The Verbal Complex in Subordinate Clauses from Medieval to Modern German

The Verbal Complex in Subordinate Clauses from Medieval to Modern German Linguistik Aktuell/Linguistics Today (LA) Li...

Author: Christopher D. Sapp

13 downloads 570 Views 15MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

The Verbal Complex in Subordinate Clauses from Medieval to Modern German

Linguistik Aktuell/Linguistics Today (LA) Linguistik AktueU!Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.

General Editors VVerneri\braharn

Elly van Gelderen

University ofVienna I Rijksuniversiteit Groningen

Arizona State University

Advisory Editorial Board Josef Bayer University of Konstanz

Christer Platzack University of Lund

Cedric Boeckx ICRENUniversitat Autonoma de Barcelona

Ian Roberts Cambridge University

Guglielmo Cinque University of Venice

McGill University

Liliane Haegeman University of Ghent

StenV!kner University of Aarhus

Hubert Haider University of Salzburg

C. Jan-Wouter Zwart University of Groningen

Lisa deMena Travis

Terje Lohndal University of Maryland

Volume173 The Verbal Complex in Subordinate Clauses from Medieval to Modern German by Christopher D. Sapp

The Verbal Complex in Subordinate Clauses from Medieval to Modern German Christopher D. Sapp University of Mississippi

John Benjamins Publishing Company Amsterdam I Philadelphia

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences - Permanence of Paper for Printed Library Materials, ANSI z39.48-1984.

Library of Congress Cataloging-in-Publication Data Sapp, ChristDpher D. The verbal complex in subordinate clauses from medieval to modem German I Christopher D. Sapp.

p. em. (Linguistik.Ak:tueU!Linguistics Today, ISSN o166-o829; v.173) Includes bibliographical references and index. 1.

German language--Verb phrase. 2. German language--Syntax. 3· Grammar, Comparative and general--Verb phrase. I. Title.

PF3315.P52S27

2011

435:6--dc22 ISBN 978 90

272 5556 3 (Hb ; alk. paper)

ISBN 978 90

272 8725 o (Eb)

© 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 written permission from the publisher.

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

Table of contents

Acknowledgements

v

CHAPTER 1

Inrroduction 1.1 The verbal complex in modern and older German 1 1.2 Previous scholarship on diachronic German verb order 1.2.1 The sentence-frame analysis 4 1.2.2 The verbal complex in MHG and ENHG 5 1.2.3 The verbal complex in contemporary German 13 1.3 Organization 14

1

4

CHAPTER 2

Factors influencing verb order in MHG 2.1 Introduction 15 2.1.1 Basic facts 15 2.1.2 The database 17 2.1.3 Analysis with GoldVarb X 19 2.1.4 Organization of this chapter 19 2.2 Complexes of two verbs in MHG 20 2.2.1 Introduction 20 2.2.2 Phonological, morphological, and syntactic factors 2.2.3 Diachronic, dialectal, and sociolinguistic variation 2.2.4 Conclusion 36 2.3 Complexes of three verbs in MHG 36 2.3.1 Introduction 36 2.3.2 Phonological, morphological, and syntactic factors 2.3.3 Diachronic, dialectal, and sociolinguistic variation 2.3.4 Conclusion 43 2.4 Conclusion 44 2+1 Summary of findings 44 2+2 The combined effect of the favoring factors 44

15

20

31

38 41

VI


CHAPTER 3

Factors influencing verb order in ENHG 3.1 Introduction 47 3-1.1 Basic facts 47 3.1.2 The database 50 3.1.3 Organization of this chapter 53 3.2 Complexes of two verbs in ENHG 53 3.2.1 Introduction 53 3.2.2 Non-favoring factors 53 3.2.3 Favoring factors 56 3.2.4 Diachronic, dialectal, and sociolinguistic variation 66 3.2.5 Conclusion 75 3·3 Complexes of three verbs in ENHG 76 3.3.1 Introduction 76 3.3.2 Non-favoring factors 77 3.3.3 Favoring factors 79 3·3·4 Diachronic, dialectal, and sociolinguistic variation 85 3·3·5 Conclusion 90 3·4 Discussion 91 3.4.1 Favoring factors in individual texts and dialects 91 3.4.2 Dialect groups in ENHG 92 3·4·3 Developments from MHG to ENHG 95 3·5 Conclusion 97 3.5.1 Summaryoffindings 97 3.5.2 The combined effect of the favoring factors 98

47

CHAPTER 4

Verbal complexes in contemporary German 4.1 Introduction 101 4.2 Previous studies 101 4.2.1 Introduction 101 4.2.2 Two-verb complexes 102 4.2.3 Three-verb complexes no 4.2.4 Complexes of four or more verbs 120 4.2.5 Verbal complexes with an intervening constituent 4.2.6 Conclusion 129 4·3 Focus and two-verb complexes in dialects 129 4.3.1 Introduction 129 4.3.2 Swabian and Austrian 130 4·3·3 Zurich German 137 4·3·4 Conclusion 149

101

124

Table of contents

4·4 Focus and three-verb complexes in Standard German 4.4.1 Introduction 150 4.4.2 Schmid & Vogel (2004)

150

150

4·4·3 Magnitude estimation experiment 153 4·4-4 Discussion 160 4·5 Conclusion 161 4·5-1 Summary of findings 161 4.5.2 From ENHG to Modern German 162 CHAPTER

5

German clause structure and the prosody-syntax interface 5.1 Introduction

165

165

5.2 The structure of German subordinate clauses 165 5.2.1 Introduction 165 5.2.2 Diachronic approaches 166 5.2.3 The OV approach to German clause structure 167 5.2.4 The VO approach to German clause structure (Zwart 1996) 5.2.5 The OV and VO approaches compared 176 5·3 Focus, prosody; and verb order 179 5.3.1 Observations that need to be accounted for 179 5.3.2 Formal syntactic analyses of focus and verb order 181 5·3·3 Accounting for prosodic factors 186 5·3·4 Focus marking and its prosodic and syntactic effects 199 5·3·5 Summary 204 5·4 Conclusion 205

175

CHAPTER 6

Conclusions 6.1 Summary of findings 207 6.2 Implications 2o8 6.2.1 The origin of Standard German 208 6.2.2 The clause structure of German 210 6.2.3 The peripheral nature of verbal complex phenomena 6.2.4 Syntactic change in German 212 6.3 Outlook 213

207

211

Bibliography Primary sources from the Bochumet· Mittelhochdeutschkotpus 215 Primary sources from the Bonner FrUhneuhochdeutschkorpus 215 Secondary sources 217

215

Inde."

223

VII

Acknowledgements

This project began as a brief paper comparing Verb Projection Raising in Swiss German with the same phenomenon in West Flemish. At the end of the paper, I asked why these two non-contiguous dialects share this construction, which exists in neither Standard Dutch nor Standard German. My advisor, Rex A. Sprouse, encouraged me to seek an answer by investigating the verbal complex from a diachronic perspective, thus I owe him a debt of gratitude for shaping this project from its beginning. I would like to thank several other people for facilitating this research. KlausPeter Wegera and Sandra Waldenberger gave me access to the Bochum Middle High German Corpus, which forms the basis of Chapter 2. I conducted the surveys discussed in Chapter 4 during stays in Tiibingen where I worked with the SFB 441, in Vienna by invitation of Werner Abraham and Richard Schrodt, and in Zurich where I received valuable aid from Claudia Bi.icheli-Berger. I would also like to acknowledge all of the participants in my surveys. Allison Burkette, Owen Davis, Robert D. Fulk. Kari Ellen Gade, Stefanie Hirscher, Barbara Vance, an anonymous reviewer, and the editors at John Benjamins provided invaluable proofreading and comments. Participants at the following conferences also engaged me in helpful discussions on the more recent stages of this research: the 2007 International Workshop on Information Structure and Word Order Variation in older Germanic at Humboldt University, the 15th Germanic Linguistics Annual Conference in Banff, the 24th Comparative Germanic Syntax Workshop in Brussels, and the 11th Diachronic Generative Syntax conference in Campinas. This book is dedicated to the memory of two friends and fellow Germanisten: Marshall Billings and Prof. Jim Knowlton.

CHAPTER 1

Introduction

1.1

The verbal complex in modem and older German

A well-known characteristic of Modern Standard German is the asymmetry between main-clause and subordinate-clause word order. In main clauses, the finite verb occupies the second position of the clause (la-d), and any non-finite verb (lb) or verbal particle (lc) will occur at the end of the clause. If there is more than one non-finite verb, these will occur adjacent to each other at the end of the clause (ld), forming a "verb cluster" or "verbal complex" (VC), in which the verbs appear in a prescribed order relative to each other (le).l (1) a. Klaus liest heute das Buch. K. reads6n today the book 'Klaus reads the book today: b. Klaus will heute das Buch lesen. K. wantsfin today the book readmf 'Klaus wants to read the book today.' c. Klaus liest heute das Buch durch. K. readsfin today the book through 'Klaus reads through the book today.' d. Klaus will heute das Buch lesen kiJnnen. K. wantsfin today the book readmr canmr 'Klaus wants to be able to read the book today: e. *Klaus will heute das Buch kiJnnen lesen. K. wantsfin today the book Catl;.0 f readinr

Throughout this work. parts of the verbal complex will be italicized In addition, extraposed or intervening constituents will be indicated by bold face.

1.

2


In subordinate clauses introduced by complementizers, on the other hand, all verbs, finite and non-finite, occur at the end of subordinate clauses (2a).2 If the subordinate clause has two or more verbs, these will form a verbal complex. Word order within the VC is fixed, at least when the complex consists of only two verbs. I will call the finite verb 1, and the non-finite verb selected by the finite verb will be labeled 2. (Any dependent verb selected by verb 2 will be labeled 3, and so on.) In Standard German, the only possible order for a two-verb complex in a subordinate clause is 2-1, i.e. the non-finite V followed by the finite Vas in (2b). (2) a.

. .. dass Klaus heute das Buch liest. that K. today the book reads that Klaus is reading the book todaY:

b.

. .. dass Klaus heute das Buch lesen will. that K. today the book read2 wants 1 that Klaus wants to read the book today:

c.

. .. dass Klaus heute das Buch wiU le.sen. that K. today the book wants 1 read2 '... that Klaus wants to read the book today:

The word-order asymmetry between main and subordinate clauses goes back to Old High German, the earliest recorded stage of the language: main clauses strongly tend toward V2, while subordinate clauses have later placement of the finite verb (Axel 2007: 6). Because verbs in subordinate clauses tend to occur late, earlier stages of German, like the modern language, exhibit verbal complexes. However, medieval varieties of German (like many contemporary continental West Germanic dialects) show considerable variation in word order within the VC. This variability can be illustrated with clusters of two verbs in Early New High German (ENHG), the language of the central and southern German-speaking areas from 1350 to 1650. In addition to the 2-1 order (3a), we find the opposite order 1-2 (3b). Furthermore, some material may occur between the verbs, resulting in an order that we willlabel1-x-2 (3c). Finally, independently of verb order, a constituent may be extra posed, i.e. placed to the right of the verbs (3d).

Unintroduced subordinate clauses, typically the complements of bridge verbs, display main-dause-like, verb-second word order:

2.

(i) Ich habe gesagt, Klaus will heute das Buch lesen. I have said K. wants 1 today the book read2

'I said Klaus wants to read the book today.'

Chapter 1. Introduction

(3) a.

das er in kainer sund verczweiffeln sol that he in no sin despair2 shall1 'that he shall not despair in any sin'

(Pillenreuth 161)

b. das der mensch alle sein lebttag nicht anders scholt thun that the person all his life.days nothing else should1 do2 'that man should do nothing else all the days of his life' (Pillenreuth 206) c.

das der mensche nicht scholt sein rew sparen ... that the person not should 1 his regret sav~ 'that one should not hold back his repentance .. .' (Pillenreuth 212)

d. Wye man fragen sol dy krancken how one ask2 shall 1 the sick 'how one should ask the sick'

(Pillenreuth 166)

Word order variation within the verbal complex is the subject of this study. Although VCs occur in both main and subordinate clauses, this study is limited to subordinate clauses. This is because in main clauses, the finite verb is always in the second position, so that VCs can occur only when there are three or more verbs in the clause, as in (ld). Given that the vast majority ofVCs contain only two verbs, especially in older texts, subordinate clauses provide the most fruitful ground for searching for complexes. Moreover, comparing VCs in main and subordinate clauses may be comparing unlike types, since main clauses have complexes of only non-finite verbs, while subordinate-clause verbal complexes usually contain a finite verb. This study investigates these VCs in the history of German, beginning with medieval German and continuing to some modern varieties of German. For medieval and early modern German, I will attempt to describe as accurately as possible the frequencies of particular word orders, the linguistic factors that favor some orders over others, and the diachronic and dialectal spread of these phenomena. For Modern German, I investigate similar phenomena, both in the standard language and in several dialects, resulting in a more complete understanding of synchronic grammar and diachronic developments. There are at least two reasons why the changes in the relative order of verbs that occur in the history of German are interesting, not only for philologists of German but also for those interested in language change and syntactic theory. First of all, many changes in word order have been argued to be related to other changes in the language, such as the loss of inflectional morphology. However, the inflectional system of German has remained remarkably stable throughout its attested history. Therefore, the diachronic developments within the German VC present an interesting case, because the change in word order appears to be motivated by factors other than morphological change.

3

4


The second reason that this phenomenon is worthy of diachronic treatment is because it has increasingly received synchronic attention. There has been a flurry of research in the last twenty years seeking the best analysis for the 1-2 and 2-1 orders in Dutch and varieties of German (see Chapter 5). Because written Standard Dutch shows variation in the VC, it can be adequately investigated using corpus studies, as in de Sutter et al. (2008). Contemporary dialects of German, on the other hand, are rarely written, thus a large-scale corpus study of variation in the VC is necessarily limited to older stages of the language. Earlier stages of German appear to allow even more variation than most of the modern dialects, thus providing additional data that the synchronic analyses should take into account. Moreover, investigating earlier stages of contemporary varieties may help explain some aspects of these orders that have not been understood. The rest of this chapter is organized as follows. Section 1.2 presents some previous scholarship on the VC in earlier stages of German. Section 1.3 outlines the organization of the rest of this book.

1.2

Previous scholarship on diachronic German verb order

1.2.1

The sentence-frame analysis

Despite an enormous body of diachronic research on the word order of German, the verbal complex has been relatively neglected. Most scholarship on syntactic change in German has concentrated on two aspects of German word order. First, the verb-second (V2) phenomenon of main clause word order has received substantial diachronic treatment e.g. by Lenerz (1985) and more recently by Axel (2007). Secondly; extraposition (both in main and subordinate clauses) has been a major topic in German grammar, being an object of research in many historical studies of word order in German. This state of aftrurs is, I believe, largely a result of the traditional sentenceframe analysis of German word order (e.g. Wollstein-Leisten et al. 1997:53-54). Under this model, illustrated in Table 1, the verbs in main clauses "frame, the rest of the sentence, with the finite verb forming the so-called "left bracket" and the non-finite verb the "right bracket': In subordinate clauses, the complementizer forms the "left bracket" and the verb duster the "right bracket': The positions before, between, and after the "brackets, are labeled the "pre-field~ "middle field~ and "post-field~ respectively.


Table 1. The sentence-frame analysis of German

maind

sub. d.

pre-field

L bracket

middle field

Rbracket

Klaus

will dass

heute das Buch Klaus heute das Buch

lesen. lesen will.

post-field

Modern German strictly limits what kinds of constituents can be extraposed, i.e. found in the post-field. However, as discussed above and illustrated in (3), earlier stages of German are much freer in this respect. Thus much of the scholarship on the history of German word order has focused on the development of the sentence frame, i.e. the decreasing frequency of extra position. The relative order of the verbs in the VC is often either ignored (because the variation is within the so-called right bracket) or as in Admoni (1990: 156), the 1-2 order is treated as an instance of extraposition (i.e. the non-finite verb is in the post-field). Lenerz (1985: 105) similarly downplays changes in the VC, noting that there is synchronic and diachronic variation but labeling this an example of"stylistic reorderings" that "do not concern the syntactic structures of German as such."

1.2.2

The verbal complex in MHG and ENHG

1.2.2.1

Behaghel (1932)

For many years, most discussion of Middle High German (MHG) grammar was limited to the poetic texts from the high point of courtly literature. Many earlier grammars such as Paul (e.g. 1966) hardly discuss the position of verbs at all. The exception is Behaghel's Deutsche Syntax, which discusses word order in older Germanic prose texts in great detail, from Old English through Old High German to ENHG. This includes a lengthy treatment of verbal complexes in subordinate clauses (1932: 86-118). Regarding two-verb complexes, Behaghel notes that both orders are possible and claims that the 2-1 order is increasingly preferred under Latin influence (1932: 87). In addition, a rhythmic principle influences verb order: the combination of a stressed word plus a verbal complex with the 1-2 order produces the sequence "Hochton, Unton, Hochton" ('stressed word, unstressed word, stressed word'), while the combination of a light word and the 2-1 order produces the sequence "Unton, Hochton, Unton" (1932:87). Finally. Behaghel finds that constructions with infinitives behave differently from those with participles, with the participial construction preferring the 2-1 order (1932: 105).

5

6


Prell (2001) Prell (2001) provides a very thorough study of MHG syntax, based primarily on prose rather than courtly poetry. He analyzes word order in main and subordinate clauses from twenty-four MHG prose texts from an early version of the Bochumer MHG corpus.3 Prelfs work forms the basis of the thoroughly revised syntax section ofPaufs (2007) MHG grammar. Of the 901 instances of two-verb complexes in subordinate clauses in Prell (2001:83), 28% have the 1-2 order. Prell finds several factors that have an effect on verb order. First, there is an increasing tendency for VCs in the 2-1 order to appear clause-finally: at the beginning of the 12th century, nearly half of the 2-1 complexes have extra position, while by the 14th less than a quarter do (2001: 84). On the other hand, VCs with the 1-2 order continue to occur with a high degree of extraposition throughout the period (2001:84), a trend that is found in several ENHG studies. Secondly, there is a strong effect of syntagm: 39% of modalinfinitive syntagms have the 1-2 order, while 24.7% of perfects and only 16.8% of passives do (2001:85). Thirdly, when the non-finite verb has a stressed separable prefix (SSP), the 1-2 order occurs more frequently than expected: there are 21 instances of the 2-1 order but 22 instances of 1-2 (2001: 87). With respectto complexes of three verbs, Prell ( 2001: 88) finds only 16 instances in subordinate clauses. Of these, 7 are in the 1-3-2 order, 5 are 3-1-2, and4 are 1-2-3. The majority of these examples involve a modal verb with the passive voice. Prell finds no instances of 3-2-1, the prevalent order in Modern Standard German. Prell concludes from this data that MHG, like Modern German, is a verb-final language with movement of the finite verb to COMP in main clauses (2001: 119). For Prell, the topological fields in the two stages of German are exactly the same, so that the structure for both is as in Table 1 above. The differences are that the postfield is a "strong" position in MHG (i.e. extraposition was frequent) but has weakened in the modern language, while the 2-1 order in the right bracket was "weak" in MHG (i.e. there was variation in the VC) but has become "strong (2001: 119). Prell suspects that these two changes may be related, with the result that Modern German subordinate clauses nearly always end in a finite verb (2001: 119).

1.2.2.2

Hammarstrom (1923) By Early New High German (ENHG), prose texts range across many dialects and genres, and thanks to an increasingly literate population, there are many texts such as personal letters that may more accurately reflect the speech of the time. Hammarstrom (1923) looks at verb placement in ENHG in both chancery

1.2.2.3

3· Although this is the same corpus from which I selected texts for my study (Chapter 2), only five texts were selected by both Prell (2001) and my study: Notker, Buch der Konige, Speculum ecdesiae, Schwarzwalder Predigten, and Muhlhauser Reichsrechtsbuch.


documents and popular literature. The earliest chancery documents, from the first half of the 14th century, have the 1-2 order 12-16% of the time. By the 16th century, this has fallen to 0-1%. The rates of 1-2 are higher for popular literature of the 16th century. ranging from 5-37%. In all time periods and text types, the rate of the 1-2 order is higher for the modal-infinitive construction than for the participial constructions.4 Hammarstri:im (1923: 150-162) also looks at subordinate clauses with three verbs, but the number of tokens is quite small. Generally speaking, he finds a preference for the 3-2-1 order in official documents but for 1-3-2 with the IPP construction. Hammarstrom's (1923: 199-200) conclusion is that the word order of the chancery documents was influenced by Latin (although he does not explicitly discuss what form that influence took), and the chancery documents, in turn, served as the model for the emerging written standard. However, Hammarstrom's selection of texts is very uneven, with the chancery documents much older than the literary texts and, as Maurer (1926: 84) points out, not distinguished by dialect.

Maurer (1926) Maurer (1926) places much more emphasis on dialect differences. He investigates a large number of both official documents and literary texts, from different dialect areas from 1300 to 1700. Unlike Hammarstrom (1923), Maurer looks only at perfect constructions. Although he does not give the results for perfects with haben 'have' in tabular form, his results for perfects with sein 'be' are repeated in Table 2. Based on these results, he concludes that the Alemannic areas (Switzerland, Alsace, Swabia, and Baden) have the highest rates of 1-2, while East Middle German (EMG) and North Bavarian (i.e. Nuremberg) have the lowest rates.

1.2.2.4

Table 2. Frequency of 1-2 order with sein perfects in Maurer (1926) Region

Percentage of 1-2 order (1400-1600)5

High Alemannic (Swiss) Low Alemannic (Swiss) Alsatian Swabian S. and M. Bavarian N. Bavarian (Nuremberg) Rhine Franconian EMGdialect EM G written language

30% 12% 50%-+ 33% 50% -+60% 8% 3% 10%-+ 30% 60% -+20% 5% -+0%

4· See Sapp (2006) for a more detailed discussioiL

s.

Adapted from Maurer (1926: 148). Maurer's table is more complicated than this, with some variation over time for some dialects.

7

8


Maurer (1926: 151) concludes that the EMG and North Bavarian dialects prefer the 2-1 order because they were already under the influence of the written language at this early stage. The tendency to have 2-1 in the written ENHG language, according to Maurer (1926: 123), is a direct result of Latin influence. Maurer (1926: 180) claims that late-medieval Latin school grammars prescribed the 2-1 word order for perfect passives like quod dictum est 'which was said', which then provided the model for ENHG word order.6 Maurer (1926: 165) finds that verb clusters translated from Latin may vary in word order when the Latin original is a simplex verb (4), but are 2-1 when the original is a cluster with 2-1 order (5). (4) a. b.

(5) a.

Latin original:

a quo lingua latina nomen accepit from REL language Latin name receives

translation:

von deme latinisch tunge ist genannt from REL Latin tongue is named 'for which the Latin language is named'

Latin original:

Qui postea dictus est Sedechias who afterward said is S.

b. translation:

geMten was Sedechias der dar rul who afterward called was S. 'who was later called Sedechias'

Furthermore, in Latin-based texts, Maurer (1926: 164) finds a higher frequency of the 2-1 order with perfects formed with sein than with haben, supporting his conclusion that the Latin perfect passive (which is formed with the verb esse 'to be') played an important role. Finally. in addition to Latin influence, Maurer (1926: 159) attributes a certain role to so-called rhythm: some combinations of verbs prefer specific orders, to maintain an alternation of stressed and unstressed syllables, and verbal prefixes affect the rhythmic structure as well. Maurer (1926: 161-162) also claims that falling intonation favors 2-1 order, so 2-1 is more likely to occur in a sentence-final subordinate clause than in a sentence-initial one. Besides the problems with Maurer's Latin hypothesis pointed out by Ebert (1981) (discussed in Section 1.2.2.6 below), there are several problems with the way his study is conducted. First of all, it does not look at modal plus infinitive constructions, which Hammarstrom and later studies show to have a higher rate of 1-2. Secondly, it mixes text types, so it is not clear how much of the variation is due to dialect and how much is due to genre. The EMG data in Table 2 are especially telling in this regard: that dialect shows both the highest and lowest

However, Burridge (1993: 115) and Ebert ( 1998: 116) demonstrate that medieval grammars of German do not make any such prescriptions.

6.


percentages of the 1-2 order, because it is divided into two categories. Thirdly, as Hard (1981: 26) points out, the number of tokens from each text is quite small. It is unclear whether the differences between Hamrnarstrom's and Maurer's results are due to the fact that Maurer takes into account texts from various dialects. or due to the problems with his study. Finally, like Hammarstrom (1923), Maurer (1926) does not give any indication of whether his results are statistically significant.

Hard (1981) Hard (1981) is a studyof17,073 clusters of three or more verbs, from 1450 to 1975. We will be concerned here only with his first period (1450-1580), from which he has 2,704 tokens. The details of his study will be discussed and compared to my results in Chapter 3, but here I will summarize his basic findings. First of all, Hard (1981:75) finds that, although ENHG has an increasing tendency to have the finite verb at the end of a two-verb cluster (2-1), it has the opposite tendency in three-verb clusters, with an increasing preference for the 1-3-2 order. Secondly, Hard (1981: 54) finds this trend mainly in the High German dialects, with Low German preferring the 3-2-1 order. Finally, after the ENHG period, Hard (1981: 174) demonstrates that the downward trend in the 3-2-1 order reverses for all constructions except the IPP, becoming the norm by the 20th century.

1.2.2.5

Ebert (1981) Ebert (1981) examines a variety of text types written by forty-four people from Nuremberg from the 14th to 16th centuries. By investigating texts from just one city, Ebert is able to pinpoint some of the social and stylistic variables that determine the variation between the 2-1 and 1-2 orders.7 He thus avoids Maurer's pitfall oflumping all types of texts together while dodging the problem of dialect differences. Moreover, Ebert uses a sophisticated statistical model (Generalized Linear Interactive Modeling) that controls the variables stress, time, style, class, education, and occupation. Ebert (1981: 219-228) finds that the following factors have an eftect on verb order. The stress of the word preceding the VC turns out to be a significant factor. When the preceding word is a noun (i.e. stressed according to Ebert), there is no clear preference for either order, but when the preceding word is a pronoun (assumed to be unstressed), the 2-1 order is strongly preferred. Time is also a significant variable, with a general decline in the 1-2 order as has been shown in other studies. The results for style are significant as well, with more formal letters having

1.2.2.6

7. Ebert (1981) treats only VCs where the two verbs are adjacent, thus excluding instances of 1-x-2.

9

10


higher rates of 2-1. Combining the factors class, education, and occupation gives the following hierarchy: administrators have the highest rates of 2-1, followed by merchants, artisans, students, nuns, and secular women. The type of syntagm is significant as well: werden + participle has the highest rate of 2-1, followed by haben + participle, modals/wet·den + infinitive, and finally sein + participle. Ebert considers other "rhythmic" factors, which have been given primary importance since Maurer (1926). First. Ebert (1981: 206) disproves Maurer's contention that sentence-final subordinate clauses show a higher rate of 2-1 and thus that the 2-1 order is linked to falling intonation. Secondly, Ebert (1981: 229) confirms Maurer's findings (1926: 159) that verbs with stressed separable prefixes are the least likely to show the 2-1 order. Thirdly, Ebert (1981: 208) finds that the alternation of stressed and unstressed syllables in the VC "may affect the choice" of orders, but the number of examples is too small for statistical analysis. Finally, Ebert (1981: 209) notices that the placement of the VC within the clause may have an effect on the ordering within the duster: when the verbs are in clause-final position, the 2-1 order is preferred. Based on these findings, Ebert concludes that Latin influence on the 2-1 order has been overstated. First of all, the preference for 2-1 is not tied to an individual's knowledge of Latin: merchants and artisans, who would not have had much schooling in Latin, show a higher rate of 2-1 than students, who received their education in Latin. Secondly, although Maurer claimed that Latin perfect passives like quod dictum est 'which was said' influenced the German passive das gesagt ist 'which is said: sein + participle is the syntagm which shows the lowest rate of 2-1. Ebert (1981: 231) maintains that this "soundly refutes" the Latin hypothesis. a According to Ebert (1981: 237), the tendency to show the 2-1 order more and more over time is an example of "change from above~ passed down from the chancery style, rather than an imitation ofLatin syntax.9

Ebert (1998) Ebert (1998) studies verb placement primarily in the language of teenagers from 16th-century Nuremberg. Ebert (1998:65-67) finds several factors that influence verb order in the writings of these individuals. First, as seen in earlier research,

1.2.2.7

8. Ebert's examples of sein + participle are all the perfect active use of sein; he excludes the sein passive (1981: 204). Other studies find that the sein passive has very high rates of 2-1, see the discussion of Ebert (1998) below and Chapters 2 and 3. This somewhat weakens Ebert's refutation of Latin hypothesis, since Maurer (1926) would predict high rates of2-1 with the sein passive, not necessarily the sein perfect. 9· However, as Burridge (1993: 117) points out, this still does not rule out indirect Latin influence, since Latin word order could have influenced chancery style.


different syntagms favor the 2-1 order to different degrees, following the hierarchy -wet·den passive > sein passive > haben perfect > infinitive constructions > sein perfect. Second, the stress or category of the preceding word affects verb order (as in Ebert 1981). Third, Ebert claims that the rhythmic structure of the non-finite verb affects word order, although essentially this is reducible to the type of prefix on the verb. Fourth, Ebert finds that for some individuals the lexeme of the nonfinite verb plays a role, but this is possibly reducible to prefix type and syntagm. The highest rates of 2-1 occur with the verbs vernehmen/vernommen 'perceive (inf./PPP)' and empfangen 'receive (inf./PPP): which have an unstressed prefix, while the lowest rates occur with the forms gewest 'been' and wet·denlworden 'become (inUPPP)'. Ge-west and worden occur only in the sein perfect, and -werden is an infinitive, thus these forms necessarily occur in the syntagms with the lowest rates of 2-1.10 Fifth, some individuals show lower rates of the 2-1 order when the finite verb is subjunctive. Finally, the 2-1 order increases over time, and the effect of some factors (the stress of the preceding word, the lexeme of the non-finite verb) diminishes over time. In addition, most of the individuals show increasing rates of 2-1 over their lifetimes. Ebert investigates not only the linguistic factors that affect verb order with these teenagers, but also the social circumstances that may have influenced them. Although individuals show higher rates of 2-1 with increased schooling (and young men more than young women), the above factors continue to favor the 1-2 order, leading Ebert (1998: 102) to conclude that children acquired the effect of those factors on VC order from the spoken language of the time, rather than learning them in school. Moreover, Ebert (1998: 116) finds no mention of VC order in school books or grammars of the time. Ebert's (1998: 154) examination of printed texts from Nuremberg reveals much higher rates of 2-1, with only syntagm having a significant efiect on verb order. Ebert (1998: 154) concludes from this that the teenagers' increasing tendency to use the 2-1 order was influenced by contemporary printed texts, but that the factors that favor the choice of word orders are genuine features of spoken ENHG.

Bies (1996) Another recent study of ENHG word order is Bies (1996). Bies compiles a corpus of over 5,000 clauses (900 of which are subordinate clauses) mainly from letters, including material from a wide range of dialects. Assuming that ENHG, like Modern Standard German, is an underlyingly SOY language, she investigates two aspects of ENHG syntax: extraposition and the 1-2 order.

1.2.2.8

On the other hand, gewesen 'bee.O: an alternative form to gewest, has the same syntagmatic distribution but a much higher frequency of 2-1.

10.

u

12


In the section on extraposition, Bies (1996: 66) points out that ENHG, like Modern Standard German, allows the extraposition of clauses, PPs, and heavy NPs. Unlike Modern Standard German, however, ENHG also allows the extraposition of non-heavy NPs. When heavy NP shift has been controlled for, Bies (1996: 39) finds that NPs extrapose to force a narrow focus interpretation. This focus-driven extraposition is lost by the Modern German period (1996:65). In her study of the 1-2 order, Bies (1996:61) confirms some of Ebert's (1981) results, establishing that the rate of 1-2 falls over time and agreeing with Ebert that ENHG is undergoing a change from above. Although Bies (1996:54) determines only a weak effect of social class, she does find a significant effect of style, with higher rates of 1-2 in "less monitored styles': She also generally confirms Ebert's hierarchy of syntagms. However, Bies disputes some other results of Ebert. First, Bies (1996:59) finds that the stress of the word preceding the VC has no significant impact on the order of verbs. She believes that this may be "due to a comparison of unlike objects across corpora" (1996:60).11 Secondly, Bies (1996:60) concludes that the placement of the VC within the clause is insignificant, i.e. extraposition has no effect on verb order. However, her data do show an effect of extraposition on word order, if one looks at different types of extraposition separately. In Bies' Table 17 (1996:61), the 1-2 order occurs more often than expected with an extraposed NP (37.1% versus the expected rate of27.3%), whereas extraposed PPs have the 1-2 order at 29.2%, close to the expected rate. Reifsnyder (2003) Reifsnyder (2003) is a study of the ENHG dialect of Augsburg, using a corpus with a wide variety of text types from the period 1500-1660. Her study covers orthographic, morphological, and syntactic variation in Augsburg, with the three syntactic variables being double negation, position of the verb within the clause (i.e. extraposition), and order within the VC. Only the results for the latter variable will be discussed here. There are three factors that affect verb order in her study. First, as in previous studies, Reifsnyder (2003: 229) finds a general decline of 1-2 over time. Secondly, Reifsnyder (2003: 224-225) confirms the importance of text type for verb order. The text types assumed to be removed from the spoken languagechronicles, city ordinances, letters from schoolmasters, official letters and reports, and printed pamphlets-have the lowest rates of the 1-2 order. Personal letters, personal narratives, and guild books are assumed to be most reflective of the 1.2.2.9

11. My own study, like Bies, finds no such effect across dialects in ENHG, but like Ebert I find the effect in particular dialects (see Chapter 3.2.3.2).


spoken language and indeed have the highest rates of 1-2. Thirdly, Reifsnyder (2003: 226-227) determines an effect of clause type: clauses starting with a whword or with the relative complementizer so have the highest rates of 2-1. For Reifsnyder (2003:245), the prevalence of the 2-1 order in official texts, as well as its increasing frequency over time, is the result of the adoption of a standard language ideology. Summary Because traditional philological study of MHG concentrated on the poetic language of the most famous literary works, the study of its prose syntax has only begun to receive much attention, for example by Prell (2001). There is a larger body of scholarship on ENHG syntax, although tew studies address word order within the verbal complex. The studies that do treat the relative order of verbs are all incomplete in some sense. Hammarstrom (1923) investigates a variety of texts over the whole ENHG period, but does not take dialect differences into account. Maurer (1926) covers a wide range of dialects but has tew tokens from each text. Moreover, these early studies do not use any kind of test for statistical significance. Hard (1981) uses a very large corpus of three-verb complexes but ignores those with only two verbs altogether. Ebert (1981, 1998) treats only two-verb clusters and only in writings from Nuremberg, and Reifsnyder (2003) does the same for Augsburg. Bies (1996) looks at two-verb dusters from a broader geographic distribution but does not give detailed information on dialect differences. Finally, only Ebert (1981, 1998) pays serious attention to the factors that favor particular orders. In attempting to establish linguistic factors that afiect word order in the VC, the following factors are discussed by more than one of these studies: syntagm type, rhythm, extra position, prefix type, dialect, sociolinguistic factors, and Latin influence. However, no study to date has tested all of these factors against each other using a large corpus and sophisticated statistical tools. The historical studies presented in Chapters 2 and 3 attempt to fill this gap in the research by including a wide variety of prose texts from many dialects, addressing clusters of two and three verbs, and statistically testing multiple linguistic and social factors.

1.2.2.10

1.2.3

The verbal complex in contemporary German

Like the previous scholarship on word order in ENHG, studies of contemporary German verb order have focused almost exclusively on extraposition. It is relatively rare to find a description of a German dialect that gives details on the relative order of the verbs. Even rarer is a discussion of the linguistic factors that influence VC order. (Those few studies that do address these phenomena will be

13

14


discussed in detail in Chapter 4 below.) Thus the chapter on Modern German seeks to broaden the synchronic description of these orders in present-day standard and dialectal German. Moreover, since the phenomena investigated are the same as for ENHG, the study of the contemporary language allows for a better understanding of the diachronic developments involved.

1.3

Organization

The remainder of this book is organized as follows. Chapter 2 treats verb order in subordinate clauses in Middle High German (1050-1350), based on a database compiled from the Bochumer Mittelhochdeutschkorpus. Chapter 3 examines similar phenomena in Early New High German (1350-1650) using the Bonner Frithneuhochdeutschkorpus. These chapters establish the frequencies of the various word orders, the linguistic factors that favor them, and their diachronic, dialectal, and sociolinguistic distributions. Chapter 4 covers Modern Standard German and several contemporary dialects of German. After reviewing existing descriptions of these varieties, the chapter presents three new studies: a questionnaire-based study of word order in Austrian dialects and Swabian, a more detailed questionnaire study of the effect of focus on these orders in Zurich German, and a Magnitude Estimation study of Standard German word order. This will allow a comparison with the data from MHG and ENHG. Note that Chapters 2 through 4 are primarily descriptive in nature, and I attempt to avoid terminology that assumes a particular syntactic analysis. Therefore, purely descriptive labels such as 1-2 and 1-x-2 are used instead of the more familiar terms Verb Raising and Verb Projection Raising, which imply an analysis like that of Haegeman (1992). Likewise, in these descriptive chapters the term extra position should not be taken to imply a specific analysis (such as rightward movement), but instead is used to simply describe the occurrence of a constituent to the right of the verbal complex. Chapter 5 builds on the descriptive and empirical work of the previous chapters to addresses the theoretical aspects of this research. The two primary goals of this chapter are to establish the best analysis of German clause structure and to determine the nature of the relationship between focus, prosody, and word order in the VC. At that point, I will make more explicit assumptions about the syntax of these word orders. Chapter 6 concludes the book, summarizing the most important findings and discussing their broader synchronic and diachronic implications for German word order.

CHAPTER 2

Factors influencing verb order in MHG

2.1

Introduction

2.1.1

Basic facts

Middle High German (MHG) is a term that encompasses central and southern dialects of German from approximately 1050 to 1350. MHG is often associated with the courtly poetry of the high Middle Ages in the German -speaking realms, and many previous descriptions of the language are based primarily on this narrow range of texts. This was the case, for example, with the first twenty-four editions of the standard Mittelhochdeutsche Grammatik edited by Hermann Paul et al., although the twenty-fifth edition of the Mittelhochdeutsche Grammatik (Paul 2007) and the new grammar by Klein et al. (to appear) are based on a much wider variety of text types, including both verse and prose texts. In the present study; only prose texts are analyzed, since rhyme and meter could have effects on word order in the verbal complex that might not be present in prose or in everyday speech. In MHG, subordinate clauses are generally distinguishable from main clauses by word order (Paul 2007: 396).1 This certainly holds true in the texts examined here: of the 1,133 subordinate clauses analyzed in this study, only 52 have the finite verb in the second position of the clause. Within the verbal complex, there are three possible orderings when the VC consists of two verbs. In the 2-1 order, the finite verb follows the non-finite verb (1), which is the dominant order according to this study and Paul (2007: 455). Also possible is the 1-2 order as in (2), in which the finite verb immediately precedes the non-finite verb. When the finite verb precedes, a non-verbal constituent may be found between the two verbs (1-x-2) as seen in (3).

As in Modern German, there are also subordinate clauses which are not introduced by a conjunction; these are verb-first or verb-second and often use the subjunctive mood (Paul 2007: 400-401 ). Because the object of this study is the clause-final VC, unintroduced subordinate clauses will not be treated further.

1.

16


(1) (wier daz volk verflvchet) daz got gefegent het. how he the people cursed, REL God blessed 2 had 1 '(How he cursed the people) whom God had blessed: (Buch der KlJnige 04va) (2) (alle die) den got gewalt ufi geriht hat verlihen. all those REL God power and rule has 1 granted2 '(all those) whom God has granted power and rule: (Buch der KiJnige OSra) (3) (daz dv vnf vergrebest) swaz wir vbelf heten an dir geta.n. that you us torgive REL we evil had1 to you done2 '( ... that you forgive us) whatever evil we have done to you:

(Buch der KlJnige 03va) A fourth logically possible order would be for the verbs to be in the order 2-1 but with an intervening constituent. This order is unattested in my MHG corpus.2 Thus MHG follows the generalization tor the West Germanic verb-raising languages discussed in Chapter 4.2.5, that no constituents can intervene between the verbs when in the left-governing (2-1) order. In addition to variation in the position of the verbs relative to each other, the verbs need not be in the clause-final position. Some constituents may follow the VC, and this extraposition can occur regardless of the relative order of the verbs (4a-b), even with the 1-x-2 construction (4c). (4) a.

Do iofeph gelebt het hvnd't iar vn zehn iar when ]. lived2 had1 hundred years and ten years 'When Joseph had lived 110 years .. : (Buch der KlJnige 03vb)

b.

alf and' din genoz hant getan. in Ierufalem ... as other your companions have 1 done 2 in ]. 'as other companions of yours have done in Jerusalem

c.

daz ein iclich man hie ... fal vrfdf habi in fimf huis ... that an every man here shall1 peace have2 in his house 'that every man here ... shall have peace in his house

(Buch der Konige 10va)

(Mahlhauser R. b2ra)

2. The only item that may come between the verbs in the 2-1 order is the negative particle ne-, which may be prefixed to the auxiliary:

(i)

(Neheine zungen ... ne fint) da iro predige gehoret tae fin. no tongues NEG are where their sermons heard2 NEG are 1 '(There are no languages) where their sermons are not heard.' (Notker 030ra)

This construction occurs in the two earliest texts in the database, Notker and Williram (which could be considered late OHG), and in four other texts. If one treats the particle as being part of verbal morphology. these are straightforward examples of the 2-1 order.

Chapter 2. Factors influencing verb order in MHG

VCs with three verbs are much rarer, but also display variation within the complex. Four of the six logically possible orders are attested in my MHG database: there is one instance of the 3-2-1 order (5), five of the 1-3-2 order (6), seven of the 1-2-3 order (7), and ten of3-1-2 (8):

(5) uuanta er geb6ran uutrdan uuolta uon armen u6rderon. wanted 1 of poor ancestors because he born3 be2 (Williram lOr) 'because he [Christ] wanted to be born of humble origins: (6) daz elliv werlt die er gefaphen Mte. folti verlorn werdfn. that all world REL he created had should 1 lost3 be2 'that the whole world that he had created should be lost: (Speculum eccL 08r) (7) wie er och Joelte wtie gemarterot an de (heiligen) cvce. how he also should 1 be2 martyred3 on the holy cross '... how he should also be martyred on the holy cross:

(Schwarzw. Pred. 009v) (8) daz daz chint befnitin folti werdin. that the child circumcised3 should1 b~ '... so that the child should be circumcised .. .'

(Speculum eccL llv)

These four orders are also reported for MHG by Paul (2007:456). In addition, extraposition can occur with these VCs, and there are three instances where a constituent intervenes within a 1-3-2 cluster: (9) daz fiv niene mahti mit handen gervoret wtiin ... that she not could 1 with hands touched3 be2 'that it could not be touched with hands .. : (Zaricher Pr. lOSva) The other two logically possible orders for three-verb clusters, 2-1-3 and 2-3-1, are unattested in my MHG database. (10) *uuanta er uutrdan 2 uuolta1 geb6ran3 (11) *uuantaer uutrdan 2 geb6ran3 uuolta 1 There are no instances of a VC with four verbs in the database, which should come as no surprise given the scarcity of three-verb complexes.

2.1.2

The database

The MHG database analyzed in this study consists of 1,133 subordinate clauses. These clauses are taken from thirteen prose texts of the Bochumer Mittelhochdeutschkorpus (BoMiKo). Although the BoMiKo has many more prose texts than this, this selection of texts attempts to represent one text from each century and

17

18


dialect in the corpus. In some cases, there is no text in the corpus for a given century/dialect, and in other cases, the text for a century/dialect was not analyzed because it is a gloss and thus might not be a reliable source for analyzing syntax. Where a century/dialect has more than one appropriate text and one of those was a sermon, I chose the sermon so that the database would have texts most reflective of spoken language. Table I. Texts in my MHG database3 Dialect

1070-1100

1150-1200

1250-1300

Bavarian

Wiener Notker

Physilogus Speculum ecdesiae

Buch der KCJnige Stadtbuch Augsburg

Ztlricher Predigten

Schwarzwalder Predigten Satzungsbuch Ntlrnberg

Swabian, ind W. Bavarian Alemannic East Franconian

Williram

(1302-1315) Middle Franconian Hessian Thuringian

Frankfurter Predigtfragmente

DieLilie Mitteldeutsche Predigten Mtlhlhauser Reichsrechtsbuch

Clauses were selected for the database if they are introduced by a subordinating conjunction or relative pronoun and contain at least one finite and one non-finite verb (except for embedded infinitival clauses).4 Of the 1,133 subordinate clauses in the database, 1,109 clauses have two verbs, 23 have three verbs, and one was excluded from analysis because it contained two finite verbs (like example (15) in Chapter 3). In the case of shorter texts, the entire text was analyzed, and from longer texts the first 100 or so relevant clauses were analyzed. The number of clauses taken from each text ranges from 28 to 107, with an average of 87.2 clauses per text. Each clause in the database was coded for the following variables: verb order (the dependent variable), syntagm type, constituent preceding the VC, constituent intervening in the VC, constituent following the VC, focus type (new vs. contrastive), focused constituent, whether an intervening constituent is part of the focus,

3· For complete philological information about these texts and others in the corpus, see Wegera (2000). 4· Because of the ambiguity between relative and demonstrative pronouns in German, apparently relative clauses with V2 are ambiguous with main clauses (Paul 2007: 396). Likewise, wande 'because' may be either a coordinating or subordinating conjunction (Paul2007: 396). Clauses introduced by a (relative) pronoun or wande were included in the database, but any V2 clauses were tagged as such.


whether an extraposed constituent is part of the focus, prefix type, scrambled object, and verb second.s Each text in the database was tagged for century, dialect, genre, and the sex, occupation, and education of the author. 2.1.3 Analysis with GoldVarb X

The analyses were conducted using the statistics package GoldVarb X (Sankoff et al. 2005). This program was originally developed for sociolinguistic studies, and allows the researcher to determine the extent of the effect of several independent variables (linguistic and sociolinguistic factors) on a dependent variable. GoldVa.rb can be a useful tool in historical linguistics as well, since time can be treated as an independent variable. Sets of independent variables are called factor groups, and each value of the variable is called a factor. For example, syntagm is a factor group with the factors passive, perfect, modal-infinitive, etc. A powerful feature of GoldVarb is the ability to recode the factors during the analysis. If the researcher is not certain whether a particular distinction will be significant, e.g. the difference between present perfect and pluperfect, they can be initially tagged as separate factors. Then during the course of the analysis, one can recode, combining these specific factors into more general ones, and test which combination has the most significant effect. There are three statistical outputs of GoldVarb X that will be utilized in this chapter. The first is statistical significance. Note that this represents the statistical significance of the entire factor group, e.g. syntagm, but does not indicate which factor or factors have the significant effect For that, one has to look at the second output, the factor weight. The factor weight is expressed as a probability between 0 and 1, with 0.5 indicating no effect. The further the factor weight is from 0.5, the greater that factor's effect is on the dependent variable. Finally, GoldVarb has a step-up/step-down analysis, which tests all possible combinations of factor groups to determine which combination results in the most statistically significant model. For details on GoldVarb, see Tagliamonte (2006).

2.1.4

Organization of this c.hapter

This chapter is organized as follows. Section 2.2 presents the results of the analysis of the MHG database for complexes of two verbs, and Section 2.3 presents the results for complexes of three verbs. This chapter is concluded in Section 2.4. 5· Two additional factors, clause type and VP coordination, had no effect and will not be discussed further here.

19

20


2.2

Complexes of two verbs in MHG

2.2.1

Introduction

My MHG database contains 1,109 subordinate clauses with one finite and one non-finite verb. Ofthose, 320 clauses (28.9%) have the 1-2 (or the 1-x-2) order.6, 7 Thus a rate of 1-2 higher than 28.9% in a given context indicates that the order is favored in that context, which should be expressed as a factor weight greater than 0.5 by GoldVarb. In Section 2.2.2, I discuss the linguistic factors that affect verb order: syntagm type, the category and weight of the word preceding the VC, extraposition, focus, and verbal prefixes. Section 2.2.3 discusses variation in the rate of the 1-2 order over time, across dialects, and sociolinguistically. Of the factors that have a significant aftect on verb order, six were selected by GoldVarb's step-up/step-down function as the most significant combination of factors: syntagm, constituent before the VC, focus, dialect, genre, and occupation.

2.2.2

Phonological, morphological, and syntactic factors

2.2.2.1

Syntagm type

Many previous studies of historical and contemporary West Germanic dialects have shown a correlation between VC order and syntagm type. In modern varieties that have some word -order freedom, there is a general tendency to have 2-1 in constructions with participles and 1-2 in those involving infinitives (Wurmbrand 2006: 237; Chapter 4 below). As shown in Table 2 below, MHG follows this trend. The syntagms are listed in order of the rate of 1-2, from least frequent to most frequent. The passives are the least likely to have the 1-2 order, occurring less often that the total rate of 1-2 and with low factor weights. The perfects have the 1-2 order very close to the expected rate of 28.7%. The modal-infinitive syntagm somewhat favors 1-2 at 35.5%, while other syntagms with an infinitive favor that order even more strongly. (In addition, there are a handful of instances of the progressive, a syntagm which will be discussed in more detail in Chapter 3.2.3.1.) The effect of syntagm is statistically significant and was selected in GoldVarb's step-up/step-down function as one of the most significant factor groups.

6.

Throughout this section, these two orders are treated together unless indicated otherwise.

7. This rate of the 1-2 order is quite similar to that found by Prell, in whose corpus 27.9% of the two-verb dusters are 1-2 (2001: 84, cited inPaul2007:455).


Table 2. Effect of syntagm on 1-2 order Syntagm

2-1

1-2

Factorwt.

sein passive progressive werden passive perfect with sein perfect with haben modal +in£ other V + inf. future

127 (90.1%) 5 (83.3%) 145 (82.9%) 52 (71.2%) 181 (70.7%) 272 (64.5%) 7 (25.9%) 0 (0%)

14 (9.9%) 1 (16.7%) 30 (17.1 %) 21 (28.8%) 75 (29.3%) 150 (35.5%) 20 (74.1%) 6 (100%)

0.233 0.355 0.362 0.526 0.532 0.602 0.887

Total

789 (71.3%)

317 (28.7%)

n/a

p < 0.001 These numbers are very similar to Prell's (2001) findings for MHG (see Chapter 1.2.2.2).8 The hierarchy from my MHG corpus (12a) can be compared to those found by Prell (12b): ( 12) a. b.

sein pass. > werden pass. > sein perf. > haben pert~ > modals sein pass. > werden pass. > pedtd > modals (Prel12001: 85, cited in Paul 2007: 455-456)

In both my MHG study and Prell (2001), the statal passive with sein has the 1-2 order considerably less often than the werden passive does. Other than that, the hierarchies are similar to that in ENHG, all showing the general pattern of passive> perfect> modal. Note also that this is likely a genuine effect of syntagm independently of the effect of the auxiliary or the type of non-finite verb (participle vs. infinitive): both the sein passive and sein perfect involve the auxiliary sein plus a participle, yet the two syntagms show quite difierent word order preferences. 2.2.2.2

Word/phrase preceding the verbal complex

Ebert (1981: 206) finds that the distinction noun vs. pronoun influences verb order in ENHG texts from Nuremberg. In this section, we examine whether the preceding word, not only noun vs. pronoun but other categories as well, has any effect in MHG. First of all, let us see what effect the previous word may have on the VC in terms of category. As shown in Table 3, a preceding non-pronominal noun phrase ("noun'') favors the 1-2 order at 36.4%, well above the expect rate of 1-2 (28.7%).

8. Recall from Chapter 1 that my database overlaps with Prell's by five texts.

21

22


Quantified NPs behave similarly. On the other hand, pronouns occur with the 1-2 order only 22% of the time, somewhat less than expected.9 Table 3. Effect of the category of the preceding word on 1-2 order Class of preceding word

2-1

1-2

Factorwt.

noun pronoun quantified NP adjective adverb prepositional phrase stranded preposition clause (infinitival or finite) nothing precedes Totallo

185 (63.6%) 202 (78.0%) 27 (64.3%) 15 (53.6%) 162 (76.8%) 170 (74.2%) 7 (77.8%) 2 (66.6%) 18 (64.3%)

106 (36.4%) 57 (22.0%) 15 (35.7%) 13 (46.4%) 49 (23.2%) 59 (25.8%) 2 (22.2%) 1 (33.3%) 10 (35.7%)

0.593 0.418 0.586 0.688 0.435 0.469 0.421 0.560 0.586

788 (71.6%)

312 (28.4%)

p < 0.001 In addition to non-pronominal NPs, adjectives strongly favor the 1-2 order. Both of the remaining frequently occurring categories, adverbs and prepositional phrases, have rates that are close to or slightly below the expected rate of 1-2. (Possible reasons for the effect of stranded prepositions and clauses with nothing preceding the VC will be discussed in more detail in Chapter 3.2.3.2.) This raises the question of why the category of the preceding word should have any effect on verb order. Ebert (1981: 207), who found similar effects in ENHG, attributed the differing effect of nouns and pronouns on verb order to phonological stress. This is based on an earlier claim by Behaghel (1932:IV, 87), that the combination of a stressed word such as a noun plus a VC with the 1-2 order produces the sequence heavy word, light word, heavy word, while the combination of a light word plus verbs in the 2-1 order produces the alternation light, heavy, light According to this claim, other combinations result in disfavored sequences of two stressed words or two unstressed words. To test this claim, I recoded pronouns, short adverbs, the negator, and da-compounds as unstressed and other categories as stressed. As can be seen in Table 4, the rate of 1-2 is somewhat higher

9· Throughout this section, "NP" includes both nouns and pronouns. When pronouns are meant to be excluded, the terms "noun" or "non-pronominal NP" will be used No distinction will be made between NP and DP.

The total for this table is lower because the analysis excludes six cases where the preceding "word" was a verbal prefix.

10.


than expected following a stressed word (34%) and somewhat lower following an unstressed word, confirming Behaghel's claim. Table 4. Effect of the stress of the preceding word on 1-2 order Stress of preceding word

2-1

1-2

Factorwt.

stressed unstressed Totaln

420 (66.0%)

216 (34.0%)

0.570

351 (79.4%)

91 (20.6%)

0.401

771 (71.5%)

307 (28.7%)

p < 0.001 Alternatively, the differing effect of nouns versus pronouns on verb order could be due to a difterence in definiteness. As seen in Table 5, when the preceding NP is indefinite (including QPs withjeder, kein, aile), the rate of 1-2 is much higher than expected at 42.4%, but when it is definite (including pronouns and proper names), the rate is close to the expected rate. Thus, definiteness could partially explain the differing behavior of nouns versus pronouns. Table 5. Effect of definiteness of the preceding NP on 1- 2 order Preceding NP

2-1

1-2

Factorwt.

indefinites (including QPs) definites

57 (57.696)

42 (42.496)

0.634

357 (72.4%)

136 (27.6%)

0.473

Total

414 (69.996)

178 (30.196)

p = 0.005 However, there are two arguments against the possibility that definiteness accounts for the noun/pronoun distinction found in Table 3. First, note that the results in Table 5, although statistically significant, have a lower p score than those in Table 3 or 4. Secondly, in GoldVarb's step-up/step-down analysis, the factor group preceding constituent is selected as one of the most significant factor groups when coded by part of speech or heaviness, but not when coded by definiteness. The log likelihood (an output of GoldVarb indicating the best model fit for a combination of factor groups) was worst with this factor group coded by definiteness (-494.350), better when coded by part of speech (-484.571 ), and best when coded by weight (-480.682). Therefore, definiteness is the least likely reason for the effect of the preceding constituent on the order of verbs within the VC, and heaviness is the most likely reason, confirming Behaghefs ( 1932) hypothesis.

The totals for this table are lower because the analysis excludes cases where no word precedes the verbal complex.

11.

23

24


2.2.2.3

Extraposition

As mentioned above, although the VC tends to be clause-final in MHG, some constituent (or more rarely, more than one constituent) may appear be extraposed, i.e., located to the right of the VC (Paul2007: 458). Based on the high frequency of extra position, Lehman claims that medieval German is an SVO language, unlike the SOV structure of Modern Standard German (1971: 19). Because the 2-1 order is frequently associated with SOV typology (Greenberg 1963: 85), one might expect non-SOV clauses (i.e. those with extraposition) to have the 1-2 order. As can be seen in Table 6, there is indeed such a correlation.12 This is a statistically significant factor group, although it was not selected as one of the most significant groups in the step-up/step-down analysis. Table 6. Effect of extraposition on 1- 2 order Extraposition

2-1

1-2

Factorwt.

extraposed argument NP/PP extraposed adjunct PP nothing extraposed

34 (55.7%) 109 (69.4%) 602 (73.4%)

27 (44.3%) 48 (30.6%) 218 (26.6%)

0.670 0.530 0.481

Total

745 (71.8%)

293 (28.2%)

p = 0.014 When there is no extraposed constituent, the rate of the 1-2 order is 26.6%, slightly less than the expected rate. Likewise, when an adjunct PP is extraposed, the rate of the 1-2 order is slightly greater than expected at 30.6%; the factor weight, being close to 0.5, indicates that extraposed adjunct PPs have very little effect on the order of the verbs. On the other hand, an extraposed argument (including both NPs and PPs) favors the 1-2 order at 44.3%, considerably higher than expected, with the factor weight indicating a strong effect. Similar results have been found for ENHG (see Chapter 3). Finally, of the 1,106 clauses in the MHG database, 218 have some extraposed constituent, yielding a rate of extraposition of 19.7%. This is consistent with the data for MHG reported in Prell (2001), who also finds higher rates of 1-2 with extra position. 2.2.2.4

Focus

A constituent is considered to be focused if it introduces new information into the discourse or is contrastive. Note that focus is not identical to stress: 12. When extraposition is broken down into the category of the extraposed constituent (NPs versus pronouns, etc.) the result was not statistically significant (p = 0.063). Note that Table 6 excludes the following low frequency types: multiple extraposed constituents, extraposed adjective, extraposed comparison, and extraposed second conjunct


although a focused constituent may be stressed, a given stress pattern can often be associated with different focus interpretations. This can be illustrated using a Modern German example, in which a single sentence with the same stress (indicated with all capitals) can answer very different questions: in (13a) only the object is new to the discourse and thus focused, while in (13b) the entire clause is focused: (13) a.

What did Karl give to the child yesterday? Gestern hat Karl dem Kind [poe das BUCH) geschenkt. yesterday has K. the child the book given b. What happened yesterday? Gestern hat [Foe Karl dem Kind das BUCH geschenkt]. 'Karl gave the book to the child yesterday.' (Stechow & Sternefeld 1988:461)

In order to test the effect of focus on word order in the VC, each clause in the database was examined within its section (or within one or two pages, if the text was not divided into sections) in order to determine what elements of the clause are contrastive or discourse-new. The clause was then tagged as having focus on the entire clause (14a), the subject, the object (14b), the verb, the VP (14c), or some other constituent. (14) a.

[Foe

daz di fvginden kint durch vnfirs h'ren willen our Lord's sake that the nursing children tor

worden irjlagfn.] were 1 killed 2 '(the day) that the infants were killed because of our Lord:

(Mitteldt. Pred. alra) b.

die !Foe finin willin ] heton getan. his will had1 done 2 'who had done his will' daz ir [Foe fin' gutdete muzit gen{zen ]. that you his good.deeds might1 enjoy2 'that you might enjoy his good deeds'

REL

c.

(Zaricher Pr. 109rb)

(Mitteldt. Pred. a1ra)

Moreover, the focused constituent was characterized as representing newinformation focus or contrastive focus. If there were no new or contrastive constituents, the clause was tagged as old information. Table 7 presents the results for focus, regardless of which constituent is focused. Both new and contrastive focus favor the 1-2 order at 33.7% and 31%, respectively, slightly higher than the expected rate of 28.7%, while clauses with all old information disfavor 1-2. However, unlike ENHG (see Chapter 3.2.3.4) where the favoring effect of contrastive focus was stronger than that of new

25

26


information, in MHG there is little difference between the two types of focus.B (Although statistically significant, this factor group was not selected as one of the most significant in GoldVarb's step-up/step-down analysis, probably because it largely overlaps with the next group.) Table 7. Effect of focus on 1-2 order (general) Focus

2-1

1-2

Factorwt.

contrastive focus new information focus old information

49 (69.0%) 490 (66.3%) 250 (84.5%)

22 (31.0%) 249 (33.7%) 46 (15.5%)

0.539 0.569 0.324

Total

789 (71.3%)

317 (28.7%)

p < 0.001 Having established that focus in general has an effect on the order within the VC, the next factor group is the particular constituent that is focused. As seen in the table below, focus on an object is the most favorable context for the 1-2 order at 50.7%, well above the expected rate of 28.7%. The other favoring focus context is focus on the entire clause, which has the 1-2 order 38.3% of the time. Two other contexts, VP focus and focus on some other constituent (e.g. an adjunct) have the 1-2 order somewhat more frequently than expected, but neither has a factor weight much above 0.5, indicating at most a slight favoring effect. Finally, verb focus and subject focus disfavor the 1-2 order. This factor group was selected as one of the most significant in the step-up/step down analysis. Table 8. Effect of focus on 1-2 order (specific constituents) Focus constituent

2-1

1-2

Factorwt.

object focus clausal focus VP focus focus on another constituent verb focus subject focus nothing (old information)

34 (49.3%) 71 (61.7%) 158 (65.6%) 74 (67.3%) 185 (73.4%) 17 (73.9%) 250 (84.5%)

35 (50.7%) 44 (38.3%) 83 (34.4%) 36 (32.7%) 67 (26.6%) 6 (26.1%) 46 (15.5%)

0.684 0.566 0.525 0.506 0.433 0.426 0.470

Total

789 (71.3%)

317 (28.7%)

p < 0.001

13. New and contrastive focus appear to behave similarly for a given focus constituent as well, although there are too few examples of each type of contrastively focused constituent to draw any firm conclusions.


To corroborate the effect of object focus on the 1-2 order, two related factors were tested. The first, the definiteness of an NP, was presented above in Table 5, which demonstrates a strong and significant correlation between indefinite NPs and the 1-2 order. Because indefinite NPs tend to be discourse-new and thus focused, while definite NPs generally represent old information, the correlation between definiteness and verb order may support the finding that focus effects verb order. The second factor that correlates to focus is scrambling; however, there were too few examples of this in the database to draw any conclusions. Finally, let us examine the interaction between focus on an object, the position of that object relative to the VC, and verb order. First let us examine cases where a constituent intervenes in the VC, i.e. the 1-x-2 order. There are 35 instances of this construction in the MHG database. Only two of these, as in (15), involve narrow focus on only the intervening constituent, so clearly, this is not a typical position for narrow focus. Perhaps crucially, both of these are ambiguous between a relative clause with 1-x-2 order and a main clause, as the relative pronoun is homophonous with the demonstrative. In 15 instances, the intervening constituent and some other constituent (usually the lexical verb) are focused, thus this construction may be used to focus the VP or a larger projection such as the entire clause. In (16), the focus, indicated in brackets, is on all of the clause except the subject pronoun. In the remaining 18 instances. there is VP or clausal focus but the intervening constituent is not part of the focus, as in (17), where the pronoun 'you' is mentioned in the matrix clause.

Jolte [poe ane vlecken] fin (15) (efn lfip flahen.) daz a lamb slaughter REL/DEM should 1 without spots be 2 'slaughter a lamb, which should be without spots .. .' 'slaughter a lamb. That should be without spots .. .' (Mitteldt. Pred. b4ra) (16) daz fi [Foe fnden oitem }olden efn lap flahen]. that they in the Easter should 1 a lamb slaughter 2 '... that they should slaughter a lamb at Easter (Mitteldt. Pred. b4ra) wir [poe vbeli] heten an dir geta.n. (17) ... swaz whatever we evil had on you done '( ... we ask you to forgive) whatever evil we have done to you.'

(Buch der KiJnige 03va) Thus it appears that the position between the two parts of the VC has two functions: the constituent there may be part of wide focus or it may be defocused. That position, however, is very unlikely to be a focus position.14 14. No statistical analysis is possible, since intervening constituents are necessarily found with only one word-order type, 1-x-2.

27

28


The relationship between focus and extraposition is less clear. Extraposed constituents may be narrowly focused (18), part of wider (usually VP) focus (19), or defocused (20). (18)

(div bihte.) die wir tun fuln [Foe untern priftern]. the confession REL we do 2 should 1 our priests '(the confession) which we should make to our priests:

(19)

die

(Mitteldt. Pred. b4va-b) dir tievil hate gebvndin mit den fvndon]. REL the devil had 1 bound2 with the sins whom the devil had bound by sin: [Foe

(Zitricher Fred. b4ra)

(20) alf fi wol [Foe verdinet] hat an dem armen manne. as she well deserved2 has 1 on the poor man 'as she [Jezebel] well deserved because of the poor

man:

(Buch der KCln ige 07vb) As seen in Table 9 below, 48 of the extraposed constituents are defocused,lS 59 are narrowly focused, and 111 are part of a wider focus projection. Neither the extra position of defocused nor of narrowly focused constituents has an effect on verb order, since the rates of 1-2 are similar in these contexts to the expected rate and the factor weights are close to 0.5. However, when an extraposed constituent is part of wider focus, the 1-2 order is favored at 40.5%, well above the expected frequency. Thus, the conclusion from Section 2.2.2.3 above that extra position of an argument has an effect on verb order must be revised somewhat: instead, it appears that extraposition of an argument that is part ofVP focus favors 1-2. Table 9. Effect of the focus of an e.x:traposed constituent on 1-2 order Extraposed constituent

2-1

1-2

Factorwt.

not focused narrowly focused part ofVP focus nothing extraposed

34 (70.8%) 43 (72.9%) 66 (59.5%) 602 (73.4%)

14 (29.2%) 16 (27.1%) 45 (40.5%) 218 (26.6%)

0.513 0.488 0.636 0.481

Total

745 (71.8%)

293 (28.2%)

p = 0.034

15. Of the defocused extraposed constituents, most are adjunct PPs, leaving only 26 cases where a non-focused argument is extraposed.. Recall from Section 2.2.2.3 above that adjunct PPs extrapose at higher rates than other types of constituents in MHG and also in Modern German.


Summing up this section, it seems that focus on an object plays some role in determining verb order. Narrowly focused, unscrambled, and indefinite NPs favor the 1-2 order, while wide focus favors the 1-x-2 orders and the 1-2 order with extraposition. However, because focus is associated with stress, it is difficult to determine whether these are effects of focus pet· se or of the phonological manifestation of focus. In Chapter 5, we will see that stress turns out to be the best way of accounting for the relationship between focus and word order. 2.2.2.5

Prefix type

The continental West Germanic languages have two kinds of verbal prefixes: unstressed prefixes such as ge- (21a), and stressed, separable prefixes (also known as converbs or verbal particles), most of which are derived from prepositions (21b ). (21) a. b.

daz got gefegent het REL God blessed2 had 1 daz im Ionathaf wa.z ab gega.nge. that him Jonathan was 1 away-gone 2 'that J. had departed him'

(cf. (1a) above)

(Buch der KiJnige 12rb)

The stressed prefixes tend to be adjacent to the verb when the verb is in the clausefinal position (usually appearing as prefixes, hence the name) but may also be stranded in clause final position when the head verb is finite and in the V2 position. Prell (2001:87) finds that stressed separable prefixes favor the 1-2 order, and Ebert (1981: 207; 1998 passim) has similar results from ENH G. Therefore, the present study tests whether the effect in MHG is significant. As can be seen in Table 10, my MHG database, like Prell's, shows a favoring eftect of stressed-prefix verbs on the 1-2 order, which occurs 40.4% of the time in that context, considerably higher than the total rate of 28.7%. In addition, verbs with no prefix also favor the 1-2 order at 38.1 %. Thus the most important distinction is between verbs with a stressed or no prefix, which favor 1-2, and verbs with an unstressed prefix, which slightly disfavor 1-2. Table 10. Effect of prefix type on 1-2 order Prefix type

2-1

1-2

Factorwt.

stressed no prefix unstressed Total p < 0.001

31 (59.6%) 203 (61.9%) 555 (76.4%)

21 (40.4%) 125 (38.1%) 171 (23.6%)

0.620 0.547 0.470

789 (71.3%)

317 (28.7%)

29

30


This state of affairs suggests that it is not prefixation per se that affects word order, but rather the syntagms involved. Past participles typically have the unstressed prefixge-, while infinitives may have no prefix, a stressed prefix, or an unstressed prefix. Therefore, the apparent preference for the 2-1 order with unstressed prefixes may be because many of these are syntagms with participles, which favor the 2-1 order (Section 2.2.2.1 above). In fact, of the 726 verbs with an unstressed prefix, 396 have the prefix ge-. The interaction between syntagm type and prefix type may explain why only one of these two factors, syntagm, was selected during the step-up/step-down analysis as one of the most significant factor groups. This kind of interaction can be tested using the cross-tabulation function of GoldVarb. The cross-tabulation of prefix type and syntagm type, given in Table 11, suggests that the factors prefix type and syntagm type overlap to a great extent but are still independent of each other. For all three factors in the factor group prefix type, the frequency of 1-2 is higher with infinitival constructions (54%, 39%, and 33% respectively) than with participial constructions (27%, 24%, and 21%, respectively). Likewise, for both past participles and infinitives, the frequency of 1-2 is highest with stressed prefixes (27% and 54% respectively), lower with no prefix (24% and 39%), and lowest with unstressed prefixes (21% and 33%). Table 11. Cross-tab of prefix type with syntagm type Prefix type stressed

word order 1-2

7 (27%)

14 (54%)

21 (40%)

no prefix

2-1

35 (76%)

166 (61%)

201 (63%)

1-2

11 (24%)

107 (39%)

118 (37%)

unstressed

2-1

451 (79%)

101 (67%)

552 (77%)

1-2

122 (21%)

49 (33%)

171 (23%)

2-1

505 (78%)

279 (62%)

784 (72%)

1-2

140 (22%)

170 (38%)

310 (28%)

Total

2-1

participial syn. 19 (73%)

infinitival syn. 12 (46%)

Total 31 (60%)

In sum, the presence of a stressed separable prefix on the non-finite verb favors the 1-2 word order, and this effect is somewhat independent of syntagm type. Having discussed four internal linguistic factors that favor the 1-2 order in MHG, in the next section I will present the variation in subordinate clause word order over time, across dialects, and sociolinguistically.


2.2.3

Diachronic, dialectal, and sociolinguistic variation

2.2.3.1

Date

As will be seen in the next chapter, the 1-2 order decreases over the course of the ENHG period from approximately 35% in the 14th century to just 11% in the 16th, disappearing by Modern Standard German. However, it appears that the decrease in ENHG begins in that period. As can be seen in Table 12, at both the beginning and end of the MHG period, rate of the 1-2 order is roughly the same as at the beginning of ENHG: about 32% and 34%, respectively. Table 12. Effect of time period on 1-2 order Century

2-1

1-2

Factorwt.

1070-1100 1150-1200 1250-130016

139 (67.8%) 253 (83.8%) 397 (66.3%)

66 (32.2%) 49 (16.2%) 202 (33.7%)

0.560 0.451 0.504

Total p < 0.001

789 (71.3%)

317 (28.7%)

On the other hand, there is an inexplicable drop in the rate of 1-2 in the middle of the MHG period. It is difficult to say whether this dip represents any linguistic reality or is merely a result of the four texts selected for this century. In any case, although the factor group is statistically significant, it was not selected as one of the most significant factor groups by GoldVarb's step-up/step-down analysis. Having seen that the 1-2 order (in the general sense that the finite verb precedes the non-finite verb) remains mostly stable throughout MHG, let us examine the difference between the 1-2 order in the narrow sense (the finite verb immediately precedes the non-finite verb) and the 1-x-2 order. As shown in Table 13, both the 1-2 order in the stricter sense and the 1-x-2 order show the same trend as in the table above, with the eleventh and thirteenth centuries roughly the same, but a drop in the twelfth century.l7 However, the frequencies of 1-x-2 and

16. Recall that the text representing this time period for Nuremberg is slightly later, having been written between 1302 and 1315.

17. Gold Varb X is not equipped for multi-variate analyses, i.e., although it can calculate the

raw nwnbers and percentages for several values of the dependent variable, it can statistically test only two values. Thus the data for 1-x-2 order are given without factor weights or statistical significance.

31

32


the 1-2 order in the strict sense remain very stable relative to each other, with the 1-x-2 order making up roughly 10-12% of the cases where the finite precedes the non-finite verb. Table 13. Effect of time on 1-2 and 1-x-2 orders Century

2-1

1-2

1-x-2

1070-1100 1150-1200 1250-1300

139 (67.896) 253 (83.8%) 397 (66.3%)

58 (28.396) 44 (14.6%) 180 (30.1%)

8 (3.9%) 5 (1.7%) 22 (3.7%)

Total

789 (71.3%)

282 (25.5%)

35 (3.2%)

Recall from Section 2.1.2 above that some subordinate clauses with the verb in the second position (but introduced by an unambiguously subordinating conjunction) were coded as potentially V2 in the database. This makes it possible to exclude such clauses in a GoldVat·b analysis. Excluding potential verb-second clauses is especially important in the case of 1-x-2 clauses, to rule outthe possibility that they involve main clause word order rather than true, unambiguous 1-x-2 (with more than one constituent preceding the finite verb). Excluding these subordinate clauses with possibly main clause word order results in a smaller number of 1-2 clauses (especially 1-x-2 clauses), as can be seen by comparing Table 14 with Table 13. The remaining 1-2 and 1-x-2 subordinate clauses must then be treated as true verbal complexes. Table 14. Effect of time on 1-2 and 1-x-2 orders, V2 clauses removed Century

2-1

1-2

1-x-2

1070-1100 1150-1200 1250-1300

139 (71.6%) 253 (86.3%) 397 (66.3%)

53 (27.3%) 37 (12.4%) 160 (28.0%)

2 (1.0%) 3 (1.0%) 14 (2.4%)

Total

789 (74.6%)

250 (23.6%)

19 (1.8%)

2.2.3.2

l)ialect

Having examined the relative stability of subordinate-clause verb order over time, let us now consider dialectal variation in MHG. As seen in Table 15, the frequency of the 1-2 order (in the wider sense, i.e. including 1-x-2) varies greatly by dialect, ranging from about 11% in the text from Thuringia up to 81.7% in one text from the Black Forest.


Table 15. Effect of dialect on 1-2 order Dialect

2-1

1-2

Factorwt

Bavaria Swabia (incl. W. Bavaria) Alemannic: SwissJs Alemannic: Black Forest is

Hesse Thuringia Is

177 (68.1 %) 175 (87.5%) 86 (86.9%) 19 (18.3%) 133 (80.6%) 24 (53.3%) 84 (64.1 %) 91 (89.2%)

83 (31.9%) 25 (12.5%) 13 (13.1 %) 85 (81.7%) 32 (19.4%) 21 (46.7%) 47 (35.9%) 11 (10.8%)

0.494 0.336 0.262 0.894 0.447 0.830 0.525

Total

789 (71.3%)

317 (28.7%)

E. Franconian CologneJs

0.434

p < 0.001

There is no clear pattern here that might correspond to traditional dialect divisions: for example, the two Alemannic texts behave very differently. There is something of a geographic pattern when plotted onto a map of the German-speaking area, as in Figure 1. The three dialects with relatively high rates of 1-2 (and factor weights above 0.5) cluster along the Rhine: Black Forest, Hesse, and Cologne. However, combining these three dialects into a western group and all others into an eastern group results in a worse model fit. Moreover, note that many of these dialects are represented by just one text, so the apparent dialectal variation may be reducible to differences between individual texts.

Figure 1. Frequency of the 1-2 order by dialect

18. Each of these dialects is represented by only one text.

33

34


Nor does one see a clear pattern when looking at the rate of 1-2 across the dialects and by century. As can be seen in Table 16, the rate of the 1-2 order increases from the 11th to the 13th century in the Bavarian texts but decreases in the East Franconian texts. With this kind of variation, it is difficult to draw any conclusions about the dialectal or diachronic distribution of 1-2 in MHG. Whatever the reason for this variation, the factor group is clearly an important one, having been selected as one of the most significant in the step-up/stepdown analysis. Table 16. Effect of dialect and century on 1-2 order

Dialect

1070-1100

1150-1200

1250-1300

Total

Bavaria

24.5%

11.7% 18.0%

41.2% 8.8%

31.9% 12.5% 13.1%

81.7% 11.1%

81.7% 19.4% 46.7%

Swabia Swiss Black Forest E. Franc. Cologne

13.1% 39.8%

Hesse Thuringia

37.0%

46.7% 35.6% 10.8%

35.9% 10.8%

2.2.3.3 Sociolinguistic factors

In several studies ofENHG syntax (Hammarstrom 1923; Ebert 1981 and 1998; Bies 1996; and Reifsnyder 2003), subordinate-clause word order varies greatly by sociolinguistic factors and text type. In my MHG database, clauses were coded for four sociolinguistic factors: the genre of the text and the sex, education, and occupation of the author. However, none of the texts in the database are written by women, and the level of education is only known for one of them. Therefore, the only factor groups that may yield any relevant results are occupation and genre. Unfortunately for the factor group occupation, the only two occupations tagged in the database are clerics and unknown. Still, these two groups have quite different word-order preferences, as seen in Table 17. The clerics have the 1-2 order somewhat more often than expected (35.7%), and the texts by authors of unknown occupation have that order somewhat less than expected (19%). Although this effect is not a very strong one, as evidenced by the fact that the factor weight in both cases is very dose to 0.5, it is statistically significant and was even selected as one of the most important factor groups in the step-up/ step-down analysis.


Table 17. Effect of occupation on 1-2 order Occupation (number of texts)

2-1

1-2

Factorwt.

clerics (7) unknown(6)

410 (64.6%) 379 (81.0%)

228 (35.7%) 89 (19.0%)

0.508 0.490

Total

789 (71.3%)

317 (28.7%)

p < 0.001 Genre was also selected by step-up/step-down as one of the most significant factor groups. All texts in the database were tagged as one of four types: chronicle, chancery document, sermon, and other religious text, and the effect of these text types on word order is presented in Table 18. The most formal text type in the corpus, chancery documents, has the lowest frequency of 1-2 at 9.8%, well below the expected rate of 1-2. Sermons, on the other hand, favor the 1-2 order at 37.4%, while other types of religious texts have roughly the same rate of 1-2 as the expected rate. The one chronicle in the database strongly favors 1-2, but it is unclear whether that is a property of chronicles per se or an idiosyncratic feature of the particular text. One might tentatively conclude from these data that since sermons are the genre closest to the spoken language and strongly favor 1-2, the 1-2 order is more prevalent in spoken MHG, while the most formal written language tends to shun that order. Table 18. Effect of genre on 1-2 order Genre (number of texts)

2-1

chancery docwnent (3) sermon(5) other religious text (4) chronicle (1)

240 271 231 47

Total

1-2

Factorwt.

(90.2%) (62.6%) (70.6%) (58.8%)

26 (9.8%) 162 (37.4%) 96 (29.4%) 33 (41.2%)

0.406 0.557 0.433 0.754

789 (71.3%)

317 (28.7%)

p < 0.001 Finally, given the fact that sermons and other religious texts were mostly written by clerics, let us examine whether the effects of genre and occupation are independent of each other. This is presented in the cross-tabulation of genre and occupation in Table 19.19 Of the texts written by clerics, sermons have a higher rate of the 1-2 order at 37% than other religious texts written by clerics at 32%. Of the texts by persons of unknown occupation, religious texts have a higher rate

19. We will set aside the one instance of a chronicle.

35

36


of 1-2 (25%) than chancery documents (10%). Religious texts by clerics use the 1-2 order more frequently (32%) than the religious texts by non-clerics (25%). Thus the two factors are indeed somewhat independent of each other and have a cumulative effect on verb order. Table 19. Cross-tab of genre with occupation Genre

word order

clerics

occup. unknown

sermons

2-1

271 (6396)

1-2

162 (3796)

0 (096)

162 (3796)

other religious

2-1

139 (6896)

92 (7596)

231 (7196)

1-2

66 (3296) 0 (096)

30 (2596) 240 (9096)

96 (2996) 240 (9096)

0 (096)

Total 271 (6396)

chancery

2-1 1-2

0 (096)

26 (1096)

26 (1096)

Total

2-1

410 (6496)

332 (8696)

742 (7296)

1-2

228 (3696)

56 (1496)

284 (2896)

2.2.4 Conclusion In this section, I have discussed the results of my analysis of word order variation in MHG complexes of two verbs, based on a database of 13 texts. Several factors were determined to affect verb order: syntagm type, the category and weight of the word preceding the VC, extra position, focus, verbal prefixes, dialect, occupation of the author, and genre. Of these, the following combination was found to be the most significant factors by GoldVarb's step-up/step-down analysis: syntagm type, preceding word, focus, dialect, occupation, and genre. The next section will test all of these factors on subordinate clauses with clusters of three verbs.

2.3

Complexes of three verbs in MHG

2.3.1 Introduction Perhaps unsurprisingly, there are few examples in my MHG database ofVCs containing three verbs. One reason for this is that MHG still has robust use of the preterit tense, so in complex syntagms such as the past tense of the passive, MHG has two verbs (the preterit passive) where later stages of the language tend to have three verbs (the present perfect passive). In my database of 1,133 clauses with a VC, just 26 have three verbs and none have four verbs. Given the small number


of tokens, this section will report the numbers, percentages, and statistical significance, but factor weights will only be reported in the few instances where the factor has a statistically significant effect. Recall from Section 2.1.1 that the following orders for three-verb complexes are attested in MHG (repeated here for the reader's convenience): the 3-2-1 order (5), the 1-3-2 order (6), the 1-2-3 order (7), and 3-1-2 (8): (5) uuanta er geb6ran uuerda.n u.uolta uon armen u6rderon. wanted1 of poor ancestors because he bom3 be2 (Williram lOr) 'because he [Christ] wanted to be born of humble origins: (6) daz elliv werlt die er gefaphen Mte.folti verlorn werdfn. that all word REL he created had should 1 lost3 b~ 'that the whole world that he had created should be lost: (Speculum eccL 08r) (7) wie er och Joelte wi:ie gemarterot an de (heiligen) cvce. how he also should 1 bez martyred3 on the holy cross '... how he should also be martyred on the holy cross: (Schwarzw. Fred. 009v)

Jolti werdin. (8) daz daz chint befnitin that the child circumcised3 should1 be 2 '... so that the child should be circumcised .. .'

(Speculum eccL llv)

Moreover, recall that a constituent may break up the VC with the order 1-3-2, resulting in what I call1-x-3-2 (9). Throughout this section, no distinction will be made between 1-3-2 proper and 1-x-3-2. Having four possible word orders presents a number of methodological challenges. The first has to do with the fact that GoldVarb X allows only binomial analyses, and having four word orders requires a multinomial analysis. To get around this limitation, I ran three separate analyses for each factor group.2o I tested the order 1-3-2 against the combined orders 3-2-1, 1-2-3, and 3-1-2. Then I tested the 1-2-3 order against the other three combined, etc. This allowed me to test for factor weights and statistical significance for each dependent variable. The second problem is the low number of tokens. There are only 26 clauses with three-verb complexes to begin with, and this is compounded by having four dependent variables. Thus in this section, statistical significance is not achieved in any of the analyses, unless otherwise noted. Other recodes were attempted in many cases, but in no case did recoding result in increased significance, thus such recodes will not be discussed further.

20.

Since only one clause has the 3-2-1 order, that order is not tested for signlli.cance at all

37

38


The third problem is one of presentation, thus I will briefly explain how to interpret the tables for this section. Taking Table 20 as an example, there are two occurrences of the order 1-2-3 with the modal+ perfect syntagm. These two instances make up 66.7% of the three total instances of that syntagm. To determine whether this is a favoring factor for the 1-2-3 order, this percentage should be compared notto the percentages of the other word orders, hutto the expected rate of 1-2-3 in the database, which is 26.9%. Thus it appears that the modal + perfect syntagm strongly favors the 1-2-3 order, since the order appears more frequently in that context (66.7%) than its overall expected rate (26.9%). Of course, to determine whether that really is a strong favoring effect, the factor weight needs to be consulted. In this section, factor weights will be reported in the body of the text when relevant. This section is organized as follows. In Section 2.3.2, I examine whether the linguistic factors that affect two-verb clusters also affect three-verb clusters. Section 2.3.3 discusses the effect of date, dialect, and sociolinguistic factors on threeverb clusters in MHG.

2.3.2

Phonological, morphological, and syntactic factors

2.3.2.1

Syntagm type

In Section 2.2.2.1 above, I showed that syntagm type plays a major role in subordinate clauses with complexes of two verbs. This is also the case for complexes of three verbs, as shown in Table 20. Several examples of modal + passive are given in (5)-(8) above, and the other three syntagms are modal+ perfect (22), modal+ modal+ infinitive (23), and other, such as modal+ causative (24):

han (22) den er von im genomen Jolte REL he from him tak.en3 should1 have 2 'which he should have taken from him:

(Stadtbuch Aug. 20va)

(23) Swenne wir alliz daz nit uollfn grundfn mvgfn if we all that (... REL ••• ) not want1 prove3 can2 'If we do not want to be able to prove all (that we have heard)'

(Mitteldt. Fred. b4va) (24) daz er daz hus nider Jol haizzen Jlahen ... that he the house down shall1 order2 strike3 'that he shall order the house struck down .. :

(Sta.dtbuchAug. 19ra)


Table 20. Effect of syntagm on three-verb complexes Syntagm

3-2-1

1-3-2

1-2-3

3-1-2

modal + passive modal + perfect21 modal + modal + in£ other Total

1 (5.3%) 0 (0%) 0 (0%) 0 (0%)

6(31.6%) 0(0%) 1 (100%) 1 (33.3%)

4 (21.1%) 2 (66.7%) 0 (0%) 1 (33.3%)

8 (42.1%) 1 (33.3%) 0 (0%) 1 (33.3%)

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

Although not statistically significant, we can perhaps draw some preliminary conclusions from the data above. First, note that as in Paul (2007:456), modal verbs governing an auxiliary plus participle (i.e. the first two syntagms in the table above) make up the overwhelming bulk of the three-verb complexes in my database.22 Secondly, as in Paul (2007:456), the modal+ passive syntagm can occur with any of the four word orders. Thirdly, there is only one instance of the 3-2-1 order in my database, and Paul concurs that the order is extremely rare in MHG (2007:456). As a final note, I find no examples of the infinitivus pm participio (IPP) or Ersatzinjiniti11 construction,23 although Paul does find two instances (2007: 456). The three orders other than 3-2-1 are about equally frequent in my study as in Prell's (2001:88). In each of these three orders (1-3-2, 1-2-3, and 3-1-2), finite verb (V1) precedes the verb it modifies (V2). This seems to be the opposite preference from that in complexes of two verbs, where 71% of the VCs have the finite verb in the final position (2-1). However, note that in almost all of the three-verb complexes in the database, the finite verb is a modal which tend to favor the 1-2 order in MHG. Thus it is an open question whether the preference for a finite verb to occur early in the three-verb complex is a property of three-verb complexes per se or due to the fact that modal verbs tend to occur before the verbs they select. 2.3.2.2

Word/phrase preceding the verbal complex

In Section 2.2.2.2 above, it was demonstrated that in VCs with two verbs, thecategory and phonological weight of the word preceding the VC influenced the order of the verbs. When a non-pronominal NP or a QP preceded, the rate of 1-2 was

21. Actually, two of these are the modal+ perfect as in (22), and the other is a verb of perception plus perfect: uuanot ergezzen haben 'believe to have forgotten' (Notker 014ra). 22.

Recall that the syntax section of Paul (2007) is written by Prell, largely based on Prell (200 1).

23. See Chapter 3 for a description and examples of this phenomenon in ENHG.

39

40


higher than expected, and that order was lower than expected when preceded by a pronoun. When recoded according to phonological weight, preceding stressed words were found to favor the 1-2 order while unstressed ones disfavored it For complexes of three verbs, this factor group is not significant for any of the word orders, neither when coded by part of speech as shown in Table 21, nor when coded by stress as in Table 22. The numbers are very small and no clear pattern can be seen in the distribution of the factors across the various word-order types. Table 21. Effect of the preceding category on three-verb complexes Cat. of prec. word

3-2-1

1-3-2

1-2-3

noun pronoun adverb prepositional phrase nothing precedes

0 (0%) 1 (33.3%) 0 (0%) 0 (0%) 0 (0%)

4 (33.3%) 0 (0%) 2 (66.7%) 0 (0%) 2 (50.0%)

4 (33.3%) 1 (33.3%) 1 (33.3%) 0 (0%) 1 (25.0%)

3-1-2 4 (33.3%) 1 (33.3%) 0 (0%) 4 (100%) 1 (25.0%)

Total

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

Table 22. Effect of the preceding stress on three-verb complexes Stress of prec. word

3-2-1

1-3-2

1-2-3

3-1-2

stressed unstressed nothing precedes

0(0%) 1 (16.7%) 0(0%)

5 (27.8%) 2 (33.3%) 1 (50.0%)

5 (27.8%) 2 (33.3%) 0 (0%)

8 (44.4%) 1 (16.7%) 1 (50.096)

Total

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

2.3.2.3 Extraposition Like the previous factors, NP/PP extraposition has an effect on VCs with two verbs but not on those with three verbs. As seen in Table 23, clauses with nothing extra posed are about evenly distributed across the three common word orders, as are clauses with extraposed adjunct PPs. Moreover, none of the tests of the word orders yielded statistically significant results. Table 23. Effect of extraposition on three-verb complexes Extraposition

3-2-1

1-3-2

1-2-3

3-1-2

extraposed argument nothing extraposed extrapos. adjunct PP

1 (33.3%) 0 (0%) 0 (0%)

0(0%) 6 (37.5%) 2 (28.6%)

0 (0%) 5 (31.2%) 2 (28.6%)

2 (66.7%) 5 (31.296) 3 (42.9%)

Total

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.596)


2.3.2.4 Focus In Section 2.2.2.4 above, I showed that focus favors the 1-2 order, while old information disfavors it. As seen in Table 24, there is no dear efiect of focus on VCs with three verbs. Clauses with new/contrastive focus are about evenly distributed across the three common word orders, and the results are not significant for any order. Table 24. Effect of focus in three-verb complexes Focus

3-2-1

1-3-2

1-2-3

new I contrastive old information Total

1 (4.8%) 0 (0%)

8 (38.1%) 0 (0%)

5 (23.8%) 2 (40.0%)

7 (33.3%) 3 (60.0%)

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

3-1-2

2.3.2·5 Prefix tvpe As with other factor groups, prefix type does not have a discernable effect on the choice of verb orders with three verbs. As shown in Table 25, the number of clauses representing any given word order and prefix type is quite small, and this factor is not significant for any order. Moreover, this factor group largely overlaps with syntagm., as the numbers for unstressed in this factor group are nearly identical to those for modal+ passive in Table 20. Finally, note that the only two instances of a VC with a stressed prefix are in the 1-2-3 order. Table 25. Effect of prefix type on three-verb complexes Prefix: type

3-2-1

1-3-2

1-2-3

Stressed No prefix Unstressed Total

0 (0%) 0 (0%) 1 (4.8%)

0 (0%) 1 (33.3%) 7 (33.3%)

2 (100%) 1 (33.3%) 4 (19.0%)

0 (0%) 1 (33.3%) 9 (42.9%)

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

3-1-2

2.3.3 Diachronic, dialectal, and sociolinguistic variation 2.3.3·1 Date Turning now to date, we find first of all that there are only two instances of threeverb complexes in the 11th century, as seen in Table 26. From the 12th to the 13th centuries, the 1-3-2 order changes little, the 3-1-2 order becomes less frequent, and the 1-2-3 order greatly increases in frequency. Only this last development is statistically significant. changing from a factor weight of 0.198 in the 12th century to 0.837 in the 13th.

41

42


Table 26. Effect of time on three -verb complexes Century

3-2-1

1-3-2

1-2-3

3-1-2

1070-1100 1150-1200 1250-130024

1 (50.0%) 0 (0%) 0 (0%)

1 (50.0%) 4 (33.3%) 3 (25.0%)

0 (0%) 1 (8.3%) 6 (50.0%)

0 (0%) 7 (58.3%) 3 (25.0%)

Total

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

s ignificance2s

n/a

p =0.568

p = 0.012

p = 0.191

2.3.3.2 Dialed

As might be expected given the rather random dialectal distribution of twoverb complexes (Section 2.2.3.2 above), the distribution of three-verb complexes is even less clear. Given the small number of tokens per dialect, for none of the four word orders is this factor group statistically significant, as shown in Table 27. However, a few remarks may still be made. First, note that the 3-1-2 order is the most frequent order in Bavaria, Swabia, and Switzerland, dialects that favor that order both in ENHG (see Chapter 3) and the modern period (see Chapter 4). Secondly, note that the 1-2-3 order is the only one found in the text from the Black Forest, an Alemannic dialect, while Swiss, also an Alemannic dialect, does not have any instances of that order. This may indicate that these differences are due to idiosyncrasies of the representative texts, rather than true dialectal variation. Table 27. Effect of dialect on three-verb complexes

Dialect

3-2-1

1-3-2

1-2-3

3-1-2

Bavaria Swabia Black Forest26 E. Franconian Cologne26 Hesse Thuringia26

1 (50.0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%)

0 (0%) 3 (30.0%) 1 (20.0%) 0 (0%) 1 (50.0%) 0 (0%) 3 (100%) 0 (0%)

0 (0%) 2 (20.0%) 0 (0%) 3 (100%) 1 (50.0%) 1 (100%) 0 (0%) 0 (0%)

1 (50.0%) 5 (50.0%) 4 (80.0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%) 0 (0%)

Total

1 (3.8%)

8 (30.8%)

7 (26.9%)

10 (38.5%)

Swiss26

24. Recall that the text representing this time period fur Nuremberg is slightly later (1302-1315). 25. Statistical significance for each order was calculated by combining the 11th and 12th centuries. 26. Each of these dialects is represented by only one text


2.3.3.3 Sociolinguistic factors In Section 2.2.3.3 above, I demonstrated that occupation and genre have an effect on word order in two-verb complexes, with clerics and religious texts favoring the 1-2 order. However, with three-verb complexes, no such preferences emerge, probably due to the very small number of tokens for each of the four word orders. As seen in Table 28, sex/occupation does not have a significant effect on any verb order. However, the 1-3-2 order is disfavored in texts by unknown authors (factor weight= 0.286). The other orders are more evenly distributed, and the factor weights are close to 0.5, indicating little favoring effect.

Table 28. Effect of sex/occupation on three-verb complexes Sex/occupation27 males/clerics (7) unknown(6) Total

3-2-1 1 (5.3%) 0 (0%) 1 (3.8%)

1-3-2 7 (36.8%) 1 (14.3%) 8 (30.8%)

1-2-3 4 (21.1%) 3 (42.9%) 7 (26.9%)

3-1-2 7 (36.8%) 3 (42.9%) 10 (38.5%)

As seen in Table 29, genre has no statistically significant effect on any word order either. Chancery documents do slightly disfavor 1-3-2 (factor weight= 0.351) while favoring 1-2-3 somewhat (factor weight= 0.635). Again, the number of tokens is quite small, and it would be hasty to draw any conclusions from this. Table 29. Effect of genre on three-verb complexes Genre chancery doc. (3) sermon(5) other religious (4) chronicle (1) Total

3-2-1 0 (0%) 0 (0%) 1 (33.3%) 0 (0%) 1 (3.8%)

1-3-2 1 (20.0%) 6 (35.3%) 1 (33.3%) 0 (0%) 8 (30.8%)

1-2-3 2 (40.0%) 4 (23.5%) 1 (33.3%) 0 (0%) 7 (26.9%)

3-1-2 2 (40.0%) 7 (41.2%) 0 (0%) 1 (100%) 10 (38.5%)

2.3.4 Conclusion

In this section, I have attempted to establish whether the factors that affect twoverb clusters also affect clusters of three verbs. Because of the small number of tokens for each word order, few significant effects could be found. However, two preliminary conclusions may perhaps be drawn. First, the 3-2-1 order is rare, while the other three orders are about equally frequent. Secondly, the 1-2-3 order increases greatly from the 12th to the 13th century. 27. Recall that all non-anonymous texts in the database are by male clerics, thus these two categories are identical

43

44


2.4

Conclusion

2.4.1 Summary of findings This chapter reports on a study of verbal complexes in subordinate clauses in thirteen MHG texts. Although few conclusions can be made about VCs containing three verbs, there are a number of interesting results for two-verb complexes. First of all, syntagm has a significant effect on verb order, with participial constructions favoring the 2-1 order and infinitival constructions favoring 1-2 (Section 2.2.2.1). Secondly, the category and phonological weight of the word preceding the VC influences the order within the complex (Section 2.2.2.2). Thirdly, focus plays an interesting role in the choice of verb orders, with narrow object focus favoring the 1-2 order, wide focus favoring 1-x-2 and 1-2 with extraposition, and old information favoring 2-1 (Section 2.2.2.4). Fourthly, the effect of dialect is significant (2.2.3.2). Finally, genre and occupation have significant effects, with clerics and religious texts independently favoring the 1-2 order (Section 2.2.3.3). The above factors in combination were found to have the most significant effect on verb order in GoldVarb's step-up/step-down function. In addition, extraposition (2.2.2.3) and prefix type (2.2.2.5) influence the relative order of the verbs, although these effects are largely the result of interactions with other factors, namely focus and syntagm. Time is also a significant fuctor, although rather than a steady development, we find that the 1-2 order dipped in the 12th century before rising again in the 13th (Section 2.2.3.1). The next chapter will test these same effects on a corpus of Early New High German, with largely the same results.

2.4.2 The combined effect of the favoring factors In this chapter, I have discussed a number offactors that favor the 1-2 order in MHG. However, none of these alone can account for the high frequency of the 1-2 order in the database. However, when all of these factors are considered together, they account for the vast majority of the occurrences of that order. As seen in Table 30, just 13 of the 317 clauses with the 1-2 order, or 4.1%, have none of the favoring factors. Thus, although no individual factor forces the 1-2 order, the relatively high frequency of 1-2 in these texts can be understood as the result of the cumulative favoring effect of a number of factors.


Table 30. All favoring factors for 1- 228 Favoring factor

1-(x)-2

Syntagm with infinitive Preceding stressed word Separable prefix verb Focus Extraposition No scrambling None ofthe favoring factors Total

170 216 21 271 28 9 13

317

Throughout this chapter, we have seen that several prosodic factors have an effect on word order within the verbal complex: a preceding stressed word, a stressed prefix, and focus (which is expressed through prosody). Table 31lists only those factors that are arguably related to prosody and focus. With only the prosodic/ focus factors considered, the majority of 1-2 clauses are still accounted for. Only 21, or 6.6%, of the clauses are not accounted for by a prosodic factor. Table 31. Prosody-related favoring factors for 1-2 Favoring factor

1-(x)-2

Preceding stressed word Separable prefix verb Focus Extraposition No scrambling None of the prosodic factors Total

216

21 271

28 9

21 317

In Chapter 3, we will see that these factors have largely the same effect in ENHG as in MHG. Chapter 4 will investigate variation in verb order in some varieties of Modern German: we will see that syntagm type and prosodic factors continue to be the major factors impacting verb order today.

28. Many clauses are represented in this table more than once, since a clause may have two or

more of the favoring factors.

45

CHAPTER3

Factors influencing verb order in ENHG

3.1

Introduction

3.1.1 Basic facts Early New High German (ENHG) is the stage of the language spoken approximately from 1350 to 1650. This period of European history represents the transition from the Middle Ages to the modern world, and is characterized (among many other developments) by the rise of the city, the invention of the printing press, and the Protestant reformation. It is a time of increasing mass literacy, including literacy in vernacular languages, and as a result this period sees a dramatic increase in the type and number of texts produced. ENHG is also an age of linguistic transition: over those three hundred years the language gradually takes on many of the phonological morphological, and syntactic features that characterize Modern Standard German. In traditional scholarship on the history of German, ENHG was not considered a separate stage of the language at all (Schmidt 1996: 27), so that texts before about 1500 were often characterized as corrupt or dialectal late MHG, while those after 1500 were seen as the New High German language forged by Martin Luther. Most recent reference works, however, recognize ENHG as a separate stage, albeit one that is characterized by much dialectal and diachronic variation. We shall see in this chapter, in fact, that subordinate-clause verb order at the beginning of ENHG is quite similar to that of MHG, while many of the later texts approximate the syntax of modern German. As in prior stages of the language, the 2-1 order (i.e. the Modern Standard German order in which the finite verb follows the non-finite verb), is a frequent word order in ENHG: (1) daz er chain todsund nymmer mer tuen wolt

that he no death-sin never more do2 would 1 'that he would never again commit a deadly sin'

(Pillenreuth 160)

In addition to the 2-1 order, there are also clauses in which the finite verb precedes the non-finite verb, either directly, resulting in the 1-2 order (2), or separated by some constituent, in the 1-x-2 order (3):

48


(2) das der mensch aile sein lebttag nicht anders scholt thun that the person all his life.days nothing else should 1 do 2 'that man should do nothing else all the days of his life' (Pillenreuth 206) (3) das der mensche nicht scholt sein rew sparen an das todpett that the person not should 1 his regret save2 on the deathbed 'that one should not hold back his repentance on his deathbed'

(Pillenreuth 212) The fourth logically possible order, in which the verbs are in the order 2-1 with an intervening constituent, is virtually unattested in my ENHG corpus.! In addition to variation within the verbal complex. its position within the clause mayvary. As in MHG, constituents maybe extraposed whether the VC is in the 2-1 or 1-2 order, even with the 1-x-2 order (3). Thus ENHG largely continues the word-order possibilities found in MHG. In complexes of three verbs, the four word orders of MHG are also found in ENHG (albeit in much greater numbers): the 3-2-1 order (4) with each verb selecting its complement verb to the left. the 1-2-3 order (5) with each verb selecting its complement verb to the right, and two mixed orders, 1-3-2 (6) and 3-1-2 (7): (4) das so darvorgesetzt ist in fragweis verstanden werden soll that REL before.set is in question understood3 be 2 should 1 'that what is set before should be understood as a question' (Eunuchus 14) (5) so er dan den menschen nicht ha.t mugen vberwinden when he then the person not has 1 can2 overcome3 'when he has not been able to overcome the person' (Pillenreuth 158) (6) als er des tages scholt begraben werden as he the day should1 buried3 be 2 'when he should be buried on that day'

(Pillenreuth 212)

t. There are only three examples in which this order does occur, all of which are from a single text, the Oxforder Benediktinerregel from the 14th century. Perhaps crucially, in all three examples, the intervening constituent is the negator, with a negative prefix on the finite verb: (i) daz sie mit missetroste virloren nit inwerde that she with false.comfort 1ost2 not NEG-be1 'that she not be lost because of false comfort' (Benediktinerregel16)

This does not necessaril ypose a challenge to the observation that no constituents can intervene between the verbs in their base order (Bobaljik 2004: 139), if one assumes that in these cases nit is part of the verbal inflection. Note that this is quite unlike the cases in MHG, where only the particle ne- but never the free-standing negator was found in this position.

Chapter 3. Factors influencing verb order in ENHG

er ... getan solt haben he done 3 should1 have2 'that he should have done'

(7) dy

REL

(Pillenreuth 159)

The other two logically possible orders, 2-1-3 and 2-3-1, are unattested in my ENHG database, as in MHG.2 In addition, with two of these orders, 1-3-2 and 1-2-3, some constituent may break up the VC. In all eleven instances involving the order 1-3-2 (some of which are V2 and thus ambiguous with main clauses), the intervening constituent appears just before the non -finite verb of which it is the complement, i.e. V 3 (8). The fact that no constituent may break up the 3-2-1 order, or appear between V 3 and in the 1-3-2 order, follows the generalization discussed above, that no constituent can intervene between verbs in the left-governing order. On the other hand, in the four instances involving the 1-2-3 order, the intervening constituent appears once between verbs 1 and 2 (9), once between verbs 2 and 3 (10), and once between all three verbs (11):

v2

(8) die sich niemals haben Schande vbergehen lassen REL REFL never have 1 disgrace come.upon3 let2 'who have never let disgrace come upon themselves'

(Bange Chronik 7r)

(9) wie si ain wuenderlich ding hiet vor dem ofen gesehen ligen how she a miraculous thing has 1 before the oven seen 2 lie3 'how she saw a miraculous thing lying in front of the overi

(Denkwitrdigk. 17) (10) das er seinen Sohn fuer vns hat lassen ein Schlachtopffer werden that he his son for us has 1 let2 a sacrifice.victim become3 'that he has let his son become a sacrificial victim for us' (Passionale 46v) (11) das jhr euch dis Capitel wollet ja lassen Iiebi angelegen vnd that you REFL this chapter will1 yes let2 dear fitting and befohlen sein recommended be3 'that you will indeed let this chapter be dear, fitting, and recommended to you' (Passionale 35r)

2. It may be safe to claim that these word orders were ungrammatical in ENHG, first because they are unattested in my corpus. and secondly because they are ungrammatical or extremely rare in modern German dialects (Chapter 4.2.3).

49

so


There is only one attested order for complexes of four verbs: 1-2-4-3, which occurs just three times. Two of these clauses exhibit the infinitivus pm participio (IPP) construction plus either another infinitive (12) or a passive (13), and one involves a modal plus causative lassen plus passive (14). (12) das Gott wunderbarlicher wyB hat lassen herfiirwallen geschinen REL God miraculous way has 1 let2 forth.flow4 appear3 'which God miraculously caused to appear to flow forth' (Gespenster 19v) (13) das sie mehr fiir ein Rilst- dann Schlaffkammer ... hette moegen that it more for an armor than sleep.chamber had 1 can2

gehalten werden held4 be3 'that it could have been considered more an armory than a bedroom'

(Beschreibung 13) (14) das die selben jre Prediger jnen sollen lassen befolhen sein that the same their pastors them shalh let2 ordered4 be3 'that they should let their pastors be ordered to them' (Sum maria 19v) Finally, as discussed in Section 3.3.2.2 below, extraposition may occur with threeverb complexes, regardless of the order of the verbs.

3.1.2 The database

My ENHG database consists of 2,906 subordinate clauses from thirty texts available on the on-line Bonner Frithneuhochdeutschkorpus. These thirty texts represent three texts each from ten dialect areas: Cologne, Hesse, Alsace, Zurich, Swabia, Nuremberg, Thuringia, Saxony, Augsburg, and Vienna. Within each dialect, there is one text each from the 14th, 15th, and 16th centuries} There are 2,737 clauses with VCs consisting of two verbs, 166 clauses with complexes of three verbs, and three clauses with four verbs. Approximately 100 clauses from each text were selected according to the following criteria.4 First of all, the database contains only finite subordinate clauses with complexes of at least two verbs, not counting the infinitive of any embedded infinitival clause. Not included in this study are participles with no auxiliary,

3· The Bonner corpus has texts from 4 time periods: 1350-1400, 1450-1500, 1550-1600, and 1650-1700. I have not included any texts from the last period, since I consider that to be the NHG era and presume that the rate of1-2 would be too low for statistical analysis. 4· Pillenreuth ~stik is written in two hands; I took 97 clauses from the first hand and 80 clauses from the second hand


Table 1. Texts in my ENHG databases Dialect

1350-1400

1450-1500

1550-1600

Cologne Hesse Alsace Zurich Swabia Nuremberg Thuringia Saxony Augsburg Vienna

Buch ]((jln Benedictinerregel Mannen Naturlehre Altvater Namen Psalter Dresden Altdeutsche Pred. ThJja Rationale

Koellwff Chronik Hortus Santitatis Chirurgie Edlibach Chronik Eunuch us Pillenreuth M1stik Rothe Chronik Sermon Vita Denkwurdigkeiten

Gegenwartigkeit Walter Ralegh Nachbarn Gespenster Beschreibung Summaria Bange Chronik Passionale Nachtmahl M.oscouia

which are fairly common in ENHG (for a discussion, see Reifsnyder 2003:231234). Also excluded are two tokens with a repeated finite verb as in (15), another construction discussed in Reifsnyder (2003: 237 -241). rache ist gehaissen ist revenge is 1 called2 is 1 'which is called revenge'

(15) der da

REL PART

(Altdeutsche Fred. 7)

Secondly, only unambiguous subordinate clauses were selected.6 Frequently, clauses with 1-2 or 1-x-2 order had to be rejected because they were ambiguous with verb-second (V2) main clauses. Many potential relative clauses were rejected, since the relative pronoun is usually homophonous with the demonstrative pronoun, resulting in clauses that are ambiguous between a main clause and a subordinate clause (16a). Thus relative clauses were included in the database only if at least one constituent intervened between the pronoun and the verb, indicating an unambiguous subordinate clause (16b).

werden vergeben sein sund be 1 forgiven2 his sin 'that one I whom will be forgiven his sin' b. der in dem himel nicht mocht sein vnter got REL in the heaven not could 1 be2 under God 'who could not be in heaven with God'

(16) a.

dem

DEM/REL

(Pillenreuth 176)

(Pillenreuth 162)

5· Complete titles are found in the bibliography, and additional information can be found on the website for the Bonn corpus: http://www.korpora.org/Fnhd!

These criteria are stricter than those used in the MHG study (Chapter 2), which included V2 clauses introduced by wande and even those introduced by an ambiguous relative/demonstrative pronoun.

6.

51

52


Also frequently ambiguous are clauses with wan. According to Reichmann & Wegera (1993:473) wan is subordinating in the meaning 'when' (17), but in the meaning 'for/because: wan may be coordinating (18a) or subordinating (18b ). (17) wan sich dy sel sol abschaiden von dem leib when REFL the soul shall1 part2 from the body 'when the soul shall leave the body'

(Pillenreuth 161)

wan er mag nymer mer sunden for he can 1 never more sin2 'for he can never sin again' (Pillenreuth 158) b. wan er kain menschen lest mer angefochten werden for he no person lets 1 more tempted3 be 2 'for he does not let anyone be more tempted' (Pillenreuth 166)

( 18) a.

With wa.n 'for/because, clauses are only included in the database in cases where it is dearly subordinating, i.e. when at least two constituents intervene between the complementizer and the finite verb, as in (18b ). Finally, if the complementizer is unambiguously subordinating but the clause is V2, as in (19), the clause is kept in the database but coded as V2. This allowed me to include such clauses in the database (under the assumption that they involve extra position), while maintaining the option of excluding them if more restrictive criteria seemed necessary later on. (19) ob er wolt leben in aynem rechten cristenlichen glauben if he would1 live2 • • • in a correct Christian faith 'whether he would live ... in a true Christian faith' (Pillenreuth 167) The analyses were conducted using the statistics package GoldVarb X (Sankoff et al. 2005). For more information about GoldVarb and on how to interpret the tables, see Chapter 2.1.3. The clauses were coded for the following factor groups: verb order (the dependent variable), syntagm type, constituent preceding the VC, constituent intervening in the VC, constituent following the VC, clause type, focus type (new vs. contrastive), prefix type, scrambled object, coordination, and verb second.7 Each text in the database was tagged for century; dialect, genre, and the sex, occupation, and education of the author.

7. Two of these factors, clause type and VP coordination, had no effect and will not be discussed further here, but they were discussed in detail in Sapp (2006). A pilot study (Sapp 2005) tested two additional variables that had no effect and thus were not coded in the complete database: position of the clause within the sentence and the alternation of stressed and unstressed syllables (see Section 2.1 below). Finally, note that some factors were tested in MHG (see Chapter 2) but not in ENHG: constituent focused, whether an intervening constituent is part of the focus, and whether an extraposed constituent is part of the focus.


3.1.3 Organization of this chapter

This chapter is organized as follows. The analyses of VCs consisting of two verbs are presented in Section 3.2. Section 3.3 treats three-verb complexes. Section 3.4 discusses the factors that influence word order in individual texts and in the various dialects, uses those data to propose dialectal divisions in ENHG, and compares the MHG results with those from ENHG. This chapter is concluded in Section 3.5.

3.2

Complexes of two verbs in ENHG

3.2.1 Introduction

In my database of 30 texts from the Bonner Friihneuhochdeutschkorpus, there are 2,737 subordinate clauses with exactly one finite and one non-finite verb. In 671, or 24.5%, of these clauses, the finite verb precedes the non-finite verb. I refer to these instances as "the 1-2 order~ and unless otherwise noted, this refers to all cases with a fronted finite verb, i.e. it includes both the 1-2 order proper and the 1-x-2 order. As a rough approximation, a rate of 1-2 higher than 24.5% indicates a favoring effect. A favoring effect on the 1-2 order should be expressed by GoldVa.rb as a factor weight greater than 0.5, with a higher factor weight signifying a stronger effect In Section 3.2.2, I discuss factors that do not have any effect on verb order, contrary to the claims of previous scholarship: the position of the subordinate clause within the sentence and the pattern of stressed and unstressed syllables in the VC. In Section 3.2.3, I discuss the five factors that do affect verb order: syntagm type, the category of the word preceding the VC, extraposition, focus, and verbal prefixes. Section 3.2.4 discusses the diachronic, dialectal, and sociolinguistic variation in the rate of the 1-2 order in ENHG. Of the factors that have a significant favoring effect on the 1-2 order, eight were determined to be the most significant combination of factors by GoldVarb's step-up/step-down function: syntagrn., constituent before the VC, focus, prefix type, date, dialect, genre, and occupation. 3.2.2 Non-favoring factors 3.2.2.1 Position of clause Using a small sample of ENHG clauses, Maurer finds that the position of the subordinate clause within the sentence affects word order (1926: 161-162). In his data, sentence-final clauses tend to have the order 2-1 and non-sentence-final

53

54


clauses 1-2. He argues that the 2-1 order is preferred in sentence-final clauses due to the falling intonation at the end of the sentence and the tendency for auxiliaries to be located in phonologically weak positions. I examined this factor in a pilot study of four texts (Sapp 2005).8 The data from that study do not support Maurer's findings. Table 2 shows that whether the clause is sentence-final or not, the rate of the 1-2 order is very similar to 27%, the overall rate of 1-2 in these four texts. The distinction between sentence-final and non-sentence-final clauses is very slight: 26% versus 28%, with factor weights very close to 0.5. Furthermore, this distinction is not statistically significant.

Table 2. Effect of the position of the clause on 1-2 order (Sapp 2005)9 Position of clause

2-1

sentence final not sentence final

87 (73%) 251 (71%)

1-2 32 (26%) 98 (28%)

Total

338 (72%)

130 (27%)

Factorwt. 0.469 0.510

p = 0.102 Thus the effect of the position of the subordinate clause within the sentence on the choice of verb orders could not be confirmed. Nor was Ebert able to confirm any effect in his study of texts from Nuremberg: "there is no consistent pattern here of influence due to occurrence in a 'Vordersatz' vs. 'Nachsatz'" (1981: 206). Since neither Sapp (2005) nor Ebert (1981) could verify this effect, it was not tested on the whole ENHG database and will not be discussed further. 3.2.2.2

Alternation of stressed and unstressed syllables

ENHG, like the other continental West Germanic languages, has a series of verbal prefixes. There are a number of stressed, separable prefixes (SSPs), like auff in (20). There are also unstressed prefixes, which are inseparable from the verb, such as ver- in (21).

8. For the pilot study, I selected clauses from four 15th-century texts from the Bonner FrUhneuhochdeutsch-Korpus: Eunuchus, Edlibach Chronik, Rothe Chronik, and Pillenreuth Mystik. 9· This analysis was conducted using GoldVarb 2001 (Robinson et al. 2001), an earlier version of the program than was used for the other analyses in this book GoldVarb 2001 does not round up percentages, so they add up to 99% rather than 100%. Moreover, in Sapp (2006) I analyzed the data with the 2-1 order as the default variable, thus factor weights greater than 0. 5 indicate a favoring effect on the 2-1 order in these two analyses.


(20) dye got der her pannherczi.gklich hat auffgenumen REL God the Lord mercifully has 1 up.taken2 'whom God the Lord has mercifully received'

(Pillenreuth 160)

(21) das er in kainer sund verczweiffeln sol that he in no sin despair2 shall1 'that he shall not despair in any sin'

(Pillenreuth 161)

Ebert (1981: 207; 1998 passim), like Maurer (1926: 159), finds that non-finite verbs with a SSP favor the 1-2 order as in (20), while non-finite verbs with an unstressed prefix favor the 2-1 order as in (21). This observation holds true in my database as well, as demonstrated by Section 2.3.5 below. Ebert (1981: 208) attempts to account for this eftect of prefix type by hypothesizing that the verb order is sensitive to the alternation of stressed and unstressed syllables within the VC.IO According to Ebert, some patterns should favor either order (SOLlen SCHREiben - SCHREiben SOL/en 'should write'), since both orders produce an alternation of stressed and unstressed syllables. For other patterns, reordering to 1-2 results in consecutive stressed or unstressed syllables (beRICHten SOLlen ~ *SOLlen beRICHten 'should report') and thus should not favor 1-2. Likewise, the 1-2 order should be favored when it results in alternating stressed and unstressed syllables. However, this explanation does not hold for the texts analyzed in Sapp (2005), as shown in Table 3. In fact, the results here are the opposite of what one would expect under Ebert's hypothesis. When the underlying 2-1 order has alternating stressed and unstressed syllables but reordering to 1-2 removes that alternation, the frequency of 1-2 order is actually higher (44%) than the expected rate of 27%. Moreover, when the 2-1 order involves consecutive stressed (or unstressed) syllables but reordering to 1-2 would produce an alternating pattern, the rate of 1-2 is lower than expected (19%). Finally. clusters like schreiben sollen, which Ebert finds to have a high frequencyof1-2, have 1-2 at close to the expected rate (22%).11 Given the failure of the pilot study to find the expected effect, and given the difficulty of coding syllable structure in a large corpus, this factor was not tested on the complete database of thirty texts.

Ebert concedes that his sample is too small to for statistical significance. He does not maintain this explanation in a later study; however, he continues to find that when the V has the pattern as in beRICHten, 2-1 is fuvored (1998: 162). Note that this pattern is exclusively found in verbs with an unstressed prefix. which do slightly favor 2-1 in my corpus.

10.

Ebert limited his conclusions to modal + infinitive clusters. I tested this again excluding syntagms other than modal + infinitive and found very similar results to those in Table 3.

11.

55

56


Table 3. Effect of syllable pattern on 1-2 order (Sapp 2005) 12 Alternating syllables

2-1

1-2

Factorwt.

both orders result in clash both orders alternate 2-1 alternates, 1-2 dashes 2-1 dashes, 1-2 alternates

14 (70%) 168 (77%) 64 (55%) 92 (80%)

6 (30%) 50 (22%) 52 (44%) 22 (19%)

0.448 0.515 0.434 0.548

Total p < 0.001

338 (72%)

130 (27%)

Thus there does not seem to be a general effect of alternating stressed and unstressed syllables within the VC. However, as discussed in Section 3.2.3.5 below, there is an effect of prefix type per se. 3.2.3 Favoring factors 3.2.3.1 Syntagm type In those West Germanic dialects (including Standard Dutch) that allow 1-2, there is a preference for 2-1 in constructions with participles, and for 1-2 with a modal plus infinitive (see Chapter 4 for details and references). As discussed in Chapter 1, several earlier studies have found a similar effect in earlier stages of German, and the current study confirms this for MHG in Chapter 2. The data from my ENHG database match this tendency. as can been seen in Table 4. The passive syntagms have the lowest rates of 1-2 at around 11%. The perfect syntagms display frequencies of the 1-2 order that are fairly close to the expected rate of24.5%, and the factor weights indicate little efiect on verb order. On the other hand, syntagms with an infinitive favor 1-2, with frequencies higher than the total and factor weights well above 0.5. The effect of syntagm was selected in GoldVarb's step-up/ step-down function as one of the most significant factor groups. In addition to the participial and infinitival syntagms, there is a relatively rare third type, the progressive tense. The ENHG progressive consists of the auxiliary sein or werden plus either the infinitive or present participle:13

was (22) do nue die Junk.chfraun vnd y-eder man slaffen when now the virgins and every man sleep-inf2 was 1 'now when the virgins and everyone were sleeping' (Denkwardigk. 15) 12.

Analysis was conducted with GoldVarb 2001 (see footnote for Table 2 above for details.)

13. The 24 instances of subordinate clauses with the progressive tense in my corpus come from only five texts. Ten of these have the infinitive as in (22)-(23). All14 instances with the present participle come from a single text, Mannen.


Table 4. Effect of syntagm on 1-2 order

Syntagm

2-1

sein passive werden passive

236 (89.4%) 446 (89.0%) 20 (83.3%) 520 (77.4%) 49 (76.6%) 177 (70.8%) 592 (65.5%) 26 (44.8%)

28 (10.6%) 55 (11.0%) 4 (16.7%) 152 (22.6%) 15 (23.496) 73 (29.2%) 312 (34.5%) 32 (55.296)

2066 (75.5%)

671 (24.5%)

progressive perfect with haben future perfect with sein modal +inf other V + inf. Total

1-2

Factorwt. 0.287 0.295 0.404 0.498 0.509 0.583 0.641 0.806

p < 0.001 (23) do sich sein sach pessern ward when REFL his case improve-inf2 was 1 'when his case was improving'

(Denkwardigk. 14)

(24) sider ich dise bekentnisse von dir habende bin since I these confessions from you having2 am 1 'since I have these confessions from you'

(Mannen 4)

(25) das ich ... frvege alleine wart sitzzende that I early alone was 1 sitting2 'that I was sitting alone early (in the morning)'

(Mannen 3)

Although the number of tokens is too small to allow one to draw any firm conclusions, the progressives appear to pattern with the participial constructions. The resulting hierarchy can now be compared to those in previous studies. The hierarchy from my ENHG corpus (26a) is largely similar to the hierarchies in Bies' (26b) and Ebert's (26c) ENHG corpora. These are also comparable to the hierarchy I find in MHG, repeated in (27) from Chapter 2. (26) a. b.

sein pas.> werden pas.> haben pf. > fut. > sein pf. >mod. werden pas. > sein pas. > haben pf. > sein pf. > fut. > mod. (Bies 1996: 58)

c.

werden pas. > sein pas. > haben pf. > fut. > mod. > sein pt~ (Ebert 1992: 5)14

(27) sein pass.> werden pass.> sein perf.> haben perf.> mod.

14. As reportedinBies (1996: 58); this is verysimil.artothehlerarchy reported in Ebert (1998:65), see Chapter 1. Ebert (1981) is not compared here because sein passives were not included The direction of the arrows here indicates the descending preference fur the 2-1 order.

57

58


The hierarchies do not correspond exactly, but the general trend of passive >perfect> modal does hold. 3.2.3.2 Word preceding the verbal complex Ebert (1981: 206) finds that a noun preceding the VC favors 1-2 while a preceding pronoun favors 2-1, and I have found similar results in MHG (see Chapter 2). However, as can be seen in Table 5, in my ENHG database there is no difference at all between the rate of the 1-2 order after a noun versus after a pronoun. In both cases, the rate is close to 24.5%, the expected rate of 1-2. This indicates that Ebert's (1981:206) claim about the effect of word stress on verbal order in Nuremberg does not hold for all of ENHG. Table 5. Effect of the category of the preceding word on 1-2 order Class of preceding word

2-1

1-2

Factorwt.

noun pronoun quantified NP adjective adverb prepositional phrase stranded or da-preposition clause (infinitival or finite) nothing precedes Total

519 (75.596) 302 (75.996) 49 (68.196) 39 (60.0%) 436 (75.096) 598 (79.0%) 6 (27.3%) 45 (81.8%) 72 (72.0%)

168 (24.5%) 96 (24.1%) 23 (31.9%) 26 (40.0%) 145 (25.0%) 159 (21.0%) 16 (72.2%) 10 (18.296) 28 (28.096)

0.502 0.498 0.594 0.675 0.509 0.453 0.892 0.409 0.548

2066 (75.5%)

671 (24.596)

p < 0.001 Although the predicted noun/pronoun distinction does not hold, this factor group is statistically significant, probably because of the effects of other parts of speech, to be explained to some degree below. (Moreover, this was selected as one of the most significant factor groups by GoldVarb's step-up/step-down analysis.) The categories that have a favoring effect on verb order are adjectives, stranded prepositions, and nothing preceding. With an adjective preceding the VC, the rate of the 1-2 order is 40%, considerably higher than the expected rate of 24.5%. I have no explanation for the fact that adjectives seem to favor 1-2; however, note that the number of tokens with adjectives is relatively low. ENHG, unlike Modern Standard German, allows preposition stranding: (28) (durch die wuisti) da anthonius inne was gewesen through the desert REL A. in was been 'through the desert that Anthonius had been in'

(Altvater73v)


When preceded by a stranded preposition (or ada-compound), the 1-2 order is more frequent than expected, at a rate of over 70%. This fact is easily accounted for by the interaction with time. Of the ten occurrences of preposition stranding, six are from the 14th century (three of which are nearly identical clauses from a single text, Altviitet·), when the rate of 1-2 is the highest, and none at all are from the 16th century; when 1-2 is rare. When no constituents occur between the complementizer or relative pronoun and the VC, the 1-2 order occurs at a rate of 28%, somewhat higher than the expected rate. This is also easily accounted for: most instances of 1-2 with nothing preceding the VC come from second conjuncts, and thus are best analyzed as instances of anacoluthon, i.e., a sentence starts as one type but after some syntactic break continues as another type. One of the most frequent kinds of anacoluthon in earlier stages of German involves conjoined subordinate clauses: although the second conjunct should have subordinate-clause word order, it sometimes has main-clause word order instead (Paul1920: 381). This is especially clear when the second conjunct contains several arguments, as in (29), for which the most likely analysis is that the finite verb wil is in the V2 position while the constituents dir and ring are in the middle field, rather than assuming that wil dir ring machen is a VC in the 1-x-2 order: (29) das ich mich in dein sel pergen mufi vnd wil dir ring that I REFL in your soul hidez must1 and want1 you humble machen alles das make 2 all that 'that I must hide myself in your soul and want to make all of that humble for (Pillenreuth 223) you' Thus apparent instances of 1-2 with nothing preceding the finite verb are probably best treated as having main-clause-like V2 word order. Recall from Chapter 2 that Ebert (1981: 207) attributed the differing effect of nouns and pronouns on verb order to phonological stress, speculating that the heavier stress of nouns compared to pronouns favors the 1-2 order. In a later, much more detailed study; Ebert (1998:65) finds stress to be a significant factor, independent of the noun vs. pronoun distinction. Likewise, in my MHG analysis presented in Chapter 2, both the part of speech and the stress of the word preceding the VC had significant effects. So, although the analysis presented Table 5 fails to show any effect of the noun/pronoun distinction in my ENHG database, perhaps the phonological weight of the preceding word plays some role. Treating pronouns, short adverbs, the negator, and da-compounds as unstressed, with all other categories considered stressed (including the significant types adjective

59

6o


and stranded preposition), stress has no effect at all.15 As can be seen in Table 6, the rate of 1-2 is nearly exactly the same whether the VC follows a stressed or an unstressed word. Table 6. Effect of the stress of the preceding word on 1-2 order Stress of preceding word

2-1

1-2

Factorwt.

stressed unstressed

1295 (75.5%) 699 (75.9%)

421 (24.5%) 222 (24.1%)

0.502 0.497

Total

1994 (75.6%)

643 (24.4%)

p = 0.777 These results, together with Bies' (1996:59) findings, indicate that the stress of the preceding word has no efiect on the order of the verbs in general ENHG. However, such an effect exists in ENHG at the dialectal level, as determined by Ebert for Nuremberg. In fact, Sapp (2006) analyzes each dialect individually, determining that the 1-2 order is favored by a preceding noun in most dialects (see Table 40 below for details). One of these dialects is. perhaps not coincidentally, Nuremberg, the subjectofEbert's studies. On the other hand, some dialects show the opposite effect, i.e. preceding nouns disfavor and pronouns favor 1-2. Thus the favoring effect of preceding nouns on the 1-2 order, which is present in much of the ENHG-speaking area, is probably cancelled out in the analysis of the entire database by texts that show the opposite behavior. Perhaps the conclusion from this should be that there was indeed a strong favoring effect of preceding nouns on the 1-2 order in MHG (perhaps due to Behaghel's "rhythmic" principle), an effect which by ENHG has become limited to certain dialects and which soon becomes obsolete in written German with the disappearance of the 1-2 order. 3.2.3.3 Extraposition Ebert (1981: 209) notes that when the VC is in clause-final position, the 2-1 order is preferred. To put it differently, there is a correlation between the 1-2 order and the extraposition of some constituent, usually an NP or a PP. This holds true in my data as well, as can be seen in Table 7. When an argument occurs to the right of the VC, the 1-2 order occurs nearly 35% of the time, well above the expected rate of24.3%. (Although statistically significant, this was not selected as one of the most significant groups in the step-up/step-down analysis.)

15. This method of determining stressed vs. unstressed words is the same as Ebert's (1998:7), so the difference between my study and his is not due to differing criteria


Table 7. Effect of e.xtraposition on 1-2 order NP/PP extrapositionl6

2-1

1-2

extraposed argument extraposed adjunct PP nothing extraposed

143 (65.3%) 152 (72.4%) 1742 (77.0%)

76 (34.7%) 58 (27.6%) 520 (23.0%)

Total

2037 (75.7%)

654 (24.3%)

Factorwt. 0.625 0.545 0.484

p < 0.001 When there is no extraposition, the rate of the 1-2 order is 23%. very dose to the expected rate. The one category of extraposed constituent that behaves difterently from other categories is adjunct PPs, such as an das todpett in (3). When an adjunct PP is extraposed, the rate of 1-2 is just 27.6%, also close to the expected rate.l7 Finally. note the rate of extra position in the corpus. Of the 2,691 clauses, 210 or 7.9% have a extraposed adjunct PP, while 219 or 7.9% have some other constituent extraposed. The total rate of extraposition is 15.9%, which has decreased somewhat from the rate of extra position in MHG (19.7%). This leads to an interesting comparison with Modern Standard German. First of all, nearly half of the extraposed constituents in my ENHG corpus are adjunct PPs. Secondly, the fact that adjunct PPs have 1-2 at a lower rate than other types of extraposed constituents suggests that they are not subject to the same restrictions on extraposition. This is not surprising, since adjuncts also extrapose more easily than NPs and argument PPs in Modern Standard German. In Lambert's (1976: 137) corpus of written and spoken Standard German, 16.5% of the sentences have an extraposed adverbial phrase, which includes adverbial PPs, while only 3.5% of the sentences have an extraposed argument PP. Bies (1996:65) concludes that focus-driven extra position of NPs and argument PPs was lost in the transition to NHG, while extraposition of adjuncts continued. The data presented in this section are consistent with that conclusion. 3.2.3.4 Focus In the previous chapter, it was determined that focus affects verb order in MHG. In this section, focus effects in ENHG are tested. Unlike in my MHG database, in the ENHG database no distinction is made between focus on difierent constituent 16. This table excludes a small number of extraposed constituents that are neither adjunct PPs nor arguments. 17. Recall from Chapter 1 that Bies' data also show that NP extraposition affects verb order,

while PP extraposition does not. In fact, her numbers are strikingly similar to mine: 1-2 occurs 37.1% of the time with NP extraposition and 29.2% with PP extraposition. versus the expected rateof27.3% (Bies 1996:61).

61

62


types, e.g. direct object focus versus focus on a verb. Rather, clauses were tagged as either having new information focus on some constituent, contrastive focus on some constituent, or no focus (i.e. all old information). For example, in (30) the context is the sufferings of Christ, thus 'I have done everything' is discourse-old, and du.rch die mensch is the new information in the clause. Examples (31) and (32) illustrate cases where contrastive focus can be detected, with the NP nicht anders in (31) contrasting with the following clause, and the two verbs in (32) contrasting with each other. (30) das ich alles lFoe durch die mensch] hab getha.n. that I all for the person have done 'that I have done everything for the sake of

man:

(Pillenreu.th 220)

(31) das der mensch alle sein Iebttag [Foe nicht anders] scholt thun, denn that the person all his life.days nothing else should do than 'that man should do nothing else all the days of his life, than ..: (Pillenreuth 206) (32) das ein cristenmensch in dem stat getar (Fleben], in dem er nicht that a christian. person in the place dares live, in REL he not gern wolt [Foe sterben] gladly would die '(It is a wonder), that a Christian dares LIVE in a place where he would not like to DIE' (Pillenreuth 211) Clauses were tagged new if an argument was new within the section of the text, or new within one or two pages if the text was not divided into sections. Clauses were tagged contrastive regardless of whether an argument was new or given in the discourse, if it could be interpreted as contrastive. If there were no new or contrastive arguments, the clause was tagged as old. As seen in Table 8, if there is contrastive focus, the 1-2 order is very strongly favored, occurring more than 53% of the time. With new information focus, 1-2 is also favored at the rate of nearly 31%, still considerably higher than the expected rate of24.5%. On the other hand, old information has a disfavoring effect on 1-2, occurring only 19% of the time. The effect of focus is not only statistically significant but was also selected in GoldVarb's step-up/step-down function as one of the most significant factor groups. Using context for determining the focus structure of a sentence in a nonliving language can be quite difficult, and that is especially true for contrastive focus. It is possible that in some cases the 1-2 order has biased me into reading the clauses with a special intonation, and there are no doubt instances of focus that I have overlooked. However, there are some additional data that lend support to the argument that focus influences verb order.


Table 8. Effect of focus on 1-2 order Focus

2-1

1-2

contrastive focus new information old information

15 (46.9%) 815 (69.1%) 1236 (81.0%)

17 (53.1 %) 365 (30.9%) 289 (19.0%)

Total

2066 (75.5%)

671 (24.5%)

Factorwt. 0.782 0.587 0.426

p < 0.001

Therefore, let us attemptto corroborate the effect ofinformation structure by looking at two additional factors: the definiteness of the NP immediately preceding the VC and scrambling. It is well known that indefinite NPs tend to represent new information while definite NPs and pronouns tend to represent old information, thus one would predict that indefinites favor the 1-2 order and definites do not. However, in my database, there is no favoring effect of indefinite NPs preceding the VC, as seen in Table 9. When the preceding NP is indefinite (including QPs withjeder, kein, or aile), the rate of 1-2 is 21.9%, and when it is definite (including pronouns and proper names), the rate is 25.8%. Note, however, that the definiteness of the preceding constituent would only be able to capture new information focus, since contrastive focus may involve a preceding definite NP.

Table 9. Effect of the definiteness of the preceding NP on 1-2 order Preceding NP

2-1

1-2

Factorwt.

indefinites (including QPs) definites

235 (78.1%) 635 (74.2%)

66 (21.9%) 221 (25.8%)

0.461 0.514

Total

870 (75.2%)

287 (24.8%)

p = 0.176 The second piece of supporting evidence that focus influences verb order is the correlation between lack of scrambling and the 1-2 order. An object is considered to be scrambled if it appears to the left of a sentential adverb or negator, like den menschen in (33), and is not scrambled if it is to the right, like the word sundt in (34): (33) so der teutel [den menschen] nit vberwinden mag if the devil the person not overcome can 'if the devil cannot overcome that person' (34) als ob du nie [sundt] habst gethan as if you never sin have done 'as if you have never committed sin'

(Pillenreuth 163)

(Pillenreuth 223)

63

64


When an object is new information or contrastive it should not undergo scrambling but remain inside the VP, whereas old information NPs tend to scramble to the left. In the database, clauses were tagged as unscrambled if an object appeared to the right of a negator or adverbial (i.e. immediately left of the verbs) and scrambled if an object was separated from the VC by some constituent. As seen in Table 10, clauses with unscrambled objects favor the 1-2 order at a rate of 38.4%, well above the expected rate. In clauses with scrambled NPs, on the other hand, the rate of 1-2 is closer to the expected rate, although there does seem to be a slight favoring effect. The fact that the 1-2 order is favored with unscrambled objects may confirm the effect of focus on that order; however, it may also be a result of the so -called rhythmic effect of an NP immediately preceding the V C. Table 10. Correlation between scrambling and 1-2 order Scrambled object

2-1

1-2

object not scrambled object is scrambled cannot tellls

45 (61.6%) 115 (71.4%) 1906 (76.1%)

28 (38.4%) 46 (28.6%) 597 (23.9%)

Total

2066 (75.5%)

671 (24.5%)

Factorwt. 0.657 0.553 0.492

p = 0.013 Finally, Section 3.2.3.2 above showed that extraposition favors the 1-2 order. This fact may also support the favoring effect of focus on 1-2, because according to Bies (1996) extraposed constituents in ENHG tend to be focused. 3.2.3.5 Prefix type In his studies of ENHG texts from Nuremberg, Ebert (1981: 207; 1998 passim) finds that stressed separable prefixes (SSP) favor the 1-2 order. The same effect is also found in my MHG study (Chapter 2). As seen in Table 11, this also holds for my ENHG corpus: when the non-finite verb has a stressed prefix, the 1-2 order occurs 35.7% of the time, well above the expected rate of 24.5%. The factor weight, well above 0.5, confirms that this factor strongly favors the 1-2 order. Verbs with no prefix have a slight favoring effect on 1-2, while those with an unstressed prefix disfavor it The effect of prefix type was selected in GoldVarb's step-up/step-down function as one of the most significant factor groups.

18. I have included pronouns in this category. Pronouns do scramble since they are always old

information, and tend to appear much earlier in the clause than other NPs, with pronominal objects often coming even before the subject When this factor is removed from the analysis, so that only the dear cases of scrambling versus non-scrambling are tested, the significance is improved top = 0.003.


Table 11. Effect of prefix type on 1- 2 order Prefix type

2-1

1-2

Factorwt. 0.636 0.576 0.442

stressed no prefix unstressed

187 (64.3%) 531 (69.9%) 1348 (80.0%)

104 (35.7%) 229 (30.1%) 338 (20.0%)

Total

2066 (75.5%)

671 (24.5%)

p < 0.001 A conceivable reason for the correlation between prefix type and word order could be an interaction with syntagm type. Past participles usually have an unstressed prefix, especially the perfective marker ge-, whereas infinitives do not necessarily. (Either form may have an SSP.) GoldVarb X has a cross-tabulation function, which allows the researcher to test for interactions of this kind. The result of the cross-tabulation of prefix type and syntagm type, given in Table 12, suggests that the factors prefix type and syntagm type are independent of each other, and that their effect on verb order is cumulative. For all three factors in the factor group prefix type, the frequency of 1-2 is higher with infinitival constructions than with participial constructions (stressed prefix: 51%> 25%; no prefix: 33% > 22%; unstressed prefix 34% > 17%). Likewise, for both types of syntagms, the frequency of 1-2 is highest with stressed prefixes and lowest with unstressed prefixes (past participles: 25% > 17%; infinitives: 51%> 34%). Table 12. Cross-tab of prefix type with syntagm type Prefix type

word order

stressed

2-1

128 (75%)

57 (49%)

1-2

43 (25%)

59 (51%)

102 (36%)

no prefix

2-1

106 (78%)

409 (67%)

515 (69%)

1-2

30 (22%)

197 (33%)

227 (31%)

unstressed

2-1

1145 (83%)

201 (66%)

1346 (80%)

1-2

235 (17%)

103 (34%)

338 (20%)

2-1

1379 (82%)

667 (65%)

2046 (75%)

1-2

308 (18%)

359 35%)

667 (25%)

Total

participle

infinitive

Total 185 (64%)

In conclusion, the presence of a stressed separable prefix on the non-finite verb favors the 1-2 word order, and this effect is independent of syntagm type. Because prefix type has a significant, independent effect, and Section 3.2.2.2 above demonstrates that Ebert's account for the efiect of prefix type does not hold in my ENHG database, another explanation must be sought. Perhaps there is a syntactic reason for this preference, such as the fact that a verb and its SSP form a complex predicate (Muller 2002:409). On the other hand, phonology could be the reason,

65

66


but not in the way Ebert claims. While Ebert claims that the prosody of prefixed verbs plays a role due to a preference for alternating stressed syllables in the VC, I suggest in Chapter 5 that SSPs may result in a clash of word stresses with the word preceding the VC. Alternatively, one might appeal to Behaghel's (1932:III, 367) Gesetz der wachsenden Glieder ('Law of the growing constituents'), which states that in German, heavier items tend to be placed as late as possible in a clause. Since a verb with an SSP is phonologically heavier than a verb without an SSP, it follows that complexes including such verbs would be a favoring environment for the 1-2 order. Having discussed four linguistic factors that favor 1-2 in ENHG subordinate clauses, in the next section I will present the variation in subordinate clause word order over time, across dialects, and sociolinguistically.

3.2.4 Diachronic, dialectal, and sociolinguistic variation 3.2.4.1 Date As discussed in Chapter 1, it is well established that the 1-2 order decreases over time, becoming ungrammatical in Modern Standard German. At the end of the MHG period, the 1-2 order occurred in 33.7% of two-verb complexes (see Chapter 2). In the ENHG corpus, there is a sharp decrease in the rate of 1-2 order from the 14th to the 16th centuries, as seen in Table 13. In the 14th century, the rate of 1-2 is 35.5%, considerably higher than the overall rate of 24.5%. By the 15th century, the rate of 1-2 has dropped to 22%, just below the overall rate. The frequency of 1-2 in the 16th century is lower still. The effect of the date of the text is significant, and in fact was selected in GoldVarb's step-up/step-down function as one of the most significant factor groups. Table 13. Effect of time on 1-2 order

Century

2-1

1-2

Factorwt.

1350-1399 1450-1499 1550-1599

544 (64.596) 783 (78.096) 739 (83.196)

300 (35.596) 221 (22.096) 150 (16.996)

0.638 0.475 0.395

2066 (7 5.596)

671 (24.596)

Total

p < 0.001 The relatively high rate of 1-2 in the 16th century in the table above may be a bit misleading, due to the presence of one text, Walter Ralegh. In Walter Ralegh, 69% of the VCs with two verbs have the 1-2 order, more than twice as high as any other 16th-century text, and more than triple the earlier Hessian texts. I believe


the extremely high frequency of 1-2 is a result of the fact that the ENHG Walter Ralegh is a translation from Dutch.19 With that questionable text removed from the analysis, the following distribution obtains, with the frequency of 1-2 in the 16th century reduced to 11.5%. Table 14. Effect of time on 1-2 order, excluding Walter Ralegh Century

2-1

1-2

Factorwt.

1350-1399 1450-1499 1550-1599

544 (64.596) 783 (78.096) 714 (88.596)

300 (35.596) 221 (22.096) 93 (11.596)

0.666 0.505 0.321

2041 (76.996)

614 (23.1 96)

Total

p < 0.001 This result is largely similar to trends found in previous studies. In Hammarstram (1923), Ebert (1981), and Bies (1996), 1-2 generally declines over time, as seen in Table 15. For all three corpora, the frequency of 1-2 is highest at the beginning of the period and lowest at the end. However, unlike my ENHG corpus and Hammarstrom's study, which show a steady decrease of 1-2 over time, both Ebert's and Bies' data have an increase in the middle of the period (1500-1550 and 1450-1500, respectively) before decreasing in the end. According to Bies (1996:50), this kind of discrepancy across corpora is "suggestive of widespread synchronic variation and the normative pressure of a change from above=' Table 15. Effect of time on 1-2 order: Comparison to previous studies Period 1300-1350 1350-1400 1400-1450 1450-1500 1500-1550 1550-1600

Hammarstr0m20

Ebert (1981)21

12.696

30.796

5.096

25.296

096

32.396 23.696

Bies (1996) 36.196 20.696 19.296 34.896 24.296

Table 14 35.596 22.096 11.596

19. Walter Ralegh had to be removed from the time, dialect, and sociolinguistic analyses, where one text could have a strong influence on the results. However, Walter Ralegh is included in the other analyses because the favoring factors are the same as other ENHG texts (see Sapp 2006

for analyses of each individual text). 20.

Percentages for imperial and princely documents. For details, see Sapp (2006).

21. As reported in Bies (1996: 52). Bies converts Ebert's ratios to percentages. Moreover, I have averaged the percentages for sentence-internal and sentence-final, to ease comparison.

67

68


Having seen the decrease over time of 1-2 in the wider sense (the finite verb precedes the non-finite verb), let us look at the difference between the 1-2 order in the narrow sense (the finite verb immediately precedes the non-finite verb) and the 1-x-2 order. As seen in Table 16, the rate of the 1-2 order in the narrow sense decreases over time, while the 1-x-2 order decreases only slightly..22 As a result, over time the 1-x-2 order makes up a larger percentage of the VCs in which the finite verb precedes, from 36 out of300 (12%) in the fourteenth century to 24 out of93 (nearly 26%) in the sixteenth. Table 16. Effect of time on 1-2 and 1-x-2 orders Century

2-1

1-2

1-x-2

1350-1399 1450-1499 1550-159923

544 (64.4%) 783 (78.0%) 714 (88.5%)

264 (31.3%) 170 (16.9%) 69 (8.6%)

36 (4.3%) 51(5.1%) 24 (3.0%)

2041 (76.9%)

503 (18.9%)

111 (4.2%)

Total

Recall from Section 3.1.2 above that some clauses introduced by unambiguously subordinating conjunctions but with the verb in the second position were tagged as V2, making it possible to exclude such clauses in a GoldVa.rb analysis. Doing so is especially important with 1-x-2 clauses, to rule out the possibility that they involve main clause word order rather than unambiguous 1-x-2. Excluding these subordinate clauses with potentially main clause word order results in a smaller number of 1-2 clauses (especially 1-x-2 clauses), as can be seen in Table 17, although a comparison with Table 16 shows that this difierence is very slight. In any event, those 1-x-2 subordinate clauses that remain cannot be accounted for by main clause word order, indicating that ENHG, like some contemporary West Germanic dialects (see Chapter 4), allows VCs with an intervening constituent (the so-called Verb Projection Raising construction). The rate of the 1-x-2 order is fairly stable throughout the period, both as a percentage of the total clauses, and as a percentage of the 1-(x)-2 orders. Thus most of the decrease over time of the non- 2-1 word orders is a result of the decrease of the 1-2 order in the strict sense rather than a decrease of the 1-x-2 order.

GoldVarb X is not equipped for multi-variate analyses, i.e., although it can calculate the raw numbers and percentages for several values of the dependent variable, it can statistically test only two values. Thus the data for 1-x-2 order are given without fuctor weights or statistical 22.

significance. 23.

Excludes Walter Ralegh.


Table 17. Effect of time on 1-2 and 1-x-2 orders, V2 clauses removed Century

2-1

1-2

1-x-2

1350-1399 1450-1499 1550-159924

544 (68.0%) 783 (80.9%) 714 (89.0%)

227 (28.4%) 152 (15.7%) 68 (8.5%)

29 (3.6%) 33 (3.4%) 20 (2.5%)

Total

2041 (79.4%)

447 (17.4%)

82 (3.2%)

3.2.4.2 lJialect As one might expect from a linguistic feature that shows great variation by linguistic context and over time, verb order in subordinate clauses in ENHG varies widely by dialect As seen in Table 18, the frequency of the 1-2 order (in the wider sense, i.e. including 1-x-2) ranges from a low of 8.6% in texts from Cologne to a high of 41.7% in texts from Swabia. The effect of dialect on word order is one of the most significant factor groups, according to GoldVarb's step-up/step-down function. Table 18. Effect of dialect on 1-2 order 1-2

Dialect

2-1

Cologne Alsace Hesse25 Thuringia Zurich Augsburg Vienna Saxony Nuremberg Swabia

256 (91.4%) 247 (88.2%) 155 (87.6%) 179 (85.2%) 192 (83.8%) 216 (76.1%) 215 (75.7%) 193 (68.7%) 226 (64.2%) 162 (58.3%)

24 (8.6%) 33 (11.8%) 22 (12.4%) 31 (14.8%) 37 (16.2%) 68 (23.9%) 69 (24.3%) 88 (31.3%) 126 (35.8%) 116 (41.7%)

2041 (76.9%)

614 (23.1%)

Total

Factorwt. 0.259 0.332 0.346 0.392 0.418 0.539 0.544 0.629 0.674 0.727

p < 0.001 The geographical distribution of the 1-2 order is clearest when seen on a map of the German-speaking area, as in Figure 1. It appears that the main division is between the northern and western dialects (Cologne, Alsace, Hesse, Thuringia, and Zurich), with a frequency of 16% or less, and the southern and eastern dialects (Augsburg, Vienna, Saxony, Nuremberg, and Swabia), with a frequency of 23% or more. With the exceptions of Zurich and Saxony, this division roughly

24. Excludes Walter Ralegh. 25. Excludes Walter Ralegh.

69

70


corresponds to the isogloss bundle between Middle German or Mitteldeutsch (with varying degrees of the High German consonant shift) and Upper German or Oberdeutsch (with the complete consonant shift). The correspondence with phonological isoglosses suggests that this north-west/south-east distribution is a genuine one. Note, however, that this distribution is somewhat different from that in MHG, where the northwestern dialects had among the highest rates of the 1-2 order (see Chapter 2, Figure 1).

Figure 1. Frequency of the 1-2 order by dialect

Looking at the rate of 1-2 across the dialects and by century, however, complicates this picture a great deal. The percentages are given in Table 19 (the raw numbers can be found in Sapp 2006). Note that since the BFnhdK has only one text per dialect per century, the percentages here represent the results for just one individual text, which may account for much of the fluctuation seen here. For example, the drop in Augsburg from 58% in the 14th century to 7% in the 15th may reflect the fact that the 14th-century text is literary while the 15th-century representative is a chronicle (see Section 3.2.4.3 below for the effect of genre). As Ebert (p.c.) points out, given the fact that word order varies sociolinguistically, a single text cannot accurately represent a dialect.


Table 19. Effect of dialect and century on 1-2 order Dialect Cologne Hesse Alsace Zurich Swabia Nuremberg Thuringia Saxony Augsburg Vienna

1350-1399

1450-1499

1550-1599

Total

5%

19%

0%

9%

6% 18%

21% 7%

69% 8%

28% 7896 43%

13% 19% 33%

12% 25% 32%

30% 12% 16%

23% 46%

21% 32%

6% 13%

58% 3496

7% 34%

4% 2%

42% 36% 15% 31% 24% 24%

Nevertheless, there are several points to be made here. First of all every dialect shows a decrease in the rate of 1-2 from the 14th to the 16th century, some with a slight increase in the 15th (the obvious exception is Hessian, see footnote 19 above). Secondly, the north-west/south-east distinction is much less clear than before. In the 14th century, Cologne and Hesse have very low rates of 1-2, while Zurich and Thuringia have considerably higher rates (although still lower than 35.5%, the 14th-century average). Then in the 15th century, Cologne, Hesse, and Thuringia have higher rates of the 1-2 order (close to the overall rate for the 15th century, 22%), while Alsace and Zurich have somewhat lower rates. In the 16th century, the south-eastern dialects have very low rates of 1-2, patterning more with the north -western dialects. Because of the complexity of the data, it is necessary to use GoldVarb to determine the best dialect groups. The best model fit resulted from the following grouping, with the percentages given in Table 20, illustrated graphically in Figure 2. Table 20. Effect of time and most significant dialect groups on 1-2 Dialect group West (Col., Hes., Als., Zur.) Central (Swabia, Nuremb.) EMG (Saxony, Thuringia) Bavarian-Austrian

1350-1399

1450-1499

1550-1599

Total

12%

15%

7%

12%

61% 43% 46%

28% 27% 21%

28% 9% 3%

38% 24% 24%

71

7:2.


Rata of 1-2 ordar ovar time by most slgnlftcant dalact groups --+--West

70% 60%

50%

J

::!!

40%

"'

~

~-

~ 30% 20% 10%

..

--Q..-.

""' ··. .... ·..... ~""' ~

---6--EMG

••'I

The Verbal Complex in Subordinate Clauses from Medieval to Modern German

Subordinate Position

From Real to Complex Analysis

Verbal Violence in Contemporary Drama: From Handke to Shepard

Farming to Halves: The Hidden History of Sharefarming in England from Medieval to Modern Times

Medieval Warhorse: From Byzantium To The Crusades (Medieval Military Library)

Syntax of the Modern Greek Verbal System

Agent Oriented Programming From Prolog to Guarded Definite Clauses

Verbal Tone in Kuria

The Cambridge Companion to Modern German Culture

Agent-Oriented Programming: From PROLOG to Guarded Definite Clauses

Verbal Art, Verbal Sign, Verbal Time

The Future in Greek: From Ancient to Medieval (Oxford Linguistics)

Medieval and Modern Greek

Agent-Oriented Programming: From Prolog to Guarded Definite Clauses

From Latin to Modern French

Congenial Souls: Reading Chaucer from Medieval to Postmodern (Medieval Cultures)

From Holomorphic Functions to Complex Manifolds

From Simple Input to Complex Grammar

From Holomorphic Functions to Complex Manifolds

Handbook To Life In Medieval And Early Modern Japan

From holomorphic functions to complex manifolds

From simple input to complex grammar

Complex Numbers from A to ...Z

Complex Numbers from A to... Z

Medieval Trinitarian Thought from Aquinas to Ockham

Medieval Trinitarian Thought from Aquinas to Ockham

Complex Numbers from A to... Z

Medieval Trinitarian Thought from Aquinas to Ockham

Complex numbers from A to--Z

Chaos: From simple models to complex systems

The Verbal Complex in Subordinate Clauses from Medieval to Modern German

Subordinate Position

From Real to Complex Analysis

Verbal Violence in Contemporary Drama: From Handke to Shepard

Farming to Halves: The Hidden History of Sharefarming in England from Medieval to Modern Times

Medieval Warhorse: From Byzantium To The Crusades (Medieval Military Library)

Syntax of the Modern Greek Verbal System

Agent Oriented Programming From Prolog to Guarded Definite Clauses

Verbal Tone in Kuria

The Cambridge Companion to Modern German Culture

Agent-Oriented Programming: From PROLOG to Guarded Definite Clauses

Verbal Art, Verbal Sign, Verbal Time

The Future in Greek: From Ancient to Medieval (Oxford Linguistics)

Medieval and Modern Greek

Agent-Oriented Programming: From Prolog to Guarded Definite Clauses

From Latin to Modern French

Congenial Souls: Reading Chaucer from Medieval to Postmodern (Medieval Cultures)

From Holomorphic Functions to Complex Manifolds

From Simple Input to Complex Grammar

From Holomorphic Functions to Complex Manifolds

Handbook To Life In Medieval And Early Modern Japan

From holomorphic functions to complex manifolds

From simple input to complex grammar

Complex Numbers from A to ...Z

Complex Numbers from A to... Z

Medieval Trinitarian Thought from Aquinas to Ockham

Medieval Trinitarian Thought from Aquinas to Ockham

Complex Numbers from A to... Z

Medieval Trinitarian Thought from Aquinas to Ockham

Complex numbers from A to--Z

Chaos: From simple models to complex systems

Recommend Documents