fact a gain in formal animacy features, and that children acquiring such "deflected" grammatical gender systems are in fact assigning more grammatically complex feature structures to expressions.

2.

2.1

(Socio-)linguistic history of Henderson Plautdietsch Introduction

Germanic languages are moving from three-way masculine-feminine--neuter gen-der systems, through intermediary stages, toward purely semantic gender systems.

and in northern Germany. Competing analyses exist of the relationship between Low Saxon/ Low German and other West Germanic languages. In one, the language is part of the Lm\' Sa.:mn-Low Franconian branch to which also belong Dutch and Flemish (e.g. Bthnologue. org). In another analysi'>, an additional relationship is drawn between the Low Saxon-Low Franconian branch and the Anglo-Frisian (e.g. Epp 1993). Related to the issue of classification are hotly debated questions of "language" and "dialect': Is Plautdietsch a dialect of Dutch, of German, or is it a language of its own? Because I do not believe these to be interesting questions in the sense that they would add to our ultimate understanding of gender in human language, l will not address them here. Where I refer to the Henderson Plautdietsch variety simply as "Plautdietsch", I do so realizing that the gender systems of speakers of other varieties of Plautdietsch may be different.

6.

70

Annemarie Toebosch

Weerman (2008) illustrates this for Standard German, Dutch and English with the example of definite articles as agreement targets:

(der, die, das) German: a three·- way system Dutch: a two-way system (de, het) English: a purely semantic. system (the) As a Germanic language, and with influences from all three of these languages, we expect Plautdietsch to be somewhere on this continuum. Weerman argues that the speed of grammatkal gender loss "runs parallel to the scale oflanguage contact that played a role in the history of the (standard) languages involved: from extensive contact for English,? to a relatively isolated position for High German:' Plautdietsch has a history of extensive language contact. It developed over the past four and a half centuries and moved around the world for the Henderson variety from modern--day Belgium, to The Netherlands and Germany, to Poland, Ukraine, and the US. 1he nature of the contact between Plautdietsch and other languages has not however been one that would have encouraged significant deflection, or gender loss specifically. Even though the contact has been extensive and the differ-ence in gender systems often significant (say, between Henderson Plautdietsch and English), Plautdietsch has not, on a large scale, been the target language of second language learners.8 I will nonetheless give a brief history of Henderson Plautdietsch, hoping this will inspire future research into the ways language contact influenced the development ofthis little-studied lanbruage. 2.2

The homelands

Plautdietsch has its roots in the minds of 16th century Mennonites 9 living in what is today Flanders (northern Belgium), The Netherlands, and northeastern Germany. During the Reformation period, these netherlandic10 Mennonites were persecuted for their religious beliefs, and many tled to what is called the Frisia Triplex: West Friesland

7· Extensive and prolonged contact between Old Norse and Old English, especially at the Danelaw border is seen as the reason for Middle English deflection. 8. Thomason and Kaufman (1988) and Van Coet'iem (1988) posit that second language learners indirectly cause deflection by spreading their nonnative, relatively inflectionally simpler grammars to new generations of LI learners.

9· Called "Waterlanders"then, Mennonites are now called '1Joopsgezinden" (lit: the baptismminded) in Duteb. Mennonites form a cultural and religious group practicing the teachings of Menno Simons and other Anabaptists. One of the tenets of Anabaptism is the voluntary baptism of adult'> rather than of babies (as done in for example Catholicism). 10. I follow Epp (1993) in using the term "netherlandic" to refer to these regions and "Netherlandic"to mean more narrowly"ofthe nation-state of'Ihe Netherlands."

Plautdietsch gender: Between Dutch and German (now called "Friesland") and Groningen in the north of 1he Netherlands, and East Friesland in northeastern Germany. (Unruh 1955). Ancestors of the Henderson Mennonites specifically came from Flanders, Belgium (Voth 1975), 11 among them those who became members of the Groningen Old Flemish Society, a branch of Dutch Mennonites. (Global Anabaptist Mennonite Encyclopedia Online at www.gameo.org). 2.2.1

Language in the homelands

Mennonites in the Frisia Triplex spoke Low Franconian dialects (Flemish or any number of Dutch varieties), Low Saxon dialects, or Frisian, with Frisian and the Low Saxon variety spoken in the Frisia Triplex being nearly indistinguishable at the time. Henderson ancestors would have spoken Flemish- which to this day has kept its three--way gender contrast 12 before coming in contact with the speech of other Mennonites (Dutch, Frisian, Low Saxon) in the Frisia Triplex. With a strong movement toward the standardization of Dutch at the time, and with contact with Flemish contributing heavily to this standard, (Burke 2005) they likely would have had some degree of fluency in Dutch. 16th century Dutch 13 and West Frisian had already lost their masculine-feminine distinction. 14 2.3

Prussia

In the 16th and 17th centuries, threatened by further persecution and encouraged by the promise of religious freedom and economic improvement, 15 netherlandic Mennonites moved from the Frisia Triplex to Prussia, 16 where they would stay for over two centuries. Others joined them fmm Belgium and other parts of'lhe Netherlands. 'lhey settled along the Vistula River, from the Baltic Sea coast to the tow-n of'Ihorn, 150 miles south. The ancestors of Henderson Mennonites settled in Przechovka ("Kleinsee" in German) near the towns of Schwetz and Kulm, just north of Thorn. (Voth). Buchheit

u. Banders counted about 2000 Mennonites in the middle of the 16th century. (van Leuvenstijn, 1997). 12.

De Vogelaer (2009) suggest'i that Flemish gender is becoming more semantic.

When I use the unqualified term "Dutch': I am referring to the standard variety spoken especially in the northwestern part of The Netherlands.

13.

'Ihe exception to this is the variety spoken on the island of Schiermonnikoog, which still has the traditional three-way gender contrast. (Versloot 2008)

14.

15. With their skill in land reclamation and farming, netherlandic Mennonites were invited to farm the wetlands in the Vistula Delta of Prussia. (Epp 1993)

In Prussia, they were first called "Dutch Anabaptists" and later "Mennonites': the term used for members of the more pacifist branch of Ana:baptism. (Epp 1993)

16.

71

72

Annemarie Toebosch

(1978) describes a sub-group that in 1765 moved westward from the Kulm lowlands to found the villages of Brenkenhoffswalde and Franzthal in the Netzebruch region between Kulm and Berlin. 2.:p

Language contact in Prussia

When the netherlandic Mennonites arrived in Prussia, the local population spoke a variety of Low GermanY Unruh (1955) and Epp (1993) state that the Mennonites adopted this West Prussian Low German as their spoken language. For many this would not have been a big jump from the Low Saxon they had alre-ady been using in the Frisia Triplex. Duerksen (1967) on the other hand states that the Mennonites in Prussia (and in particular those in Prechkovka) continued to speak Dutch for the majority of their time in Prussia, perhaps as late as 1800 in some homes. Since many moved to New Russia before 1800, this would mean that some would have spoken Dutch throughout their time in Prussia. Of relevance to the current study is what Duerksen meant by the term "Dutch": Flemish, Dutch, Low Saxon, or some mix form. It is unclear. Although the exact state of the gender system of Henderson ancestors in Prussia is unknown, we can assume that, with influences from Flemish, Dutch, Frisian, and Low Saxon, it was in flux. What we know is that during their years in Prussia, they continued contact with the homeland languages, came in contact with the local language, and were influenced by the language of church and writing.

Low Saxon/Low German: Mennonites in both Prechovka and Necr.ebruch stayed in dose contact with the Old Flemish in Groningen, continuing their contact with netherlandic Low Saxon. They were also in continuous contact with the local Low German-speaking population.

Low Fmrzconian: Voth further describes frequent contact between the homelands and Pr:1.echovka in the form of pastoral visits and financial aid from the home front. Though Mennonite visitors from the Frisia Triplex would at that time have spoken a Low Saxon variety, others still living in Belgium and other parts of the Netherlands spoke Low Franconian Flemish (with three grammatical genders) or Dutch (which had already lost the masculine-feminine distinction), so language contact between Prussian Low German and Low Franconian varieties continued.

Language ofchurch and writing: Dutch and High German Dutch was the written language and language of the church of netherlandic Men-· nonites for the majority of their time in Prussia. By law, High German became the

Merchants in the north-German Hanseatic League had left their Low German mark on the region, which. was now settled by many Low German farmers. (Epp 1993)

17.

Plautdietsch gender: Between Dutch and German 73 language of instruction from 1772 on. In church, the change to High German happened at different times in different Mennonite communities, earlier in the more progressive Frisian branch of the church and later in the more conservative Flemish church. Henderson ancestors in the Flemish Pr:z.echovka church would have changed their written and church language to High German in 1775, though they kept their most treasured Dutch texts 18 dose through the centuries, even after they moved to America. (Duerksen). 2.4

New Russia

After the death of King Frederick the Great of Prussia in 1786, and following intermittent challenges to their pacifist beliefs, renewed hardship fell on the Mennonites in the form of unaffordable taxes and land leases. In 1788, they started moving to areas of New Russia (now Ukraine) north of the Black Sea. There they were assured freedom of religion, exemption from military service, and land. 1he first wave of immigrants founded the Chortitza Colony (also called the "Old Colony"). After 1.800, a second wave of Mennonites formed the Molotschna Colony (or "New Colony"). Ancestors of Henderson Plautdietsch speakers belonged to the latter group. Voth describes how a significant amount of them moved in 1821 and settled in AlexanderwohLI 9 The Netzebruch Mennonites didn't move to the Molotschna Colony urrtil1834, and founded the village of Gnadenteld dose to Alexanderwoht2° 2.4.1 Language contact in New Russia Old Colony and New Colony Low Gennan are distinct varieties. A look at the literature suggests that differences between them, brought over from Prussia, are largely phonological in nature. Corbett (1.991.) describes how grammatical gender contrasts can be phonologically motivated. I have no evidence however to suggest that the sound differences between Old Colony and New Colony varieties would have resulted in divergent gender systems.

Russian: In New Russia, Henderson ancestors came in contact with local varieties of Russian/ Ukranian, though few seem to have learned it or used it. It is doubtful therefore that

Most notably the Biestkens-Bijbel byNicolaes Biestkens (printed from 1560 to 1723),and the writings of Menno Simons.

18.

19.

Instrumental in this move was Peter Wedel, elder ofthe Przechovka church.

'Ihe leader of this move was Wilhelm Lange, Lutheran-turned-Mennonite elder of the Brenkhoffswalde and Gnadenfeld churches. 20.

74

Annemarie Toebosch

the strongly grammatical gender system of Russian would have affected their speech. In addition, they were among the first to leave Russia for America. Those Mennonites who left Russia two or more generations later had much more extensive contact v.ith Russian, especially through the educational system.

High German: In Alexanderwohl and Gnadenfeld, Henderson ancestors were repeatedly exposed to High German. Gnadenfeld especially became a progressive stronghold and center of education, and much emphasis was placed on the learning oflanguages. Of particular relevance here is the fact that Henderson ancestors came in close contact with German Lutherans in the region. We know on the other hand that High German continued to be taught (both in New Russia and America), evidence that it was still a foreign language to the general population.

Karolswalde: Additional contact would have also been the norm with a Low German variety called Karolswalde (according to Epp 1993 not a variety of Plautdietsch). Epp states that Mennonites in the Kulm area in Prussia spoke the Kulmerland dialect of Low German. Though it is not explicitly stated in the literature on Henderson Mennonites, some Henderson ancestors may have spoken or at least have come in frequent contact with this dialect,21 both in Prussia and in New Russia. A sister group of Mennonite Kulmerland speakers, more isolated in their cultural history, developed a variety called Karolswalde (after their New Russian town) that underwent less change (Epp 1993), and would possibly have retained more of its netherlandic (Flemish, Dutch, Low Saxon) roots. It would be interesting to see how the gender system of Karolswalde compares to that of Henderson Plautdietsch and other Plautdietsch varieties. 22 2.5

United States of America

Initially, Low German was frequently the language of school and church in the New Russian colonies. From the 1820s on, there was a focus on teaching High German in schools, a foreign language to the children. Few people learned Russian. Then in the mid to late 19th century the freedoms of Mennonites threatened to be severely curtailed, most importantly by the threat of compulsory military service and change to Russian as tl1e language ofinstruction. Many decided to leave.

21. 'Ihe Global Anabaptist Mennonite Encyclopedia Online (gameo.org) mentions that there were family relations between the Karolswalde and Gnadenfeld Mennonites. 11.

There are today Karolswalde speakers lhtng in Kansas and South Dakota.

Plautdietsch gender: Between Dutch and German

Among the first to leave were groups of Mennonites headed for Nebraska. Voth tells the story of delegates scoping out the land in the preceding years, among them representatives of both Molotschna and Chortitza Colonies. Molotschna Mennonites from the Alexanderwohl vicinity settled in and around Henderson, Nebraska, while others left tor Kansas (among them speakers of Karolswalde)_23 Close ties were main-tained between the Nebraska and Kansas Metmonites.

2.5.1 Language contact in America Identifying themselves predominantly as German, though not generally fluent in High German, Henderson Mennonites kept their language alive for many generations. AntiGerman sentiment during the war forced their language underground, according to many sealing its fate. English started to be used more and more in the home, though often side by side with Plautdietsch. Today, Henderson Plautdietsch, four and a half centuries in the making, is dying a relatively rapid death. In 1978, Buchheit describes the last generation of fluent speakers, mostly over 30 years of age. I interviewed several of the same people he did, now in their sixties, seventies and eighties. Many of the younger speakers among them no Ionger feel confident in the language, most of the older speakers do. All of them are proud of their language, and mourn the imminent loss of an expression of their culture.

2.6

Summary

In the past four and a half centuries, the language of Mennonites now living in Henderson, Nebraska was influenced by a variety oflanguages. The gender systems of these languages ranged from (relatively) strongly grammatical (like High German and Flemish) or mixed (like Dutch) to purely semantic (English). 'Ihe next section lays out the current gender system of Henderson Plautdietsch and compares it to those of its closely related languages.

3· 3.1

Gender contrasts in Plautdietsch Mixed gender systems

When the ancestors oftoday's Plautdietsch speakers left their homelands, Early Modern Dutch, the dominant language spoken in the netherlandic Mennonite homeland at

Voth identi.fi.es those as «the Dutch of Volhynia': Volhynia referring to their initial New Russian settlement.

23.

75

76

Annemarie Toebosch

the time, had lost the distinction between masculine and feminine gender in nouns as part of a general process of det1ection. Geerts 1966 shows that the rhetorician Hendrik Spieghel no longer made a regular masculine-feminine distinction when he wrote his 1584 Twe-spraack, particularly interesting since, as a language purist. he would have tried to speak and write in the traditional system.24 'Ihe common25--neuter distinction of Modern Dutch nouns as seen in certain agreement targets is the result of this gender loss. Consider for example the demonstratives in (la-b) and attributive adjectives in (le---d). (1)

NEUTER:

COM.'\.I:ON:

a.

dic-e klok-c 26 that clock

b.

dat-N huis-N that house

c.

een27 oude-c klok-c clock an old

d

een oud-N huis-N an old house

Dutch pronouns by comparison have not lost the masculine-feminine distinction (e.g. masculine hij 'he' versus feminine zij 'she'). 28 Audring (2006) posits that the solution to the mismatch between the two-way common-neuter contrast in Dutch nouns and the unchanged three-way contrast in pronouns is a reinterpretation of pronoun gender on semantic grounds (as proposed in Wurzel1986). One instance of semantic gender assignment can be seen for specific targets in Dutch nouns with animate referents.29 In (2) for the nouns jongen "boy" and jongetje "little bo~ attributives (in this case demonstrative) and relative pronouns are assigned

24. Twe-spmack vande Nederduitsche letterkunst by Hendrik LaurenS'.l. Spieghel is lhe lirst prescriptive grammar of Dutch, published in 1584-7. 25. The term "common" gender is used in the literature to refer to the common class of historically masculine and feminine gender. I adopt the term while realizing that the notion of "combining masculine and feminine" has no reality in the minds of Dutch speakers with this system. They simply have one class X (of non -neute:t; or, uter 'either' forms) and one classY (of neuter'not either' forms).

I use the feature terms like common (C), masculine (M),feminine (F), neuter (N), animate (A), inanimate (I), and female (FEM) purely descriptively in this section. In Section 4, I ~ill

26.

explain them in terms of tor mal features. 27.

Note in (le-d) that Dutch indefinite articles have no overt gender contrast.

Attributive targets are more susceptible to gender detlection than are pronouns. (Priestly 1983 & Corbett 1991)

28.

29. Recall from the introduction that even strongly grammatical gender systems have a basis in meaning. The semantic features associated with gender agreement are those of

Plautdietsch gender: Between Dutch and German grammatical gender: common in (a) versus neuter in (b). Personal pronouns on the other hand (or the demonstratives that stand in for them) have semantic gender, as evidenced by die in (2b ). The label A in (2) stands for animate. (2)

a.

Die-c jongen-c die-c in de hoek zit, that boy that in the corner sits

GRAMMATICAL

die-A huilt omdat 'ie-A bang is. that one cries because he scare is

SEMANTIC

"That boy who sits in the corner, he's crying because he's scared:' b.

Dat-N jongetje-N dat-e in de hoek zit, GRAMMATICAL that boy that in the corner sits die-A huilt omdat 'ie-A bang is. that one cries because he scare is

SEMANTIC

"That little boy who sits in the corner, he's crying because he's scared:' Mixedgendersystemssuchastheonein(3)areexplainedbyCorbett's(1979)Agreement Hierarchy, which posits that rightward targets (types of targets to the right of other types of targets) are more likely to be assigned semantic gender than are leftward targets: attributive< predicate< relative pronoun< personal pronoun

Figure 1. Agreement Hierarchy Tied in is the fact that rightward targets are more likely to have gender inflection. In diachronic terms, leftward targets are more likely to lose their gender inflection than are rightward targets. The idea is then that deflection on the one hand and semantic gender spread on the other are expected to occur in opposite directions across targets and that the spread of one is dependent on the spread of the other. DEFLECTION attributive< predicate< relative pronoun< personal pronoun

SEMANTIC GENDER

Figure 2. Deflection and Semantic Gender Spread

individuation and animacy, described as a feature continuum in Sasse (1993), and as a formal feature geometry in Harley and Ritter (2002) and (the latter) in Toebosch (2007).

77

78

Annemarie Toebosch

Dutch, with relatively deflected attributive targets has developed semantic gender assignment in pronouns, thus following a predictable pattern of language change. Plautdietsch has a mixed gender system similar to that of Dutch, though with less attributive deflection and with semantic gender having spread farther leftward in some of its varieties. 3.2

Grammatical gender 30

For this study, I interviewed 29 people living in Henderson, Nebraska, who identify themselves as speakers of Plautdietsch.31 I collected three types of data from each speaker: a questionnaire with background information (age, family history, education, linguistic history, language preferences and identity) a 5-to-10 minute taped conversation with 1 to 3 other Plautdietsch speakers a list of specific elicited forms (noun phrases consisting of nouns, articles, attributive adjectives, and possessive determiners, as well as sentences)-3 2 In terms of grammatical gender, we find a number of different systems. 31% of speakers have a consistent (as in, the same per noun for all speakers), three-way gender contrast33 The remaining speakers can be classified as those with varying degrees of gender attrition. I then identify two general gender systems: Strong Grammatical Gender (SGG): 9 speakers (31%) have a strong three··way masculine-feminine-neuter contrast, consistent across attributive agreement targets. Weak34 Grammatical Gender (WGG): 20 speakers (69%) have varying degrees of grammatical gender loss in attributive targets

30. For a detailed description of the ways in which gender and case interact in various varieties ofPlautdietsch, see Nieuweboer (1998).

31. I use the term Plautdietsch here, though many speakers in the Henderson comrrrunity use the term "Low German" or just "German" for their language. 32. Forms were taken from Rempel (1984) and Neufeld (2000) and were selected based on the gender listed in these works. 33· Spontaneous speech reveals that speakers in this group are generally the most fluent (never searching for words or hesitating about forms). Careful systematic analysis of taped conversations is needed to establish the effects of English-Plautdietsch bilingualism and proficiency on gender system of the speakers interviewed.

34· By "strong" and "weak" I mean relative to the other varieties of Plautdietsch studied. There are of course languages \-\i.th stronger and weaker grammatical gender systems than the ones described here.

Plautdietsch gender: Between Dutch and German 79 Evidence for a separate neuter gender comes from definite articles and attributive adjectives. For all Plautdietsch examples, I use the widely accepted orthography proposed in Epp (1996).35 (3)

NEUTER: a.

daut-N 36 kjlienet-N 37 Uag-N

the c.

MASCULINE/l'EMININE:

little

b.

eye

de-M k.jliena-M Baul-M

the

little

ball

de-·I' kjliena-·I' Niis··I'

the

little

nose

Evidence for the masculine-feminine distinction comes from indefinite articles, as in (4). (4)

NEUTER/ MASCULINE: a.

b.

CCil··N

Uag-N

an

eye

l'EMININE:

c.

'ne-·F

NiiS··l'

a

nose

een-M Baul-M a

ball

We find the same masculine--neuter versus feminine grouping for possessive deter·· miners, as shown in (5).

(5)

NEUTBR/MASCUUNE:

a.

sien/ii.ah-N Uag--N

his/her b.

c.

siene/iiahre--F 38

his/her

Nas-·l' nose

sien/ii.ah-M Baul-M

his/her

35·

eye

FEMININE:

ball

For more on proposed Plalttdiet'lch spelling systems, see Nieuweboer (1998, 1999).

Daut is also a demonstrative ('that'). Other definite article forms recorded are det and et, recorded in DPs like et Massa ('the knife') and det Schwien ('the pig').

,36.

In various descriptions of Plalttdietsch, the t1nal -tin neuter attributive adjectives only occltrs after indefinite articles. In Henderson Plautdietsch it occurs independent of definiteness.

37·

I believe the feminine Sltffix here to be -e, and -r to be part of the stem -ltahr, dropped in word-final position (hence the net.tter/masculine form -ltah and not -ltahr).

,38.

so Annemarie Toebosch 3.2.1

Grammatical gender loss

The target for which grammatical gender contrast is most commonly lost is the defi-nite article (as compared to the indefinite article and attributive adjective). In 30% (6/20) of the WGG group. the definite article contrast is fully lost (only one form is left: de).39 1he remaining 70% (14/20) ofWGG speakers has varying degrees of neuter gender loss in definite articles, using de with some nouns where SGG speakers use daut. No dear semantic patterns are found here. Compare this with indefinite articles, where only 16% of the WGG group (3/19)40 has lost all contrast (only een or an English-sounding a is left). A closer look at indefinite articles however reveals that their gender assignment is semantically determined for an additional42% (8/19) ofWGG speakers and one SGG speaker (11.%: l/9). For these speakers (9 in total- 32%), the group of feminine nouns has been augmented by nouns with female referents, so what seems like a grammatical gender contrast is in fact semantically determined. We'll look at this more in the Section 5. 15% of speakers in the WGG group (3/20) has lost all grammatical gender con-trast in attributive adjectives (only the -a suffix is left). 1he remaining WGG speakers have varying degrees of grammatical gender loss in attributive adjectives (using -a with certain nouns, where SGG speakers use --et), without clear semantic patterns. In total, only 2 speakers (7%) have lost all attributive gender contrast (as in English), and 4 (14%) have a two-way attributive gender contrast. The rest {79%) has maintained a three-way contrast. with 14 (48%) WGG speakers showing varying degrees of attrition as compared to SGG speakers. 3·3

Semantic gender

As in Dutch, the gender assignment of pronouns (or the demonstratives that stand in for them) is semantically determined in Plautdietsch. For all speakers, pronouns ~ith animate referents have semantic gender. 41 Examine the data in (6).

39· Although speakers in this group no longer have a definite article form daut ("the"), they do have a demonstrative form daut ('that'), which they can use to modify any noun. Recall that I consider grammars to exist only in the minds of speakers (Footnote 4). My synchronic data sample cannot then establish any actual "'oss" of gender features. A diachronic sample is needed to establish such a loss. In other words, I use the terms "loss" and "attrition" loosely here, to mean that I ass-ume that gender structure was present in earlier stages of the language or individual speakers' grammars. 40.

One speaker did not answer the indefinite article part.

41. For the remainder of this paper, I will talk about animates with the understanding that an additional human-nonhuman contrast needs to be drawn. I elicited only a few human versus nonhuman animate pronoun forms, from 21 speakers. Of those speakers, about a quarter

Plautdietsch gender: Between Dutch and German (6)

ANIMATES: a.

de-F/'ne-F Grootmutta the/a grandmother Dee·-A wea meed. SEMA..."''nc that one was tired "She was tired:'

b.

de-Mieen-M Voda the father Dee-A wea meed SEMANTIC that one was tired "He was tired."

c.

daut-Nieen-N l\1aakje the girl Dee-A wea meed SEMANTIC that one was tired "She was tired:'

As illustrated in (6 ), all nouns with animate referents, independent of their grarnmatical gender, are replaced by the same pronominal form dee.42 Inanimate nouns on the other hand are replaced by pronouns with matchinggnunmatical43 gender (with femi-nine and masl~uline grouping as one). Compare the data in (6) with those in (7), noting in particular the difference in pronoun use between (6c) and (7c).

groups all animates together, where-ds the rest groups nonhuman animates with inanimates. A larger number of contrasting forms need to be collected to get a dearer picture. .p.. Contrastive male and female subject forms exist (e.g. hee 'he' versus see 'she, as opposed to inanimate et'it'), but even upon frequent prompting, speakers seem hesitant to use them. When used upon prompting, speakers will quickly revert back to dee. Further data will have to uncover when exactly these sex-differentiated subject forms are used in spoken Henderson Plautdietsch. As we saw in example (5), we do find clear examples of female versus male in possessive adjectives, which (as in some varieties of High German) have two controller nouns: a 'possessor' noun such as Maakje 'girl' assigning semantic gender (female tiah 'hef, as opposed to non-female sien 'his') and a 'possessed' noun assigning grammatical gender (e.g. neuter tiah Uag versus feminine iiahre Ntis). We find further evidence for the male-female distinction in object pronouns (ahm 'him' versus aith 'her').

43· It is important to note that the classification of the gender of these forms as 'grammatical' is actually too simplistic since the gender assignment of these forms is dependent on referent animacy (in this case inanimate). In Section4, l will discuss this observation in light of the feature geometry proposed.

81

82

Annemarie Toebosch

INANIMATES:

(7) a.

de-F/'ne-F Niis

the

nose

Dee·-1' wea lank. that one was long "It was long." b.

GRAM...\fATICAL

de-Mieen-M Baul the ball Dee-M wea lUlld. GRAMMATICAl. that one was rotmd "It was round."

c.

daut-Nieen-N Uag the eye Daut-N44 wea too. GRAMMATICAL that one was closed "It was closed:'

Indefinite article targets

3.3.1

An interesting case of semantic gender is found in WGG Plautdietsch indefinite article targets (as opposed to other attributive agreement targets, which are assigned grammatical gender). 42% of WGG speakers group females with feminines for indefinite articles. 45 Remember that neuter normally pairs with masculine tor indefinite article targets (see Example (4)). We find distributions however like those in (8), where neuter nouns with female referents (like Diiwtje 'sweetheart' and Maakje 'girl') pair with feminine nouns. FEMININE (F)/FEMAI.B (FEM):

(8)

a.

'ne-F Niis

a

b.

(definite: de-F Niis)

GRAMMATICAL

nose

'ne-FEM Duwtje (definite: daut-N Duwtje) SEMANTIC a dove-little 'a (female) sweetheart'

44·

Recall that the form daut can be a definite artide and a pronounJdemonstrative.

45·

Feminine (F) is a grammatical gender feature and female (FEM) a semantic one.

Plautdietsch gender: Between Dutch and German

In terms of semantic gender then, we can divide Plautdietsch speakers into the following two groups: Weak Semantic Gender (WSG): 19 speakers (68%) have semantic gender in personal pronouns only. Strong Semantic Gender (SSG): for 9 speakers (32%), semantic gender has spread to attributives (indefinite articles).

3·4 Conclusion In sum, all varieties of Plautdietsch studied have mL'!:ed gender systems: certain targets are assigned grammatical gender while other targets are assigned semantic gender (based on referent animacy). In one variety (SGG-WSG: 8i28), speakers have a strong, three-way contrast (consistently reflected across attributive targets), and semantic gen-der in pronouns only. In a second variety (WGG-WSG: 11/28), speakers have varying degrees of grammatical gender loss and semantic gender in pronouns only. In a third variety (WGG--SSG: 8i28), speakers have varying degrees of grammatical gender loss and semantic gender in attributives.46 Recall that change in gender systems is predicted to happen as follows: gender det1ection and semantic gender spread occur in opposite directions across targets and are dependent on each other. Audring proposes that, at least for Dutch, deflection has resulted in a mismatch between nouns (as reflected in attributives) and personal pronouns, which in turn has resulted in semantic gender assignment in these pronouns. A mismatch does not seem to have been the cause of semanticization in (especially SGG) Plautdietsch, where personal pronouns have semantic gender but attributives still exhibit a three-way grammatical gender contrast across target types.4·7 Semantic gender assignment in pronouns seems to have developed for other reasons, which I will discuss in Section 6.

46. A single SGG speaker had one instance of semantic gender in indefinite art ides. Because this was such a rare occurrence, I am ignoring this data point here. Future research may show that there is indeed a fourth variety of speakers with strong grammatical gender and semantic gender in attributives (SGG-SSG). "!his would be significant, as it would prove that semantic gender spread to attributives is not tied to gender attrition (see also Section 5.2.1). 47. It is unlikely that there would have been a development where attributive deflection and resulting mismatch resulted in pronoun semanticization at an earlier stage in the language after which attributive gender would have been gained back.

83

84

Annemarie Toebosch

4· 4.1

Gender-animacy Introduction

Toebosch (2007) proposes a unified analysis of grammatical and semantic gender in terms of hierarchically organized (rather than unorganized bundles of) features. Specifically, in an attempt to arrive at the smallest set of features able to account for gender contrasts in language, she proposes a geometry of the four features [C), [F], 48 [animate], and [female]. This is illustrated in Figure 3.49 [class)

[C]

GRAMMATICAL=======* GENDER

Figure 3. Gender-animacy geometry50

The sub-geometry of grammatical gender features (represented by the unshaded disc) and the sub-geometry of semantic gender features (represented by the shaded disc) together make up one geometry. This gender-animacy geometry5 1 constrains

48. The feature [F) is the grammatical gender feature previously referred to as feminine, [C) is common. 49· Toebosch (2007) gives detailed motivations for the 4 features and their hierarchical organization. 50. The image has been flipped horizontally from its original version in Toebosch (2007) in order to fit the representation in Corbett's Agreement Hierarchy of semantic gender assignment in rightward types of targets.

51. In the remainder of the paper, I will use the term 'gender' for grammatical gender and 'animacy' for the set of semantic features associated with gender. I will refer to the systematic relationship between gender and animacy proposed in Figure (3) as 'gender-animacY.

Plautdietsch gender: Between Dutch and German

possible gender-animacy specifications in language, accounting, among other things, for the fact that pronouns with animate referents cannot be neuter in languages like Dutch and Plautdietsch. I adopt [class], proposed in Harley and Ritter's (2002) geometry of person and number, as the root node of gender-animacy. Possible other sub--geometries of [class] are those constraining numeral and verbal classification systems (see in this regard Unterbeck 1999). The features and geometry in Figure (3) have the same basic properties as those proposed by Harley and Ritter (2002). These properties are:

.M.onovalence Property 1: Formal notation: [Fj5 2 or [0 F], where [0 F] means 'unspecified for F: All features are monomlent. They are either present or absent for an Description:

Example:

expression X. Features do not have values (+or - ). A linguistic expression either has the feature [ C], or it does not have the feature [C].

Entailment Property 2: Formai notation: [F2] ~ [Fl], or, [F2] ~ [Fl], or, [Fl[F2]] 53 lfX is specified for a feature [F2], then it is specified for all features Description: [Fl]. In terms of a set hierarchy, [F2] is a subset of [Fl). In terms of a tree hierarchy, [Fl] dominates [F2]. In terms of markedness relations, [Fl] is unmarked with respect to [F2]. A linguistic expression specified for the feature [female] also has the Example: feature [animate]. Property (2) has the contrapositive in (2'):

Property 2': Formal notation: Description: Example:

Contraposition [OFl] ~ [OF2] IfX is unspecified for [Fl], then it is unspecified for all [F2]. A linguistic expression unspecified for [C] is also unspecified for [F] (feminine), [animate], and [female].

52.

By [F] here I mean ''a formal feature: not feminine.

53·

These three different notations represent the same basic relationship between elements.

85

86

Annemarie Toebosch

If feature geometries in general generate potential linguistic expressions (X), or, sets of potential feature specifications, the gender-animacy geometry generates the set of potential gender-animacy expressions illustrated in Figure 4.

X3

Figure 4. Potential gender-animacy expressions

Each area in Figure 4 represents a contrastive gender-animacy feature specification. The geometry thus allows for a six-way gender-animacy contrast in language. 54 Let's map this onto the traditional features of [masculine], [feminine], and [neuter]. As mentioned above, the unshaded sub-geometry analyzes the traditional notion of 'gender' in terms of the features [C] and [F]. Feminine forms are then specified for the feature [F] (and thus [C]) and masculine forms only for the feature [C]. The traditional gender feature [neuter] is analyzed as the lack of gender features. This analysis does not of course preclude neuter forms from having sound. Their phonological features map onto grammatical features higher in the structure.

Corbett (1991) describes much more elaborate gender contrasts in for example Bantu languages. He does not formalize these elaborate contrasts. To account for these levels of contrast, we would (i) need to add additional features to the geometry, or (li) follow Harley and Ritter (2002: 514) in analyzing these languages as having"an open-ended set oflexically determined classes" rather than "a closed set of grammatically determined classes" In the latter analysis, (some) gender expressions in these languages are analyzed as being unpredictable, or, not generated by a formal feature geometry.

54

Plautdietsch gender: Between Dutch and German 87 Let's say that neuter forms have the terminal feature [class], associated with a unique phonological string. Like gender, animacy (the sex and animacy of referents, to be precise) is encoded in the grammar in terms of the two f{)rmal semantic features [female] and [animate] . .P..xpressions with female referents have the specification [female], non-.female animate expressions (male/sex-uspecified animates) the specification [animate]. Inanimates are analyzed as lacking animacy features. I propose that inanimates have an additional referential property, call it ENTITY, encoded in the grammar as [entity]. I will discuss this in more detail in Section 5.3. To sum up. Table 1 lists and describes the feature specifications of potential expressions generated by the gender-animacy geometry. Note that feature specifications are given only in terms of terminal features, or, features not entailed by other features of an expression. 55 Expressions that have [C] (as a terminal fe-dture, or through entailment by [animate]) are traditionally called masculine in language with contrastive feminine gender, and common in languages without feminine. Expressions with terminal feature [animate] have either male or sex--unspecified (i.e. non--female) animate referents.

Table I. Potential gender-animacy expressions l'liATURE

TR.'I.DITIONAL DllSCRIPTION

SPECIFICATION

XI X2 X3 X4

xs X6

[C] [F] [animate] [F] [animate] [female]

ANU.IACY

GP.N.DER

inanimate inanimate inanimate non-female animate non-female animate female

masculine/common feminine neuter masculine/common feminine feminine

All gender assignment is then in a sense both grammatical and semantic. If an expression has semantic gender, then it has grammatical gender. Furthermore, the lack of formal semantic structure ([female] andior [animate]) is in itself 'meaningful': if you're not animate, then you're an inanimate entity.

55· I assume that entailed featlues are predictable (from UG) and as stLch not present in the derivation of an expression. Thus, only terminal feattLres are interpreted and pronotmced.

88

Annemarie Toebosch 4.2

Notes on explanatory adequacy

Several general observations made in the literature about gender in language can be explained by the gender-animacy geometry. Let's look at them in turn. 4.2.1 Animate--inanimate asymmetry Nouns with inanimate reference tend to have more flexible gender contrc:tst than nouns with animate reference. (e.g. Dahl2000a) The gender-animacy geometry explains this generalization as follows. Inanimate forms have more flexible gender because a lack of animacy features ([animate] and/or [female]) and the gender features entailed by those features allows for any gender specification (masculine, feminine, or neuter). Related to this observation is the fact that nouns with female referents tend toward feminine gender (i.e. group together in one gender traditionally called feminine), nouns with male or sex-unspecified animate referents to masculine gender. We can explain this as follows: animate forms have more fixed gender because their animacy features entail specific gender features. 4.2.2 Animacy and gender deflection Lexical categories more likely to have semantic gender (most notably personal pro-nouns) are more likely to have overt gender marking. (e.g. Corbett 1991: 260) The explanation for this provided by the gender-animacy geometry is as follows. Grammatical gender is morphologically overt (maps onto certain phonological strings). If a particular target has semantic gender (i.e. has [female] and/or [animate]), then it has gender marking by entailment (see Property 2).56 On the flipside, lexical categories more likely to be gender-deflected are less likely to have semantic gender (most notably attributives). In an analysis where semantic gender features (i.e. animacy) entail grammatical gender features, this is predicted: the absence of [F] and [C] implies the absence of [female] and [animate] (see Property 2'). 4.2.3 Defa.uit gendt-"'1" It is a well--known phenomenon that in languages with a feminine--masculine gram-· matical gender contrast, expressions will default to masculine as opposed to feminine. 5"7 This is predicted by the fact that the feature [F) (i.e. feminine) is marked compared to the feature [C] (common/masculine). I do not speak of the markedness of neuter, as

_56. Recall also that tb.is geometric analysis doesn't preclude neuter forms from having sound. Tb.eir phonological features map onto grammatical features that are higb.er in the structure. 57· Corbett's comprehensive (1991) study on gender describes no language where feminine is the default gender when compared to masculine.

Plautdietsch gender: Between Dutch and German 89 this is not a feature (rather, it is the absence of [C]) in the present analysis. If neuter is found to be the default gender with respect to masculine/common, I suggest this is due to [C] being marked with respect to some higher non-gender feature. Interestingly, Corbett and Fraser (2000) discuss the fact that "neuter" is the default gender of the type of Russian verbs that lack prototypical nouns as gender assigners, suggesting that what seems to be "neuter" gender agreement is in fact not gender agreement at all.

4.2.4 Gender deflection and the lack offominine--neutet· pa.it·ings In Germanic languages we find the following gender declension groupings: masculine-neuter (versus feminine): e.g. in New High German genitives (Duke 2005), and in Plautdietsch indefinite articles feminine-masculine (versus neuter): e.g. in Dutch (all targets except for personal pronouns), and in Plautdietsch definite articles But not: feminine- neuter versus masculine 'Ihe gender--animacy geometry predicts this as follows. Masculines are structurally between feminines (with the most structure) and neuters (with the least structure). We can see this in Table 2.

Table 2. Structural distance between neuter and feminine FEMININE

[C]

MASCUUNJ!

[C]

I

[F]

In other words, in cases of gender loss, feminines have more structure to lose to be (pronounced as) neuters than they do to be (pronounced as) masculines. 59

58. In Section 5.3, I reanaly:1~ neuter in term'> of an additional formal semantic feature.lhis does not change the explanation of Observation 3. 59· Gender-animacy does not of course rule out that feminines deflect to join the group of neuters or that neuter gain in structure to join feminines (the latter found in for example Russian), and phonological factors play a role here; it just predicts a higher likelihood of masculine-feminine and masculine-neuter collapse. Pu.rther study needs to determine exactly how unlikely feminine-neuter pairings are as compared to feminine-masculine and masculineneuter pairings.

90

Annemarie Toebosch

In conclusion, I have shown how a relatively simple feature-geometric analysis is able to explain the tendency of inanimate expressions to have a relatively wide range of genders, the negative correlation between the likelihood of being gender-deflection and having grammatical gender, the tendency of masculine and not feminine to be the default gender in language, and the incidence of feminine--masculine and masculine--neuter, but not feminine-neuter pairings in Germanic and other languages. In the next section, I examine how gender-animacy accounts for gender contrasts in Plautdietsch and show that it explains different types of hybrid gender assignment in this language.

5·

Gender-animacy in Plautdietsch

Recall that we can divide Plautdietsch speakers into the following groups, depending on their extent of grammatical gender assignment (gender) on the one hand and semantic gender assignment (animacy) on the other. 'Ihe varieties are: SGG-WSG: speakers with strong grammatical gender (reflected in attributives)+ semantic gender in pronouns only WGG-WSG: speakers with (varying degrees of) grammatical gender loss + semantic gender in pronouns only WGG-SSG: speakers with (varying degrees of) grammatical gender loss + semantic gender in attributives

I will examine the feature specifications of Plautdietsch gender and animacy expressions in the next the remainder of this section.

5.1

Gender

Recall that in one variety of Plautdietsch (SGG ), speakers have a three-way masculinefeminine--neuter contrast in nouns, reflected consistently across attributive agree-ment targets. I suggest that we find the SGG feature specitkations of nouns in

Plautdietsch gender: Between Dutch and German

Figure 5 and their attributive modifiers in Figure 6. Each specification is given with an example word.

Uag

Figure 5. SGG Plautdietsch nouns

a.

DEFINITE ARTICLES

daut

c.

INDEFINITE ARTICLES

een

Figure 6. SGG Plautdietsch attributive targets

b.

ADJECTIVES

kj/fenet

91

92

Annemarie Toebosch

Remember thatindividual attributive target types have only a two-way contrast in Plautdietsch: e.g. definite article de versus daut. We know from looking across target types that Plautdietsch has a three-way contrast. I adopt a theory of Separation, which assumes that a single phonological string (like Plautdietsch de) can correspond to two (or more) morpho-syntactic feature specifications. Separation is described in theories like Distributed Morphology (Halle & Marantz 1993, 1994) and Lexeme-Morpheme Base Morphology (Beard 1995). An alternative analysis would have one and the same feature specification, say [C], for de and kjliena. This would simplify the grammar in one way but create agreement problems in another.

5.1.1 Gender attrition In the most extreme case of gender attrition, Plautdietsch speakers (2 in total) have full gender deflection in attributives. Their system, which generates no expressions with terminal gender features, looks like that of English. I will discuss this in the next section. Another group of speakers (4 in total) have two-way gender contrast. An example of such a system is one where the masculine/feminine, but not the common-neuter contrast has been fully lost (as in Dutch). This grammar would generate the following types of gender expressions:

a.

NOUNS

b.

daut

Uag

c.

DEFINITE ARTICLES

INDEFINITE ARTICLES

d.

een

Figure 7. WGG Plautdietsch: common-neuter system

ADJECTIVES

kPiena

Plautdietsch gender: Between Dutch and German

The remaining WGG speakers have gender-animacy systems that show attrition for particular nouns (when we compare them to those of SGG speakers), yet have still retained a three-way contrast. They would therefore generate the same types of gender-animacy expressions as SGG speakers (seen in Figures 5 and 6, though perhaps not for those specific nouns).

5.2

Animacy

Recall that gender assignment ofPlautdietsch pronouns is semantically determined. Evidence comes from neuter nouns, where we find animate-inanimate contrasts as in (9). (9)

IN ANIMATES:

ANIMATES: a.

daut-Nieen-N Maakje girl the Dee-A we a meed. that one was tired "She was tired:'

b.

c.

daut-N/een-N Uag eye the Daut-·N we a to. that one was closed "It was dosed."

daut-N/een-N Kjind child the Dee-A wea meed. that one was tired "He/she was tired:'

I suggest that targets like dee in (9a) and (9b) are specified for [animate]. This is a terminal feature for all non--female animates, such as sex--unspecified dee in (9b) and male subject he "he" and object ahm "him" described in Footnote (42 ). I suggest that dee in (9a), like dee in (6a), has the additional feature [female] and thus groups with female forms like subjectse "she'' and object iiah "her".60 Inanimates like in daut in (9c) form the third contrasting group: expressions with no animacy features. I will leave for the next section the thorny issue of the feature specification of neuter controller nouns like 1VIiia~ie and Kjirzd, which seem to have one feature specification reflected in attributive targets and another reflected in personal pronouns. 'The contrasting animacy specifications are then given in Figure 8.

6o. The homophones dee in (9a) and dee in (9b) are thus analyzed as having different feature specifications.

93

94

Annemarie Toebosch

daut

et

Figure 8. Plautdietsch pronouns

Possessive determiners deserve some special attention. Recall from Footnote (42) that these forms have two controller nouns: the 'possessor' noun which assigns semantic gender to the stem (Example (5)- male sien 'his' versus female iiah 'her') and the 'possessed' noun assigning grammatical gender to the suffix (masculine null suffix versus feminine -e). What we have then are two morphemes (stem and suffix), each with their own controller noun, that are merged at some point in the derivation. Figure 9 shows the various specifications of each morpheme as they are controlled by different nouns. Specifically, (12a) shows all specifications when the stem controller noun is male (i.e. has a male referent), (12b) when it is female. Note that the null suffixes are represented by 0, and specifications are given only in terms of terminal features of stem and of suffix. Examples are given below each specification, with arrows indicating agreement relationships.

1. MALEstem a. MASCULINE suffix

b. NEUTER suffix

Ex: sien (A/ber(s) Bau/

Ex: sien (A/berfs) Uag

Plautdietsch gender: Between Dutch and German c. FEMININE suffix

Ex: slen (Albert's) NCis 2. FEMALE stem a. MASCULINE suffix

b. NEUTER suffix

Ex: iJah (Elsie's) Baul

Ex: iJah (Elsie's) Uag

c. FEMININE suffix

Ex: iJahre(Eisie's) NCis

Figure 9. Plautdietsch possessive determiners

95

96

Annemarie Toebosch

5-3

Hybrid nouns

Perhaps the most difficult part of this gender-animacy discussion is the analysis of what Corbett refers to as hybrid nouns: nouns that seemingly have evidence of one specification in one type of target and of another specification in another type of target. We have seen two types of hybrid gender assignment in Plautdietsch: all varieties: nouns like 1'1Makje and Diiwtje assign grammatical gender to attributives and semantic gender to pronouns. (review Example 6c) WGG--SSG: nouns like Jrfiiakje and Diiwtje assign semantic gender to pronotms and indefinite articies and grammatical gender to other targets. (review Example 8b) 5·3·•

Pronouns versus attributives

A rare formalization of hybrid gender agreement is proposed by Wechsler & Zlatic (2000). They explain such gender assignment as a mismatch between the specification of index gender features and concord gender features. They stipulate that DP-internal gender agreement (gender agreement between a head noun and its dependents- e.g. daut-l•vfiiakje) is restricted to the sharing of concord gender features (which are head features), whereas gender agreement between noun and coreferring pronoun (e.g. Miiakje-dee) does not have this restriction (i.e. it is a sharing of all, including crucially index, gender fealures). 61 Wechsler and Zlatic furthem10re explain Corbett's Agreement Hierarchy by stating that index teatures have a direct link to semantics (whereas concord features do not - they are directly linked to declension), 62 and that gender assignment of pronouns is therefore more likely to be semantically determined. One way to align Wechsler and Zlatifs analysis with the one proposed here is to suggest that gender features are concord features and animacy features index teatures. This mapping is represented in Figure 10 by the dotted lines.63

Plautdietsch has no gender agreement on verbs, so I do not discuss in this paper how my analysis deals with verb target agreement. 61.

62.

Not relevant to my analysis of gender-animacy in Plautdietsch, I omit declension from

my representations. 63. 'Ihe arrows represent mere correspondence relationships. I do not formalize in this paper specific e,1tailment relationships between referential information and formal animacy features (though MALE -7 [animate] and PEMALF. -7 [female] seems intuitive), or between concord and declensioiL Data giving e"idence of the latter are presented in for example Corbett & Fraser (2000).

Plautdietsch gender: Between Dutch and German [class)

[C]

FEMALE

concord

1-1

index

1-e:l==se=rna=n=ti=cs=/r=e~=e=re=nt=ia=l=pr=o=pe=rt=ie=s=:=!

Figure 10. Gender-animacy relationships

For this combined analysis to be explanatory, several questions need to be answered. First, how does an analysis that stipulates that index gender(= animacy) is not shared DP-intemally explain the semantic gender agreement in WGG Plautdietsch indefinite article targets? Second, why are the DP-intemal targets of Miiakje neuter, and not feminine (or masculine for that matter)? Why pronoun targets of Miiakje have specification [female] is clear. Pronoun targets are not restricted to concord features, so they will share the noun's animacy information (in this case [female]). DP-intemal targets on the other hand cannot have the index feature [female], but why do they not have specification [F] (in my analysis entailed by [female])? And in fact, DP-intemal targets of some hybrid nouns are feminine (e.g. the Spanish title Majestad "majesty': a hybrid noun, has male-masculine pronouns, and feminine DP-intemal targets). Wechsler and Zlatifs analysis merely states that there is a mismatch between the concord and index feature information of hybrid nouns, but because they do not encode specific entailments between features, they have no way of predicting why we find the specific types of mismatches we do (for example, neuter vs. feminine in Miiakje). What's more, Wechsler and Zlatifs analysis fails to capture an important generalization about the meaning of hybrid nouns: many of them are diminutives, augmentatives, honorifics, pejoratives, in short, expressions of speaker point of view.

97

98

Annemarie Toebosch

I suggest we look at things a different way. Consider the noun Mii.akje again. This noun has (among other things) the referential properties FEMALE, ANIMATE, YOUNG, and ENTITY. ANIMATE is grammaticalized64 as [animate], FEMALE as [female]. Let's say furthermore that an entity in the world, conceptualized as the property ENTITY/THING/ OBJECT assigned to the referent of an expression, is encoded in the grammar as a formal semantic feature [entity]. Let's furthermore propose the following entailments: [female[animate [entity]]]. This is represented in Figure (11). [class) [entity)

'

ENTITY

[F)

FEMALE

Figure I I. Gender-animacy Agreement Network

Now all female animate expressions have specification [female] (that is, the terminal feature [female] entailing [animate] and [entity]); nonjemale animates (males and sex-unspecified animates) have specification [animate]. Inanimate expressions lack sex/animacy features and have the specification [entity]. Plautdietsch nouns like Grootmutta have specification [female], Voda specification [animate], and Uag specification [entity].6S I in fact suggest all nominal expressions (animate or inanimate) have

64. I use the term "grammaticalized" synchronically here (as in "meaning made grammatical" by a speaker) rather than diachronically (i.e. "language adding formal structure over time"). 65. Since [entity] does not entail any gender features, it can have any gender: Uagwith specification [entity] is neuter (has no gender); Baul with specification [C][entity] is masculine; Niis with specification [F] [entity] is feminine.

Plautdietsch gender: Between Dutch and German 99

a feature [entity] .66 We now have a combined class of neuter-inanimates67 linked to one set of referential properties ENTITY/YOUNG/THING/OBJECT. Speakers often group the concepts YOUNG, PRE-PUBESCENT, SEX-LESS with INANIMATE, thus assigning expressions with PRE-PUBESCENT/SEX-LESS/YOUNG referents the terminal feature [entity] and not [animate]. A noun like Miiakje "girf' thus has conflicting referential information: FEMALE + YOUNG. This results in conflicting feature information: specification [female] (which entails [animate]) and specification [entity] (which excludes [animate]). I suggest that the entailed feature [entity] and not just the terminal feature [female] is expressed (both through agreement and on the noun itself). Specifically, I suggest that the diminutive suffix -je of Miiakje is a Plautdietsch expression of the specification [entity], and the stem Miiad 68 of the specification [female]. This is represented in Figure 12.

ENTITY

FEMALE

index

semantics

Figure 12. Hybrid nouns: Miiakje

66. The feature [entity] is then equivalent of Chomsky's (1995) feature [N] of nouns and adjectives (in Chomsky's analysis nouns are [+N] [-V] and adjectives [+N] [+V]). 67. In their preliminary analysis of gender and animacy, Harley & Ritter (2002) also propose a neuter-inanimate category, though they do not suggest specific corresponding semantic properties. 68. Mitadje is an alternative pronunciation for Mitakje and since [d] is present in many of Mitakje's cognates (Dutch rneid, English maid, German Mitdel), I am selecting /d/ as the underlying sound. Note that Mitad of course has features (such as number features) in addition to gender-animacy features.

too Annemarie Toebosch

If agreement between expressions is a sharing of feature structure, then all targets of Miiakje would have the specification [female] (the most marked of the index features) and be feminine. We know however that attributive targets are neuter (daut Miiakje, where daut is used with inanimates like Uag as well). I propose that the noun Miiakje has in fact only one specification, [entity], the specification corresponding which the referential property YOUNG/ENTITY/THING and expressed as the diminutive suffix -je. Maakje shares its specification with its DP-internal targets (by entering into a syntactic relationship with them). Since [entity] does not entail [F) or [C), and since the diminutive suffix -je that Maakje gets its specification from does not have any additional feature [F) or [C), these DP-internal agreement targets are neuter. Pronoun agreement targets on the other hand have feature information corresponding to all referential information of Miiakje. With [female] being the most marked of these features, pronouns have specification [female]. I have now done away with Wechsler and Zlatit's stipulation that index features are not shared DP-internally. I have shown that we can explain the Plautdietsch hybrid gender without it: pronoun targets of Maakje have specification [female] (which entails [animate] and [entity]) corresponding to all referential information. Furthermore, the geometry in Figure 11 generates two morphemes with distinct genderanimacy specifications (feminine Maad[ female] and neuter -je[entity]) that correspond to conflicting referential information (MALE/ ANIMATE VS. YOUNG/INANIMATE/ENTITY, respectively). The noun Miiakje upon being formed receives the specification [entity] ofits diminutive morpheme -je, 69 which it shares with its DP-internal targets. The specifications of Miiakje's different types of gender agreement targets are given in Figure 13, and contrasted with those of non-hybrid nouns. Note that the co-reference relationship between noun and pronoun is indicated with subscript i. a. MiJak}e1 PRONOUNS

ATTRIBUTIVES

ENTITY

FEMALE

69. In work in progress, inspired by Wiltschk.o and Steriopolo's (2007) morphological analysis of diminutives, I am exploring the possibility of a morpho-syntactic explanation for the fact that the noun Mttakje has the feature information of -je and not of Mttad, thereby adding further explanatory power to my analysis of hybrid gender.

Plautdietsch gender: Between Dutch and German

b. GrootmuttaJ PRONOUNS

ATTRIBUTIVES

FEMALE

FEMALE

c. Uagk PRONOUNS

ATTRIBUTIVES

ENTITY

ENTITY

Figure 13. Hybrid noun versus non-hybrid noun targets

A remaining question is this: does the proposed analysis explain that pronoun targets of hybrid nouns are more likely to have semantic gender assignment than do attributive targets of these nouns (i.e. Corbetfs Agreement Hierarchy)? The answer is that the gender Agreement Hierarchy is now understood in a different way: targets in a syntactic configuration with the gender assigning noun (DP-internal targets) have the semantic information of only one noun morpheme, targets not syntactically related to the noun (i.e. pronouns) have the semantic information of both noun morphemes. In this sense, we correctly predict that pronouns typically have more semantic information. This is an important reframing of the Agreement Hierarchy: all targets have semantic gender (eliminating Wechsler & ZlatiC's stipulation), syntactic targets (i.e. DP-internal targets) just have "less ofif' (as in, they encode the semantic information of only one noun morpheme). An important question is how this analysis works for nouns without overt diminutive morphology. A perhaps radical proposal is that nouns like Plautdietsch Kjind "child" (with animate co-referring pronouns and neuter attributives) have the same analysis as nouns like Miiakje: Kjind has two conflicting referential properties ANIMATE and YOUNG, encoded in the grammar as the distinct feature specifications [animate] and [entity], respectively, where the latter is a null morpheme (as opposed to an overt morpheme like -je).

101

102

Annemarie Toebosch

Careful cross-linguistic study of hybrid nouns needs to be done to see what the connection is between these nouns and what in Lexeme-Morpheme Base Morphology (LMBM) singles out as "expressive L-derivations': that is, morpheme categories that retlect speaker attitude: e.g. diminutives, augmentatives, pejoratives, and honorifics (Beard 1995). A quick examination of Corbett's (1.991.) discussion of hybrid nouns reveals that many of the examples given indeed have diminutive, pejorative, and honorific meanings (e.g. Spanish Majestad "majesty': Polish lajdak "scoundrel")/0 It could be the case that these expressive morphemes are all encoded as THING··[entity], but I leave this for future study. In addition to a cross-linguistic comparison of the meaning of hybrid nouns, not so trivial details need to be explained of how exactly a specification [female] or [animate] is derived as a pronoun (as opposed to for example a DP--internal modifier). Finally, we haven't yet explained the peculiar case of noun "hybridity" we find in certain varieties of Plautdietsch: semantic gender assignment in some types of attributives and not in others. I propose an analysis of this in the next section. 5.3.2

Indefinite articles versus other attributive targets

Recall that in WGG-SSG, indefinite article targets of neuter nouns with female referents are assigned to the feminine gender class. Review Example (8), repeated here as (10). (10)

FEMININE (P)/FBMALE (FBM):

a.

'ne-P Nas a

b.

(definite:
GRAMMATICAL

nose

Diiwtje (definite: daut-N Diiwtje) a dove-little 'a (female) sweetheart' 'ne-FBM

SEMANTIC

The contrast where definite articles are assigned grammatical gender and indefinite articles (seeming) semantic gender is a problem for most theories of gender. For example, Corbett's Agreement Hierarchy treats all attributives as one. Other theories, such as Wechsler and Hahn (to appear) explain certain cases of semantic (in this case number) agreement in attributive targets by saying that their controller nouns lack concord features and thus default to semantic gender assignment, 71 but their theory does not I believe explain a situation where one and the same controller (say Plautdietsch

70. Words like Landuma abok "snake" could be expressive (perhaps pejorative?), though this is not stated.

71. Perhaps of interest is the fact that their example controller is again an expressive (an honorific).

Plautdietsch gender: Between Dutch and German

Duwtje in (lOb)) assigns different types of gender (grammatical versus semantic) to different types of targets (definite articles versus other attributives) in the same syntactic position (DP-internal). Wechsler and Hahn would say that this agreement is not technically semantic, but rather that because the controller nouns somehow lack concord gender, agreement is established by defaulting to the semantics (the default gender of the semantic property FEMALE is feminine). We have no reason to believe Duwtje lacks concord gender. In fact, we see its concord gender reflected in definite article and adjective targets. Following the analysis in 5.3.1, Having explained away Wechsler and ZlatiC's stipula-tion that index gender(= animacy) is not shared with DP-intemal targets, I suggest then thatforWGG-SSG Plautdietsch speakers, nouns like Diiwtje"female sweetheart" have the specification [entity], its definite article and adjective targets the specification [entity], and its pronoun targets the specification [female]. 'Ihe indefinite article target behaves unlike other DP-internal targets: like pronouns it has specification [female). All examples recorded of this indefinite versus other attributive target hybridity occur with nouns that also exhibit attributive--pronoun hybridity (again, diminutives like Diiwtje, Miiakje, Friitje "little woman': Brntje "little bride"). 'lhis is predicted by Corbett's Agreement Hierarchy: if we find semantic gender in an attributive target, we also find it in a pronoun target of a particular noun. Interestingly, this ty-pe of definite-indefinite contrast seems to happen only in speakers with significant amounts of grammatical gender attrition (a weakening of grammatical gender morphology). We do not find this contrast in SGG Plautdietsch speakers. As discussed in Audring (2006), gender attrition (a loss of arbitrary gender contrasts) leads to semantici7.ation (specifically here, a link to the referential sex information FEMAI.E). vVhy we find this semanticization in indefinite articles and not definite articles or adjectives is unclear. It is not surprising that we find this type of semanticization in a language island, such as Plautdietsch, which exhibits e.lCtensive attrition, having had bilinbrualism accompany it throughout its history. Future study needs to be done then on this and other settlement lanbruages before we jump to the conclusion that definiteness plays a role in semanticization.

5·4

Conclusion

In this section I have given a feature-geometric analysis of the Plautdietsch gender system. I have analyzed animate nouns as having specifications [female] or [animate]. Inanimate nouns on the other hand are specified for a higher feature [entity]. Table 3 gives the contrastive specifications of the non--hybrid nouns examined and of their agreement targets. Note the co--indexation. Note furthermore that WGG Plautdietsch speakers will have reduced gender contrast. DA stands for definite article, lA for indefinite article, and ADJ for adjective.

103

104

Annemarie Toebosch Table 3. Plautdietsch gender-animacy expressions: GENDER

ANIMACY

SPHCU'ICATION

EXAMPI.E

NOUN

TARGET PRO

feminine

[F[female]]

FEMALE

[F ][entity]

INANIMATE

masculine

NON-FEMALE

[C[animate]]

ANIMATE INANIMATE

[C][entity]

INANIMATE

[entity]

Mutta; 'mother' Nilsk 'nose' Vod!J. 'fathei·· Baulk 'ball'

Uagk 'eye'

ATTR

dee/see; 'she' deek 'it' dee/hee, "h" e deek 'it' d!Jut/et1 'it' J

DA

IA

ADJ

de 'the' de 'the' de 'the' de 'the' ®ut 'the'

~e

kjliena 'little' kjlietra 'little' kjliena 'little' kjlietra 'little' kjlienet 'little'

'a

'ne

'a

een

'a

Cetl

'a

een

'a

Furthermore, I have analyzed the gender of hybrid nouns as words that express an additional referential property ENTITY as [entity] in some cases reali7.ed overtly on the noun itself (-je). Plautdietsch hybrid noun specifications and that of their targets are given in Table 4. EXPR stand for expressive morpheme, here a diminutive suffix. Table 4. Miiakje and its targets EXAMPLE

SPECIFICATION MORPHEME BASE

[entity] [female]

NOUN PRO

EXPR

-je 'little' Mllad 'maid'

TARGET OTHER DA

IA

Mllakje 1

dattt

'girl'

'the'

eetr 'a'

kjlietret 'little'

dee/see; 'her'

WGG-SSG Plautdietsch speakers have indefinite article target than een [entity] .

6.

AllJ

'ne[lemaleJ

rather

Acquisition and variation

Let me end by explaining the analysis in Section 5 in terms of acquisition and variation. I assume that human beings have a genetically determined language acquisition

Plautdietsch gender: Between Dutch and German

program consisting of a cognitive system and two performance systems (Chomsky, 1995). The cognitive system consists of a finite number of hierarchically organized linguistic features (including for example [common[animate]]) a set of procedures that generate linguistic expressions from subsets of these features (e.g. the operation merge) The performance systems interacting with the cognitive system retrieve and use lin-guistic expressions. They are the Conceptual--Intentional system which, simply put, allows a speaker/signer to interpret linguistic expressions the Articulatory- Perceptual system which allows a speaker/ signer to perceive and articulate linguistic expressions With language acquisition program being an invariable property of human beings, the acquiring child gets it 'for free: The child is left the following tasks (adapted from Toebosch 2007): 1.

2. 3.

Assigning properties to referents (e.g. the referent of Mi:iakje "girl" has, among other things, the property YOUNG), Recognizing contrasting forms in the linguistic input (e.g. daut Uagversus de Niis) Assigning to each contrasting form the simplest possible distinct feature specification

All Plautdietsch speakers studied here were exposed as children to the speech of adults with similar grammars. Where their grammars are similar, I assume this to be the result of similar early-life input. Variation results then from something other than this early exposure, likely variation in degree of exposure to English. I will not explore in this paper the mechanisms of attrition other than to say that it involves a change, in a speaker's lifetime, in the assignment of feature specifications to contrasting forms (i.e. a change in (3) above). Language variation can now be reduced to differences in the assignment of feature specifications, reflecting differences in referent characterization and contrast level. In the remainder of this section, I consider similarities in and differences between gender-animacy systems of Plautdietsch and related Germanic languages in terms (changes in) feature specification assignment. Lefs start by looking at gender systems that are (relatively) strongly grammatical. If we take Jan to be a boy acquiring a variety of German, he is exposed to, tor example,

105

106

Annemarie Toebosch

the data in (11). He assigns each contrasting form a distinct specification. Co-indexing indicates a shared referent.72 (11)

Gender contrasts in the input: Jan a. b.

der[CJ Jungen/Tischuc1 the boy/table · ·

Ich mag ihni[CJ" I like him/it

Ich mag sieHFJ" die[FJ Frau/fasse,J ["' • ' the

c.

German

woman/cup

I

like

her/it

das MiidcheniFensterk kh mag esk. the girl/window I like her/it

Note that sex/animacy is not contrastive in the input in (11),74 so based on data such as these alone Jan would not select [female] or [animate] and would consider for example the property FEMAI.E of the referent of Frau. in ( 11 b) to be a purely semantic property, much like the property SNART of the referent of the English noun genius (review Footnote 2). However, as noted in the introduction, all gender systems have a semantic core, and Jan will also be exposed to for Example (12). ( 12)

Animacy contrasts in the input: Jan - German

a.

J~/Georg/KarVOtto 1 [animate] kh sah i~[animate]

I b.

saw him

Birgit/Johanna/Petraj[female] Ich sah si.ej[female] I saw her

In (12), Jan encounters data where male forms contrast with female fonns. I suggest then that all grammars (even those with strong grammatical gender systems) select at least one animacy feature. 'Ihe full gender--animacy specification is given in (13).

72. I asstLme of course that a child acquiring language does not actually make conscious choices in gender assignment, and wording like Jan assigns... and Jan selects... should be interpreted accordingly.

73· I am only listing gender features in these examples. Note that I analy1.e all nouns as having the additional formal semantic feature [entity]. Por inanimate and hybrid nouns, this is an additional terminal feature not listed here; for non-hybrid animates, [entity] is entailed by sex/animacy features. 74· In many varieties of German, sex/animacy is in fact contrastive. In those varieties, the pronouns sie «she" and ihn "him" are commonly used for animates (so, Mtldchen is a hybrid noun), while distance-neutral demonstratives den, die, das (masculine, feminine, neuter "it") are used for inanimates. See also data point (13e) in this regard.

Plautdietsch gender: Between Dutch and German

(13)

Gender-animacy contrasts in the input:Jan- German a.

b.

der[animate] JungenJJanJOttoi[animat boy the

j

Ich sah ihni[animate)" I

saw him

n•··y•· Ich sah ihn,.[C'J' Tisch,[<']'t--_·;~,, der,,,,[P"tt'tvJ • i.e -Jt j , Ll)J ..J' saw it I table L

-H

-1

I

..J

the c.

dieffemale j FraukiBirgit:cfPetrak[female] Ich sah siek[femalej· saw her I woman the

d.

die[F][eniityj Tassei[rl[entityj Ich sah siei[F][eniityj" saw it I cup the

e.

Ich sah esm[entity]" . M"d h 'Fensterm[entity] ..l J a c en, uas[entity] saw her/it I girliwindow the

'Ihe interesting data point is of course Miidchen in (13e). Why is it exactly that Jan's input has pronoun es and notsie co-referring with Miidchen? I assume that Miidchen is like Miiakje. It is made up of morphemes with conflicting referential information: Miid with specification [female] corresponding to FEMALE, and chen[entity] corresponding to YOUNG!TNANIM.ATE. Miidchen and its DP-internal targets have specification [entity]. In Plautdietsch, pronoun targets of the equivalent noun Miiakje have specification [female) corresponding to all referential information. I suggest pronoun targets of i\:Iiidchen do not. vVhy? I assume that a general economy condition, requiring that expressions have the simpiest possibie structure, is imposed on the grammar and that as a result expressions, like pronoun targets ofMiidchen, are assigned the simpler structure [entity] (as opposed to [entity[animate[female]])). For a boy Willem acquiring Dutch defaulting to this simpiest str-ucture is not possible. Here's why. Willem finds that the specification [C] suffices to account for the two-way gender contrast in his input (see Example (1)). He is however also exposed to a three-way pronominal contrast (e.g. hij "he", .zij "she': het "it"). Recall that Audring's (2006) proposed solution to the noun-pronoun mismatch in is a reinterpretation of pronoun gender on semantic grounds. I suggest that this semantic reinterpretation is in effect an overriding of the need for structural simplicity. Willem then analyzes these pronominal forms as having specification [female]-.zij "she'', [animate]-hij "he': and [entity]-het "it': agreeing with for example the nouns meisje "girf', jongetje "boy'', and huis "house': respectively. Diena, a girl acquiring Plautdietsch (specifically the SGG variety), finds no dear mismatch in the system as a whole. She looks across the different attributive target classes, and finds a three-way contrast for inanimate nouns. However, individual target classes of these nouns show syncretism resulting in a two-way phonological contrast for each attributive target (see Examples (3) and (4)). Diena, unlike Jan acquiring the variety of German with strong grammatical gender, has evidence that the arbitrary

107

108

Annemarie Toebosch

gender system is weakening as evidenced by a loss in phonological contrast (different from the noun-pronoun mismatch described above), and chooses the more structurally complex assignment [female] for the noun Miiakje over the structurally simpler neuter-inanimate assignment [entity]. Why exactly? When language loses contrast in one area (here, phonological contrast), it gains it back in another (here, semantic con-trast: an increase in semantic feature specification), perhaps by some general equilibrium principle. In sum then, a loss in syntactic structure (gender deflection) is a gain in formal semantic structure. Finally. take Nancy to be a girl acquiring English. Nancy has no evidence of any sort of arbitrary gender contrast. She assigns specifications [female], [animate], and [entity] to for example girl-she, boy-he, and table-it, respectively. Attributives do not enter into any sort of gender agreement relationship with nouns.

7· Conclusion Van Haeringen, in his celebrated 1956 work, positioned Dutch between English and German?5 The analysis presented here positions Plautdietsch between Dutch and German. Speakers of Plautdietsch, Dutch, and German all encode the semantic properties of se.x/animacy (i.e. FEMALE and ANIMATE) as the specifications [female] and [animate]. Speakers of (some varieties of) German have no evidence of different ieveL~ of gender contrast in types of expressions (there is a three-way contrast level in nouns and all types of targets). As a result, they have no need for semanticization, or, they obey a condition on structural simplicity (and have pronoun targets of hybrid notms be as simple as possible: neuter). Speakers of Dutch have e"Vidence of different levels of gender contrast in types of expressions (two-way for nouns and attributive targets, three-way tor pronoun targets). As a result, they have a need for semanticization and override a condition on structural simplicity (and have pronoun target specifications corresponding to all referential properties of co-referenced hybrid nouns). Speakers of some varieties of Plautdietsch (SGG Plautdietsch) have evidence of a three-way contrast across all targets, with syncretism (a two-way contrast) in specific targets. The loss of phonological contrast (different from the noun-pronoun mismatch described for Dutch) is evidence that the arbitrary gender system is weakening and causes semanticization as in Dutch. The gender system of other varieties of Plautdietsch (WGG Plaudietsch) exhibits relatively more gender attrition. In these varieties, semanticization has spread to the indefinite article targets of hybrid nouns.

75·

For recent studies on same topic, see Htini.ng e.a (2006).

Plautdietsch gender: Between Dutch and German

Further research needs to be conducted to see if and how definiteness affects this process. In conclusion, I hope the research presented here will inspire future work in other areas ofPlautdietsch, shedding further light on the structure of this language, its structure, use, its development, and the way bilingualism has aflected this development.

References Aksenov, A.T. 1984. K probleme ekstralingvisticeskoj motiv--a.di grammaticeskoj kategorii roda. Vopro~y jazykoznanija 33: 14---25. Audring, Jenny. 2006. Genusverlies en de betekenis van voornaamwoorden. In Nederlarzds tussen Duits er1 Engels, Matthias Hii.ning, Ulrike Vogle, Ton van der Wouden & Arie Verhagen (eds), 71---88. Leiden: Stichting Neerlandistiek LeiderL Beard, Robert 1995. Lexeme-M01pheme Base Morphology: A General Theory ofinjlection and Word Formation. Albany NY: SlJNY Press. Buchheit, Rober1 H. 1978. Mennonite Plautdietsd1: A Phonological and Morphological Description of a Settlement Dialect in York and Hamilton Counties, Nebraska. Ph.D. dissertation. Lincoln NB: University ofNebraska. Boroditsky, Lera, Schmidt, Lauren & Phillips, Webb. 2003. Sex, syntax, and semantics. 1n Language in Mind: Advances in the Study of Language and Cognition, Dedre Gentner & Susan Goldin- Meadow (eds), 61-79. Cambridge MA: The MIT Press. Chomsky, Noam. 1995. The .Minimalist Program. Cambridge MA: The MIT Press. van Coetsem, Frans. 1988. Loan Phonology and the two Transfer 'T)tpes in Language C01ztact. Dordrecht: Foris. Corbett, Greville & Fraser; Norman. 2000. Default genders. ln Gender in Grammar and Cognition, Barbar-a Unterbeck & Matti Rissanen (eds), 55-97. Berlin: Mouton de Gruyter. Corbett, Greville. 1991. Gender. Cambridge: CUP. Corbett, Greville. 1979. The agreement hierarchy. ]oumal of Lirzguistics 15: 203-224. Dahl, ()sten. 2000a. Animacy and the notion of semantic gender. In Gender in grammar and cog1zition, Barbara Unterbeck & Matti Rissanen (eds), 99-115. Berlin: l\·1outon de Gruyter. Due.rksen, Jacob. 1967. Transitions from Dutch to German in West Prussia. Mermonite Life 22(3): 107-109. Duke, Janet. 2005. Gender systems and grammaticali7.ation: Evidence from Germanic. In Unguistik: Impulse und Tendenzen - Grammatica/isierung im Deutschen, 1brsten Leuschner, Tanja Mo1telmans & Sarah De Groodt (eds), 31-57. Berlin: Mouton de Gruyter. Epp, Reuben. 1993. 1he Story ofLow German atzd Plautdietsch. Hillsboro KS: The Reader's Press. Epp, Reuben. 1996. The Spelling £~f"Low German and Plautdietsch: Towards an Official Plautdietsch Ortho!,>raphy. Hillsboro KS: The Reader's Press. Geerts, Guido. 1966. Genus en geslacht in de Gouden Eeuw. Een bi_jdrage tot de studie van de

nominale klass~tikatie en daarmee samenhangende adnominale j7exievormen et! pronominale verschijnselen in Hollands taalgebruik van de zeventiende eeuw. Bmssel: Belgisch Interuniversitair Centrum voor Neerlandistiek. Halle, Morris & Marantz, Alec. 1993. Distributed morphology and the pieces of inflectiorL In The View from Building 20. R~says ir1 Linguisti£~~ in Honor of Sylvain Bromberger, Kenneth Hale & Samuel Jay Keyser (eds), 111-176. Cambridge MA: TI1e MIT Press.

109

no Annemarie Toebosch Halle, Morris & Marantz, Alec. 1994. Some key features of d.istrlbuted morphology. In MITWPL 21: Papers on Phonology and Aforp1:ology, Andrew Carnie & Heidi Harley (eds), 275-288. Cambridge MA: MITWPL Harley, Heidi & Ritter. Elizabeth. 2002. Person and number in pronouns: A feature-geometric analysis. L.anguage 78: 482-526. Hockett, Charles. 1958. A Course in Modern Li,lgui.~tic.~ New York NY: MacmlllaiL van I.euvenstei)n. Arjan. Ruljssendaal. Els, Mooljaart, Marijke, Gelrnaert, Dirk & Tempelaars, Rob. 1997. Vroegnieuwnederlands. In Geschiedenls van de Nederlandse Taal, Maa.tten van den Toorn, Willie Pijnenburg. Arjan van Leuvensteijn & Johannes van der Horst (eds). Amsterdam: Amsterdam University Press. Neufeld. Eldo. 2000. Plautdietsch Grammar: An aid to speaking, reading. and writing Netherlandic-· Mennonite Plaudietsch [Lincom Studies in Germanic Linguistics 8]. Munich: Lincom. Nieuweboer, Rogier. 1998. The Altai Dialect of Plautdiitsch (West--Siberian Mennonite Low German). Ph.D. dissertation. University of GroningeiL (Republished in the Lincom Studies in Germanic Linguistics series in 1999 by Lincom, Munich). Priestly, Tom. 1983. On 'drift' in Indo-European gender systems. Journal ofIndo-European Studies 11: 339-363. Rempel, HermaniL 1984. Kjenn fie Nodl Plautdietsch?: A Mennonite Low German Dictionary. rev. edn 1995. Winnipeg: Mennonite Literary Society. Sasse. Hans-Ji.irgeiL 1993. Syntactic categories and subcategories. In Syntax. An International Handbook of Contemporay Researdl, Joachim Jacobs, Arnim von Stechow. Wolfgang Sternefeld & Theo Venneman (eds), 6<16--686. Berlin: de Gruyter. Thomason. Sarah & Kaufman, Terrence. 1988. Language Contact, Creoiization, and Genetic Linguistics. Berkeley CA: University of California Press. Toebosch. Annema.t·ie. 2007. A feature--geometry of gender and animacy: The case of deficient pronouns in Dutch. .MorphOn, 1---36. <wllrw.morphologyonline.com>. (18 June, 2007). De Vogelaer, Gunther. 2009. Changing pronominal gender ln Dutch: Transmlssl.on or diffusion? In Language Variation, European Perspectives IJ [Studies ln Language Variation 5], 'Isiplakou, Stavroula, Marilena Karyolemou & Pav'los Pavlou (eds), 71-80. Amsterdam: John Benjarnins. Voth. Stanley E. (ed). 1975. Henderson l\.1ennonites: Prom Holland to Henderson. Henderson NB: Service Press. Unruh, BenjamiiL 1955. Die niederlandisch-niederdeutschen Hintergriinde der mennonitischen Ostwanderungen im 16., 18., und 19. .Tahrhundert. Karlsruhe: Selbstverlag. Versloot, ArjeiL 2008. Mechanisms of Language Change: Vowel Reduction in 15th Century West FrisiaiL Ph.D. dissertation. Rijksuniversiteit Groningen. Unterbeck, Barbara 1999. Gender: New light on an old category. In Gender in Grammar and Cognition, Barbara Unterbeck & Matti Rissanen (eds),xvi-xlvi. Berlin: Mouton de Gruyter. Wechsler, Stephen & Zlatic, Larisa 2000. A theory of agreement and its application to SerboCroatiaiL Language 76: 799-832. Wechsler, Stephen & Hahm, Hyun-Jong. 'Ib appear. Number markedness and polite plurals. In a speciallssue of Morphology, Jonathan Bobaljik. Andrew Nevin~, Hazel Pearson & Uli Sauerland (eds). Weerman. Fred 2008. Deflection and age of OIL~et. Paper presented at Jransmi.~sion and Diffusion. Nijmegen: Max Planck IIL~titute for Psycholingulstlcs. Wurzel, Wolfgang. 1986. Die wiederholte Klassl.fi.kation von Substantiven. Zeltsd1r!{t for Phonetfk, Sprachwissenschaft und Kommunikatlon~forsdlung 39: 76···96.

Anaphors in contact The distribution of intensifiers and reflexives in Amana Gern1an* Michael T. Putnam The Pennsylvania State University

This paper presents a novel sketch of a research program into the morpho-syntactic/semantic. characteristics of intensifiers and reflexives in a moribund Sprachinsellanguage, Amana German. As demonstrated in this pilot study, Amana German has (apparently) tmdergone a shift in the syntactic distributional properties of its intensifier and reflexive inventory. As a result, Amana German in this regard more strongly resembles Dutch and modern English rather than German. Following Gast (2006), I interpret this as a shift involving the morpho-syntactic/semantic realization of recognizing two distributional types of transitive predicates; namely, typically 'other-directed' (e.g. [t·Oo]) and typically 'self-directed' predicates (e.g. [···Oo]).

1.

Introduction

Recent scholarship in the field of contact linguistics has witnessed a (significant) increase in the application of theoreticaV generative frameworks to this field of inquiry. Since the application of established formalisms in language contact situations is in many regards still gaining ground, there are certain areas of lanbruage contact that remain sorely under-researched from a generative perspective. Although the morphosyntax of languages that exist in (sometimes extreme) contact situations has been heavily researched, the morpho-semantic level of language has thus far to the best of my knowledge received extremely short shrift. in the contact linguistics literature. In this chapter, my goal is to pursue and develop a line of research that I believe shows

*I am deeply indebted to my friends and informants in the Amana Colonies for providing me with preliminary data and discussion of these iss-ues. I also owe a great deal of thanks to Werner Abraham, Volker Gast, Elly van Gelderen, and Jon Coffee for lengthy discussions of these matters. The us-ual disclaimers apply.

111

MiChael ~1: Putnam much promise. The chief aim of this study is to sketch out the beginnings of a research program into the morpho- semantics ofa moribund Sprachinsellanguage on the verge of extinction; namely, Amana German. Amana German, also known as Koionie-Deutsch, is a middle-Hessian dialect spoken in the 7 Amana Colonies in East Centraliowa. 1 This study engages in a brief (yet succinct) analysis of the inventory and properties of reflexives in Amana German. As will be discussed in Section 2, there are clear distinctions concerning the distribution of retle.xives in Gennan, Dutch, and English (with stronger similarities existing between the latter two languages). Since Amana German (hereafter AG) hails from a Hessian dialect, it stands to reason that the original dia·· lect was similar - if not identical - to the distribution of intensifiers and reflexives found in modem German. The data presented in Section 3 from three AG speakers (by means of dosed questionnaire) reveal. however, that noted differences do in fact exist between modern German and AG with regard to the distribution of intensifiers and reflexives in particular contexts. It is especially noteworthy to acknowledge that AG seems to have adopted a reflexive system that is more similar to modern Dutch and English than what is found in modern German, in particular with AG's sensitivity to the two distributional classes of transitive verbs that are 'other-directed' and 'selfdirected' predicates. Section 4 concludes this paper and offers a brief discussion and outline of related research that could unveil even more promising results in tl1e area of morpho-semantics in contact linguistic/attrition studies. Onc.e again, I acknowledge from the beginning that the purpose ofthis paper is to provide an outline for fruitful research in the near future and not to provide a comprehensive analysis of the topic at hand in this chapter. Such a detailed study, in my opinion, is simply an undertaking that is too grand in scope and content for a mere volume chapter. This caveat notwith ·· standing, I feel that the initial findings brought to our attention here merit further serious query.

z. Intensifiers and reflexives in Germanic Languages: Dutch, English, and German 'Ihe purpose of this section is to familiarize ourselves with the lexical inventories and distributional properties of intensifiers and reflexives in Germanic lanbruages. In particular for this study, we will restrict ourselves to a detailed sketch of the three most relevant (and closely-related) languages involved in this study; namely, Dutch, English, and German. The analysis adopted here is in most respects an exegesis of the work of

1. For a more detailed history and depiction of the sociolinguistic landscape of the Amana Colonies, the reader is referred to Rettig (1976) and Webber (1993).

Anaphors in contact: The distribution of intensifi.ers and reflexives l.n Amana German 113

Kiparsky (2002) and Gast (2006) (with more reliance on the latter) and is cast within an Optimality-Theoretic (OT) framework that employs violable (soft) constraints in its explanation of the facts concerning the distribution of intensifiers and reflexives in these at{)rementioned languages. Actually, it is worth mentioning that the choice of formalism is somewhat arbitrary at this stage of this research program; however, to maintain reciprocity with the already established studies by Kiparsky and Gast, it makes sense to work within the framework of these prior studies. Furthermore, agreeing with Gasfs observation that when one looks at a wider array of cross--linguistic data concerning reflexivity, a theoretical analysis that proposes that certain principles in the domain of reflexivity based on the distribution of grammatical alternatives concerning elsewhere-phenomena, pronominals, and anaphors is well-motivatecJ.2 Before taking a closer look at these three individual languages with respect to their reflexive inventories and their distributions, it is necessary to establish the pri-mary violable constraints that will surface through the analysis and discussions that follow. However, even before introducing the constraints, the semantics of certain transitive verbs must be brietly introduced and discussed. Follmving original observations made by Everaert (1986), Koster (1997), Reinhart & Reuland (1993), Konig & Siemund (2000), and Donaldson (for Dutch) (1997) and implemented in analyses by Kiparsky (2002) and Gast (2006), transitive verbs can be classified into (among others) two classes with regard to the intended direction (not in a spatial sense) of activities. According to this line of reasoning, transitive verbs can be classified as 'typically otherdirected' verbs (e.g. -t-On-VERBs) and 'typically self-directed' verbs (e.g. ---On-VERBs). Following Konig & Siemund (2000) and Gast (2006 ), I, too, concede that the two types of predicates form distributional classes, rather than purely semantic ones (taken from Gast (2006: 189; from Donaldson (1997: 203ff) ). (1)

+0D-VERBS

aangrijpen 'to attack; afmnselen 'to beat: haten 'to hate: lachen in 'to laugh about; praten met 'to talk to' (2)

-On-VERBS

aankleden 'to dresS, overeten 'to overeat: scheren 'to shave: verbergen 'to hide; wassen 'to wash' As we shall see immediately below, the [±Oo] distinction will play a decisive role in our analysis of not only Dutch, English, and German that follow in this section, but also in the inventory and distribution of intensifiers and retlexives of AG.

2.

See also Burzio (1991, 1996, and 1998) for seminal research in this area.

114 MiChael ~!: Putnam

Concerning the constraints employed here, Gast (2006; Sec..tion 7.7.3) expands and improves upon the system originally developed by Kiparsky (2002) with the following constraints to model the distribution of intensifiers and reflexives in Germanic languages (and beyond): (3)

"0THER·-DIRECTED BINDINGi*0DBIND

'Ihe complements of an [+0D] -predicate P may not be botmd by the subject of P. (4)

BINDING DoMAINiBINDoM

A pronotm has a compatible antecedent in a designated domain D. (5)

fEATURAL EcoNOMY/fEe

Avoid Pronominals (6)

MoRPHOLOGICAL EcoNOMY/Mnc

Avoid morphologically complex pronouns (7)

PARSE

The output is not empty

'Ihese c.onstrc:rints act upon the inherent semantic characteristics of pronominals and anaphors. Following previous work by Chomsky (1981, 1986), Reinhart & Reuland (1991, 1993), and Burzio (1991, 1996, 1998), Kiparsky (2002) combines these notions in the development of a comprehensive typology of pronominals, which admits wide cross-linbruistic applkability. Although for our limited study we can for the time being forego a lengthy discussion of Kiparsky's comprehensive typology, there are a few details that are paramount for our discussion at hand. First, both Dutch and German have [+LocAL] (weak) anaphors in their pronominal/anaphor inventories- zich and sich respectively. Although (modern) English lacks an overt (weak) [+LocAl.] ana·· phor, Gast postulates that a null operator (0) does exist and functions as a enclitic that must be adjacent to the predicate: ( [+LocAL], [+CLy]). Second, following Eckardt (2001), Hole (2002), and Gast (2006), I employ the identity function In in contexts of intensification to derive forms such as himself in English and sich selbst" in German. 'Ihe identity tl.mction Io takes a nominal denoting an individual x as its arbrument and maps it onto itself: Io(x) == x. 3 For example, the DP in (Sa) can be interpreted as (8b ): 4

3· For a detailed exposition on the ID function, see Gast (2006; Chapter 3). As pointed out by Gast (2006: 4-5), although the operation illustrated in (8b) above does not appear to be truth-conditionally meaningful, the ID function becomes highly significant when it interacts with focus; the crucial detail that Gast wishes to convey here with regard to this analysis is the ass-umption that the intensifier is the only component of a sentence in focus. A...'l a result, "the effect of such focusing is the evoking of a set of alternative functions 4·

Anaphors in contact: 1he di~tribution of intensifiers and reflexives in Amana German 115 (8)

a. b.

[DP the farmer himself] Io(the farmer) =the farmer

The In function plays a significant role in Gasfs (2006) study as well as mine here (which, of course, is fundamentally based on the former). When functioning as a marker of reflexivity, a syntactic--morpho--semantic interaction takes place. 'Io see this, ids return to the [±OD] distinction that is argued to exist between tr"c:lilsitive predicates. Here we observe a widely-held claim regarding -t-Oo-verbs: typically otherdirected predicates do not allow any of their internal arguments to be bound by the local subjects. In other words, the binding of an object pronoun by a co-argument is prohibited. Gast demonstrates that the In function lexically represented by the SELF-morpheme - is used here to prevent a violation of this syntactic distributional constraint. Consider the following examples in Dutch (take from Gast 2006: 5; (13)): (9)

a. "'Jar1 haat zich. Jan hates

b.

ANPH

fan haat [zich--zelf]. Jan haat [ANPH--SBLF]

1he distribution of antecedent and anaphor in (9a) is ungrammatical; the predicate haten 'to hate' belongs to the class of +On-verbs that are typically regarded to be 'other directed' and as such prohibit the appearance of the 'simple' reflexive zich. The 'solution' to this problem is delivered by means of the application of the ID func-tion in (9b); the anaphor zich does not occupy the traditional 'object position, but is rather embedded under a SELF-morpheme, which is interpreted by the In function. As explained by Gast (2006: 5), "as a consequence, binding of zich by the local subject Jan becomes possible, since zich is no longer in the object position of the predicate haat. Rather, it takes the position of the detenniner his in a sentence like John; loves [his; wife] (cf. Safir 1996 for a similar argument):' The structure of zichzelfis provided in Fibrure 1 below (from Gast (2006: 193)): DP 1

ID(x)

~ DP2

I

zich

J:l..'"T '

zelf

~ X

I

zich

Ay[ID(y)] I

''

I

zelf

Figure 1.

which are of the same semantic type as the identity function, that is, a sent of alternative

functions from individual'i to individuals."

116

MiChael "[ Putnam

As we will discuss below in the following sections, the proposed alterc:ttion in the structure of D P1 qua the Io function has the advantage of avoiding violations of the *OoBIND constraint. which requires the complements of an [-1-0o]-predicate P to not be bound by the subject of P, as illustrated in Figure 2 (from Gast (2006: 1.93)):

Jan;

wherei '-~

Figure 2.

In the configuration Figure 2 (DP 1), the reflexive zich occupies the equivalence of a determiner position (i.e. a specifier position), hence not crucially violating the *ODBIND constraint Although the anaphor zich inherits (syntactically) its referen-tial index from ]ani, the object of the verb haat (DP 1j) has a lexkal intention of its own (j), and is, as a result, not bound by ]anr As we shall see below, although Dutch, English, and German make use of the Io function to construction intensifiers (~ith a SELF--morpheme), they crucially differ with regard to the ranking of the *OoBIND constraint; Dutch and .English rank this constraint relatively higher than German. As could be expected, AG also employs the Io function in creating reflexive intensifiers; however, what remains to be seen can be summarized in the following guiding research questions: Ql: Does AG, which exhibits an anaphor/pronominal inventory very similar (if not identical) to modern middle Hessian dialects of German, also have a relatively low ranking of the *OoBIND constraint as similarly argued to exist in modern German? Q2: If the answer to Q1 is negative, does AG have a morpho-syntactic/ semantic reaction to the [±Oo] predicate distinction as evidenced in Dutch and English? Q3: If the answer to Q2 is in the affirmative, what is the cause/motivation for such a change (again, assuming that the original AG inventory and constraint rc:tnking (strongly) resemble those assumed to exist tor modern German)? We will return to these questions in Section 4 once a working analysis within the aTframework employed here has been established for AG (in Section 3), Dutch, English, and German.

Anaphors in contact: The distribution of intensifiers and reflexives in Amana German 117

2.1

Dutch

The following three subsections will offer insight into the distribution of anaphors in Dutch, English, and German. The choice to include Dutch is that it sits "in between" Dutch and German; i.e. although it does have a weak [+LocAL] anaphor zich (which modern English lacks), its zich+SELF-Morpheme reflexive intensifiers (e.g. zichzelf) surface in more environments than its German counterpart (e.g. sich selbst). In a framework that employs violable constraints this, of course, boils down to the crucial ranking of a small number of constraints to arrive at this distinction. The data and tableaux in these three subsections are taken directly from Gast (2006; Chapters 7 & 8) unless otherwise noted. Concerning Dutch, Gast (2006: 192) arrives at the following ranking for the constraints introduced in the previous section (i.e. (3)-(7)): (1O)

Constraint ranking for Dutch: BINDoM >>*On BIND>> PARSE>> FEe>> MEc

Here we can motivate the ranking for these constraints. Consider (11) below and the competition in Tableau 1 that follows: (11)

Jan; haat PRON(i)" Jan hates __

Tableau 1.

Jan; haat PRONm·

BINDOM

*ODBIND

Jan. haat zichr;l·

"!

Jan; haat hem(;)·

"!

F.Ec

,.

~Jan. haat zichr;1zelf

Jan; haat hem(ilzelf

M.Ec

"!

,."

The winning candidate [Jan; haat zich(i)zelf] of this competition is the ANPH+SELFMorpheme. This is due to the *OoBIND constraint outranking the economy constraints FEe and MEc. As mentioned in the previous section, the predicate haten 'to hate' is generally regarded to be a +On-predicate, thus the first candidate displaying zich cannot win. The final candidate exhibits the form hem(iJzelf ultimately loses due to a fatal violation of the economy constraint FEe. Let's now move to datum (12) and its corresponding tableau, Tableau 2: (12)

Jan; wast PRON(i)" John washes

us Michael T. Putnam Tableau2. Jan. wastPRONm· ~Jan; wast

*ODBIND

BINDOM

F.Ec

M.Ec

zichCil.

Jan; wast hemr;l· Jan. wast zichmzelf

*!

Jan; wast hemr;lzelf

*!

*!

,.

Once again, the ranking of*OoBIND above the economy constraints FEe and MEe calls the shots; since wassen 'to wash' is classified as a -On-predicate, it does not violate *On BIND. The other losing candidates eventually fall prey to the economy constraints FEe and MEe. Lastly, let us consider Example (13) and its corresponding competition represented in Tableau 3: (13)

Jan; zette de vaas voor John sets the vase before

PRON(iJ

neer. down

Tableau3. Jan; zette de vaas voor PRONm neer. ~Jan;

BINDoM

*ODBIND

F.Ec

M.Ec

zette de vaas voor zichr;l neer.

Jan; zette de vaas voor hemr;l neer. Jan. zette de vaas voor zichCilzelf neer.

*!

Jan; zette de vaas voor hemr;lzelf neer.

*!

*!

,.

As suggested by (13), sentences involving exceptional case marking (ECM) and raising-constructions will also deliver zich. Gast (2006: 194) makes the following observation: "Zich is optimal simply because there is nothing which prevents it from being used, and since it should be used whenever this is possible because of its economic nature. There are only two reasons justifying the use of a pronoun other than zich: (i) the input requires a pronoun that functions as a direct object of a [+Oo]-verb, or (ii) zich cannot be used because it is too distant from its antecedent, thus violating BINDoM:' In Tableau 3 Jan and zich are not co-arguments, therefore *On BIND is not relevant.

2.2

German

Although German shows a similar anaphor/pronoun inventory to Dutch, the complex ANPH+SELF-Morpheme rarely surfaces except in context involving contrastive focus.

Anaphors in contact: The distribution of intensifiers and reflexives in Amana German 119

As a result, the markedness constraint banning complements of an [+Oo]-predicate P may not be bound by the subject of P (e.g. *OoBIND). In German *OoBIND is ranked very low with respect to other economy constraints, as demonstrated by the constraint ranking for German in (14) (Gast 2006: 203). (14)

Constraint ranking for German: BINDoM >> FEe>> MEe >> *OoBIND

This key difference in constraint ranking between Dutch and German can be seen in the following examples. (15)

Hans; betrachtet John considers

PRON(i)"

Tableau4. Hans; betrachtet PRONw

BINDOM

FEe

MEe

*On BIND

~Hans. betrachtet sic~il'

Hans. betrachtet ihnr;J·

*

*

*!

Hans; betrachtet sich(i)selbst. Hans; betrachtet ihnr;lselbst.

*! *!

Example (15) represents a co-argument reflexive, which is naturally violated by the winning candidate (i.e. a violation of*OoBIND has been marked). The losing candidates, however, all receive fatal violations at the hands of higher-ranked economy constraints, namely, FEe and MEc. The same situation applies to Example (16) below; sich appears in the winning candidate because it is the most economic candidate. (16)

Hans; sah eine Schlange neben John saw a snake next to

PRON(i)"

TableauS. Hans; sah eine Schlange neben PRONw·

BINDOM

FEe

MEC

*OnBIND

~Hans. sah eine Schlange neben sic~n·

Hans; sah eine Schlange neben ihmw.

*!

Hans; sah eine Schlange neben sichrilselbst. Hans; sah eine Schlange neben ihmwselbst.

*! *!

*

120

Michael T. Putnam

(17)

Hans; bat

John

Maria, PRON(j) nicht zu beliigen.

asked Mary

not

to

lie

Tableau6. Hans1 bat Maria, PRONm nicht zu beliigen. Hans. bat Maria, sichm nicht zu beliigen.

BINDOM

Hans1 bat Maria, lh~~cilselbst nicht zu beli.igen.

MEc

*OnBIND

*!

*

..-Hans1 bat Maria, ~ 11 nicht zu beli.igen. Hans. bat Maria, sichc1lselbst nicht zu beli.igen.

FI!C

*!

*

*!

Example (17) denotes a situation where sich is not the winner; the first (loser) candidate that exhibits sich violates BINDoM (i.e. due to the [+LocAL] nature of sich, both sich and sich selbst must be bound in a locaL minimal domain). The pronoun ihn 'himAcc' is the optimal choice due to the fact that it only violates one economy constraint FEe whereas ihn selbst would violate two (e.g. FEe and MEc). As a result of *On BIND being ranked below FEe and MEc, the [±On]-distinction is not morphosyntactically attested or relevant in German.

2.3

English

Finally, before moving onto our discussion and analysis of AG, we must consider modern English. Modern English, which lacks a weak [+LocAL] anaphor altogether, now exhibits two series of pronouns: (i) simple pronouns such as him, her, etc. and (ii) the complex pronouns himself, herself, etc. 5 The simple pronouns are referentially independent, whereas the complex pronouns can be regarded as expressions denoting the identity function, applied to simple pronouns. 6 In addition to the constraints mentioned in Section 2 (e.g. (3)-(7)), Gast maintains that we need an additional constraint

5· The development of the reflexive and pronominal paradigms of modern English is extremely complex and cannot be done justice in this chapter. For more detail on the matter see van Gelderen (2000). 6. Therefore, as suggested by Gast (2006: 216; (34)), reflexive SELF-forms are interpreted as if they occupy OP-positions: (i) (ii)

[[himself0 p]] = In([[him]]) [[herselfnp]] = In([[her]])

As such, Gast assumes that 'full' (intensifying) SELF-forms are also used as expressions of an In function, and that In exhibits allomorphic variation; it can be realized as a fully fledged SELF-form in some contexts and as an affix -self in others.

Anaphors in contact: The distribution of intensifiers and reflexives in Amana German in order to properly analyze the English data, a constraint which he calls *CATEGORY MISMATCH/*CM: (18) *CATEGORY MISMATCH/*CM Elements of a given class C can only occur in syntactic positions allowing elements of category C (e.g. a verb cannot be inserted in a OP-position). The purpose for introducing (18) is due to the complex pronominal system for modern English mentioned above; *CM ensures that pronouns will only appear in contexts where pronouns are properly licensed. Relatedly, the *CM constraint insures that the clitic weak reflexive in English cannot surface as the object of a preposition (cf. (22) and Tableau 9. As we see below in (19) (Gast 2006: 216), *CM occupies the highest ranking among the other constraints in modern English. (19)

Constraint ranking for English: *CM >> BINDoM >> *OoBIND >> FEC >> MEc

We can see these constraints in action in Tableau 7, which analyzes Example (20): (20)

John; hates PRON(i)"

Tableau 7. John; hates PRON1.,. John; hates himr;l· .,..John; hates himc;lself. John; hates 0w·

*CM

BINDOM

*0DBIND *!

Fl!c * *

Ml!c *

*!

Given that the predicate hate is a [+Oo]-predicate, this would eliminate the losing candidates that have a simple pronominal him and the final candidate that exhibits only a null enclitic reflexive. Simply put, Io(him) does not violate *OoBIND. The constraint *OoBIND also plays a critical role in evaluating (21) below: (21)

John; washes PRON(i)"

TableauS. John. washesPRON111 • John; washes hiffitn· John; washes hiffit;lself. .,..John; washes 0cn·

*CM

BINDoM

*OnBIND

Fl!c *! *!

Ml!c *

Since the null-reflexive in English is still regarded to be [+LocAL], it does not violate *OoBIND. The other candidates exhibiting more complex morphological units,

121

12.2 Michael T. Putnam

i.e. simple and complex pronouns, are eliminated by the economy constraints (e.g. FEe and MEe). Finally, let's consider Example (22): (22)

John; saw a snake near PRON(i)"

Tableau 9. John 1 saw a snake near PRONiil.

*CM

BINDOM

...,John1 saw a snake near himul· John1 saw a snake near h~ilself. John. saw a snake near 0ril"

*OnBIND

FEe

MEc

* *

!*

*!

The final candidates bearing the null-reflexive violates *CM, because this enclitic cannot be licensed in this position (recall, the enclitic must be immediately adjacent to the verb; here it would be the object of the preposition). The simple pronoun him is seen in most dialects of English to be 'more economical' that himself. In summary, the typology of two general types oflanguages emerge with regard to their treatment of the economy constraints such as FEe and MEe and *OoBIND. Languages such as Dutch and English recognize the [±Oo]-predicate distinction morpho-syntactic/semantically and rank *OoBIND higher (e.g. *OoBIND » {FEe, MEe}). In German, we witness the exact opposite effect; namely, where economy constraints outrank *OoBIND and thus suppress this effect (e.g. {FEe, MEe} » *OoBIND ).

3·

Reflexives in Amana German

At this juncture, it should be relatively clear what is at stake with regard to any discussion of AG's treatment of reflexives and their syntactic distribution; namely, does AG exhibit a constraint ranking system that is 'faithfuf to its German-roots and rank *OoBIND lower than the economy constraints, or has this dialect in its moribund state undergone an internal or external change that finds the economy constraints ranked below *OoBIND? From the outset, it is worth noting that at this stage in the research, a clear-cut answer cannot be derived; however, with this (temporary) shortcoming being stated, there does appear to be are-ranking of the constraints slowly taking place. From a preliminary standpoint, let us adopt the following constraint ranking for AG: (23)

Constraint ranking for AG: BINDoM >> *OoBIND >> FEC >> MEC

(24)

Hans(l) wascht PRON(I)"

Anaphors in contact: The distribution of intensifiers and reflexives in Amana German 123 Tableau 10.

Hansm wiischtPRONrw

BINDOM

*ODBIND

F.Ec

M.Ec

~Hansm wiischt sichm.

*!

Hansm wlischt sichmselbst

*!

Hansm wlischt ihn,il. *!

Hansm wlischt ihn,;Jselbst

,.

The observant reader will note that the proposed constraint ranking for AG in (23) is identical to the constraint ranking posited for Dutch in Section 2.1 (cf. (10)). In (24), the fact that wasche 'to wash' is generally regarded to be a -Go-predicate avoids a *Oo BIND violation. The other losing candidates exhibit pronominal forms that violate the economy constraints. The data presented in (25) and (26) below, however, pose a significant challenge to the purported constraint ranking in (23). (25)

Hans(JJ gleicht PRON(i)'

Tableau 11.

Hans(i) gleicht PRONay

BINDOM

*ODBIND

Hansm gleicht sichm.

*!

Hansm gleicht ihnm.

*!

F.Ec

,. *

~Hansm gleicht sichwselbst.

Hansm gleicht ihn,;Jselbst (26)

M.Ec

*!

,.

F.Ec

M.Ec

Hans(JJ hasst PRON(;J·

Tableau 12.

Hansw hasst PRONm· ~Hansm

hasst sichr;r

~Hansm hasst

BINDOM

*0D8IND *!

*

sichmselbst.

Hansm hasst ihnm.

*!

Hansm hasst ihnr;;;elbst.

*!

*

Clearly, both gleiche 'to like' and hasse 'to hate' should be predicates generally classified as +Oo-verbs; however, as we see in Tableau 11, this constraint ranking excludes a grammatical candidate (based on the observation in Figure 3 below that both sich and sich selbst are possible for AG speakers with this predicate). To probe further whether the [±Oo]-predicate distinction is (overtly) recognized in the AG grammar, I returned to my informants with a list (based on Donaldson's (1997) list for Dutch

124 Michael T. Putnam

predicates (of which an abbreviated form appears in (1) and (2) in the introduction of this chapter)) generally regarded to be +Go- and -Go-verbs respectively. The informants' task was to identify whether sich (weak reflexive), sich selbst (intensifier), both, or neither of these elements are preferred to appear with these predicates. The results are presented below in Figure 3 and Figure 4.

+OD-predicates 'to attack' 'to beat' 'to hate' 'to know/be familiar with'

AG-1

AG-2

AG-3

both both both both

both both both sichselbst

sich selbst both both both

'to laugh about' 'to talk to' 'to care for/to take care of 'to amuse'

sichselbst sichselbst sichselbst

both sichselbst sichselbst

sich selbst sich selbst sich selbst

both

sichselbst

both

Figure 3. +Oo-predicates As illustrated in Figure 3, the tokens (N = 8) of typical +Go-predicates could be optionally marked with the status of both (i.e. indicating that either sich or sich selbst could suffice as a grammatical element) or only with the intensifier (sich selbst). As could be expected, the presence of sich selbst with predicates that were classified as ambiguous (i.e. with the status of both) were also judged to be focused for discourse purposes.? In this evaluation task, almost half (N = 11/24 = 45.8%) of these +Go-predicates in AG, however, can only appear with the intensifier sich selbst. This percentage represents a substantial increase based on the following observations: First, considering that standard German only allows for sich selbst to appear in focus positions - as appears to be the case in AG when these +Go-predicates were judged to be grammatical with both sich or sich selbst - the fact that some (i.e. almost half) of these verbs can only occur with the intensifier sich selbst clearly demonstrates that the morpho-syntactic/semantic recognition of the [±Go]-predicate distinction is underway in AG. Second, as demonstrated in Figure 4 below, -Go-predicates are overwhelmingly marked with only the weak reflexive sich.

7· Anecdotally, two of the informants pointed out that "sich selbst would be used in these instances only to emphasize that the action involved the subject and not someone else" (from informant AG-1).

Anaphors in contact: The distribution of intensifiers and reflexives in Amana German

+OD-predicates 'to get dressed' 'to move' 'to turn around' 'to overeat' 'to shave' 'to hide' 'to defend' 'to disguise' 'to wash'

AG-1

AG-2

AG-3

both sich sich neither both sich sich sich sich

sich sich sich neither sich sich sich sich both (but with a strong preference for sich)

sich sich sich neither sich sich sich sich sich

Figure 4. -Oo-predicates

Based on the same number of tokens (N = 8),8 AG speakers only marked these predicates as grammatical with either the intensifier or the reflexive 3 times (N = 3/24 = 12.5%). What these figures demonstrates is that a trend towards recognizing the [±On]-predicate distinction in AG is taking place and not that some form of flippant sich/sich selbst usage is underway. These figures clarifY this fact. Returning to the failed tableau (Tableau 11), the harmonic ranking of these constraints should therefore not be consider as rigid, but overlapping and stochastic. Therefore the result of this 'failed' competition is the result of two scenarios: (i) due to its moribund status, it is not possible (at this time) to achieve a completely consistent/harmonious ranking in AG, and (ii) perhaps hasse 'to hate' is ambiguous on the [±On] -predicate scale.9 In light of the evaluation task and data provided in Figures 3 and 4, I view the latter (i.e. option (ii)) to be the more likely scenario, although more detailed research in the future will bear this out. Finally, consider the AG translation of (27) in (28): (27)

John has many friends around him/*himself.

(28)

Der Johann(iJ hat viel Freunde urn PRON(iJ rum.

8. Due to the fact that AG lacks a reflexive 'to overeat' predicate, I removed this apparent lexical gap from further consideration in this study. 9· Volker Gast (p.c.) alludes to the possibility that rather than a static, polar binary distinction between +Oo- and -Oo-predicates perhaps it would be more appropriate to interpret the self vs. other-directedness of predicates along a scale with numerical values. In a situation of observable grammatical change (such as this one with AG), the re-ranking of the *OoBIND and the economy constraints could interact with the degree of ±Oo-predicates in a stochastic rather than an absolute OT-model. At this point in time, however, research into the gradient ranking of self vs. other-directedness of predicates is lacking.

125

12.6 Michael T. Putnam

Tableau 13. Der Johann,11 hat viel Freunde urn PRONm rum. ~oer Johannm

BINDOM

*ODBIND

F.Ec

MEC

hat viel Freunde urn sichm rum.

Der Johannm hat viel Freunde urn sichmselbst rum.

"! "!

Der Johannm hat viel Freunde urn ihn,il rum. Der Johannm hat viel Freunde urn ihn,ilselbst rum.

"!

*

The PRON-element, which functions as the object of the preposition um 'around: is clearly not affected by the *OoBIND constraint, while the rest of the candidates once again fall prey to the economy constraints. Taking these observations together, it appears- at least from a very preliminary view- that AG has undergone some form of morpho-syntactic/semantic change with regard to rules that regulate the distributional properties of its intensifiers and reflexives.

4·

Conclusions &: directions for future research

In conclusion, let us return to the three guiding research questions that we mentioned at the end of Section 2:

Ql: Does AG, which exhibits an anaphor/pronominal inventory very similar (if not identical) to modern middle Hessian dialects of German, also have a relatively low ranking of the *OoBIND constraint as similarly argued to exist in modern German? Q2: If the answer to Ql is negative, does AG have a morpho-syntactic/semantic reaction to the [±Oo] predicate distinction as evidenced in Dutch and English? Q3: If the answer to Q2 is in the affirmative, what is the cause/motivation for such a change (again, assuming that the original AG inventory and constraint ranking (strongly) resemble those assumed to exist for modern German)? Regarding Ql, data from our AG informants seem to indicate (e.g. (25), Tableau 11) that "OoBIND is ranked higher (in most cases at least) than the economy constraints in AG. This, of course, stands in contrast to what one finds in modern German (cf. Section 2.2) where the exact opposite effect can be observed. As a result of this change (qua Q2), we do witness a morpho-syntactic/semantic reaction to the [±Oo] predicate distinction as evidenced in Dutch and English. 10 Regarding the answer to

10. Although it is generally assumed in the OT -literature that learning algorithms take place via markedness constraint demotion (see Kager 1999: Chapter 7), it is unclear if a similar process is involved in language contact.

Anaphors in contact: Tb.e di.~tribution of intensifiers and reflexives in Amana German

the internal vs. external cause debate (qua Q3), at this point only mere speculation can be offered here. In other words, it is impossible to ascertain whether continued (sociolinguistic) pressure exerted by the dominant culture/language (i.e. English) has functioned as the cause or catalyst f{)r this change. A~ a matter of fact, perhaps a word of caution should be issued before advancing any sort of ad hoc assumptions that external influence must be at play here. This theory can easily be dismissed when one looks at Dutch (cf. Section 2.1); here exists a language with an almost identical anaphoripronominal inventory as AG with a similar constraint ranking regulating its syntactic distribution of these elements. Therefore, at this stage it is not productive to engage in such a 'windmill' argument. What is needed at this time are more studies of a similar nature focusing on more Gennan(ic) Sprachinsellanguages throughout the world. Lastly, looking forward, studies into the morpho--syntactic/semantic properties of these Sprachinsellanbruages also have the potential to provide informative perspec·· tives into the argument structure of inherent and non-inherent reflexive predicates (see Reinhart & Siloni 2004, 2005; Dobrovie-Sorin 2005; and Schafer 2008) in attrition environments. 11 As for the study at hand, more detailed data are also necessary from a more diverse pool of AG informants in order to obtain a more accurate depiction of this apparent linguistic change that is underway. Through this study, as well as future related ones, we could learn a great deal more about the interaction of these sorts of morpho-syntactic/semantic shifts in attrition environments.

References Burzio, Luigi 1991. Tb.e morpb.ological ba.~l~ of anaphora. ]ourr1al of Llr1guistics 27: 81-105. Burzio, Luigi 1996. Tb.e role of the antecedent in anaphoric relations. In Current Issues in Comparative G-rammar, Robert Freidin (ed), 1-45. Dordrecht: Kluwer. Burzio, Luigi. 1998. Anaphora and soft constraints. Inls tl:e Best Good Bnougl:? Optimality and Competition in Syntax, Pilar Barbosa, Danny Fox, Paul Hagstrom, Ma.1tha McGinnis & David Pesetsky (ed~). 93-113. Cambridge MA: The MIT Press. Chom.~ky, Noam. 1981. Lectures on Govemment ar1d Blr1dlr1g. Dordrecht: Foris. Chom.~ky, Noam. 1986. Barriers. Cambridge MA: 'Ihe MIT Press. Dobrovie-Sorin, Carmen. 2005. 'Ihe SE-anaphor and its role in argument reali7ation. In 7fte Biack14--ell Companion ro Syntax, M. Everaert & H. van Riemsdijk (eds), 118-179. Oxford: Blackwell. Donaldson, Bruce. 1997. Dutch. London: Routledge. Eckb.ardt, Regine. 2001. Reanalyzing selbst. Natural Language Semantics 9(4): 371-412. Everaert, Martin. 1986. 111e Syntax of Refle.xivlzation. Dordrecht: Foris.

n. For example, SE/weak-anaphors are generally assumed to be able to appear in the fol-· lo\Ving en\'ironments: reciprocal readings, anticausative, middle, and passive voice uses; see Kemmer (1993).

127

12.8 Michael T. Putnam

Gast, Volker. 2006. The Grammar of Identity: Intensifiers and Refkxives in Germanic Languages [Routledge Studies in Germanic Linguistics]. London: Routledge. van Gelderen, Elly. 2000. A History of f:nglish Reflexive Pronouns- Persotl, Self; and Interpretability [Linguistik AktuelliLinguistics 1bday 39]. Amsterdam: John Ben)amins. Hole, DanieL 2002. Agentive seibstin German. In Sinn und Bedeutung VI- Proceedings of the 6th Ant1ual Meeting of the Gesellschaft fur SemanNk, Graham Katz, Sabine Reinhard & Philip Reuter (eds), 133-150. Osnabriick: Institute of Cognitive Science. Kage1; Rene. 1999. Optlmailty Theory. Cambridge: CUP. Kemmer, Suzanne. 1993. Tile Middle Voice [Typological Studies in Language 23]. Amsterdam: John Benjamins. Kiparsky, Paul. 2002. Disjoint reference and the typology of pronouns In More than Words A Festschrift for Dieter Wunderilc.h [Studia Grammatica], Ingrid Kaufmann & Barbara Stiebels (eds), 179··226. Berlin: Akademie Verlag. Koster, Jan 1997. Anaphora and the uniformity of grammar. In Atomism and Binding, Hans Bennis, Pierre Pica & Johan Rooryck (eds), 235-250. Dordrecht: Foris. Konig, Ekkehard & Siemund, Peter. 2000. Intensifiers and reflexives A typological perspective. In Reflexives Fortns and Functions [Typological Studies in Language 40], Zygmunt Frajz.. yngier & Tract Walker-Curl (eds), 41-74. Amsterdam: John Benjamins. Reinhart, Tanya & Reuland, Eric. 1993. Reflexivity. Linguistic Inquiry 24: 657-720. Reinhart, Tanya & Siloni, Tal. 2004. Against the unaccusative analysis of reflexives. In Tile unaccusativity puzzle, Artemis Alexiadou, Elena Anagnostopoulou & Martin Everaert (eds), 159··180. Oxford: OUP. Reinhart, Tanya & Siloni, Tal. 2005. The lexicon-syntax parameter: Reflexivization and other arity operations. Linguistic Inquiry 36: 389···436. Rettig, Lawrence. 1976. Amana Today: A History of the Amana Colonies from 1932 to the Presetlt. Amana IA: Amana Historical Society. Safl.r, Ken. 1996. Semantic atoms of anaphora. Natural Language and Linguistic Theory 14: 545-589.

Schafer, Florian. 2008. The Syntax of (Anti-)Causatives: External Arguments in Change-ofState Contexts [Lingulst!k Aktuell/L!nguist!cs ~lbday 126]. Amsterdam: John Benjamins. Webber, Philip. 1993. Kolonie-Deutsch: Life and Language in Amana. Ames IA: Iowa State Press.

Lexical developments in Texas German* Hans C. Boas & Marc Pierce University of Texas at Austin

This paper offers new insights into on -going research on lexical borrowing in language contact situations by presenting a typology oflexical borrowings in Texas German (TxG), a critically endangered dialect that will go extinct within the next 30 years. We show that the lexicon ofTxG has not undergone any dramatic changes over the past four decades and that the dialectal origins of words that were still apparent when Gilbert collected his data in the 1960s can still be found today. We also argue that TxG should be classified as "stage 2" in Thomason and Kaufman's (1988) 5-stage borrowing scale ("slightly more intense contact"), which is characterized by lexical borrowing and slight structural borrowing in combination with conjunctions and adverbial particles.

1.

Introduction

One of the most prominent features of contact languages is the borrowing of lexical material. In casual contact situations, languages exhibit relatively slight lexical borrow-· ing, while in long·· tem1 intensive contact situations extreme structural borrowing may occur (Thomason & Kaufman1988: 74-76). Thus, in the case of (American) English contact with Finnish, tor instance, only one or two words have been borrowed from Finnish into English (sauna and possibly sisu 'guts'), while the much more intensive contact with French during the Norman period fundamentally altered the structure of English (the stress system of Old English differs radically from the stress system of Modern English, for instance, which is traditionally attributed to contact with French). In communities with high levels of bilingualism that are subject to socioeconomic and/or political pressures, one speaker group may eventually shift to the more prestigious language. As the language with the lower prestige is used less over time, its lexicon erodes.

""An earlier version of this paper was presented at the Germanic Linguistics Annual Conference 15 (Hanft: Alberta, Canada, May 2009). We are grateful to the conference participants for helpful comments and discussion; to an anonymou.'i referee for comments on an earli.er draft; and to Mike Putnam for his assistance in his role as editor.

130

Hans C. Boas & Marc Pierce

This paper investigates lexical borrowings in Texas German (hereafter TxG) to establish how far lexical change has progressed over the past century. 1 We first outline the status and history of TxG. Next, we present an overview of the literature on lexical borrowing in TxG and apply it to earli.er studies ofTxG. 'Ihe summary of previous accounts of lexi.cal erosion serves as the basis for our analysis of the present--day data in Section 4. We focus on a select number of phenomena to illustrate the types of changes taking place in the lexicon ofTxG. Section 5 discusses the question oflexic.al erosion in present-day TxG, and Section 6 presents our conclusions.

2.

The history of Texas German

The German language has a long history in Texas. Promises of land grants and transportation to Texas attracted a significant number of immigrants, mainly from northern and central Germany, beginning in the 1840s.2 By 1850 there were 8,266 German-born immigrants living in Texas (Jordan 1975: 48), and by 1860 there were approximately 30,000 Texas Germans, both immigrants and their American-born children (Jordan 1975: 54). German immigration to Texas eventually slackened, but the number of Texas Germans continued to increase: Eichhoff (1986) estimates that there were approximately 7 5,000-100,000 Texas Gennans in 1907, Kloss ( 1977) states that in 1940 there were approximately l. 59,000 Texas Germans, and Nicolini (2004: 42) suggests that at the beginning of the twentieth century approximately 1/3 of all Texans were of German anc.estry. For the first several decades of German settlement in 'H!xas, the 'Iexas Germans were relatively isolated, thanks to a number of political and social factors, ranging from the anti-slavery views held by most German settlers to deliberate attempts at self-sufficiency (see Salmons 1983 and Benjamin1909, respectively, on these points). This isolation, coupled with serious attempts at language maintenance, allowed for the development and spread of TxG: there were 145 church congregations offering German-language church services as of 1917 (Salmons & Lucht 2006: 168); there were numerous German-language newspapers and periodicals, some with very healthy circulation numbers (Texas Vorwiirts, published in Austin, had a circulation of approximately 6100 in 1900, according to Salmons & Lucht 2006: 174); there was

There is a long history of research into TxG, ranging from the pioneering studies ofEikel (1954) and Gilbert (1972) to more re'~ent works like Fttller and Gilbett (2003), Nicolini (2004), Salmons and Lucht (2006), Boas (2009), and Boas et al, (2010).

1.

See Biesele (1928), Moore (1980), Salmons (1983), Augspurg--Hacker (1984), Nicolini (2004), Boas (2005a, 2009), Salmons and Lucht (2006), and Boas et al. (2010) for additional overviews of German immigration to Texas and the changing status of German in Texas. 2.

Lexical developments in Texas German

a wide range of German literature written in Texas; there were German-language schools and numerous social organizations, including choirs, shooting clubs, and so on (see Nicolini 2004: 46-49 for further discussion of such groups). 'lhis situation eventually changed dramatically, starting with the passing of an :English--only law tor 'Iexas public schools in 1.909 (Salmons 1983: 188). World War I, especially following America's entry into the war in 1917 and the resulting increase in anti-German sentiment, along with the passage of another English-only law for public schools in 1918 (Salmons 1.983: 188), led to the stigmatization of German and the beginning of its decline. 3 World War II reinforced the stigmas attached to Germany, Texas Germans, and the German language. Institutional support for the widespread maintenance and use of German was largely abandoned, with devastating consequences for TxG. German--language newspapers and periodicals stopped publishing (Das Wochenblatt, published in Austin, stopped publishing in 1940) or switched to English as the language of publication (the Neu-BraunfeL~e1' Zeitung was the last to switch to English, in December 1957);4 some German-language schools closed and German instruction was dropped in others; and German--speaking churches replaced German-language services with English-lanbruage ones.5 After World War II, the increasing migration of non-German speakers to the traditional German enclaves and the general refusal of these newcomers to learn German led to the large-scale abandonment of German in the public sphere. The increased use of :English in the public domain pushed German even further into the private domain. At the same time, younger Texas Germans left the traditional Gennanspeaking areas for employment, education, or military service (Jordan 1977; Wilson 1977), and consequently switched to English as their primary language, which in turn weakened their command ofTxG. Also, Texas Germans increasingly married partners who could not speak German, and in such linguistically mil!:ed marriages, English typically became the language of the household. Children raised in such households are typically monolingual in English, or have only a very limited command of TxG, normally a few stock phrc:tses like prayers or profanities (Nicolini 2004; Boas 2005b). Finally, the development of the American interstate highway system starting in 1956 made the once-isolated TxG communities much more accessible. This new accessibility cut both ways, as it was now easier both tor non--German speakers to visit or live in the

3· This situation was not unique to Texas; similar situations arose in various other states \Vith a strong German presence. 4· See Salmons and Lud1t (2006: 17 3--·1 78) for further discussion of the Germa.Il--laiJ!,>uage press in Texas. 5· Some German-la.Ilgllage services were retained, especially on holidays like Good Friday and Christmas (Nicolini 2004: 101), and at least one church still offers a German-language service on 'fifth Sundays' (Roesch 2009).

131

132 Hans C. Boas & Marc Pierce

originally German-speaking communities, and for German-speakers to accept employment in more urban areas. Both of these possibilities led to the spread of English at the expense of German. Despite these factors, in the 1960s approximately 70,000 speakers of TxG remained in the "German belt:' which encompasses the area between Gillespie and Medina Counties in the west. Bell and \Villiamson Counties in the north, Burleson, \Vashington, Austin, and Fort Bend Counties in the east, and DeWitt. Karnes, and Wilson Counties in the south. Today, however, only an estimated 8---1.0,000 Texas Germans, primarily in their sixties or older, still speak TxG fluently (Boas 2003, 2005b, 2009), and English has become the primary language for most Texas Germans in both private and public domains. With no signs of this language shift being halted or reversed and fluent speakers almost exclusively above the age of 60, TxG is now critically endangered and is expected to become extinct within the next 30 years. 6

3·

The Texas German Dialect Project

In 2001 Hans C. Boas founded the Texas German Dialect Project (TGDP) in order to record, document, and analyze the remnants of TxG; preserving TxG is seen as a less crucial (and probably impossible) task. 7 Over the past eight years, members of the TGDP have recorded three different types of data. The first type of data consists of English word lists and sentences taken from the Linguistic Atlas of Texas German (Gilbert 1972) and from Eikel (1954). An interview begins by eliciting personal information (date and place of birth, level of education, language spoken at home when growing up, etc.) from the interviewee. Interviewers read the English words and sen-tences to the informants who are then asked to translate these words and sentences into TxG. The interviews last about 20-60 minutes, are recorded using various devices, and normally take place at the informants' homes, nursing homes, museums, or local churches. The use of word lists and sentences enables the comparison of the current recordings with data collected over four decades ago, and also provides well-focused and well-controlled data sets giving information about the distribution of specific phonological, morphological, and syntactic features in present--day TxG.

6. Our assessment of the current situation therefore differs sharply from the claim of Nicolini (2004: 165) that "Interviews mit alten Texanem lassen den Schluss zu, dass die deutsche Sprad1e amEnde des 20. Jahrlmnderts lebendiger ist, als es in der gerrnanbtischen Forschung gemeinhin gesehen wircl." 7· See Boas et al (201 0) for more extensive discussion of the TGDP, including technical details.

Lexical developments in Texa~ German 133

The second type of data seeks to capture the informants' daily use of TxG. An eight page questionnaire serves as the basis for sociolinguistic interviews conducted in German. At the beginning of an interview, interviewers speak (standard) German to the informants to begin eliciting personal information about the informants in TxG (date and place of birth, place of origin of informants' ancestors, languages spoken with the parents at home, etc.). During this first phase of the interview, informants are typically aware of the recording device and pay attention to their speech. However, they soon become more relaxed, forget about the recorder, and begin to respond to questions in TxG. Th.e second section of the interview consists of about 140 questions in German about a wide range of topics, including childhood activities, the community, religion, education, living conditions, tourism, government, language, and current activities. These questions are designed to produce casual, relaxed conversation in which informants are given the chance to respond freely in TxG without being asked to produce specific linguistic structures. Allowing informants to speak freely also makes it possible to examine linguistic features of TxG that were not noticed by previous studies that did not include such elicitation methods. 'Ihese interviews typically last 40 to 60 minutes. The third type of data seeks to capture the informants' use ofTxG when participating in activities with other Texas Germans. In order to record this type of data, we chose card-playing activities, dinner preparation, and farm chores. After filling out the consent torms, informants are given wireless microphones which are linked to a recorder. Interviewers leave the area and do not get involved in the 60-100 minute long recording activities. The three scenarios enable the collection of data in a variety of environments that involve different usages of TxG. The three types of spoken data are augmented by a written biographical question-naire. This questionnaire elicits information about age, date ofbirth, level of education, domains of language use (TxG and English), and language attitudes, among others. The biographical data are used to create the metadata records for each informant and each interview to be included in the digital on-line archive. Since the beginning of the project, members of the TGDP have interviewed more than 350 speakers, totaling more than 650 hours of data, available in the on-line Texas German Dialect Archive (TGDA), available at <www.tgdp.org>. We rely heavily on this data in the following sec-tions, and turn now to an analysis of the evidence involving recent lexical borrowings.

4· 4.1

Lexical borrowing

Social contexts supporting borrowing

Thomason and Kaufman (1988) maintain that borrowing can be classified according to the types of settings in which it takes place, and propose a borrowing scale of different levels, each representing an increasing intensity of contact as well as an increasing

134 Hans C. Boas & Marc Pierce

typological distance. Perhaps the two most important social features of their borrowing scale are intensity of contact and cultural pressure. Winford (2003: 30) presents the following abridged version of Thomason and Kaufman's scale:8

Table l. Thomason & Kaufman's (1988) borrowing scale (Winford 2003: 30) Features

Stage 1

Casual contact

Lexical borrowing only

2

Slightly more inteiJSe contad

3

More intense contact

4

Strong cultural pressure

5

Very strong cultural pressure

Slight strudurdl. borrowing; conjunctions and adverbial particles Slightly more structural borrowing; adpositions, derivational affJXes Moderate structural borrowing (major structural features that cause relatively little typological change) Heavy structural borrowing (major structural features that cause significant typological disruption)

Winford notes that stage 1 is the most typical scenario, where there is only marginal contact with other languages. Individual words are borrowed, but speakers of the recipient language typically do not achieve fluency in the donor language. Examples include native American words such as skunk or teepee borrowed into American English during the English colonization of North America or the intro-duction of English loanwords such as bat and song into Japanese (with phonological adaptation) beginning in the mid-nineteenth century (Winford 2003: 30-31). Stage 2 usually involves some degree of bilingualism in the recipient language community. Winford (2003: 33) makes particular reference to the languages of immi-grant groups and ethnic minorities in this context, which are often absorbed into a larger host community. According to Winford, the languages of such groups are particularly open to lexical borwwing from the dominant language. During the first half of the 20th century TxG speakers were at this second stage, first becoming bilingual while at the same time borrowing more and more words from English into TxG, and eventually shifting entirely to the donor language, i.e. English. This process can be promoted by social factors such as the asymmetry in power and prestige of the languages involved. In the case of TxG, Boas (2009) argues that its massive loss of prestige following World War I was one of the main factors that curbed the use of German in the public domain and even among some TxG families who decided to raise their children in English. Over time, English crept into more and

8. See 'Ihomason and Kaufman (1988: 74-76) for a significantly more detailed version of this scale. For a detailed critique of this borro\o\oing scale, see Wolsemuth (2009).

Lexical developments in Texas German 135 more private domains, eventually replacing TxG as the main means of communication in families. As one New Braunfels informant states: "As the oldest kid in the family, I was raised in German, but my four younger siblings were all raised in English. Once my brothers and sisters were old enough to carry on conversations, I talked to them in English, because they did not understand any German. Soon after that, I also switched to English when talking to my parents:' We return to this point below and show that by the mid-twentieth century TxG had already adopted a substantial amount ofloans from English and was on its way towards stage 3 of Thomason and Kaufman's model. Stage 3 is typical of more intense contact situations (more bilinguals, attitudes favoring borrowing, etc.). At this stage, both basic and non-basic vocabulary is borrowed. Some structural features are also borrowed at stage 3, but typically without significant typological change in the recipient language (Thomason 2001: 70). Below we arbrue that in the middle of the 20th century TxG exhibited features characteristic of Thomason & Kaufinan's stages 2 and 3. Stage 4 is characterized by intense language contact where very heavy bilingualism is the norm among speakers of the borrowing language. Strong cultural pressures pro·· mote the borrowing process, leading to lexical borrowing in all sections of the lexicon, as well as moderate structural borrowing, leading to minor typological changes in the recipient language (e.g. the beginning of word order changes). Finally, stage 5 takes place under very strong cultural pressure, which promotes heavy structural borrowing that eventually leads to typological disruptions in the recipient language (Winford 2003: 30). Factors like "need" and "prestige" also promote lexical borrowing.9 Speakers often need new words when they are exposed to new areas of cultural knowledge or technical innovation. Such needs clearly motivated the borrowing of words such as armadillo, electricity, and county commissioner from English into TxG: armadillos are not native to Germany, electricity was not yet used by the time the majority of Gennan immigrants left for Texas, and German immigrants were not familiar with the concept of county commissioner and consequently lacked a native word for it. Borrowing words from English into TxG thus allowed its speakers to fill gaps in their lexicon easily without having to invent new words. Vve now turn to the question of how loanwords are integrated into the redpient language and whether there are any constraints on borrowing. 4.2

Structural constraints on borrowing

A number of structural factors restrict the degree and type of lexical borrowing. Following earlier accounts by Whitney (1881) and Haugen (1950), Muysken (1981)

9· See Poplack et al. (1988) and Treffers-Daller (1994) for discussion of other factors influencing the degree of borrowing.

136 Hans C. Boas & Marc Pierce

proposes a hierc:rrchy of borrowability which captures the general observation that open-class items such as nouns and verbs are more easily borrowed than dosed-class items such as pronouns and subordinating conjunctions (see also Poplack et al. 1988 and Poplack & Meechan 1.998 on this point). Nouns> adjectives> verbs> prepositions> coordinating conjunctions> quantifiNs > deterrniners > det.:>rminers >free pronouns> clitic pronouns> subordinating conjunctions

F4,JUie I. Hierarchy ofborrowabUity

Syntagmatic constraints between lexical items also determine whether lexical items are borrowed. For example, Winford (2003: 51-52) observes that categories like verbs or prepositions are not as easily borrowed as nouns and adjectives because they govern other categories and assign case to them. Similarly, lexical categories with greater morphological complexity in their paradigms are borrowed less frequently than those without such complexity (Winford 2003: 52-53). Before determining tl1e make-up of tl1e lexicon in present-day TxG, we discuss the relevant literature on English borrowings into TxG from the 1960s and 1970s to establish what types of words were borrowed earlier and to what degree. 4·3

Earlier accounts of borrowings into TxG

Gilbert (1965) provides the first in--depth study ofloanwords in TxG, based on data he had collected himself as well as data from other published sources. According to Gilbert ( 1965: 104), early written sources ofTxG (letters and diaries starting in the 1840s) show the "enthusiasm with which newcomers absorbed the English language and culture during the first few years or months after their arrival" in 1exas, along with the fre-quent use of English terms f{)r a number of plants and animals, utensils, and particular types of activities such as cam pen 'to camp'.l0 Gilbert (1965) further notes that more and more words from the semantic domains of higher culture, government, technology, and contact with the world in genercil were eventually borrowed into TxG, especially after the decline in prestige of German. A handful of conjunctions such as because, but, and except were borrowed into TxG, and have replaced the corresponding German forms (Gilbert 1965: 110). Gilbert's inventory of borrowings includes loan translations such as der .Feuerpiatz 'the fireplace'; loan extensions such as die Kann 'the buckef, 'the can of food'; and hybrid compounds such as die peach pie, among others. These loan words were integrated to different degrees into the TxG phonological system. In Gilbert's view, the oldest loanwords such as der Norder 'northerly wind'

10. This particular example is now also frequently used in standard German, alongside the older term zelten 'to camp' (from da..~ Zeit 'tent').

Lexical developments in Texas German 137 were totally assimilated phonologically, while later loanwords like der Blcmket or der Television were only partially assimilated or were left unas similated (Gilbert 196 5: 110 ). Age clearly influenced the degree of integration, as fewer younger people attempted to integrate borrowed words phonologically into TxG. 1hey instead tended to aim for ''a more or less faithful reproduction of the English model, even if the process entails the injection of a gross foreign element into nearly every utterance, thereby endangering the integrity of the language as a system" (Gilbert 1965: 110). Finally, despite the apparent ease with which Texas Germans borrowed words from English, Gilbert (1.965: 110) observes that the actual number of .English words in Tx:G as spoken at the time of writing "remains small, probably less than 5 percent~ In a slightly later study, Meister (1969) offers an analysis of English loanwords in TxG that includes more detailed information about pronunciation. Based on the data collected tor Gilbert ( 1972 ), Meister documents extensive dialectal variation across the German-belt. which he treats as a refle.lC of the vocabulary brought to Texas by Gennan immigrants. For instance, based on Gilbert's sentence The animal died out in the pasture (later published in Gilbert 1972, Map 10), Meister (1969: 8---9) investigates the different lexical variants of to die and their dialectal origins. In addition to the more common form gestorben, Meister discusses the distribution of krepiert and verreckt (both denoting quite miserable ways of dying), and concludes that the distribution of these three forms in TxG largely reflects their distribution in the donor dialects of TxG. On the other hand, the most common way to express 'to die' in TxG was tofgega.ngen, which does not appear in all traditional German dialects, according to Meister (1969: 9). Meister further shows that Gilbert's TxG data contain both native German words and their (borrowed) English counterparts. He indicates that there is a con-tinuum of borrowing English words and replacing native German words with them: at one end of the continuum we find large numbers of English loanword variants and few Gennan counterparts for a given word or phrase, e.g. the creek (Gilbert 1972, Map 137), which is used by almost all of Gilbert's informants, and which is pho·· nologically integrated into German. Other English variants borrowed into Tx:G include branch and gully, which are used to a lesser degree (particularly in the eastern part of the German-belt). German words occasionally used for creek include Bach, Bruch, Wtlssergang, Graben, Strom, and Fluss, all of which refer to different types and sizes of flowing water (Meister 1969: 44). 11 On the other end of the continuum we find a strong use of German words, with English loanwords replacing German words only occasionally, e.g. drawer (Meister 1969: 27-28), where ninety percent of

n. Other examples where T:li.-G speakers predominantly prefer to use English loanwords include icicles, pumpki11, candy, and tank. among others (Meister 1969). For the most part, these loanwords are phonetically integrated into TxG, which suggests that they were bor-· rowed into TxG comparatively early on.

138 Hans C. Boas & Marc Pierce

Gilbert's informants preferred German Schubiade (or some lexical variant thereof) to the English loanword drawer. Wilson (1977) emphasizes the strong regional differences in the TxG lexicon. Focusing on the area around Giddings and La Grange (about 60 miles east of Austin) he observes that the local variety of Tx:G is influenced by Wendish (e.g. der Bobbak 'boogeyman'), Saxon (e.g. schmoochen 'to smoke'), and English (e.g. der Belt 'the belt: among many others). With respect to English borrowings, Wilson (1977) maintains that the most common type of English loan words in 'TxG are words tor new concepts with which the settlers were not familiar (die Roach, der Airplane, etc.). In Wilson's view, the settlers maintained their German vocabulary to a large degree and therefore did not feel a need to replace more words. In tact, he points out that some German words such as Luftschiff'airship' (although meaning 'airplane' in TxG), Auto 'car' and Ka.ugummi 'chewing gum' existed alongside English words. As for phonology, the phonological integration of English loanwords is not always complete. Wilson (1977: 53) views the use of [v] for [w] ('it was vunderful') as one of the defining characteristics of the "German accenf' of the area surrounding Giddings and La Grange. At the same time, Wilson reports considerable phonological interference from English in the speech of younger speakers without much knowledge of German. Wilson (1977: 57) concludes that "Texas German is essentially good standard German." 12 Jordan ( 1977) offers a more general account of TxG spoken in the western Hill Country, i.e. to the west of New Braunfels. With respect to lexical changes, he point~ out that the time between 1845 and 1945 was a century of tremendous change. Along the lines of earlier observations made by Gilbert and Wilson, he identifies technological developments like trucks, road graders, telephones, phonographs, radio, television, and so on, as the impetus for many English borrowings into TxG. In describing the differences between Standard German and TxG Jordan (1977: 61) notes that "while German built up its new vocabulary in Europe, the German Texans borrowed the needed terms from English, and hundreds of English words slipped in by default~ Besides words describing unknown objects or processes, TxG also borrowed words for which there were already German words available, such as das Rope (Standard German das Seil) and die Fence (Standard Germander Zaun). Jordan also reports the borrowing of nouns such as car and cotton, as well as compound nouns made up of English and German words such as Stacheldrahtfence

12. Compare here the statement by Gilbert (1965: 102) that TxG "deviates in certain characteristic ways from Contemporary Standard German as described by Si.ebs and Duden. Nevertheless it is sufficiently intelligible to the speaker of Standard German to be classed as a colonial variety of the standard language and not as a separate entity. Mutual intelligibility is still very goo d."

Lexical developments in Texas German 139 'barbed wire fence' and Schweinepenne 'pig pen: Besides nouns, Jordan notes the borrowing of multi word expressions such as Vieh aufrmmden 'round up cattle: die Kuh dehornen 'dehom the cow: and den Draht stretchen 'stretch the wire: among many others. Verbs such as cra.nken 'to crank' were also borrowed (and are normally weak). Many borrowings were apparently phonologically integrated, or at least exhibit what Jordan (1977: 63) calls "a distinct Gennan flavor and a strong German intonation:' At the end of his paper Jordan (1977: 68-71) provides an extensive list of almost 300 English loanwords from various semantic domains such as administration, tech-· nology, agriculture, nature, transportation, and education, albeit without any exact indications of pronunciation, stating only that "the pronunciation is a somewhat Germanized fonn of the local Texas English norm" (Jordan 1977: 68).

4·4

Evaluation of earlier accounts

One of the challenges in evaluating earlier accounts of borrowings into TxG is that they only offer anecdotal evidence. For instance, Jordan's (1977) impressive list of borrowings fmm various semantic domains lacks detailed infi.)rmation on the degree of phonological integration, while Meister (1969) offers phonetic transcriptions, but only covers a small set of English borrowings. Ideally, we would like to have an exhaustive corpus ofborrowings into TxG accompanied by precise phonetic transcriptions. Despite the absence of such a c.orpus, it is obvious that borrowings from English into TxG have been somewhat limited in size and type. ru noted above, Gilbert (1965: 110) estimates the amount of English lexical material at the time of writing at about 5%, and all the analyses discussed above seem to agree that borrowings affect almost exclusively the semantic domains of administration, education, technology, agriculture, telecommunication, and tnmsporta·· tion. among others. The degree of phonological integration suggests that the majority ofborrmvings were readily integrated into TxG, with some variation between speakers. Returning to Thomason and Kaufman's (1988) borrowing scale discussed above, the question arises as to how we should classify the TxG lexicon at this stage. Besides lexical borrowings, which are indicative of stage 1, previous analyses also mention slight structural borrowings. f-or example, Gilbert ( 196 5: 109-110) reports that "English verbal constructions of the forms 'he goes; 'Does he go?' and 'He is going"' have exact coun-terparts in German. Similarly, the function and syntactic position of English for seems to have influenced its use in sentences such as Was solln mir fighten for? What should we fight for?' (Gilbert 1965: 110). Besides slight structural borrowings, Gilbert also reports borrowing of conjunctions such as because, but, and except, but only to a limited degree (see above). His claim that these conjunctions "have been incorporated into vari-ous syntactic constructions" (Gilbert 1965: 1.10) suggests that they did not replace their German counterparts across the board. Instead, they appear to be only tied to specific syntactic constructions and multi-word expressions, indicating tl1at tl1e borrmving of

140

Hans C. Boas & Marc Pierce

these closed-class elements presumably took place not as individual words, but instead occurred when specific multi-word expressions (idiomatic phrases, particular syntactic constructions, etc.) were borrowed into TxG and were later reanalyzed. In sum, the historical data indicate that the TxG lexicon of the mid-twentieth century exhibited characteristics of Thomason and Kaufman's second stage, which includes slight structural borrowing as well as borrowing of conjunctions and adverbial particles (besides, of course other lexical borrowing at stage 1). None of the previous analyses up to the 1970s offers any evidence for more intense structural borro-wing characteristic of the higher stages of Thomason and Kaufman's model. 13

5·

lexical borrowings in present-day TxG

To compare the historical TxG data with more recent data, we now tum to the interview data recorded by the TGDP "'ith 52 speakers in New Braunfels between 20012006. A full- fledged comparison and analysis of all of the relevant data would go far beyond the scope of this work, so we focus on a few illustrative examples representative of the overall trends. All New Braunfels speakers exhibited fluent speech during open-ended interviews, and were typically at ease with a wide range of topics. The only types of words that were problematic were trom the semantic domains in which heavy borrowing from English was already described by earlier research (education, administration, nature, technical innovation, etc.). Consider the following data from open-ended sociolinguistic interviews where a question or a short hesitation signals that the speaker does not know the appropriate Gem1an word. 14 (I)

a.

Aber die habn zusammengr,pielt in ihr ihre but they have together-played in their their uh Schulyard. (1-27-1-11) uh school-yard 'But they played together in their schoolyard~

b.

. .. da war ein - ein uh Highway Patrolman da. (1-32-1-2) there was a a uh highway patrolman there 'There was a highway patrolman:

c.

Ich hab uh business studiert. (1-45-1-4) I have uh business studied 'I studied business:

1here is also some very minor bonowi.ng from Spanish into TxG (largely in Medina County), whidt we do not address here.

13.

14. The numbers following each example are the unique file numbers referencing the files in the on--line TGDA. See Boas (2006) for details.

Lexical developments in Texa~ German

d.

Wie sagt man

naturalized:? (1-51-1-1)

naturalized how says one 'How do you say naturalized?'

We now discuss some specific examples from the translation task. The first n.vo examples, Beerdigung 'funeral' in 'Table 2 and Fussboden 'tloor' in Table 3, come from the relatively basic semantic fields "life and death" and "areas in the house". Gilbert's (1972) data for the two words show very little dialectal variation, and no English loanwords. The TGDP data are almost identical with Gilbert's data, except for t'i
Gilbert (1972) TG DP-Informants

Total-TGDP

Beerdigung

Begrabnis

Beerdigung and. Begrabnis

'1 (27%)

9 (60%)

2 (13%)

24, 25, 29, 30, 32, 34, 35, 71, 72, 76, 79, 80, 84, 88, 107, 123, 124, 125, 129, 139, 159, 165, 167

27, 28, 33, 77) 78) 82) 83, 128, 138, 153, 160, 161,170, 171, 173

96

23 (59%)

15 (38%)

funeral

None

75

26, 60, 62, 85, 108, 110, 155, 161, 168, 169, 172, 174.

1 (3%)

12

'H1hle 3. Pussboden ('floor') (Gilbert (1972), Map (122) FuBbod.en Gilbert (1972) TGDP-Informants

Total-TGDP

15.

Boden

FuB

floor

1 (7%)

14 (93%)

24, 26, 27. 30, 32, 33, 35, 72, 75, 76, 79, 80, 85, 88, 123, 125, 129. 139, 153, 160,161,165,171, 172,174

25, 28, 29, 3'1, 62, 71, 77, 78, 82, 83, 84, 96, 107, 108, 110, 124, 128, 138, 155, 159, 167, 168, 169,170,173

60

164

25 (48%)

25 (48%)

1(2%)

l (2%)

Numbers in the table refer to speaker numbers in the TGDA.

None

141

141

Hans C. Boas & Marc Pierce Next, consider the various TxG realizations of peca.n tree reported by Gilbert (1972) and how TGDP speakers differ from Gilbert's informants. Table 4 shows that the hybrid compound Pecanbaum 'pecan tree' continues to be used more often than the native German Nu.ssbctum 'nut tree'. 1he most significant point revealed by the TGDP data is the great number of speakers who did not provide an answer to this word. In comparison, a much smaller number of present-day speakers did not remember the translation of garden rake, as Table 5 illustrates. At the same time, the TGDP informants exhibit a much greater degree oflexical variation for this word that can be found in Gilbert's (1972) data. Table4. Pecan(nuss)baum 'pecan tree' (Gilbert 1972, Map 140) Pecanbaum

Gilbert (1972)

Nussbaum

10 (67%)

1 (6%) 32, 35,

Informants

25, 26, 28, 29, 30, 33, 75, 88, 110, 125, 128,

Total-TGDP

II (79%)

2 (14%)

TGDP-

Nussbaum and Pecanbaum

Baum

None

138

24, 27, 34, 60, 62, 71, 72, 76, 77, 78, 79, 80, 82, 83, 84, 85, 96, 107, 108, 123,124,129,139, 153, 155, 159, 160, 161, 164, 165, 167, 168, 169, 170, 171, 172, 173,174

4 (27«Xl)

1 (7%)

38

Table 5. Gartenrechen 'garden rake' (Gilbert 1972, Map 108) Gartenrechen

Gartenreche

Gilbert

Rechen

Gartenrake

14 (93%)

Harken

Other 16

None

1 (7%)

(1972)

TGDPinformants

Total-TGDP

24, 25, 29, 32, 76, 79, 82, 83, 123, 138, 160, 171

72,96, 125, 129

26,27, 80, 88,110, 124, 139, 155, 159

12 (31%)

4 (10%)

9 (23%)

30, 35, 164, 173 71, 75, 84, 153, 128, 165, 167, 170 172, 174

8 (21%)

2(5%)

4 (10%)

28,33, 34,60, 62, 77, 78,85, 107, 108, 161, 168, 169 13

This category includes lexical variants such as Garte11harge, Garte11reke, Garte11rech, and Garte11rache.

16.

Lexical developments in Texas German 143

Finally, consider the loanwords creek and candy in Tables 6 and 7. The data demonstrate that both words were well established in TxG some four to five decades ago, although in different ways. While creek had been phonologically integrated into TxG for the most part, candy retained its 'Texas English pronunciation. 'Ihis is still generally the case in the TGDP data, although some speakers now use the English pronunciation of creek, while others also use comparable German equivalents such as Fluss 'river' and Bach 'creek, brook'. Table6. Bach ('creek') (Gilbert (1972), Map (136)

Gilbert (1972) TGDP--Informants

Total-TGDP

Krik

Krike

13 (87%)

2 (13%)

Creek

Fluss

24, 25, 29, 30, 33, 35, 71, 72, 76, 79, 80, 82, 83, 84, 85, 88, 96, I 08, 110, 123, 124, 125, 129, 138, 139, 153, 155, 159,160, 171, 172, 173, 174

26,27, 77, 161, 164

34, 60, 75, 107, 128, 169, 170

33 (72%)

5 (11%)

7 (15%)

Bach

Other

28

32,62, 78, 165, 167, 168

6

Table 7. Bonbons ('candy') (Gilbert (1972), Map (139) Zucker

None

24, 25, 26, 27, 28, 29, 30, 32, 33, 3'1, 35, 60, 71, 72, 75, 76, 77, 78, 79, 80, 82, 83, 84, 85, 88, 96, 107,108,110,123,124,125,128,138,139,153, 155, 159,160, 161,164,168, 169,171,172, 173, 174

170

62, 129, 165, 167

47 (98%)

1 (2%)

4

c-andy Gilbert (1972) TGDP-Informants

Total-TGDP

15 (100%)

This brief comparison demonstrates three important points. First, the number of informants who do not remember a specific word differs from item to item. Based on the limited data discussed above, lexical loss appears to be lowest for native German words belonging to core semantic domains, e.g. Fuss baden 'floor: as well as for wellestablished English loanwords such as creek or candy. It is important to keep in mind that the rate of lexical reduction in 'LxG is somewhat minor when compared with other studies on d}i.ng languages (e.g. Dorian 1973; Ruoff 1973; Mithun 1989, and Holloway 1997, among others). This is particularly true when the different factors discussed above (age, intensity of contact with other speakers, etc.) are taken into account, as these factors point to parallels among elderly monolingual speakers who have not used words from certain semantic domains for a long time. Second, there are apparent

144 Hans C. Boas & Marc Pierce

exceptions to the trend that loanwords from the dominant language become more prevalent in language death situations, as noted above with the example of sink, where, although all of Gilbert's (1972) informants used the English loanword, a great number ofTGDP informants use German hybrid compounds such as Kichensink 'kitchen sink.' instead. At this point it is not clear whether these new forms signal an increased aware·· ness of TxG identity or whether their occurrence is triggered by other factors. Third, although it seems possible to observe different types oflexical developments in the data, it is unclear whether it is possible to arrive at systematic patterns that would allow tor predictions about whether a particular type of word will develop one way or another. Clearly, further research needs to explore this issue by systematically indexing all of Gilbert's (1972) lexical data and comparing it with the present-day TGDP data.

6. Lexical erosion in present-day TxG? The final thematic section of this paper considers lexical erosion and decay in presentday TxG. Lexical loss is one of the most obvious characteristics of dying languages, and it is relatively easy to identify. Absolute lexical loss usually affects words for objects that are no longer culturally relevant, such as shoemakers awl (Gilbert 1972, Map 119 ), while partial lexical loss typically takes place inlong-te.rm language contact situations with widespread bilin!,rualism. In these situations, vocabulary from the dominant language often replaces native lexical items, a process known as relexifi.cation (Craig 1997: 262). An examination of the relevant literature reveals a wide range of opinions on lexical loss. It seems uncontroversial that lexical loss is typically subject to age gradation, in that older tluent speakers tend to remember more words than younger tluent and semi-fluent speakersP 'Ihe causes of lexical loss, on the other hand, are significantly more controversial especially because, as Thomason (2003: 704-705) observes, multiple causation is often at work in contact-induced language changes. 18 While authors such as Woolard (1989) propose a "loans to loss" model where extensive borrowing from the dominant language over long periods results in language loss, authors like Cook (1989) claim that such convergence towards the dominant language does not occur in a moribund language. In this view, simplification such as lexical reduction is a result of internally motivated processes in the speech of semi-speakers. Thus, it is

Dorian (1973: 414) observes, for instance, that semi-speakers often "feel sure their elders had 'more word'! for things' than they have themselves."

17.

t8. For instance, in some cases phonological changes are intimately tied to lexical changes, as discussed in significant detail with respect to TxG in Boas (2009).

Lexical developments in Texas German 145 difficult to establish universal patterns capable of predkiing specific patterns of lexical loss. To avoid these problems, Andersen (1988) focuses on the role of the semi-speaker, highlighting the different situations in which the moribund language is still used; while Sasse (1992) argues that semi-speakers show the highest degree of lexical loss because of imperfect acquisition, usually caused by inadequate input and/or the limited opportunity to use the moribund language. Similarly, Wolfram (2002: 776) points out that "the lexical inventories of speakers of moribund varieties will depend on their experience in different situational domains, with frequently used vocabulary in common domains the most persistent as the language dies~ 1he situation-bound use of vocabulary is certainly the main factor influencing le.lCical decay in the TxG community. Unfortunately, the TGDP members have not been able to interview any semi·· fluent speakers in the New Braunfels area. This is a sit· uation typical of dying languages: "'Ihe 'native speaker' population itself may not agree on who falls within that category: some people may claim speaker status when others would not accept them as such; some may say they are not speakers when others would include them as speakers" (\Vatson 1989: 41).1his situation is even more pronounced in the New Braunfels area, as many lo<.-als believe that TxG is "only" a dialect, and not comparable in prestige to Standard German. Boas (2009) reports that he had difficulty finding informants in New Braunfels, as many potential informants refused to be interviewed (even though Boas had been introduced to them by other TxG speakers who he had already known for some time), claiming that they were Deu.tschverderber 'corrupters of German~ and do actually not speak any German at all Boas (2009) states that he has had numerous such encounters with speakers ofTxG during his fieldwork. However, on one occasion Boas interviewed a semi--fluent speaker in Doss (about 10 miles northwest of Fredericksburg) whose lexical knowledge appeared to be very limited. Although he appeared to be able to understand almost everything that Boas said in German, he had difficulties remembering German words. Having decided to end the interview because Boas thought that he had exhausted the informant's knowl· edge ofTxG, Boas happened to ask the informant about his upcoming hunting trip. As it turned out, the informant had been an enthusiastic hunter since his childhood, often going hunting with his father and brother, and they always spoke TxG while hunting. As a result, the informant could carry on an elaborate conversation about deer hunt· ing, including different weather conditions, types of terrain and bruns, finding good hunting spots, shooting and butchering deer, and making sausage. This example illustrates that lexical loss is difficult to measure, and that we should be careful not to conclude automatically that a semi-speaker has limited competence in the language. 'Ihis point is even more important when we analyze the lexical knowledge of Boas' (2009) New Braunfels area informants, who are technically fluent speakers ofTxG. They have no problem following and participating in German conversations covering a wide variety of topics, although occasionally they simply do not

146 Hans C. Boas & Marc Pierce

remember a particular word or phrase. This is even more pronounced in the elicitation tasks where they are asked to translate English words, phrases, and sentences into TxG. In this case, most TxG speakers are capable ofwhat Tsitsipis (1989: 135) describes as manipulating creatively formulaic and lexical material as well as producing contextually appropriate figurative speech. Although their communicative competence lacks certain aspects, their overall knowledge of the language as well as the social and cultural contexts allows them to compensate for it. Perhaps one of the most persistent problems encountered during fieldwork on TxG is that almost all informants are more than 60 years old, and have often not spo·· ken any TxG in years or even decades. Despite their infrequent use of TxG, they are typically able to carry on long open-ended conversations about a broad range of topics, including growing up, their ancestors, attending school, ranching and farming, experiences in the military, meeting their spouse, and their communities. None of the New Braunfels informants had serious difficulties translating the Gilbert and Eikel data into TxG, except for a few words and phrases. As a result, they either did not translate the word or phrase at all, or they substituted English words for it Boas (2009) reports that many consultants asked him to supply a particular word or phrase in German, and when he told them, they usually remembered it immediately and felt somewhat embarrassed that they forgot it. As one informant put it: "I just haven't spoken any German in such a long time so I forgot how to say junk in German~' Another New Braunfels speaker explained: "After my husband passed away in the late 1960s I stopped making sausage so I forgot almost all the words that go with it." These examples illustrate that there is little actual differenc.e between what Tsitsipis (1981: 117) calls "passive competence'' and "active competence~' The small number of cases where speakers initially did not know a word or phrase, but later on remembered it could thus be classified as "language recovery" in the sense of Hill (1979: 72) and Dressler (1981: 14). All New Braunfels informants exhibited fluent speech during open-ended interviews, being typically at ease with a wide range of topics. The only types of words that turned out to be problematic were from the semantic domains in which heavy borrowing from English was already described by earlier research, i.e. education, administration, nature, transportation, technical innovation, etc. That is, all New Braunfels informants continue to have an excellent passive knowledge of TxG, but their active knowledge sometimes lags when it comes to particular German words or phrases that they have either forgotten or that they never acquired in the first place (presumably because they acquired an English word or phrase instead of the TxG equivalent). In our view, this "momentary" lexical loss is not necessarily indicative of lexical erosion, but is instead a normal attrition caused by old age and/or not using the Ian·· guage, or certain "parts" ofthe language, for an extended period of time. As the example of the Doss informant illustrates, it is quite possible to exhibit a fairly significant reduction of one's lexical inventory, except for words from particular semantic domains such

Lexical developments in Texas German 147 as hunting, which have continued to be used throughout one's life. To be sure, we need to distinguish between the failure to acquire TxG completely and language attrition due to old age and/or disuse of the language. The written questionnaires reveal that all informants, not only the New Braunfels infi.)rmants, but also the informant from Doss, were raised speaking German at home and among family, neighbors, and friends. 'Ihis means that the type of lexical reduction is not caused by inadequate acquisition, but by other factors such as age, the time that the language has not been used, and the intensity of contact with other speakers. 19 As such, we suggest that the type of le.xicalloss exhibited by the New Braunfels informants is parallel to that of elderly monolingual speakers who have not used words from particular semantic domains (such as children's games or high school math terminology) for decades and thus have problems remembering them. In addition, a third of the New Bmuntels informants speak a little German only once a month, while another third speak it only once a year. For these informants, the opportunities to speak are typically limited to greetings at church or small talk with family, friends, or neighbors. 'Ihe New Braunfels speakers thus appear to have retained a passive knowl·· edge of basically all words, but have problems with remembering some of them. It is therefore important to keep in mind that in contrast to the one informant from Doss the lexical reduction among the New Braunfels informants is very minimal.

7·

Conclusions

This paper has shown that the lexicon of TxG has not undergone any dramatic changes over the past four decades. More specifically, we have demonstrated that the different dialectal origins of words that were still apparent when Gilbert collected his data in the 1960s can still be found in the present-day data collected by the TGDP. Based on Gilberfs (1972) data we argued that TxG has borrowed words primarily from particular semantic domains such as education, administration, telecommunication, transportation, ranching and farming, and nature, among others. In addition, a few prepositions and conjunctions were borrowed. As such, TxG can be classified as "Stage 2" according to 1homason and Kaufman's (1988) 5-stage borrowing scale ("slightly more intense contact"), which is characterized by lexical borrowing and slight structural borrowing in combination with conjunctions and adverbial particles. The comparison of Gilbert's data with the TGDP data shows that there have been comparatively few changes in the TxG lexicon over the past four decades. The small

19.

See de Bot and Clyne (1994) and Schmid (2002) on these points.

148 Hans C. Boas & Marc Pierce

changes observed in the present-day data do not appear to follow any systematic pattern, which has led us to argue that they are item-based. Moreover, we do not see the current changes to the TxG lexicon as signs oflexical erosion.

References Andersen, Roger W. 1988. Determining the linguistic attributes of language attrition. In The Loss of Language Skills, Richard D. Lanbert & Barbara F. Freed (eds), 83--118. Rowley MA: Newbury House. Augspurg-Hacker, Dagmar. 1984. Deutsche Auswanderung nach Texas im 19. Jahrhundert. Ph.D. dissertation, University of Bochum. Benjamin, Gilbert G. 1909[1974]. The Germans in Jexas. New York NY: Appleton & Co. Biesele, Rudolph L. 1928. The History of the German Settlements in Texas, 1831-1861. Austin, TX: Bockmann-Jones. Boas, Hans C. 2003. Tracing dialect death: The Texas German dialect project. In Proceedings of the 28th Meeting of the Berkeley Linguistics Society, Julie Larson & Mary Paster (eds), 387-398. Berkeley CA: University of California. Boas, Hans C. 2005a Texas German dialect. In Germany and the Americas. Thomas Adam (ed.), I 029-1035. Santa Barbara CA: ABC-CLIO. Boas, Hans C. 2005b. A dialect in search oflts place: 'Ihe use of'Iexas German in the public domain. In Transcontinental Encounters: Central Europe Meets the American Heartland, Craig Stephen Cravens & David John Zersen (eds), 78-102. Austin TX: Concordia University Press. Boa~, Hans C. 2006. From the field to the web: Implementing best -practice recommendations in documentary lingui.sti.cs. Language Resources and Evaluation 40(2): 153-174. Boa~, Hans C. 2009. 1he Life and Death of1e.xas German. Durham, NC: Duke University Press. Boa~. Hans C., Pierce, Marc, Roesch, Karen, Weilbacher, Hunter & Halder, Guido. 2010. The 'Iexas German dialect archive: A multimedia resource for research, teaching, and outreach. Journal of Germanic Linguistics 22(3): 277--296. Cook, Ew~.g-- Do. 1989. Is phonology going haywire in dying languages? Phonological variations in Chipewyan and Sarcee. Language in Socie~v 18: 235-255. Craig, Colette. 1997. Language contact and language degeneration. In The Handbook ofSocioiinguistics, Florian Coulmas (ed), 257-270. Oxford: Blackwell. de Bot, Kees & Clyne, Michael. 1994. A 16-yearlongitudinal study oflanguage attrition in Dutch immigrants in Australia }ournai of Multicultural and Multilingual Development 15: 17-28. Dorian, Nancy. 1973. Grammatical change in a dying dialect. Language 49: 413-438. Dressler, Wolfgang. 1981. Language shift and language death -A Protean challenge for the linguist Folia Linguistica XV: 5---28. Eichhoti, Ji.irgen 1986. Die deutsche Sprache in Amerlka. In Amerika und die Deutschen, Frank Trommler (ed), 235---252. Opladen: Westdeutscher Verlag. Eikel, Fred 1954. The New Braunf-els German dialect Baltimore, MD: Johns Hopkins University. Fuller, Janet, & Gilbert, Glenn. 2003. The Linguistic Atlas of Texas German revisited. In German Language Varieties Worldwide: Internal and External Per~pectives, Williarn Keel & Klaus Mattheier (eds), 165---176. Frankfurt: Peter Lar~.g.

Lexical developments in Texa~ German 149 Gilbert, Glenn. 1965. English loanwords in the German of Fredericksburg, Texas. .American Speech 40: 102-112. Gilbert, Glenn. 1972. Linguistic .Atlas ofTexa.~ German . .Austin, TX: University ofTexa~ Press. Haugen, Einar. 1950. The analysis oflinguistic borrowing. Lartguage 26:210-231. Hill, Jane. 1979. Language death, language contact, and language evoluti.on. ln .Approaches to Lan,'5uage: .Artthropological issues, \\'illiam C. McCormack & Stephen A. Wurm (eds), 47-78. 'Ihe Hague: Mouton. Holloway, Charles. 1997. Dialect Death. The Case of Brule Spanish [Studies in Bilingualism 13]. Amsterdam: John Benjamins. Jordan, Terry. 1975. German Seed in Texas Soil. Austin, TX: University of Texas Press. Jordan, Terry. 1977. The German element in Texas. Rice Universifcv Studies 63(3): 1--- 11. Kloss, Heinz 1977. Die deutsche Schriftsprache bei denAmischen. In Deutsch als Muttersprache in Katmda: Berichte zur Geget1wartslage, Leopold Auberger, Heinz Kloss, & H. Rupp (eds), 97--98. W1esbaden: Steiner. Meister, George J. 1969. English lexical influence on Texas German. MA thesis, The Unlversity of Texas at Austin. Mithun, Marianne. 1989. The incipient obsolescence of polysynthesis: Gayuga. In Investigating Obsolescetlce: Studies in Language Contraction and Death, Nancy Dorian (ed), 243--258. Cambridge: CUP. Moore, Barbara J.R. 1980. A Sociolinguistic Longitudinal Study (1969--1979) of a Texas German Communlty, Including Curricular Recommendations. Ph.D. dissertation, The University of Texas at Austin. Muysken, Pieter. 1981. Halfway between Quechua and Spanish: The case for relexification. In Historicity and Variation itl Creole Studies, Arnold Highfield & Albert Valdman (eds), 52--78. Ann Arbor MI: Karoma. Nicolini, t...1arcus. 2004. Deutsch in Texas. Munster: LIT Verlag. Pop lack, Shana & Meechan, Marjory. 1998. Introduction: How languages fit together in codemixing. International journal of1Jiling<.lalism 2(2): 127-138. Poplack, Shana, Sankoff, David & Miller, Christopher 1988. The social correlates and linguistic processes oflexical borrowing and assimilation. Linguistics 26: 47-104. Roesch, Karen A. 2009. Texas Alsatian: Henri Castro's Legacy. Ph.D. dissertation, University of 'Iexas at Austin. Ruoff, Arno. 1973. Grundlagen und Methode1t der Untersuchung gesprochener Sprache. Ti.ibingen: Niemeyer. Salmons, Joe. 1983. Issues in Texas German language maintenance and shift. A1onatshefte 75: 187-196.

Salmons, Joe, & Lucht, Felicia. 2006. Standard German in Texas. In Studies in Contact .Unguistics. Essays in Honor of Glenn G. Gilbert. Linda Thornburg & Janet Fuller (eds), 165-186. Frankfurt: Peter La.ng. Sasse, Hans J. 1992. Language decay and contact-induced similarities and differences. In Lang-uage Death, Matthias Brenzinger (ed.), 59-80. Berlin: Mouton de Gruyter. Schmid, Monika. 2002. First Language Attritim~, U5e and Malnte1wnce. 7he Case ofGerman-jews in Anglophone Countries [Studies in Bilingualism 24]. Amsterdam: John Benjami.ns. Thomason, Sarah. 2001. Language Contact: An Introduction. Washington, D.C.: Georgetown University Press. Thomason, Sarah. 2003. Contact as a source oflanguage change. In The Handbook of Historical Linguistics, Brian D. Joseph & Richard D. Janda (eds), 687---712. Oxford: Blackwell.

150

Hans C. Boas & Marc Pierce Thomason, Sarah & Terrence Kaufman. 1988. Language Contact, Creolization, and Genetic Linguistics. Berkeley, CA: University of California Press. Treffers-Daller, Jeanine. 1994. Mixing T~·o Languages: French-Dutch Contact in a Comparative Perspective. Berlin: Mouton de Gruyter. Tsltslpis, Lukas. 1981. Arvanltika language change in speech communities in Greece. In f'oiia Slavic.a: Studies in Balkan Linguistic.~ to Honor E.P. Hamp on his Sixtieth Birthday, Howard Aronson & Blll J. Darden (eds), 370-383. Columbu.~. OH: Slavl.ca. Tsitsipis, Lukas. 1989. Skewed performance and fi.tll performance in language obsolescence: The case of an Albanian variety. In Investigating Obsolescence: Studies in Language Contractiotl and Death, Nancy Dorian (ed), 117···138.Cambridge: CUP. Watson, Seosamh. 1989. Scottish Irish Gaelic: The giant's bedteJJ.ows. In Investigating Obsolescence: Studies in Lat1guage Comraction and Death, Nancy Dorian (ed.), 41···60. Cambridge: CUP. Whitney, William Dwight. 1881. On mixture in language. Transactions of the American Philosophical Association 12: 5-26. Wilson, Joseph. 1977. The German Language in Central Texas Today. Rice University Studies 63(3): 47-58. Wmford, Donald. 2003. An Introduction to Contact Linguistics. Oxford: Blackwell. Wohlgemuth, Jan 2009. A Typology of Verbal Borrowings. Berlin: Mouton de Gruyter. Wolfram, Walt. 2002. Language death and dying. In The Handbook of Language Variatlotl and Change, Jack K. Chambers, Peter Trudgill & Natalie Schilling-Estes (eds.), 764-787. Oxford: Blackwell Woolard, Kathryn A 1989. Language change and language death as social processes. In Inves-tigating Obsolescence: Studies In Language Contraction and Death, Nancy Dorian (ed.), 355--367. Cambridge: CUP.

Gender assignment of English loanwords in Pennsylvania German Is there a fetninine tendency?* B. Richard Page Pennsylvania State University

Previous investigations have reported a feminine tendency in the assignment of grammatical gender for English loanwords in Pennsylvania German as well as in other German Sprachinseln languages located in Anglophone areas (e.g. Reed 1942). This study investigates whether or not there is a feminine tendency in Pennsylvania German using a corpus based on the Comprehensive Pennsylvania German Dictionary. A statistical analysis reveals no evidence of a feminine tendency in Pennsylvania German. Monomorphemic English loanwords and monomorphemic native nouns in Pennsylvania German are significantly more likely to be assigned masculine gender than monomorphemic nouns in Standard German. Unlike Standard German, there is no significant difference in gender assignment in Pennsylvania German between monosyllabic and polysyllabic now1s for either the native lexicon or English loanword<>.

1.

Introduction

1he assignment of grammatical gender in German is notoriously complex. Until relatively recently, most research considered gender assignment in German to be largely arbitrary as famously asserted by Bloomfield's statement that "[t]here seems to be no practical criterion by which the gender of a noun in German, French, or Latin could be determined" (Bloomfield 1933: 280; see also Maratskos 1982). German has three grammatical genders: masculine, neuter and feminine. In many instances, there seems to be no discernible rule that accounts for gender assignment. Consider, for example, the nouns Lojfel 'spoon (masc.): A1esser'knife (neut.)' and Gabel'fork (fern.)'.

*I would like to thank Marc Pierce, Mike Putnam and the patticipant'i of GLAC 14 for their comments and s-uggestions. Most especially I thank Andrea Lohr for her help in compiling the corpus based on 'lhe Comprehensive Permsylvar1ia Germar1 Dictionary.

152.

B. Richard Page

Despite their semantic and phonological similarities, each is assigned a different grammatical gender. Due in large part to the work ofKopcke and Zubin (e.g. Kopcke 1982; Zubin & Kopcke 1986), the generally accepted view today is that gender assignment in German is not arbitrary but is governed by phonological, morphological and semantic regularities, most of which are probabilistic in nature. For example, Kopcke (1982) finds a general tendency tor monosyllables to be masculine as well as subregularities based on the phonological shape of the syllable. An example of a subregularity is a tendency for monosyllables ending in a sequence of non·· sibilant fricative plus /t/, such as Luft 'air (fern.): to be feminine. Of the 55 monosyllabic German nouns ending in such a sequence, Kopcke finds that 12 are masculine, 7 neuter and 36 feminine. Kopcke shows that the gender of most of the exceptions is consistent with another gender assignment rule motivated by semantics or morphology. Thus Knecht 'servant, laborer' is masculine because of a semantic rule assigning masculine gender to nouns referring to occupations without signifying natural gender (Kopcke 1982: 72, 97-98). Steinmetz (1986) finds that apparent exceptions to the regularities identified by Kopcke and Zubin can be accounted for by simply counting the number of rules that apply to a given noun. lf one rule indicates that a word should be masculine but two rules indicate that the same noun should be feminine, the noun will be feminine. Work by Steinmetz and others also assumes that there is a default gender, or gender hierarchy, that plays a role in gender assignment in German and other Germanic languages. For German, masculine is the most preferred gender and neuter is the least preferred (Steinmetz 1986, 2006). Of c.ourse, grammatical gender must also be assigned to nouns that are borrowed into German. In an analysis of a corpus based on the German news magazine Der Spiegel, Onysko (2007: 178) finds that the assignment of gender for English loanwords operates in essentially in the same manner as the assignment of gender for native lexical nouns: The interaction of a default hierarchy of gender assignment and various semantic, morphological, and phonological rules is able to account for the vast majority of masculine, feminine and neuter anglicisms .... [A] large portion of the masculine anglicisms received gender by default as no specific gender associations apply."

In light of Onysko's findings, it is surprising that previous investigations have reported a "feminine tendency" in the assignment of grammatical gender to English loanwords of German language enclaves (Sprachinseln) in the United States (Aron 1930), Canada (Wacker 1964: 114) and Australia (Clyne 1968, 1969, 2003). Haldeman (1872: 24-25), Buffington (1941) and Reed (1942) agree that there is a feminine tendency for English loanwords in Pennsylvania German as well. In contrc:lst, there is a masculine tendency in the assignment of grammatical gender to English loanwords in Standard German, at least for monosyllables (see Carstensen 1980; Onysko 2007). Thus, masculine is

Gender assignment of English loanword~ in Pennsylvania German 153

widely-viewed as the default gender for native lexical items and English borrowings in Standard German, whereas previous research indicates that feminine is the default gender for English borrowings in German linguistic enclaves in Anglophone areas. This study investigates whether or not there is indeed a feminine tendency in Pennsylvania German using a corpus based on the Comprehensive Pennsyivania German Dictionary. In the following section, there is an overview of previous research on gender assignment of English loanwords in Pennsylvania German. Section three describes the corpus and presents the findings of this study. Section four concludes the paper.

2.

Previous research on gender assignment of English loanwords in PG

The claim of a feminine tendency in the assignment of gender to English loanwords is quite old. Haldeman (1872: 24) states that speakers of Pennsylvania German (PG) are likely to use the feminine gender for English loanwords because they use the same pronounciation for the English article the [di:] and the PG feminine and plural article die [di:]. As evidence, Haldeman lists 30 English loanwords, all of which are feminine. Two of the most prominent researchers of PG in the twentieth century also find a tendency fi.)r English loanwords in PG to be feminine. Buffington (1941) compiled a 30,000-word corpus taken from the prose of three nonsectarian Pennsylvania German writers: Pumpernickle Bill (pseudonym for vVilliam Troxell), Solly Hulsbuck (pseudonym for I·farvey Miller), and Boonastiel (pseudonym tor Thomas Harter). In Buffington's corpus, over half of the English loan words that are lexical nouns are feminine. Like Haldeman, Buffington attributes the feminine tendency to the phonetic similarity of"PG die and English the. The most well-know study of gender of English loanwords in Pennsylvania is Reed 1942. Reed identifies four categories, listed in ( l ), that play a role in gender assignment. ( l)

Categories of gender assignment for EnglL'lh loanwords in PG (Reed 1942: 25-26) a. b. c. d.

Nouns that have taken over the gender of the (Pa.) German nouns which they displace. Nouns having a type of suffu that normally characterizes a particular gender in German. Nouns whose gender is determined by the sex of a 1iving being'. Nouns that have been given the feminine gender, because the English definite article [CJi:/oJ] resembles phonetically the German feminine definite article [di:/dJ ].

'Ihe nouns listed in (2) illustrate that it is not always easy to identify the noun that is displaced. For example, Reed attributes the variation between neuter and masculine

154 B. Richard Page

gender for the loanword Bocki to the difference in gender between der Spazierwagen and das Wiiglein. The second difficulty is ascertaining the gender of the noun that is replaced. As illustrated by Standard German Gerste 'barley (fern.)~ dialectal Gersten 'barley (masc.)~ PG Ga1"Scht or Gat"Schde 'barley (masc.), it is entirely possible that the noun that is displaced has a different gender in PG than in other varieties of German. Thus, if one finds an English loanword with an unexpected gender based on the gender of its presumed PG equivalent in Standard German, there are two possible sources tor the discrepancy: either (1) the displaced noun has not been correctly identified or (2) the displaced noun has a different gender in PG than in other varieties of German. It may be simply impossible to rule out these possibilities for unexpected gender assignments. (2)

Norms that have taken over the gender of the German nouns which they displace (Reed 1942: 26-27)

Engl. Loan. Haendel (m) Beiler (m) Floor (m) Barli (m) Boggi(m) Schpring (f) Grab (f) Pickter (n) Bocki (n)

'handle' 'boiler' 'tloor' 'barley' 'buggy' 'source of a creek' 'crop, harvest' 'picture' 'buggy'

Pa. German

Std. German

Schtiel (m) Kessel (m) Bodem(m) Garschd (m)

Stiel (m) Kessel (m) Boden (m) dial. Gersten (m) dial. Spazierwagen (m) QueUe (f) Ernte (t) Bild (n) dial. Waglein (n)

Sd~.padsierwagge

Gwell (f) Aern (f) Bild (n) Weggli (n)

(m)

As examples of nouns that are assigned gender on the basis of their suffix, Reed's examples include both derivational morphemes and non--morphemes that share the same phonological shape. 1hus, polysyllabic English loans ending in -er in PG tend to be masculine regardless of whether the -er ending constitutes a morpheme: Meschgider 'mosquito (masc.)~ Kaunder 'counter (masc. )' and Beinder 'binder (mas c.)'. The Pa. German derivational morpheme -ingis cognate to Standard German-ung, and polysyllabic nouns ending in-ing are feminine: meeting 'meeting (fern.)'. Polysyllabic English loanwords ending in -ie/-y [-i:) also tend to be feminine: Bocki 'buggy (fern.): Schtory 'story (fern.): Paerdy 'party (fern.): The third category comprises English loanwords that are assigned gender on the basis of sex. Reed (1942: 28) lists examples such as Paep 'father (masc.); Bom 'bum (masc.): and Draemb 'tramp (masc.)'. In the fourth category, Reed has placed English loan words that are assigned feminine gender but the assignment of gender cannot be accounted for by the first three categories. For example, PG Fens 'fence' is feminine, but the noun that it presumably displaces, Ger. Zaun, is masculine. Fens does not refer to a living being, and it does not have a suffix that is associated with a particular grammatical gender. Reed argues that

Gender assignment of English loanword~ in Pennsylvania German 155

the gender of English loan words like PG J<'ens that don't fall into one of the tl.rst three categories tend to be feminine. He asserts that the basis for this feminine tendency is the phonetic similarity of English the [cH/dd] and the PG feminine definite article die [dilda]. Reed lists the examples ofthe feminine tendency in (3). (3)

English loans that are feminine by default

Maetsch Fens Rann, Ihum Grick Lehg

'match' 'fence' 'freshet, brook' 'creek'

Dieboh Schprie Blaus Flub

1ake'

Suhd

(Reed 1942: 28)

'depot' 'spree' 'blouse' 'Hue' 'suit' (clothes)

Springer (1980) points out that feminine gender for PG B/aus may be attributable to Biuse 'blouse (0: which is found in Palatine dialects. Similarly, the gender of PG Grick is not surprising when one considers that Ger. Bach is feminine in the Palatinate and elsewhere (Springer 1980: 200). PG Rann, Runn 'freshet, stream (fern.)' (cf. English nm) could also be attributable to Bach being assigned feminine gender in Palatine German. In addition, Lambert (1977) points out that PG Run could also be the reflex of Ger. Runs 'run of water (masc.): Ger. Runs also has a feminine variant Runse (Duden 1996). As Springer (1980) argues, the existence of a feminine tendency in PG is questionable. Efforts to identify a default gender for loanwords often take a statistical approach. In corpus studies of gender assignment tor English loanwords in German, Carstensen (1980), Fink (1968), Glahn (2002), and Yang (1_990) all find a preponderance of masculine nouns whereas Viereck (1.986) finds a tendency for English loanwords to be feminine. As Gregor (1983: 62) points out, the finding of a tendency to assign a particular gender to loanwords may have more to do with the underlying corpus than with a default gender. Indeed Pop lack, Pousada & Sankotf ( 1982: 21---3) doubt whether default gender plays a role in gender assignment for loanwords. Corbett (1991: 78) notes that the vast majority of German loanwords in Russian are masculine. However, there is no need to view this pattern as evidence that masculine is the default gender in Russian. A large percentage of the German loanwords end in a consonant, and Russian gender assignment rules assign masculine to nouns ending in a consonant (Corbett 1991: 78). Thus, a bias toward a particular gender in the lexicon may simply reflect the outcomes of gender assignment rules based on form and semantics rather than indicate a detault gender or gender hierarchy. Nevertheless, studies of gender assignment in Germanic languages have demonstrated the usefulness of a gender hierarchy or default gender (e.g. Steinmetz 1986; Rice 2006; Onysko 2007). Moreover, there is evidence that the gender hierarchy for a language can change over time. Thus, Steinmetz (2006) argues that neuter was originally the preferred gender in Germanic, but the preferred gender shifted to masculine

156 B. Richard Page

in the West Germanic languages. If one accepts Steinmetz's argument that masculine is the default gender for West Germanic languages, reports of a feminine tendency in Pennsylvania German and other German language islands would appear to be evidence that a gender shift has taken place, perhaps due to contact with English.

3·

Statistical analysis of gender assignment

Establishing the existence of a feminine tendency is a challenge. Reed's approach of categorizing the gender of loanwords on the basis of semantic analogy, form and natural gender is problematic. One must correctly identify the PG noun that is dis·· placed and its gender in order to determine whether a loanword is feminine because of analogy or because of a feminine tendency. As illustrated above, identifying the noun displaced and its gender is not a trivial matter. Statistical approaches are also problematic. The results of a statistical count may simply reflect the outcome of gender assignment rules in PG and not the existence of a feminine default gender. In order to show that there is a default gender, it is necessary to have a large sample size and to control for morphological and phonological shape. TI1is study compares the distribution of grammatical gender for Englsh loanwords in PG with the distribution of grammatical gender of native vocabulary in PG and in Standard German. The null assumption is that gender distribution would be the same for English loanwords and native vocabulary once there is a control for morphological and phonological shape. If there is a feminine tendency, one would expect teminine grammatical gender to be more prevalent among loanwords than native vocabulary in the absence of alternative explanations. The primary corpus used in the analysis is the Comprehensive German Dictionary. The dictionary draws from oral, literary and lexicographical sources. The dictionary also includes English loanwords, though Beam excludes those from the late twentieth century. By our count, the dictionary's eleven volumes contain entries for over 16,000 nouns including over 1000 nouns that are loanwords from English. In analyzing the corpus, forms with multiple genders were counted more than once. Thus, der Bocki, die Bocki and das Bocki 'buggy' were counted as one example each of masculine, feminine and neuter gender assignment On the other hand, a form with multiple spellings but a single gender would only be counted once. For example, the PG word der Buh/Bu/ Bub!Bup 'boy: which is a native PG noun, would count as one occurrence of masculine gender assignment in the native vocabulary. 3.1

Gender assignment of monosyllabic English loanwords

Previous research on Standard German shows a strong tendency for monosyllabic English loanwords to be masculine (Carstensen 1980). Therefore, it is surprising that

Gender assignment of English loanwords in Pennsylvania German 157

nine of the ten loanwords that Reed (1942) listed as examples of the feminine tendency are monosyllabic. Using a x2 goodness-of-fit test, the distribution of gender among monosyllabic English loans was tested against the distribution of gender among monosyllabic nouns in the native PG lexicon (see Schiller & Carama:tza 2003). The results are presented in (4) and (5). (4)

Distribution of gender among monosyllables in native PG lexicon.

Total Inanimate (5)

Mas c.

Fem.

Neuter

537 (.4324) 459 (.4161)

516 (.4155) 467 (.4234)

189 (.1522) 177 (.1605)

Distribution of gender among monosyllabic English loanwords in PG.

Total Inanimate

Mas c.

Fem.

Neuter

x2(z)

86 (.4866) 75 (.4658)

70 (.3977) 66 (.4099)

20 (.1136) 20 (.1242)

3.14 2.35

p = .21 p = .31

If there is a feminine tendency as outlined by Reed ( 1942 ), one would expect a greater percentage of feminine monosyllabic loanwords than feminine monosyllabic native nouns. As shown in (4) and (5), that is not the case. In fact, there seems to be a slight tendency toward masculine gender in monosyllabic loanwords. However, the differences in gender distribution are not significant. The next comparison tested the distribution of gender of polysyllabic English loanwords against polysyllabic native nouns using a x2 goodness-of-fit test. In both cases, there is a higher percentage of masculine nouns than among monosyllables. As is the case with monosyllables, the difference in gender distribution between English loanwords and the native lexicon is not significant for polysyllabic nouns (x2(2) = 2.49, p = .29).1 The results are displayed in (6). ( 6)

Distribution of gender of polysyllabic English loanwords tested against PG native lexicon.

Polysyllables

Masc

Fern

Neut

Tot.

x2(2)

Native English loans

427 (.5531) 77 (.5203)

248 (.3212) 56 (.3784)

97 (.1256) 15 (.1014)

772 148

2.49

p= .29

In sum, there is no significant variation in the distribution of grammatical gender among English loanwords in comparison to native PG nouns. For monosyllabic and

The parenthetical nurnber in "X2 (2)" indicates the degrees of freedom for the X2 goodness of fit test.

1.

158 B. Richard Page

polysyllabic loanwords, masculine is the most common gender. Polysyllabic English loanwords are more likely to be feminine than native polysyllabic nouns, but the difference is not significant. 3.2

Gender distribution in PG tested against Standard German

The findings in Section 3.1lead to further exploration of the question of gender shift. If Haldeman, Reed and Buffington are correct about the feminine tendency, it is possible that feminine has already established itself as the default gender for native and nonnative nouns in PG. If true, the feminine tendency should be evident in a comparison of Standard German nouns with English loanwords in PG. For evidence of a gender shift in PG, gender distribution in native nouns and English loanwords in PG was tested against the distribution of gender in monomorphemic German nouns using a 2 goodness-of-fit test as shown in (7).

x

(7)

Distribution of gender ofPG monomorphemic nouns tested against German (Baayen et al. 1995) monomorphemic nouns based on Celex database.

Category German

PG Native PG Eng. loan

Masculine

1758 (.4274) 964 (.4786) 163 (.5031)

Feminine

1567 (.3810) 764 (.3793) 126 (.3889)

Neuter

788 (.1916) 286 (.1420) 35 (.1080)

Total x2(2)

x2(1)

x2(1)

masc

neut.

4164 2014 324

38.24" 21.61 * 16.21 * 7.58**

31.98* 14.61 *

*p < . 001, **p < .01

'Ihe x2 goodneSS··Of.·fit tests reveals that the distributions of grammatical gender are significantly different between PG native nouns and German (p < .001). The distribution of grammatical gender was also significant different between PG English loanwords and German (p < .001). Therefore, the proportion of masculine and neuter nouns in PG were both tested individually against their respective proportions in Standard German (columns "x2 (1) masc:' and "x2 (1) neut"). For both native nouns and English loans, PG has a significantly higher proportion of masculine monomorphemic masculine nouns than Standard German (p < .001 for PG native nouns, p < .01 for PG English loanwords). In addition, there is a significantly lower incidence of neuter in PG than in Standard German (p < .001). 3·3

Phonological shape and gender assignment

Phonological shape plays a major role in gender assignment in Standard German. Kc.ipcke ( 1982) finds that over 64% of monosyllabic German nouns are masculine. In contrast. over90% of nouns ending in --e [a] are feminine (Durrell2002: 9).ltis therefore not surprising that previous investigations of English loanwords in German indicate

Gender assignment of English loanword~ in Pennsylvania German 159

a strong masculine tendency for monosyllables (e.g. Carstensen 1980; Onysko 2007). Pennsylvania German has undergone widespread apocope, as illustrated in (8), which has greatly increased the number of feminine monosyllables in PG. This raises the question of whether phonological change has resulted in different gender assignment patterns for English loanwords in Pennsylvania German than in Standard German. (8)

Apocope in PG

P.G.

.St~I...G.~n11~n

WQ.~

Gwell Harf Frog

QueUe

'source, spring (f.)' 'harp (f.)' 'question (f.)'

Harte Frage

A comparison of monosyllabic Standard German nouns to monosyllabic English loanwords and native nouns in Pennsylvania German is given in (9). The figures tor Standard German are taken from Kopcke (1982), who used the 1967 Leipzig edition of Der grofte Duden, a standard reference dictionary, for his corpus. Distribution of gender among monosyllabic nouns in German and PG was tested against Standard German using a x2 goodness-of-fit test. Monosyllabic nouns in PG are significantly more likely to be feminine in PG than in German (p < .0001). This is true for native nouns and for English loanwords in PG. (9)

Distribution of gender in PG monosyllables tested against Standard German using a£ goodness-of-f1t test.

Category

Masculine

Feminine

Neuter

x2(2)

x2(l)fem.

German Native nouns in PG English loans in PG

940 (.6412) 537 (.4324) 86 (.4866)

205 (.1398) 516 (.4155) 70 (.3977)

321 (.2190) 189 (.1522) 20 (.1136)

784.9"" 99.1,.

784.8* 97.4*

*p < .0001

The effect of syllable count on the assignment of gender to English loanwords in PG was also tested using the x2 test of independence. Previous research suggests that monosyllabic English loanwords are more likely to be masculine than are polysyllabic loanwords (Carstensen 1980). The gender distribution for monosyllabic and polysyllabic English loans are displayed in (10). (1 0)

Gender among monosyllabic and polysyllabic monomorphemic English loanwords in PG.

Syllable Structure Monosyll. Polysyll.

Mas c.

Fern

Neut

Tot.

86 (.4866) 77 (.5203)

70 (.3977) 56 (.3784)

20 (.1136) 15(.1014)

176 148

t6o B. Richard Page

x

The 2 test of independence shows that the differences in gender distribution in (10) are not significant (x2(2) = .3496, p = .84). Unlike Standard German, syllable structure appears to play no discernible role in gender assignment of English loanwords in PG. Thus, apocope has indeed resulted in gender assignment patterns for .English loanwords in PG that differ significantly from those in Standard German.

4·

Di.'icussion and conclusion

An examination of the Comprehensive Pennsylvania German Dictionary reveals no statistical evidence for a feminine tendency in the assignment of gender of English loanwords in Pennsylvania German. 'When gender assignment of English loanwords is tested against gender distribution in the native PG lexicon, there is no significant difference in the assignment of gender for either monosyllables or for polysyllables. However, when compared to monomorphemic nouns in Standard German, monomorphemic English loanwords in PG are significantly more likely to be masculine. Monomorphemic native PG nouns are also significantly more likely to be masculine than monomorphemic nouns in Standard German. The masculine tendency in PG may be attributable to the corpus. An in-depth study ofmonomorphemic PG nouns and their German cognates should reveal whether there are numerous examples of gender shift. in PG. The data reported above suggest the possibility that a number of neuter and feminine nouns in German may have masculine cognates in PG and perhaps in other dialects as well. One example noted by Reed is Ger. Gerste 'barley (fern.): PG Garschd 'barley (mas c.): dialectal Gersten 'barley (masc.)'. The results in (7) show that that monomorphemic nouns in Pennsylvania German are significantly more likely to be masculine and significantly less likely to be neuter than in Standard German. It is unclear whether this difference is attributable to a default gender and a resulting shift of nouns from masculine to neuter in the history of Pennsylvania German and its source dialects. 'Ihis is a matter for further study. Though PG and Standard German share a masculine tendency, it is clear that their gender assignment rules differ. Testing the distribution of gender for monosyllables in PG against Standard German reveals a significant increase in the assignment of feminine gender to monosyllables in PG (see 9). Syllable structure does not affect the assignment of gender in English loanwords in PG. In this regard, PG differs from Standard German. In Standard German, there is a strong preference to assign masculine gender to monosyllabic English loanwords (Carstensen 1980). This difference in gender assignment between Standard German and PG is attributable to apocope in PG, which greatly increased the number of monosyllabic feminine nouns. Thus, the only differenc.es in gender assigmnent patterns between PG and Standard German ascertainable in this study are due solely to internal developments in PG with parallels

Gender assignment of English loanwords in Pennsylvania German

in Palatine German and other dialects spoken in Germany, Switzerland and Austria. A statistical analysis reveals no evidence of a feminine tendency in gender assignment of English loanwords in Pennsylvania German.

References Aron, Albert W. 1930. The gender of English loan-words in Colloquial American German. Language6: 11--48. Baayen, R. Harald, Piepenbrotk, Richard & Gulikers, Leo. 1995. The CELEX lexical database (CD- ROM). Philadelphia PA: Linguistic Data Consortium, University ofPensylvania Bloomfield, Leonard. 1933. Language. New York NY: Henry Holt. Buflington, Albe11 F. 1941. English loan words in Pennsylvania GermaiL In Studies in Honor oflohn Albrecht Walz, Fred 0. Nolte, Harry W. Pfi.md & George J. Metcalf (ed.s), 66---85. Lant<J.ster PA: Lancaster Press. Carstensen, Broder. 1980. The gender of English loan-words in GermarL Studia Ar1glica Posnaniensia 12: 3-- 25. Clyne, Michael G. 1968. Deutscher Idiolekt und deutscher Dialekt in einer zweisprachigen Siedlung in AustraJien. Wirkendes Wi.Jrt 18: 84-95. Clyne, Michael G. 1969. Inha]t, Klangassoziation und Genus in der deutschen Sprache bei Ein-und Zweisprachigen. Zeitschrift fiir Photletik, Sprachwisseuschaft ;md Kommunikationsforschung 22: 218-224. Clyne, Mi.chael. 2003. Dynamics of Language Contact. Cambridge: CUP. Comprehensive Pennsylvania German Dictionary. 2004-2007. C. Richard Beam, Joshua R. Brown, Jennifer L. Trout & Dorothy Pozniko Beam (eds), 11 Vols. Millersville PA: Center for Pennsylvania German Studies. Corbett, Greville. 1991. Ge~tder. Cambridge: CUP. Durrell, Martin. 2002. Hammer:S Germar1 Grammar tmd Usage. 4th edn. Chicago IL: McGraw-Hill. Duder1 Deut.sches Ur1iversalworterbuch, 3rd edn. 1996. Wissenschaftlichem Rat und den Mitarbeitern der Dudenradakti.on (eds). Mannheim: Duden. Fink, Hermann. 1968. Amerikanismen i.m Wortschatz der deutschen Tagespresse, dargestellt am Beispiel drei.er tiberregionaler Zeitungen. Ph.D. dissertation, University of Mainz. Glahn, Richard. 2002. Der Bitifluss des I:inglischen auf gesprochene deutsche Gegenwartsspradte. Frankfurt: Peter Lang. Gregor, Bernd. 1983. Genuszuordrmng: Das Genus englischer Lehnworter im Deut.sdzen. [Linguistische Arbeiten 129). T'libingen: Max Niemeyer. Haldeman, SamuelS. 1872. Pennsylvania Dutch: A dialect of South German with an Infusion of English. London: Trtibner & Co. Kopcke, Klaus- Michael. 1982. Untersuchungen zum Genussystem der deutschen Gegenwartssprache [Linguistische Arbeiten 122]. Ti.ibingen: Max: Niemeyer. Lambert, Marcus Bachmann. 1977. Pennsylvania-German Dictionary. Originally published 1924 under title: A Dictionary of" the Non-English Words ofthe Pennsylvania-German Dialect [Publications of the Pennsylvania German Society 30]. Exton PA: Schiffer. Maratsos, Michael. 1982. The child's construction of grammatical categories. In Language Acquisition: The State of the Al-l, Eric Wanner & Lila R. Gleitmann, 240-266. Cambridge: CUP.

161

162.

B. Richard Page

Onysko, Alexander. 2007. Anglicisms in German: Borrowing, Lexical Productivity, and Written Codeswitching [Linguistik-Impulse & Tendenzen 23]. Berlin: Walter de Gruyter. Poplack, Shana, Pousada, .Alicia & Sankoff, David. 1982. Competing influences on gender assignment: Variable process, stable outcome. Lingua 57: 1-28. Reed, Carroll E. 1942. The gender of English loan words in Pennsylvania German. Ameri;:an Speech 17: 25-29. Rice, Curt. 2006. Optimizing gender. Lingua 116: 1394-1417. Schiller, Niels 0. & Caramazza, Alfonzo. 2003. Grammatical feature selection in noun phrase production: Evidence from German and Dutch. Journal of Memory and Language ·18: 169--194.

Springer, Otto. 1980. The study of the English of the Pennsylvania Germans. In Perspectives Otl American Engli..~, Joey Lee Dillard (ed.), 195---204. The Hague: Mouton. Steinmetz. Donald, 1986. 1\vo principles and some rules for gender in German: Inanimate nouns. Word 37: 189--217. Steinmetz. Donald. 2006. Gender shifts in Germanic and Slavic semantic motivation for neuter? Lingua 116: 1418-1440. Viereck, Karin. 1986. The influence of English on Austrian German. In English in Contact with Other Languages, Wolfgang Viereck & Wolf-Dietrich Bald, 159-178. Budapest: Akaderniai Kiad6. Wacker, Helga. 1964. Die Besonderheiten der deutschen Schriftsprache in den USA. [Duden Beitriige 14). Mannheim: Bibliographisches Institut. Yang, Wenliang. 1990. Anglizismen lm Deutschen: Am Beispiel des Nachrichtenmagazlnes Der Spiegel [Reihe germanistische Linguistlk 106]. Tlibingen: Max Niemeyer. Zubin. David & Kopcke, Klaus-· Michael 1986. Gender and folk taxonomy: The indexical rela-· tion between grammatical and lexical categorization. In Noun Classes and Categorization: Proceedings of a Symposium on Categorization and Noun Classification, Eugene, Oregon, October 1983, C. Craig (ed.), 139-180. Amsterdam: John Benjamins.

SECTION

3

Syntax I - Verb clusters

Synchrony and diachrony of verb dusters in Pennsylvania Dutch* Mark L. Louden University ofWisconsin-Madison

This paper presents the major synchronic facts about verb dusters in modern Pennsylvania Dutch (Pennsylvania German) and indicates how they have developed historically. Although Pennsylvania Dutch is descended from primarily Palatine German dialects, the behavior of verb dusters in the modern language is distinct from what is found in Ew·opean German dialects. Focusing on three- and four-verb dusters in subordinate clauses, it is observed that Pennsylvania Dutch observes a strict rule whereby a maximum of one lexical verb may occur within a clause; additional lexical verbs are located to the right of the clause periphery. The analysis presumes that non--lexical verbs in verb dusters, specifically the finite auxiliary hawwe 'have' and a non--finite modal, form a single syntactic unit.

1.

Introduction

One of the most vexing problems of the syntax of modern Continental "\Vest Germanic (Dutch, Frisian, German, and Luxembourgish) dialects is the analysis of what are known as verb clusters, that is, structures consisting of a finite verb with at least one,

*I presented papers dealing with much of the substance of this article at the Germanic Linguistics Annual Conference---13 at Penn State University in 2007 and at the Philipps-Universitiit Marburg in 2008. I am grateful to both audiences for helpful feedback. The present paper has also benefited significantly from discussions with Shannon Dubenion-Smith, Jtirg Fleischer, and G5z Kaufmann. I also appreciate greatly the suggestions for changes from an anonymous reader, nearly all of which were incorporated into the final version of this paper. Finally, thanks are due to series editor Werner Abraham for his thoughtful and useful comments. Any errors of analysis, however, are mine alone.

166 Mark L. Louden

but often two or three additional nonfinite forms. 1 An example of a three-verb cluster from standard German is given in (1). (I)

Susanne weiH,

dass er nicht hat l Susanne knows that he not has 'Susanne knows that he did not want to

gehen wollen 3 2 want go go'

The major challenge in the analysis of verb clusters is accounting f{)r the linear order of the verb forms: although all dialects of Dutch, Frisian, German, and Luxembourgish, as well as genetically related extraterritorial varieties such as Afrikaans, Mennonite Low German (Plautdietsch), Hutterite German, and Pennsylvania Dutch, are underlyingly verb--final (OV) languages, the surface order of elements in verb clusters typi·· cally deviates fi·om what we assume the underlying structure to look like. That is, in subordinate clauses, while the underlying order of verbal elements proceeds from most to least embedded (i.e. in numerically descending order), tb.is is not what we find in most varieties at the surface. Taking the example in (1), a three--verb cluster in which the second element is an infinitive where we would normally e.xpect a past participle ("lnfinitivus pro Participio," IPP), all six logically possible orderings of the three elements in this structure are attested in Continental \Vest Germanic varieties. 2 1he main purpose of this paper is a descriptive one, namely to present basic facts about two-, three-, and four-element verb clusters in Pennsylvania Dutch, a North American minority language descended from mainly Palatine German dialects

1. The secondary literatlue on verb dusters in Germanic and other languages is an extensive one. Susi WtLrmbrand has compiled a thorotLgh bibliography of works up to 2006, which is accessible online: <.http://wurmbrand.uconnedu/Bibliographies/vc-bib.html>. WtLrmbrand 2006 offers a good overview to the topic, including a review of the various generative theoretical accounts for duster formation. Den Besten & Edmondson 1983 and Zwart 1996 are classic papers on verb dusters in Continental West Germanic. Selected works specifically on variolLS German-related varieties include the following: Schmid & Vogel 2004 for multiple, mainly High German dialects; L5tscher 1978 and ScMnenberger 1995 for Swiss German; Patocka 1997 for Austrian German; DtLbenion-Smith 2008 and 2010 for West Central German; Kaufmann 2007 for Mennonite Low German; and Sapp 2006 for Early New High German. Abraham 2009 considers data from Dutch and multiple southern German varieties, including dialects ~;poken in northern Italy, and interprets differing duster orders as indicative of a typological continuum between OV and VO basic word orders. 2. For modal IPP constructions, Schmid and Vogel (2004) document all possible orders except 2-3--1, though it should be noted that some orders are limited to very specific focus/ stress conditions. For the varieties Wurmbrand (2006: [9]) investigated, only 2 ··1·· 3 is lacking. She notes that 2-·3··1 is the norm in Afrikaans and marginally possible in West Flemish, both varieties of which were not considered by Schmid and Vogel since their study was limited to dialects of German. Likewise, 2··1·3 is found in German dialects that Wi.umbrand did not consider.

Synchrony and dlachrony of verb clusters in Pennsylvania Dutch 167

brought to Pennsylvania during the colonial period.3 We begin with a discussion of data from modern Pennsylvania Dutch varieties spoken by Old Order Anabaptist sectarians from i'i·VO major regions, Lancaster County, Pennsylvania, and the Midwest (e.g. 0 hio and Indiana). Although a number of lexical differences distinguish regional and generational varieties of Pennsylvania Dutch from one another, there is remark-able uniformity across them at the syntactic level, including with respect to verb clusters, which allows us to make descriptive generalizations that capture the grammars of most if not all individual sectarian Pennsylvania Dutch speakers. The data from modern Pennsylvania Dutch, discussed in Section 2, are followed in Section 3 by historical evidence of verb clusters derived from Frey 1941, which allows us to identify a limited amount of change from older to modern Pennsylvania Dutch. In both sections, the focus is on three--verb clusters in subordinate clauses. The paper concludes with discussion of a number of implications, as well as an incipient change in main-clause verb clusters in varieties of Pennsylvania Dutch spoken by younger Midwestern speakers.

2.

Verb clusters in modern Pennsylvania Dutch

As mentioned above, there is a remarkable degree of uniformity across modern Pennsylvania Dutch varieties in terms of which orderings of elements in verb clus-ters are grammatical. Further, for Pennsylvania Dutch speakers, unlike what has been documented in many European German varieties, there is no scale of acceptability across multiple orderings of the same verbal elements. That is, for individual Pennsylvania Dutch speakers there is typically only one grammatically possible order of elements for each verb cluster type. Let us begin with two superficially similar three-verb structures, the first (2) an IPP construction and the second (3) involving a verb of perception. (2)

heere hab kenne 2 1 3 they have known that them hear have can 'They knew that I could hear them'

(3)

gheert hab schwetze 3 2 1 they have known that them heard have speak 'They knew that I heard them talking'

Sie

Sie

hen

hen

gwisst, as

gwisst, as

ich sie

ich sie

3· In Louden 2006 I give an overview of the current sociolinguistic situation of Pennsylvania Dutch, also known as Pennsylvania Germa11 The two terms are synonymous, the latter being preferred by scholars and some language advocates; most speakers themselves use the former.

168 Mark L. Louden

Although both these constructions involve a finite verb and two nonfinite elements, the structures underlying them differ, specifically as regards the structural position of the linearly final verb form in each. In (2), the infinitive to the immediate right of the finite verb, kenne, is the complement of the finite verb hab. In (3), on the other hand, the infinitive schwetze is the complement of the past participle gheert, so the adjacency of hab and schwetze is coincidental. That the two sentence-final infinitives in (2) and (3) occupy different structural positions is underscored by the fact that no word may intervene between the finite verb and its complement to the right, as in (2), whereas that is not the case for structures like (3). Consider (4), in which the finite verb is followed by the simple 0-V constituent, Deitsch schwetze 'to speak Pennsylvania Dutch'. (4)

Sie hen gwisst, as ich sic gheert hab they have knovm that them heard have Deitsch schwetze Pennsylvania-Dutch speak 'They knew that I heard them speaking Pennsylvania Dutch'

Key to understanding the structural difference between 3-1-2 and 2-1-3 clusters in Pennsylvania Dutch is the recognition that this language, like most spoken varieties of German, makes productive use of the so-called post-field (Nachfeld), that is, the syntactic "space" located beyond the rightmost clause periphery (rechte Klammer 'right bracket'), thereby minimizing the number of elements that may appear in the clause proper, that is, the inner field (Mittelfeld) and the right bracket. 4 Essentially, the modern Pennsylvania Dutch inner field contains only arguments (including subjects, objects, and a limited number of prepositional phrases that are verbal complements) and negation; other elements, including non-argument adveibials (especially PP--adjtmcts) and some PP-- and all non--PP complements (e.g. infinitival phrases, sub-ordinate clauses), must be located in the post-field. Consider the brief text in ( 5), from the introduction to a book of children's Bible stories written in Pennsylvania Dutch (Vella Deitsh 1997: 7). The right periphery of each clause is marked with T

4· The descriptive "topography" of the German clause is characterized by a verbal or clausal frame (Verbalklammer, Satzklammer) that defines the following structure: prefield (Vorfeld)left bracket (linke Klammer) - inner field (Mittelfeld) - right bracket (rechte Klammer) post-field (Nachfeld). In terms of X-bar phrase structure, the prefield is [Spec, CP] and the left bracket is [C, CP]. "Ihe right bracket consists of two verbal head positions, [Y, VP] and ['1: TP], in that order. "Ihe inner field contains all non-verbal elements between the left and right brackets. In main clauses the finite verb is located in the left bracket ([C, CP]), while in subordinate clauses it is in ['I: TP], which ma1f<s the right periphery of the clause. The post-field begins immediately after rt: TP].

Synchrony and diaduony of verb clusters in Pennsylvania Dutch 169 (5)

Kinner glekhe children like

I

Sd1tories glese hawwe I zu sie. stories read have to them

Die Schtories sin vun die Biwwell these stories are from the Bible un waare gschriwwe I bei etlichi Leit vun die Gegend. and were written by several people from the area Mir wisse I as unser Schprooch net gleiche is I iwwerall not same we know that our is everywhere language in alli Heemet. in each home Dir sollet frei sei I fer eier eegni Wadde neiduh, you (pl) shm.lld free be to your own words insert so wie dir wellet I as like you want un die Kinner aa noch meh verzeele I weeich die Schtories. more tell about the stories and the children also yet 'Children like to have stories read to them. These stories are from the Bible and were written by several people from the area. We know that our language is not the same everywhere and in every home. You should feel free to insert your own words as you please and to tell the dlildren more about the stories:

Returning to sentences (3) and (4), we can reasonably argue that the infinitive

schwetze 'talk, speak' is located to the right of the clause periphery:5 (3')

Sie hen gwisst I as ich sie gheert hab I schwet:r.e

(4')

Sie hen gwisst I as ich sie gheert hab I Deitsch schwet:r.e

Other examples of clauses with postposed infinitivals are given in (6)-(8): (6)

... as sie gange sin I (Balle) schpiele that they gone are ball play 'that they went to play (ball)'

(7)

ich en gholfe hab I (Gscha) wesche ... as him helped have dishes wash that I 'that I helped him wash (the dishes)'

5· Note that a subordinate clause in Pennsylvania Dutch (and German) that follows the main clause to which it is semantically "subordinate" is strictly speaking right -adjacent to the main clause and not enibedded within it, as is the case with as ich sie gheert hab 'that I heard tb.em'vis-a-vis sie hen gwisst 'they kneW. l-Ienee the terms "subordinate" and «embedded" with respect to clause types are not synonymous.

170

Mark L. Louden (8)

... as ich gegliche hab I (Deitsd1) schwetze that I liked have PD ~;peak 'that I liked to talk (Dutd1)'

We may thus formulate the following preliminary rule about the structure of sentences (3)-(4) and (6)-(8):

Preliminary verb cluster rule: Verbal complements conta.ining an infinitive in Pennsyl-vania Dutch must be located to the right of the clause periphery. So if the 2 --1-· 3 order in sentences such as ( 3) has the structure 2--1 I 3,6 what do we make, then, of the 3-1-2 IPP structure in (2)? (2)

Sie hen gwisst I as ich sie heere hab kenne

If this structure were to behave according to the preliminary rule above, then we would expect the following 2--1 I 3 structure, which is ungrammaticaL (2') *Sie hen gwisst I as ich sie kenne hab I heere

In order to get at the structure in IPP constructions such as (2), let us begin by considering (9) and (10). (9)

(10)

Sie wissel as ich sell Buch will they know that I that book want 'They know that I want that book' Sie

hen

gwisst I as

ich sell

Buch hab

welle 2 they have known that that book have want "Ihey knew that I wanted that book' 1

In (9) and (10) welle is not functioning as a modal verb, rather as a main verb meaning 'to want (s.th.)'. Yet in the perfect tense, welie and all other modal--cum--main verbs show the IPP effect, that is, they appear as infinitives rather than participles (contra German, wollen ""gewollt, etc.). In perfect constructions in subordinate clauses such as ( 10 ), the only grammatical order of the verbal elements is 1-2 (i.e. *... as ich sell Buch

welle hab).

Wurmbrand (2006: [10]) mentions that others (e.g. Rutten [1991], Robbers [1997], and Wurmbrand [2001]) have also analyzed 2-1-3 orders as being the result of extraposition (i.e. 2-1 I 3). Note also the complementary analys-is by Kroch and Santorini (1991) in the formalism of Tree Adjoining Grammar.

6.

Synchrony and diachrony of verb clusters in Pennsylvania Dutch

Let us now consider (11)-(15). (11)

Sie

wisse I as

ich gehe will 2 they know that go want 'They know that I want to go'

(12)

Sie

hen

ich gehe hab welle 3 1 2 they have known that go have want 'They knew that I wanted to go'

(13)

ich gange bin 2 they have knoWl.l that gone am 'They knew that I went'

(14)

ich gange bin schwimme 2 1 3 gone am swim they have knoWl.l that 'They knew that I went swimming'

(15)

ich gehe hab welle schwimme 4 2 3 1 have want swim they have knoWl.l that go 'They knew that I wanted to go s"Wimming'

Sie

Sie

Sie

hen

hen

hen

g~isst

I

as

gwisst I as

gwisst I as

gwi-.st I as

Sentence (14) resembles (6) in that the infinitivalcomplementof gange'gone: schwimme 'to swim: is located in the post-fi.eld. Both sentences share a 2--1 I 3 structure. How, then, may we analyze the 3-1-2-4 structure in (15), which would appear very difficult to derive, e.g. by rightward movement into the post-field. One solution to this problem would be to assume that hab and welle are both located under the same structural node, the same node under which bin is located in (14). (14')

Sie hen g·wisst I as ich gange [bin] I schwimme

(15')

Sie hen gwisst I as ich gehe [hab welle] I schwimme

This would mean that the 3-1-2 IPP structure is actually a 2-1 structure in which the "1" position is occupied by a complex element.? Thus a sentence like (2), where we began our analysis, would be structured as follows: (2)

Sie hen gv;isst I as ich sie heere [hab kennel 2 1

7· In structural terms these [hawwe + modal] compounds would be complex heads located under the D: TP] node.

171

172

Mark L. Louden

tar

To be sure, it is not possible to go so as to argue that ha.b and welle (and other forms of hawwe 'have' and modals) comprise a single morphological word, since they occur discontinuously in main clauses, as in (16). (16)

Ich hab gehe welle schwimme 'I wanted to go swimming'

On the other hand, semantics supports the assumption that hab and welle occupy two branches under a single, functional verbal syntactic node, cf. the Gennan and .English one--word translations of these two Pennsylvania Dutch words, namely woUte and wanted. Further, there is an interesting morphosyntac..tic behavioral property of [hawwe +modal] that suggests an affinity between the two elements. When the auxiliary haww-e is inflected for the subjunctive mood, the stem of the modal infinitive is homophonous with the present subjunctive stem for that modal. Cf. examples for welle in ( 17). (17)

a.

Mir welle gehe we want go 'We want to go'

b.

Mir hen gehe welle we have go want 'We wanted to go'

c.

Mir wedde gehe we would-·want go 'We wm.lld want to go'

d.

Mir hedde gehe wedde would-want we would-have go 'We would have wanted to go'

This phenomenon of subjunctive mood being marked on both the finite auxiliary and modal infinitive applies to all modals in Pennsylvania Dutch, not just welle, as shown in (18). ( 18)

mir dafe ~ mir deifde; mir hen dafe ,.. mir hedde deifde 'we are/would be permitted to; we were/would have been permitted to' mir kenne "' mir kennde; mir hen kenne "' mir hedde kennde 'we are/would be able to; we were/would have been able to' mir misse "" mir missde; mir hen misse ..., mir hedde missde 'we have/would have to; we had/would have had to' mir solle/selle ~ mir sedde; mir hen solle/selle.., mir hedde sedde 'we are/would be supposed to; we were/would have been supposed to'

This use of "subjunctive modals" is the norm in modem Lancaster and Midwestern sectarian Pennsylvania Dutch, however it is not attested in older varieties of the

Synchrony and diaduony of verb clusters in Pennsylvania Dutch 173 language, suggesting that it might be a relatively recent innovation. Regardless of the antiquity of this phenomenon, it does lend morphological support for the existence of a syntactic unit [hawweFL.'l + IPP].8 Let us now consider our analysis of the apparent 3-1-2-4 cluster in (IS') as a 2··1 I 3 structure, given again here, in light of the preliminary rule formulated above, also repeated below. (IS')

Sie hen gwisst, as ich gehe [hab welle] 3 2

2

I

schwimme 4 3

P1-eliminary verb cluster ru.ie: Verbal complements containing an infinitive in Pennsylva·· nia Dutch must be located to the right of the clause periphery. If we analyze gehe as the complement of welle or [hab "M'elle], then this rule must be modified. Since all verb fonns in subordinate clauses in Continental \Vest Gennanic dialects are located in the clausal right bracket, the follo~ing generalization, based on the difference between lexical and non--lexical verbs (i.e. auxiliaries and modals), cap·· lures the facts of verb clusters in modern Pennsylvania Dutch most accurately:

Verb cluster rule: Ihe right bracket in a Pennsylvania Dutch clause may conta.in no more than one lexical verb; additional lexical verbs must occur to the right of the clause periphery. 3. Verb dusters in earlier Pennsylvania Dutch We tum now to the historical development of verb clusters in Pennsylvania Dutch. The earliest, most thorough description of Pennsylvania Dutch syntax in general is contained in the doctoral dissertation of]. William Frey (Frey 1941). Included in this description is an extensive discussion of verb clusters in main and subordinate clauses. 1hough based on his native dialect from eastern York County, Pennsylvania, Frey's observations about Pennsylvania Dutch syntax are in line with what we know from prose texts produced by native speakers from across the Pennsylvania Dutch-speaking area and dating as far back as the middle of the 19th century. In what follows we

8. Hirg Fleischer (p.c.) has brought to my attention references to apparently simila:r"subjunctive modal infinitives" in European German dialects, namely in Moselle Franconian (Labouvie 1938: 105) and Lower Alemannic (Noth 1993: 330). Gt>z Kaufmann (p.c.) has doctLmented the phenomenon in the (originally Rhenish-Palatine) HunsrUckisch spoken in southern Brazil. Pennsylvania Dutch, however, is not directly descended from input dialects from any of these areas, which suggests that these similarities are due to parallel development.

174 Mark L. Louden

examine what Frey documents for verb clusters in earlier Pennsylvania Dutch and compare his data ~ith parallel structures in the modem language. The data are divided into four groups, according to verb type: (i) modals; (ii) causatives; (iii) auxiliaries; and (iv) verbs of perception.

3.1

Verb clusters with modals

In IPP verb clusters containing a form of hawwe +modal infinitive+ lexical infinitive, the J-1··2 order attested in the modern language (MPD) is also what we find in earlier Pennsylvania Dutch (EPD). (19)

EPD/MPD: kh wees I as er net kumme hot kenne I 3 l 2 'I know that he could not c.ome'

(Frey 1941: 234)

Frey also documents the same structure for the verb brauche 'need' in the "2" slot, which differs fi·om what we find in modem Pennsylvania Dutch. (20)

EPD: Ich wees, as er net kumme hot brauche 3 1 2 MPD: Ich wees I as er net gebraucht hot I kumme 3 1 2 'I know that he did not need to come'

(Frey 1941: 234)

What has happened here is that bra.uche is no longer regarded as a modal verb in modem Pennsylvania Dutch, henc.e the need tor its complement (here, kumme) to be located outside the clause. In his dissertation, Frey also lists 3-1-2 structures in which the finite verb ("1") is a modal, "2" is hawwe, and in the "3" position is a lexical verb. Cf. (21). (21)

EPD: Ich wees, as er en gfa:nge soll hawwe 3 1 2 'I know that he should have caught him'

(Frey 1941: 234)

No such inversion is possible in the modem lanbruage, which arranges the elements in a straightforward J. 2··1 way. (21 ')

MPD: Ich wees I as er en gfimge hawwe solll 3 2 I

Clearly, the modern structure is again obeying the verb cluster rule formulated above: after the modal soll, both hawwe and gfange 'caught' may remain within the clause since only gfange is lexical, thus the maximum of one lexical verb in the right bracket is not exceeded.

Synchrony and diaduony of verb clusters in Pennsylvania Dutch 175

One final comment about verb clusters containing modals in earlier Pennsylvania Dutch is in order here. Frey (1941: 233) notes that when modals occur without a complement and in the perfect tense in subordinate clauses, the order is always 1-2, as is still the case in the modern language. (22)

3.2

EPD/MPD: Sie hot ihre Kinner griege misse I de bescht Weg I as sie hot kenne I 1 2 'She had to get her children the best way (that) she could'

Verb clusters ""ith causatives

In earlier and modern Pennsylvania Dutch, the verbs losse 'lef and mache 'make' may function as causative verbs and take infinitival complements. In earlier Pennsylvania Dutch, these two verbs behaved identically to modals, participating in IPP construe·· tions and occurring in 3-1-2 structures. (23)

EPD: Ich wees, as er ihn geh hot losse 3

1

(Frey 1941: 234)

2

'I know that he let him go' (24)

EPD: Ich wees, was ihn schreiwe hot mache 3 1 2 'I know what made him write.

(Frey 1941: 234)

In modern Pennsylvania Dutch, losse and mache, like bra.uche above, no longer behave as modal verbs (i.e. they do not show the IPP effect), but the order of the elements in the verb clusters in which they appear depends on whether they are interpreted as lexical verbs or not. Ct~ (23') and (23"), and (24') and (24"). (23')

MPD: Ich wees I as er ihn geh glosst hot I 3

2

1

'I know that he (ju'>l:) let him go' (23")

MPD: Ich wees I as er ihn glosst hot I geh 2 1 3 'I know that he gave him permission to go

(24')

MPD: kh wees I was ihn schreiwe gmadrt hot I 3

2

1

'I know what moved him to ~Tite' (24'')

MPD: Ich wees I wer ihn gmadrt hot I schreiwe 2 1 3 'I know who compelled him to ~•ite'

176 Mark L. Louden

In clauses in which there is clear agency behind the action expressed by the verb, as in (23'') and (24"), the causative verb is regarded as lexical and its complement is postposed. Where agency is deemphasized, the causative is not lexical, therefore its complement is allowed to remain adjacent to it within the right bracket of the clause, as in (23') and (24').

3·3

Verb clusters with auxiliaries

Frey (1941) identifies a number of auxiliary verbs in three--verb clusters of the form 3-1-2, including ha·wwe 'have' and sei 'be' in the past perfect tense. Cf. Examples (25)

and (26). (25)

EPD: Ich wees, as er sell geduh hot ghatt 3 1 2 'I know that he had done that'

(Frey 1941: 234)

(26)

EPD: Ich wees, as er heem gange is gwest 3 1 2 'I know that he had gone home'

(Frey 1941: 234)

Not surprisingly, in modern Pennsylvania Dutch these clusters have the order 3-2--1, since the participles ghatt and gwest are not le.xical forms. Cf. (25') and (26'). (25')

MPD: Ich wees I as er sell geduh ghadde hot I 3 2 I

(26')

MPD: Ich wees I as er heem gange gwest is I 3 2 1

Frey mentions that some of the speakers he interviewed also inverted the finite and participial forms of hamt-e in two-verb clusters, as in (27). (27)

EPD: Wie waer sell, wammer noch Schpuke do rum hedde, wie sie als fer alders hen ghatt? (Frey 1941: 232) 1 2 'How would that be if we still had ghosts around here as they used to have in the old days?'

Frey (1941: 233) does comment, however, that such 1-2 orders are less common than the 2-1 order (ghatt hen), which is the norm in modern Pennsylvania Dutch. Aside from perfect auxiliaries, Frey (1941: 236) mentions another auxiliary verb that participated in 3-1-2 structures in earlier Pennsylvania Dutch, namely griege 'get,

Synchrony and diaduony of verb clusters in Pennsylvania Dutch 177 receive'. Although this is a native (Palatine German-derived) construction, its semantics parallels that of the English construction get + direct object + past participle (e.g. to get something done). Cf. example (28). (28)

EPD: Un wie sie alles ausgepaeckt: hot grigt, (Frey 1941: 236) 3 1 2 hot sie die Hensching net finne kenne. ~nd when she had gotten everything unpacked she could not find the gloves'

Since griege in this construction is not interpreted as a lexical verb in modern Pennsylvania Dutch, its participle remains in the right bracket with its complement (here, the participle a.usgepaeckt), as in (28'). (28')

MPD: Un wie sie alles ausgepaeckt grigt hot I ... 3 2 1

'!here is one final verbal auxiliary in earlier Pennsylvania Dutch that occurred in 3-1-2 structures, namely the passive auxiliary wan-e (ct: German werden ). Frey (1941) points out, however, that there was variation with this verb, as Examples (29) and (30) show. (29)

EPD: Ich wees, as er doot gmacht is warre 3 1 2 'I know that he was killed'

(Frey 1941: 234)

(30)

EPD: kh wees, as en Faecktri gschtaert warre is 3 2 1 'I know that a factory has been sta1ted'

(Frey 1941: 235)

Speaking of frequency, Frey (1941: 235) comments that the inverted (3-1-2) order fi.)r warre in three-verb clusters is (was) more common than the non--inverted order (3-·2··1, as in Example [30]). But wa.rre also used to invert on occasion in two--verb clusters, when used as the lexical verb meaning 'to become'. In this case, though, Frey (1941: 233) states that the 2-1 order was the more frequent of the two. C£ Example (31). (31)

EPD: Ich hab net gwisst, as er grank is warre/warre is 1 2 2 1 'I did not know that he became ill'

(Frey 1941: 233)

In modem Pennsylvania Dutch, there is still more variation between (3- )1-2 and (3··)2··1 orders with warre than with any other verb cluster type, though younger speakers especially favor the latter. The likely reason for this variation has to do with the phonetic shape of the verb. The infinitival and past participial forms of wan-e are

178 Mark L. Louden

homophonous, and unlike in standard German, which has two different participles for werden depending on whether the verb is lexical or a passive auxiliary (i.e. geworden vs. worden), Pennsylvania Dutch has never distinguished the two phonetically. Because of this identity between infinitive and past participle, some speakers may unconsciously assign warre to the lPP (modal) class of verbs, which invert (i.e. take 1·· 2 orders) in both two- and three-verb clusters.

3·4

Verb clusters with verbs of perception

Finally, the verbs of perception sehne 'see and heere 'hear: both in earlier and mod·· ern Pennsylvania Dutch, participate in 2··1 I 3 structures, though in the older variety, according to Frey (1941: 235-236) these verbs could either appear as infinitives in the perfect (i.e. IPP) or as past participles. Examples are given in (32) and (33). Frey documents the same variation for the verb heife 'help' (cf. [34]). (32)

EPD: Ich wees, as er mich heere/gheert hot I singe 2 1 3 'I know that he heard me singing'

(33)

EPD: Ich wees, as er mich sehne/gsehne hot I kumme 2 1 3 'I know that he saw me c.oming'

(ibid)

(34)

EPD: Ich wees, as er ihm net helfe/gholfe hot I schaffe 2 1 3 'I know that he did not help him work'

(ibid)

(Frey 1941: 235)

In the modern language. only the participial forms occur with sehne, heere, and heljewhen they take an infinitival complement. This suggests that today only true modal verbs, as opposed to other non-lexical and lexical verbs, may participate in I PP structures.

3·5

Summary: Change from earlier to modern Pennsylvania Dutch

Taking stock of what we have just observed, the following general observation can be made about the difference between earlier and modern Pennsylvania Dutch regarding verb dusters. In the earlier language, 3--1··2 structures included a more diverse range of verb types and verb forms; in today's Pennsylvania Dutch, by contrast, only IPP modal constructions take the order 3-1-2 in subordinate dauses. Other, formerly 3-1-2 clusters have been reanalyzed as either 2·1 13 or 3--2··1 structures. This is summari7..ed in the table below.

Synchrony and dlachrony of verb clusters in Pennsylvania Dutch 179

Table I. Three-verb clw.ters in earlier and modem Pennsylvania Dutch ModernPD

EarlierPD Modals: 3-1-2 (INF- V- IPP)

3-1-2 (INF- V- IPP)

kumme hot kenne

kumme hot kenne

I 3 (PTC - v I INF)

3-1-2 (INF- V- IPP)

2--1

kumme hot brauche

gebraucht hot kumme

3-1-2 (PTC- VMOD- IPP)

gfange soli hawwe

>

3-2-1 (PTC- INF- VMOD)

gfange hawwe soli

_Q_\\JJ.s.a.U'!:~_s_:

3-1-2 (INF- V- IPP)

3-2-1 (INF- PTC- V)

geh hot losse sd:reiwe hot mad:e

geh glo~t hot schreiwe gmad:t hot 2-1 I 3 (PTC -

v I INF)

glosst hot geh gmacht hot schreiwe Ay;xJ!ia.ri.~:

3-1-2 (PTC- V- PTC)

3-2-1 (PTC- PTC- V)

geduh hot ghatt gange isgwest ausgepaeckt hot grigt gmacht is warre9

geduh ghadde hot gange gJ-Vest is ausgepaeckt grigt hot gmacht warre is

Recalling the verb cluster rule formulated in Section 2, no more than one lexical verb may be located in the right bracket of a clause in modern Pennsylvania Dutch: additional verbal complements must be postposed, yielding 2-1 I 3 structures. We also note that the formerly modal verb brauche has been reanalyzed as a lexical verb, hence its participation in the 2-1 I 3 pattern. Finall}; it is dear that 1-2 inversion in subordinate clauses in the modern language of the hab welle type is limited exclusively to modals (the only class of non-lexical verbs that may appear in IPP constructions): non-modal auxiliaries in both their historical IPP (e.g. hawwe, losse, and mache) and participial forms (e.g. ghatt, gwest, grigt, and warre) may no longer occur in 1--2 inverted structures. Finally, we note that the historical 2-1 I 3 pattern with verbs of perception and helfe is maintained in modern Pennsylvania Dutch, the only difference being that causatives, as non-modals, may no longer take an IPP form in the perfect.

9· We recall here that Frey (1941: 235; see Ex. [30] above) mentions that a minority of his consultants also produced 3-2-1 clusters with warre, suggesting that the change to the modern pattern was already underway at the time of his study.

t8o Mark L. Louden

Table 2. 2-1 I 3 structures in earlier and modern Pennsylvania Dutch

ModemPD

EarlierPD

sehne. heere. he{te 2-1 I 3 (INF/PTC - v I INF) heerelgheert hot singe

sehne/gseh,le hot kumme helfe/gholfe hot schaffe

4·

>

2--113 (PTC-

v IINF)

gheert hot singe gsehne hot kumme gho(fe hot schaffe

Discussion

The major observation that emerges from a consideration of the synchrony and diachrony of three-verb clusters in Pennsylvania Dutch subordinate clauses is that the number of verbs that occur in 3-1-2 clusters has decreased since the 1930s and 1940s, when]. William Frey conducted his fieldwork. Whereas non-finite forms (both infinitives and past participles) of apparently all non-lexical verbs (modals, auxiliaries, and perception verbs) could invert with the finite verbs that dominated them (finite forms of the perfect auxiliaries ha1+we and sei, as well as finite modals, e.g. as in Ex. [21]), today only modal verbs may do so (with a finite form of hawwe). Further, these modal verbs only ever appear as infinitives in the perfect tense, even when used lexically, and are today the only verbs that show the IPP effect The reanalysis of formerly 3-l-2 clusters under the constraint of the modem verb cluster rule formulated in Section 2 has led to the expansion of 3-2-1 and 2-1 I 3 structures. We recall, however, that in order to account for apparent 3-1-2-4 structures such as (15), we need to assume that the finite form of hawwe and its modal IPP comple-ment form a structural unit [hawweFIN + IPP]. Following this assumption, there are in essence no "four-verb clusters" in Pennsylvania Dutch; 3-1-2-4 orders are in fact 2-1 I 3 structures. (15)

Sie hen gv.'i.sst, as ich gehe hab welle I schwimme 3 [1 2] 4 3 2

That means, then, that 3-1-2 clusters (e.g. gehe hab welle) are themselves really just 2-1 structures, as discussed earlier. The system that reveals itself is thus a simple one for modern Pennsylvania Dutch verb clusters in subordinate clauses: the underlying 2--1 order is preserved in all surface orders, and if the number of lexical verbs within the clause exceeds one, then the most deeply embedded infinitive ("3") is extraposed.

Synchrony and diaduony of verb clusters in Pennsylvania Dutch Table 3. Two- and three-verb dUl>iers and 2-113 structures in modern Penru.ylvania Dutd1 2-1

PTC + VAUX INF + VMOD INF + [VAUX + IPP]

3-2-1

2--113

PTC INF INF PTC PTC PTC

PTC -

VAUX

VMOD VAU'.!:

VAUX

I INF

... gange bin ... gheerthab ... gehewill ... gehe hab welie ... gfange r.awwe soil ... geh glosst hot ... geduh ghadde hot ... gebraucht hot kumme ... glosst hot geh ... gheert hot singe

As mentioned at the outset of this paper, the main goal here has been a documentary one, that is, to give a thorough description of the Pennsylvania Dutch data from both synchronic and diachronic perspectives. A future task is to consider the theoretical implications of these data in the context of the larger discussion about verb clusters in Continental West Germanic. In particular, the most radical claim here is that the finite form of ha.-w'l+-e and its modal IPP complement form a syntactic, if not a morphological unit of structure. While it may be possible to derive 3--1--2 orders via movement or reanalysis (cf. the theoretical discussion going back at least to Haegeman & van Riemsdijk 1986), 3-1-2-4 structures remain a problem. The elegance of a solution based on a unit [hawweFIN + IPP], which is supported by the fact that these two lexical items only ever occur in "inverted" (apparent 1--2) order in subordi-nate clauses, as well as by the behavioral phenomenon related to subjunctive inflection, is compelling. It remains to be seen, however, whether there are parallels in closely related European German dialects. 10 In any case, the analysis here should be evaluated against the backdrop of the larger generative theoretical discussion on verb clusters. Additional tasks for future work include a closer look at both the historical record of Pennsylvania Dutch, as well as at what is perhaps a change in progress in verb clusters among younger sectarian speakers. Regarding older data, it will be important to compare Frey's data from adult speakers who were living in the 1930s and 1940s with evidence from speakers from earlier generations. Quite fortunately, we have hundreds of Pennsylvania Dutch prose and poetic texts produced by native speakers as far back

10. My own suspicion is that the micro-level phenomena described here are probablyuniqtLe to Pennsylvania DtLtch, or at the very least not the result of inheritance from Palatine German source dialects. It seems unlikely that the many 1-2 structures in eadier Pennsylvania Dutch, which almost certainly wotLld have also been found in Palatine dialects, all took the syntactic form [1 + 2].

181

182

Mark L. Louden as the middle of the nineteenth century and earlier. Owing to the vernacular, oral character of the language, most Pennsylvania Dutch texts were produced without regard for prescriptive norms either coming from within the Pennsylvania Dutch community or from without (e.g. from standard German). These texts therefi.)re hew quite closely to the naturally spoken language. My preliminary comparison of Frey's data with data from older speakers has yielded relatively few differences thus far, but those I have found are quite interesting. 'Ib take one example, we can consider the idiolect of Edward H. Rauch, a native Pennsylvania Dutchman who was born in Lancaster County; PA, in 1820 and died in Carbon County; PA, in 1902. Over the course of his long career as a political activist, newspaper editor, and language advocate, Rauch produced a number oflengthy prose texb in Pennsylvania Dutch that comprise a rich corpus for syntactic analysis, one collection of which is Rauch 1868. In terms of verb clusters, the data from this collection are generally in line with what Frey found among speakers who would have been born two or three generations after Rauch. That is, Rauch's data, like those in Frey 1941, include more types of (3- )1-2 clusters than what we find in the modern language. There is at least one inversion construction in Rauch, however, that Frey did not document, namely with the verbs geh 'to go and kumme 'to come: See Examples (35) and (36). (35)

un

alli

Mol getriet hot er, wann mer anne

is gange 2

(Rauch 1868: 3) and every time treated had he when one to--there is gone 'and he treated every time one went there' (36)

wie

(Rauch 1868: 4)

mer awwer an sei Haus

sin kumme ... 2 as/how we but at his house are come 'but when we came to his house .. .'

Compare these examples to (37) and (38). (37)

wie

es am

Readinger Singerpescht 11

gange is

(Rauch 1868: 15)

2

as/how it at-the Reading Singers--Festival gone 'how it (things) went at the Reading Singers Festival'

is

u. Rauch is engaging in wordplay here, substituting the noun Pescht )Jiague' for Fescht 'festival:

Synchrony and diachrony of verb clusters in Pennsylvania Dutch 183 (38)

sidder der Johnson frei kumme is

(Rauch 1868: 12)

2

since the Johnson free come 'sinc.e Johnson was set free'

is

In Rauch's idiolect, the participles gange and kumme invert with the auxiliary sei (here, is), but only when the verbs geh and kumme are used non--figuratively, that is, meaning literally 'to go' and 'to come: as in (35) and (36). In (37) and (38), where geh and kumme have figurative meanings, 1-2 inversion does not occur. This is an intriguing fact suggesting semantic restrictions on inversion in earlier Pennsylvania Dutch that should be investigated in further research. Just as important as adding to our understanding of the historical record of verb clusters in Pennsylvania Dutch is noting possible changes in progress among today's youngest speakers. To date, I have not observed any variation in contemporary Pennsylvania Dutch in the order of elements in three-verb clusters in subordinate clauses, the primary focus of the present paper, however there is variation with main clauses. See Examples (39) and (40) from VellaDeitsh 1997, pp. 61 and 63, respectively. (39)

laafe hot kenne, 3 1 2 who never not properly walk has can

Der laahm M.ann, wo the

lame

man

nie

net recht

hot nau laafe un schpringe kenne 1 3 3 2 has now walk and run can

"Ihe lame man, who never was able to walk properly, was now able to walk and run: (40)

duh kenne fer ihre Sache reclrt mache, 3 2 they have nothing more do can to their things right make Sie

hen

nix

meh

1

kenne sehne, was Blatz gnumme hat 2 3 see what place taken has the other people have can

awwer die annri Leit

hen 1

but

'They could do nothing more to make amends, but the other people could see what had happened: In three-verb dusters in main clauses, in earlier and modern Pennsylvania Dutch, the normal order is 1 3-2, as in (39) and in the first main clause in (40). Some younger speakers, however, produce 1 2-3 orders (cf. the second clause in [40]). Pending further investigation, it seems likely that pragmatic factors are at work here. That is, in a focused or othenvise discourse-prominent clause, the main (lexical) verb is extraposed

184 Mark L. Louden

for emphasis. This would mean that the 1 ... 2-3 order in such sentences would be represented more accurately as 1 ... 213. To be sure, this extraposition of the lexical verb is not mandatory according to the modern verb cluster rule, since there is only one lexical verb in the clause ("3~ in [40] sehne 'to see'), yet it may well be that the grammar allows for optional extraposition for pragmatic reasons. Likewise, since the main clause 1 2 I 3 order is analogous to the 2-1 I 3 order in subordinate clauses (albeit with non- IPP constructions), the diachronic trend toward increasing produc.ti.vity of the latter structure might suggest an eventual reanalysis of 3-·1· 2 clusters in subordinate clauses. Such a reanalysis is purely speculative at this point, in the absence of further data.

References Abraham, Werner. 2009. Methodological considerations on grammar variation. The right periphery as an OV /VO deciding parameter more so than the left periphery: Gradience in the verb cluster. In Describing ar1d Modelling Variation in Grammar, Andreas Dufter, Jtirg Fleischer & Guido Seiler (eds ), 21-58. Berlin: Mouton de Gruyter. den Besten, Hans & Edmondson, Jerrold A. 1983. The verbal complex in continental West Germanic. In On the Formal Syntax of the Westgermania: Papers from the 3rd Groningen Grammar Ialks [Linguistik Aktuell!Linguistics Today 3], Werner Abral1am (ed.), 154-216. Amsterdam: John Benjamins. Dubenion-Smith, Shannon A. 2008. Verbal Complex Phenomena in the West Central German Dialects. Ph.D. dissertation, University of \l\-1sconsin-Madison. Dubenion-Smith, Shannon A. 2010. Verbal complex phenomena In West Central German: Empirical domain and multi-causal account Journal of Germanic. Unguistics 22: 99-191. Frey, John Wllll.am. 1941. The German Dialect of Eastern York County, Pennsylvania. Ph.D. dissertation, University of Illinois. Haegeman, Liliane & van Riemsdijk, Henk. 1986. Verb projection .raising, scope, and the typology of rules affecting verbs. Linguistic Inquiry 17: 417-466. Kaufmann, Go1. 2007. 'Ihe verb duster in Mennonite Low German: A new approach to an old topic. Lingulstische Beridtte 210: 147-207. Kroch, Anthony S. & Santorini, Beatrice. 1991. 1he derived structure of the West Germanic verb-raising construction. In Principles and Parameters in Comparatiw Grammar, Robert Freidin (ed.), 269-338. Cambrtdge MA: The MIT Press. Labouvie, Erich. 1938. Studien zur Syntax der Mundart von Dillingen an der Saar. Marburg: Elwert. Lotscher, Andreas. 1978. Zur Verbstellung im Zi.irichdeutschen und in anderen Varianten des Deutschen. Zeitschrift fUr Dlalektologie und Linguistik 45: 1-29. Louden. Mark L. 2006. Pennsylvania German in the twenty-first century. In Sprachlnselwelten. The World of Language Islamis, Nina Berend & Elisabeth Knipf-Koml6si (eds), 89-107. Frankfurt: Peter Lang. Noth, Harald. 1993. Alemannlsches Dlalekthandbuch vom Kaiserstuhl und seiner Umgebr.mg. Freiburg: Schillinger Vedag. Patocka, Franz. 1997. Satzgliedstellung In den balrischen Dialekten Osterreichs. Frankfi.ut: Peter Lang.

Synchrony and diachrony of verb clusters in Pennsylvania Dutch Rauch, Edward H. 1868. Pennsylvanis1: Deitsh. De Campa in Breefa fum Pit Schwefflebrenner un de Bevry, si Alty. Lancaster PA: Rauch and Cochran. Robbers, Karin. 1997. Non-finite Verbal Complements in Afrikaans: A Comparative Approach. Ph.D. dissertation, University of Amsterdam. Rutten, Jean. 1991. Infinitival Complements and Auxiliaries. Ph.D. dissertation, University of Amsterdam. Sapp, Christopher. 2006. Verb Order in Subordinate Clauses: From Early New High German to Modern German. Ph.D. dissertation, Indiana University. Schmid, Tanja & Vogel, Ralf. 2004. Dialectal variation in German 3--verb dusters: A surfuce-· oriented optimality theoretic account Journal of Comparative Germanic Linguistics 7: 235--274.

Schonenberger, Manuela. 1995. Constituent order in the VP: Verb raising and verb projection raising. In Topics in Swiss German Syt1ta.x, Zv1 Penner (ed), 3·17···41 1. Bern: Peter Lang. Vella Deitsh. 1997. Velia Laysa: Biwwel Shtoahrisja Kinnah (Let's Read: Bible Stories for Children). Sugarcreek OH: Schlabach Printers. Wurmbrand, Susi. 2006. Verb clusters, verb raising, and restructuring. In The Biackweil Companion to Syntax, VoL 5, Martin Everaert & Henk van Riemsdijk (eds.), 229-343. Oxford: BlackwelL Zwart, C. Jan-Wouter. 1996. Verb dusters in continental West Germanic dialects. In Mlcroparametric Syntax and Dialect Variatlotl [Current Issues in Linguistic Theory 139], James Black & Virginia Motapanyane (eds), 229-258. Amsterdam: John Benjamins.

185

Looking for order in chaos Standard convergence and divergence in Mennonite Low German* Goz Kaufmann Freiburg, Germany

In this paper I analyze the ongoing converging and diverging processes between Mennonite Low German (MLG) and Standard German in six Mennonite colonies in the Americas. 'Ihe statistically detectable differences in the informants' behavior with regard to five linguistic phenomena and the comparison between all colonies and different age -gender--subgroups in three of these colonies offer new and promising insights into the stru<..tural and sociolinbsuistic conditioning of converging and diverging.

1.

Introduction

Structural convergence and divergence between languages and language varieties have been on the research agenda for quite some time now, and the research to date has produced many valuable insights (cf., for example, the articles in Auer, Hinskens & Kerswill 2005). In spite of this, many problems have not yet been solved. Auer and Hinskens (1996: 4) mention some of these: Empirically rich, well--dol."Uillented and quantitative investigations of processes of convergence and divergence, and not just oftheir outcomes, are rare, investigations into the links between social changes and the linguistic developments they can trigger are even rarer. Our goal in the present study is to overcome some of these problems. Starting ·with the first part of Auer and Hinskens' critique, one should not forget that in the last few decades many quantitative studies have been published. The problem, however, is that these studies frequently show methodological and/or analytical flaws. I will discuss only two of them: firstly, tl1ere are still studies which do not test (all of) tl1eir

*I would like to thank Peter Auer, an anonymous re"iewer, and especially Claudia Bucheli Berger for many valuable hints and discussions.

188 GOz Kaufmann

results statistically leaving it up to the reader's good will to accept or not to accept supposed differences between different subgroups as "significant" (cf., for example, Woolhiser 2005 & Cheshire, Kerswill & Williams 2005). Secondly, a less obvious, but equally serious problem is the fact that many studies simply compare raw frequency data abstracting away from possibly influencing linguistic and sociolinguistic factors (ct:, for example, Woolhiser 2005: 252-253 and 256). If we do not control the linguistic context when dealing with context-sensitive structures and if we do not use a sociolinguistically balanced sample with regard to age, gender. social class etc., we are bound to skew the results of our analyses. In the study presented here, we carry out a rigorous quantitative screening of five linguistic phenomena in a data set of more than 14.000 sentences (cf. Section 3). This screening will always be done by means of statistical tests. Besides this, we will only use raw frequem:y data in strictly controlled contexts (cf. Section 5). Otherwise, we will use an index, which despite its simplicity will account both for different linguistic contexts and for a possibly unbalanced sample (cf. Sections 6.1 through 7). The second part of Auer and Hinskens' critique mentions "the links between social changes and the linguistic developments they can trigger:' I would like to comment on three points, which are connected to these links. Two of them are more general (a and c) and one is more methodological in nature (b): a. b.

c.

the question of whether and which contact situations can be compared in a meaningful way the question of how different linguistic phenomena in the same contact situation should be compared the question of what we can learn from the speaker's frequently heterogeneous behavior with regard to these phenomena

As for (a), Thomason and Kaufman (1988: 36) write: Then, under conditions of roughly similar intensity of contact within each type [borrowing and substratum interference, respectively; G.K.], we can begin to make general predictions about the kinds of expected interference, based on linguistic criteria like markedness and the typological distance between the source language and the language undergoing the interference.

Due to the fact that Thomason and Kaufi.nan (1988: 4) stress the utmost importance of social factors for the outcome oflanguage contact throughout their book, it may come as a surprise to the reader to learn that they compare, tor example, literary Indic-Dravidian language contact dating back thousands of years (idem: 79) with present--day Low German as spoken in Nebraska (idem: 81---82), allotting to both cases the label of slight structuml borrowing thus implying a "roughly similar intensity ofcontact." In our opinion, pairing such widely differing cases under the same heading

Looking for order in chaos 189

is problematic. Like Thomason and Kaufman, we will also compare different contact situations, but the comparison of six Mennonite colonies is far less problematic because the similarities between these speech communities are so big that the differences (the role of Standard German, f{)r example) can be used as tools of analysis. A second point on which we differ from Thomason and Kaufman is the direction of our investigation. Thomason and Kaufinan want to predict the linguistic outcome of language contact (intetjerence) by means of an analysis of the markedness of the phenomena in question and of the typological distance between the contact languages. We will replace this theoretical (deductive) approach by a predominantly empirical (inductive) one. In other words, in order to detect differences in markedness or cognitive complexity of linguistic structures, we will rely on the speaker's linguistic behavior, rather than on the linguist's pre--empirical analysis of notoriously elusive concepts. With regard to (b) and indirectly to (c), 'Ihomason and Kaufman (1988: 61) stress the necessity of comparing the development of several phenomena in the same contact situation: The appropriate methodology, then, requires examination of a conta<..t situation as a forest rather than as a collection of isolated trees. In order to support a claim that feature x arose in language A under the influence of language B, we need to show that features a, b, c, y, z at least some oi whid1 belong to a s"Libsystem different from the one x belongs to also arose in A under the influence of B.

As Thomason and Kaufman have already shown that linguistic influence is hardly ever restricted to a single and isolated phenomenon, we can take this point for granted. What we want to do is compare the infonnant's different behavior towards several contact phenomena in order to learn more about the nature of these phenomena (and about the "nature'' of the informants). Mattheier's (1996: 44) reasoning comes pretty close to this goal. He writes that accommodation differences between varieties in contact are "not affected simultaneously:" Diese Akk.ommodation erfaEt aber, nach allem, was wir wissen, nicht alle Differenzmerkmale der Varietaten gleicb:T.eitig, sondern folgt bestimmten, noch weitgehend unerforschtenlinguistischen Regelma6igkeiten. Die Aufgabe ist nun fur den Linguisten in erster Linie, die Merkmale zu bestimmen, bei denen der AkkommodationsprozeB beginnt, und dann die linguistischen ( und auch S07iolinguistische.n) Prozesse zu identifizieren, in denen diese Entwicklung zu einer anderen ode.r auch eiuer neuen Varietat ablaufen [bold print; G.K.).

Mattheier sees the reason for the lack of simultaneity in the working of"hitherto rarely investigated linguistic reguiarities" and wants to identify "the linguistic (and also sociolinguistic) processes leading to different or new varieties~ This resembles the major points in Auer and Hinskens's critique mentioned at the beginning. Kaufmann (2010: 489), too, makes a similar point: "[... ] in studies of convergence and non-convergence neither

190

Goz Kaufmann

ianguages nor speech communities should be analyzed as if they were monoiithic blocks" (cf. also Milroy 1992: 148). The reason for this demand is that convergence and divergence would only a1fect all structures and all speakers in the same way if all structures of the target language were of the same production and processing complexity and if the socio--psychological and sociolinguistic dispositions of all members of the con-verging or diverging speech community were identical. As this is probably never the case, there are two basic research questions which Kaufinann (2010: 482) formulates and which will also guide this study: d.

e.

\Vhat are the structural reasons for the fact that some linguistic structures remain (relatively) stable in a situation which is otherwise marked by converging structures? 1 \Vhat are the sociolinguistic reasons f{)r the fact that some speakers do not converge or even diverge from a prestigious source language in a generally converging group?

Besides the different complexity of linguistic structures and besides the differing dis-positions of members of a speech community, there are obviously other re-dsons which may explain the linguistically "chaotic" picture we often encounter in contact situations. Speakers might, for example, not be able to identify correctly the prestigious2 variants, or they might be concerned with the "purity" of the phonetics and the lexicon

1. Cheshire, Kerswill, and Williams (2005: 135) make a similar point: "[tjhe phenomenon of dialect convergence presents us with ar1 opportunity to examine ar1 issue that is not yet weil understood in variation studies: the extent to which iir1guistic variation in different components of language patterr1s in similar ways." However, they deal with phenomena of different linguistic

relevance thtLS compromising the validity of their far-reaching conclusions (idem: 166). The only example for syntax they (idem: 159-164) present are emphatic pronoun tags (ct: idem: 159, Extract 4 (line e): I haver1't even tried itme),and this seems to be a rather exotic phenomenon compared to the variables they are treating in phonology (idem: 145-149; fronting of It/ and /dl and pronunciation of /ail) and morphosyntax (idem: 150-154; for example, negative concord and relative what). 2. I will use the concept prestige in the "classical': i.e. Labovian sense. This means that I will connect change from above and the reactions to it with the existence of variants with an overt prestige. In so doing I do not follow Milroy (1992), who partly (but not completely [cf. idem: 160]), distinguishes prestige explanations from identity--based explanations. In our opinion, these two types of explanations complement one another. I also do not see why we should abandon the idea of prestige variants being predominantly used by women (at least intra-ditional, male--dominated societies like the Mennonite ones) in order to compensate for the impossibility of socially mo\'ing upward in other ways (cf. idem: 153). Men in those societies do not need such a "compensation" (cf. the analyses in Sections 6.1 through 6.3). Likewise, supra--local norms preferred by women, which Milroy alludes to (cf. idem: 157), are rather frequently prestige norms.

Looking for order in chaos

of their language but not "care" too much about the "purity" of its syntax. And last but not least, researchers might erroneously consider some variants as prestigious when they are not. The structure of the article is as follows: Section 2 will furnish the relevant information about the Mennonites; Section 3 will introduce the data set and Section 4 will describe the five phenomena investigated. In Section 5 the results of a balanced subset of the Mennonite informants ti·om all six colonies will be presented. Sections 6.1 through 6.3 'Vvill sharpen these analyses for the three colonies in Brazil and Paraguay adding detailed sociolinguistic interpretations. In Section 6.1 the new indexing method will be discussed. Finally in Section 7, the results of these three colonies will be compared, the potential of diverging and converging behavior as means of marking one's identity will be analyzed, and some theoretical implications will be formulated.

2.

The Mennonites

'Ihe ancestors of all Low German-speaking Mennonite communities in the Americas emigrated from Russia or the Soviet Union where Mennonites had lived since the end of the 18th century. The Mennonites originally came from east Holland, Frisia, Flanders, and what is today northwest Germany, where they had formed Anabaptist communities during the time of the Reformation. Due to religious persecution most of them emigrated to West Prussia during the 16th century where they first adopted the local variety of Low German and later High (Standard) German for religious and formal purposes. When the Prussian government started to impose stricter rules on the Mennonites in the 18th century, they started to look for other places to live and accepted an invitation by Catherine II of Russia to settle in the Ukraine. There they lived for a century in almost complete isolation, but at the end of the 19th century, Russian officials started to change their policies towards the Mennonites by intro-ducing laws to ensure a certain measure of integration. 'Ihis led the more conservative Mennonites to emigrate to Canada around 1870. \Vhen the situation tor German-speaking people in Canada also became more difficult during and after World "\tVar I, it was again the more conservative members who decided to move to Mexico, where most of them settled in the state of Chihuahua ( Ciudad Cuauhtemoc; today roughly 40,000 people). Others found a new home in Paraguay setting up the colony Menno (today 9,000 people). Mennonites from Mexico founded several sister settlements, namely Santa Cruz de la Sierra 3 in Bolivia (roughly 30,000 people),

3· Besides the Mennonites who came from Mexico, a smaller contingent left the Paraguayan colony Menno in the 1950s. The Bolivian informants in this project belong to this Paraguayan group and not to the Mexican one.

191

192

GOz Kaufmann

various communities in Belize, and one in Seminole, Texas, USA (4,000 people). The Mennonites who had stayed in the Ukraine in 1870 accepted the new situation and introduced a more elaborate school system sending future teachers to Germany to study there and sometimes inviting teachers from Germany to teach in their colonies. But in spite of their efforts and due to their economic success, these Mennonites taced severe problems after the foundation of the Soviet Union and especially when Stalin came to power. Around 1930 many of them wanted to leave the Soviet Union. Some of them managed to emigrate to Canada, some moved to Paraguay, where they set up the colony Fernheim (4,000 people) just twenty kilometers away from their breth·· renin Menno, and some went to Brazil, where they first lived in the state of Santa Catarina and later in Parana and Rio Grande do Sul (Col6nia Nova; 1,000 people). The changes betw·een 1870, when the first Mennonites left Russia, and 1930, when the second wave left the Soviet Union, are very important for our study, because the Mennonites who left in 1930 did so with a much higher degree of formal learning and a much better command of Standard German. And in spite of the fact that conservative Mennonites frequently express their aversion towards all things worldly, the more modern Mennonites often proved to have a strong (linguistic) int1uence on their more conservative brethren. This can clearly be seen with regard to the two colonies in Paraguay, where Fernheim had a strong influence on Menno (cf. for the lexicon Kaufmann 2003a).4

3·

The data set

The data on whic.h the follo~ing analyses are based were collected in order to learn more about the sequencing of verbal elements in clause-final clusters in Mennonite Low German (MLG ), i.e. three of the five phenomena analyzed are mere side·· products of this research design. The 313 intormants were asked to translate orally and on the spot46 stimulus sentences from English, Spanish, or Portuguese into MLG. They did not have access to a written version of the stimulus sentences. These sentences were created in order to allow the analysis of three independent linguistic factors, (a) the type of the finite verb, (b) the number of verbal elements, and for embedded clauses (c) the type of the embedded clause. For clusters with two verbal elements (Two vERBs) the goal was to elicit the temporal auxiliary han (English have) selecting a past

4· Another cultural and linguistic source of influence on Menno was exerted by a second more progressive Mennonite group which left the Soviet Union together with the German army at the end ofWorl.d War II and arrived in the Paraguayan Chaco in 1947. Before founding their own colony Neuland, these Mennonites spent some time in Menno (cf. Warkentin 1998: 307-308).

Looking for order in chaos 193

participleandmodalverbs selectinganinfinitivewithout tu (English to). For clusters with three verbal elements (THREE VERBs), the main focus was on the temporal auxiliary han selecting a modal verb with an infinitive. Besides these cluster-stimulating sentences, some stimulus sentences aimed at eliciting embedded clauses with a single verbal element frequently resulting in translations with finite dun(e) (English to do) plus infinitive (Do-suPPORT). The different cluster types were distributed over six isolated main clauses and four types of embedded clauses, namely relative clauses (RELATIVE), conditional clauses, causative clauses, and complement clauses. All main verbs in the stimulus sentences required a complement which frequently featured a definite article (ARTICLE). Many readers will rightly object that such a metl10d is unable to elicit natural language data (cf. the detailed discussion of this method in Kaufmann 2005). However, one should not forget tl1at the amount of free speech necessary to elicit enough comparable syntactic data of this kind from hundreds of speakers would be difficult to obtain. The informants were chosen in order to obtain enough tokens from two gender and three age groups. Although the Mennonite colonies are certainly not completely homogeneous with regard to social characteristics such as profession, education, and income, these characteristics do not appear to play such an important role as in modern western societies. For this reason, they were not used as an analytical factor. Church denomination, however, plays a role in Me>..ico and the United States. For these two countries the informants were additionally selected with regard to three types of churches.

4·

The five linguistic phenomena investigated

The tlve linguistic phenomena investigated 5 will be analyz.ed in order to place every Mennonite informant on a scale whose poles represent, on the one hand a variety without any trait from or similarity with Standard German and on the other hand a variety which uses Standard German variants exclusively or coincides completely with these variants. In order to do so the phenomena chosen had to fulfill three conditions: firstly, they had to appear in a robust number in the data set; secondly, they had to show variation in the colonies in Brazil and Paraguay; and thirdly, they had to

5· In Kaufmann (2004), three of these variables were already investigated for two colonies, Pernheim in Paraguay and Brazil. However, besides analyzing fewer phenomena and fewer colonies, that a1tide only worked with frequencies, ·which might have skewed some of the results, and did not compare the variables in a coherent way.

194 GOz Kaufmann

be marked by a clear distinction between the MLG non-standard variant(s) and the variant( s) coinciding with or being directly borrowed from Standard German. 6 Before we can introduce the five phenomena, however, we have to ask ourselves what "Standard German~' means to Low-German speaking Mennonites in North and South America. Put differently, does it make sense to speak about the occurrence of Standard German variants in MLG? With regard to the general role of Standard German, which in most Mennonite colonies is used in formal domains, the major difference is between the conservative Mennonites who left. Russia in the nineteenth century and the more progressive Mennonites who left the Soviet Union in the twentieth century. The conservative Mennonites, especially in Mexico, the United States, and Bolivia, employ a hagiolectal form of Standard German tl1at derives from older religious texts and is largely uninfluenced by the modem European standard variety. Due to this, it is not very probable that their MLG is influenced by this standard variety or that they share the European attitudes towards the linguistic variants in question. Indicative of this state of affairs is the fact that in the past many Mennonite teachers in these colonies were severely reprimanded when they tried to teach a slightly more modern type of German (cf. Sawatzky 1986: 247-250 ). The biggest struggle arose when some teachers wanted their students to pronounce the word Vater (English father) with a "standard" /a/ instead of the traditional Mennonite pronunciation /o/. Due to the lack of European Standard German influence, these colonies are not the primary focus of this article. For the Mennonites in Brazil and Paraguay, however, the role of Standard German is quite different. The ancestors of the Brazilian Mennonites and the Mennonites in Fernheim not only left the Soviet Union with a rather solid competence in European Standard German, but they also spent on average one year in Germany itself (between 1929 and 1931; ct: Klassen 2001: 94-108) before leaving for South America. And unlike the Brazilian Mennonites (ct: Section 5), tl1e Mennonites in Fernheim have never lost tl1eir close contact to Germany and Standard German. ln their first years in Paraguay, they used school materials they had brought with them from Germany (cf. Warkentin 2007: 42) and during the time of the Third Reich (cf. Warkentin 2007: 44; and for Hitler's impact on Fernheim Kaufmann 2004: 262-263) they received support and donations from the Deutsche Schulverband in Paraguay and other German organizations. Since World War II they have remained in close contact with the German embassy in Asuncion and especially with the Zentralstelle fur das Auslandsschulwesen

6. With this third condition we automatically exclude more indirect cases of influence which Thomason and Kaufman (1988: 61) describe: "It has sometimes been claimed that a

particular change cannot be due to foreigtz itzterference ber. ause the putative source language does rwt exhibit exactly the same structure that ht:JS beer1 inrwvated:'

Looking for order in chaos 195

(Central Agency for Schools Abroad), which was founded in Cologne in 1968 and which has sent teachers from Germany to all Mennonite colonies in the Paraguayan Chaco ever since (cf. Warkentin2007: 46; Warkentin 1998: 297-298). Enlightening in this respect is the entry for "Deutschlehrer" (teachers of German) in the Lexikon der Mennoniten in Paraguay (2009: 107---108) in which Warkentin writes: Dennoch bleibt Deutsch cine wichtige Unterrichtssprache, denn sic ist neben >Plattdeutsch in den> Mennonitenkolonien die Umgangssprache. Gottesdienste, Jugendveranstaltungen und Kulturabende werden in der Regel auf Deutsch gehalten. Der Schriftverkehr wird noch gr6Htenteils in deutscher Sprache gefill111. Da viele der mennonitischen Siedler noch nie in Deutschland gewesen sind, unterscheidet sich ihr Hochdeutsch teilweise von dem in der Bundesrepublik Deutschland gesprochenen. Es ist daher wichtig, class die Deutschlehrcr in den Sekundarschulen nach Moglichkeit ein oder mehrere Jahre in der Bundesrepublik gclebt und studiert habcn. Auf diese Weise wird der Kontakt mit dcr Sprache des Mutterlandes aufrecht erhaltcn. Von ganz groHem Wett sind in dieser Hinsicht auch Lehrer au s Deutschland, die als Auslandsdienstlehrkriifte oder Programmlehrkrafte einige Jahre in den Mennonitenkolonien unterrichten. Sic verbessem als native speakers nicht nur die Aussprache der Schiller, sondern enveitern deren Wortschatz [... ] [bold print; G.K.].

Not only does Warkentin refer to "Deutsch" (German from Germany) as a "wichtige Unterrichtssprache" (important language of instruction) and, even more importantly, as an "Umgangssprache" (vernacular language) in the Mennonite colonies in Paraguay; he also stresses the point that Mennonite teachers of German in secondary schools ("Deutschlehrer in den Sekundarschulen") should spend one or more year(s) studying in Germany in order to improve the students' "Hochdeutsch" (Standard German), which differs partly ("unterscheidet sich [... ] teilweise") from the standard variety spoken in Germany, as well as to maintain contact with the language of the mother country ("Kontakt mit der Sprache des Mutterlandes"). Besides this he again mentions tl1e importance of the constant presence of teachers ti·mn Germany, who as native speakers play a decisive role in improving the students' ".4.ussprache" and "T,:\lort~chatz" (pronunciation and vocabulary). In light of these facts we feel justified to speak of possible influences of Standard German (variants) on MLG. 1his is definitely valid for all informants in Fernheim and for the oldest informants in Brazil. One of them was born in 1924 and two in 1938, i.e. at a time when European Standard German still played an important role in Brazil. Since the oldest infl)rmant in Menno was born in 1941 and the second-oldest in 1946, we are also confident that all informants there entered a school system which had already been strongly influenced by the contact with Fernheim (and the founders of Neuland; cf. Footnote 4), i.e. by the contact with modern approaches to studying European Standard German (cf Section 6.1).

196 GOz Kaufmann

At this juncture, we can now introduce the five phenomena: The first phenomenon is the so-called DO-SUPPORT, the use of MLG dun(e) (English to do) as an auxiliary. This lexical insertion occurs in practically all German dialects, but has an extremely low overt prestige in Standard German? In comparison to Standard English, the use of dun(e) in MLG is not emphatic as in I do love you. It rather seems that there are aspectual and/or syntactic reasons leading to its insertion. Langer (2000: 295) sees in Early New High German tun a polyfunctional auxiliary. Louden (1992: 221-224) categorizes the Pennsylvania German due as a marker of iterative/habitual aspect, but also mentions the possible syntactic function of maintaining the typical German OV-structure by inserting finite due in non-introduced clauses. Such an insertion keeps the main verb in its original clause-final position (e.g. Rhenish Franconian: Isch du dem Mann !li"ET helfe (I do the man NOT help == I won't help the man) instead of Standard German Ich helfe dem Mann NICHT). 8 Since our data was elicited focusing mainly on verbal clusters in embedded clauses, most matrix clauses feature semantically simple verbs like the copula which are not compatible with DO-SUPPORT. Because of this, the analysis of DO-S"lJPPORT in the colonies in Brazil and Paraguay is carried out for six embedded clauses9 which contain only one verb in the English, Spanish, and Portuguese stimulus sentences. An example from the data set is given for Fernheim: (2) is the standard variant without dun(e), (3) the non-standard variant with dttn(e). ( 1)

No ves que estoy prendiendo la luz?

(Don't you see that I'm tuming on the light() (2)

Standard

F-1-3

Siehts du nich, daut ik daut Iich anmoak?

(3)

Non-Standard F-5-39 Siehts du nich, daut ik daut Licht anmake du?

The second phenomenon is the variation in the masculine definite ARTICLE in noun phrases serving as direct objects. 1he Mennonites in the colonies in Brazil and Paraguay use basically two forms, namely den, which in Standard German represents accusative case, and dem, which in Standard German represents dative case. As many

7. But compare the standard use of tun as a pro-verb in clauses with a topicalized main verb as in Schlagen tue ich den Hund nicht (Beat do I the dog not= I will definitely not beat the dog).

Besides redudng the structural distance from "deep" to "surface structure': there exists another cognitive advantage with regard to thl-s auxiliary. In the negated Rhenish Franconian main clause, the sentential negation particle net appears directly before the negated main verb helfe. In Standard German nicht has no superficial connection to helfe because this finite form must be moved to second position (head of CP).

8.

9· As duf!(e) in embedded clauses is not needed to keep the main verb at the end of the clause, its rather frequent occurrence in embedded clauses could be a case of syntactic analogy.

Looking for order in chaos 197

Low German dialects in Germany have just one oblique case generalizing either the form of the dative or the accusative (cf. Fleischer 2004: 215), there are two ways to explain this variation. It could either be proof of a not yet completed generalization (cf Thiessen (2003: xix), who for MLG in Canada indicates variation between dem and den for the masculine object article and between dem and daut for the neutral object article) or the consequence of contact with Standard German. This contact could have introduced a new form into an already generalized system leading to linguistic insecurity (cf. also the more detailed explanations in Section 5). As the colonies in Brazil and Paraguay have strong contact with Standard German and as Standard German does not allow dem in direct objects, only den as in (5) (332 tokens out of a total of 534 tokens; 62.2%) was considered a standard variant Dem as in (6) (177 tokens; 33.1%) or otherwise deviant forms like reduced de in (7) (21 tokens) or daut (4 tokens; cf. for these two forms in the "Iexan colony Kaufmann 2008) were consid-ered non-standard. Five clauses were used for this phenomenon. 1he example comes from the Brazilian colony: (4)

Tern certeza que ele concertou a cadeira?

(Are you sure that hes repaired the chair?) (5)

Standard

B-1-6

Bis du di sichet; daut heiden Stuhl haft torechtgemoakt.

( 6)

Non --Standard B -l-7

Bis du di sicher, daut her dem Stuhl torechtgemoakt haft.

(7)

Non-Standard B-7 -50 Bis du sicher, daut hei de Stuhltored1tgemoakt haft.

The third phenomenon concerns the introductory element in relative clauses (RELATIVE). Standard German uses relative pronouns which coincide in gender and number with the noun phrase they refer to and are case-marked in compliance with their syntactic role within the relative clause leading, for example, to das ist (sind) der 1viannldie Frau!das Kind!(die Kinder), der!die!das!(die) mich gestern besucht hat (haben) (=this is (these are) the man/the woman/the child!(the children) who visited me yesterday). Many Upper and Middle German dialects use this relative pronoun together with the particle wo, which formally coincides with the interrogative pronoun (English where). This leads to a rather complex relative marking with der wo/ die wo!das wol(die wo). But in almost all of these dialects it is also possible to drop the relative pronoun, marking relativity just by the use of the particle wo. MLG has a similar strategy normally marking relativity by the invariant particle waut, which also coincides with an interrogative pronoun (English what). In the colonies in Brazil and Paraguay 973 out of 1170 tokens used as subject or direct object (83.2%; nine clauses) were marked in this way. The relative pronoun der!die/dauti(die) was used in 164 tokens (14%). Other forms include waut da (10 tokens; cf. Fleischer 2004: 224),

198 GOz Kaufmann

wont (10 tokens), and der waut (3 tokens; cf Fleischer 2004: 219). Translations of a relevant stimulus sentence are given for Menno: ( 8)

Quien es el hombre que podria haber salvado la vida de mi hermano?

(Who is the guy who could have saved my brother's life?) (9)

Standard

M-1--8

Wer is die Person, die daut Lewe von min Bmder hat konnt rate?

(10)

Non-Standard .M-2-16 Wer is der Mann, waut min Bmder sin Lewe hat rate konnt?

(11)

Non--Standard M-5--34 Wer is der Mensch, der waut mine Sister ihr Lewe gerat haf?

All relativizers except (9) were counted as non-standard variants. That the relative pronoun in (9) is directly borrowed from Standard German and not the consequence of an internal language change in MLG can be shown by two facts: firstly, Fleischer (2004: 218-219) mentions East Pomeranian (a variety close to the base varieties of MLG), where the relative pronoun is only attested in the oblique position, i.e. very low in the accessibility hierarchy (and not in higher positions like subject or direct object as in MLG); and secondly, the introduction of the relative pronoun in MLG does not follow the path one would expect in a more internally caused language change (cf. Fleischer 2004: 230), i.e. the innovative and with regard to parsing less complex [+case] strategy is not more frequent in lower positions of the accessibility hierarchy than in higher ones. In the stimulus sentence 1his is the journey 1 am inviting my mother on, which has the relativizer in an oblique context and, therefore, was not used for the analyses in this article, the relative pronoun (with prepositions) occurs in 11.4% of the 131 South American tokens, while for the subject and object ftmctions analyzed here, it occurs in 14% of the cases. Even between these two structural cases, there is a significant difference in the "'""Tong" direction in some of the colonies. In Femheim, the standard relative pronoun is used in 29.8% of the tokens as a subject and in 19.2% as an object (p = 0.027*Nalue 4.8/df 1 [in the oblique function 22.2%]), 10 in Brazil

to. For nominal scale variables like token frequencies, Pearson's Chi-Square was used. For interval scale variables like competence and use of the contact languages, a One-Way ANOVA was used (if more than one independent variable was tested a General Factorial ANOVA was applied). "Ihree values will be presented: the significance of the calculated value (Value or Value F), the value itself, and the degree of freedom of the calculation. Correlations between interval scale values were calculated with a Bilateral Pearson Correlation. In all indications for the level of significance, one asterisk *mean.<; that the probability for a Type I error is between I% and 5% (0.01 s: p s: 0.05). 'I'wo asterisks *'·mean that this probability is smaller than 1% (0 < p < 0.01 ), and three asterisks **"that it is virtually 0% (p 0). One asterisk

Looking for order in chaos 199

the percentages are 8.4% for subjec..t and 4.7% for object (p = 0.099("")/Value 2.7/df 1 [oblique 1.8%]), and in Menno the difference is not significant (15.4% for subject; 11.4% for object [oblique 16.7%)).11 'Ihe fourth phenomenon, the sequence of TWO VERBS in clause-final dusters in embedded clauses, has already been extensively studied (cf. for a detailed analy-sis of MLG Kaufmann 2007, and in general EvERS 1975; L5tscher 1978; Haegeman & van Riemsdijk 1986, and Koopman & Szabolcsi 2000 ). Standard Gennan only allows a strictly left-branching, non--raising sequence of finite and non--finite verb as in Wenn er DAS nAus jetzt verkaufenv2 muflv 1 [ ... ](If he THE HOUSE now sell must [... ]=If he has to sell the house now[... ]). MLG- as many West Germanic continental varieties - has two more variants, which are traditionally called Verb Projection Raising (Wann hei nu mut:.1 DArJT nOs verkope,2 { ... ])and Verb Raising (Wann hei DAUT Hus nu. mutv1 verkopev2 [ ... ]). 'Ihese two non--standard variants are distinguished by the position of the complement which may (Verb Raising) or may not scramble (Verb Projection Raising). Eleven clauses were selected for this variable. As example, the translations of a stimulus sentence in Brazil are given; (13) counts as standard variant; (14) and (15) are non-standard variants. (12)

Se ele tiver que vender a casa agora, ele vai ficar muito triste.

(If he has to sell the house now, he'll be very sorry) B-1--3

Wann hei SIN Hus nu verk()pe mut, wird hei sehr tril.rig were.

(13)

Standard

(14)

Non--Standard B- 7--52 Wann hei nu mut DAUT Hus verkt}pe, wird her sehr tril.rig bliewe.

(15)

Non-Standard B-l-4

Wann hei DAUT Hus nu mut verk()pe, wird her sehr tril.rig sene.

The fifth and final phenomenon, the sequence of THREE VERBS in clause-final clusters in embedded clauses, is separated from TWO VERBS, because the processing of a cluster with three verbal elements is more difficult than the one with two ver·· bal elements. Kaufmann (2007: 185--·193) could show that there exists a structural

in brackets (• l indicates a statistical tendency where the error margin lies between 5% and 10% (0.05 < p < 0.1). n. It goes without saying that a detailed analysis of the more than 3.000 tokens for relative markers elicited in the Mennonite colonies would be highly interesting for typological research even though indirect objects were not elicited. Pleischer (2004: 233) writes, for example, that in YiddiSh "[a]s a last step, was [vos in Yiddish; G.K.] developed into a more ger1ercd subordination marker:' a strategy also extant in MLG (79 tokens for waut instead of daut in object clauses).

200

Goz Kaufmann connection between the two phenomena, but we will presently see that due to different levels of complexity this does not mean that all informants behave in a similar way. Even in Standard German, the higher complexity of THREE VERBS leads to obligatory, but rather uncharacteristic cases of partial verb raising in the context of modal verbs in the perfect tense: wenn er (DIESEN BRIEF) wirklich hatv 1 (DmSEN BRIEF)

schreibenv3 wollenv2 [... ] (If he (1'HIS LE1'TER) really has (THIS LETI'ER) write want .... ] = If he really had wanted to write this letter [... ]) instead of expected Wenn er DIESEN BRIEF wirklich schreibenv3 gewolltv2 hatv1 [ ... ]with the governing verb always to the right of the governed verb. 12 Besides these two standard variants (cf. (17) and ( 1.8) below), there are four non-standard variants in MLG. Three of them, i.e. (20) through (22 ), are characterized by two cycles of verb raising, leading to superficially strictly right.. branching sequences: wann hei (mESEN BRIEF) wirklich hatvl (I>IESEN BRlEP) welltv2 (DIESEN BRlEF) schriewev3 [ ... ]. Like in TWO VERBS and like in (17) and (18), tl1e complement may or may not scramble in these variants. The sequence in (19) is one in which tile finite verb appears in the middle of the two non-finite verbs leading to the sequence wa.nn hei DlESEN BRlEF wirklich schriewev3 hatv1 welltv2 [ ... ],which Kaufmann (2007: 189) does not analyze as the result of verb raising but of head movement of welltv2• Five clauses with three verbal elements could be used for the colonies in Brazil and Paraguay. The translation examples come from Menno: (16)

Si realmente hubiera querido escrihir esta ca1ta, habrfa enc.ontrado tiempo.

(If he really had wanted to write this letter, he would have found the time) (17)

Standard

M.S..39 Wann her wirklkh hat DIE SEN BRIEl' sdtriewe wollt, wii.rd her die Tied datu gefunge hawe.

(18)

Standard

M.. }.. 2

(19)

Non .. Standard M· 2.. 12 Wenn hei wirklich DEN BRIEl' schriewe hat wollt, dann hat her Tied gehat.

(20)

Non-Standard .M-4-27 Wann hei -wirklich hat wellt DIESEN BRIEF schriewe, wiird her ook Tied gehat hawe datu.

(21)

Non-Standard M-2-14 Wann dei wirklich hat DIESEN BRIEF wolh schriewe, wird her Tied gefunge datL1 hawe.

(22)

Non .. Standard M-3 .. 19 Wannhei DIESEN BRIEl' unbedingt hat wollt schriewe, hat her ook Tied gekreie.

Wann hei DIESEN BRIEF wirklkh hat schriewe wollt, dann wilrdhei die Tied ook geftmge have.

The second unexpected fact in Standard German is the appearance of the infinitive of the modal verb functioning as a past participle (wolle11 instead of gewollt). This well ..studied, but still not completely understood phenomenon, normally called infi11itivus..pro.. participio (cf., for example, Schmid 2000 or Vanden Wyngaerd 1994), does not exist in MLG. MLG uses the past participle of modal verbs, albeit without the typical German prefix ge-·. 12.

Looking for order in chaos 5·

General analysis of all six colonies

In this section I present data from all six Mennonite colonies in order to show the general converging tendencies of MLG in the colonies in Brazil and Paraguay where Standard German is not just a hagiolect. Table 1 gives an overview of some sociolinguistic facts in the colonies. For the duration of schooling, competence in and use of the contact languages, a balanced sub-sample for each colony was randomly chosen with regard to gender (2 levels), age (3 levels), and in the USA and Mexico with regard to church denomination (3levels). For the USA and Mexico nine informants for each age-gender-subgroup were drawn (3 for each of the 3 church levels; ideally a total of 54 informants), in Paraguay and Brazil a maximum of five informants for each of the six age-gender-subgroups could be selected.B In the case of Bolivia, the least intensively studied colony, all informants were included. The reader should keep in mind that all Bolivian informants are men and that just one of them belongs to the younger age group. Table 1. Language constellation in six Mennonite colonies

(Stand= Standard; Maj.= Majority) USA

Mexico

Bolivia

Brazil

Menno

Fernheim One-Way AN OVA

INFORMANTS

n-informants

52

51

8

30/25

30

30

201

SCHOOL

in years

8.4

7.9

p = 0*** 12.4 7.9____,-___,;;=.:..-___:.:::..:..:...___J_~.:....__

___:__ _ ___r__ _ _ _ _ _

COMPETENCE

Low German

p = 0***

StandGerman

p = 0***

Maj. Language

p = 0*** USE

Low German

19.5

p = 0***

Stand German

0.3

p = 0***

Maj. Language

p = 0***

With regard to language use this number could not be reached in Brazil due to a lack of data (25 instead of 30 informants). For some informants in Brazil we unfortunately do not have all the sociolinguistic information we would like to have (c£ also Table 5).

13.

201

202

Goz Kaufmann

In 1able 1 competence in the contact languages is measured with a 14-point-scale and language use on a 30-point-scale. In comparison to the results of the five linguistic phenomena discussed in Table 2, these data are subjective in nature. 14 All seven complexes are significantly different on an extremely high 0***-level. In view of these differences, one might speak of currently rather unrelated speech communities, although all of them share part of their history and many of their members still maintain strong (family) links to members of other colonies. With regard to the basic language constellation, three types of speech communities can be distinguished. All of them share one characteristic, though: MLG is still the most important in--group language, both with regard to linguistic competence and language use. 1he most "traditional" colonies are Mexico and Bolivia. In Me.xico competence in MLG is ahead of the second strongest language (Standard German) by 7.5 points (Bolivia 6.2 points ahead of the major-ity language Spanish), while use of MLG is completely unrivalled in both colonies (Me.xico 21.6 and Bolivia 23.7 points ahead of Spanish). This "traditional" outlook is also confirmed by the duration of schooling; Mexico and Bolivia have the lowest figures in this respect Different from these two colonies, Standard German has impor-tant functions in the two Paraguayan colonies, where it is already infringing on MLG. In Fernheim, competence in MLG is ahead of Standard German by only 1.1 points (Menno 2.7), while use of MLG is still dearly ahead of Standard German (12.1 points in Fernheim and 17.2 points in Menno). Both colonies have a comparable and above average duration of (German) schooling, which explains their high values for Standard German. The situation for MLG is most precarious in the USA and in Brazil, because in these colonies not only competence in MLG but also its use is threatened, but this time the threat does not come from Standard German but from the respective majority language. In Brazil, MLG is ahead of Portuguese by 1.3 points in compe-tence and by only 6.4 points in language use; in the USA, MLG is ahead of English by 2.2 points in competence and 9.6 points in use. In spite oftl1ese similarities, there

With regard to competence, two questions were asked. The first one referred to an informant's s-ubjective evaluation of his competence in each language. Did he consider his competence as very good (7 point'i), good (five points), OK (3 point'i), bad (1 point) or non-existent (0 points)? For the second question the informant had to rank the languages he knew with regard to his competence. Which language did he speak best (7 points), second-best, thirdbest (the values for intermediate languages depended on the nuniber of languages named), and worst (1 point). The two questions correlate strongly in all colonies; a fact which heightens our confidence in the reliability of the data (cf. for more detail'> Kaufmann 1997: Chapters 6.3.1.1). With regard to use, most questions referred to domains and interlocutors within the ethnic group, especially within the informant's family. If only one language was named for one type of domain or interlocutor, this language received 5 points; if two languages were named without further qualification, 2.5 points were allotted (cf. for more detail'i Kaufmann 1997: 6. 3.1. 3). 14.

Looking for order in chaos exists a big difference between the two colonies which has to do with the fact that the US-American Mennonites belong to the conservative Russian-Canadian and notlike the Brazilian Mennonites - to the more progressive, exclusively Russian group. 1he ancestors of the BrciZilian Mennonites reached America with a more elaborate idea of schooling and higher cultural aspirations. Mennonite schooling in Brazil is, there-fore, still of a much higher standard and lasts longer than in the USA. The fact that this high quality of schooling in Brazil has not led to better results in Standard German has to do with the aftermath of ousting Standard German from all schools in Brazil during the time of the F..stado Novo (1.937---1.945). 15 Only in the 1970s did German re-enter Brazilian schools and it has never gained the importance it used to have. The linguistic consequence of the banning of Standard German from Mennonite schools in BrciZil can be clearly seen when one compares these Mennonites with their brethren in Fernheim. There, German was not banned during World War II (cf. Warkentin 2007: 44-45), and consequently Standard German plays a much more important role (cf. Kaufmann 2004). As for the five linbruistic phenomena, it was neither possible nor necessary to consider the factor church in Mexico and the USA. 16 Otherwise, a similar procedure was applied, i.e. the tokens of five informants for each age-gender-subgroup and each clause were selected; in Bolivia, again, all tokens available were selected. Because of the fact that random sampling was carried out for each clause separately, the tokens selected were not from the same speakers for all clauses (different from Table l where all data came from the same informants), i.e. we are analyzing tokens rather than informants. As we are comparing six colonies in Table 2 and not just the ones in Brazil and Paraguay, the number of eligible clauses for each phenomenon had to be reduced. For TWo VERBS seven clauses could be compared (ideally 210 tokens per colony [6 age-gender-subgroups/5 informants!? clauses]), for RELATIVE markers six clauses (ideally 180 tokens), for the definite masculine ARTICLE five clauses (150 tokens) and for THREE VERBS and Do--suPPORT four clauses (120 tokens).

Portuguese was declared the sole language of instruction in 1937; teachers who were not able to teach in this language lost their jobs. In 1942, eventually, the use of German in the public sphere was also forbidden (cf. Doll2002: 22-23).

15.

'Ihis distinction is no longer necessary because only two of the five phenomena (Doand TWO VF.RBS) show anything close to a s-ubstantial amount of variation in Mexico and the USA. It is not possible because due to the low level of proficiency in the majority language many of the most conservative Mennonites in these two countries (especially older and middle-aged women in Mexico) were not able to do the (complete) translation of the stimulus sentences.

16.

SUPPORT

203

204

GOz Kaufmann Table 2. Results for five linguistic phenomena in six Mennonite colonies USA

Mexico

Bolivia

Brazil

Menno

TWO VERBS (n)

208

210

53

210

159

163

Standard(%)

25.5

THJU!J! VERBS (n)

116

110

22

101

82

80

3.

0

RELATIVE (n)

180

180

46

2.2

2.2

Do-suPPORT (n)

120

120

30

Standard(%)

61.7

ARTICLE(n)

148

134

39

138

11.5

3

5.1

5.1

0

0

10.3

26.8

dem (%)

511 p = 0***

180

170

180

117

114

117

936 p = 0***

618 p = 0***

Standard den(%) de/daut(%)

1003 p = 0***

Standard(%) Standard(%)

Fernheim Chi-Square

127

128

714

67.7

52.3

p = 0***

4.7

6.3

27.6

41.4

Table 2 shows us that not only the role of the contact languages in these six colonies is far from identical, but also that the variety of MLG spoken differs greatly. All five phenomena analyzed show extremely significant differences. Four of the five phenomena behave in a similar way; the only exception being ARTICLE. Femheim leads in the use of the standard variant(s) in three of the four phenomena (the exception being THREE VERBS where it comes second). On three counts Menno comes in second and once in first place. The Mennonites in the United States are on three counts furthest away from the standard variant(s) (the exception being RELATIVE markers where Bolivia and Mexico show marginally lower percentages). Brazil is in third place three times and once in fifth (Do-sUPPORT), while Mexico comes fifth twice, once fourth and once last Bolivia comes in fourth place twice, once third and once last. If one compares these results with the indices for competence in and use of Standard German in Table 1 one will undoubtedly see linguistic convergence at work. For both competence in and use of Standard German, Fernheim comes first, Menno second, Brazil third, and Mexico fourth . Only Bolivia and the USA change places at the bottom end of these indices. The coincidence with the sequences of the four linguistic phenomena is almost perfect. Also with regard to the fact that Menno shares its conservative Russian-Canadian origin with the colonies in the USA, Mexico, and Bolivia and Fernheim shares its more progressive, exclusively Russian origin with Brazil, these results show an influence of Standard German on MLG. The Brazilian colony lost its Standard German roofing during the time of the Estado Novo and, consequently, its

Looking for order in chaos

dialect variety has lost many Standard German characteristics. Menno, on the other side, gained Standard German roofing via the intluence from Fernheim, and its dialect variety shows strong Standard German traits. The exception to this otherwise clear-cut picture is ARTICLE (cf. also Kaufmann 2004: 292---296). Here, surprisingly, it is the most Standard-German--like colonies in Paraguay which show the highest frequency of non-standard dem. This does not mean, however, that the colonies in Mexico, the USA, and Bolivia are converging to Standard German with regard to this phenomenon. On the one hand, we can see that there is some deviation from den in the USA. The difference to the situation in Paraguay is that this deviation does not lead to dem, but to daut and especially to de. In the USA seems to exist more of a "gender" than of a "case problem'' (but cf for a more detailed analysis Kaufmann2008: 93---96). On the other hand, if we take a closer look at indirect objects, we would get a completely different picture. In this case, the Paraguayan colonies are closer to Standard German. 1his has to do with the tact that the colonies in Mexico and the USA have generalized den as the masculine object article. In Paraguay, therefore, we are either seeing a generalization of dem or a transition period on the way to a more standard-like distribution of den and dem. Dem would then be a more or less recent innovation leading to insecurities in its correct application. Another interesting point is that with regard to the other four linguistic phenomena we may claim that the fact that Menno is more standard-like than its sister colonies in the USA, Mexico, and Bolivia is the consequence of Standard German int1uence. With regard to the mascu-line definite ARTICLE, however, one is tempted to say that it is not so much Standard German, which has influenced the dialect variety in Menno, but the MLG variety spoken in Fernheim. After all, by using dem the Mennonites in Menno are diverging from Standard German and converging to Fernheim MLG (cf. for a comparable triple-con-stellation the situation in Lake Constance Alemannic in Auer & Hinskens [1996: 14]). Granted, one could also think that dem was introduced into Menno MLG via Standard German and that the informants in Menno also show a kind of hypercorrection. A counter-argument to this hypothesis can, however, be found in Bolivia. The ancestors of the Bolivian informants left Menno precisely at the time when the Mennonites in Menno decided to improve schooling (cf. Footnote 3). This means that the Bolivian Mennonites did not have any contact with more modern ways of learning Standard German, but they had had more than twenty years of contact with Fernheim. And this still shows in their low, but nevertheless existing use of dem (as opposed to the Mennonites in the USA and Mexico). Interestingly, loosing Standard German as the Brazilian Mennonites did seems to be equivalent to loosing dem. A possible reason for this might be due to the fact that den marks the more frequent case accusative or that n is by far the most frequent, thus unmarked nasal in German varieties (cf. for Standard German Kohler 1995: 222).

205

206

Goz Kaufmann Some more comments need to be made in order to understand the nature of variation with regard to the five linguistic phenomena. The highest degree of variation between the colonies is found with regard to TWO VERBS (70.2% between the highest and the lowest value f{)r the standard variant), the lowest with regard to DO-SUPPORT (30.6%) and REI..ATIVE markers (26.1 %). But whereas DO·· SUPPORT or rather the lack of it shows little variation at a high level, RELATIVE markers show little variation at a low level; relative pronouns are virtually nonexistent in the USA, Mexico, and Bolivia. The same is true for the standard variants with regard to THREE VERBS. Likewise in Mexico and the USA, the use of dem is nonexistent. One may, therefore, conclude that relative pronouns, only partially raised variants with regard to THREE VERBS and dem in direct objects are not native to MLG. Where they occur, they were probably borrowed from Standard German. For DO··SUPPORT and Two VERBS, on the other hand, one may conclude that all variants already existed in MLG before contact with Standard German. In this case, the variant(s) coinciding with tl1e Standard German variant(s) may just have increased its/their prestige after contact (cf. Mattheier 1996: 34). As tor complexity and markedness, a look at the Paraguayan colonies is enlightening. The informants in these cdonies almost succeeded in suppressing the non-standard variants with regard to TWO VERBS and DO-SUPPORT (less than 10%), but they use the standard variants in THREE VERBS less than half of the time and RELATIVE pronouns in less than 30% of the cases. 'Ihis hierarchy fits the expected scale of comple.xity involved in each of these phenomena. While there are only three extant variants with regard to TWO VERBS and just one of them is allowed in Standard German, there are sil!: possible variants with regard to THREE VERBS and Standard German allows some variation in this respect, i.e. two of the six variants are considered standard; a fact, which besides the higher number of verbal elements, may complicate the identification of the stan·· dard variants. The lower occurrence of standard variants with three verbal elements is tl1en the consequence of a higher complexity. The introduction of relative pronouns into a variety featuring a single relative particle seems to be even more complex (if not tor listeners, certainly for speakers), because the form of the relative pronoun depends both on elements in the same clause (case) and on an element in the matrix clause (gender, number). In comparison, the suppression of dun(e) does not create any processing problem, because firstly, an element is dropped and not added, and secondly, in spite of the fact that this suppression causes head-movement of the main verb, this movement is nothing new for MLG speakers. The values for the standard ARTICLE form den lie in the middle of this hierarchy. This intennediate position can be explained by two opposing factors: on the one hand, the use of dem instead of den does not change anything with regard to the syntactic processing complexity, because itis a kind of par·· adigmatic morphological exchange with no syntagmatic consequences; on the other hand, with regard to dem and den the speaker has to be aware of the correct context in which to use either form, because both fonns exist in Standard German.

Looking for order in chaos

6.

Specific analyses of the three colonies in Brazil and Paraguay

Many of the Mennonite informants in Paraguay have spent part of their life outside of the colonies investigated, mostly in Canada, Germany, or the Paraguayan capital Asuncit'in. 'lhe average percentage of the informants' life-time spent outside the colony is 14.3% for the informants in Menno and 13.9% for the informants in Fernheim. The reasons for this temporary emigration are in the case of Canada the frequent economical crises in Paraguay and the fact that many Mennonites especially in Menno still possess a Canadian passport In the case of Germany and Asuncion the reasons are studying and ,.vorking in Mennonite-related businesses, respectively. A look at the results of these partly non-local informants shows that their more complex biographies have had an influence on their linguistic behavior. With regard to the general sum of all five phenomena (cf_ for this sum the explanations in Section 6.1), the infor-mants from Menno who have spent more than 10% of their life in Canada rank on places 3, 12, 24, 25, 27, 30, 33, 36, 38, 39, 40, i.e. five of the eight most non-standard informants (among them the three most reluctant users of standard variants) belong to the "Canadian" subgroup. 17 In Fernheim the result is similar: the ranks of the seven "Canadian" Mennonites among the 37 informants are 5, 7, 24, 25, 30, 36, 37, i.e. three of the seven most non- standard informants (among them the two most reluctant users of standard variants) belong to this subgroup. In Menno the ranks of the "German:' and "Asunci6n"-Mennonites are 6, 8, 10, 12, 21, 30, 32; in Fernheim l, 3, 9, 10, 11, 12, 21, 31, 34. Although this distribution is less one-sided than the one of the "Canadiali'' group, these informants were likewise not considered because they definitely had much more contact with Standard German than the other informants (but cf. the results in Table 6). As the Brazilian informants for which we have the necessary information (cf. Footnote 13) have spent on average only 5.8% of their life outside the colony, i.e. much less than the Paraguayan informants, and as there is no clear pattern for the place of stay outside the colony, all Brazilian informants were included in the f{)llowing analyses. 6.1

Menno, Paraguay

Since the role of Standard German in the colonies in Menno and Brazil started changing roughly sixty years ago, it is a reasonable hypothesis that the consequences of these changes should still be visible in the MLG varieties spoken there today. Table 3 shows the results for the six age-gender-subgroups in Menno. As we compare subgroups of the same colony in the follovnng analyses, we will include the data of all informants

"l\¥o of the 42 informants in Menno had less than fifteen tokens for the five phenomena and were, therefore, excluded from this analysis.

17.

207

208 GOz Kaufmann Table 3. Language constellation and results for five linguistic phenomena and their sum for six age-gender-subgroups oflong term residents in Menno (inform.= informants; Stand. = Standard) young men

young women

middle men

middle women

older men

older One-Way women ANOVA

4

4

4

23

13,3

13,8

12,3

ns

12.5 10.5 7.5

13.5 10.2 8.2

13.5 9.8 9.1

ns ns ns

22.1 3.3 4.6

25.8 2.7

24.6 1.9 2.7

ns ns ns

INFORMANTS n-informants

4

4

3

SCHOOL in years

12,3

11,8

12,7

COMPETENCE Low German Stand German Spanish

12.1 11.5 7.8

13 11.6 8.5

13 9.9 7.6

USE Low German Stand German Spanish

21.6 6.6 1.8

23.5 4.8 1.7

21.9 5.3 2.8

LINGUISTIC PHENOMENA TWO VERBS (n) interval scale THREE VERBS (n) interval scale RELATIVE (n) interval scale Do-suPPORT (n) interval scale ARTICLE(n) interval scale

31 +0.016

36 +0.021

24 +0.028

34 -0.049

30 +0.068

34 +0.074

189

12

16

9

12

8

12

69

- 0.238

p= 0.034"

+0.034 36

36

24

- 0.148 23

24

16

15

16

10

35

27

- 0.088

+0.069

23

19

19

124

- 0.33

+0.006

p=O"""

14

17

17

35

ns

193

p=O"""

- 0. 361

89

p=O"""

Total n-tokens Standard(%) interval scale

83

121

98

50

55.4

54.6

63.3

61.1

- 0.074

- 0.041

- 0.047

+0.017

p=O"""

Total n-inform. interval scale

4 - O.Q75

Ranking

118

127

4

10

3

13

4

3

- 0.046

+0.02

12

4

17

3

15

14

11

21

5

16

18

20

22

1-

7

19

I

117

4

664

22

p= 0.007""

I

8 9

Looking for order in chaos

and not just the ones from the balanced data set in 1ables 1 and 2. Due the exclusion of informants with more complex biographies we, nevertheless, end up with just 23 long term Menno residents. Although this leads to rather low numbers of three to four informants per subgroup, 18 we are contl.dent that the analyses are valid, because if we take the concept of more or less homogeneous social (age--gender-) subgroups seriously, tokens are as characteristic for these subgroups as informants. And tokens, we do have enough. Before we start analyzing, however, we will have to introduce the new indexing method mentioned in Section 1. Simply adding up frequency data for standard and non-standard variants in a non-balanced data set can result in skewed analyses, because not all informants translated every sentence. That such incomplete data constitute a real problem can be easily demonstrated. The probability of using dun(e), for example, depends among other things on the type of the main verb present in the clause. Likewise, the probability of using the standard variant(s) with nvo VERBS and THREE VERBS depends on the finite verb and the type of embedded clause (cf. for two verbal elements Kaufmann 2003b: 184, 'Iable 2 and 187, Table 3). 'Iherefore, the probability for the occurrences of (non- )standard variants depends on the clauses translated and these clauses are not identical for all informants. In order to solve this problem, the mean probability for the occurrence of the standard variant( s) for each clause in each of the colonies in Brazil and Paraguay was calculated. 'Jhis was done with the help of the balanced data sets. 19 For the translation of the clause }u.an no cree que conozcas bien a tus amigos (=John doesn't think that you know your friends well), for example, most informants of the balanced sample in Menno (86%) used the standard variant without dun(e). For the clause No esta aqui porque esta ayudando a tu padre(= He:~ not here, because he is helping your father out), on the other hand, this share dropped to 52%. In order to integrate this difference in our analyses, an informant using the standard variant in the first clause obtains for this clause a positive value of -t-0,14 (1-0,86).

18. Due to the now low number of informants in the Paraguayan colonies, we had to change the cut-off points for age a little bit in order to maintain an even distribution. The young Mennonites in the two Paraguayan colonies are between 17 and 20 years old (Brazil: 14 and 25), tl1e middle--aged ones between 21 and 33 years (Brazil: 26 and 40), and the older ones between 38 and 64 years (Brazil: 42 and 75).

In the balanced data set all informants had a chance to be selected, i.e. even the ones who have spent some time outside the colony. This might have caused some slight skewi.ng of the results. If we had taken out the partly non-local informants from the balanced set of Table 1 and 2, we would, however, have been left with only 23 informants in Menno and 21 in Fernheim, i.e. the highest possible number for the six age-gender-subgroups would have been 3 or 4 (i.e. a total need of 18 and 24 informants, respectively) instead of 5. The decision, therefore, was between a more solid basis with a certain risk oi skewing or a less solid basis v.ith no risk of skewing.

19.

209

210

Goz Kaufmann This value represents the difference between the observed value ( l for the use of the standard variant) and the expected value (the mean probability of the balanced sample). In the second clause, an informant using the standard variant obtains a higher value of +0,48 (1-0,52), because this informant uses the standard variant in a clause where its usage is less common. For using the non--standard variant with drm(e), the informant receives a negative value of ···0,86 (0-0,86) in the first and of ···0,52 (0-0,52) in the second clause; the first value being more strongly negative because the use of the non-- standard variant is rarer in this clause. This method has two advantages: from an analytical point of view, we can now compare all translations by all informants regardless of the linguistic phenomenon involved and regardless of the precise linguistic context in which a certain variant occurs. The value (between 1 and +1) ·will always correctly mark the token as more or less standard--like. From a statistical point of view, we are no longer dealing with frequency data, but with interval-scaled data and can, therefore, use parametric tests; a much stronger way of detecting significant differences between subgroups. A look at the sociolinguistic background information provides a very homoge-neous pattern f{)r long term residents in Menno; no significant difference exists with regard to length of schooling, use of, and competence in the contact languages. Because of this, differences in the linguistic behavior between the subgroups cannot be connected to differences in the language constellation as in Tables 1 and 2. 'Ihey are rather the result of differences in the socio--psychological dispositions of the informants. We will start our analyses by looking at the five phenomena individually. With regard to young informants, young women show a markedly higher use of Standard German variants in three of the five linguistic phenomena (THREE VERBS, REI.ATIVE, and ARTICLE). \Vith regard to the other two phenomena the results of the two young subgroups are comparable. These striking differences can be interpreted in three different ways: firstly, one could simply argue that young men exaggerate their competence in and use of Standard German (cf. Auer & Hinskens 1.996: 22). In this case one could connect the low frequem:y of standard variants with young men's interior Standard German competence and use values. However, our observations in Menno support such a conclusion only partly. Although it is not clear whether young men really speak Standard German as well and as often as young women, there is no doubt that they do not speak it worse or less often than older Mennonites. Secondly, one might consider the possibility that young men are generally fluent in Standard German and use it quite frequently, but that their standard variety is more MLG-like, i.e. has more MLG borrowings than the standard variety of young women. If this were correct, young men would not be able to borrow many standard variants into MLG (cf. tor a possible mixture of downgrade convergence and upward convergence the situation in Switzerland in Auer and Hinskens [1996: 12]). Since we did not collect Standard German data, we cannot analyze this hypotl1esis empirically. Besides these two points, a

Looking for order in chaos third possible explanation comes to mind, namely different sociolinguistic dispositions towards accepting Standard German variants in MLG. From this point of view, the results would coincide with Labov's (2001: 274) claim: "In linguistic change fmm abfwe, women adopt prestige forms at rt higher mte them men~ That the contact with prestigious Standard German is a change from above is clear. Interestingly, the two phenomena which show a significant difference as for gender (collapsing the three age groups) were considered linguistically complex: RELATIVE (p = 0***/Value F 13.2/total df 192; women use more standard variants) and ARTICLE (p = 0.001~·"/Value F 1.2.8/ total df 88; women use more standard variants). 20 The high complexity means that a conscious effort is required to converge to these standard variants, a linguistic effort we would rather e.xpect from women than from men. Lefs look further into the linguistic behavior of men. In four of the five phenomena analyzed, we see a steady reduction ofthe Standard German variant(s) from older men to younger men (TWo VERBS (non-significant for the sb:: subgroups), THREE VERBS, RELATIVE, and ARTICLE). For these four phenomena, then, we are again faced with the same seemingly contradictory result: young men do not show a lower competence in and use of Standard German but refrain from using the prestigious Standard German variants in MLG. This looks much like Labov's (2001: 297) "retreat of lower wvrking class males fmm a female-dominated change:' Granted, the young Mennonite men hardly belong to the working class, but in relation to the traditional roles of men and women, they share many characteristics of this class (ct~ Kaufmann 2004: 290···291). However, one phenomenon defies this interpretation. Contrary to the other phenomena, young men lead the usage of the standard variant with regard to Do-suPPORT and there is a steady rise in the use of this standard variant between older men and younger men. As the retreat from a female···dominated change must be a conscious process and as the insertion or non-insertion of dun(e) is a rather superficial, at least partly lexical process, one would expect that young men insert the non-standard auxiliary dun(e) more often. The answer to this riddle may lie in the subgroup of older men. 'Ihey show a marked "rejection" ofthe standard variant in this case, although tl1ey are the most frequent users of tl1e standard variant(s) with regard to complex THREE VERBS. Excluding older men from the analysis, the difference between the oilier five age--gender--subgroups is no longer significant. The result for DO··SUPPORT, then, has to be explained with regard to older men and not with regard to younger men. The older men investigated were born between 1946 and 1959, i.e. tlley belong to the first generation which enjoyed better schooling in Menno, and perhaps no-suPPORT

20. Do-suPPORT shows a statistical tendency, aL'io with women using more standard variants (p o.osi•ltvalue P 3.1/total df (123); TWO VERBS and THREE. VERBS do not show any difference.

1.11

212

GOz Kaufmann

was not a major focus in school in these early years or they, too, retreated from a prestigious standard variant. That this is not just an ad-hoc hypothesis can be proven by the fact that older men in Fernheim also show the significantly highest use of nonstandard dun(e) (cL Table 4), and many teachers from Fernheim taught in M.enno in the 1960s (cf. Warkentin 1.998: 309---31.0). But the attitude towards dun(e) seems to have changed dramatically, because among the three significant differences between the age-groups (c.ollapsing the gender groups), DO-SUPPORT shows the clearest difference (p = o~**/Value F 10.5/total df 123/younger informants use the standard variant most (steady rise); RElATIVE p == O.OOl**iValue F 7/total df 192/older informants use the standard variant most. middle-aged iniormants least; ARTICLE p = 0.044*/Value F 3.3/total df 88/older informants use the standard variant most, middle-aged informants least). Most probably, the steady drop of the insertion of dun(e) has to do with the presence of teachers from Germany mentioned in Section 4. 1hese teachers are sure to have commented on the use of tun/dun(e), which is completely banned from Standard German in Germany. The exceptional behavior of younger and older men still requires some further reflection. Differently from Table 2, we cannot explain this behavior by just evoking different levels of complexity. Older men have no problem whatsoever in identifying and borrowing the standard variants \\ith THREE VERBS, and differently from Fernheim it is not very probable that these men used the standard sequences before contact with Standard German (cf. the extremely low values for Menno's sister colo-nies in the USA, Mexico, and Bolivia in Table 2). So, why should they not be able to identify and borrow the rather simple standard variant with regard to DO-SUPPORT? Looking at the behavior of older women in 'Iable 3, we see that they use the standard variants more frequently than older men in three of the five phenomena. But they do lack behind older men with regard to THREE VERBS. What we would expect is that a Mennonite who is able to borrow the standard variant( s) in a complex linguistic phenomenon should be able to do the same with regard to less complex phenomena; or put differently, those intormants who for different reasons (advanced age, lack of contact, conscious resistance to standard variants) do not apply Standard German variants in one phenomenon should avoid them across the board. This expectation is not borne out, though. The problem then is this: In order to put our analyses on a broader basis, it would be desirable to fuse these five phenomena into one standard/non-standard-dimension gauging the general amount of standard variants for every age-gender-subgroup (and eventually for every informant). But does this kind of fusion make sense if the results for the five phenomena are so different? 21

A Bilateral Pearson~s Correlation corroborates our impression: there is only one significant correlation between the five indices for the informants, and this correlation is even

11.

Looking for order in chaos 213 Convergence, indeed, seems to be a rather heterogeneous mechanism. So should we abandon the idea of summing up the five phenomena (more precisely, all tokens of the five phenomena) into one standard/non-standard-dimension? For many researchers, doing this without reaching a satisfactory level of statistical reliability may constitute a serious breach in scientific confidence. However, if we do add up the indices for all tokens, the results for the age-gender-subgroups and for the individual informants make a lot of sociolinguistic sense. The corrected model of the General Factorial AN OVA for the sum of the tokens of the five phenomena is extremely significant (cf. the line 1i:Jtal n-·tokens;22 p:::: o~·**/ Value F 7.2/df 5). 1he most important independent variable is gender (p == 0.001**/ Value F 12.2/df 1; women more standard-like than men) and after gender follows the interaction of age and gender (p :::: 0.003~·" /Value F 5.7/df 2). Age is a little less signifi·· cant (p = 0.011*/Value F 4.5/df 2; with middle--aged informants least standard--like). Thus, our explanations using gender as an important factor for the linguistic behavior in Menno is confirmed. If we compare informants instead of tokens (cf. the line Total n--infonn.); the corrected model is also highly significant even for this low number of informants (p = 0.007**/Value F 4.8/total df 5).23 Coming back to tokens, the difference between young men (most non-standard subgroup) and young women (most standard subgroup) is impressive confirming the male "retreat [... ]from a femaledominctted change" (Labov 2001: 297). An unsolved enigma is the behavior of middleaged women. All eight younger and older women, but no middle--aged woman, are

negative (DO·· SUPPORT and THREE VERBS: -0.59**; this trdilslates into a co--variation of 34.8%). With regard to this surprising re~;ult, one might say that DO··SUPPORT is more lexical in nature than THREE VERBS and, therefore, older men might use it as a conscious marker of their iden-tity as ~>peakers of "pure" MLG. Be this as it may, it is a striking fact that out of ten possible correlations, there is only one significant result. Pernheim shows exactly the same result: there is just tb.is one negative correlation (Do-suPPORT and THREE VERBS; -0.53*/co-variation of 28.1 %). In Brazil there are two positive, but rather weak correlations (Two VERBS and THREE VERBS; +0.4*/co-variatio.n of 16% and Do-suPPORT and THREE VERBS; +0.39*/co-variation of 15.2%). lhe information about the absolute frequency of standard variants in the block 1btal n-tokens (Standard [%]) is given for a (possibly slightly skewed) comparison of the overall linguistic behavior between the colonies in Brazil and Paraguay. As the newly formed index used in the follo\\oing tables depends on the different average probability for the usage of standard variants in each clause and in each colony, this information cannot be used for a comparison between the colonies (differently from the balanced frequency data in Table 2). 22.

23. With a1142 infor.mants tb.e significance for the sum of tokens is p O*•·* (Value F 6.9/ df 5; gender: p 0.003'.*/Value P 9.1/df 1; age: p 0.004"*/Value F 5.6/df 2; age and gender: p o.ou•·Nalue P 4.5/df 2). For informants the significance for the corrected model is p 0.03" (Value F 2.8 df 5).

214

GOz Kaufmann among the nine most standard informants (cf. the line ranking; the only exception among these nine informants is one older man; the five most frequent users appear in light and broad shading), while two middle-aged women are among the five informants which use standard variants least (dark shading; the other three informants are men; the two most non--standard informants are young men). One reason tor this might be the tact that the whole subgroup of middle-aged women has the highest share among the "Canadian" Mennonites (42.9o/; young women have a share of 25%; older women of 20%) and it is also the only subgroup with not a single informant among the "German" or ':Asunci6n"--Mennonites (both young and older women have one informant each in these subgroups). In view of the high probability that the contact between middle-aged women is more intense than between them and the members of other subgroups, one could cogitate that the middle--aged female Mennonites who have spent their whole life in Menno are nevertheless under a stronger influence from Canadian types of linguistic behavior than the other subgroups, this influence not being counter-balanced by a more German-like behavior.

6.2

Fernheim, Paraguay

Table 4 (next page) lists the sLx age-gender-subgroups for the 21 long term residents in Fernheim. Similar to the situation in Menno, the language data show a very homogeneous pattem. Only the length of schooling shows a significant difference; middle-aged and older men have the lowest figures. 'Ihe linguistic data, however, is again highly dependent on the age and gender of the informants. The General Factorial AN OVA for the individual indices of the 21long term residents (cf. Total n-irifotm.) shows a highly significant difference (p = 0.004~*/Value F 5.8/df 5). The results for the corrected model for tokens is also extremely significant (cf. Total n--toke::-ns: p = 0***/Value F 10/df 5). Like in .Menno, gender is the most important independent variable (p = O***/Value F 27.8/df 1; women use more standard variants than men) followed by age and gender. This interaction is highly significant (p = 0.002**/Value F 6.1/df 2). Age again is a less significant variable (p = 0.018*/Value F 4.1/df2; again with middle-aged informants using the least amount of standard variants). 24 Looking at the results in more detail, one can see that women show a steady decrease in the use of standard variants, and this fact clearly isolates older women as the most standard informants from all other subgroups. They occupy the first three ranks of all informants. TI1e highly significant interaction between age and gender is due to the contrary result for men. Older and middle-aged men do not use many standard variants, while the younger men approach the standard variants more.

24. With all 37 informants the significance for the sum of tokens is p = O***(Value F 6/ df 5; gender:p = O***Nalue F 18.7/df 1; age:p = O.Oll*Nalue F 4.5/df2; age and gender: ns). For informants there is no significant difference.

Looking for order in chaos 215 Table4. Language constellation and results for five linguistic phenomena and their sum for six age-gender-subgroups oflong term residents in Fernheim (Inform.= Informants; Stand. = Standard) young men

young women

middle men

middle women

older men

older women

One-Way AN OVA

3

4

21

INFORMANTS n-informants

4

3

3

4

SCHOOL in years

11.3

0.047*

11.3

COMPETENCE Low German Stand German Spanish

12.8 12.3 6

11.2 12 7.3

13.3 11.7 5.3

14 11.4 6.5

13.3 10 8.3

13.5 12.4 8.3

ns ns ns

20.6 7.5 1.3

22.3 6 1.7

21.6 5.9 2.2

ns ns ns

26

36

192

USE Low German Stand German Spanish

17.3 ll 1.7

9.4 17.8 1.9

24.7 2.2 3.1

LINGUISTIC PHENOMENA TWO VERBS (n) interval scale THREE VERBS (n) interval scale

40

26

28

36

p = 0.005**

- 0.083 17

8

6

-0.05

- 0.284

- 0.425

RELATIVE (n) interval scale

36

27

26

Do-suPPORT (n) interval scale

24

ARTICLE(n) interval scale

15

- 0. 197 18

ll

7

15

36

24

34

183

p = 0***

- 0. 16

64

p = 0.02*

24

15

22

119

p = 0.003** 12

12

17

ll

17

Total n-to.kens Standard(%) interval scale

91

88

59.9 -0.024

61.5 -0.003

48.9 - 0.144

Total n-inform. interval scale

4

3

-0.024

132

3

124

83

124

8 15

642 62

p = 0*** 4

3

4

21

p = 0.004**

- 0.147

Ranking

84

p = 0***

- 0.434

4 7 ll 14

13

18 21

3 12

216

Goz Kaufmann

'lhe most interesting result, however, is the difference between the informants' behavior in Femheim and Menno. Although in both colonies women use standard variants more frequently than men 25 and although at least in absolute numbers women in Femheim still use more standard variants than women in Menno (68.4o/ vs. 65.5%; non--significant), the direction of the linguistic behavior in apparent time is completely different While younger women in Menno exhibit a much higher usage of standard variants than older women (middle-aged women do not fit in this respect), women in Femheim show a steady decrease in the use of standard variants. This leads to a statistical tendency for the difference between young women in the two colonies (in Femheim 61.5%, in Menno 73.2% of standard variants; p = 0.067<*>/Value 3.3/ df 1; middle-aged women: non-significant; older women p = 0.084< )fValue 3/df 1 with Fernheim women using more standard Vdl'iants ). And while men in Menno show an ever more marked rejection of standard variants, young men in Fernheim approach these variants. The reason for this might be precisely the fact that the use of standard variants is not characteristic of young women in Femheim and, therefore, does not need to be rejected wholesale by young male informants.1he result of this partly more standard-like linguistic. behavior (THREE VERBS, RELATIVE), however, showes again a marked difference between young men and young women. Another reason for the absence of standard variants among the middle-aged and older male informants in Fernheim may be their relative lack of f()rmal instruction. In any way, the explanation for the marked difference between Menno and Femheim is probably the different length of contact with Standard German. The introduction of Standard German as something more than a hagiolect is a rather recent innovation in Menno; it dates back to the 1950s. Because of this, our informants still react to the standard variety and its variants. ·women have adopted the new prestige variants eagerly, while men have diverged from Standard German features. For the Fernheim informants, on the other side, Standard German has always been a matter of source; therefore, no recent change from above has taken place. Therefore, women in Femheim seem to be innovating away trom Standard German, and this looks rather like a change trom below. Labov (2001: 280) characterizes this kind of change as gender-dependent: "[w)omen have

been found to be in advance of men in most of the linguistic changes in progress studied by quantitative means in the past severa.l decades." It fits into this picture that the two most non-standard phenomena for young women are complex ones, namely THREE VERBS and RELATIVE. Convergence to these standard variants would require a lot of attention, and it seems that women only pay that much attention in a change from above. Interestingly, older men in Femheim behave in an opposite fashion; they are

25. For tokens, women in Fernheim have a value of +0.064 and men of -0.078 (One--Way ANOVA:p = O***Nalue F 26.6/total df641). In Menno the results are +0.065 for women and -0.035 for men (p = O*** Nalue F 12.8/total df 663).

Looking for order in chaos 217 exclusively standard-bound for these two phenomena. Summarizing these differences, one can say that there is a strong influence of Standard German on MLG in Menno, whereas the informants in Femheim seem to keep Standard German and MLG more apart. This means that the existence of a roofing standard variety, even one with many important functions, does not automatically cause standard convergence in the roofed non-standard variety.

6.3

Coltmia Nova, Brazil

The last colony this study focuses on is the Brazilian colony of Colonia Nova in Rio Grande do Sui. These Mennonites arrived in Brazil at the same time as the Femheim Mennonites and belong to the same emigration group. The major difference between the two groups is the fact that the Brazilian Mennonites lost Standard German roofing during the time of the Estado Novo. Table 5 shows the results for these informants. Table 5. Language constellation and results for five linguistic phenomena and their sum for six age-gender-subgroups in Brazil (inform. = informants; Stand. = Standard)

young men

young women

middle men

middle women

older men

older One-Way women ANOVA

INFORMANTS

n-informants

9

9

9

9

9

11

56

6 11.8

8 15.4

11 11.5

42

SCHOOL

n-informant in years

7 12.6

5 12.4

5 16

ns

COMPETENCE

Low German Stand German

12.2

Portuguese

9.3

p = 0.001** p = 0.004**

4.7 11.5

7.5

p = 0***

USE

n-informant Low German Stand German

25

p = 0.001** ns p = 0***

Portuguese LINGUISTIC PHENOMENA TWO VERBS

(n)

interval scale

84

74

87

79

-0.056

-0. 131

73

100

497

p = 0*** (Continued)

218

GOz Kaufmann Table 5. Language constellation and results for five linguistic phenomena and their sum for six age-gender-subgroups in Brazil (inform.= informants; Stand= Standard) (Continued)

THJU!J! VERBS (n) interval scale

RELATIVE (n) interval scale

Do-suPPORT (n)

young men

young women

middle men

middle women

older men

older women

One-Way ANOVA

12 -0.05

18 - 0.115

27

31

33

39

160 p = 0.067(*)

78 - 0.077

73 -0.05

80

80

79

98

488 p=O***

49

45

51 -0. 102

49 - 0.114

50

65

309 p = 0.014*

36

37

42 - 0.268

33 -0.016

35 -0.003

259 44,8 -0.048

247 41,7 -0.059

287 39 - 0.087

272 37,9 -0.078

270

337

1672 45.9% p=O***

9 -0.047

9 -0.061

9 - 0.083

9

9

11

56 p=O***

12 17 22 30 31 33

19 20 23 25 28 37 48 52 55

14 24 26 39 43 45 47 49 56

2 5 6 8 15 18 21 34

4 7 9

interval scale

ARTICLE(n) interval scale

Total n-tokens Standard(%) interval scale

Total n-inform. interval scale

Ranking

36 38 40 42

35 218 +0.044 p = 0.007**

11

13

16 27 32 35 46

In comparison to the Paraguayan colonies, the results for the contact languages indicate a clear case of a threatened minority language. Competence in and use of MLG is steadily diminishing, while the opposite is true for the majority language Portuguese. Although there is only one young female informant, who could be called a semi-speaker in Dorian's (1977) sense, the differences between older and young informants are, indeed, impressive: young women have Portuguese indices which are for competence 4.8 and for use 5.5 points higher than their MLG indices, whereas

Looking for order in chaos older women have higher MLG indices in both cases; the difference for competence is 6.5 points and for use 22.5 points. The picture with regard to Standard German competence (use of this variety is not significantly different) is different. It follows the disappearance and reappearance of Standard German in school. As a consequence of removing Standard German from school, the middle--aged informants exhibit a dra-· matic drop of competence in comparison with the older informants, but the younger informants, who again had contact with Standard German in school, almost reach the level of the older informants . .ru for the sum of all tokens, the Brazilian result is as significant as the results in the Paraguayan colonies (corrected model: p = 0***/Value F 15.3/df 5; for the informants (ct: tl1e block Total n-iriform.) the corrected modelis p = 0***/Value F 6.7/ df 5). In comparison to the Paraguayan colonies, though, the decisive variable in Brazil is age and not gender. While gender and the interaction of age and gender do not reach a significant level, age is extremely important (for the tokens: p = 0***/Value F 37.7/ df2; the older informants showing by far the highest use of standard variants). Thus, loosing Standard German as the language of instruction in Brazil did not only have an effect on language competence in this variety but also on the linguistic shape of MLG. The first nine most standard-like informants are among the older Mennonites (cf. the line ranking). Although only one of them received schooling in German, it seems that the Standard German traditions were not abruptly given up with the prohibition of German schooling, but rather petered out over the years. 'Ihe tour youngest informants in the older subgroups (all under 50 years) exhibit the biggest distance from the use of standard variants in these subgroups (ranks 32, 34, 35, and 46). With this result, they blend in with the middle--aged and young informants, who do not show any significant difference with regard to age and gender (taking out the older informants). In comparison to the subjective data of competence and also to tl1e situation in Menno, the (re)introduction of German as a school subject and partly as a language of instruction in the 1970s does not seem to have had any measumble effect on the MLG of the young informants. 'Ihis is probably the conse-quence of the fact that in the absence of Standard German Portuguese had practically taken over all formal domains. In spite of this seemingly dear picture, it is necessary to have a closer look at the individual results in Brazil, because the fact that gender is not significant for the sum of all tokens does not mean that this is also true for the individual phenomena. In fact, besides the always extant effect of age, four of the five phenomena do show a gender effect. \Vith regard to DO-SUPPORT and ARTICLE, young women exhibit the most standard---like behavior of all subgroups. These are also the only phenomena where women in general show a tendency towards a higher use of the standard variants than men (Do-suPPORT: p = 0.073(*)/Value F 3.2/total df 308; ARTICLE: p = 0.052(*)/ Value F 3.8/total df217). So it seems that here (young) women do show the reaction

219

220

GOz Kaufmann we expect in a change from above. Interestingly, these are also the only phenomena where there is no significant difference in the behavior of young and older informants, only the middle-aged informants show a significant non-standard deviation. Therefore, one could say that the middle-aged informants' drop in the use of standard variants is due to the disappearance of Standard German, but the reintroduction of this prestige variety did only have an effect on young women with regard to these two, not very complex phenomena. Men show significantly higher indices for TWO VERBS and REI.ATIVE ('Two VERBs: p = 0.03l*iValue F 4.7/total df 496; Rm.ATIVE: p = O*~··; Value F 31.3/total df 487). Only THREE VERBS do not show a gender difference. Once again, there is an interrelationship with age, because differently from DO-SUPPORT and ARTICLE we now have a steady divergence from the standard variant, i.e. younger and older informants show a maximum difference. For TWO VERBS (and for THREE VERBS) we can detect this divergence for women, for RELATIVE for men. For the two verbal phenomena, then, we have a female type of reaction we expect in a change from below; they lead linguistic innovation. If the situation were comparable with the one in Femheim, we should see women approaching the standard variant of TWO VERBS, but whereas young women are the standard-closest subgroup in Femheim, the Brazilian young women are the subgroup furthest away from the standard variant. This difference could mean that even the syntactic sequence of just two verbal elements is less under the speakers' control than the lexical (non- )insertion of dun(e); i.e. these word--order phenomena seem to be cognitively more complex than the lel!.i·· cal insertion rule. Analyzing Two VERBS and THREE VERBS from the point of view of men, one can detect something like a rejection of a female led c.hange because after a significant drop of the use of the standard variants between older men and middle·· aged men, young men stop diverging any further from the standard variants, while young women continue the process of divergence. Two more things must be explained concerning the male linguistic behavior in Brazil: firstly, what does the steady male drop of relative pronouns mean, and secondly, why do young Mennonites refrain from using non-standard dem? With regard to RELATIVE we are faced \'oith a similar constellation as the one for DO-SUPPORT in Menno (cf. Section 6.1). It first looks as if young men were deviant because they show the lowest value, but it is actually older men who are deviant. If we exclude them from the analysis, the result for the other five subgroups is no longer significant. The high use of relative pronouns instead of the relative particle waut among older men is mostly likely connected to their upbringing and their involvement in formal decision taking which at this point was still connected to Standard German. Their rather special position can also be seen in the fact that they constitute the only Brazilian subgroup which does not show a significant difference in the use of standard variants when compared with the same subgroup in Fernheim. At least

Looking for order in chaos

with regard to the overall frequency of standard variants, older men seem to be the last linguistic link connecting the two colonies, which were founded by the members of the same emigration group. As for ARTICLE we have already written in the discussion of 'Table 2 that the comparatively high use of the standard variant in the Brazilian colony is probably not the result of a conscious effort but of the fact that with the disappearance of Standard German the more marked dative form dem also started disappearing. Somewhat puzzling is the dramatic non-standard behavior of middleaged men. Here, as in Menno and Fernheim (cf Tables 3 and 4 and Kaut!nann 2004: 294---296), we see that the use of dem is the most marked male characteristic. The reason behind this is unclear.

7·

Conclusion

'Ihe major goal of these concluding remarks is to lvrap up the results provided in Sections 6.1 through 6.3 and to answer some of the questions and hypotheses formulated in Sections 1, 4, and 5. Comparing the absolute percentages of standard variants elicited for all five phenomena, we can see once again the important role Standard German plays in the varieties of MLG spoken in the three colonies. Fernheim exhibits the highest share (61.6%), closely followed by Menno (58.6%), the colony which got under Standard German influence in the 1950s. Between the two Paraguayan colonies and the Brazilian colony. however, there is a large gap. The Brazilian informants, whose parents and grand--parents belonged to the same emigration group as the founders of Fernheim, but who lost Standard German roofing during World War II, only use 45.9% of the standard variants (for the 3 colonies: p = 0***/Value 80.9idf2; cf. also the results in Table 2). In order to make a valid judgment about the processing complexity of the five phenomena, we will have to look once more at the behav-ior of women. Cheshire, Kerswill, and Williams (2005: 143) write that "indeed, the linguistic behaviour of female speakers is sometimes taken as a diagnostic of change in progress [... Theretore, if we apply this proposition to our data, we should be able to distinguish the five linguistic phenomena (cf., once again, the quote of Cheshire, Kerswill & Williams (2005: 13.5) in Footnote 1). If (young) women adopt a Standard German variant fast, this variant appears to be sociolinguistically prominent. If they do not adopt a Standard German variant, one has to analyze why they are not sensitive to this variant. Table 6 shows whether a significant difference or at least a statistical tendency exists between the two genders for the five phenomena in the three colonies (collapsing age groups) and indicates the rank of young women among the six age-gender-subgroups.

r

221

222

Goz Kaufmann Table 6. Gender differences for the five linguistic phenomena in the colonies in Brazil and Paraguay THREE VERBS

RELATIVE

TWO VERBS

Menno young women

ns second

women first

ns ns

women second

women second

Fernheim young women

ns

women

women first

women second

women first

women first

women first

Brazil young women

ARTICLE Do-SUPPORT

The colonies and linguistic phenomena in Table 6 are grouped according to the behavior of young women. We can separate the table into two parts: on the left- hand and lower part we have the five blocks where young women in Fernheim and Brazil (not in Menno) show non-standard results (fifth and sixth places are marked in dark shading). On the right-hand and upper part we find the nine results where young women in the three colonies show above-average standard-like results (first and second places are marked in light shading). Only one result (Two VERBS in Menno) did not show a statistical tendency for the six age-gender-subgroups). The remarkable role of young women is obvious: they rank first in five blocks (33.3%), second in four blocks (26.7%), and fifth and sixth in two blocks each (13.3%). Only in one of the fourteen "significant" results do they rank in an intermediate position (fourth place for RELATIVE in Brazil). Looking for general gender differences in the fifteen blocks (cf. the lines starting with the name of the colony), we have four non-significant results and two results where men use standard variants more often than women (dark shading). Interestingly, four of these six "exceptions" are found in the left-hand and bottom part (80% of these 5 results), while in the larger section there are only two exceptional non-significant results (20% of these 10 results) . Thus in this part, the behavior of young women is confirmed by the female behavior in general (light shading). The most homogenous results can be found for ARTICLE and Do-suPPORT. In all six blocks women are closer to the standard and young women rank either first or second. Since there is no real Standard German roofing in Brazil and as the roofing in Fernheim has always been a matter of source, we suggested in Sections 6.2 and 6.3 that women in these colonies will not be disposed to invest much energy into convergence towards Standard German variants. Therefore, the fact that women in Fernheim and Brazil are nevertheless closer to the standard than men could be seen as proof that convergence in these two phenomena is cognitively not very costly. With regard to semi-lexical Do-suPPORT this explanation seems fair enough; with regard to the morphological form of the ARTICLE, however, the explanation is less clear. We had analyzed this phenomenon as being of medium complexity (cf. Section 5) and have already pointed

Looking for order in chaos

several times to the marked male behavior with regard to it (cf. also Kaufmann 2004: 294-296). Looking on the two columns on the left-hand side the separation is dearly between Menno on the one hand and Fernheim and Brazil on the other hand. We, theret{)re, conclude that these two phenomena exhibit the highest degree of complexity for the speakers, because only the extraordinary effort invested by strongly standard-driven (young) women in Menno succeeds in converging towards the standard variants. This result coincides with our classification of THREE VERBS and REI.ATIVE as complex (morpho)syntactic phenomena (cf. Section 5). The breaking point in Table 6 is TWO VERBS. Here Fernheim women are still standard--oriented, while the Brazilian women have lost Standard German as a pivot. Judging from the results in these two colonies, TWO VERBS seems to be more complex than ARTICLE, because the roofless Brazilian women start innovating in the non-standard direction at this point. On the other hand, we should not forget that the disappearance of dem in Brazil may also have purely structural reasons (cf. Sections 5 and 6.3). "Unfortunately': in Menno neither of the two values for TWO VERBS is significant and, therefore, it seems that this phenomenon is sociolinguistically inconspicuous there. The biggest surprise in this study was the fact that although the sum of the individual phenomena displays an expected outcome in these standard-with-dialect situations, strong correlations between the five phenomena (cf. Section 6.1 and Footnote 21) were not t{)und. Consequently, ongoing convergence and divergence appear to be rather heterogeneous processes which do not seem to be driven only by the linguistic complexity of the phenomena but also by other factors. Table 7 constitutes one more attempt to solve this enigma. In it the reader will find the results of a cluster analysis (with six clusters) in which the values for the five phenomena were analyzed for 76 informants from Menno and Fernheim; the Brazilian situation was judged to different for inclusion. 1he raw index data for each phenomenon were transformed into z-scale-data, which guarantee the comparability of the individual results and exclude possible skewing by the different numbers of tokens per pheno-menon. This transformation was done for each colony separately in order to respect the colony-internal means. If only one of the five phenomena displayed missing or too few data for an informant, the cell was filled with the neutral value 0. This was necessary for nine informants in THREE VERBS. As the cluster analysis calculates the indi-vidual behavior and not the one of age-gender-subgroups, all Paraguayan informants and not just the long term residents (cf. Tables 3 and 4) were included. A first point in relation to Table 7 is whether the linguistic clustering coincides with extra-linguistic factors. Interestingly, the distribution of the informants with regard to the two colonies is not significant, i.e. in spite of the differences between Menno and Fernheim, it seems that comparable types of speakers exist. A.s gender played the most important role in both colonies, it does not come as a surprise that the clustering coincides strongly with this factor. The distribution is highly significant

223

224 GOz Kaufmann

Table 7. Results of a duster analysis for informants from Menno and Fernheim (inform.= informants) Cluster 1

Cluster4

Cluster6

Cluster 3

Cluster 5 Cluster 2

INFORMANTS n-informants

4

9

27

7

20

9

3 4

10 10

4 5

COLONY Menno Fernheim

2 2

4 5

16 11 GENDER

Male

4

6

17

4

5

Female

0

3

10

3

15

8

34.5

31.7

2

2

AGE in years

28.9

26.6

RESIDENCE GROUP 5

Canada Germany

0

1

Asundon

0

0

own colony

3

3

2 4

6 0

3

0

5 10

5

11.3 7.2

11.3 7.4

SCHOOL in years

12.8

12.3

STANDARD GERMAN Competence

Use

8.8 5.5

11.1 5.6

10.9 3.9

10 4.6

LINGUISTIC PHENOMENA THREE VERBs

- 0.74

RELATIVE

- 0.52

Two VERBs

- 2.52

ARTICLE

-0.07

Do-SUPPORT

-0.22

Total n-inform.

-0.1 7

- 0.8 1

(p = 0.004**Nalue 17.4/df 5; higher shares in light shading) and only cluster 3 could be considered gender-neutral. Due to this result the clusters in Table 7 are ordered with regard to gender dominance starting on the left-hand side with an exclusively male cluster. Besides gender, the residence groups show a highly significant distribution

Looking for order in chaos

(p = 0.005**/Value 32.7/df 15). If we finally add age (p = 0.015*/Value F 3.1/total df75) and duration of schooling (p "'0.023*/Value F 2.8itotal df 75) as significant distributions - the values for competence in and use of Standard German are not significant-, we can characterize three pairs of cluster. Clusters l and 6 contain predominantly long term residents, are predominantly (or exclusively) male and have a comparable dura-tion of schooling. Their only sociological difference is age. Also the second pair shows just one sociological difference. Clusters 2 and 5 are both predominantly female and they have a comparable average age and schooling. 'Iheir major difference lies in the fact that in duster 5, 40% of the informants have spent part of their life in Germany or Asuncion, while in cluster 2 this share is only 22.2%. Clusters 3 and 4 are the least convincing pair: they both contain predominantly Mennonites who have spent part of their life in Canada (68.8%; the other four dusters just contain 8.3% of "Canadian" informants), which proves the strong intluence of this factor on MLG. However, this is the only comparable feature. Otherwise they are either predominantly male or genderneutral and show strong differences in age and schooling. If we analyze the results for the linguistic phenomena of the six clusters, we realize that there are six extreme values (bigger than+ 1 or smaller than -1 ). 'Ihese values could be regarded as candidates for identity marking. Two clusters, however, do not show a single extreme value, and quite unsurprisingly these two clusters contain most of tlw 76 informants, i.e. 47 or more than 60%. Looking first at duster 6, the biggest duster, the hierarchy for usage of standard variant~ is DO·· SUPPORT> TWO VERBS >RELATIVE> THREE VERBS> ARTICLE. With the exception Of ARTICLE, which might be explained by the high proportion of men in this cluster (it contains 45.9% of all male informants, but only 25.6% of all female informants), this sequence mirrors our complexity hierar-chy perfectly. So, 27 informants, i.e. 35.5% of all Paraguayan Mennonites studied, seem to function according to complexity. For the twenty informants of cluster 5, the results are not as clear-cut as with cluster 6: firstly the distance of 1.1 between the extreme values is slightly smaller than the one in cluster 6 (1.32); secondly there are two outliers (ARTICLE and RELATIVE), the other three results cluster together. The standard--like behavior with regard to ARTICLE could again be explained by the unbalanced gender mixture (cluster 5 contains 38.5% of all female informants, but only 13.5% of all male informants); the only below--average result of the cluster in RELATIVE may be explained with the complex and foreign nature of relative pronouns in MLG. In comparison to the rather homogeneous behavior of these two clusters, the six extreme values which occur in the otlwr clusters can be considered highly marked options. The Canadian clusters 3 and 4 each show one extremely below-average result. Interestingly, the distance to the most standard--like phenomenon in both dusters is rather similar (duster 3 shows a distance of 2.21; duster 4 of 2.45). The more standard-like and rather young duster 4, which shows a statistical tendency for more schooling in comparison to cluster 3 (p == 0.082(*)/Value F 2.5/total df 15), inserts non-standard dun(e) very often. 'Ihis might not be a marked option with regard to the oldest cluster 1, but definitely constitutes

225

226

GOz Kaufmann

a marked option in comparison to the other young and middle-aged clusters 2, 5, and 6, which show above-average values for this phenomenon. Likewise the most extreme of all values, TWO VERBS in cluster 3, seems to be a dear non-standard act of identity; all other clusters show positive ·values for this phenomenon. Whether the difference in schooling is responsible for the choice of one or the other non--standard variant in these two clusters, is unclear. 1he four other extreme valnes will be analyzed in a slightly different manner. If we compare the different clusters with regard to their behavior as for the five phenomena, we see that five of the fifteen possible duster pairings show only one or two significant differences, i.e. the members of the clusters do not show a different behavior with regard to three or four of the phenomena investigated. We will thoroughly analyze two of these cases. In these cases the few differences occur in one of the "homogeneous" cluster pairs mentioned at the beginning of the discussion of Table 7. 'Ihe female clusters 2 and 5 only show one significant difference, namely RELATIVE, but this difference is indeed huge. Interestingly. the more female and less "German"- and ':-\sunci6n"connected cluster 2 shows by far the highest use of relative pronouns, while cluster 5 uses relative pronouns only rarely (p = 0*""*/Value F 239.3/total df28). If we restrict our analysis to women, the absolute difference between the two values rises from 2.69 to 2.76 (p = 0***/Value F 198/total df22), and if we continue restricting the analysis to women in Menno it reaches 3.12 (p == 0**""/Value F 128.8/total df 10; for women in Fernheim the difference is only 2.42/p =0**~/Value F. 92.2/total df ll ). This result could be regarded as a marker of identity by some women in Menno, because the introduction of relative pronouns in MLG is sure to be a very c.onspicuous Standard German influence (cf. the discussion of Table 2), i.e. a perfect means for marking on~·s "stan-dard" identity. Besides this, the informants in these two clusters do not show a second difference as for the five phenomena (this remains true restricting the comparison to women, or women from Menno) stressing the symbolic importance of the use of relative pronouns instead of a relative particle. It also fits our expectations that it is women from Menno, who show a more marked difference, because in comparison to women from Fernheim they react more strongly to Standard German. Unfortunately, we cannot explain the difference in beha\ior by the difference in the cluster characteristics. For female informants from Menno, the distribution of residence groups is not significant. In duster 5 there is just one woman, who has lived in Germany, in cluster 2 there are only Canadian and long term residents. Interestingly, only four of the 23 women in these two clusters have lived in Germany or Asuncion, the two places strongly connected to the use of Standard German, while all six men in these two clusters have had such an experience. As these clusters have by far the highest values for the use of stan-dard variants (cf. the line Total n-inform in Table 7), we may say that men only reach this standard-like level with direct and intensive exposure to Standard German, while women can do so "i-vithout this kind of exposure symbolically marking their "standard" identity by the use of standard variants.

Looking for order in chaos

The two male clusters 1 and 6 are statistically different with regard to the complex THREE VERBS and the not complex DO-SUPPORT. The exclusive male cluster 1 is a highly interesting cluster because it shows the most extreme behavior of all clusters. Three of the iive phenomena show an extreme distance to the standardized mean, i.e. half of the six extreme values in Table 7 are concentrated in this duster. As it is also the cluster with the highest average age, we can see the dramatic changes in the linguistic and attitudinal development in Paraguay. These four informants can combine the most standard-like behavior with regard to THREE VERBS with the least standard--like behavior with regard to .ARTICLE and DO- suPPORT. As complexity cannot possibly explain this behavior, we may have to reckon with the possibility that these four older informants evaluate these variants ditierently from Germans in Germany and younger Mennonites in Paraguay. The different behavior of the men in cluster 6, which we have already analyzed, proves this point. This very young subgroup (on average the men are 24 years old) follows our expectations as for complexity. The three rather complex phenomena THREE VERBS, ARTICLE, and RELATIVE show negative results, while the least complex phenomenon DO--SUPPORT shows a strongly positive value. 1would like to conclude this article by repeating the most important result: there are many different paths of convergence and divergence, and it is this fact that has caused the lack of correlations between the five phenomena investigated (cL Section 6.1 and Footnote 21). Interestingly, however, the general tendency tor convergence or divergence can be dearly seen once we sum up the results for the five phenomena. This may be seen as a proof that the informants are aware of the (lack of) prestige of all variants; they just mix them in unforeseeable ways. A perhaps rather common-place rule for convergence and divergence in contact situations of the kind studied here might, therefore, look like this: Convergence to prestigious variants is easier when cognitively simple linguistic stmctures are involved. One can converge, however, even to complex linguistic stmctures provided the wish or the necessity to do so is strong enough ( cf. young women in Menno with regard to THREE VERBS and RELATIVE), and one can resist to ':onverge even to simple linguistic structures if one has the wish and the liberty to do so (cf. men in Menno and Fernheim with regard to ARTICLE) or if one does not succeed in identifying the prestigious variants any more (cf. young women in Brazil with regard to Two VERBs).

A corollary to this rule is that the general structural context of a phenomenon can interfere with its complexity and its sociolinguistic meaning in complicated ways. We have seen this tor ARTICLE in Brazil (cf. Sections 5 and 6.3), where the disappearance of the non-- standard variant might just be a structural consequence of the lack of Standard German input completely unrelated to any conscious decision of the speakers; and we can see this for the relationship between TWO VERBS and THREE VERBS (cf. Section 4). For these two phenomena Kaufmann (2007: 185---193 and 198---204)

227

228

GOz Kaufmann

could show that multiple verb raising with three verbs leads to verb raising with two verbs and under certain conditions even to verb raising with just one verb. Leads seems to be more adequate for this structural relationship than implies, because due to different levels of complexity these processes rarely exhibit complete synchrony. This means that using non--standard verb raising with just one verbal element is a late and rare innovation in a non-standard variety without contact to its standard variety, while not using multiple verb raising with three verbal elements is a late innovation in standard convergence. Finally. I would like to quote some rese-archers who apparently have had similar experiences with the heterogeneous and somehow "chaotic" converging behavior in contact situations: Although it L-. clear from our Belize and St. Lucian examples that language changes are taking place, it is also clear that not all change is in the same direction, towards the same target[ ... ]. (LePage & Tabouret Keller 1985: 199) Fiir gewisse Phasen der Entwicklung ist es dabei nur schwer moglich, Regelhaftigkeiten zu erk.ennen, die die Obernahme bestimmter Varianten steuern und die Obernahme anderer Varianten verhindern. (Mattheier 1996: 41)

References Auer, Peter, Hinskens, Frans & Kerswill, Paul (eds). 2005. Dialect Change: Convergence attd Divergence in European Languages. Cambridge: CUP. Auer, Peter & Hinskens, Frans. 1996. 'Ihe convergence and divergence of dialects in Europe. New and not so new developments in an old area In Sociolinguistica 10: International Yearbook of Buropean Sociolinguistics, UlriCh Ammon, Klau.~ ). Mattb.eier & Peter H. Nelde (ed~), 1-30. 'ltibingen: Max Niemeyer. Cheshire, Jenny, Paul Kersv.111 & w1lliams, Ann. 2005. Phonology, grammar, and di.~course in dialect convergence. In Dialect Change: Convergence and Dh•ergence in Buropean Languages, Peter Auer, Frans Hinskens & Paul Kerswill, 135-167. Cambridge: CUP. Doll, johannes. 2002. Geschichtstexte versteh.en in einer anderen Sprache: Eine vergleichende Untersuchung der Leseleistung von Schillern aus deutsch-brasllianischen Begegnungsschulen und deutschen Gymnasie.n. Frankfurt: Peter Lang. Dorian, Nancy. 1977. The problem of the semi-speaker in language death. The International Journal of the SociologyofLangr..tage 191(12): 23-32. Evers, Arnold 1975. 'Ihe Transformational Cycle in Dutch and German. Ph.D. dissertation, Unlversi1y of Utrecht. FleisCher, Jiirg. 2004. A typology of relative dau.~es in German dialects. In Dialectology meets Typology: Dialect Grammar from a Cross-Lingusitic Perspective, Bernd Kortmann (ed), 211-243. Berlin: de Gruyter. Haegeman., Liliane &van Ri.emsdi_ik, He.rik. 1986. Verb projection raisl..ng, scope, and the typology of rules affecting verbs. Linguisticlnquiry 17(3): 417--466.

Looking for order in chaos Kaufmann, Goz. 1997. Varietiitendynamik in sprachkontaktsituationen: Attitiiden rmd Sprachverhaiten ruftlanddeutscher Afennoniter1 in Mexiko rmd den USA Frankfu1t: Peter Lang. Kaufmann, Goz. 2003a. Des Plattdeutschen Wanderjahre oder die lexikallschen Folgen der mennonitlschen Flucht nach Amerika. In ~'>tandardfragen': Soziolinguistische, kontaktlinguistische und sprachgesd:ichtliche Aspekte, Jannis Androutsopoulos & Evelyn Zi.egler (eds), 139-160. Frankfurt: Peter Lang. Kaufmann, G07. 2003b. The verb cluster in Mennonite Low German. In German Language Varieties Worldwide: Internal and External Perspectives, Klaus J. Mattheier & William Keel, 177--198. Frarikfurt: Peter Lang. Kautinann, Goz 2004. Eine Gruppe Zwei Geschichten Drei Sprachen: RuBlanddeutsche Mennoniten in Brasilien und Paraguay. Zeitschrift fur Diaiektologie und Linguistik LXXI/3: 257--306. Kaufmann, Goz. 2005. Der eigensinnige Informant: Argernis bei der Datenerhebung oder Chance zum analytischen Mehrwert? In Corpuslinguistik it1 Lexik und Grammatik, Friedrich Lenz & Stefan Schierholz (eds), 61---95. Ttibingen: Stauffenberg. Kautinann, Goz. 2007. The verb cluster in Mennonite Low German: A new approach to an old topic. Linguistisdte Berichte 210: 147--207. Kautinann, Goz 2008. \'\There syntax meets mo11,hology: Varianten des bestimmten Artikels und die Variation satzfinaler Verbcluster im Plattdeutschen texanischer Mennoniten. In Dialektale Morphologie, dialektale Syntax: Beitriige zum 2. Kongress der Intemationalen Gesellschaft fiir Dialektologie des Deutschen, Wien, 20.---23. September 2006, Franz Patocka & Guido Seiler (eds), 87--119. Wien: Prasens. Kaufmann, Goz. 2010. Non-convergence despite language contact In Language and Space. An International Handbook of Linguistic Variatiotl: Theory and Methods, Peter Auer & Jiirgen Erich Schmidt (eds), 478---493. Berlin: Walter de Gruyter. Klassen, Peter P. 2001. Die Mem1ot1iten in Paraguay, Band 1: Reich Gottes und RJ?ich dieser Welt. Bolanden-Weiherhof: Mennonitischer Geschichtsverein e. V. Kohler, Klaus J. I 995. Binfiihrung in die Ph011etik des Deutschen. Bedin: Erich Schmidt. Koopman, Hilda & Szabolcsi, Anna 2000. Verbal Complexes. Cambridge MA: The MIT Press. Labov, William. 200 l. Principles ofLinguistic Change: Social .Factors. Oxford: BlackwelL Langer, Nils. 2000. 7...ur Verbreitung der Tu.n-Periphrase im Friihneuhochdeutschen. Zeitschrift for Dialektologie und Linguistik LXVH/3: 287-316. Le Page, Robert B. & Tabou.ret-Keller, Andree. 1985. Acts of Identity: Creole-based 1\pproad1es to Lang-uage and Bthnicity. Cambridge: CUP. Lexikon der Mennoniten in Paraguay. 2009. Verein ftir Geschi.chte und Ku.ltur der i•.,.fennoniten in Paraguay (ed.). Asu.nci6n. Lotscher, Andreas. 1978. Zur Verbstellung im :zurichdeutschen und in anderen Varianten des Deutschen. Zeitschrift fiir Dialektologie und Linguistik XLV(I): 1-29. Louden, Ma.rk L. 1992. German as an object-verb language: A unification of generative and typological approaChes. In On Germanic Linguistics: Issues and Methods, Irmengard Rauch, Gerald F. Carr & Robert L. Kyes (eds), 217-231. Bedin: Walter de Gru.yter. Mattheier, Klaus J. 1996. Varietatenkonvergenz: Oberlegungen zu. einem Baustein einer 'Iheori.e der Sprachvariation. In Sociolhlguistic.a 10: Internati011al Yearbook of European Sociolinguistics, Ulrich Ammon, Klaus J. Mattheier & Peter H. Nelde (eds), 31-52. Tlibingen: Max Niemeyer. Milroy, James. 1992. Social network and prestige arguments in sociolinguistics. In Sociolinguistics Today: International Penpectives, Kingsley Bolton & Helen Kwok (eds), 14.6--162. London: Rou.tledge.

229

230

GOz Kaufmann Sawatzky, Harry Leonhard. 1986. Sie suchten eine Heimat: Deutsch-MennoniNsche Kolonisierung in Mexiko 1922-1984. Marburg: N.G. Elwert. Schmid, Tanja. 2000. Die Ersatzinfinttivkonstruktion im Deutschen. LinguisNsche Berichte 183: 325-351.

'Ihiessen, Jack 2003. Mennonite Low Germa,1 Dictlo,lary/Mennoniti.~ch-Plattdeutsches lt\l()rterbuch. Madison WI: University of Wisconsin Press. 'Ihomason, Sarah Grey & Kaufman, 'Ierrence. 1988. Language Contact, Creoll.1.atlon, and Genetic Linguistics. Berkeley CA: University of California Press. Vanden Wyngaerd, Guido. 199·1. IPP and the Structure of Participles: Groninger Arbelten zur germanistiscr.en Linguistik. Groningen: Rijksuniversiteit Groningen Warkentin, Jakob. 1998. Die Deutschsprachiger~ Siedlerschulen in Paraguay im spanr~ungsjcld staatlicher Kultur- und Emwicklungspolitik. Miinster: Waxmann. Warkentin, Jakob. 2007. Erzieimng und Blldung im Raum der Schule, Verein fur Geschichte und Kultur der Mennoniten in Paraguay (ed). Asunci6n. Woolliiser, Curt 2005. Political borders and dialect divergence/convergence in Europe. In Dialect Change: Convergence and Divergence in European Languages, Peter Auer, Frans Hinskens & Paul Kerswill (eds), 236-262. Cambridge: CUP.

SECTION

4

Syntax II - The syntax of Cimbrian German

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy- and what they (do not) betray about language universals and change under areal contact with Italo-Romance* Werner Abraham Universitiit \Vien

This article is on Cimbrian German, an old enclave dialect in Upper Italy surrounded by Italo-Romance dialects. Next to clear traces of German clausal syntax, it shows Romance characteristics, which could be due to borrov.ing from the surrounding Italo-Romance majority dialects. Pertinent literature to date has assumed that the mix of stmcl1u·al prope1ties of German and Italian are indeed due to the century·· long isolation of the German island dialects and their relationship to the majority Italo-Romance dialects. The position presented in this paper focuses on the exclusive orality of Cimbrian and the specific stmctural changes oral-only variants are subject to as opposed to written, standard vernaculars. More generally, the methodological tenet is pursued that single changes subject to ambiguous interpretation need to be disambiguated by careful alignment with the major set of properties- i.e. a minority stluctures that are commonly found in ltalo-Romance dialects may receive interpretations that are typical of oral-only German. The methodological null-hypothesis, then, is that cl1ange occurs under the narrowest accompanying structural conditions accessible within one and the same lanbruage rather than by borrowing from the social majority language.

*I read papers on this topic in the years of 2007 and 2008 to ditferent audiences: in V1enna, Munich, Rome, Oslo, and Leeds. One of the anonymous reviewers reminded me of the importance to consider contact··induced change. Special thanks are due to Ermenegildo Bidese (Trento ), Anthony Rowley (Munich), Leo Tyroller (Munich), Agnes Kolmer (Zurich), and Alessandra Tomaselli (Verona). Claudio di Meola (Rome) is thanked for his administrative support. 1he usual provisos apply.

234 Werner Abraham

1.

Introduction

Cimbrian German, a linguistic enclave dialect in Northern Italy, shows both traces of German clausal syntax and deviations from it that would seem to have allowed contact influences from ltalo-Romanc.e and Italian. 'lhis has led to the conclusion represented in the literature on the topic that the sentential syntax of Cimbrian has yielded to the Romance majority environment during centuries of immediate social contact and equally long isolation from German oral as well as written enmding. By contrast, the present approach suggests specific structural properties of change that Cimbrian has undergone are due not to contact transfer, but rather to results of the exceptionless oral code that Cimbrian speakers have been exposed to. More generally, we adopt the position that changes under contact will be possible only given certain structural openings in the receiving language for any foreign influence and, more generally still, that such influences have to be alignable with the main typological configuration of the receiving languages such that the newly developing structural properties have autonomous, i.e. not contact induced, status. In particular and explanatory to the latter, we shall assume that changes under exclusive oral encoding may be fundamentally different from those of a written code (cf., with a far wider typological scope, Abraham 2006a,b; Abraham & Conradie 2001). In what follows, data from Cimbrian will be discussed against the background of the historical changes from older OV to VO in modern Romance in the context of areal linguistic contact with the environing Italo-Romance dialects and Standard Italian. The following specific questions come to the forefront: Which structural components will change earlier than others and on which of the two clausal peripheries or elsewhere within clause structure? Are there genuine universals of diachronic change different from synchronic derivational ones? Or else, given our object of investigation being a collection of dialects which have hardly ever known stabilization through written encoding, do we have to distinguish processes of parsing facilitation under cognitive economy principles, and if so which precisely? What would such "universals of linguistic change" be in the first place? Before a more general background is introduced, the ensuing discussion needs to be seen between the striking preservation (the inertia against change) of German-Old TyTolean, or Old Bavarian, characteristics and the question whether the heterogeneity of Italo--Romance (henceforth IR) triggered deviations from German V2 1 and OV; or which other force, produced the change in Cimbrian that unfolds before our eyes.

1. A common feature of all Germanic languages, except English, is the so-called V2 property: the finite verb is the second constituent, following an arbitrary, single, clause-initial ('atomic')

Spoken syntax in Clmbrian of the lingulstlc island~ in Northern Italy 235

1.1

Cimbrian German: What is this anyway?

Cimbrian is the collective name for the dialects of the inhabitants of the "Seven Communities" as well as the "Thirteen Communities" in the Province of Verona in the South of Northern Italy (A.siago). In what follows, I will also include three more dialects from the North in this discussion: those found in the Fersental (in the province of Trentino), the dialects of the villages of Luserna (likewise in Tl-entino), and those in Pladen to the North. The decision to include these dialects in this study is based on the simple reason that the speakers of these dialects all speak minority dialects of Old T}'rolean or Old Bavarian descent in the midst of Italian-speaking majority populations. Yet. in his important Laut- and Flexionslehre des Zimbrischen, Kranzmayer (1981: 3) does not include the dialects of the Fersental nor any others but those of the so-called Seven Communities, the Thirteen Communities, and those of adjoining Folgaria and Lavarone. He calls only the latter the Cimbrian linguistic enclaves proper, separating them from the remaining dialects further to the North on clear soundhistorical distinctions: According to Kranzmayer, the status of Cimbrian vowels and consonants is considerably older than those of the Fersental or Pladen, which are not that different from today's South Bavarian. 'Ihe original German Tyrolean (fi·om South and North Tyrol, moving to today's settlements between the XIIth and XIVth centuries; see Rogger 1979) speaking population of the 7 and 13 Communities had a much wider extension (see Map 1) staying autonomously German for a long period of time. This area included the entire border chain of the Alps (Dolomites) between Etsch (Italian Adige) and the mountain chain of the Brenta reaching down to the plain of the Piave River (see Map 1 and its legend). Today's Cimbrlan linguistic enclaves are the remainder of this originally undivided linguistic area. According to modern terminological usage, Cimbrian does not include the German dialect of Pladen (ital. Sappada), which is far to the North and, thus, outside of the Cimbrian speaking area in the South (contra the classification Grewendorf & Poletto 2005 have undertaken). For the most exhaustive description of Cimbrian, see Schweizer 2008 (and its review by Abraham 2009b ).

constituent. Unless XP is a wh-phrase, the instantiations ofV2 yield a declarative clause. IfXP is a wh-phrase, the clause is interrogative: [XP (i) [V + n [ ••• (e ) ... ]]].German and its dialects are uniquely also Vlast. The Vlast slot is open only for atomic + finite verbal elements. German has an OV clause structure meaning that the subject of the clause may remain in its VP-internal position. See Haider 2009, ch. l.

236 Werner Abraham

Map 1. Areas to the East of Lake Garda and to the North of Verona where German Cimbrian was spoken in the 16th century (outer boundaries), around 1820, and is spoken in today's truncated smallest enclaves (from Kiihebacher 1972)

1.2

Presuppositions and consequences

Cimbrian German or, more appropriately, Cimbrian Tyrolean-Bavarian, respectively, has existed for centuries under social and linguistic contact with Romance (Ladinian, Friulian, Italian and its dialects in Northern Italy). Thus, Cimbrian German/Tyrolean provides an especially appropriate pattern for synchronic as well as diachronic change (loss, new adoptions) triggered under areal linguistic contact. Due to the fact that Cimbrian has not been subjected to any media or schooling influences from Standard German, as well as the fact that the vernacular has lived on only orally inside immediate communities, this dialect functions as a dependable source of its original linguistic status (as well as its oral linguistic quality for the last 200-300 years). 1.3

Assumptions about linguistic change to be confirmed, or disconfirmed, by Cimbrian German

The following assumptions are kept as short and succinct as possible in this introductory section of the paper. More specific and detailed motivations will be added later in due course.

•

Spoken syntax in Clmbrian of the lingulstlc island~ in Northern Italy 237

The radical assumption is that areal linguistic contact never takes place in terms of arbitrary transfer of grammatical elements i.e. as a simple copying mechanism of a foreign pattern. Consider the Cimbrian ·venire-futures khimp gamocht (Rowley 2003: 237), literally 'it-comes-made' to mean "it will be done" using the venitive auxiliary instead of Standard German werden "become·: Now, replacing German werden with Cimbrian kommen "come" may be a lexical calque from IR. vVhat counts, however, is the inchoative aspect shared by both lexical items, werden and kommen, which is the leading criterion for the transfer (compare Lotscher 1993 for Alemannic). In the High Alemannic dialect of the Montafon/Vorarlberg (Abraham 1965), both the venitive and the andative futures are in use with a slight, but definite meaning distinc-tion (Abraham 1965; Lotscher 1993). 1.3.1

(l)

a.

got

ga

regna

andative remote future

it goes toward rain

b.

s krmt

ga

regna

venitive nearby future

it comes toward rain

See also Bayer 1993, Bhatt & Schmidt 1993, and Abraham 2004. Any structural transfer or diachronic change is subJect to principles of least resistance - i.e. of strongest economy in the process of transfer. Such control criteria are the subject of theoretical implications in modern linguistic literature concerning linguistic change in van Gelderen 2004. Other modern literatures (e.g. Heine & Kuteva 2005; Good 2008) have nothing to say about linguistic criteria of economical processing.

1.3.2

1.3.3 Taking 1.3.1 and 1.3.2. above as points of departure, we might speak about "linguistic change triggered by areal linguistic contact, but controlled by autonomous regularities': It remains to be seen exactly what 'autonomous regularities' look like.

1.3.4 Should one consider that "linguistic change by areal linguistic contact'' is con-trolled by leading principles different fi·om "autonomous linguistic change" in a single receiving language such as Cimbrian German? Kuteva & Heine (2008) highlight areal differences where grammaticalizing buffer :~.ones receive independent status (such as the mL1C of Pro- as well as Post-article morphemes in a medial area of Scandinavia). However, so far a general clarification has not been reached in the literature regarding whether or not processes of diachronic grammaticization in foreign linguistic criteria are simply adopted. Given 1.3.1 above, I cannot see this as a realistic premiss for linguistic change.

238 Werner Abraham

StnH.:tural change has nothing to do with arbitrary (though, quite naturally, socially useful) lexical change. Pure lexical influence will always be strong and will encounter far less resistance in the receiving language. 1.3.5

1.3.6 Even iflexical elements undergo change under social contact, syntactic structure need not be subject to alteration. One prominent example is cliticization in Cimbrian and lit IR has proclitics, whereas Cimbrian has never given up enclisis of its pronominallexicals- regardless of whether the host category is CP/Comp (as in German proper) or TP/T0 (as in more strongly contact affected dialects of Cimbrian German) even when the clitic is a calque from IR. 1.3.7 We take it as a given that the description and explanation of linguistic change according to 1.3.1.-3. above presupposes a grammar-theoretical framework which is based on principles of least resistance and strongest procedural economy in correspondence with Universal Grammar and modern generative syntax (expanded by criteria of Distributive Morphology).

1.4 What precisely are the main structural phenomena of German (D) and of Italo.. Romance (IR) - those that need to be considered most prominently given that one cannot exclude change of Cimbrian German under contact with IR? 1.4.1 Verbal government direction and verbal bracket: Verbal government is leftward oriented in D, counter to verbal rightward government in IR. Furthermore, the very existence of the verbal bracket (V2 as well as Vlast) and, concomitantly, the wide middle field between the two clausal poles, V2 and Vlast, presupposes verbal leftward government. This relates to the following clause structural details:

il. "(AUXn_ 11 >) object > PPV" in D, whereas "(Auxfin >) PPV > object" in IR. [AUX(iliaryverb), fin(ite), PPV =perfect participle of the verb] i2. principled serialization in a multi· membered verbal complex; i3. main-(V2&V1881 )/embedded clause asymmetries (V2 and V1ast as opposed to only i4.

V13_.1 or Yearly as in English or French). Is Vlast in Cimbrian satisfied, or has Cimbrian already surrendered to Romance SVO? Are there structural interface forms between German SVOV and Romance SVO?

Let us consider (il)-(i4) more closely in the framework of the clausal structure in Standard German as in Figure 1. In contrast, Figure 2 sketches the corresponding clause structure of Italian. Figures 1 and 2 do not have the status of invariable templatic structural descriptions. Contrastive clausal accent determines whether a constituent is

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 239

in basic or in derived position. Only subjects are unaccented in the Fore Field/SpecCP; all other clausal functions go into this position under contrastive accent. DISCOURSE THEMA/RHEMA FORE FIELD lepSpec.CP

STRICT V2 ~ FIRST 'IBRACKET [C"

THEM A

lnal ciiUcs

-ts

Aber

sle Wleso

-st

RHEMA

DP, full pronouns (es)mirdes dumirdes

lw adverblals

lw

SIE

glbt ~enn

lslm

-

=>

MIDDLE FIELD

~(pronom-

Ob Ob

=> ~

Vlast ECHO THEMA SECONDVBRACKET PosTFIELD 1/0 Ill rl!tJt dlslocatlon heut? ~AG-TS heut? ~AG-ST

heute.

AB

fJaBGEBEN?

Figure 1. The computable relation between clausal structure and discourse function in German (the two shaded fields mark the poles of the verbal bracket)

Germanic and Romance are essentially distinguished by the structure of the Fore Field in these languages: German does not make use of an extension such as the one found in the first column in Figure 2. Rather, German projects at most one extra categorial node beyond SpecCP: 2 a properly governed about-topic (not represented in Figure 1) taken up pronominally in the Middle Field (MF). It is perhaps redundant to say that Romance has no double-polar verb bracket (strict V2 as well as Vlast) as German and Dutch do. This latter property marks and distinguishes most prominently the two language groups in terms of grammar (government direction) and discourse properties (see Abraham 1997, 2005a). TOPIC-Focus-TOPIC lrOP p XP-YP-ZP-

PROCLISIS

Vfin!Comp-TP(-encllsls)

fu nctlonal morphology

lexical oomplements MIDDLE FIELD

pronominal diUcs

lep C [TP IIP(Neg)

VAux (MF (mai)]

l,.p V-IIP1-IIP2 ... JD

Figure 2. The computable relation between clausal structure and discourse function in Romance (with a reduced Middle Field)

For this, see also more closely Figure 3 below. The differences between the clausal structures of German and of Italian are quite clear: 1.

2.

2.

IR has only one single verbal field (i.e. it does not have asymmetrical structures in dependent and independent clause structures, as opposed to German). IR has radical proclisis of weak pronominals (the only exceptions being infinitive and imperative enclisis); German, by contrast, has enclisis without exception.

Not considering hanging topics, which are outside of governance by the clausal predicate.

240 Werner Abraham

3.

Where German most clearly and unambiguously distributes discourse themata and rhemata by left-right serialization in the Middle Field and has only one single position for contrastive rhemata, Spec,CP, Italian furnishes three Fore Field positions ("expanded CP" in line with Ri7zi 1997). It also tends to put discoursefunctional material in right dislocation. By contrast, Gemtan has accent options beyond the discourse-timctionally decidable positional slots in the Middle Field and in clause-inceptive position to distinguish thema and rhema. IR, on the other hand, does not make equally systematic use of such options. IR has rightward verbal government (right column in Figure 2), German, by con-trast, has leftward verbal government (DP moving into the Middle Field or into Spec,CP). IR has sentence negation to the left of finite V, after the subject, in independent clauses, while German places negation to the right of finite V. IR has virtually no Middle Field elements, since it does not project a verbal bracket (Y:O&OY.). See (2) below.

4.

5. 6.

(2)

VAux [MF(mai)] [yp V-NP1-NP2 ... )

The discussion in the follov.ing section proceeds with these structural differences in mind.

1.5

Conclusions for Cimbrian

Pronominal clitics: German always has them .~.f.te.r. V2 in the Wackernagel position/WP ('end isis'); in Romance, by contrast, they are always before the finite V (i.e. in 'prodisis'). Other than in the neighboring IR contact dialects, the Cimbrian of Lusema is not a pro-drop-language, and it uses expletive es "it" enclitically. Yet, expletive pronouns also occur in the neighboring Friulian dialects (e.g. with the weather verb corresponding to German es regnet "it is raining").

1.5.1

Strict V2 and subject inversion: The option of a non-subject in sentence-initial position exists in Standard German, but only in parts of Cimbrian - in other words, the subject in SpecCP, the Fore Field, is not only strongly preferred, but is the only option in most Cimbrian dialects.

1.5.2

Thema-rhema-discourse order in Romance in terms ofleft or right dislocation (cleft constructions; CP--expansion): See Figure 2 where the dislocation of a thema to the right periphery in Italian does not have a clause--external pendens function, but is clause-integrated.

1.5.3

Spoken syntax in Cimbrian of the linguistic !.~lands in Northern Italy 241 2.

The facts- illustrations from Cimbrian dialects

ad 1.4.1.(il) German Aux-DO-PP (clause bracket) is replaced by (what looks like lR) Aux-PP-DO '!here is no clausal bracket (S(V)O(V)) in Cimbrian. However, see (3a,b) below. Does this mean that (contact-induced?) change begins at the sentential right periphery (as has been claimed by Grewendorf & Poletto 2005)?3 Why is only the right periphery targeted? In what f(.)llows all illustrations refer to Luserna (following Bidese eta!. 2005: 82) except if signaled otherwise (such as for '[Ruana]'): [do = pronominal direct object, DO = nominal direct object] 4 (3)

a.

Haute die Momme [yp [y hat gebiischt] die Piattn] A11x-Participle-DO heute die .Mamma hat gewaschen die Teller 'today mother rinsed the boards'

b. *Haute the Momme [yp ltat [yp die Piattn [y gebaschtlmt]] 'today the mother has the boards rinsed' ,.Aux-DO-Participle (4)

Sa hom

khott

ke

dar Gianr1i hat net

gebOllt gian

pit

se

sie haben gesagt dass der G. hat nicht gewollt gehen mit ihnen 'they said that G. did not want to come with them' Aux··neg·Participle··PP (5)

a.

I

hott

r1iemad

gesek

Aux-DO-Participle

ich habe niemand gesehen 'I did not see anyone' b.

han-ich khoome gflSeecht habe ich kaum gesehen 11ave ·I hardly seen'

[Ruana] Aux-neg-Participle

3· Grewendorf & Poletto (2005), on searching for universals in the process of change from OV to VO, focus on whether the right or the left periphery are easier entries for structural changes and on exactly whiCh of such changes. Given that the left periphery host<~ illocutive and discourse representations, while the right periphery is the genuine location of lexical verbal representations and information, it would seem interesting to see whiCh of these two peripheries admits structural intrusion more easily and under which categorial pressure for change this occurs. See, for a discussion of typological hybrids between OV and VO, Santorini (1994a,b) as well as Haider (2009). See also Grewendorf & Poletto (this volume). 4· Obviously, I have focused on the syntactic data assembled in the illustrations and corpora presented in the literature (most importantly, Bidese, Poletto, Rowley, and Tomaselli). The latest work by Schweiz.er (2008) had not been available when the final touches had been laid on the present paper (end of 2007). Tyroller (2003) was consulted to ensure that the present empirical conclusions are well founded.

242. Werner Abraham

Bidese et al. (2005: 82) identify Cimbrian "middle field inserts" (more precisely: "Aux + Participle- inserts") as being restricted to short categories such as: pronouns, negation, monosyllabic adverbials, and verbal particles as well as short quantifying morphemes. These are hardly categorially unambiguous, unifiable material, but, rather, phonetically light elements that do not interrupt the prosodic unity of predi.cative Aux + Participle in any prominent fashion. 5 What does this mean tor the typological question of"SVO or SOV"? I \\
(7)

a.

Biar zeteren nete wir geben-nach nicht 'we do not give in'

b.

se nette ghenan vuar dass sie nicht gehen wetter 'that they not go further on'

a.

Noch in de erste Lichte von deme Tage noch beim ersten Tageslicht

az

[Lusema] Bidese et al. (2005: 82)

0 0

0

vuar =verbal particle

hevatt-se-sich aile erheben sie sicl1 alle/stehen sie alle auf 'already ~ith first daylight they get up'

b.

'az se sich legen ir1 Kiete class sie sid1 legen in RU:he 'that they calm down'

(== beruhigen)

az

(8)

de Consiliere ghen nette on in de Sala dass die Beamten gehen niclrt hinauf in den Saal 'that the civil workers do not walk up to the hall'

(9)

'az diese Loite richten-sich ... . dass diese Leute richten sich .. . 'that these people accommodate-themselves .. .'

5· It is well known that pronouns and quantified indefinites behave differently from lexical NPs (see Abral1am & Wiegell993). Given this, it is no small wonder that pronouns and quantified NPs lag back in linguistic change (irrespective of whether or not under contact) in diachronic development and that equilibrium evolved and something new could evolve from it. Clearly also the left periphery does not look that clean any longer when one thinks of the change ofV2 in Old English and when compared with the formation ofV2 in the southern linguistic enclave dialects.

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 243 (10)

umbrume di

andar

Lentar saint net

contente

denn die andem Dorfer sind nicht zufrieden 'since the other villages are not pleased' (11)

umbrume dear .4far

hat-sich gamachet gro,lz

denn die Sache hat sich gemacht 'since the istme has been blown up'

gro:6

These illustrations warrant the following generalization: The asymmetry holding between dependent and independent clauses characterizing Standard German is suspended, albeit it appears reconstructed just for phonetically light inserts (clitic pro-nouns, explicit and implicit negative elements). 6 This appears to invite the conclusion that such split Aux-PP structures do not simply represent VO (V inside ofVP), but that V raises to a higher position (to the first V-bracket neighboring Compas in German). Cf. Figure 1 above. We shall return to this later. ad (i4) above: As a rule, Vlast is not satisfied in Cimbrian; see the predicate periphrastics in (12)-(13): The past participle links directly to the finite auxiliary both in Cimbrian German and in Italian, as shown in (12)-(13). Does this mean that Cimbrian has yielded to IR SY.O/ Adv or Italian (X)VSO? Furthermore, consider the differenc-e of the pronominal positions between Cimbrian (12)-(13) and Italian (14): Italian has proclisis, whereas Cimbrian has enclisis, entirely in line with German. (12) shows subject inversion typical of Gem1an, with all instances of rightward verb valence (DO before SUBJ); Vl in the declarative as in (13) may be tound in German dialects ((s.EXPL) Is amol a Konig gwesen, der ... '(it.EXPL is once a king been'= "There was once a king who ..."); (13) is totally void ofanyverbai bracket (pure VO). (12)

Gheistar hat gessat

dain

Manestar iz

gestern hat gegessen deine Suppe 'yesterday the girl has eaten your soup' (13)

Hat gahakat iz

Holtz dain Vatar

[Giazza] Diarlja das Dirndl Bidese &Tomaselli (2005: 79) ... VI thetic declarative [Giazza]

hat gehackt das Holz dein Vater 'your father has chopped the wood'

In fact, since my fundamental claim L'l that prosodic regularities (trochees and dactyles, at least in German) are at the bottom of constituent -internal order, this phenomenon receives a natural explanation. Light, monosyllabic morphemes represent the weak components of the unbow1d, mmpheme-independent, trochaeic and dactylic s(trong)-w(eak) patterns a lot more readily than longer unbound words. Such prosodically motivated order plays a role also in (often monosyllabic) multi-membered verbal complexes in the right clausal periphery of oral encoding (see Abraham 2001, 2009). 6.

244 Werner Abraham (14)

( 15)

Lo 1Janno comprato al mercato i miei genitori es haben gekauft am Markt die meinen Eltern 'my parents have done errands in the market' c.f. Italian La taia la legna to papa

's 'it.BXPL

hat gehackt das Holz dein Papa has chopped the wood your father'

Note that these data allow for no identification of prosodic and accent properties with respect to distribution between them a and rhema status of object and subject positions with respect to the perfe<..t participle in (12)···05). Any ofthe subje<..ts in clause--final position may be topic/thema just as much as any of the earlier objects may have the status of a clausal rhema- neither can one exclude subjects under contrastive accent. ad 1.5.1 above: genuine pronoun ditics (ditic doubling) as well as pronouns in WP, also V60-enclisis is counter to IR Pro-Vfin' with IR V-proclisis strictly retained. The left periphery of the clause remains more consistently resistent to change. Bidese et al. (2005: 83)

(16)

az sai-der getant diar es sei-dir1 getan diri 'it be-you done to you'

(17)

Hoite de 1\-fuatar hat-se gakhoofet de Ojar in Merkaten die Eier am Markt.DAT heute die Mutter1 hat-siei gekaufl: 'today mother bought -she the eggs at the market'

[Roana]

(18)

Giantli hatt-ar-mi gaboorset, bear hat-ar-dicll telephotJaart G. hat-er mich gefragt, wer hat-er did1 angerufen 'G.1 has-he 1 me asked whoJ has-heJ called you'

[Roana]

Cimbrian has no empty pronominal subject (it is not a pro-drop-language) as opposed to Italian although occasionally pro--drop is documented in 17th century Cimbrian (Meid 1985: Katechismo 1602): see the gloss "0" for "zero" in (21). (19a) compares the Cimbrian and the Italian patterns with one another. (19)

a.

Gianni was hat er gekauft? 'G., what has he bought?'

b.

Haute han-i gaarbatat ine Baije heute habe ich gearbeitet in Walde 'today have I worked in the forest'

Gianni che cosa ha comprato?

[Giazza]

expletive marking

[Giazza]

(20)

Haute iz regat Haute regat--z heute es regnet heute regnet--s 'today it rains today rains--it'

(21)

Disa Kirka, barume cheu-0-sik Heilega, unt Catholica? [Katechismo 1602] diese Kirche, waiUm nennt 0 sich heilige and katholische 'this church, why 0 calls itself holy and Catholic?'

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 245

ad (iii) above - strict V2 as well as subject inversion

3·

Non-subjects in the Fore Field (=SpecCP) (i.e. the Cimbrian German subject may invert in V2-dedaratives): Grewendorf & Poletto (2005: 121) show this using (22)-(23)

3.1

(22)

Hainte tuit de Marne de Kinder ins Pette. heute tut die Mamma die Kinder ins Bett 'today does the mother the kidc; into bed'

(23)

a. *HaUte geat dar Gianni vort.

[Pladen German]

[Cimbrian: lllsema]

heute geht der Gianni fort 'today goes the G. out'

b.

Haiite geat-a vort ttiamat. heute geht-da fort niemand 'today goes-he out nobody'

See the distribution in (23a) and (23b ): Pure subject inversion exists in Pladen German, but not in the remainder of Cimbrian except in Southern Lusema: It is difficult to ascertain whether or not (geat)-a "(goes- )he" contains a coreferent clitic pronoun or an agreement suffix at the end of the verb. Bidese concludes (Bidese 2008a,b) that today's Cimbrian should have agreement status for this elitic element. However, it is not always easy to find unambiguous distributions in the data to warrant clear conclusions on this matter.' Instead of subject-NP-inversion, there is only pronominal clitic inversion - clearly with variants across the Cimbrian-speaking communities ("more modern" Giazza as opposed to to "more conservative" Lusema; cf. Bidese, Poletto & Tomaselli 2006: 81):

3.2

(24)

Berme di andarn drai Len tar habent gahiirt as6, wennials die andern drei Dorfer haben gehOrt so, haben-se-sich manegiart ... haben sie sich angestrengt 'when the other 3 villages have heard so have they made great effort· [Bidese & Tomaselli 2005: 81]

7· geat--a must not be read as 'geht--er' = "goes--he'': but as 'geht--da' = "goes··PART(ICLE)': where da is a temporal adverb signalling the specific contextual characteristics of the activity describe by the predicate: alie sutzta lalltnda di klokkng'alle Sonntage liiuten--da die Glocken' = "every Sunday the bells ring--there" (vgl. Panieri/Pedrazza/NicohtsSi Baiz 2006: 315···317). Thanks to Ermenegildo Bidese for help on this. For clitic equivalences in Northern German, see Abraham & Wiegel199 3.

246 Werner Abraham

(25)

Am

boute gan Ljetsen hen-se

getrust" gien

einmal in Giazza haben--sie miissen gehen ... 'once to G. have-they must go' (26)

In Sontaghe regat--z- V2 In Sontaghe iz regat-V3

(")SubjNP--Inversion [Giazza]

'on Sunday rains--ition Stmday it rains' 'on Sunday rains-it/on Sm1day it rains' (27)

(28)

*Haute geat dar Giani vort 'today goes the G.

away'

Htiute geat-ari vort,

dar Gianii

'today goes-he away, the

*SubjNP-Inversion [Luserna] SubjClitic-inversion [Luserna]

G:

Compare again (24) with (26): As soon as the pronoun appears clitically reduced in Cimbrian, it is enclitical with V2 as in (24). Example {27) differs from (28) in that the verbal bracket may host only prosodically short elements as the elitic pronoun -ar "he': not, however, full nominals as dar Giani "the Giani.PROPER NAME': 3·3

Strict V2 given up in favor of Romance topic/thema-V3/4

(29)

(30)

Gheistar-1 in Giani-2 htJt galu.iktJt iz Holtz ine Balje/in Bait-- V3

[Giazza]

gestem der Gianni hat gehackt das Holz in Wald 'yesterday the G. has chopped wood in the forest'

Bidese et al. (2005: 82)

De muotar-.l gheistar-2 kam Abato-3 hat lroft

iz

Mel- V4

die Mutter gestern in Abato hat gekauft das Mehl 'the mother yesterday in A. has bought the flower'

3·4

Conclusion

The core of the V2--rule, i.e. finite V in independent declarative clauses strictly in clause--second position, Comp/C0 as in Figure 1., as well as subject inversion if a non-subject occupies SpecCP appears, was weakened in Cimbrian through V3 and subject inversion only for the pronominal subject. 1he direct left periphery of the clause is unaffected with the proviso that the discourse topic/thema remains far to the left, apparently in WP, as in Standard German.

4·

ad (i2) above: Serialization in the multi-layered verbal complex

The question is whether the scope extended by the clausal head (of the predicate carrying agreement) is directed to the left ("head finality") or to the right ("head initiality")? The data are however more complicated than this. Notice that the verbal complex in Dutch (fundamentally SVOV just like German) exhibits the opposite pattem when compared with German: Dutch displays the ordering t_..,c;R-2-3 (in line with VO?),

Spoken syntax in Clmbrian of the linguistic island~ in Northern Italy 247

whereas German has (1AGR)3-2(1AGR) (as expected with OV). This leads us to condude that Dutch, as an SVOV type, nevertheless setializes as aVO language. However, this conclusion is not correct for two reasons: First, even German tends to place the finite component of the V-complex early on, (1 AGR ), often in a position identical to that of Dutch. Second, we concur with Haider (2007) that V-complexes are a privilege of OV languages and that true VO types such as English and Romance are not amenable for assembling multi--layered V-complexes, in the first place. 'Ihis, then, leaves us with the conclusion that head-initial order in the V-complex is the result of process and parsing facilitation. See a brief survey ofV-complex orders across a variety oflanguages in 4.1. below (Cinque 1999, 2007; cf also Hr6arsd6ttir 2009: 38).

Auxiliary-main verb serialization

4--I

«Aux1>> is always «finite Aux»; «Aux2 Aux3» may be modal verbs or other auxiliaries in the verbal complex;
(31)

a.

Aux 1 (Aux2) (Aux 3) V

SVO-languages (Italian, English, ... ) if at all possible, in the first place no representation

b. •-Aux3 AuJS Aux1 V c.

(V) Au.x1 Au~ Aux:3 V (West-Flemish, Dutch)

d.

(V) Aux3 Aux2 Aux 1 V (Standard German)

Clause bracket (seemingly under OV > VO) Clause bracket in OV /\t1ast -languages

Compare the more general serialization of mood, tense, and aspect morphemes in (32) below (Cinque 1999, 2007; Hr6arsd6ttir 2009; Abraham 2009a).

4--2

Speech act morpheme serialization (mood, tense, aspect, modality)

'Ihe generalization is marked down in (32) (following Cinque 1999). See German in (33), Dutch in (34), and the Cimbrian illustrations in (33)-(35). (32)

a.

b. c.

mood-la tense-lb aspect-2 V V mood-la tense-lb aspect-2 V aspect-2 tense-lb mood-la

(Yoruba/Nigeria) (Comox) (Korean) OV tVlast -language

v\lhatdo V-complexes look like in German and Dutch? Cf. German (33) with the serialization 4-3--2( l) as well as the opposite order in equally leftward governing Dutch (34) with 1.. 2--3-4. Cf Abraham 2009 tor more details. [sMALL CAPS for default accent in the V-complex].

248 Werner Abraham

(33)

er (hatte-1 AGR-TENSE-MOOD) d.as sehen-4v konnen-3MODAUTY this see could he had (hiitte-1) wollen-2ASPEC. had wanted

(34)

dat hij dit had-lTENSE(-MOOD) wiJlen--2ASPFCT kunnen--3MODALITY zien-·4y that he this had wanted can see

It is important to see that scopal order is reflected by verbal government direction only in German, not, however, in Dutch and Cimbrian: See German 4-3-2 (irrespective of finite 1), whereas Dutch and Cimbrian-Luserna (Bidese et al. 2006: 82) pursue the order 2-3. 1his makes us conclude that the linear order in V-complexes is not determined by V-govemment direction and that it is variable under facilitating parsing ease. 1he same holds for Cimbrian. Compare (34) above with (35)--(36a-c)-(37). (35)

Sa hom khott ke dar Gianni hat net gebOHt gian pit se. [Lusema] sie haben gesagt class der G. hat not gewollt gehen mit ibnen 'They have said that G. has not wanted go with them: Au.x60-Neg MV- Vinfin PP 1 2 3

(36)

a.

I hon gamiast suachn s Puach ich habe gemusst suchen. 's Buch I have must seek the book 2 3

b.

bail deist ano gaben pet de hend za belln helven aushorn de 1schupp

[Fersental] Rowley (2003: 279)

2 3a 3b weil der ist auch n.od1 gewesen mit den Hii.nden zu wollen helfen ausheben die Fichte 'because he is also been with his hands to will help dig out the fir tree' c.

(37)

[... ] bo er se

recht garichtet hot gahop 3 1 2 wo er sie recht gerichtet hat gehabt 'where he her well pointed has had'

[Luserna] mage hom oogerilaf 3 2 1 ich frage-mich zu sehen wer mich mag haben angerufen 'I ask myself to see who me may have called' Bidese et al. (2006: 86)

I

vorsmaar

zega

bar me

As stated already above, I share Haider's (2007) view that any such multi-layered V-complex serialization as 4--3--2--lfor German or 1-2--3--4 in Dutch and Cimbrian occurs only in languages of the right-branching type. Left-branching types, such as English, do not have such complexes in the first place. For variants among the princ.ipally SVO--vernaculars in Scandinavia see Hr6arsd6ttir 2009.

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 249

43

May we take the Cimbrian V-complex to be an instance ofverballeftward government? Does this mirror Scandinavian SVO with enclisis? Or are the facts indicative of Cimbrian-autonomous clause-process facilitation, which is something I would also assume for Dutch and Yiddish? Only the OV-type has V-complexes of the German--Cimbrian kind. Notice that modal verbs, no doubt paradigmati-cally manifest only in Germanic as opposed to Romance, play a prominent role in the multi-layered V-complex.

4·4

The OV-type has DP-extraposition (not only for objects) almost exclusively for definite nominals. We have no evidence for Cimbrian object right disloca-tion of indefinites (which could be indicative for a true VO-type)? Consider the Bavarian-Austrian dialects with only oral encoding, where such dislocation is not observed. Thus, in Cimbrian change starts with Vlast in the right sentence periphery. Cf. (38) and (39).

(38)

Gheistar in Giani [,1p hat gahakat [DO-DEF iz Holtz ine Balje!in Bait]] [Giazza] gestern der Giani hat gehackt das Holz in Wald 'yesterday the G. has chopped the wood in the forest' Bidese et al. (2006: 8)

(39)

De muotar gheistar kam Abato fvp hat koft fDo-DEi' iz Mel]] die Mutter gestern in Abato hat gekauft das Mehl

'the mother yesterday in A. has bought the flower'

5· GP-ditici7.ation next to the variant TP-clitici7.ation 5-1

German pronominal cliticization to Comp

Consider the column WP in Figure 2. for Fersental (Roana, Luserna): There is e"Vidence for Comp-CL (next to the variant Vfin -CL), albeit not for the 13 communities, where only Vfi 0 -CL exists. See again Figure 2 as well as (40) below with data from Bavarian-Austrian. In Alemannic we do not find such subjunction inflection (with rare exceptions as documented in Dal Negro (2004: 168, Example (20)-(21) in the author's account of the High Alemannic retreat dialect in North Italian Formazza, Wallis). Cimbrian subjunction inflection (Comp-CL) is illustrated in (40)-(41), while illustrations for Vfi0 -CL-variants follow in (42) as well as 5.2. below. (40)

... , ob-ts/-st(es.2PIJdu.Iso) mir des sag-ts!-st

... if-you/-you

sag-ts/-st des tell-you that

heut- *ob es!du me this tell-you/you today *if you/you

2.50

Werner Abraham (41)

a.

Bas-er kat, ... was-·er sagt 'what--he says'

(42)

[Roana] Bidese et al. (2005: 84ff.)

b.

ben--ig--en r10ch vinr1e wenn-ich-ihn noch finde 'if"- I-him still find'

c.

az-da-ich gea au dass-da-ich gehe (hin)auf 'that.-there·-I go up'

a.

die 'latta hat-se gelwaft der Vater hat--sic gekauft 'the father has-her bought

b.

Habban-sa mich gaboorset, ba gehe-bar haint haben-sie mich gefragt, wo(hin) gehen-wir heute 'have-they me asked where go-we today'

"De r Tatta se hat gekoaft the father her has bought

More generally, with respect to German and Dutch as well as West- Flemish, see Haegeman 1993; Hoekstra 1993, and Zwart 1993. 5.2 In contrast to the data in 5.1., there is also Vfi 11 -CL in dependent clause structures- this even more in conservative Luserna opting for Comp-CL as a variant. See the tt)llowing illustrations with dependent clauses and pronominal cliticization to the finite V though not in clause Vlast position. See the short form Vfin··CL for this kind of data. (43)

a.

I woas, ke der Tatta ("se) hat-se net gekoaft ich weiB, dass der Vater (sic) hat-·sie net gekauft 'I know that the dad (them) has-them not bought:

b.

Gianni hat-ar-mi gaboorset, bear hat-ar dich telephonaart G. hat-er-mich gefragt, wer hat-er dich angerufen 'John has-he-me asked who has-he you called:

[Lusema]

[Roana]

The clitic pronoun (hat)-se "(has)-she" in the dependent, Camp-introduced ke "thaf' sentence in (43a) projects in to, not in Comp/C0 as in Modern Gennan. We conclude that V6n··CL is an overgenerali1.ation following the German pattern of the C:Lsuffix-ation to finite V. This is counter-evidence to the otherwise prominent Cl.-attraction to Comp. This clearly shows that V6n is in TP -but not in CP. Furthermore, the fact that (hat)-ar "(has) he" likewise does not project in Comp appears to be because the eli tic pronoun has the status of a resumptive pronoun. Notice that bear (== German wer) "who" in Spec,CP is coreferent with ·-ar "er/he". Its diagnostic value for Comp--CL as an option for Vfin -CL appears to be minimal.

Spoken syntax in Clmbrian of the linguistic island~ in Northern Italy

5·3

Just for Luserna: Comp-CL in relative and dependent interrogative clauses

It is interesting to see that the CL-question offers a further complication. Cf. (44a-c). (44)

a.

b.

an Libe Bidese et al. (2005: 86) lw-t geet 's Baibe, bo-da-r--ett s Weib, wo-die-ihr-ihn hat gegeben ein Buch' 'the-woman where/who-she-her-him has given a book: dar man, bo-dar-en

(er) hat geet

an Libar

der man, wo-die-ihm (er) hat gegeben ein Buch 'the man where/who--they-him (he) has given a book' c.

I

boas net, ber--me

hat ogruaft

ich weiH not, wer-mich hat angerufen 'I know not who-me has called:

d.

I

voorsmaar zega

bar-me

mage hom

oogruaf

ich frage--mich COMPL/zu sehen wer-mich kann haben angerufen 'I ask-me to see who-me may have called:

In relative clauses such as (44a,b) and embedded interrogatives as (44c,d), the German CL-position turns up again no doubt. as opposed to the quotes in (43a,b). This makes our conclusion with respect to (43a,b) even stronger: Luserna with CompCL is the more conservative ("more Modern German") diticization version, while (43a,b) are exceptions to the rule. In other words, Luserna represents the more conservative German version deviating to Vfin--CL, the shorter clause structure, only by exception. See 5.4. 5-4

The deeper link between pronoun cliticization and predicate position

The diachronic development of subjunctive inflection in (dialectal--oral) South German, as well as the reduction of the weak pronouns to become cliticized to Comp, warrants the assumption of a elitic constituent in ist own right, CLP - i.e. one between CP and TP (mirroring the development of the Cimbrian pronouns ever since 1602 (see Katechismus, Meid 1985) as discussed in Bidese 2008a,b)). The original pronoun in Spec,CLP reduced to a Comp/V--sutfrx in CL0 and percolated higher up to Comp or V, respectively, in the highest clausal node depending on the status of clausal dependency. Consider Figure 3 below. This holds for Standard and South German, for Cimbrian, and for Italian and its dialects in the same fashion, although not under identical verbal government direction in its diachronic emergence (sees for this Poletto & 'lbmaselli 2000). Other than for Modern German, we posit AgrSP above TP, but below CLP, for IR. Finite V raises to AgrS 0 in IR, as opposed to C0 in German in independent clauses (the evidence being dialectal subjunction inflexion and, most importantly, complementary distribution between C0 in dependent and V0 in independent clauses). All of

2_51

252

Werner Abraham

these operations result in enclisis in German, counter to proclisis in IR - the eroded IR-pronoun raises no further than to AgrS 0• It is important to see that Cr.P in German is bound to finite V more strictly than in IR. See Romance proclisis in (45a,b) with the adverbial insert (Poletto & 'J.bmaselli 2002; their (1.)···(2) on p. 11.), which is totally impossible with Cr.P in Gem1an. (45)

a.

Il

fort bien parler

en

du

a

he has must thereof.CL Adv

b.

French

speak

El me sempre dizi he me.Cr. always.Adv told

Triestinian

c: ~

dt wenn-a-ngsehen hat!Hat-a-ngsehen

CuTP/\'IlPP

V-CL

~

Spec

CJ!1WP' ~

CLc'/WP 0 Ia

ital.la ho >'sf:a/la veda( dew vedere

AgrSP ~

Spec

...--........, Agrft

.AgrS0

Vrm ko!-oi-o

TP _.........., Spec 'I''

ital. lwl-o/-o

_,..............._,

T'

AgrOP

(Vf.,,)

~

Spec

AgrO'

·---~ Agroc• vista.PP

VP

~

V'

Spec

~

tv 0

DP

ved-/dev1

V' ~

DP

V0

gesehen.PP

Figure 3. Clitic constituent, CuTPiWPP, and V-sentence in German and IR

Modem German raises the finite verb to C0, whereas the clitic pronoun, -a-n "er-ihn': raising from CL0 to C1\ remains in enclitic position. On the other hand, Italian la remains in proditic position with respect to the finite verb in AgrS0• SpecAgrSP may host adverbials as in (45a,b ). The Italian PP, vista, has been raised to AgrOP due to the requirement of gender and number agreement. By contrast, the Gennan PP in predicate periphrasis remains in situ inside VP. It has not preserved its original Early

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 253

High German agreement with the object inherited (just copied?) from Late Latin (consider EOHG .. .habem figboum.M.Acc gipflanzotan.M.Acc).

5·5

Typological comparison

The illustrations so far permit the following Germanic comparison with IR. See Figure 4.

Strl ct V2 (Vfinltoln C): fr ("gerne) Continental

Predicate bracket

LAuxtc"-PPNfi.,): _<.hat>/ dass..ge-

redet

tan

+

+

Null subject: tiamo

-

Expletive subject: fs frostelt mkh

En-vs.

C0 -pronouns/

V-

proclisis fr("s-) sagt=l!

enclisis (counter

complax

to vfi,-cllsls)

+

+

+

+

+

+

+

+

+

+

+

+

+

+

-

+

-

+(obj.)

n.a. n.a. n.a.

-

West German lc

(German, Dutch) Scandinavian

+

Lusarn/

+1-

Roan a

remaining-

+I-

Ombrian

English French

Italian

-

-

+

-

[un.a." = nnot applicable"]

Figure 4. Typological comparison under the V2-criterion (expanded and refinded beyond Tomaselli (2004: 537))

This is what Figure 4 says: Comp-CL-languages have the long clausal structure of Modern German, CP, whereas the ordering Comp + V6n-CL characterizes the shorter clausal structure of Romance. Only Comp-CL-languages share with Modern German the predicate bracket resulting in asymmetry between dependent and independent clause structures. Romance is symmetrically divided between dependent and independent structures, just as much as English is. Scandinavian, on the other hand, has a small Middle Field, despite its VO-status in both dependent and independent clauses. See however (47 a-c) below. Is this Scandinavian structural exception to clausal symmetry mirrored in the Cimbrian of conservative Luserna and Roana as opposed to the more modern remainder of Cimbrian standing (as a microvariety) for the OV-VO macro-transfer from Proto- Germanic to modern VO inScandinavian (with a remainder of OV in the small Middle Field)? We feel encouraged to think so by Brugmann's admonition about the diachronic impenetrability of syntax (his 'UndurchHissigkeit der Syntax': "auf syntaktischem Gebiet sind Entlehnungen meistens viel weniger leicht zu erkennen als in anderen Gebieten der Grammatik" (= 'on the syntactic level, more often than not, loans are less easily

254 Werner Abraham

traceable than on other levels'; Brugmann 1917: 54). This position is subscribed to by the Veronese enclave dialectologists insisting on highly constrained Romance influence on Cimbrian ("beschrankter Einfluss romanischer Dialekte auf das Zimbrische''; see Bidese & 'Ibmaselli 2005: 88). 5.6

See (46) below as an interim result about diachronic loss ofV2 across different structural steps: [the arrow,>, means "results in."]. Referenc.e is made to Figure 2, the illustrations (15)---(17), (27)···(28) as well as (35)···(39) and, finally, (44a---d).

The relative head and, respectively, the thema pendens (hanging topic) in the depen-· den.tinterrogative as well as relative clause, w-V6n-Pron.i (ct: "[0 p der Manni [M00 der/ wer; hat-er; ... ]]"), leads to structural ambiguity, namely: resumptive pronoun + w- Vfin Pron; ("[DP der Ma.nnJ lcp der/weri ha.t-er,T) and, finally, by grammaticalizing reanalysis: declarative clause such as "deri (Mann) ha.t-er;···"- which terminates in loss ofV2. Note that this never collocates with prodisis (CL before finite V60 ). In other words, Cimbrian CP-clausal structure holds as in Figure 5 for Modem German. From this follows SVO-clause order as in the Germanic languages of Scandinavia; see (47a-c) (from Wiklund 2009: 27). Compare Cimbrian (Sa,b), (12), and (15) ["( ... )"for alternative positions].

(46)

a.

1rulega hej(Ji }otz keypt b6kina probably has J. bought the book

Icelandic: subjed inversion

b.

. .. ad .l6n hefdi tritlega keypt b6kina (small) predicate bracket ... that J. has probably bought the book

c.

Olle sa att han (inte) hade (inte) liist boken 0. says that he (not) had (not) read book.the

Swedish: (small)

predicate bracket

Although there are embedded clausal structures deviating from pure VO in Scandinavian, Cimbrian also has regular structures not in line with otherwise VO, as in Au.x6n-PP-DO, such as AU4m-pro-do/neg-PP, opening a small Middle Field for pronouns, and, furthermore, pronominal adverbials as well as negative elements; consider again (3a,b) above. This leads to the assumption that such short elements are lexically merged in this discourse-thematic position and should be evaluated as a remainder from original ov.&

One could also consider a separation of the two positional variant.s,Auxfin-PP-DO as well as Au:Jrsn -pro-do/neg-PP, because of featural differences and Chomskys DlSLOC-mechani.sJIL Cf. Chocano 2007, completing this program for Scrambling in German and the Scandinavian Germanic languages (dative shift).

8.

Spoken syntax in Cimbrian of the linguistic !.~lands in Northern Italy 255

6.

Clear grammatical calque from Italian?

The infinitive as a elitic host - which is impossible in German ( CL only in WP, i.e. in right--adjunction to Comp; see Figure 2 appears to be a grammatical interference from Italian (IR ?) beyond doubt; after all, infinitives have no syntactically identifiable position). The clitic suffixes in the following illustrations have been highlighted. [SMALL CAPS for contrastive accent]. (47)

Non so do~·e metter--Iii (i libriJ nicht weiB (ich) wohin legen--die (die Bucher) 'not know (I) where put--them (the books)'

(48)

Penso di mangiar--la; piu tardi (la torta) (ich) glaube zu essen--sie spater (die Torte) '(I) believe to eat-her later (the cake)'

(49)

Vado a portar-glie--la; io (la torta) (id1) gehe zu bringen-ihnen-sie ich (die Torte) '(I) go to bring--them--it1 I (the cake/

(50)

Se saint gant grussan-se; alle1 sie sind gegangen grii.Ben--sie alle 'they are gone to greet--them all'

(51)

ribalan--se aus zu reiben--sie aus 'thus rub--they out'

(52)

net khoas andarst habent--s gebellt khoofan nicht kein anderes haben--sie gewollt kaufen 'not no other have-they wanted buy'

(53)

tse puosse--mi zu beichten-mir 'to confess-me'

(54)

tse vorkhovosei d'ouksanl zu verkaufen--sie d'Ochsen 'to sell--them the oxen'

Castagna (2005: 104---105)

[7 Communities]

[13 Communities]

See the Italian patterns in (47)---·(49) as opposed to (50)---(54) in Cimbrian German. We might assume - instead of simply taking it for granted that it is a grammatical interference from neighboring Italian-, the assumption that it is an overgeneralization of German V-enclisis (counter to Romance V--prodisis), no doubt under disregard of the V--finiteness condition that holds in German. Notice that under this assumption a small Middle Field opens, which offers another argument against the claim of pure

2.56 Werner Abraham

verbal rightward governance in Cimbrian (but be aware also of the counter assumption manifested by Aux-Participle-0 in 1.4.1. ad (il)). Consider also my claims with respect to "historical change" in Section 1.3 above.

7·

lhe Cimbrian phenomena in the light of South German dialect characteristics -late contact transfer or no contact transfer at all?

The question arises as to how the Cimbrian phenomena observed so far are indicative of contact influence in view of the differences, and equivalences, between Cimbrian and Standard German and its Southern dialects. 7.1 Clitic doubling (CD) does not exist in Italian (as opposed to dialects of Italian and French- see (49)- and, more prominently and important, in the Friulian and Ladinian dialects of the immediate neighborhood of Cimbrian, where they do exist), neither does it in Cimbrian. However, this does not have to count heavily in favor of the assumption that this phenomenon is contact induced,9 given the fact that the same holds for West Flemish and the Dutch dialects (ct: Abraham & Bayer 1993 Introduction). 'Ihus, the conclusion is that it is an autonomous development. 10 Subject-CD: (55)

dar habant-za; de

Baibar; gaerbatet

d'Ekhare

da haben-sie die Weiher bearbeitet the .Acker 'there have-they the women worked the field~' (56)

Bir; morgan gea-bar;

au1

Saagra

[7 Commw1ities] Castagna (2003: 106f.)

[13 Commw1ities]

wir morgen gehen-wir zum StraBenfest 'we tomorrow go-we to the street party'

Objed-CD: (57)

Aso hats-mar; gapreart

gargenuin miar1 och

so hats-mir geschienen richtig mir 'this way has-it-me seemd alright me also' (58)

tse vorkxovo-se;

d'Ouksan/

[7 Communities]

auch [13 Communities]

zu verkaufen-sie die OdlSen? 'to sell--them the oxen'

9·

Observation provided by Gildo Bklese (p.c.).

10. An anonymous reviewer points out that ditic doubling in the South Bavarian and German enclave dialects looks different from those in the Dutch dialects. However, this is of no avail in the present context. It is quite clear and common knowledge that parallels between the enclave dialects and those of Dutch hold for subjects, not, however, for objects.

Spoken syntax in Cimbrian of the linguistic !.~lands in Northern Italy 257

(59)

Gestam hast-eni gikhoaft du in Giornal/ gestern hast--ihn/sie gekauft: du den Journal/die Zeitung? 'yesterday have--you--him/them bought you the paper'

[Lusema]

A pertinent question that arises from such data concerns the exact narrative or dia-logue function of sentential elements displaced from their base positions in the clause (such as subjects or objects in paraphrasal position in the middle of the sentence: see, e.g. Da haben sie, die Weiber, bearbeitet the Acker 'There have they, the women, worked on the fields') compared to those appearing in a hanging post--position (thema post-pendens, at the end of either a clause or constituent: Gestern hast sie gekctuft - du, die Zeitung? 'Yesterday have (you) heri bought - you, the papern. The illustrations on which we base our conclusions are not marked for special prosodic or accentual charc:tcteristics.

7.2

Dative-object differentiation (DOM)

Triggered by case syncretism, objects are often identified by prepositional marking in some languages of Romance and its dialectal vernaculars. This can also be found in South German dialects, most prominently in Alemannic and Southern varieties of Austrian-Bavarian (see Seiler 2003 or already Mayerthaler & Mayerthaler 1998, who highlight this phenomenon as a typical trait of Southern substandard German; see also Rowley 2003: 265). Clearly, DOMin Cimbrian may have developed under IR contact, but, given DOM in the dialects of Southern German, it may also be an autonomous development. We do not know anything about the age of its emergence. (60)

Du muasst des tJ/in deina Pmu vaschmibn. du musst das an/in deiner Frau verschreiben 'you must that at/in your wife prescribe'

(61)

Er giit dr Opfl a miar instead ofa diar. er gibt den Apfel an mir statt an dir 'he gives the apple at me instead of at you

(62)

The laicht-s enkh bol

en enkh ondera

der leiht -es euch wohl in euch anderen 'he loans-it you well in-you others'

[Bavarian] Seiler (2003: 15)

[Alemannic]

[Fersental] Rowley (2003: 265)

(63)

Kloob-ich Gi1:mtti h1it-ar-me ghet net ad 1:mder. Bidese et al. (2006: 86) glaub-ich Gianni hat-er-mir gegeben nicht in an andern 'believe- I G. has-he-me given not in another one'

(64)

Ghestar han-ich ghet an libar mne Pieren [Roana] gestern habe-ich gegeben ein Buch in dem Piero Bidese et al. (2006: 84) 'yesterday have-I given a book in the.DAT I~

2.58 Werner Abraham

Processing facilitation

8.

If my assumption is correct that in a variety of phenomena in Cimbrian as well as Austrian-· Bavarian dialects processing facilitation eases the parsing of oral encod·ing, then this would open the door f{)r a new interpretation of the data. See Abraham (2006a,b) as well as Abraham & Conradie (2001) for a fundamental discussion. All phenomena discussed in what follows are shared between Cimbrian and South German, Alemannic, and Bavarian-Austrian. Each of the characteristics observed and discussed below is not only a core property of Cimbrian, but it also shows that Cimbrian is fully alligned with South German. Moreover, these characteristics hold for oral codes only and, consequently, have autonomous status, which cannot plausibly be transferred from environing Romance dialects and the majority language, Italian. 11 8.1

Double perfect before the background of the South German preterite demise (Oberdeutscher Priiteritumsschwund, OPS)

OPS is firmly anchored in Cimbrian. Simple preterite forms are not observed in Cimbrian. (Dal Negro notices just two items in the conservative Wallis dialects of Agaro and Saley: chontun "konnten" (instead of hon chonna) and wan "waren" (instead of sen gsii); Dal Negro 2004: 242; see also Frei 1970: 315; cf. most recently Poletto 2009 for identical double forms in certain dialects of Northern Italy). (65)

I hon--s gahot tsechen id1 habe-es gehabt gesehen 'I have had seen'== "I had seen"

(66)

I pin gaben khemmen id1 bin gewesen gekommen 'I am been come'= "I had come"

8.2

[Fersental] Rowley (2003: 231)

Obligatory tun "do" as finite auxiliary in the declarative dause

There is a solid amount of genuine declarative clauses in Cimbrian with the light finite tun instead of the finite full predicate lexeme as is the case in Southern substandard German (Abral1am 2006a). This holds also for Cimbrian in both independent and dependent clauses as {67)-{69) below show. The result is a principled verbal bracket; the full verb (next to light tun) never fills finite position in TP or CP. This will leave the full verb--predicate in clause--final position for discourse·-·structur.U re-asons characteris-tic of Modern South German (the purpose is to fill the narrative foregrounding position

n. Any parallel claims oflinguistically genetic autonomy in the sense of HMer's (1956) are not immediately present here.

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 259

in its structural default site ("Aux-filter"; see Abraham & Conradie 2001; Abraham 2006a,b; Abraham & Bayer 1993 Introduction). Cimbrian shares this characteristic in full alignment with South German. (67)

Dos tua e

gern

song

[Palai] Rowley (2003: 201)

das tue ich gerne sagen 'this do I you willingly tell' (68)

S is a

Bail

as

bortr1

tucl

es ist eine Welle, dass ich warten tue 'it is a while that I wait do' (69)

Pariarn

tuat--s mer en Zboasesker

reparieren tut-s mir ein Zweiandsechziger 'repair does-it me a sixty-twoer'

8.3

Subjunction inflection

Subjunction inflection (subjunction suffigation mirroring the inflecting suffixes of the finite verb in clause-finalilate position) is a common feature among the AustrianBavarian dialects. Cimbrian shares this phenomenon undividedlyP (70)

Betzn.-so

sa du sechen tanast

[Fersental] Rowley (2003: 239)

wenn-du sie du sehen tiitest 'if--you them see would do' (71)

Dass-so olbe

lamentarn

muast!

dass-du allweil lamentieren must 'that -you always lament must'

8.4

Serialization of elements in the V-complex

The most remarkable characteristic is the serialization between the two most deeply embedded predicate components i.e. (2> )3 > 4 in (72)-(7 3) below which inverts the order of Standard Written German. See the glosses, which enable the reader to compare with Standard German as well as Dutch. See also (35)-(37) in Section 4.2 above. (72)

As

o hot gameicht lon

spiln

de kinder [Palai] Rowley (p.c. Leo Toller)

1 2 3 4 dass er hat gemacht lassen spielen the kids 'that he has made let play the kiM

12.

See for this topic also Hoekstra 1993 as well as Zwart 1993 under typological premisses.

260

Werner Abraham

(73)

Der hot bol

gameicht belln

helven

2 3 4 I der hat wohl gemocht wollen helfen 'this one has wanted will help'

The ordering 3 > 4 is a serialization unshared by South German dialects. Yet, little if anything can be concluded from this with respect to structural neighborhood to IR since lR has no such V-comple.xes in the first place (to all appearances due to the fact that modal verbs have far less usability in such complexes due to the fact that they lack the systematic change between epistemic and root infinitival forms). See Abraham 2006b for a discussion of this issue.

9·

Concluding generalizations

We have confronted the assumption that Cimbrian is the result of typological change under triggering conditions of contact with Italian (Friulian, Ladinian), much in line with prior claims in the literature to this effect. By contrast, we have assumed throughout autonomous linguistic change of an exclusively oral preserved dialect code (Abraham 2006a). For either position, there are pros and cons. In what follows I weigh the consequences of those pros and cons and weigh them out against one another. 9.1 'lhere are several infallible conclusions we can establish here with respect to the clausal structure of Cimbrian and it~ comparison with Modem Standard German and the South German dialects. For example, as long as the clitic form of pronouns attaches to the clausal subjunction as well as to finite V2 in the independent declarative- i.e. in WP - anything that ranges above XP-Vfin (true V2) amounts to a thema pendensi.e. to CP-dislocation. Not until Pro in WP is given up in favor ofV6n--proclisis does CP-expansion in Romance develop in the sense of Rizzi (1997). CP-expansion is unknown to German and its dialects (Grohmann 2003). Some of the V3-occurrences, such as (7 4), may have to be evaluated as clause left-dislocated apropos, i.e. as hanging themata/topics.B (74)

Adv »subject>> Vnnit

Rowley (2004: 529); see illustrations in Section 1.1.-1.5.),

9.2 Strict V2 in German need not fully cover V2 in Romance. Scandinavian, Yiddish, and Cimbrian also have V2 in the dependent clause. Yet, a small Middle Field has

13.

Note that V3 has already been documented for Old High German ('lbmaselli 1995).

Spoken syntax in Cimbrian of the linguistic !.~lands in Northern Italy 261

been maintained (cf Haider 2000, 2004; 1bmaselli 2004) just like in Cimbrian. Cf the discussion in 74 above. 9·3 One may assume (with Pintzuk's and Krach's Double Base Hypothesis) that areal linguistic contact of bilingual Cimbrian speakers triggered historical change with the result that original (German) Cimbrian OV gave way to (Romance) Cimbrian VO almost trivially. We have questioned this simple assumption. Whatever the answer, there would be sufficient interesting problems to be solved, among which: What structural components yield to contact induced digrc:llllmaticity early, and which do later? 14 Are universals of change involved, and if so, which (any of van Gelderen's (2004) economizing principles)? Or, alternatively, does clausal process facilitation under oral encoding play a determining role? To come closer to an answer consider the following cognitive economy principles: V-early, including V2/3, sinc.e V.-early warrants early identification of clausal functions as opposed to V-last, thus ascertaining relief of the working memory and that clause parsing is facilitated. This in turn prompts the question: What would 'linguistic change universals' be in the first place if not for something like economy principles? Taking this in account, consider the processing facilitation to be in line with van Gelderen's (2004) economy prindples (such as the reduction of phrases (XP) to head (X0 ) (ct: (75)), most prominently which can be summarized as "make a fortiori a simple word from a (complex) constituent" as in (76). (75)

XP> X 0

(76)

a. b.

[pp

[DP

wes.GEN [p wegen]] > 'whose ways' >

lsPECCP

weswegen] "because of this"

[Subj-DPi Schuhei dilnt\ [lauf-]] > [Subj-DPi [sc Schuhei di.\nnilauf-]]] shoes thin 0 r.subi··DP nm > shoes thine>f.shoes-run

The concept of Late Merge, i.e. to "insert as a functional morpheme a fortiori as late as possible into the proc.ess of syntactic derivation, also plays a role in these observed changes. In other words, insert as a ready--made lexeme in the appropriate syntactic (functional) slot without it further triggering any derivative processes". "Late Merge" follows from the principle of linguistic change in (75), but is more strongly motivated

14. Most remarkably, this question does not turn up in major recent work on linguistic change under contact (Heine & Kuteva 2005; Kuteva & Heine 2008). It may not be too presumptuous to assume that both the question and an appropriate answer and solution to the problem presuppose a sound and powerful syntactic mechanism of description and functional explanation (see, for example, such claims in Abraham 2005a).

262

Werner Abraham

theoretically in Minimalist syntax. Notice that "XP > .XO" is immediately plausible and more or less theory-independent 15 Changes under the economy trigger will often yield analytic transparency (separation of simple functional morphemes instead of a single more complex one) and, by consequence, ease clause processing most prominently under the restrictions and limitations that the working memory is subject to in oral online-processing. 16 These two criteria for change - structural simplification under the economy criterion and analyticity to satisfy categorial and functional transparency are manifested in the observable cyclic changes (as "Jespersen's Cycle" in Romance and Germanic). I take for granted that the parallel existence ofbilinguality in a society as Cimbrian triggers pressure towards analyticity; in any case, if the linguistic code is that of dialectal orality unsupported by a written norm.

9·4

What would be the consequences and implications of a shift from German V2 (S/0/Adv >> Vftn-postpro >> Adv/0/S- ~nftn as well as _Aux >>do » PP) toward Romance syntax (S >> clitic- V:in >> 0, where also _Aux >> PP >>DO)? C[ Bidese, J Poletto & Tomaselli (2002: 89) and others prior to the latter.

9.4.1 Dependent w-constructions (also relatives and dependent interrogatives) resist symmetrization between independent and dependent clausal structuring longer than other dependent clausal structures: WP--pronominal enclisis as well as V60 ··clitic are restricted to dependent word interrogatives and relative clause formation. In other words, the more conservative ("more German.") Lusema German has CP-structures (w(o)--enclisis in relative and interrogative clauses): (77)

lcoMP WO·· [wp ·-CL [ (IP ... ]]])

15. Analogy, as claimed by Fischer 2008 as a fundamental trigger for grammaticization, is no doubt strongly change-motivating and may also be added to such principles of economy underlying language change. Other mechanisms claimed as candidates for the universality of linguistic change such as "(inter)subjectification" still await sufficient empirical confirmation or may be an epiphenomenon before a deeper explanation (see Abraham 2005b, 2009). To the extent that metaphorical and metonymical processes are claimed as explanations for grammaticizing developments (as, iteratingly, forwarded by Heine 2003 and, more recently, Kuteva & Heine 2008), the absence of descriptive measures in terms of mereological feature hierarchies may be detrimental to intersubjective observability of similarity relations (see Abraham 2004 and, a long time prior to the latter,Jakobson 1972). For a list of powerful candidates explanative of morphologi.cal change, see Leiss 2008.

t6. For an inventory of element'> and mechanisms characterizing"oral encodinl:f as opposed to "written encoding'; typically motivating syntactic change, see Abraham 2006a.

Spoken syntax in Cimbrian of the linguistic islands in Northern Italy 263

In contrast, the dialect of Roana has only a TP-structure just like Romance (only V-enclisis (i.e. no Comp-enclisis in WP) despite of Camp-occurrence: (78)

[cOMPas

b

NP

h V-CL] ... ]])

9.4.2 DP-inversion (S/0/Adv- V6,. >>Adv/0/S-V) is lost prior to subjectclitic inversion, i.e. ~fin--ciitic >> --clitic-- ~'in~-. What is the driving force behind this? Behaghel's principle of constituent heaviness is hardly to be made responsible since it has no explanative quality.

9-4-3 What exactly are the deeper links and differences between S/0/.4.d1'-Vfin -cliticAdv/0/S and S--clitic-- Vfin--O? What what kinds of structural reduction exhausts the counter positions? At this point, I leave these questions for future research. 9·5

Contact-induced grammaticalization: The research status quo

9·5·1 The present claim The traditional account of the VO-OV mix in Cimbrian is that all serializations different from Modem German are the result of contact with IR varieties (among others strongly advocated by Krach 1989 ('Double Base Hypothesis') and by Rowley 2003: 251, 289, (291). My counter claim to Cimbrian change due to L-contact has been strong: no change has been observed beyond what has proven to be possible in Germanic without contact (as, for example, in Scandinavian Germanic). More generally, change (between typologically different languages such as SVO(rw V2)-IR and SVOV(V2 and Vlast)-Cimbrian \\
Central claims ofchange under contact: Copying offoreign structures under sociolinguistic pressure?

Frequential copying of foreign structures: Does Cimbrian copy SVO of the majority languages as a consequence of bilingual use of Cimbrian and administratively dominant Italian? Among the properties that may be copied, Johanson 2008 empha-sizes frequency of use, observing how in contact situations speakers may start using

264 Werner Abraham

a certain structure of their native variety more frequently until they end up matching the frequency of use of a corresponding structure in the other variety ("Frequential copying"). Frequency of use is central to most contributions to contact-induced change, e.g. as a means by which pragmatic markedness is removed from a construction (cf. Heine 2008). Pietsch 2008, for example, describes how a linguistic structure, which is weakly represented in terms of frequency can be reinforced in a situation of contact with another linguistic variety in which a corresponding structure is more strongly entrenched. Sociolinbrui!>1ic conditions: Another point pervasively made in the literature is the importance of the sociolinguistic conditions in which contact occurs, as an important variable that one should ideally be able to control in order to ensure mutual comparability between different contact scenarios. Noonan 2008, for example, is concerned precisely with the sociolinguistic factor and its influence in determining the outcome of contact. He presents the case of a family oflanguages spoken in Nepal, the Tamangic group, whose members have been in contact with a number of non-affiliated languages under different sociolinguistic circumstances, and he focuses on the different outcomes that such diverse circumstances have produced among different members of the group. Autonomous variation in contact situations: How accessible different grammatical domains are to the influence of a majority contact language from a purely grammatical perspective is demonstmted in Comrie 2008. 'Ihis paper is concerned with the borrowability of a core grammatical domain, i.e. inflectional morphology (IM). Comrie presents the view that IM is one of the least borrowable features of a language. He argues that within IM itself there exists a borrowability hierarchy, with agglutinative morphology (AM) being more borrowable than fusional morphology (FM). What is even more important is Heine's (2008) claim, based on the analysis of a good number of typologically diverse language contact situations, that when a language changes its word order preferences in a contact situation, it never actually substitutes a new and hitherto unknown word order for the old one. In these situations, the lan-guage undergoing word order change merely replaces the previous ordering by an already possible, alternative word order, which was previously reserved for pragmatically marked expressions. 9·5·3 Cimbrian as arz autonomous grammatical calque of!R? The main insight gained from independent discussions - such as those in the previous section is that when a language changes its word order preferences in a contact situ-ation, it never actually substitutes a new and hitherto unknown word order for the old one. Rather, it merely replaces it by an already possible, alternative word order, which was previously reserved for pragmatically marked expressions (Heine 2008). 1his fully

Spoken syntax in Clmbrian of the lingulstlc island~ in Northern Italy 265

endorses our findings concerning Cimbrian changes, which turned out to have been possible in other Germanic languages in any of their historical stages. No output under sociolinguistic pressure has been recognized. Frequent copying of foreign structures has not been detected to have played any role in the diachrony of Cimbrian through the centuries (Bidese 2008b). The changes that took place in Cimbrian Germanic exploit the options compatible with V2 and Vlast (for the full range and implication along descriptive and explanatory layers other than the V-module see Haider 2009).17 Recall that IR definitively does not license from V2 and Vlast. What Cimbrian calqued from IR are options open within Germanic in the first place. "Calque: on this view, needs reinterpretation: IR may have triggered developments in Cimbrian on I R-features that have interfaced. with Germanic in the first place. There is one issue that has not been traceable in the literature; namely, an equiva-lent of our claim with respect to parsing in an oral-only linguistic society. 'Io the extent that indications to this particular status are missing altogether may lead one hypothesize that this assumption is self-evident and is commonly shared.

10.

Summary

'Ihe previous discussion concerns itself with the sources and triggers of change and the question of why they happen in the first place. If the tasks of historical linguistics can be described as, i. describing the changes as parameter setting, ii. explaining the morpho-syntactic changes, and iii. reconstructing on the basis of syntactic evidence (see Crisma & Longobardi 2009: 7). Here we have found answers to each of these questions in ways, it appears, that reach beyond those offered by others. Thus, the reason why OV-related structures may appear in Scandinavian dependent clauses and in other Germanic languages remains unclear under the account involving remnant movement (openly admitted in Hr6arsd6ttir's 2009 on phenomena in Swedish), while the explanation in the present discussion also answers (iii) next to (ii): parsing facilitation under extra strain on the working memory in oral online speech. In other words, we reconstruct a trigger of change that must have been active tar down in the history of Cimbrian as a release

17. Quite obviously, claims such as" [... ] the replica (changing minority) language may acquire a category for which previously there v.ras no or no appropriate equivalent:' (Matthews & Yip 2009: 374; secondary quote from Heine & Kuteva 2005: 124) do not invalidate our position of an autonomous language change ln Clmbrian.

266 Werner Abraham

that has nothing to do with the exposition of the dialect to IR once the enclaves of Cimbrian had become so small that dependency on the emironing majority languages had become inevitable. At the same time, we respect the inertia factor in Cimbrian that strikes our eyes so compellingly on each of the linbruistic details that Cimbrian shares with all dialects of South German, Austria, and Swit7.erland. The only difference with the status of Cimbrian is that all of the mainland dialects are in continued direct contact with, and under the strong normative supervision by, the standard written language and its media and school commitments. Cimbrian, which has been without such standard support and normative coercion over centuries of complete linguistic isolation, has always been open for processing facilitation as a natural limitation levied on online speech. We have discussed, then, a situation which is absolutely singular in the sense that Winteler (1.876) had already singled out as change occurring in speech communities under conditions of near--absolute isolation and internal homogeneity (i.e. independent of the influence of another linguistic variant with a heterogeneous corpus). We thus have a remarkable situation of change where corpus heterogeneity is not the source of change and where change and inertia hold a telling relative propor-tion enables us to draw safe conclusions concerning the typological characteristic of Cimbrian, the language under discussion. Is Cimbrian both VO and OV- thus, is it a typological mix? If so, (see Bidese 2008a,b for such a claim) it must have started dropping its original German OVtypology on the right periphery much less so, on the other hand, on the left periphery of the clause, where Cimbrian is more or less_ V/Aux(_ V) like Standard German and its dialects. Is there a rationale for such a fundamental right vs. left periphery asymmetry with respect to the calquing accessibility toward foreign structures? However, given that older stages of German exhibited an VO ordering just as much as OV this appears to boil down to the following crucial question: Is M6cheno syntax the result of rules fi·om one single grammar or of two grammars that are in competiton with one another (Double Base Hypothesis)? Cognola (2009: 12) claims that the syntactic rules ofM6cheno (the dialect of the village Palu in the FersentaVFersina Valley) stem from one single gr.unmar and that this is for the following reasons (79)---(81) are Cognola's (2009) illustrations (see also Poletto & Tomaselli 2000). 10.1

iv. Verbal arguments leave VP moving beyond; i.e. higher than, manner adverbs as in Modern German- see (80) (Cognola's (14)). (79)

Ber

hot sc1lia galezer1 s:puach?

'Who has well

v.

read

the book?'

1he order of PPs or manner adverbs (cf. schia "well" in (iv) above) shows that the past participle has not moved; this leads to the conclusion that OV reflects

Spoken syntax in Cimbrian of the linguistic !.~lands in Northern Italy 267

the order of merge previous to movement of V: manner adv - PP - (DO) -past

participle - (DO) = OV. vi. Arguments ofV move outofVP to higher manner adverbs as in Modern German. vii. All V-arguments remain above manner adverbs, but below sentential adverbs; this is different from Modern German: rnatmer adv past participle PP/DO = VO. viii. For sentences beginning with a focused XP (such as a DO or PP), the only word order admitted is VO. See (81) (Cognola's (28)). (80)

A PUACH hot-er schua kaft in de bodeig 'a book has--he already bought in the shop'

ix. Counter to the evidence in (v), a focused subject or a focused comitative PP allow fi.)r either VO or OV. Cf. (82) (Cognola's (31.)). (81)

Df:R MARIO hot schua 'the

M.ario

s'puach

kaft

ont rret der Nar1e

has already the book bought and not the John'

x.

Both OV and VO orders are possible in sentences without extraction from the phase below CP. In other words, the CP in M6cheno remains unaffected by extra(..tion. This holds for independent dedaratives and sentences with a focused subject or a comitative PP, which are moved to CP from the higher phase, i.e. from outside of CP. xi. The rules confirming VO synt.ax are different from VO in otherwise also OV-Old High German (Hinterh0lzl2009, 2010) and OV--Old English (1aylor & Pirrtzuck 2008, 2009) emerging from the syntax-infi.)rmation structure interface encoding old information after the past participle, which is in line with Romance and Latin IS (all orders divergent from OV yielded marked IS-constructions in Latin). The left periphery of the clause encapsulates discourse identifiers like thema and rhema - these distinctions are preserved (or show to be even extended through Cimbrian V3 - i.e. through several thematic/topic positions before the first verbal bracket, Comp ). By contrast, however, the clausal right periphery is dissolving in that clause processing is facilitated, thereby yielding early finite V (V2/3/4 before verbal complements). This makes subject-object-adverbial facilitating early on considerably distinguishable to online demands under oral encoding as opposed to any written code, which allows ready and constant verification by means of Backtracking and Forward looking (see Abraham 2006a,b; Abraham & Conradie 2001). Typological mixes such as those claimed above for Cimbrian are frequent in Germanic and even in French (though not evidenced in the history of Italian). See Hr6arsd6ttir's (2009) discussion of Old Icelandic as well as the history of French. Earlier French had overt restructuring effects with respect to long-DP movement

10.2

268

Werner Abraham triggered by meclio-passive se. See (83a,b) from Hr6arsd6ttir (2009: 71; illustrations due to Roberts 1997: 445). (82)

a.

Nuls om

mortals no.I

pod

penser

no man mortal not-it could think 'No mortal man could think

u:

b.

Elle le

commenqa

11

desirer

she him began to desire 'She began to desire him:

However, this word order option was lost in French along ~ith the loss of coherent infinitives (Roberts 1997), a structure characteristic that Italian never lost. As a result of maintaining this distinction, Italian retained the possibility oflong clitic movement. Another OV-trait is reported for quantifiers in both languages; i.e. (older) French (Kayne 1994; Roberts 1997: 441) and (Old) Icelandic (Hr6arsd6ttir 2009: 72). (83)

a.

Jean a

tout

lu.

Modern French

John has everyting read 'John has read everything: b.

Il

na

rien

voulu

que je dise.

he not has nothing wanted that I say 'He did not want me to say anything:

Marie a

tottt voulu

faire.

Mary has all wanted to do 'Mary has wanted to do evel)1hing: (84)

a.

J6n

hefur ekkert

lesilJ.

John has nothing read 'John has not read anything: b.

1-Iann hefur ekkert

viljalJ

aC1

eg seglJi.

he has nothing wanted that I 'He didn't want me to say anything:

say

Negation and quantifier movment may even permit extraction out of tensed subjunctive clauses as demonstrated by (84b) and (8Sb) in either language. Hence, such con-structions proved to be unaffected when OV word order construals disappeared with ordinary common appellative now1s, non-quantified, non-negated object or adverbial material. In fact, negation seems to act as sort of a minimality blocker for movement of the finite verb across it. (85)

a.

Marfa hefur ekkert Mary

has

sagt Joni

nothing said to John

b. Mar{a (ekkert) saglJi (*ekkert) J6ni Mary

said

nothing

to John

Spoken syntax in Clmbrian of the lingulstlc island~ in Northern Italy 269

(86b) mirrors the development in French and Italian. It leaves Cimbrian however completely untouched: Negation never obtains a position before the finite verb -although the weak elitic negation did in Middle High German. All of this adds an interesting perspective on the comparative syntax of not only the two languages. Data from French, (Old) Icelandic, and Cimbrian demonstrate that the shift from an underlying OV to VO-o:rdering can be considered a step in the development of (historical) Germanic languages. See, once again, (87), the former (.5) above. (86)

a.

I

hon

niernad

Aux--do--Participle

gesek

ich habe niemand gesehen 'I have no one seen' b.

han-ic/1

khoorne gaseecht

Aux-neg-Participle

gesehen habe ich kaum 'have --1 hardly seen'

This Cimbrian evidence is even more eye-catching since Aux-Participle separation as in French, Icelandic, and Cimbrian is ungrammatical in modern Italian (Alessandra Tomaselli, p.c.). See (88a-c). Interference, or grammatical calque, is not an option to explain the phenomenon in Cimbrian. (87)

a. b.

Ganni ha letto tutto Ganni ha letto niente Maria ha volute fare tutto

=I= =I= =I=

*Gianni ha tutto letto *Gianni non ha niente letto *Maria ha volute tutto fare

Thus, since interference under contact with the majority language may safely be excluded, the typological change ot~ or the split between, OV-to/and--\'0 (VO-to--OV, obviously, does not exist), may be considered as a development driven by a mechanism of facilitation of clause processing under the very specific oral-only code conditions (Abraham & Conradie 2001; Abraham 2006a,b; in a fundamentally ditferent explanatory vein (assuming rudimentary VO for all languages, much in the light of Kayne 1994) Hinterh6lzl2006; Hr6arsd6ttir 2009). Parsing VO is easier than that of OV since the distinction between SU and OBJ is recognized earlier. This will ease the task of the working memory and make room for other tasks left from Vlast, the only category eventually determining the valence of the clausal arguments (Abraham 200 l, 2006a). In this sense, then, the VO-OV split and linguistic change in Cimbrian can be understood as the result of clause processing facilitation under simultaneous optionality of several variants (options ranging from more conservative to more modern, process facilitating varieties). Under these desiderata, linguistic change of a single construction can be obsevered in a situation where two equally valid and simultaneous options have existed for at least two generations of older and younger speakers,

10.3

270

Werner Abraham

such as for Cimbrian strictV2 and V3/4 and serialization variants in verb complexes). Notice that, under such a simultaneous co-existence of two competing stylistic, sociolectal variants, the phenomenon of sudden change receives a plausible explanation and acceptability (cf. Lightfoot 1979; Kroch 1989; Hinterhillzl2004). More specitlcally, the change of OV >> VO (as for Old Engiish >>Me./ ModSte.) represents online process facilitation. From this we conclude that for purely oral language use, uninfluenced, furthermore, by standard speaking media and under the complete absence of school education of Standard German, the change to VO is expected (counter to Hinterh6lzl2004, who assumes light verb-sentence-over-0 to be the result of obligatory VP-intraposition as the fw1damentallinguistic change; consider Abraham 2009). 18

10.4

More specifically, Romance- Italian VO does not necessarily have to be assumed as a grammatical superstratum. Cimbrian is not simply VO, i.e. is not Romanicized German, given, most prominently, the OV-corollary of pronoun enclisis as well as subject-thema-inversion (clausal postposing of the subject). 19

10.5

Again and beyond all doubt, we conclude that linguistic change under IR contact is possible only under the assumption that autonomous regularities open the door for such a change in the first place. Recall that Cimbrian pronominal enclisis to infinitival forms is possible when the speaker of Cimbrian overgeneralizes enclisis of finiteness onto non-finiteness. Under such a measure, a small Middle Field is opened. In other words, these phenomena appear to force the conclusion that changes under contact are never arbitrary. Rather, it appears that change under contact "force doors already half-open.".

10.6

Assume someone speculating that the diachrony of Cimbrian is due to contactinduced change as a change of micro-varieties because of the similarity of change from Proto-Germanic to Scandina\ian as macro-varietal change.

10.7

Then, Fersental Cimbrian has Comp-CL and, accordingly, the long clausal structure being on par with Standard German. Comp-CL-languages have the clausal bracket with V2 as well as Vlast, thus presenting asymmetry between dependent and 10.8

18.

C£ also Bauer 1995 in this context.

19. While Grewendorf & Poletto (2009) seem to have a conclusive account for the Cimbrian mix ofVO and OV, the phenomenon of sub.iect-verb inversion remains unaccounted for.

Spoken syntax in Clmbrian of the lingulstlc island~ in Northern Italy

independent clause structure ( Cognola 2009). Cimbrian outside ofFersental, however, with Comp + Yfin-CL, shares with IR the short clausal structure. This, however, is by itself not an unambiguous clue that leads to the assumption that Cimbrian is on a diachronic path toward Romance clausal structure (with the highly conservative exception of the Fersental). Cf. Padovan 2009 as well as Bidese & Padovan 201.0, who arrive at an identical conclusion. Consider that Standard German has TP-clauses as well: the topic-empty thetic/presentative clauses. The short structure, thus, is not foreign to the computational system of German in the first place. 10.9 Cimbrian has retained sufficient solid properties of German SVOV to warrant the conclusion that it is not structure-neutral change under contact, but sentence processing measures that guide this change quite independently of this contact. Recall the German clausal structure traits with exceptionless pronominal enclisis and the small verbal bracket (with "light" element~). We might say that these characteristics as well asAux-PP +objects are not due to Romance contact, but to SVO with process facilitating the S-O-division (as in Afrikaans; see Abraham & Conradie 2001).

Romance, as opposed to German and all its dialects, shows sentence embed-ding symmetry between dependent and independent clause structure just as English does. May we assume that change in Cimbrian- as a type of micro-variation- mirrors the transfer of OV-VO from Proto-Germanic to Scandinavian, which shows (almost exclusively) SVO (recall Figure l above) i.e. the Greenbergian VO-type, but otherwise concurs with German thereby projecting a macro-variant of Germanic as an entirety? This would seem to act as a correlation from which we might gain a large and quite important linguistic diagnostic field of investigation which, given the lack of linguistic fossils from the early periods of Germanic, we have dearly missed to date.

10.10

Language change in the sense of the aforegoing is seen as performance-based 10.11 variation feeding grammar variants one of which eventually becomes the new norm selected by, since alignable with, the axiomatics of an overall grammatical design. 1here must not be exceptions in such an axiomatically computed grammar design. Performance-based variants ideally emerge from spoken, dialogetic speech- thus, discourse on an oral-only basis forms the major data inventory on which to base one's analytic conclusions. Parsing restrictions play a crucial role next to other discourse properties and constraints reflecting limitations on the working memory. Recent work on older language stages in the Germania confirm the validity of this methodological course (cf. the 'Introduction' in Hinterh6lzl & Petrova 2009; likewise, Hinterhc.>lzl 2009). Again, this provides a clear counter position to the assumption that results from

271

272

Werner Abraham

variation in word order (SOV ne.xt to SVO) by contact of an original SVOV language with pure Romance variants exhibit an SVO order. This is a claim that is plausible at first sight to pursue if it comes to describe Cimbrian (Rowley 1986, 2003, 2004, 2006 as well as others in this limited field of work on linhruistic enclaves).

Cited references Abraham, Werner. 1965. Die Mundart von Tschagguns im Montafon. Ph.D. dissertation. University of Vienna. Abraham, Werner. 1997. The base structure of the German clause under discourse functional weight: contentful functional categories vs. derivative functional categories. In: W. Abraham & E. v. Gelderen (hg.) German: syntactic problems problematic syntax. 11-42. [Linguistische Arbeiten 374]. Thbingen: M. Niemeye~: Abraham, Werner. 2001. SVOV in German and Dutch. Interface between discourse prominence and subject identification In Refkctions on Language and Language Learning. In Honour of Arthur van Essen, Marcel Bax & Jan-Wouter Zwart (eds.)155-170. Amsterdam: John Benjamins. Abraham, Werner. 2004. The grammaticallzation of the infinitival preposition toward a theory of'grammaticalizing reanalysis: Journal ofComparative Germanic Linguistics 7(2): 111-170. Abraham, Werne1: 2005a Deutsdre Syntax im Sprachenvergleich. Grundlegungeiner typologischetl Syntax des Deutschen [Studien ztu· deutschen Grammatik 41]. Ttibingen: Stautfenbw·g. Abraham. Werner. 2005b. An intersubjective note on the notion of'subjectificatio.ri In Orga-nizing Grammar: Linguistic Studies in Honor of Henk van Riemsdijk, Hans Broekhuis. Norbert Corver, Riny Huybregts, Ursula Kleinhenz & Jan Koster (eds). 1--16. Berlin: Mouton de Gruyter. Abraham, Werner. 2006a. Dialect and typology: Where they meet and where they don't. In Types of Variation: Diachronic, Dialectal, and Typological Interfaces [Studies in Language Companion Series 76], Terttu Nevalainen, Juhani Klemola & Mikko Laitinen (eds), 243-268. Amsterdam: John Benjamins. Abraham, Werner. 2006b. Re7
Spoken syntax in Clmbrian of the linguistic island~ in Northern Italy 273 Abraham, Werner & Conradie, C. Jac. 2001. Priiteritumschwrmd rmd Diskursgmmmatik in

gesamteuropiiischen BezUger1: i\reale Ausbreiturrg, heterogene Bntstehung, Parsing .~OY.'ie diskursgrammatische Grandlagen and Zusammer1hiinge. Amsterdam: John Benjamins. Bayer, Josef 1993. zum in Bavarian and scrambllng. Linguistische Berichte Sonderheft 5: 50-70, Dialektsyntax, Werner Abraham & Josef Bayer (eds). Bhatt, Chri.sta & Schmidt, Claudia Maria. 1993. Die am infinitive-Konstruktion als Aspektphrase im Kolni.schen und im umgangssprachlichen Deutsch. Linguistische Berichte Sonderheft 5:71--98, Dialektsyntax, W Abraham & J. Bayer (eds). Bidese, Ermenegildo. 2008a. Die \Vackernagelposition als CLP. Ein neuer Interpretationsvorschlag zur Position morphophonologisch reduzierter Pronominalelemente in den ger-manischen Sprachen unter besonderer Berticksichtigung des Zimbrischen. Romtagung der Germanisten/DSWI 1-1. Feb. 2008. Handout. Bidese, Ermenegildo. 2008b. Diachrone Syntax des Zimbrischen. Ti.i.bingen: Narr. Bidese, Ermenegildo, Dow, James R. & StoLz. Thomas (eds). 2005. Das Zimbrische zwischen Germanisch and Romanisd1 [Diversitas Linguamm 9]. Bochum: Brockmeyer. Bidese, Ermenegildo, Poletto, Cecilia & Tomaselli, Alessandra. 2006. The relevance of lesserused languages for theoretical linguistics: The case of Cimbrian and the support of the TITUS corpus. In LULCL -- Lesser Used Languages. Computer Linguistics. Proceedings ofthe Lesser Used Languages and Computer Linguistics Conference Balzano, 27--28 October 2005, Isabella Ties (ed),77--96. EURAC research. Bidese, Ermenegildo & Padovan, Andrea. 2010. Erodierte Sprachstrukturen und Grammatik-theorie. Zur Morphosyntax der Halbsprecher in der zimbrischen Enklave Lusern. Talk and handout at the conference Deutsche Sprachwissenschaft in Italien, Rome, Februa1y 4--6, 2010. Bidese, Ermenegildo & Tomaselli, Alessandra. 2005. Formen der 'Herausstellung' and Verlust der V2-- Restriktion in der Geschichte der zimbrischen Sprache. In Das Zimbrische zwischen Germanisch tmd RomanL~ch [Diversitas Linguarum 9], Ermenegildo Bidese, James R. Dow & 'Ihomas Stolz (eds), 71-92. Bochum.: Brockmeyer. Brugmann, KarL I 917. Der Ursprung des Schelnsubjekts es in den germanlschen and romanischen Sprachen. Berichte iiber die Ferhandiungen der koniglichen siichsischen Gesellschaft der Wissenschaften zu Leipzig, Philosophisch-historische Klasse 69(5). Leipzl.g: Teubner. Castagna, Adriana. 2005. Personalpronomlna und Klitika im Zlmbrischen. In Das Zimbrische zwischen Germanisch und Romanisch [Dlversitas Linguarum 9], Ermenegildo Bldese, James R. Dow & 'Ihomas Stolz (eds.) 93-113. Bochum: Brockmeyer. Chocano, Gema. 2007. Narrow Syntax and Phonological Form. Scrambling in the Germanic Lang-uages [Linguistik Aktuell/Linguistlcs Today 109]. Amsterdam: John Benjamins. Cinque, Gugli.elmo. 1999. A.dverbs and Functional He,ads: i\ C'ros.~-linguistic Perspectit'e. Oxford: OUP. Cinque, Guglielmo. 2007. The fundamental left-right asymmetry of natural languages. lVIs, Universita Ca' Foscari di. Venezia. Cognola, Federica. 2009. On the mixed OV/VO syntax of Mocheno. Talk and handout at the 37th ()sterreichische Linguistentagung, Syntax Workshop, Salzburg Dec. 7, 2009. Comrie, Bernard 2008. Inflectional morphology and language contact, with special reference to mixed languages. In Siemund & Kintana (eds), 15-32. Crisma, Paola & Longobardi, Giuseppe. (eds). 2009. Introduction. In Historical Syntax and Linguistic :lhe.ory, 1-16. Oxford: OUP. Dal Negro, SiMa. 2004. 7he Decay of a Language. 7he Case of a German Dialect in the Italian Alps. Bern: Peter Lang.

274 Werner Abraham Fischer, Olga. 2008. On analogy as the motivation for grammattcallzatton. Studies in Language 32(2): 336-382.

Fret, Gertrud. 1970. Walserdeutsch in Saley. Wortin1:altliche Untersuchung zu Dialekt und Weltsicht der altertumlichen Siedlung Salecchio!Saley (Antigoriotal). Bern: Haupt van Gelderen, Elly. 2004. Grammaticalization as Bccmomy [Lingulstik Aktuell/Ungulstl.cs Today 71]. Amsterdam: John Benjamins. Good, Jeff (ed.). 2008. l.Amguage Change and IJniverStlls. Oxford: OUP. Grewendorf, Gi.inther & Poletto, Cecilia. 2005. Von OV zu VO: Ein Vergleich zwischen Zimbrisch und Plodarisch. In Das Zimbrische zwischen Germaniscil und Romaniscil [Diversitas Linguarum 9], Ermeneglldo Bidese, James R. Dow & Thomas Stolz (eds). Bochum: Brockmeyer. Grewendorf, Gi.inther & Poletto, Cecilia. 2009. Was ist eine Verbzweitsprache? Talk and handout at the Tagung Deutsche Sprachwissenschaft in Italien, Rome, February ·1···6, 2010. Grohmann, Kleanthes K. 2003. Prolific Domains. On Anti-locality of Movement Dependencies [LinguistikAktuell/Linguistics Today 66]. Amsterdam: John Benjamins. Haegeman, Liliane. 1993. Some speculations on argument shift, clitics, and crossing in West-Flemish. Linguistische Berichte Sonderheft 5: 131-160, Dialektsyntax, W. Abraham & J. Bayer (eds). Haider, Hubert. 2000. OV is more basic than VO. In The Derivation of VO and OV [Linguistik Aktuell/Linguistics Today 31], Peter Svenonius (eel.), 45-68. Amsterdam: John Benjamins. Haider, Hubert 2004. How to turn Germanic into Icelandic and derive the OV-VO contrasts.

Journal of Comparative Germatzic Linguistics 8: 1··· 53.

:t Kiss & Henk van Riemsdijk (eds), 200·1, Verb Clusters. A Study of Hungariatz, German, and Dutd1 [Linguistik Aktuell/Linguistics 68]. Amsterdam: John Benjamins. Latzguage 38(2): 647--651. Haider, Hubert. 2009. German. What it Means to be OV. Cambridge: CUP. Heine, Bernd. 2003. Grammatica:tization. In The Handbook of Historical Linguistics, Brian D. Joseph & Richard D. Janda (eds), 575-601. Oxford: BlackwelL Haider, Hubert. 2007. Review of Katalin

Heine, Bernd. 2008. Contact-induced word order change without word order change. In Siemund & Kintana (eds), 33-60. Heine, Bernd & Kuteva. Tania 2005. Language Contact and Grammatical Change. Oxford: 0 UP. Hinterhtil:ll, Roland. 2004. Language change versus grammar change. What diachronicic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic C.rammar [Linguistik Aktuell/Linguistics Today 72], Eric PuB & Carola Trips (eds ), 131-160. Amsterdam: John Benjamins. Hinterholzl, Roland. 2006. Scrambling, Remnant Movement, and Restructuring in West Germanic. Oxford: OUP. HinterhOlzl, Roland. 2009. The role of information structure in word order variation and word order change. In New Approaches to Word Order Variation in Germanic, Roland Hinterholzl & Svetlana Petrova (eds) 45-66. Berlin: Mouton de Gruyter. HinterhOlzl, Roland. 2010. Information structure and unmarked word order in (older) C:.ermanic. In Information structure. Theoretical, typological, and experimental perspectil'e.-;, Malte Zimmermann & Caroline Jlery (eds), 282-306. Oxford: OUP. Hinterh0l71, Roland & Petrova, Svetlana (eds). 2009. New Approaches to Word Order Variation in Germanic. Berlin: Mouton de Gruyter. Hoekstra, Eric. 1993. Dl.alectal variation inside CP as parametric variation. Linguistische llerirhte Sondt'T'heft 5: 161···1i9, Dialek~yntax, W. Abraham & J. Bayer (eds).

Spoken syntax in Cimbrian of the linguistic island~ in Northern Italy 275 Otto. 1956. Stammbaumtheorie, Wellentheorie, Entfaltungstheorie. PBB 77: 1-37; 424-476. PBB 78 (1956): 93-136. Hr6arsd6ttir, 'Ihorbjorg. 2009. Restructuring and OV order. Working Papers in Scandinavian Syntax 83: 37-82. Jakobson, Roman. 1972. Two aspects of language and two types of aphasic disturbances. In Selected Writings, Vol. 2: 239-259. Johanson, Lars 2008. Remodeling grammar: Copying, conventionali7.ati.on, and grammaticallzation. In Siemund & Kintana (eds), 61-·80. Kayne, Richard S. 1994. The Antisymmet1:v of Syntax. Cambridge MA: The MIT Press. Kranzmayer, Eberhard. 1981. & 1985. Laut-- and Flexionslehre der deutsch-zimbrischen Dialekte [Beitrage zw· Sprachinselforschung Band 1, Teil1 & 2]. Wien: Verband der wissenschaftlichen Gesellschaften (lsterreichs. Kroch, Andy. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1: 199---244. Ktihebacher, Egon 1972. Die geschichtliche Entwicklung der deutsch- italienischen Sprachgrenze. Der Schlem 46: 33-- 52. Kuteva, Tania & Heine, Bernd 2008. On explanatory value of grammaticalization. In Good (ed.), 215--230. Leiss, Elisabeth. 2008. Grammatical complexity and affix ordering. A typological and diachronic approach to the theory of word structllre. Paper at the Morphology Conference, Workshop on Affix Ordering, Vienna Jan. 30 Feb. -1, 2008. Lightfoot, David. 1979. Principles ofDiachronic Syntax. Cambridge: CUP. Lotscher, Andreas. 1993. Zur Genese der Verbdopplung bei gaa, iaa, aafaa ("gehen", "lassen", "anfangen") im Schweizerdeutschen Linguistisc.he Berichte Sonderheft 5: 180---200, Dialekt-syntax, VV. Abraham & J. Bayer (eds). t...1atthews, Stephen & Yip,Virginia. 2009. Contact--induced grammaticalization: Evidence from bilingual acquisition. Studies in Language 33(2): 366-395. (Special issue:). Clancy Clements & Shelom.e Gooden (eds.) Language Change In Contact Languages: Grammatical Hofler~

and Prosodic Considerations]. Mayerthaler, Willi & J:vlayerthaler, Eva. 1998. Aspects of Bavarian syntax or 'Every language has at least two parents: In Development and Diversity. Lang-uage Variation across Jime and Space. A Festschrift for Charles-James N. Bailey [The Summer Institute of Llnguistl.cs and 'Ihe University of 'Iexas at Arlington. Publication 93], Jerrold A. Edmondson, Crawford Feagin & Peter Miihlhausler (ed~.) 371-429. Dallas TX: SlL. Meid, Wolfgang. 1985. Der erste zimbrische Katechismus. Dar kl6ane catechismo vor dez beloseland.

Die zimbrische Version aus dem Jahre 1813 a11d 1842 des picc.olo c.atechismo ad uso del regno d'Italia of .1807 in kritischer Ausgabe [Innsbrucker Beitrage 1ur Sprachwissenschaft: 48]. Innsbruck: Institut fur Sprachwissenschaft der Universitat Innsbruck. Noonan, Michael. 2008. Contact- induced change: The case of the Tamangic languages. In Siemund & Kintana (eds), 81-106. Padovan, Andrea. 2009. Diachronic clues to grammaticali7.ation. Phenomena in the Cimhrian CP. 'Ihlk and handout at the Workshop Dialektologie als Briicke zwischen synchronischer und diacltroner Linguistik at the Universlty of Verona, December 17-18, 2009. Panieri, Luca, Pedrazza, Monlka, Nicolussi Bai1, Adelia, Hipp, Sabine & Pruner, Cristina. 2006.

Bar lirnen z' sdlraiba un zo re.da az be biar. Grammatica del cimbro di J.,userna. Grammatik der zimbrischen Sprache von Lusem. Regione Autonoma Trentino-Alto Adi.ge/Autonome Region Trentino/Stidtirol: Istituto ClmbroiKulturinstitut Lusern.

276

Werner Abraham Pietsch, Lukas. 2008. Prepositional aspect constructions in Hiberno-English In Siemund & Kintana (eds), 213-238. Poletto, Cecilia 2009. Double auxiliaries, anteriority, and terminativity..fournal of Comparative Germanic Linguistics 12: 31-48. Poletto, Cecllia & 'lbma~elli, Alessandra. 2000. I:interazione tra germanico e romanzo in due 'isole llnguistl.che: Cinbro e ladino a confronto. In lsole linguisNche? Per un' a,1alisi del sistemi in contatto, G. Marcato (ed.) 163-176. Padova: Unipress. Poletto, Cecilia & Tomaselli, Alessandra. 2002. La sintassi del soggetto nullo nelle is ole tedesco-fone del Veneto. InLa dialettologia oltre 112001, G. Marcato (ed.), 237---252. Padova: Unipress. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 231---337. Dordrecht: Kluwer. Roberts, Ian. 1997. Directionality and word order change in the history of English. In Pare-meters of morphosyntactic change, Ans van Kemenade & Nigel Vincent (eds.), 396--·126. Cambridge: CUP. Rogger, Igino. 1979. Dati storici sui mocheni e I loro stanziamenti. In La valle del Fersina e ie isole lingulstiche tedesche del Trentino. Atti del convegno diS. Orsola, 1-3 settembre 1978, Giovan Battista Pellegrini (ed. ), 153-173. San Michele all~dige: Mus eo degli usi e costume della trentina. Rowley, Anthony R 1986. Fersef!tal (Val Fersina bel Trient/Northernlta~v). Untersuchungen elner Sprachinselmuf!dart [Phonai 31, Monographien 18]. Ttibingen: Niemeyer. Rowley, Anthony R. 2003. Liadlt as de sprodl. Grammatica della lingua mochena Grammatik des Deutsch-Fersentaierlschen. Lusern: Istituto Culturale Mocheno-Cimbro. Kulturinstitut fur das Fersental and Lusern Kulturinstitut Bernstol-Lusern Rowley, Anthony R. 2004. Syntax des Fersentalerischen. In Linzerschnitten. Beitriige zur 8. bayer-isch-osterreichischen Dlaiektologentagung, zugleich 3. Arbeitstagung zu Sprache and Dlalekt in Northernosterrelch in Llnz, September 2001. Stephan Gaisbauer & Hermann Scheuringer (ed.), 529-536. Lin7~ Ada:tbert-Stift:er-lnstltut des Landes Oberoster.reich. Rowley, Anthony R 2006. Bar lirnen z~hraiba un zo reda az be biar. Grammatica del cimbro di Luserna. Grammatik der zimbrischen Sprache von Lusern. Luse.rn: Centro Stampa e Duplicazlonl della Regione Autonoma Trentino-Alto Adige/Druckerei and Ve.rvielfiiltigungsdienst der autonomen Region Trentino-Stidtirol. lstituto cimbro- Kulturin.~titut Lusern Santorini. Beatrice. 1994a. Jiddi.~ch als gemischte OV /VO-Sprache. Linguistische Berichte Sonderheft 5: 230-245, Dialektsyntax, W. Abraham & ]. Bayer (eds). Santorini, Beatrice. 1994b. Some sinilarities and differences between Icelandic and Yiddish. In Verb Mo1•ement, Norbert Hornstein (ed.), 87-106. Cambridge: CUP. Schweizer, Bruno. 2008. (ed. by James R. Dow). Zimbrische Gesamtgrammatik. Vergleichende Darstellung der zimbrischen Dialekte [ZDL- Belheft 132]. Stuttgart: Franz Steiner Verlag. Seiler, Guido. 2003. Prapositionale Dativmarkierung lm Oberdeutschen. Wlesbaden: Steiner. Siemund, Peter & Kintana, Noemi (eds). 2008. Language Contact and Contact Languages [Hamburg Studies on Multilingualism 7]. Amsterdam: John Benjamins. Taylor, Ann & Pintzuck, Susan 2008. The position of objects in Old English: The role of information status. Talk and handout at the 'Ihursday Night Colloquium, University of Tromsoe, Nov. 20, 2008. Taylor, Ann & Pl.ntzuck, Susan. 2009. The effect of Information structure on ob_iect position in the history of English. In information Structure and Syntactic Change. [Oxford StudiF.s in the History ofEnglish 1], Marie-Jose L6pez-Cou.~o. Bettelou Los & Anneli Meurman-Solin (eds). Oxford: OUP.

Spoken syntax in Cimbrian of the linguistic island~ in Northern Italy 277 Tomaselli, Alessandra 1995. Cases ofV- 3 in Old High German. GAGL 33: 93-127. (Also in Ciau.<>e Structure and Larzguage Change, Adrian Battye & Ian Roberts (eds), 345-369. Oxford: OUP. "lbmaselli, Alessandra. 2004. Il cimbro come laboratorio d'analisi per la variazione linguistic a in diacronia e si.ncronia. In Vizriis linguis. Stud/ offerti a Blio l'.iosele in occasione dei suo settantesirw compleanno, 533-549. Pubblica:Z:roni. dell' Universita degli Studi di Verona, Facolta di Lingue e Letterature Strani.ere. Verona: Edizi.oni Fiorini. Tyroller, Hans. 2003. Grammatisd:e Beschreibung des Zimbrischen in Luserrz. Stuttgart: Steine.r: \Viklund, Anna- Lena. 2009. In search of the force of dependent V2: A note on Swedish. Working Papers in Scarzdhzavian Syntax 83: 27---36. Winteler, Jost. 1876. Die Kerenzer Mundart des Kantons Glarus. Leipzig: Winter. Zwart, Jan-Woute1: 1993. Clues from dialect syntax: Complementiz.er agreement. Linguistische Berichte Sonderheft 5: 246---270, Dialektsyntax, W. Abraham & J. Bayer (eds).

Bibliographical titles belonging to the field. beyond the author quotes in the present article Abraham, Werner. 1987. Burz.io triift Wulfila. Zu den distributionellen Eigenschaften von wairdan 'werden' und wisan 'sein' im gotischen Passiv [Groningen Papers in Theoretical and Applied Linguistics --- TTT 9]. Groningen: Rijksuniversiteit Groningen. Abraham, Werner. 1991. The grammaticalization of the German modal particles. In Gram-maticalization [T)'pological Studies in Language 19--2], E. Closs Traugott & B. Heine (eds), 331--380. Amsterdam: JohnBenjamins. Abraham, Werner. 2009. Review of Silvia Dal Negro, 200,1, The Decay ofa Language. The Case of a Gennan Dialect in the Italian Alps, Bern: Peter Lang. Linguistische Bericht<:~ 217: 107--115. Bauer, Brigitte. 1995. The .Emergence and Development of SVO Patteming in Latin and French. Oxford: OUP. Beninca, Paola. I 988. Piccola storia ragionata della dialettologia itaiiana [Monografia 3]. Padua: Quaderni patavini dilinguistica. Beninca, Paola. 1994. La variazione sintattica: Studi di dialettologia romanza. Bologna: I1 Mulino. Bidese, Ermeneglldo. 2004. Tracce di embedded clauseklammer nel clnhro settecomunigiano. In I dialetti e ia montagna..Atti del convegno di SappadaiPiodn (Belluno), 2-6 iuglio 200.3, Marcato, Gi.anna (ed). Padova. Bosco, Haria. 1999. 'Christlike unt korze Dottrina': lJn'analisi sintattica della lingua clmbra del XVI secolo. Tesi discussa nell'aa, Universita degll studi di Verona Cmschina, Silvio. 2006. Informational focus in Sicilian and the left periphery. In Phases of Interpretation [Studies in Generative Grammar 9I], M. Frascarelli (ed), 363-386. Berlin: lVIouton de Gruyter. Geiser, Remigius. I 999. Grundkurs in klassischem Zimbrisch. . Giusti, Giuliana. 2006. Parallels in clausal and nounsl periphery. In Phases of Interpretation [Studies in Generative Grammar 91], Mara Frascarelli (ed), 163-186. Berlin: i\·1outon de Gruyter. Hornung, Mari.a 1977. Pladener W6rterbud1. Glossario Sappadino. Wien: Edition Praesens. Kranzmayer, Eberhard. 1956. Historische Lautgeographie des gesamtbairischen Diaieictraumes [Studien zur osterreichisch--bairisd1en Dialektkunde 1]. Wien.

278 Werner Abraham

Nieuweboer, Rogier. 1998. The Altai dialect of Plautdiitsch (West Siberian Mennonite Low German). Ph.D. dissertation, Unlversity of Groningen. (Book edition 1999. The Altai Dialect ofPlautdiitsch (lVest Siberian .Mennonite Lew German) [Lincom Studies in Germanic Linguistics 07]. Munich: Lincom). Poletto, Cecilla. 2006. Parallel phases: A study on the high and low left periphery of 0 ld Italian. In P1:ases ~flnterpretaticm [Studies in Generative Grammar 91], M. Frascarelll. (ed.), 261-292. Berlin: Mouton de Gruyter. Rapelli, Giuseppe. 1999. Bibliografia clmbra. Cw·atorium Cimbricum Veronese (ed.). Renzi, Lorenzo & Vanelli, Laura. 1983. I pronomi soggetto in alctme varieta romanz~. In Scritti in onore G.B. Pellegrini, Giovan Battista Pellegrini (eel..), 121---145. Pisa: Pacini. Renzi, Lorenzo & Paola Beninca 2000. La venetlzzazione della sintassi nel dialetto cimbro. In L~ole linguistiche? Per un' analisi del sistemi in contatto, G. Marcato (eel), 137--162. Padova: Unipress. Scardoni, Sara. 2000. La sintassi del soggetto nel cimbro parlato a Giazza. Unlversita degli studi di Verona, tesi discussa nell'a.a. 1999/2000. Schmeller, Johann Andreas. 1838 [1986]. Die Cimbern der VII und XIII Communen auf detl venedischen Alpen und ihre Sprache. Mi.inchen. In: Abhandlungen der phllosophischphilologischen Classe der koniglidl bayerisdum Akademie der Wissenschaften II(3): 557-708. Facsimile-Edition Mi.inchen: Curatorium Cimbricum Bavarense. Schmeller, Johann Andreas. 1855/1984. Die Cimbern der VII and XIII Communen in den venedlschen Alpen und lhre Sprache. 1855: Wiener Akademieverlag. 1984. Miinchen, Curatorium Cinbricum Bavarense. Facsimiledruck der Ausgabe von 1855. Schwelzer, Bnmo. 1939. Zlmbrisdre Sprad1reste. Teil1: Texte aus Giazza (Dreizehn Gemelnden ob Verona). Nach dem Voiksm1mde au_fgenommen und mit deutsd1er Obersetzung herausgege-ben. Halle/Saale: Max Niemeyer. Schwelze1; Bruno. 1952. Zimbrische Gesamtgrammatik. Band V: Syntax der zlmbrlschetl Dialekte in Norditalien. Diessen am Ammersee. Tlppskrlpt: Marburg/Lahn - Institut fi.ir die Erforschung der deutschen Sprache. Semolini, Jlederico. 2001. Testi cimbri nelle bibliotheche venete. Dattiloscritto, Universita di Verona. Thtine, H.M. & ~lbmaselll. Alessandra (eds). 1999. Te.si di linguistica tede.
Diachronic clues to grammaticalization phenomena in the Cimbrian CP Andrea Padovan University ofTrento

Saying that a minority language proves to be quite a fascinating object of research when it comes to describing how it is affected by a stronger (standard) language is quite obvious. However, the way in which elements (both lexical and functional) borrowed from the stronger language infiltrate the weaker one is far less evident. Observing the Cimbrian CP, I will take into account both the complementizer ke (Italian loanword, competing with the earlier az) and other (CP-related) particles that I take to have undergone a process of reanalysis from a diachronic point of view, putting forward that the "expansion" ofloanwords (especially functional ones) does not take place randomly. I also v.ill show that synchronic data especially when it comes to the so-called semi-speakers--- can confirm this hypothesis.

Key words. Complementation, minority language, cartography.

1.

Introduction 1

In minority languages, functional loanwords go through a path of grammaticalization that can quite obviously be accounted for as the consequence of the continuous influence from a stronger language in contact In general, speech act particles or functional elements like complementizers infiltrate the weaker language undergoing a reanalysis that deeply affect their semantic and syntactic status. In fact as far as some enclaves in Italy are concerned- it already has been observed (see Stolz (2005: 48-58) among

t. Many thanks to my excellent Cimbrian informants Piorem,o Nicolussi Castellan, Maria Luisa and Andrea Nicolussi Golo who have always put up with my numerous questionnaires. Heartfelt thanks to Paola Benindt, Gildo Bidese and Cecilia Poletto for discussing with me the a~pects of this paper at length. l also am thankful to the organi.1.ers and audience at the conference on "Dialektologie als Brli.cke zwisd1en synchronischer und diachronisd1er Linguistik" held at the University of Verona in December 2009 for valuable discussion. All errors are my own.

28o

Andrea Padovan others) that e.g. Italian speech act elements like allom, 'then or periJ, 'yet, but' can take on a more conjunction-like behavior and a borrowed complementizer- which, by the way, already has an original functional status- can affect the complementation system of the minority language in such a way that it may even become the predominant element of the whole class, especially in younger speakers' grammar. In (1) I display some examples taken from Stolz (2005) containing Italian lexical items found in Albanian, Slavic, and Cimbrian minorities in Italy. (l)

a.

b.

(Arberesh, Stolz (2005)) allura keta tirin sembre me furken then these spin.3.PL always ~ith a distaff per6 but Lor Since per6 but al6ra then pero but

mos uharr6! not DAT forget.IMPER! ka si bila tak• ndelidi.end (M.olise Croatian dialect) you then have become so smart ti cinim na pat I make a deal with you Mm-sa khott di lai.\t (Cimbrlan, Bell otto ( 1978)) have.they said the people bar sain gest sem ... we were there

In a sense, it might be the case that such loanwords are inserted into the edges of phases and not at the level of narrow syntax. The examples in (1) involve the high phase CP but in Ladinian (thanks to Cecilia Poletto for pointing this to me) low adverbials like za, 'already' and furse 'perhaps' (borrowed from Italian) are possibly inserted into the edge of the lower phase vP. Thus, the fact that borrowed adverbials start out being inserted in edges might prove to be an interesting line of investigation to be pursued in future work. In this paper I focus on grammaticalization phenomena within the Cimbrian2 complementizer layer and in particular on the expansion of the "Italian' complemen·· ti:z.er ke. Diachronic research proves to be very useful when it comes to describing the path that these elements might have gone through even when the time span between the oldest documents available and contemporary everyday language amounts to about a hundred years. As far as the syntactic analysis of Cimbrian complementizers is concerned, I refer to Grewendorf and Poletto (this volume). By comparing the tales gathered by

The variety taken into account here is Luserna Cimbrian: Lusema is a small village in the Province of Trento, the capital of the Italian region Trentino--Alto Adige/Stidtirol; it is located ca. 25 km southeast ofTrento. 2.

Diachronic clues to grammaticali7..ation phenomena in the Cimbrian CP

Bacher in 1905 with modern usage, I concentrate in particular on the distribution of the two complementizers ke and az (az being the "original" germanic complementizer). I claim that ke has not infiltrated the Cimbrian CP randomly but along a precise hierarchy from top to bottom. The paper is organi:l.ed as follows: In Section two, I sketch the main syntactic properties of the two series of complementizers in Cimbrian making reference to Grewendorfand Poletto's analysis. In Section three, I emphasize the diachronic change C-elements seem to have undergone, observing the data taken from old Cimbrian sources and comparing them with modern usage. I focus in particular on ke and its tendency to overlap with az~ I then turn to z:~ega, an interrogative element that seems to behave - at least syntactically - in a way similar to ke. Section four is devoted to the analysis of the data: in Subsection 4.1 I seek to give a formal account for the gram-maticalization path of a complementizer like ke, claiming that in contexts of language attrition a functional loanword - a complementizer in this case - can expand in such a way that it can even become the only element of the class where it belongs. In Subsection 4.3 I discuss other elements of the CP--layer that can be thought of as heads in modern Cimbrian, however data exists suggesting these may have previously resided in specifier positions. This last subsection looks far more like work in progress because it draws just tentative conclusions.

2.

2.1

Complementation systems in Cimbrian Two series of Complementizers

Grewendorf and Poletto (this volume), (hereafter G&P), offer a detailed account of Cimbrian complementi:~..ers: here, after listing the properties of the two series of com-plementizers, I outline briefly what G&P's analysis boils down to. The variety of Cimbrian spoken in Luserna has two series of complementizers that differ as to which word order they trigger. The first one displays no asymmetry between main and embedded clauses (exactly like the Romance varieties Cimbrian is in contact with) whereas the second one still preserves the originaJ3 syntactic asymmetry between the two word orders. The elements belonging to the first category are ke, 'that: umbromm, 'because: benn, 'when: bia, 'as' etc. while the elements belonging to the second one are az, 'that: bal, 'when/if: bo, relative complementizer, intanto az, 'while' etc. (for a complete list of

3· The adjective 'original' might not be felicitous here: I use it for explanatory purposes only, as if there had been an early stage of the Cimbrian language when all embedded clauses dis-played a different word order w.r.t. matrix clauses, what might well have been the case.

281

282

Andrea Padovan these items see Panieri (2006)). Since az and ke are the prototypical elements of either class, throughout the paper I will focus on these two items only. a.

b.

Ke is the declarative complementizer; it is a loanword from Italian (or even Trentino dialect) and is typically selected by strongly assertive verbs such as 'saY, perception verbs such as 'see' and some weakly assertive/non-factive verbs such as 'believe: Az could be classified as a modal complementizer because it is typically selected by volitional and factive verbs. It also triggers subjunctive morphology in the embedded clause except when it expresses a temporal meaning in conjunction with temporal adverbs and prepositions such as dopo az, 'after (that): intanto a.z, 'while' etc.

This twofold system somehow recalls the behavior of Southern Italian diale<..is, see Section 2.1. In Table 1, I list the most common verbs occurring with either complementizer. Table I. ke and az-selecting verbs Verbs selecting ke

Verbs selecting az

nonfactive kh6n, 'say' auhalten, 'claim' soln sichar, 'be sure' kontarn, 'telf vorhoasan, 'promise' gioam, •believe' tiian dnt, 'regret'

volit"ional/jactive/negated forms b6lln, 'want' parlrn, 'seem' dubitarn, 'doubt' sperarn, 'hope' net gloam, 'not to believe' net vorstian, 'not to understand'

2.2

Syntactic. aspects of ke and az

G&P point out that embedded clauses introduced by ke exhibit the same properties of the syntax of root clauses; on the other hand, az-type complement clauses have their own features with respect to the list of properties displayed in (a), (b) and (c) below. All examples in this section are taken from G&P's article. However, there might be some slight orthographic change, because the guidelines of a recently introduced spelling reform have been followed. a.

Prefixes. Cimbrian has a class of separable verbs like other Germanic languages. The prefixes of this class of verbs can either precede or follow nonfinite forms but always have to follow auxiliaries in main clauses. (2)

a.

hon au-geh0a1t di arb at have up-given the job b. hon gehoart au di arb at have given-up the job c. *I au hon gehOart di aibat ka Tria

ka in ka in

Tria Trento Tria Trento

Diachronic clues to grammaticalization phenomena in the Cimbrian CP 283

In a ke-type clause the same properties displayed in (2a-b) are valid: there is no asymmetry between root and embedded clauses whereas the syntax triggered by az allows for one more possibility i.e. the prefix can also precede the auxiliary. (3)

b.

... az--ar--en ~ hat geriiaft ... that--he-him up has called

Negation. In main and ke-type clauses negation net comes after both finite verbs and auxiliaries. In az-- type clauses it obligatorily precedes main verbs but need not precede auxiliaries. (4)

a.

Dar is net vortgont he :is not away-gone

b.

I boas ke dar is net vortgont I know that he is not away-gone

c. *I boas ke dar net is vortgont d.

Dar hat gebOllt azz-e net vortgea He has wanted that-I not away-go.SUBJ

e. "Dar hat gebollt azz-e vortgea net

f.

... azz-a--dar

{net} hat {net} khott zo khemma ... that-he--to.you (not) has (not) said to come

Clitic. pronouns and the particle da. Object clitics can be enclitics onto either the Vfin or az. Likewise, the particle da can be found in enclisis to complementizers, typically in relative clauses, but also to a.z; in main clauses it is sometimes com-bined with the Vtin to (possibly) instantiate habitual aspect in some contexts (Sa), progressive aspect in others (Se ), and a sort of "double"4 expletive together with 'z in existential sentences (Sf): (5)

4·

a.

Da soin vortgont ena az-ta-s niamat barn they are away-gone before-da-it nobody notice-sunr

b.

I gloabe ke dar gebat-mar-s I think that he gives-to.me-it

c.

Aile sunta Mn-da gelautet di klokkng the bells Every Sunday have-da rung

d.

Bal-da rivan di khindar sperrbar di rur when-da arrive the kids, close-we the door

e.

'z gian-da laut ka miss it.go.PLUR.da people to Mass "People are going to Mass"

f.

'z soin-da ploas bar ka Tria "there are lots of bars in Trento"

For the reason why da is dubbed "double expletive" see § 3. 3.2.

284 Andrea Padovan

To sum up, G&P put forward that Vfin is in complementary distribution with rtz- type complementizers; only in main and ke-type clauses can it reach a high position moving past negative markers and gathering up clitics on its way upward as we can see in (6). On the other hand, ke is merged in a position even higher than Force thus allowing verb movement through the left. periphery as is the case of root clauses. ( 6)

[SubordP

ke

[ForceP

az/Vfin -tafditicsk. [TopicP •• • [WackP tkHGroundP tJ lrr [AspP prefix [NegP net] ·· · (.,,.p]]]]]]]]]]

[TopicP •• • [FocusP [FinP ti

With this background, I will now seek to give a formal account of the "pressure exerted" by a strong standard on a minority language in isolation. In particular, the diachronic evolution of both ke and az can shed light as to whether functional elements actually infiltrate the functional layers from top to bottom.

3·

The diachronic data collected

As mentioned before, the data taken into account here span over a period of about a hundred years. The oldest texts stem from Joseph Bacher's work "Die deutsche Spmchinsel Lusem" published in Innsbruck in 1905. The author repeatedly conducted his fieldwork in Luserna gathering orally handed-down tales he transcribed using a particular phonetic alphabet to render the idiosy-ncrasies of Cimbrian pronunciation. I then concentrated on Bellotto (1978), who edited an up-to-date version of Bacher's corpus getting first rid of the phonetic symbols, making thus the text more accessible to readers. Moreover, he slightly modified some syntactic constructions that had been used early on and that must have been perceived as obsolete in the 1970s. As for modern Cimbrian, the examples displayed here belong to a corpus of data I have collected by means of administering questionnaires to a group of tluent speakers. However, when it comes to minority languages it is important to recall that speakers can be divided into different groups according to their competence, which can be more or less (in some cases even severely) eroded. In particular, there are minority language speakers who are referred to as "semi·· speakers" if they have a restricted code avail· able and are able to speak the weaker language only in limited contexts. However, at the same time they make themselves understood in the community since their grammar, though poor, allows them to build simple sy-ntactic constructions. This means that they do not produce isolated words only: not that such speakers do not exist in minorities, 5 but they are usually discarded from the group of informants for obvious reasons when syntactic properties are being observed. I decided to include the relevant semi-speakers' production in this comparison between old and modern Cimbrian as these particular speakers somehow represent a

5·

These speakers are sometimes referred to as "terminal speakers':

Diachronic clues to grammaticalization phenomena in the Cimbrian CP 285

further (in a sense "extreme") step in the grammaticalization path w.r.t. younger fluent speakers who still preserve the main features of the standard language even though they manifest a great deal of innovation as well. 6 In the following subsections I will show in which way ke seems to have "evolved" overlapping the usage of az.

3.1

The "evolution" of ke

The number of contexts where ke can surface is increasing to the detriment of az. In (7) we can see it has become obligatory in contexts where it was not in Bacher's tales (7a---b) vs. (7c). 7 (7)

a.

is toat (DSL) bal-se hat gesekk ke soi pual when-she has seen that her beloved is dead

b.

on bal dar har gesekk, 0 de sichl bil the sickle will and when he has seen net khemmen ... (DSL) not come (off) ...

c.

dar Hihls hat gisekk •(ke) soi sun studjart mearar (CL) that his son studies more has seen John

Consecutive clauses were introduced either by az or ke in Bacher's corpus; nowadays only ke seems possible. (8)

6.

a.

dar armen muatar hat-s-ar getihlt ant a:1..-ar sai gestorbet 'z (DSL) khinn as-ta de laut Mm gemuant, se khint narr«t "the poor mother regretted so much that her son was dead that people thought she could go crazy"

b.

on disar giirtner hat getant aso letz, ke dar man hat gemocht avangen zo lacha starkh (DSL) "and this gardener made it so bad that people could have begun to laugh out loud"

c.

'z khinn hatt sovl gegaillt ke sa Mms geh6art fin atz bege (P) "the baby was crying so loud that people heard it even down the street"

For details on the coiKept of senli -!speaker see Dorian ( 1977, 1981)

7· From now on and when necessary I will use the following acronyms to make reference to the sources I took the examples from: DSL, Bacher's "Deutsche Sprachinsel Lusern"; CL, the CimbroLang corpus associated with a project on Cimbrian senli -speakers being carried out at the university of Trento; Bl978, Bellotto's "I racconti di Luserna"; P, Panieri (2006).

286 Andrea Padovan

Weakly assertive/non-factive verbs such as believe select both ke and az but in different contexts: the presence of az is connected with embedded irrealis mood that manifests itself in subjunctive morphology. Moreover, the negative counterparts of non-factive verbs select rtz (as is the case of Italian bridge verbs). Therefore, there seems to be a connection between az and an irrealis mood head in IP - in a split-IP framework a la Cinque (1999, 2006) - at least in older speakers' grammar. (9)

FinP > ·· · [IP MoodPspeech act> MoodPevaluative > MoodPevidential > ModPepistemic > TP(Past) > TP(Future) > MoodPirrealis > ModPalethic > AspPhabitual > ·· ·

As for the connection between negative marking of the matrix verb and the selection of the complemetizer there are some Romance varieties in which negation gives rise to the same effects, see among others Blasco-Ferrer ( 1986) and Damonte (2006) on the Sardinian complementi1..ers ca selected by verba dicendi and chi which has a modal status. (Hl)

non creu chi issu nci bessat con custu tempus I don't think that he out go--sUBJ with such (bad) weather (Sardinian, Blasco --Ferrer ( 1986))

In ( 10) the presence of the negation triggers the selection of chi, while the non-negated form creu, 'I think' requires the declarative complementizer ca. Giorgi and Pianesi (1997) had pointed out that it is possible to account for the relation between clause--typing and mood selection assuming that a low CP--head is endowed with a Mood feature. Rizzi (1997 and subsequent works) has pinned down the features ofthis projection- FinitenessP - that faces "inward" (i.e. toward the complement clause) and relates the CP--layer with tense and mood. Therefore, it is not unreasonable to maintain that az is merged in that projection. The fact that it ends up in Force is not a problem since we know complementizers can move through the left periphery to check features. On the other hand, the fact that ke can never trigger subjunctive selection and the order *['lop> ke] is impossible suggests that ke cannot be merged in Fin°, but it necessarily occupies a much higher position and all data presented so far bear out this hypothesis. However, this assumption turns out to be a problem when it comes to younger speakers' usage of complementizers: Some of them are liable to use ke in all contexts where az is expected. Semi-speakers show an even stronger tendency to extend the use of ke even in those contexts where it is impossible in both older and younger speakers' competence. (11)

a.

bar ham gegloabet ke 'z i!.i spiit (CL) we have thought that it is late

b.

bar ham gegloabet az--ta sai spiit (CL) we have thought that.da is-sUBJ late

Diachronic clues to grammaticalization phenomena in the Cimbrian CP 287 c.

gloabe ke 'z renk haint (CL) think that it rains tonigl1t

d.

reng (CL) gloabe az-z Mint think that it tonight rai.n-SUBJ ke 'z is--ta khent eparmnmas [---ilTealis] (CL) arrived somebody it seems (to me) that it is

e.

'z parirt (me)

t

az-ta sa:i 'z parirt (me) khent eparummas [t-i.rrealis] (CL) it seems (to me) that it be-SUBJ arrived somebody

g.

doi tatta hatmar net khOtt azzar hat gekoaft 'zhaus (CL) your father has.me not said thaLhe has bought the house

h.

miar (CL- semi-sp.) bill ke du kisto pit 8 want that you come v.ith me

(lla) and (11c) are used when the speaker has a strong presupposition concerning the truth values of the complement clause. On the contrary, (11b) and (lld) express uncertainty, since the speaker has restricted knowledge of the facts or no direct evidence as is the case of (llf).9 Now, one might wonder whether ke has been slowly taking on the properties of az becoming a modal complementizer: In my opinion the answer is negative. But before addressing this issue in Section 4.1, I introduce another complementi:t.er-like element that deserves closer investigation.

3.2

Zsega

Embedded interrogative clauses can display a comp-like element. zsega, literally 'to see, 10 co--occurring with (embedded) wh--elements. A fully lexical z'sega also exists: it still preserves its original meaning, 'in order to see/observe: and is often spelled 'zo sega: at least in Bacher's version of the tales.

8. The form 'kisto' contains a clitic used in interrogative and V2 contexts, thus it is not expected here but semi-speakers' competence is eroded even as far as verbal paradigms are concerned. 9· In this case Mood.,.,;d.ntiol instead of Moodirrealis might be activated in IP and trigger subjunctive morphology.

More specifically this item L-; formed by the infinitive marker z(o) that selects the special infinite form always ending in -a, sega (cfr. the full infinitive seng), which is used only in these constructions.

10.

288

Andrea Padovan (12)

dar man isse-se gevuntet aso guat sem in disan gart zo sega da geat als viirsnen (DSL) that man was so happy to see (=noticing) (in that garden) dais going everything well

However, functional zsega co--occurring with embedded commonly observed especially in colloquial language. (13)

a. b.

wh-element~

is much more

geat (DSL/B 1978) gevorst z'St.-ga bo-dar ... Mm se asked zsega where-he goes-sUBJ ... have they vors in gotfnhearn z'sega umbromm 'z khemmen--da zsega why it grows-da god

aSk

khuane mear [baimarn] no more grape c.

iz vii.rpasart akarotz ... on hat gevorst z'sega be da a carriage and he asked zsega whether-da had passed

Zsega can be followed both by ke-type and az-type wh-elements. (14)

a.

i vorsmar z'sega bettana 'z khinta I wonder z'sega what--sort--of woman is coming

ka uns [zsega + wh + matrh.: word order] to ·us b.

moi tatta hatmar gevorst z'sega ber-da my dad has-me asked zsega who hat ~geruaft [zsega + wh-da +embedded word order] (P) has phoned

Furthermore, all embedded interrogatives are perfectly fine without z'sega, i.e. it is an optional element. However, some older speakers still perceive a subtle difference: Let us take example (14b) repeated here as (15a---b). (IS)

a.

moi tatta hat mar gevorst z'sega ber-da hat ~geruaft has phoned my dad has-me asked zsega who

b.

moi tatta hatmar gevorst ber-da hat ageruaft (CL) has phoned my dad has-me asked who

The presence of z'sega entails a presupposition: The utterer of (15a) is aware ofherihis dad knowing already who might have phoned; in this case the wh-element scopes over a (small) set of given (and kno~--n/familiar) individuals. At any rate, younger speakers have a more restricted competence of z'sega and tend to leave it out. 'Ihrough testing the reciprocal position of zsega and topicalized NPs, I discovered that the sequence Top > z'sega proved to be impossible as will be shown in the next sec.tion. To sum up, it is reasonable to assume that z'sega is merged in a very high position.

Diachronic clues to grammaticalization phenomena in the Cimhrian CP 289

Analysis: The ongoing changes in the Cimbrian CP

4·

4.1

Ke

The diachronic data I presented in Section 3 confirm G&P's analysis: old and modern Cimbrian pattern alike regarding the position of ke (and all other ke-- type complemen-tizers), which is merged in a high structural position over Force allowing for finite verb movement to occur in the left periphery because all target positions are available. Now, we might wonder why ke cannot instantiate lower CP projections much like Italian che; however - at least in initial stages of contact - it is not unreasonable to assume that functional items borrowed from a stronger standard start out surfacing in the highest functional layers of the weaker language. If the pressure exerted by the standard is strong enough, in the course of time the functional loanwords begin to expand, eventually taking on features formerly proper to their native homologous counterparts. 1herefore, the question is: Does this process of expansion (and reanalysis) occur randomly? In other words, can these items lexicali7.e whatever functional projection as soon as they have acquired new features? In gener.U the answer seems to be negative. Let us observe for a moment the situation oflanguage attrition between Standard Italian and Calabrian dialects described by Egerland (2003): We notice that there is a precise path that this expansion process must follow. In the Calabrian varieties taken into account, restructuring verbs select a comple-menti;.;r.er and a finite verb instead of the usual construction with infinitives. 11 (16)

a. he begun to sing (standard Italian construction) b. he begun that he sings (dialectal jorm)

Egerland shows that there is a general loss of finite forms ( 16b) while infinitival forms (16a), typical of the Italian syntax, are becoming common. In other words, dialectal speakers begin to use constructions typical of the stronger standard. Anyway, this infiltration seems to be rigidly regulated. In fact, :Egerland shows that the expansion of infinitival fonns does not come about randomly, but along an implicational scale that matches the hierarchical structure of the sentence in the spirit of Cinque {1999, 2006). In the four major restructuring contexts tested by means of questionnaires, the results are consistent with the (sub-- )hier.rrchy in (17). (17)

mental ability modal'>> deontic modals > aspectual verbs> ergative verbs of movement.

n. This phenomenon is typically associated with Balkan languages but is widespread in Southern Italian dialects as welL

290

Andrea Padovan Therefore, a top-downwards infiltration of Italian constructions into dialectal usage takes place: There is a first subgroup of informants that use infinitives just with mental ability modals preserving the dialectal construction in the other three cases; another subgroup uses infinitives with mental ability modals and with deontic modals while dialectal constructions are maintained in the other two cases and so on. 'Ihis implica-tion holds tor all contexts. Let us now go back to Cimbrian: the expansion of ke seems somehow to resemble what we have seen so far. If we observe the behavior of ke in Old Cimbrian and in the Cimbrian of contemporary fluent speakers, we can assume that this comple-mentizer starts out as a general subordinating element in the spirit of G&P. However, as for younger speakers (even more so for semi-speakers), we have to account for the fact that ke seems to take on features matching with lower positions of the CP--field: Fin but also Force, because as we have seen a subgroup of informants use ke as a relative complementizer. (18)

Grammaticalization path of the Cimbrian c.omplementizer (preliminary): if a complementizer loses old or acquires new features it can be directly merged in other (lower) functional heads: [s, ordP {ke} h:arceP {ke} ·· · [FocusP [FinP {ke} [!P [vp]]]]]

Note that I am not assuming that grammaticalization involves downward movement (though at a diachronic stage) that should be avoided in general; lexi.cal items becom-ing functional items show the opposite pattern ending up occupying higher functional projections. However, recall that (i) ke already has a f-unctional stat-us; (ii) the process of expansion in an area that could be conceived of as a "bridge" between two clauses can in principle entail what looks like downward movement but it is just the instantiation of a different head of the CP--layer. This hypothesis is not unproblematic from other viewpoints, though: If ke is really merged in the bottommost CP projection, then we should expect (i) that verb movement to Cis blocked; (ii) that clitics and the particle da cannot encliticize to the complementizer; (iii) that the asymmetry between matrix and embedded clause syntax gets lost. In fact, (i)-(iii) are the hallmarks of the production of semi-speakers and some young fluent speakers. Moreover, subjunctive morphology is virtually absent in ke- type clauses even when ke is selected by a volitional verb: this leads to assuming there is no longer an Irrealis Mood activated in the IP and long distance-checked by FinP. The final step of this process might be the loss ofV2 in general: this is just a tentative (but not unreasonable) idea that deserves more data in order to be borne out. Therefore, I leave the question open fur future research. To sum up, there is evidence for arguing that a gradual, top-down infiltration of the CP-layer has occurred but, still, it is difficult to show that the CP gradually becomes "deactivated": In other words, if it is indeed the case that ke begins to be merged in Force, the target position of the Vfin is no longer available. Therefore, even if the verb can reach

Diachronic clues to grammaticali7..ation phenomena in the Cimbrian CP Fin, its movement upward is blocked and ditic gathering cannot occur. Accordingly, I would like to restate (18) as (19) in terms of inactivation of the relevant heads. Note that I am not claiming for an unsplit CP: Sticking to a strictly cartographic approach amounts to assuming that CP-Iayer projections must be present in order for the hierarchy of features to be represented in the syntax. As a result we have a "simplified" CP but, crucially, not an unsplit one i.e. some projections are present but may not be accessible. 12 (19)

Grammaticalization path of the Cimbrian complementizer (revised): when a complementizer undergoes a process of expansion giving rise to more restrictive syntactic operations within the CP-layer some projections may not be active or accessible: ("SubordinatorP"

[ForceP

ke

ke

((Fon:eP ·· • {[FocusP} [FinP)

({[FocusP} [FinP)

[;p (yp])]]]) or

lrp [yp]]]]]

Roughly speaking, there might just be a topmost head, say Subordinator or Force, under which subordination and clause-typing are subsumed. This head is followed by an inactive C P-layer (in round bracket in (19)) that does not allow for syntactic operalions to occur at tl1is leveL Verb movement may be restricted to IP.

4.1.1 A parailel l will now just touch on an interesting paralle~ namely the situation of language con-tact between a variety of Arabic and an Indo- European language. Khuzistani Arabic is a minority language spoken in Western Iran (see Matras (2007); examples cited from there) in contact with Farsi. Among the functional elements borrowed from Farsi, there is the factual comple-mentizer ke, 'tl1at: (20)

a.

rayl-<'lc caMab ~ala kh<'>l-si tJdr-in ke know-2sG.F COMP husband-2sG.F on everything lied.3sG.M 'you know that your husband lied on everything'

b.

hna ... marta ke huwwa w Yazi <:lb<:ln uxu and wife--3SG.M here ... son brother Ghazi REL he (.Matras (2007: 145)) 'Ghazi's nephew who is here with his wife .. ;'

Matras points out that in a first stage ke is just a factual complementizer that does not appear innonfactual (subjunctive) complements where an Arabic historical relativizer

However, it might be the case that Focus and Topic fields are still active: in fact the syntax/ pragmatics intetface is likely to be preserved in semi--speakers' competence; in (19) Foci are enclosed in braces to indicate their plausible optionality. Anyway, I \-\111 administer further questionnaires focusing on this issue.

12.

291

292

Andrea Padovan (l-lalJi) is found instead. He also assumes a gradual process of convergence between the Farsi and the Khuzistani Arabic complementation systems. The former has an identical marker for complement clauses and relative clauses; this model is copied in the latter with the final result of extending the usage of ke even in relative clauses,

which seems to be a general tendency attested in contemporary language although the form l-lalJi is still preserved.

4.2

Zsega

Let us now take up the questions we left open in Section 3.2 regarding the position in which z'sega surfaces. Recall that there are two classes of wh-elements in Cimbrian behaving differently w.r.t. the word order they trigger. G&P claim that wh-items triggering matrix word order are merged very high, namely in a referential, topic--like phrase above Force (I refer the reader directly to their chapter in this volume for the details of the analysis). Z'sega co-occurs with both classes of wh-items; this leads to assuming tl1at z'sega is merged in the topmost projection of CP. (21)

lxp z'sega [RefP Topic--whiDPSubject [ForceP 'z vfin ... [FinP ]]]]

z'sega looks somehow like an "interrogative" subordinator in the same fashion as ke is a general subordinator (cf. also Kolmer (2007), which discusses the complementizer nature of z'sega). Moreover, the reciprocal position of z'sega and topicalized items confirms that it can never show up under a Topic. (22)

a. *dar hat--mar gevorst [moin auto ]10P z'sega bo--dar he has.me asked my car zsega where.da.I hatt -en gekoaft have.it bought b.

dar hat-mar gevorst [moin auto]TOP bo-da:t· hatt-en gekoaft

c. *dar ham--mar gevorst [moinpruader]1'0P z'sega benn-i he has.me asked my brother z'sega when.I hil.nn-en gesekk have.him seen d.

darham-margevorst [moin pnUlderhoP benn-ihil.nn-en gesekk 13 (CL)

13. There is just one counterexample in my corpus, but I decided to disregard it because it is an instance of ellibedded yes/no question that might entail a different structure. Por this reason, (i) might play a minor role here. (i)

i vorsmar [doi autohoo z'sega azzen hat gekoaft dar Mario I wonder your car:ro~ z'sega whether-it: has bought the Marlo

Diachronic clues to grammaticalization phenomena in the Cimhrian CP 293 Leaving out zsega., 1bpics are perfectly fine in embedded interrogative contexts, see (22b) and (22d). On the one hand, the fact that that the order Top> z'sega is ungrammatical leads to excluding that z'sega occupies Int(errogative)P in Rizzi's (2001) sense, and it is, therefore, not comparc:tble to Italian se, "whether" On the other hand, zsega. and ke seem to both be merged in the topmost projection of CP. However, "Subordi-natorP" is too loose a label to host declarative and interrogative subordinators. In the spirit of Bhatt and Yoon (1991), SubordinatorP is the topmost CP-layer projection, which is active in those languages that do not contlate subordinating and clause-typing; this work put forward that languages displaying embedded V2 (e.g. Yiddish) split the Comp category in a Subordinator and a Mood head, which allows for V6n movement to CP. Certainly, ke and z'sega are both merged very high in the structure. All the same, the fact that they occupy the same position is questionable to say the least. It might then be the case that this uppermost projection has a finer--grained internal structure that deserves more investigation. Julien (2008) exploits the idea ofSub(ordinator) to account for embedded V2 in Scandinavian languages: in particular, she assumes that a subjunction phrase is always merged over Force. Embedded V2 is taken to involve movement of either the subject or a topicalized XP to Spec, Force and the V6n to the head of Force, ForceP encoding the illocutionary force of the sentence. In non-V2 embedded clauses, the subordinator is base-generated in SubP 14 anyway, but no Forc.e is merged because there is no illocutionary force to be instantiated and the sentence displays the usual embedded word order. 1hus, Julien claims for an unsplit CP where Sub is merged directly over Fin and V6n remains stranded in IP; in other words, Sub and V6n do not compete for the same position, differently from Cimbrian az. Anyway, I believe she is right in taking Force not to conflate both illocu·tionary force and clause typing. Force could then be conceived of more as a field than a single projection, on top of which a SubP is always merged. 4·3 4.3.1

Old specs, new heads

Cimbrian 'Be'

Van Gelderen (2004, 2009) deals with diachronic changes in the (English) CP: she claims that economy principles such as Lctte Merge Principle (LMP) and Head Preference .P1·inciple (HPP) are responsible tor the reanalysis of elements as higher in the structure and as heads. In particular, LMP and HPP, which in a nutshell means 'Merge as late as possible' and 'Be a head rather than a phrase' respectively, are both at play in the grammaticali7.a-tion path of wh-·pronouns such as Old English hwreper, 'whether'.

14.

See also Haegeman (2006) for a similar definition of Sub.

294 Andrea Padovan

'

(23)

[cp wh; [c" 0 ... [vp ti ~ [cp wh [c· 0 ...

[___________________________________!!

via LMP I stage

listage

[cp

[c· wh via HPP III stage

In the first stage, the usual movement of the interrogative pronouns to a Spec of CP is represented; in the second stage the pronoun is reanaly:t.ed as an interrogative marker directly merged in the Spec; finally in the third stage the head is reanalyzed as a question indicator: at this stage new wh-items are likely to appear and the cycle goes back to stage one again as can be seen in in (24).

(24)

a.

Hw.e(jer para twe3 ra dyde paes f<eder willan? (stage I) which of the-GBN two did the-GEN father will?

b.

Nast ou hwae(ler beoo paes rican mannes ban (stage II; V to CO) not-know you whether be of that rich man's bone

c.

oa cwredon ... hwreoer renig man him they said... whether any person him mete brohte (stage III, head) food brought

(van Gelderen (2004: 94))

In Cimbrian the -wh--item be (alternatively spelled bi) introduces an embedded yes/no question and it looks like it is a phonologically weak form of bia, 'how'. Note that be and bia were interchangeable in Bacher's corpus while both Bellotto's re-edition and modem spoken Cimbrian use the two items with different functions.

(25)

a.

Be da di Loserner ham getdnt zo khemma zo wissa, be da Ia pesta no is bait vor (DSL) "How people from Lusern happened to know, whether the pest is still far away"

b.

zboa drai tage darnA soin-sa bidar gekheart zo ghiana au z'sega be 'z steat pet'n proat (DSL)

c.

zboa drai tage darnA soin-sa bidar gekheart zo ghiana au z'sega bia 'z steat pet'n proat (Bl978) ''two or three days after they came back to see how the bread was like"

In Example (25a) which is the title of a tale the first be functions as a full-fledged wh-item meaning 'how, in which way'; the second one just introduces an embedded yes/no question. Example (25l~) displays an interesting change in the 1978 version of the original tale transcribed by Bacher i.e. be can no longer be used as equivalent to bia. Thus, it seems that be!bi has an evolution similar to Old English hwa~per, starting out as a wh-element and ending up being a head due to ec.onomy reasons, as proposed

Diachronic clues to grammaticalization phenomena in the Cimbrian CP 295

by van Gelderen. C[ also Alber (1994) who proposed that wh-elements like wia or wos are reanalyzed as heads in some Austro- Bavarian varieties. Cimbrian 'da' Cimbrian, like most Germanic languages has an e.xpletive pronoun, 'z, which behaves in a way similar to German Vo~feld--es.

4-3.2

(26)

a. b. c.

'z renk "it rains" 'z tiiatmar Ant "I am sorry"

'z givalltmar di milch (P) "I like milk" (personal constr.)

z

also is used in presentative/existential clauses together with the particle da, which surfaces enclitic onto the finite verb so in, 'be'. (27)

a.

'z is-ta no vil snea at di techar there is still a lot of snow on the roofs

b.

'z soin-da net Yil sbemm in halt there aren't many mushrooms in the woods haiir soin-da net vii sbemm in balt 15 (P) this year there aren't many mushrooms in the woods

c.

As we have seen, the particle da nowadays has a restricted usage (cf. § 1.2 ). However, in earlier stages of Cimbrian as far as I can see it was in free variation with 'z at least in ergative/ unaa.~usative wnteJ..i:s; therefore, it must have been an expletive element. In the following examples we compare as usual Bacher's and Bellotto's usage. (28)

a. b.

c.

d.

dar hear haltet-en aso gearn ke da soin vortgant di bochan sovil as-be de urn (DSL) dar hear haltet-en aso gearn ke 'z soin--da vortgant di bochan sovil as-be de urn (B 1978) "the m-mer cared for him so much that the weeks went by like they were hours" dar man isse-se gevuntet aso guat sem in disan gart zo sega da geat als viirsnen (DSL) dar man isse-se ge\'Untet aso guat sem in disan gart w sega 'z geat als viirsnen (B1978) "that man was so happy to see (=noticing), in that garden, da is going everything well"

15. Notice that in (27c) no "Vorfeld-z" is present when there is a topicalized element.

296 Andrea Padovan

Along the lines of G&P, modern da. is a particle (possibly) connected with habitual and progressive aspect. They argue it is merged in Ground0, a projection belonging to the Topic layer, it must nonetheless have been a there-expletive early on (note that it is etymologically and semantically connected with the locative adverbial da). As I have few data available, I tentatively assume it was used only in ergative/unaccusative contexts or with post-verbal full DP subjects or quantifiers. 16 Eventually 'z must have won out over da becoming the only expletive pronoun. Da seems to have undergone a grammaticalization path similar to be because it started out as a specifier being then reanalyzed as a head. But why does this binary system tend to stabilize preserving just one of the two elements? Such "binary" systems of expletives are not uncommon. 1his coexistence is attested in Old High German: FuB (2009) shows that do!da 'there/then began to be used as expletive before Vorji?.ld--es developed. The gradual loss ofdoida (which, by the way, begun to be associated with V2, whereas it had a pragmatic function, at least at an earlier stage) seems to have given rise to the development of es. Apart from German dialects, Modern German still preserves da, but only in specific contexts such as idioms and existentials - and with restrictions. All Germanic V2-languages have expletives but they differ as to what kind of subject is allowed to co-occur with an expletive. Gennan and Icelandic are the most permissive ones whereas Scandinavian lanbruages and Dutch rule out all kind of definite subjects. Platzack (1998) and Boeckx (2001), among others, argue that expletives have different functions depending on the language: First, it is necessary to distinguish between the EPP per se, which is checked in FinP, and a general "visibility requirement" that the complementi7.er layer is endowed with, due to its discourse--related nature. According to this requirement, lexicalization of some G- projection must occur. On the one hand, languages such as Swedish or Dutch have to check both the EPP and [+finiteness] features 17 whereas Icelandic just has to fulfill the visibility requirement forcing lexical elements to instantiate CP--layer projections. This is the reason why no expletive need be present if the relevant CP--layer projection is lexicalized by some constituent (29c-d). On the contrary, Swedish always displays lexical expletives both in first or post verbal position (29a-b ).

My older fluent speakers consistently reject both (28a) and (28c). Thus, this construction is ill-formed in modern Cimbrian; consequently, Bellotto's change is not a stylistic variant but the only possible form. 16.

17. EPP and finiteness are taken to be strong in languages such as Swedish or DutCh and weak in Icelandic (and German).

Diachronic clues to grammaticalization phenomena in the Cimbrian CP 297 (29)

a. *(det) har kommit manga linguister hit idag (Sw.) many linguists here today it have come b.

Idag har *(det) kommit mdnga lingui'lter hit

hafa komia margir malvfsindamenn hingad i dag (Ic.) c. *(pad) here today it/there have come many linguists d.

i dag hafa (*pa<J) komi
e. •·(es) sind viele Linguisten heute gekommen (Ger.) heute sind (..es) viele Linguisten gekommen

(Boeckx (2001: 52)

z

Old and modem Cimbrian differ inasmuch as nowadays is the only expletive element and da is just a head used in the contexts described above but, exactly like Icelandic, the visibility condition is the only criterion to be met. Recall that disappears in interrogative and topicalization contexts, as can be seen in (30):

z

(30)

a.

'z iz kalt 18 it is cold

b.

iz kalt? is it cold? 'z ista khent di miirnma it is··DA arrived the mother

d.

ista khent di mamma? is-DA the mother arrived?

e.

'z soinda ploas bar ka Tria it are lots of bars in Trento

f.

ka Tria soinda (*'z) ploas bar

Contexts like ( 30b ), which typically display total absence of the pronominal subject. also block the presence of da, cf. *izta kalt? (but other weather verbs like reng, 'to rain' have obligatory in questions, ct: renk ("''z)?, 'is it raining?: thus impersonal verb soin, 'to be: is totally subjectless in interrogative and topicalized contexts). On the other hand, in presentative/existential constructions like (30c-f) da is incorporated into the verb. As a tentative line of research we could say that da, though no longer an expletive element, has retained some personal features (at least 'singular' and 'plural') that match with the real expletive element 2:. Also recall that the relative complementizer bo shows up as boda (i.e. along with an incorporated ··da) only when

z

Ct: Icelandic var forward there.

18.

(~palJ)

kalt?, 'was (it) cold?' in

Sigur~sson

(2008) and the analysis put

298 Andrea Padovan

full NP subjects are present in the relative clause, otherwise bo just incorporates weak subject pronouns (both subject and object): (31)

a.

dar mann boda di khindar Mm gesekk the man that the children have seen

b.

dar mann bo bar hAm gesekk the man that we have seen

c. *dar mann boda barhamgesekk Tentatively, we could conceive of da as a default pronominal clitic in the topic·· subfield. Anyway, this particle seems to have complex properties that are definitely not accounted for in this last section, which just tries to put forward a line of investigation that will be dealt with more thoroughly in future research.

References Alber, Birglt. 1994. Iruilzi per l'l!slstenza di uno split-CP nelle lingue germaniche. In Teo ria del linguaggio e analisi linguistica, Gianluigi Borgato (ed.). Padova: Unipress. Bacher, Jose£ 1905. Die deutsche Sprachinsel Lusern. Innsbruck: Wagner'sche UniversitatsBuchhandlung. Hellotto, Alfonso. 1978. 1raa:onti di Luserna, Roana: Istituto dl. cultura Cimbra Dal Pozzo. Bhatt, Rakesh & Yoon, James. 1991. On the compositl.on of COMP and parameters of V2. In Proceedings of the lOth Wt:st Coast Conference otl Formtil Linguistic_~ (WCCFL), Dawn Bates (eel.), 41-53. Blasco- Ferrer, Eduardo. 1986. La iingua sardtz contemportmea. Cagliarl: Edizlonl della 'Iorre. Hoeckx, Cedric. 2001. On the co-occurence of expletives and definite subjects In Germanic. In Issues in Formal German(ic) Typology [I.Inguistlk Aktuell!Lingulstics 'Ioday 45], Werner Abraham & C. Jan- Wouter Zwa.1t (eds). Amsterdam: John Henjamins. Cinque, Guglielmo. 1999. .ttdverbs and Fu11ctional Heads. Oxford: OUP. Cinque, Guglielmo. 2006. Restructuring and Functionai Heads. Oxford: OUP. Damonte, Federico. 2006. Complementatori e complementi congiuntivi in alcuni dialetti sardi. In Quaderni di Lavoro dell1\Slt 6, Andrea Padovan & Nicoletta Penello (eds). . Dorian, Nancy C. 1977. The problem of the semi-speaker in language death. Linguistics 191: 23-32. Dorian, Nancy C. 1981. Language Death. Ihe Life Cyde of a Scottish Gaelic Dialect. Philadelphia PA: University of Pennsylvania Press. Egerland, Verner. 2003. Acquisizione, "-ariazione e cambiamento. Paper presented at the conference IX Giornata dl Dialettologia, University of Padua Fu:6, Eric. 2009. Expletive do/da 'there, then' and the rise of V2 in German. Paper presented at the conference Dlaiektoiogie als Briicke zwischen synchronischer und diachronisd1er Linguistik, University of Verona.. van Gelderen, Elly. 2004. Grammatic.alization as Ec.onomy lLinguistik Aktuell/Linguistics Today 71]. Amsterdam: .John Benjamins.

Diachronic clues to grammaticali7..ation phenomena in the Cimbrian CP 299 van Gelderen, Elly. 2009. Renewal in the left periphery: Economy and the complementizer layer. Transactions of the Philological Society 107: 2 131-195. Gi.orgi., Alessandra & Pianesi, Fabio. 1997. Tense and .Aspect. From Semantics to A1orphosyntax. Oxford: OUP. Haegeman, Liliane. 2006. Conditionals, factives and the left periphery. Lingua 116: 1651-1669. Julien, Marit. 2008. Embedded V2 in Norwegian and Swedi.sh. Working Papers in Scandinavian Syntax 80: 103-161. Kolmer. Agnes. 2007. z'sega im Cimbro: Vom Verb zum Komplementierer? Ms, University of ZUrich. Matras, Yaron. 2007. The borrowabllity of structural categories. In Grammatical Borrowing in Cross-Linguistic Perspective [Empirical Approaches to Language Typology 38], Yaron Matras & Jeanette Sakel (eds). Oxford: OUP. Panieri, Luca. 2006. Bar lirnen zsdtraiba un zo reda az be biar. Lusern: Kulturinstitut Lusern. Platzack, Christer. 1998. A visibility condition tor the C-domain. Working Papers in Scandinavian S_vntax 61: 53--99. Rizzi, Luigi. 1997, The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281-- 337. Dordrecht: Kluwer. Rizzi, Luigi. 2001, On the position 'Int(errogative)' in the left periphery of the clause. In Current Studies itl Itaiian Syntax Essays ojfered to Lorenzo Renzi, Guglielmo Cinque & Giampaolo Salvi (eds), 287---296. Amsterdam: Elsevier. Sigurasson, Halld6r. A. 2008. On EPP effects. Ms, University of Lund. Stolz, Thomas. 2005. Italianisierung in den alloglotten Sprachen Italiens. In Das Zimbrische zwischen Germanisch und RomanisdJ, Erminegildo Bidese, James Dow & Thomas Stolz (eds), 114--128. Bochum: Brockmeyer-Verlag.

Hidden verb second The case of Cimbrian Gunther Grewendorf & Cecilia Poletto University of Frankfurt/University ofVenice

The aim of the work i'> to provide some detailed insight into the mechanisms that regulate movement to the left periphery on the empirical basis of an up to now rather poorly investigated German variety, i.e. Cimbrian. We first show that Cimbrian still possesses the V2 property in the sense that the inflected verb moves to the left periphery of the clause. Five empirical arguments are discussed in favor of this hypothesis concerning the existence of a prefield-expletive of the German 'es' type, the order of clitics, of negation, of separable prefixes and of the particle da with respect to the inflected verb. We propose that V3 instances, which apparently violate the V2linear restriction, are to be analyzed as invohing Topics located higher than the C projection where the inflected verb moves. We also show that wh-elements are to be split into two clas.'les, which are moved to two different specifiers in the layered left periphery of the clause. This split is a function of tl1e internal structure of the wh-items, whicl1 can contain or not a lexical restrictor, which is in turn related to two different interpretations (i.e. d.e dicto and d.e re) of the wh-item itself. Keywords: left periphery; Cimbrian; wh-items; verb second; Topics

1.

Introduction

In this work we examine the left periphery of Cimbrian, a German dialect with very peculiar grammatical features spoken in some Veneto and Trentino villages in NorthEastern Italy. Given that Cimbrian is an endangered language, and is already dying out in most of the villages where it used to be spoken, we will restrict our empirical investigation to the variety of Luserna, the only one where Cimbrian is still actively spoken by the majority of the population. 1

We thank Fiorenz.o Nicolussi for his help and patience with the Cimbrian data and we are grateful to Ede Zirrunerrnaim fur ha'fi.ng drawn our attention to the distinction between de re and de dicto readings of wh-elements. Thanks also go to the audience of the IGG conference held in Siena in Februar 2009 for helpful comments. We would also like to thank Gildo Bidese 1.

302

Gunther Grewendorf & Cecilia Poletto Cimbrian is usually assumed to have lost the V2 property, because it allows f{)r V3, V4 etc. orders and it does not display DP-subject inversion. However, we claim that Lusema Cimbrian is still a V2 language, in the sense that the inflected verb moves to the CP domain in main declarative clauses. We will show that Cimbrian maintains some of the typical properties ofV2languages, as subject (clitic) inversion in declara-tive clauses, a first position e.l!:pletive pronoun similar to German es, and the typical asymmetry between main and embedded clauses (although restricted to one subclass of embedded clauses). The sections from 3 to 6 are devoted to a detailed analysis of these properties: in Section 3 we describe the peculiar type of subject inversion Cimbrian has maintained through the centuries; in Section 4 we analyze the complementizer system showing the existence of two series of complementi:l.ers only one of which blocks V to C. In Section 5 and 6 we hypothesize that the two classes of complementizers are located in different positions, and bring arguments to map the e.xact position of each class. Once we have a clear idea of the way the Cimbrian CP is layered, we investigate the properties of wh--items, different types of subjects and 'lbpics and the position to which they are moved in the left periphery. In Section 7 we show that Cimbrian wh-items display a split between two classes: complex wh-phrases are always moved very high in the structure of the CP to a position which we define as ReferentialP (RefP). Simple ·wh-words have two possibilities: they either move to RefP or to a much lower Operator position, the choice between RefP and OpP is related to the interpretation of the wh-word. We claim that the movement of wh-items to the low OpP or the high RefP position is related to the distinction between the de die to and the de re reading of wh--elements, which is associated with the presence of a "Restrictor" in the internal structure of the wh-item itself: wh-elements which have an internal restrictor (and complex wh-phrases always have a lexical one, namely their internal noun) move to RefP, wh-elements which do not have an internal restrictor move to the low OpP. Bare wh-words can be inter-preted as including a null lexical noun or not, therefore they have both possibilities. In Section 8 we deal with Topics and show that the linear V2 restriction typical of Germanic languages is not a direct consequence ofV to C, in the sense that all and only those lanbruages where the verb moves at least to FinP ban two (or more) con-stituents in front of the verb. Some V2 lanbruages do allow V3 (and more), but only if

for pointing out to us interesting phenomena of Cinibrian and helping us find our way through the vallies and mountains on the border between the Romance and Germanic world. Many of our examples are due to our own field work, some others are reported from Panieri et al. (2006). Por the purpose of the Italian academy, Cecilia Poletto is responsible for Sections 1-4.2 and 8, and GUnther Grewendorf is responsible for Sections 4.3.-7.

Hidden verb second: The case of Cim brian 303

the first XP of the V3 sequence is a left dislocation. We observe that these are precisely those languages where clitic pronouns can double left dislocated XPs. This empirical generalization can be explained by adopting a theory of doubling in terms of splitting ofthe DP in two subparts: the ditic pronoun (which encodes Case, the highest projection in the internal structure of the DP) and a TopicP internal to the structure of the DP (we follow here Giusti (2006), who assumes that each DP, being a phase, possesses an internal left periphery similar to the one found in CP). Languages where the DP can be split leaving the ditic behind allow for movement of a pure TopicP to the left periphery of the clause, while languages without clitic doubling must move the whole DP to the left periphery. A possible explanation for this correlation is found at the end of Section 8.

2.

Luserna Cimbrian as a V2language

Before starting with our discussion, we defme the V2 property in technical terms as movement of the inflected verb to a position which in other contexts is occupied by a complementizer. Therefore, not only those languages where the verb reaches Force are V2, but any language in which the verb reaches at least the FinP projection, which we take to be the lowest position where a complementi:t.er can be merged. Our definition ofV2 is therefore rather wide: movement of the inflected verb to any projection in the C domain conforms to it. If this only occurs in a subset of main clauses like interrogative main clauses (as for instance in English, French or Northern Italian dialects) we will define the language as a residual V2 language, following Rizzi's (1991) original proposal. If verb movement occurs also in declarative main clauses, the language will be a full V2language (as all Germanic languages except English). Roberts (2004) proposes that the V2 linear restriction is a consequence of an .EPP feature with which the head of FinP is endowed. Any element moving to the left periphery can satisfy the EP P feature because FinP is not specified tor any Focus, Topic or operator feature and thus can attract any XP. In turn, any XP moving to the left periphery has to pass through the specifier of FinP, and given that there is only one SpecFin, there can only be one XP in front of the inflected verb. This analysis very elegantly ties the V2 linear restriction to subject inversion by means of a sort of"bottle neck" through which any XP must move to reach higher projections in the CP domain. However, as it is, it clearly does not account for those languages like Old Italian, central Rhaetoromance, Mocheno, Sappadino and, as we will see, Cimbrian, where the verb moves to Fin but more than one element can be found to its left. In Section 8 we will come back to this problem and suggest a possible way to reconcile Roberts' analysis with V3 cases data. In this section we concentrate on the peculiarity of the Cimbrian left periphery.

304 Gunther Grewendorf & Cecilia Poletto

Cimbrian has not attrc:tcted much attention in generative studies until recent years. The work available, (see Bidese (2008), Tomaselli and Bidese (2007), and Scardoni (1999)) is mainly based on data from the dialect of Roana, which is nowadays a non-V2 variety. The arguments the afi.)rementioned authors present to show that Roana Cimbrian has completely lost the V2 property are based on the three standard correlates ofV2: (a) the lack of the V2linear restriction, according to which the inflected verb has to be realized as the second constituent of the dau se, (b) the im possibility of any kind of subject inversion in main declarative clauses and (c) the lack of asymmetry with respect to verb movement between main and embedded clauses. The dialect of Luserna is similar to the one of Roana in having lost the linear V2 restriction, since several constituents can freely occur in front of the inflected verb, as the following sentences show: (1)

Gestarn dar pua hatt gisekk in has. a hare yesterday the boy has seen

(2)

Haute die mome hat gekoaft die oala al merca. today the mother has bought the eggs at the market

Notice, however, that this test is not sufficient to exclude the possibility of verb raising to the C domain, as other languages where the V2 linear restriction is relaxed have been analyzed as V2 languages. This is the case of Old Italian, which, in spite of V3 (and V4 etc.) orders (cf. (3)), displays systematic subject inversion in main declara·· tive clauses and an asymmetry between main and embedded clauses with respect to subject inversion, ditic placement and pro drop phenomena (see Beninca (2006) for a detailed discussion on this). We provide some examples ofV3 in (3) and of subject inversion in (4): (3)

a.

Et

dall'

altra parte Aiaces era uno cavaliere franco.

and on the other side

b.

A.

·was a

knight

courageous (BL, Rett., p. 94, r. 7)

E la reina Artemidora di Alicamasso, che in adiuto and the queen A. of Alicarnasso who in help

mescolo di Serses era venuta, francamente si of Serses was c.ome courageou~ly herself mingled.3sG nella battaglia. into the battle (4)

(BG, Or., p. 92, r. 1)

a.

quali denari avea Baldovino lasciati loro. which money had.3sG Baldovino left them

b.

. . .primieramente avea ella fatta a lui ingiuria. (BL, Rett., 116) ... for first had.3sG she done to him injury

(Doc. fior., 437)

Hidden verb second: The case of Cimbrian 305

The same is true of other German dialects spoken in the Italian Alps like Sappadino (Plodarisch) and Mocheno, which also display regular subject inversion, but allow for V3 orders: (the Sappadino data in (5) are reported from Poletto and Tomaselli (2001) and the Mocheno data in (6) are reported from Cognola (2008): (5)

(6)

a.

Hainte tuit de marne de kinder ins pette. Today does the mother the children to bed

b.

bolde s holz gehokt. Geischter der Hons hot me has in--the wood the wood hacked Yesterday the H.

a.

Van Nane ho de Mario klofft. has the M. spoken of.the N.

b.

Gester en de boteig der Mario hot a puach kaft. has a book bought yesterday in the shop the M.

The (a) examples display subject inversion, while the (b) examples show that nonetheless V3 is possible. V3 orders are also possible in Old High German, as noted (among others) by Fuss (2003), (2005), (2008) as the following example illustrates: 2 (7)

a.

(lsid01; 278) Dnthtines uuordu sindum himila d1ifestinode. (Robinson 1996: 36) word are heavens made fast of Gods

b.

Inti ubil man fona ubilemo tresouue bringit ubilu. and bad man from bad treasure brings bad-things

(Tatian)

Therefore, a language which displays V3 is not necessarily a non V2language, but sim-· ply one where several1bpic positions can be filled simultaneously. 'Ihtts, the appear·· ance ofV3 does not prove anything with respect to the movement path of the inflected verb: a language can have I to Fin movement (as attested by subject inversion) \Vithout respecting the V2 linear restriction, as the languages above show. If the loss of the

2.

Fuss distiguishes among six cases ofV3 orders in Old High German:

a

V3 order after dislocated topics V3 order after two adverbial expressions (that in most cases can be taken to form some kind of unit, e.g. [XP [XP] [XP]] Vfin ... ) c. V3 order induced by intervening sentence adverbs d. V3 order induced by intervening personal pronouns e. V3 induced by intervening short adverbs (mostly thcJ 'then') f. V3 after preposed adverbial clauses

b.

We refer to hL~ work for a detailed analysis of the different cases. For the purpose of the present work it is sufficient to note that "iolations of the V2linear restriction exist in Old Germanic.

3o6 Gunther Grewendorf & Cecilia Poletto

V2 linear restriction is not enough to conclude that the verb does not move to C0 in a given language, we can use the second diagnostic for V to C movement, namely subject inversion. Therefore, any analysis which ties subject inversion and the V2 linear restriction as the one proposed by Roberts (2004) encounters a problem with languages like these. The situation of Cimbrian concerning V3 orders has changed with time: Bidese (2008) shows that in the first and second Catechisms (dating respectively 1602 and 1813) the V2linear restriction is respected (the only exception found is the one with pronominal subjects we mention in the next section), while nowadays V3 sequences are rather common, as the following examples show: (8)

a.

merca. Hai.\te die mome hat gekoaft die oala al today the mum has bought the eggs at.the market

b.

Gestam die mome hat gebii.scht di piattn. yesterday the mom has washed to dishes

However, we will not conclude from examples like these that Cimbrian is not a V2 language for the reasons mentioned above: there exist languages which do not respect the V2linear restriction but which nonetheless display subject inversion and movement of the inflected verb to the C domain.

J.

The problem of subject inversion

From the above discussion, we can conclude that the lack of V3 orders is a suf.· ficient but not a necessary condition of V2 languages. Therefore. we need to resort to subject inversion to show whether Lusema Cimbrian really has movement of the inflected verb to the C domain. 'Ihis test is also not without problems since in Luserna Cimbrian the possibility of having subject inversion is related to the type of subject: no DP subject inversion of the Germanic type is allowed as the ungrammaticality of (9a, b) shows: (9)

a. *Haute geat dar Giani vort. today goes the G. away b. "Gestarn hatt dar Giani gisekk in has. seen a hare yesterday has the G. Haute dar Giani geat vort. today the G. goes away

d.

Gestarn. dar Giani hat gisekk in has. has seen a hare yesterday the G.

However, inversion is possible and indeed obligatory when the subject is represented by a subject clitic pronoun as in the following examples:

Hidden verb second: The case ofCimbrian 307 (10)

a.

Gestarn hatt-ar gisekk in has. a hare yesterday has--he seen

b. *Gestarn ar hatt gisekk in has. yesterday he has seen a hare

Bidese (2008) notices that the ban against DP-subject inversion is already attested in the first Cimbrian document, the Catechism dating 1602. He reports the following example attesting inversion with subject pronouns and claims that subject pronoun inversion (illustrated in the following example) is common in the text while inversion of a DP subject is virtually non e.lC:istent: (11)

und darome saint za zbelf artkheln. and therefore are they twelve articles

(Kat. 1602: 247, Bidese (2008: 34))

Working on the same text, Beninca and Renzi (2000) also confirm the absence of DP subject inversion and report a single example ofinversion of a non clitic subject, which is a quantified noun phrase. (12)

unt hia saint iere painighe alle ghegogelt. all cancelled and hier are their sins

(Kat. 1602: 1272-3, Beninca & Renzi (2000: 148))

Contemporary Lusema Cimbrian also displays a special treatment of quantifiers (at least bare ones) as opposed to usual noun phrases: on a par with subject clitics they allow subject inversion, as the following examples show: (13)

a.

'Z

b.

'Z

hat-ta niamat telefonaart. it has-part. nobody phoned hat niamat telefonaart. it has nobody phoned

Moreover, cases of inversion of full pronouns become possible if the full pronoun is doubled by a subject clitic,3 as is the case in constructions which probably have to be analyzed as cases of focused subject pronouns: (14)

a. ""Gestarn hatt er gisekk in has. yesterday has he seen a hare

b.

Gestarn er hat gisekk in has. yesterday he has seen a hare

c.

Gesta:rn hatt-ar er gisekk in has. yesterday has-he he seen a hare

3· An additional afb>urnent in favour of V2 is the fac..t that an embedded clause in first position triggers subject inversion (although limited to !>'Ubject ditics) as it does in standard German and other V2 languages: (i)

Bhlda rivan di khindar, sperr--bar di ttir. when arrive the kids, dose .. we the door

3o8 Gunther Grewendorf & Cecilia Po1etto If real DP inversion is lacking, we could conclude that Cimbrian is not a real V2 language, and never was, as DP inversion was never possible (at least from the first attestation we have). However, this would be wrong: as we noticed above on the basis of Bidese's (2008) work. the first attested document (the Catechism dating 1602) as well as the sec.ond (the Catechism dating 181.3) were consistently V2 with respect to the linear restriction. The only cases of V3 found were due to the fact that the text is a translation fi·om Latin where there are two constituents preceding the inflected verb. Bidese further notes that in the first Catechism, the violation of the V2 linear restriction consists in the addition of a subject pronoun to sentences where the origi·· nal Latin text had none (Latin being a null subject language, lexical subject pronouns are restricted to cases of Focus or Topic). In these cases the position of the subject pronoun inserted in the Cimbr:ian trmslation is located to the left of the inflected verb and not to the right, as exemplified in the following example: (IS)

Im funften iz schaffet. in-the fifth it orders ...

(Kat 1602:529, in Bidese (2008: 42e))

The same type of violation is found in Old High German and Old English texts and has been studied by van Kemenade (1987), Tomaselli (1995), Fuss (2003), (2008) among others. (16)

a.

Dhaz ir chichundida. this he announces

(Isidor Contra Iudaeos)

b.

erino p011Wl ih fircbnussu. iron portals destroy

(Isidor Contra Iudaeos)

This has been interpreted by the authors mentioned above as evidence that those pro·· nouns were either 'VVackernagel eli tic" or weak forms and occupy a dedicated special position in the CP layer.4 1his possibility has completely disappeared from the synta..lC: of Luserna Cimbrian: when a clitic occurs in main clauses starting with any XP, the subject eli tic must fol·· low the inflected verb. We follow Bidese's proposal here and assume that modern Cimbrian clitics are real heads and behave as such, while Older Cimbrian pronouns in V3 contexts are rather weak pronouns and have a dedicated position in the clause (see Cardinaletti and Starke (1999)), which is however a specifier position.

4· Several authors have worked on the problem ofV3 with subject pronoun in Old Germanic languages. The first seminal work is van Kemenade (1987) on Old English, Tomaselli (1995) concentrates on Old High German.

Hidden verb second: The case ofCimbrian 309

Modem Cimbrian also has weak pronouns, and they appear on the left of the inflected verb, contrary to clitic heads: 5 (17)

a.

Dar khint. he (weak pron.) comes

b. ""Haute hat dar

der Giani gisekk. today has he (weak pron.) the G. seen

c. "Ar khint. he (dit.) comes d.

Haute hat-ar der Giani gisekk. today has-he (clit.) the G. seen

With Bidese we conclude that: 1. 2.

The syntax of DP inversion has not changed in Cimbrian, as DP subject inversion was never possible. The syntax of pronouns has changed because clitic heads have been developed.

However, the fact that in Older Cimbrian the V2 linear restriction was respected leads us to assume that these two properties (i.e. subject inversion and V2linear restriction) are not really two sides of the same coin: Cimbrian shows that the linear restriction has been lost, while DP subject inversion has remained impossible. Even more, Older Cimbrian shows that a language can obey the linear V2 restriction without having DP subject inversion. 'Iheref{)re, we should be cautious in interpreting lack of DP subject inversion as evidence for the non V2 character of a language. This is not only confirmed by the fact that in XVII century Cimbrian the linear restriction is respected but no DP subject inversion is allowed, but also by the fact that in Old Italian the linear restriction is not respected, though DP subject inversion is found. This shows that what has standardly been assumed to be a cluster of properties associated with the V2 parameter is not to be considered as such. In a more general vein this is not a surprising result, as other parameters have also been taken apart in recent literature. Work on the pro drop parameter has shmvn that what we used to view as properties associated with this parc:IIlleter can be found separately in different languages and that clusters of properties being present or absent together are rather to be conceived in probabilistic terms than as a real yes/no split.

5· lhe fact that real ditics cannot appear in first position in Cimbrian constitutes independent evidence for its V2 status, as larJ.b>uages like the Northern Italian dialects (which are not V2 and possess subjec..t ditic heads) do not bar1 subject ditics in first position: (i)

El

vien.

He comes

(Venetian)

310

Gunther Grewendorf & Cecilia Poletto

Concerning the reason for the rather astonishing fact that DP inversion is banned from a V2 language, we postpone the discussion of our proposal to Section 7, where we will have a more precise idea of what the left periphery of Cimbrian looks like. More generally, given that the tests to ascertain the V2 status of Cimbrian do not appear to be as clear--cut as we would like them to be, we intend to further investigate the third empirical correlate assumed to be a reflection of the V2 property, namely the asymmetry between main and embedded clauses. As we will see, the system of complementi1.ers found in Luserna Cimbrian is rather interesting and reveals that for one subclass of complementizers, the asymmetry typical ofV2 languages is still opera-tive. This in turn will shed light on the diachronic development of languages losing the V2 property, which we interpret to mean, in syntactic tenns, V to C movement, as standardly assumed.

The complementizer ~-ystem of Cimbrian

4·

The system of Cimbrian complementizers can be split into two classes: we will refer to them as "ke-type complementizers" and "az-type complementizers" using the two complementizers which most frequently occur in embedded declarative clauses. Here is the list of the complementizers belonging to each class reported by Panieri et al. (2006: 338---342): (18)

Ke-type

ke, 'that'; benn, 'when'; bia, 'as'; umbromm 'because'; bia nd, 'why';

Az- type

az, 'if/that'; bal, 'when/if'; benn, 'if'; ir1tdnto az'while'; dnka az 'even if'; dopo az 'after';Jirl az 'until'; ma az6 'unless'; bo 'relative c.omplementizer'.

'Ihe clauses following ke--type complementizers behave as main clauses in various respects, while clauses introduced by a.z-type complementizers display different properties. We illustrate the point with respect to four different properties, which we will discuss in turn. 4.1

Separable prefixes

On a par with other Germanic languages, Cimbrian has a set of separable prefixes. However, they are not unmovable as they are in standard German, but appear in at

6. Notice the combination of a Romance adverb with the Germanic complementizer rather than with the Romance one. Younger speakers tend to use intanto ke, dopo ke,fin ke instead, and this could be the key to the loss of the Germanic complementizer system. However, we will not investigate this phenomenon any further here.

Hidden verb second: The case of Cirn brian 311

least two positions. As already shown in Grewendorf and Poletto (2005), separable prefixes can either precede or follow the past participle in a declarative clause, but always follow the auxiliary or a simple main verb in main clauses: (19)

a.

hon au-gehort die arb at ka Tria. have up-given the job in Trient

b.

hon gehmt--au di arb at ka Tria. have given-up the job in Trient

c. *I auhon gehmt die arb at ka Tria. in Trient up-have given the job (20)

a.

hon o.ffe-geton die ture. have open-done the door

b.

hon geton-offe die ture. have done-open the door

c. *I offe hon get on die ture. open have done the door

Embedded clauses with ke do not differ from main clauses and display the same two possibilities: (21)

a.

die arb at an menta. Dar hat-mar khott ke dar hat ogeheft he has-me told that he has pref.-begun the job on Monday

b.

die arb at an menta. Dar hat-mar khott ke dar hat geheft 0 he has-me told that he has begun pref. the job on M.onday

Notice that in sentences like ( 19)-(21) the prefix can never cross the inflected auxiliary (or any inflected verb), as shown by (22): (22)

hat geheft die arhat an menta. a. *Dar hat-mar khott ke dar o he has--me told that he pref. has bet,'l.m the job on Monday di arbat an menta. heft b. "Dar hat-mar khott ke dar o on Monday job the begins pref. he he has-me told that

This rather interesting oscillation between a pre- and a postparticipial position of the prefix might be interpreted in the following way. Assume that Cimbrian is not different from German with respect to the position of separable prefixes, which encode aspectual features and therefore must be located in some Aspectual projection in Cinque's (1999) hierarchy. The pre- or postparticipial position of the prefix cannot be due to its optional movement in front of the past participle, given that in general verbal prefixes are unmovable, as German clearly shows. Rather, we surmise that the distinction between the two languages is to be attributed to verb movement: Cimbrian is a VO language; therefore, it must be different from German with respect to the movement possibilities of the verb in general and of the past participle too. Thus, we propose that the oscillation found in (19) to (21) is due to movement of the past participle, which

312 Gunther Grewendorf & Cecilia Poletto

can remain lower or rc:rise higher than the prefix. 'Ihe following structure illustrates the two possible orders: (23)

a. b.

[cp.. ·lrr.

[AsoP prefix lvp past pa1tidple DPob!]]]] b···lrr· pa..~t participle [AspP prefix [yp past·part:ieiple DPobJ]]]]

Notice furthennore that embedded clauses of the az-type also display prefixes before or after the participle. In addition to that, they have a third option, which is impossible with ke·· type embedded clauses: the prefix can be located in a position higher than the auxiliary (or the main inflected verb). (24)

(25)

a.

Dopo az--ar hat 0 geheft di arb at an menta. after that--he has pre f. begun the job on Monday

b.

Dopo az-ar hat geheft 0 di arb at an menta. on Monday after that-he has begun pre f. the job

c.

Dopo az--ar 0 hat geheft di arb at an menta. on Monday after that-he pre f. has begun the job

a.

dar mann bo da hat o-geheft a naiiga arbat. job the man that part. has up--taken a new

b.

dar mann bo da hat geheft-o a naiiga arbat. the man that pa1t. has taken-up a new job dar mann bo da 0 hat geheft a naiiga arb at. the man that part. up has taken a new job

The empirical generali7.ation we can state is that az-type clauses have a syntax dif.· ferent from main and ke-·type clauses. Elaborating on this empirical observation, we can assume that in az-type clauses the inflected auxiliary remains in a position lower than the prefix, while in main and ke-type clauses, it always raises higher and crosses the prefix. 'Ihere is independent empirical evidence that this hypothesis is correct: both higher and lower adverbs of the aspectual and modal type (as analyzed by Cinque (1999)) can occur higher than the auxiliary in az-type clauses, thus attesting that the order prefix--auxiliary is due to lack of movement of the auxiliary, not to prefix rc:ris·· ing. 'Ihe f(.)llowing examples show the case in point: while in main clauses the adverb za occurs to the right of the inflected verb, it occurs to its left in az- type embedded clauses. Given that adverbs do not move from their merge position (unless they are focused, which is not the case here), we can conclude that the preverbal position of adverbs which usually occur postverbally shows that in this type of embedded clauses the verb does not move as high as it does in main clauses. (26)

a.

. .. az ar za vorl is gont . ... that he already away is gone

Hidden verb second: The case ofCimbrian 313

(27)

b.

is gont. ... az ar vort za ... that he pref. already is gone

c.

gont. ... az ar vort is za ... that he pref. is already gone

d.

... az

e.

. .. az ar vort furse is gont . ... that he pref. maybe is gone

t

... az

a.

Dar hat za geriiaft. he has already phoned

b.

Dar hat-mar khott ke dar hat za geriiaft. he has-me told that he has already phoned

vort is gont. ar furse ... that he maybe prt. is gone

ar vort is furse gont . . . . that he pref. is maybe gone

We can draw the following tentative conclusion: in az-type clauses, the inflected auxiliary can remain lower than in main and ke- type clauses. In what follows, we present additional tests which confirm this conclusion. 4-2

Negation and verb movement

Another test which is often used to determine the position of the verb in VO languages with the V2 property like the Scandinavian languages is the relative ordering of the inflected verb and the sentential negative marker. In Mainland Scandinavian the verb is usually analyzed as remaining in a lower position (inside the VP), given that the order is Neg- V. while in Icelandic the fact that the order V-Neg is possible is analyzed as raising of the verb higher than the negative marker to some 1° projection. If the tentative conclusion presented above is correct, tb.en we predict that the clauses selected by the twn complementi:l.er types should also differ with respect to the posi-tion of the standard negative marker? 0 nee again, we can observe that ke- type clauses pattern with main clauses; in both cases the negative marker obligatorily follows both main and auxiliary inflected verbs (and always precedes the past participle): (28)

a.

boas ke dar is net vortgont. know that he is not away-gone

b. "I boas ke dar net is vortgont. I know that he not is away--gone

7· Notice incidentally that the negative marker uet in Cim.brian seems etymologically and syntactically similar to the German 'nkht' type, and not to the higher one used in Italian, as it does not trigger negative concord. We assume here that it occupies the same position as German 11icht.

314 Gunther Grewendorf & Cecilia Poletto

c. *I boas ke du net geast ka Tria. to Trient I know that you not go

(29)

d.

Dar khiit ke dar steat net dahuam. he says that he stays not at-home

a.

Dar is net khent. he is not come

b. *Dar net is khent. he

not is come

With az-type complementizers the situation is different and more complex: in the case of main verbs, negation must precede the inflected verb: (30)

a.

Dar hat gebollt azz-e net vortgea. he has wanted that-I not away-go

b. *Dar hat gebOllt azz-e vortgea net. he has wanted that-1 away-go not The constrast between (28)/(29) and (30) clearly shows that main verbs in az-type clauses cannot raise to cross negation, while main verbs in main and ke-type clauses must do so. (31)

hebat geboellt az-ar-me net oruaf, ma dar hat-s getont. had wanted that-he-me not phones, but he has-it done

a.

oruaf net, ma dar hat-s getont. b. *I hebat geboellt az-ar-me had wanted that-he-me phones not, but he has-it done An interesting difference is found as far as auxiliary and modal verbs are concerned; in this case negation can either occur before or after the auxiliary: (32)

(33)

(34)

a.

... azz-a-dar net hat khOtt zu kemma. . .. that-he-to-you not has said to come

b.

hat net khott zu kemma. ... azz--a-dar ... that-he-to-you has not said to come

a.

Onka az-ar hat net ogeheft a naii.ga arhat, issar herta toebig. job, is-he always nervous even if-he has not begun a new

b.

Onka az-ar net hat ogeheft a naii.ga arhat, issar herta toebig. job, is-he always nervous even if-he not has begun a new

a.

Bal dar net bill

if b.

he

gian, schikh-en vort. not wants go, send-him away

Bal dar bill net

gian, schikh-en vort.

In this case we propose that auxiliaries and modals can but need not raise higher than the position of the negative marker. This difference between auxiliaries and main verbs is well known in the literature on verb raising: already Pollock (1989) notes the same

Hidden verb second: The case of Cirn brian 315

difference between infinitival auxiliaries, which can (but need not) raise higher than negation in French, and main verbs, which cannot move past the negative marker pas. vVe can conclude that the second test also goes in the same direction as the first one: in az-type clauses the inflected verb seems to be located lower than in ke-lype clauses and main clauses, where the verb must move past the negative marker net. 4·3

The particle da

Another test showing that we are on the right track in assuming that in az- type clauses the verb does not raise as high as in main and embedded clauses introduced by ke has to do with the position of the particle da. 8 In main clauses the particle is ahvays located after the inflected verb, as shown by the following example: (35)

a.

gelaii.tet die klokkng. han--da Aile sunta the bells every Sunday have-part. rung

8. This particle is homophonous •Nith the locative element da 'there; though the fact that the two can cooccur shows that they are not the same item. We will not investigate the semantic imp01t of the particle here, leaving it to future research. Here we limit ourselves to providing some information on its distribution. Da is a patticle occurring in relative clauses (on the subject, object and other arguments)

(i)

hat geschenkt in Gianni. Dar libar bo da-r to G. the book that part.--he has given

interrogative clauses (ii)

da-r hat gisekk in pua. boas net bo the boy I know not where part.-he has seen

declarative clauses (iii)

di milch di bake. 'Z gen-da it give-part. the milk the peasants

The pattide da is not a locative: as it can cooccur with a locative instance of da (iv)

hat gelek. Dar libar bo da der Giani da there has put the book that part. the G.

Da senres as a host of ditics (v)

putzt ist kronk. bo da-se S beibe the woman that patt...them cleans is sick

Da is incompatible \-\'i.th weak pronouns, but co-occurs both with ditic and tonic pronouns (vi)

a.

herta lesst worma geat in pett. Dar libar boma the book that-one always reads when--one goes to bed

rna herta lesst v\'orma geat in pett. b. *Dar libar bo da the book that part. one always reads when-one goes to bed

316

Gunther Grewendorf & Cecilia Poletto

b.

Alle sunta latitn-da die klokkng. every Sunday ring-part. the bells

c. *Alle sunta da lai.itn die klokkng. every Sunday part. ring the bells

If the complementizer is of the ke-type, the particle da is again located immediately after the inflected verb, as shown by the following examples: (36)

Dar Mario hatt khOtt ke aile sunta han-da gela(itet has said that every Sunday have-part. rung the M. die klokkng. the bell-.

(37) *I boas

ke da khint di nona. know that part. comes the granny

In the case of az-type complementizers (like bal., in the example below) the particle is located immediately after the complementizer itself: (38)

Bal-da rivan di khindar, sperr-bar di ttir. close-we the door when-part. arrive the kid<;,

Again, the empirical generalization we can state groups main and embedded ke--lype clauses together, setting az-type clauses apart: da is located after the inflected verb in main clauses and in clauses introduced by ke but immediately after the complementizer in az--type clauses. We can interpret this fact along the lines suggested above: the position of verbal prefixes, the position of the negative marker and the position of the particle da consistently show that the verb raises higher in ke-clauses than in az-clauses. 1he test on da takes us even further in the interpretation of the data: the fact that da is enclitic onto the verb in main clauses and embedded ke--type clauses and enclitic onto the complementizer in az-type clauses suggests that the position occupied by the inflected verb in main and embedded ke-type clauses is the same as the one occupied by the complementizer in az-type clauses: in other words, verb second occurs in main clauses and in a subset of embedded clauses.

4·4 Clitic pronouns Another test showing the complementary distribution of intlected verbs and az-lype complementizers is the position of object clitics: with ke-- type complementizers object clitics must be in enclisis to the inflected verb, with az-type complementizers object clitics occur in enclisis to the complementizer itself (or to the particle da when it is present creating a duster).

Hidden verb second: The case ofClmbrian 317 (39)

a.

Da so in vortgont en a az-ta-s niamat barn. they are away--gone before that--part.--it nobody noticed

ena az niamat ba:rn-da-z. b. *Da so in vortgont they are away-gone before that nobody noticed-part.- it c.

(40)

I gloabe ke dar gebat-mar-s. I think that he gives-to.me-it

a. *Dar hat-mar khOtt ke dar en sich morng. he has-to.me said that he him sees tomorrow

b.

Dar hat-mar khott ke dar sich-en morng. he has-to.me said that he sees-him tomorrow

The tentative conclusion we reach on the basis of the contrast in ( 39) and ( 40) is that the finite verb in ke--clauses oc.cupies the same position as the complementi7.er a.z. & discussed in Section 2, Cimbrian also has subject clitics, which are obligatorily in enclisis to the verb in main clauses. They never occur in first position in V2 clauses, where either tonic or weak pronouns are used: (41)

Er/Dar khint. he comes

(42) *Ar khint. he comes

& expected by the V2 pattern, subject ditics occur in enclisis to a.z-type complementi7.ers, confirming the idea that the verb in main clauses occupies the same position occupied by the complementizer in az-type clauses: (43)

'Z tiiat mar ont azz--ar sai za vorgont. It does me sorrow that--he is already away--gone

Given that we analyze ke-type clauses as embedding a whole main clause structure after ke--, we expect that no subject clitics are possible immediately after ke-Lype com-plementizers, since this position corresponds to the prefield position in V2 clauses. This prediction is borne out: no subject clitics are found after ke; tonic or weak pronouns are used instead like in main clauses: (44) *I boas ke ar khint. know that he comes The last empirical generalization we formulate is the following: the first position to the right of ke in embedded clauses and the first position in main clauses cannot host clitics, the first position after az can host clitics.

318

Gunther Grewendorf & Cecilia Poletto 5·

Verb and complementizer movement

Summing up what we have discussed so far, we can state that elements like da, negation, separable prefixes and object ditics occur after the inflected verb in main clauses and Ice-type clauses, while they occur before the verb in az-lype clauses. We have suggested that this is a reflex of the well-known asymmetry between main and embedded clauses in V2languages of the German type: if the verb raises to the C domain in main and ke-type clauses only but not in az-type clauses, then we expect it to cross: (a) separable prefixes, (b) negation, (c) object clitics, (d) the particle da. Therefore, we assume that in az-type clauses the complementizer is located in the same position where the verb ends up in main and ke-type clauses. However, if we adopt the by now standard idea of a split--CP, 9 this is not enough and we have to determine precisely the C0 position target of verb movement i.e. the position of az. There are two plausible positions where az/the inflected verb can surface: FinP and ForceP. This gives rise to the three possible analyses for the surface structure illustrated below: (45)

[ForceP ke hoptcP ·· · [FocusP [FinP az/V [IP •• • lwackP da/clitics] ·· · [AspP prefix [NegP net] ... lvp]]]]]]]

(46)

lsubordP ke h'orceP az/V hopicP •• • [wackP clitics] [GroundP ••• [JI.spP prefix [NegP tte~) ... [vp]])]])]]])

(47)

Is,

·rdP

ke [ForceP az!V hopicP ···l:rocusP

[FinP

da hopic: ·· · h>ocusP

[FinP

hr

lrp lwackP da/clitics) [AspP prefi.'l:

[NegP net] ... [yp]]]]]]]]

Az--type complementizers can either be high or low in the structure ofthe CP: if they are low complementizers located in Fin°, this means that clitics of the Wackernagel type and the particle da, which occur in enclisis to the complementizeriinflected verb, must be in some IP position, as shown in (45). According to this analysis, complemen·ti1.ers of the ke-type are located in ForceP. The alternative is that az/the inflected verb are in Force: in this case Wackernagel clitics and the particle da could be located either in IP (as in (47)) or in CP (as in (46)). If az occupies the Force position then complementizers of the ke-type are located in a projection even higher than Force, which we call here SubordinatorP. 10 Notice that the two alternatives make distinct predictions concerning the position of Topics and Foci with respect to the complementizer: if az is a low complementizer,

9· We assume here Rizzi's (1997) original structure with the modification proposed in Benind. and Poletto (2004). to. See Bhatt and Yoon (1991) on the distinction between complementi.:1.ers that act as mood-indicators and complementizers that act as pure subordinators.

Hidden verb second: The case ofClmbrian 319

'Ibpics and Foci should precede it; if az is a high complementizer, it is expected to be followed by Topics and Foci. The following examples show that az-type complementizers such as bo are high complementi1..ers, given that Topics and focused elements occupy a position lower than these complementizers: (48)

Dar libar bo da i in Gi1mi za on get. to-the-G. the book that part. I already have given

(49)

a.

Dar libar bo da-r IN GIANNI hat get. the book that part.--he THE G. has given

GIANNI dar hat get. b. *Dar libar bo IN the book that THE G. part.-he has given

Although (46)/(47) seem prima facie to be more complex than structure (45), as the additional projection SubordinatorP must be postulated, the order with respect to 'lbpics and Foci shows that it is the correct one. 'Iheretore, we exclude (45) on the basis of the examples above. Furthermore, we can also exclude (47) on the basis of the following argument If da and clitics were located in IP, then we would predict that some specifiers can intervene between the complementizer and the clitic cluster. (50)

az-ta-r-en that-part.--he--him

(51)

a. *Fin az--o--ar net riiaft. until that-perfl1e not phones b. *Fin az net at• otiiaft:. until that not he perf.phones ar otiiaft. c. *Fin az furse until that maybe he perf.phones

'Ihe examples above show that this is never the case, as complementizers and clitics always form a single unit: no prefix, adverb or negation can intervene between a.z and a subject clitic. There are also phonological phenomena of assimilation between the complementizer and the particle da: for instance az + da = azta (z is pronounced as a voiceless sibilant /s/ and the voiced consonant of the particle becomes voiceless as wel1). 11 Moreover, if we adopt an antisymmetric framework in which right adjunction is not allowed (see Kayne 1994), we cannot obtain the order az-da-subject clitic-object clitics through clitici7.ation.

n. An additional indication comes from the fact that native speakers write the sequence complementizer-da-clitics as one single word.

320

Gunther Grewendorf & Cecilia Poletto If we adopt structure (46) we solve both problems: dct and Wackemagel clitics are in the CP domain and az moves from Fin° to Force0 crossing the positions of da (here represented as GroundP) and adjoins to the left of the clitics in WackP creating a cluster which cannot be split by any specifier: 12 (52)

lsubordP ke lForceP

az-da-ar hopicP az-dfil-aT' lwacJr.P clitics arv-dtl-al' lGroundP ardfil·

L_______________________________L____________________________________________________________L__________________________________

hoptcP az..[FocusP a,z [FinP az (rp[AspP

prefix lNegP net] ... [yp ]]]]]]]]]]]

i

In this way, we capture the tact that Topics and Foci are lower than the complemen·ti:z.er, and the fact that the clitic cluster is enclitic to az. Moreover, we also capture the fact that the class of az-complementizers does not allow for verb movement (as the order with negation and adverbs considered above shows), which in principle would be possible if the Fin position were empty. Lack of verb movement only in these constructions shows that the complementi1...er is not directly merged in Force but must start out in Fin and then move to Force. There is independent evidence for complementizer movement in the left clausal periphery. Watanabe (1993) assumes CP--recursion in order to account tor the fact that embedded topicalization in English targets a position to the right of the complemen·tizer in non-wh-clauses and to the right of the fronted wh-element in wh-clauses: (53)

a. b.

. .. [cp leo that b Topic [co 0 [AgrsP ••• ]]]]] . .. [cp wh-phrase leo 0 b Topic leo 0 [AgrsP ••• ]])]]

Furthermore, Watanabe makes two crucial assumptions about the selection of clause types: From the two types of clauses to be selected by a verb, namely wh-clauses and non-·wh--dauses, the former are characterized by the presence of a wh-·phrase in Spec of the topmost CP, while the latter are charc:tcteri:z.ed by an empty Spec of the topmost CP. The argument for complementizer movement derives from Watanabe's analysis of factive complements which he develops on the basis of these assumptions. He argues

Empirical evidence for the existence of a left--peripheral Wackernagel position can be derived from an observation by Hubert Haider (see Haider 2009) according to which there is a garden path effect with the scrambled noun Marga in (i) but not with the pronoun in (ii), which may be attributed to the fact that there exists a left --peripheral syntactic position whid1 is exclusively designed for pronouns: 12.

(i)

well Marga Kollegen vorgestellt bekamen. since Marga colleagues introduced got

(ii)

well es Kollegen vorgestellt bekamen. since it colleagues introduced got

Hidden verb second: The case ofCimbrian 321

that factive complements also require CP-recursion with the Spec of the lower CP being occupied by a factive operator: (54)

John regrets

b

that bOp [AgrS he fired Mary]]]

While this analysis allows him to accow1t for the impossibility of adjunct extraction out of factive complements as well as for the impossibility of topicalization within factive complements, it conflicts with the idea that the definiteness of factive complements as represented by the factive operator is selected by the matrix verb, which would require a structure like (55). (55)

b

Op [co that

[Ags.P ••• ]]]

However, (55) is not compatible with the assumption that a non-wh-complement is characterized by an empty Spec of the topmost CP. To solve the conflicting demands from the theory of selection and the theory of clause types, Watanabe assumes that CP recursion is created by complemerrtizer movement, as indicated in (56):

1he assumption of complementizer movement also provides him with an account of the COMP-- trace effect observed with embedded topicalization: (57) "Which athletes1 do you think that [pictures of til are on sale?

If one assumes that the trace of the complementizer that inherits the inertness of tha.t for government, then the CP recursion analysis can explain the ungrammaticality of (57) along the lines of Rizzi's (1990) analysis of the that-t effect. (58) ?*Whid1 athletes do you think lcp that 1 lcp [pictures of]i ti [Agrs.P ti are on sale]]]

In (58) it is the trace t; of the subject that is the offending trace since it cannot be governed by the trace\ of the complementi:r.er, which would also account fi.)r the contrast between the ill--formed example (57) and the significantly better example (59): (59) ??VVhich athletes1 do you think that [pictures oft 1] Mary bought?

Empirical evidence that there exist indeed nvo complementizer positions located on the left and on the right of Topics comes fi·om Paoli (2003), who provides examples with two complementizers and a Topic in Old Romance: (60)

a.

Sire, je te adjure par le vray Dieu que ta Lord I you beg for the true God that your fille Tarsienne, que tu ne la donnes that you not her give daughter T. a mariage a autre que a moy. to marriage to anybody that to me

(apoll f48b, Paoli (2003): 261)

32.2 Gunther Grewendorf & Cecilia Poletto b.

Onde dize Josepo que en casa de so padre que le Where says ]. that in house of his father that him llamaron primeramiente Ciro. called for-first C.

(GenEst 177r.2)

A theory-internal argument for complementizer movement is offered in Poletto (2000: Section 2.2.3). Poletto shows that in many Northern Italian diale<..ts, invariable and deic·· tic subject elitics are generated in the left clausal periphery and cluster enclitically with the complementizer. If this kind of clustering is analyzed as adjunction and if a Kaynean (1994) approach to adjunction is adopted, then the only way to analyze the complementizer duster is to assume that the subject clitic merges in a position higher than the complementizer and that the complementi.7..er moves in order to adjoin to the subject clitic. Paletta (2000) also notices the presence of low complementizers in the Northern Italian dialects, which occur to the right of subjects in modal non realis contexts: (61)

a.

Gnun c a s'bogia. Nobody that he moves 'Nobody move!'

b.

A venga che gnung cH a fasa borde!. It needs that nobody that he makes noise 'It is necessary that nobody make noise:

Turin (Piedmontese)

Moreover, in Northern Italian dialects, wh--items occur to the right of Topics, which occur to the right of declarative complementizers (on a par with standard Italian as originally noted by Rizzi (1997)). However, in some dialects a complementizer occurs to the right of Topics and wh-items, showing that there is indeed a low complemen·· ti7..er position (as noted by Rizzi (1997)) and that the declarative complementizer can occupy it: (62)

No so Gianni a chi che el ghe parla. to whom that he him talks Not know G. 'I do not know to whom John is talking:

This goes hand in hand with Belletti's (2008) findings on cleft clauses (see below). In his account of the particle system of Welsh, Roberts (2004) shows that at the relevant level of abstraction, the system of Welsh particles can be analyzed as an instance of the same parameter value that is operative in the case of full verb second in the Germanic languages. In both languages, Fin must have a lexical realization, which is achieved by merger of particles in Welsh and by verb movement in the Germanic V2 languages (since the latt.er do not have particles that can be inserted into root Fin). In embedded clauses of the Germanic languages, the need to lexicalize Fin can be satisfied by generating a complementizer in Fin, which blocks verb movement in embedded clauses as a general case of Merge preempting Move. Since selected Force

Hidden verb second: The case of Cim brian 323

requires PF-realization, typical complementizers like English that undergo raising from Fin to Force. Roberts concludes that all the Germanic languages, including English, have Fin-to-Force movement in clauses where Force is selected (i.e. in all embedded dauses). 13 Proceeding from the assumption that the overt complementizer that expresses both finiteness and (declarative) Force, Rizzi and Shlonsky (2007) suggest that the complementizer that is first merged in Fin, to express finiteness, and then moves to Force to check the Force feature. 'Ihe assumption that that is generated in Fin provides them with a new account for the that-- trace effect in English. Since movement of the wh-subject cannot proceed through the criteria! subject position (SubjP) 14 because a phrase meeting a criterion is ti·ozen in place ("Criteria! Freezing'), Rizzi and Shlonsky assume that the Subject Criterion can also be met by a head-- head relation between the head ofSubjP and a Fin head endowed with an expletive-like element (comparable to "-i" in the French que-qui alternation). Since the complementizer that in Fin cannot simultaneously function as the head of the declarative (i.e. as a clausal argument) and as an expletive--like surrogate subject that formally meets the Subject Criterion, the wh-subject cannot be extracted across the complementi:~.er that in Fin. In her analysis of contrastive clefts, Belletti (2008) assumes that the complement of the copula is a truncated CP which lacks the Force projection (due to the absence of an independent declarative Force). She then shows that the defted constituent occupies the Focus position of the truncated Cl', as shown in (63 ): (63)

E' [cp fOTc-e ... [FocP Maria; [che [TP Gianni abbracciava t;JJ]] Gianni kissed (it)-is Maria that

If this analysis is on the right track, it implies that the complementizer che is generated in a position lower than Force. Consequently, Belletti assumes that a complementizer always originates in a low position in the left clausal periphery (Fin) and then receives its (declarative) Force by moving into the head of Force. Belletti points out that under the copy theory of movement. multiple appearances of overt complementizers, as e.g. found in languages such as Turinese and Ligurian (cf: Paoli 2003), can be interpreted as pronounciation of several copies of the complementizer chain.

To account for the presence of that in CP-recursion contexts, Roberts as.<>umes that the verb is in Foe such that the complementizer cannot have raised from Fin to Force. He therefore suggests that the complementizer is merged in Porce in the complements of bridge verbs (again to satisfy the requirement that embedded Force has a PP-realization and to represent the fact that these embedded clauses have assertive illocutionary force).

13.

14.

Rizzi and Shlonsky (2007) assume that the SubjP is located between AgrsP and FinP.

324 Gunther Grewendorf & Cecilia Poletto

6.

Subject clitic inversion in embedded contexts

Additional independent evidence that ke is a subordinator base-generated higher than ForceP and that a.z reaches Force0 by movement is provided by the distribution of the expletive pronoun 'z, which has the typical properties of CP expletives (it behaves like the German "Vorfeld-es"). In main clauses 'z occurs in first position where no other element is fow1d to the left of the inflected verb. If any XP is located in front of the inflected verb, 'z disappears. 15 (64)

a.

'Z han--da gelailiet die klokkng alle sunta. it have-part. rung the bells every Sunday

b.

Aile sunta laiitn-da die klokkng. every Sunday ring--part. the bells

The most plausible analysis of 'z is that it is located in SpecForce: given that Cimbrian does not display the V2linear restriction (see Section2), we can only account for the fact that expletive targets the first position of the clause by assuming that it is located in the highest specifier, namely Spec Force. If we assumed that it is located in SpecFin, then Focus and Topics could precede it, which is not the case. The fact that expletive 'z can occur in embedded clauses introduced by ketype complementizers, but not by az--type complementizers shows that ke is higher than Force:

z

(65)

a.

Dar Mario hatt khott ke 'z han-da gelailiet die the M. has said that 'z have-part. rung the klokkng alle sunta. bells every Sunday

b. *Dar Mario hatt gebOllt az 'z han-da gelaiitet die the M. has wanted that 'z have--part. rung the klokkng alle sunta. bell~ every Stmday

(65) illustrates that the expletive pronoun 'z can only occur with ke--type clauses but not ~ith az- type clauses. On this basis we adopt the follo~ing structure: (66)

[SuborclP [Suborcl 0 ke] [ForceP 'z

[Tot~cP ••• [FocusP [FinP]]]]]

15. Notice that Cinibrian has Romance "free" s-ubject inversion and contexts: (i)

'Z hat-ta gerliaft die momma. it has-part. phoned the mum

~

occw·s also in these

Hidden verb second: The case ofCimbrian 315 1his explains why ke-type clauses and main clauses behave exactly the same: in both cases there is no complementizer blocking the CP layer, and the inflected verb can raise to Fin and then up to Force (in which case we have a construction with Vorfeld 'z or V2), or raise only to a lower projection in the CP yielding V3 by allowing preverbal 1bpics or a preverbal subject. But even in the case ofV3, clitic objects and the particle da will still attach enditically to the verb rather than to the complementizer ke. (67)

a.

Dar hat-mar kh6tt ke dar sich-en morng. has-to.me said that he sees-him tomorrow

he

b. "Dar hat--mar he (68)

a.

khOtt ke--en dar sich morng. has--to.me said that--him he sees tomorrow

Dar Mario hatt khott ke aile sunta han--da the M. has said that every Sunday have--part. gelai.\tet die klokkng. rung the bells

b. ""Dar Mario hatt khOtt ke-da the

M.

has said

alle sunta han that-part. every Sunday have

gelalltet die klokkng. rung the bells Our analysis of the two types of complementizers implies an interesting prediction con-cerning the appearance of subject clitic inversion. Since the finite verb cannot reach a left -peripheral head position if az is generated in Fin, we can predict that subject clitic inversion should not be possible in embedded clauses introduced by complementizers of the az-type but should occur in embedded clauses with ke-type complementizers, even in the presence of a preverbal Topic. As the examples in (69) and (70) show, this prediction is in fact borne out: (69) "I boas

az gestarn hatt --ar--z-- en gett. know that yesterday has--he--it--him given

(70)

boas ke gestarn hatt --ar--marL gett. know that yesterday has--he--me--it given

We are thus provided with further evidence tor the claim that verb second structures can be present despite a violation of the linear V2 restriction. We conclude that complementizers of the az- type reach the Force projection, but tl1ey must have been merged lower (in Fin°) in order to gather up clitics and the particle da and block verb movement to any position higher than its usual IP position. 'Iheretore, the whole CP layer is unavailable to verb movement: neither Force nor Fin nor any intennediate Topic or Focus head can be targeted by the verb.

326 Gunther Grewendorf & Cecilia Poletto 7·

Verb movement and the position of wh-elements

It has been argued that fronted wh-elements target different positions in main and embedded interrogatives. Pesetsky (1989) claims that in English, wh-·movement tar·· gets SpeclP in matrix questions 16 while the wh-element moves to SpecCP in embedded clauses. In order to establish this claim, he adduces topicalization evidence according to which Topics land to the left of a moved wh-phrase in matrix questions (71) and to the right of the moved wh-phrase in embedded questions (72): (71)

a. b.

?'Ihis book, to whom should we give? "To whom this book should Bill give?

(72)

a. b.

?I wonder to whom this book we should give? ""I wonder this book to whom we should give?

In a similar vein, Rizzi (2001) points out that Italian shows an asymmetry between matrix and embedded questions w.r.t the licensing of Focus. While wh-·elements in matrix questions occupy SpecFocP and are thus correctly predicted to be incompatible with a fronted Focus, wh-elements in embedded questions can co-occur with a fronted Focus: (73)

QUESTO hanno altro)? detto (non qual cos' a. *A chi they have said (not something else) to whom THIS detto (non qualcos' altro)? b. *QUESTO a chi han no THIS to whom they have said (not something else)

(74)

a.

Mi domando A GIANNI che cosa abbiano detto wonder what they-have said I TO GIANNI (non a Piero). (not to Piero)

b. "?Mi do man do che cosa A GIANNI abbiano detto wonder what TO GIANNI they-have said (non a Piero). (not to Piero) Rizzi concludes from the contmst between (73) and (74) that the position occupied by wh-elements in embedded questions like (74a) is lower in the left periphery of the

Pesetsky assumes that the subject stays in a low position in this case and receives Case from the auxiliary do. He considers the need to assign Case to the low subject as the trigger for do- insettion.

16.

Hidden verb second: The case of Cimbrian 327

clause than FocP while the position of the wh-element perche, 17 which can be followed (but not preceded) by a Focus and, unlike other wh-elements, does not require inversion, is taken to be in a position IntP, which is located higher than FocP. Poletto (2000) presents e\-idence from Northern Italian dialects which shows that there are up to four different left-peripheral target positions for wft.-movement depending on the type of the wh-elements, their interaction with a specific class of clitics, and their interpretation. As tar as embedded wh--questions are concerned, the situation in Cimbrian is somewhat reminiscent of the situation usually found in standard French and of what Poletto has observed in Nortem Italian dialects. While matrix wh-questions require subject clitic inversion as shown in (75), the examples in (76) seem to suggest that embedded wh-questions are characteri1..ed by the absence of subject clitic inversion: (75)

a.

gimacht? Baz han-sa what have-they done

b. *Baz (76)

a.

sa hi\n

gimacht?

boaz net baz sa Mn gimacht. know not what they have done

b. *I boaz

net baz

hdn sa

gimacht.

This is an empirical generalization which also applies to Old Romance: in Old Italian and 0 ld French there is a sharp asymmetry between main and embedded clauses with respect to the possibility of subject clitic inversion. However, the situation in Cimbrian embedded wh--questions is more complex. As the contrast in the embedded interrogatives of (77) shows, there are wh-elements which occupy a position higher than the wh-element in (76a) since they appear in front of the CP-expletive 'z and require enclitic attachment of the particle da to the finite verb: 18 (77)

a.

'z soinda gestant dahiiarn laili boaz net biavl stayed at home know not how many people it are in reng. peng because of the rain

b. *I boaz net biavl laut da soin gest1int dahiiam peng in reng.

17.

'Ihe same assumption is made for the interrogative complementizer se ('if').

18. Notice that this has nothing to do with a different interpretation of the wh-iten1, as the context in which these sentences have been elicited has been kept constant, as the presence of the same matrix verb confirms.

32.8 Gunther Grewendorf & Cecilia Poletto

As our familiar test with negation shows, the high position of the wh-element correlates with the high position of the finite verb: (78)

a.

Dar hat mar gevorst biavl geld za gem net he has me asked how much money they give not bodrum dar bonka. back the bank 'He has asked me how much money they will not refund to the bank:

b. *Dar hat mar gevorst biavl geld za net gem bodrum dar bonka. The question that arises at this point is what the structural difference between embedded wh-questions such as (76a) and the embedded wh-questions in (77) and (78) depends on. 'Ihe obvious difference between (76) and (77) is that the embedded question in (76) is introduced by a bare wh-element while the embedded questions in (77) and (78) are introduced by a complex wh-phrase with a lexical restriction. This difference suggests a generalization like (79): (79)

Generalization (preliminary) on the target position of wh-elements in embedded interrogatives Wh--phrases v.ith a lexical restriction obligatorily occupy a higher position in the left periphery than bare wh--elements.

However, that this generalization cannot be correct. can be seen from the examples in (80), where it is a bare wh-element that occupies a high position: (80)

a.

Bar bizzan, bem du redest net il. we know whom you address not pref.

b.

Bar bizzan net, benn 'z khemmen-da di khindar vo we know not when it come--part. the d1ildren from schual hai.\t. school today

Even one and the same bare wh-element may have the option of either occupying a high position or a low position, as the following examples show on the basis of the tests related to the CP-expletive and the position of negation: (81)

a.

bas 'z hat-da gelaiitet. Dar hat mar gevorst zega he has me asked (to-see) what it has-part. rung

b.

hat gelaiitet. Dar hat mar gevorst 7..ega bas da he has me asked (to-see) what part. has rung net. Dar hat mar gevorst :r.ega bas gevallt-en he has me asked (to-see) what pleases-him not

d.

bas--en net gevallt. Dar hat mar gevorst zega he has me asked (to-see) what-him not pleases

Hidden verb second: The case of Cirn brian 329 (82)

a.

ber 'z khint-da. Dar hat mar gevorst zega he has me asked (to-see) who it comes-part.

b.

khint. ber da Dar hat mar gevorst zega he has me asked (to-see) who part. comes

c.

ber da vorsmar zega oschauget net di diarnen. not the girls wonder (to-see) who part. watches

d.

net oschauget die diarnen. ber da vorsmar zega the girls wonder (to-see) who part. not watches

These observations suggest that (79) has to be revised along the lines of (83): (83)

Generalization (revised) on the target position of wh-elements in embedded interrogatives a. b.

Wh-phmses with a lexical restriction obligatorily occupy a high position in the left periphery. Bare wh--elements may either occupy a high position or a low position in the left periphery.

At this point, two crucial questions arise: (a) v\lhat exactly are the target positions of high and low wh-movement? (b) Which property of a bare wh-element determines whether it lands in a high position or in a low position? As for the target position of high wh--movement, our assumption that the CPexpletive 'z occupies SpecForceP implies that complex wh--phrases move to a position higher than ForceP. 1he exact fi.mctional status of this position will emerge once we have found an answer to the second question. Low wh-movement may either target the specifier position of FocP or a position lower than FocP. Rizzi (2001) assumes that the position where wh-elements move in embedded clauses is a whP located below FocP. Since we have not tested yet if Cimbrian allows co-occurrence of Focus and fronted wh-elements in embedded clauses we will leave open the question which of these options has to be assumed for Cimbrian and call the low landing site of whelements provisionally ( OperatorP). Although tl1e contrast in (76) and (77) may be taken to suggest that it is the morphological complexity of wh-elements that is responsible for the obligatoriness of high wh--movement, it should be clear from the examples (80)--(82) that morphological complexity cannot be the triggering factor for high wh-movement. The conclusion to be drawn from these examples is rather that the specific factor that licenses high movement of bare wh-elements is generally present with complex wh-elements. For that reason, the crucial factor cannot be 0--linking since there are complex wh--phrases which are clearly non-D·-linked but nevertheless have to undergo high wh-movement in Cimbrian. 1his can be seen from the behavior of the aggressively

330 Gunther Grewendorf & Cecilia Poletto

non-D-linked "what the hell" phrases in (84) and (85) (using the position of negation and clitics as a test): (84)

a.

Dar hat mar gevorst 7.-ega che cavolo du isst net. he ha<> me asked (to see) what the hell you eat not

b. "Dar hat mar gevorst zega (85)

a.

che

cavolo

du

net isst.

ber cavolo hat-dar Dar hat mar gevorst zega he has me asked (to see) who the hell has--you get an punk. given a punch

b. "Dar hat mar gevorst zega ber cavolo dar hat get an punk.

Moreover, there are bare wh-elements which are clearly D-linked in a specific context but still have the option of undergoing low wh-movement. This can be seen from the examples in (86) where the wh-pronoun still has two options despite the checklist context that turns it into a D-linked element. (86)

Question: He has asked me who will come, Piero or Giovanni? a.

her da khint, dar Piero Dar hat mar gevorst zega he has me asked (to-see) who pa1t. wmes, the P. odar dar Giovanni. or the G.

b.

Dar hat mar gevorst zega ber 'z khint-da, dar Piero odar dar Giovanni.

An interesting account for similar wh--contrasts in Bellunese (Veneto) can be found in Poletto (2000: Section 3.3.5). Munaro (1995, 1997) has observed that in Tignes d'Alpago (Northern Veneto) complex wh-elements need to undergo overt fronting while wh--pronouns have the option of being fronted or staying in situ: (87)

a.

Quanti ledest? libri a-tu How many books have-you read

b. "A-tu ledest quanti libri? have-you read how many books (88)

(89)

a.

Qual a-tu compn\? which have-you bought

b.

A-tu compra qual?

a.

Quando parte-tu? leave-you when

b.

Parte-tu quando?

Hidden verb second: The case of Cirn brian 331

While the wh-pronouns corresponding to where and how have a preference for the in-situ option, the in-situ position is obligatory for the wh-pronouns corresponding to what and who/whom: (90)

fat? a. *Che a-tu what have-you done

b. (91)

A-tu fat

che?

a. *Chi laore-lo? who works-he 'Who is working?'

b.

E-lo chi che laora? 19

While Munaro (1997) attributes the split between wh-element~ to their internal structure, Poletto (2000) suggests that the split among wh-words might have to do with the difference between arguments and adjuncts. However, the fact that complex wh-elements may act as arguments as well as adjuncts and that bare wh-elements likewise cannot be associated with selectional properties in a uniform way, casts doubt on an account in terms of argumental status. In what follows we will briefly outline in what direction we think a promising approach to the Cimbrian wh-split might be found. 20 The basic idea of our suggestions will neither resort to the property of D-linking nor will it focus on the morphological make-up of wh-elements; it will rather make use of a distinction between di1ferent readings of wh-complements that Groenendijk and Stokhof (1984) appeal to when discussing the question of whether (92b) is a valid inference from the premise (92a): (92)

a. b.

John knows who walks. John knows which girl walks.

Groenendijk and Stokhof point out (see also Heirn 1994) that the conclu.sion is valid if the argument is to be understood in the following sense. If we know of a set which of its elements have a certain property, then we also know this of every subset of that set. Since the set of girls is a subset of the set of individuals, it follows that John knows

19. The requirement of a cleft construction here depends on a different factor: this is the subject of a transitive verb. With unaccusative verbs you can indeed get

(i)

Vegne-lo chi? comes-he who? '\ftlho is coming?'

We refer to Munaro and Pollock (2004) for a treatment of this distinction. 20.

For an elaborate presentation of this approach see Grewendorf and Poletto (2009b).

332 Gunther Grewendorf & Cecilia Poletto

which girl walks if he knows who walks. The crucial point of this argument is that the conclusion is taken de re. However, the wh-complement of (92b) may also have a de dicto reading and with that reading, the conclusion may not be valid, as can be seen from the following reasoning. Assume that only one individual walks and that this individual is a girl. Now, if John knows of this individual that it is the individual that walks but fails to believe that this individual is a girl, then the premiss (92a) is true but (92b) is no longer a valid conclusion. In other words, the conc.lusion is not valid unless it is read in the tollowing way: In order for John to know which girl walks, he should believe of every individual which is in fact a girl and walks, not only that this individual walks but also that she is a girl. 1he distinction benveen the de 11?. and the de dicto reading of wh-elements can be used to mark an important difference between wh--elements that may help us to disentangle their varying syntactic behavior in Cimbrian. It should be dear that whenever wh-phrases exhibit a quantificational restriction, they can have a de dicto reading irrespective of whether or not they are D-linked. In this way we can separate the class of wh-phrases that obligatorily undergo high wh-movement without making use of the notion of D-link.ing. It may be objected that our distinction just resorts to a semantic characterization of morphological complexity, thereby still facing the problem of the ambivalent behavior of bare wh-elements. However, if we incorporate contextual restrictions on quantification, we can assume, again f(.)llowing Groenendijk and Stokhof (1984, Section 1.6), that contextual restrictions function as a "hidden" common noun in the wh-term, and thus allow for de dicto readings with respect to such a hidden common noun. It thus follows that depending on the context, bare whelements may also have de dicto readings. If we now interpret the existence of a de dicto reading as a constraint on the topi-· cality of wh-elements, we can correlate the topicality of a wh-element ~ith a high functional Topic projection that represents the property of individuation in the sense of S7.abolcsi and Zwarts (1997). Our Cimbrian examples show that wh-elements which are characterized by the availability of a de dicta reading move to a functional projection that obviously precedes ForceP, which is the projection that hosts the CPexpletive z. Let us theretore assume that the Topic projection that is targeted by de dicto wh-elements is higher than ForceP. Still following Groenendijk and Stokhof's intuition that this reading hides a covert N, we can hypothesize that the syntactic translation of the de die to reading with bare wh-items which behave as Topics includes precisely a phonetically null NP in its struc.tme, which is not present otherwise. 'Ihe internal structure of topical bare wh-items in the de dicto interpretation would then be different from the case where they are non--topical. Following Poletto and Polloc.k (2009) we assume that the internal structure of wh-words contains at least

Hidden verb second: The case of Cirn brian 333

the following three projections: a Disjunction phrase, an Existential phrase and a Restrictor phrase. (93)

[DisiP [ExistP [RestrictorP]]]

Romance varieties provide evidence tor the existence of these three projections because they display two types of doubling of bare wh-words in simple questions (see Poletto and Pollock (2009) on specific arguments). 1he first type involves a clitic whitem (which obeys the usual tests tor clitichood) sitting in DisjP and a tonic counterpart being merged in the Restrictor position and then being moved to the Existential phrase. The second type involves a weak pronoun (which is clearly bimorphemic) located in ExistentialP (and moving to DisjP) and a tonic counterpart, which occurs in the RestrictorP. 21 Suppose that the Restrictor is precisely the element which enables the bare whitem to be interpreted as de dicto, (i.e. as topical), then the null assumption is that in this c.ase there is a phonetically empty noun sitting in the RestrictorP, which is not there if this interpretation is not triggered. The empty noun must evidently be different for each wh-word, being a null THING for 'what; PERSON for 'who; PLACE for 'where' and TIME tor 'when'. 22 Notice that the existence of these empty nouns has already been proposed by Kayne (2005) in recent work on constructions which have nothing to do with interrogative wh-items and their internal stmcture. If the idea of an empty noun is correct, we conclude that bare wh--words can have the same structure as complex wh-phrases modulo the presence of a phonetically empty noun, or a different structure. The two possible structures of a wh-word are illustrated in (94) for the wh-item 'what': (94)

a.

[DisjP

b.

[DisjP

weak pronj weak proni

[ExistP [ExistP

tj [RestrictorP THING]]] ti]]

21. The two internal structures of wh-items cotTesponding to clitic doubling and weak pronoun doubling proposed by Poletto and Pollock (2009) are the following:

(i)

bsjP

(ii)

bsjP

cl wh [Exist}' tonic whj [Restri torP tj]] weak pronj [Exist!' t; [RestrictorP tonic wh]]]

'Ihe stmcture of a non doubling wh-word is the following: here the tonic pronoun raises to the highest projection in its internal structure (iii)

[DisjP

tonic wh;

[Exist?

tj

[RestrictorP

tj]]]

Following this reasoning the item corresponding to 'why' should have an empty REASON inside, but we suspect that this wh-item is more complex than that (see Shlonsky (2009) on different interpretations of 'v.ily' which would in our view correspond to different internal structures of the wh-item). 22.

334 Gunther Grewendorf & Cecilia Poletto

One could further speculate that being a "pure operator" corresponds precisely to the syntactic property in (94b ), namely the absence of any possible restrictor, which leaves the bare wh-item with a purely functional structure without a lexical category.23 This type of assumption turns upside down Cardinaletti and Starke's (1.999) hypothesis that weak and eli tic pronouns lack one or two layers of functional structure respec-tively. Here the lacking portion in the internal structure of the wh-item is the lexical part, not the functional one. Evidently, the pruning mechanism which "cuts out" the highest projections, as proposed by Cardinaletti and Starke, does not work here, but pure operators are indeed more functional than other wh--items. The sheer fact that the item corresponding to 'what' in Romance becomes the complementizer, a purely functional element, is an argument in favour of this approach (see Grewendorf (2010) for a detailed discussion of German 'was' as a pure operator occupying a D and not an N position). We reier tl1e reader to Grewendorf and Poletto (2009a) for additional arguments in favor of this analysis. Here we would like to mention only one empirical argument supporting our proposal, namely the fact that bare wh--items must behave as 1opics (and occupy the high position in the CP layer in Cimbrian) if a complement of the null noun is phonetically realized, forcing the null noun to be present as well: (95)

a.

hat -dar get Dar hat-mar gevorst ber vo se an pwlk. He has-me asked who of them has-you given a punch

b. ""Dar hat-mar gevorst ber

vo se

dar hat get

an punk.

The fact that the object clitic pronoun must be enclitic on the verb shows that the wh-item is in the high Topic position. On the null hypothesis that only an NP can take a PP as a complement, the presence of the PP implies that of the NP with the empty noun PE,RSON in its head. Therefore cases like those in (95) are forcedly 'Ibpics rather than pure operators.u Moreover, our assumption that wh-elements with specific semantic/pragmatic properties target a left-peripheral Topic projection receives support from Stowell and

In the minimalist framework we are forced to assume that the NP is just not there if it is not occupied by the empty noun, because projections do not exist "per sf!' but as a reflex of lexical items extracted out of the lexicon. In a strict carthographic approach as the one taken in Cinque (1999), (2001) and subsequent work, it is assumed that functional projections are always there, though nothing occupies them, because they are simply part of our "hardware':. Anyhow, the NP is not a functional projection, but a lexical one, therefore we tend to think that ·when interpreted as a pure operator a wh-word does not have any lexical structure, but only functional layers. 23.

14. In a similar vein, the presence of adjectives which are related to the NP should imply a topical treatment of wh-words.

Hidden verb second: The case of Cim brian 335

Beghelli's (1994) analysis of quantificational phrases with specific reference. Stowell and Beghelli argue that such quantific.ational phrases undergo covert movement to a left-peripheral position which they call "Referential Phrase" (RefP) and which they locate above the root CP. The Cimbrian data in (77) and (78) can be taken to show that left--peripheral movement of wh-· phrases that is motivated by their topical properties also occurs in an overt way in natural languages. A similar observation is made by Thornton (1995), who shows that in child language, referential wh-phrases may undergo overt movement to a position that she identifies as Stowell and Beghelli's Referential Phrase. Our analysis of the wh-split in Cimbrian may provide us with an alternative account of a contrast between wh-constructions in Fassano reported in Poletto (2000: sect. 3.4). Poletto observes that subject clitic inversion is obligatory for some wh-elements but not for others, as illustrated by (96) and (97): (96)

(97)

a.

pa? Can vasto when go-you interr.marker

b.

tu vas? Can pa when interr.marker you go

a.

(Pera di Fassa)

Che compresto pa? interr.marker what buy-you

tu compre? b. *Che pa what interr.marker you buy Proceeding from the assumption that verb movement depends on the type of wh--item, Poletto (2000) appeals to Cardinaletti and Starke's (1999) theory of weak elements and claims that weak wh-elements such as che ('what') need to enter a Spec- Head relation within a left--peripheral Agreement-projection (AgrCP) whose head hosts the complex formed by the inflected verb and the enclitic morpheme. Poletto speculates that this requirement might be due to the absence of the "highest" functional projection within the wh-phrase, which she considers as "corresponding to a TopP inside the DP functional projections" (Poletto 2000: 81). Movement to the SpecAgrC position is required in order for the features of this (absent) topical projection to be recovered from the Spec-head relation with the AgrC-head. The idea that the contrast between (96) and (97) should be related to the topical nature of wh-phrases can be shaped in a different way on the basis of our account for the wh-split in Cimbrian and the theoretical mechanism suggested in Grewendorf (2010). We can claim that the absence of a wh-internal Topic feature turns the whelement into a pure wh-operator that has to overtly fulfill the wh-criterion in a Spechead relation with the finite verb. Topical wh-elements do not need to fulfull this

336 Gunther Grewendorf & Cecilia Poletto requirement overtly and undergo overt movement to a higher Ibpic projection the head of which does not attract the finite verb.

8.

V2, V3 and subject positions

At the end of this analysis of the CP positions and of the elements which occupy them in Cimbrian, we would like to brietly return to the problems we left. unsolved in Section 3. If Cimbrian was and still is a V2 language in the sense that the verb moves to the C domain, how come it does not allow DP subject inversion? The same state of a1la.irs is observed in residual V2 languages like French or Northern Italian dialects, where I to Cis found in main interrogative clauses; in this context subject eli tic inversion is possible, but DP subject inversion is not. Anyhow, if we are here on the right track, Luserna Cimbrian is not a residual, but a full-fledged V2 language, as attested by the following empirical arbrumerrts: (a) obligatory subject clitic inversion in main declara·· tives and embedded clauses introduced by ke-type complementizers, (b) the presence of the CP expletive 'z corresponding to German es, (c) subject clitic inversion triggered by embedded clauses in front of main clauses, (d) the asymmetry between main clauses and embedded clauses introduced by a.z-type complementizers. The literature on Romance languages offers at least nvo possibilities to deal with this problem. 1he first is the one proposed by Benindt (1984) for Old Romance languages and extended to modern Romance languages, where D P subject inversion of the Germanic type is always banned (except for some Rhaetoromance dialects, which in fact are V2). Benidt assumes that movement of the inflected verb to the C domain automatically licenses a subject pro in the SpecAgrS/SpecTP position, which then cannot be occupied by a subject DP. In this case, the subject DP can only be either right or left dis·· located in Romance, and notably this is also true of Cimbrian. However, Cimbrian is not a pro drop language in the original sense oftl1e tenn: when the subject DP is not present, there must be a subject pronoun (clitic, tonic or weak) even when the verb is in C. Pro drop is not possible even for the expletive pronoun or quasi arguments found with meteo-verbs (though Cimbrian allows for low subjects in the vP phase), hence an analysis in terms of pro drop, as suggested by Benindt for Old Romance, becomes rather implausible. An alternative solution, which has also been proposed by other authors still on the empirical basis of Romance, (see among others Beninca and Cinque ( 1985), Giupponi (1.988) Barbosa (1.995), and more recently for Germanic lanbruages Platzack (201.0), is to assume that definite DP subjects are located too high in the structure to be inverted. Barbosa, as well as other authors, does not make any distinction between different

Hidden verb second: The case of Cim brian 337

types of subjects, and claims that subjects are generally "left dislocated'~ As we have seen in Section 3, Cimbrian shows that actually we have to make a distinction at least between DPs, subject clitics and bare Quantifiers. 25 Subject clitics morphologically express a nominative case feature, and sit in the lP position where nominative is assigned (either AgrSP or TP, depending on the theory we assume) before undergoing movement to the dedicated position, which is, as we have shown (see the discussion concerning structures (45) to (47)), in the Topic field below ForceP in Cimbrian. The following structure illustrates the point: (98)

[SubordP [Subord 0 ke] [ForceP 'z Vfin [TvpicP [WackP

subj.clit. [Focusi' [FinP]]]]]]

Therefore, they invert because the inflected verb in V2 contexts passes through the WackernagelP position of (subject) ditics on its way to Force. As for the distinction between DPs (which cannot invert) and QPs (which can invert), we would like to again exploit the perspective proposed by Beghelli and Stowell (1997), who show that different types of quantifiers occupy different positions according to their interpretation. As seen above, Beghelli and Stowell (1997) offer arguments to the effect that referential DPs (on a par with specific quantifiers) are located in the highest position, namely RefP. As shown above in Section 3, (at least some) quantifiers can indeed be inverted in Cimbrian, while DPs cannot. Following this line of reasoning we propose that quantifiers occupy a different (and lower) position with respect to referential DPs and that this lower position is crossed by the inflected verb on its way to Force. As for the exact position of quantifiers, we assume with Poletto (2000), 26 (2007) that quantifiers are located inside the Focus field

2;. At present we do not know how indefinite nouns behave. 'Ihis is an empirical question, which we hope to solve in the near future on the basis of fi.eld work 26. Arguments in favour of this analysis reported in Poletto (2000) are the following: (a) bare QP subjects in the N!Ds do not trigger number agreement, they have a defaul third person singular reading even when they are morphologically marked as plural, (b) bare QP subjects must be into nationally focused when they are in preverbal position, otherwise they are realized in postverbal position, (c) in dialects like Franco-Proven<;:al where Focus constructions are substituted by cleft clauses, preverbal quantifi.ers are systematically realized in the Pocus position of a deft clause:

(i)

r: ha pa gnun ke m a vit. Rodoretto di Prali (Francoproven<;:al) It has not nobody that me has seen 'Nobody saw me?

Moreover, in Old Italian object quantifiers systematically precede the past participle, while this L~ not the case for DPs, which shows that object QPs also have a special position in the low pmtion of the sentence structure.

338 Gunther Grewendorf & Cecilia Poletto

in a dedicated position. Given that in the left periphery Focus is lower than Force, quantifiers invert, as illustrated in (99): (99)

[SubordP f.subord 0 ke) h:arceP

•z Vfin f.'lbpicP [WackP h:acusP [ QP) [FinP])]])]

Still following this way of reasoning, we assume that referential DPs reach a position higher than Force and this is the reason why they never invert If Barbosa and other authors are ri.ght in assuming that DP subjects are located in a sort of Topic position, the only position which satisfies these two conditions (being a Topic and being higher than Force) is the one we have discussed in the previous section for topical wh-items. (1 00)

[SubordP [Subord 0 ke] [RefP

Topic-wh/DPsubject [Fared' 'z vfin

hopicP[FocnsP [FinP]]]]]]

A specific claim about the relation between Topics and subjects is made by Rizzi (2004), who assumes that "subject and 'Ibpic have something in common: the event as somehow involving that argument. Following a rather standard terminology, we will call this relation "aboutness"" (Rizzi (2004: 122)). Rizzi further notices that Topics and subjects are not c.om pletely alike as they differ in terms of D-linking: subjects are not necessarily D-linked, while Topics are. Anyhow, definite DPs are indeed D-linked, being known to both speaker and addressee. We believe that this is directly reflected in the syntax of Cimbrian in the sense that definite DPs move to the Topic. position RefP represented in (99).271his allows us to capture the fact that DP subjects never invert: the inflected verb stops in Force located lower than RefP. Therefore, we preliminarily conclude that the position of topical wh--items and referential DP subjects in Cimbrian is the RefP position located above ForceP. 28 Let us now come back to the second problem we left unsolved in Section 3, namely the possibility of reconciling Roberts' analysis ofV2 with the existence ofV3 (and by V3 we also mean V4, VS and V,., any number of XPs in front of the inflected verb) in languages which nonetheless display V to C movement.29

At this point we still do not know whether the relevant notion here is definiteness, or rather, as suggested by Beghelli and Stowell, specificity. The answer partially depends on the behavior of indefinite DPs, which can be specific. As we do not have data on indefinite DPs, for the moment we leave the question open.

17.

18. The concept of "aboutness" is semantic in nature. According to Beninca and Poletto (2004) an aboutness relation can be encoded either by Hanging Topic (HTP) or by Left Dislo-

cation in standard Italian. Here we assume that RefP is one of the positions where aboutness can be encoded. though not the only one. 19. ThL"> clarification is not trivial at all Fuss (2008: 242) argues on the basis of Axel (2007) that nwnerous V3 cases which are attested in Old High German are still possible in modern German. However, modern German clearly does not have the freedom of stacking one Topic after the other as Old Romance or Cimbrian do.

Hidden verb second: The case ofCimbrian 339

We briefly outline a sketch here which we think is promising, but will need further confirmation by work on other V2 languages and different diachronic stages of them. 30 Let us first recall the problem: Roberts analysis assumes that any XP moving to the left periphery has to pass through the "bottleneck" of SpecFin because this projection is endowed with an EP P feature which can be satisfied by any XP. As FinP has only one specifier, there can only be one XP to the left of the inflected verb in Fin°, although the XP can continue to move further to check its Operator, Focus or Topic feature. How come then that languages like Old Italian, Sappadino, Mocheno and Cimbrian allow fi.)r any number of XPs located in front of the inflected verb? In a nutshell, we intend to propose that this is due to the fact that Topics are allowed to skip the SpecFin position in some languages but not in others and that this must be tied to the presence of clitics which double Topics. Our argument runs as follows: in languages where V3, V4 etc. sequences are allowed, not all orders of the XPs preceding the inflected verb are possible. It is imprecise to say that any type of XP can be located in front of the inflected verb in the languages mentioned: the only possible sequence is the one provided by an operator (like wh or Focus) preceded by one or severallbpics. Poletto (2002) examines modern Rhaetoromance dialects, which have much in common with Old Romance and are still clearly V2, and shows that the only possible sequences are precisely: (101)

a. b.

Topic*wiJ31 Topic*Focus

Any other sequence (more than one Focus, more than one wh or operator, one or more Focus or operator followed by one or several Topics) is excluded.1he violation of the V2 restriction is thus limited to Topic XPs, which occur in front of a single A'-movedXP. 1his violation seems to be connected to the existence of resumptive elitic heads: Old Italian, Rhaetoromance, Sappadino, Cimbrian, Mocheno32 are alllanbruages which have clitic heads and where clitic heads can double 1opics. Bidese (2008) and Poletto and Tomaselli (2008) propose the following empirical generalizion: (102)

V2 is relaxed when clitic doubling appears.

It is evidently possible to reconcile Roberts' account with languages like Cimbrian by saying that 1bpic elements in the CP do not move, but are base generated there. This is however not what we would like to say, becau.se we adopt the recent standard "iew first proposed in Cecchetto (1999) that "Ibpics do move from the IP.

30.

31·

Here the asterisk means that there can be more than one 1bpic.

32·

Por a detailed analysis of the Mocheno left periphery see Cognola (2008).

340 Gunther Grewendorf & Cecilia Poletto

In more technical terms, we can state that the possibility of clitic resumption somehow permits a Topic to circumvent the restriction proposed by Roberts. If Topics are resumed by a clitic in the IP, they can move directly to the Topic positions in the left periphery bypassing the Spec of FinP, while operator-like items like wh-elements and Focus always have to go through SpecFin; so in any V2language there can be only one operator-like item in the CP. However, Topics must move through SpecFin only if the language does not have doubling elitic heads. How does the presence of a ditic allow 'Ibpics to skip SpecFin? We capitalize here on an intuition that we have already developed in the preceding section on wh·items, namely the one that the type of movement an XP undergoes is determined by its internal structure. Moreover, given that DPs are phases, they are internally built in a fashion parallel to CPs. Suppose that clitic doubling is to be analyzed as a torm of "splitting" of a bigger constituent (usually referred to as "big DP" in Uriagereka (1995)) as proposed by Kayne (1975), Uriagereka (1995), Belletti (2006), Grewendorf (2008), and Poletto (2008) among others. The structure of a "big DP" as proposed by Uriagereka is illustrated in (103). (103)

[[DP [Do d) [DP]]

We adopt this idea in the version developed in Poletto (2008), who assumes that eli tics are the head of a Case projec.tion KP and that KPs/DPs are phases and have an internal left periphery similar (though probably not identical) to the one of the CP. (1 04)

[KP [Ko cl]

·· · [.1\;p!::P [r~:::us.P [Gender}'

[N\:mber [NP]]]]]]

The left periphery internal to a DP, whose highest projection is a KP where the clitic head sits, contains Topic and Focus projections.33 If eli tic doubling is a form of splitting of the KP illustrated in (104), then the element moving to ~lbpicP in languages which have clitic doubling is not the entire KP, but a lower projection. The KP projection containing the clitic is left behind and only a lower projection is subextracted. We propose that the subpart of KP moved out to the CP left periphery is precisely the internal TopicP, which leaves the clitic stranded in IP. (105)

Ucp fropicP] ..• [FinP [1p [SpecAgrP [KP [r«'cl] ~~~~ Agr-0 ... [VP]]]]]]]34

33· Giusti (2006) shows that in some languages the specifiers of the 1bpic and Focus phra'!es internal to the DP are available to movement of ad.iectives. 34· We have represented the KP as being in a SpecAgr position, notice however that nothing crucial for our analysis depends on this asswnption.

Hidden verb second: The case of Cim brian 341

'Ibis means that the constituent moved to the TopP in CP in languages with clitic doubling is a "pure TopicP" in the technical sense that the constituent moved is not a DP/KP but precisely a TopP which perfectly matches the TopP hosting it in CP. In languages where no eli tic doubling is possible (because there exist no clitic heads), the only option is to pied··pipe the whole big DP/KP to the CP left periphery.35 If the whole DP/KP is pied piped to CP, it must pass through SpecFin to reach the left periphery; if only the subpart of KP including TopicP is moved, then this pure TopicP can directly reach its corresponding position in the CP bypassing SpecFin. This explains Bidese's empirical generalization: when clitic heads start doubling left dislocated XPs, then V3 becomes possible even in V2 languages. In our terms, extraction of a pure TopicP from the whole KP/DP becomes possible and the pure TopicP is the only element which can skip SpecFin and reach its final position in the CP left periphery. As for the reason why pure "Iopic constituents have the property of skipping FinP, we capitalize on an observation concerning Topics made by Rizzi (2004): he notices that Topics do not trigger minimality effects, while Focus and whitems do. Rizzi explicitly discusses a hypothesis in terms offeature emptyness: Topics do not contain other features than those that make them 'Ihpics (i.e. aboutness and D-linking in Rizzi's terminology). Suppose that Topics are indeed inert to any sort of feature checking except their own; in intuitive terms, a Topic is in a sense sometl1ing external to the predication and as such cannot check any feature internal to it Therefore, the EPP feature 36 in FinP cannot be checked by a constituent whose highest pro· jection is TopicP, as is the case in ditic doubling languages. However, the whole DP/KP of non clitic doubling languages is still a "normaf' argument and as such is not inert to feature checking and can satisfy the EPP feature in FinP. The link between clitic doubling and V3 is not immediately visible, though our analysis captures it in a rather straightforward way. If our analysis is correct, it follows that only V2 languages with clitic doubling of Topics allow for V3 orders. This means that there might be V2 languages which have already developed ditic heads, but do not have doubling of Topics yet, and even in tl10se cases V3 would be blocked. Our analysis also predics that there should be no V2 languages where V3 is systematically possible and which have no clitic doubling of Topics. To bear out tl1is prediction requires further empirical work on other languages

35·

The technical execution of the idea is represented in the following derivation: (i)

(ii)

[[KP [K'

[[xp

cl]

[TopicP [FocusP [FP [NP]j]J]j

[TopicP [FocusP [FP

[Np]]

rx' [KP [K, cl] H~ f~ f-l':P- fm:]-]]]]]]

36. At present the big mystery of what an EPP feature exactly is remains. However, this does not bear on our analysis, as we claim that Topics are inert to all features of the clause except their own.

342 Gunther Grewendorf & Cecilia Poletto of the Romance and Germanic type and of other language families where our empirical generalization can be tested.

9·

Conclusion

In this article we have discussed the left periphery of Cimbrian and the elements which can occupy it. From this work some general consequences for the structure of the left. periphery and for V2 emerge: first of all, these data confirm the idea that V2 is not a unitary phenomenon in the old sense of a parameter triggering a cluster of phenomena which include the linear restriction, subject inversion and the main versus embedded clause asymmetry. Cimbrian is different from German, as it allows V3 (and V*) orders and displays restrictions on inversion, but still maintains one class of embedded structures where the asymmetry is visible. Various research perspectives are still open. First of all, the analysis of complementizers of the az-type could be extended to other languages which do not display any mixed complementizer system, like standard German, a problem we do not discuss here. The spreading of the comp lementizer ke and the progressive loss of the az comp lementizer will lead to the loss of the main versus embedded asymmetry in sentence stnH.:ture. 1his in turn will probably weaken the evidence native speakers have for the V2 phenomenon (recall that subject inversion is reduced to clitics and that the linear V2 restriction is not respected in Cimbrian). The loss of the "Germanic" type of complementizer might be one of the factors which will eventually lead to the entire loss of any correlate ofV2 (in our terms, loss of any V to C), hence, also of the cases of subject clitic inversion in declarative clauses, and of expletive 'z. Given what we said above, the loss ofV2 seems to derive from the interplay of at least three factors: (a) the loss of the V2 linear restriction due to the development of clitic doubling, (b) the activation of a high Topic position above ForceP which hosts topical wh-items and referential DPsubjects, preventing DP subject inversion, (c) the creation and spreading of a new class of complementizers which erases the original asymmetry in verb movement with respect to main and embedded clauses. We also have further empirical work awaiting us in order to verify if the empirical generalization that clitic doubling and V3 orders are connected is correct. The main thread of our argumentation has been the assumption that the type of movement of an XP to the left periphery depends on its internal structure, both with respect to the final landing site (low OpP or high RefP as is the case for wh-items) and with respect to the path the moved XP follows (through FinP or directly to the 'Ibp P in CP as is the case for left dislocated DPs depending on whether they are eli tic doubled or not). However, there still are several loose ends which we hope to develop in the future concerning the status of Topics, and in particular whether tl1ey are really inert

Hidden verb second: The case of Cirn brian 343

to feature checking internal to the predicate they are associated with. Another possible development of our idea would be to consider the internal structure of subjects and see whether the fact that they occupy different positions- a WackernagelP (as subject eli tics do in Cimbrian), a low Operator position (as bare Quantifiers do) or a high RefP (as definite DPs) is reflected in their internal structure. We leave all these questions open and hope to have shown that the study of Cimbrian is particularly revealing for the syntax of the left periphery.

References Axel, Katrin. 2007. Studies on Old High Germatl Syntax: Left Senterzce Periphery, Verb Plac.ement and Verb-Second [Linguistik Aktuell!Linguistics Today 112]. Amsterdam: John Benjamins. Barbosa, Pilar. 1995. Null Subjects. Ph.D. dissertation, MIT. Beghelli, Filippo & Stowell, Tim. 1997. Distributivity and negation: The syntax of each and every. In Ways of Scope Taking, Anna Szabolcsi. (ed.), 71-109. Dordrecht: Kluwer. Belletti, Adriana. 2006. Extending doubling to non-lexical domains: Complete vs. partial copying + deletion and related reconstruction issues. In Form, Structure and Grammar. il Fest~chrift Presented to GUnther Grewendorf on Occasion of HL~ 60th Birthday, Patrick Brandt & Eric Fuss (ed.~), 129-136. Bedin: Akademie Verlag. Belletti, Adriana. 2008. Answering Strategies: New Information Subjects and the Nature of Clefts, Ms, University of Siena. Beninca, Paola. 1984. I:interferen:za sintattica: Di un aspetto della sintassi Iadina considerato di origine tedesca. Quaderni Patavini di Unguistica 5: 3-15. Beninca, Paola 2006. A detailed map of the left periphery of medieval Romance. In Negation, Tense and Clausal .4.rchitecture: Cross-linguistics Investigations, Raffaella Zanuttim, Hector Campos, Elena Herburger & Paul Portner (eds), 53-86. Washington DC: Georgetown University Press. Beninca, Paola & Cinque, Guglielmo. 1985. Lexical subjects in Italian and the pro-drop parameter. Paper presented at the Comparative Generative Grammar Fiesta, Salzburg, August 1985. Benincit, Paola & Renzi, Luigi. 2000. La venetizzazione della sintassi nel dialetto cimbro. In Isole linguistiche, GiannaMarcato (ed.), 137-162. Padova: Unipress. Beninca, Paola & Poletto, Cecilia. 2004. Topic, focus and V2: Defining the CP sublayers. In The Structure ofCP and IP, Luigi Rizzi (ed.), 52-75. Oxford: OUP. Bhatt, Rakesh M. & Yoon, James. 1991. On the composition of COMPand parameters ofV2. In Proceedings of the lOth West Coast Conference on Formal Linguistics (WCCFL), Dawn Bates (eel.), 41-53. Stanford CA: Leland Stanford Junior University (Center for the Study of Language and Information). Bidese, Ermenegildo. 2008. Die diachronL~che Syntax des Zimbrischen. Ttibingen: Gunter Narr. Cardinaletti, Anna & Starke, Michael. 1999. The typology of structural deficiency: On the three grammatical classes. In Clitics in the Languages c~fEurope, Henk van Riemsdijk (eel), 145-233. Berlin: Mouton de Gruyter. Cecchetto, Carlo. 1999. A comparative analysis ofleft and right dislocation in Romance. Studia Linguistica 53: 40-67. Cinque, Guglielmo. 1999. Adverbs and Functional Heads. Oxford: 0 UP.

344 Gunther Grewendorf & Cecilia Poletto Cinque, Guglielmo. 2001. 'Restructuring' and the order of aspectual and root modal heads. In Current Studies in Italian Syntax - Essay Offered to Lorenzo Renz~ Guglielmo Cinque & Giampaolo Salvi (eds ), 137-155. Amsterdam: Elsevier. Cognola, Federica 2008. OVNO syntax in Mocheno main declarative clauses. To appear in Sele;.:ted Proceedings ofthe .34th 1,1contro di Grammatic.a Ge,lerativa, Paola Benindt, Federico Damonte &Nicoletta Penello (eds), Padm'R: Unlpress. Fuss, Eric. 2003. On the historical core of V2 in Germanic. Nordic .Journal of LinguL~tic.~ 26(2): 195--231. Fuss, Eric. 2005. Tile Rise of Agreeme11t. A Formal Approach to the Syntax and Grammati-cali.zation of Verbal Injlectio11 [Linguistik Aktuell!Linguistics Today 81]. Amsterdam: John Benjamins. Fuss, Eric. 2008. Word Order and Language Change. On the Interface between Syntax and Morphology. Habilitationsschrift, Universitiit Frankfurt. Giupponi, Elena. 1988. Pro drop Parameter und Restrukturierung im Trentino. Ph.D. dissertation, University of Vienna Giusti, Giuliana. 2006. Parallels in clausal and nominal periphery. In Phases of Interpretation. Mara Frascarelli (ed. ). 163-184. Berlin: Mouton de Gruyter. Grewendort; Gunther. 2008. The left clausal periphery. Clitic left dislocation in Italian and left dislocation in German. In Dislocated Elements in Discourse. Sy11tactic, Semantic, and Pragmatic Perspectives, Benjamin Shaer, Philippa Cook, Werner Frey & Claudia Maienborn (eds), 49-94. London: Routledge. Grewendort; Gunther. 2010. Wh-movement as topic movement. To appear in Functional Heads. Studies in honor of Guglielmo Cinque, Laura Bruge, Anna Cardinaletti, Giuliana Giusti. Nicole Munaro &Cecilia Poletto (eds), Oxford: OUP. Grewendort; Gi.inther & Poletto, Cecilia. 2005. Von OV zu VO: Ein Vergleich zwischen Zimbrisch und Plodarisch. In Das Zlmi:Jrische zwischen Germanisch und Romanisch, ErmeneglldoBidese, James R. Dow&'Ihomas Stolz(ed.~). 114-128. Bochurn: Brockmeyer. Grewendorf, Giinther & Poletto, Cecilia 2009a. The hybrid complementi:z.er system ofCimbrian. In Proceedings XXXV Tncontro di Grammatica Generativa [Studies in Linguistics 3], Vincenzo Moscati & Emilio Senidio (eds), 181-194. Siena: Centro Interdlpartlmentale di Studi Cognltivl sul Linguaggio, Universita degli studi di Siena. . Grewendort:, Gunther & Poletto, Cecilia. 2009b. Topic or operator? 'I'wo types of wh-items in Cimbrian. Ms. Universities of Frankfurt and Venice. Groenendijk, Jeroen A.G. & Stokhof, Martin. 1984. Studies on the Semantics of Questions and the Pragmatics of Answers. Ph.D. dissertation, University of Amsterdam. Haider, Hubert. 2009. Gardenpaths mlt Marga. Talk delivered at the workshop in honor to Marga Reis, Thbingen. Heim, Irene. 1994. Interrogative semantics and Karttunen's semantics for know. Ms, MIT. Kayne, Richard. 197 5. French Syntax. Cambridge MA: The MIT Press. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge MA: The MIT Press. Kayne, Richard. 2005. On parameters and principles ofpronounciation. In Organizing Grammar. Studies in Honor ofHenk van R.iem~di,tk, Hans Broekhuis, Norbert Co rver. RinyHuybregts, Ursula Kleinhenz & Jan Koster (ed.~). 289-299. Berlin: Mouton de Gruyter. van Kemenade, Ans. 1987. Syntactic Case and Morphological Case in the History ofBnglish. Dordrecht! Forts. Munaro, Nicola. 1995. On nominal Wh-phrases in some norfuern Italian dialects. Rivista di Grammatica Generattva 20: 69--100.

Hidden verb second: The case of Cirn brian 345 Munaro, Nicola. 1997. Proprieta Strutturali e Distribuzionali dei Sintagmi Interrogativi in Alcuni Dialetti Italiani Settentrionali. Ph.D. dissertation, University ofPadova. Munaro, Nicola & Pollock, Jean- Yves. 2004. Qu'est -ce que (qu)-est -ce que? A case study in comparative Romance interrogative syntax. In Handbook of Comparative Syntax, Guglldmo Cinque & Richard Kayne (eds), 221-258. Oxford: OUP. Paoli, Sandra. 2003. COMP and the Left Periphery: Comparative Evidence from Romance. Ph.D. dissertation, University of Manchester. Panieri, Luca et al. 2006. Barlirnen z'schriha tm reda az be biar: Grammatica del ambro di Luserna, Regione Autousma Trentino Alto Adige, Istituto Cirnbra Lusern. Pesetsky, David. 1989. Language--particular processes and the earliness principle. Ms, MIT. Platzack, Christer. 2010. Backward binding and the C-- T phase: A case of syntactic haplology. To appear in Functiorwl Heads. Studies in honour of Guglielmo Cinque, Laura Brage, Anna Cardinaletti, Giuliana Eiusti, Nicole Munaso, & Cecilia Poletto (eds.), Oxford: OUP. Poletto, Cecilia. 2000. The Higher Functional Field: Evidence from Northem Italian Dialects. Oxtord: OUP. Poletto, Cecilia. 2002. On V2 and V3 sequences in Rhaetoromance. In Syntactic Aiicrovariation, Sjef Barbiers, Leonie Cornips & Susanne van der Kle~i (eds). . Poletto, Cecilia. 2007. Tra tutto e niente: Ordini OV in Italiano antico. In Miscellanea di studi linguL~tici ojJerti a Laura Vanelli da amici e ailievi padovani, Roberta Maschi, Nicoletta Penello & Piera Rizzolatti (eds), 33--48. Udine: Forum Editrice. Poletto, Cecilia. 2008. Doubling as a spare movement strategy. In Aiicrovariatiotl itl Syntactic Doubling, Sjef Barbiers, Olaf Koeneman, Marika Lekakou & Margreet van der Ham (eds), 38--68. Bingley: Emerald. Poletto, Cecilia & Pollock, Jean-- Yves. 2009. Another look at Wh-questions in Romance: The case of Mendrisi.otto and its consequences for the analysis of French Wh-in situ and embedded interrogatives. In Romance Languages and Li11guistic Theory 2006. Selected Papers from 'Going Romance', Amsterdam, 7-9 December 2006, Daniele 'Ibrck & W. Leo Wetzels (eds), 199-258. Amsterdam: John Benjamins. Poletto, Cecilia & Tomaselli, Alessandra. 2001. La Sintassi del Soggetto Nullo nelle isole Tedescofone del Veneto: Cirnbro e Sappadino a Confronto. In La dialettologia oltre il2001, Gianna Marcato (ed.), 237-252. Padova: Unipress. Poletto, Cecilia & Tomaselli, Alessandra. 2008. Die Syntax der Pronominalobjekten und die Form des Partizips. Konservative Merkmale in der Sprachgeschichte des Zimbrischen. 'Ib appear in Parrallela III, atti del T'erzo Convegno Internazionale Deutsche Sprachwissenscltaft in Italier., Claudio di Meola (ed.). Pollock, Jean- Yves. 1989. Verb movement, universal g.rammar and the structure ofiP. Linguistic Inquiry 20: 365-424. Rizzi, Luigi. 1990. Relativized MinimaWv. Cam bridge MA: The MIT Press. Rizzi, Luigi. 1991. Residual verb second and the Wh-cri.terion. Technical Report in Formal and Computational Ling<Jistics 2, University of Geneva. Reprinted in Parameters and Functional Heads, Adriana Belletti. & Luigi Rizzi (eds), 63-90. Oxford: OUP, 1996. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Eleme12ts of Grammar, Liliane Haegeman (ed.), 281-337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the positi.on 'int(errogative)' in the left periphery ofthe clause. In Current Studies in Italian Syntax Essays Oj_fered to Lorenzo Renzi, Guglielmo Cir1que & Giampaolo Salvi (eds), 287--296. Amsterdam: Elsevier.

346 Gunther Grewendorf & Cecilia Poletto Rizzi, Luigi. 2004. Locality and the left periphery. In Structures and Beyond, Adriana Belletti (ed.), 223-251. Oxford: OUP. Rizzi, Luigi & Shlonsky, Ur. 2007. Strategies of subject extraction. In Interfaces Recursion = Language?, Ulrich Sauerland & Hans-Martin Gii.rtner (eds), 115-160. Berlin: de Gruyter. Roberts, Ian. 2004. 'Ihe C-system in Bryt:honl.c Celtic languages, V2, and the EPP. In 1he Structure ofCP and IP, Luigi Rizzi (ed.), 297-328. Oxford: OUP. Scardonl. Sara. 1999. La Sinta.ssl. del Soggetto nel Cimbro Parlato a Giazza. Ph.D. dissertation, Universita degli Studi di Verona. Shlonsky, Ur. 2009. \\'here is Why? Ms, University of Geneva. Stowell, Tim & Beghelli, Filippo. 1994. The direction of quantifier movement Paper presented at the GLOW Conference, Vienna. Szabolcsi, Anna & Zwarts, Frans. 1997. Weak islands and an algebraic semantics for scope taking. In Ways ofScope Taking, Anna Szabolcsi (ed.), 217---262. Dordrecht: Kluwer. Thornton, Rosalind. 1995. Referentiality and Wh··movement in child English: Juvenile D-linkuency. Language Acquisitiotl 4: 139-175. Tomaselli, Alessandra. 1995. Cases of verb third in Old High German. In Clause Structure atid Language Change, Adrian Battye & Ian Roberts (eds), 345-369. Oxford: OUP. Tomaselli, Alessandra & Bidese, Ermeneglldo. 2007. Diachronic development in isolation: The loss ofV2 phenomena in Cimbrian. Li11guistische Berichte 210: 209-228. Uriagereka, Juan. 1995. Aspects in the syntax of clitic placement in Western Romance. Linguistic Inquiry 26: 79--123. Watanabe, Akira. 1993. AGR-based Case Theory and its Interaction with the A-bar System. Ph.D. dissertation, MIT.

Revisiting the Wackernagelposition The evolution of the Ciinbrian pronominal systen1 Ermenegildo Bidese Universita di Trento

The present contribution reconstructs the development of the personal object pronouns of Cimbrian, a German dialect spoken in Northern Italy which evolved many centuries in close contact with northern Italy's Romance dialects. With reference to their functional status and their clausal position we discover that Cimbrian's object pronouns started from a German model and have over time become closer to a Romance one. In the older Cimbrian texts, these elements are clearly recognizable as full phrases (XP), occupying the traditional WackernagelposUion; in modern wTitings they behave as heads (X 0) and appear only in an 'adverbal' position, i.e. enclitic to the finite verb, similarly to the syntax of Romance object pronouns. The fact that they cannot be realized as proditic to the finite verb - like the Romance ones - shows however that the original Germanic. syntax limits the influence of that Romance. Attempting to explain this phenomenon, this current study suggests revisiting the structure of the

Wackernagelposition.

1.

Introduction 1

Cimbrian is a German dialect spoken today only in the secluded mountain village of Lusem (Italian Lusema) in the province of Trento in northeast Italy.2 Yet, in the past, Cimbrian was commonly used - in a few villages until the eighties of the last

I sincerely thank Werner Abraham and James R. Dow, who read the first draft of the present paper, both for reviewing the content and revising my English. I am also grateful to Mid1ael T. Putnam for giving me the opportw1ity to publish this study.

1.

2. Por the discussion about some aspects of the syntax of Luserna-Cim:brian see the contributions by Grewendorf & Poletto and by .Andrea Padovan (this volume) as well as the extensive syntactic discussion and comparison of Cirnbrian 'A'ith to day's Bavarian by .Abraham (also this volume). A classic description of Lusern's Cimbrian can be found in Bacher ([1905], 1976). Tyroller (2003) has also offered a general description of the grammar of Lusern's Cimbrian. Cf. also Panieri, Pedrazza & Nicolussi Baiz (2006).

348 Ermenegildo Bidese

century- in a much more extensive area including a part of the provinces ofVicenza and Verona in the region of Veneto.3 The Cimbrian communities in those provinces formed an administratively autonomous federation tolerated for a long time by the sea-republic of Venice, in Vicenza in the so called Fedemtion of the Seven Municipalities, and in Verona in the Federa.tion of the Ihirteen Municipalities. Generally, it is assumed that the Cimbrian variety spoken in the Seven Municipalities was the most conservative of the three, especially with regard to aspects of both morphology and the lexicon.4 In any case, only in this latter variety was a form of'literature' produced such that we can approach Cimbrian from a diachronic perspective analysing the Cimbrian syntal!: during the last 400 years. 5 The present contribution deals with the evolution of a particular aspect of Cimbrian syntax, namely the functional status and the clausal position of the personal object pronouns. The issue is particularly interesting because the German dialect of Cimbrian has developed in close proximity ·with northern Italy's Romance dialects for many centuries. German distinguishes itself significantly from Italian and northern Italy's Romance dialects with regard to the nature of personal pronouns and their positions in the clause. It is a well-known fact that the paradigm of German's personal pronouns presents only one series, that of the 'full forms; with the exception of the third person singular neuter es that behaves idiosyncratically. By contrast, Italian and northern Italy's Romance dialects clearly show two pronominal series: one 'full form' and another clitic one.6 It is true that many German dialects present several cases of pronominal fonns reduced morphonologically attaching to the final verb in the main clause and to the complementizer in the embedded one, but they are to be considered as allomorphs of the full forms that means as maximal projections (XP) like the full pronouns. By contrast, Italian and northern Italy's Romance ditic pronouns are

J. Today, in the Venetian villages, only few speakers or semi-speakers can be found, as there is no longer an integrated community of people who use the Cimbrian langtLage in everyday commtmication Yet, many institutions and cultural associations are involved in the revitalisation of the Cimbrian language. For a general introduction into the geographical and historical main questions about the formation, the evoltLtion, and the linguistic vitality of the Cimbrian enclaves see Bidese (2004).

4· See the classic contributions of Schmeller (1838) and Kranzmayer (1981 ..·1985) and, most recently, Panieri (2005, 2008, and 201 O) as well as Abraham (this volume) extracting from the data very explicit syntactic differences. In the fifties of the last century, the Bavarian linguist Bruno Schweizer pro\'ided the most complete description of the grammar of all the Cimbrian varieties then still spoken and documented in written. The manuscript comprehending five volumes remained unpublished for fifty years. Recently James Dow edited and published Schweizer's legacy (cf. Schweizer 2008 ). See the re\'iew of this monumental work in Abraham (2009). 5·

See Bidese (2010b).

6.

See Poletto & Tomaselli (1995).

Revisiting the Wackernagelposition 349

functional heads? Considering object pronouns and their forms, this prompts the following questions: (1) How do Cimbrian object pronouns behave, or, rather, how do Cimbrian object pronouns evolve under the imminent influence of the neighboring Romance dialects? (2) How can such an evolution under imminent language contact be adequately explained? The oldest known Cimbrian document is the Cimbrian translation of Cardinal Bellarmino's Italian catechism Dottrina christiana breve ("short Christian doctrine"), published in 1602.8 Other relevant historical texts of the Seven Municipalities' Cimbrian are a collection of 'baroque lyrics' from the 17th and 18th centuries, 9 a fic-titious dialogue about a visit to the market bet\¥een 1:\'V'o persons from Asiago (the largest of the Seven Municipalities )10 as well as a novena in Cimbrian translated for a religious holiday in honour of the Holy Mary 11 from the 18th century. For the 19th century we have to mention the Cimbrian translation of a further catechism, Pic-coio Catechismo ad uso del regno d'Italia ("small catechism for the Italian Kingdom"), published in 1813, and in a second, slightly revised, edition from 1842,12 as well as a narration about the construction of Roanas (one of the Seven Municipalities) bridge written in 1895 by the teacher Domenico Zotti and published in 1906 by the Paduan university professor and ethnologist Aristide Baragiola.13 Another source of the synt:lX of this phase is the collection of tales found in Schweizer (1939). Further texts have been produced in the eighties of the last century, among them worth mentioning the collection of Costantina Zotti's memoirs. 14

7· Cardinaletti (1999: 63): ':A.t the end of the derivation, weak pronouns are (deficient) maximal projections occurring in specifier positions, whereas clitic pronouns are heads, adjoined to the functional head." Note that Cardinaletti & Starke (1999) proceed on the assumption that there are three types of pronouns in natural languages: strong pronouns, weak pronouns, and clitics. 8. Two partially diverging and differently preserved original copies of the first Cimbrian catechism can be consulted respectively in lnnsbruck (Ferdinandeum, sig. H~ 906, nr. 3) and in Vienna (()sterreichische Nationalbibliothek, sig. 62790-A Rara). The critical edition of this text was provided by Wolfgang Meid (cf. Meid 1985a). 9·

See Heller (1988) for the critical edition.

10.

See Stefan (1998).

n.

See Stefan (2000).

For both editions see Meid (1985b ). A digital version of the two texts can be retrieved in http:/ /titus. uni- frankfurt.de/texte/etcs/ germ/zimbr/kat 1813d/kat18.htm and http://titus. uni -frankfurt.de/teA.1e/etcs/germ/zimbr/kat 1842d/kat 18.htm.

12.

13. See Baragiola (1906). We use the new edition of the text provided by the Cirnbrian Cultural Institute (d Lobbia & Bonato 1998).

14.

See Zotti (1986).

350 Ermeneglldo Bidese

For our investigation of the diachronic syntax of Seven Municipalities' Cimbrian it seems advisable to select both catechisms (abbreviations being Cat.1602 and Cat.1813) because of similarity in style, and the narration about Roana's bridge (abbreviation Bar..l906) because of the length of the story and the narrative style; lyTic, however, appeared less suitable for our syntactic investigation. Furthermore, from a syntactic point a view, Bar.1906 represents a dear innovation in comparison with the two catechisms. Later narrations present the same aspects Bar.1906 already shows. In what tollows, each of these literary works will be characteri7.ed with respect to their linguistic properties to be critically discussed.

2.

2.1

Tite diachronic syntax of Cimbriau object pronow1s "!he Cimbrian catechism of 1602

In the first documented phase of Cimbrian diachronic syntax, object pronouns systematically fill two positions: I.

Immediately to the right of the finite VerbiV2, or else, if this position is already occupied by the subject pronoun (under subject-verb-inversion), to the right of that (in the far-left middle field). This can be observed in the following clausetypes: (a) in declarative main clauses (cf. 1---3); (b) in interrogative main clauses (cf. 4 and 5); (c) in subordinate clauses governed by a 'bridge verb' (cf. 6); and (d) in subordinate clauses introduced by the causal complementizer barume!barome "because" (cf. 7 and 8): (1)

unt dez ezzen, unt rusten ist unzDAT noat and to eat and to dress is (for) us necessary.

(2)

De belt hatten ghepittet the world has--him implored.

(3)

derzua schaffet er urlZnAT ••• in addition orders he (to) us ...

(4)

Ber hat se aufgheleghet? who has them ordered?

(Cat.1602: 597)

(5)

Disa bia kimet: si unz..iJA'i' zo sainan vorghebet? these, how became they (for) us to be forgiven?

(Cat.l602: 741)

(Cat.1602: 423---4) 15

(Cat.1602: 1345)

(Cat.1602: 502-3)

15. 1he nunibers indicate the lines of the original text in the standard edition of Meid (1985a).

Revisiting the Wackernagelposit!on 351

II.

(6)

darome bizzebir, er mak unzACC erhoern therefore know--we, he can us hear.

(Cat.l602: 395)

(7)

barume er [ ... ] bil uzACC holik because he [... ]wants us (to be) saintly.

(Cat.1602: 555)

(8)

Barome mit der Vorte enhalteber unzAcc von sunten because with the fear abstain -we llS from sins.

(Cat.1602: 92)

Generally, immediately to the right of the complementizeriComp or, if this posi-· tion is already occupied by the subject pronoun, to the right of those. This is typical of those subordinate clauses introduced by other complementizers than barume!barome "if" (cf. 9···12) or by a relative pronotm (cf. 13 and 14 ): (9)

daz urlztJAT ist zoakemt dazselbe liberle that U'l is sent this the same little book.

(C!it.1602: 66)

(Hl)

Daz er dik burt erhueren that he you will hear.

(11)

dazar u.znAT ghebe dez bol that--he us may give the well--being.

(CaU602: 376)

(12)

bia ber unsACC haben zo botragan how we us have to behave.

( Cat.1602: 492)

(13)

beile tmZnAT habent ghebet daz sainen who us have given the life.

(Cat.1602: 526)

(14)

derse hat in himel ghefL1ert who--them has in (the) heaven conduced.

(Cat.1602: 1352)

(Cat.1602: 1267)

From a morphonological point of view we can distinguish two pronominal series in the first Cimbrian Catechism, the tormer with 'full forms; and the latter with morpho·· nologically reduced fonns (cf. for example the fonn unz "us" as in 9 and the form uz "us" in 11). The crucial question is whether the morphophonologically reduced pronominal forms are to be considered as eli tics like the Romance ones or as 'allomorphs' of the full forms. We plead for the second solution, for the following syntactic reasons: a.

'Ihe morphophonologically reduced pronouns do not exclusively occupy an 'adverbal' position as the Romance clitics generally do, 16 since they clearly show structural adjacency to the complementizer or the relative pronoun (see above 9-14 );

16.

See Poletto & Tomaselli (2009), and Tomaselli (2010).

352 Ermeneglldo Bidese

b. c.

d.

2.2

In Cat.1602, there are no examples of the typical Romance phenomenon of object clitic-doubling (l;'ho visto luiJ"himditic (I) have seen him") to be foundP In Cat.1602, in special constructions with dislocated elements like the 'leftdislocation/LIY, in which a pronominal resumption of the left-dislocated nominal elements is required, a demonstrative pronoun appears as a resumption element (contrastive left-disiocation in German: den neuen Lehrer, den habe ich heute gesehen/"the new teacher, this have I today seen") against the Romance leftdislocation that requires a ditic element (clitic left--dislocation in Italian: il nuovo insegnante l'ho visto oggi/"the new teacher, himditk (I) have today seen"). 18 1his means: The pronominal resumption strategy of dislocated elements pursued in Cat.1602 stands for a typology that is typical of German, but not of Romance. 19 Cat.1602 dearly presents the German Verb--Second-rule,20 which is traditionally assumed to be connected with the so called Wackernagelposition,21 which represents, in the traditional sentence field-analysis (Satzfeldanalyse), the top position of the MittelfeltP 2 reserved for pronominal elements. In this respect, it has often been proposed to introduce for the first position of the left edge of the Mittelfeld, a Pronominalfeld; the term Wackernageiposition also indicates the special status of the Mittelfeld's left border. 23

The Cimbrian catechism of 1813

In contrast to the first Cimbrian catechism, the Catechism of 1813 shows a consider·· able innovation in light of the position of the personal pronominal object Only in a residual context is it possible to find pronominal objects as in Cat.1602: more precisely, in embedded contexts adjacent to the complementizer; in other words, when the complementi1..er is the declarative conjunction az "that, so that" (d. 15-17):

17.

See Anagnostopoulou (1999).

See for this topic Anagnostopoulou (1997) and for an in-depth analysis of the Cimbrian data Bidese (2008).

18.

19.

See Poletto & Tomaselli (2000) and Pili (2001).

2.0.

See Bosco (1999) and Bidese (2008).

2.1.

See den Besten (1983), Travis (1984), Tomaselli (1990) and Kiparsky (1995).

2.2.

See for example WC'Illstein-Leisten. Heilmann & Stepan (1997: 57).

See Abraham (1997), (2005) and this volume. See also Zifonun., Hoffman & Strecker (1997: 1557).

23.

Revisiting the Wackernagelposit!on 353 (CaU813: 533) 24

(15)

Az uzDAT dar iinzar Gott schenke alle de sain gnizien so that us our God grants all his graces.

(16)

az U:ZUAT kemme ghet bas barpitten that us may be given what we ask.

(Cat.1813: 384)

(17)

az ar mar nommet alla de bool that you (from) me would accept all the good">.

( (',at.1813: 477)

With all other subordinate conjunctions or with relative pronouns, object pronouns are realized enclitically to the finite verb, therefore 'adverbally' (cf. 18-20 and 21): (18)

biabarhicensenaach how we forgive--them.

(Cat.1813: 396)

(19)

se dar benne bar srunUZACC net gapaichtet since we have-us confessed

(Cat.1813: 465)

(20)

b~EL ar hebetmar voar hemmest gamacht which you have-(for)-me l.mtil now made.

(Cat.18L3: 362-3)

As tor the object pronouns appearing in these contexts, we assume that they are to be considered as clitics like the Romance ones, but with the relevant difference that the Cimbrian clitics are always in an enclitic position, never in a proclitic one. This assumption about the nature of these pronouns is confirmed by the fact that there are several clauses in Cat.1813 with a clitic duplication of the pronominal (only indirect) object (object clitic-doubling). See (21)-(23): (21)

bibel se' habentmart seii miari iibel gataant how often they have-meclitic they me harm done.

(22)

brumme gal6.ichentachi euchj because they like-youclitic you.

(23)

tort z' kimmetacht euchj zo kemmen ... since it belongs-(to) youclttlc (to) you to be ...

se

(Cat.1813: 375---6)

(C,at.1813: 369)

(Cat.1813: 505-6)

As for the two other phenomena (left dislocation and V2), we assume that their development is connected with that of the pronominal system of Cimbrian. It can be established beyond doubt that in Cat.1813 there are no more examples of contrastive left--dislocation, with a D-·pronoun as resumptive element. By contrast, there are many examples of clitic left--dislocation, above all in interrogative contexts. 'Ihis develop-ment of the dislocation and resumption strategy from a contrastive left-dislocation to

24.

The numbers indicate the lines of the original text in the standard edition ofMeid ( 1985b).

354 Ermeneglldo Bidese

a clitic one is to be judged as a signal that the syntax of the left periphery is going to change radically. In addition to the appearance of resumption clitics in Cat.1813, we also have evidence for the disappearance of the phenomenon of contrastive leftdislocation. Unlike clitic left-dishJCation, the contrastive one allows only one fronted element. Now, in Cat.1813 there are dues as to the realization of two elements in front of the V2-clause. See (24) below, which is not to be considered as a left-dislocation, but, rather, as a hanging topic. 25 Nevertheless, it is a hint at a modified left-periphery in Cat..1813 as compared to Cat..l602. (24)

[Dar Sun von Gotte me Herren] 1 [nukhentensich man] batar1galazt zo sfunan Gott? (Cat.18.l.3: 157-8) the son of God the Lord, making-himself man, has -he ended being God?

We see what appears to be a change of the left periphery confirmed by the fact that the V2-rule is on its way out in Cat.1813. For example, there are optative sentences showing two elements in front of the finite verb (cf. 25 and 26): (25)

[Aso] [de liba Vrau ... un ... ] pitten voriiz iinzarn Herren Jesu Christ (Cat.1813: 555-6) so the beloved Lady ... and ... may pray for us with our Lord Jesus Christ.

(26)

[Aso] [Gott ... ] vii.ariiz in de iinzarn dineste so God ... may lead-us in our activities.

( CaU813: 557)

The same can be found in interrogative sentences, in which the WI·I·-word is preceded by another element, generally a prepositional phrase, as in (27) and (28): (27)

Un [nach den viarzk taghen] baz hatar gatant? and after forty days what has-he made?

(28)

[In minsd1en boart], bazhatsid1 zo tii.nan zo volghen allen disen Comandamenten? (Cat.1813: 232-3) in few words, what must be done in order to follow all these commandments?

(Cat.1813: 194)

All these observations confirm the idea that a change in the Cimbrian pronominal system and the corresponding syntactic phenomena took place between Cat.1602 and Cat.1813, even though in the latter text the signals of syntactic variation are not that evident

25.

See Bidese & Tomaselli (2005) and (2007).

Revisiting the Wackernagelposit!on 355 2.3

The Cimbrian of Dez Dink vo' der Pru.cka

In the third Cimbrian text analyzed here, the narration of the events concerning the construction of Roana's monumental bridge Dez Dink vo' der Prucka ("1he affair of the bridge"), with all the syntactic variations which in CaU813 we only had weak indication tor, clearly come to light The chain of changes follows the following steps --see (a-c): 'Ihe V2-rule is now violated systematically (cf 29-31):

a.

(29)

[In Doi Zait] [dear erste Deputato] hat kott 'me Loite at that time the first deputy has said (to) the people.

(30)

[Afde noin Oarn] [de Klocka] hat get Avviso at 9 oClock the bell has given alarm.

(31)

[In Tak saiten zbenonzbozek von Agosten] [dear Ponte] ist ganget all~a at the day 22nd of August the bridge was up for auction.

(Bar.1906: 110)

(Bar.l906: 115)

In Bat:1906 there are no clauses with clitic left-dislocation, but we saw that this typical Romance structure already has appeared in Cimbrian in Cat..l813. Other Cimbrian texts from this period (tor example Schwei7.er 1939) confirm this. The typical Germanic form of left dislocation, contrastive left-dislocation (~ith a o-pronoun as resumptive element and only one element fronted) was also a characteristic only of the first documented phase of Cimbrian represented by Cat.1602. Like Cat.1813, Schweizer's (1939) texts only show the clitic resumption of dislocated elements and the possibility to have more than one element in left dislocation. 27 & for object clitlc--doubling, it can be noted that, in Bar.1906, this construction expands to other contexts than those in CaU813. Unlike CaU813, object clitic-doubling now captures nominal phrases too (as opposed to only pronominals in Cat.1813) (cf. 32 and 33). Likewise, doubled elements can now be found in preverbal position (cf. 33 and 34):

b.

(32)

biar haben-z-eni gamachet segen de Braveni we have-it -(to) themcHtic let see to the fops.

(Bar.1906: 115)

(33)

me dessen ~·onme Roane; degnarn--se-sich gnanca segen--sent that that of Canove deign-they not even to see-itclitic'

(Bar.1906: 110)

26.

Pagination corresponds to that of the new edition of Baragiola's text by Lobbia & Bonato

(1998).

27.

(Bar.l906: 105) 26

For more data see Bidese (2008).

356 Ermeneglldo Bidese

(34)

miar; importar-z-mar; nicht zo sterben to me matters--it-(to) meclitic nothing to die.

(Bar.l906: 111)

Consistently with this syntactic development, the only position the object pronow1s occupy in Bar.1906 is the 'adverbaf one, i.e. enditically to the finite verb. Thus, in this Cimbrian text, there are no more examples of subordinate clauses in which the object pronouns are realized adjacent to the complementizer (cf. 35-37): (35)

baREL gebent--ilz zua who combat-us.

(Bar.1906: 112)

(36)

ba sd1on biar haben--z-·et! gatant how terrific we have--it--(to) them made (=take them in).

(Bar.1906: 107)

(37)

as6 bia der Komaun bil-se tradiren how the town council wants-them to betray.

(Bar.1906: 109)

Actually, this text contains no e.xamples of subordinate sentences introduced by az, therefore we can not be sure whether in this particular context it was also possible in this text to reali1.e object pronouns adjacent to the complementizer as residually attested in C..at.18l3 (see above 15---17). We conclude that it can not be completely excluded for Bar.1906 either. In any case. it would be a matter of a very residual possibility. To sum up, the diachronic analysis of the Cimbrian pronominal system from the first attested document in this language, the 1602 C-atechism, to the story Dez Dink vo' der Prucka, shows that over some three hundred centuries Cimbrian moved away from a syntax which is typologically German and got closer to a Romance type. Whereas the pronominal objects in Cat.1602 can be considered as morphonologically reduced pronouns and also as allomorphs of the full forms, those in Bar.1906 must be ana-lyzed as clitics in accordance with the categorical status of ditics in northern Italy's Romance dialects. Regarding the position in the clause, whereas the original position of the pronominals was the Wackernagelposition, on the left periphery of the Mitteifeld, that in Bar.1906 is 'adverbal: i.e. structurally adjacent to the finite verb. Furthermore, the examined data seem to suggest that this development of the Cimbrian pronominal system went hand in hand with the following changing phenomena in three distinct syntactic areas: (a) the syntactic change of the left clausal periphery: from contrastive left-dislocation to clitic left-dislocation, concomitant with the loss of the strict V2-rule; (b) the syntactic development of the structural center of the clause: the emergence of object clitic-doubling; and, finally, (c.) the grammaticalization of the intermediate area between the left dausal periphery and the structural center of the clause: the change of personal pronouns from morphonologically reduced forms to ditics. The following section is devoted to the analysis of three aspe<.:ts of the diachrony of Cimbrian.

Revisiting the Wackernagelposition 357

3·

3.1

The diachronic syntax of Cimbrian personal object pronouns: An explanation The left periphery of the clause between Cat.l602, Cat.l813 and Bar.1906: Romance expanded CP or German unexpanded CP?

As discussed previously in this chapter (cf: 2.1 above), from a typological perspective, the syntax of the left clausal periphery in Cat.1602 shows significant resemblance to that ofV2··languages like German. Cat.1602's Cimbrian demonstrates the crucial linear restrictions that strict V2-languages are characterized by: (a) In root contexts, the finite verb is in strict second position; (b) In root contexts, subject-verb-inversion appears; (c) word order asymmetry is present between main and embedded clauses for 'light' elements (negation, the reflexive, and object pronouns as well as verbal partides) 28 and a left dislocation strategy of nominal elements and their resumption all of which are typically German(ic). Following Giorgi & Pianesi's (1997) Theory of Scattering and its application to V2-languages in Poletto & Tomaselli (1999) and Poletto (2001), we may assume that in V2··languages and also in Cat.1602, functional features are not 'scattered' across more than one head, but, rather, that they are bundled onto only one. This means that Cat.1602's Cimbrian does not present a CP-split language/9 but that its CP-unexpanded, since the relevant functional features are encoded in the lowest of the layers Rizzi assumes for the CP, i.e. in FrNP. Let us now try to determine the features that appear to be involved in order to explain the V2-phenomenon. A classical distinction drawn already in the eighties of the last century provides a difference between languages, in which the CP manifests a particular wealth of features in languages like German and others in which it is the IP that has such complexity. 30 Tomaselli (1.990, Chapter 5) strove to specif)r this difference assuming the CP of Con· tinental West-Germanic V2-languages to be characterized by a [+ pronoun]-feature that is filled by the verbal morphology of the finite verb in root contexts. This triggers movement of the finite verb to C0, which is not the case in languages that do not share this general feature. Moreover, Tomaselli argued that the linear restriction of only one phrasal constituent before the finite verb, i.e. the main phenomenon of V2-rule, can be explained by assuming compulsory lexicalization of a feature [+declarative(/- interrogative)] that, once it is licensed, prevents other constituents

See Bidese & Tomaselli (2005) and, for a detailed analysis of the data, Bidese & 'lbmaselli (2007: 211-214).

28.

29.

See Rizzi (1997).

30.

See den Besten (1983) and Rizzi (1982).

3.58 Ermeneglldo Bidese

from moving into the CP. After Rizzi's split-CP hypothesis at the end of the nineties, Haegemann (2000) and Roberts (2004) proposed to account for the particular phenomenology of the V2 languages by assuming: (a) a [+ Fin]-feature in the FrNP that allows the rise of the finite Verb into the head of this projection in the V2-lanbruages; (b) an EPP-.feature-' 1 whose mandatory fulfilment (and deletion) has as a consequence the blocking of further movement to SPECFINP. Following Tomaselli's intuition of a [+ declarative]-feature for the German SPEC--CP (see above) and Grewendorfs (2002) evidence in support of the fact that only discourse--bounded DPs can be left--dislocated in German, i.e. only those, which allow a definite, specific interpretation (cf. exan1ple 38 versus 39 from Grewendorf 2002: 91)32 are possible: (38) *Bis zu zehn Bierkriige, die kann sie tragen. As much as ten beer mugs, these can she carry.

(39)

Bis zu zehn dieser Bierkriige, die kann sie tragen. As much as ten of these beer mugs, these can she carry.

We suggest to interpret the EPP-feature assumed in the FINP of CaU602-Cimbrian (and of German) as a[+ d]--feature, i.e. as an operator for the grammatical properties of declarative and discourse-boundedness, and, as suggested by Grewendorf (2002), of definiteness, in ways similar to 'wh' standing for interrogatives. The diachrony of Cimbrian confirms the correctness of this assumption. In fact, the original German, i.e. characteri1..ed by the V2 syntax of the Cimbrian left periphery (:::: Cat.1602) still showed the construction of contrastive left--dislocation (ct~ 40 and 41 ), which is standardly assumed to involved-linking or referentiality33 and to realize them in the CP~ 4 (40)

die andere sibnai diei lernt tmz ... the other seven (rules), these teach us ...

(41)

Dain bili deri ghesceghe ... Your will, this may happen ...

(Cat.1602: 493---4)

(Cat.l602: 363)

31. Originally, the Extended Projection Principle [EPP] indicated a [d]-feature (for definiteness) in the specifier position of the functional category IP and expressed the need for every clause to have a subject, since a nominative element (usually) functions as the clausal subject. In Minimalism, the EPP-feature \\o
See Boeckx & Grohmann (2004) too.

33·

See Cinque (1990) and Anagnostopoulou (1997).

34·

See alc;o Anagnostopoulou ( 1997).

Revisiting the Wackernagelposit!on 359

In Cctt.1602 there is an example of a contrastive right-dislocation (cf. 42) and of a lt:ftdislocation both in embedded contexts, too (cf. 43): (42)

Deri hatz ghemachet Christo unzer Herei This has--it made Christ our Lord.

(43)

Bil koden, das der son Gottezi ( ... ), deri ist ghemakt man (Cat.l602: 200---1) (It) wants (to) say(= it means), that the son of God( ... ), this is(lhas) become man.

(CaU602: 371)

Crucially, as we have seen, this typical German(ic) construction disappears in the following diachronic phases of Cimbrian and makes room for the Romance structure of the clitic left--dislocation (cf. 24 above). Another diagnostic proof of this evolution is the loss ofthe V2--rule (<.f. above 25----28 tor C'at.l813 and 29---31 for Bar.l906), which reveals that the unexpanded Cat.l602-CP underwent a slow expansion in Cat.l813 and Bar.1906. With regard to the left clause-periphery, all of these observations point to a significant modification of the distribution of features between early Ca.t.1602, Cctt.18.l3 and later Bctr.1906. As for the syntax of the pronominal elements, the modifications we have just observed in the C-system do not remain unaffected, as we will see in 3.3.

3.2

The middle of the clause between Cat.1602, Cat.1813 and Bar.1906

As was to be expected, the changes in the CP-layer keep pace with a shift in the structure of IP. The rise of the object clitic-doubling in the period from Cat.1602 to CaU813 provides evidence for this. Based on ideas ofSportiche (1996) we assume a Clitic Phrase (Cr.P) or, better, several clitic phrc:tses whose heads (Cr.O) are filled by the clitics themselves.35 As Sportiche argues, the reason for this assumption lies in the fact that a pronominal clitic does not change the thematic properties of the predicate, since the eli tic stands for an object of the predicate. This means that the link of the clitic with the verb does not affect the transitivity of the latter. 'Ibis intuition can be expressed assuming that the clitic occupies the head of a phrase with a specifier/ head-relationship existing between the clitic and an argument position XP*(cf. 44 from Sportiche 1996:235---236 and his French example): (44) (45)

35·

Cli ... [Y ... XP*i ... ] ...

Marie lesi aura presentes XP*i aLouis.

See also Poletto & Tomaselli (2009).

360

Ermeneglldo Bidese As for the question of the position ofthe clitic phrase(s) in the clause, examples from Romance provide evidence for the fact that the CLP is located below CP, but higher than, or within, IP. In light of Sportiche's suggestion, the very disparate phenomenology of clitic constructions in many languages, including clitic doubling, can be explained syn-chronically by modulating the following parameters: (a) the movement of XP*from IP to SPECCLP can occur overtly or covertly; (b) the clitic head can be overt or covert; (c) xp~·can be overt or covert. 'Iherefore, the structural process of clitic doubling can be explained assuming a clitic head realized overtly and a DP that fills the argument position XP*overtly. Movement of the XP*to SPECCLP can take place synchronically either overtly or covertly. From a diachronic perspective it seems obvious tl1at the covert stage precedes the overt one. This is exactly what the diachrony of Cimbrian provides evidence tor. As shown above, object clitic--doubling emerges in Cat.1813 (cf. 21-23 above, here reproduced as 46-48): (46)

bibel se Mbentmar1 seii miart iibel gataant how often they have-medttic they me harm done.

(47)

bnunme se galaichentachi eikhi becau..e they like-youclittc you.

(48)

tort kimmetachl euchi zo kemmen ... since it belongs-(to) youclitic (to) you to be ...

z

(Cat.1813: 375-6)

(Cat.1813: 369)

(Cat.l813: 505---6)

Quite obviously, all these examples are sentences with clitic doubling of only pro-nouns (pronominals in dative case), a noteworthy fact we will discuss below. For our reconstruction it is important to point out that the movement of the doubled full pronouns to SPECCLP occurs covertly, since in all examples the doubled pronouns appear after the verb. In the next phase, represented by Bar.1906, there are also doubled DPs (cf. 32 and 33 above, reproduced here as 49 and 50). 'Ihis means that the possibility of XP*moving to SPECCLP can now occur overtly too (cf: 33 and 34 above, reproduced here as 50 and 51): (49)

biar haben--z--en1 gamachet segen de Bravent we have-it-(to) themciitic done see (to) the 'fops:

(Bar.1906: 115)

(50)

che dessen vonme R6anet degnarn-se-sich gnanca segen-sent that that of Canove do-they (not) deign even to see-itclltic

(Bar.1906: 11 0)

(51)

miar1 itnpOltar-z-m,lrt nicht zo sterben to me matters-it-(to) mecllttc nothing (inot) to die.

(Bar.1906: 111)

The question why, in the diachrony of Cimbrian, we first find doubled pronouns (Cat.1813) followed by double DPs (Bar.1906) rests on the rise of the object cliticdoubling itself and is connected with the property licensed by this construction.

Revisiting the Wackernagelposition 361 Sportiche ( 1996) arbrued that NPs capable of doubling must be characteri7.ed by the property of 'specificity'.36 Consequently, the clitic phrase has to be interpreted as an operator for specificity. We assume that it licensed by a[+ specificity)-feature which is filled by a specificator/head-relationship between this clitic and the specific DP. Developing Sportiche's proposal further, Anagnostopoulou (1999) argues that clitic doubling constructions are specifically sensitive to 'referentiality'. Thus, she identifies a scale of referentiality starting out with referential indefinite NPs (as a student of physics) on the lowest level and ending on its top with anaphoric pronouns. The latter represent the strongest degree of referentiality and only admit a strictly discourse-bound reading. In view of language acquisition processes, it has been argued that children exhibit the opposite tendency, starting from the structure that shows the most restrictive load of referentiality, i.e. from anaphoric pronouns, and then going down the proposed scale step by step extending referentiality to the NP. 37 An identical course of diachronic emergence may be assumed with regard to the diachronic emergence of clitic doubling. This means that clitic doubling of pronominal elements may be assumed to occur first Only then may this structural process be extended to NPs. Such an explanation is in line with Sportiche's observation: Languages licensing clitic doubling of NPs will always also admit the doubling of pronouns, but not conversely. Applying these arguments to the diachrony of Cimbrian we are able to explain why, as seen in the section devoted to the reconstruction of the historical evolution of Cimbrian, the emergence of object clitic-doubling in the history of this German dialect goes hand in hand with the changes in the left periphery of the clause, in particular with the loss of the German strategy of left-dislocation, tor the Romance one, and how this is an explanation for the weakening of the V2--rule. 'What we postulate for the history of Cimbrian, according with the analysed texts, is a radical modification of the functional feature system in the syntactic areas meant to express and to realize specificity or discourse--bounded ness. More precisely, we assume that, with the expan-sion of CP, the specificity feature [+ d] in FINP got lost or became weak, whereas it arose or became strong in the CLP. More evidence in support of this explanation can be offered by analysing the evolution of the pronominal elements from morphonologically reduced forms to clitics, which involves the syntactic area between CP and IP.

36. lhis explanation tallies with what Linder (1987) discovered for subject ditic-doubling constructions in Rumantsch: lhey are subordinated to a restriction of definiteness/specificity. Cf. also Grewendorf (2002) and PuS (2005). 37· See Anagnostopoulou (1999: 787) which makes reference to Bel¥\-ick (1985).

362. Ermeneglldo Bidese

3·3

The intermediate area between the left periphery and the center of the clause structure from Cat.1602 to Bar.1906

As we have seen in Section 2, the personal object pronouns of Cimbrian underwent a modification of their syntactic position from the Wackernagelposition (right-adjacent to the finite verb in main clauses and to the complementizer in embedded clauses) to an 'adverbaf one, similar to the Romance clitics. As early as in the traditional description of the German clause (Satifeldanalyse "sentence field analysis"), many scholars have proposed to introduce particular pronominal positions in the very left periphery of the Mitteljeld. 'Ihanks to the theoretical proposal ofSportiche (1996), we are now capable of describing this syntactic area more appropriately, applying Sportiche's description to the diachrony of Cimbrian. For the first documented period of Cimbrian, i.e. for Cat.1602, we assume the head of the clitic phntse (Cr.O) to be silent or 0. In this phase, pronominal elements are still XPs, which means that they have to be moved from VP to a Spec-position, hence to the position we assume to be the Wackernagelposition, or SPECCLP. A~ morphonologically reduced forms, in this stage pronouns may lean against the finite verb in root contexts or against the complementizer in embedded clauses, at Spell- Out. The following structure (52) represents the syntax of personal object pronouns in the first documented periodofCimbrianlanguagehistory (:::: Cat.1602) (WP = Wackemagelposition):

___...__...__...__.....-....__....__.....__,...__.

(52)

FinP

Spec

Fin'

• "'i· ,'(' .....:--;~...

F1n'' •-.Fl.o, +<•,

-,

--------- --/--_.._________ . .

--- ',

..-.._....

ClP

Spec(= WP)

I

prlmil_unsi

Cl'

---------------

Cl0

I

--

( ... )

-------VP

--

I ti

-----------------------

The second diachronic phase is characterized by the coexistence of two grammars (cf. 2.2 above).38 With the subjunction az 'that: pronouns behave as in the first stage, probably because of the frequent use of such patterns. By contrast, with subordinate conjunctions other than az 'that: a fundamental innovation can be noticed in comparison with CaU602: In Cat.1813, object pronouns are realized enclitically to the finite verb. We have documented evidence in support of the hypothesis that these pronouns

38.

For this possibility in the linguistic theory ct: Kroch (1989), Pintzuk (1999) and

Fuf}

(2005).

Revisiting the Wackernagelposition 363

are no longer morphonologically reduced forms, but clitics. This means that they are now syntactic heads und must be realized in the head of the CLP that evidently underwent a process of grammaticalization from silent to active. In doing so Cat.1813's pronominal eli tics fill the function of referentiality operators, since this feature is no longer realized in the F.rN°, but now in CL0 (cf. structure 53): FinP

(53)

--

__.....----------. /

Spec

Fin'

/

/-------....._____ ClP

FinOHl /

Spec

__.......----------

0'

------~( ... ) "~'· V + Cl

•

VP

I

The finite verb no longer goes up to FrN° ( == V2-rule), but only to CL0 in order to license tl1e strong feature [+ d] in iliis head and to support the ditic lexically lvith a consequent incorporation of the eli tic into the verb (hence 'adverbal' position). This substantial S}'ntactic innovation becoming visible no sooner than in Cat.l813, expanded further in the later analysed phase of Cimbrian diachrony that has been analyzed here: ln Bar. J 906 those phenomena, representing the old German syntax, are in definite demise.

4

Conclusion

This paper has set out (a) to reconstruct, on the basis of original empirical data, how the Cimbrian syntax of object pronominals evolved ti·om their first documented stage in Cat.1602 to the threshold of the twentieth century (cf. above Section 2); and (b) to offer a theoretically based explanation for this development (cf. Section 3 above). The outcome ofSection2 shows that Cimbrian object pronouns have tmdergone a remarkable evolution: from morphonologically reduced forms, which were realized just as today's German personal pronouns in the Wackernageljmsition, to enclitics occupying an 'adverbal' position. In doing so, the syntax of Cimbrian object pronouns moved indisputably closer to that of the elitics ofNortl1 Italy's Romance dialects, but without ever converging totally with the status of the Romance exceptionless proclitics, since in

364 Ermeneglldo Bidese

any stage of Cimbrian history proclitics can also be shown to appear. 39 In Section 3, a theoretical connection between such evolution of the pronominal elements and other substantial transformations of Cimbrian syntax was established regarding both the left periphery of the clause- i.e. the change of the left-dislocation strategy and the weakening ofthe V2·· rule (cf. 3.1) and the structural center of the clause, i.e. the rise of clitic doubling (cf. 3.2). Based on Sportiche (1996), it was argued in 3.3 that the evolution of the pronominal forms was due to a process of grammaticalization that affected the intermediate area between the left periphery and the center of the clause structure, i.e. the clitic phrase. Whereas its head was silent in Cat.1602's stage, with morphonologically reduced pronouns realized in the Specifier-position (=the traditional Wackernagelposition), in the subsequent periods of Cat.1813 and in Bar.1906 the pronominal head become active and hence able to base-generate clitics. The movement of the finite verb to CL0 secures that the clitics are preserved lexically, and it explains the only possibility of enclisis in Cimbrian. Crucially, this formation of a class of enclitics went hand in hand with the loss of the strict German V2-rule. Indisputably, Cimbrian is a very marginal phenomenon in the panorama of the European languages, but thanks to the fact that it evolved in tension between German and Romance, the investigation into its diachrony can help clarify the basic structures and historical changes of either one.

References Abraham, Werner. 1997. The base structure of the German clause under discourse functional weight: Contentful functional categories vs. derivative functional categories. In German: Syntactic Problems Problematic Syntax, Werner Abral1am & Elly van Gelderen (eds), 11···42. Ttibingen: Niemeyer. Abrallam, Werner. 2005. Deutscfo.e Syntax im Sprachenvergleich. Grundlegung einer typologl..
39·

See also the contribution of Werner Abraham in this volume, and 'lbmesdli (2010).

Revisiting the Wackernagelposition 365 [1905]1976. Die deutsche Sprachir1sel Luserr1. Innsbruck: Wagner.- Reprint Wien: Verband der wissenschaftlichen Gesellschaften Osterreichs. Baragiola, .Aristide. 1906.ll tumulto delle donne di Roana per ii ponte (nei dialetto cimbro di Camporovere). Padova: Fratelli Salmin. Berwick, Robert C. 1985. The Acquisition ofSyntactic Know ledge, Cambridge MA: The MIT Press. den Besten, Hans. 1983. On the interaction of root transformation and lexi.cal feletive tules. In On the Formal Syntax of the Westgermania [LinguistikAktuell/Linguistics Today 3], Werner Abraham (ed.), 47--138 Amsterdam: John Benjamins. Bidese, Ermenegildo. 200·1. Die Zimbern und ihre Sprache: Geographische, historische und sprachwissenschaftlich relevante Aspekte. In Alte Sprache11. Beitriige zum Bremer Koiloquium tiber Aite Sprachen und Sprachstujim (Breme11, Sommersemester 2003), Thomas Stolz (ed.), 3---42. Bochum: Brockmeye1: Bidese, Ermenegildo. 2008. Die diadtronische Syntax des Zimbrischen. Ti.i.bingen: Narr. Bidese, Ermenegildo (ed.). 2010a Il cimbro negli studi di linguistica. Padova: Unipress. Bidese, Ermenegildo. 201 Ob. Alle fonti scritte del cimbro: la 1etteratura' cimbra come esempio di genesi d\ma tradizione scrittoria alloglotta. In Bidese 2010a, 61--85. Bidese, Ermenegildo & Tomaselli, Alessandra. 2005. Formen der Herausstellung und Verlust der V2--Restriktion in der Geschichte der zimbrischen Sprache. InDas Zimbrische zwischen Germanisch und Romanisch. Ermenegildo Bidese, James R Dow & Thomas Stolz (eds). 71--92. Bochum: Brockmeyer. Bidese, Ermenegildo & Tomaselli, Alessandra. 2007. Diachronic development in isolation: The loss ofV2 phenomena in Cimbrian. Linguistische Berichte 210: 209--228. Boeckx, Cedric & Grohmann, Kleanthes K. 200-1. Sub Move: Toward a unified account of scram-bling and D-Unking. In Periph,?ries: Syntactic Edges and their Effects, David Adger, Cecile de Cat & George Tsoulas (eds), 341--357. Dordrecht: Kluwer. Bosco, Haria. 1999. Christlike unt korze Dottritla: Un'analisi sintattica della lingua cimbra del XVI secolo. In Tesi di linguistica tedesca, Eva Maria Thi.ine & Alessandra Tomaselli (eds), 23-39. Padova: Unipress. Cardinaletti, Anna. I 999. Pronouns in Germanic and Romance languages: An oveni.ew. In van Riemsdi)k, 33-82. Cardinaletti, Anna & Starke, Michal I 999. The typology of structural deficiency. A case study of the three classes of pronouns. In van Rl.emsdijk, I 45-232. Chomsky, Noam. 2001. Derivation by phase. In Kim Hale: A Life in Language, Michael Kenstowi.cz (ed.), 1-52. Cambridge MA: The MIT Press. Cinque, Gugli.elmo. 1990. 'J.)tpes of.4.' Dependencies. Cambridge MA: The MIT Press. Fufi, Eric. 2005. The Rise of Agreement. A. Formal Approach to the Syntax and Grammaticalization of Verbal Inflection [Linguistik Aktuell/Linguistics Today 81]. Amsterdam: John Ben.iamins. Grewendorf, Gi.inther. 2002. Minimalistische Syntax. Ti.ibingen: Francke. Gi.orgi., Alessandra & Pianesi, Fabio. 1997. Ti:;nse a11d Aspect: Prom Semantics to Morphosynta_-.:. Oxford: OUP. Haegeman, Liliane. 2000. Negative preposing, negative inversion, and the split CP. In Negation 011d Polarity, Laurence R. Horn & Yasuhiko Kato (eds), 21-61. Oxford: OUP. Heller, Karin. 1988. Barocke Dichtung aus den 7 Gemeinden. Zimbrische Texte aus dem 17. und 18. .Tahrhu11dert. Wien: Verband der wissenschaftlichen Gesellschaften C)sterreichs. Kiparsky, PauL I 995. Indo- European 0 rigins of Germanic Syntax. In Clause Structure and Language Change, Adrian Battye & Ian Roberts (eds.), 140--169. Oxford: OUP. Bacher~ Josef.

366 Ermeneglldo Bidese Kranzmayer, Eberhard. 1981-1985. Laut- und Flexionslehre der deutschen zimbrischen Mundart: das sind die Mundarten in den 7 Vicentinischen Gemeinden, den 1.3 Veroneser Gemeinden und den deutschen Orten im Jrentinischen (mit Ausnahme des Fersentales und des Nonsberges). 2 voll, Maria Hornung (ed.). Wien: Verband der wissenschaftltchen Gescll~chaften Osterreichs. Kroch, Anthony. 1989. Reflexes of grammar in patterns oflanguage changes. La,1guage Variation and Cluznge 1: 199-244. Linda; Karl P. 1987. Grammatisc.he Utztersuc.hungen zur CharakterL~tik des Rdtoromanischen in Graubunden. 1tibingen: Nan: Lobbia, Nico & Bonato, Sergio. 1998. Il Potzte di Roana. Dez Dink vo' der Prucka. Roana: Istituto di Cultttra Cimbra. Meid, Wolfgang. 1985a. Der erste zlmbrische Katechismus CHRISTLIKE UNT KORZE DOTTRINA. Die zimbrische Version aus dem Jahre 1602 der DOTTRINA CHRISTIANA BREVE des Karditzais Bellar min in kritischer Ausgabe. Einleitung, italieniscr.er utzd zimbrischer Text, Obersetzung, Kommentar, Reproduktlonen. Innsbruck: Institut flir Sprachwissenschaft der Universitat Innsbruck. Meid, Wolfgang. 1985b. Der zweite zimbrische Katechismus DAR KL6ANE CATECHISMO VOR DEZ BELOSELAND. Die zimbrische Version aus dem Jahre 1813 und 1842 des PICCOLO CATECHISMO AD USO DEL REGNO D'ITALIA von 1807 in kritischer Ausgabe. Einleltung, italienisciler und zimbrischer Text, Obersetzung, Kommentar, Reproduktionen. Innsbruck: Institut fur Sprachwissenschaft der Universitat Innsbruck. & . Panieri, Luca. 2005. Ii cimbro del sette comuni. Raccolta di contributi storico lingulstlci. Camposampiero: Edizioni Del Noce. Panieri, Luca. 2008. La questione 'ctmbra' alla luce della linguistica diacronica. In La llngulstica germanica oggi. Bilanci e prospettive. A ttl del 34! Convegno associazione italiana filologia germanica. Claudia Hiindl & Chiara Benati (ed~). 275-304. Genova: EGIG. Panierl, Luca. 2010. I:identita linguistica ctmbra sotto la lente della filologia germanica In Bldese 2010a. 25-40. Panierl, Luca. Pedrazza, Monica. Nicolussi Hal2., Adelia et al. 2006. Bar lirnen z5chraiba un zo reda az be biar. Grammatica del cimbro di Luserna/Grammatik der zimbrischen Sprache von Lusern. Regione Autonoma Trentino-Alto Adlge- Autonome Region Trentino-Siidtiroli Istituto Cimbro - Kulturinstitut Lusern. Pill. Diana. 200 l. La dislocazione con rlpresa in cimbro: Sintassi romanza e sintassi germanica a confronto. Quaderni di lingue e ietterature 26: 111-129. Pintzuk, Susan. 1999. Phrase Structures in Competition: Variation and Change in Old English Word Order. New York NY: Garland. Poletto, Cecilia. 2001. The left-periphery ofV2-Rhaetoromance dialects: A new view on V2 and V3. In ~vntactic Microvariation, Sjef Barbiers, Leonie Cornips & Susanne van der Kleij (eds ). . Poletto, Cecilia & Tomaselli. Alessandra. 1995. Verso una definizione di elemento clitico. In Studi di grammatica tedesca e r_omparativa, Dolci, Roberto & Giuliana Giusti (eds), 159-224. Venezia: La Tipografica Poletto, Cecilia & Toma~elli, Alessandra. 1999. Towards a Typology of V2 [MS. XV. IGG, Siena] Poletto, Cecilia & 'lbma~elli, Alessandra. 2000. I:interazlone tra germanico e romanzo in due "is ole linguistid1e': Cimbro e ladino centrd.l.e a confronto. Inisole lingu~'tiche? Per un'analisi

Revisiting the Wackernagelposition 367

dei sistemi in contatto. i\tti del convegno di Sappada/Plodn (Beiluno), l-4 luglio 1999, Gianna Marcato (ed), 163-176. Padova: Unipress. Poletto, Cecilia & Tomaselli, Alessandra. 2009. Dle Syntax der Pronominalobjetrte und die Form des Parti.zips. Konservative Merkmale in der Sprachgeschkhte des Zimbrischen. In Perspektiven Drei. Akten der .3. J'agung Deutsche Sprachwissenschaft in Italien (Rom, 14.-16.2.2008), Claudia Di Meola, Livio Gaeta and .Antoni.e Hornung, et al. (eds). Frankfurt: Peter Lang. van Riemsdijk, Herik (ed.). 1999. Ciitics in the LMtguages ofEurope. Berlin: Mouton de Gruyter. Rizzi, Luigi.1982. Issues in Italian Syntax. Dordrecht: Foris. Rizzi, Luigi 1997. The fine structure of the left periphery. In Elements of Grammar, Liliane Haegeman (ed.), 281--337. Dordrecht: Kluwer. Roberts, Ian. 2004. The C--system in Brythonic Celtic languages, V2, and the EPP. In The Structure ofCP and IP, Luigi Rizzi (ed.), 297--328. Oxford: OUP. Schmeller, Johann Andreas. 1838. Ueber die sogenannten Cimbern der VII und XIII Communen auf den Venedischen Alpen und ihre Sprache. In Denksdtriften der baye1: Akademie der Wissenschaften 15 Abhandlungen der phUos.-phUol. Klasse 2, 555---708. Reprint in Schmeller, Johann Andreas. 1984. Die Cimbern der VII und XIII Communen rmd ihre Sprache, Richard J. Brunner (ed). Landshut: Curatorium Cimbricum Bavarense, 13---162. Schweizer, Bruno (ed.). 1939. Zimbrische Sprachreste, TeU 1: Texte aus Giazza (Dreizelm Gemeitldetl ob \lerona). Nach dem \loiksmunde aujgenommen und mit hochdeutscher Obersetzung herausgegeben. Halle: Niemeyer. Schweizer, Bnmo. 2008. Zlmbrische Gesamtgrammatik. Vergleichende Darstellung der zimbrischen Dialekte, James R. Dow (ed). Stuttgart: Steiner. Spmiiche, Dominique. 1996. Clitic constructions. In Phrase Structure and the Lexicon, Johan Rooryck & Laurie Z1ring (eds), 213---276. Dordrecht: Kluwer. Stefan, Barbara. 1998. Das zimbrische Marktgesprach mit Rechenexempel des D. Piermodesto Dalla Costa von 1763. In Wi.1rt, Text, Sprach.e und Kuitur. Festschrift fUr Hans Schmeja zum 65. Geburtstag, Peter Anreiter & Hermann M. C)lberg (eds.), 179-197. lnnsbruck: Institut fur Sprachwissenschaft. Stefan, Barbara. 2000. f..'ovena vun unzar liben Vraun. Die zimbrische Mariennovene des D. Giuseppe Strazzabosco mit Obersetzung und Kommentar. Innsbruck: lnstitut fur Sprachwissenschaft. "lbmaselli, Alessandra. 1990. La sintassi del verba fin ito nelle ling-ue germaniche. Padova: Unipress. "lbmaselli, Alessandra. 2010. La rilevanza del cimbro per gli studi di. sintassi nel quadro teo rico della grammatica generativa. In Bidese 2010a, 41-59. Tra"is, Lisa deMena. I 984. Parameters and Effects of Word Order Variation. Ph.D. dissertation, MI1: Tyroller, Hans. 2003. Grammatische Beschreibung des Zimbrischen von Lusern. Wiesbaden: Steiner. Wollstein-Leisten Angelika, Heilmann, Axel & Stepan, Peter et al. 1997. Deutsche Satzstruktur. Grundlagen der syntaktischen Analyse. 'Ilibi.ngen: Stauffenburg. Zifonun, Gisela, Hoffman, Ludger, Strecker, Bruno et al. I 997. Grammatik der deutschen Sprache. 3 vols. Bedin: de Gruyter. Zotti, Costantina. 1986. Bin 'zimbrisches' Sprachdellkmal vom Sudrand der Alpen. Die Brinnerungen der Costantina Zotti (1904-1980) aus Toballe in den Sieben Gemeinden. Zimbrischer 'Text mit neuhodtdeut~cher Obersetzung. Sachbearbeitung von Norbert i\·1umelter. Bo1.en: Landesverband fiir Heimatpflege in SiidtiroL

SECTION

5

Syntax III - The syntax of Pennsylvania German

Changes in frequency as a measure of language change Extraposition in Pennsylvania German* Gesche Westphal Fitch Universitiit Wien

Samples of Pennsylvania German (PG) are compared with samples of Standard German (SG) and Palatinate (PL) dialects. Analyses ofthe ti·equency ofthe occurrence of extraposition show that all dialect groups differ significantly from each other. PG has the highest rate of extraposition. Howeve1; the Pl. group also has a higher extraposition rate than SG. Concerning adverbials, speakers ofPG extrapose significantly more elements than the two comparison samples. SG and PL are not different from each other in this respect. A high frequency of event-related adverb placement that corresponds to an English surface order ('mirror order') and a low frequency of the order corresponding to tmmarked German in PG is presented. This frequency distribution is not found in the comparison samples. Crucially, the mirror orders do occur, abeit rarely, in PL. It is argued that PG displays frequency changes in which previously marked variants corresponding to English surface structure rise in frequency which is possibly facilitated by pre-existing tendencies in PL.

1.

Introduction

Pennsylvania German (PG) developed out of the dialects spoken by German immigrants to the US some 300 years ago, who mainly came from what is now the Palatinate, but also Switzerland, Alsace, Hessen, north Baden and \V'tirttemberg1• By 1825, PG

''Ihis report is a summary of my master's thesis submitted to the Preie Universitil.t Beilin. A special thanks to Tonjes Veenstra and Carol Pfaff for their guidance with this project. I would like to thank Tecumseh Fitch for his advice \\oith statistics, Mark Louden, Joseph Salmons, and participants other German Language and Immigration conference, Madison, ·wisconsin, for suggestions and comments. .All errors are of course my own. By 1970, more than one third of the population of Pennsylvania consisted of people of German ancestry (Louden 2006).

1.

372 Gesche Westphal Fitch

had become a distinct lanbruage. 1oday's speakers of Pennsylvania German are almost exclusively members of religious groups that are culturally distinct from mainstream America. While originally only a small fraction of Pennsylvania Gennan speakers were strongly religious (sectarian), most non-sectarian speakers have now shifted to :English, so that now, through "maintenance by inertia" (Louden 2003), virtually all PG speakers are sectarians. Speakers today are bilingual, that is, they are perfectly able to communicate and interact with the surrounding society in English, but at the same time, PG serves as a strong marker of cultural distinctness. Hence, unlike many other languages of immigrants that have disappeared within a few generations, PG offers a valuable opportunity to study effects of bilingualism over multiple generations. PG is spoken in a relatively stable bilingual setting. This makes it unusual, since the typical linguistic scenario in immigration shows a shift to the laTibruage of the new country (L2) within three generations. Such a quick shift generally leaves little time for structural influence of the immigrants' first language (Ll) by the L2. Influence of L1 structure on the L2 is obvious during transition as non-standard syntax. Examining PG for structural changes can potentially lead to insights into the cross--language compromise strategies that speakers utilize in a stable bilinbrual situation, where Ll and L2 are in contact over multiple generations, and changes are more likely to be bi-directional. As is well known, a core syntactic difference between English and German is the relative position of the verb (V) to the object (0). English has a clear VO constellation: [buy[ the computer]NP]VP' while German has an OV constellation: [[den Computer]NP kaufen]vp· In English, this applies to all clause types, while the situation is more complex in German: In embedded clauses, the finite verb or verb cluster follows the object and marks the right edge of the clause. In main clauses, the second constituent must be the finite verb (with the exceptions of imperatives and yes/no questions, where it is in first position). If the finite verb is an auxiliary, note that the untensed verb still follows the object: Du musst" den Computer kaufen (You must buy the computer). It has long been argued that the complementizer of the embedded clause and the finite verb in main clauses fill the same position. 1he word order in main clauses is taken to be a derived one, involving verb movement. Pennsylvania German has retained its core OV properties despite prolonged contact with English: the V2/l rule in main clauses is robust, as is the OV order in embedded clauses. However, it is argued here that the more peripheral properties of the OV constellation in PG may be more diverse than is often described in the literature. One e.xample of peripheral OV qualities that will be discussed extensively here is extraposition. Consider the examples below. :Extrapositions are marked in boldface throughout.

(I)

a.

Ich habe den Mann gesehen in der Stadt I have the man seen in the town 'I saw the man in town'

Extraposition in Pennsylvania German 373 b.

Ich habe den Mann in der Stadt gesehen I hm.•e the man in the town seen

In sentence (1b), the participle gesehen 'seen is the right boundary of the clause, that is, no material should occur after it. 'Ihe two verbal elements are sometimes called a bra£:ket (German: Vet·bklctmrner), a term that I also will use here.'Ihis is modelled after Ernst Drach's Stellungsfelder, originally published in 1940, which divides the clause into three fields separated by said verbal brackets (Drach 1963). Other useful terms derived from this mode! are Mitteijeld (middle field, the area between the brackets) and Naciifeld, (post--field, the area after the bracket). A bracket also is taken to be present in an embed·· ded clause, between the complementiser and the verbal elements at the end of the clause. (I)

c.

dass ich den M.ann in der Stadt gesehen habe tht:zt I the rmm in the town seen have '... that I saw the man in town"

Clauses in which words occur in the Naciifeld as in (I a), are regarded as substandard. Of course, this does not mean that such violations do not occur: ( 2)

So konnen sie

So can

Frieden schafien im

they peace

create

in.DAT

Land

country

'That is how they can create peace in this c.otmtry' ('heute: German TV news programme, 02.09.2006)

It is unclear what the reasons for e.xtraposition are; possible causes given in the extra·· position literature are wide-ranged: Archaic style, particularly high literary style, or dialectal influence (Hentschel & Weydt 1994). Weiss (1975) argues for unpreparedness as the major factor for extraposition in Standard German. HinterhOI:rl (2004) illus·· trates that in earlier stages of German, the right verbal boundary was by no means as impermeable as it is in today's prescriptive Standard German: extraposition occurred in half ofbrackets in Old High German and 80% of brackets in Middle High German. Looking at the glosses of the sentences containing extrapositions in Examples (la) and (2) above, there is a striking surface similarity to English word order. 'Ihe occurrence of extraposition in Pennsylvania German has of course not gone unnoticed: Also in terms of grammar, German .. speakers will recognize frequent occurrences of extraposition (Ausklammerung), whid1, while less productive in the written standard language, is widespread in spoken German. In Pennsylvania German, extraposition is not random but highly regular, applying to virtually all noncomplement prepositional phrases (PPs) but almost never to complements (other prepositional phrases and nominal objects). (Louden 2006)

'Ibis highly rebrular phenomenon of e.xtraposition is even known to occur in written form, as the following example from a Pennsylvania German letter cited in Costello (1986) illustrates.

374 Gesche Westphal Fitch

(3)

Wen du es liegen las e wcil If you it lie let a while 'Ifyou let it lie for a while'

(Costello 1986)

However, given that extraposition occurs in PG, the question becomes whether this is entirely internally motivated or can be attributed to language contact. Contemporary descriptions of German dialects, such as the Palatinate dialect (Green 1990) and the S·wiss German spoken in Bern (Hodler 1969), report e.xtraposition as a characteristic feature. A hypothesis to be pursued in this paper is that extraposition is an inherent part of the German dialects forming Pennsylvania German, but that its rise in frequency is accelerated by the existence of parallel surface structures in English and German. This idea was put forth by Reed (1979), though it has never, to my knowledge, been supported with empirical data: [... ] Pennsylvania German is dearly intluenced by word-structures and syntactic rules peculiar to English, sinc.e its speakers are bilingual and to some e.xt:ent also subject to the formal effects of written English. In such instances where the rules of "W"1'itten English contradict those of German, optional variations of spoken German, are, in this dialect, actually reinforced. (Reed 1979, 242)

A similar point has been made by Silva--Corvalan (1994) for Spanish--English bilinguals. She finds little evidence of a direct structural influence of English on the Spanish spoken by second- and third-generation immigrants, but a lot of evidence for a rise in structures that occur in English with a similar surface word order: The permeability of a gran1mar to foreign influence does not depend on its structural weaknesses but rather on the existence of superficially ( ... ) parallel structures in the languages in contact. (Silva-Corvah1n 1994, 93)

If parallel surface word orders exist, it is plausible that contact-induced language change would affect the frequem:y with which a certain structure is used, with parallel structures rising in trequenc.:y and non-parallel structures decreasing in frequenc.:y. An example Silva-Corvalan uses is the extension of the copula estar to contexts in which ser is usually used in standard varieties of Spanish. Both are copulas, but ser is (roughly) used for concepts that are not susceptible to change; changeable concepts take esta.r. Hence, the sentence Qu.e inteligente estas is an unusual thing to say, meaning 'How intelligent you are (today)'. In the Spanish varieties Silva-Corvahin examined, she found a rise of estar in the contexts that usually demand the use of ser: (4)

Pero yo estoy inteligente y muy guapo y note puedo tener to-do 'But I am (estar) intelligent' and very handsome and I can't have everything (Silva-Corvalan 1994, 93)

It would seem to be obvious that English, a language with only one copula, is helping to simplify the Spanish system with two. Note however, that the same falling frequency of ser is found in other varieties of Spanish that are not in contact with a single copula

Extraposition in Pennsylvania German 375 language but where normative influence of the standard variety is weaker, for example Mexican Spanish. The influence of English may thus be amounting to an acceleration of pre-existing internal tendencies. Henry (2002) also sees frequency changes as a major mechanism of language change in general: It has often been fotmd that where there are two alternants, one gradually declines in frequency, until there is a stage where it is no longer used. (Henry, 2002, 276).

1hinking of language change as a shift in usage frequencies rather than only focusing on new structures and features seems a promising way of discovering subtle changes in language usage. I will argue here that a rise in the frequency of extraposition is indeed present in Pennsylvania German and will present a comparison of speaker recordings from Pennsylvania German, Standard German and Palatinate dialects (the closest present-day approximation of the base dialects of Pennsylvania German). This analysis of the frequencies of extraposition in each variant, as well as analyzing adverbial syntax as a possible further indicator of ongoing lanbruage change, demon-strates the frequem:y shift under way in Pennsylvania German.

2.

Methods

Samples of spoken language from three groups were analy:r.ed: the Palatinate region (PL), Pennsylvania German (PG) and Standard German (SG). Six speakers were analyzed from each group, with 200 to 250 verb phrases for every speaker. I used the Deutsches Spracharchiv, an online data bank of transcriptions made available by the Institut fur Deutsche Sprache (IDS) in Mannheim. 'Ihe data bank, compris-ing several different individual corpora, is available on the Internet: http://dsav-oef[ ids-mannheim.de/DSAv. 'Ihe six speakers trom the Palatinate were deliberately chosen from throughout the region (i.e. Nordpfalz, Vi>rderpfalz and Siidpfalz) as there is no single dialect of the region that can be called the base dialect of Pennsylvania German. For Standard German, I used the Pfeffer corpus of the IDS, which originally served the purpose of documenting spoken varieties of contemporary and urban German and now offers good examples of varieties that are very close to Standard German. I transcribed recordings of six Pennsylvania German speakers myself from the Moellecken Collection with kind permission of the Mal!: Kade Institut, University of Madison, Wisconsin. Unless stated otherwise, all examples stem from these transcripts. Following Zahn (1991), I chose to use verb phrases as a basic unit for analysis. In a first step, the transcript was broken down into clauses, the main distinctions being made between (i) subordinate clauses and (ii) main clauses with (a) only a finite verb in second position or (b) with a finite verb in second position and a verbal element on

376 Gesche Westphal Fitch

the right clause boundary ("bracket" clauses). Main clauses with only a finite verb in second position were excluded from e.xtraposition analysis, as there is no overt right clause boundary to mark an extraposition. All statistical analyses were run in JMP 7 .0.1, a statistics programme from the SAS Institute (www.jmp.com).

3·

Results

'Ihe total sample size was 3,915 clauses (PG: 1,475, Palatinate 1,403, SG 1,037). Of these clauses, 1,272 were e.lCcluded as they merely had the finite verb in second position, i.e. no verbal right clause border. One hundred twenty clauses had more than one verb, so that there was a bmcket structure, however, there were no other words in the clause, so extraposition would not have been possible - these clauses were excluded as well. Finally, there were 15 clauses that were fragmented or otherwise unclear. The 2,508 remaining clauses had a clear bracket structure- 716 were embedded clauses with a complemeti1..er as the left bracket element, and 1,792 were main clauses with a purely verbal bracket structure. In all, there were 499 instances of linguistic material occurring after right clause borders. Discounting discourse markers, afterthoughts and elaborations, I found 392 instances of extraposition from V2 clauses, and 69 instances of extraposition from embedded clauses. 'lable l below shows the raw number of extrapositions for each group as well as the percentage of extraposition of all extraposable brackets. Table I. Extrapositions by dialect group Dialect group Pennsylvania German Palatinate Standard German

Nr. of extrapositions total

Nr. of bracket structures total

% Extrapositions of

235 176

935 965 608

25.2 18.1 8.2

50

all bracket strudures

Clearly, the Pennsylvania German group has the highest rates of extraposition compared to both Palatinate speakers and Standard German speakers. Nevertheless, verbal complements, subjects and objects are not affected by extraposition, in agree·· ment with observations by Louden (2006). Whether speakers would judge such extra·· positions as ungrammatical is, however, an open question. 1he e.xtraposition rate for Standard German is comparable to Zahn's (1991) study on extraposition in spoken German, where the rate was 10.7%. To test whether the dif· ferent rates between the groups is also statistically significant, I performed an analysis of variance (ANOVA) of the rates of extraposition. The results were highly significant

Extraposition in Pennsylvania German 377

(F(2,15): = 34.3, p < 0.0001). That is, the distribution of rates of extraposition is highly unlikely to be due to chance, and the hypothesis that this distribution is due to a genuine difference between the groups is more plausible. Likewise, the Kruskal-Wallis test, which makes no assumptions concerning the distribution of the data, found highly significant differences between all groups (p =0.0009). In order to answer the question where the differences found by the AN OVA stem from, a Tukey- Kramer test between the means of all groups was performed, revealing that all three groups differ significantly from each other. OAO~-------------------------,

•

035

!a:

•

030

II

•

~ 0.25

l

0.20

j

0.15

~

0.10

c

0.05 0.00

PG

PL

SG

dialed group

Figure 1. Comparison of mean rates of extraposltion

Here we have statistical evidence that there is a significant difference in extraposition rates- both compared to Standard German but also to Palatinate variants, for which extraposition is typical Unfortunately with this finding alone we have no good argument for the cause of these difference - it could be that Standard German has had a normative influence on PL speakers, leading to decreasing rates of extraposition, or it might be the PG group that has undergone change, or possibly even both. An indepth diachronic analysis would certainly be worthwhile, though far beyond the scope of this study. As an alternative, I looked at the distribution of event-related adverbials in the sample.

4

Why look at event-related adverbs in this context?

Event-related adverbs modify verbs, unlike sentence adverbs that modify a whole clause. Regularities in the ordering of these adverbials have been attributed to the

378 Gesche Westphal Fitch

headedness of the verb phrase; in German the unmarked ordering of event-related adverbials is Time Place Manner, while in English it is the mirror-image order Manner Place Time, see the Standard Gennan Example (5) below. Of course, permutations of this order are possible, especially if a particular element is stressed. For a structural explanation for these regularities, see Cinque (1999) or Hinterh6l71 (2002, 2004). German adverbials are placed after the tensed verb and the object in second position, but before the right verbal bracket structure (if there is one), while in English they are always placed after the verb and object. Example (5) below of Standard German and English illustrates the contrast. (5)

Klaus kommt heute mit der Bahn nach Hause. Klaus comes today with the train to home 'Klaus is coming home on the train today'

'Iheretore, two aspects of adverbial syntax are worth examining in Pennsylvania German: (1) the position of the adverbial elements relative to the verb, and also (2) the orderings of the adverbials relative to each other.

5·

Methods

The three most common adverb types in the data were Time, Place, and Direction. The mirror image orderings hold true for these three adverbials as well in SG: ( 6)

Ich gehe am go on.DAT I

Montag in Berlin zum Arzt Monday in Berlin to.DAT doctor

'I will go to the doctor in Berlin on Mondtl)l' All clauses containing any adverbials of this type were counted and analyzed. 'Ihere are three sites relative to the bracket that adverbs can occur.

6.

Results

In all 1,519 clauses contained at least one temporal, locative or directional adverb, that is, 39% of the data sample was included in the analysis of adverbs. 'Ihe Palatinate group contributed the highest number (700, or 46% of adverb clauses), the count for Pennsylvania German was 434 (29%) and 385 (25%) for Standard German. The distribution of the adverbs over the three possible locations in the clause is summarized in the table below.

Extraposition in Pennsylvania German 379

Table 2. Distribution of adverbs in the clause

B

M

A

BM

group

Clauses total (in%)

PG PL

434 (100) 160 (36.9) 159 (36.6) 65 (15) 700 (100) 268 (38.3) 242 (34.6) 38 (5.4) 385 (100) 98 (25.5) 243 (63.1) 13 (3.4)

SG

BA

26 (6.0) 97 (13.9) 27 (7)

13 (3.0) 33 (4.7) 2 (0.5)

MA

BMA

8 (1.8) 3 (0.7) 14 (2) 8 (1.1) 1 ( <0.3) 1 (<0.3)

While almost two-thirds of adverbs in Standard German are placed in the Mittelfeld exclusively, this is the case in the Pennsylvania German and Palatinate group for only about one-third of the adverbs (36.6% and 34.6% respectively). The comparison of the percentages of adverbs in the Mittelfeld (one-way AN OVA. F (2,15) = 17.54, p = 0.0001, Kruskal-Wallis p = 0.0078) showed that the differences between the groups were significant. A more detailed comparison between the groups with a Tukey- Kramer test showed that the Palatinate and Pennsylvania German groups did not differ from each other, but that both were different from Standard German.

0.9

*

0.8 0.7

'5 0.6

:!!

~ 0.5

:i

~ 0.4

c

ns

*

nn

"'

~ 0.3

0.2 0.1 0.0

PG

Pl

SG

Group

Figure 2. Comparison of adverb placement in the Mittelfeld

However, when conducting the same statistical tests on adverbs in the Nachfeld, it turns out here that Palatinate groups together with Standard German are both significantly different from Pennsylvania German (one-way AN OVA, F(2,15), = 10.50, p = 0.0014, Kruskal-Wallis p = 0.0035).

38o Gesche Westphal Fitch 0.25

*

ns

*

0.20

'6

:! -6., z

0.15

'$.

c.,

0.10

Cl

::::E

0.05

0.00

PG

PL Group

SG

Figure 3. Comparison of adverb placement in the Nachfeld

So while adverb placement in the Nachfeld does occur in the Palatinate and Standard German groups, the frequency is relatively low- in contrast, in Pennsylvania German there is a high frequency of adverbs to the right of the verbal bracket When focusing on the ordering of adverbials relative to each other, again interesting differences between the groups became evident. Filtering the data for occurrences of more than one of the adverbials Time, Place, and Direction in the Mittelfeld or Nachfeld of a clause led to a total of 96 clauses, 49 of which were produced by the Palatinate group, 30 by the speakers of Standard German and 19 by Pennsylvania German speakers. Most examples of the mirror order (analogous to the English order), were found in Pennsylvania German, with the speakers showing no evidence at all of disfluencies: (7)

Wei, now bin ich schon da drunne siebe joar Well, now am I already there down seven years

'Well, now I've been down there seven years already' SG: 'Nun bin ich schon sieben Jahre da unten' (8)

wel, ich seen eich in de karich de sundag Well, I see you in the church the Sunday

'Well, I see you in church on Sunday' SG: 1ch sehe euch am Sonntag in der Kirche'

However, the Palatinate group also has some, albeit infrequent examples of mirror orders:

Extraposition in Pennsylvania German 381 (9)

Habe ich im Saargebiet selbigmals geschafft Have I in.DAT Saargebiet then worked

'I worked in Saargebiet back then' SG: 'Ich habe damals im Saargebiet gearbeitet'

The fewest mirrored orders were found in Standard German, and some dearly resulted from unpreparedness, as the were accompanied by disfluencies, or prompted by the interviewer: (I 0)

Ich ging zweimal die Woche zum Turnverein [mhm]abends I went twice the week to.DAT gym club in the evening 'I went to gym dub twice a week' SG: 'Ich ging zweimal die Woche abends zum Turnverein'

Table3. Mirror and base orders by dialect group (in absolute numbers) Dialect

Mirror order

Base Order

8 7 4

11

19

40

47 30

PG PL SG

26

Total

As summarized in Table 3, Pennsylvania German has an overall lower number of adverb combinations than the comparison groups - but a high relative frequency of mirror orderings (42%). In contrast, Standard German and Palatinate have very similar base order frequencies of 87% and 85% respectively, and a low mirror order frequency (13 and 15%). 100 90 80

70 60

>

so 40

30 20 10 0

PG

Pl

SG

Dialect •

%Mirror Order

•

Figure 4. Frequencies (in%) for base and mirror orders

%Base Order

382 Gesche Westphal Fitch

7·

Summary

By analyzing frequencies ofboth extraposition and adverb placements in Pennsylvania German, a picture has emerged of an ongoing lanbruage change that shares some traits with Palatinate variants (high rates of extraposition, low rates of adverb placement in the Mitteifeld), but also has unique traits, such as the decrease of adverb orderings typical for Palatinate and Standard German. The frequen<..-y with which extraposition occurs was found to be significantly different for all groups while it was very low for Standard German, the rates for the Palatinate group were found to be higher and the rates for Pennsylvania German higher still. We can say that Pennsylvania German is indeed different from both Palatinate and Standard German, but if anything, more similar to the Palatinate when it comes to extraposition. Without a diachronic analysis, it is hard to say whether the rate of extraposition has fallen for Palatinate or whether Pennsylvania German has increased its frequency of extraposition. Furthermore, it is logically possible that Pennsylvania German has either expanded a natural tendency of its base dialects for extraposition, or that the bilingual coexistence of English and German has led to a rise of a variant that has a surface order more similar to English. In an attempt to find possible answers to this puzzle within the synchronic data available, an analysis of adverb placement and ordering was undertaken. 'lhe fre·· quency in extraposed adverbial elements was found to be statistically significant, that is, the Nachfeld in Pennsylvania German seems to have 'opened up' to adverbs more than in the other two sample groups. yet the low frequency of adverb placement in the Mittelfeld in Pennsylvania German is shared with the Palatinate dialects and differs from Standard Gem1an. Even though Pennsylvania German still retains V2/1 and OV word order, there appears to be a loosening of placement restrictions after the right clause boundary in this dialect. A decrease in the frequency of German adverb ordering compared to English-like adverb ordering was found. These developments may be accelerated by bilingual language contact, but they derive from structures and tendencies already present in base dialects, or, more precisely, are present in other present-day dialects that developed from the same base dialects. I hope this report has shown that questions about changes induced by language contact cannot be answered without taking into account changes in the frequencies of syntactic structures. Examining bilingual situations can be particularly enlightening when contemplating the language that a speaker develops, that is, the syntactic system that an individual uses to map concepts onto actual phonological oul]mt (and vice versa). In the presence of two languages speakers can show a lot of tlexibility, building syntactic bridges to lighten the cognitive load of maintaining and traversing between

Extraposition in Pennsylvania German 383

two syntactic systems. I have argued that changes in adverbial orderings and extrapositions thereof are such bridges between Pennsylvania German and English. Which principles constrain the building of these bridges are yet unclear, but it seems likely that peripheral, not core syntactic areas are affected. For example, in PG adverbials are extraposed often, while subjects and direct objects do not seem to be extraposable. Hence, the development of extraposition in PG is based on structural rules. Extending our knowledge of how bilingual strategies come about and what rules constrain them may enrich our understanding of how the human language faculty works in general, and in particular, offer insights into which structural properties in linguistic systems are susceptible to change when used by bilinguals.

References Cinque, Gugli.elmo. 1999. .Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford: OUP. Costello, John R. 1986. Diglossia at twilight: German and Pennsylvania 'Dutch' in the midnineteenth century. In Studies on the Languages and the Verbal Behavior of the Penm>ylvania Germans I. Zeitschrift f.ir Dialektologie und Linguistik, Werner Enninger (ed.), 1-24. W1esbaden: Franz Steiner. Drach, Erich. 1963. Grundgedanken der deutschen Satzlehre. Darmstadt: Wissenschaftliche Buchgesellschaft. Green, W.A.I. 1990. The dialects of the Palatinate (Das Pfalzische). In The Dialects of Modem German. A Linguistic Survey, Charles Vl Russ (ed.), 241-264. London: Routledge. Henry, Alison. 2002. Variation and syntactic theory. In The Handbook of Language \lariation and Change, J.K. Chambers, Peter Trudgill & Natalie Schilling-Estes (eds), 267-282. Oxford: BlackwelL Hentschel, Elke & Weydt, Harald. 1994. Handbuch der deutschen Grammatik. Berlin: Mouton deGruyter. Hodler, Werner. 1969. Berndeutsche Syntax. Bern: Franke Verlag. Hinterh0l71, Roland. 2002. Parametric variation and scrambling in Engli.sh. In Studies in Comparative Germanic Syntax, Walter Abraham & Jan- Wouter Zwart (ed.~), 131-151. Amsterdam: John Benjamins. Hinterholzl, Roland. 2004. Language change versus grammar change: vVhat diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar [Linguistik Aktuell/Linguistics Today 72], Eric FuB & Carola Trips (eds), 131---160. Amsterdam: John Benjamins. Louden, Mark L. 2003. Minority language 'maintenance by inertia': Pennsylvania German among nonsectarian speakers. In Standardfragen: soziolinguistische Per!>pektiven auf Sprachgeschichte, Sprachkontakt und Sprachvariation, Jannis K. Androutsopoulos & Evelyn Ziegler (eds), 121--137. Frankfurt: Peter Lang. Louden, Mark L. 2006. Pennsylvania German in the 21st century. In Sprachinselwelten. The

World of Language Islands. Entwicklung und Beschreibung der deutschen Sprachinseln am

384 Gesche Westphal Fitch

Anfang des 21. }ahrhunderts. The developmental stl.lges and the description of German language islands at the beginning of the 21st century, Nlna Berend & Ellsabeth Knlpf-Komlosi (eds), 89-107.1:rankfurt: Peter Lang. Reed, Carroll E. 1979. 'Ihe syntax of Pennsylvania German. Orbis 28: 242- 256. Silva-Corvalan, Carmen. 1994. La,1guage Conflict a,1d Change: Spanish in Los £1ngeles [Oxford Studies in Language Contact]. Oxford: Clarendon Press. Weiss, Andrea.~ E. 1975. Syntax spontaner Gesprache. Einfl.u1~ von Situation und Thema auf da.~ Sprachverhalten. Sprache der Gegenwart [Schriften des Instituts fur deutsche Sprache 31]. Diisseldort! Piidagogischer Verlag Schwann. Zahn, Giinther. 1991. Beobaciltungen zur Ausklammerung und Naclifeldbesetzung in gesprocilenem Deutsch [Erlanger Studien 29]. Erlangen: Palm und Enk.e.

From preposition to purposive to infinitival marker The Pennsylvania German fer ... zu construction* Kersti Borjars & Kate Burridge The University of Manchester/Monash University

The development from an allative preposition to a purposive marker to an infinitival marker is a common one cross-linguistically. In this paper, we look in some detail at this change in Pennsylvania German. We demonstrate that fer has completed this change in that it now occurs in a full range of complements. It has gone paired with a loss of the original infinitival marker zu. Functionally, the element has undergone a drastic change and given the loss of zu, one might expect that there would be equally far·- reaching structural changes. 'Ihe standard accotmt of this type of change is one of reanalysis, but in this paper, we arbrue that the change is likely not to have involved reanalysis.

Key words: Allative; grammaticalization; infinitival marker; Lexical-Func.tio nal Grammar; Pennsylvania German; pw·posive; reanalysis

1.

Introduction

In this paper, we will consider the development offer in Pennsylvania German (PG) from a benefactive preposition to a purposive to a general infinitival marker. We will provide data on the distribution offe1' and its interaction with zu in PG, in particular

*This paper has depended on the kindness and generous support of so marry members of the Mennonite community in Waterloo County. To all these people. we owe a special debt of gratitude. We have particularly depended on the generosity of NarK)' Martin and Melvin K and Nora Mattin and their families. Thanks are due also to Martin Durrell, Winifred Davies and Cecile Somers for help with data from various Germanic languages and non-standard varieties. We are grateful to Mark Louden for the helpful comments we received from him. KEB is also grateful for the financial support of the Economics and Social Science Research Council (grant no ROOO 23 7820) and also to the British Academy who provided the funding which made it possible for us to start this joint project.

386

Kersti Borjars & Kate Burridge the variety spoken by Plain PG speakers in Waterloo County, Canada (WCPG). 1 The use of the benefactive preposition has developed in similar ways in a number of varieties of Germanic, and we provide a brief overview of such varieties in order to place the WCPG changes in a Germanic context. Though the spread of fer... zu clauses from adjunct to subject and complement functions is common to a number of varieties of Germanic, in PG it has taken place at the same time as a related and very pervasive change in the language, namely the loss of the infinitival marker zu. As far as we are aware, this is unique within Germanic. Explanations for the development from preposition to purposive to infinitival marker have been proposed in the literature and we evaluate these in the light of the PG data, with special focus on the phenomenon of grammaticalization and the role of reanalysis within it We also consider how the change can best be modelled.

z. Germanic FOR ••• TO infinitivals in purposives and beyond It is not uncommon in Germanic languages for adverbial clauses expressing purpose

to be marked by the prepositionfor!ferlfiir in combination with an infinithm.l marker. In English, a purposivefor... to exists alongside the in order to construction as in (l) (contrary to what Huffines (1990: 103) claims in her discussion of the PG construction: "'for to' plus infinitive is an obsolete English construction to express purpose"). (1)

a. b.

Osc.ar gave Sarah a banana for her to eat when she got hungry. For him to win, he needs to play at his best.

Standard English only allows the for... to purposive when there is an overt subject.2 There are, however, non--standard varieties of English which have a for... to con-struction also in subjectless constructions, possibly the most well-known one being

In discussing these constructions, we will sometimes use fer ... zu as a cover term for the Germanic ones, even though the pronunciation and spelling varies greatly and even though the zu element may sometimes be missing. Similarly, we will use FOR ••• TO to refer more generally to the construction type. This means that the terminology is in principle ambiguous between any purposive construction with fUr/fer (as opposed to urn) and an actualfer... zu construction (as opposed to a fer construction. for instance), but we hope that it is always dear which meaning is intended. 1.

2. In fact, the example which Huffines (1990: 103) provides to show that the construction is obsolete in English sounds old-fashioned exactly because it lacks a subject: I'm goif1g to

Louisiana for to see my Suzia,la.

From preposition to purposive to infinitival marker 387

Ozark (Chomsky & Lasnik 1977; Chomsky 1981.: 281, 300, Koster & May 1982: 121.), but a similar phenomenon in Belfast English is better described (Henry 1992, 1995).3 Potentially of more interest to the topic of this paper because of its geographical proximity to WCPG dialects is Ottawa Valley English (OVE, Carroll 1983). However, we will later argue against influence from :English being a major factor in the development of the fet: .. zu construction in WCPG or any other variety of PG (pace Reed (1948), Costello (1978) and van Ness {1994: 436) and in partial agreement with Louden (1988)). In Standard German, the preposition um is used in the purposive construction, similarly, standard Dutch has om ... te and Frisian om ... ta. There are, however, anumber of non-standard varieties of Gennan which have a for... to purposive; for instance Palatine German (Pfiilzisch) (Henn 1980: 53), Frankish (Lockwood 1968: 154) and Swiss German (cf. Reed 1948: 244, Weber and Dieth (1964: 32)). 4 Another Germanic language which has a purposive FOR ••• TO construction is Luxembourgish (Bruch 1973: 102-104; Christophory 1974: 93). (2)

a.

Si kommen fir eis rosen ze maen, net fir eis 'T.e hellefen. they c.ome for us angry to make not for us to help Christophory (1974: 93) 'They come to make us angry, not to help us:

b.

Fir ze stierwe brauch een net vill, awer fir ze liewen. for to die need one not much but for to live 'One doesn't need much to die, but to live one does: Luxemburger Worterbuch

Luxembourgish is of special interest to us here because just like in PG, the fir... ze constructions is no longer confined to the purposive construction. In (3) we provide examples offir... ze as complements of adjective, noun and verb, respectively (Bruch 1973: 103). However, as the examples show, ze is still obligatory.

3· It is actually remarkable how little data from Ozark appears to be available, given the role it plays in the debate. Later discussions, like Chomsky (1981) and Koster and May (1982) refer only to Chom~ky and Lasnik (1977) for data, as does Carroll (198:') in her explicit comparison of OVE and Oz.a1'k. Still, Chom~ky and Lasnik give fewer than ten example se.mences ( 1977: 454, 500-1 ).

4· Both Lockwood (1968) and Ebert (1978: 31) trace this back to the French pour/pour a infinitival construction. This certainly could not be the explanation for all the varieties we are considering in this paper and, given what we shall have to say in this paper abotLt how common this pattern of development is typologically, we see no reason to invoke French, or as we shall see, English, influence.

388 Kersti Borjars & Kate Burridge (3)

3· 3.1

a.

Ed1 si frou frr ze raschten. I am happy for to rest 'I am happy to take a rest:

b.

fir z>•1essen:I give me something for to-eat 'Give me something to eat:

c.

En huet probeiert fir op de Bam ze klammen. he has tried for on the tree to climb 'He tried to climb the tree:

Gef mer eppes

Penm.-ylvania Germanfer... zu Background

The Mennonites, like many other Anabaptist groups, were persecuted in Europe and fled to the US, where they initially settled mainly in Pennsylvania. Some of this migration took place in the late 17th century, but the bulk of it occurred in the 18th century. From Pennsylvania, groups of Mennonites moved to other parts of the Americas and the Waterloo County Mennonites, on whose language we will focus below, migrated to Canada in tl1e beginning of the 19th century. The community with which we work consists of what we may refer to as "Plain Mennonites': or in Huffines' (1990) terms "sectarians". 1he term we use refers to the fact that these groups still adhere to rigid codes of clothing and thereby distinguish themselves from the modem Mennonites, whose style ofliving and church buildings fairly closely resembles other modem radically reformed Christian communities. The group defined by the tenn Plain Mennonites here includes both the Old Orders and the Markham Mennonites (sometimes called the "Modern Plain" because they have accommodated more to modern ways). vVe include the Markham Mennonites because with the Old Orders they define a community that is homogenous linguistically. All speakers we have recorded have grown up as monolinguals until the age of about six, when they started school. Even though Plain Mennonites run their own schools, the lanbruage of instruction is English, and for many children this will be the first time tl1at they are required to interact in English. Since the community does not have radio or television, or use the internet, young WCPG people's contact with English is more limited than for many young speakers of minority lanbruages in English speaking countries. All adult WCPG speakers are bilinbrual PG and English. 5 In what

;. A detailed discussion of the historical background of the Waterloo County Mennonites can be found in Fretz (1989).

From preposition to purposive to infinitival marker 389

follows, we will draw comparisons with other varieties of PG described in the literature. Since these are all spoken in the US, we will refer to them as USPG, though they are in fact studies of individual communities. 3.2

Early Pennsylvania German

As far back as there are written records of PG, a purposive je1~ .. zu construction c.an be found. This is not surprising, given that the variety of European German spoken in the area from which the Mennonites migrated, Palatine German, did have a fer... zu rather than an um ... zu purposive. 'Ihe oldest texts we have had at our disposal are from an anthologyofpG texts published in 1905, but compiled by 1875 (Horne 1905).6 The book contains a very brief section entitled 'Pennsylv<mia German Grammar' (1905: 221.---225), but clause structure is not discussed. The prose texts included in the book do, however, give ample examples of the fer... zu construction in purpose clauses. The status of the texts in Horne (1905 [1875]) is not entirely clear; some texts, tor instance, appear to be translations and in some cases we find expressions which are more likely to have belonged in High German than in the dialect spoken at the time. However, since neither the High German nor the English sentences corresponding to these examples would have contained a FOR ... TO construction, we can assume that the use of fer... zu in the texts reflect then current PG use. An example from Horne is provided in (4). 7 (4)

hut's g'num for de Englisha leit Un grawd so and exactly sud1 people has-it taken for the English d'umshtenda, wawra. we se zu led'r-a un'r to beat under the.circumstances as they were 'And it took exactly this kind of people to beat the English in the circumstances (Horne 1905: 136) as they were:

It should be pointed out here though that in the same text, there are also examples of um ... zu purpose clauses. This is most likely to have been the influence from High German.

6. We have a copy of the third edition of this work, which was published in 1905. The first edition was published in 1875 and the second edition in 1895. The preface to the second edition indicates that no essential changes have been made and the preface to the third edition only refers to 'many additional illustrative features' (Horne 1905: 6).

7· Note, PG spelling is not standardized and there is wild variation. We have retained the oliginal spelling in all examples throughout this paper. In our examples we adopt a system roughly based on that of Standard German (and found, for example, in pedagogical texts like Buffington & Barba 1965).

390 Kersti Borjars & Kate Burridge

'Ihe texts in Horne (1.905) also provide some examples ofj'er... zu in complement clauses, this includes use of fer without zu as in (5). It is interesting to note that these early examples of complement clauses withfer... (zu) clearly retain some of the purposive meaning- the situation expressed by the complement is presented as not yet reali1...ed, but with an expectation that it be reali1.ed in the future. 8 (5)

a.

Se wore ols so shlim fer danse. they were always so eager for dance 'They were always so eager to dance:

(Horne 1905: 130)

b.

Er hat die Insching all eiglauda fer kumma. he has the Indians all invited for come 'He invited all the Indians to come:

(Horne 1905: 201)

Keeping in mind that the texts in Horne (1905) were compiled by 1875, we can condude from these examples thatin this variety of USPG, the first steps in the expansion offer( ... zu) and the loss of zu were taken by the late 1800s. The earliest reliable examples of WCPG date from the 1930s and take the form of poetry written by the Old Order Mennonite Ben Sauder.9 Admittedly poetry is not ideal f{)r the study of syntactic change, but in this case it is all we have. Moreover, it is unlikely that considerations of rhythm or rhyme in Sauder's informal poetic style would greatly influence the use offer and zu. Generally the distribution in Sauder's poetry can be assumed to reflect then current usage. In Sauder's work, purpose is overw·helmingly expressed withfer... zu. However, as (6) illustrates, zu could sometimes be omitted. (6)

Dass der Vodda het ihm Gelt vasc.henkt, Fier so en groosi that the Dad has him money given for so a big Trip zu macha; Un mich nu griega far die Erwat schaffa. now get for the work do trip to make and me 'That his Dad gave him money to make such a big trip and then to hire me to do the work: (Ben Sauder's poetry)

Complement clauses of nouns, adjectives and verbs usually appear with single zu, though there are also e.xan1p les with fer ... (zu). Subject clauses, whether sentence initial or extraposed, are commonly introduced by fer as in (7).

An anonymous referee has pointed out to us that danse could potentially be argued to be a noun here. However, there is no evidence to suggest that such complements are any different when they follow fer than when they are preceded by zu. So, though tbe category status of danse is an interesting issue, it is not directly relevant to the historical change we are studying. We \\-ill return to the issue of the categorical status of the complement of fer in Section 4. 8.

9· Der Nachbar an de Schtroas, published 1955 by the Folklore Society.

From preposition to purposive to infinitival marker 391

(7)

a.

b.

ke Shand. is doch gar Fer Sauder zu haysa for Sauder to be-called is but absolutely no shame 'To be called Sauder is no shame at alL' ke groos:i Schand fa zafr:idda sei certainly no great shame for content be

E<J :iss gewiss it

is

im a kalda Land. in a cold country 'It's certainly no great shame to be content in a cold country:

(Ben Sauder's poetry)

We have not been able to find a pattern in the use ofjer~ .. zu,fa and zu in this data. In (8), the relevant clauses function as extra posed subjects, but all three options occur and presumably reflect the variation existing at this time. (8)

a.

Es is.;; net dawaert, tier die Schteddia zu saachen. for the townfolk to say is is not wotih 'It's not worth telling those not on the land:

b.

es v'lekht dawaert, fer des Ding browiera. Waerr for the thing try would.be it perhaps worth 'Ita perhaps be worth trying the thing?

c.

Es iss nimmi dawaert, Schulde zu mache. to make debts it is never worth 'It is never wotih making debts:

(Ben Sauder poems)

More recent collections of Canadian PG date from around the 1970s and include Buehler's autobiography in Pennsylvania German (1977), two short taped interviews of Old Order Mennonites (recorded in 1969; Karch & Moelleken 1977) and a collection of sentences written by Buehler. 10 All speakers here were elderly at the time of writing and recording and these texts cont1rm the picture provided by the earlier sources; in the purposive there is still considerable variation betweenfer... zu, zu and fer clauses. There are e..xamples of jet~ .. (zu) in subject clauses and now also as complements of adjectives, nouns and verbs. We can conclude then that in \VCPG, as in USPG,jer as a clause introducer had started to spread from purposive to subject and complement clauses at this stage. Given that this construction has developed in such similar ways in the two varieties, it is not unlikely that this spread had started before the Waterloo County Mennonites moved north. However, in principle it could be two independent developments.

'Ihe first comprises a collection of PG sentences compiled by Alan Buehler, a former Old Order Mennonite (unpublished manuscript).

10.

392. Kersti Borjars & Kate Burridge

The data also shows that even though fer could occur in a clause without zu, there was a lot of variation and zu was in regular use certainly into the 1970s. Pedagogical grammars of (US)PG published some decades later than Horne and Sauder, such as Frey (1981 [1942]), Bu.ffingtonand Barba (1965) and Haag (1982) mention the fer(. .. zu) construction in purpose clauses, but do not discuss any extended use. Nouns and adjectives are described as taking the infinitive with zu as their complements, as are verbs, with the exception of a set of verbs (inc.luding the modals) which take the infinitive without zu. When discussing complementation, Buffington and Barba (1965) do not mention the fer(. .. zu) construction and fer is not included in the list of subordinating conjunctions (1965: 91). However, the texts used in Buffington and Barba do contain some examples which show that fer(. .. zu) had started to spread beyond adverbial clauses to other clause types also in the varieties that they describe. Nonetheless, the fact that the wider use has not yet entered grammatical descriptions suggests that it was still sporadic. 3·3

The distribution offer(... zu) clauses in contemporary (WC)PG

With respect to modern USPG, Louden (1988, 1997) provides a thorough discussion of the distribution of the fer( ... zu) and states that "'fer' is generalized in all contexts where overt comp lementizers are required" (1988: 209). Van Ness's sketch of PG grammar describes fer as a general complementizer (1994: 436). Both authors also state that in USPG, it is only older speakers who show remnants of the earlier variation between fer... zu and zu. Huffines (1.990) carried out a study of the use of the fer(... zu) construction in different contexts by different sub-groups ofPG speakers in Pennsylvania. According to her results, there is still some use offer... zu and zu, but only amongst the non-sectarian groups, which include some non-native speakers ofPG. As we shall see, the picture is strikingly similar in WCPG, so that there appears to be little variation across Canadian and US Plain PG. There are, however, some interesting minor differences in distribution, but these will not be relevant to our argument and will not be discussed here. Our corpus ofWCPG comprises conversations, stories and elicited data from the late 1980s until the present time.l 1 The examples that follow, unless otherwise specified, are from naturally occurring data; elicited sentences are always indicated. What is striking is that the data do not contain a single example of a construction involving both fer and zu, neither in purposives nor in subject or complement clauses. When confronted with fu... zu sentences like those reported above from the older literature,

n. An anonymous referee asked us to quantify the corpus. Howeve1; since our data is not organised in one format in one database we cannot specify the size of the corpus in any useful way.

From preposition to purposive to infinitival marker 393

our language consultants find them contrived- at best old fashioned or extremely high style. In contemporary WCPG non-finite adverbial, subject and complement clauses typically appear with either bare infinitive or single fer, though as we shall see, older speakers still use zu on its own in certain non-finite environments. We will return to this issue in Section 3.5. Adjuncts of purpose are introduced by fer, as in (9). 12 (9)

a.

Een owed hot en arme mann en schupkarrich g'numme one evening has a poor man a wheel-barrow taken fer zu erne nochbar blatz geh fer riewe holle. for to a neighbour place go for turnips fetch 'One evening a poor man took a wheel barrow to go to a neighbouring farm to fetch turnips:

b.

waere. Ich nemm die schuh zum schumacher fer g'fixt: I take the shoes to.the cobbler fer mended be 'I take the shoes to the cobbler (in order for them) to be mended:

Natuarlly occurring fer... (zu) constructions with a subject are very rare, but they are sometimes produced in response to specific translations tasks. vVe will not discuss this further here, but just point out that this makes PG fer strikingly different in distribution from the Present Day English for in for... to constructions, which makes it less likely that the spread offer in PG is the result of contact with English. 1his is an issue we will discuss further in Section 4. Fer clauses are common in subject position, particularly as extraposed subjects, as in (10). (10)

a.

Fer datt anner laafe waer sadde dumm. for there towards walk would-be sort-of stupid 'To walk there would be sort of stupid:

b.

Es waer hatt uff ien fer uns sene gee. it was hard on him for us see go 'It was hard on him to see us go:

(elicited)

They also occur as the complement of an adjective and a noun, as in (11 ). (11)

a.

Ich bin verschtaunt fer dich sene do. am astonished tor you see here I 'I am astonished to see you here:

b.

Sie hat e muschde fer e track mache. she had a pattern for a dress make 'She had a pattern to make a dress from:

12. The zu \V"h.ich follows the first instance offer in (9a) is not part of the fer... zu construction, but forms part of a PP zu erne nochbar blatz.

394 Kersti Borjars & Kate Burridge

Fer clauses also occur as the complements of verbs. In fact, they appear to have spread to any type of verb that can take a non-finite complement (except modals), though with some verbs fer is used only under specific contextual circumstances. Subject and object control verbs withfe1" clauses are provided in (l2a) and (l2b ), respectively. (12)

a.

No hot sic so hatt browiert fer alles sauwer halde. for everything clean keep then has she so hard tried 'Then she tried so hard to keep everything clean:

b.

Ich hab ien verschwetzt fer geh. I have him persuaded for go 'I persuaded him to go:

(elicited)

Sentences of the type usually referred to as raising constructions also involve fer clauses. In ( 13a ), we have an example of a Subject Raising clause and in ( 13b ), we find an example of Object Raising. (13)

a.

Er scheint fer en ehrliche man sei. he seems for an honest man be 'He seems to be an honest man:

(elicited)

b.

Ich hab ien als ghalde fer e bissel arm sei. for a little stupid be have him always held 'fve always considered him to be a little bit dense:

(elicited)

It should be pointed out here that some older speakers would use zu rather than

fer in (13a). We will return to issues of how the context may influence the choice of complement type in Section 3.5 and how fer complementation has spread through semantic classes of predicate in Section 3.4.1. 3·4

Historical development

From benefactive to infinitival marker Fe1"... zu is the marker of purposive in the German dialects of the Palatinate from

3·4·1

which PG has developed. As would be expected there is evidence that PG also had a fer... (zu) purposive as far back as we have data. Even in the earliest preserved texts, there is evidence offer... (zu) having spread to other constructions, including clausal subjects and complements of adjectives, nouns and verbs. However, given that this has not happened in the original German dialects of the Palatinate, we can assume that fer as an introducer of clauses first appeared in purposive clauses and that the wider distribution has developed from there. Additional evidence for this chronology is plentiful. In earlier stages of German, the allative preposition zu had come to be used as a marker of purpose, it was certainly used in this way in Old High German (see

From preposition to purposive to infinitival marker 395

Haspelmath 1989 and references there). By the stage of Middle High German, zu. was spreading to complement clauses. At some stage, it had lost its purposive meaning and a fresh marker was required to distinguish the purposive from other clauses. In most dialects, um took on this role and this use remains in Standard German today. In the dialects from which PG developed, the benefactive preposition fer was used instead. 1he development from a benefactive combining with a noun phrase to a purposive marker combining with a clause is a common one typologically (see examples in Haspelmath 1.989: 293···5). 'Ihis change is commonly attributed to reanalysis, a pro·· cess by which the structure of a string is reinterpreted as illustrated in (14). (14)

[E kind waar in de schtor g'schickt child was in the store sent

a a.

[fer rode

peffer]

[zu kaate]

b.

[fer [rode peffer zu kaafe]] for red pepper to buy ~child was sent to the store for red pepper, to buy (it)/to buy red pepper~

It should be pointed out that a semantically, rather than structurally, motivated anal· ysis is also plausible. The semantic development involved is natural: a purpose can be viewed in essence as an action carried out to benefit someone, self or other - the notion of goal underlies both. On this view, the structural change would be motivated by the semantic development. In the case of the benefactive, the goal is an individual or a set of individuals who benefit and hence a benefactive combines ·with a noun phrase. A purposive is aimed at bringing a situation about and hence it combines with a clausal unit. A second stage in the development that leads to the current distribution in PG is that fer comes to introduce clauses which function as subjects (many of which are extraposed) and as complements of adjectives, nouns and verbs. We then have three stages in the development, as in (15). (IS)

a. b. c.

fer+ 'NP benefactive fer + clause purposive fer+ clause general infinitival

There was a stage in the predecessor of PG in which only (lSa) existed. 1here was then a stage at which (lSa) and (15b) existed, but not (15c). 1his is the situation still in the dialects from which PG originated. In current PG all options in (15) exist side by side. Historically, we may then want to posit an explanation within which (lSc) developed from (15b), which in turn developed from (lSa). An alternative would be to assume that (lSc) developed directly from (lSa), which of course still remains

396 Kersti Borjars & Kate Burridge

in the lanbruage today. 'Ihough the latter alternative seems less plausible to us, this would in fact appear to be what is assumed under the traditional reanalysis approach sketched in (16) (cf: Jespersen 1922; Ebert 1978: 12-13; Harris & Campbell1995: 62; Los 2005). Even Haspelmath, who rejects the traditional claim that reanalysis plays a crucial role in grammaticali7.ation, gives the development offor in English as an example of reanalysis without grammaticalization (1998: 324-5). (16)

a. b. c.

[to get a new job] is nice [for him] [it is nice [for him] [to get a new job]] [it is nice [for [him to get a new job]]]

A reanalysis account f{)r the development of (l6c) makes a direct connection between the prepositional use, (l.Sa), and the complement use, (lSc), with no refer·· ence to the purposive use which historically precedes the complement use. We are not excluding the possibility that the structural connection between ( 15a) and (15c) may have had some role to play in the development of fer as a general infinitival marker. However, for a number of reasons, this strikes us as a less likely path of development than an alternative in which the infinitival develops from the purposive. For one, it would involve the preposition twice developing into a complementizer, once with an accompanying semantic shift from benefactive to purposive and once with a shift from benefactive to basic infinitival. Furthermore, an environment in which an infinitival subject, such as to get a new job in (16), has been e.xtraposed to follow a benefactive preposition phrase modifying an adjective is crucial to the account in ( 16). Hence this account would also predict that jer... zu complement clauses would first appear with adjectives. In the earliest PG texts, we already find a full range of infinitivalfer... zu clauses in complements and as subjects and so we do not have firm evidence of the chronology of events. Yet, if ( 16) were to explain the development in PG, you would expect examples like (17) to be plentiful and yet this is one of the few we have and it is elicited. (17)

Es is ungweenlich fer de John harti Bkher zu lese. it is unusual for the John hard books to read 'It's tmusual. for John to read difficult books:

Though the distribution of for ... to clauses in English is different fi:om that of jer... zu in PG, the early development can be assumed to have been similar in that all three stages in (15) have existed. English offers more promising data to decide on whether the infinitival use developed fi:om the prepositional or the purposive use since the change took place during the Middle English period for which ample textual material remains. Indeed, de Smedt (2007, 201.0) shows that the early instances of the infinitival use do not appear in the environment required for ( 16) to be the correct explanation. Instead he shows that an account in which the infinitival jot'

From preposition to purposive to infinitival marker 397

developed from the purposive for (the 'reinforcingj(1r' in his terms) does make the correct predictions. We would then hold that the most likely development for PG is one in which the benefactive developed into the purposive from which, in turn, the infinitival use developed. The second phase of this development can be assumed to have started in environments in which there may be some uncertainty as to whether the fer clause was adverbial and truly had a purposive meaning or whether it could be construed as a complement of the verb. An example would be (18). In Present Day English such contexts could be disambiguated through the use of in order to. (18)

Sie hen hot gaschafft, fa Sacha inventa. 'They worked hard in order to invent thingsiThey worked hard inventing (Ben Sauder's poetry) things:

C'.Amtrary to de Smedt (2010), we do not believe that the prepositional traits that remain in the English infinitival for is due to essentially accidental similarity in form and dis·· t:ribution, but to that which Hopper (1991) terms 'persistence'. 1he purposive marker, the reinforcing for in De Smedt's terms, derived from the preposition and some of its original prepositional properties glimmer through in its infinitival use. 'Ihis is clearer in English than in PG. At a more general level, we then assume that a reanalysis such as that sketched in ( 16) did not play a crucial role in the development of the PG infinitival marker and hence that this is a form of grammaticalization without reanalysis. This is the general point made by Haspelmath (1998), though at a more speciiic level he would appear to assume that reanalysis is a correct analysis of the development of the English for (1998: 324-5). We would then argue that fer has gone through the three stages in (15) in turn. A more subtle division of the semantic changes that elements may undergo when developing from benefactive to purposive and beyond is provided by Haspelmath (1989: 298-9). This is represented in (19). (19) ~

benefactive irrealis potential

~ ~

purposive realis non-factive

~ ~

irrealis directive realis factive

~

We have already discussed the development from benefactive to purposive. 'Ihe pur·· posive refers to an intended future realization. Irrealis directive describes the modality of the complements of manipulative (ask, order) and desiderative (want, prefer) verbs; like the purposive, there is a clear future expectation. Irrealis potential is the modality of predicates such as be possible, be necessary and to evaluative predicates such as funny or interesting. Here the realization is not specifically associated with the future, but as potentially occurring any time. Realis non·· factive refers to the meaning of complements of verbs of thinking and speaking. With these predicates, the situation is presented not as potential, but as real, though without speaker commitment

398 Kersti Borjars & Kate Burridge

to its truth. Realis factive, finally, refers to the complement of verbs of cognition and evaluative predicates with a factive interpretation. Haspelmath presents this as a path of grammaticalization and an element which has gone through all stages of this path has desemanticized fully into an infinitival marker. 'Ihe examples in (20) from WCPG show that fer can occur in all these types of predicates and hence has developed from a dedicated purposive marker to the fully generalized infinitival marker. Semantically, fer has then undergone a radical shift. (20)

Irrealis directive: a.

Sie ecschpect dich fer kumme. she expects you for c.ome 'She expects you to c.ome:

Irrealis potential: b.

able fer kumme. Sie hot g'saagt sie waer she would.be able for c.ome she has said 'She said she would be able to come:

Realis non-factive:

c.

Ich hab gClenkt fer in schteddel gee. I have thought for intown go 'I thought of going into town:

Real is factive:

d.

Ich bin versdttaunt fer dkh sene do. I am astonished for you see there 'I am astonished to see you there:

The loss of zu There is evidence that zu was beginning to disappear in certain environments in PG already in our earliest sources from the late 19th century. Buffington and Barba (1965: 79) state "The tendency among present-day speakers and writers of PG (here American KEB/KB] is to omit the zu more and more frequentl~ 13 The early evidence ofWCPG that we have at our disposal also shows some loss of zu. Ben Sauder's poetry and Alan Buehler's writing contain many examples of bare infinitives with verbs normally demanding a zu infinitive. However, we still find fer... ztt clauses and other

3.4.2

Horne's 1875 manual also has a number of examples showing a surprising absence of any infinitive marking.

13.

Er was ready ahead geh mit der bisness. he was ready ahead go with the business 'He was ready to go ahead with the business'

(Horne 1910/1875: 202)

From preposition to purposive to infinitival marker 399

zu complements in general use into the 1970s. As previously mentioned, in WCPG and other varieties of PG fer now occurs in all constructions except as complement of modal verbs and zu is completely lost amongst younger speakers. This has been a remarkably fast change; it has all happened in about 200 to 300 years. We can compare this to the spread of the to infinitive in English. The to infinitive also had a purposive meaning, but appears to have spread to complement functions as early as the early Old English period (e.g. Miller 2002: 187-189). During the Middle English period the directional force weakened and the use of to infinitives increased. But the change was slow even into the 1500s to was still competing with the bare infinitive (although showing clear signs of winning). As Visser (1963-73: §901) concludes, "it took a long time for the particle to to be reduced fi'om a preposition expressing motion, direction ... to a semantically empty sandhi form, function-ing as a mere sign of the infinitive''. In contrast, the final stages of the loss of zu have been very rapid- in only 30 years it has gone from being in productive use to be being completely lost. 3·5

Remnants of zu and variation in complementation types

Given the loss ofjer... zu, there are then in principle three options for non-finite complementation in WCPG, as in other forms of PG: (i) bare infinitive, (ii) zu-infinitive and (iii) fer--infinitive. 'Ihere are certain fixed expressions in which all speakers use only zu; like nix zu mache, nix zu duh 'nothing to do', but only the older generations use zu productively. There are construction types in which older speakers will use only zu, but younger speakers prefer fer. An example is non-finite wh-const:ructions, as in (21), another is with raising verbs like scheinen 'seem'. (21)

a.

Sie hot nimmi gewisst was zu mat:he; sie waar she was she has never known what to do end vun ierem bendel. am cord at.the end of her 'She never knew what to do; she was at the end of her tether:

b.

Er scheint en ehrliche man zu sei. he seems an honest man to be 'He seems to be an honest man:

(elicited)

The bare infinitive is the only possibility with modal verbs for all speakers, but it is also used with verbs of movement or perception, where there is an implication of simultaneity, as in (22). (22)

kind ghaert ("fer/"zu) heile. Ich hab's c.ry have.the child heard I 'I heard the child crying:

(elicited)

400

Kersti Borjars & Kate Burridge For older speakers, who have both zu and fer as an introducer of non-finite dauses, there are semantic associations ~ith the two. Differences are not always apparent when taken out of context, but a particular context can influence the choice in a significant way. For example, when presented with the simple sentenc.e 'I have thirty cows to mill(, all these speakers agreed that either zu or fer was possible. However, when the speakers were provided with more context, clear preferences emerged. If a person was assumed to be in the middle of milking already and was explaining that they had thirty more cows to milk betore they were done, then the speaker overw·helmingly produced the sentence in (23a). If, on the other hand, the person had just met someone for the first time and was explaining the nature of farm life, including the fact that there were thirty cows to be milked every day, then the speaker was more likely to produce (23b ). (23)

a.

Id1 hab

b.

lch hab thirty kieh fer melge. I have thirty COWS toifor milk 'I have thirty cows to milk:

thirty kieh zu

melge. (elicited)

'Ihe event of milking the cows in the first scenario is associated with the immediate future - some cows had just been milked and another thirty were about to be. The orientation in the second scenario, however, is less time specific; the speaker is making a more general statement. Both zu. and fer are associated with future, as opposed to the simultaneity associated with the bare infinitive, but with zu the association is more direct. The examples in (24) with the different preferred markers illustrate this. It is interesting to note that the English translation of (24b) also requires for. 14 (24)

a.

Sie verwart en brief. she expects a letter 'She expects a letter'

b.

Sic hofft fer en brief. she hopes for a letter 'She hopes for a letter'

If we consider the original meaning of the two elements, this semantic tendency is not surprising: zu is in origin an allative preposition where the meaning direction in space has been extended to direction in time; fer as an infinitival marker has developed

14. 'Ihe difference between zu and fer for these speakers of WCPG can be related to that between to and for to in English, as argued for by Wierzbicka (1988: 111-132). She maintains that, whereas both to and for to are future oriented, to has much firmer expectations of effectiveness: "for-to versions sound more helpless and less confident" (1988: 120). See also Jespersen 1909-42:5, Bresnan 1979 and Dixon 2005.

From preposition to purposive to infinitival marker

from a benefactive via a purposive. Though purpose implies some degree of futurity, it is not as direct- it is a consequence of the fact that a purpose may be realized in the future. In a system which has both zu and fer, we would expect zu to be associated with direci future andjerwith a more general future. 'Ihis kind of persistence in meaning is very common in grammaticali7..ation (e.g. Hopper 1991). For speakers who have lost zu, there are only two options: bare infinitive or fer constructions. Given what we have just said about the remnant meaning offer, we would expect it to have an association of futurity absent in the bare infinitive. Furthermore, we might expect it to have taken over a more general future meaning in a system that does not also contain zu. This is indeed what we find. Verbs for which there is a clear implication of simultaneity between the higher and the lower verb have a bare infinitival complement. Examples are perception verbs, as in (22) above. Verbs whose meaning imply futurity with respect to the lower verb take fer, as in (25). (25)

a.

Ich hab ien verschwetzt fer geh. have him persuaded for go I 'I persuaded him to go:

b.

Sie ecschpect dich fer kumme frie you for come early she expects 'She expects you to come early:

(elicited)

There are few environments in which there is a choice between a bare infinitive and a fer infinitive. Some verbs generally take a bare infinitive, but a fer complement can be used if there is a context which removes the simultaneity. The verb gieiche 'to like' prefers a bare infinitive and the enjoyment can be said to coincide with the activity, as in (26a). 15 However in (26b ), the context implies futurity between the liking and the meaning of the lower verb. In this case, fer is possible, though not obligatory. 16 As the idiomatic translation shows, this parallels the English alternation between ··ing and to-infinitives. (26)

15.

a.

schwetze. kh gleich zu leit to people talk like 'I like talking to people:

b.

Deedscht du gleiche fer en teacher sei een daag? for a teacher be one day you like wmild 'Would you like to be a teacher one day?'

(elicited)

The zu here is a preposition which introduces the PP zu leit.

According to Costello (1978: 39), Southeastern Pennsylvania German generally uses fer with gleiche.

16.

401

401

Kersti Borjars & Kate Burridge There are also other, more subtle semantic differences which influence the choice between bare infinitive and fer infinitive, such as self- versus other-orientation and the degree of confidence the speaker expresses with respect to the lower clause. Some of these effects are similar to those noticed for English by Wierzbicka ( 1988) and an explanation along similar lines appears appealing. However, this is not the place to go into further details. As Louden (1988: 210-2) points out. the meaning of PG bare infinitive corresponds closely to that often ascribed to the English gerund complement. Wierzbicka (1.988: 59---97), in her semantic account of English complementation, describes --ing complements as having "simultaneous orientation': which is exactly the way we have described the bare infinitive in WCPGP Similarly, the English translation ofPG examples with fer will gener.Uly involve a to--infinitive. To Louden (1988), this is evidence of influence of English on PG. He argues that the similarity in distribution between the two forms in the two languages has been 'the catalyst for the system-internal change we find when we compare PPG and CPG: where C(ommon)PG is assumed to be the ancestor of P(lain)PG. We would be more cautious with positing any kind of influence here. In both systems, there is evidence of a particular form showing a remnant of its original meaning and of this remnant meaning to some extent determining its distribution. This is a common phenomenon in grammaticalization. The explanation for the similarity in distribution between PG bare infinitive and English -ing, on the one hand, and PG fer and English to, on the other, may just lie in the fact that the develop-ment of the particular fonns within the system have been following common paths of grammaticalization. 3.6

Structural changes

So far, we have seen that fer has undergone a radical change in meaning and that zu has essentially been lost. We will now turn to consider potential structural changes to fer. Under relatively standard assumptions about clausal structure, a functional element can be associated with any of three levels of the clause; with the clause as a whole; with the inflectional system; or it can be part of the verbal complex. 18 'Ihe associated

17. See Miller (2002: Ch2) for a discussion of infinitives and -ing forms in the history of English. As he points out, bare infinitivals in Old English tended to have a progressive aspectual meaning.

'Ihis view of clause structure developed from Chomsky (1986), but is now also current in other frameworks (e.g. Lexical-Functional Grammar, Bresnan 2001 and Head-driven Phrase Structme Grammar, Sag & Wasow 1999). Within the Chomskyan tradition, work by Riui (1997) and Cinque (1999) has led to a proliferation of clausal functional categories, which is irrelevant to our purposes. The c-structure representations in (27) and (36) are thus 18.

From preposition to purposive to infinitival marker 403

structure is as in (27). We have used the traditional categories C(omplementizer) and !(inflection) here, but we will return to the categorial status offer below. (27)

CP

~

C

IP

~

I

VP

In these terms, it seems clear that in the original fer... zu construction, jer occurs in C- it is associated with the full clause - and zu is found in I or below - it is associated with the verb phrase or possibly it is part ofV.l 9 1he question is then: as fer has taken over the functions of zu, has it also taken over its struch1ral position, i.e. has fer been "lowered" to the I position or to the actual verbal complex? Examples such as (28a) - elicited from an older WCPG speaker - can be used to argue that the original zu occurred inside the verbal complex. The corresponding (28b) which is ungrammatical for all speakers shows that fer does not have an identical distribution to the original zu. In terms of the tree in (27), the question is then whether fer in current WCPG appears in C or in I. (28)

a.

Es is hatt die sache uffzupicke. it is difficult the things up.zu.pick

b. *Es it

c.

is hatt die sache uffterpicke. is difficult the things up.fer.pick

fer die sache ufiPicke. Es is hatt it is di:ffk.-ul:t fer the things up.pick: 'It is diffk.-..ilt to pick the things up:

(elicited)

The linear order between the major elements of the clause in non-finite subordinate clauses with both fer and zu was as in (29). (29)

fer

SUBJECT NEGATION ZU VERB

If the position offer has changed as zu has disappeared, to parallel the change in function, it would now occur to the right of the subject and the negation. It is interesting

abstractions over order and do not have anything to say about the ordering of derivational procedures. l~or Standard German, most analyses assume that zu is either part of I or V (Bed1 1983; Beukema & den Dikken 1989; Giusti 1991 and Kiss 1994); the one dissenting voice is Wilder (1988), but to our mind, his argwnents are weak. llm is assumed to be a complementi1..er by some (van Riemsdijk 1984), whereas others have argued that is a preposition (Tappe 1984; Koster & May 1982; Giusti 1991; van Gelderen & Abraham 2004).

19.

404 Kersti Horjars & Kate Burridge

here to compare the response to a translation task of an older speaker (30a) with that of a younger speaker (30b). (30)

a.

b.

Er scheint en ehrliche man zu sei. he seems an honest man to be 'He seems to be an honest man:

(elicited)

Er scheint fer en ehrliche man sei. he seems for an honest man be 'He seems to be an honest man:

(elicited)

Judging by these two examples, for speakers who use fer instead of zu, fer does not occur in the same position as zu. would, but remains in a higher position. Non-finite constructions with the semantic subject in the syntactic subject position of the non-finite verb are not common but they do occur and when they do, the subject can only follow fer, as in (31). (31)

is hatt fer ien die sache uffpicke is difficult for him those things pick. up 'For him to pick those things up is difficult:

(elicited)

A sentence like this then indicates that fer has not changed its position. Similarly, the negation net follows fer as in the sentences in (32).20 (32)

a.

b.

c.

Es waer dumm fer's net kaafe. it would.be stupid for.it not buy 'It would be stupid not to buy it:

(elicited)

Sic ecschpect ien fer net glei do sei. she expects him for not soon there be 'She expects him not to get there soon:

(elicited)

Er seemt fer net en guter man sei. He seem for not a good man be 'He seems not to be a good man:

(elicited)

On the basis of this data, we conclude that the position offer within the clause has not changed even though its function has. We return now to the issue of its categorical status. The tree in (27) assumes that an element that combines with a full clause is of category C. Given that benefactive fer was a preposition, this would still involve a change of category. The evidence for the change in category is essentially that its complement has changed: fer combines with a clausal rather than nominal element However, there is evidence that the constituent

20. 'Ihe speaker in (32c) is 10 years old and the use of seemt instead of scheir1t is typical of the more extensive use of words borrowed from English that we find in younger speakers.

From preposition to purposive to infinitival marker 405 introduced by fer behaves like a preposition phrase. In elliptic c.onstructions, a combination ~ith da (defoor in (33)) is used, as it would be with a preposition phrase. 21 (33)

fer e dash mache wann de gaul is not good fer a dash make when the horse ready is defoor/ *foor es. ready is da.for for it 'It is not good to make a dash (across the road) when the horse is not ready for it: Es it net not

is net gL1t

We conclude then that fer has not changed its position since zu disappeared, but that it is at the highest clausal level, just as it was in the originaifer... zu purposives. In fact, it can even be argued that it has not changed its category, but that it remains a preposition as it was in its original use as a benefactive.

4·

Conclusions

When the first PG speakers established themselves in the US, we can be reasonably certain thatjer... zu was used only in purposive adverbial clauses. 'Ibis is still the dis·· tribution in the dialects from which PG originates. In the earliest sources to which we have had access, fer... zu appears also in subject clauses and in complements of adjectives, nouns and verbs. Towards the end the 19th century, zu started to disappear from purposive clauses. In other environments zu remained in productive use by most speakers into the 1970s. Current young speakers do not use zu in any constructions bar a couple of set phrases. This is clearly the case in WCPG, but to judge by works on other varieties of Plain PG, the situation is similar across the geographically dispersed but linguistically homogeneous group. An aspect of this change which appears unique amongst varieties of German is the loss of zu. The speed with which this has happened is remarkable. 22 The development offer from preposition to purposive marker to infinitival marker is, on the other hand, by no means unique. Indeed, it is the change that German zu and English to have undergone. In this paper, we have argued that reanalysis plays less of a role in this

21.

We are grateful to Mark Louden for having drawn our attention to this.

Phonological reduction and omission of unstressed material may be more expected in a language that is spoken only and has no written form. 'Ihe nature of the PG speech community may also be playing a role here. Trudgill (1995), for example, argues that grammaticali?.ation may spread exceptionally quickly in isolated close-knit: speech communities and that is certainly how one would want to describe these communities. 21.

406 Kersti Borjars & Kate Burridge

development than has been assumed in the literature. We maintain that the second stage of the change, which gave rise to the infinitival fer, did not involve reanalysis at all. We have also suggested that the development of the purposive from the benefactive need not have been driven by reanalysis, but could have been motivated by semantic tactors. The historical path from allative or benefactive to purposive to infinitival marker is typologically widespread. The purposive meaning desemanticizes and a new marker is drafted in to reinforce the purposive meaning. 1he benefactive or allative preposi-· tion is a common option due to the semantic connection. The purposive meaning of the new marker can then in tum be weakened, leading to the need for reinforcement. Indeed, Poutsma (1923) and more recently Miller (2002: 188) argue that the Old English bare infinitive ending in --an originally had a purposive meaning, which was weakened for it to be reinforced by to, which in tum was semantically weakened and reinforced in purposive environments by for. After the wane offor as a purposive marker, Present Day English now uses in order to as a dedicated purposive marker. Even though this is a common change typologically, one might look for local explanations for it. A suggestion found in the literature is that it is due to influence from English (Reed 1948: 243-244; Costello 1978: 36-44, van Ness 1994: 436). However, there are a number of reasons why this is not a plausible explanation. Firstly, though for at one time was the marker of purposive in English and to some extent still serves this purpose (see examples in (1)), it has not developed into a general infinitival marker, but occurs only in infinitivals with subjects (this has led to it being described as a 'subject marker: see de Smet 2010). As we have seen, in PG examples of fer clauses with a subject are very infrequent. 'Ihe distribution of fer i.n PG then shows little resemblance to that of for in English. As Louden (1988: 209) points out: "That 'fer: rather than 'zu: which is closely cognate to AE [American English, KEB/KB] 'to: should be expanded testifies to the apparent lack of influence from AE: There are varieties of English such as Ottawa Valley English (OVE, Carroll1983) that have a more extensive distribution of the for... to construction. The connection between OVE and WCPG may appear particularly tempting because of the geographical proximity, but the WCPG community does not speak OVE, nor are they likely to come into contact with people who do. Furthermore, the distribution of for... to in OVE is still much more limited than the WCPG equivalent. The distribution offet: .. zu appears also to be common to all varieties of plain PG and there is absolutely no evidence that it has spread from Canada. As we saw in Section 2, it is also the case that fi.r... ze in Luxembourgish has a similar distribution to that offer... zu in PG without having had the kind of influence from English that PG has had. 'Ihere are then any number of reasons to refuse the claims offer... zu as a contact phenomenon (pace Reed 1948: 243-244; Costello 1978: 36-44, van Ness 1994: 436). The type of historical change fer has undergone (and before that zu.) is an instance of gr.:unmaticali7.ation. Grammaticalization is generally assumed to involve parallel

From preposition to purposive to infinitival marker 407

changes to meaning, morpho-syntax and to phonology. Lehmann (1995 [1982]), for instance, argues that desemanticization is often accompanied by a reduction in 'syntactic scope: that is to say that a grammaticalized element would tend to combine with a smaller syntactic unit than the original element In terms of the tree in (27), tb.is would lead to an expectation that an element originally found in C would appear in I once it has grammaticalized (it should be pointed out that Tabor & Traugott 1998 argue against this as a generalization of grammaticalization). In fonnal approaches to grammaticali7.ation such as Roberts and Roussou (1999, 2003) or van Gelderen (2004), grammaticalization is essentially viewed as frozen movement. The association with a different position in the tree, in particular a functional category rather than a le.lC:ical one, plays a role in accounting for changes also to function. Our conclusion on the basis of the development of PG fer is that it has changed drastically in meaning, but that its structural position and maybe even its categorical status - has not been affected. This is then evidence in favour of modelling grammaticalization in such a way that change in different dimensions can be represented separately. Lexical--Functional Grammar (Bresnan 2001) is a parallel correspondence theory of grammar in which different types of linguistic information are represented separately, with mapping relations defining the way in which they relate to each other. Certain relations are expected to be more common, see for instance the StructureFunction Association Principles posited by Bresnan (2001: 101-9), but non-one-toone mappings are permitted, indeed expected as part of historical change. With respect to PG fer, we can say that its representation in f(unctional)-structure has changed but its c( onstituent)-structure representation has not. Hence the mapping between the two has also changed. A representation of the f.-structure of the original purposive fer is fi.)und in (34a)and thatofzu in (34b). (34)

a.

fer iPRED ='fer ' hYPE =purposive jE

b.

zu

(i ADJUNCT)

i == J,

Here fer combines with a clause (caMP) to form a constituent that has purposive meaning and belongs to the set of adjuncts that its mother hasP This ensures an adjunct function, but allows for further adjum.ts to be associated with the mother. 1he infini·· tival zu, on the other hand, is a particle which does not add meaning itself; the arrows indicate that functional information from the node is passed up to make it part of the same [--structure as the mother. This interpretation of the function of zu is similar to that adopted in Head-driven Phrase Structure Grammar of English, where to is analysed as a raising verb without its own semantics (Pollard & Sag 1995: 143).

23.

We ignore the details of the analysis of subjectless clauses here.

408 Kersti Borjars & Kate Burridge

By contrast, in the more recently evolved fully desemanticized general function, fer would have the same [-structure as zu in (34b), as in (35). (35)

fer '1' = J..

The position in the c-struc.ture offer, on the other hand has not changed. It is associated with a tum~tional node at the highest level of the clause; fer is in the C position in both (34a) and (35), giving the structure in (36) for both. The only difference is whether or not there is a zu under the lower node. Indeed, we argued in Section 3.6 that its category has not changed since its days as a benefactive preposition, in which case C should be substituted by P in (36).24 (36)

CP

~

C

IP

fer

I

~

VP

16 (zu) Naturally. it may well be that fer is still changing as far as its structural position goes and maybe in a few decades, fer will show more of the properties which we associate with I elements like zu. Ongoing changes to clause structure may contribute to this change. 'lhe normally verb final nature of PG subordinate clauses means that clause initial fer remains separated from the clause final verbal complex. However, word order changes in PG are seeing violations of the so-called verbal brace (Sa.tzrahmen). Im.Teasingly examples such as (37) are found, in which elements like arguments of the verb and adjuncts are appearing after the verb(s), thereby bringing discontinuous verbal constihtents closer together. 1he relaxing of verb final order in this way means thatfe1' often now appears adjacent to the verb, as in Example (37). This may well also speed up a structural change which associates fer more closely with the verb. (37)

Awwer er saagt im blatz fer saage "oh yes oh yes but he says in place for say

.,

no saagt er immer "oh el. then says he always oh ei 'But he says in place of saying 'oh yes: he always says 'oh ei~

24. We are ignoring detailed theory specific uses of clausal functional categori.es here and seeing C as the functional category associated with the whole clause and I as the functional category associated with the verb phrase.

From preposition to purposive to infinitival marker 409

As yet there are no signs ofPG recruiting a new purposive marker, but with the spread ofjer and the associated weakening of its purposive meaning this is a predictable next stage. We expect that some kind of reinforcing element will eventually be pressed into service and via a kind of"Jespersonian cycle of purposive': this will evolve into a new dedicated purposive marker.

References Bed1, Gunnar. 1983. Studierl iiher das deutsche Verbum ir~fi.nitum. Tiibingen: Max Niemeyer. Beukema, Frits & den Dikken, Marcel. 1989. The position of the infinitival marker in the Germanic languages. In Ser1tential Complementation and the Lexicon, Dany Jaspers, Wim Klooster, Yvan Putseys & Pieter Seuren (eds), 57···75. Dordrecht: Foris. Bresnan, Joan. 1979. Theory of Complementation in English Syntax. New York NY: Garland. Bresnan, Joan. 2001. Lexical-functional Syntax. Oxford: Blackwell. Bruch, Robert. 1973. Luxemburger grammatik in volkstiimlichem abriss. Luxembourg: Editions de la Section de Linguistique de l'Institut gr.-d. Buehle1~ Alan. nd. Stories in PG. Ms. Buehle1~ Alan. 1977. The Pennsylvania German dialect and the life of an old order Mennonite. Camhrige, Ontario: Publication of the Pennsylvania Folklore Society of Ontario. Buffington, Albert F. & Preston, Barba A. 1965. A Permsylvania German Grammat~ 2nd edn. Allentown PA: Schlechter. Carroll, Susan. 1983. Remarks on for-to infinitives. Linguistic Analysis 12: 415-451. Chomsky, Noam. 1981. Lectures on Govemment artd Binding. Dordrecht: Foris. Chomsky, Noam. 1986. Barriers. Cambridge MA: 'Ihe MIT Press. Chomsky, Noam & Lasnik, Howard. 1977. Filters and control. Linguistic Inquiry 8: 425-504. Christophory, Jul. 1974. Mir sclrwiitze Letxe/Juergesch. Luxembourg: Imprimerle Saint-PauL Societe Anonyme. Cinque, Guglielmo. 1999. A.dver/Js and Functhmal Heads: A Cross-linguistic Perspective. Oxford: 0 UP. Costello, John R. 1978. Syntactic change and second language acquisition: The case for Pennsylvania German.. Linguistics 213: 29-50. deSmet, Hendrik. 2007. For... to-lnfinitives as verbal complements in Late Modern and Present-Day English: Between motivation and change. Brzglish Studies 88: 67-94. deSmet, Hendrik. 2010. Grammatical interference. In Gradualness, Gradierzce and Grammaticalization, Elizabeth Closs Traugott & Graeme Trousdale (eds), 75-104. Amsterdam: John Benjamins. Dixon, Robert M. W. 2005. A New Approach to English Grammar, on Semantic Principles. Oxford: Clarendon Press. Ebert, Rober P. 1978. Historische Syntax des Deutschen. Stuttgart: J.B. Metzlerische Verlags buchhandlung. Fretz, J. Winfield. 1989. The Waterloo Mennonites: A Community it1 Paradox. Waterloo: \Vilfrid Laurier University Press. Frey, J. William. 1981[1942]. A Simple Grammar c1'Pennsylvrmia German, 2nd edrt Lancaster PA: Brookshire Publications.

410 Kersti Borjars & Kate Burridge

van Gelderen, Elly. 2004. Grammaticalization as Economy [Linguistik AktuelliLinguistics Today 71]. Amsterdam: John Benjamin.s. van Gelderen, Elly & Abraham, Werner. 2004. The grammaticali7.ation of the infinitival preposition - toward a theory of 'grammatl.calizing reanalysis~ Journal of Comparative Germanic Linguistics 7(2): 111-170. Gl.ustl. Gl.uliana. 1991. Zu-lnfinltivals and sentential structure in German. Rivista di IJnguistica 3:211-234. Haag, Earl C. 1982. A Pennsylvania German Reader and Grammar. University Park PA: The Pennsylvania State University Press. Harris, Alice C. & Campbell, Lyle. 1995. Historical Syntax in Cross--linguistic Perspective. Cambridge: CUP. Haspelmath, Martin 1989. From purposive to infinitive ··· a universal path of grammaticization. Folia Linguistica Hl..~torica X(l---2): 287--310. Haspelmath, Martin. 1998. Does grammaticalization need reanalysis? Studies in Language 22(2): 49-85. Henn, Beate. 1980. Pfalzlsch. Di.isseldorf: Piidagogischer Verlag Schwann Henry, Alison. 1992. Infinitives in a for-to dialect. Natural Language and Linguistic Theory 10: 279-301. Henry, Alison. 1995. Belfast English and Sta11dard English. Oxford: 0 UP. Hopper, Paul 1991. On some principles of grammaticization. In Approaches to Grammaticalization [Typological Studies in Language 19-1], Elizabeth Closs Traugott & Bernd Heine (eds), 17···36. Amsterdam: John Benjamins. Horne, Abraham R. 1905[1875]. Pennsylvania German Afanual, 3d edn Allentown PA: T.K Horne. Huffines, Marion Lois. 1990. Contact phenomena in language maintenance and shift: the Penn-· sylvania German infinitive construction American Journal of Germanic Linguistics a11d Literatures 2: 95-- 108. Jespersen, Otto. 1909. A Modern English Grammar on Historical Principles. Heidelberg: C. Winter. Jespersen, Otto. 1922. Language: Its Nature, Development and Origin. London: G. Allen & Unwin. Karch, Dieter & Moellcken, Wolfgang w: 1977. S!edlungspfiilzisch in Kreis Waterloo, Ontario, Ki;mada. 1iibingen: Max Niemeyer. Kiss, Tibor. 1994. Bemerkungen zum Vorkommen des 2. Status. Linguistische Berichte 154: 461-484. Koster, Jan & May, Robert. 1982. On the constituency of infinitives. Language 58: 116-143. Lehmann, Christian. 199 5[ 1982]. Thoughts on Grammaticalization. Munich: Lincom. Lockwood, William Burley. 1968. Historicai German Syntax. Oxford: Clarendon Press. Los, Bettelou. 2005. The Rise oftheto-Infmitiveas Verb Complement. Oxford: OUP. Louden, Mark. 1988. Bilingualism and Syntactic Change in Pennsylvania German. Ph.D. dissertation, Cornell University. Louden, Mark. 1997. Linguistic structure and sociolinguistic identity in Pennsylvania German society. In Languages and Ln:es: Essays in Honour of Werner Enninger, James R. Dow & Michele Wolff (eds), 79-92. Bern: Peter Lang. Luxemburger Warterbuch, Vols. 1-5 (I 1950-54, II 1955-62, III 1965-70, IV 1971-75, V 1977). Luxembourg: P. Linden Hofbuchdrucker. Miller, Gary D. 2002. Nonftnite Structures in Theory and Change. Oxford: OUP. van Ness, Silke. 1994. Pennsylvania German. In The Germanic Languages, Ekkehard Konig & Johan van der Auwera (eds), 420-438. London: Routledge.

From preposition to purposive to infinitival marker 411 Pollard, Carl & Sag, Ivan. 1995. Head-driven Phrase Structure Grammar. Stanford CA: CSLI. Poutsma, Hendrik. 1923. The Infinitive, the Gerund and the Participles of the English Verb. Groningen: P. Noordhoff. Reed, CarrolL 1948. 'Ihe adaptation of English to Pennsylvania German morphology. Ameri;:an speed: 23: 239-244. van Riemsdijk, Henk. 1984. 0 n pied- plpcd lnfinltives in German relative clau.~es. In Studie.~ in German Grammar, Jindfich 'loman (ed), 165-192. Dordrecht: Foris. Ri:z:zi, Luigi 1997. The fine structure of the left periphery. In Elements of Grammar: Handbook itl Generative Syntax, Liliane Haegeman (ed.), 281---337. Dordrecht: Kluwer. Roberts, Ian & Roussou, Anna. 1999. A formal approach to grammaticali:zation Linguistics 37: 1011"" 1041. Roberts, Ian & Roussou, Anna. 2003. Mittimalist Approaches to Grammaticalization. Cambridge: CUP. Sauder, Ben. 1955. Der nachbar atl de schtroas. St. Jacobs: Pennsylvania German Folklore Society of Ontario. Tabor, Whitney & Traugott, Elizabeth Ooss. 1998. Structural scope expansion and grammaticalization In The Limits of Grammaticalization [Typological Studies in Language 37], Anna Giacolone Ramat & Paul]. Hopper (eds), 229-272. Amsterdam: John Benjamins. Tappe, Hans Thilo. 1984. On infinitival clauses without COMP. In Sentential Complementation, Wim de Geest & Yvan Putseys (eds ), 227-237. Dordrecht: Foris. Trudgill, Peter. 1995. Grammaticalisation and social structure: Non-standard conjunctionformation In East Anglian English. Grammar and Meaning: Essays in Honour of Sir Johtl Lyons, Frank Palmer (ed.), 136---147. Cambridge: CUP. Visser, Frederic Theodor. 1963---73. An Historical Syntax of the English Language, 3 Vols. Leiden: E.]. BrilL Webe1; Albert & Dieth, Eugen 196·1. Ziirichdeutsche grammatik; ein wegweiser zur gutetl mundart, 2nd cd. Zi.irich: Schwei1..er Spiegel. Wierzbkka, Anna. 1988. The Semantics ofGramtr.ar [Studies in Language Companion Series 18]. Amsterdam: John Benjamins. Wilder, Chris. 1988. On the German infinitival marker zu and the analysis of raising construction~. Lingua 76: 115-175.

SECTION

6

Pragmatics & Conversation analysis

Word choice, turn construction, and topic management in German conversation Adverbs that are sensitive to interactional positioning EmmaM.Betz Kansas State University

Using conversation analysis, thi<> study examines the relationship between word choice and turn construction. In Siebenburger Siichsisch, a Romanian settlement variety of German, certain adverbs possess functional variants that exhibit position-sensitivity v.'ith respect to their placement in turns in interaction. This study investigates the lexical items ietttz(t)tnow' and ientzter!'now'. In interaction, speakers use these variants as resources to mark the status of a turn constructional unit or tum as complete or incomplete. Moreover, the use of ientz(t) and ientzter is sensitive to the placement within the larger sequence: Turn-final ientzter is tied to the management of topics in interaction, specifically to promoting a new action trajectory. This study outlines an intersection of turn construction, topic development, and word choice and expands the body of comparative work in conversation analysis.

1.

Introduction

Conversational interaction relies on an intricate machinery for constructing units of talk, for inspecting these units within the context in which they occur, and for responding to them. The regularities of turn and sequence organization are the means by which participants establish who gets to speak when, where, and for how long and thus organize actions and topics, negotiate meaning, and accomplish understanding. A central 'task' at any point in the organization of turn-taking is that participants recognize the actions and structures underway and project the possible completion of a turn or turn constructional unit (i.e. a 'TCU; Sacks 1974) in progress, thereby also projecting appropriate places for speaker change. Participants to interaction accomplish this in real time, using prosody, the regularities of grammatical structure, and lexical markers as resources (Auer 1996, 2004; Couper-Kuhlen & Ford 2004; Selting 1995, 1996, 2000, 2001, 2005; Selling and CouperKuhlen 1996, 2001).

416 Emma M. Hetz

Grammatical and prosodic resources have received considerable attention in interactional linguistic research on the organizing of turn- taking in every-day interaction in German. 1his study, by contrast, is concerned with lexical resources for organizing turns and sequences: with the connection between lexis and syntax and between lexis and discourse. 1 'Ihrough analyzing different forms of a tempo-ral marker in one variety of German, I show how lexical resources are used and understood by interlocutors as indicators of the status of the utterance, topic, and sequence underway: They systematically index a turn constructional unit or turn either as still in progress or as designedly complete. Data excerpt 1 provides a first example of the phenomenon. 2 (1)

nenjtziihr1 [43_av4/3_56:55], talk about the age of an acquaintance's children ==> 1 Anna: tsk![da WO-IRE NENitziihn.=ientz iam oprial. tsk![they WEJRE nineTEEN.=now in .!!pril. tsk![they TURJNED nineTEEN.=-just now in 2 Gabi: 3

[0 m:hm0

gpril.

]

(.)

==> 4 Anna: OCH71ihn oder NENJt71i.hn >wores=ientzter. EIGHteen or NINEteen >were-tthey=now.

EIGHteen 5

6

i[ch weis ] I[know ] I[ don't ]

or

NINEteen chia PRT

>they turned=(just)rww.

net< se: not< the:y

sa= (look)=

know< the:y (look)=

Gabi: [0 m:hm, 0 ]

The ne.xt section outlines previous research on the use of grammatical, prosodic, and lexical resourc.es for managing turn--taking. 'Ihe data used to illustrate some of these resources come from Siebenbiirger Sachsisch, a settlement variety of German. Note that the regularities described in Section 2 hold for Colloquial Standard German as well. The phenomenon that is the focus of this study, however, has no equivalent in Colloquial Standard German. Section 3 provides further preliminaries: It describes the methodology and data used.

In developing my analysis, I have benefitted from discussions with Amanda Huensch and from audience questions and comments at the 2008 convention of the National Communication Association in San Diego, CA. In 2006, during a data session at the University of Illinois, Makoto HayaShi first drew my attention to the lexical variants that are the subject of this study. Juliane Edenstrom and one anonymous reviewer have provided valuable feedback on earlier drafts of this article. All remaining errors are mine. 1.

The lines containing the target forms ier1tz and ientzter are highlighted through arrows. See appendices A and B for transcription conventions and abbreviations used.

1.

Adverbs in German conversation 417

z. Resources for constructing turns: Grammar, prosody,lexis There exist, first, grammatical resomces for shaping sentential TCUs and thus for signaling the status of a turn. Typologically, German exhibits verb-second ordering in declarative main clauses and wh-questions. 'Ihis verb-second ordering is displayed regardless of the type of constituent that occupies the first position. In exc.erpt 2, the syntactic first position is occupied by a direct object ( 2)

Verb -second ordering in German: ge rmrmistik [0 05- B_4 96] l1h dieses briecht det INA nkhen . .hh this kind needs the ((name)) not. [O][V][ S ] .hh this kind LillA doesn't r1eed.

K:

Composite verb forms and complex verb phrases are ordered according to a syntactic rule called Satzklammer ('sentence brace; English terminology following Auer 1996) in main clauses. Data sample 3 provides an example of a sentence brace structure; the parts highlighted through bold print form the brace itself. (3)

Strong syntactic closure in German: der schober [0 03-B_378] M: hias ientz ian pension gonge, he+has now in p~_nsion gone, [S] [V]

[PPT]

he just retired, In a declarative sentence in German, the finite verb (V) in second position represent~ the left brace of the sentence brace structure, followed by the so--called MiUe!feld ('inner field'). Adverbials and prepositional phrases in German are typically contained in this inner field (cf. Frey & Pittner 1998; Pittner 1999; Weiss 1975; Zifonun, Hoffmann, Strecker & Ballweg 1997 ). 3 'Ihe right brac.e constitutes the right boundary of the inner field and may be filled with elements such as infinitives, predicate complements, separable verb prefixes, or a past participle (PPT). Thus, the second part of a composite verb generally appears in clause-final position in German. Because the right brace in a syntactic structure (clause or sentence) typically constitutes the actual end of a sentential TCU and is also oriented to as such by recipients in spoken interaction, the ends of such units are strongly projectable based on syntax alone. In short, there are certain clearly defined and syntactically driven word order rules in German. Based on these, interactants can both predict and recognize points of possible syntactic completion and thus identify places at which turn transition is interactionally 'le!,>itimate:

3· For accounts of expansions beyond the right boundary of a sentence in German, see Auer 1996,2006, Couper-Kuhlen & Ono 2007, Edenstrom 2008,2009, and Vorreiter 2003.

418 Emma M. Betz

Speakers also have a variety of prosodic resources available to indicate that a TCU that is projectable as nearing completion is indeed constructed to be complete at the projectable point. These signals of 'designed' completion in German include characteristics of volume, speed, stress, and pitch (e.g. Auer 1996; Gilles 2001; Kern 2007; Selting 1996, 2000, 2001, 2005). Certain pitch configurations, combined with gram·· matical and pragmatic features, mark transition relevance. They are typically absent fi·om turns designed as not yet complete. In German, tenninal falling or rising pitch with respect to the speaker's pitch range signal turn yielding. whereas final level or only slightly rising pitch is used to signal turn holding (Gilles 2001; Kern 2007; Selting 1996, 2000: 510, 2001; for research on English, see Ford, Fox & Thompson 1996; Local, Kelly & Wells 1986). Moreover, pitch peaks close to a possible completion of a TCU seem to function as "a carrier whose syntactic properties and placement can have it heard as projecting upcoming completion of tum" (Schegloff 1998: 241; see Auer 1996 for German). Lastly, an upstep or downstep in pitch, in combination with syntactic cues, often marks a prosodic break, and thus, for example, the beginning of a new structure (Auer 1996; Selting 2001; see also Walker 2004). We can thus note that prosodic features, just as features of syntax, are used by participants for signaling and identifying actual transition relevance. Besides syntax and prosody, lexical markers are important in organizing units in conversation. Studies investigating the connection between lexis and turn organization in German have described turn--final markers (e.g. response pursuit tokens, Harren 2001) and turn-initial markers, which connect turns and sequences, for example by signaling the beginning of an extended contribution (e.g. Egbert 2002, 2003; Zifonun et al. 1997: 899; Taleghani--Nikazm 2008). Egbert (2002, 2003) outlines the use of German iibrigens, which, placed turn--initially, projects a new and discontinuous topic to come.4 In its main syntactic position in the inner field of a verb-second sentence, however, iibrigens indirectly refers to a previous sequence or interaction beU.veen the same interactants, thereby invoking a common interactional history. In the latter use, iibrigens is consequently classifiable as a modal particle rather than as an adverbial. With respect to lexical TCUs, Sacks, Schegloff and Jefferson (1974) note that, in constructing a turn-at-talk, lexical TCUs (e.g. really? in English) appear overwhelmingly as first or only TCUs in their turn and their occurrence elsewhere is limited to doing something special. These differences in distribution can offer one kind of evidence for position-sensitive grammars. Studies such as Egbert 2003 and Sacks et al. 1974 highlight the connection between word semantics, word choice and interaction by showing that the meaning of a lexical item, indeed its word class, often crucially depend on its placement in a tum and sequence.

4· This is similar to 'by the way' in English (Sd1egloff & Sacks 1973).

Adverbs in German conversation 419

While the connection between lexis and turn construc..tion (that is, the organization of single turns and the connection between turns) has received some attention in German, the connection between lexis and TCU construction (that is, the construction ofbasic units within a turn) has not. 'Ihis study looks at one particular lexical item within TCUs, specifically within sentential TCUs, in Siebenbiirger Sachsisch. Data excerpt 4 provides another e.l!:ample of ientz(t) and ientzter used in spontaneous interaction. Both forms carry the semantic value of a temporal adverb. They refer to the immediate present or to a point in the past constructed as close and immediately rei-· evant to the present and are thus translated as 'now' or 'just now' (cf. example 1).5"We can note, however, that ientz and ientzter differ in their placement in the interactional units in which they are used: TCU-final and TCU-initial placement can be contrasted. (4)

branessle [53.54....avl3/l4 ....44:12], talk about harvesting edible nettles l Maia: OV OVER nau (.) g_y wttse glech BU- BUT PRT (.) then when+they immediately BU- BUT (.j right when they come 2

ereu'l ku, -w-use si gri:B sien "nuer.0 c.ome ~· when+ they this ta:ll are oonJy.o .!!1!~

=> 3

*M: head *M: dismissive shake hand gesture* gaut. niet IEN*tzter. *ientz siese* n[immi a[ny more] good. not now. now they're not

not 4 Erna:

when they're this ta:ll ~only (then).~

now.

now they're not"

g[oodany] more. [ ojao [ oyeso

1 I

Besides documenting a distributional difference for two semantically congruent versions of the temporal marker 'now' (ientz(t) and ientzter), the present study provides a sequentially grounded analysis of their use in interaction. Previous conversation analytic research has indicated that, on the one hand, different tokens used in similar environments typically perform different interactional functions (e.g. Gardner 1995; Zimmer·· man 1993; Golato & Betz 2008), and that, on the other hand, homonymous tokens used in different tum and sequence position usually serve quite different functions (e.g. Alm 2007; Barske 2009; Betz & Golato 2008; :Egbert 2003; Heritage & Raymond 2005).

5· Ientz(t) and ientzter are of Germanic origin and thus cognates to the Standard German jetzt rather than borrowings from Romanian (Braun-Santa, Haldenwang, Richter & 'Ihudt 1972: 220; Grimm & Grimm 1874-1960 (1971); Ney 1984; Schullerus 1925: 389-390). Both go back to Middle High German ietze, iezuo, iezunt, iezent; the colloquial forms are assumed to have formed through metathesis (Braun-Santa et al. 1972: 220).

420 Emma M. Betz

Given these findings, it stands to reason that ientz(t) and ientzter, two different tokens with the same semantic content, fulfill different interactional function andior exhibit a systematic structural-distributional difference in Siebenburger Sachsisch. In Section 4 below, I show that speakers of this ·variety of Gem1an indeed use ientz(t) and ientzter as position-sensitive tokens. 'Ihe tom1er indexes the status of the current turn as 'design-· edly incomplete; the latter marks it as 'designedly complete: Moreover, I argue that the description of these tokens needs to inc.lude an analysis of the larger sequenc.e in which they occur. since it can be shown that ientzter is involved in topic management. The following section provides a brief description of the methodolo!:,'Y and data used. The analysis in Section 4 inc.ludes the presentation and discussion of ientz(tj and ientzter within the turn, sequence, and action contexts in which they appear. I dose in Section 5 with a discussion of the phenomenon of position--sensitive adverbs in a broader conte..lCt: I present a similar phenomenon in another variety of German, Bavarian German, and locate the implications of my findings at the intersection of morphology, word choice, and pragmatics. I also discuss implications for the classification oflexical variants and raise questions concerning the grammaticalization paths of adverbs and particles in German.

3·

Methodology, data and transcription

This study is a micro-analysis of sentence-level resources and their use in interaction. It aims to describe the interplay between the linguistic structure of utterances, specifically the choice and linear ordering of lexical elements, and the in--situ tasks that they are deployed to fulfill in discourse. To this aim, this study adopts a view of grammar as embedded within the larger organization of social conduct. In this conceptualization of grammar, "the linguistic shape of an utterance is intertwined with the changing relationships among participants over interactional time" (Schegloff, Ochs & Thompson 1996: 44). Guided by the view that grammar shapes and is shaped by interaction, and using the methodology of conversation analysis (CA) (Atkinson & Heritage 1984; Boden & Zimmerman 1991; I-Iutchby & Wooffitt 1998; Lerner 2004; Prevignano & Thibault 2003; Sacks et al. 1974), this study analyzes the use of temporal adverbs in German as determined by aspects of turn-taking and sequence organization (see e.g. Auer 2005; Ford 1993, 2001; Ford & Fox 1996; Ford, Fox & Thompson 2002; Jefferson 1993; Schegloff 1996, 2002, 2007). Within an interac.tionallinguistic. view of linguistic structures, word ordering in German within and outside the sentence brace is primarily a resource for partici-· pant~ to structure actions and display understanding (e.g. Auer 1996; Edenstrom 2008, 2009; Schwitalla 2003). In the description of the German sentence brace (Sec.tion 2), I already suggested that adverbs may appear in different syntactic positions. In Section 4 below, I connect structural adverb placement and interactional meaning by analyzing

Adverbs in German conversation 421

representative examples of ientz(t) and ientzter in the environment of turn and sequence in Siebenbiirger Sachsisch. This study is based on non-elicited recordings of every-day interactions in Siebenbiirger Sachsisch, a Romanian settlement variety of German. The speakers recorded grew up in different parts of Siebenbiirgen (Transylvania) and now live in Southern Germany or Austria. 6 The recordings include 3.5 hours of telephone and 5 hours of face-to-fac.e interaction at such occasions as family meals and festivities, afternoon coffee, telephone chats, and arrangement--making on the phone. Using Jeffersonian transcription conventions (see Appendix A), the data were closely trdll·· scribed to include such features as audible pauses, simultaneous talk, and characteristics of speech delivery, for example pitch and loudness of voice (Atkinson & Heritage 1984: ix---xvi; Jefferson 1983, 1.985; Sacks et al. 1974). For each line of German data, three lines of transcript are provided: the German original, an interlinear English gloss, and an idiomatic English translation (Duranti 1997). Where the English gloss and idiomatic translation are identica~ the gloss is omitted. Capital letter abbreviations for syntactic and semantic features largely follow Leipzig glossing rules (Comrie, Haspelmath & Bickel2008) and are provided in Appendix B. Embodied actions (eyegaze, gesture, posture) are selectively included in the transcripts and appear as narrative descriptions above the original utterance. In all excerpts, the line containing the target phenomenon is marked with arrows. My data yielded a total of7l examples of ientz(t) and ientzter, with 33 from tele·· phone calls and 38 from face-to-face interactions. 61 were ientz(t) tokens (28 telephone, 33 fac.e-to-face), and 10 were ienttter tokens (5 telephone, 5 face-to-face). The phenomenon presented here cannot be found in Colloquial Standard German interac·· tion (there exists only one form of the word jetzti'now'), but preliminary observations suggests that it exists in other "V--arieties of German as well. One of these - Bavarian German - will be presented in Section 5.2.

4-

Analysis: ientz(t) and ientzterin Siebenbiirger Sachsisch

In this section, I present data segments that are representative of my collection. For each data segment, a brief analysis highlighting how interactants use ientzt (Section 4.1) and ientzter (Section 4.2) in Siebenbiirger Sachsisch is provided. In the analysis itself, each section moves from a distributional analysis focused on individual reus and turns to the analysis of the lexical token in its broader sequential environment. The latter includes a discussion of, for instance, co--participant orientation and topic development Section 4.3 provides a summary of findings that hold for all cases in

6.

The author herself is a native speaker of this variety of German.

422. Emma M. Betz

my collection and discusses apparent exceptions or marked uses. Finally, Section 4.4 discusses two data excerpts in which speakers deploy ientz(t) and ientzter 'strategically' in interaction to construct an extended turn-at-talk, a syntactic 'pivot construction~ 4.1

Use of ientz(t)/'now' in spontaneous conversation

Data examples 5···8 illustrate the syntactic and turn positions in which ientz(t) may occur: TCU- and sentence-initially in a sentential TCU without a sentence brace (example 5), TCU- and sentence-internally in a sentence with strong syntactic closure (example 6) and without (example 7), and in a rightward syntactic expansion, i.e. outside the sentence brace (example 8). The first segment is drawn from a conversation between Maia and her daughterin-law Anke at a family gathering. Maia has been observing an interaction between two family members, Dragu and Dan, both non-native speakers of German and from different native language backgrounds (Romanian and English). She highlights the noteworthiness of this encounter for Anke. At the point at which Maia draws Anke's attention to the encounter between Dragu and Dan, their conversation has come to a halt. What Maia identified as noteworthy is thus no longer observable for Anke. Maia uses the adverbs ientz/'now' and i:nteren!'a minute ago' (lines 3···4) to contrast the current interactional situation with what was observable just previously. Jentz is used at the begitming ofMaia's tum and in the syntactic front-field of the first TCU. (5)

miatden huendjen [5l_avl3.14_24:50.25:29]. 1 Maia: dies zwi:n menschen haesi. these two: people right here. 2

(1.5)

((A: shifts position, leans forward to look at Dragu and Dan))

nau niet miadem. now speaks+he PRT not wit:h+him. now hes not speaking with him.

=> 3 Maia: ientz riate

4 Maia: over i:nteren, (uallebit) but a minute ago:, (both)

miat den huendjen. with their hands.

5 Anke: der dragu uch der dan. the ((name)) and the ((name)). dragu and dan. 6

(0.7)

((Maia nod'> several times, smiling))

In the next data sample, ientz(tj- realized once with and once in the elided form without the final plosive (Schwitalla 2003: 38···9) is used in clause-internal position as part of an insertion to the main telling underway. Maia and 'Hmja, who are friends, are discussing the situation for pastors in their respective parishes. Maia notes that their

Adverbs in German conversation 423

long-time pastor, Schober, who had been with the church for 17 years, just recently retired and was replaced by another pastor, Schuster. In her turn in line 2, Maia uses the adverb ientz(t) twice in sentence-internal position, after the subject and finite verb (nii huse!'then they have' and hias/'he has~ respectively), but before the participle gongel'gone (line 3), which constitutes the right brace and thus the possible syntactic completion of the repaired unit-in-progress. (6)

der schober [26_0 03-B_378] 1

Maia:

=> 2

.hhh

ba

der

kirch,

.hhh

with

the

church, seventeenyears.

siven:l.lin gar.

nii husen ientzt-- hias ientz ian then have+they+him now- he+has now in

so they have (him) now- he (has) t1ow retired, pension pension

3

gonge, nau husen undern gone, PRT have+they+a other

so they hired somebody ugestualt,==schu'lter ias uchen sochs? hired,=((name)) is also a saxon?

4

else==schuster (he is) also a saxon? (=-from Siebenburgen) Data sample 7 illustrates another use of ientzt in clause-internal position. In contrast to example 6, ientzt here occurs in a declarative sentence without strong syntactic closure, i.e. without a sentence brace. The topic is a family member's drinking problem. Anna uses the lexical item ientzt in line 3 when, in response to her interlocutor's inquiry (not in transcript), she describes another family member's reaction to the problem. (7)

drinking [12_av1i2_4_19:10] 1

Anna:

0 n0° oPRTo

sai she

drantj drinks

uch too

PRTo

she

drinks

as :W.fJJl

0

2

=> 3

Anna:

(0.3) *(G:gazeshiftawayfromA, unbelieving fadal expression) 0 sa *drantj ja ientzt yQ),0 0 she *drinks PRT now tQ_Q_0

"she 4 5

Anna:

6

Gabi:

7

*drinks PRJ' now as .WJ:.ll0

(0.3) ((A: gaze away from G; G: gaze shift back to A)) [((2syll))] [ ] [ad1 >es]i,< [oh >s]o,<

[oh >is th]at true,< (0.3) ((A shifts ga:1.e to G))

424 Emma M. Hetz In addition to clause-initial and clause-internal positions, ientz(t) is also used at the beginning ofTCU-expansions, that is, after an interactional unit has been marked as syntactically and pragmatically (though not necessarily prosodically) complete. The following excerpt is taken from a telephone interaction between Anna and her friend Gabi. The topic of conversation is the age of an acquaintance's children. 'Ihis excerpt and its continuation is discussed in greater detail in Section 4.2. (8)

nenjtzahn [42_av4/3_56:55] 1 Anna:

2 Gabi:

.H ias ja endjen ohnen uarbet . .H he+is PRT always without work. .H he is (as we knowj always out of work.

e§.i; ~

is that~ 3

(0.4)

4 Gabi:

de kentj sien over g:riS; the children are though biG; the children are

grOWN though;

5 Anna:. 6 Anna:

hhh <de ke:ntj sien nau> gor kliener wa aus hans. all smaller than our ((name)). all younger than hans.

7 Gabi:

m:[hm [(denn der (1 syll)) wor nau -l,nenjot:zehn.0 = [(because the ( )) was PRT J-nineoteen."=

8 Anna:

[(because ())(just) turned -l-nineOfeen.""= 9

=niet, =not, =(right!see).

10 => 11 Anna:

(0.3)

12 Gabi:

tsk! [da WO]RE NENJtziihn.=ientz iam oprial. tsk! [they WE]RE nineTEEN. ::::now in .!!.pril. tsk! [they TUR]NE.D nineTf~N.=just 1ww in gpril. [0 m:hm0 ]

13

(.)

The third party referred to in line 1 is a close acquaintance of Anna's; thus, the informa-tion G-cibi provides in line 4lies in Anna's domain of knowledge and makes confirmation or disconfirmation relevant (cf. Kamio 1979; Labov & Fanshel1977). This is not immediately forthcoming, and after a delay in line 5, Anna provides a comparison to her own children's age (line 6). By describing the children in question as 'younger than' her own youngest son, she effectively disconfirms and corrects Gabi's candidate understanding

Adverbs in German conversation 415 in a non-straightforward way. Research on repair resolution has shown that otherrepairs (especially of assumptions and expectations) tend to be followed in third positions by information and/or action receipts such as ach, acht-repeat, or achso (Golato & Betz 2008). However, the correction in the present example receives only a minimal receipt by Gabi (line 7), and Anna's ensuing pursuit of a clearer action receipt (lines 8---9) evidences that Gabi's response was not sufficient: Anna further specifies the exact age of the children in question, thus supporting her claim to epistemic authority. Her turn is formulated syntactically as a continuation of the prior tum with denn/'because' (line 8), and closes with the tum--final response--pursuit niet,i'(you) see; (line 9; Harren 2001). Again, no response is immediately forthcoming. In overlap with a delayed repeat by Gabi of the earlier minimal response token (line 12), Anna now reissues her earlier age estimate (line 1.1), but specifies the earlier nau/'just recently' as ientz iam opriaL!'just now in april~ 'Ihis specification of the temporal reference is synta<..iically realized as an expansion, with ientz/'now' as the first element of this expansion. Through this increment (see e.g. Auer 1996, 2006; Couper-Kuhlen & Ono 2007; Edenstrom 2008, 2009; Ford et al. 2002), which expands and recompletes the potentially complete prior TCU, Anna further pursues an overt receipt of the new information. 'lhe continuation of this interaction is analyzed in detail below (Section 4.2). For the present purposes, it should suffice to note that the temporal marker ientz may be syntactically post-positioned, and that in this position, it appears as ientz rather than ientzter. For my collection of 61 examples of ientz(t), the following general observations can be made: All but 5 instances occur in turn-initial or turn-internal position. They occur in clause-initial or clause-internal position, and they indicate that the TCU underway is not yet complete. When speakers use them at points of possible syntactic completion (3 examples in my data), it can be shown that they rely on the basic prop-erty of ientz(t) as indicating designed incompleteness to construct an extended turn at talk (see Section4.4 for a discussion of such examples). 1hus, ientz(t) is used strategically in those examples: Its use relies on and thereby confirms its prototypical function as marking a turn as designedly incomplete. 7 1he next section will outline the use of the form ientzter in Siebenbti.rger Sachsisch. 4.2

Use of ien.tzter!'now' in spontaneous conversation

Data examples 9---·13 illustrate that ientzter, in contrast to ientz(t), is used in designed unit-final position. In addition to uncovering a nearly complementary distribution in the syntactic placement of ientz(t) and ientzter in SiebenbUrger Sachsisch (see excerpt 9 for a particularly clear illustration), I outline a specific interactional function for

7· The remaining two exceptions will be addressed in Section 4.2, Footnote 13.

426 Emma M. Hetz

which ientzter is put to use. Ientzter not only signals unit (TCU or turn) completion and can thus serve as one resource for incipient speakers to determine legitimate places for speaker change, but ientzter overwhelmingly also signals the relevance of speaker change (see excerpt 1.0). I show that ientzteris typically used at sequential and conversational boundaries, that is, at points at which the current topic is closed, a new topic or departure is launched, or a return to an earlier topic or action is attempted (see excerpts 11-13). 'Ihe first excerpt is taken from an interaction between Maia and her daughter-in--law Anke at a family gathering. The most recent topic recipes for nettle salad is closed in line 1 with a second assessment and double saying of the acknowledgement token chia/'yes' by Maia. When Maia's granddaughter Erna arrives at the table, Maia reopens the topic (lines 3---4). After a minimal response by Erna (line 5), Maia uses an assessment as a second pre-- telling move (line 6; d. Scheglotf 2007: 37) and this time receives a non-minimal response. In lines 8-11, Maia then provides advice about when nettles should be harvested for eating: before they mature. After indicating how a young, tender plant may be recognized by its size,8 Maia also provides a temporal reference point for her interlocutor: She contrasts the best time of the year for harvesting nettles with the present (line 11; ientzter and ientz refer to the last days of]une). (9)

branessle [53.54_av13/14_44:12] Maia:

gaut des BRAnlisle.=chia chia good the NEttles.=yes yes (they) taste good NEttles do.=yes yes (2.4)

2

3

sclunocke taste

Maia: ((to A:))

((gaze disengagement; Erna enters picture. Anke and Maia look up toward her.))

HOste schi GE:se:n iih spina:t HAve+you already EA:te:n uh spi:nach

HAve you ever EA:te:n uh (cooked) greens .h os

4

branessle?

11 from nettles? .h madefrom nettles? 5

Erna:

0

na(h) 0 =hh

ono(h)""-=hh 6

Maia:

.h tschmockt: fenj,

.h it tastes good,

8. A hand gesture indicating plant size accompanies line 10 but is not included in the transcript.

Adverbs in German conversation 427 7

Erna:

oh ich glivent. (hun nuer nod1 nie[t ge:se.) 0 oh I beh!:_Ve+it. (have only yet no[t ea:ten.) 0 0 oh I'm sure.((I')ve just not ha:d(it) yet.) 0

0 0

8

::::>

[OV- OVER [BU-- BUT

Maia:

9

nau (.) du wuse glech ereusku, PRT (.) then when-1-they immediately come Ylb (.)right when they come J&P,

10

wuse si gri :B sien °nuer.0 when·tthey this ta:ll are 0 only.o when they're this ta:R 0 only (then). 0 •M: head •M: dismissive shake hand gesture•· niet IEN.*tzter. *ientz siese* n[immi not now. now they're not a[ny more] good. not now. now they're rwt" g[ood any] more.

11

[ ojao [ oyeso

12 Ern a: 13

Maia: .hhh

14

Maia:

over du kust=uch(t) but then can+you=also( +the) but then you can also (the) gJ.ntschen uch &t[em-DIM also li[ ttle M.r:.m. as well

15

gaut.

u-] a-] a--]

[wuse noch gunz no- k]lin uch [when+they still very st- s ]mall and [when·t·they are stiR very st- s ]mall and (=> [when+they still]) 16

17

Erna:

uch: gang sie; a:nd young are; a:ttd young;

In her turn in lines 8-11, Maia uses the temporal marker 'now' twice, in two different forms: once as ientzter and once as ientz. The shape of the marker can be correlated with different placement within Maia's extended utterance. While ientz appears at the beginning of a new sentential TCU, that is, at a point at which the unit underway is designedly incomplete, ientzte1'completes the immediately preceding TCU and exhibits unit-final intonation. It should also be noted that the two TCUs follow each other closely, and that ientzter, although it completes a TCU, is not followed by a change of

428 Emma M. Betz

speaker or a continuer. 9 After ientzter, Maia continues her tum. The next unit in her utterance provides a restatement or elaboration of the point just made: that one should not harvest nettles 'now: since they are not palatable (gaut, line 11) any more this late in the year. Example 9 clearly illustrates a distributional difference between ientz and ientzter with respect to the unit underway: Jentz is used in non--final position while ientzteris used TCU-finally. 1he data in excerpt 9, however, also allow for the question whether the restatement in line 11 (beginning with ientz) was prompted by a lack of response by Erna and thus serves to recomplete Maia's tum. More generally, these data raise the question whether ientzter marks designed TCU completion only, or whether it makes speaker change relevant. Based on my collection of ientzter, I argue that ientzter indeed contributes to marking the relevance of speaker change in interaction. Examples 1.0-13 serve as representative evidence. We first revisit data sample 8 nenjtziihn (reprinted and extended as example 10). Recall that Anna and Gabi are discussing an acquaintance's family situation- specifically the age of his children. In response to and contrasting with Gabfs gris/'grown-up', used in a candidate estimate of the children's age (line 4), Anna announces that the children are kliener/'younger' (line 6). In two unsuccessful tries (lines 8-9 and 11), Anna pursues Gabi's receipt of this as new and corrected information. Consider now the sequel to extract 8: In her tum in line 14, Anna qualifies the age estimate she had provided in line ll., and then she performs an epistemic back-down in lines 15···17. The latter takes the shape of a disclaimer of the specific knowledge she had claimed in lines 6-11. (1 0)

nenjtzahn [43... av4/3 ...56:55] 4

Gabi:

de kentj sien over griS; tf1e children are grOWN though;

5 6

Anna: hhh Anna: <de ke:ntj sien nau> gor kliener wa aus hans.

all younger than hans. 7

Gabi:

8

Anna:

m:[hrn [(denn der (lsyll))) wor nau J-nenj•tzehn.•=

[(because (name)) (just) turned J-nine"teen."= 9

=niet,

==(right/see), 10

(0.3)

9· Erna's acknowledgement token ja/'yefi in line 12 is directed to Maia's just preceding hand gesture, which conveys the content of the last unit in Maia's turn non-verbally.

Adverbs in German conversation 429

11

Anna: tsk! [da WO]RE NENJtziihn. =ientz iam oprial. tsk! [they :W.ElRE nine.TE.EN:. =JlQ:W. in .\lpril. tsk! [they TUR]NED nine_'fE.f.,""{.=just" .11J!.W. in _qpril.

12

Gabi:

13

=>

[0 m:hm 0

]

(.)

14 Anna: OCHzlihn oder NENJtzlihn >wores=i~nt.~er.

EIGHteen or NINEteen >were+they=nQw. EIGHteen or NINEteen >they turned=(just)m.l't i[ch weis .l chia net< se: sa== I[ know ] PRT not< the:y (look)= I[ don't ] know< the:y (kwk)=

15

16

Gabi:

[0 m:hm,o]

17 Anna: =chia grod gena:u

>wa aulte nau wor.< =PRT just exa:ctly >how old+he PRT was.< =just exa:ctly >how old he (actualM was.< 18 Anna: .hhh wai ich weiss der hanzi wor klein l1hh (cause) I know the ((name)) was little .hhh (cause) I know hanzi was little ((Story about Anna's son follows.)) In line 14, Anna qualifies her earlier statement by combining a partial repeat of elements from the prior turn (the verb phrase and subject, the numeralnenjtziihn, and the temporal reference to the present) with the additional ochziihn as another pos·· sible age estimate. 'Ihe repeated adverb 'noW, however, takes a different shape in line 14 (ientzter, vs. ientz in line 11). We can again note that the two adverbs appear in different syntactic and sequential environments. Ientz introduces an increment to a possibly complete TCU, while ientzteroccurs in TCU-final position and exhibits unitfinal intonation. Similar to segment 11, ientzter appears in TCU-· but not tum- final position. Here, however, we can observe a clearer orientation by the co-participant to a possible transition-relevance point: After ientzter and in overlap with Anna's turn continuation, Gabi provides a minimal response token. Considering the action of the larger sequence (Anna's pursuit of a response) and the placement of ientzter, this example suggests that ientzter contributes to making speaker transition relevant. Let us consider yet another example. Data example 11 is taken from the beginning of a phone conversation between Gabi and her sister Christiane. 'Ihey are coordinating afternoon plans. Gabi was expected to be at Christiane and her husband Dragu's house at the time of this conversation, and she has called to provide an update about her whereabouts. In lines 19-20, she announces that her visit to Christiane's house is still part of her afternoon plans but will come to pass later than expected.

430 Emma M. Betz (11)

gorte [14_0 02_571]

2

3

4

Chr: wli biaste. where are+you. where are you. Gabi: ba der muetter, with the mothe~ at mother's,

a mer kun ient:z do:r. look we come now there. look we're going there now. Gabi: mer hu schin, we have already.

Chr:

( ... )

glider ian de gorte? where gQ-1-youPL into the garden? where are you gjf to the garden?

12 Gabi: wor

13

(0.2)

14 Chr:

no 0 =n f A::i ech hun nau geworte sillt PR'f 0 =niO::i I have PRT wait:ed+youSG should 0 (well/but)" n'i'O::i I waited for you to 0 0

15

hA:r kun nau sieder nimmi kun hE:re come PRT are-tyouPL not any more come come o:ver and you didn't come

16

nau hummer geslit miatem dragu PRT have+we said with+the ((name)) so we, dragu at1d I, said we'll

17

nau e:se mer, t(r)li kummer zer muetter. now e:at we, then come+we to-1-the mother. e:at tww, then we7l go over to mother's.

18

(0.4)

19

=>

Gabi: no: ed1 wul nau km:t.= PRT I wanted PRT _g)_me.= (welJ) I did want to come.

20

::::over nau nie(t) gluad ientzter. =but PRT not just now. =but ttot just)&..

21

(.)

22 Chr:

no 0 trli tkummer osem gor iten har. 0 PR'f 0 then icome+we from+the gariden here. 0 ( weli)0 then we'll i come (back) here from the gar i den. 0

Adverbs in German conversation 431

With regard to its placement within the clause and turn, we can note that (nie(t) gluad) ientzter is used in an expansion of the first TCU of Gabi's turn (line 19) to specify the time for her visit as 'not just yet: The adverb constitutes the last element of this tum and exhibits final falling intonation. With regard to the larger sequence, we can note that ientzter is used at a point at which the trajectory of the topic and interaction is determined: After an insertion sequence initiated by Gabi (lines 4-11, omitted from transcript) and related to the circumstances of her call, Gabi steers the interaction back to the 'here and now' of her co--participant (line 12) and then of herself (line 20). In line 1.2, she attempts an understanding check by reissuing previously received infor·· mation (line 3) for confirmation. Christiane's emphatic rejection of this candidate understanding (line 14) marks a mismatch between Christiane's expectations and Gabi's displayed knowledge, and it is followed by detailed accounts (e.g. reiterations of prior arrangements). In lines 19···20, Gabi first responds to the implicit accusation Christiane's rejection carried. She then shifts from past plans to the immediate present, thus effectively marking a sequential boundary. After a short gap in the conversation, speaker change occurs: Christiane aligns with the sequential and topical shift by announcing a revised plan f{)r the afternoon (line 22 ). In addition to providing another example of turn-final ientzter, excerpt 11 evidences that ientzter tends to occur at sequential boundaries, that is, at points at which speakers manage the direction of the topic. 1his observation holds across my collection of unit-final ientzter. Excerpts 12 and 13 provide further examples of ientzter occurring tum--finally at topic and sequence boundaries. In both, ientzter appears as the last element of an announcement tum, at the beginning of either a new topic (as in 12) or a new aspect of a larger topic (as in 13; the topic is health). (12)

elin [25_0-03-B_125]; Lhasaskedabout M'sgeneral well-being. 1

Maia:

o:h marlene. ech ha. ich bian nuer elin ha? o:h marlene. look here. I am only alone here?

o:h marlene. you know. I am all by myself here? 2

::::>

de

kentj

the

children are not now- now

sien niet gluat-· gluat niet

not"

"problem with cough" "J-hajen.==ie-.i==( (cough) )"iientzter,

3

*J-here. =no-J-=((cough)) "i now. (.)

4

5

Maia:

mhhh mtsk mhhh mtsk

ent nau bian ich nuer elin. and PRT am I only alone.

mhhh mtsk so I'm just by myself. 6

Lene:

no ne hiaste PRT PRT have+ you

kom denje friden. at last your peace. well ther1 you ht:zve your peace t:zt last.

432 Emma M. Betz 7

Maia:

.hh chia hehehe

hehehe

.hh yes 8

Lene:

•na.•

•( well/so. )• 9

Maia:

no [nau, PRT [PRT,

well [then, [•(dias• ]der sd1a) wa geideter. [•(it+is• ]PRT nice) how goes+it+for+you.

10 Lene:

[~(thats 0]

(13)

nice) how are you.

zer kor1troB [4_0-02_281.286]; .M's low results on tests for blood clotting indicators and the intervals at which she has regLllar check-ups are discussed. Maia:

2

eli wi [al. he wa[nts (it like that). ((he= M's physician))

Tanja:

[ 00

m:h,••

[oom:h,"" 3

Maia:

chia. chia.=

yes. yes.4

Tanja: ==aha,(.)

m:[hm.

=uhuh,(.) m:[hm. [chia.

Maia:

[yes. (0.4).

6 7

Maia:

si niedrich ias llh h nau giin ich chia wiele PRT because+it so ]QW is llh h PRT go

.hh h (so then) I go because its so .l121¥.

=>

uch endjen nuer ba zwa: wochen=ientzter,

8

also these days on~v every two: weeks=now, 9

(.)

10 Maia:

.H .H

mies must

kontroll ba ich mare giin zer I tomorrow go to+ the check-up for

.H I have to go tomorrow to den (.) darm? the (.) intestines?

11

my(.) intestinal check-up? (.)

12 13

M.aia:

14

Tanja:

maren noch L.;;t[er tomorrow again on[ce

tomorrow once mo

[re [•mmhm• [ 0 m mhm 0

Adverbs in German conversation 433

In both examples, Maia's announcement (lines 1-3 and 7-8 respectively) serves as a topic proffer and makes a response relevant. This is evidenced by the subsequent interaction, in which such a response is pursued: In both examples, there is a delay in the co-participant response after ientzter. At this point, the first speaker reissues her initiating action (the first pair part, FPP) (example 1.2) or continues the topic-initiating turn syntactically (example 13). With the respective second try at a topic proffer or redirection, Maia pursues an uptake to the proffer or an alignment with the redirection of the topic. In example 1.2, a response ensues. In example 13, the new topic receives an uptake only after the third try (line 13), and this uptake is minimal and thus dispreferred (Schegloff2007: 173). For both examples, we can note that ientzter is used turn- finally at a sequential and topical juncture, that is, at a place at which speakers do visible interactional work to launch a new topic or a departure from the current topic. A connection between adverb placement and the management of topical and interactional sequences has recently been documented by Edenstrom (2008) in a study of syntactically post--positioned temporal markers in German. Edenstrom argues that temporal adverbials in certain syntactic position systematically function as covert markers of sequential and topical boundaries. In other words, they are used to facilitate a return to a previous sequence or launch into a new action trajectory. Structurally, the temporal markers analyzed by Edenstrom all constitute syntactic right-expansions (see tootnote 3), that is, they are added after a turn constructional unit has reached a projectable syntactic and pragmatic (though not necessarily prosodic) completion point, for example after the right brace in a German sentence. Many of the examples of ientzter in my collection are of this kind: 'Ihey are used in sentences that exhibit a sentence brace and occur outside of this brace (cf. examples 12. 13).1° These ientzter-turns all constitute FPPs and mark sequential and topical boundaries, that is, they are deployed systematically to control or redirect a course of action: With an ientzter-turn, speakers nominate or pursue a topic or action (9 branessle, 10 nenjtzcihn, 12 elin), or they redirect a topic (11 gorte, 13 zer kontroll; see also example 18 drinking). We can now describe the prototypical use of ient.zter in Siebenbtirger Sachsisch. In my collection, ientzternever occurs in turn- or TCU-initial position. Overwhelmingly,

to. Recall that predicate complements may constitute right boundaries, although they provide weaker syntactic closure than non--finite verb elements. Examples 9 and 11 seem less clear examples of syntactic right-expansions, but the larger turns in which ientzter is used as the final element are clearly built incrementally; as continuations of turns-at-talk. In excerpt 10, ientzter is produced as the final element in a TCU v.'ithout sentence brace. Again, however, the syntax of the larger turn (lines 11---14) emerges in a piecemeal fashion similar to rightward e:A11ansions (see also Edenstrom 2009, Couper-Kuhlen & Ono 2007, Vorreiter 2003 ongiue-ons and Ford et al. 2002 on extensions in English).

434 Emma M. Betz

ientzter is used in unit-final position (8 out of 10 exarnples), 11 typically in tum-final position outside of the sentence brace and thus of a TCU, and it exhibits stress and terminal intonation (Selting 1996, 2000). The analysis of unit-final ientzter-tums within the larger sequential conte..~ts in which they occur shows that all are sequence-initiating moves (F.PPs). They typically constitute announcements, noticings, or informings, in other words, typical first moves for topic proffers, nominations, or shifts (Button & Casey 1984, 1985, 1989; Jefferson 1984; Sacks 1987; Scheglofl.2007). Indeed, all unit-final ientzter··tttrns are used at sequential boundaries to manage topic development Here, they seem to be instrumental in pursuing the uptake of the new topic or topic direction. How can we account for ientzter (but not ientz) fulfilling this function in interaction? First of all, we have noted- based on a purely distributional analysis- that ientzte1' has its 'home' in unit--final position, carrying terminal intonation. It thus syntactically and prosodically marks designed completion and a potential transition point. Secondly, it has been shown that temporal adverbs used in non-canonical position, 12 for example in rightward syntactic expansions, serve discourse functions beyond the level of the sentence. These include the marking of sequential and topical boundaries (Edenstrom 2008). At such boundaries, the distribution of speakership and recipiency and the trajectory of the current topic are negotiated. In my data, ientzter occurs exactly at such sequential places at which speakers do visible interactional work to steer the development of the topic and sequence. Thirdly, it has emerged from my data that in all examples but one (excerpt 9 l1ranessle) of unit-final ientzter, turn·· transition is clearly relevant before or at the point at which ientzter is uttered: V\r'here speaker change does not occur after ientzter, its absence is treated as problematic. and a response is actively pursued by the ientzter-·speaker (see excerpts 12 and 13). 'While ientzter used at the end of a sequence-initiating and topic--directing turn cannot be shown to alone make speaker transition relevant (as, for example, a response pursuit would; Harren 2001), my data suggest tl1at it contributes to signaling this relevance as part of a bundle of features. Different types of sequence initiating moves differ in their 'sequential implicativeness' (Schegloff & Sacks 1973; Scheglotf 2007), that is, in the strength with which they make a (specific type of) response relevant. Hence,

n. I define 'unit-final' as constituting either the last element of a designedly complete TCU or (a part of) a syntactic e>.:pansion after possible TCU completion. Such increments can be prosodically integrated (see excerpt 13) or emerge after a prosodic break (see 12). Note that the two examples in my data in which ientzter occurs in TCU-medial position are discussed in Section 4.3 below. In my data of ientz(t) and ientzter, the adverb occurs in clause-initial or medial position in 61 instances (59 of ientz(t) and 2 of ientzter) and in designed clause-final or turn-final positi.on in 10 instances (8 ientzter and 2 ientz(t)). We can thus speak of the unit-final position as 'marked; or 'non-canonical' for this temporal marker. 12.

Adverbs in German conversation 435

actions such as informings, noticings, and announcements that are done as declaratives differ in their sequential implicativeness from the same (and other) actions that are done as questions. In sequence-initiating moves that are not grammatically packaged as questions, sequential strength - and thus the strength of the topic-directing move can be increased by adding, tor example, a response pursuit marker in turn-· final position. Ientzter in w1it-final position seems to function similarly: In my collection, all examples but one (example 18 drinking below) constitute informings, announcements, or noticings that are packaged linguistically as declaratives. The addi-· tion of the two--syllable temporal marker, typically as a syntactic increment and often in a second try at eliciting a response, actually enhances the strength of the sequential implicativeness of a turn and thus the likelihood that the co-participant will take up the new topic or topic departure. 13

13. This understanding of the use of ier1tzter sheds light on a small number of ientzt (2 examples out of 61) that occur turn-finally and thus constitute exceptions in my ientz(t)-collection. Similar to ientzter in its prototypical use, both constitute sequence-initiating moves, but they instantiate actions different from the ones described for ientzter. Grammatically, they constitute wh-questions. Consider data example 14 below:

(14)

dege.siangth!t [4.1_0-02_281.286] 1

Maia:

osi

git

schnieller veiba. (h)

that way it will be 2

WELL: let- >uh< '·">

over faster. (h)

Tanja: NAJA: llis- >lih< llisemer wot- it

3

wa

let~

gidet,~hllh: (.)

what- uh how are,~huh: (.) 4

Maia: =.hh tgitmer niet licht.

=.hh I'm der

nau,"~

dot

not talk about this aey more,'"

i dir nau ientzt?,.~h'·" i J!illl. now?,.Jt,, miat dem

not too bad. with

,~vMASC

gesiangt·-

heal-

gesiangthit?=rnhh

myF&.\1 health?=mhh (.)

6

.H[H

7

Maia:

8

Tanja:

9

Maia: =ech hun nimmi geho:t, da anfltl.le dii?

[ehm.= [uhm.= =l have

rwt f:a:d (them) any more, those attacks (I have)? ((=asthma attacks))

I have described unit-final ientzter as a temporal marker that serves to promote a topicchanging move by enhancing the strength of the sequential implicativeness of the first turn in the sequence: an announcement, noticing, or informing done as a declarative. If we take tb.is as its main interactional function, then the use of ientzter at the end of a topic-initiating

436 Emma M. Betz

4·3

Summary of findings for ietttz(t) and ietttzter in Siebenbiirger Sachsisch

The present analysis is based on 71 examples of a temporal marker in spontaneous interaction in Siebenbiirger Sachsisch, a settlement dialect of German. The adverb occurs in two forms: ientz(t) and ientzter. There is no observable semantic difference between these two lexical variants. 'Ihe variants are not the result of a phonetic reduc-· tion process, whereby ient:zter would be reduced to ientz(t): While -er endings may be reduced in German when the syllable is unstressed, this reduction typically results in an unstressed schwa--sound and not in complete elision of the syllable (Schwitalla 2003: 38-9). 'Ihe preceding analysis has show""Il that the difference between the two tokens is functional: Ientz(t) and ientzter are position-sensitive, that is, the form ofthe adverb depends on its place in the TCU, turn, and sequence-in-progress. The shorter variant occurs overwhelmingly in designedly non-final position in a turn constructional unit, usually in the syntactic front- or syntactic inner field of a sentential TCU. The longer variant never occurs in unit-initial position. It may occur in TCU-medial position (see below for a discussion of those instances), but it typically occurs in turn-final position, either TCU-finally or as an increment. Ientzter occurs much less frequently than ientz(t) (10 vs. 61 instances in my collection). The latter observation confirms that the sentence--initial and sentence--internal positions, described as the canonical places for temporal adverbs in a sentence (Weiss 1975; Frey & Pittner 1998; Pittner 1999), are also the unmarked positions for the temporal adverb carrying the semantic content 'now, at this moment' in Siebenbiirger Sachsisch. In these positions, the adverb appears as ientz(t). The form ientzter, by contrast, is reserved for the marked, unit-final, position in a limited number of well-defined sequential environments. This study has shown that, where such lexical variants are available, speakers use tb.em as a resource for constructing tmits and sequences in conversation, and specifi·· cally for signaling the current status of the larger unit underway: By using ientz(t), a speaker marks this element as designedly non-final and the TCU as still incomplete. By using the variant ientzter, a speaker marks the adverb as a TCU- or turn-final element and thus the larger structural unit as designedly complete. The use of the marked form of the temporal marker can be further specified: Speakers use ientzter in sequenceinitiating moves at sequential boundaries, and they tend to use them in syntactically

move that takes the shape of a grammatical question may not be interactionally necessary; a response (the answer to the question) has already strongly been made relevant. With the caveat that the number of examples of ie,ltz(t) in this position and environment L'> very small, I would argue that the sequential factors that motivate the use of Jentzter in all other examples in my data are not present here; thus, we see le,1tzt. We can maintain that the unmarked positions for ientz(t) in Siebenbiirger S!lchsisch are unit-initial and -medial.

Adverbs in German conversation 437

marked positions: in increments. In such positions, ientzter advances the uptake of a new topic or topic trajectory. l-Ienee, ientzter operates on the discourse-level. I have two examples of ientzter in which this form of the adverb does not occur in unit-final position. For this use of ientzter in sentence-internal position, only a few observations can be made. In contrast to all other ientzter in my data, these occur in responsive moves (second pair parts to questions). There is no indication in these examples that recipients orient to ientzter as marking the end of a TCU or turn. In both examples problems in epistemic intersubjectivity surface, and they are addressed through an explicit contrast of a prior state--of--affairs with the immediate 'here and now' ("ientzter"). It is possible that ientzter in medial position appears systematically in turns designed to correct a mismatch in expectations or assumption between interactants. In these turns there is a "heightened focus" (Tanaka 2001: 566) on the current state of atfairs carried by a two--syllable, stressed form of the tem-poral adverb. This suggests that ientzter in medial position serves a function very different from ientzter in final position. However, more data is needed to explore this function. My tl.ndings on the prototypical use of ientzter, specifically its discourse-level function, confirm findings by Edenstrom (2008) on post-positioned temporal markers in Colloquial Standard German. Beyond this, however, my analysis shows that in at least one other variety of German, this systematic interactional use of a time adverb for promoting a new topic or topic direction is actually reflected in the lexicon. That is, there exist two forms of the adverb 'now' (with the semantic content 'now; 'at this moment; or 'just now') that are - in their respective prototypical use - complementary in their distribution and distinct in their interactional use. Thus, the present analysis locates the principled use of ientz(t) and ientzter in Siebenbiirger Sachsisch at the interface of word choice and interaction, of morpho-lexis and pragmatics. In the following section, I present two further examples of ientz(t) and one of ientzter, including another segment in which both variants of the temporal adverb are used by the same speaker in the same turn-at--talk. These excerpts constitute intriguing examples of the 'strategic' use of ientz(t) and ientzter in interaction and thus further support my analysis of their prototypical functions.

4·4

Additional support: The use of ientz(t) and ientzt-er in syntactic pivot constructions

This section presents further data to support this distributional and sequential analysis of ientz(t) and ientzter as interactionally position-sensitive adverbials: l discuss two examples where something notable occurs in the negotiation of TCU boundaries. In these examples, it can be said that ientzter is used strategically to promote the specific interactional goal of extending a turn in progress beyond a point of possible syntactic

438 Emma M. Betz and pragmatic completion. In order to do this, speakers rely on the basic functions of ientz(t) and ientzter as indicators of the status of the unit underway. Turn extensions in German rely on various syntactic and prosodic mechanisms, one of which is the withholding of prosodic signals of turn completion (Betz 2008: 40···44; cf. Local & Walker 2004). Examples 1.7 and 18 below illustrate that speak·· ers of German may also withhold the signaling of completion (and thus, of transition relevance) through lexical means: by uttering ientz(t) and not ientzter at a point of possible syntactic--pragmatic completion. Three examples in my collection appear at first pass to constitute examples of ientz(t) uttered in unit-final position. However, in all these examples, a close analysis reveals that within the unfolding unit ientz(t) is used in a position which is not designed to signal completion: Ientz(t) is followed by a turn extension, or is part of a syntactic pivot construction (c[ Betz 2008; Noren 2007; Scheutz 2005; Selling 1998: 239-240; Walker 2007), that is, it occurs in a construction in which two potentially independent structures are connected (or merged) by way of a shared element to form an extended TCU. In pivot constructions, signals of designed completion are withheld at the points at which a transition from one syntactic construction to the next occurs. The shared element can then syntactically and prosodically be part of two different constructions. Example 15 illustrates the basic structure of a syntactic pivot construction in English: Here, the NP the bone can, based on syntax and prosody, be said to belong to two difterent syntactic constructions (it is an object NP in lmt l('d) lo:ve the bone and a subject NP in the bone was so:: beautif) and thus constitutes the structural center of an extended TCU. Prosodic. design, notably the lack of prosodic breaks at the potential syntactic boundaries, is essential in allowing the whole construct to be uttered and understood as one unit. (15)

the bone [NB.IV.3-185s;··pJV005] 1 Emm: but I('d) l:ove the bone was so:: 2 beautif eh the pink was exquisite

(taken from Walker 2007: 2219)

The central element of a syntactic pivot construction can also be a prepositional phrase. Consider example 16, in which the temporal expression 'every sunday' can be heard as belonging to two different constructions, 'but she goes every sunday' and 'every sunday she goes to church: (16)

rosi [O 03_344)

Lene:

et kl:d=over en jAde suandje kidedande Kirch. she cO:mes=though on Every sunday comest·shet·int·the CHURch. but she gO:es gvery sunday she goes to CHURch.

We now take a closer look at some such constructions that involve the use of ientz(t) and ientzter. I show that the choice between ientz(t), which indexes 'unit-in-progress'

Adverbs in German conversation 439

and ientzter, which signals 'designed unit completion: is a lexical strategy that contributes to making possible syntactic pivot constructions actually bearable as one unit. In excerpt 17, Bert and Maia are discussing the weather in their respective hometowlls. Maia's turn in lines S-6 is the focus of analysis here. Built to be contiguous with her prior turn in which she announces that it rained last night (HA hodet he:ntj gerent., line 3), Maia's turn is potentially complete after ientzt with respect to syntax and pragmatics. It is, however, not prosodically complete. Moreover, the potentially last element in this construction is the adverb ientzt, the lexical item which, as I have shown, indexes the current TCU as incomplete. Maia's turn continues with a closely latched-on verb, the second element in a German declarative sentence. (17)

tv.•adder [2___ 0 05--B..JJOS] 1 Bert: 0 esi0 ko[mesch, et wor] esi lidrt, ~so 0 str[ange, it was] so bad, 2

Maia:

3

Maia: HA hodet HE:ntj gerent. HERE has+ it l!!.;§.t. night rained. HERE it rained l~;M. night.

4

Bert: chia? really?

=>

=>

[hhh

Maia: .h enta HEtj !niet. h and-1-then today !not. atul then todt:ty (it did) not. bes I e>j=.J,i~tzt--=(h)odet ugefongen.=ech.< until> ilook=.J,nQw.=(h)as+it r.tarted.=look.< (rain) until >just=now=it started.=look.<

6

7

Maia: ikntz wa=mer hi:m=sie ku fu-1-det nau u. i now that we're ho:me its .J,sttlrting.

8

Bert: chia; yes;

9

(0.5)

10

Maia: m.h[h

11

Bert:

[no dias chia gaut. [wen alright then.

By omitting prosodic markers of TCU completion, the speaker produces HEtj niet bes ej=.i.~_ntztl'(it did) not (rain) today until just now' and ej==i?..ntzt==(h)odet ugefongen.!'just now it started: under one intonation contour and thus as one inter-actional unit. (Ej=)~ntzt appears as both unit-final and unit-initial element in this construction. Thus, by omitting prosodic signals and using i?..ntzt at a potential

440 Emma M. Hetz

syntactic boundary, the speaker signals non-completion prosodically cmd lexically at this point in her evolving turn. In data sample 18, Gabi and Anna are talking about a member of Anna's family. This third party used to criticize her own son - Anna's cousin - k)r his drinking habits. In line 2, Gabi self-selects to inquire about the most recent developments in this family matter. (18)

drinking [av1/2...4....19:10] 1 Anna: [>(no en danj)< [>(PRT a thing)<] [

=>

]

*(A: gaze to G)

2

Gabi: [>no nau ien]tz< wot .1!P.I~:~.;~;htse nau ientzter, hh* [ >PRT PRT no]w< what ~m;.r.sshe PRT now, hh [ >PR't PRT tw ]w< what" does she sa:y PRT tww. hh

3

Anna: 0 no 0 sai drantj ud1 •PRT0 she drinks .f.Q.!l. •(we0.) 0 she drinks as well

4

(0.3) •·(G: ga7j! shift away from A, unbelieving fadal expression) Anna: 0 sa *drantj ja ientzt .Y~;h0 0 she *drinks PRT now .t.QQ.0 "she *drinks now as ~11·

After emerging from overlap with Anna in line 2, Gabi continues her turn. 'Ibis turn continuation could either be a syntactic continuation or a new beginning (cf. Scheglotl1987). No pause, cut-off, or other prosodic break occurs that suggests that wot/'whaf constitutes the beginning of a new TCU: Contrast 'no nau ientz, wot spdichtse'/'so now, what does she say' with the actual data: no nau ientz wot spri:ichtse!'so now what does she say'. 'Ihus woti'what' constitutes a prosodic continu-ation of the unit that was initiated in overlap. It can also be a syntactic continuation: In Siebenbiirger Sachsisch (as in English), a wh-question may take additional elements (other than conjunctions) before the question word; that is, it may have a filled, prosodically integrated pre-front-field (cf Eisenberg 1989: 411-12). The question pronoun, therefore, is a syntactically possible next element after ientz in the linear progression of the turn. The use of ientz over its variant ientzter contributes to marking this syntactic place as non-TCU-final. Gabi's turn reaches a possible syntactic and pragmatic, but not prosodic, comple·· tion with ~i:i:chtse!'says she' (line 2). No terminal falling or rising intonation (both are possible here) occurs at the end of 5P-.ri:i.:chtse. Thus, wot .$J.J.ti:i.:chtse is bearable as the last constituent element of the TCU ientz wot Wt.ii:chtse/'now what does she say' and

Adverbs in German conversation 441

also as the beginning of wot ~ii:chtse nau ientzter!'what does she say now'. The clause wot wra:chtse is part of two possibly independent constructions and therefore a pivot element (see 19). 14 (19)

Pivot strudure and shape of adverbials: drinking >no nau ientz< wot §P.l~_;!;;btse wot sprJ:chtse nau ientzter, hh >no

nau

ientz< no-w

(.~o)

wot ~ra:chtse nau ientzter, hh -what does she say no-w? hh 15

I argued in Section 4.1 and 4.2 that ientz(t) is marked as a non-final token, while ientzter is marked as a unit-final element. In the present example, both the particle nau and the adverb ientz are recycled in post-overlap (after ientz), post-pivot position (after the verb) in this tum (see 20), but the temporal adverb now takes a different shape: Jentz is recycled as ientzter. (20)

Shape of recycled elements: drinking nau ienster, hh ientz< wot sprii:chtse (so) no-w -what does she say now?hh

>no nau

This example allows for the following general observations to be made: 1.

Jentz occurs at a point at which the TCU could be complete but is prosodically not

2.

marked as such. If the goal of the speaker is to suppress signals of unit completion, the choice of ientz over ientzter contributes to withholding such a marking. 'Ihe choice of ientz over ientzter is a choice not to mark the TCU as complete and wot as a new beginning. The first part of Gabi's talk is uttered in overlap. The elements in overlap (except fi.)r the position-sensitive no that connects this tum to prior talk) are recycled. The sentence--initial ientz is recycled tum--finally as ientzter, the variant marked for designed completion. The speaker thus marks the turn as complete after ientzter, and it is also understood as such: Speaker change occurs.

Both excerpts presented in this section constitute examples of a 'strategic' use of ientz(t) and ientzter in interaction. This use relies on the systematic difference that the two forms of this temporal marker exhibit in indexing the status of the unit-in-progress. Excerpts 17 and 18 thus lend further support to the preceding analysis of the basic interactional functions of ientz(t) and ientzter.

14.

For a more extensive discussion of this data example, see Betz (2008: 56-61).

15. I translate the idiomatic gloss here more freely. Specifically, I deviate from the original German stress pattern in trying to convey the syntactic pivot in English.

442

Emma M. Betz 5·

Summary and implications

This section offers a short summary of my findings and a discussion of their implications tor the study of syntactic and lexical regularities of spoken interaction (Section 5.1). I place the phenomenon of position-sensitive adverbials in its broader context by, on the one hand, pointing to other adverbials in Siebenbiirger Sachsisch that possess le.xical variants, and, on the other hand, taking a closer look at Bavarian German as another variety of German that exhibits a similar phenomenon (Section 5.2 ). In this section I also ask how the lexical variants documented in the present study should be linguistically classified: Should they be viewed as two fonns of one word or as two separate lexemes? Furthennore, I problematize a semantics-driven understanding of temporal adverbials and call for an inclusion of syntactic--pragmatic aspects into the definition of adverbials (Section 5.3).

5-1

The interface of morphology, word choice, and pragmatics

Certain temporal adverbs in Siebenbi.irger Sachsisch possess two position--sensitive lex-ical variants, a one-- and a twn--syllable variant: ientz(t)l'now' and ientzter!'novl. These are nearly complementary distributed and convey information about the stalus of the current turn-at-talk as well as the larger interactional sequence. The use of ientz(t) signals that a TCU is designedly incomplete, while ientzter used within a TCU signals that it is designedly complete. Moreover, ientzter is used at spedfic sequential junc-tures at which it promotes the uptake of a topic-changing move. 1hus, the placement of the adverbs ientz(t) and ientzter at a specific point in a sentence is determined by the regularities of turn and sequence design rather than by syntactic rules. When looking exclusively at the distribution of ientz(t) and ientzter in Siebenbiirger Sachsisch, 16 we can maintain the view gained tb.rough traditional sentence field analysis as well as the generative view that the unmarked or base position of temporal adverbs in a German sentence is in the inner field. However, only a sequential analysis reveals that in the case of ientz(t) and ientzter tb.is is due to a specialized use of the adverbs in interac-tion: lentzter is limited to sequence--initiating moves at sequential junctures at which a speaker sees the need to employ additional strategies for promoting a new topic or topic departure. Here, the adverb typically appears outside the sentence brace. Thus, ientzter functiom at the discourse rather than the sentence level. My findings add to existing interactional linguistic research that documents the complex connection between linguistic form and interactional demands. More specifically. it lends further support to studies that view syntactic increments as a

16. A repeat of the nu.nibers for convenience: My data yielded 61 examples of ientz(t) and 10 of ientzter. 10 of these (8 of ier1tzter, 2 of ier1tz(t)) appeared in unit-final position.

Adverbs in German conversation 44.1 systematic practice that fulfills describable functions in spoken discourse (cf Auer 1996, 2006; Couper-Kuhlen & Ono 2007; Edenstrom 2008, 2009; Ford et al. 2002; Kern & Selling 2006; Vorreiter 2003 ). In addition to showing a connection between the ordering of elements in a turn (syntax) and the local interactional goals a speaker pursues (pragmatics), this study establishes a connection betw·een pragmatics and word choice in Siebenbti.rger Sachsisch. Two observations are important here: 1.

2.

While previous studies have focused on the importance of prosody and syntax for the construction of units in conversation, this work is a first step toward expand-ing the discussion to the area of lexis/word choice. 'Ihe findings trom this study thus have general implications for the study of units in conversation. They offer insights into how word choice signals the status of a unit underway and thereby provides interactants with important infi.)rmation f()r negotiating turn-taking. If we view ientz(t) and ientzter as tw-o forms of one word with one of the nvo as the 'basic' fonn, l"l this work also has implications for the study of morphological processes in Gennan: By documenting how the status of a discourse unit can be indexed through the choic.e of one lexical variant over another, the discussion of ientz(t) and ientzter illustrates that the process of affixation serves not only to mark lexical category or add grammatical information but also to add functional-interactional information to lexical items. The meaning of such a le.lC:eme can thus only be grasped by combining a description of its semantic referent with a description of the constructions and intemctional contexts to which the lexeme is bound (cf. Deppermann 2006). My findings on the typical use ofientzterin particular corroborate findings by Edenstrom (2008) on adverbs of time used in right-expansions in Colloquial Standard German. Edenstrom shows that post--positioned temporal markers are crucially different from their homonyms in clause-internal position. The former function at the discourse level: They promote new action trajectories and thus secure intersubjectivity (ibid.). Aspects of sequential placement thus create different 'meanings' and functions for one lexical item and lead to 'functional polysemy: 18 My analysis

17. The existence of two adverbs in Siebenb!irger Slichsisch with the semantic content 'now' and different functional-interactional roles raises questions of grammaticalization and lexical

classification of these items. Outlining a grammaticalization path (Ban & OhlschUiger 1998; Hopper & Traugott 1993; Traugott 1997; Wischer & Die\\--al.d 2002) or pragmaticali:l.ation path (Auer & G\i.nthner 2005; Streeck 2002) for ientz(t) and ientzter is beyond the scope of this paper; however, the near-complementary distribution and the functional specialization of the two forms suggest that they sbould be analyzed as two forms- or functional variants- of one word rather than as two different lexernes. For a similar argument regarding the context-bound description of the 'meaning' of response particles, see Betz and Golato (2008) for German and Sorjonen (200 I) for Finnish.

18.

444 Emma M. Betz

shows that, for at least one variety of German, this systematic interactional use of a time adverb for marking sequential and topical shifts is actually reflected morphologically. That is, there e..xist two forms of the adverb 'now' (with the semantic content of'now') that are nearly complementary in their distribution and distinct in their interactional use. 'Ihus, this study sheds light on a phenomenon that lies at the interface of morphology, word choice, and pragmatics.

5.2

Lexical variants in different varieties of German: A broader view

This shtdy presented a detailed analysis of the adverbials ientz(t) and ientzter in Siebenbi.i.rger Sachsisch. This lexical item was selected because it was the first for which I noted the existence of functional variants that had no equivalent in my Colloquial Standard German data. In Siebenbii.rger Sachsisch, however, such lexical variants exist for a number of adverbials. They include the following 'pairs: The members of each pair differ in the number of syllables.

1. 2. 3. 4.

ist and ister ('once') 19 ualdist and ualdister ('sometimes; 'every now and then') irest and irester ('somewhere') nirest and nirester ('nowhere')

The examples in my data include adverbs of time and place. Preliminary analyses of small collections for these tokens show tendencies similar to those described for ientz(t) and ientzter in terms of a complementary placement within turn-constructional units. However, although a preliminary distributional analysis yielded similar results to ientz(t) and ientzter, it stands to reason that a sequential analysis would uncover different or additional interactional functions for these adverbials. A larger collection is necessary to carry out such an analysis. Further research is also needed to determine if such variants exist in other varieties of German and to uncover the functions they serve. Below, I present data from Bavarian as a first step toward this goal. In Bavarian German, there exist two temporal adverbials with the semantic content 'now:' eytz and eytza(t). Data example 21 exemplifies their use in spontaneous Bavarian German interaction: 20

19.

See example 13 zer kontroll, line 13, for an instance of turn-final

ister.

20. I thank Veronika Drake and Derek Drake for making data segments from their Bavarian telephone data available to me. 'Ihe data I had available (four conversations, approximately 1 hour in total), yielded 44 analy1-
Adverbs in German conversation 445 (21)

kr4fee [2007-10-20_H0.10_13:08] 1

Hans:

(0.2) > Rainer is aber sdJ,kcht (0.2) >(for) ((name)) (it) has though ]2ru,lly (uhm) (0.2) Rainer was in b_{!d.

ehm (uhm/the) gange gone

2

vor ner woch.< 0 ::::eh0 a week ago.< 0 ::::(uhlhuh) 0

shape a week ago.< 0 =(uh!huh) 0 (0.1)

3

=>

4

5 6

Hans:

wei cause

S·· (i!:htS

cause

it- [is he well ttow.

em i-· [goes+ it him

Om a: Oma:

na

PRT

_ey.t:za.] now.] ]

[ja::

wos: I(.)

Lvea::

what:] (.)

eigentlich scho wieder gWANdert. actually already again WALKed.

(he was) actually already out and aBOUT again. (0.6)

7

=>

8

Oma:

hhh

9

Oma:

eytz giiits=em fai wieder ganz guet; now goes+it=him PRT again pretty good;

10

Oma:

=[kaffee derf er (bi ]ssla) trinke. oi tass, ]ttl e) =[coffee can he Oi drink. one cup,

11

Han.'>:

now he5 actually quite well again;

[~!Ji?e

hes allow]ed a (little). one cup,

[mit

seim :l.ahn,]

[with his (0.2)

12 13

Hans:

kaffee trin [kt e]r au scho wida. coffee drin[ks h]e also already again.

coffee he's 14 15

tooth,]

[also ] already drinking again.

[() ]

Oma: (0.3)

Similar to Siebenbii.rger Sachsisch, the placement of eytz and eytza(t) within TCUs in Bavarian German follows describable rules: Eytz is used in TCU-initial and medial positions, while eytza(t) is used overwhelmingly in TCU-final and also tum-final position (see lines 4 and 9 in Example 21). Eytza(t) also performs interactional tasks similar to ientzter: It occurs in sequence-initiating turns, and specifically at topical boundaries at which speakers pursue a new topic or the departure from the current topic. In the above example (21 kaffee), Hans proffers a new topic in lines 1···2 by

446 Emma M. Hetz

proposing talk about the health of a common acquaintance. After a lack of uptake in line 3, he continues his turn grammatically ('cause, line 4), abandons this continuation, and then makes a second try at topic proffering. This second attempt, uttered in overlap with Oma's delayed response to lines 1.-2, takes the shape of a polarized yes/ no--question (note the falling final intonation that adds epistemic polarity and gives the question the character of an assertion; cf. Koshik 2005), with ey_tza (line 4) as the unit-final element. This temporal adverb contrasts with the earlier post-positioned temporal reterence vor ner woche (line 2), and, similar to turn--final ientzter, it seems to increase the sequential force of the FPP. Speaker change occurs after eytza, and Oma embraces the new topic (lines 5-l 0). 1he function of the two-syllable variant of the adve.rb 'now' as a turn-final lexical element that promotes the uptake of a new topic or topic departure holds across the two varieties investigated. However, my data suggest that eytza.(t) in Bavarian can serve additional functions not observable in Siebenbiirger Sachsisch. In my data, three examples of eytzat seem distinctly different with respect to phonetic shape and sequential placement from all other instances of eytza(t). 'Ihey appear between dif.· ferent, topically related sequences, but they constitute the only turn element or are at most preceded by a conjunction ('buf or 'and'). The token itself exhibits level intonation and lengthening on the second syllable, and in all three examples, silence follows (and may also precede) the eytza- turn. After this silence, the same speaker continues. Consider the tollowing excerpt: (22)

besuch [2007--10--20... H0.5 ___04:17], tabout Hans' sister's upcoming visit to Oma 1

Hans: wefs zu mir gsagt hat se: hoat vor dass

because she said to me she: is planning to 2

bloB ze dir kummt, wefs sonst immer so lang

only vi~"it"you, because she usually stays 3

bei die gro<ern is, dass du dich

with the grandparents so long, that you 4

scho:

(0.2) .hh dass scho angst hat dass du:

alre1idy: (0.2j .hh th1it she5 1ifraid you 5

dich nich iih:: ausreichen:d besucht fi.\hlst. feel like you're not uh:: (being) visited enough.

6

haha [hahaha] [ha iih ]==(tz) .HH= haha [h,lhah!l] [ha uh .l=(tzj .HH=

7

8

Oma:

[aso. [J see.

.l lhm m:,] ] [hm m:,]

Hans: =drum hat's goar net Y.Qrghabt.=ne, .hh

=that'S why she didn't even plan to do it.=you know, .hh

Adverbs in German conversation 447 9

Oma:

m[hm. m[hm.

10 Hans:

(0.5)

11

=>

[ [

Hans:

<e:ytza:> <no::w>

13

(0.7)

14

Hans:

.h ja no isst(s) dann NAmittoch da:tm,

12

.h yeah so she71 eat then in the afterNOON then, (.) bei die groBeltern. (.) with the graruiparents. Due to the small number of examples, the function of this 3rd position eytza. cannot be described here. Additional data may reveal a specific function within the achievement of sequence and topic closing (e.g. Button 1987, 1990; Holt & Drew 2005; Schegloff 2007) or in the management of silence in turn-taking at points where a topic has come to completion and no new topic has been proposed (cf. Golato 2006; Sacks et al. 1974). 5-3

Defining and classifying position-sensitive lexical tokens

I conclude by contextualizing the phenomenon of interactionally position··sensi·· tive adverbs within its broader linguistic context. In this context, the classification and definition oflexical items such as ientz(t)!ientzter and eytz!eytza(t) as adverbials merits further discussion. Both ientz(t) and ientzter carry the semantic value of 'now' and they overwhelmingly appear in syntactic positions that have been described as typical tor adverbials. 'Ihe analysis of ientz(t) and ientzter, however, has shown that the meaning of the two words, especially of ientzter (and in particular ientzter in the syntactically marked post-position), cannot be described without reference to its systematic use in interaction. Hence, the description and classification of such a lexical item needs to include discourse-pragmatic features. 'Ihe additional observation that a number of adverbials in German- includingjetz(t)/'now' in Colloquial Standard German- may appear as sentence particles (Helbig 1988; Weydt & Hentschell983; Auer & Giinthner 2005) raises the question whether ientzter should be analyzed as a particle rather than as an adverbial, at least in certain functions. 21

21. See, e.g. M51lering (200 1) and Diewald and Fischer (1998), who show that particles in German are definable via tb.eir interactional functions. See also Sorjonen (200 1), who call'> for an understanding of the semantics of particles and response tokens as inherently interactional.

448 Emma M. Betz In my Siebenbiirger Sachsisch data, however, all ientzter and most ientz(t) clearly retain the semantic quality of the adverb, that is, they explicitly refer to matters of time, for example by establishing a contrast between 'then'/ 'a long time ago' and 'now'I 'just recently: Moreover, Siebenbii.rger Sachsisch uses a separate lexical item, nau, which is etymologically connected to other Germanic words for 'now; as a dis-course particle. The present study thus classified ientz(t) and ientzter as adverbs of time, with the understanding that this classification cannot be based on semantics and syntax alone. A description of their meaning needs to include the discourse-pragmatic level. This view reflects the complex connection between form and inter-actional placement that emerged from investigating the use of temporal adverbs in German in every-day interaction.

Appendix A Transcription conventions used Based on the Jeffersonian transcription system (Jefferson 1983, 1.985; Sacks 1.974: Appendix; Atkinson 1.984: ix----:xvi), with some additional conventions taken from GAT and GAT 2 (Selting et al. 1998, 2009).

(0.5) (.)

.hh hh haha (h) (hh) word

WORD owordo

i J.. iwordi (word)

start of overlap (simultaneous talk by two/more speakers) end of overlap latching between turns (i.e. two utterance by different speakers follow each other without an intervening beat of silence) or continuation of a speaker's turn across trmscript lines silence; length of silence timed relative to the delivery of the surrounding talk micro pause (less than 1/10 of a second). audible inbreath audible outbreath laughter; different vowels (i.e. e, i, a) indicate the quality oflaugh tokens laughter within a word emphasis, usually higher pitch higher volume relative to the surrounding talk enclosed passage quieter than the surrounding talk rising pitch on following syllable falling pitch on following syllable enclosed passage markedly higher in pitch than surrounding talk transcriber's uncertain hearing

Adverbs in German conversation 449 ( ) (( ))

>word< <word>

? word!

unintelligible stretch of talk. transcriber's additional comments/ transcription of events increase in tempo relative to surrounding talk slowing down in tempo relative to surrounding talk lengthening of the sound before the colon abrupt ending or cut-off (glottal closure) TCU-final falling intonation TGLJ--final intonation, falling to mid TCU-final continuing, slightly rising intonation TCU-final rising ('question') intonation accentuated delivery marks location of or the beginning and end of embodied actions (described above the frrst line of speech)

AppendixB Category abbreviations and other abbreviations used DIM

FEM FPP MASC NP 0 p

PL PPT PRT

s SG SPP

v VP

diminutive form feminine inflection first pair part of an adjacency pair (e.g. the question in a questionanswer sequence) masculine inflection noun phrase object person plural form past participle form (modaV response) particle subject singular form second pair part of an adjacency pair (e.g. the answer in a questionanswer sequence) verb verb phrase

References Aim, Maria. 2007. A.lso darUber liisst sich }a streiten! Die Analyse von 'also' in der Diskussion zu DL
Wlksell International

450 Emma M. Betz Atkinson, J. Maxwell & Heritage, John (eds). 1984. Structures ofSocial Action. Studies in Conversation Ana~vsis. Cambridge: CUP. Auer, Peter. 1996. On the prosody and syntax of turn-continuation~. In Prosody in Conversation, Eli7..abeth Couper-Kuhlen & Margret Selting (eds ), 57-1 00. Cambridge: CUP. Auer, Peter. 2005. Delayed self-repal.rs as a structuring device for complex turns in conversation. In Syntax and Lexis in Conversation. Studies on the Use of Linguistic Re.~ources in Jlllk-inlnteraction [Studies in Discourse and Grammar 17], Auli Hakulinen & Margret Seltlng (eds), 75--102. Amsterdam: John Benjamins. Auer, Peter. 2006. Increments and more. Anmerkungen zw· augenblicklichen Diskussion i.iber die Erweiterbarkeit von 1hrnkonstruktionseinheiten. In Grammatik und Interaktion. ArnulfDeppermann, Reinhard Fiehler & Thomas Spranz-Fogasy (eds). 279---294. RadolfZell: Verlag fur Gesprachsforschung. Auer, Peter & Gtinthner, Susanne. 2005. Die Entstehungvon Disktusmarkern im Deutschen ... ein Fall von Grammatikalisierung!' In Grammatikalisierung im Deutschen. Torsten Leuschner. Tanja Mortelmans & Sarah de Groodt (eds). 335-362. Berlin: Mouton de Gruyter. Auer, Peter & Ronfeld, Barbara. 2004. Prolixity as adaptation: Prosody and turn-taking in German conversation with a fluent aphasic. In Sound Patterns in Interaction [l}'Pological Studies in Language 62], Elizabeth Couper-Kuhlen & Cecilia Ford (eds), 171-200. Amsterdam: John Benjamins. Barske, Tobias. 2009. Same token, different actions: A conversation analytic study of social roles. embodied actions. and ok in German business meetings. Journal for Business Communication 46(1): 120--1-19. Barz, Irmhild & 6hlschlager, GUnther. 1998. Zwischen Grammatik und Lexikon. li.ibingen: Niemeyer. Betz, Emma. 2008. Grammar and Interaction: Pivots in German Conversation [Studies in Discourse and Grammar 21]. Amsterdam: John Benjamins. Betz, Emma & Golato, Andrea. 2008. Remembering relevant information and withholding relevant next action~: The German token ach}a ..Research on Language and Social interaction 41(1): 58-98.

Boden, Deirdre & Zimmerman, Don H. (ed~). 1991. Talk and Social Structure: Studies in Rthnomethodology and Conversation Analysis. Cambridge: Polity Press. Braun-Santa, Roswitha, Haldenwang, Sigrid, Richter, Gisela & Thudt, Anncliese (eds).1972. Siebenburgisch-Siichsisches Worterbuch, Band 4: H-J. Berlin: Mouton de Gruyter. Button, Graham. 1987. Moving out of closings. In Talk and Social Organization, Graham Burton & John R. E. Lee (eds.), 101-151. Clevedon: Multilingual Matters. Button, Graham. 1990. On \'llfieties of closings. In Interaction Competence, George Psathas (ed.), 93-148. Lanham: University Press of America Button, Graham & Casey, Neil. 1984. Generating topic: The use of topic-initial elicitors. In Structures of Social Action, J. Maxwell Atkinson & John Heritage (eds), 167-190. Cambridge: CUP. Button, Graham & Casey, NeiL 1985. Topic nomination and topic pursuit Human Studies 9: 3-55. Hutton, Graham & Casey, NeiL 1989. 1bpic initiation: Business-at-hand. RP.search on Language and Social Interaction 22: 61-92. Comrie, Bernard, Haspclmath, Martin & Bickel, Balthasar. 2008. Leipzig Glossing Rules: Conventl.ons for interinear morpheme-by-morpheme glosses (revised February 2008). ( 30 May 2008). Couper-Kuhlen, Elizabeth & Ford, Cecilia (eds). 2004. Sound Patterns in Interaction. Cross-linguistic Studies from Conversation [Typological Studies in Language 62]. Amsterdam: John Benjamins.

Adverbs in German conversation 451 Couper-Kuhlen, Eli:l,abeth & Ono, Tsuyoshi. 2007. Incrementing in conversatio.u A comparison of methods in English, German and Japanese. Pragmatic.~ 17(4): 513-552. Couper-Kilhlen, Elizabeth & Selting; Margret (eds). 1996. Prosody in Conversation. Cambridge: CUP. Deppermann, Arnulf 2006. Construction Grammar- eine Grammatik fur die Interaktion? In Grammatik 1md lnteraktion, Arnulf Deppermann, Reinhard Fi.ehler & 'Ihomas SpranzFogasy (eds), 43-66. Radolfzell: Verlag flir Gespriichsforschung. Diewald, Gabriele & Fischer, Kerstin. 1998. Zur diskursiven und modalen Funktion der Partlkeln aber, doch und ja in Instmktionsdialogen. Linguistica 38: 75--99. Duranti, Alessandro. 1997. Linguistic Anthropology. Cambridge: CUP. Edenstrom, Juliane. 2008, Nov. The post-positioning of temporal adverbs in German conversa-tions. Paper presented at the Annual Convention of the National Communication Associa-tion, San Diego CA. Edenstrom, Juliane. 2009. Word Order in Spoken German: Syntactic Right--expansions as an Interactionally Constructed PhenomenoiL Ph.D. dissertation, University of Illinois at Urbana--Champaign. Egbe11, Maria. 2002. Syntaktische Merkmale von tibrigens in seiner Hauptposltion: Im Mittelf. eld des Verb--Zweit--Satzes. Zeitschrift fur Germanistische Linguistik 30: 1--22. Egbe11, Maria. 2003. Die interaktionelle Relevanz einer gemeinsamen Vorgeschichte: Zur Bedeutung und Funktion von ubrigens in deutschen AlltagsgespriicheiL Zeitschrift ftir Sprachwissenschaft 22(2): 189--212. Eisenberg, Peter. 1989. Grundriss der deutschen Grammatlk. Stuttgart: Metzler. Ford, Cecilia. 1993. Grammar in Interaction: Adverbial Clauses in American English Conversa-tions. Cambridge: CUP. Ford, Cecilia & Fox, Barbara. 1996. Interactional motivations for reference formulation: He had. This guy had, a beautiful, thirty--two O:lds. In Studies in Anaphora [Typological Studies in Language 33]. Barbara Fox (eel), 145---168. Amsterdam: John Benjamins. Ford, Cecilia. 2001. At the intersection of turn and sequence: Negation and what com.es ne)..i. In Studies in Interactional Ling-..listics [Studies in Discourse and Grammar 10], Margret Selting & Elizabeth Couper-Kuhlen (eds), 51-80. Amsterdam: John Benjamins. Ford, Cecilia, Fox, Ba.rbara & Thompson, Sandra. 1996. Practices in the construction ofturns: 'Ihe 'TCU' revisited. Pragmatics 6: 427-454. Ford, Cecilia, Fox, Ba.rbara & Thompson, Sandra. 2002. The Language of Turn and Sequence. Oxford: OUP. Frey, Werner & Pittner, Karin (1998). Zur Posltion\erung der Adverbiale im deutschen Mlttelfelcl Unguistische Berichte 176: 489-535. Gardner, Rod. 1995. On Some Uses of the Conversational Token Mm. Ph.D. dissertation, University of Melbourne. Gilles, Peter. 200 L Di.e Intonation der Weiterweisung: Ein Beitrag zur konversationsanalytisch orientierten Erforschung von Regi.onaHntonation am Beispiel des Hamburgischen und Berlinischen. Zeit.~hrift ftir Germanistische Linguistik 29: 40-69. Golato, Andrea. 2006. Nov. action and topic shifts in conversation: The case of 3rd position naja in German. Paper presented at the A1mual Convention of the National Communication Association, San Antonio, TX. Golato, Andrea & Betz, Emma 2008. German ach and ac/-zso in repair uptake: Resources to sustain or remove epi.stemic asymmetry. Zeit.~hrift fiir Sprad1wissenschaft 27: 7-37. Grimm, Jacob & Grimm, Wilhelm. 1854-1960(1971). Deutsches Worterbuch, Band .10. Leipzig: Hirzel. Vol 10, entry jetzt, adv. ui> (1 L Nov. 2010).

452 Emma M. Betz

Harren, Inga. 2001. Ne? in Alltagsgespriichen- Interaktive Funktionen und Posltionierung in Turn und Sequen7. MA thesis, Carl von Ossietzky Universltat Oldenburg. Helbig, Gerhard. 1988. Le.'tikon deutscher Partikeitl. Leipzig! VEil Verlag Enzyklopadie. Heritage, John 1984. A change-of-state token and aspects of Its sequential placement. In Structures of Social A:ticm. Studies in Conversation AnalysL~, J. Maxwell Atkinson & John Heritage (eds), 299-345. Cambridge: CUP. Heritage, John & Raymond, Geoffrey. 2005. The terms of agreement: Indexlng epistemic authority and subordination in talk-in--interaction Social Psychology Quarter~v 68(1): 15--38. Holt, Elizabeth & Drew, Pau12005. Figurative pivots: The use of figurative expressions in pivotal topic transitions. Research on Language and Social Interaction 38: 35---61. Hopper, Patti J. & Traugott Elizabeth C. 1993. Grammaticalization. Cambridge: CUP. Hutchby, Ian & Wooffitt, Robin 1998. Conversation Anarysis. Cambridge: Polity Press. Jefferson, Gall. 1983. Two explorations of the organization of overlapping talk in conversation Til burg Papers in Language and Literature 28. Jefferson, Gall. 1984. On stepwise transition from talk about a trouble to inappropriately next-positioned matters. In Structures of Social Action. Studies in Conversation Analysis, J. Maxwell Atkinson & John Heritage (eds ), 191-222. Cambridge: CUP. Jefferson, Gall. 1985. An exercise in the transcription and analysis oflaughte1: In Handbook of Discourse Anarysis, Vol. 3, Teun A. van Dijk. (ed.), 25-34. London: Academic Press. Jefferson, Gail. 1993. Caveat speaker: Preliminary notes on recipient topic-shift implicature. Research on Language and Social Interaction 26(1): 1-30. Kamio, Akio. 1979. On the notion of 'speake1~s territory of information': A functional analysis of certain sentence- final forms in Japanese. In Explorations in linguistics: Papers in Honor of Kazuko Inoue, George D. Bedell, Eiichi Kobayashi & Masatake Muraki (eds), 213--233. Tokyo: Kenkyuusya. Kern, Friederike. 2007. Prosody as a resource in children's game explanations: Some aspects of turn construction and recipi.ency. Journal of Pragmatics 39: lll-133. Kern, Friederike & Selting, Margret. 2006. Einheltenkonstruktlon im 'ltirkendeutschen: Grarnmatlsche und prosodische Aspekte. Zeit.~chrift fUr sprachwissenschaft 25: 2.39-272. Koshik, Irene. 2005. Beyond Rhetoricai Questions: Assertive Questions in Everyday Interaction [Studies in Discourse and Grammar 16]. Amsterdam: John Benjamin.~. Labov, William & Fanshel, David. 1977. 1herapeut!c Discourse: Psychotherapy as Conversation. New York NY: Academic Press. Lerner, Gene (ed.). 2004. Conversation Anaiysis: Papers from the First Generation [Pragmatics and Beyond New Series 125]. Amsterdam: John Benjarnins. Local, John & Walker, Gareth. 2004. Abrupt-joins as a resource for the production of multi-unit, multi-action turns. Journal of Pragmatics 36(8): 1375-1403. Local, John, Kelly, John & Wells, William. 1986. Towards a phonology of conversation: Turntaking in Tyneside English. journal ofLinguistics 22: 411-437. Mollering, Martina. 2001. Teaching German modal particles: A corpus-based approach. Language Learning & Technology 5(3): 130-151. Ney, Karin 1984. Rumiinische lransferenzen in vier siebenburgisch-siichsischen Ort:smundarten dP.s Kreises lrermannstadt, Rumanien. Marburg: Elwert. Noren, Nlklas. 2007. Apokoinou in Swedish Talk-in-lnteraction: A Family of Methods for Grammatical Construction and the RRsoivingofLocal Communir.ative Projects [Linkoping Studies in Arts and Science 405]. Linkoplng: Llnk<:iping Unive.rsity. Pittner, Karin 1999. Adverbiale im Deutschen: Untersuchungen zu iilrer Stellung und Interpretation. Tlibingen: Stauffenburg.

Adverbs in German conversation -53 Prevlgnano, Carlo & Thibault, Paul J. 2003. Discussing Conversation Ana~vsis: The Work of Emanuel A. Sd:egioff. Amsterdam: John Benjamins. Sacks, Harvey. 1974. An analysis of the course of a joke's telling in conversation. In Explorations it~ the Ethnography of Speaking, Richard Bauman & Joel Shen.er (eds), 337-353. Cambridge: CUP. Sacks, Harvey. 1987. On the preference for agreement and contiguity ln sequences in conversation. In Talk a,1d Social Orgtmizatlon, Graham Button & John R.E. Lee (eds), 54-69. Clevedon: Multilingual Matters. Sacks, Harvey, Schegloff, Emanuel A. & Jetierson, Gail 1974. A simplest systematics for the organization oftmn-taking tor conversation. Language 50( 4): 696--735. Scheglofl~ Emanuel A 1987. Recycled tmn beginnings: A precise repair mechanism in conver-sation's turn- taking organization. In Talk and Social Organization, Graham Button & John RE. Lee (eds), 70---85. Clevedon: Multilingual Matters. Schegloff, Emanuel A. 1996. 1\.trn organization: One intersection of grammar and interaction. In Interaction and Grammar, Elinor Ochs, Emanuel A Schegloff & Sandra Thompson (eds), 52-133. Cambridge: CUP. Schegloff, Emanuel A. 1998. Reflections on studying prosody in talk-in-interaction. Language and Speech 41: 235-263. Schegloff, Emanuel A. 2002. Accounts of conduct in interaction: Interruption, overlap, and turn-taking. In Handbook of Sociological Theory, Jonathan H. Turner (ed.), 287-321. New York NY: Kluwer Academic/PlenUIII.. Schegloff, Emanuel A. 2007. Sequence Organization in Interaction. A Primer In Conversatlo,z Analysis. Cambridge: CUP. Schegloff, Emanuel A., Ochs, Elinor & Thompson, Sandra 1996. Introduction. In Interac-· Non and Grammar, Elinor Ochs, Emanuel A Schegloff & Sandra Thompson (eds), 1--51. Cambridge: CUP. Schegloff, Emanuel A. & Sacks, Harvey. 1973. Opening up closings. Semiotica 8: 289-327. Scheut1~ Hannes. 2005. Pivot constructions in spoken German. In Syntax and Lexis in Conversation [Studies in Discours and Grammar 17], Alili Haktillnen & Margret Seltlng (eds), I 03-128. Amsterdam: John Benjamins. Schullerus, Adolf 1925. Siebenburgisch-Siidtsische.s Wiirterbuch, Band 2: D-E Berlin: Mouton deGruyter. Schwitalla, Johannes. 2003. Gesprochenes Deutsch. Berlin: Erich Schmidt. Selting, Margret 1995. Prosodie im Gesprach. Aspekte einer interaktionellen Analyse am Beispiel der Burger- Verwaltungs-Kommunlkation. Thbingen: Niemeyer. Selting, Margret. 1996. On the interplay of syntax and prosody in the constitution of turnconstructional units and turns in conversation. Pragmatics 6(3): 357-388. Selting, Margret. 1998. Fragments of TCUs as deviant cases of unit production in conversational talk. In Str.sdies in Interactional Linguistics [Studies in Discourse and Grammar 10], Margret Selting & Elizabeth Couper-Kuhlen (eds), 229-258. Amsterdam: John Benjamins. Selting, Margret. 2000. 'Ihe construction of units ln conversational talk. Language in Society 29: 477-517. Selting, Margret. 200 l. Fragmenls of units as deviant ca.~es of unlt -production in conversational talk. In Studies in interactional Linguistics [Studies in Discourse and Grammar 10], Margret Sclting & Elizabeth Couper-Ktihlen (eds), 229-258. Amste.rdam: John Benjamin.~. Selting, Margret. 2005. Syntax and prosody as methods for the construction and identification of turn-constructional units in conversation In Syntax and Lexis in Conversation [Studies in

454 Emma M. Betz

Discourse and Grammar 17], Auli Hakulinen & Margret Selting (eds), 17-44. Amsterdam: John Benjamin~. Selting, Margret & Couper-Kuhlen, Elizabeth (eds). 2001. Studies in Interactional Linguistics [Studies in Discourse and Grammar 10]. Amsterdam: John Benjamins. Selting, Margret, Auer, Peter, Barden, Birgit. Bergmann, Jorg, Couper-Kuhlen, Elizabeth, Gunther, Susanne, Quasthoff, Uta, Meier, Christoph, Schlobinski, Peter & Uhmann, Susanne. 1998. Gesprach.sanalytisches Transkriptionssystem (GAT). Lingui.~tische Berichte 173: 91-122. Selting, Margret, Alter, Peter, Barth--Weingarten, Dagmar, Bergmann, Jorg, Bergmann, Pia, Birkne1~ Karin, Coupel'·Kuhlen. Elizabeth, Deppermann. Arnulf~ Gilles, Peter, Glinth·· ner, Susanne, Hartung, Martin, Kern, Friederlke, Mertzluift, Christine, Meyer, Chris·· tian, Morek. Miriam, Oberzaucher, Frank, Peters, Jorg, Quasthoff, Uta, Schutte, Wtlfrted. Stukenbrock. Anja & lihmann. Susanne. 2009. Gespriichsanalytisches Transkriptions-· system 2 (GAT 2). Gesprac.hsforschung Online-ZeitsdJrift zur verbalen Interaktlon 10: 353--402. <www.gespraechsforschung--ozs.de>. Sorjonen, Marja-Leena. 2001. Respotlditlgin Conversation: A Study ofResponse Particles In Fimlish [Pragmatics and Beyond New Series 70]. Amsterdam: John Benjamins. Streeck. Jurgen. 2002. Grammars, words, and embodied meanings: On the uses and evolution of so and llke. Journal of Communlcatiotl 52(3): 581-596. Taleghani-Nlkazm, Carmen. 2008, Nov. How and when are new conversational matters brought into the conversation? Paper presented at the Annual Convention of the National Communication Association, San Diego, CA. Tanaka, Hiroko. 2001. Adverbials for turn projection in Japanese: Toward a demystification of the "telepathic" mode of communication. Language in Society 30: 559--87. Traugott, Elizabeth. 1997, July. The discow'Se connective 'after all': A historical pragmatic account Paper presented at the lOth International Congress of Linguists, Paris, France. (16 July 2009). Vorrelte~. Susanne. 2003. Turn continuations: Towards a cross-linguistic dassl.fi.cation. Interaction and Linguistic Structures [InLiSt] .39. (7 December 2007). Walker, Gareth. 2004. On some interactional and phonetic properties of increme.nts to turns in talk-in-interaction. In Sound Patterns in Interaction rrypological Studies in Language 62], FJizabeth Couper- Kuhlen & Cecilia Ford (eds ), 147-169. Amsterdam: John Benjamins. Walker, Gareth. 2007. 0 n the design and use of pivots in everyday English conversation. journal ofPragmatics 39(12): 2217-2243. Weiss, Andreas. 1975. Syntax gesprochener GespriidJe. EinjlUsse von Situation und 1hema aufdas Sprachverhalten. Dusseldorf: Padagogischer Verlag Schwann. Weydt, Harald & Hentschel, Elke. 1983. Kleines Abtonungsworterbuch. In Partikeln und Interaktion, Harald Weydt (ed.), 3-24. Tiibingen: Niemeyer. Wischer, Ilse & Diewald, Gabriele. 2002. New Reflections on Grammatlcalization [Typological Studies in Language 17]. Amsterdam: John Benjamins. Zifonun, Gisela, Hoffmann, Ludger, Strecker, Bruno & Ballweg, Joachim. 1997. Grammatik der deutschen Sprache, 3 Vols. Berli.n: Mouton de Gruyter. Zimmerman, Don H. 1993. Acknowledgment tokens and speakership incipiency revisited. RRsearch on Language and Social interaction 26(2): 179-194.

Texas German discourse pragmatics A preliminary study of the English-origin discourse markers of course, see, and now Hunter \.Yeilbacher University of Texas at Austin

This article seeks to clarify the role that English-origin pragmatic discourse markers play in the speech of Texas German (TxG). The data in this study reveal that these elements function to lighten the cognitive load of the speaker by pragmatically indicating that the speaker is processing the upcoming utterance. This observation may be taken to indicate that for many TxG speakers English is (or has become) the pragmatically dominant language, however it does not rule out that these discourse markers are lexical items found in a unified mixed-code system.

1.

Introduction

Although the many Texas German communities represent various Spmchinseln, they have by no means been isolated: 1exas German (hereafter TxG) developed in contact with :English, its remaining speakers are bilingual, and so the existence of English words and phrases in TxG is hardly surprising. A striking characteristic ofTxG speech is the prevalent use of English-origin pragmatic elements such as discourse markers (hereafter DMs) including well or you know, as well as other discourse--pragmatic phrases or 'particles' from English (Salmons 1990, Goss & Salmons 2000, Boas & Weilbacher 2007). Data from sociolinguistic interviews collected by the Texas German Dialect Project (TGDP) (Boas 2006) reveal a multitude of DMs and similar discourse regulating devices from English that are regularly used by TxG speakers. Bilingual DMs have been the focus of many studies because their use fulfills a conversational purpose and a cultural convention simultaneously (see Brody 1987; Salmons 1990; Serra 1998; Fuller 2001; Hlavac 2006, among others). The analysis of English-origin pragmatic elements in TxG raises a number of important questions for research into borrowing, code-switching, and discourse particles because certain usages do not

456 Hunter Weilbacher

easily fit into existing theoretical paradigms. Consider (1), which contains two Englishorigin pragmatic elements: (l)

So meine Kusine von Dallas und wir gehen zu Industry, die erste so my cousin from Dallas and we go to Industry, the first

deutsche Stadt, un sure enough, das war unser Ur-urgrossvater. German dty and sure enough. that was our great -great -grandfather. 'So my cousin from Dallas [and I] went to Industry, the first German town, and sure enough, that was our great -great -grandfather: (1-60-l-4 -a) i

In (1), the speaker has been talking about researching her genealogy and found an important piece of infonnation warranting further investigation. She uses so, which in this case is most likely English so, 2 to move the topic onto what she and her c.ousin did as a result of receiving this new information (going to a particular town). She then uses sure enough to express that they were able to verify that information, namely that their ancestor had been a part of that community. This example points to the central question of this study: are these cases of emblematic code--switching, which has often been viewed as a linguistic behavior that indexes a certain social identity (Auer 1998, 1999, 2005; Muysken 2007), are these borrmvings because of their frequency and degree of integration (Myers-Scotton 2002) or convergence (Salmons 1990), or is this evidence tor a mixed--code system (Maschler 1998; Serra 1998)? Perhaps what is most relevant about these data are that they show just how complicated the situation is. 'Ihis study first discusses code-s\vitching versus borrowing and then examines the use of English see, now, and ofcourse in TxG to see what their use can tell us about bilingual discourse prc:tgmatics. This study is structured in the following way: in the second section, I offer a brief overview of Texas German and the TGDP. I then turn, in Section 3, to a discussion of previous work on language contact situations, focusing primarily on studies that deal with the integration of pragmatic elements from one language into another.

TxG data are cited in the following format: the first number is that of the interviewer (1 is Hans C. Boas); the second number is the speaker; the third is the number of the interview \\oith that speaker; the last number is the segment of the interview; 'a' indicates an audio file, V indicates a video file. Thus ( 1-1-1-14-a) is Hans Boas interviewing speaker# 1 for the first time, and this is the 14th segment of that interview. See Boas (2006) for detaiL<~. 1.

1. There is a great deal of overlap between the pragmatic properties of so in English and German. 'Ihis instance is most likely English so for two reasons: (1) the pragmatic purpose it serves in directing the succession of events in her statement is more a property of English so (compare German tmd so), and (2) the speaker pronounces so in this utterance like English [so] rather than German [zo].

Texas German di~course pragmatics 457 In Section 4, I present and discuss examples from the TGDP corpus of English pragmatic markers used in TxG speech. Section 5 summarizes the findings of this study and presents an outlook for future research.

2.

Texas German

Beginning primarily in the 1830s, emigrants left German--speaking Europe tor 'Iexas. Wide variation in these settlers' homeland dialects contributed to the development of Texas German as a new and distinct (although by no means homogenous) new world dialect. Regional varieties ofTxG continued to evolve tluoughout the 19th and early 20th centuries, reflecting the contact between different European dialectal varieties in communities settled primarily by immignmts from different parts of Germany (Boas 2009).3 English-only laws and strong anti-German sentiment following World War I effectively stunted the development of these regional TxGs, and prompted the switch from German to English as an institutional language (Salmons 1983; Guion 1996). These laws included the closure of German schools, which were only partially revived when the laws were repealed after the war. In the 1940s, reactions to World War II caused another substantial loss of prestige for the German language, and transmission of Tx:G to younger generations slowed dramatically after 1945 (Gilbert 1972; Salmons 198 3; Boas 2009 ). 1he changing demographics of Texas, due in large part to an influx of English speakers from other parts of the United States into previously 'German' areas of Te.xas, primarily because of the US military bases around San Antonio, resulted in a gradual shift from TxG to English in many social domains such as the market or church. Many Texas Germans aL~o married people from outside the TxG communi-ties or moved away, and even for many farnilies in which both parents were speakers of TxG, the aversion to being identified as a 'German' was so great because of World War II that they began to speak only English to their children. Most of the TGDP's informants were born between the two World Wars, grew up in a diglossic situation in which English was tl1e prestige language and German was stigmatized, and have therefore predominately spoken English for most of their lives. With an estimated 6-7000 remaining speakers and virtually no transmission to younger generations, TxG is criti-cally endangered and effectively moribund (see Boas 2009).

3· Early immigrants to the area between the lower Brazos and Colorado rivers in the 1830s and early 1840s came principally from Low German dialectal regions, while many immigrants to the organiz.ed settlements of New Braunfels and Fredericksburg were from Middle German s1'eaking areas such as Hesse and Nassau Oordan 1966: 31-48).

458 Hunter Weilbacher

2.1

The Texas German Dialect Project and the data

Since 2002, researchers tor the Texas German Dialect Project have interviewed over 350 fluent and semi-fluent speakers ofTxG throughout central Texas. Audio and video files of the interviews are housed in the on--line Texas German Dialect Archive (TGDA) at http://www.tgdp.org. Interviews typically consist of two English-to--German elicita-tion tasks (about 350 words and phrases), open-ended interviews, and biographical questions. The data used tor this study are from open-ended interviews ~ith speakers of TxG conducted by the TGD R which have been transcribed and translated and then stored in the TGDA. 'Ihe interviews typically contain the speakers' answers toquestions prompted by the interviewer, often including stories about life experiences, local history, and changes in lifestyles and language use that they have personally experienced. 'Ihe interview segments examined for this study represent only interviews that have been transcribed and translated and are therefore searchable ·with an xml plaintextconcordancer interface. To date, 3142 open-ended interview file segments from 88 speakers have been transcribed and are searchable using the concordancer. Using the concordancer, I searched the corpus for the targeted words, and then downloaded the sound files belonging to the file segments. I then listened to the file segment and isolated the portion of the conversation in which the instance occurs and noted the conversational context While it is possible to count the number of tokens of see, now, and of course in the TGDA corpus, it immediately becomes apparent that classifying a given instance as a borrowing or a code-- switch is extremely difficult. For example, part of a statement or conversation may be considered an extended codeswitch if the target word or phrase occurs ·within large chunks of English. For the purposes of this study, I have excluded instances like Example (2), where you see is part of a larger switch into English: (2)

Me in GroB--- mein um urn my great my

was war mein Greatgrandfather, der kam what was my great -grandfather he came

von Deutschland und der war hier for Civil War and he v."!ls here for Civil War from Germany

in the Civil War. in the Civil War

'My great- my urn- what was my great-grandfather, he came from Germany and was here for the Civil War: Der war hier in the South . .. you see und der hat he was here in the South . . . you see and he had und die haben denn ihm appointed .. . and they have then him appointed .. . 'He was here in the South, you see, and he- they then appointed him .. .' (l-33--l-23--a)

Texas German di.~course pragmatics 459 lhis study focuses on instances of English-origin pragmatic markers that are seamlessly integrated into the German utterance, as in Example (3), in which the speaker employs the use of English see to focus explanatory information in her following statement. (3)

Wir sind nach New Braunfds gegangen. See wir sind- bei Converse we are to New Braunfels gone. see we are in Converse da warn nidrt viel. there were not many

'We always went to New BraWlfeL'l. See, we- in Converse there weren't many [dancehalls]:

(1-29-1-23-a)

lhese TxG data are not easily analyzed according to any one theoretical approach, because some authors' approaches would identify see in (3) as a code-switch rather than a borrowed DM. Before presenting a more in-depth analysis of the data, I give a brief overview of research on language contact that results in borrowing, code-· switching, or the creation of a mixed--code system.

3·

Code-switching to borrowing

In recent years, a consensus in the literature on language contact has developed in which borrowing and code-switching are seen as existing on a continuum. The extent of a foreign element's usage in a speech community, its frequency in use, and the level of phonological or morphosyntactic integration into the native system it exhibits indi-cate the degree to which that element has been borrowed. A code-switch requires the changing of languages, whereby a speaker exits the primary language structure of the utterance to employ linguistic elements that are anchored in the other system and then returns (or not) to the original linguistic system. That is, in a code-switch, the speaker relies on elements from the other linguistic system outside of what Myers-Scotton (1993, 2002) calls the 'matrix system: In this view, let's see in Example (4) would likely be seen as part of a code-switch: (4)

Well, mir hamm in well, we have in well well

in uh in uh

lets see let's see

in neunzehnhunde1tw1 .... in nineteen-hundred-and ....

ne1.m1.mdreiBig hat er die Farm verkauft. nine--and--thirty had he the farm sold 'Well, we were in let's see well in 1939 he sold the farm:

{l--1--1--12--a)

460

Hunter Weilbacher Example (4) raises a question that is central to this study. The speaker's hesitation before let's see may indicate that he has excited the 'matrix language' of the conversation while he is in the process of recalling specific information (the year the farm was sold). However, the presence of weil (especially the first instance) is not necessarily a code--switch for several reasons. Matras' (1998) indicates that DMs such as well are particularly borrowable because they may lack semantic or pragmatic equivalents in the other language in addition to having unique or useful pragmatic functions.4 As noted by Brinton (1996: 36), ""'ell is the most frequent of the discourse particles due to its wide range of discourse functions. Well can signal acknowledgement, hesita-tion, reservation, reluctance, or introduce an explanatory comment. among its other conversational uses (ct: Schourup 2001; Brinton 1996, among others). This multitude of discursive functions makes well an element that is very likely to be borrowed. Its frequency in TxG speech also indicates that it is treated like a native element. Clyne (2003: 73) notes that while '"integration' and 'frequency' are the main criteria employed to establish 'borrowing' 'code-s\vitching' is not a uniform category, and different scholars have emphasized different aspects of the phenomenon?' 'lhe presence of elements from another language in the speech of individuals or groups has largely been interpreted as code-switches or borrmvings depending on the sociolinguistic or pragmatic functions those clements serve. It is also possible for a single element to be both a borrowing and a code-switch synchronically, even in the idiomatic speech of an individual speaker, depending on the communicative purpose that the speaker wishes to achieve and his motivation for doing so. Brody (1987: 510) notes that "the boundary between code-switching and borrowing is indeterminate for any speech community, and may even be manipulated for effect within the speech of individuals~ In clarifying this difference, she points out that "the particles borrowed from Spanish have been sufficiently incorporated into the Mayan languages to qualify as borrowed, but the use of a particle in any particular case maybe considered a codeswitch" (1987: 510). Many authors, such as Auer (1998), have examined bilingual conversational structure in terms of its levels of social and cultural embeddedness, claiming that code-s\vitching "is related to and indicative of group membership in particular types of bilingual speech communities" (1998: 3). In a later article, Auer (2005: 404) dis-cusses code-switching as a symbolic gesture that indexes social identity. For Auer, code-s\vitches involve the juxtaposition of two codes that usually occur at major syntactic boundaries, e.g. at clause or sentence level (1999: 312). The difference between code-switching and borrowing with regard to DMs seems to be further obscured by Muysken (2007). He claims that "while borrowing

4·

Matras (1998) is discussed in more detail below in Section 3.1.

Texas German di.~course pragmatics 461 and insertion generally involve elements from a dominant language put into a community language, this does not hold for discourse marker switching" but then goes onto say that "the frequency of discourse marker switching probably is the reason for the fact that in so many lanbruages disc.ourse markers have been borrowed. Frequently S\\
Mixed c.ode systems

Maschler (1998: 125) defines 'mixed-codes' as those which use two languages "so that a third, new code emerges, in which elements from the two languages are incorporated into a structurally definable pattern." When new and distinct pmgmatic functions develop, the emergence of a new mixed-code system can be arbrued for (see Maschler 1998; Serra 1998; Fuller 2001). According to Serra (1998: 119) mixed-codes are to be accounted for on the basis of bilingual rather than monolingual parameters, so that "all the connectives in the new system f{)rm a specific linbruistic confibruration, using and exploiting the dynamic relationship between the two languages from which they are taken:' Consider the following Example (5) from TxG: (5)

Du muss beinuh Deutsch sprech because also das jetzt You must almost German speak because well that now heutzutag maclrts nichts mehr. makes-it nothing more. today 'You really had to speak German because - well now today it doesn't make a difference anymore:

(l-l-l-14-a)

Example (5) shows that the speaker incorporates the English conjunction because into the utterance without any apparent reliance on English as a system or any reason to do so. This same speaker (Speaker 1) in the same conversation alternates between the English conjunction and its German equivalent weil, using weil three times and because five times. Similarly, Speaker 62 appears to indiscriminately alternate between because (5 instances) and weil (4 instances) in the same conversation. The following two Examples (6) and (7) are from Speaker 62: (6)

Und ich frag sie alsmal was And I ask her sometimes what

was Werte in Spanisch sind, what words in Spanish are,

weil

ich her was und denn will ich wissen was es ist. because I hear something and then want I to--know what it is.

~nd I ask her sometimes what words in Spanish are because I hear something and then I want to know what it

is:

(l-62-l-22-a)

462 Hunter Weilbacher

(7)

Aber es ist because es ist nicht pasteurized. but it is because it is not pasteurized. 'But it is because it is not pasteurized:

(1-62--1-8--a)

An analysis of the searchable TxG data shows that because is far less frequent than the use of the German conjunction weii. Table 1 juxtaposes the usages of weil and because by speaker (in bold) and instance. Speakers who use both forms are italicized.

Table 1. The distribution of v,:eil and because in the TG DA corpu'l

because

weil

Instance

Speaker

Instance

Speaker

Instance

1

3

42

6

43

10 1

3 12

44 45 51 54 55

5 7 3 6

80

2 21 24 25 27 28 29 30 32 33

Speaker

34

35 36 37 38 39 40

6

17

4

57

19

58

7 17 6 2 4

2 14 1 3 4 10

59

47

61 62 63 64 71 76 79

18 4

11

27 17

7 27

56

4

1 5 1 2

82 83 84 85 86

90

92 94 96 97

115 118 129 134 135

6 3 1 6 1 1 6 1 1 4

Speaker

Instance

1

5

2 3

4 1 8 2

7 25

30 36

4

59 60

2 3 5

1

62

4 1"'

64

76

4

77

6 1 4

82 83 84

2 1, 1* 2 1

"These uses of bec:a·use seem to trigger an extended code-switch. supporting that they themselves are code-switches.

Table 1 demonstrates that only four speakers use because instead of wei/, but the infrequency of their use of the English conjunction lends little to the argument that for them, because is a borrowing rather than a code-switch. Of the other speakers who use both forms, however, only Speaker 59 ovenvhelmingly uses weil, and while Speakers 2, 25, 36, 83, and 84 seem to prefer weil to because, their use is not as disproportional. Speaker 30 uses both forms equally and Speaker 82 uses both forms once and uses because once in what can be considered a code-switch trigger - that is, an element that initiates an extended switch into English and so can be reasoned to be a code--switch itself Auer (1998: 17), tollowing Clyne (1967), states that 'nonce borrowings; infrequent

Texas German di.~course pragmatics 463 single lexical elements, may trigger the inclusion of other language materials around them. Clyne (2003: 169) proposes that "lexical transfers function as trigger-words all across language dyads and triads:' 1he reasons behind this "facilitation;' as Clyne calls it, range from phonetic environments to sociolinguistic motivations such as identification of topical information with the 'realm' of the other language. Speakers 1, 62, and 64 statistically prefer because to weil, but what can they indicate about the definition of borrowings versus code-switches? If the determining factor is whether the element under analysis in this case a marker of subordinate idea units--- is drawn from the other linguistic system, then it is said to be a code--switch. These uses seem to behave more like synonyms, which seems to indicate a degree of convergence of the two systems for some speakers. This notion may be supported by Example (8) from Speaker 1, who alternates between weil and because: (8)

Und auch die Eltern hamm ver And also the parents had for

denn at die Zeit angefimgen then at that time begun

mehr Englisd1 zu sprechen becah more English to speak becau

weil mir mussten das in because we must that in

die Schul Englisch the school English das Deutsd1e the German

so sind die mehr von so are they more from

weggekomm. come away--come

1\nd even the parents sta1ted to speak more English at that time because we had to speak English at school- so they moved away from the German more and more: (1-1-1-9-a) Example (8) comes after Speaker 1 has used weil once and then beca.use once. Here, he starts to use because but then corrects himself and uses weil, although he subsequently uses because three more times and well once more in the same conversation. The question here is, does Speaker 1 sometimes try to prevent his reliance on the other linguistic system and his usage of English because, or does the English conjunction occupy the same space in his mental German system as does weil, where Example (8) is only indicative of his awareness that because is in fact English and his German speech is being observed and recorded by a linguist? This analysis of because and weil in our TxG data show how complicated it is to dearly distinguish between code-switches, borrowings, and convergence. When an entire utterance is in the other language, it is clear that a code-switch has occurred. However, when single elements like DMs frequently occur in speech that is otherwise entirely in the matrix language of the conversation, it seems undesirable to view

464 Hunter Weilbacher

them as code-switches. For the purposes of this study, I will consider examples like (9) instances of code-switching: (9)

Ich denke, dass der Krieg ... World War Two, hat das ne M.ase World War Two has that a lot I think that the war... angebracht, weil wo ich zu Schule gegang bin, dann ward to school gone am then became on--brought because when I Deutsch verboten. Un uh ... it just forced a lot of people to talk German forbidden And uh ... it just forc.ed a lot of people to talk

English. Course w- English

mir ham es immer noch zu Haus gesprochen. Course (we) - we have it always still at home spoken

'I think that the war World War II, had a lot to do with it, because when I went to school, then it was forbidden to speak German. 'Course we we still spoke it at home: {l--55--1--9--a) The TxG data reveal countless code--switches like Example (9). 'Ibis study, however, will exclude code--switches and focus on instances of English--origin DMs that are integrated into the utterance and can therefore be considered to be part of the TxG DM system. 3.2

Motivations for borrowing DMs

Matras (1998) attributes borrowing of DMs to the fact that the donor language can be classified as the pragmatically dominant language in contact situations. As mentioned above, Matras (1998) offers a hypothesis for the fusion of different grammatical operations in language--contact situations that he calls the principle of pragmatic detachability: Grammatical elements that organize the speech event are perceived as gesture-like, situation--bound devices and are therefore detachable from the content message of the utterance. (Matras 1998: 309) Matras (1998) establishes several scales which rank the likelihood that DMs may be borrowed according to the communicative functions inherent to those particular DMs. The integrated pragmatic detachability hierarchy predicts that DMs higher on the scale (operational in nature) are more likely to be borrowed than those lower on the scale (more content--oriented). Thus, elements that are not easily analyzed in terms of lexical meaning (e.g. well) are more easily borrowed than highly lexical items like unde1'stand? ('do you understand?'). Two previous studies have applied Matras' (1998) pragmatic detachability hierarchy to borrowed DMs in German--American dialects. Fuller (2001) applies Matras' (1998) hierarchy to Pennsylvania German to predict when, based on their pragmatic functions, specific DMs will be replaced and which

Texas German di.~course pragmatics 465

might be preserved. 'Ihe distribution of English and German DMs leads Fuller (2001) to conclude that Pennsylvania German is in a process of'turnover' to an English-based system. Boas and Weilbacher (2009) compare Fuller's (2001) findings for Pennsylvania German to TxG data from the TGDA corpus, but find that Matras' (1998) hierMchy fails to make the correct predictiom about the distribution of bilingual DMs in TxG, as many native-Gennan DMs are attested with considerable frequency. Although Matras' (1998) hierarchy may not be universally applicable in terms of turnover from a native DM system to a borrowed system, the number of English--origin DMs in TxG does seem to support his point that 'gesture--like' elements like DMs are highly borrow-able. Matras (2000) offers an cognitive-based explanatory scenario for why the use of borrowed DMs may result in a change in the DM system of a minority language community: ... bilingual speakers, in an unc.onsc.ious effort to reduce the mental effort which is necessary to monitor and direct the hearer's responses and reactions to the speaker's utterance, -will simplify such monitoring- and -directing operations by eliminating the language-specific options available to them, thereby automatic.ising the choice of expressions. This c.ognit:ive motivation, I argue, is so strong that it will at times override the social and communicative constraints on the discourse, leading to counterstrategic, accidental, or tmintentional choices (i.e. slips). Such cases of situation-bound, local Fusion of the systems may lead to language change especially in minority language communities, whkh may establish a permanent licensing for speakers to prefer the use of one system for monitoring--and--directing operations over the choice among two. (Matras 2000: 511-512)

Contrary to studies by Salmons (1990) and subsequently by Goss and Salmons (2000), Boas and Weilbacher (2009) show that the English DM system has not replaced the German DM system in Tx:G. Consider the following Example (10): (Hl)

(laughs)

a.

Das haben sie immer verkehrt gekriegt. that had they always confused got 'They always got that confused:

b.

Nun, id1 wei6 nicht, gar nid1t wo meine Ureltern herkamen Well, I know not, at--all not where my great--parents here--came 'Well, I don't know where my great--grandparents came from at all: {l--43--1--4--a)

In Example (10), speaker (a) digresses into a story about how her last name sounds like the Gennan expression for 'I am called' which would confuse people when they asked her name. At this point the other speaker interrupts, answering the original ques·· lion which was about where her great--grandparents came from, and introduces her statement (regains the floor) with the Gennan DM nun. Boas and Weilbacher (2009)

466 Hunter Weilbacher

find that the German D.Ms ja, halt, eben, mai, and doch are still prevalent in TxG (see Boas & Weilbacher 2009 for a discussion of German DMs in TxG). The use of DMs from both languages is a good indication that the DM system in TxG might best be viewed as a mixed-code. I now turn to a discussion of see, of coufse, and now in our TxG data to see what their use indicates about their degree of integration and the DM system in TxG as a whole.

4·

See, of course, and now in Texas German

As bilinguals, many TxG speakers use the English DMs see, of course, and now in both German and English. In some cases, these DMs occur as part of extended code-switches or as integrated elements that escape easy classification. Take for Example (11), which shows the integration of several English elements: (11)

'Oh ihr misst noch lange lange leben because mit der kann oh you-PL must still long long live because with him can sin kein finden mehr was'- you know. Now ich sag dis, da no find anymore what you know Now I say this there are verschiedene Deutsche, was von Deutschland sin, die wohn hier. here different Germans what from Germany are they live Aber die wollen nkh mit so was gebut they want not with so what PAST·

you you

know geborde1t wern. know - bored become

"Oh you all have to live for a long time because you can't find anyone anymore who can'- you know [translate]. Now I say this, there are other Germans, who are from Germany, who live here. But they don't want to be bored with something like this: (1-59-1-21-a) Example (11) shows the use of English because, you know, now, and the inclusion of an English verb, to bofe, which has been incorporated into the German verb structure. The use of these English elements highlights the degree to which many TxG speakers mark their speech. In (11), Speaker 59 is quoting from a personal conversation she had which took place in English. The point of the statement is that someone was telling her that she, and the few other Texas Germans who can translate old family correspondences, need to stay alive tor a long time because the native European Germans who live in the area do not want to be bothered with such a task. Her use of because in this instance may be due to the fact that she is recalling something that was said to her in English. Her use of the English DMs you know and now fulfill the pragmatic objec-tives of appealing for understanding and focusing her next statement, respt~tively, and

Texas German discourse pragmatics 467

indicate that these elements are part of her TxG DM system- that is, they have been borrowed from English. Previous studies (Salmons 1990) and (Boas & Weilbacher 2007) have examined you know in TxG. Salmons (1990) also examines of course in German-American dialect~, but does not specifically comment on how it fwH.."tions in TxG. l now turn to of course, now, and see from conversational data collected by the TGDP to characterize their use and what this reveals about TxG speech.

4.1

Of course as a discourse marker in Texas German

Of course has been examined as a discourse particle in terms of the rhetorical moti·· vations for its use, e.g. as a concessive element (Quirk et al1985), or as a hedge that "operates on the subjective plane to modulate or attenuate the pragmatic force of an utterance" (Lewis 2003: 81). As mentioned above, Salmons ( 1990) examined of course as a DM in American German dialects, including TxG, but only proposes that it occurs frequently. that it seems to focus on upcoming text, and that it occurs utterance initially or after a conjunction in sentence-initial position. 20 of the 32 instances of of course in the TGDA corpus occur initially. For example: (12)

Die sind bei Indianola reingekommen they are by Indianola in-come

of course Indianola is ··· of course Indianola ist

ja gibts nicht mehr. yes gives not more 'They came in through Indianola - of course - well, Indianola doesn~t exi.<>t any more: (1-33-1-22-a)

However, the TGDA data include five instances that are not easily classified as being utterance-initial. Consider Examples (13) and (14): (13)

Ja, ja, die ist ja - 'course sie Ieben jetzt nich mehr. yes yes she is yes 'course they live now not more 'Yes, she is - of course, none of them are alive anymore: ( 1-1-1-6-a)

(14)

Well, welhe von die - welhe von die Amerikaner Kinder was nich well, some of the - some of the American children what not deutsch warn - die - course die konnten Englisch und die English and they German were - they - 'course they could konnten kein Deutsch sprechen. could no Gennan speak 'Well, some of the American kids who weren't Gennan of course, they could speak English but they couldn't speak German: ( 1-1-1-9-a)

468 Hunter Weilbacher

These examples do include a pause or hesitation, but are still incorporc:tted into the structure of the utterance. There are six other examples that are integrated into the statement: (15)

Und das - well - ~ir haben course andere Kieh auch gehabt. and that well we had 'course other cows also had (1-94--1--20--a) ~nd that well of course we also had other cows:

(16)

Un denn Abends of course die Kieh z'melken und And then evenings ofcourse the cows to-milk and die - alles mu'lS gefuttert werden. the - eveq1hing mu~ fed be ~nd then in the evenings of course the cows had to be milked and the (1--1--1--7--a) everything had to be fed:

There is one final example of of course used in a rather peculiar way at the end of a statement: (17)

Ich glaub in Corpus un Galveston da baut er I believe in Corpus and Galveston there builds he seine Sd1litterbahn of course. his Schlitterbahn of course 'I think in Corpu~ and in Galveston he's building his Schlitterbahn ofcourse: (1-59-1-19-a)

These speakers have integrated English of course to fulfill the same discourse pragmatic function in TxG. 4.2

Now as a discourse marker in Texas German

As noted by Schi1frin (1987: 230), no·wS as a DM "marks a speaker's progression through discourse time by displaying attention to an upcoming idea unit, orientation, and/or participation framework?' 'Ihe TGDP data contain two instances in which now heads an extended code--switch into English. There are 12 instances from nine different speakers in which now functions as a borrowed DM. For example: (18)

Das war eine ha:tte Zeit, hauptsii.chlich fii:r Letrte in die Stadt. Now, that was a hard time especially for people in the city Now welche wa:t· auf'n Farm, die haben ihr eigen Garten gehabt, die some were on-a farm they have their own garden had they

Now can also be a temporal adverb. There are several occurrences of English 11ow in the TGDA corpus in which it is used as a temporal adverb, but only uses of the DM 11ow is included in this study.

5·

Texa~

German discourse pragmatics 469

so was. haben ihr eigenes Fleisch gehabt, und alles and everything so--what had meat have their own 'That was a hard time, especially for people in the city. Now, some people were on farms, and they had their own gardens and their own (1-59-1-16-a) meat and everything like that:

In (18), the speaker has been talking about how well everyone ate and how much people could grow in their gardens, but then shifts into a digression about how things were during the Depression. She then uses now to highlight that people who had their own farms during this time still were able to eat well in contrast to peo-ple in the city. ln Example (19), now is used in conjunction with English and and German aber ('buf). (19)

ich war. Die hat ieber den River gewohnt von uns - von wo was from us - from where she has across the river lived zsammen kommen bis spiiter in uh An now aber wir sind nie until later in uh and now but we are never together come wa in die Hochschul waren. wenn wa, WO when we where we in the high-school were 'She lived on the other side of the river from where we lived. Now, we never got (1-21-1-12-a) together until much later when we were in high school:

In Example (19), the speaker is telling a story about when he met his wife. After explaining that she lived across the river from him, he uses now to emphasize that in spite of the fact that they knew each other from a young age, they did start dating until high school. There is another example of now in the TGDA corpus that is difficult to classify: (20)

Oh Gott nochmal. Das weiss ich now aber he was a subject of the that know I now but he was a subject of the Oh God again Duke of Nassau. And how he came here I don't know. Wie er Duke of Nassau and how he came here I don't know how he hergekomm is, das uh . . . no er . . . diih. Ah de. i weiss nich. doh Uh tha- I know not. no he here-came is that uh Ah das is uh ... das is suh bisschen uh uh a little cloudy now. uh uh a little cloudy now little that is a Ah that is uh Das ist. uh su lang zurick un dann naja. is. uh too long back and then whatever.

It

'Oh God. I only know that he was a subject of the Duke of Nassau. And how he got here I don't know. And how he came here, that uh, that is ... Ah, I don't know. That's a little cloudy now. It is just too long ago and then, well, whatever: (1-24-3-1-a)

470 Hunter Weilbacher

In (20), the speaker is responding to a question about how his grandfather came from Nassau to Texas. He uses now to focus that the only thing he knows about that aspect of his grandfather's history is that he was from Nassau, a point that had already been established in the conversation, and not how he got to Texas. 'J.he use of now in this utterance to orientate to the upcoming idea seems to also trigger the switch into English. It is difficult to know what the underlying mechanisms at work are in this instance, but it appears that the mental effort required to rec.all that information has something to do with the switching of codes. The speaker then seems to be having trouble not only finding the information, but difficulty finding the words. He then gives up, dismissing the burdensome search as unimportant. 1his example seems to point to Matras' (2000) argument that speaker's may employ elements from one linguistic system to 'lighten the cognitive load~ I now turn to see as it appears as a DM in the TGDA data.

4·3

See as a discourse marker in Texas German

Much like now, see is used to focus information in the upcoming segment of speech. Verbs like see may function as DMs when they are used as clearly pragmatic elements to direct discourse (Fraser 1990: 388). 1he TGDA contains 51 examples of see used as an utterance-initial DM in TxG from 26 different speakers. For example: (21)

See der is da see he

dort

in der Schule gegang in in

is there there in the school gone

Deutsd.lland fier sowas. Germany for SO··SUch 'See he had actually gone to school in Germany to study [carpentry]: (1-2-2-25-a) (22)

zehn Das war ungeflihr vierunfufzig - das war ungefahr that was approximately four-and-fifty - that was approximately ten Jahr - zwischen die Zeit. Zwar - see it hat schon 'ne Masse see it had already a lot between the time In-fact years wie das Deutsche hat schon angefangen, ein ausgemacht out-made - how - the German had already started a bisschen weniger zu werden. little-·bit less to become 'That was about forty there were about ten years in between. In f.'tct see it had already made a difference, how German had already started to be [spoken] less: (1··1·1··17·a)

Texas German di~course pragmatics 471

(23)

See die see the

die Eltern hamm an die Seit rum gesessn, haben the parents have on the side around sat have

aufgepaBt und die Kinder haben gedanzt. supervised and the children have danc.ed 'See the parents always sat arotmd on the side, kept an eye on things, and the kids danced: {l--2--l--14--a)

In addition to the uses of 51 instances of see like those in Examples (21)---(23), there are three instances from two different speakers which co-occur with well. There are also five instance from four different speakers in which see occurs utterance-finally as a type of tag--question. 'lhe data also contain eight instances of lets see from seven different speakers, two occurrences of lemme see from two speakers, and one instance of oh let me see. These examples create a problem for classification: as mentioned in Section 3, elements like let's see may indicate that the speaker has left the matrix language of the utterance and has therefore switched codes. On the other hand, these phrases, used as DMs to indicate that the speaker is in the act of searching for infonnation, may be speech mannerisms that are not necessarily in a separate 'code: but rather evidence that TxG at its current stage is a mixed--code.

5·

Conclusions and outlook

1he TxG data analyzed in this paper raise some important questions for the study of language contact situations. In particular, the differences between borrowing and code--switching as defined in the literature seem to be rather arbitrary, and TxG at its current stage presents its own set of problems in that many of the remaining speakers have predominately spoken English from age six onward, and the majority are over 60 years of age. Because of the frequency and level of syntactic integration of the dis-course--pragmatic elements discussed here, I have reserved the term 'code--switching' for longer sections of te.xt that are clearly in English. However, the data reveal that some elements, such as lets see, are still very difficult to classify because they may exist as borrowed linguistic gestures, or nativized elements of the linbruistic system like shorter, less lexical DMs such as now. Furthermore, without a way to test where an ele-ment is located in a speaker's mentalle.xicon, it seems desirable to view these elements in light of their cognitive functions that seem to lessen the load while pragmatically indicating that the speaker is processing the upcoming utterance. This may indicate that for many TxG speakers, English is the pragmatically dominant language, but it does not rule out that these are part of one mixed--code system.

472. Hunter Weilbacher

To date, the TGDP has interviewed over 350 speakers ofTxG. The data used for this study are from interviews that have been transcribed and are therefore searchable with a concordancer, which represents less than a third of the total data collected. As existing data is transcribed and new interviews are recorded, this study can be revisited in the future.

References Aue.r, Peter. 1995. The pragmatics of code-switching: A sequential approach. In One speaker, 1\.vo Languages: Cross-disciplinary Perspectives on Code-switching, Lesley Milroy & Pieter Muysk.en (eds), 115-135. Cambridge: CUP. Auer, Peter. 1998. Introduction: Bilingual conversation revi.~ited. In Code-switching in Com•ersation: Language, Interaction, and Identity, Peter Auer (ed.), 1-25. London: Routledge. Auer, Peter. 1999. Prom codes witChing via language mixing to fused lects: Toward a dynamic typology ofbilingual speech. International Journal of Bilingualism 3(4): 309-332. Auer, Peter. 2005. A post-script: Code-switching and social identity. Pragmatics 37: 403-410. Boas, Hans C. 2006. From the field to the web: Implementing best-practice recommendations in documentary linguistics. Language Resources and Evaluation 40(2): 153-174. Boas, Hans C. 2009. The Lifo and death of Texas German. Durham NC: Duke University Press. Boas, Hans C. & Weilbacher, Hunter. 2007. How universal is the pragmatic detachability scale? Evidence from Texas-German discourse markers. 11te Proceedings of the Texas Linguistic Society IX Conf-erence, 33--58. Boas, Hans C. & Weilbacher, Hunter. 2009. 1he unexpected survival ofGerman discourse markers in Texas German. Proceedings of the Chicago Linguistics Society 42nd Conference, 1--15. Brinton, Laurel J. 1996. Pragmatic Mf.lrkers in English: Grammaticalizatton and Discourse Functions. Berlin: Mouton de Gru}1:er. Brody, Jane. 1987. Particles borrowed in Spanish as discourse markers in Mayan languages. Anthropological Linguistics 29( 4): 507-- 521. Clyne, Michael. 2003. Dynamics o_fLanguage Contact. Cambridge: CUP. Fuller, Janet. 2001. TI1e principle of pragmatic detad1ability in borrowing: English--origin discourse markers in Pennsylvania German. Linguistics 39: 351-369. Gilbert, Glenn 1965. English loanwords in the German of Fredericksburg, Texas. American Speech 40(2): 102-112. Goss, Emily & Salmons, Joe. 2000. The evolution of a bilingual discourse marking system: Modal particles and English markers in German-American dialects. International Journai ofBilingualism 4(4): 469-484. Guion, Susan G. 1996. The death of Texas German in Gillespie county. In Language Contact across the North Atlantic, P. Sture Ureland & lain Clarkson, 243-263. Thbingen: Niemeyer. Halliday, Michael A.K. & Hasan, Ruqaiya 1976. Cohesion in Engfislz. London: Longmans. Hlavac, Jim. 2006. Bilingual discourse markers: Evidence from Croatian-English code-switching. Journal of Pragmatics 38: 1870-1900. Jordan, Terry G. 1977. The German element in Texas. Rice Unil'ersity Studies 63(3): 1-11. Lewis, Diana M. 2003. Rhetorical motivations for the emergence of discourse particles, with special reference to English ofcourse. InPtzrtides [Belgian Journal of Linguistics 16), Ton van der Wouden, Ad Foolen & Piet Van de Craen (eds), 79-91. Amsterdam: John BeiJ:iamins.

Texas German discourse pragmatics 473 Maschler, Yael. 1998. 0 n the transition from code-S\Iritching to a mlxed code. In Code-Switching in Conversation: Language, Inft:"'raction and Identity, Peter Auer (ed.), 125-149. London: Routledge. Matras, Yaron. 1998. Utterance modifiers and universals of grammatical borrowing. Linguistics 36: 281-331.

Matras, Yaron. 2000. Fusion and the cognitive basis for billngual discourse marking. International Journal of Bilingualism 4(4): 505-528. Muysken, Pieter. 1995. Code--switching and grammatical theory. In One Speaker, TWo Latl·· guages: Cross-disciplinary Perspectives on Code--switching~ Lesley Milroy & Pieter Muysken, 177--198. Cambridge: CUP. Muysken, Pieter. 2007. Mixed codes. In Hattdbook ofMultilinguali..~m and Multiiingual Commu-nication, Peter Atter & Li Wei (eds), 315---340. Berlin: Mouton de Gruyter. Myers-Scotton, Carol. 1995. A lexically--based model of code--switching. In One speaker, Two Languages: Cross--disciplinary Perspectives on Code-switching, Lesley MUroy & Pieter Muysken (eds), 259-275. Cambridge: CUP. Myers-Scotton, CaroL 1993. Duelling languages: Grammatical structure in code-switching. Oxford: Clarendon Press. Myers-Scotton, Carol. 2002. Contact Linguistics: Bilingual Encounters and Grammatical Outcomes. Oxford: OUP. de Rooi,, Vincent A. 2000. French discourse markers in Shaba Swahili conversations. International Journal ofBilingualism 4(4): 447-467. Salmons, Joe. 1983. Issues in Texas German language maintenance and language shift. Monatshefte 75: 186---195. Salmons, Joe. 1990. Bilingual discomse marking: Code switching, borrowing, and convergence in some German--American dialects. Linguistics 28: 453- 480. Schtlfrin, Deborah. 1987. Discourse Afarkers. Cambridge: CUP. Schourup, Lawrence. 200 l. Rethinking wei/. Journal of Pragmatics 33: 1025-1060. Ser.ra, Celicia Oesch 1998. Di.scourse connectives in bilingual conversation: The case of an emerging Italian-French mixed code. In Code-Switching in Conversation: Language, Interaction and Identity, Peter Auer (ed.), 101-122. London: Routledge. 'lexas German Dialect Archive. 2009. .

Index

A Ablaut 36 Adverb(ial) 7, 129, 134, 147, 168, 242, 2)2, 254, 267-268, 280, 296, 305, 371, 375, 377··378, 380, 38}, 386, 392··393, 397, 405, 417···418, 433, 437, 441··-442, 444> 447 l\ffrkates 36-38, 40, 43, 46, 55-57 .-\greement hierarchy 77, 84, 96,101-103 .-\lemannic dialect (s) 2, 50,173,205,237,249, 257··258 Allative 385, 394, 400, 406 Allomorphs 33--36, 40, 43"·-44, 46, 48, 53> 55, 58-)9, 60-61, 348, 351, 356 t\mana German 5,111-112,122 Ambisyllabic structures so Animacy 5, 67-69, 77, 81, 83---88 Auslautverharumg 5, 13, 15, 52 see final devolcing

B Bersntol 5, 31

c c. artography

277 Cimbrian 3, 5--7, 233---238, 240-249, 251, 253-267, 269--272, 279--282, 284-285,289-297 Clitic--Doubling (CD) 3.52--353. 355-356, 359-361 Cliticization (TP-, CP-, pronotm-) .238, 249-251, 319 Code-switching 2, 16, 30, 455, 456, 458--461, 464, 470--"471

Complementation 279, .280-281, 292, 392, 394, 399, 402 Complementizer 6, 279, 280-284, 286-287, 290--292, 296--297, 302--303, 310, 313--314, 316--325> 327> 334> 336, 342· 348, 350---352, 356, 362, 372, .392, 403 Constraints 33-35, 47-48, 52-53, 55-56, 58, 60-61, 113-114, 117-123, 12)-126, 135-136, 2 71, 465 Violable constraints 113,117 Conversation Analysis (CA) 5, 7, 78, 133, 415, 420

D Dative-object Dttferentiation (DOM) 257 Defmite article 70, 79-80, 82, 89, 91-92, 10.2-103, 109, 153> 155, 193,196, 203, 205 Deflection 70, 76---78, 83, 88--90, 92, 108 Discourse Marker (DM) 7-8, 76-;8, 83, 88-90, 92, 108 Do-support 193,196, .203-.204, .206, 208, 2ll-2:13, 215, 21.8-2:10, 222, 224--·225, 227 Dominant language 2-3, 127, 134, 144> 45.5, 461, 464, 471 Double Base Hypothesis 3, 6, .261, .263, 266 Dutcll 5, 14, 50, 68-76, 78, 8o, 83, 85, 89, 92, 99, 107, 108, 111-119, 122-123, 126-127, 165--·170, 172--183, 239, 246--·250, 253> 256, 2.59, 296, 387

E Eastern \Visconsin 13-15, 30-31 English 6-8, 13-16, 18, .20-22, 24-25, 27, 29-31, 40, 68---jO, 75, 78, 80, 92, 99, 105·--106, 108, 111·--114., 116--117, 120·--122, 126--127, 129, 131·--141, 142·--144> 146, 151-161, 1f2, 177, 192-19+ 196, 197, 202, 234> 238, 242, 247-248,253,267,270, 293-294, :10:1, 308, :120, 323, 326, 391 396-397, 399-402, 404-407, 417--418, 421--·422, 433, 438, 440-441, 455-459, 461--471 Epenthelic vowel 42-45 Event-related adverb 7, 371, 377-379 Rxtraposition 7, 170, 184, 249, 371-377

F Feature geometry 69, 77, 81, 86 Final devoicing 5, 13--15, 50---52 see Auslautverhartung Final laryngeal distinctions 13-15 Finallarynge al neutralization ;, 13-16, 24--31 Finnish 129, 443 First (Ll) language attrition 1-4, 30, 147, 281 First language (Ll) 2--4, 14, 30, 70, 372 Flemish 69, 71-75, 166, 247, 2)0, 256 Fricatives 15-16, 21, 34-44, 46, 49---53, 55--58, 6o, 152 Frisian .5o, 69, 71--73, 165--166,387

476 Index

G Gender 5-6, 67-90, 92-94. 96-108, 151-161, 187-188, 19J, 197. 201, 203, 205-209, 211-2.25, 252, 340 Attrition (gender) 78, 83, 92,103, 108 Common gender 76, 158 Feminine 67---69, 71---~, 76, 78-82, 84-90, 92, 94· 97··98, 100, 102···104, 106, 151-161 Granunatical gender 5, 67-70, 72-74> 76-78, 80-81, 83-84> 87-88, 90, 94. 96, 102-103, 106-107, 151-152, 154> 156-158 Hybrid gender 67, 69, go, 96,100 Masculine 6, 67-69, 71-~, 76, 78-79, 81-82, 86-90, 92, 94-95. 97-98, 104> 106, 151-160, 196-197. 203, 205 Neuter 67-69, 76-82, 85-90, 92, 97-102, 106, 1o8, 151·-153. 155·--160, 348 Semantic gender 5, 67-69, 76···78, So, 82---84, 87--88, !)0, 96, 101-103 Strong grammatical gender (SGG) 5, 74, 76, 78, 90, 1o6-107 Weak grammatical gender (WGG) 78, So, 82-83, go, 92·-gJ, 96---97, 102, 104, 108 Generative syntax 238 German 1-7,13-22, 24-·25, 27-31, 33. 36-37. 40, 42, 44-45. 47-48, so, 52, 61, 67-75. 78, 81, 89-90, 99. 105-108, 111-114> 116-120, 12.2-124, 126-127, 129-147· 151-156, 158-161, 165-170, 172-173. 177-178, 181-182, 187-189, 191-208, 210-212, 214-217, 219-227, 233"-241· 243. 245-263, 265-267, 269-271,281-282,295-296, 301-303, 305-308, 31.0-311, 313. 318, 322-324. 334> 336,338,342.347-348, 352, 355···3.59. 361·-364.

371-383, 385-389, 391, 394-395.401, 403. 405, 415-42.2. 433. 436-439. 441-445· 447-448. 455-459. 461-470 see Standard German Gramrnaticaliza.tion 6-7, 26:;, 279-280, 285, 290-291, 296, 356, 363"-·364. 386, 396-398,401-402, 406---407, 420, 443

H Henderson Plaut:d.ietsd1 4-5, 67-81 Hybrid Nouns 69,96-97, 99-104, 106,108 Gender 67, 69, 90, 96, 100 I Identity function 114-115, uo Incomplete acquisition 2 Indefinite article(s) 69, 76, 79-80, 82-83, 89, 91-92, 96---97· 102---103 Infinitival marker 7, 315, 38.5·--386, 394···398, 400, 405-406 Inflllitivus pro Participio (IPP) 166-167, 170-171, 173-175, 178-181, 184 Intensifiers 5, 111-114, 116-117, 126 Italo--Romance (IR) J, 233---234 Ivy Hypothesis 2

Low Franconian 69, 71-72 Low German 6, 31, 68-69, 72-74. 78, 187-188, 191-192, 194> 197. 208, 215,217 I.use.rr1a 6, 235, 240-24.2, 245-246, 248-251, 253. 257> 262, 280-281, 284-285, 301--304> 306·-JOS, 310, 336,347 M Manitowoc 13, 16-19, 21, 23-26 Mennonite I.ow German (MLG) 6, 166,187, 192, 194-202, 204-207, 210-211, 213, 217-219, 2.21, 225-226 Middle Field!Mittelfeld (MF) 168, 238-240, 242, 253.255. 260, 270, 350, 352, 356, 362, 373, 379- JSO, 382,417 Minimalism I Minimalist Program (MP) 262, 334> 358 Minority language 166, 218, 26,5, 279··-280, 284,291, 388, 465 Mod1eno (Fersentalerisch) 4-5. 33-51, 53. 56-61. Morphology 3, 5, 92, 101-103. 152, .238-239. 264. 282, 286-287, 2!)0, 348, 357> 420>442>444

0 K Kiel 15-17, 23-24, 26 L

Left periphery 241-242,244, 246, 266-267, 284> 286, 289,301-303,310, 320, 322-323. 326, 328-329. 338-341354.356-359. 361-362, 364 Lexical borrowing 5, 129-130, 133-135,139-140,147 I.e:xical erosion 130, 144, 146,148 I.e:xical Functional Grammar (LFG) 7, 402, 407

Object pronouns 7, So-81, 347-350, 353> 356-357. 362-363 Old English 129, 242, 267, 270, 293-294.308,399.4o6 Optimality Theory 5, 113 OV (Object- Verb) 234-235, 241-242, 246-247> 249, 253-254· 261, 26J, 265-2~, 3~. :;82 Ozaukee (Wisconsin) 13, 16, 19-21, 27-29

p Palatinate German (PL) 7, 155, 371. 374··382, 394

Index 477 Past p-articiple(s) 33"··34, 36···48, 5J··56, 59·-61, 177-178, 180 Pennsylvania Dutch (Pennsylvania German) 4> 15:1, 160-161, 1();-169, 172-183, }~2-383 Phonetics 5, 13-vh 190 Phonology J··s, 13, 52, 138, 190, 407 Plautdietsch 4---5, 67--70, 73-75, 78-83, 8;, 89-105, 107-109, 166 Pomerania 15-16, 19, 27, 198 Pragmatics 7, 291, 420, 437, 439, 442-444 Predicates Self-directed 5, 111 Other--directed 5, 111--112, 114, 125 Prokosch's Law so-;2 Prosody 415, 417-418, 438, 44:1 455-456 Purposive 7, 385-387, 389, 391-399, 401, 405-407, 409

R Reanalysis 7, 180-181, 184, 254> 277, 289, 293, 385-386, 395-397' 405-406 Reflexives 5, 111-114, 122, 126 Romance 7, 14, 38-39, 233-234, 236, 238·-·240, 246·-·247, 249> 252···25_5, 257···258, 260---263, 267, 2j'l···272, 281, 286, 302, 310, 321, 324-> 327> 333···334, 336, 342, :147-349, 351-:153, :155-357, 359-.364

s Second language (L2) 3, 30,372 Semi-speaker 2, 4, 144-145, 218,279, 284···287, 290-291, 348

Semantics 3, 36, 96---97, 99, 103,112···113,152, 155, 172, 1}7, 407, 418, 442, 447-448 Siebenbiirger Siichsisch 7, 415-416, 419-421, 425, 433, 436-437, 440, 442-446, 448 Sociolinguistics 1-4, 6, 34, 112, 127, 133, }40, 16;, 187··188, 190-191, 201, 210-211, 213, 221, 223, 227, 264··265, 4.55, 460,463 Sonorants 37, 40-41, 43-44, 46, ;o, 52, 57-58, 6o Sprachinsel 1-8, 111-112, 127, 151-152, 284-285, 455 Standard German 7, 13---14, 16, 19, 24, 27··30, 36·-37· 42, 44···45> 47-·48, 70,124, 133,138, 145> 151, 153···154, 156, 158-160, 166, 1il:1, 189, 192-207, 210-212, 216-.217, 219-227, 237-238, :.!.40, 243, 246, 256, .259-260, 266, 270-2j'l, 373, 375-382, 39.5, 403, 4}6, 421, 437· 444, 447 see German Structural borrowing 129, 134-135, 139-140, 147> 188 Subject inversion 6, 240, 243> 245-246, 254, 270, 302-325, 327, 335, 336, 342, 35<~ 357 Subsegment 33, 35, 43, 59 Syllable structure monosyllable words 151··1.52, 156---160, 242--·243 polysyllabic words 151, 154157 Syntax 3-4, 6-7, 111, 165, 173, 190-191, 233-234, 238, 253, 262, :.!.66-267, :.!.69, 280, .282-283, .289-291, 308-309, 312, 338, 341 347···348, 350, 3.54> 356--3.59. 362-364, 372, 375> 378, 407, 416·-·418, 433, 438··439, 443, 448

T Texas German (TxG) 5, 7, 129-133, 137, 455-459, 466-468, 470 Texas German Dialect Project (TGDP) 132, 455, 458 Topics 196, 239, 244 246, 254, 260, 26;, 2j'l, 28,h 288, 291··298, 301--303, 305, 308, 318-322, 324-326, 332··342, 3.54, 415--·416, 418, 420-421, 423-424> 426, 431, 433-435, 437, 442, 444-447> 463 Trentino (dialec.t) 33, 38-40, 235, 280, 282, 301 Turn Construction Unit (TCU) 7, 415---416, 419, 436, 444 1)'Iolean dialects (Meran, Passe.ie.r, Upper Vinschgau) 40-42, 234-236

v VO (Verb-Object) 240-243, 236-219, 253-254> 263, 266--267, 269, 272, 372 Verb cluster 6, 165-167, 170, 173-184, 192-196, 199,372 Verb-last (Vlast) 235, 238-239, 242, 247> 249-250, 261, 263,269-270 Verb--second (V2) 5, 234. 301, 316, J22, 325· 350, 352, 357, 37.5··-376, 417-·418 Verbal fran1e (Verbalklammer) 168

w Wackernagelpositio11 347, 352, 356, 362-364 West Germanic dialects/languages 50, 69, 165-166, 173,181, 357 l\Tn--items 292,294, 301--302, 322,332-334,338,340,342

Studies in Language Companion Series A complete list of titles in this series can be found on the publishers' website, www.benjami 124 MALCHUKOV, Andrej and Anna SIEWIERSKA (eds.): Impersonal Constructions. A cross--linguistic perspective. Expected .Marcl12o11 123 PUTNAM, Michael 1'. (eel.): Studies on Gemtan-Language Islands. 2011. xii, 477 pp. 1.21. Cl-ANCY, Ste11en J.: The Chain of Bc.ing and Having in Slavi•- 2010. xvi~ 1.97 pp. 1.21. BRII.,habelle (ed.): Clause Linking and Clause Hierarchy. Syntax and pragmatics. 2010. viii. 632. pp. 120 ROTHSTEIN,Bjom and Rolf TlllEROFF (eds.): Mood in the Languages of Europe. 2010. xvi. 647 pp. 119 STATffi,Katerlna, Flke GRHWEII.F.R and Ekkebard KONIG (eds.): Grammaticalization. Current views and issues. 2010. vii, 379 pp. 118 M"OHLEISEN, Susanne: Heterogeneity in Word-Formation P..rtterns. A corpus-based analysis of surTJXation with-ee and its productivity in English. 2010. xiii, 245 pp. 117 SI1EVAK, Olga: Constituent Order in Classical Latin Prose. 2010. xv, 318 pp. 116 NORDSTROM, Jackie: Modality and Subordinators. 2010. xvii, 341 pp. 115 HASKO, Victoria and Renee PERELMUTTER (eds.): New Approaches to Slavic Verbs of Motion. 2 X, 392 pp. 114 ROBY, David Brian: Aspect and the Categorization of States. The case of ser and estar in Spanish. 2009. xiii, 191 pp. 113 COMRIE, Bernard, Ray FABRI, Flizabeth HUME, Manwel MIFSUD, Thomas STOLZ and Martine VANHOVE (eds.): Introdudng Maltese Linguistics. Selected papers from the 1stinternational Conference on Maltese Linguistics, Bremen, 18-20 Octobet; 2007. 2009. xi, 422 pp. 112 DUFTER, Andreas and Daniel JACOB (eds.): Focus and Background in Romance languages. 2009. vii, 362 pp. 111 POLGUi!RE, Alain and Igor A. MEL'CL'K (eds.): Dependency in Linguistic Description. 2009. xxii, 281 pp. 110 DIMMENDAAL, Gerrit J. (eel.): Coding P..utidpant Marking. Construction types in twelve African languages. 2009. xvi. 389 pp. 109 NARROG,Heik~ Modalil'f in Japanese. lhe layered structure of the clause and hierarchies of functional •.ategories. 2009. xxii, J.77 pp. 1.08 BARDDAI., J6hanna and Shobhana I. CHEI.I.IAH (eds.): The Role of Semantic, Pragmatk. and Discourse Factors in the Development of Case. 2009. n:, 432 pp. 107 BUTLER,
97 96

95 94 93 92 91 90 89 88 87

86

85 84 83 82 81 8o 79

78 77

76 75 74 73 72 71 70 69 68

DOUJ:NGER. Stefan: New- Dialect Formation in Canada. Evidence from the English modal auxiliaries. 2oo8. xxii, 35 5 pp. ROMEO, Nicoletta; Aspect in Butmese. Meaning and function. 2008. xv, 289 pp. O'CONNOR, Loretta: Motion, Transfer and Transformation. The grammar of change in Lowland Chontal. 2007. xiv, 251 pp. MIES1'AMO, Matti, Kaius SINNEMAKI and Fred KARLSSON (eels.): Language Complexity. Typology, contact, change. 2008. xiv, 356 pp. SCIIAILEY, Andrea C. and Drew KHLENTZOS (eds. ): Mental Slates. Volume 2: Language and wgnitive structure. 2007. x. 362 pp. SCHALI.EY, Andrea C. and Drew KliLENTZOS (eels.): Mental States. Volume 1: Evolution, function, nature. 2007. xii, 304 pp. FIUPOVIl:, I.una: Talking about Motion. A crosslinguistic investigation of le.xicali7.ation patterns. :zoo7. X. 182. pp. MUYSKEN,Pieter (ed.): From Linguistic Areas to Areal Linguistics. 2008. vii, 293 pp. STARK,Flisabetb,EiisabethLF.ISS and Werner ABRAHAM (eds.): Nominal Deterntination. Typology, context constraints, and historical emergence. 2007. viii, 370 pp. RAMAT, Paolo and FJisa ROMA (eds.): Europe and the Mediterranean as linguistic Areas. Convergencies from a historical and typological perspective. 2007. xxvi, 364 pp. VERHOEVEN, FJisabeth: Experiential Constructions in Yucalec Maya. A typologically based analysis of a functional domain in a Mayan language. 2007.xiv, 380 pp. SCHWARZ-FRIBSEL, Monlka, Manfred CONSTEN and Mareile KNEES (eds.): Anaphors in Tel.i. Cognitive, formal and applied approaches to anaphoric reference. 2007. xvi, 282 pp. BUTLER, Christopher S.,Raquel IDDALGO DOWlii'ING and Julia LAVID (eds.): Functional Perspectives on Grammar and Discourse. In honour of Angela Downing. 2007. XXX. 481 pp. WANNER.J..eo (ed.): Selected Lexical and Grammatical Issues in the Meaning-Text Theory. In honour of Igor Mel' cuk. 200 7. xviii, 380 pp. HANNAY, Mike and Gerard J. STEEN (eds.): Structural-Functional Studies in English Grammar. In honour of Lachlan Mackenzie. 2007. vi, 393 pp. ZIEGELER, Debra: Interfaces with English Aspect. Diachronic and empirical studies. 2006. xvi, 325 pp. PEETERS, Bert (ed.): Semantic Primes and Universal Grammar. Empirical evidence from the Romance languages. 2006. xvi, 374 pp. BIRNER,Betty J. and Gregory WARD (eds.): Drawing the Boundaries of Meaning. Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn. 2006. xii, 350 pp. LAFFUT, An: Three--Participant Constructions in English. A functional-cof,nitive approach to caused relations. 2006. ix, 268 pp. YAMAMOTO, Mutsumi: Agency and Impersonality. Their Linguistic and Cultural Manifestations. :zoo6. X, 152 pp. KUUKOV,Leonld,Andrej MALCIIUKOV and I'eter de SWART (eds.): Case, Valency and Transitivity. 2006. XX. 503 pp. NEVALAINBN, Thrttu,JubaniKLEMOLA and Mikko LAITINEN (eds.): Types of Variation. Diachronic, dialectal and typological interfaces. 2006. >iii, 378 pp. HOLE,Daniel,Andre MEINUNGBR and Wemer ABRAHAM (eds.): Datives and Other Cases. Between argument structure and event structure. 2006. viii, 385 pp. PIBTRANDREA,Paola: Epistemic Modality. Functional properties and the Italian system. 2005. xi~ 232 pp. XIAO,Richard and Tony McENERY: Aspect in Mandarin Chinese. A corpus-based study. 2004.x, 305 pp. FRAJZ¥NGIBR, Zygmunt, Adam HODGES and DavidS. ROOD (eds.): Linguistic Diversity and Language 'Iheories. 2005. xii, 432 pp. DAHL, Osten: The Growth and Maintenance of Linguistic Complexity. 2004. x. 336 pp. I.F..FEBVRE, Claire: Issues in the Study of Pidgin and Creole Languages. 2004. xvi, 358 pp. TANAKA, Lidia: Gende1; Language and Culture. A stndy of Japanese television interview discourse. 2004 xvii, 233 pp. MODER, Carol Lynn and Aida MARTINOVIC-ZIC (eds.): Discourse Across Languages and Cultures. 20 04. vi, 366 pp.

67 LURAGHI, Silvia: On the Meaning of Prepositions and Cases. The expression of semantic roles in Ancient Greek. 2003. xii, 366 pp. 66 NARIYAMA, Shigeko: Ellipsis and Reference Tracking in Japanese. 1.003. xvi, 400 pp. 65 MATSUMOTO,Kazuko: Intonation Units in Japanese Conversation. Syntactic. informational and functional structures. 200 3.1:viii, 215 pp. 64 BUTLER, Christopher S.: Structure and Function- A Guide to 1hree Major Stru<.."tural-Punctional 'Iheories. l'art 2: From clause to discourse and beyond. 2003. xiv, 579 pp. 63 BUTLER, Christopher S.: Stru<."ture and Function- A Guide to Three Major Structural-Functional Theories. Part. 1: Approaches t.o the simplex clause. 2003. xx. 573 pp. 62 FIELD, Fredric: Linguistic Borrowing in Bilingual Contexts. With a foreword by Bernard Comri~. xviii, 2 55 pp. 61 GODDARD, Clift' and Anna WIERZBICKA (eels.): Meaning and Universal Grammar. Theory and empirical findings. Volume 2. :zoo2.. xvi, 337 pp. 6o GODDARD, Clift' and Anna WIERZBICKA (eels.): Meaning and Universal Grammar. Theory and empirical findings. Volume 1. 2002. rvi, 337 pp. 59 SHI, Yud1i: 1he Establishment of Modern Chinese Grammar. The formation of the resultative construction and its elt'ects. 2002. xiv. 262 pp. 58 MAYLOR, B. Roger: Lexical Template Morphology. Change of state and the verbal prefixts in German. 2002. X. 27 3 pp. MEL'CUK,Igor A.: Communicative Organization in Natural Language. The semantic-communicative structure of sentences. 2001 . .xii, 39 3 pp. 56 FAARLUND, Jan Thrje (ed.): Gmmmatical Relations in Change. 2001. viii, 326 pp. DAHL, Osten and Maria KOPTJEVSKAJA-TAMM (eds.): Circum-Baltic Languages. Volume 2: Grammar and Typology. 2001. xx, 423 pp. 54 DAHL, Osten and Maria KOPTJEVSKAJA-TAMM (eels.): Circum-Baltic Languages. Volume 1: Past and Present 2001. XX. 382 pp. 53 FISCIIER,Oiga,AnetteROSENBACII and Dieter STEIN (eds.): Pathways of Change. Gr.mnnaticalization in English. 2000. x, 391 pp. TORRES CACOULLOS, Rena: Grammaticization, Synchronic Variation, and Language Cont&."t. A study of Spanish progressive -ndo wnslructions. 2ooo.:~..vi, 255 pp. ZIEGELER,Debra: Hypothetical Modality. Grammaticalisation in an L2 dialect 2000. XX. 290 pp. ABRAHAM, Werner and Leonid KULIKOV (eels.): Tense-Aspect. Tnmsitivity and Causativity. Essays in honour ofVIadimir Nedjalkov. 1999. xxxiv, 359 pp. 49 BHAT,D.N.S.: The Prominence of Tens~. Aspect and Mood. 1999· xii, 198 pp. 48 MANNEY, Linda Joyce: Middle Voice in Modern Greek. Meaning and function of an inflectional category. 2.ooo. xiii, 262 pp. 47 BRINTON,I.aurel J. and Minoji AKIMOTO (eds.): Collocational and Idiomatic Aspects of Composite Predicates in the History of English. 1999. xiv, 283 pp. 46 YAMAMOTO, Mutsumi: Animacr and Reference. A cognitive approach to corpus linguistics. 1999. xviii, 278 pp. 45 COWNS, Peter and David LEE (eels.): The Clause in English. In honour of Rodney Huddleston. 1999· XV, 342 pp. 44 HANNAY, Mike and A. Machtel.t BOLKESTEIN (eels.): Functional Grammar and Verbal Interaction. 1998. xii, 304pp. 43 OLBERTZ, Hella, Kees HENGEVELD and Jesus SANCHEZ GARcfA (eds.): The Structure of the Lexicon in Functional Gramma& 1998. xii, 312 pp. 42 DARNELL, Michael, Edilb A. MORAVCSIK, Michael NOONAN, Frederick J. NEWMEYER and Kadlleen M. WHEATLEY (eels.): Functionalism and Formalism in Linguistics. Volume II: Case studies. 1999. vi, 407 pp. 41 DARNELL, Michael, Edith A. MORAVCSIK, Michael NOONAN, Frederick J. NEWMEYER and Kathleen M. WHEATLEY (eds.): Functionalism and Formalism in linguistics. Volume I: General papers. 1999· vi, 486 pp. 40 BIRNER,BettyJ. and Gregory WARD: Information Status and Noncanonical Word Order in English. 1998. xiv, 314 pp. 39 WANNER, l-eo (eel.): Recent Trends in Meaning .. Text 1heory. 1997. x.x. 202 pp.

38 HACKING, Jane F.: Coding the Hypothetical. A comparative typology of Russian and Macedon ian conditionals. 1998. vi, 156 pp. 37 HARVEY, Mark and Nicholas REID (eds.): Nominal Classification in Aboriginal Australia 1997. x, 296 pp. 36 KAMIO,Akio (ed.): Directions in Functional Linguistics. 1997. .xiii. 259 pp. 35 MATSUMOTO, Yoshiko: Noun--Modifying Constructions in Japanese. A frame semantic approach. 1997. viii,204pp. 34 HATAV, Galla: The Semantics of .Aspe<-1: and Modality. Evidence from English and Biblical Hebrew. 1997· X, 224 pp. 33 VELAZQUEZ -cASTILLO, Maara: 'Ihe Grammar of Possession. Inalienability, incorporation and possessor ascension in Guarani. 1996. xvi, 274 pp. 32 FRAJZYNGIER, Zygmunt: Grammaticalization of the Complex Sentence. A case study in Chadic. 1996. xviii, 5 01 pp. WANNER, Leo (ed.): Le.xical Functions in Le.xicography and Natural Language Processing. 1996. XX, 355 pp. 30 HUFFMAN,Alan: The Categories of Grammar. French lui and Ie.1997· xiv, 379 pp. :z9 F..NGBF!.R.G-·PEDERSEN, Elisabeth, Michael FORTF.lSCTJE, Peter HARDF!.R.,I.ars HF!.I.TOFI' and Usbeth Falster JAKOBSEN (eds.): Content, Expression and Structure. Studies in Danish functional grammar. 1996. xvi, 510 pp. 28 HERMAN, J6zsef (ed.): Linguistic Studies on Latin. Selected papers from the 6th International Colloquium on Latin Linguistics (Budapest.23-27 March 1991). 1994· ix, 421 pp. 27 ABRAHAM, Werner, T. GIV6N and Sandra A. THOMPSON (eds.): Discourse, Grammar and Typology. Papers in honor ofJohn W.M. Verhaar. 1995. XX. 352 pp. 26 UMA, Susan D., Roberta CORRIGAN and Gregory K. IVERSON: The Reality of Linguistic Rules. 1994. xxiii, 480 pp. GODDARD, Cliff and Anna WIERZBICKA (eds.): Semantic and Lexical Universals. Theol"y and empirical findings. 1994. viii, 510 pp. 24 BHAT, D.N.S.: The Adje<.tival Categorr. Criteria for differentiation and identification. 1994· xii, 295 pp. 23 COMRIE, Bernard and Marla POUNSKY (eds.): Causatives and Transitivity. 1993. x, 399 pp. 22 McGREGOR, WDliam B.: A Functional Grammar of Gooniyandi. 1990. xx, 618 pp. 21. COI..EMAN, Robert (ed.): New Studies in Latin Linguistics. Proceedings of the 4th International Colloquium on Latin Linguistics, Cambridge, April1987. 1990. x, 48o pp. 20 VERIIAAR, S.J., John W.M. (ed.): Melanesian Pidgin and Tok Pisin. Proceedings of the First International Conference on Pidgins and Creoles in Melanesia 1990. xiv, 409 pp. 19 BLUST, Robert A.: Austronesian Root 1heory. An essay on the limits of morphology. 1988. xi, 190 pp. 18 WIERZBICKA, Anna: 1he Semantics of Grammar. 1988. vii, 581 pp. 17 CALBOLI, Gualtiero (ed.): Subordination and Other Topics in Latin. Proceedings of the 1hird Colloquium on Latin Linguistics, Bologna. 1···5 Apri1198_). 1989. xxix, 691 pp. 16 CONTE, Maria--FJisabeth, Janos Sander PETOFI and HIDel. SOZER (eds.): Text and Discourse Connectedness. Proceedings of the Conference on Connexity and Coherence, Urbino, July 16--21,1984. 1989. xxiv, 584 pp. 15 JUSTICE, David: The Semantics of Form in Arabic. In the mirror of European languages.1987. iv. 417 pp. 14 BENSON, Morton, Hvel.yn BENSON and Robert F. ILSON: Lexicographic Description of English. 1986. xiii, 275 pp. 13 REESINK,GerP.: Structures and their Functions in Usan. 1987. xviii, 369 pp. 12 PINKSTER,Harm (ed.): Latin Linguistics and Linguistic Theory. Proceedings of the 1sllnternational Colloquium on Latin Linguistics, Amsterdam, April1981. 1983. xviii, 307 pp. PANHUIS, Dirk G.J.: The Communicative Perspe<.tive in the Sentence. A study of Latin word order. 1982. viii, 172 pp. 10 DRF.lSSI.r"IR, Wolfgang lJ., Willi MAYERffiAI.ER, Oswald PANAGI. and Wolfgang lJIIrich WlJR7.F!.I.: LeitmotifS in Natural Morphology. 1987. ix, 1.68 pp. 9 I.ANG, Ewald and Joho PHEBY: The Semantics of Coordination. (English trans!. by John Pheby from the German orig. ed. 'Semantik der koordinativen Verknllpfung', Berlin, 1977). 1984- 300 pp. 8 BARTH, E.M. and J.L. MARTENS (eds.): Argumentation: Approaches to Theory Formation. Containing the Contnbutions to the Groningen Conference on the 1heory of Argumentation, October 1978. 1982. xviii, 333 pp.

6

4 3

PARRET, HennBD, Marina SBISA and JefVERSCHUEREN (eels.): Possibilities and Limitations of Pragmatics. Proceedings of the Conference on Pragmatics, Urbino, July8-14.1979. 1981. x, 854 pp. VAGO, Robert M. (ed.): Issues in Vowel Hat1no1ty. Proceedings of the CUNY Linguistics Conference on Vowel Harmony, May 14,1977. 1980. xx, 340 pp. IIAIMAN, John: Hua: A l'apuan Language of the Eastern Highlands of New Guinea. 1980. iv, 550 pp. LLOYD, Albert L.: Anatomy of the Verb. "Ihe Gothic Verb as a Model for a Unified lheory of Aspect. Actional Types, and Verbal Velocity. (Part I: "Iheory; Part II: Application). 1979. x, 351 pp. MALKIEL, Yakov: From Particular to General Linguistics. Selected Essays 1965-1978. With an introduction by the author, an index rerum and an index nominum. 1983. xxii, 659 pp. ANWAR, Mohamed Sam!: BE and Equational Sentences in Egyptian Colloquial Arabic. 1979. vi, 128 pp. ABRAHAM, Werner (eel.): Valence, Semantic Case, and Grantmatical Relations. Workshop studies prepared for the 12th International Congress oflJnguists, Vienna. August 29th to September 3rd,1977. 1978. xiv, 729 pp.