Clause Structure and Word Order in the History of German (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated] 9780198813545, 0198813546
This volume presents the first comprehensive generative account of the historical syntax of German. Leading scholars in
122 26 4MB
English Pages 432 [423] Year 2018
Cover
Clause Structure and Word Order in the History of German
Copyright
Contents
Series preface
Preface and acknowledgements
List of figures and tables
List of abbreviations
List of contributors
1: Introduction
1.1 Aims of the volume
1.2 The syntax of German, the topological model, and the generative analysis of German clause structure
1.3 Overview of the history of German
1.4 Organization of the volume
Part I: The Left Periphery
2: Introduction to Part I
2.1 The evolution of the verb-second rule
2.2 Verb-initial declaratives
2.3 Distribution of information-structural categories in the prefield in later stages of German
2.4 The rise of adverbial connectives
3: Origins of verb-second in Old High German
3.1 Introduction
3.2 Proto-Germanic sentence grammar: major characteristics
3.3 The pre-verb-second grammar of Old High German: extensions at the left periphery
3.3.1 Dislocated topics and left-peripheral adverbial clauses
3.3.2 Left-peripheral sentence particles
3.3.3 Pronouns to the left of the fronted finite verb
3.3.4 Summary
3.4 Characteristics of the verb-second grammar of Old High German
3.4.1 Sketch of the Modern German verb-second grammar
3.4.2 An innovative feature of the Old High German left periphery: the filling of the prefield
3.5 Diachronic issues
3.5.1 Continuity and discontinuity in the C-domain
3.5.2 Origins of the different ways of filling the prefield
3.6 Summary
4: Verb-initial declaratives in Old High German and in later German
4.1 Introduction
4.2 On the native status of V1 declaratives
4.3 The syntactic analysis of V1 declaratives
4.4 The distribution of V1 declaratives in OHG
4.4.1 General observations
4.4.2 Verb-initial order in fully rhematic utterances
4.4.3 Verb-initial order with non-accusative verbs
4.4.4 V1 and narrative inversion in Germanic
4.4.5 Sentence particles as blockers of XP movement
4.4.6 V1 and rhetorical structure
4.5 Excursus: the variation between V1 and thô-V2
4.6 The later development of V1 declaratives
4.7 Conclusion
5: The prefield after the Old High German period
5.1 The prefield in Modern German
5.1.1 Syntactic constraints
5.1.2 Information-structural constraints
5.2 The prefield in older stages of German
5.2.1 Middle High German
5.2.2 Early New High German
5.3 On the doubly filled prefield in Early New High German
5.3.1 The data
5.3.2 Implications for the syntactic analysis
5.3.3 Excursus on post-initial connectives
5.4 Summary
6: Adverbial connectives
6.1 Introduction
6.2 The rise and development of aber: from restitutive VP-adverb to CP-connective element
6.2.1 OHG afur as a restitutive and repetitive adverb and its replacement by wieder
6.2.2 From an adverb to a connective adversative element
6.2.2.1 The component of contrast
6.2.2.2 The syntactic position of OHG afur/aber
6.2.2.3 Excursus: Modern German aber
6.2.2.4 OHG afur/aber in the first position
6.2.2.5 Aber in OHG as a translation for Latin second-position autem, vero, and enim
6.2.2.6 Aber in later texts of ENHG and in NHG: CoorP and gapping
6.2.3 Summary
6.3 Allerdings: from VP-adverb to epistemic speech-act adverb and connective element
6.3.1 Allerdings in Modern German
6.3.1.1 Allerdings as an epistemic speech-act adverb
6.3.1.2 Allerdings as an adverbial connective
6.3.2 Allerdings in ENHG: from adverbial to connective element
6.4 Summary
Part II: The Middle Field
7: Introduction to Part II
7.1 The Wackernagel complex (WC)
7.2 Order of NPs
7.3 Negation
8: The Wackernagel complex and pronoun raising
8.1 Introduction
8.2 The Wackernagel complex (WC)
8.2.1 Pronominal clitics
8.2.2 Double agreement
8.2.3 Complementizer agreement
8.2.4 Partial pro-drop
8.3 The diachrony of the Wackernagel complex
8.3.1 Potential diagnostics
8.3.2 Diachronic evidence
8.3.2.1 Synaloepha in Otfrid’s Gospel Book
8.3.2.2 Assimilation
8.3.2.3 Clitic cluster
8.3.2.4 Merged forms
8.3.2.5 Reanalysed forms
8.3.3 Summary
8.4 Emergence of Complementizer Agreement (CA)
8.4.1 A new explanation
8.4.2 Diachrony of Complementizer Agreement
8.5 Pronouns in MHG
8.6 Conclusion
9: Serialization of full noun phrases in the history of German
9.1 Introduction
9.1.1 Remarks on the German middle field
9.1.2 Serialization of noun phrases in Modern German
9.2 German free word order in a broader historical context
9.3 Constituent order in Old High German
9.4 Constituent order in Middle High German
9.5 Constituent order in Early New High German
9.5.1 Perpetuation of the rigid word order state
9.5.2 The rise of the ‘new’ word order freedom
9.6 Reasons for the changes in word order
9.6.1 Reasons for the less flexible Old High German word order
9.6.2 Reasons for the ‘liberation’ of word order in Early New High German
9.7 Summary
10: History of negation in High and Low German
10.1 The negation particle
10.1.1 Old High German
10.1.1.1 The preverbal clitic negation particle
10.1.1.2 Negation strengtheners’ and the beginning grammaticalization of a verbindependentsecond neg-particle
10.1.2 Old Low German
10.1.2.1 The preverbal clitic negation particle
10.1.2.2 ‘Negation strengtheners’
10.1.3 Middle High German
10.1.3.1 The preverbal clitic negation particle
10.1.3.2 The bipartite negation particle
10.1.3.3 The verb-independent negation particle
10.1.4 Middle Low German
10.1.4.1 The preverbal clitic negation particle
10.1.4.2 The bipartite negation particle
10.1.4.3 The verb-independent negation particle
10.1.5 Early New High German
10.1.6 Modern (High and Low) German
10.2 Indefinites in the scope of negation and Negative Concord
10.2.1 Indefinites in the scope of negation and Negative Concord (NC) in OHG
10.2.2 Indefinites in the scope of negation and NC in OLG
10.2.4 Indefinites in the scope of negation and NC in MLG
10.2.5 Indefinites in the scope of negation and NC in ENHG and Modern (High and Low) German
10.3 Analysis
10.4 Conclusion
Part III: The Right Periphery
11: Introduction to Part III
11.1 Mixed OV/VO patterns and the theoretical analysis of word order variation
11.2 The verbal complex
12: The OV/VO alternation in early German: Diagnostics for basic word order
12.1 Introduction
12.2 OV/VO diagnostics
12.2.1 Unambiguous cases
12.2.2 Elements that resist extraposition
12.2.3 Order in the verbal complex
12.2.4 Immobile complex verbs
12.2.5 Quantitative evidence
12.2.6 Basic order = unmarked order
12.2.7 The significance of translations: deviations from the source text
12.3 Conclusion
13: OV versus VO in Old High German: The case of thaz-clauses
13.1 Introduction
13.2 Overview of surface patterns and their derivation in OHG
13.2.1 Analysis within a basic OV order
13.2.1.1 Evidence for pure OV order
13.2.1.2 Orders derived by extraposition within a basic OV structure
13.2.2 Ambiguous surface orders
13.2.3 Surface orders involving obligatory verb fronting
13.3 Conclusion
14: Prosodic and information-structural factors in word order variation
14.1 Introduction
14.2 Word order variation in older Germanic
14.3 Information-structural conditions on word order in OHG
14.3.1 The position of background versus new-information focus
14.3.2 The position of contrastive information
14.3.3 Theoretical implications
14.3.4 Discussion and supporting evidence
14.4 Conclusion
15: Periphrastic verb forms
15.1 Periphrastic verb forms as an example of grammaticalization
15.2 Periphrastic verb forms in the history of German
15.2.1 The stock of periphrastic verb forms
15.2.2 Some forms in detail: the Old High German forms
15.2.3 The future forms: a new wave of grammaticalization in Middle High German
15.3 Permutation in the verbal complex
16: On the history of the IPP construction in German
16.1 The rise of the IPP construction
16.2 Historical origin and theoretical accounts of the IPP construction
16.2.1 IPP from ge-less past participle
16.2.2 IPP as a true infinitive
16.3 Related phenomena: IPP as one repair strategy among others
16.4 Order of verbs in IPP constructions and syntactic analysis
16.5 Summary
17: The ACI construction in the history of German
17.1 Introduction
17.1.1 The ACI in Latin
17.1.2 The ACI in Modern German
17.2 The ACI in Old High German
17.3 The ACI in Middle High German
17.4 The ACI in Early New High German
17.5 Conclusion
Sources
References
Index
Recommend Papers
![Word Order Change (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198747307, 0198747306](https://ebin.pub/img/200x200/word-order-change-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198747307-0198747306.jpg)
![Verb Movement and Clause Structure in Old Romanian (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198736509, 0198736509](https://ebin.pub/img/200x200/verb-movement-and-clause-structure-in-old-romanian-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198736509-0198736509.jpg)
![Syllable and Segment in Latin (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780199660186, 0199660182](https://ebin.pub/img/200x200/syllable-and-segment-in-latin-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780199660186-0199660182.jpg)
![The Development of Latin Clause Structure: A Study of the Extended Verb Phrase (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198759522, 0198759525](https://ebin.pub/img/200x200/the-development-of-latin-clause-structure-a-study-of-the-extended-verb-phrase-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198759522-0198759525.jpg)
![Verb Second in Medieval Romance (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198804673, 0198804679](https://ebin.pub/img/200x200/verb-second-in-medieval-romance-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198804673-0198804679.jpg)
![Parameter Theory and Linguistic Change (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780199659203, 0199659206](https://ebin.pub/img/200x200/parameter-theory-and-linguistic-change-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780199659203-0199659206.jpg)
![Constructionalization and Constructional Changes (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780199679898, 0199679894](https://ebin.pub/img/200x200/constructionalization-and-constructional-changes-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780199679898-0199679894.jpg)
![Syntactic Reconstruction and Proto-Germanic (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198712299, 0198712294](https://ebin.pub/img/200x200/syntactic-reconstruction-and-proto-germanic-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198712299-0198712294.jpg)
![Clause Structure and Word Order in the History of German (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198813545, 0198813546](https://ebin.pub/img/200x200/clause-structure-and-word-order-in-the-history-of-german-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198813545-0198813546.jpg)
- Author / Uploaded
- Agnes Jäger (editor)
- Gisella Ferraresi (editor)
- Helmut Weiß (editor)
File loading please wait...
Citation preview
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Clause Structure and Word Order in the History of German
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OXFORD STUDIES IN DIACHRONIC AND HISTORICAL LINGUISTICS GENERAL EDITORS
Adam Ledgeway and Ian Roberts, University of Cambridge ADVISORY EDITORS
Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge RECENTLY PUBLISHED IN THE SERIES
The Rise and Fall of Ergativity in Aramaic Cycles of Alignment Change Eleanor Coghill Portuguese Relative Clauses in Synchrony and Diachrony Adriana Cardoso Micro-change and Macro-change in Diachronic Syntax Edited by Eric Mathieu and Robert Truswell The Development of Latin Clause Structure A Study of the Extended Verb Phrase Lieven Danckaert Transitive Nouns and Adjectives Evidence from Early Indo-Aryan John J. Lowe Quantitative Historical Linguistics A Corpus Framework Gard B. Jenset and Barbara McGillivray Gender from Latin to Romance History, Geography, Typology Michele Loporcaro Clause Structure and Word Order in the History of German Edited by Agnes Jäger, Gisella Ferraresi, and Helmut Weiß For a complete list of titles published and in preparation for the series, see pp. –
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Clause Structure and Word Order in the History of German Edited by AGNES JÄGER, GISELLA FERRARESI, A N D HE L M U T WE I ẞ
1
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
3
Great Clarendon Street, Oxford, OX DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © editorial matter and organization Agnes Jäger, Gisella Ferraresi, and Helmut Weiß © the chapters their several authors The moral rights of the authors have been asserted First Edition published in Impression: All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press Madison Avenue, New York, NY , United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: ISBN –––– Printed and bound by CPI Group (UK) Ltd, Croydon, CR YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Contents Series preface Preface and acknowledgements List of figures and tables List of abbreviations List of contributors . Introduction Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
vii viii ix xi xv
Part I. The Left Periphery . Introduction to Part I Svetlana Petrova
. Origins of verb-second in Old High German Katrin Axel-Tober
. Verb-initial declaratives in Old High German and in later German Svetlana Petrova
. The prefield after the Old High German period Augustin Speyer and Helmut Weiß
. Adverbial connectives Gisella Ferraresi
Part II. The Middle Field . Introduction to Part II Gisella Ferraresi and Agnes Jäger
. The Wackernagel complex and pronoun raising Helmut Weiß
. Serialization of full noun phrases in the history of German Augustin Speyer
. History of negation in High and Low German Anne Breitbarth and Agnes Jäger
Part III. The Right Periphery . Introduction to Part III Eric Fuß
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
vi
Contents
. The OV/VO alternation in Early German: Diagnostics for basic word order Eric Fuß
. OV versus VO in Old High German: The case of thaz-clauses Svetlana Petrova and Helmut Weiß
. Prosodic and information-structural factors in word order variation Roland Hinterhölzl and Svetlana Petrova
. Periphrastic verb forms Augustin Speyer
. On the history of the IPP construction in German Agnes Jäger
. The ACI construction in the history of German Augustin Speyer
Sources References Index
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Series preface Modern diachronic linguistics has important contacts with other subdisciplines, notably first-language acquisition, learnability theory, computational linguistics, sociolinguistics, and the traditional philological study of texts. It is now recognized in the wider field that diachronic linguistics can make a novel contribution to linguistic theory, to historical linguistics, and arguably to cognitive science more widely. This series provides a forum for work in both diachronic and historical linguistics, including work on change in grammar, sound, and meaning within and across languages; synchronic studies of languages in the past; and descriptive histories of one or more languages. It is intended to reflect and encourage the links between these subjects and fields such as those mentioned above. The goal of the series is to publish high-quality monographs and collections of papers in diachronic linguistics generally, i.e. studies focusing on change in linguistic structure, and/or change in grammars, which are also intended to make a contribution to linguistic theory, by developing and adopting a current theoretical model, by raising wider questions concerning the nature of language change, or by developing theoretical connections with other areas of linguistics and cognitive science as listed above. There is no bias towards a particular language or language family, or towards a particular theoretical framework; work in all theoretical frameworks, and work based on the descriptive tradition of language typology, as well as quantitatively based work using theoretical ideas, also feature in the series. Adam Ledgeway and Ian Roberts University of Cambridge
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Preface and acknowledgements The passion for diachronic generative syntax of German we share with a number of scholars, many of whom were members of the former network ‘Historische Syntax (HiSyn)’ funded by the German Research Foundation (Deutsche Forschungsgemeinschaft, DFG), whose support is hereby gratefully acknowledged. The present volume is one of the major results of the intense and fruitful collaboration within this network. About half a dozen research meetings over the course of several years provided the platform for concentrated discussion and sustained stimulation of diachronic generative syntactic research on German, one visible outcome of which is this book. We gratefully acknowledge the support and input from our guest researchers, Karin Donhauser, Elvira Glaser, Rosemarie Lühr, and Marga Reis, as well as from those members of the network who could not contribute a chapter to this volume but, by sharing their thoughts and ideas or with their helpful comments on earlier versions of the papers, contributed a great deal to this volume: Ulrike Demske, Jürg Fleischer, Agnes Kollmer, and Oliver Schallert. We would like to thank all the authors for their valuable contributions and their willingness to revise their papers time and again, rendering this, we believe, a highly consistent volume. We are indebted to the series editors Adam Ledgeway and Ian Roberts for their kind support, to two anonymous reviewers for their extremely constructive and helpful comments, as well as to everyone who assisted us throughout the project, including Julia Steer and Vicki Sunter from Oxford University Press, Kim Birchall, Manikandan Chandrasekaran, Ruth Durbridge, Sophia Oppermann, and Susanne Trissler. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß Cologne/Bamberg/Frankfurt am Main, February
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
List of figures and tables Figures . St. Gallen Cod. ; , –
. Development of double prefield filling between and
Tables . The topological model (or field model) of the German clause
. The distribution of V versus thô-V according to the lexical class of the verb
. Clauses with two competitors for the prefield in MHG (Bavarian)
. Clauses with three competitors for the prefield in MHG (Bavarian)
. Clauses with two competitors for the prefield in MHG (East Central German)
. Clauses with three competitors for the prefield in MHG (East Central German)
. The ENHG corpus
. Clauses with two competitors for the prefield in ENHG
. Clauses with three competitors for the prefield in ENHG
. Additional texts for the period –
. Percentage of double prefield filling in declarative main clauses between and
. Position of aber in Isidor, Tatian, and Notker
. Corpus of texts from to the early twentieth century
. Corpus of written and spoken NHG
. Allerdings in the first position in main clauses
. Occurrences of allerdings in the St. Galler Tagblatt () and in the Berliner Morgenpost ()
. Synaloepha in Otfrid’s Gospel Book
. Disyllabic versus trisyllabic verbs
. Pronominal subjects versus adverbs
. Conjunctions
. Two types of CA dialects
. Variation of relative order of IO and DO in OHG
. Variation of relative order of IO and DO in OHG, varying dependency from Latin
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
x
List of figures and tables
.
Order of objects in Otfrid’s Gospel Book
.
Order of objects in OLG texts
.
Texts used (Bonner Frühneuhochdeutschkorpus)
.
Relative order of IO and DO in ENHG texts
.
Ratio of clauses including ni (or ne) in OHG negated clauses
.
Ratio of clauses including ni or ne in OLG negated clauses
.
Ratio of clauses including en/ne in MHG negated clauses
.
Bipartite neg-particle en/ne . . . niht in MHG negated clauses
.
Ratio of neg-particle niht in MHG negated clauses
.
Ratio of clauses including en/ne in MLG negated clauses
.
Bipartite neg-particle en/ne . . . niht in MLG negated clauses
.
Neg-particle niht as the only neg-marker in MLG negated clauses
.
Distribution of negation patterns A, B, and C in OHG
. Distribution of negation patterns A, B, and C in OLG
. Distribution of negation patterns A, B, and C in MHG
. Distribution of negation patterns A, B, and C in MLG
. Features of German neg-markers: Multiple Agree approach
. Features of Low German neg-markers: Pairwise Agree approach
. Features of High German neg-markers: Pairwise Agree approach
.
Position of major constituents in embedded clauses of the Tatian: OHG versus Latin
Number and percentage of word order patterns in thaz-clauses in the corpus of Weiß ()
.
Examples for periphrastic verb forms with first attestation
.
Ordering variants in the MHG verbal complex, depending on type of periphrastic verb form
.
IPP verbs and verb class in the history of German
.
Verb types governing ACI in ENHG
.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
List of abbreviations ABL
ablative
ACC
accusative
ACI
accusativus cum infinitivo
ACP
accusativus cum participio
Adv
adverb
AgrXP
agreement phrase
AP
adjective phrase
AspP
aspect phrase
Aux
auxiliary
C
complementizer
CA
complementizer agreement
COMP
complementizer
ContrP
contrastive topic phrase
CoorP
coordination phrase
CP
complementizer phrase
DA
double agreement
DAT
dative
DBH
Double Base Hypothesis
DEP-IO
output is dependent on input
DL-Top
dislocated topic
DO
direct object
DP
determiner phrase
DS
Deep Structure
DU
dual
ECM
Exceptional Case Marking
EME
Early Middle English
ENHG
Early New High German
EPP
Extended Projection Principle
FamP
familiarity topic
FEM
feminine
FinP
finiteness phrase
FocP
focus phrase
FOFC
Final-over-Final Constraint
ForceP
force phrase
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
xii
List of abbreviations
FUT
future tense
GEN
genitive
I
inflection
IDENT-IO
output is identical to input
IMP
imperative
IMPERF
imperfect tense
INDEF
indefinite (pronoun)
iNeg
interpretable negative feature
INF
infinitive
INFL
inflection
INTERJ
interjection
IO
indirect object
IP
inflection phrase
IPP
infinitivus pro participio
iQ
interpretable quanitificational feature
KontrP
Kontrast phrase
L
lexeme
LAT
Latin
LF
logical form
LSB
left sentence bracket
MASC
masculine
ME
Middle English
MF
middle field
MHG
Middle High German
MLG
Middle Low German
Mod
modal
NC
Negative Concord
NEG
negation
Neg-Doubling
Negative Doubling
NegP
negation phrase
neg-particle
negation particle
Neg-Spread
Negative Spread
NHG
New High German
n-indefinite
negative indefinite
NOM
nominative
NP
noun phrase
NPI
negative polarity item
NSR
northern subject rule
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
List of abbreviations O
object
OBJ
object (in glossing examples)
OE
Old English
OHG
Old High German
OLG
Old Low German
ON
Old Norse
OT
Optimality Theory
OV
object-verb order
PAST
past tense
PERF
perfect tense
PIE
Proto-Indo-European
PF
Phonetic Form
PL
plural
PostF
post field
PP
prepositional phrase
PPI
participium pro infinitivo
PRES
present tense
PRET
preterite
PTC
particle
PTCP
participle
QR
quantifier raising
REFL
reflexive (pronoun)
RSB
right sentence bracket
S
subject
SBJV
subjunctive
SceneP
scene setting phrase
SG
singular
ShiftP
aboutness-shift topic phrase
SOV
subject-object-verb order
Spec
specifier
SUBJ
subject (in glossing examples)
SUPERL
superlative
SVO
subject-verb-object order
T
tense
TopP
topic phrase
TP
tense phrase
uNeg
uninterpretable negative feature
uQ
uninterpretable quantificational feature
xiii
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
xiv
List of abbreviations
V
verb
V
verb-initial
V
verb-second
V
verb-third
Ve
verb-final
Vfin
finite verb
Vinf
non-finite verb
VMain
main verb
VO
verb-object order
vP
little v phrase
VP
verb phrase
VPR
Verb Projection Raising
v-Prt
verb particle
VR
Verb Raising
WC
Wackernagel complex
WP
Wackernagel position
XP
phrase with variable categories
YP
phrase with variable categories
¬Op
negative operator
, ,
first, second, third person
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
List of contributors K ATRIN A XEL -T OBER is professor of German Linguistics at the University of Tübingen, Germany. Her research focuses on the synchronic and diachronic syntax of German. She has published the books Studies on Old High German syntax: left sentence periphery, Verb placement and verb-second (Benjamins, ) and (Nicht-)kanonische Nebensätze im Deutschen: synchrone und diachrone Aspekte (de Gruyter, ) as well as several articles on sentence structure, complementizers, null subjects, and modal expressions in historical and modern German. A NNE B REITBARTH is lecturer in historical German linguistics at Ghent University, Belgium. She has published widely on issues in the historical syntax of—and language change in—High and Low German, and also Dutch and English. Besides a monograph on the development of negation in Old Saxon and Middle Low German (OUP, ), she is the (co-)editor of several volumes on language change in the domains of negation and polarity, as well as on continuity and change in grammar. She is one of the PIs of the infrastructure project ‘Parsed Corpus of Historical Low German’, and PI of a research project on the post-cyclic development of negative markers in Continental West Germanic. G ISELLA F ERRARESI is professor of German linguistics and German as a foreign language at the University of Bamberg, Germany. She has (co-)edited several volumes on language change and language contact, and published two monographs and many articles on language change, grammaticalization, Gothic syntax, language contact, and second-language acquisition. Her current field of research includes topics such as particles and clause structure, connectives, and aspectuality from a diachronic and acquisitional perspective. E RIC F U ẞ graduated from the Goethe-University Frankfurt and held positions at the Universities of Stuttgart, Frankfurt, and Leipzig. He is currently senior researcher at the Institute for the German Language in Mannheim, Germany. He has written three monographs and (co-)edited several volumes of articles. His primary research interests are language change, linguistic variation, and the interface between syntax and morphology. R OLAND H INTERHÖLZL is associate professor of German linguistics at the University Ca’Foscari in Venice. His research interests comprise diachronic and comparative Germanic syntax and issues in the interface between syntax, phonology, and pragmatics. He has published a monograph on comparative syntax entitled Scrambling, remnant movement, and restructuring in West Germanic and co-edited a volume on Language change and information structure. His current interests concern the properties of embedded V, the syntax and pragmatics of modal particles, and the role of subjects and expletives in the clause. A GNES J ÄGER is professor of historical German linguistics at the University of Cologne. Her research interests include diachronic syntax and its interfaces with semantics and morphology as well as theories of language change. She has published several articles and a monograph on the history of negation in German and co-edited a volume on language change at the
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
xvi
List of contributors
syntax–semantics interface. Her current research focuses on the historical development and dialectal variation of comparison constructions. S VETLANA P ETROVA is professor of historical German linguistics at Bergische Universität Wuppertal, Germany. She has studied the role of information structure in language variation and change, with a particular emphasis on early Germanic, and has a long-standing interest in corpus linguistics and formal diachronic syntax. Her current fields of research comprise the meaning and use of various types of nominal expressions, including pronouns, in the history of German, as well as the development of the determiner system of German. A UGUSTIN S PEYER is professor of German linguistics at Saarland University, Saarbrücken. His research interests focus on the syntax of historical German and of German dialects. The main topics of his research are serialization, subordination, and infinitival syntax. He has written textbooks on the Germanic languages and the history of the German language as well as a monograph on prosodic aspects of the development of topicalization in the history of English, co-edited a volume on co- and subordination and is co-editor of a series within the ZDL monographs on dialect syntax. H ELMUT W EI ẞ is professor of historical German linguistics at the Goethe-University Frankfurt a.M. He is one of the leading experts in the syntax of German dialects and has co-edited the Syntactic Atlas of Hessian dialects (http://www.syhd.info/startseite/) and a volume on Bavarian syntax. He is the author of the Syntax des Bairischen () and he has published many studies in journals and books on topics such as complementizer agreement, negative concord, possessive constructions, and pronominal syntax.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
1 Introduction A G N E S J Ä G E R , G I S E L L A F ER R A R E S I , AN D H EL M U T W E I ẞ
. Aims of the volume The scientific enterprise of historical German syntactic research goes back at least as far as to the nineteenth and early twentieth centuries with the seminal works of the Neogrammarians and other scholars including Grimm (, Vol. IV); Paul (, ), and Behaghel (–). Since then, a number of detailed studies on historical German syntax have appeared (Wunder ; Bolli ; Näf ; Borter ; Robinson ; Dittmer and Dittmer among others) as well as descriptive overviews of diachronic German syntax and the syntax of individual historical stages of German (e.g. Dal ; Betten ; Admoni ; Ebert , ; Prell ; Schrodt , and most recently the descriptive introductory textbook by Fleischer and Schallert , which briefly also touches on some theoretical aspects).1 Since the mid-twentieth century, the framework of generative grammar has proved very successful and influential within syntactic theory. Already since the late seventies and increasingly since the nineties, it has been fruitfully applied to the study of different stages of various Indo-European languages (Hale a). Especially the Old Romance (Vance ) and the Old Germanic languages, notably Old English (OE)—but also Old Norse (ON) etc.—have been thoroughly studied and discussed (see Lightfoot , , ; Sigurðsson ; Kiparsky ; Kroch and Taylor ; Pintzuk ; Delsing ; Hróarsdóttir a; Roberts and Roussou ; van Gelderen b; Roberts ). This research during the last forty years has greatly enriched our understanding of historical syntax and mechanisms of syntactic change. For a long time, there were only scarce studies on historical German syntax within the generative framework (Lenerz ; Tomaselli ). However, over roughly the last decade, there has been a notable rise in new research on historical German syntax
1 For an overview of different strands in recent historical syntactic research of German inspired by typological, sociolinguistic, pragmatic, corpus-linguistic, and generative impulses see also Fleischer ().
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Agnes Jäger, Gisella Ferraresi, and Helmut Weiß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
in a generative perspective with a number of detailed studies on various aspects of the historical German syntactic structure (Demske ; Breitbarth , ; Fuß ; Axel ; Jäger ; Sapp a; Axel-Tober ; Schlachter ; Ferraresi etc. For a recent overview see also Axel-Tober ). This volume presents a state-of-the-art survey of this thriving new line of research by leading scholars in the field, combining it with new insights into the syntax of historical German couched in a generative framework. It is thus the first comprehensive and concise generative historical syntax of German—comparable for instance to Fischer et al. () for historical English—covering numerous central aspects of clause structure and word order and tracing them throughout various historical stages of German, making the crucial results of this recent research available in one volume to a wider international audience. It is thus a contribution to the field of German historical linguistics as well as to diachronic generative syntax. Each chapter aims to combine two crucial aspects of research: one is a solid empirical basis and valid descriptive generalizations with reference to the topological model widely used in more traditional approaches to German syntax; this includes quantitative data surveys as well as a wealth of historical examples.2 The other is a detailed discussion of theoretical analyses of the phenomenon in question. The volume therefore constitutes a valuable reference work for scholars interested in the history of German and diachronic German syntax—both from a more traditional point of view and one of formal syntactic accounts—as it provides a convenient overview of current research on the major issues concerning the diachrony of German clause structure. Since the book is written in English and uses the widely known theoretical framework of generative grammar, it makes the material also accessible to others, including scholars working on the history of other Germanic languages as well as those working on syntactic change in general or on one of the specific syntactic phenomena covered in the various chapters. However, the contributions are not only aimed at researchers in the field, but in giving a basic overview of the respective topics and relating them to more descriptive and traditional accounts, the book is also suitable for academic teaching, either as a central text in courses on historical German syntax or as a source for individual chapters on specific topics covered in courses on syntactic change in general or in other Germanic languages. In order to provide the necessary context and background, we will characterize the main traits of German syntax; introduce the topological model, which is frequently used in more traditional approaches to German syntax; and relate it to generative analyses of German clause structure in the following, thus making the work more accessible both to non-Germanists working in the generative framework as well as to scholars who may be trained in more traditional approaches to the structure of German but less familiar with the generative approach to it. In addition, we will give a
2 All examples from the same historical source texts are consistently quoted from identical editions throughout the volume, as listed in the sources section at the end of the book.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction
brief overview of the history of German and characterize its historical stages. Finally, the organization of the volume is laid out in relation to the discussed theoretical framework.
. The syntax of German, the topological model, and the generative analysis of German clause structure German is a Germanic language and thus closely related for instance to the Germanic Scandinavian languages, but in particular to other West Germanic languages, notably English and Dutch, which therefore constitute an important point of reference in comparison with German and its development, also in this volume. As far as the syntax of Modern German is concerned, it is generally assumed that the basic word order is SOV. (This can be argued to have been the case in German since the beginning of attestation, see Chapters and for discussion.) The SOV order is evidenced in subordinate clauses introduced by complementizers such as dass ‘that’ in the (a) examples of ()–(). It is also the order that children overwhelmingly produce in the early stages of first language acquisition (Schulz and Grimm : ). Furthermore, separable verbal particles, which remain in the basic verbal position, always follow subject and objects even when the finite verb has been fronted, and thus betray the basic word order, cf. auf ‘up’ in (a) and (b). Also non-finite verbs are positioned after subject and objects irrespective of whether the finite verb in the clause has been fronted or not, as illustrated in (a) versus (b)–(d). () a. dass das Kind den Apfel isst that the child the apple eats ‘that the child eats the apple’ b. Das Kind isst den Apfel. the child eats the apple ‘The child eats the apple.’ () a. dass das Kind den Apfel aufhebt that the child the apple up.picks ‘that the child picks up the apple’ b. Das Kind hebt den Apfel auf. the child picks the apple up ‘The child picks up the apple.’ () a. dass das Kind den Apfel essen möchte that the child the apple eat wants ‘that the child wants to eat the apple’ b. Den Apfel möchte das Kind essen. the apple wants the child eat ‘The child wants to eat the apple.’ c. Vielleicht möchte das Kind den Apfel essen. maybe wants the child the apple eat ‘Maybe, the child wants to eat the apple.’
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß d. Möchte das Kind den Apfel essen? wants the child the apple eat ‘Does the child want to eat the apple?’
As visible in the (a) versus (b) examples of ()–(), German, however, displays a so-called asymmetric word order in main and subordinate clauses (OV/V asymmetry): while prototypical3 subordinate clauses show the basic OV or verb-final (verb end/Ve) word order as discussed above and illustrated in the (a) examples of ()–(), the finite verb shows up in the left periphery of the clause in ordinary declarative main clauses, with exactly one constituent (Satzglied) preceding it. Although this constituent is the subject in many cases, as in (b) and (b), the verb may instead be preceded by another constituent such as the object, as in (b), an adverbial, as in (c) etc., largely due to information-structural conditions (see Chapters and ; see also Musan ). German can therefore be characterized as a verb-second (V) language, like Dutch but unlike English. (Again, this feature can be argued to characterize German since its earliest stages, while competing syntactic patterns can also regularly be observed at that time, see Chapters –.) Fronting of the finite verb to the left periphery of the sentence is also observed in imperatives and in questions, as in the yes/no question in (d). In these instances, we can witness so-called verb-initial (V) sentences. The fact that the parts of the predicate in V and V clauses as well as complementizers and the predicate in subordinate clauses generally show up at the left versus right periphery of the clause, respectively; and, so to speak, frame the sentence, along with certain regularities of the syntactic areas or fields (Felder) before, inbetween and, after those elements are descriptively captured in the so-called topological model or field model ((Stellungs-)Feldermodell) of the German clause. This was originally developed by Drach (), later extended, and is widely used in more traditional approaches to German syntax (e.g. Zifonun et al. ; Pittner and Berman ; Wöllstein ; Duden ). Therefore, it also constitutes an important point of reference for the organization of, and the analyses within, this volume. For illustration, the examples discussed above as well as two additional sentences are analysed according to the topological model in Table .. The central descriptive observation regarding verbal placement, viz. the common split into a left-peripheral and a right-peripheral position is reflected in the concept of verbal or more generally sentence frame or brackets (Satzklammer, also Verbale Klammer, Satzrahmen). The parts of a complex predicate in a V or V clause take up the left and the right sentence bracket (linke/rechte Satzklammer), respectively. As discussed above, this holds for combinations of several verbs, e.g. auxiliary or modal and main verb, as in Table . f–h, but also for the verb stem and a separable
3 Verb-final order is one feature of prototypical or canonical subordinate clauses, generally found in dependent clauses introduced by complementizers, relative pronouns/adverbs, or interrogative pronouns/ adverbs. However, less frequently, non-canonical subordinate clauses are also attested in which the finite verb is in left-peripheral position (for a detailed account including the historical development of these patterns see Axel-Tober ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction
T . The topological model (or field model) of the German clause prefield a. b. c. d. e. f. g. h. i. j.
Das Kind Das Kind Den Apfel Vielleicht
left sentence bracket
middle field
right sentence bracket
dass isst dass hebt dass möchte möchte Möchte
das Kind den Apfel den Apfel das Kind den Apfel den Apfel das Kind den Apfel das Kind das Kind den Apfel das Kind den Apfel das Kind du den Apfel
isst
den wenn
aufhebt auf essen möchte essen essen essen essen möchte hast
post field
den das Kind essen möchte
verbal particle, as in Table . d. Complementizers, e.g. dass in Table . a, c, and e, are also positioned in the left sentence bracket. They prevent the finite verb from occurring in a left-peripheral position. This complementary distribution suggests that these elements take up the same syntactic position. As can be seen in Table ., the left sentence bracket thus consists of—at most—one single word: the finite verb or a complementizer. It may also be empty, as in the relative clause in Table . i (‘which the child wants to eat’). The right sentence bracket, on the other hand, may encompass several verbal elements. These then form a verb cluster or verbal complex (Verbalkomplex) in the right sentence bracket, as in Table . e and i (on the diachrony of complex predicates and verbal complexes see Chapter , on changes in the relative order of the verbs within the right sentence bracket see also Chapter and Sapp a). Just as the left sentence bracket, the right sentence bracket may be empty, as in Table . b in which the predicate consists of the single verb isst in the left sentence bracket. The fact that German is a V language is captured in the topological model by stating that the finite verb needs to be in the left sentence bracket in a declarative clause and that the syntactic area before it, the so-called prefield (Vorfeld), needs to be taken up by exactly one constituent (Satzglied), as in Table . b, d, f, and g. Indeed, the prefield is generally restricted to at most one single constituent in Modern German. (This was not generally the case in earlier stages of German, and there are still deviating patterns in Modern German, see Chapters and .) Relative or interrogative pronouns and adverbs such as den in Table . i can also be assumed to be positioned in the prefield as they represent full constituents—in this case, den ‘which’ is the object of essen ‘eat’. They form full syntactic phrases that may also encompass more than one word (e.g. die Frau, [deren Kind] den Apfel essen möchte ‘the woman [whose child] wants to eat the apple’), which again speaks in favour of analysing them as prefield elements rather than as left sentence bracket elements,
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
for instance.4 Furthermore, in present-day dialects as well as historically, combinations of relative or interrogative pronouns and adverbs with complementizers occur.5 As the complementizers take up the left sentence bracket, the relative or interrogative pronoun or adverb must be in prefield position. Unlike the prefield, the syntactic area between the two sentence brackets, the so-called middle field (Mittelfeld), may encompass not just one but several constituents, as illustrated in Table . a, c, e, g, h, and j. In clauses that display the basic SOV word order, i.e. subordinate clauses introduced by complementizers such as in Table . a, c, and e, the predicate is placed in the right sentence bracket while all constituents are positioned in the middle field. Within the middle field, constituents may secondarily appear in an order differing from the basic order due to phonological and semantic-pragmatic reasons. This syntactic re-ordering is also referred to as ‘scrambling’. (On the diachronic development of constituent order in the middle field see Chapter .) Unstressed pronouns are always raised to the left edge of the middle field and in this position, which is called the Wackernagel position (WP), they may interact with the finite verb or the complementizer in the left sentence bracket giving rise to new verbal inflection or, in German dialects, to complementizer agreement (see Chapter on the emergence of these phenomena). Just as the prefield, the middle field may also be empty, for instance if a clause merely consists of a subject in prefield position and a finite verb in the left sentence bracket (e.g. Das Kind isst. ‘The child eats.’). The post field (Nachfeld), i.e. the syntactic area following the right sentence bracket, is empty in most cases in Modern German, as in Table . a–i. One common exception is complex sentences: subordinate clauses frequently occur in the post field—rather than within the middle field or prefield—of their matrix clause, as illustrated with the relative clause in Table . i, which forms the post field of its matrix clause in Table . j (‘if you have the apple [which the child wants to eat]’).6 Occasionally, other constituents—such as adverbials in the shape of complex PPs—occur in the post field, whereas for arguments of the verb and ‘light’ elements— such as simple pronouns or adverbs—it is virtually impossible to be placed in the post field in Modern German. (The placement of constituents in the post field was less restricted in historical German, as discussed in Chapters –; see also Schlachter on information-structural conditions.) As mentioned above, all constituents are generally assumed to be part of the middle field in the basic order. Secondary placement of constituents in the post field is therefore referred to as
4
But see Reis () for an analysis of relatives and interrogatives as occupying the left sentence bracket, and Weiß () for arguments against it. 5 In generative terminology, this phenomenon is generally referred to as doubly filled COMP, see Bayer () or Weiß (a) among others, and Coniglio, Linde, and Ruette () for Old High German (OHG) relative clauses. 6 The alternative order with the relative clause centre-embedded within the middle field next to its head noun is possible but somewhat dispreferred in Modern German (wenn du den Apfel, den das Kind essen möchte, hast).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction
extraposition (also called Ausklammerung/exbraciation), secondary placement in the prefield as topicalization.7 The organization of the present volume is roughly related to the topological model: Part I deals with the prefield and left sentence bracket; Part II with the middle field; and Part III with the right sentence bracket and partly the post field, but also with the question of possible diachronic changes in the order of the verb in its right sentence bracket position relative to other constituents. As discussed above, the topological model nicely captures many central descriptive syntactic facts of German. The generative syntactic analysis of the clause is of course more detailed in depicting the full phrase structure, dependencies between basic and surface positions of elements by syntactic movement etc. (For comprehensive generative analyses of Modern German syntax see Grewendorf ; von Stechow and Sternefeld ; Haider ; Grewendorf ; Sternefeld ; Haider a among others.) Furthermore, as it is widely used in syntactic research on various languages, it makes the analysis of German clause structure and its development comparable to that of other languages. However, the topological model and the generative syntactic analysis should not be viewed as opposing approaches. Rather, there are crucial correspondences between certain positions in the phrase structure and the topological model so that insights from this more traditional approach are also captured within the generative analysis (den Besten ; Sabel ). As German constitutes an SOV language, the German VP is generally assumed to be head-final. Similarly, the German IP is usually taken to be head-final, as the finite verb occurs in final position in canonical subordinate clauses (Grewendorf , ).8 Thus, the clause structure in German looks roughly as given in (), illustrated with the sentence from (b)/Table . b. This phrase structure of the German clause is also largely assumed throughout the present volume, while individual chapters vary for instance in additionally assuming further functional subprojections within the CP or IP layer (following the split-IP/split-CP models/cartographic approach, see Pollock ; Rizzi ; Cinque ).9
7
In addition to the parts of the topological model introduced above, the so-called pre-prefield (Vorvorfeld) and post-post field (Nachnachfeld) are assumed for material that is not integrated in the clause, including constituents co-occurring with anaphoric pronouns in the prefield as in the left-dislocation construction (Linksversetzung) in (i) or the right-dislocation construction (Rechtsversetzung) in (ii). (i) [Das Kind], das möchte vielleicht den Apfel essen. ‘The child, (s)he maybe wants to eat the apple.’ (ii) Vielleicht möchte das Kind den essen, [den Apfel]. ‘Maybe the child wants to eat it, the apple.’ 8 Other authors, for instance Haider (, a), have argued explicitly against the assumption of an IP and in favour of a VP-internal subject position in German. 9 A different stance is taken under Kayne’s () universal base hypothesis according to which the syntactic structure of all languages is underlyingly head-initial (SVO). An analysis based on this approach is presented in Chapter . Similarly, a head-initial IP/TP is assumed in Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
()
CP NPi Das Kindi
C’ C isstj
IP ti
I‘ I tj
VP NP
V tj
den Apfel
prefield
left bracket
middle field
right (post field) bracket
The level of CP is exploited in generative syntactic analyses of German in order to explain the derivation of V and V sentences, as already proposed by den Besten (). The head C is not only the position that is taken up by complementizers in subordinate Ve clauses but also alternatively by the finite verb in V and V clauses in which the verb moves here from its right-peripheral position, thus capturing the complementary distribution of both types of elements. In V clauses, a further constituent is moved to the specifier of CP. SpecCP in other words corresponds to the prefield, C to the left sentence bracket, V and I to the right sentence bracket, while everything after C and before V corresponds to the middle field and everything after V/I to the post field. Just as verb fronting to the left periphery—as well as topicalization of constituents—is derived by moving the respective elements, extraposition is also generally viewed as the result of syntactic movement.10 While typical German V clauses can easily be captured in a structure like (), occurrences of V (verbthird) declaratives, which are rarely found in Modern German but were more common in earlier stages of the language, can be taken as evidence for a split CP layer along the lines of Rizzi () and Grewendorf (), see Chapters and for discussion.
. Overview of the history of German On the basis of written records, the history of German can be traced back over more than a millennium (see Polenz –; Besch et al. –; Salmons among others). It involves a number of intriguing phonological, morphological, semantic, and of course syntactic changes. The earliest attested stage of the language leading eventually to present-day German is referred to as Old High German (OHG). 10 As far as extraposition is concerned, however, the movement analysis is not undisputed, see for instance Haider (a); Schallert ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction
It dates from around to . The textual evidence from this period was overwhelmingly produced in monasteries and for the most part consists of translations of religious and other Latin texts. Much of OHG is evidenced in glosses, which are hardly of use for investigating historical syntax (but see Glaser ). Among the larger texts of classical OHG, there is the relatively free translation of Isidor’s treatise De fide catholica (around ), the translation of Tatian’s gospel harmony—which is closer to its Latin original in deviating from the Latin word order at most within one line of text but not across lines—(around ), and the Gospel Book by Otfrid von Weißenburg (between and ). While the first two constitute direct translations from Latin so that possible influence of the original needs to be taken into account, the latter is a poetic text in which word order may be influenced by metre and rhyme. Most other texts from the classical OHG period consist of merely a few lines or pages. From the late OHG period, there are, however, a number of lengthier translations and commentaries by Notker (Notker Labeo/Notker Teutonicus, around –), e.g. of the psalms and of Boethius’ De consolatione philosophiae. In view of the textual evidence, which consists largely of translations or poetic texts, investigating OHG syntax is a methodologically challenging task (for a detailed discussion see Fleischer ). Historical texts from all periods of German including OHG, but without linguistic annotation, can be accessed online in the TITUS database.11 A valuable new resource in which linguistically annotated versions of all OHG texts are now available is the Referenzkorpus Altdeutsch.12 The central reference books on OHG include the grammar by Braune () with the additional volume on syntax by Schrodt () and the dictionaries of AWB and Schützeichel (). From the ensuing Middle High German (MHG) period, which dates from around to and can easily be phonologically distinguished from OHG by virtue of the typical reduction of unstressed vowels to schwa (Nebensilbenabschwächung, e.g. dia gesti > die geste ‘the guests’), we have a somewhat larger bulk of attestation including famous examples of courtly poetry such as the Nibelungenlied or the works of Hartmann von Aue, Gottfried von Straßburg, Wolfram von Eschenbach, Walther von der Vogelweide and others, but notably also various larger original prose texts, which are methodologically far more suitable for syntactic investigations. Besides earlier online corpora such as the TITUS database mentioned above or the Mittelhochdeutsche Begriffsdatenbank with only very limited search options, the new Referenzkorpus Mittelhochdeutsch has become available very recently, a comprehensive and
11
Further freely available but considerably smaller online corpora with texts from all historical stages of German include the Kali Korpus, which is mostly annotated for verbs only, and the Deutsche Diachrone Baumbank, a small, syntactically parsed corpus consisting of three texts. Another small corpus is the Tatian Corpus of Deviating Examples, which consists of , clauses from the OHG Tatian that differ in word order from the Latin orginal. These are morphosyntactically as well as information-structurally annotated. For a recent overview of historical German corpora see also Dipper (). 12 The Referenzkorpus Altdeutsch as well as the Referenzkorpus Mittelhochdeutsch mentioned below have only become available very recently, mostly during the finishing stages of this volume, and could therefore not be used systematically for the research presented here. Together with the state-of-the-art overviews in this volume, however, these corpora will constitute a valuable basis for future research on historical German syntax.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
linguistically annotated MHG corpus balanced for time periods, dialect areas, and genre. The MHG dictionaries of BMZ and Lexer (–) are also conveniently available online as is the new MHG dictionary MWB, which is still in the process of being compiled. The standard MHG grammar is Paul (). The amount of textual evidence and also the variety of attested genres increases even further during Early New High German (ENHG), which dates from around to . ENHG can be distinguished from MHG phonologically by characteristic diphthongizations and monophthongizations such as [i:] > [aɪ ̯] in mîn > mein ‘my’ or [uo̯] > [u:] in guot > gut ‘good’ etc. During the ENHG period, printing with movable type was invented, which greatly boosted the production and distribution of texts. Due to the spreading of literacy, we also have texts from a broader variety of social groups including the citizens of the sociopolitically and economically increasingly important towns and cities. A number of partly annotated ENHG texts are available online in the Bonner Frühneuhochdeutschkorpus, which is also balanced for time periods and dialect areas. In the near future, it will be superseded by the Referenzkorpus Frühneuhochdeutsch. The ENHG dictionary FWB is under preparation. The ENHG reference grammar is Ebert et al. (). The period of German from around until now is referred to as New High German (NHG). During this stage, a cross-regional uniform standard language developed which was mostly used as a written language and secondarily learned in school. Only since the middle of the twentieth century, this non-dialectal variety of German began to be acquired as a first language, so Modern German or present-day German (Gegenwartsdeutsch)13 is sometimes considered a separate period from NHG (Weiß c). Although early NHG from over years ago is quite different from the current stage of the language, in many ways, language change is impeded by the standardization of the language (Weiß , b, c). However, the regional dialects still pertain, mostly as spoken varieties. Modern dialects therefore form a crucial point of comparison for diachronic linguistics as they have not only preserved certain linguistic patterns extinct from the standard language today but have also developed further independently and are thus more progressive than the standard language in some respects (Weiß , b). Texts from the early NHG period are available online in the corpora of Deutsches Textarchiv and German Manchester Corpus, which is balanced for time periods, regions, and genres. OHG, MHG, and ENHG constitute the predecessors of NHG and thus also of Modern Standard German. Therefore, these historical varieties are at the centre of interest in historical and diachronic German linguistics and also in this volume. At the same time, there are regional varieties of German spoken today but also in earlier times that are referred to as Low German. In High German, which encompasses Central and Upper German, the second (or High German) consonant shift that changed voiceless stops into affricates or fricatives depending on the phonological environment, as in Apfel ‘apple’, Affe ‘ape’, and partly voiced stops into voiceless stops, as in Tag ‘day’ etc., has taken place to a larger or smaller extent, most consistently in
13
In this volume, we will use the term ‘Modern German’.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction
the South. In contrast to this, there is no evidence for this phonological change in Low German spoken in the North. For Low German, too, there is textual evidence for historical language stages. Old Low German (OLG, also referred to as Old Saxon), the earliest stage dating from around –, is attested notably in one longer text, viz. the Heliand epos (around ) and fragments of the Genesis (around ), besides a number of minor texts and glosses. Middle Low German (MLG) dates from around – and is thus, despite what the terminology might suggest, only partly contemporary to MHG, but also overlaps with the ENHG period. It was widely used as a lingua franca throughout the Northern half of Europe at the time due to the Hanseatic trade. This situation has created a rich corpus of legal and administrative texts besides religious and other texts. As mentioned above, from the late seventeenth century onwards, NHG developed into a cross-regional standard language. From the middle of the seventeenth century onwards, NHG replaced MLG as the written language also in the Low German area so that there is hardly any Low German textual evidence anymore, but Low German survived as a mostly spoken regional variety. The central handbooks on historical Low German include Gallée’s () and Holthausen’s () OLG grammars, Sehrt’s () dictionary of Heliand and Genesis, Lasch’s () MLG grammar, the MLG dictionary by Schiller and Lübben (–), and the still incomplete MLG dictionary by Lasch and Borchling (ff). Historical Low German corpus texts are available from the TITUS database and the Referenzkorpus Altdeutsch mentioned above. Furthermore, the Referenzkorpus Mittelniederdeutsch/Niederrheinisch, an annotated corpus of MLG which is balanced for time periods, regions, and genres, as well as the Corpus of Historical Low German, a parsed corpus of OLG and MLG, are currently under preparation. In the present volume, some reference to the diachronic syntax of Low German is made where we have relevant research results. The main emphasis, however, is on the direct predecessors of NHG, viz. OHG, MHG, and ENHG, as discussed above.
. Organization of the volume As indicated in Section ., this book is thematically divided into three parts corresponding to the subdomains of the German clause: the first part is devoted to the left periphery or, in terms of the topological model, the prefield and the left sentence bracket. It deals with such central questions as the origin and development of the verb-second word order in German and competing patterns, notably verbinitial and verb-third clauses, and accordingly the question of the structure and complexity of the C-domain in the history of German. In this context, the development of adverbial connectives is also discussed. The second part is concerned with the middle field, i.e. the domain after the C position and before V. Immediately following C is the position of Wackernagel elements that could already be found in historical stages of German—a phenomenon related to cliticization as well as complementizer agreement. The diachronic discussion of the order of NPs in the middle field reveals that the Modern German relative freedom of word order influenced, not only by structural and prosodic but also by
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger, Gisella Ferraresi, and Helmut Weiß
semantic and pragmatic factors, generally accounted for in generative syntactic theory in terms of scrambling, is a more recent development. The middle field is also the central locus of clausal negation which, however, distinctly changes its syntactic behaviour over time. The third part of the volume deals with the right periphery of the clause, notably the intriguing question of OV or VO basic order in historical German including a discussion of the role of prosody and information structure. Furthermore, the development of periphrastic verb forms and of verb cluster formation is investigated, also covering the diachrony of infinitivus pro participio (IPP) and accusativus cum infinitivo (ACI)—phenomena indicative of verb cluster formation. Each part begins with an introductory chapter which provides an overview of the main questions and phenomena dealt with and puts the individual contributions into context.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Part I The Left Periphery
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
2 Introduction to Part I S V E T L A N A P E T R OV A
The first part of this volume deals with the diachronic developments in the left periphery of the clause in German. In terms of the traditional topological model introduced in Chapter , this is the prefield and the left sentence bracket, in which the lexical complementizer in subordinate clauses and the finite Verb (Vfin) in main clauses are realized. As discussed in Chapter , in generative terms, the left sentence bracket corresponds to the head C of the functional projection CP dominating the entire clause, while the prefield is equivalent to the specifier of this projection, SpecCP. These structural properties account for the verb-second rule in Modern German, restricting the number of phrases preceding Vfin in main clauses to exactly one maximal projection, irrespective of its category type and grammatical function. Given that Vfin is in the right sentence bracket in basic order (see Chapter and Part III) but occupies the left sentence bracket in main clauses, Modern German is generally classified as an asymmetric OV/V language, like Dutch and Frisian (den Besten ; Haider a). It is intriguing how German developed its notable OV/V asymmetry. According to the traditional literature (Braune ; Wackernagel /; Delbrück ; Fourquet ; Behaghel – among others), main and dependent clauses differ in terms of verb placement in the entire history of German, including the earliest stages of the attestation. It is also observed that there is massive evidence for verbsecond orders in main clauses already in the earliest records. This is taken to indicate that the OV/V asymmetry typical for Modern German developed before the time of the earliest written records and is already established at the beginning of the attestation. But some notable diachronic differences are observed as well. For example, it seems that overt filling of SpecCP is less categorical in declarative clauses in early German, leading to a high frequency of verb-initial orders in OHG, as observed by Braune (); Ruhfus (); Diels (); Behaghel (–, Vol. IV); Maurer (, ) among others. Additionally, orders in which two or more constituents precede Vfin in a matrix clause are attested (Behaghel –, Vol. IV: ff), challenging the status of the verb-second constraint in historical German. Finally, it is important to investigate how filling the prefield interacts with information structure and discourse coherence over time, i.e. whether there is change Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Svetlana Petrova . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Svetlana Petrova
in the selection of material in the left periphery of the clause. The individual chapters in Part I provide detailed analyses of these various aspects relating to the rise of the verb-second rule and the explanation of the variation that is found in the domain preceding Vfin in main clauses in historical stages of German.
. The evolution of the verb-second rule In Chapter , Katrin Axel-Tober examines the origins of the verb-second rule in German. She investigates the extent to which the various components of the Modern German verb-second rule, namely (i) verb movement to C and (ii) overt filling of SpecCP via XP-fronting or merge of expletive es ‘it’ are established in the earliest phase of the attestation, i.e. in the OHG period. Axel-Tober shows that the first component of the verb-second rule is essentially related to the emergence of the category of complementizers as a functional category heading the clause. She refers to the seminal work by Kiparsky () who argues that the rise of this category is an important innovation that takes place in the grammar of Proto-Germanic and sets it apart from its Indo-European relatives. There is, however, variation as to how categorically verb movement to C takes place in different varieties of early Germanic. For example, for OE (Old English), it is argued that verb fronting to C is restricted to operator contexts, i.e. to main clauses whose preverbal domain hosts a wh-phrase, the sentential negator ne, or a short adverb of the type of þa, þonne ‘then’ (see van Kemenade ; Kiparsky ). These are exactly the contexts in which Vfin precedes pronominal subjects, while in non-operator contexts, Vfin surfaces below such categories that may be preceded by additional material themselves. Additionally, it has been observed that in OE non-initial root conjuncts, the verb tends to remain in situ (Kiparksy ). In OHG, however, there is massive evidence for verb movement to C in all root contexts. Main clauses in which the verb arguably remains in its basic position are extremely rare (see Axel : as well as Cichosz ). Next to the establishment of the category of complementizers, another potential reason for verb movement has been related to the presence of sentence particles in the left periphery of the clause. The inventory of such particles, generally regarded as markers of clause type and illocution, is best represented in Gothic, attested from the fourth century (Eythórsson ; Roberts ; Ferraresi , ). It has been observed that in Gothic, the verb moves to the left in clauses containing such a particle (Eythórsson ). It has also been assumed that the verb moves because it is attracted by the particle, base-generated in some functional projection in the left periphery. This assumption, however, cannot be maintained. Roberts () has shown that in Gothic, there are cases in which the verb moves to the left but targets another projection, below the one in which the particle is located. This is basically confirmed by data from OHG, which also displays some representatives of the old Germanic system of sentence particles (foremost eno and ia). Axel () has shown that in clauses with particles, the verb targets a functional projection which can be different from the one that hosts the particle. This implies that verb movement in these clauses takes place on independent grounds, and that the particle cannot be considered the trigger of the movement operation. There might, however, be an
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Introduction to Part I
indirect relation between verb movement and the illocutionary potential of clauses containing a sentence particle in the left periphery. In sum, OHG provides massive evidence for generalized verb movement to C in main clauses. Walkden (), detecting this picture in OLG (Old Low German) and also ON (Old Norse), assumes that this is a Proto-Northwest Germanic property that is essential for verb-second grammar. OHG displays fundamental similarities to Modern German also regarding the second component of the verb-second rule: filling of SpecCP. Axel-Tober shows that the position to the left of Vfin in main clauses is open for XPs of all category types, grammatical functions, and informational statuses. More precisely, we find evidence for both operator movement and formal movement of phrases to SpecCP in OHG. Operator movement affects constituents endowed with the features +wh, +topic, +focus. Formal movement, by contrast, takes place when the left-most VPinternal constituent is fronted to SpecCP, without bearing a special semantic feature (Fanselow , ; Frey b). The only class of category that is not attested in SpecCP in OHG yet is the expletive es, which emerges towards the MHG period. As a consequence, existential constructions and impersonal passives, which are among the core contexts displaying expletive es in the prefield in Modern German, appear as verb-initial clauses in OHG (see also Chapter ). Axel-Tober outlines a diachronic scenario accounting for the evolution of expletive es via reanalysis of correlative es in a superficially ambiguous structure. But at the same time, there is data challenging the verb-second character of OHG because more than one phrasal constituent can precede Vfin in a main clause. The most significant patterns violating the verb-second property are those in which a left peripheral sentence particle and another constituent, for example a wh-phrase or an XP, are to the left of Vfin, but also those in which a pronoun intervenes between Vfin and another clause-initial XP. One way of explaining these patterns is to analyse them as instances of an older, Indo-European grammar, which according to Kiparsky () was inherited in Germanic and which provided various extensions to the left of C, for example special positions designated for topics and focus. The idea that the left periphery of the clause hosts several structural positions is worked out in Rizzi’s () influential cartographic approach, splitting the C domain into a cascade of functional projections designated for various clausetype and information-structural features. Axel () argues that Rizzi’s model of the structure of the left-periphery is suitable to account for the non-verb-second orders found in the OHG attestation. She assumes that in these cases, Vfin occupies the lowest functional projection in the complex left periphery, which is Fin, while SpecFinP and other specifiers further left host additional material, yielding the respective surface orders. But this system is already reorganized during OHG times. Axel-Tober outlines the basic diachronic developments in the structure of the left periphery, addressing various reanalysis scenarios accounting for the loss of particles and intervening pronouns in the left periphery. Interestingly, however, clauses with a complex left periphery are also attested in the later stages of High and Low German (see Speyer b; Petrova ; and Chapter of this volume). In Section ., I will return to this issue.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Svetlana Petrova
. Verb-initial declaratives It is a well-known observation that in OHG, verb-initial order represents a highly frequent word order option in all types of matrix clauses, including declarative ones.1 The native status of this word order pattern is indisputable, as it occurs independently of foreign original texts and may be detected in practically any of the early Germanic varieties (see Petrova for a summary and discussion). However, while this pattern is still the canonical one in interrogatives and imperatives, its use is considerably restricted in declarative clauses in Modern German (see Önnerfors for a summary and discussion, and Weiß b for verb-initial declaratives in modern dialects of German). From the point of view of surface order, the presence of verb-initial declaratives challenges the status of OHG as a verb-second language. But verb-initial order can be derived under a specific interpretation of the verb-second constraint. Adopting an analysis according to which the different components of the verb-second rule can apply independently, we can derive verb-initial clauses by assuming that the verb is fronted to C, while movement of an XP to SpecCP fails to apply. There might be different reasons why no XP undergoes fronting to SpecCP, either because this movement operation is less categorical in OHG, or because the landing site of this movement operation is not available but occupied by a silent operator (Santorini ). The literature on verbinitial declaratives mainly adopts the first scenario, trying to identify various factors which block movement of an XP to an empty prefield. There are some structural reasons, for example the non-availability of expletive es in SpecCP in OHG, leading to superficial verb-initial order in the equivalents of sentences requiring expletive es in Modern German, or the extraposition of heavy (lexical or clausal) subjects, a category that prototypically occupies the prefield of the clause (Axel : §..). A further reason for verb-initial order on the surface is argument structure. As outlined in Sasse’s () cross-linguistic survey, clause-initial order is a property of verbs displaying a reduced argument structure, as for example non-agentive and intransitive verbs as well as passive constructions. In such cases, subjects represent underlying objects and therefore display the syntactic behaviour typical of VPinternal arguments. This factor certainly plays a role in OHG as well, as a great number of clauses displaying verb-initial order involve existential predicates, motion verbs, and passive constructions. But verb-initial declaratives in OHG are also attested beyond these contexts, for example in clauses involving pronominal subjects of transitive verbs. This amounts to the conclusion that verb-initial order in declarative clauses correlates with a number of heterogeneous syntactic, semantic, and discourse-related factors. Chapter by Svetlana Petrova provides an overview of the various contexts and conditions triggering verb-initial order in declaratives in OHG. This chapter also summarizes previous proposals on the explanation of verb-first declaratives and argues in favour of an analysis according to which this order is a formal correlate
1 Note that clauses with empty topical arguments in the prefield (‘topic-drop’, see Trutkowski for a recent discussion on Modern German) or elliptical conjuncts in coordination do not classify as verb-initial declaratives in the strict sense of this term.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Introduction to Part I
of all-focus sentences, i.e. of sentences asserting the entire situation, including all participants, as new in the context. It is claimed that such sentences display properties of thetic judgements in a broad sense of this notion (Sasse ), assuming that the division into a topic and comment of the utterance fails to apply in these sentences, even though they may contain some discourse-given material. The development of verb-initial orders in declaratives in the later stages of German is subject to some debate in the literature. Maurer (, ) observes that these orders drop in frequency towards the end of the OHG period, and claims that they are completely missing in the entire MHG attestation, until they re-enter the system of German during the sixteenth century, probably as imitations of Latin humanistic style. Önnerfors () presents data challenging Maurer’s claims and showing that verb-initial order was continuously available in declaratives. Coniglio () studies the conditions triggering this order in MHG. He argues in favour of an underlying similarity in the pragmatic properties of modern and historical German verb-initial declaratives, which depends on the fact that they constitute sentences in which a topic-comment-structure fails to apply.
. Distribution of information-structural categories in the prefield in later stages of German Speyer (b) has shown that although the prefield of strict verb-second languages is considered open for maximal projections of any category type, grammatical function, and information-structural value, the movement of phrases to SpecCP is not random but rather governed by various interrelated constraints. Very generally, these constraints amount to the following hierarchy: frame-setter >> contrast >> aboutness topic. This means that in Modern German, the constituent that is most likely to appear in the prefield of a declarative main clause is the one that acts as a frame-setting adverbial, specifying the local and temporal domain for which the main predication holds (Chafe ). If no frame-setter is available in the clause, an element bearing contrastive (narrow focus) interpretation classifies as a candidate to appear clause-initially, and if no such element is available, the aboutness topic, i.e. a referential expression referring to the individual which the sentence is about (Reinhart ), is moved to the prefield. In Chapter , Augustin Speyer and Helmut Weiß investigate how informationstructural categories compete for the prefield of a main declarative clause over time. They observe that the hierarchy described by Speyer (b) for Modern German differs from what is known about the positional realization of information-structural categories in OHG, in which aboutness topics rank relatively high as candidates for the prefield (Hinterhölzl and Petrova ). Speyer and Weiß start out in the MHG period, comparing the constraints for moving categories to the prefield in two dialectal areas: Bavarian and East Central German. The investigation reveals that the constraints governing the positioning of information-structural categories are basically the same as in Modern German. In the following period, in Early New High German (ENHG), these principles are maintained. But the two periods differ in the way in which the prefield is filled. For example, the prefield in MHG displays a certain preference for connectives (see also Chapter )
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Svetlana Petrova
over argument expressions bearing a special information-structural specification. In other words, the expression specifying the discourse relation that the sentence establishes to the previous context is a good candidate to fill the prefield in MHG but not in ENHG. Another difference relates to the use of the option of merging categories in the prefield versus moving categories from the middle field to the prefield. The first strategy is preferred in MHG, while ENHG more often opts for movement. A further phenomenon that Speyer and Weiß discuss are main clauses with complex (multiply filled) prefields in ENHG. The interpretation of these orders in terms of clause structure is challenging, given that there has been a continuous process of consolidation of the verb-second constraint in the grammar of German since late OHG times. Note, for example, that sentence particles no longer play a role, i.e. a strict continuation of the OHG system in which sentence particles give rise to verb-third structures cannot be assumed. Fuß () explains multiple prefields as the result of base-generation in or movement of XPs to different specifiers of one and the same functional head. The loss of such options in later German is explained as the result of a change in the feature specification of the closest specifier, which becomes a multifeature specifier hosting expletives and preventing further movement to the outer specifiers (this is also known as the so-called ‘bottleneck effect’, see Roberts ). Speyer and Weiß refer to previous work by Speyer (a), couched in the framework of Optimality Theory (OT), and propose an analysis within a historically constant, cartographic model of the left periphery. They observe that the categories in the multiply filled prefield in ENHG adhere to the hierarchy of informationstructural categories in the prefields, i.e. frame-setting adverbials precede focus, and focus precedes aboutness topics. Thus, the orders found in the data correspond to the functional projections in the structure of the complex left periphery proposed by Rizzi (), provided that the upper TopP position is reinterpreted as a position hosting frame-setters. But recall that these are exactly the principles that according to Speyer (b) govern the distribution of categories in the prefield of verb-second clauses since MHG times. Assuming that Modern German has a cartographic left periphery as well (see Grewendorf ), the diachronic differences are derived without a change in clause structure, but rather by assuming that an additional— OT-based—constraint, restricting the number of phrases in the left periphery, is ranked higher in Modern German than in the previous stages. This leads to a drastic reduction in the number of main clauses involving complex (multiply filled) prefields today.
. The rise of adverbial connectives In Chapter , Gisella Ferraresi investigates the emergence of adverbial connectives in German, another class of elements that may occur in the left periphery of the clause. This special category comprises elements like allerdings ‘indeed’, immerhin ‘anyway’, or aber ‘but’, which specify temporal, clausal, adversative, and other relations between units in the discourse. In terms of semantic function, adverbial connectives resemble other clause-linking devices such as subordinators (temporal nachdem ‘after’, causal weil ‘since’ etc.) and inter-sentential conjunctions (coordinating und
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Introduction to Part I
‘and’, disjunctive oder ‘or’), but differ from these regarding their positional flexibility; while both subordinators and conjunctions appear in strictly fixed syntactic domains, adverbial connectives can appear both clause-internally, i.e. in the VP, as well as in the prefield. A special class of connectives dubbed post-initial connectives, e.g. aber ‘but’, übrigens ‘by the way’, can also right-adjoin to XPs in the prefield (see Breindl and the discussion in Chapter of this volume). Ferraresi describes the evolution of adverbial connectives via a process of upward reanalysis from adverbials situated low in the VP to sentence adverbs marking the right edge of the upper middle field, which comprises the Wackernagel domain and scrambling domain, and to a higher CoorP exponent. For example, in Modern German, aber ‘but’ is such an element. In the first position, it is syntactically and semantically a coordinating element. In a lower position between the upper and the lower middle field, aber is only semantically a coordinating element; syntactically it is an adverbial one. In addition, aber can be used post-initially, adjoining an element in the prefield. Ferraresi presents evidence suggesting that aber displayed the adverbial functions already in OHG times. But in addition, the lexical equivalent of Modern German aber (OHG afur; MHG aber) displays a variety of meanings, lost today, in which aber is clearly located lower in the VP. One of them is the so-called restitutive meaning, corresponding to English back or Latin rursum; another expresses repetition of events, corresponding to English again and Latin iterum. In Modern German, both the restitutive and the repetitive meaning of OHG afur are expressed by wieder ‘again’. Importantly, the two modern versions of wieder differ in terms of scope and basic position in that restitutive wieder behaves as a low manner adverb modifying V, while repetitive wieder shares the domain of temporal adverbials, taking a higher position above V and its arguments. Ferraresi shows that the different versions of OHG afur share the semantic and syntactic properties of Modern German wieder. Ferraresi concludes that the attestation of aber supports a reanalysis scenario from a low adverb to a high (sentential) adverb and then to a head of a coordination phrase. The majority of adverbial connectives is much younger as their grammaticalization takes place during the ENHG period. One instance is the connective allerdings which originates from the genitive of the DP aller dinge ‘of all the things’ and represents a quantifying adverb meaning ‘completely’ in ENHG times. Ferraresi presents evidence showing that allerdings surfaces below negation in ENHG, which suggests that this element is situated low in the VP. In Modern German, this use of allerdings is obsolete, but allerdings has developed towards an epistemic adverbial meaning ‘for sure, in any case’. This use is related to the function of allerdings as an enforcer of the truth value of the proposition, which can also appear independently, as an elliptical answer confirming a previous utterance. In this configuration, allerdings has been reanalysed as a connective element. However, unlike aber, allerdings is only semantically a connective, also in the first position; syntactically it is still an adverbial element. In sum, adverbial connectives exemplify various diachronic scenarios of upward-reanalysis from low VP-adverbs to high sentential adverbs, and then further to elements occurring in the left periphery of the clause, acting at the extrapropositional level of meaning.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
3 Origins of verb-second in Old High German K A T RI N AX E L - T O B E R
. Introduction This chapter focuses on the structure of the left sentence periphery of the oldest attested stage of the German language, of OHG (for an overview of the historical periods of German see Chapter ).1 It will be shown that even though OHG syntax still showed some features inherited from Proto-Germanic, it had already developed many of the essential ingredients of a verb-second grammar. The chapter is structured as follows: in Section ., I will give an overview of the major characteristics of ProtoGermanic sentence grammar that are relevant for the diachronic assessment of the left sentence periphery in OHG. This section also contains some well-known diachroniccomparative evidence from Gothic and OE. Section . deals with aspects of the nonor pre-verb-second grammar of OHG, i.e. with sentence particles and verb-third effects triggered by prefinite pronouns, dislocated phrases, etc. Section . focuses on the precursor of the modern prefield, i.e. SpecCP,2 and the different mechanisms of filling this position in comparison to Modern German. In Section ., some diachronic issues will be discussed, viz. the development of the different XP-positions in the formerly complex C-domain and the origin of topicalization via movement and the so-called expletive prefield es. Section . contains a short summary.
. Proto-Germanic sentence grammar: major characteristics Since the work by den Besten () and others, the generative analysis of the verbsecond phenomenon in Germanic is based on the assumption that the sentence is a projection of the so-called COMP or C-position—the position where complementizers 1 This chapter is a condensed and updated version of material that was published in Axel (: Chapters –). 2 On correspondences between individual syntactic fields of the German clause and generative phrase structure see Chapter .
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Katrin Axel-Tober . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
are base-generated (see also Chapter ). In sentences without complementizers, the finite verb moves obligatorily into this position and an XP is moved to SpecCP or basegenerated there. In more recent literature, the obligatory filling of SpecCP is sometimes not regarded as an essential ingredient of verb-second. Roberts (: , ), for example, argues that only obligatory V-to-C movement is required for a positive setting of the verb-second parameter. In his very influential paper on ‘Indo-European origins of Germanic syntax’ Kiparsky () proposes the reconstruction of the Proto-Germanic left sentence periphery in (). ()
CP topic
CP spec comp
XP
XP V
C′ S t ... (pro)
t ...
(Kiparsky 1995: 153) As can be seen in (), Kiparsky assumes that there were two operator positions to the left of COMP. At the left edge there was a TOPIC-position that was, however, not targeted by XP-movement, but was a base-generated position that belonged to a kind of left-dislocation/hanging-topic construction. Clause-internally, there was an (overt or silent) resumptive pronoun (= pro in ()) that was associated with the topic-XP. Further to the right, there was a position for focused constituents (= SpecCP in ()). Kiparsky argues that the structure with the two left-peripheral operator positions had been inherited from Proto-European.3 He furthermore provides evidence that there must have been several systematic contexts for verb movement in ProtoGermanic.4 Kiparsky claims that the Proto-Germanic system is best retained in OE, which already exhibited V-to-C movement in operator contexts such as interrogatives (a), negated declaratives (b), and after certain demonstrative adverbs (c):5 () a. Hu begæst þu weorc þin? how go.about you work your ‘How do you go about your work?’ (ÆColl. – after Kiparsky : , ex. (a))
3 As evidence, Kiparsky refers to Hale’s (b) and Garrett’s (, ) work on the left periphery in Vedic Sanskrit and Hittite. 4 Other branches of Indo-European, notably Old Romance, also developed a verb-second(-like) syntax, see Vance () on Old French and Poletto () on Old Italian. 5 In these and the following examples, the finite verb is highlighted by underlining.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober b. Ne mæge we awritan ne mid wordum asecgan ealle þa wundra NEG can we write nor with words express all those wonders ‘We can neither write nor express with words all those wonders.’ (ÆLS . – after Kiparsky : , ex. (b)) c. þa ondwyrdon hie him tweolice. then answered they him doubtingly ‘Then they answered him doubtingly’ (Orosius .– – after Kiparsky : , ex. (a))
In other syntactic contexts, however, there are still residues of verb-final order in OE (see also W. Koopman ; Pintzuk ; Haeberli and Pintzuk ): () He þa his here on tu todælde he then his army in two divided ‘He then divided his army in two’ (Orosius . – after Kiparsky : , ex. ()) Moreover, as will be discussed in Section ., the distribution of pronouns has been taken as evidence that operator-initial sentences and non-operator-initial sentences were structurally different. Work on Gothic syntax by Eythórsson () and Ferraresi (, ) has confirmed some of Kiparsky’s () claims. Also in Gothic there is evidence for more than one XP-position at the left periphery. In example (), for instance, a topic precedes a wh-phrase. () (καὶ ὁ βραχίων κυρίου τίνι ἀπεκαλύφθη;) jah arms fraujins hvamma andhuliþs warþ? and arm.NOM Lord’s whom revealed became ‘and to whom has the power of the Lord been revealed?’ (John : – after Ferraresi : , ex. (d)) Ferraresi adopts Rizzi’s () finely structured C-domain and proposes that the topic occurs in the topic projection and the wh-phrase in the focus projection. She also argues there was a system of left-peripheral clitic and non-clitic particles (see also Eythórsson , ) that were associated with different projections in the C-domain. This can be seen in (). The outmost projection, C in () corresponds to the force projection in Rizzi’s model; here we find the Gothic particle an—a clause-type marker—marking interrogatives. Topics and contrastive topics were marked by the particle iþ and so forth. It would be beyond the scope of this chapter to go into further detail. What is important to note, however, is that such a system is quite different from what we find in verb-second grammars. In verbsecond-systems, clause-typing is generally marked through word-order differences and not particles. Topics and focused phrases are traditionally assumed to target the same projection in verb-second languages, namely the SpecCP-position or, if a cartographic approach (a split C-domain) is adopted, there are constraints to the effect that only one specifier position may be filled to the left of the fronted finite verb.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
() C C C C C C an iþ -u -uh þan nu/auk C = realized as a complementizer or clause-type marker C = Top C = ‘probably’ Foc C C C = ‘Mod realization to be further defined’ (after Ferraresi : ) Furthermore, there was an interaction between clitic particles and verb movement in Gothic (Eythórsson , ; Ferraresi , ; Roberts ). The enclitic interrogative particle ‑u needed material to its left to attach to, as can be seen in (). If there was no fronted XP, the verb had to be fronted to serve as a host for the clitic. This not only happened in independent interrogative clauses as in (a), but also in subordinate interrogatives as in (b). () a. (δύνασθε πιεῖν) maguts=u driggkan? can=PTC drink ‘can you drink?’ (Mk : – after Ferraresi : , ex. (a)) b. (ἄφες ἴδωμεν εἰ ἔρχεται Ἠλίας) let ei saihvam, qimai=u Helias let that see comes=PTC Elias ‘let us see whether Elias will come’ (Mt : – after Ferraresi : , ex. (a)) Regarding verb movement in Gothic more generally, it was shown by Eythórsson (, ; see also Fuß ) that there was a tendency for verb fronting contrary to the Greek original in operator contexts such as imperatives (a), wh-interrogatives (b), and negated clauses (c). This is reminiscent of the OE system, indeed even of the Modern English phenomenon of so-called residual verb-second (Rizzi ). () a. (καθαρίσθητι) wairþ hrains become.IMP cleansed ‘be cleansed’ (Mt :; Mk :; Lk : – after Eythórsson : , ex. (a)) b. (τί ἄρα τὸ παιδίον τοῦτο ἔσται;) hva skuli þata barn wairþan? what shall that child become? ‘what is this child going to be?’ (Lk : – after Eythórsson : , ex. (c)) c. (οὐκ ἐγκακοῦμεν) ni wairþam usgrudjans NEG become.PL discouraged ‘we do not become discouraged’ (. Cor :A – after Eythórsson : , ex. (d))
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
Even though there was a tendency for verb movement in operator contexts, Gothic did not have generalized V-to-C movement as would be the case in a fully fledged verb-second grammar; in Modern German, the finite verb has to be fronted to C in all types of sentences that are not introduced by a complementizer (overt or silent).6 In other words, finite clauses with verb-final order are ungrammatical: () *Leider der Wolf die Großmutter verschlungen hat. unfortunately the wolf the grandmother devoured has ‘unfortunately, the wolf devoured the grandmother’ As to the status of V-to-C in OHG, there is ample evidence that it was already very much generalized at that early stage of the language. In sentences that are not introduced by complementizers, relative pronouns, or the like, the finite verb is generally fronted to verb-second or verb-first position. In translational texts, verb fronting is also systematically realized against the Latin sources: () (Ex abundantia enim cordis. ós loquitur.) fon ginuhtsami thes hérzen sprihhit ther mund of abundance of.the heart speaks the mouth ‘out of the abundance of the heart, the mouth speaks’ (Tatian , –)7 In prose, examples with verb-final order in clauses without complementizers (), are relatively rare.8 () a. (filia mea male a demonio uexatur.) min tohter ubilo fon themo tiuuale giuuegit ist my daughter severely by the devil shaken is ‘my daughter is severely possessed by a demon’ (Tatian , –) b. (testatur idem propheta . . . ) Dher selbo forasago auh in andreru stedi chundida, dhazs . . . the same prophet also in other place proclaimed that ‘the same prophet also elsewhere proclaimed that . . . ’ (Isidor , )9 In her study of (translated) prose, Axel (: ) finds ‘fewer incontestable examples than . . . explicitly or implicitly assumed in the literature’. This has been confirmed by Cichosz’s () quantitative investigation. In her corpus, only % of non-conjoined declarative main clauses in translated prose show verb-final order
6 In some accounts, it has been proposed that subordinate clauses introduced by relative or interrogative pronouns in SpecCP contain silent complementizers in C, see Chapter . 7 Throughout this volume, Tatian is quoted after the recent edition by Masser giving page and line number of the original manuscript (also available online under http://www.e-codices.unifr.ch/de/list/one/csg/). 8 See Lenerz (: –, b); Schrodt (: §); Axel (: –), and Schlachter (: –) for more examples. 9 Some of the English translations of the Isidor examples have been adapted from Robinson ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
(Cichosz : ). In original prose this number is much smaller (= %) (Cichosz : ).10 Cichosz points out that this contrasts with OE, where % of nonconjoined declarative clauses show verb-final order (Cichosz : ). We may thus conclude that in OHG, finite sentences with verb-final order are much less frequent than those with verb-first or verb-second order. Walkden (: ), who furthermore discusses the situation in OLG and ON, even claims that ‘Proto-Northwest Germanic had generalized . . . verb-movement to Fin0 . . . in ordinary declarative clauses.’11 The main focus of the remainder of this chapter is on XP-positions at the left periphery and on XP-movement into the left periphery.
. The pre-verb-second grammar of Old High German: extensions at the left periphery At the left-sentence periphery of OHG, we find a range of phenomena which may be interpreted as residues of a pre-verb-second grammar, namely verb-third effects triggered in particular by hanging nominatives, left-peripheral adverbial clauses, and prefinite pronouns as well as a residual system of sentence particles. .. Dislocated topics and left-peripheral adverbial clauses Kiparsky’s () reconstruction of the Proto-Germanic left sentence periphery comprises a left-peripheral position for dislocated topics. Evidence for such a leftperipheral topic position is also present in OHG. In the eighth and ninth century texts, most examples include so-called pendent or hanging nominatives (Havers ). In this construction, the dislocated phrase bears nominative case, whereas the resumptive pronoun (a personal or demonstrative pronoun), which either occurs in the clauseinternal argument position as in () or is topicalized to SpecCP, bears the case assigned by the governing predicate. The finite verb linearly shows up in third position; structurally it must be in C since, as was mentioned above, the grammar of OHG was already (largely) subject to generalized V-to-C movement.12 () ( . . . & uos igitur nunc quidem tristitiam habebitis.) Inti [ir uuarliho13]i [CP [SpecCP nu] habet iri gitruobnessi.] and you.NOM indeed now have you sorrow ‘and you will indeed now have sorrow’ (Tatian , –) 10
Even though verb-final order in the main declarative clause is much more frequent in translational texts than in original texts, its occurrence cannot always be attributed to Latin influence. In (b), for example, the corresponding Latin source sentence has verb-first order. 11 Fin is the lowest functional head in Rizzi’s () proposal for the fine structure of the C-domain. 12 Henceforth, underlining in examples highlights the finite verb in C. Superscripts are used for coindexing. 13 In OHG, adverbs such as wārlīh(h)o ‘truly, in truth, really, indeed’ and giwisso ‘certainly, surely, really, indeed’ could be used as discourse connectives. This use is still very productive in Modern German with elements such as freilich ‘sure enough, to wit’, schließlich ‘finally, eventually’ etc. Such post-initial discourse connectives can be argued to form one complex constituent together with the preceding phrase (see also Axel : – and Chapters and of this volume).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
The hanging-nominative construction is of old age. As Havers () has shown, it can be found in many old Indo-European languages, but it also occurs in most of the modern Indo-European languages including the Germanic verb-second languages. Besides the hanging nominative, there are also OHG examples in which the left-peripheral topic-phrase does not bear nominative, but a non-nominative case that agrees with the resumptive pronoun: () (Abstulit seuo spolium leoni) [Témo léuuen. dér grece nemeus hîez.]i . . . the.DAT lion.DAT that Greek Nemeus was.called [CconnP [SpecCP démo]i . . . nám er dia hût] that.DAT took he the skin ‘he took the skin off the lion, who was called “Nemeus” in Greek’ (Notker, Consolatio , –) This construction might be the precursor of the modern so-called German left dislocation construction. In present-day German, two types of left-peripheral dislocation construction have to be distinguished: the so-called Freies Thema (Altmann ) or hanging topic left dislocation (Frey ) and the Linksversetzung (Altmann ) or German left dislocation (Frey ). The two constructions differ with respect to the degree of integration of the dislocated phrase into the rest of the clause. In the hanging topic construction, the dislocated phrase is less integrated than in the German left dislocation construction, which is reflected in the following properties (Altmann ; Frey ): firstly, a hanging topic is an independent prosodic unit and is usually set off by a pause from the rest of the clause, whereas a dislocated topic and the following clause exhibit progredient intonation; secondly, with hanging topics the form of the resumptive pronoun varies (e.g. d-pronoun, personal pronoun, or superordinate term), whereas dislocated topics are always resumed by weak d-pronouns; thirdly, if the dislocated phrase is an NP, it often occurs in the nominative case (i.e. as a hanging nominative) instead of in the case of the resumptive element. In the left dislocation construction, by contrast, there is obligatory case agreement. Given these characteristics, it is clear that in the written language, there are cases that are ambiguous between an analysis as hanging topic or German left dislocation. As for OHG, examples such as () with non-nominative dislocated phrases and case agreement can be found sporadically. Without prosodic information, it cannot be decided whether such examples conform to the hanging topic or to the left dislocation pattern. Note also that such examples only occur in late OHG texts and even there they are still rare. The hanging nominative was more frequent. It may well also be the case that the distinction between the two patterns did not exist in OHG and developed only later. Besides constructions with dislocated topics, OHG grammar had also developed the topicalization construction of the modern type via A-bar movement to SpecCP (cf. Section . and Chapters and ), which means that nominative, accusative, or oblique NPs as in () and () are also attested in nondislocated position directly in front of the finite verb (= SpecCP).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
Furthermore, some types of constituents always occur at the periphery of the sentence.14 Notably adverbial clauses are never integrated into the matrix clause in OHG. The preposed temporal clause in (a) occurs in a left-peripheral position and is resumed by the adverb dhanne in the matrix-internal adjunct position. The position directly in front of the finite verb, i.e. the SpecCP-position, is occupied by the topicalized PP mit imu.15 As can be seen in (b), this left-peripheral placement also shows up in those cases where the matrix clause is a subordinate clause itself. Here, the conditional so-clause is placed to the left of the complementizer taz. This placement would be ruled out in Modern German.16 In contrast to (a), there is no overt resumptive adverb. It should be noted that the overt realization of the resumptive adverb was optional—both in the case of independent matrix clauses as in (a) or with subordinate matrix clauses as in (b). () a. (quando certa lege et gyro uallabat abyssos, . . . cum eo eram cuncta componens) [dhanne ir mit ęrcna euua abgrundiu uuazssar umbihringida . . . ], when he with certain law abysses water encircled [mit imu ]i uuas ih dhannei al dhiz frummendi with him was I then all this making ‘when he encircled the water abysses with a certain law . . . , I was then carrying all of this out with him’ (Isidor , ) b. (Nemo miratur flamina chori. túndere litus frementi fluctv.) neíst tés nîomannen vuúnder [sô der NEG.is that.GEN nobody.DAT wonder when the uuínt uuáhet]i táz tiu uuélla Øi án den stád sláhet wind blows that the wave at the shore crashes ‘nobody is surprised that the wave crashes at the shore when the wind is blowing’ (Notker, Consolatio , –)
This paragraph is based on Axel-Tober (: –). See also Axel (: –). In late OHG and in MHG prose texts, potential examples for adverbial clauses in the matrix SpecCPposition (i.e. with the surface order ‘adverbial clause – matrix finite verb’ and no resumptive adverb) are very rare. The OHG prose texts from the eighth and ninth centuries do contain a fair number of examples. As is argued in Axel (, ) and Axel-Tober (: –), however, the matrix clauses in the respective examples exhibit the typical characteristics of verb-first declaratives (see also Chapter of this volume), which strongly suggests that the adverbial clauses are peripherally adjoined to verb-first CPs instead of occupying the SpecCP-positions of verb-second clauses in these cases. Note also that there must have been a ban on moving silent resumptive adverbs to SpecCP. Otherwise surface orders ‘adverbial clause – matrix finite verb’ would be predicted to have been much more frequent. This can be argued to be a reflex of a more general ban prohibiting the movement of phrases with silent lexical heads (Sternefeld : ). 16 In Modern German, extensions of the left periphery to the left of the complementizer are only marginally possible provided there is a clause-internal resumptive pronoun or pro-adverb. In OHG and MHG, by contrast, adverbial and certain types of free relative clauses are attested in this position quite systematically (see also Axel-Tober : Chapter ). 14 15
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
Kiparsky () provides parallel examples from OE.17 Example () gives the structure for the OHG example (a) along the lines of Kiparsky’s proposal. The adverbial clause is not embedded, but a base-generated adjunct to the matrix CP. The clause-internal adjunct position is occupied by a resumptive adverb whose overt realization is optional (this is indicated by pro in Kiparsky’s structure in () in Section .). () [CP [CP dhanne ir . . . abgrundiu umbihringida]i [CP[PP mit imu]i [C’ [C uuasj] ih dhannei/ Øi al dhiz frummendi ti tj ]]] This way of clause combining has Indo-European roots in the so-called correlative diptych, which is discussed e.g. by Haudry (). A Latin example with an adverbial relative clause is given in (). () [CP [DP Quibus [NP diebus ]]i Cumae liberatae sunt obsidione ], which days.ABL Cuma released is from.the.siege [[ isdem diebus ]i . . . Tib. Sempronius . . . prospere pugnat ] the.same days.ABL T.S. successfully fights ‘T. S. won a victory in the same days in which Cuma was released from the siege’ (Livius , , – cf. Haudry ; after Bianchi : , ex. ()) The correlative pattern is also typical for MHG adverbial clauses (Prell : ). The modern, integrated type of adverbial clauses that may occur in SpecCP and in other matrix-internal positions only develops in ENHG times (Axel-Tober : Chapter ). To sum up, in OHG the left periphery could be extended by dislocated phrases and unintegrated clauses. As was shown above, these constructions probably have ProtoGermanic, partly even PIE (Proto-Indo-European) roots. In the following sections, how such extensions can be analysed within a finely structured C-domain will be discussed. .. Left-peripheral sentence particles What we also find at the left periphery in OHG are residues of a system of sentence particles. For example, the particle eno/inu occurred sporadically at the left periphery of interrogative clauses in Tatian (), in Isidor, and in some other eighth and ninth century sources.18 () (numquid pulmentarium hab&is) eno habet ir uuaz muoses. PTC have you anything food.GEN ‘have you anything to eat?’ (Tatian , )
17
See also Donaldson () and Bech and Meklenborg Salvesen () for verb-third effects after initial adverbial clauses and other types of subordinate clauses in Old French and in OE. 18 For a brief overview of OHG textual attestation see Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
The following example from Otfrid contains the particle jā, which was used in declarative clauses to emphatically assert the truth of the utterance (Delbrück ). () ia uuúrtun tóte man ovh lés · quéke sines wórtes PTC became dead men also INTERJ alive his.GEN word.GEN ‘after all, dead people were reawakened through his word’ (Otfrid IV , ) As can be seen in () and (), the finite verb is also fronted to C in sentences introduced by particles. In all the attested examples there is V-to-C movement in sentences with particles. This is a major difference from Gothic where there was no V-to-C movement with non-clitic particles. In contrast to Gothic, V-to-C movement in OHG thus already seems to be an independent mechanism obligatorily triggered in sentences without complementizers very much as in the modern language with its fully fledged verb-second grammar. In Modern Standard German, there are no clause-typing particles anymore.19 Instead clause type is marked through word order differences: yes/no-interrogatives have verb-first order and declaratives verb-second order.20 In declaratives, the prefinite position (= SpecCP) has to be filled via movement or base-generation of an XP (see Section . and Chapter ). Sentence particles no longer occur in this position, in fact this class of lexemes is no longer attested after OHG times. Eno/inu died out during the OHG period. The particle ja also did not survive until Modern German.21 The OHG sentence particles are also attested in sentences with dislocated topics: () (nonne cor nostrum ardens erat In nobis) eno [unser herza]i niuuasizi brinnenti In uns PTC our heart.NOM NEG.was.it.NOM burning in us ‘did not our hearts burn within us?’ (Tatian , –) As can be seen in (), the position of the sentence particle was to the left of the dislocated topic. In a Rizzian-style split C-domain, we may take the sentence particles to occur in the Force projection, a clause-typing projection at the outmost left periphery: ()
[ForceP eno/jā Force0 [DL-TopP XP DL-Top0 . . . [CP XP C0 . . . ]]]
Further to the right, there was a position for dislocated topics (= SpecDL-TopP in ()), which was followed by the traditional SpecCP position at the right edge of the
19
As Weiß (b) shows, many Modern German dialects require the use of a clitic form of the particle denn in (wh-)interrogatives. This particle, however, occurs lower in the clause than the OHG particles (viz. in the Wackernagel position). 20 See Section . and Chapter on verb-first declaratives. 21 The Modern German modal/discourse particle ja is probably of a different origin and is subject to different grammatical conditions than the OHG sentence particle. It is usually unstressed and may not occur at the left periphery in front of the preposed finite verb, but is restricted to a clause-internal position at the left edge of the middle field.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
C-domain that corresponds to the Modern German prefield position. We will see later that the filling of SpecCP in OHG was already very similar to Modern German. .. Pronouns to the left of the fronted finite verb As was shown in the last section, in OHG texts we find several phenomena at the left periphery that are not characteristic of a verb-second grammar. A further case in point is verb-third effects triggered by prefinite pronouns, a phenomenon that no longer occurs in Modern German. In contrast to the extensions of the left periphery discussed so far, the left-peripheral pronouns occur in a position between the fronted XP and the finite verb. The examples in () show pronouns following different fronted XPs: a fronted object-XP, fronted subject pronoun, and a fronted participle (see also Lippert ; Tomaselli ; Harbert ; Axel : Chapter for more types of examples). So it looks as if the position to the left of the pronoun is the Modern German prefield (= SpecCP), whose major characteristic is that it can be filled by an XP of any category or syntactic function (see also Chapter ). () a. (Portas aereas conteram) [Erino portun]i ih ti firchnussu, iron portals I.NOM destroy ‘I will destroy iron portals’ (Isidor , ) b. (Et statuam eum in domo mea) Endi [ih]i inan chistiftu ti in minemu dome and I him.ACC install in my house ‘and I will install him in my house’ (Isidor , ) c. (remittuntur tibi peccata tua.) [forlaaz senu]i dhir uuer dant dino suntea ti forgiven you.DAT become your sins ‘your sins are forgiven’ (Monseer Fragmente I, –) As is well known in the generative literature since van Kemenade (), verbthird effects triggered by personal pronouns are also typical for OE and ME (Middle English): () [hiora untrymnesse]i he sceal ti ðrowian on his heortan their weakness he.NOM shall atone in his heart ‘He shall atone in his heart for their weakness’ (CP , – after Pintzuk : ) In OE, however, prefinite pronouns almost exclusively occur after fronted topics as in (). In so-called operator contexts with e.g. a fronted wh-phrase or demonstrative adverb, the pronouns were placed in post-finite position. For OE and ME, it has therefore been argued by Kroch and Taylor () that verb fronting targets a lower position in topic-initial versus operator-initial sentences. They propose that the
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
finite verb is only raised to the I-domain in topic-initial sentences. Since pronouns can be argued to be positioned at the left edge of the I-domain, they show up in front of the finite verb in this case. In the ‘operator’ contexts, however, pronouns appear after the finite verb, which can therefore be taken to have moved to C in such cases.22 In contrast to OE and ME, the number of OHG examples with prefinite pronouns is very small in the first place (see Axel : Chapter ), which might suggest that this configuration is a syntactic archaism (Eythórsson ). In the vast majority of examples, regardless of whether SpecCP is filled by a ‘normal’ XP or a wh-phrase and the like, OHG pronouns are attested in the same position as their Modern German counterparts, namely in the slot immediately following the finite verb (or the complementizer in subordinate clauses) in C, i.e. in the Wackernagel position:23 () a. (Probauimus . . . ) Chiuuisso chioffanodom uuir nu hear dhazs . . . surely revealed.PL we now here that ‘surely we have now revealed here that . . . ’ (Isidor , ) b. (quando huc uenisti.) uuanne quami thú héra when came you here ‘when did you come here?’ (Tatian , ) Even though prefinite pronouns may be an archaism in OHG, it is nevertheless worthwhile to check in the scarce data available whether pronoun placement is subject to the same syntactic restrictions as in OE. If yes, this would provide further evidence for the hypothesis established on the basis of the OE data that V-to-C movement in early West Germanic was restricted to ‘operator’ contexts (interrogatives, negated clauses, or after fronted demonstrative adverbs) and did not take place in topic-initial sentences. In the OHG data, however, we do not find such a sharp division between topic-initial and operator-initial contexts. Prefinite pronouns are also sporadically attested in wh-interrogatives and in negated sentences: () a. (quid me que̦ritis Interficere . . . ) [zihiu]i mih suoh& ti zi arslahanne . . . why me.ACC seek to kill ‘why do you seek to kill me?’ (Tatian , –)
22 Variants of this ‘two-structure hypothesis’ (Walkden : ) have been proposed by Pintzuk (); Eythórsson (); Fuß () and others. See Walkden (: –) for a recent summary of the different proposals. See also Axel-Tober () for a short comparison of English and German regarding the diachrony of the verb-second parameter. 23 See Chapter on the diachronic development of the Wackernagel position in German.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober b. (Generationem eius quis enarrauit?) christes chiburt [huuer]i sia24 chirahhoda ti? Christ’s birth.FEM who it.FEM.ACC reported ‘the birth of Christ, who reported it?’ (Isidor , ) c. (nihil respondit) [neo uuiht]i im ni ant uurta · ti nothing.ACC them.DAT NEG answered ‘he did not answer them anything’ (Monseer Fragmente XXIV, )
Schlachter (: f) discusses the verb-third examples with prefinite pronouns in Isidor and claims that they involve verb movement to a sentence-medial functional head, namely to Foc, the head of the focus phrase.25 To support her analysis, she argues that V-to-Foc movement is not only attested in main, but also in subordinate clauses introduced by complementizers: () (Numquid de illo salomone creditur prophetatum?) Neo nist zi chilaubanne [CP dhazs [FocP fona dhemu salomone sii that from this Solomon be [InflP ti [VP dhiz chiforabodot]]]] this prophesized ‘that this was prophesized about Solomon’ (Isidor , – bracketing after Schlachter : , ex. ()) Axel (: –), on the contrary, argues that subordinate clauses with a nonfinal finite verb as in () do not involve movement of the finite verb to a sentencemedial functional head, but are instances of so-called verb projection raising (see Chapters –). This variant of non-strict verb-final order is still attested in the modern dialects (Weiß and Schwalm ) and to some extent even in Modern Standard German (see also Prell : – on MHG).26 Her main argument is that
24 This personal pronoun is actually a resumptive pronoun which is co-indexed with the left-peripheral XP to the left of the wh-phrase (a dislocated topic). 25 For a similar analysis of OHG thaz-clauses including a sentence-medial functional projection targeted by the finite verb see Chapter . 26 Schallert () discusses examples such as the following as instances of verb projection raising in Modern Standard German:
er hätte besser nutzen können. (i) (Das war . . . ein Kredit), den that was a loan which he had better used could ‘that was a credit that he could have made better use of ’ (Die Zeit: December , Nr. – after Schallert ) According to the traditional analysis, verb projection raising is viewed as rightward movement of the verbal projection. This analysis, however, is controversial. Sternefeld (: –) proposed that the respective reorderings of material in the verbal complex merely resulted from an inversion of branches in the syntactic tree (see also Chapter of this volume).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
we do not find independent evidence for a leftward movement of the finite verb in subordinate clauses (Axel : –). In contrast to early English, examples with stranded verbal particles etc. in subordinate clauses are not attested. A further problematic aspect in Schlachter’s () account is that it posits a structural asymmetry between main clauses and subordinate clauses as the latter contain a CP, while the former generally do not. In a recent analysis of OE, Walkden (: –) raises a number of objections against the widespread analysis according to which verb-third effects arise from verb movement to a sentence-medial functional head. He comes to the same conclusion as Axel () for OHG, namely that the examples with verb-third or sporadic verbfourth order must be a local phenomenon of the C-domain. Walkden refers to recent work by Westergaard and Vangsnes () on (modern) Tromsø Norwegian. In this verb-second variety, subjects occur prefinitely in wh-questions when they are contextually given, thereby giving rise to verb-third orders. Walkden thus suspects that the verb-third effects in OE and OHG might have been sensitive to information-structural aspects in a similar way. He hypothesizes that the material intervening between the sentence-initial constituent and the finite verb was discourse-given. The prominence of pronouns, notably of subject pronouns, would then receive a natural explanation, as (subject) pronouns are prone to be discoursegiven. He adopts the cartographic proposal for discourse functions by Frascarelli and Hinterhölzl (), according to which the C-domain contains projections for ‘aboutness-shift topics’ (ShiftP), for contrastive topics (ContrP), and for ‘familarity topics’ (FamP): () ForceP > ShiftP > ContrP > FocP > FamP* > FinP (adapted from Frascarelli and Hinterhölzl : ex. ()) Walkden (: ) thus argues that ‘[s]econd position elements in V clauses, including pronouns, occur in SpecFamP, as they are discourse-given, and surface after all left-peripheral material apart from the finite verb; the position can be iterated’. One advantage of the information-structural approach is that it predicts that verb-third/fourth effects could not only be triggered by pronouns, but also by (discourse-given) full XPs. This is more in line with the empirical facts. In OHG there are different ‘classes’ of verb-third constructions (see Axel : – for an overview). However, there is as yet no empirical investigation on the information-structural status of verb-third-inducing elements. For OE the respective evidence has been provided by Bech (). A further question for future research is the relation between the OHG verb-third patterns and the ones in later periods of the language (see Speyer a on ENHG and Modern Standard German and Müller on Modern Standard German, see also Chapter of this volume). To sum up, there is compelling evidence that verb-third effects by pronouns and other material in OHG are a local phenomenon of the C-domain and do not result from verb movement to a functional projection below the C-domain. Whether verbthird is sensitive to information-structural effects parallel to Walkden’s () proposal for OE, is still an open question.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
.. Summary The OHG prose texts from the eighth and ninth centuries provide some residual evidence of a formerly complex C-domain. The linear order of elements in the C-domain can be summarized as follows: ()
particle > dislocated topic > fronted XP/wh-phrase > pron > finite verb > pron precursor of prefield
At the outmost left periphery, there was a position for sentence-typing particles followed by a position for dislocated topics. Further to the right there was a position for fronted XPs and wh-phrases. (Note that there is no evidence that ‘normal’ XPs and wh-phrases targeted different positions in the C-domain.) This is probably the precursor of the modern prefield position. Pronouns could either be placed directly to the right of this position and to the left of the finite verb or to the right of the finite verb. The latter position seems to correspond to the Wackernagel position in Modern German at the left edge of the middle field.
. Characteristics of the verb-second grammar of Old High German .. Sketch of the Modern German verb-second grammar In the last section we have seen that there is still residual evidence for a complex C-domain in early OHG.27 If we compare these findings to Modern German, it should be noted that according to standard assumptions (see Chapter ), there is only one XP-position to the left of the fronted finite verb (= SpecCP/prefield) in sentences without left dislocation constructions and the like; cf. (). Verb movement to C is obligatory in sentences without complementizers or other types of subordinating conjunctions or phrases. There is no left-peripheral position for pronouns and no position for sentence particles. Unstressed pronouns always show up after the finite verb in the Wackernagel domain, which is situated to the right of C (and to the left of the so-called scrambling domain): ()
(dislocated XP) > prefield (SpecCP) > finite verb > pron (= Wackernagel position)
Sentence typing is not encoded by means of sentence-initial particles, but by word order and other morphosyntactic means (e.g. verbal mood). Yes/no-interrogatives and most imperatives have verb-first order; declaratives have verb-second order. In sum, the hierarchical structure of the Modern German sentence can be modelled as follows:28
Section . is a condensed and revised version of Axel (a). See also Axel (: –). As discussed in Chapter , it is controversial whether Modern German sentence structure also includes an I-projection besides CP and VP. Haider (b, a) and Sternefeld (), argue that there is no compelling evidence for an I-projection in German, have proposed models of German sentence structures without it. Note that this particular problem is not really relevant to the topics discussed in this chapter. 27 28
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German ()
The CP-VP model of Modern German sentence structure (along the lines of proposals by Haider b, a; Sternefeld )29
prefield CP pronoun domain (Wackernagel domain)
C’ C°
VP VP
scrambling domain
VP VP VP VP
VP DPnom V’ DPacc
V0
In declaratives, the filling of SpecCP is obligatory in present-day German.30 In topological terms, SpecCP is referred to as the prefield (see Chapter ), sometimes even as the ‘joker field’. The latter term captures well the generalization that SpecCP can be filled by a single phrasal constituent of any category or any syntactic function. In minimalist, feature-based accounts the attraction of CPs to the clause-initial position in German is split up into different operations. Fanselow (, , ) proposes such a feature-based account and argues that SpecCP is targeted by two different types of movement, namely by true A-bar movement (i.e. operator movement) or by so-called ‘stylistic fronting’. Furthermore, SpecCP can be filled by merging the expletive es. In sentences with operator movement, the fronting of the XP is triggered by a feature with semantic-pragmatic content in C; this can either be a [+wh]-feature as in interrogatives (a), a [+focus]-feature (b), or a [+topic]-feature (c).
29 In this structure, an accusative DP is moved to SpecCP. Note that this type of XP-movement is, of course, not restricted to accusative DPs. Any phrasal constituent of any category can move to SpecCP. 30 With the exception of so-called verb-first declaratives (see Önnerfors for an overview). The most frequent subtypes today are narrative, enumerative, and explanatory verb-first declaratives (see Chapter ). See also Weiß (b) on verb-first declaratives in Modern German dialects.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
() a. [Was]i isst Merkel ti am liebsten? what eats Merkel at.the best ‘What’s Angela Merkel’s favourite food?’ b. (A: Was hat er gelesen? ‘What has he read?’) B: [Ein BUCH]i hat er ti gelesen. a book has he read ‘He has read a book.’ (adapted from Fanselow : , ex. ())
[+wh]-feature in C
[+foc]-feature in C
c. (A: Soll ich dir was über Hans erzählen? ‘Shall I tell you something about Hans?’) B: [Diesen Verbrecher]i hat man endlich ti verhaften können. this.ACC criminal has one finally arrest could ‘One has finally been able to arrest this criminal.’ (adapted from Fanselow : , ex. ()) [+top]-feature in C When C has no semantic or pragmatic feature, Fanselow argues that it carries an EPP (Extended Projection Principle) feature that attracts the closest category below C (most often the subject or a high adverb). This type of movement is called ‘stylistic fronting’ by Fanselow or ‘formal movement’ by Frey (b). The following B-examples occur in a null/wide-focus context. As can be seen, SpecCP can for instance be filled by the subject or a scene setting adverb in this type of context. () (A: Was gibt’s Neues? ‘What’s new?’) B: [Peter]i hat ti einen Unfall gehabt. Peter has an accident had ‘Peter has had an accident.’ B’: [Gestern]i/ [Glücklicherweise]i wurden ti die Einbrecher verhaftet. yesterday fortunately became the burglars arrested ‘Yesterday/Fortunately, the burglars were arrested.’ Examples with fronted indefinites or quasi-arguments as in () provide clear evidence for the existence of the phenomenon of stylistic fronting in Modern German as the fronting of such elements cannot have been triggered by a feature with semantic/pragmatic content. The only feature that might serve as a trigger for this movement would be a topic feature, but indefinites and expletives cannot be topics. () (A: Was ist los? ‘What’s wrong?’) B: [Niemand]i kommt ti zur Party nobody come to.the party ‘Nobody is coming to the party.’ B’: [Es]i regnet ti it rains ‘It is raining.’ Besides operator movement and stylistic fronting, the SpecCP position can be filled by merging the expletive pronoun es. This type of es is also referred to as
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
Vorfeld-es (prefield-es). This typically happens in presentational and existential sentences as can be seen in (a). Note that es is not the subject. The verb spielen carries plural inflection and agrees with Wiener Philharmoniker, which is the subject. () a. Es spielen die Wiener Philharmoniker. it.SG play.PL the Vienna Philharmonics.PL ‘The Vienna Philharmonic Orchestra is playing.’ b. Es ist getanzt worden. it is danced become ‘There was dancing.’ A further context for expletive es is impersonal passives as in (b). In German, it is possible to passivize intransitive verbs. Since the subject is suppressed by the passive morphology, there is no argument to front to SpecCP. In this case, expletive es is merged in SpecCP in order to satisfy verb-second. In some dialects, this expletive element is not as common as in the standard language or does not even exist at all. In Bavarian, for example, the participle would be moved to SpecCP via stylistic fronting instead (there is no prefield-es available to be merged in SpecCP and the participle is the highest frontable constituent below C): () tanzti is ti worn danced is become ‘There was dancing.’ (Weiß : )
Modern Bavarian
In the standard language, this word order is only possible if there is contrastive focus on the participle. .. An innovative feature of the Old High German left periphery: the filling of the prefield Interestingly, if we compare the filling of the prefinite position in OHG to that in Modern German, it turns out that OHG already behaved very much like a verbsecond language; not only operator movement, but also stylistic fronting is already attested in the early texts. As was shown in Section ., wh-movement is of old—probably even PIE—origin. In OHG, we find evidence not only for short (a) (= b),31 but also for long wh-movement out of complement clauses, (b): () a. (quando huc uenisti.) uuannei quami thú ti héra when came you here ‘when did you come here?’ (Tatian , )
31
See Petrova and Solf (b) for detailed data on verb placement in OHG interrogative clauses.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober b. (ubi uis paremus . . . ) uuári uuili [thaz wir ti garauuemes] . . . where want that we prepare.SBJV ‘where do you want us to prepare . . . ?’ (Tatian , )
Besides wh-phrases, focused or topicalized phrases also occur in SpecCP in OHG. In the Tatian example (a), the fronted XP (= ein) is focused as it corresponds to the wh-phrase in the preceding question. Example (b) from Notker’s late OHG Consolatio is somewhat different, here we are probably dealing with a contrastive focus as the fronted dative object (= herbeste ‘autumn’) is contrasted with temo lenzen ‘the spring’ in the following sentence. () a. (uuaz ist mir nohnu uuan. ‘What do I still lack? [Tatian , ]) (unum tibi deest.) [ein]i ist ti thir uuan. one.NOM is you.DAT missing ‘one thing you lack’ (Tatian , ) b. [Hérbeste]i gáb kót ti tîe êrâ náls temo lénzen autumn.DAT gave God the honour not the.DAT spring.DAT ‘God has given this award to the autumn, but not to the spring’ (Notker, Consolatio , –) There is furthermore ample evidence that SpecCP could be filled by a topicalized constituent in OHG. In () from the Isidor translation, for example, the PP umbi dhiz is likely to serve as an aboutness topic that refers back to content of the preceding clause. ()
( . . . dhazs dher aerloso man . . . christ, got endi druhtin uurdi chinemnit ‘that the impious man . . . was named Christ, God and Lord’) (Unde et in translatione LXX non habet) [Umbi dhiz]i nist auh so ti chiscriban in dhero sibunzo tradungum about this NEG.is also so written in the Septuagint ‘about this is not written thus in the Septuagint’ (Isidor , )
To sum up, there is indisputable evidence that the prefinite position was already the target of different types of operator movement in OHG. What is more interesting is the question of whether stylistic fronting already existed at that early stage of the language. This would actually be a typical characteristic of verb-second: the SpecCP position has to be filled obligatorily—even if there is no semantic or pragmatic trigger. Stylistic fronting would be an innovative feature and a clear sign of an emerging verb-second grammar. It has been hypothesized that in PIE and Proto-Germanic, the left periphery could only be targeted by operator movement (see Section .). In Gothic, XP-movement to the C-domain was always operator movement (see Section .).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
Interestingly, already in the eighth and ninth century OHG texts, indefinite XPs are attested in SpecCP as, for example, the negative indefinite pronoun neoman in (a). This constituent cannot have been topicalized. In example (b) the modal sentence adverb odouuān occurs in SpecCP. Such modal sentence adverbs cannot function as topics. () a. (nemo bonus nisi unus deus) [neoman]i nist ti guot nibi ein got. no.one NEG.is good except the.one God ‘no one is good except God alone’ (Tatian , ) b. (creder&is forsitan & mihi) [odouuan]i giloubtit ti mir perhaps believed.SBJV me ‘perhaps you would have believed me’ (Tatian , ) The fronting of indefinites or of modal sentence adverbs can neither have been triggered by a topic nor by a focus feature. This type of movement is semanticopragmatically vacuous, which shows that an important step towards verb-second has been undertaken in OHG: the innovation of stylistic fronting. What has not yet evolved in OHG is an expletive pronoun to be merged in SpecCP in presentational and existential constructions (see also Lenerz a; Fleischer and Schallert : –). As can be seen in (), these constructions show verb-first order in OHG: () a. (Ecce sponsus uenit,) See quimit der bruti gomo. INTERJ comes the bridegroom ‘behold, the bridegroom is coming’ (Monseer Fragmente XX, ) b. (vidua autem quædam erat In ciuitate illa) uuas thar ouh sum uuitua In thero burgi was there also some widow in that city ‘there was a certain widow in the same city . . . ’ (Tatian , –) Verb-first order is also attested in examples with subjectless constructions: example (a) contains a zero-place time expression and there is no overt quasiargument or the like to front to SpecCP. Similarly, (b) contains an extraposed subject clause and a correlative/placeholder is not present. In both types of examples, expletive es would have to be merged in SpecCP in Modern German, but since this lexeme has not evolved in OHG, the sentences exhibit verb-first order.32 32 See also Chapter and Hinterhölzl and Petrova (, ) for an information-structural account of verb-first declaratives in OHG.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
() a. (hora erat quasi sexta;) uuas tho zit nah sehsta. was PTC time near sixth ‘it was about the sixth hour’ (Tatian , –) b. (Nam uera sententia est.) Íst áleuuâr [dáz man chît . . . ] is very.true that one says ‘it is a true sentence that one says . . . ’ (Notker, Consolatio , ) The earliest unambiguous instances of expletive es are only attested in MHG. A well-known example is the following verse from the Nibelungenlied, which contains a presentational construction: () ez whs in Bvregonden ein vil edel magedin. it grew in Burgundy a very noble maid ‘there grew up in Burgundy a noble maid’ (Nibelungenlied (C) I, ) To sum up, the filling of SpecCP in OHG is quite innovative. It is not restricted to operator movement, but also comprises stylistic fronting, which is an essential ingredient of a verb-second grammar. On the other hand, there are more instances of declarative verb-first order than expected for a proper verb-second language. This is partly due to the fact that there is no expletive pronoun yet to fill SpecCP.
. Diachronic issues In the last two sections, two contrasting sides of the OHG grammar were presented: on the one hand, it was shown that there are still residues of verb-third (sometimes even verb-fourth) orders triggered mainly by prefinite pronouns and unintegrated adverbial clauses. This might be taken as evidence of a formerly complex C-domain, a hypothesis which is also supported by comparative-diachronic evidence from Gothic and OE. On the other hand, it was argued that the filling of the (immediately) prefinite position, the precursor of the modern prefield (SpecCP), was already very much verb-second-like in that it was not only targeted by operator movement (i.e. by movement of wh-phrases, topics, and focused constituents), but also by stylistic fronting (Fanselow , , ) or formal movement (Frey b), which was semantico-pragmatically vacuous. If we look at these two contrasting sides from a diachronic perspective, the following two questions arise: what was the reason for the reduction of the formerly complex C-domain? And what are the origins of the different ways of filling the prefield (operator movement, stylistic fronting, merge of expletive es)? These questions will be discussed in the following subsections.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
.. Continuity and discontinuity in the C-domain The schema in () summarizes the observations that have been presented in the last two sections from a diachronic viewpoint (the arrows indicate aspects of diachronic developments, the non-arrow lines mark points of diachronic continuity): ()
OHG left periphery (8th/9th century)33
particle > disloc. topic > operator-XP/non-oper.-XP > pron >
finite verb > pron … (Wackernagel)
? Mod odern nGerman German an lleft ft periph phery Modern left periphery (c. . since ce MHG HG) (c. since MHG) disloc. XP > operator-XP/non-oper.-XP/es >
finite verb >
pron … (Wackernagel)
At the left-most edge, there was originally a position for sentence-typing particles. This position disappeared, in fact, the whole lexeme class was lost. Already in OHG, these particles were only used sporadically and in a restricted range of texts. In MHG, they are no longer attested. The next position further to the right was occupied by dislocated topics. The phenomenon of dislocation as such is attested continuously in the history of German. It is, however, unclear whether the grammar of left-dislocated constituents is a case of complete diachronic continuity.34 Since the OHG texts contain unambiguous examples of the hanging topic construction, but only ambiguous potential candidates for German left dislocation, it is not clear whether the latter construction already existed at that early stage of the language. It might be a later innovation, but this is a question that has to be left to future research. The position following the position for dislocated phrases in OHG was probably the (precursor of the modern) prefield. The syntax of this position is partly characterized by continuity, partly by diachronic developments. As for continuity, there is no evidence that topicalization and wh-movement targeted different landings sites at the OHG left periphery as it has been proposed for OE. We have furthermore seen that already in early OHG, this position was the landing site for both different types of operator movement (wh-movement, topicalization, and the fronting of focused
33 This schema only shows the linear order of left-peripheral positions in independent clauses. Subordinate clauses introduced by complementizers exhibit a more reduced left periphery. Cf. Walkden (: ) for a theoretical explanation of this fact. 34 The fact that a complete diachronic continuity is doubtful is indicated by the question mark in the schema in (). As Frey () has argued, hanging topics and XPs dislocated via German left dislocation occupy different positions in Modern German.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober
constituents) and for stylistic fronting. So the development of stylistic fronting—a very important innovation towards a proper verb-second grammar—must have taken place in pre-OHG times. Since we do not have any data, it is, of course, almost impossible to develop an account of how this type of syntactic movement evolved. It may, however, be speculated that there is a diachronic relation between the lexical loss of sentence-typing particles and the evolution of stylistic fronting, i.e. the generalization of XP-movement. After XP-movement had become generalized, yes/ no-interrogatives and declaratives were distinguished primarily by word order (verbfirst versus verb-second), which made the use of particles superfluous.35, 36 This might be the reason for their disappearance. Already in early OHG, sentence-typing particles are only used sporadically and unsystematically. A further essential diachronic innovation that took place in this position and that contributed to the further consolidation of verb-second was the innovation of an expletive pronoun to be merged there in order to satisfy verb-second in cases where no XP was fronted. How this element might have evolved will be discussed in the next section. The position further to the right of the emerging prefield could still host pronouns in OHG, but this position was lost later. From MHG times onwards until today, pronouns are placed in the Wackernagel position following onto the finite verb (or complementizer in subordinate clauses).37 .. Origins of the different ways of filling the prefield As has become clear from the discussion above, the different ways of filling the prefield in German have different time depths. Wh-movement (and maybe also the fronting of focused constituents) can be speculated to even have PIE roots (e.g. Delbrück : ; Kiparsky ). Furthermore, Kiparsky () has provided evidence that a type of dislocation construction with a dislocated phrase basegenerated at the left periphery and resumed by a clause-internal resumptive pronoun must have existed in PIE and Proto-Germanic. Topicalization via movement as in examples like () in Section .., might thus be a younger phenomenon. Interestingly, in the OHG Tatian we find some examples that suggest that topicalization via movement might have evolved from a construction with a dislocated topic. In (a) the DP unser herza ‘our heart’ is resumed by a resumptive pronoun (iz) in postfinite position. The finite verb with the proclitic negation particle ni is in C.38 The clause is a yes/no-interrogative introduced by the sentence-typing particle eno. In (b) we find exactly the same construction: eno followed by an XP (got in this case), followed by the negated finite verb. But there is no resumptive pronoun. The development of an interrogative particle in certain dialects (see fn. in Section ..) is probably a later innovation. 36 Cross-linguistically, interrogative clauses are only rarely marked by verb order. With a few exceptions, this strategy is restricted to European languages (see Dryer and Haspelmath : Chapter ). In German, the typologically unmarked pattern has thus been replaced by the marked pattern. Thanks to Oliver Schallert (p.c.) for pointing this out to me. 37 See Chapter on the emergence of the Wackernagel position in German. 38 For detail on the placement of negation particles in historical German see Chapter . 35
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
() a. (nonne cor nostrum ardens erat In nobis) eno [unser herza]i niuuasizi brinnenti In uns PTC our heart.NOM NEG.was.it.NOM burning in us ‘did not our hearts burn within us?’ (Tatian , –) b. (deus autem non faci& uindictam electorum suorum) eno got nituot giriht sinero gicoronero PTC God.NOM NEG.does justice his.GEN elect.GEN ‘and will not God give justice to his elect?’ (Tatian , –) Sentence (b) is in fact ambiguous between two analyses. On the first analysis, got is base-generated at the left periphery and there is a null resumptive pronoun in the clause as shown in (a). This analysis is possible since OHG was a partial null subject language (Axel : Chapter ). On the second analysis, there is no empty pronoun, but a trace. The XP has moved to the left periphery as shown in (b). () a. (eno) goti ni tuot Øi giriht sinero gicoronero b. (eno) goti ni tuot ti giriht sinero gicoronero (a) could be taken to represent the older structure. Both analyses are motivated by independent evidence. They provide synchronic evidence for a diachronic scenario in which topicalization via base-generation and resumption was reanalysed as a movement construction.39 () [goti ... [VP Øi ...]]
[goti ... [VP ti ... ]]
A reanalysis of an ambiguous surface structure might also lie behind the evolution of the prefield expletive iz/es in late OHG/MHG. This element might have arisen out of a reanalysis of correlative quasi-argumental es in SpecCP (see Axel b for a more detailed account). In this case as well, the OHG Tatian contains examples that may be taken as evidence for such a reanalysis. (a) is an example with a putatively extraposed infinitival clause. It can be hypothesised that there is an empty placeholder/correlative in the matrix-internal argument position. In (b), iz occurs in the prefield to the left of the finite verb in C. () a. (oportuit ergo te committere) gilampfj Øi thir tj [zibifelahanne]i behoved it you.DAT to.commit ‘it behoved you to commit . . . ’ (Tatian , ) 39 In order to further support this hypothesis we would, however, need more evidence that not only subject drop, but also object drop was possible in OHG or in pre-OHG. A further problem is that in the scenario sketched above, a construction with a base-generated topic was reanalysed as a movement construction. This scenario violates the widespread assumption in diachronic research that syntactic change usually involves the reverse development (movement > base-generation/merge) since movement is less economical than base-generation/merge (e.g. Roberts and Roussou : ; van Gelderen : –).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Katrin Axel-Tober b. (oport& enim haec fieri) iz gilimpfiti ti [sus zi uuesanne] it behoves so to be ‘it behoves to be so’ (Tatian , )
The sentence in (b) is structurally ambiguous: Either the iz is a placeholder/ correlative that has been fronted to SpecCP as in (a), or the sentence has an empty correlative in the argument position. In this case, iz would not be an argument coindexed with the extraposed clause, but an expletive element merged in the prefield/ SpecCP position with a structure as in (b). () a. izj gilimpfitk tji tk [ sus zi uuesanne]i b. iz gilimpfitj Øi tj [ sus zi uuesanne]i In Modern German, the realization of the placeholder/correlative is still optional if it occurs in the middle field: ()
. . . weil es/ Ø mich gefreut hat [beizutragen] because it me pleased has to.contribute ‘because it has been a pleasure for me to contribute’
Thus sentences with the overt placeholder/correlative es in the prefield/SpecCP are still ambiguous (see also Sternefeld : –): ()
Es hat mich gefreut beizutragen. (ambiguous: correlative or prefield es)
The OHG example (b) may thus be an early instance of prefield es. Unambiguous evidence, however, can only be found in MHG texts. Both topicalization via movement and the expletive prefield iz/es might thus have arisen by reanalysis of ambiguous surface structures. What is unclear, however, is the evolution of stylistic fronting40—the further essential ingredient of a verb-second grammar.
. Summary To sum up, in this chapter I have demonstrated that in OHG we still find some characteristics of a non- or pre-verb-second grammar. These include residual and partly productive features of a non-conflated C-domain arguably inherited from Proto-Germanic and PIE. On the other hand, there is ample evidence that the precursor of the prefield position already existed in OHG and that it was already a target for both operator movement and stylistic fronting. A major step towards the generalization of XP-movement and thus towards a verb-second grammar was the innovation of stylistic fronting. This must have already taken place in pre-OHG.
40 The question is: how did it come about that an EPP-feature developed that required an XP to be moved to SpecCP even when there was no semantic trigger?
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Origins of verb-second in Old High German
Its evolution remains unclear. A further important innovation was the development of topicalization via movement. This syntactic operation might have evolved from a reanalysis of an old type of left dislocation construction. The prefield iz/es can be speculated to have arisen from a reanalysis of a fronted correlative/placeholder iz. Thus, all these innovations seem to have conspired to shape the German verb-second syntax as we know it today.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
4 Verb-initial declaratives in Old High German and in later German S V ET L A N A P E TR O V A
. Introduction As outlined in Chapter , the word order pattern typical for main declarative clauses in Modern German—namely the verb-second order—is firmly established already during the OHG period (Axel , a). Still, main declarative clauses in OHG frequently display verb-first order (V) on the surface. Consider the examples in () in which the finite verb appears in clause-initial position, with no constituent preceding: () a. (apertum est autem ilico ós eius) gioffonota sih thô sliumo sîn mund opened REFL then immediately his mouth ‘His mouth opened immediately’ (Tatian , ) b. (& surrexit) árstuont siu tho úf arose she then up ‘She arose then’ (Tatian , ) The existence of this pattern is widely accounted for in traditional work on the historical syntax of German (see Braune ; Ruhfus ; Diels ; Behaghel –, Vol. IV; Maurer , among others), and it is also a subject of debate in the later research. Basically, V is attested throughout the entire history of German. For example, it is a historically robust pattern in non-declarative root clauses, such as yes/no-questions and clauses expressing requests and desire. Some examples from the OHG period are given in (), for equivalent structures in Modern German, the reader is referred to Reis (: ).
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Svetlana Petrova . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
() a. (proph&a es tú) Bist thu uuîzago are you prophet ‘Are you a prophet?’ (Tatian , ) b. Stant úf, . . . gihori mír, \joh nim thin bétti mit thir stand up listen to.me and take your bed with you ‘Stand up, listen to me and take your bed with you’ (Otfrid III , ) c. Si sálida gimúati Sálomones gúati be.SG.SBJV.PRES bliss glory.DAT.SG Solomon.GEN.SG piety.DAT.SG ‘May Solomon’s glory and piety be blessed’ (Otfrid Ad Salom. ) The intriguing point is, however, that in OHG, V order is also a relevant pattern attested in declaratives. In Modern German, this is a rare option, strongly restricted in usage (see Section . and the references therein). But in OHG, the picture is different. Here V declaratives are considered a regular pattern in the early and the classical period, declining in frequency towards late-OHG times. This raises questions regarding the status and the interpretation of V order in declaratives in OHG in general, and also regarding the subsequent decline in the use of this pattern in later periods of German. All these questions have been intensively discussed in the literature, but there is still much debate on individual aspects of the analysis, and a number of issues require further research. The present chapter will outline the current state of the discussion, focusing on the questions of the native status, the syntactic analysis, and the distributional properties of V declaratives in OHG. In the final part, the later development of V declaratives will be sketched.
. On the native status of V1 declaratives The early philological literature has treated V declaratives in OHG on a par with all other word order patterns found in the records. Robinson (), however, has raised the question of whether this pattern is of native origin or rather an imitation of foreign style. Focusing on the OHG translation of the Isidor treatise, Robinson () observes that verb-initial declaratives are less frequent in the vernacular part than in the Latin original. Additionally, he discovers that Latin V orders are retained in the translation of biblical citations more often than in those parts of the treatise in which these citations are explained. Sharing the opinion raised by Matzel (: ) that the Isidor translation is more faithful to the Latin word order in the citations than in the explanatory passages (see Fleischer for a summary of the discussion), Robinson concludes that V is a non-native pattern, retained in the translation to indicate foreign speech. Robinson’s explanation faces the following problems: first, note that V is not a speciality of the Isidor translation but a pattern found in the entire OHG attestation.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
It can be shown that in other vernacular translations of the period, V is also regularly employed, independently of the word order of the Latin original. This is illustrated in () from Tatian. In addition, V is also attested in texts composed in the vernacular language, as in Otfrid’s Gospel Book, exemplified in (). Of course, it might be objected that here, the surface order is determined by metrical factors, given the fact that this is a poem written in end-rhyming verses. But in the examples in (), the rhyme does not affect the position of the verb and the subject. Lötscher (: ) even considers that the V pattern is the most natural (or the unmarked) one in matrix declaratives in Otfrid’s Gospel Book, from which all remaining root patterns are derived for various pragmatic purposes: () a. (Phariseus autem coepit . . . dicere) bigonda ther phariseus . . . quedan began the Pharisee speak ‘The Pharisee began to speak’ (Tatian , ) b. (Et pastores erant In regione eadem) uuarun thô hirta In thero landskeffi were then shepherds in that region ‘There were shepherds in that region’ (Tatian , ) () a. Wúahs thaz kind in édili \ mit gote thíhenti grew the child in nobleness with God.DAT.SG prospering ‘The child grew and strengthened in spirit with God’ (Otfrid I , ) b. Gisáhun sie thaz wórt thar \ joh irkántun iz sar saw they the word there and recognized it immediately ‘they [the shepherds] saw and immediately recognized therein the story [that had been told to them before]’ (Otfrid I , ) Second, consider that V declaratives are also found in all remaining early Germanic languages (see Petrova , ; and Petrova and Solf for an overview and examples). According to Eythórsson (: ) and Ferraresi (, ), there are many examples of V declaratives in the Gothic bible. In OE and OLG, we also find plenty of evidence for V in matrix declarative clauses, see () and (), see also Pintzuk (a: ) for OE; Erickson () and Linde () for OLG. Note that verb-initial order in root clauses is also a well-known property of ON prose syntax, accounted for in various grammar books (e.g. Nygaard : §; Heusler : §) and reflected in the syntactic literature (Sigurðsson ; Leiss among others). () a. Secgge ic þe nu eac say I you now also ‘now I also tell you’ (Blickling Homilies XVII – after Petrova : )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
b. Lag thar ên felis bioƀan laid there a stone upon ‘there was a stone upon it’ (Heliand ) These observations suggest that V in main declarative clauses in OHG is a native pattern which the vernacular language shares with all other early Germanic varieties.
. The syntactic analysis of V1 declaratives First of all, it is crucial to distinguish V declaratives from equivalents of Modern German constructions like topic-drop or subject-drop (see Fries ; Trutkowski ), understood in the sense that a discourse-given constituent is omitted in the prefield of the clause, yielding an apparent verb-initial order on the surface, see (a). Such cases are attested in OHG as well, as demonstrated in (b), see also the discussion in Axel (). Note, however, that the term ‘V declaratives’ is not used to refer to such cases. Importantly, in V declaratives as in ()—in contrast to (b)—all argument positions licensed by the governing predicate are overtly filled, foremost that of the subject.1 () a. Q: Was ist mit Peter? what is with Peter ‘What about Peter?’ A: Kann erst später kommen Can only later come ‘[he] can’t come till later’ (Axel : ) b. Q: Gilóubist thu so ih zéllu\ thiu minu uuórt ellu? believe you as I tell the my words all ‘Do you believe in all my words?’ A: sint drúhtin quad si fésti \ in mines hérzen brusti are Lord said she safe in my heard.GEN.SG breast ‘[they] are safe in my heart’s breast’ (Otfrid III , –) Let us address the structural representation of V declarative clauses in OHG. One obvious way of explaining these orders is to subsume them under the common analysis of Germanic verb-second put forward by den Besten (). According to this analysis, verb-second results from two components that apply simultaneously (see also Chapter ), namely (i) verb movement to C (verb fronting) and (ii) XP-movement to SpecCP
1
In Axel (), OHG is considered a partial pro-drop language in which pro is licensed in clauseinternal position, c-commanded by an element with person and number features in C. In main clauses with the verb in C, pro is always post-verbal.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
(XP-topicalization).2 Applying this account to V orders in Modern German leads to the uncontroversial assumption that only one component of the verb-second rule applies in these cases, namely verb fronting to C, while SpecCP remains empty, yielding verb-initial order on the surface. In Chapter , it was argued that the components of the verb-second rule are established already in the earliest OHG attestation (see also Axel , a). This implies that the analysis of Modern German V clauses can be properly extended to OHG, suggesting that the surface order in V declaratives results from the incomplete application of the verb-second rule, in that the verb moves to C but no XP undergoes movement to SpecCP. In line with this analysis, the surface order in V clauses such as those in () can be represented in the way illustrated in (): () a. [CP C gioffonotai [VP sih thô sliumo sîn mund ti]] b. [CP C árstuonti [VP siu tho úf ti]] Evidence for verb movement to C as a component of the verb-second rule comes from the fact that the finite verb is found to the left of pronominal arguments (including reflexive pronouns) which occupy the Wackernagel domain, i.e. the left edge of the middle field immediately below C (see Chapter on the Wackernagel complex in historical German), or above light (monosyllabic) adverbs which are situated between the higher VP (the so-called ‘scrambling domain’) and the lower VP where the argument expressions are base-generated. The condition in (ii) is more problematic. While it is obvious that no constituent is fronted to SpecCP, it is unclear what the reasons for this might be. One option is that the prefield remains empty, as assumed in the analysis of Modern German V orders. But it has also been assumed that the prefield of V clauses hosts a silent operator making SpecCP unavailable for additional movement operations (see Santorini on this proposal). As will become clear from the remaining part of this chapter, the first option is the preferred one in the previous literature on V in OHG. Considering the pros and cons of these two analyses must be left open for further research. Several factors have been considered as blockers of XP-movement to (the empty) prefield. First, there are simply structural factors, for example the lack of expletives as prototypical prefield placeholders in particular constructions, or the extraposition of heavy subjects to the post-field of the clause (Axel ). These conditions, however, do not apply in each individual case. As will be shown in the next section, a series of pragmatic factors have been considered as blockers of XP-movement, all dwelling on the assumption that in general, filling of SpecCP is less categorical in OHG than in Modern German. Consider for example the optionality of the operation of stylistic fronting in the generative literature on Scandinavian V (Holmberg b, ).
2 A special case is base generation in SpecCP, applied to the Modern German expletive es and to other expressions which are ungrammatical in clause-internal position (see Frey b). As for the presence of expletives in historical German, Axel () convincingly argues that this category is not established before the MHG period (see also Chapter of this volume).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
This term relates to a process whereby particles, participles, or other non-nominal constituents undergo movement to satisfy the V constraint, if this operation cannot apply to any nominal constituent in the clause. In work on German, stylistic fronting has been reinterpreted as movement of a non-operator (Fanselow ), or simply of the highest middle field element, to SpecCP, also called formal movement by Frey (b) (see Chapter ). Note that in OHG, we find surface V orders even in clauses containing elements that are appropriate candidates for stylistic fronting or formal movement in Modern German, for example high adverbials (a) or participles (b). This suggests that stylistic fronting does not necessarily apply in OHG to satisfy the V-criterion: () a. (Factum est autem dum irent / & ipse intrauit in quoddam castellum) uuard tho uuarlihho mitthiu sie fuorun / thaz hér selbo happened then truly as they went that he himself gieng in suma burgilun went in some town ‘now it came to pass, as they went, that He Himself entered a certain village’ (Tatian , ) b. (Factum est autem in alio sabbato) uuas thó giuuortan in anderemo sambaztag was then happened in another Sabbath ‘It happened on another Sabbath’ (Tatian , ) These facts allow for a diachronic scenario according to which the consolidation of the various components of the verb-second rule does not take place simultaneously but rather proceeds successively, in a gradual fashion. While verb fronting is already firmly established at the beginning of the attestation, XP movement to the prefield remains optional, and subject to various constraints. The following section outlines various proposals of modelling the variation in the left periphery of main declarative clauses in OHG.
. The distribution of V1 declaratives in OHG .. General observations Numerous facts related to the distribution of V declaratives in OHG can be found in traditional work on historical German word order (Braune ; Ruhfus ; Diels ; Behaghel –, Vol. IV; Maurer , among others), summarized in Schrodt (: f). A basic observation is that V declaratives in OHG are attested in various, rather heterogeneous contexts, and that V is used to satisfy a series of different syntactic, semantic, and discourse-pragmatic requirements. First, it is recognized that V is preferred with a number of semantically nonrelated classes of predicates, such as existential be, as in (), verbs denoting the appearance or disappearance of a referent, as in () and (), verbs of saying, as in (), inchoative predicates, foremost the verb beginnan, as in (), etc.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
()
Svetlana Petrova (vidua autem quaedam erat / in ciuitate illa) uuas thar ouh sum uuitua / in theru burgi was there also certain widow in this town ‘A certain widow also lived there in this town’ (Tatian , )
() (Et ecce angeli accesserunt) giengung thô zuo gotes engila went then to Lord’s angels ‘The angels of the Lord appeared’ (Tatian , ) () (& decessit / ab illa angelus) arfuor tho / fon Iruther engil departed then from her the angel ‘The angel left her’ (Tatian , ) () (ait) quad her thô spoke he then ‘Then he spoke’ (Tatian , ) () (Phariseus autem coepit . . . dicere) bigonda ther phariseus . . . quedan began the Pharisee speak ‘The Pharisee began to speak’ (Tatian , ) Second, V is considered a typical text-opening strategy, regularly applied at episode onsets. An example is provided in the translation of the Latin introductory formula factum est ‘it happened’ by way of a V construction, as in example (). V order is actually also widespread in opening sentences in later German, for example in folksongs and in poetry (Maurer : ff). Lenerz () argues that this function of V is retained in some colloquial genres in Modern German as well, for example in jokes (see also Simon ). Additionally, some authors relate V to various effects of foregrounding and expressivity, in that they observe a tendency to use this order in clauses that convey a very important, or unexpected event or action (Fourquet ). Consider () and (): () (& sensit corpore / quod sanata ess& a plaga) furstuont siu thó in ira lihhamen / thaz siu heil uuas fon perceived she then in her body that she healed was from theru suhti this plague ‘And then she felt in her body that she was healed from the plague’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
() (& facta est / tranquillitas magna) uuard tho gitan/ mihhil stilnessi was then done great calm ‘And there was a great calm’ (Tatian , –) In the more recent literature, there have been numerous attempts at deriving these various functions of verb-initial declaratives from a single, very general property. In the following section, an overview of the accounts proposed in the literature will be provided. As will become clear, it is extremely difficult to capture verb-initial order in OHG in terms of one single, straightforward criterion. More work is needed to provide an adequate picture of the various uses of this order in OHG. .. Verb-initial order in fully rhematic utterances Lenerz () argues that V in OHG is typical for sentences which are fully rhematic, i.e. which convey only new information. He assumes that the initial position of the verb results from the fact that no thematic information suitable to be placed pre-verbally is available in the clause (also Ramers ). He also observes that the cases in which we prototypically find only new information in the sentence are presentational sentences and existential constructions, which introduce new referents to the discourse. In these cases, Modern German would use the expletive es in the prefield. But as the expletive had not yet evolved in OHG times (Axel , a), the consequence is V on the surface. Several observations suggest that this explanation is too restricted. First, verbinitial declaratives with existential be can also contain given information, e.g. in thero landskeffi ‘in that region’ in () (= (b)), while novel information is conveyed in the subject expression hirta ‘shepherds’. Moreover, we encounter plenty of evidence for verb-initial declaratives whose subjects are thematic, e.g. pronouns, as in () or anaphoric definite DPs, as in (): () (Et pastores erant In regione eadem) uuarun thô hirta In thero landskeffi were then shepherds in that region ‘There were shepherds in that region’ (Tatian , ) () (& uenerunt) quamun sie thó came they then ‘Then they came’ (Tatian , ) () (Phariseus autem coepit . . . dicere) bigonda ther phariseus . . . quedan began the Pharisee speak ‘The Pharisee began to speak’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
Therefore, we need an explanation that can account for V order in clauses with both novel as well as given information. .. Verb-initial order with non-accusative verbs According to the second hypothesis, VS-orders in general—and V order in Germanic in particular—are typical for intransitive non-accusative verbs whose subjects are underlying objects, thus realized in the canonical clause-internal position, see Lenerz (). However, there are verb-initial declaratives with transitive verbs selecting accusative objects as in () and (): () (Dicebat autem & parabolam ad illos) quad thô zi In glîhnessi Inti bilidi told then to them parable and picture ‘Then he told to them a parable’ (Tatian , ) () (Acceperunt autem corpus ihesu) Intfiengun sie tho thes heilantes lichamon took they then the Saviour’s body ‘They took then the body of the Saviour’ (Tatian , ) Therefore, the explanation of V as a consequence of the argument structure of particular predicates must also be supplemented by additional factors accounting for the presence of this pattern in clauses with transitive predicates. .. V1 and narrative inversion in Germanic In the context of the remaining Germanic languages, V declaratives are associated with the phenomenon of narrative inversion. It has been claimed by Santorini () for Yiddish and Sigurðsson () for Icelandic that V order in declaratives implies a close relation of the utterance to the previous statement.3 Therefore, V declaratives are considered to be typical for discourse-continuative contexts and excluded from discourse-initial ones. But this generalization cannot be adopted for OHG in which V is regularly attested at the beginning of new texts or text passages. As in Modern German, V is a common text-opening device, used in the first sentence of texts in OHG. But it is also employed inside running texts to mark the beginning of new episodes or new paragraphs. Consider the example in Figure ., taken from the Tatian translation.
3 Interestingly, this continuative function of V is also very common in Modern German dialects (Weiß b: f).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
F . St. Gallen Cod. ; , – (St. Gallen, Stiftsbibliothek, Cod. Sang. ; p. —Tatians’ Gospel Harmony; http://www.e-codices.unifr.ch/en/list/one/ csg/. Reproduced with permission)
A new chapter of the New Testament begins in the middle of the second line of the extract in the facsimile. In the Latin part of the text (in the lefthand column), different graphical means are employed to mark this narrative break, such as the capitalization of the first letter of the new discourse section and the use of a marginal note to refer to the new passage of the New Testament. But in the OHG translation (in the righthand column), no graphical means can be found at all. The only device marking the beginning of a new narrative sequence is the initial placement of the verb (gieng ‘went’) in the sentence. This illustrates the discourse-structuring potential of V orders as a text-opening device in OHG. .. Sentence particles as blockers of XP movement Axel (, a) observes that V declaratives frequently contain the clause-internal tho. She argues that, next to its function as a temporal adverbial, this element displays properties of a sentence particle encoding information on sentence type and sentence mood. This function is well known for a class of elements in the early Germanic languages, as for a series of enclitic and phrasal particles in Gothic (see Ferraresi : Chapter for discussion and analyses). In OHG, the interrogative particle inu/enu and the affirmative particle ja are considered representatives of this system, see () and ():4 () (Aut non legistis in lege) Eno ni lasut ír in thero euuu PTC NEG read.PL.PRET you.PL in the.DAT.SG law ‘Didn’t you read in the law’ (Tatian , ) () iá bin ih scálc thin PTC am I slave your ‘I am truly your servant’ (Otfrid I , ) Axel (, a) assumes that by analogy, tho is used as a narrative-emphatic particle in declaratives. In this case, as sentence type and sentence mood are expressed via this particle, it is not really necessary to move a constituent to SpecCP
4
See also Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
via stylistic fronting. From this perspective, the loss of V in declaratives is related to the breakdown of the particle system in later stages of Germanic, with a subsequent generalization of XP movement in declaratives in German. Note, however, that there is a substantial syntactic difference between tho and the class of old Germanic sentence particles. While the latter is settled in the C-domain of the clause, the former is possible in clause-internal position as well. In view of this, the status of tho as another representative of the system of old Germanic sentence particles is uncertain. .. V1 and rhetorical structure Petrova (, ) and Hinterhölzl and Petrova () present a discourse-based approach aiming at a unified explanation of all verb-initial declaratives in OHG. They adopt previous observations according to which V is the canonical pattern in presentational sentences and existential constructions in OHG and extend this interpretation to the remaining types of V declaratives, claiming that all cases represent instances of thetic judgements. Thetic judgements, in the terminology introduced by Kuroda (), apply to utterances that lack a topic-comment structure (see also Sasse , ; Drubig ), in contrast to categorical judgements, which display a bipartite division into a predicational base, or topic, and a comment on that topic. Recall that next to their function of introducing novel referents, V declaratives occur in contexts involving an overall change in the narrative setting, for example the establishment of a novel time span, the shift to another place of action, or the introduction of another speaker. Applying the model of Asher and Lascarides () in which discourse structure is represented in terms of rhetorical relations between utterances, Hinterhölzl and Petrova () claim that V declaratives instantiate prototypical relations of coordination, which rest upon the temporal succession of events in the discourse. This is in contrast to relations of subordination, subsuming those types of discourse relations whereby an utterance elaborates on a previous one with which it temporally overlaps. The incompatibility of V declaratives with rhetorical relations of subordination and the lack of a topic-comment division in the sentence are taken to correlate with each other. Elaborating on a previous utterance usually involves the resumption of a particular referent or entity as the subject of the predication in the subsequent sentence, which attributes to the creation of a topic-comment division in it. In turn, expressing that a sentence should be interpreted outside the rhetorical relation of subordination, for example as an instance of asserting a new situation as a whole, requires the suppression of a possible topic-comment interpretation, especially in case some given material is involved. In this way, the common function of V order in declaratives is explained as a correlate of theticity in a general sense, i.e. as a strategy of disabling the topic-comment interpretation of the utterance. But note that this account dwells on a very broad definition of theticity, and that the predicates attested in clauses with V order satisfy this notion to a different extent. Also, in the contexts described as constitutive for V declaratives, another construction can also be found, namely a verb-second clause introduces with the adverbial thô ‘then’. The competition between V and thô-V will be addressed in the next section. It
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
might be suggested that the V variant is plainly optional in all of the cases in which it is found, for which a canonical verb-second order is also available.
. Excursus: the variation between V1 and thô-V2 V declaratives in OHG share the distributional properties described above with another frequently attested pattern, the one in which the temporal adverb thô5 is placed clause-initially, followed by the inflected verb in second position. This pattern will be referred to as the thô-V construction. It can be seen from the examples below that this pattern also occurs in clauses with existential be (), as well as with the entire list of V-sensitive verbs such as motion verbs (), verbs of saying (), and inchoative predicates (). () (homo erat In hierusalem) tho uuas man In hierusalem then was man in Jerusalem ‘There was a man in Jerusalem’ (Tatian , ) () (accedentes serui / patris familias autem) thó giengun scalca zuo / thes híuuiskes fater then went slaves to the families father ‘Then the slaves of the father of the family approached’ (Tatian , ) () (Dixit autem maria) thô quad maria then said Mary ‘Then, Mary said’ (Tatian , ) () (& repleti sunt omnes / in sinagoga ira) thó uuvrdun sie gifulte alle / in theru samanungu gibuluhti then became they filled all in the synagogue anger ‘Then, all in the synagogue became filled with anger’ (Tatian , ) Obviously, there is variation between V and thô-V in the same contexts in OHG. Nevertheless, the two constructions are not fully equivalent. Investigating a sample from the OHG Tatian, Donhauser and Petrova () identify a series of conditions favouring the use of one pattern over the other, foremost the lexical class of the verb and the temporal properties of thô. One observation is that the thô-V construction only occurs in affirmative sentences, while V order is also attested with negative sentences. Here the finite verb 5 This is the cognate of OE þa which also regularly triggers subject-verb inversion in OE (see van Kemenade ; Fuß ; Fuß and Trips ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
T . The distribution of V versus thô-V according to the lexical class of the verb lexical class of the verb
V
tho-V
existential be uuard tho motion verbs perception verbs verbs of saying phase verbs (incl. inchoative predicates) passive and mediopassive constructions varia
.% .% .% .% .% .% .% .%
– –
.% – .% .% .% – .% .%
total of examples
%
%
(Donhauser and Petrova )
forms a clitic complex with the negative particle ni, which moves to C as a whole (see Jäger and also Chapter of this volume). But if we look at affirmative sentences, which may have both orders, we find additional factors governing their distribution. First, consider the role of the lexical classes of the predicates that select these orders, defined in column of Table .. It can be seen that V declaratives obviously impose no restrictions on the lexical type of the predicate used in the sentence, while thô-V strongly prefers one particular class of verbs, namely verbs of saying, accounting for .% of the occurrences in a total of clauses. Second, Donhauser and Petrova () observe a difference in the temporal properties of thô, depending on its position relative to the verb. Consider () versus (). In (), thô introduces a novel, indefinite time interval which is established as the topic time6 of the episode, namely in anderemo sambaztag ‘on another Sabbath’. In (), by contrast, thô resumes the already pre-established time span: the topic time of the episode. In other words, thô refers to a definite, given time span, which makes it similar to topics as frame-setters (Chafe ): () (Factum est in alio sabbatum autem) uuas thó giuuortan in anderemo sambaztag was then become in another Sabbath ‘It happened on another Sabbath’ (Tatian , ) () (homo erat in hierusalem) tho uuas man In hierusalem then was man in Jerusalem ‘A man lived in Jerusalem at that time’ (Tatian , ) 6
See Klein ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
The data analysed by Donhauser and Petrova () shows that novel thô prototypically appears clause-internally, thus giving rise to V on the surface, while the frame setting adverbial thô is canonically realized in the prefield of the clause, leading to subject-verb inversion. These distributional properties of thô are consistent with previous observations on the role of information structure in word order variation in OHG. It has been argued that expressions lacking properties of sentence topics resist movement to SpecCP in OHG. We can extend this observation from expressions referring to individuals to expressions referring to time spans; novelty and indefiniteness of the time interval denoted by thô prevent its placement in SpecCP, while canonical properties of topicality such as givenness and identifiability favour its fronting to the prefield of the clause. These observations on the distribution of V versus thô-V should, however, be verified by the analysis of other texts of the OHG period.
. The later development of V1 declaratives In this section, we will finally address the relation between historical V declaratives and their putative Modern German formal counterparts. In Section ., it was outlined that in Modern German, V declaratives occur only sporadically. The literature has distinguished the following three contexts in which V declaratives can be found in Modern German (see Lenerz ; Auer ; Önnerfors ; Simon ; Reis ; Scheutz ): (i) narrative V declaratives, for example at the beginning of jokes (),7 (ii) enumerative V declaratives (), and (iii) explanatory V declaratives (): () Treffen sich zwei nach den Faschingstagen meet REFL two after the carnival’s days ‘Two people meet after the carnival’s days’ (Reis : ) () [Die Tendenz geht nach unten.] Kommt noch hinzu, dass . . . the trend goes down comes still on top of that that ‘[There is a downward trend.] Still on top of this has to be added that . . . ’ (Reis : ) () [Sein Tod bewegte viele.] Hatte doch seine Ära den Wiederaufstieg his death moved many had surely his era the upturn begründet founded ‘[Many were moved by his death.] For surely his era was the foundation of the upturn’ (Reis : ) 7
A subtype of this function is also described by Simon () for Bavarian, who observes that V is used to start a new dialogue or to continue an old one after a break. As for the narrative function of V in modern German dialects in general, see Weiß (b).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova
The question arises what the origin of these constructions is, and more specifically, if there is a historical relation between V declaratives in Modern German and the same pattern described for OHG. This question is justified in the face of the observation that V declaratives are frequent in texts from the classical OHG period, i.e. from the ninth and tenth centuries, while already in late OHG texts, they decline significantly. In the comprehensive work by Notker, who is one of the latest writers of the OHG period (see Chapter ), verb-initial declaratives are already highly infrequent. Maurer (, ), who studies the development of V declaratives in the history of German, puts forward the so-called ‘discontinuity hypothesis’, claiming that there is a gap in the attestation of V declaratives over time. He argues that towards the end of the OHG period, and especially during the early MHG period, V declaratives are completely missing from the documents, until they re-enter the system only later, in the fifteenth century, presumably as imitations of foreign scientific style. During this renovation process, V declaratives are found preferably with verbs of saying, and only later with other predicates. Finally, during the sixteenth century, V declaratives are found in native prose, independently of any foreign influence. Maurer’s discontinuity hypothesis is discussed controversially in the literature (for an overview and literature, see Coniglio : ). It is partially supported by Axel (a: Section ) who observes that discourse-internal V declaratives in OHG such as those in examples (), (), and () cannot be properly translated into Modern German by way of V orders (see Axel a: ). She concludes, in the vein of Maurer () and Behaghel (–, Vol. IV), that there is no historical connection between these types of V declaratives in OHG and the Modern German representatives of this pattern. As for discourse-initial V declaratives, Axel (a) shares Lenerz’s () opinion that there is some historical continuity, retained in informal, colloquial style. Maurer’s () discontinuity hypothesis is challenged by Önnerfors () and Coniglio (), who present examples for V declaratives from various subperiods of MHG, suggesting that this pattern was constantly present in the history of German. Consider the examples in () and (): () nu segen was de sanc si. ist iz sanc allir sange now let’s say what the song is.SBJV.PRES is it song of-all songs ‘Now, let’s say what the song is. It is song of songs’ (Trudp HL , – after Önnerfors : ) () sprichet sanctus Augustinus says Saint Augustine ‘St. Augustine says’ (Tauler, Red , , – after Coniglio : ) Önnerfors () and Coniglio () argue that a significant factor responsible for the drastic drop in the frequency of V declaratives is related to genre and register. Assuming that V declaratives are typical for colloquial style rather than
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Verb-initial declaratives in OHG and later German
for written language, they appear inappropriate for the register of court prose and poetry which constitutes the overwhelming proportion of the MHG corpus (see Chapter ). Finally, Coniglio () observes an underlying similarity in the pragmatic properties of Modern German and historical V declaratives. Recall that in Section .., the absence of a topic-comment division was considered as the underlying property of V declaratives in OHG. The same property is considered to hold for Modern German V declaratives, according to Önnerfors (). Coniglio concludes from this that V declaratives are a historically constant pattern, displaying the same form and function from OHG to modern times.
. Conclusion Although basic properties of V are established in matrix clauses already in OHG, we find considerable evidence for verb-initial declaratives in the data. In this section, we looked at the syntactic and pragmatic properties of such clauses. We argued for an analysis according to which the finite verb is moved to the left periphery of the clause but the prefield remains empty. We interpreted these facts as indications for the gradualness with which the components of the V rule were generalized in the history of German. In this respect, V declaratives point to a stage in which verb fronting to C is firmly established but XP movement to SpecCP is not yet obligatory. In attempting to explain this situation, we looked at the factors that block movement to SpecCP in V declaratives in OHG. We investigated various contexts in which V occurs and surveyed the accounts proposed in the literature. As V in OHG correlates with rather heterogeneous syntactic, semantic, and pragmatic factors, it is difficult to derive a single, very general factor accounting for all occurrences of this pattern in the data. Further work is needed to clarify the inconsistencies of previous analyses. Also, additional investigations are needed to shed light on the development of the pattern from OHG to modern times.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
5 The prefield after the Old High German period A U G U S T I N S P E Y E R A N D H E L M U T W E Iẞ
. The prefield in Modern German .. Syntactic constraints The prefield is defined in the traditional topological or field model (see Chapter ) as the zone immediately preceding the left sentence bracket, i.e. the position of finite verb forms in main clauses (e.g. Dürscheid ; Wöllstein ). In generative terms, it is identical with SpecCP (see e.g. Thiersch ; den Besten ; Sabel ). We deliberately use the pre-Rizzian cover term for the left periphery, the reason for which will become clearer when we turn to the possible analyses of the prefield. A property which has been described for the prefield in classical topological literature as well as generative literature is that it usually contains only one constituent (in the following referred to as the ‘one-constituent property’). Whereas in the field model this is simply a descriptive generalization, in generative grammar—at least at the stage of the ‘Barriers’ model (Chomsky )—this property directly follows from the fact that it is the specifier of a single projection, hence only one phrase can occur in this position. The advent of a more fine-grained analysis of the left periphery, in the wake of Rizzi (), removed the explanatory power of the SpecCP scheme for the observable one-constituent property of the prefield, as now more than one specifier position is available in the left periphery. As a consequence, the one-constituent property has to be derived by an additional mechanism. Several proposals concerning this mechanism have been made. Grewendorf (), for instance, assumes that the prefield constituent has to pass through the lowest projection in the split C architecture, SpecFinP, before moving to an information-structurally more suitable projection in the split CP architecture such as SpecTopP or SpecFocP. As the constituent leaves a trace in SpecFinP, further movement of other constituents to the left periphery is excluded. On the other hand, Frey (a, ) suggests that only phrases bearing a contrastive feature can be moved directly to the left periphery via operator movement Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Augustin Speyer and Helmut Weiß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
(SpecKontrP in this case, which is probably identical to Rizzi’s SpecFocP, although Frey does not commit himself to this). In other cases, the left-most constituent of the middle field (that is, everything below Fin in Frey’s conception) is moved to SpecFinP by the operation of formal movement (the idea essentially goes back to Fanselow , see also Chapters and of this volume on stylistic/formal movement). By looking more closely at the data, one can observe that the generalization that the prefield always contains only one constituent does not hold in all cases. There are several cases reported in which the prefield contains more than one constituent (see Müller among others). A violation of the one-constituent property occurs only rarely, but obviously it does occur. Several attempts have been made to explain such exceptions from the one-constituent property against the background of a syntactic analysis that allows for only one phrase in the prefield. One approach is to say that cases of multiple constituents in the prefield are in fact instances of ‘remnant movement’, i.e. movement of a partially emptied verbal phrase to the prefield. Such a partially emptied VP can contain more than one phrase in complement or adjunct position, and these phrases can therefore be fronted to the prefield together (Müller ); consider [CP [VP Zum zweiten Mal die Weltmeisterschaft t1]2 errang1 [IP Clark t2 t1]] ‘for the second time the world championship won Clark’ (adapted from Müller ). Other researchers analyse multiple constituents in the prefield as exactly that and assume that more than one phrasal position in the C architecture is filled (e.g. Jacobs ; Speyer a; Chapter .. of this volume). Later on in this chapter, we will return to this question. .. Information-structural constraints Another property of the prefield or SpecCP is that syntax does not determine what phrase is moved to that position. As the movement to SpecCP is not A-movement, constraints on A-movements such as locality do not hold. In other words, any phrase— as long as it is a phrase—can be moved to the prefield (see also Chapters and ). The selection of which phrase is moved to the prefield in the end is mostly a matter of information structure. It can be observed that only phrases with certain information-structural properties appear in the prefield (see e.g. Hoberg ; Jacobs ; Speyer b, a): • phrases that determine the temporal or spatial situation in which the proposition is evaluated with respect to its truth value (‘scene-setting’ in Speyer b; see (a); definition following Jacobs ); • phrases that refer to an entity which stands in a partially ordered set relation to at least some other element in the discourse (‘contrast’ in Speyer b; definition following Prince . The clauses in (b) illustrate this; the set consists of potential windows in the building); • phrases that refer to an entity that the rest of the sentence is about (‘topic’ in Speyer b; see (c); definition following Reinhart ). In order to be a candidate for the prefield, the topic must either be discourse-old (preferably at the same time being a centre, see Speyer ), or it must be of some macrostructural relevance, that is, it must function as a topic for at least one more sentence (see Speyer a).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
() a. Auch hier sind also weitere Grabungen notwendig also here are consequently further excavations necessary ‘Thus, also here further excavations must be made.’ (Kortüm, Untersuchungen ) b. Die aus den früheren Untersuchungen bekannten Fensteröffnungen the out the earlier surveys known window.openings würden . . . zum Dachstuhl gehören. Weitere Fenster wurden would to.the truss belong further windows became nicht festgestellt . . . not identified ‘The windows, known from earlier surveys, would belong to the truss. Other windows were not found’ (Kortüm, Untersuchungen –) c. Denn . . . , Moralismus‘ ist eine Überbewertung . . . der Moral . . . for moralism is an over.appreciation of.the morals Sie will die Moral zum alleinigen Maßstab . . . machen . . . she wants the ethics to.the sole scale make ‘For moralism is an over-appreciation of righteousness. It strives to make righteousness to the exclusive guideline’ (Küng, Moral ) A clause that contains no phrase of any of the information-structural types outlined above is likely to fill the prefield with an expletive, that is, an element base generated in the prefield that has no semantic content and whose sole purpose seems to be to turn the clause into a verb-second (V) clause, as can be seen in (a). Note that this type of expletive (the Vorfeld-es/prefield-es, see also Chapter ) cannot appear in any other position, as illustrated in (b). This indicates the properties identified above, namely that it is an element base-generated in the prefield (as the prefield, or SpecCP, is part of the highest projection in the tree, its content cannot be moved further upwards in the tree), and, on a more descriptive level, that its raison d’être is strongly connected to the prefield position as such. () a. Es werden immer weniger /-Artikel verschenkt it become always less random articles given.as.present ‘One gives less and less random stuff as a present.’ (Berliner Morgenpost, October , – after Speyer a: ) b. */-Artikel werden es immer weniger verschenkt random articles become it always less given.as.present A question that presents itself is what happens if a clause contains more than one phrase which would conform to the information-structural demands for movement into the prefield. In principle, there is competition as to which one is moved to the prefield in the end. A survey of German newspaper texts and texts of a similar stylistic level reveal that there is a ranking involved (Speyer b); if one of the competitors is a scene-setting element, the prefield will most likely (in roughly % of
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
cases) be filled by the scene-setting element. If no scene-setting element is involved, but there is competition between a topic and a contrastive element, there is a slight tendency for the contrastive element to win out. This ranking is demonstrated in (). There are two sentences, each of which contains a scene-setting element (marked with { . . . }), a contrastive element (marked with / . . . \), and a topic (marked by underlining). The set opened up by the contrastive elements is different publications of the author Hans Küng; the topic of the whole passage is the author himself and his activities leading to the institution of his foundation. In both clauses, the prefield is filled with the scene-setting element. () {Im Jahr } . . . hielt ich / den Vortrag ,Warum brauchen wir in.the year kept I the presentation why need we globale ethische Standards‘\ . . . {Noch im selben Jahr} veröffentlichte global ethical standards yet in.the same year published ich / das programmatische Buch, Projekt Weltethos‘\ . . . I the programmatic book project world-ethos ‘In the year , I spoke publicly on “why we need global ethical standards”. In the same year, I published the programmatic book “Ethos of the World”.’ (Küng, Moral ) So, on the whole, there seems to be a ranking of prefield fillers of the following form in operation in Modern German if a clause contains more than one phrase that could be moved to the prefield: () scene-setting element > contrastive element > topic Note that the ranking of scene-setting elements over the other two classes of elements is pronounced, whereas the ranking of contrastive elements over topics is much less clear-cut.
. The prefield in older stages of German If one adopts a diachronic viewpoint, a question that suggests itself is whether the syntactic property of hosting exactly one constituent as a rule, the informationstructural properties and the prefield ranking described above were in operation in earlier stages of German as well, or whether a change occurred. The question is the more urgent as the prefield in itself (as the whole V-property of Germanic languages) is a relatively recent development, in that it does not seem to have been part of PIE syntax, and it is doubtful whether, within the Germanic branch, it is a very old innovation either (see e.g. Speyer b). In fact, the prefield in OHG shows some properties that differ strongly from the modern setup of the prefield, as discussed in Chapter . Yet we see that something like the one-constituent property is developing, and it can be shown that, on the whole, if the prefield is filled with only one constituent, it follows a ranking not unlike the one determined for Modern German, with the difference, however, that contrastive elements are not preferred as prefield fillers. Scene-setting elements are highest in ranking, but topics can beat them in the competition more easily than in Modern German (see Speyer b). So it looks as if
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
the ‘old’ property of the prefield as classical topic position seems to be still in operation (see Hinterhölzl and Petrova ; also Chapter of this volume). In this chapter, we will concentrate on the prefield after the OHG period. By MHG times, the one-constituent property seems to have been fairly established, although exceptions seem to be more numerous in MHG and ENHG (Early New High German) than in Modern German (see Section .. on ENHG). We will concentrate on the information-structural content and the prefield ranking. .. Middle High German In order to gain an approximative view on the state of the prefield in MHG, we analysed declarative clauses from a sermon by Berthold von Regensburg (sermon Von den siben planêten, in Pfeiffer’s edition Vol. I: –). The numbers are given in Table . and Table .. Table . is to be read as follows: the clauses that conform to the ranking identified for Modern German are in columns to and are printed in bold face. The clauses that violate this ranking are given in columns to . The percentages are aligned such that each competition case occupies a row by itself. In the tables, S stands for scene-setting elements, C for contrastive elements, T for topic. In Table ., the clauses that conform to the expected ranking are in column and in bold face; the other possibilities are in columns and . Only the number of tokens is given in this table. Examples for some of the competition cases are given in (). Here we relate cases with three competitors: (a) represents the ‘normal’ case in which the scene-setting element is in the prefield (note that dô is clearly temporal-referential here and not simply a cohesive device as e.g. also in ()), (b) represents the ‘special’ case in which the contrastive element is in the prefield. The information-structural properties are marked as in example ().
T . Clauses with two competitors for the prefield in MHG (Bavarian) S>C number of tokens %C:T %S:T %S:C
S>T
C>T
C>S
T>S
T>C
T . Clauses with three competitors for the prefield in MHG (Bavarian)
number of tokens
S > C, T
C > S, T
T > S, C
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
() a Dô er daz . . . volk fuorte . . . , {dô} gap er in /zweierleie when he the people led then gave he them twofold wîsunge\ guidance ‘At the time, when he led the people (sc. of Israel from the desert into the promised land), he gave them two kinds of guidance’ (Berthold I , –) b und /einez\ lesen wir {bî der naht} and one read we at the night ‘and we read the one at night’ (Berthold I , ) It can be noticed immediately that no ranking between contrastive elements and topics is established, not even the slight preference for contrastive elements visible in Modern German, whereas scene-setting elements are already as much preferred for the prefield as in Modern German. The total number of sentences showing competition captured by the prefield ranking is fair: out of sentences, forty-three show competition of the relevant kind. It should be mentioned that in quite a number of sentences the prefield is occupied by an element not included in the ranking, most often an adverbial element that seems to serve mostly as a cohesive device, such as alsô ‘in this way’ or ‘in the same way’ or ‘consequently’, cf. (). As a matter of fact, if there was competition between a cohesive device and some other prefield filler, which was the case in thirteen clauses, in twelve cases the cohesive device won out. () Und alsô hât uns got {alle naht und tac} in sîner huote and so has us God all night and day in his custody ‘and consequently, God keeps watch over us all days and nights.’ (Berthold I , –) Note that this cohesive adverbial may oust ‘better suited’ prefield candidates; in (), the topic (got ‘God’) and an element that can be analysed either as scenesetting or as contrastive element at the same time (alle naht und tac ‘all days and nights’) appears in the middle field.1 A similar study was made on a MHG sermon from the East Central German area (Mitteldeutsche Predigten). The numbers point in the same direction, although the tendencies are less pronounced, as can be seen from Table . and Table .. Note that clauses in which competition occurs at all are infrequent (mostly there is only a topic as potential prefield filler), so even larger portions of text yield only a few examples. Therefore the calculations should be treated with some caution, but as they point in the same direction, one feels justified in drawing conclusions. Competition between scene-setting and contrastive elements was rare, so these numbers are not meaningful. With the other competition cases, it should be noted that—on the whole—there are clear tendencies in favour of scene-setting elements over topics 1 Elements in which the assignment to one of the three classes of preferred prefield-fillers is ambiguous were excluded from the database from which the ranking was inferred.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
T . Clauses with two competitors for the prefield in MHG (East Central German) S>C number of tokens %C:T %S:T %S:C
S>T .
C>T
C>S
T>S
T>C
.
T . Clauses with three competitors for the prefield in MHG (East Central German)
number of tokens
S > C, T
C > S, T
T > S, C
(although the percentage is lower than in the Bavarian text) and of contrastive elements over topics (the tendency is even clearer here as compared to Bavarian). In cases with three competitors, scene-setting elements are in general similarly successful as in Bavarian. In the East Central German text (Mitteldeutsche Predigten), a similar phenomenon occurs as in Bavarian, namely that the prefield is relatively often occupied by a cohesive device (in this case mostly dô, which indicates the discourse relation of narration, and nû, which corresponds more or less to ‘yet’), most notably in cases in which a prefield filler would be available. In twenty-four out of twenty-six cases, a cohesive device wins over other types of prefield fillers. This phenomenon can be interpreted as a strategy to cope with the V-constraint which might be either particular to the two dialect areas from which the texts come or might be the general state of the MHG language in this respect, and which is different from the strategy that Modern German is following. Whereas in Modern German the prefield is filled by actual movement of some phrasal element that is part of the clause proper into SpecCP, in the MHG texts there is a tendency to base-generate elements in SpecCP rather than to fill this position by movement. The fact that verb-first clauses are not infrequent in OHG might indicate that at that time, the V-constraint was not yet generalized; MHG might reflect a stage of the evolution of the V-constraint in which the link between V and declarative sentence mood was already firmly established, but the movement operations filling the prefield were not yet as freely available as they are in Modern German. In some ways, this might be a continuation of the hypothetical pre-OHG state-of-affairs where particles indicating sentence mood (and probably discourse particles) regularly, if not compulsorily, appeared in the left periphery (Axel : ff and references therein; see also Chapters and of this volume).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
.. Early New High German A similar, but broader, investigation was conducted with texts from the fifteenth century, that is, the early part of the ENHG period. Texts of two comparable dialect areas, viz. Upper Saxon (again an East Central German dialect) and Central Bavarian, were selected. In contrast to the MHG investigation, texts of three text types were chosen from each dialect area, namely chronicle, narrative, and sermon. The texts constituting the corpus are given in Table .. For each text type in each region, clauses were analysed, giving a total of clauses. If a text turned out to be too short, a second text of the same text type was added. In the sample, there was no significant difference between the dialect regions, so the numbers are combined. The first observation is that it is the same three information-structural types of elements that are found in the prefield as in Modern German. So here the modern state-of-affairs is already present. An innovation compared to MHG is that cohesive devices do not play such a central role, although one Bavarian text (Die Denkwürdigkeiten der Helene Kottanerin) tends heavily towards filling the prefield with discourse connectives, mostly dô, which indicates the discourse relation of narration. In forty-nine cases, dô was used in the prefield although other potential prefield fillers were present. In the Upper Saxon texts combined, it was only eight cases. As dô
T . The ENHG corpus Region
Text type
Text
Upper Saxon
chronicle Urkunden des Oberlausitzer Hussitenkrieges und der gleichzeitigen die Sechslande angehenden Fehden. Pilgerfahrt des Wilhelm von Thüringen narrative Ritter, Bürger, und Bauern
Das Leben der heiligen Dorothea von Johannes Marienwerder sermon Predigt (Anonymous, probably from Leipzig) Central chronicle Jörg Kazmair’s Denkschrift Bavarian über die Unruhen zu München narrative Die Denkwürdigkeiten der Helene Kottanerin sermon
Das ,Beichtbüchlein‘ des Thomas Peuntner
Date of composition
Number of analysed clauses
–
not before
first half fifteenth century between and
first half fourteenth century mostly –
not before ; probably c.
first half fifteenth century
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
T . Clauses with two competitors for the prefield in ENHG
number of tokens %C:T %S:T %S:C
S>C
S>T
C>T
C>S
T>S
T>C
T . Clauses with three competitors for the prefield in ENHG
number of tokens
S > C, T
C > S, T
T > S, C
occurs only in the prefield in the texts, it could be regarded as a prefield expletive.2 So the strategy identified at the end of the last section—to fill the prefield by basegenerating cohesive devices rather than moving something out of the clause proper to SpecCP—seems to linger on in Bavarian longer than in East Central German. Turning to the competition cases, we see a slight difference from Modern German. The numbers of competition cases are reported in Table . and Table ., which are to be read as explanations for Tables . and . respectively. Examples for some of the rankings are given in (). In (a), a scene-setting element is positioned in the prefield outranking a contrastive element, in (b), a scene-setting element stands in the prefield, while the topic is in the middle field, and in (c) a contrastive element is in the prefield, outranking a topic. () a. (/Herzog S.\ kam . . . ) {Am andern Tag frue} kam /Herzog Ernst\ . . . duke S. came at.the other day early came duke Ernst ‘(Duke S. arrived . . . ). The next day, in the morning, Duke Ernst arrived’ (Kazmair, Denkschrift }) b. {Vff myttewochin . . . } fuer der furste . . . zcu Venedie on Wednesday travelled the prince to Venice ‘On Wednesday, the prince travelled to Venice’ (Pilgerfahrt ) c. /Judas\ Judas /Petrus\ Peter
2
ging went nam took
mit with dar there
/Petro\ zu dirre Peter to this an /daz ewige on the eternal
herrschaft. government leben\, life
In modern Bavarian, where the prefield -es is lacking, da still occurs in this function, see Weiß (: ). On different properties of clause-initial versus clause-internal thô (OHG predecessor of dô), see also Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
Judas /der\ nam dar an / den ewigen tot\. Judas he took there on the eternal death ‘Judas went with Peter to this discipleship. Peter gained from it eternal life, Judas gained from it eternal death.’ (Predigt Leipzig , –) So, the ENHG sources indicate that indeed there was a continuous development between MHG and Modern German, namely a gradual establishment of the modern V-syntax and as a consequence a change in the ranking of prefield fillers. The modern version of the V-constraint (obligatory prefield filling by movement in declarative clauses) gained ground; whereas in MHG, V clauses in which the prefield is not filled by movement of some phrase out of the VP-domain into the left periphery are common, in ENHG they are only exceptions. In the cases in which the prefield is filled by movement, the modern ranking (scene-setting element > contrastive element > topic) is on the way to being established in ENHG; whereas in MHG, the OHG state-of-affairs—where the competition of topics and scene-setting elements often ends up with the topic as the winner, and contrastive elements do not play a role at all—is still traceable. In information-structural terms, one might interpret this as a development of the prefield from a position specialized for topics to a less specified position that can host any element that has a context-linking quality (to which scene-setting elements and contrastive elements belong). In structural terms, we see that the association between the filling of SpecCP and the sentence mood ‘declarative clause’ is already more or less compulsory by MHG times. So we might assume that in MHG at the latest, the declarative sentence mood was implemented as a feature that could be satisfied either by base-generation of elements (as it presumably was in pre-OHG times) or by movement of some phrase to the left periphery (as is the normal case today). In the course of the development, the movement option became more and more favourable. It is not clear why this should be the case; one could assume that base-generation is a costly operation (which in optimality-theoretic (OT) terms could be implemented as a violation of a highly ranked faithfulness constraint DEP-IO(DS-PF-pair)3) whereas movement is less costly (which, in OT terms could be implemented as a violation of another lower-ranked faithfulness constraint IDENT-IO(DS-PF-pair)4). This would hinge on base-generation being more costly than movement, however, which is evidently not true. At a stage of the language in which sentence mood was indicated almost compulsorily by particles, there would be no violation of DEP-IO(DS-PF-pair), as the mood particle is ‘meaningful’ and therefore not pure expletive.
3 DEP-IO stands for ‘the output is dependent on the input’. DS-PF specifies what the input and the output is: the input is a Deep Structure, the output a Phonetic Form. The optimal candidate under this constraint is one in which all material in the output corresponds to material in the input (which does not mean that all material from the input must be represented in the output). 4 IDENT-IO stands for ‘the output is identical to the input’. For the other abbreviations see preceding footnote. The optimal candidate under this constraint is one in which there is no difference between input and output (that is: the Phonetic Form directly spells out the Deep Structure).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
. On the doubly filled prefield in Early New High German .. The data In Chapter , we saw that the one-constituent property holds in most OHG declarative clauses. In the preceding sections of this chapter, we saw the one-constituent property at work in MHG and ENHG. In fact, V declarative clauses—clauses in which the prefield is filled by one constituent—are the rule. Yet the implementation of the one-constituent property is not absolutely categorical; whereas in Modern German, most if not all counterexamples are only apparent exceptions (in the sense that the prefield is filled via remnant movement of a partially emptied VP, see Müller ), in ENHG clear violations of the one-constituent property appear every now and then. A case is counted as a clear violation if the order of the constituents in the prefield does not correspond to the base-generated order that the constituents would have if they were in the middle field and no scrambling had applied. Examples are given under (); they are cited from the Bonn ENHG corpus. (a) is an example in which a PP functioning as local adverbial (information-structurally more precise: scene-setting element) occurs together with an NP functioning as subject and representing the topic of the sentence in the prefield; in (b) an adverbial connector (see Chapter ) and a personal pronoun subject share the prefield. The constituents are indicated by square brackets. () a. [Jm . Jar] [der Großfuerst DEMETRI] hat den maechtigen in.the th year the archprince Demetri has the powerful Tatarischen Khuenig MAMAI geschlagen Tatarian king Mamai beaten ‘In the th year, Archprince Demetri gained victory over the powerful Tatarian king Mamai’ (Herberstein, Moscouia r, –) (Bavarian) b. [Dar vm-] [du] solt nit allein mercke- vnd verston therefore you shall not only notice and understand ‘Therefore, you shall not only notice and understand (sc. that . . . )’ (Brunschwig, Cirurgia r B, –) (Alsatian) The problem of doubly filled prefields in ENHG was investigated at length in an earlier study (Speyer a), on which this section is based. In this study, eight texts from four dialect areas (Bavarian, Alsatian, Ripuarian, Upper Saxon) were used that are available via the Bonn ENHG corpus. For each dialect area, two texts were chosen: one from the period between and , and one from the period between and . In order to assess the further development, an analysis of texts from the period between and was added in the present chapter. The latter texts are listed in Table .. In Speyer (a), all declarative main clauses following the one-constituent property were identified from a text sample, likewise those in which more than one constituent occupies the prefield were gathered from the sample. So it was possible to assess the percentage of declarative main clauses with doubly filled prefield out of the total of declarative main clauses that have a filled prefield. These percentages are reported in Table .. For the clause count we refer to Speyer (a); the separate clause count of the latest period is additionally given in
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
Table .. If we estimate the rate of clauses with double prefield filling, we get different results for each dialect area (Table ., Figure .). T . Additional texts for the period – Author Bavarian
Text
Clause With double count prefield
Georg Scherer
Christliche Leichpredigt, Bey der Käyserlichen Besing- und Begängnuß, [ . . . ] () (only part I) Alsatian Johann Michael Wunderliche und wahrhafftige Moscherosch Gesichte Philanders von Sittewalt () (pp. , .–, .) Ripuarian Matthias Quad Imago Flandriae. Erinnerung an den Leser über gegenwärtige Figur () Johannes Gropper Briefwechsel II (–) (pp. –) sum of both texts Upper Jacob Böhme Morgenröte im Aufgangk () Saxon (pp. –)
T . Percentage of double prefield filling in declarative main clauses between and
%
– – –
Bavarian
Alsatian
Ripuarian
. . .
. . .
. . .
Upper Saxon . .
10 9 8 7 6 5 4 3 2 1 0 1350–1449
1450–1549
1550–1650
Period Bavar.
Alsat.
Ripuar.
Up. Sax.
F . Development of double prefield filling between and
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
While it is impossible to do either a χ2-test for all data (some cells are ) or a Fisher’s test (the numbers are partly too big), a χ2-test for Alsatian versus Bavarian showed weak significance (p = .), and a Fisher’s test for Alsatian versus Upper Saxon showed significance (p = .). It is obvious that some dialect areas are more permissive with respect to multiple prefield filling, whereas others are more restrictive to begin with. It is noteworthy that the rate decreases significantly in the more northern dialects which had a relatively high frequency in the first period, whereas in the Southwest it even seemed to increase over that period. Bavarian kept a relatively low frequency. Upper Saxon started with a relatively high frequency, but it went to zero in the observed timespan, being the only dialect that did away with double prefield filling completely. It is perhaps no coincidence that Modern German shows virtually no double prefield filling, since its main source is the dialect which is the most restrictive with respect to prefield filling. In terms of syntactic change, we can say that some change must have happened in the Northern dialects and possibly Bavarian (although it is possible that the underlying structure remained unchanged, and it is only a change in frequency of possible outputs, which could be modelled in a (stochastic) OT framework as a change in the relative value of the constraints involved), a change whose outcome is a more restrictive version of the one-constituent property, whereas in Alsatian the oneconstituent property is on the whole not obeyed as strictly as in other areas. The exceptions become even more numerous in the latest period. .. Implications for the syntactic analysis The mere presence of sentences with more than one phrase in the left periphery challenges the standard view of German clause structure that goes back to den Besten () (see Chapters and of this volume). According to him, the prefield is equivalent to SpecCP, as shown in (). This means that only one position is available for movement to that position. ()
CP XP
C’ C0
prefield
left sentence bracket
IP/VP middle field etc.
Consequently, the few instances of multiple prefield filling in Modern German are often accounted for by remnant movement of a partly emptied VP as discussed above, thus salvaging den Besten’s model (Müller ). This analysis is possible as long as the prefield fillers occur in the linearization in which they would be base generated within their VP, as we have to assume that no scrambling of the VP-internal constituents has occurred prior to remnant movement. With Müller’s examples—if we choose the reading in which für die Geschworenen ‘for the jury’ is an immediate constituent of VP—this analysis works quite well, cf. ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
() [CP [VP t2[Großes Gewicht] [für die Geschworenen] t1]3 hatte1 great weight for the jury had [IP [ein Bericht]2 t3 t1]] a report ‘A report was thought to be extremely relevant for the jury’ (Müller : ex. (h)) The analysis does not work equally well, however, for several ENHG examples such as (a). Here the adverbial appears before the subject. If we assume that the subject is in some functional projection higher than VP, the order of the adverbial before the subject would be an instance of scrambling.5 But remnant movement of scrambled constituents is excluded. Thus it appears that not all cases of double prefield filling in EHNG can be accounted for by Müller’s remnant movement hypothesis. The only way out is to concede that the prefield consists of more than one syntactic position, although it is most often the case that only one of these positions is the target of movement or otherwise overtly filled. Den Besten’s analysis with one structural prefield position, on the other hand, covers the majority pattern and modern grammaticality judgements about the left periphery very well, so one should be hesitant to forgo this analysis. But is there a way to reconcile the ENHG findings with den Besten’s analysis? There are basically two ways to achieve this: one is to allow multiple specifiers in the C-architecture (cf. Fuß ). Historically, German would then have had multiple specifiers in C, just as for instance Bulgarian does today (see Bošković ), but some mechanism normally prevented more than one of these specifiers being filled. According to Fuß (), there was only a single CP in OHG, but the C-head could project ‘multiple specifiers hosting fronted XPs’ (Fuß : ). In the unmarked case in OHG (as well as in NHG), C was endowed with an EPP feature, so it could only attract the closest XP in the middle field which is then raised to SpecCP. In cases, however, where C was endowed with an additional feature (e.g. a topic or focus feature), more than one XP was raised to the prefield where the XPs occupied different SpecCP positions. The syntactic change whose eventual result was the loss of multiple XP fronting comprised a change of the feature endowment of C because the fronting of expletive thô, which emerged (roughly) in the OHG of the ninth century, led to the reanalysis of XP fronting as a semantically vacuous movement whose only function is to check the EPP feature of C. The second option of analysing the findings is to follow a cartographic approach to the left periphery of the clause, such as outlined by Grewendorf (). A sketch is given in ().
5 The analysis would still apply, however, if a VP-internal subject position were assumed which is preceded by the respective adverbial in VP-adjoined position.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
()
Augustin Speyer and Helmut Weiß
ForceP SceneP FocP TopP FinP IP
Here, we have more than one C-projection (and thus more specifiers, i.e. phrasal landing positions). As indicated above, the way to restrict movement to more than one C-position outlined by Grewendorf () is that movement of any phrase necessarily goes through SpecFinP. The phrase would move on to a suitable landing site, but the trace that it leaves would prevent further movement of some other phrase to SpecFinP and thus into the left periphery. An alternative was suggested by Speyer (b), where long movement leaving out FinP is in principle possible, but dispreferred; therefore, long movement occurs only very rarely. This dispreference was modelled using Stochastic Optimality Theory (Boersma and Hayes ) in that normally a high-ranked constraint -VF (‘Only one constituent is in the prefield’) preventing more than one constituent in the C-architecture does its job; only occasionally a paradox ranking occurs, in that the constraint -VF receives a lower value than other constraints that take care to move for example topical or contrastive material to the prefield. These are the instances of multiple prefield filling. It is perhaps noteworthy in that connection that the order of referential phrases in multiple filled prefields are consistent with the serialization forced by the cartographic structure in (). The syntactic change mentioned above would then be caused by the increase (in Upper Saxon and Ripuarian) or decrease (in Alsatian) of the median value of the constraint -VF in that model. Such changes in value can be thought of as a feedback effect: if for some reason (or just randomly) violations of a given constraint occur less frequently than expected, language learners may assume a higher value of this constraint, which then leads to even fewer violations in production etc. (similarly for violations occurring more frequently: language learners assign a lower value, which shows in even frequencies of violations in production). When the value of -VF reaches such a high number that violations are practically non-existent, the language learner would lose the evidence for more than one position in the prefield and reanalyse the sentence structure as a conflated CP-pattern with just one CP (alternatively, if we follow Grewendorf ’s analysis, the reanalysis consists of a ban on long movement over a trace in FinP). The few instances of double prefield filling which such a language learner might occasionally hear would be analysed as remnant movement; this language learner would admit doubly filled prefields as instances of
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
remnant movement in his own grammar and occasionally even use them—but only those that are compatible with the remnant movement analysis. This would explain why modern instances of multiple prefield filling are usually compatible with a remnant movement analysis. .. Excursus on post-initial connectives It is hard to find clear and independent empirical evidence for the existence of a splitCP in German. The examples discussed so far involve two phrases preceding the finite verb in Fin (or C) and this can be captured either with a remnant VP analysis or with assuming multiple specifiers (see above). There is, however, another kind of data which may provide better evidence for a split-CP in German: () a. Peter nämlich liebt die Gefahr Peter namely loves the danger ‘Peter namely loves danger.’ b. Peter allerdings bevorzugt . . . Peter however prefers ‘Peter, however, prefers . . . ’ c. Die Brigitte nun kann der Hans nicht ausstehen the Brigitte now can the Hans not stand ‘Brigitte, however, Hans cannot stand.’ In the sentences in () there are also two constituents which precede the finite verb, but in these cases the second one is an adverbial connective (see also Chapter ). Adverbial connectives which can appear in this special position between an XP in the prefield and the finite verb in Fin/C are called nacherstfähig (‘capable of post-initial position’) in German descriptive literature (e.g. Breindl ), and sentences such as in () are often considered to be V-clauses (see Breindl : ). As shown in Volodina and Weiß (), there are in principle two possibilities to analyse the syntax of connectives in post-initial position: (i) the XP and the connective form a single phrase which is built in the middle field and then raised to SpecCP; (ii) the adverbial connective is merged in Top and the preceding XP is moved from its base position to SpecTopP. The first analysis would be in accordance with the V-property of German, the second one with a cartographic structure. Although it is hard to decide which analysis is the appropriate one only on the basis of simple sentences such as (), more complex sentences (see ()) and the semantics of the connectives—they mark topic change (Onea and Volodina )—speak in favour of the second analysis (Weiß ): () a. Peter nämlich, der liebt die Gefahr Peter namely, the loves the danger ‘Peter namely, he loves danger.’ b. Die Antwort freilich, die Thales gibt, ist . . . the answer admittedly which Thales gives is ‘The answer, admittedly, which Thales gives is . . . ’
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer and Helmut Weiß
(a) is a left dislocation structure where, according to the assumptions made by Grewendorf (), an XP is raised to SpecTopP via SpecFinP leaving behind a resumptive d-pronoun in that position, see (a). A comparable analysis is possible for (b): the complex DP [die Antwort, die Thales gibt], presumably forming a big DP (Grewendorf ), is raised first to SpecFinP, where the upper part of the big DP is extracted and raised to SpecTopP, see (b). () a. [TopP Peter [Top0 nämlich] [FinP der [Fin0 liebt . . . b. [TopP Die Antworti [Top0 freilich] [FinP [ ti die Thales gibt [Fin0 ist . . . According to this analysis, the left periphery of German sentences would be more complex than the common V-analysis assumes.6 Interestingly, post-initial connectives already occurred in the earliest OHG texts (see also Chapter for further details). The adverb/conjunction aber ‘but’ is probably the lexeme which is attested earliest in this position: Example (a) is from the OHG Isidor, which, however, mirrors the word order of the Latin original (filius autem datus est nobis). This is mostly the case with OHG examples. However, at least since MHG times, we find examples from texts which are not translations from Latin (see (b)), so we can conclude that the post-initial positioning of these adverbial connectives reflects a genuine option of German syntax (see also Chapter on this point). () a. Sunu auur uuard uns chigheban, huues nibu gotes sunu? son but was us given whose if.not God’s son ‘The son is given to us, whose if not God’s son?’ (Isidor , ) b. der selbe rîter aber sprach the same knight but said ‘The same knight, however, said’ (Parzival , )
. Summary The one-constituent property of Modern German declarative main clauses was basically in operation already in MHG in that most declarative main clauses conformed to this property. The decision regarding which phrase is moved to the prefield is information-structurally governed in Modern German. In MHG and ENHG, roughly the same information-structural tendencies hold, namely that scene-setting elements are the preferred prefield fillers in cases of competition with other potential prefield fillers (contrastive elements, topics). In contrast to Modern
6
For more details of, and problems with, this structural proposal see Volodina and Weiß (); Weiß (); and Catasso (). For an alternative treatment of post-initial connectors such as aber see also Chapter ....
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The prefield after the Old High German period
German, the competition is not so clear-cut in disfavour of topics in MHG. Another difference between MHG and ENHG is that in MHG, the prefield is more often filled not by movement, but by base-generation of elements indicating the logical connection to the preceding sentence; in ENHG the development towards the Modern German filling by movement is visible. The one-constituent property is not so established that it does not allow for exceptions; these exceptions can be captured by a cartographic analysis.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
6 Adverbial connectives GISELLA FERRARESI
. Introduction Adverbial connectives are elements such as allerdings ‘indeed’, immerhin ‘anyway’, and aber ‘but’, which are comparable to the Modern English however and indeed (Quirk et al. ) and to OE soþlice ‘truly’ and witodlice ‘certainly’ (Lenker ). Their function is to express a specific semantic relationship between the clause in which they are contained and a preceding or a succeeding clause, and they thus resemble subordinating conjunctions. In the following examples, the adverbial connective danach ‘after that’ in (a) introduces a temporal relation, and the adverbial connective deshalb ‘because of that’ introduces a causal relation in (b): () a. Maria ging nachmittags einkaufen. Danach ruhte sie sich aus. Maria went in.the.afternoon shopping after.that relaxed she REFL out ‘Maria went shopping in the afternoon. After that, she relaxed.’ b. Maria hatte lange gearbeitet. Deshalb war sie so müde. Maria had long worked therefore was she so tired ‘Maria had been working for a long time. That is why she was so tired.’ In most cases, the same kinds of relationships can also be expressed through subordinating conjunctions, as in the following examples (again, (a) and (b) represent the temporal and the causal relationship, respectively): () a. Nachdem Maria eingekauft hatte, ruhte sie sich aus. after Maria shopped had relaxed she REFL out ‘After Maria had gone shopping, she relaxed.’ b. Maria war so müde, weil sie lange gearbeitet hatte. Maria was so tired because she long worked had ‘Maria was so tired because she had been working for a long time.’
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Gisella Ferraresi . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
The difference between adverbial connectives and subordinating conjunctions is that the former connect two sentences that are syntactically independent of each other. This can also be seen in examples () and () as the adverbial connective does not trigger verb-final order. Moreover, in contrast to subordinating and coordinating conjunctions—but similar to adverbs—adverbial connectives can take different positions within the clause. For example, they can remain in the middle field, as in (), and some, such as aber and allerdings, can appear in the prefield following a constituent, as in ():1 () a. Maria hatte lange gearbeitet, sie war deshalb so müde. Maria had long worked she was therefore so tired ‘Maria had been working for a long time. That is why she was so tired.’ b. Maria ging nachmittags einkaufen. Sie ruhte sich danach aus. Maria went in.the.afternoon shopping she rested REFL afterwards out ‘Maria went shopping in the afternoon. She rested afterwards.’ () a. Maria und Jan waren verabredet, Maria aber kam zu spät. Maria and Jan were agreed Maria however came too late ‘Maria and Jan had an appointment. Maria, however, came too late.’ b. Maria wollte schon immer nach Kanada. In späteren Jahren Maria wanted already always to Canada in later years allerdings interessierte sie sich mehr für asiatische Länder. however interested she REFL more for Asian countries ‘Maria had always wanted to go to Canada. In later years, however, she was more interested in Asian countries.’ There are adverbial connectives with an adversative meaning, such as aber ‘but, however’ and allerdings ‘though’, which have no corresponding subordinating conjunction: () Maria ist nett, aber sie kommt immer zu spät. Maria is nice but she comes always too late ‘Maria is nice, but she is always late.’ In the first position, aber has been analysed as a coordinating conjunction such as und ‘and’ and oder ‘or’ since it does not trigger verb-subject inversion there (Eisenberg : ). Diachronically, the majority of adverbial connectives developed during the ENHG period, but several, such as aber ‘but’, emerged earlier in OHG and are also more grammaticalized, as will be shown later.2 The adverbial connective aber, for example, is nowadays used as a coordinating element. 1 This position is called post-initial position (Nacherstposition) in Pasch et al. (), see also Chapter of this volume on post-initial elements. For a general overview of fields and positions within the topological model (field model) of the German clause see Chapter . 2 The adverbial connective aber is the most grammaticalized of the adverbial connectives and can have the meaning of a coordinative conjunction, such as ‘but’, as well as of an adverbial connective, such as
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
Except for those containing a deictic element,3 adverbial connectives arose from adverbials in a rather cyclical change from VP to sentence adverbs, similar to the rise of English adverbial connectives described in Swan (), Traugott (), and Lenker (). In this sense, their reanalysis could be seen as an upward reanalysis (Roberts and Roussou ). This chapter is organized as follows: Section . discusses the emergence and development of the connective aber ‘but, however’ from the original OHG and MHG adverb afur/avur with restitutive and repetitive meaning (Section ..). Aber ‘but’ then continued to develop into a coordinating element (Section ..). In Modern German, however, aber is only syntactically a coordinating element in the first position; semantically, it is still an adverb. In later positions, it should also syntactically be considered an adverbial element, which is discussed in Section .... In the second part of the chapter, I consider the rise of the adverbial connective allerdings, whose emergence as a connective dates to the ENHG period. Allerdings developed as an adverbial connective from a quantifying adverb. At the beginning of the second part of the chapter, the use and syntactic analysis of Modern German allerdings is presented in Section .. before the diachronic data of the ENHG completive adverbial aller dinge, from which allerdings derives, as well as its shift from a completive to a sentence adverb and later to a connective element are described in Section ... I will show that allerdings underwent a cyclical change from a VP-adverb to a CP-element and that aber changed to a coordination phrase (CoorP) exponent similar to the manner in which English adverbial connectives are analysed by van Gelderen (). Pragmatic as well as prosodic factors played a role in the reanalysis during the different stages of the development.
. The rise and development of aber: from restitutive VP-adverb to CP-connective element In this part of the chapter, I discuss the single steps in the grammaticalization of aber from a VP-adverb to a CP-connective element, which is representative of the process by which many adverbs in different languages are grammaticalized to connective elements.
‘however’. Moreover, just like und ‘and’ and oder ‘or’ but unlike other adverbial connectives, such as danach ‘afterwards’ and deshalb ‘therefore’, aber does not syntactically trigger subject–verb inversion when it occupies the first position, as in the examples in (). 3 The adverbial connectives with a deictic element (da-, des-, -dem ‘this/that’), such as deswegen ‘because of this’ or danach ‘after that’, and trotzdem ‘in spite of this’, represent a special class of adverbial connectives. Their rise and development also proceeded in a different manner (Ferraresi ), but they are not discussed further in this chapter.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
.. OHG afur as a restitutive and repetitive adverb and its replacement by wieder In its original function, OHG afur/aber4, 5 was an adverb that was ambiguous in its repetitive meaning of ‘again, one more time’ and its second restitutive meaning of ‘back’, as the following examples from Otfrid show: () sie fárent thines férehs mit selb stéinonne; nu súachist they sought your life with even to.stone now you.search sie afur thánne? REPETITIVE them again there ‘Last time, they wanted to stone you to death; are you now searching for them there again?’ (Otfrid III , ) () joh bráhta sa afur tháne zi themo ira héiminge RESTITUTIVE and brought her again there to the her hometown ‘and brought her back to her home town’ (Otfrid I , ) In (), the event of searching is repeated, whereas in (), Mary and Joseph are returning back from the census, thereby restoring the previous state of being in the same place. In Modern German, aber (< OHG afur/avur) has lost both meanings and has been replaced by wieder ‘again, back’.6 The difference between the repetitive and the restitutive interpretations is that in the former case, an entire event happened for a second time (and two events are thus described in the sentence); in contrast, with the restitutive meaning, a state that had previously been altered was restored by an intervening event. In this case, the event in question is singular (Kamp and Roßdeutscher ; Fabricius-Hansen ; Pittner , ). According to Pittner’s () analysis, the different interpretations of such adverbs as wieder ‘again’ result from different scopal properties rather than from different lexical elements. Pittner argues that there is one single lexical entry with underspecified semantics. Syntactically, in the repetitive interpretation, the entire eventuality (i.e. the VP with all its arguments) is within the adverb’s scope. The restitutive interpretation, on the other hand, is only triggered when the inner state (i.e. the minimal verbal domain V) is within the scope of the adverb (Pittner ). I assume that OHG afur/abur can be analysed as Pittner proposes for wieder ‘again’ in that it can be interpreted as being repetitive or restitutive as a single lexical
However, aber with this meaning could still be found in the NHG adverb abermals ‘again, once more’. In OHG, there was no single standard spelling; instead, there were different notations for aber, i.e. afar, avar, afur, avur, aber. Etymologically, aber derives from PIE *apo, *apu ‘away, off ’ (Pfeifer : ). 6 NHG wieder derives etymologically from the local preposition wider ‘against’, as is shown in the example from Tatian (: ): Ther mit mir nist ther ist uuidar mír ‘He who is not with me is against me’, which was reanalysed as an adverb with the meaning of ‘again, back’ in the eighth century similar to the English again(st) (Pfeifer : ). The preposition wider is still used in Modern German. 4 5
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
item depending on its syntactic position.7 This is shown by the fact that the OHG adverb abur in (a) can only have a repetitive meaning in initial position, similar to NHG wieder in the prefield in (b), which contains the entire eventuality within its scope: () a. (Iterum assumit eum diabolus in montem excelsum valde) Abur nam inan ther diuual thô in hohan berg thrato again took him the devil then in high mountain quite ‘The devil took him up into an exceedingly high mountain again.’ (Tatian , –) b. Wieder hat Hans das Auto repariert. Again has Hans the car repaired. ‘Hans has repaired the car again.’ (Pittner : ) Abur in (a) signals that the event does not take place for the first time, but is repeated. Accordingly, we find the same sentence in the preceding text, line : thô nam inan ther diuual ‘Then, the devil took him’. As an adjunct to the V-position (as in (a)), the reading of abur is ambiguous and could be either restitutive or repetitive, as the NHG example with wieder (b) also shows. Both abur in (a) and wieder in (b) are ambiguous because of the syntactic ambiguity. The position of the object can be analysed either as the base position in the VP, in which case abur in (a) and wieder in (b) are adjoined to the V, or as a higher position to which the object has been scrambled. In this case, abur and wieder would also be adjoined to a higher verbal projection. In the former case, the restitutive interpretation is triggered, whereas in the latter case, the repetitive interpretation is possible: () a. (Vocaverunt ergo rursum hominem qui fuerat coecus) gihalotun then man abur ther blint uuas called the man back/again who blind was ‘They called the blind man again.’ (Tatian , –) b. Hans hat das Auto wieder repariert. Hans has the car again repaired ‘Hans has repaired the car again.’ (Pittner : )
7 The verb type also plays a role in the interpretation. The unmarked restitutive interpretation as a process-related adverb is triggered by telic verbs of state change (e.g. a verb such as ‘open’). However, with a non-cumulative predicate, such as auflachen ‘to give a laugh’, which is inherently minimally quantified, only the repetitive meaning is possible; the restitutive reading would not make sense since no previous state or change of state is relevant (Fabricius-Hansen : ):
(i) Er lachte zu Hause wieder auf. ‘He burst into laughter again at home.’
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
The comparison with the Latin text reveals that the OHG Tatian translators chose the order DO-Adv, whereas the Latin order is Adv-DO. In Latin, rursum can also have the meaning of ‘again’ and ‘back’.8 Semantically, both adverbial readings trigger a presupposition: the repetitive reading implies an event of the same type that has already occurred, and the restitutive reading presupposes a previous state that has been restored after an intervening change of state (Kamp and Roßdeutscher : ; Fabricius-Hansen : ; Pittner ). The state of the car of being repaired in (b) contrasts with the car’s preceding state of being out of order. The use of aber/wieder also implies that the car had broken down before. The previous state before the changing event is identical with the one asserted in the clause (Fabricius-Hansen : ; see also Kamp and Roßdeutscher ). These two interpretations imply a contrast that emerges only when related to the concept of similarity or identity of a particular dimension. The similarity is implied by the repetitive reading, the identity by the restitutive reading. The feature of contrast expressed through the adverb aber also characterizes the adversative conjunction (Lang , ). It can be assumed that a change in the syntax and semantics of aber began during this stage of OHG since we find examples in Tatian in which the restitutive meaning is given by uuidar/wieder, as shown by example (): () (nam remisi vos ad illum) ich santa iùuuih uuidar ziimo I sent you again/back to.him ‘I sent you back to him.’ (Tatian , ) Interestingly, in parallel gospel verses, as in the following ones that describe how Pontius Pilate went back into the judgement hall, Tatian has abur, as in example (a), and Otfrid uses widar, as in (b): () a. (Introivit ergo iterum in pretorium Pilatus) Ingieng tho abur in thaz thinchus Pilatus went then back into the judgement.hall Pilate ‘Pilate then went back into the judgement hall.’ (Tatian , ) b. Giang Pilatus wídari / mit ímo tho in then sólari went Pilate back with him then into the judgement.hall ‘Pilate then went back with him into the judgement hall.’ (Otfrid III , ) Moreover, the Latin word regeneratio ‘reincarnation’ is translated in Tatian (, ) with aburburt, whereas Notker (Psalter , ) uses uuidiburt. There are also some examples in which both adverbs are used together, as in the following example (). In (a) and (b), the restitutive and the repetitive meanings
8
The etymology of rursum is *reuorsom < *red-wrt-to- ‘turned to the back’ (Sergio Neri p.c.).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
are realized via uuider/wieder and aber, respectively. This situation can be found not only in OHG (a) but also later in MHG (b): () a. (postestatem ponendi animam suam et iterum sumendi eam) uuanda er keuualt sînin lib ze lâunde aber uuider ze because he the.power his soul to yield again back to nemenne hábeta to.take has ‘because he has the power to yield his soul and to take it back again’ (Notker, Psalter , ) b. daz er aber wider begunde . . . that he once again began ‘that he once again began . . . ’ (Tristan ) In (a), wider nehmen is a translation of the Latin verb sumere ‘take to oneself ’, to which wider confers a restitutive meaning, whereas the repetitive adverb iterum ‘again’ is rendered via aber. As can be seen in the examples in (), the repetitive uuider is always closer to the verb than is the restitutive aber, whose different scopes are reflected in its respective interpretations. Similar to the OHG and MHG examples in () in which both afur and wieder are used in the same sentence with different meanings, in NHG (where wieder has replaced afur in both restitutive and repetitive meaning), it is possible to utter wieder twice in the same sentence. The wieder closer to the verb, i.e. the deeper one, thus realizes the restitutive reading, whereas the higher wieder realises the repetitive meaning, as in the following example: () Er hat das Auto wieder[repetitive] wieder[restitutive] repariert. he has the car once.again again repaired ‘He has once again repaired the car again.’ (Pittner : ) However, afur is still used as a repetitive adverb in later texts, as in the MHG example in () and in the ENHG example in (): () die porte wurden zuo getân, the gates were closed dâ sî durch was gegangen: when she through was passed und er was alsô gevangen and he was thus captured daz im aber diu ûzvart such.that to.him again the way.out anderstunt versperret wart. once.again blocked was ‘The gates were closed after she had passed through, and he was thereby captured as that the way had once again been blocked to him.’ (Iwein –: twelfth century)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
() (manifestavit se iterum) offenbarte sich aber revealed himself again (Luther-Bibel , (Joh ,)) The different interpretations of wider/wieder and afur/aber in this stage of MHG are also reflected in compounds with the same verb, where afur bears the repetitive component and widar the restitutive one: () widarsagen ‘reply’ versus afursagen ‘repeat’ widarbringan ‘bring back’ versus afurbringan ‘bring again’ Only after the end of ENHG and during the NHG period did wieder completely replace aber in its adverbial function. The scopal properties of restitutive afur/aber and later of uuider/wieder are similar to the properties of manner adverbs (Pittner ), which are process-related adverbs according to Frey and Pittner’s () and Pittner’s () proposal.9 Manner adverbs are closest to V and modify only the minimal verb domain. Repetitive afur/aber and uuider/wieder, by contrast, take a syntactically higher position and behave like event-related adverbs, such as adverbs of time. The relative order of the adverbial classes mirrors the relation between the syntactic positions of the heads that the adverbs modify and specific semantic entities, i.e. their interpretation is determined by the semantic entities to which they are related. Frey and Pittner () and Pittner () distinguish five adverbial classes: () (I) frame, (II) sentence, (III) event-related, (IV) event-internal, and (V) process-related adverbs. The relative order of the single adverbial classes is not syntactically determined. Adverbs’ position and c-command relations depend on their semantic relations to the rest of the sentence: () Base positions of adverbs: (i) Frame adverbs c-command the base positions of sentence adverbs. (ii) Sentence adverbs c-command the finite verb and the base positions of event-related adverbs.
9
According to Frey and Pittner (), Pittner (), and Frey (), adverbs are to be analysed as adjuncts. In other frameworks, such as in Cinque (), adverbs are in the specifier positions of functional heads, which have a rigid order, whereas Haider () argues for a system in which adverbials that follow the head of the phrase to which they relate are considered to be embedded and adverbials that precede the head are considered to be adjoined. Similar to Frey and Pittner (), according to Haider, adverb ordering is semantically conditioned and does not follow from syntactic structuring. See Pittner () for different syntactic and semantic tests (e.g. the position and interpretation of indefinite pronouns or the relative order of different adverb types) that show that the restitutive wieder behaves like process adverbs whereas the repetitive wieder behaves like event-related adverbs.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi (iii) Event-related adverbs (e.g. time, cause) c-command the base positions of the highest argument and the base positions of eventuality-internal adverbs. (iv) Event-internal adverbs (e.g. instrument, comitative, mental attitude) are minimally c-commanded by the argument to which they are related (i.e. no other argument can intervene). (v) Process-related adverbs (e.g. manner) minimally c-command the verb or verbal complex.
Such adverbs as aber and later wieder are unselective adverbs that occur both as event and process adjuncts. In Cinque’s () adverb hierarchy, the different classes of adverbs are ordered according to the heads in whose specifiers they are contained: () [franklyMood-speech [fortunatelyMood-evaluative [allegedlyMood-evidential act [probablyMood-epistemic [onceT(Past) [thenT(Future) [perhapsMood-irrealis [necessarilyMood-necessity [possiblyMood-possibility [usuallyAsp-habitual [againAsp-repetitive(I) [often[intentionallyMood-volitional [quicklyAsp-celerative(I) [alreadyT Asp-frequentative(I) [no longer Asp-terminative [stillAsp-continuative [alwaysAsp-perfect [justAsp(Anterior) [soon retrospective Asp-proximative [brieflyAsp-durative [characteristicallyAsp-generic/ [almost Asp-prospective [completelyAsp-SgCompletive(I) [tuttoAsp-PlCompletive progressive [wellVoice [fast/earlyAsp-celerative(II) [againAsp-repetitive(II) [oftenAsp-frequentative(II) [completelyAsp-SgCompletive(II) (Cinque : ) Restitutive afur occupies the specifier of quite a low aspectual phrase (Asp-repetitive(II)), whereas repetitive afur is in the higher aspectual phrase (Asprepetitive(I)). Later, towards the end of the OHG period, afur/aber was no longer a process-related adverb, as the examples in () reveal, but became reanalysed as an event-related adverb and lost the property of being unselective. Wieder became used in its place in the restitutive meaning and later also in the repetitive one, thereby entirely replacing aber in the adverbial function. Syntactically, the reanalysis of afur from a VP adverb to an event-related adverb can be seen as upward movement (Roberts and Roussou ) or as a late merge in the sense of van Gelderen (b: ). In the following section, the reanalysis of afur/aber from an adverb to a connective element is discussed. .. From an adverb to a connective adversative element The connective function of afur/aber appears to have first developed in OHG since Notker translates the Latin adversative conjunction sed ‘but’ with afur/aber in some cases. In Isidor and Tatian, however, sed is translated with oh/ouh, from which the focus particle auch ‘also, too’ derives. There is an interesting correlation between afur/aber ‘but’ and auch ‘also’, which Umbach (, ) and Sæbø () have discussed for Modern German and English, as both items imply contrast. Before going into the diachronic development of aber from an adverb to an adversative
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
coordinating element, this point is briefly discussed and clarified in order to better understand the diachronic process of grammaticalization that led to the reanalysis of afur as a conjunction. As seen in the discussion of the restitutive and repetitive adverb, contrast is an essential component of both readings and is also the source of the diachronic reanalysis. ... The component of contrast According to Lang (, ), Umbach (, ), Sæbø (), Repp (), and others, the adversative coordinating element but signals a contrast in both its corrective and contrafactual uses. English has one single lexical item, but, for both meanings, whereas other languages have distinct lexical items. German has aber for the contrafactual meaning and sondern10 ‘but’ for the corrective: () a. John is not big, but athletic. b. John ist nicht dick, sondern kräftig. Corrective sondern is used as a conjunction in Notker (Schrodt : ). With corrective but/sondern, the proposition in the first conjunct is denied and opposed to the affirmed true proposition in the second conjunct. In the contrafactual use of aber/but, the contrast is between two states-of-affairs. The state-of-affairs denoted in the second conjunct denies the conclusion that can be inferred by relating the two states-of-affairs denoted in the two conjuncts. The utterance q but p corresponds to q (and therefore ¬p), but actually p (Lakoff ), whereby therefore ¬p is not uttered but corresponds to an inference based on actual or world knowledge. This can be seen by comparing examples (a) and (b), in which the two sentences are conjoined once via the conjunction aber/but and once via the conjunction und/and: () a. John John b. John John
ist is ist is
klein, small klein small
aber but und and
er he er he
spielt plays spielt plays
Basketball. basketball. Basketball. basketball.
In (a), the inference ¬p carries the meaning that ‘small people normally do not play basketball’. The second conjunct contradicts this inference and receives the greater argumentative value. The same inference also arises with the conjunction and in (b); however, the contrast with the inference is explicitly lexically realized with but. The different readings (e.g. contrastive and concessive) of the relation between the conjuncts are due to the interplay between diverse factors, such as syntactic position, prosody, and the information structure (Lang , ; Umbach , ; Repp ). Lang () identifies the requirement imposed on both conjuncts (or on general elements) connected by coordinating conjunctions
The conjunction sondern ‘but rather’ derives from the preposition suntar ‘without’, was later used as an adverb sonder ‘aside’ (Pfeifer : ), and has been used as an adversative coordinating element since the late OHG period (Schrodt : ). 10
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
as the similarity-plus-dissimilarity condition. This is reminiscent of the identity and similarity condition in the interpretation of the repetitive and restitutive meanings of the adverbs aber/wieder, as seen in Section ... As far as the coordinating structure is concerned, according to Lang (), the coordinated elements must be semantically independent (i.e. not in a relation of hyponymy (one subsuming the other)); moreover, they must have a ‘common integrator’ (i.e. they must be subsumed under the same concept).11 However, the most relevant differences between und/and and aber/but are ) that the focus of the second conjunct influences the contrast induced by but and ) that if a but sentence is an appropriate answer to a question comprising two conjuncts, one of the conjuncts will be confirmed and the other one will be denied.12 Without going into the details concerning the focus relations between the conjuncts connected via aber/but, what is relevant for the discussion of the reanalysis of aber is that aber induces contrast in every step of grammaticalization. With the restitutive adverb, the contrast is between states; with the repetitive adverb, it is between events; and with the coordinating element, it is between utterances depending on the focus relations between both conjuncts, as will be discussed in Section .... For example, in a sentence like the following, which contains wieder, the inference raised in the first conjunct is that John has not run for a while due to illness; however, in the second conjunct, the inference is denied and the contrary is affirmed (i.e. that he is running again), thereby restoring the previous situation: () (John läuft regelmäßig. ‘John regularly goes running.’) John war krank, nun läuft er wieder. John has been ill now runs he again ‘John has been ill, but he is now running again.’ With the repetitive meaning, the event expressed in the sentences containing afur/ wieder is presupposed to have already occurred; with the restitutive meaning, the state expressed in the sentences is recovered. The presupposition arises in the same sentence containing afur/wieder, whereas with the contrastive meaning, the inference arises dependent on the first conjunct.
For example, in (i), the sentence sounds odd since ‘drink’ subsumes the meaning of ‘Martini’; in (ii), the common integrator makes clear that port is a drink and not a harbour (Umbach : ): 11
(i) ??John had a drink, and/but Mary had a Martini. (ii) John bought a beer, and/but Mary bought the port. For example, with and, the answer to the question ‘Did John clean up his room and wash the dishes?’ must contain both positive conjuncts; however, with but, it must contain one negative conjunct (Umbach : ): 12
(i) (ii) (iii) (iv) (v) (vi)
[yes] John cleaned up his room and [yes] he washed the dishes. ??[yes] John cleaned up his room, but [yes] he washed the dishes. ??[no] John did not clean up his room, but [no] he did not wash the dishes. [yes] John cleaned up his room, but [no] he did not wash the dishes. [yes] John cleaned up his room, but [no] he forgot to wash the dishes. [no] John did not clean up his room, but [yes] he did wash the dishes.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
In the following section, the syntax of OHG afur is compared with the positions of aber in Modern German in order to establish whether both elements belong to the same class and exhibit the same syntactic behaviour. ... The syntactic position of OHG afur/aber Returning to OHG, we find that afur/aber can take different positions: preverbal in the very first or in the second position after a constituent and post-verbal in the middle field. However, a different usage of afur/aber can be found in OHG texts (see also Schrodt : ). Table . displays the frequency of afur/aber in the different positions in the texts in relation to the corresponding Latin items if these are also realized in the Latin text.13 As Table . reveals, in Tatian, afur/aber in the first position always corresponds to iterum ‘again’ or rursum ‘back’ but never to sed when there is a correspondent item in Latin, as example () demonstrates: () (Iterum assumit eum diabolus in montem excelsum) Abur nam inan ther diuual thô in hohan berg again took him the devil there to high mountain ‘The devil took him up into a high mountain again.’ (Tatian , –) In Isidor, there is only one case with auur in the first position without any Latin correspondence, as seen in the following example: () (Idem quoque in psalmis) Auur auh umbi dhazs selba quhad dauid in psalmom but also because that same says David in psalm ‘Therefore, David also says the same thing in the psalm.’ (Isidor , ) Only in Notker aber translates Latin sed in the first position: () (Sed dominus suscepit me) Aber Got unterfiêng mich but God protected me ‘But God protected me.’ (Notker, Psalter , ) Aber can be found in the second position after the first constituent, or X (as translation and in the same position as the Latin particles enim, autem, vero, and ergo, which always follow the first constituent), in Isidor and Tatian, but not in Notker—although the string XP-aber is also used in Notker. An example from Isidor
13
In some cases in the Latin text, verbs with the prefix re- are translated in OHG with aber/wieder, as are the Latin renatus fuerit and OHG abur giboran uuerde ‘was born again’ (Tatian , ). These cases are subsumed in Table . under ‘no distinct lexical item’.
PREVERBAL
POSTVERBAL
(aber XP V)
after first (XP-aber V)
st
LATsed (st position)
LAT iterum rursum denuo (in different positions)
LAT autem igitur enim vero ergo (nd position)
LAT LAT LAT no distinct sed iterum rursum lexical denuo item (in different positions)
LAT autem igitur enim vero ergo
LAT no distinct lexical item
LAT sed (st position)
LAT iterum rursum denuo (in different positions)
LAT autem igitur enim vero ergo (nd position)
LAT no distinct lexical item
Isidor
–
–
–
–
–
–
Tatian
–
–
–
–
–
–
–
–
Notker
–
–
–
–
–
–
–
–
–
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
T . Position of aber in Isidor, Tatian, and Notker
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
is provided in (), where OHG avur is used to translate autem and appears to the right of the first constituent, sunu ‘the son’: () (Quod enim homo factus est, nobis profecit, et ideo nobis natus est. Filius autem datus est nobis, cuius nisi dei filius?) Huuanda chiuuisso, dhazs ir man uuardh uuordan, because surely that he man is made unsih hilpit, endi bidhiu uuard ir uns chiboran. us helps and because was he us.DAT born Sunu auur uuard uns chigheban, huues nibu gotes sunu? son but was us.DAT given whose if.not God’s son ‘Because verily He has made man, He helps us, and because a Son was born unto us. The son is given to us, whose if not God’s son?’ (Isidor , ) In post-verbal positions in the middle field, aber in Isidor’s text is used to translate different elements from Latin, such as autem in example (a); whereas in Tatian, it is only used to translate iterum and rursum (b): () a. (Sepulchrum autem eius in tantum est gloriosum) Sn grab ist auur so drado îruuirdhic his sepulchre is however so much glorious ‘However, his sepulchre is so glorious.’ (Isidor , ) b. (Et respondens Ihesus dixit eis iterum in parabolis) Thô antlinginti ther heilant quad in abur in râtisson there answering the healer said them again in parables ‘And Jesus answered them again in parables.’ (Tatian , –) In Notker, aber can also occupy inner-clause positions that differ from the Latin order. In the following example, post-verbal aber is used to translate Latin sed in the first position: () (sed confidite) uuésent aber balt be however confident ‘However, you should be confident.’ (Notker, Psalter , ) Since afur/aber differs in meaning and function in different positions in the various texts, I will discuss them separately in Sections ... and .... First, however, I briefly present the data and analysis of Modern German aber in order to compare them with the historical data and the analysis. ... Excursus: Modern German aber Syntactically, Modern German aber ‘but, however’ differs from English but since it can appear not only preverbally in the first position of the second conjunct (before any other constituent (a), much like English but) without triggering subject–verb inversion, but also (like adverbs)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
post-verbally in the middle field (b) and preverbally after one single constituent (c), similar to English however and OHG afur: () Was machen denn deine Eltern? ‘How are your parents doing?’ a. Mein Vater ist krank, aber meine Mutter geht arbeiten. my father is sick but my mother goes.out work ‘My father is sick, but my mother goes to work.’ b. Mein Vater ist krank, meine Mutter geht aber arbeiten. c. Mein Vater ist krank, meine Mutter aber geht arbeiten. (Lang and Adamíková : ) In the first position of the second conjunct, as in examples (a) and (b), aber can be analysed as a coordinating element exponent of CoorP (a) like und ‘and’ (c) (Repp ; Vincente ; Büring and Hartmann among others): () a. [CP connect [CoorP aber/but [CP connect b. [[CP1 Mein Vater ist krank [CoorP aber [CP2 geht arbeiten]] my father is sick but goes work ‘My father is sick, but he goes to work.’ c. Mein Vater ist krank [CoorP und [CP2 geht arbeiten. my father is sick and goes work d. *[[CP1 Mein Vater ist krank [CoorP und [CP2 aber geht arbeiten]] In the first position, aber is incompatible with und ‘and’ and competes for the same position (d), which is similar to English, where the sequence *and but is excluded. With coordinating elements, it is possible to have different kinds of ellipses in either the first or second conjunct (Hartmann for an overview). For example, in (b), the subject in the second conjunct is identical to that of the first conjunct and can be elided. In a coordinating structure, there is an asymmetric c-command relation, as binding data show (Hartmann : ff): () [die Bewerberi kennen die Jury nicht], [CoorP aber [sie kennen the candidates know the jury not but they know die Gerüchte über sich selbsti/*siei/einanderi]] the rumours about themselves/them/each other ‘The candidates don’t know the jury, however they know the rumours about themselves/them/each other.’ In positions other than the first one, such adverbial connectives as aber behave syntactically like adverbs but semantically like coordinative elements (Ferraresi ; Büring and Hartmann call these elements ‘particles’): aber can co-occur with asyndetic coordination in all positions (); however, aber can only occupy a position lower than the first one with a realized und ‘and’ (): () a. Mein Vater ist krank, aber meine Mutter geht arbeiten. my father is sick but my mother goes work ‘My father is sick, but my mother goes to work.’
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
b. Mein Vater ist krank, meine Mutter geht aber arbeiten. c. Mein Vater ist krank, meine Mutter aber geht arbeiten. () a. *Mein Vater ist krank und aber meine Mutter geht arbeiten. b. ?Mein Vater ist krank und meine Mutter geht aber arbeiten. c. Mein Vater ist krank und meine Mutter aber geht arbeiten. Unlike Büring and Hartmann (), who assume that adverbs occupy different functional specifier positions, I follow Frey and Pittner () in assuming that adverbs are adjoined to maximal projections. In the middle field, as in example (b), the position of aber correlates to different informational structures and seems to demarcate the division between the background and the focus domain (see also Umbach , ; Lang , ; Sæbø ; and Repp : ): () a. Meine Schwester hat mir ja (aber) gestern (aber) my sister has me.DAT yes (but) yesterday (but) ein Buch gegeben. a book given ‘However, my sister gave me a book yesterday.’ b. *Meine Schwester hat ja (aber) mir gestern ein Buch gegeben. If we assume that weak/non-stressed pronouns appear in the Wackernagel position (WP) (see Chapter of this volume), aber must—like modal particles—appear to the right of the WP in (a) in the left-most position of the middle field. Pronouns cannot follow aber or a modal particle such as ja, cf. (b) (Lenerz ) unless they are heavily accented (i.e. they receive narrow focus). A constituent in the prefield such as the one in example (c) receives a contrastive aboutness topic interpretation (Sæbø : ; Repp : ; Pasch et al. 14). The topichood of the constituent preceding aber in [DP aber], as in example (c), can be tested in left-dislocation and hanging-topic left-dislocation constructions, which, according to Frey (), have different informational status. Only left-dislocated constituents are true topics. As examples (a) and (b) reveal, only in (a)—a left-dislocated structure—is the presence of aber acceptable. By contrast, in (b), aber sounds odd with a hanging topic (Repp : ; Ferraresi ): () a. Meine Schwester aber, die hat mit dem Nachbarsjungen gespielt. my sister but she has with the neighbour’s boy played. ‘But my sister—she played with the neighbour’s boy.’ b. ??Meine Schwester aber laut Max wird sie erst morgen my sister but according.to Max will she only tomorrow kommen. come ‘But my sister—according to Max, she will come tomorrow.’ 14 Breindl () and Catasso () briefly discuss the pragmatics of the elements preceding aber in this position.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
According to Frey () and Repp (: ), the NP meine Schwester in the left dislocation is an aboutness topic in SpecCP and forms a syntactic chain with other elements in the clause. This is shown in example (), in which a left-dislocated element allows for binding with an element in the clause: () Seine1 Frau, die verwöhnt jeder Linguist1 mit großer Freude. his wife this coddles every linguist with great joy ‘His wife—every linguist coddles her with great joy.’ (Frey : ex. ()) Unlike left-dislocation constructions, hanging topics lie outside the clause. They do not display connectivity effects with the following clauses, e.g. they cannot bind elements in the clause (): () *Seine1 Frau, sie verwöhnt jeder Linguist1 mit großer Freude. his wife she coddles every linguist with great joy (Frey : ex. ()) We can assume that the adverbial connective aber ‘however’, which is acceptable only after left-dislocated elements but not after hanging topics, can only be adjoined to inner clause positions.15 By contrast, according to Hinterhölzl and Petrova (), aboutness topics in OHG occupy an extrasentential position outside the CP, with V as a neutral order (as will be further discussed below). Only after the reanalysis and the generalization of V order as a neutral order do aboutness topics become a part of the sentence. According to Frey (), left-dislocated elements are in SpecCP in Modern German and constitute the C-domain in a cartographic approach (Rizzi ) together with ContrP and FinP: () CP[
ContrP[ FinP[
C-domain
TopP[…
I-domain
The prefield in this model can therefore be realized via different positions in the C-domain. Complementizers are hosted in C. The specifiers of ContrP and FinP are A’-positions in which elements can move from the middle field or (in some cases) are base-generated (e.g. expletives). Movement to these positions is justified either for formal reasons (formal movement) or for informationstructural reasons (contrast).16 Elements with this feature [contrastive] move to SpecContrP.
15
Catasso () proposes that aber in post-initial position is base-generated in a lower TopP and moved to a higher TopP. 16 See also Chapters and .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
TopP is an A-projection, the highest position in the middle field preceding sentence adverbials in which elements with topical features that are moved from a lower middle field position are licensed, as in the following example: () (Ich erzähle dir etwas über Karl. ‘I will tell you something about Karl.’) Ich habe gehört, dass den Karl1 erfreulicherweise (??den Karl1) I have heard that the.ACC Karl.ACC fortunately nächstes Jahr eine polnische Gräfin t1 heiraten wird. next year a polish countess marry will ‘I have heard that a Polish countess will marry Karl next year.’ (Frey : ex. ()) The object, den Karl, has moved to TopP before the sentence adverbial erfreulicherweise ‘fortunately’ since Karl has already been introduced in the discourse and therefore has topical features. This topical element is unacceptable in the position after the adverbials. As () shows, aber can also adjoin to TopP: () (Ich erzähle dir etwas über Karl. Es hat länger gedauert. ‘I will tell you something about Karl. It has lasted longer.’) Nächstes Jahr wird den Karl1 aber eine polnische next year will the.ACC Karl.ACC however a polish Gräfin t1 heiraten countess marry ‘A Polish countess will marry Karl next year.’ Such topical elements as den Karl can also undergo formal movement to the preverbal position SpecFinP, to which aber can also adjoin: () Den Karl1 aber wird erfreulicherweise nächstes Jahr the.ACC Karl.ACC however will fortunately next year eine polnische Gräfin t1 heiraten a polish countess marry ‘Fortunately, however, a Polish countess will marry Karl next year.’ Unlike NPs such as den Karl in the example above, neither negative NPs, quantified NPs, nor sentence adverbials such as leider ‘unfortunately’ can be topics, even if they can move to the prefield via local movement, as example () illustrates: () Leider1 hat keiner t1 dem alten Mann geholfen. unfortunately has nobody the old man helped ‘Unfortunately, nobody helped the old man.’ (Frey : ex. (c)) Frey () thus concludes that negative NPs, quantified NPs, and sentence adverbials can move neither to SpecContrP nor to SpecTopP, but only to SpecFinP. This can be clearly seen in examples such as (a) that display long extraction to ContrP for information-structural reasons: genuine A’-movement to ContrP is not grammatical with sentence adverbials such as leider ‘unfortunately’, in contrast to argument fronting, with which long extraction is perfectly grammatical (b):
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
() a. *Leider1 sagte Karl, dass t1 keiner dem alten Mann unfortunately said Karl that nobody the old man ‘Unfortunately, Karl said that nobody helped the old man.’ (Frey : ex. (d)) b. Die Kaffeetasseni aber denke ich, dass Otto the coffee.cups.ACC however think I that Otto.NOM ‘However, I think that Otto has washed the coffee cups.’ (Frey : ex. (b), slightly modified)
geholfen hat. helped has
gespült hat. washed has
To consider the position of aber once again, it is clear that it can appear not only after elements that have been moved to ContrP, as in (b), but also after sentence adverbs in the prefield, as in (). As previously stated, these are not contrastive topics in ContrP but can only be moved via formal movement to SpecFinP according to Frey’s () proposal: () Unter der Leitung von erfahrenen Skilehrern konnten sich under the guidance of experienced skiing.instructors could REFL die Flawiler auf den Brettern [ . . . ] üben. the Flawiler on the skis practise Leider aber konnte das Schülerskirennen nicht durchgeführt unfortunately however could the learners.ski.race not undertaken werden. be ‘The Flawiler could practise skiing under the guidance of experienced ski instructors. Unfortunately, however, the beginners’ ski race was not able to take place.’ (St. Galler Tagblatt, May ) In summary, I assume that aber always has a semantically connective function. However, as it is in the very first position of the second conjunct, it is syntactically an exponent of CoorP (a). In lower positions, aber behaves syntactically as an adverbial element adjoining different clause-internal positions. In the prefield, aber can adjoin to SpecCP (b), SpecContrP (c) or SpecFinP (d),17 and in the middle field, aber can adjoin to TopP (e) or a lower position: () a. b. c. d. e.
[CP connect [CoorP aber/but [CP connect [CP meine Schwester [CP aber [CP [ContrP die Kaffeetassen ContrP[ aber [CP [ContrP [ FinP[ leider [FinP aber [CP [ContrP [ FinP Fin[wird] [TopP den Karl [TopP aber
(ex. (b)) (ex. (a)) (ex. ()) (ex. ()) (ex. ())
In the following sections, we return to OHG and discuss the different positions of afur/avur, which—like Modern German aber—also appear in the first position of the second conjunct or lower. In the next two sections, I therefore consider the question of whether OHG afur/aber is the same element as Modern German aber. 17
See, however, Chapter .. for an alternative analysis of post-initial connectors.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
... OHG afur/aber in the first position As stated in Section ..., only in Notker does afur/aber correspond to Latin sed in the first position. In contrast, in Isidor and Tatian, Latin sed is translated with oh/ouh for the contrafactual meaning (> auch ‘also, too’) and with nisi for the corrective meaning. Without going into detail on the use and development of the particle oh/ouh, according to Sæbø (), the meaning of aber/but is comparable with that of auch/also in that both imply the existence of another alternative.18 The difference between both elements is that with aber/but, the negation of the proposition inferred from the first conjunct is contained in the second conjunct. ‘With auch/also the proposition resulting from substituting the alternative for the topic of the sentence will be supported by the actual information state.’ (Umbach )19 In the first position in Isidor and Tatian, afur is still a repetitive/restitutive adverb always used to translate Latin iterum and rursum and thereby triggers subject–verb inversion (a) and also co-occurs with inti ‘and’ (b): () a. (Iterum dico vobis) Abur quidu ih iu again say I you.DAT ‘Again, I say to you’ (Tatian , ) b. (Et iterum dico vobis) Inti abur quidu ih íú and again say I you.DAT ‘And again, I say to you’ (Tatian , ) According to Behaghel (–, Vol. III: ), in instances in Otfrid such as the following, avur indicates the restoration of a previous state20, i.e. it has restitutive meaning: () tház er uns fon góte quam / ioh áuur sos er uuóolta that he to.us from God came and again as he wanted zi imo fáran scolta to him go shall ‘that he came to us from God and shall go back to Him as He wanted’ (Otfrid IV , )
18
According to Sæbø (), aber triggers a presupposition; however, I assume an implicature. Umbach (: ): ‘In both an alternative is added to the set of alternatives-under-discussion (clean up the room, wash the dishes). However, the additional alternative results in a false proposition when combined with the background part, thus requiring negation. In adding an alternative to those under discussion, but is comparable to and/also: 19
(i) John cleaned up his room, but he didn’t wash the dishes. (ii) John cleaned up his room. He did not also wash the dishes.’ 20 ‘Es bezeichnet die Wiederherstellung eines früheren Zustands’ [‘It denotes the restoration of a previous state’].
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
We now investigate whether aber in Notker can be considered a coordinating element like Latin but as Schrodt (: ) proposes. There are several arguments against this analysis, the first being that aber in Notker co-occurs together with unde ‘and’: () Sámo so uuir ne-uuízzîn uuaz uuir betoien uuir den ungesihtigen Got as.much as we not-know what we pray we the unseen God pétoien unde aber sie geséhen íro Góta pray and by contrast? they see/consider their God ‘Just as we do not know to whom we are praying when we pray to the unseen God, they experience their God in the same way.’ (Notker, Psalter , ) In Modern German and English, aber in the first position and but are coordinating elements and incompatible with other items of the same class. However, it is not entirely clear whether aber in Notker also has the meaning of ‘again’ or ‘by contrast’ as occurrences such as the following are ambiguous between both meanings when compared with the Latin: () a. (et iterum sumendi eam) unde aber uuider ze nemenne and however/again? again to taking ‘and to take it again’ (Notker, Psalter , ) b. et ad ea redeundo unde áber iruuíndendo and however/again? going.back (Notker, Psalter , ) Moreover, unlike Latin sed (), aber in Notker never precedes subordinating conjunctions of embedded sentences: () (Sed quoniam nondum tempus est firmioribus remediis) Uuánda aber nóh zît neíst stárcheren lâchenes since ABER no time NEG.is stronger remedies.DAT ‘But since it is not yet time for stronger remedies’ (Notker, Consolatio , –) Furthermore, whereas Latin shows the subordinating conjunction si following a contrastive conjunction such as sed in conditional clauses, Notker uses a V construction that leaves out the conditional conjunction and places aber in an inner-clause position: () (Sed si uelis replere indigentiam naturę quod satis est nihil est quod fortunę affluentiam petas) Uuíle du áber dés keuágo sîn dés tiu natura. want you however that.GEN satisfied be that the.DAT nature.DAT
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
bedárf sô nefórderôst tû nehéin úrgúse dero fortunę suffices so not.demand you no exuberance the.GEN fortune.GEN ‘If, however, you are satisfied with that which suffices in nature, do not demand more of fortune.’ (Notker, Consolatio , –) Similarly, aber does not precede wh-elements that introduce subordinate clauses with other subordinate clauses, which also contrasts with Latin: () (Sed quantum ornamentis nostris decesserit uides) Uuáz úns áber únserro êrôn enfáren sî dáz síhest tu what us however our brilliancy skirted is that see you ‘You see what part of our brilliance has been lost.’ (Notker, Consolatio , –) According to Axel () and Axel and Wöllstein (), adverbial clauses in OHG are not integrated into matrix clauses. Moreover, complement clauses should be considered correlative constructions. Based on the data discussed above, I conclude that aber in Notker is not yet syntactically a coordinating conjunction but rather still an adverbial element. Now that we have considered aber as a translation of sed in the first position, I will discuss aber as a translation for autem, vero, and enim in the so-called post-initial position in the following section. ... Aber in OHG as a translation for Latin second-position autem, vero, and enim In Latin, such particles as autem, vero, and enim, which always appear after the first constituent, structure the discourse by means of marking topical elements or parts of the discourse as foreground or background (see also Ferraresi and Goldbach for Latin and Gothic). According to Scholten (: , ) and Dittmer and Dittmer (: ), afur/avur as well as nu or thô (which were the correspondent forms in OHG in some cases) had a connective-structuring function in the preverbal position (see also Axel : ff). However, the use of aber is also not homogenous for function. As shown in Table . in Section ..., there are some instances of aber in postinitial position in Isidor and Tatian that correspond to Latin X(P)-autem; however, these are still adverbial occurrences. Similarly, the following example can also be seen as an occurrence of the adverbial use of aber: () Druhtin ávur zi ímo sprah Lord again to him spoke ‘The Lord spoke to him again.’ (Otfrid V , ) As previously discussed, I analyse aber in Isidor and Tatian as an adverbial element. In Notker, where aber has a semantically connective function, Latin autem following an XP is never translated with X(P)-áber:
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
() a. (Consilium autem domini manet in ęternum) Aber sîn rât uuéret iêmer ABER his advice keeps always ‘However, his advice is always valid.’ (Notker, Psalter , ) b. (In contrariis autem quibus est medium) Áber in díen contrariis tíu medium hábint ABER in the.DAT opposites.DAT the middle have ‘However, at opposite ends, where the middle is’ (Notker, Consolatio , ) c. (Tu autem domine susceptor meus es) Aber dû got pist mîn infángare but you God are my protector ‘But you, God, are my protector.’ (Notker, Psalter , –) According to Axel (: ), the string [XP avur] in () forms a single constituent much as she proposes for similar strings with other adverbials, such as uuarliho ‘indeed’ (see Chapter of this volume). Axel () compares the string [XP uuarliho], which also occurs in left-dislocated positions in OHG, with the Modern German left-dislocated construction with connective adverbials in () (see also Chapter of this volume): () a. (/ . . . & uos igitur/nunc quidem tristitiam habebitis/) / . . . Inti [ir uuarliho]/ nu habet ir gitruobnessi / and you surely now have your trouble (Tatian , ) b. [CP [ir uuarliho]i [CP nuk [C’’ [habetj] [S iri tk gitruobnessi tj]]]] () Meyer indessen/allerdings/schließlich/nämlich, der sagte zu Peter . . . Meyer on the other hand/however/finally/namely, that said to Peter . . . ‘Meyer, ADV, he said to Peter . . . ’ (Axel : ) However, for OHG, it is unclear if this type of construction is a genuine left dislocation similar to the Modern German left-dislocated structure in () or a nominative pending construction (Axel : ff; Chapter of this volume). As discussed by Axel (; Chapter of this volume) and also shown by Hinterhölzl and Petrova (), the V-order—although still generalized in OHG—should be seen as a later generalization of German after the positions in the left periphery had been integrated and become part of the clause. This generalization followed a reanalysis of these positions as neutral in terms of their discourse-informational status. According to Hinterhölzl and Petrova (), there is still a syntactic position for aboutness topics in Modern German in the preverbal position, as proposed by Frey () and discussed in Section .... Subjects in this position are unmarked, but preverbal objects still have a marked discourse-informational status (see also
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
Frey ). () exemplifies the different steps of the reanalysis of the prefield positions in OHG (Hinterhölzl and Petrova ): () a. Phase I: [Aboutness] [ForceP (=CP) Vfin [TP . . . ]] Topic + V b. Phase II: [ForceP (=CP) [Aboutness] Vfin [TP . . . ]] c. Phase III: [ForceP (=CP) [Aboutness]i Vfin [ti [TP . . . ]] Hinterhölzl and Petrova () assume that topics are extrasentential in Phase I and are only included in the clause structure in Phase II. Axel (; Chapter of this volume) as well as Hinterhölzl and Petrova () assume that the prefield in OHG consists of different split positions that are occupied by different specific information-structural elements. These positions have collapsed diachronically and been grammaticalized in Modern German to form a single neutral position with generalized XP-movement (see Chapter ). In OHG, however, aber had not yet become an exponent of the category CoorP. Data from EHNG support the hypothesis that the reanalysis of aber as a syntactically coordinating element only occurred in the late ENHG period, as will be discussed in the following section. ... Aber in later texts of ENHG and in NHG: CoorP and gapping As in OHG, aber was still an adverbial element in ENHG since it could have a repetitive meaning (a), but it could also display contrastive meaning (b): () a. Und als der doctor die selben aber und aber äfert and while the doctor the same again and again repeats ‘And while the doctor repeats the same thing again and again’ (Tünger, Facetiae – ) b. Aber sicher wiltu sprechen however surely want.you read ‘But you surely want to read.’ (Tauler, Sermon vso A, – ) The fact that aber is also syntactically an adverbial element in its adversative meaning can be shown through examples in which aber can co-occur with und (a). This was the case up to the seventeenth century, as the text by Leibniz (b) exemplifies: () a. Und als aber der wirt ain kalb [ . . . ] in die stuben and as however the innkeeper a calf in the living.room truog [ . . . ], und aber der zunfftmaister mit schlaff beladen brought and however the guild.master with sleep charged was [ . . . ] was ‘However, as the innkeeper brought a calf into the living room but the guild master was asleep . . . ’ (Tünger, Facetiae – )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi b. Solchen zu dienst und zur ergözligkeit, und aber these.DAT for service and for delightfulness and however zugleich [ . . . ] zum gemeinen besten [ . . . ] des Vaterlandes at.the.same.time for.the common best the.GEN fatherland.GEN soll dieses gemeinet seyn. should this meant be ‘These are intended not only for their service and pleasure but also for the common use of the fatherland.’ (Leibniz, Gedächtnis – /)
In later ENHG texts, the string und aber can no longer be found. Moreover, we can find ellipsis phenomena in the texts of this period (a) for the first time and more frequently later in NHG (b) and (c): () a. Er sagt ‘Mein herr bekümmert euch nit mit disem fatzman he says my lord distress you not with this clown dann dise wort sind im nur ein gespoett.’ as these words are him only a fake ‘Aber mir nit’ sagt herr Reichart but to.me not says Mr. Reichart ‘He says, “My Lord, do not let this clown distress you since these words are only a joke to him.” “But not to me,” says Mr. Reichart.’ (Wickram, Nachbarn , – ) b. Der Zwek ist wohl zu loben; aber nicht die Mittel the purpose is surely to praise but not the means ‘The purpose is surely praiseworthy, but not the means.’ (Mitternacht, Soldat – ) c. Zum Beten und Singen zwingen Sie meine Tochter nicht; to pray and sing force you my daughter not aber zur Liebe but to.the love ‘You do not force my daughter to pray and to sing, but to love.’ (Gellert, Betschwester – ) In the literature on ellipses, this type of elision by which the finite verb and some other material in the second conjunct are elided is called ‘gapping’ (Hartmann ). Both conjuncts build a contrast pair, and the elided finite verb in the second conjunct must be semantically identical to the overt antecedent verb in the first conjunct in tense, aspect, and mood (Repp : ff). The difference between gapping in conjuncts coordinated with and and those coordinated with but is that, in the latter case, there is an explicit or implicit negation that triggers the contrast between elements in the coordinate adjuncts depending on the topic and the focused constituents. The negation in all abovementioned examples is overtly realized and narrow focused. Independent of the analysis of gapping, what is interesting here is
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
that aber as coordinating element connects two sentences with a Force projection. This leads Repp (: ) to define gapping as the copying21 of the anchoring of the proposition to the factual world from the non-elliptic antecedent. Only after the reanalysis of aber as a coordinating element in CoorP with a Force projection are such ellipsis phenomena possible. As Roberts and Roussou () and van Gelderen () have proposed, the reanalysis from adverb to connective element via the rise of different CP-elements took place in the following steps: () [CP [IP [IP afur [VP ]] ] > topicalisation [CoorP [ CP afur [IP [IP t [VP ]]
reanalysis [CoorP afur [CP [IP [VP ]] Here, the topicalized adverb has been reanalysed as connective element by movement. The data seem to confirm this analysis. In the first position, the reanalysis from adverb to adverbial connective has taken place in a structure in which the adverb avur is topicalized for information-structural reasons, a scenario proposed by van Gelderen () for complementizers such as for. A similar situation also seems to be the starting point for the grammaticalization of VP adverbs such as allerdings as adverbial connectives in ENHG, as will be further discussed in Section ... .. Summary In the first part of this chapter, the development of aber from an adverb to a connective element was discussed. In OHG, aber was used as a VP-restitutive adverb, and in later texts, it was used as an IP-repetitive adverb. During this period, this unselective adverb lost its scope over the TP and was reanalysed as a clausal connective element. In a further step, aber was reanalysed as an exponent of the CoorP category, yet remained a syntactically adverbial element in positions other than the first. After having considered aber, one of the oldest adverbial connectives, the next part of the chapter is devoted to the connective allerdings ‘however’, which arose during the ENHG period, as is the case for most such connectives.
. Allerdings: from VP-adverb to epistemic speech-act adverb and connective element In this section, I discuss the grammaticalization of the adverbial connective allerdings ‘however’. Its development is very similar to that of aber but differs in that its adverbial meaning is still extant (albeit not very productive) in Modern German.
21 Repp (: ) proposes an account of gapping in terms of copying combining syntactic with semantic-pragmatic principles.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
T . Corpus of texts from to the early twentieth century th century
th century
th century
th century
th century
,
,,
,,
,,
,
texts words
T . Corpus of written and spoken NHG
words
St. Galler Tagblatt (A)
Berliner Morgenpost (L)
Freiburger Korpus (FR –)
,,
,,
,
The data presented in this section are taken from a corpus of seventy-three texts of various literary genres covering the period from to the early twentieth century and comprising a total of ,, words.22 For the Modern German written language, the DeReKo corpus (German Reference Corpus) was consulted: the entire volume of the St. Galler Tagblatt newspaper and the issues of the Berliner Morgenpost were analysed. The spoken-language data were collected from the Freiburger Korpus (dialogues –) to the extent listed in Table .. Before delving into the diachronic aspect of allerdings in Section .., the data from Modern German are presented in Section .. in order to compare them with the data of ENHG and NHG. .. Allerdings in Modern German As stated above, allerdings in Modern German can be a sentence adverb as well as an adverbial connective. I first consider the sentence adverb in Section ... followed by the adverbial connective in Section .... ... Allerdings as an epistemic speech-act adverb As a sentence adverb, allerdings has the meaning of ‘indeed, certainly, of course, absolutely’ and can assume sentence positions in the prefield (a) and in the middle field (b). It can also be used elliptically as a response to a yes/no question (c): () a. Allerdings komme ich. certainly come I ‘Of course I will come.’
22
The complete list of the corpus texts can be found in Ferraresi ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
b. Ich werde allerdings kommen. I will in any case come ‘I will certainly come.’ c. A: Kommst Du? B: Allerdings! come you in any case A: ‘ “Will you come?” B: “Absolutely!” ’ The use of allerdings as an adverb is quite rare, especially in full sentences. In this function, allerdings is mostly found as an elliptical response, as in (c). Allerdings as an adverb receives a pitch accent on its final syllable in all positions. We now turn our focus to the class of adverbs to which allerdings belongs. At first sight, allerdings ‘indeed, certainly, of course, absolutely’ seems to group with such adverbials as auf jeden Fall ‘in any case, absolutely’ and tatsächlich ‘indeed’, which are classified as ‘verificational’ adverbs in Haftka (). By using these adverbials, the speaker affirms the truth value of the proposition (‘it is true that holds: p’). According to Haftka (), verificational adverbs do not belong to the same class as sentence adverbs such as möglicherweise ‘probably’ since verificational adverbs occupy a different, lower position in the sentence, as example (a) illustrates: () a. Damals war möglicherweise tatsächlich at.that.time was possibly really nicht immer der Gärtner der Mörder. not always the gardener the murderer ‘At that time, it was possible that the gardener was actually not always the murderer.’ b. sentence adverbs > verificational adverbs > sentence negation > event adverbs Haftka thus proposes the hierarchy in (b), which deems verificational adverbs to be lower than sentence adverbs in marking the focus domain. With this focus type, the assertion operator is contrasted with the only other element in the set, i.e. with the negated proposition ¬p similar to the VERUMfocus (Höhle : ).23 The same ordering as in (b) can also be found in Italian (a), in which the relative position of the sentence adverb probabilmente ‘possibly, probably’ (classified by Cinque as epistemic in Spec Moodepistemic) and of the verificational adverb veramente ‘truly’ is similar to the relative order of the German epistemic adverb wahrscheinlich (which Frey believes to occupy a position in the middle field higher than the base-position of any other element in the middle field in its neutral
23 Höhle () assumes that VERUMfocus is an illocutionary epistemic operator through which the speaker, focusing the finite verb, asserts the truth of the proposition:
(i) Maria IST gekommen. Maria is come ‘Maria has definitely come.’
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
use as sentence adverb, see also Section ...) and the verificational adverb tatsächlich ‘indeed’ (b): () a. Gianni Gianni riparato repaired
ha has la the
b. Hans hat Hans has vollständig completely
probabilmente veramente di nuovo completamente possibly indeed again completely macchina. car
wahrscheinlich tatsächlich das Auto wieder possibly indeed the car again repariert. repaired
In Cinque’s () adverb hierarchy (which does not consider verificational adverbs), verificational adverbs would take a position between Moodepistemic and AspcompletiveII. In Frey and Pittner (), these adverbs belong to the group of sentence adverbs. Unlike the verificational adverbials auf jeden Fall ‘in any case, absolutely’ and tatsächlich ‘indeed’, the adverb allerdings ‘indeed, certainly, of course, absolutely’ is not compatible with epistemic sentence adverbs (): () a. *Hans hat wahrscheinlich allerdings das Auto Hans has probably indeed the car wieder vollständig repariert. again completely repaired. b. ??Allerdings hat Hans wahrscheinlich das Auto wieder vollständig repariert. c. *Wahrscheinlich hat Hans allerdings das Auto wieder vollständig repariert. Only the sentence with allerdings in the prefield (b) sounds somewhat acceptable as this is also an unmarked position for the connective allerdings. The adverb allerdings seems to behave more like an epistemic speech-act adverb along with sicherlich ‘surely’ and wohl ‘possibly’ (Krifka ; Zimmermann ). Consider the following example with the different epistemic adverbs sicher ‘surely’, sicherlich ‘surely’, wohl ‘possibly’ and allerdings ‘in any case’. All of these adverbs are acceptable in a declarative sentence: () Es wird sicher/sicherlich/wohl/allerdings regnen. it will sure/surely/possibly/in any case rain ‘It will surely/possibly rain./It will rain in any case.’ Since younger speakers nowadays do not possess a grammaticality judgement regarding the use of allerdings as an adverb in the middle field (as this use of allerdings is no longer productive), it is necessary to have a look at NHG texts from the nineteenth and early twentieth centuries, in which there were still some instances of this use of allerdings, as in the following example from Kafka: () »Das läßt sich hören«, sagte der Onkel. this let REFL hear said the uncle doch erwarten.« »Erwarten wird er mich surely expect expect will he me
»Aber Mack but Mack nicht«, sagte not said
wird dich will you Karl, Karl
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
»aber er wird allerdings hinkommen.« but he will surely come ‘ “I am pleased to hear that,” said the uncle. “But Mack is waiting for you.” “He is not waiting for me,” said Karl, “but he will surely come.” ’ (Kafka, Amerika – –) In the antecedent of a conditional clause, however, the epistemic adverbs seen in () are not all grammatical (Krifka ; Zimmermann ; and Repp : ff): () a. Wenn es sicher/vielleicht/*sicherlich/*wohl regnet, if it certainly/perhaps/surely/possibly rains nehme ich einen Schirm mit. take I an umbrella with ‘If it is certainly/perhaps/surely/possibly going to rain, I will take an umbrella.’ b. Wenn es *allerdings regnet, nehme ich einen Schirm mit. As epistemic speech-act adverbs that differ from other epistemic propositional adverbs (e.g. sicher and vielleicht), sicherlich and wohl cannot occur in the antecedent of conditionals (a) since this type of sentence cannot embed a speech act (Krifka ). Like epistemic speech-act adverbs, the adverb allerdings also cannot occur in the antecedent of conditionals (b). Moreover, epistemic speech-act adverbs cannot be negated in correction and differ from propositional adverbs in this respect (a) and (b). Allerdings again behaves like an epistemic speech act and not like an epistemic propositional adverb (c): () a. Er hat nicht sicher/ *sicherlich/ *wohl he has not certainly/ surely/ possibly das Haus verlassen, sondern vielleicht. the house left but perhaps b. Er he das the
hat nicht vielleicht/ möglicherweise has not perhaps/ possibly Haus verlassen, sondern definitiv. house left but definitely
c. Er hat he has sondern but
nicht *allerdings das Haus verlassen, not surely the house left vielleicht. perhaps
In Repp’s (: ) syntactic analysis, such speech-act adverbs as well as the corrective negation occupy the position of StrengthP directly after the Force position in Rizzi’s model (comparable with ΣP in Laka ) and take scope at LF.24 24 Repp (: ff) models her syntactic structure by integrating considerations by Searle and Vanderveken (), according to whom ‘speech acts come with certain sincerity conditions. [ . . . ]
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
Krifka () assumes a similar category ActP, which is the syntactic realization of the speaker’s commitments (e.g. in an assertion).25 According to Krifka (), ActP hosts response particles such as yes and no, which are anaphors of ActP in English. Adverbs such as maybe and right used as responding elements are remnants of elliptical clauses: ()
A: Does Ede steal cookies? a. B: [ActP ASSERT [TP Maybe [TP he stoles cookies]]] b. B: [ActP yes [ActP ASSERT [TP Maybe]]]
With right, the speaker makes the same speech act with a full sentence, as in (a): ()
A: [ActP ASSERT [NegP Ede did not steal the cookies]] a. B: [ActP ASSERT [TP This is right]] b. B: [ActP Right ]
According to Merchant (), fragment answers are derived via A’-movement and subsequent remnant deletion of the TP at PF. I propose a parallel analysis of adverbial allerdings in an elliptical answer, as in (), with allerdings in SpecCP. In the responsive use of allerdings, the speaker confirms the proposition with a strong commitment, as in (): () a. Andy Rufener, geht man richtig in der Annahme, dass Sie Andy Rufener goes one rightly in the supposition that you über die :-Niederlage bei den ZSC Lions frustriert sind? on the – defeat against the ZSC Lions frustrated are A.R.: Allerdings. absolutely. ‘ “Andy Rufener, is it correct to assume that you are frustrated by the – defeat of ZSC Lions?” A.R.: “Absolutely.” ’ (St. Galler Tagblatt, November ) b. [ActP ASSERT [CP Allerdings1 [TP bin ich über die :-Niederlage bei den ZSC Lions t1 frustriert]]]. ‘ “Of course I am frustrated by the – defeat of the ZSC Lions.” ’ This elliptical use is also the starting point for the reanalysis of allerdings as a connective element.
A sincerity condition for an assertion would be that the speaker believes the truth of the proposition in question.’ According to Repp, the commitment of the speaker can have different degrees of strength, which correspond to epistemic modality at the speech-act level. 25 Giorgi and Pianesi () and Giorgi () also assume that events are evaluated not only by means of the coordinates of the subject but also by means of the speaker and that these coordinates are realized syntactically in the CP. Ferraresi (), who discusses exclamative sentences, also shows that the speaker commitment is realized in the CP.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
... Allerdings as an adverbial connective In its connective function, allerdings has the meaning of ‘but, however’ and mostly bears its pitch accent on the first syllable, even if there are some speakers who also accept allerdings in this function accented on the last syllable like its adverb form. Like the adverb, the connective can assume different positions in the clause, including the prefield (a) and the middle field (b), in addition to others in which the adverb does not appear, such as the post field (c), the post-initial field (d) and the extra-sentential position (e): () a. prefield: Maria kommt morgen, allerdings hat sie gesagt, dass sie Maria comes tomorrow however has she said that she erst später ankommt. only later arrives ‘Maria will come tomorrow; however, she has said that she will arrive later.’ b. middle field: Maria kommt morgen, sie hat allerdings gesagt, dass sie Maria comes tomorrow she has however said that she erst später ankommt only later arrives c. post field: Peter hatte zugesagt. Er ist nicht gekommen allerdings. Peter had agreed he is not come though ‘Peter had agreed; however, he did not come.’ d. after an XP in the prefield (post-initial): Maria und Peter haben beide zugesagt. Maria allerdings Maria and Peter have both agreed Maria however kann erst abends kommen. can only at.evening come ‘Maria and Peter have both agreed; however, Maria can only come in the evening.’ e. outside of the clause (extra-sentential): Peter hat zugesagt. Allerdings: Was will er hier überhaupt? Peter has agreed however/though what wants he here at.all ‘Peter has agreed. But what does he want to do here?’ I assume that connective allerdings is syntactically an adverbial element adjoined to different maximal projections and is semantically a coordinator similar to the adverbial connective aber (see also Büring and Hartmann ). Thus, it cannot cooccur with und ‘and’ (a) and (b): () a. *[[CP1 Mein Vater ist krank [CoorP und [CP2 allerdings geht er arbeiten]] my father is sick and however goes he work b. *[[CP1 Mein Vater ist krank [CoorP und [CP2 geht allerdings arbeiten]] c. [[CP1 Mein Vater ist krank ],[CP2 allerdings geht er arbeiten]]
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
However, unlike aber, allerdings is not a coordinative conjunction exponent of the category CoorP as it triggers verb-subject inversion (a) when it occupies the first position in CP. In summary, NHG allerdings has two different meanings and functions. The older use, from which the connective meaning arises, is the adverbial one with the meaning of ‘in any case’. This is an epistemic speech-act adverb that is mostly used as a responsive element in NHG as a remnant of an elliptical sentence. In its connective use, allerdings has a meaning similar to aber ‘however’; it is semantically a coordinator but syntactically an adverbial element in every position. In the subsequent sections, the diachronic development of the adverbial connective from the adverb is discussed. .. Allerdings in ENHG: from adverbial to connective element The adverbial connective allerdings originally developed from the genitive DP aller dinge ‘of all things’, a quantifying VP adverbial with the meaning of ‘completely’. One example from Luther’s bible is provided in (a) with a translation in (b) in which aller dinge is translated with ganz ‘completely’: () a. Vnd thet das dem HERRN wolgefiel aller dinge wie sein and he.did that the.DAT Lord.DAT well.liked completely as his vater Vsia gethan hatte. father Usia done had ‘And he did what the Lord liked in the very same way as his father Usia had done.’ (Luther-Bibel , (Kön , )) b. Und er tat, was dem HERRN wohl gefiel, ganz wie sein and he did that the.DAT Lord.DAT well liked completely as his Vater Usia getan hatte. father Usia done had (Luther-Bibel , (Kön , )) In Cinque’s () adverbial hierarchy, a ‘completion adverb’ (Cinque’s terminology) such as completely is the lowest adverb in the hierarchy to occupy the Specifier of the Aspectual Sg CompletiveII-Projection (see Cinque’s adverb hierarchy in example (), Section ..). The adverb became unified as allerdinge as early as the sixteenth century, and the first examples of allerdings date to the seventeenth century. A completive adverb, allerdings can be negated, as in the following instances, with allerdinge in (a) and allerdings in (b): () a. wollet mir demnach verzeihen / daß ich mich nicht want me therefore forgive that I me not allerdinge offenbahre completely disclose ‘May you therefore forgive me for not revealing myself completely.’ (Bucholtz, Wunder-Geschichte – )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
b. das habe ich gemerket, dasz sie mir nicht allerdings trauen. this have I noticed that they me not completely trust ‘I have noticed that they do not trust me completely.’ (Opitz , – seventeenth century – after Grimm and Grimm ff, Vol. : ) Such VP adverbs of completion as allerdings are classified as degree adverbs (for a classification of German adverbs, see Eisenberg : ; Pittner ) and quantify telic predicates (i.e. verbs with an incremental theme or that denote a change in state). Syntactically, they cannot be topicalized. In the corpus data, there are no occurrences of topicalized completive allerdings. As early as the sixteenth century, however, aller dinge/allerdinge underwent a transformation to a higher epistemic speech-act adverb with the meaning of ‘indeed, certainly, of course, absolutely’,26 as in example () from Luther’s bible: () Aller dinge mus die Menge zusamen komen any case must the crowd together come ‘In any event, the crowd should assemble.’ (Luther-Bibel , (Apg , )) For the reanalysis, I assume that allerdings, being an unselective adverb much like aber, changed the syntactic object in its scope since it was first used as a quantifier over telic predicates, later on with atelic verbs as an intensifier and later displayed scope over speech acts (Ferraresi ). Texts from the seventeenth-century corpus display an elliptical use of the epistemic adverb allerdings, as the following example reveals: () Solte ich aber den / der mich so verachtet / [ . . . ] wieder should I though the.one who me so.much despises again annemen müssen [ . . . ]? Allerdings! (sagte die Königin) und ob ja ihr accept must absolutely said the queen and if PTC you nicht wollet/ not want ‘ “However, should I again accept him who despises me? Absolutely,” said the queen, “even if you do not want me to.” ’ (Ulrich, Syrerinn – – ) Similar to the elliptical use of the adverb in Modern German, I assume an elliptical structure in which allerdings is topicalized in the prefield as a remnant answer and the known, already-realized information from the previous context is elided at PF. In the corpus of ENHG and Modern German, the epistemic adverb allerdings at first only rarely appeared in the prefield in the seventeenth century (three out of instances) and triggered subject-verb inversion. As Table . shows, the number
26 In MHG, the genitive case frequently had an adverbial function, as demonstrated by the adverbs keineswegs ‘in no way’ and jedenfalls ‘anyway’.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
T . Allerdings in the first position in main clauses th century th century th century th century th century Total allerdings tokens allerdings V S allerdings XP V S
(. %) (. %)
(. %) (. %)
(. %) (. %)
(. %) (. %)
(. %) (. %)
of sentences in which allerdings occupies the prefield only increased in the eighteenth century (thirty-eight instances out of ): During this period, there were also some instances of allerdings co-occuring with und ‘and’: () Manche Mutter hätte ihn [ . . . ] recht gern zum Schwiegersohne some mother had him right gladly to son.in.law angenommen, und allerdings war er, [ . . . ] Das, was man gewöhnlich taken and in any case? was he that what one usually eine gute Partie zu nennen pflegt. a good marriage to call used ‘Some mothers would have been happy to have him as a son-in-law, and he was certainly what used to be called a good candidate for marriage.’ (Lewald, Clementine – ) In the second half of the nineteenth century, the sequence und allerdings was no longer attested. How did the reanalysis of allerdings as a connective take place? As an epistemic adverb, allerdings in the CP confirms the truth of the preceding proposition, as in the following example, in which the propositional anaphor das ‘that’ refers to the aforementioned proposition die Wirkung der Häßlichkeit wird gehindert: () Wird die Wirkung der Häßlichkeit, durch die aufeinanderfolgende is the effect the.GEN ugliness.GEN by the successive Enumeration ihrer Elemente, nicht eben sowohl gehindert enumeration its.GEN elements.GEN not right in.the.same.way precluded als die Wirkung der Schönheit durch die ähnliche as the effect the.GEN beauty.GEN by the similar Enumeration ihrer Elemente vereitelt wird? enumeration its.GEN elements.GEN precluded is Allerdings wird sie das. surely will she this ‘Is the effect of ugliness not hindered by the successive enumeration of its elements in the same way as the effect of beauty is confounded by the similar enumeration of its elements? Of course it is.’ (Lessing, Laokoon – eighteenth century)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
The connective function developed in sentences in which the speaker commits himself/herself to the propositional truth of a preceding sentence by both asserting and contradicting it. The elliptical use of allerdings was the precondition for its reanalysis as a connective since it allowed for the addition of new information, as in the following example: () Ich: Wer des Feuers genießen will, muß sich den Rauch I: who the fire enjoy wants must REFL the smoke gefallen lassen. like let. Er: Allerdings. Aber weil der Rauch bei dem Feuer he: certainly however since the smoke with the fire unvermeidlich ist, durfte man darum keinen Rauchfang inevitable is should one because.of.this no chimney erfinden? invent ‘I: “Whoever wants to enjoy the fire also has to endure the smoke.” He: “Of course; however, since smoke inevitably comes with fire, should one not invent a chimney?” ’ (Herder, Briefe – – eighteenth century) In (), the preceding assertion is confirmed through elliptical allerdings and is simultaneously contrasted, as the presence of aber shows. The new, added information represents a restriction on the preceding proposition. This type of concessivity is documented in texts as early as the seventeenth century. In this configuration, allerdings can be reanalysed by the listener as an adversative element in the sentence (), as shown in example (): () a. speaker: [ allerdings p ]CP [coorP aber [q]CP b. hearer: [ allerdings aber q]CP () Der alte Deutsche, auch in seinen rauhen Wäldern, er kannte das the ancient German, even in his harsh forests he knew the Edle im Weibe und genoß an ihm die schönsten Eigenschaften nobility in women and enjoyed in her the best qualities seines Geschlechts, Klugheit, Treue, Mut und Keuschheit; of.her gender cleverness fidelity audacity and chastity allerdings aber kam ihm auch sein Klima, sein genetischer however but came to.him also his climate his genetical Charakter, seine ganze Lebensweise hierin zu Hülfe. character his whole way.of.life in.this.matter to aid ‘The old German—even in his wild forests—he recognized the nobility of women and enjoyed the finest qualities of their sex: cleverness, fidelity, audacity and chastity; however, the climate, the man’s genetic character, and his entire way of life helped him in this matter.’ (Herder, Ideen – eighteenth century)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
In (), allerdings can either be interpreted as elliptical (i.e. as confirming the preceding proposition) or as a topicalized adverb in the second connect. After the reanalysis, allerdings as a connective loses its function as an epistemic speech-act operator in the same sentence. The exponent of the same syntactic position in which the adverb allerdings is realized is a different lexical item with the same function (i.e. zwar ‘truly, indeed’): () Es ist, bei Gott, wohl eine große Freiheit, mir zu widersprechen! it is by God surely a great liberty me to contradict Und wer mir widerspricht, hat sich wohl sehr zu bekümmern, and whoever me contradicts has himself well much to take.care ob ich verdrüßlich werde, oder nicht! Allerdings zwar sollte if I irritated become or not however truly should ein Widerspruch, als womit mich Hr. Klotz verfolgt, in die a contrariety as with.which me Mr. Klotz hounds in the Länge auch den gelassensten, kältesten Mann verdrüßlich machen. long.run even the calmest coldest man annoyed make ‘By God, it is surely a great liberty to contradict me. And whoever contradicts me should take note as to whether or not I become irritated! However, such a contradiction as the one with which Mr. Klotz bothers me would truly irritate even the calmest and coldest man after a while.’ (Lessing, Tod – eighteenth century) In (), allerdings has been reanalysed as a CP-adjunct. This is shown by the fact that no inversion follows after the end of the nineteenth century and that there is a pause (signalled by a comma) between allerdings and the connected clause: () Die glücklichen Erben, [ . . . ] schenkten mir, [ . . . ] den drittbesten Anzug, the lucky heirs gave me the third.best suit den der Verstorbene [ . . . ] zu tragen pflegte. [ . . . ] Allerdings, which the deceased to wear used however die Hose war bedeutend zu weit. the trousers were definitely too large ‘The lucky heirs gave me the third-best suit that the deceased used to wear. However, the trousers were definitely too large.’ (Busch, Schmetterling – ) The analysis of the corpus data has shown that the use of allerdings as an adverb— while still extant in Modern German—decreased in the nineteenth century (Ferraresi ). In Modern German, allerdings can also appear outside the rest of the clause; however, its function there is different from the one in (): () Nur noch only just an dem der when the
fünf Tage five days Mutter für Mother for
Zeit time ihre her
für die Vorbereitungen jenes Tages, for the preparation of.the day ganzjährige Aufopferung für die Familie all.year.round devotion to the family
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
T . Occurrences of allerdings in the St. Galler Tagblatt () and in the Berliner Morgenpost () total number of words
allerdings
word frequency
extrasentential
St. Galler Tagblatt ()
,,
,
.%
.%
Berliner Morgenpost ()
,,
,
.%
.%
gedankt werden soll. Allerdings: Was soll man schenken? thanked be should however what shall one give? ‘Only five more days left for the preparation of the day when Mother is to be thanked for her endless devotion to her family. However, what kind of present should we get her?’ (Kleine Zeitung, May ) In (), ‘the trousers’ (a part of the suit) are recovered from the context, and the connection is still textual. In (), allerdings lies outside the CP and introduces an utterance. In historical corpus texts, allerdings has only been documented in extrasentential positions (as in ()) since the twentieth century; however, a comparison with the Modern German corpus shows that allerdings is also quite rarely used in the extrasentential position in modern times: The St. Galler Tagblatt, with a total of million words in the – volumes, contains about , instances of allerdings, which corresponds to a frequency of .%. In , only sixteen out of , occurrences of allerdings appeared in the extrasentential position outside the clause, which corresponds to a frequency of .%. The reanalysis of allerdings from an adverb to a connective took place after the number of occurrences of allerdings in initial position had increased. At the same time, other adverbs such as zwar27 developed and became used more frequently as epistemic adverbs, whereas allerdings became obsolete in this meaning. In the reanalysis process, the prosodic shape of allerdings also changed. Allerdings as an adverb bears its pitch accent on the last syllable, and the information in focus is the assertion of the truth value. The falling accent signals the end of the turn or, at least, that whatever follows is less relevant. In the connective function, this changes. Accent retraction (‘stress shift’) is induced since allerdings no longer bears the pitch accent, and the bitonal rising tone (L+H) signals the incomplete utterance (see also Zifonun et al. : ): The rising tone is typical in questions and, more generally, expresses the fact that the utterance is incomplete: either the speaker intends to continue, or expects an answer or a confirmation from the hearer. (Féry : ) 27 The adverb zwar originally had the same meaning as epistemic allerdings: OHG zi wāre ‘in truth, truly’ > MHG ze wāre, zewāre, zwāre ‘verily, forsooth’ (Pfeifer : ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi
Stress shift also reveals that allerdings and aber in () are in the same phonological phrase. The main accent of the first element can be shifted if it is near the main accent of the next element as long as both are in the same phonological phrase (Wiese : ). Stress shift cannot take place if both elements are not in the same phonological phrase (Wiese : ff, ff). Along with the reanalysis of allerdings, a semantic change also occurred: allerdings as an adverbial connective induces the pragmatic implicature that—contrary to expectation—q is true even if p is true. This is the concessive interpretation that allerdings has in Modern German when it is used together with aber: () Wir haben den zweiten Sitz erwartet, aber allerdings we have the second seat expected but however auch ein bisschen darum gezittert. also a little therefore trembled ‘We expected the second seat; however, we also trembled a little.’ (St. Galler Tagblatt, February ) Other languages have different adversative elements, e.g. the Italian ma però ‘but however’, which can realize different semantic features, such as the adversative and the concessive ones, which are realized in German through aber and allerdings.
. Summary In this chapter, the rise and development of so-called adverbial connectives have been discussed. The first part of the chapter provided a detailed consideration of the adverbial connective aber, which developed during OHG into an adverbial connective. In OHG and MHG, avur/afur was a VP adverb with the meaning of ‘again’ and came to be reanalysed as a higher restitutive adverb. Later on, the adverb was reanalysed in the first position as an exponent of the category CoorP. However, unlike genuine coordinating elements such as und ‘and’, aber can occupy different positions in the sentence being, syntactically, an adverbial element. In every step of the reanalysis, the feature of contrast which characterizes aber plays an important role. A similar development can be ascertained for the adverbial connective allerdings ‘however’, which was discussed in the second part of this chapter. Allerdings is a more recent adverbial connective than aber since most currently existing Modern German adverbial connectives emerged during the ENHG period. Allerdings was originally a completive quantifying adverb that occupied the lowest position in Cinque’s () hierarchy (example ()). It was then reanalysed as a confirming sentence adverb with the meaning of ‘indeed, in any case’. In its elliptical use, allerdings triggers a concessive implicature by which the continuation of the discourse is interpreted as a contrast, especially in answers as a confirmation of a preceding proposition. This context triggers the reanalysis of allerdings as an introductory element to the contrast clause. The different grammaticalization stages of both aber and allerdings can be roughly schematized as follows (Ferraresi ):
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Adverbial connectives
() Grammaticalization stages of adverbial connectives VI. conjunction speech act ↑ III. adverbial connective event ↑ II. sentence adverb IP ↑ I. adverb VP Adverbial connectives also appear to go through these stages in other languages.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Part II The Middle Field
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
7 Introduction to Part II G I S E L L A F E R RA R ES I A ND A G N E S JÄ G E R
In this part of the volume different aspects concerning word order regularities and change in the German middle field are investigated. As already discussed in Chapter , the middle field is the syntactic domain between the left sentence bracket, i.e. C in generative terms, taken up by complementizers or—in verb-first and verbsecond clauses—by the finite verb, and the right sentence bracket, the position of non-finite verbs and of the finite verb in verb-final clauses, i.e. V. In the generative analysis, the middle field corresponds to the portion of the sentence where the arguments of the verb as well as adjuncts are merged but which also contains a number of functional projections dealing with, among other things, finiteness and agreement. Unlike the prefield, which is largely restricted to only a single constituent (or at most two constituents) (see also Part I), the middle field may comprise any number of syntactic constituents. Therefore, the restrictions and possibilities of word order are a central issue. In Modern German, the order of constituents in the middle field is influenced by different factors. For example, semantic as well as pragmatic, viz. discourse-informational factors can trigger scrambling of elements in the middle field from their base position. Another factor influencing the order of NPs in the middle field is the so-called Gesetz der wachsenden Glieder (‘Law of increasing constituents’, Behaghel /), according to which the weight of a constituent plays a role in the order, the lighter constituent preceding the heavier one. As far as the history of the middle field is concerned, it is therefore crucial to investigate which syntactic patterns may be observed and what determines the order of syntactic constituents in the middle field in earlier stages of German. Besides questions regarding nominal arguments, a wider range of phenomena concerning word order in the middle field has to be considered, such as e.g. pronominalization, since the order of arguments changes as soon as these are realized as pronouns—probably due to prosodic reasons. This has the effect of weak pronouns being cliticized. Phenomena such as pronominalization or cliticization are subsumed under the term of the Wackernagel complex. Furthermore, negation is also marked by elements which originate in the middle field. Accordingly, the Modern German Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Gisella Ferraresi and Agnes Jäger . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi and Agnes Jäger
negation particle nicht ‘not’ is placed in the centre of the middle field, dividing it into a higher and lower part, and marking the outer boundary of VP. The syntactic properties of the negation particle as well as the availability and types of Negative Concord, however, have changed significantly during the history of German.
. The Wackernagel complex (WC) As mentioned in Chapter , the part of the middle field furthest to the left is called the Wackernagel position (Haider : ). It corresponds to the position directly following the finite verb in C in main clauses and the subordinating conjunction in embedded clauses; it is here that elements such as clitic/weak pronouns appear, which constitutes a very remarkable feature of German (and of the other Continental West Germanic languages). The term Wackernagel position goes back to the seminal paper by Wackernagel (/), according to which clitics in earlier Indo-European languages such as Sanskrit, Hittite, and Ancient Greek are positioned after the first word in the sentence, while this first position is not syntactically defined. Hale () points out that in Sanskrit, for example, the group of pronominal clitics, which take the second position in the sentence after wh-elements in questions and relative clauses, are no longer in the second position when precisely one topicalized constituent precedes a wh-element. The reference to the second position in the clause is thus a problematic criterion for defining the Wackernagel position as a syntactic position. Moreover, there seems to be some indeterminacy in the meaning of the term ‘Wackernagel position’, since for Sanskrit it is the position clitics after the first word of the clause occupy, with pre-wh-phrases excluded from the count, whereas in West Germanic the Wackernagel position is, according to the literature, occupied by clitics after the finite verb or complementizers. In Modern German, different elements such as personal and reflexive pronouns, the indefinite pronoun man ‘one’, and modal particles seem to compete for the position after the finite verb in main and the complementizer in subordinate clauses. Besides the positioning at the left edge of the middle field, cluster formation is a typical feature of Wackernagel elements as well as grammaticalization/syntactization in the form of double agreement or complementizer agreement, all of which is investigated for historical German by Helmut Weiß in Chapter . In Modern German, pronouns take the inverse order in the middle field compared to full NPs, accusative pronouns preceding dative ones (Lenerz : ). In older stages of German, however, the order corresponding to the base order, viz. dative before accusative pronoun, is regularly attested depending on the particular pronouns involved. For instance the SG neuter accusative pronoun iz/ez ‘it’ and the SG masculine accusative pronoun i(h)n ‘him’ tend to follow the dative pronouns (Behaghel –, Vol. IV: ; Fleischer ), the datives of the SG and SG pronouns mir/dir ‘me/you’ and to a lesser extent of the SG feminine ihr ‘her’ show a strong tendency to precede accusative pronouns (Fleischer ). One question to be treated concerns therefore the factors governing the different orders in the sequence of personal pronouns.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part II
Another relevant issue is the status of pronouns in older German. For Modern German dialects a tripartite classification into full, reduced, and clitic pronouns has been argued for (Weiß , ). Reis (), Behaghel (–, Vol. IV), and more recently Tomaselli ()—and van Kemenade () and Roberts () for OE—argue that in the sequence XP – pronoun – Vfin in OHG the pronouns are clitics and that therefore the sequence does not violate the verb-second rule, because in terms of the classification proposed in Cardinaletti and Starke (), this would amount to saying that pronouns in older German are defective X. However, in contrast to the Romance languages and also to German dialects such as Bavarian, OHG does not show a double set of full and clitic pronouns. Arguments in favour of the existence of clitic pronouns are, among others, cases of contraction of the pronouns and the finite verb such as thenkistu ‘think-you’ as well as contracted forms with complementizers like theiz < thaz iz ‘that is’ discussed in Braune (: ff, ff). In Chapter , Weiß presents further arguments in favour of the existence of clitic subject pronouns in OHG which are related to the phenomenon of double agreement and the rise of complementizer agreement in later stages and varieties of German. Complementizer agreement refers to inflectional morphology occurring on complementizers agreeing with the φ-features of the subjects of their embedded clauses. This may be analysed as the effect of movement of a functional head out of the IP domain (I, T, or AgrS to C, den Besten ; Zwart ) or as independent feature valuation of T and C (Haegeman and van Koppen ). Although some kind of agreeing complementizer also occurs in a few other languages (Arabic dialects: Lewis ; Lubulusu: Diercks ; Wolof: Torrence ), this special form of complementizer agreement is a noteworthy trait of Continental West Germanic dialects, which interestingly also share the properties of OV order and verb cluster formation (see Part III of this volume) absent from North Germanic languages and English (Bayer ; Haegeman ; Zwart , ; Law ; Carstens ; Weiß a). Furthermore, in Continental West Germanic dialects with complementizer agreement, this property correlates with double agreement, i.e. two different agreement patterns in the verbal inflection depending on the syntactic position of the pronominal subject (i.e. pre- or post-finite) or the finite verb. In the second variant of double agreement the verb shows regular inflectional morphology in final position and the complementizer agreement morpheme in the second position. This variant of double agreement correlates with partial pro-drop (for example in some Bavarian and Franconian varieties in the SG and PL). Investigating the historical origin of Wackernagel complex properties, Weiß argues that they are already present in OHG. Evidence for this comes from vowel ellipsis (synaloepha) which applies to finite verbs in second position or to complementizers when followed by a pronoun as opposed to an adverb (de Boor ). Also assimilation, clitic clusters, merged, and reanalysed forms can be regarded as a result of the phonological interaction within prosodic units which applies in the Wackernagel complex. In MHG—probably due to phonological changes—the evidence for the Wackernagel complex is less clear.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi and Agnes Jäger
Reanalysis of pronouns following the verb as part of the verbal inflection leads to double agreement in many cases. Weiß therefore argues that double agreement not only correlates with complementizer agreement but is also a precondition for it. The starting point for both phenomena is the existence of subject clitics. Weiß proposes that, after subject cliticization in C, children may associate the new inflection with the position rather than with the verb itself. This leads to the production of inflection also on the complementizer. As a consequence, double agreement is realized in those languages where complementizer agreement is triggered via analogical extension. This explains the two phenomena without assuming inherent φ-features of C, which would also leave the singularity of complementizer agreement typologically unexplained.
. Order of NPs In sentences with ditransitive verbs that realize subject (S), direct object (DO), and indirect object (IO) as full NPs, all permutations of these elements are—more or less marked—possible options in the middle field. Deviations from the unmarked canonical order S > IO > DO are derived through scrambling, i.e. local movement operations (Ross ). There are many different factors conditioning the linearization of elements in the middle field in Modern German: definite NPs tend to precede indefinite ones, pronominalized arguments precede full argument NPs, and non-focused (background) material precedes focused constituents (see Lenerz ; Höhle ; Uszkoreit ; Reis ; Büring among others). Moreover, discoursestructural criteria such as new/old information can change the canonical order; for example if the DO realizes a new piece of information, it can precede the IO (Lenerz : ). A rather different perspective on this issue is proposed by Haider (), who argues that German has three verbal classes with different underlying orders. In the first one (verbs like verweigern ‘deny’ or abgewöhnen ‘break someone loose of something’) the unmarked order is IO > DO, in the second one with verbs like aussetzen ‘expose’ or unterziehen ‘subject someone to something’ the unmarked order is DO > IO. In the third class, which has no underlying order, there are verbs like geben ‘give’ or empfehlen ‘recommend’. Different verbal classes with different underlying order seem to have existed already in older Germanic languages such as Gothic (Ferraresi ) and ON (Barðal ). Further factors which influence the ordering of constituents in the middle field are animacy and agentivity, with a tendency towards the order animate before inanimate, and referent with higher agentivity before referent with lower agentivity (Siewierska ). Some orderings are thus due to the semantics of the constituents in question. This is true also of the ordering of adverbials in the middle field (Frey and Pittner ; Pittner ). Some of the ordering tendencies go back to PIE times, whereas others are considerably younger, as discussed by Augustin Speyer in Chapter . The tendency, for instance, to put old information—so-called ‘themes’ in the terms of the Prague School—before new information—or ‘rhemes’—can be found in other Indo-European
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part II
languages as well, e.g. Latin and the Slavic languages, and is thus presumably also a property of PIE (Lühr ). The tendency of definite noun phrases to precede indefinite noun phrases, on the other hand, is more recent in that it might be linked to the development of the robust morphosyntactic marking of definiteness and indefiniteness by the respective articles (Diessel ). Since Modern German shows quite a flexible order, which can be influenced by the aforementioned phonological, semantic, and pragmatic factors, it would be a plausible assumption to suppose that Modern German continues the ‘free’ order of PIE. For historical German, Behaghel (–, Vol. IV) identifies an unmarked word order S > IO > DO > adjuncts, which can be modified due to the tendencies theme > rheme and the ‘law of increasing constituents’. However, Speyer (Chapter ) shows that OHG has strict IO > DO order, which is not influenced by informationstructural effects. From this, it follows that OHG does not continue the ‘free’ order of PIE mostly governed by information-structural ordering principles. MHG texts show a similar situation. The question Speyer pursues in his contribution is if and when there is a change in the ordering principles in the history of German and whether the role of scrambling is gaining or losing importance. Only in late ENHG texts around , the findings suggest that after a period of relatively fixed word order, the order DO > IO starts to occur with some frequency and the influence of pragmatic and semantic factors increases. One hypothesis explaining this type of change would be the change of the importance of one ordering principle, viz. the animacy principle, responsible for the order IO > DO. In fact, in most cases, the dative is associated with the recipient or the benefactive, which is always an animate actant, whereas the accusative is associated with theme or more generally with actants, which are not necessarily animate. According to recent psycholinguistic research (Dowty ; Primus ), a universal principle of thematic hierarchy which assigns the most prominent (structural) cases to Proto-Agent and Proto-Patient is thus reflected in the animacy principle.
. Negation Negation can be considered a middle field phenomenon in so far as, for a large part of the history of German, the negation particle, i.e. Modern German nicht ‘not’ and its historical predecessors nieht, niht etc. as well as their variants in the dialects, take up a position in the middle field. The same holds for negative indefinite pronouns and adverbs such as niemand ‘nobody’, nichts ‘nothing’, nirgends ‘nowhere’, nie ‘never’, etc. and their predecessors, barring cases where they are topicalized, i.e. moved to the prefield. As mentioned above, the Modern German negation particle nicht is generally placed in the centre of the middle field.1 It is often taken as a diagnostic means for the outer boundary of VP. Thus, it is preceded by high adverbials, the subject, and also definite
1 This holds for unmarked sentential negation. In the case of constituent negation or narrow focus of negation, nicht arguably forms part of the relevant constituent.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Gisella Ferraresi and Agnes Jäger
NPs that may be taken to have scrambled out of VP (Webelhuth ; Büring ), while lower adverbials, prepositional objects, etc. occur after the negation particle. As discussed in detail by Anne Breitbarth and Agnes Jäger in Chapter , the syntactic position of the negation particles in German has changed diachronically in accordance with the so-called Jespersen Cycle just as in many other languages, cf. Willis et al. (). Breitbarth and Jäger argue that OHG as well as OLG are at stage I in terms of Jespersen’s Cycle: sentential negation is basically always marked by the preverbal clitic negation particle ni. This particle always attaches to the finite verb and accordingly occurs together with it in the left or right sentence bracket. One way of accounting for this syntactic behaviour is to assume that ni originates in the head position of a functional projection NegP that dominates VP and is thus in fact also positioned in the middle field. The clitic negation particle is then displaced from its original position by verb movement. In Late OHG, a second negation particle is beginning to be grammaticalized.2 Among other elements, it is especially nieht (< niowiht ‘nothing’)—the predecessor of Modern German nicht—that occurs as a second negation particle together with the negation clitic. It can be analysed as occupying SpecNegP, i.e. a fixed position in the middle field. This Jespersen stage II pattern with a bipartite negation particle is still evidenced in MHG and is also common in MLG. In contrast to previous accounts in the literature, however, Breitbarth and Jäger show that the bipartite negation particle is not the majority pattern in MHG. In fact, the investigated MHG texts for the most part display Jespersen’s stage III with just the verb-independent adverb-like middle field negation particle for the most part, as Modern German still does. Low German, on the other hand, kept the preverbal negation particle longer: while there is also already evidence for Jespersen’s stage III in MLG, this pattern only starts to become more common than the bipartite negation particle around . Around the same time, ENHG basically always shows the Jespersen stage III pattern, with only some very rare exceptions in certain dialects. Considering Jespersen’s Cycle in terms of NegP, the change resides in a change in the lexical filling of the head and specifier position of this projection, while the actual syntactic structure remains unchanged with respect to negation. Besides the negation particle, another major means of marking negation consists in the use of negative indefinite determiners, pronouns, and adverbs, usually referred to by the cover term ‘n-indefinites’. If an indefinite occurs in the semantic scope of negation, three basic patterns are possible: (A) Negation can be marked both by a negation particle and an n-indefinite resulting in a so-called Negative Concord (NC) construction, a cross-linguistically very common pattern, or (B) just by the negation particle (with the indefinite occurring in a non-neg-marked form), or finally (C) just by an n-indefinite without an additional negation particle. 2 For Low German, the development cannot be traced exactly as there is an attestation gap of years during the decisive period. Before the gap, i.e. in later OLG, there is no evidence for an adverbial use of nieht. When attestation starts again in the thirteenth century, nicht is already firmly established as a negation particle.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part II
In OHG as well as in MHG, all three basic patterns are attested. In OHG, pattern (A), i.e. NC, is the most common pattern on average while pattern (B) is also fairly common but pattern (C) is extremely rare. In OLG, only patterns (A) and (B) are attested, however, with a strong preference for pattern (B). In the investigated MHG corpus texts, pattern (C) is the most common one. NC is found considerably less often and pattern (B) hardly occurs any more in negated clauses with indefinites. Again, the development is slightly delayed in Low German: while pattern (B) is not evidenced at all in MLG, pattern (A), i.e. NC between the n-indefinite and preverbal en, is preferred over pattern (C) until the beginning of the fifteenth century. Afterwards, pattern (C) is increasingly used but even around , there is still evidence for the preverbal neg-particle co-occurring with an n-indefinite in Low German. In OHG as well as OLG and MHG, NC is found in the form of so-called ‘Neg-Doubling’ between the clitic neg-particle and an n-indefinite, but generally not in the form of so-called Neg-Spread between several n-indefinites. Furthermore, Neg-Doubling between ni(c)ht and an n-indefinite is missing both in MHG and MLG. However, in classical MLG (mid-fourteenth to mid-sixteenth century), there is also evidence for Neg-Spread. This type of NC thus appears to be a later development. It is also attested in some present-day High German dialects. NC poses a challenge to compositional semantics as several neg-markers co-occur in a clause that is interpreted as containing a single semantic negation. This challenge is met if one assumes that only one element bears an interpretable negative feature and all other occurrences of negation constitute uninterpretable negative features that are syntactically checked against the interpretable one (Weiß a; Zeijlstra ; Penka among others). As Breitbarth and Jäger demonstrate, this analysis can also be applied to historical High and Low German. Summing up, the issues covered in Part II regarding the middle field concern a wide range of syntactic phenomena, most prominently the position and order of nominal and pronominal arguments but also the specific semantically and pragmatically relevant phenomenon of negation. It should be noted that, while the study of the prefield is very much focused on whether a certain type of element occurs in a certain position, study of the middle field is generally more about the relations between elements and the groups that they form.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
8 The Wackernagel complex and pronoun raising HELMUT WEIẞ
. Introduction German (like the other Continental West Germanic languages/dialects) possesses a syntactic property that is absent in the Northern Germanic languages and English: it has a very special Wackernagel position (as will become clear in the following). This means that weak (personal) pronouns (see fn. in Section .. for further specifications) show a very strong tendency to appear immediately after the finite verb in root sentences or after the complementizer in embedded sentences. The position where pronouns are raised to is called the ‘Wackernagel position’ (WP) (Haider : ). This position marks the left edge of the middle field in the sense that it constitutes the left-most position available for constituents raised within the middle field. At the same time, however, the Wackernagel position seems to have access to the left periphery, because phenomena like (the origin of) complementizer agreement indicate that items raised to the Wackernagel position interact with the finite verb, or the complementizer respectively, which both occupy C (or Fin). Pronoun raising is independent of the syntactic function of the raised pronoun and is required even in cases when the non-pronominal equivalent would show up rather low in the sentence structure—compare ihm and dem Otto in (a) versus (b): () a. Hat ihm denn gestern der Hans nicht geholfen? (Standard German) has him then yesterday the Hans not helped ‘Didn’t Hans help him yesterday?’ b. Hat denn gestern der Hans dem Otto nicht geholfen? has then yesterday the Hans the Otto not helped ‘Didn’t Hans help Otto yesterday?’ The appropriate word order rule was first described by the Swiss Indo-Europeanist Jakob Wackernagel in his famous paper Über ein Gesetz der indogermanischen Wortstellung (‘On a law of Indo-European word order’), where he demonstrated Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Helmut Weiß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
that clitic(-like) elements tend to appear in second position in the sentence or near it (‘an zweiter oder möglichst nahe bei der zweiten Stelle,’ Wackernagel /: ). The data from classical languages (Sanskrit, Ancient Greek, Latin) in Wackernagel (/) demonstrate that second position means the position after the first word or the first constituent1—as is shown in the Latin examples in () for the pronoun te ‘you’. () a. Di te deaeque ament Gods you goddesses.and love ‘Gods and goddesses love you.’ (Noel : ) b. Di deaeque te adiuvant Gods goddesses.and you help ‘Gods and goddesses help you.’ (Noel : )
(Latin)
In German, the Wackernagel position does not always correspond to the position after the first constituent (as in (a)), but is defined in yet another sense. Descriptively speaking, it constitutes the left edge of the middle field (which is the position immediately after C/Fin in a generative account).2 This could even be the third position as it is always the case in root sentences with regular V order (a) and, at least in dialects, also in embedded sentences with doubly filled COMPs (b). () a. Gestern hat ihm der Hans geholfen yesterday has him the Hans helped ‘Hans helped him yesterday.’ b. warum dass=e=da=n geem han why that=I=you=him given have ‘(the reason) why I gave it to you . . . ’
(Standard German)
(Bavarian)
The examples in () also show that pronouns in German are post-verbal in V-sentences but preverbal in verb-final sentences. This is reminiscent of the so-called Tobler-Mussafia law, which explains the proclitical versus enclitical distribution of pronouns in older Romance languages: it states that pronouns, although they are proclitics in non-V sentences, encliticize onto the finite verb in V sentences because they cannot themselves occupy the first position in a sentence (Benincà : f). However, there are several fundamental differences: first, Old Romance pronouns occupy a preverbal position not only in complementizer-introduced embedded sentences but also in root sentences with, for example, direct objects in the left periphery of the sentence (see examples , , and in Benincà ), something which German pronouns never do. Second, clitic pronouns in German (at least in dialects)
1 According to Anderson (), Wackernagel’s original understanding of ‘second position’ was only the first one, whereas the second interpretation of ‘second position’ is the result of later research. 2 For a general overview on syntactic fields and positions in terms of the topological model of the German clause see Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
are always enclitical, irrespective of whether the clitic’s host is a finite verb or a complementizer, that is, one can observe the same phonological processes (e.g. epenthesis) in both environments (see Section ..), which is not the case with Old Romance. Third, in German, pronoun raising to the WP is associated with accompanying morphosyntactic properties like double agreement or inflected complementizers (see Sections .. and ..) which are unknown to have occurred in Old Romance. In generative syntax, there are several contradicting approaches with respect to the exact nature of pronoun raising to the WP and why pronouns behave differently from nouns: For Sternefeld (: ) it is an instance of scrambling, whereas Lenerz () or Haider () distinguish the two movements. For some linguists, pronoun fronting is syntactic movement, which is either feature-driven and targets specifier positions of TP (Yoshida ), or not feature-driven targeting ‘an outer specifier of vP’ (Müller : ). For others, however, pronoun fronting is just a kind of PF-cliticization (Breitbarth ),3 or a combination of scrambling and PF-cliticization (Weiß , ).4 In the following, I will not address such questions, but mainly be concerned with the emergence and historical development of the WP in German. First, I will outline four properties connected with the WP and which form the Wackernagel complex (WC) (Section .), before I try to give evidence for the existence of the WC since OHG times (Section .) and formulate a hypothesis about how complementizer agreement—one of the WC-features—may have emerged (Section .). In Section ., I will present some new data concerning pronoun raising in MHG.5
. The Wackernagel complex (WC) Wackernagel’s original observation was about a fundamental word order regularity. Wackernagel’s law states that expressions like pronouns, particles, or light verbs are clitics and behave syntactically alike in that they appear in the WP. Interestingly, in German dialects (or in Continental West Germanic in general) there are three additional syntactic features connected to, or associated with, the WP, viz. double agreement, complementizer agreement, and partial pro-drop, which go far beyond the original scope of Wackernagel’s law. Together with pronominal clitics they constitute the WC. The WC sets the (Continental) West Germanic languages in an interesting contrast to the Northern Germanic languages (plus English). It is well known that some fundamental syntactic differences have evolved between both groups of Germanic languages (e.g. OV versus VO—compare Haider a—or the existence or Breitbarth (: ): ‘this “position” is probably not a syntactic position as it does not serve any syntactic function, and the elements it hosts are rather varied . . . assuming PF-“cliticization” (or just positioning) to the right of C (a common phenomenon across languages).’ 4 See also Corver and Delfitto () for a similar approach. They claim that ‘pronoun movement consists of a two step operation, consisting of a DP-movement step and a subsequent D-movement step’ (Corver and Delfitto : ). The difference to Weiß’s () approach is that in their system the second step is also feature-driven. 5 I would like to thank Melanie Hobich and Anja Schenk for their help with the preparation of the text. 3
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
non-existence of verbal clusters), but the Wackernagel complex as another distinctive feature has gone unnoticed so far, although the properties as such are well known. .. Pronominal clitics Pronominal subjects and objects often occur as clitics in the WP, cf. ():6 () a. Gesdan han=e=da=n geem yesterday have=I=you=him given ‘I gave it to you yesterday.’ b. warum dass=e=da=n geem han why that=I=you=him given have ‘(the reason) why I gave it to you . . . ’
(Bavarian)
The WP in the dialects is much more distinctly expressed than it is in Standard German. The main reason for this is that dialects possess clitic pronouns which are absent in Standard German, and clitics are syntactically much more restricted than non-clitic pronouns: they are not allowed to appear in any other position in the sentence, whereas reduced pronouns are possible elsewhere, e.g. in the prefield or in a lower position in the middle field.7 .. Double agreement In Standard German, the finite verb has the same inflectional marker regardless of the syntactic position in which it occurs. In some dialects, however, verbs may carry a different inflectional marker in C than in their base position (cf. Zwart ; Weiß a). This phenomenon is often referred to as double agreement (DA). There are two variants of double agreement dialects, because the appearance of the C-agreement (CA) inflectional marker may be governed either i) by the position of the pronominal subject, or ii) by the position of the finite verb. DA I: in the first variant, the finite verb bears the regular suffix in embedded clauses and subject-initial main clauses, while in main clauses with subject–verb inversion a different morpheme is used. This kind of DA occurs primarily in Dutch dialects (Zwart ), but it is not completely absent in German dialects either
6 In general, pronominal clitics may differ with respect to the host positions they cliticize onto: in Romance languages, for example, they are T/v-oriented, whereas in (some) Slavic languages and Continental West Germanic dialects, they are C-oriented, cf. Rivero (); Cardinaletti and Starke (), and Roberts (). 7 On the syntactic level, we observe two three-way distinctions wrt. pronouns: there are three possible positions for pronouns (prefield, WP, elsewhere) and three phrase structures (DP, φP, ambiguous). On the morphological level, this corresponds to a tripartite system consisting of full, reduced, and clitic pronouns (Weiß , ). German dialects, however, mostly provide only two distinct forms (this mismatch between syntax and morphology is described in greater detail in Weiß , ). Clitics as morphological expression of weak pronouns, in the sense of Cardinaletti and Starke (, ), are just one (though very common) possibility; another one is null or reduced pronouns. The latter ones do not only occur in the WP, but in the prefield as well.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
(Weiß a). In some of them, for instance in the East Prussian dialect of the Samland, we observe that when the pronominal subject follows the verb, the inflectional marker ‐e disappears: () a. wî / jû make we / you make
(Samland, East Prussia)
b. wat mak we / je what make we / you In the Samland dialect, the change of the inflectional marker is confined to whquestions. A variety without this additional restriction is the Upper German Swabian dialect, e.g. the Swabian spoken in the Westallgäu (Gruber )—cf. ()—or that of Stuttgart (cf. Frey : ). The inflectional marker in the SG is ‐ed (just like in the PL and PL, since Swabian has an Einheitsplural ‘uniform plural’, i.e. in the plural there is no person distinction) which gets reduced to -e when the pronominal subject follows. () a. mer gloubed we believe (Gruber : )
(Swabian)
b. gloube mr believe we (Gruber : ) DA II: the regular ending is restricted to verbs in end position, while in second position (i.e. C/Fin) the verb bears the CA morpheme. This distribution is found in varieties of Central Bavarian for the PL, cf. (), and of North Bavarian for the PL, cf. (): () a. dasma mia aaf Minga fahrn/*ma that.PL we to Munich go ‘that we go to Munich.’ b. mia fahrma/*n aaf Minga we go.PL to Munich ‘We go/are going to Munich.’
(Central Bavarian)
c. fahrma/*n mia aaf Minga? go.PL we to Munich ‘Are we going to Munich?’ () a. wents neat tiets ā tepā wāet if.PL not you too with were ‘if you wouldn’t be there, too.’ (Pfalz : )
(North Bavarian, Sangerberg)
b. weits iwet pruk khumt, seatses wīetshaus when.PL over.the bridge come, see.PL.the pub ‘You’ll see the pub when you go over the bridge.’ (Pfalz : )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
.. Complementizer agreement Continental West Germanic dialects show a very special grammatical peculiarity in that complementizers, in addition to the verb, may agree with the subject as well. This is exemplified in () where the complementizer that is inflected for the PL. This kind of agreement is a typologically unique characteristic of Continental West Germanic dialects (Corbett ). () a. dán d’Láit häien that.PL the people hear ‘that the people hear’
(North Bavarian, Egerland)
b. dan Valère en Pol morgen goan that.PL V. and P. tomorrow go ‘that Valère and Pol go tomorrow’
(West Flemish)
Most German dialects, however, exhibit defective systems: CA is mostly confined to the SG and/or the PL (Bohn and Weiß ). In some other dialects such as Central Bavarian, it additionally applies in the PL (Weiß a). Complete CA paradigms are reported mainly for West Flemish dialects like the dialect of Lapscheure (Haegeman ; Zwart ), but exist also in German dialects, especially in Eastern Central and High German dialects. For example, in the dialect of the Sechsämterland, a variety of North Bavarian, CA lacks overt inflectional markers only in the SG and SG—just like modal verbs do (Weiß a). () a. wálst (du) because.SG (you) b. wáln mer PL c. wálts diets PL d. wáln si PL If we considered only the SG, we might get the impression that CA is a consequence of subject cliticization, but the PL and PL clearly show that both are independent phenomena—at least synchronically (see Section . for the diachronic connection between the two features). .. Partial pro-drop Though Standard German no longer allows referential pro-drop, many German dialects still do—but only partially. Thus pro-drop is possible in the SG and PL in some Bavarian and Franconian varieties, while it is disallowed in the PL and PL in Franconian, see () and (): () a. wennsd pro mogsd if.SG like.SG ‘If you like.’ b. wennds pro megds if.PL like.PL ‘If you like.’
(Central Bavarian, likewise North Bavarian and East Franconian)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
() a. waaln *(mer) graad besamn senn (East Franconian) because.PL we at.the.moment together are.PL ‘Because we are together at the moment.’ b. waaln *(se) grad besamn senn because.PL they at.the.moment together are.PL ‘Because they are together at the moment.’ As the examples in () show, CA as such does not license pro-drop. However, in many German dialects we can observe another correlation. Consider the distribution of pro-drop in Central Bavarian in (): here, pro is licensed only in the SG, PL, and PL. () a. *morng bin pro wieda gsund tomorrow be.SG again healthy ‘I’ll be healthy again tomorrow.’ b. morng tomorrow c. *morng tomorrow
bist pro be.SG is pro be.SG
wieda again wieda again
gsund healthy gsund healthy
d. morng tomorrow e. morng tomorrow
sama pro be.PL sads pro be.PL
wieda again wieda again
gsund healthy gsund healthy
(Central Bavarian)
f. *morng san pro wieda gsund tomorrow be.PL again healthy These are exactly those persons in which the inflectional markers are pronominal, i.e. they contain a part which historically goes back to a reanalysed subject clitic (such as the t in the SG). All other inflectional markers lack such a pronominal element. So we can observe a strong correlation between pro-drop and a special type of agreement morphology.8
. The diachrony of the Wackernagel complex As discussed above, the WC forms a micro-typological particularity of Continental West Germanic dialects and languages within the Germanic language group. The Nordic languages and English do not possess a WP, for reasons which are not entirely clear. As mentioned, a further crucial difference between the two subgroups lies in the base order (VO versus OV), which may be one reason why English and the Nordic languages never fully developed the WC-properties, although at least OE 8 The reader is referred to Weiß (a) for more details on CA and pro-drop in recent dialects (including the discussion of a possible exception to the abovementioned rule). For the diachrony of null subjects in German the reader may consult Axel and Weiß (, ); Fleischer and Schallert (); Volodina and Weiß ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
sporadically exhibited DA (Weiß a).9 The fact that both properties are connected is suggested by the following observation: there are Bavarian varieties spoken in Northern Italy which first developed from OV to VO languages and then lost their WP (Weiß and Wratil ). From a diachronic perspective, the crucial question is when these WC properties developed.10 A necessary, though not sufficient precondition is the existence of C-oriented pronominal clitics (Section ..). The early Germanic languages seem to have had clitic personal pronouns in the original sense of Wackernagel (/). In OE (Harbert : ), which exhibits SVO in root sentences with NP subjects, pronominal subjects tend to appear after the verb (i.e. VS), unless there is a topicalized constituent which allows the subject pronoun to take up a preverbal position (i.e. XSV). Such second position clitics are attested for early OHG as well (Axel , see also Chapter of this volume). The special kind of second-position clitics which we see in modern Continental West Germanic dialects/languages (i.e. always cliticized onto the finite verb or the complementizer, see Section .) has obviously in part developed already in Runic times: the pronoun eka ‘I’ ‘only occurs attached to the right of the verb’ (Harbert : ), as Eythórsson () has shown. In Gothic (Ferraresi ), object pronouns seem to follow Wackernagel’s law with the exception of ita ‘it’, which is always right adjoined to the finite verb in root sentences and to the complementizer in embedded sentences (Ferraresi : ). With respect to OHG personal pronouns, Axel () has collected and reevaluated the observations and findings made in the secondary literature. Her final judgement is: . . . the OHG pronouns are comparable neither to the X -clitics which occur in recent dialects of German and in other Germanic languages nor to the X -clitics in the Romance languages. (Axel : )11
The main reason why Axel () thinks that OHG lacked clitic pronouns, is that OHG does not exhibit distinct inventories for full and reduced/clitic pronouns: In OHG texts, there is no evidence for a systematic opposition between reduced/clitic forms and full forms even though there are some indications for a phonological cliticization in those cases where the pronoun occurs at the left edge of the middle field, i.e. in the position immediately after the finite verb or complementizer. (Axel : )
However, this reasoning is not entirely convincing since the clitic forms of pronouns in modern dialects developed out of phonologically reduced forms and in most cases it is possible to relate the clitic to the full form even synchronically (e.g. Bavarian /ǝ/ < /i:/ ‘I’). Modern German dialects often have two forms of personal
OE had the same kind of DA as Swabian (cf. example ()), but also in PL (Höhle : ). In the following, I will restrict myself mostly to the history of High German and more or less neglect Old Frisian, Old Low German, and Old Low Franconian (the predecessor of Dutch). 11 The claim that OHG possessed Romance-like V-oriented clitics is based on the rare occurrence of V clauses with pronouns between a fronted XP and the finite verb in C. These sentences are attested only for early OHG (see Chapter ..). 9
10
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
pronouns, i.e. full versus non-full forms. The non-full forms could be reduced or clitic forms, where the clitic is bound to the WP, while the reduced pronoun can also appear in the prefield or somewhere lower in the middle field (e.g. after an adverb). According to the pronoun cycle developed in Weiß (), clitic pronouns derive from reduced ones, which in turn derive from full forms through weakening.12 Therefore, if we can prove that reduced or otherwise phonologically changed forms are restricted to the WP in their distribution, we can argue, as I will do in this section, that there is evidence for the existence of at least some of the WC properties already in OHG times, whereas for others, especially CA, it is hard to find evidence before ENHG times. .. Potential diagnostics The four WC properties described in Sections ..–.. allow us to propose diagnostics which can be employed in diachronic investigations to decide whether WC properties were already present at a particular stage of the language or not. An important and obvious diagnostic feature is the position of Wackernagelelements13 at the left edge of the middle field: ()
Adjacency to the right-most C-element (= weak evidence)
In root sentences, Wackernagel elements should immediately follow the finite verb (cf. Axel and Chapter of this volume for generalized verb raising to C in OHG root sentences) and the complementizer or relative pronouns in embedded sentences.14 In case more than one Wackernagel element occurs, they form a cluster—just like a single clitic with its host. Typical cluster effects are: • Adjacency: all Wackernagel elements must be adjacent and nothing may intervene (= weak evidence) • Serialization: S > IO > DO (= weak evidence, because phonological aspects may play a role in ordering) • Phonological-prosodic interactions of the Wackernagel elements with one another and/ or with the C-host (e.g. reduction, elision, assimilation, epenthesis) (= strong evidence) • Grammaticalization/syntactization processes such as double agreement (= strong evidence) or complementizer agreement (= strong evidence)
12 The situation in Dutch and Flemish dialects is a little more complex as the strong form often looks entirely different; however, it can be related diachronically to the reduced and clitic forms (Dutch jullie ‘you-PL’ < ghilieden ‘you people’, Harbert : ). Dutch/Flemish pronouns differ syntactically as well since they can be doubled much more freely than their counterparts in German dialects (Craenenbroeck and Koppen ). 13 Since OHG lacked clitic modal particles, we take into account only personal pronouns. 14 Doubly filled COMPs as known from modern dialects (Bayer ; Weiß ) are also attested for OHG, but only for relative clauses, where a relative pronoun could appear together with the relative complementizer the (cf. Axel : ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
.. Diachronic evidence ... Synaloepha in Otfrid’s Gospel Book The first piece of evidence which I will present comes from the Gospel Book by Otfrid von Weißenburg (c. to , South Rhine Franconian), one of the major texts from OHG15 (which also served as the main text base for Schrodt , the syntax part of the OHG grammar). In Otfrid’s Gospel Book we can observe a phenomenon which is highly relevant for our purpose. Otfrid applies a poetical figure called synaloepha, i.e. elision of a final vowel before a following vowel-initial word (de Boor ; Somers Wicka : ).16 As a poetical figure to avoid hiatus (the clash of two vowels on a word boundary), synaloepha could in principle occur everywhere in a sentence if the poetical metre requires or allows for it. In hiatus sequences, there are three forms attested: ) Short forms with elision of the final vowel Type: hort er < horta er ‘heard he’ – want er < wanta er ‘because he’ ) Forms with so-called subitem, i.e. dot below the vowel: the final vowel is marked with a subitem17 Type: hortạ er – wantạ er ) Full forms without elision of the final vowel Type: horta er – wanta er Table . shows the numbers according to the study of de Boor ().18 He distinguishes four contexts: disyllabic versus trisyllabic verbs like horta ‘heard’ or sageta ‘said’, respectively, followed either by a pronominal subject (like er ‘he’) or an adverb (with an initial vowel) like io ‘ever’.
T . Synaloepha in Otfrid’s Gospel Book Type
Short forms
Forms with subitem
Long forms
horta er
/
/
/
sageta er
/
/
/
horta io
/
/
/
sageta io
/
/
/
For a brief overview over the textual evidence of OHG see Chapter . Somers Wicka () provides an in-depth investigation of this phenomenon, but her analysis of the material is formulated mostly within the frameworks of Lexical Phonology and Prosodic Phonology. 17 According to Axel (: ), the subitem below a vowel constitutes ‘a sign for the reader not to pronounce it’. 18 In this and the following tables, each cell contains two numbers of which the first one refers to manuscript V and the second one to manuscript P (de Boor ). 15 16
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
The numbers show interesting differences: • First, the difference between disyllabic and trisyllabic verbs shows that metrical requirements do play a role. Elision occurs significantly less frequently after trisyllabic verbs, cf. Table .. T . Disyllabic versus trisyllabic verbs Type
Short forms
Forms with subitem
Long forms
horta er/io
/
/
/
sageta er/io
/
/
/
• Second, there is a significant difference between disyllabic verbs followed by a pronominal subject or an adverb: although the adverb also begins with a vowel and has a short, clitic-like form, the final vowel of the preceding verb is mostly preserved, whereas it is elided in the vast majority of cases when the following word is a pronominal subject, cf. Table .. T . Pronominal subjects versus adverbs Type
Short forms
Forms with subitem
Long forms
horta er
/
/
/
sageta er
/
/
/
horta io
/
/
/
sageta io
/
/
/
Apart from verbs in second position, complementizers are also possible hosts for pronominal clitics in Continental West Germanic dialects. If a WP in the modern sense already existed in OHG, we would expect that OHG complementizers were just as suitable as hosts for clitics which allowed for phonological processes like vowel deletion (= synaloepha) as verbs in second position. This seems indeed to have been the case (cf. Table .): the subordinating conjunction wanta ‘because, for’ shows the familiar behaviour as it tends to lose its final vowel when followed by a pronominal subject, but not before an adverb.19 The other subordinating conjunction is oba ‘if ’, which follows the expected tendency, although the contrast between long and short
19 Cf. Tables and in de Boor (: ). De Boor (: f) further mentions that it makes a difference whether wanta is placed at the beginning of a line (Auftakt) or at the beginning of a half-line (Innentakt): all long forms in ms. V are placed at the Auftakt. This shows that metrical requirements also play a role, because both positions differ with respect to stress assignment: the first syllable of a line is unstressed, while the first syllable of the second half-line always gets stress.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
forms is much stronger in ms. P than in ms. V.20 As expected, the coordinators odo ‘or’ and inti ‘and’ show the reverse behaviour.21 T . Conjunctions Type wanta er
Short forms
Forms with subitem
Long forms
/
/
/
wanta io
/
/
/
oba + X
/
/
/
odo + X
/
/
/
inti + X
/
/
/
The findings of de Boor’s () study convincingly show that synaloepha mostly applies in a Wackernagel constellation, showing that pronominal subjects form a prosodic unit with the preceding verb or complementizer, but not adverbs like io ‘ever’, although they fulfil all phonological requirements (i.e. being short and beginning with a vowel). Therefore, we can conclude that in Otfrid’s Gospel Book, synaloepha is syntactically bound to the WP. That this is not only a peculiarity of Otfrid’s poetic style is witnessed by Otfrid himself in his letter to Luitbert, the Archbishop of Mainz, in which he discusses problems of writing in such an uncultivated and undisciplined language as South Rhine Franconian (the language he wrote). Speaking of synaloepha, he explicitly states ‘that we too do this excessively in our ordinary speech, if we but listen carefully’.22 ... Assimilation Assimilation is another form of phonological interaction within prosodic units. A prominent case of assimilation is the so-called i-umlaut which occurred in OHG (as well as in other Old Germanic languages): an [a] was fronted to [e], if the following syllable contained an [i, i:, j], e.g. OHG gast versus gesti ‘guest(s)’. Umlaut happened within words where the inflectional or derivational morpheme contained an umlaut-triggering vowel (Braune : ). Interestingly, umlaut occasionally also applied in the Wackernagel constellation: when a pronominal subject or object followed the verb in C, it could trigger umlaut (Somers Wicka
20 Cf. Table in de Boor (: ). As for oba, de Boor does not differentiate between pronominal and non-pronominal items following the subordinator, but it is in most cases a pronoun. 21 The numbers for odo + X and inti + X in Table . are taken from Tables and in de Boor (: f), respectively. Note that inti ‘and’ is somehow between subordinating and coordinating conjunctions, because it shows considerably more short forms than odo does. A possible explanation for this is probably that inti could also be used as a subordinating conjunction—a use that is attested since late OHG times and that disappeared only at the beginning of NHG (see Ferraresi and Weiß ). 22 The English translation of James Marchand, which is unpublished, is quoted after: http://www. harbornet.com/folks/theedrich/hive/Medieval/Otfrid.htm.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
: ). So gáb could be changed into géb (a), dránk into drénk (b), and wás into wés (c):23 () a. géb gave b. drénk drank c. wés was
imo versus gáb imo (x) him ih I iz it
versus dránk ih versus wás iz (x)
The umlaut phenomenon demonstrates that the sequence verb plus pronominal clitic could form a phonological unit and its occasional occurrence may be evidence for its existence in spoken OHG. ... Clitic cluster In case more than one Wackernagel element is present, they form a clitic cluster in modern dialects. As noted in Section .., cluster formation is often accompanied by phenomena such as vowel assimilation, vowel deletion, or epenthesis. For the disyllabic pronouns inan ‘him.ACC’ (a), imo ‘him.DAT’ (b), and iro ‘her’ (c) it is attested that they could lose their initial vowel when they followed another pronoun (Braune : §, Anm. ): () a. tranc ër-nan drank he-him.ACC ‘He drank it.’ (Christus und die Samariterin ) b. In bérge thẹ er mo zéinti on mountain which he him.DAT showed ‘On the mountain which he showed him.’ (Otfrid II , ) c. sô begriffet er ro gnuoge so take he (of) her enough ‘So he took enough of her.’ (Memento Mori , ) At least for the first two, the short form—which also occurred after verbs in second position—is attested with considerable frequency (cf. Braune : §, Anm. : ‘treten oft Kürzungen der Wortform ein’ – ‘shortening of the word form often occurs’). ... Merged forms A third kind of phonological process that occurred in the WP is merging. The complementizer thaz (or probably the older form the, cf. Schreiber ) may have merged with ih/iz ‘I/it’ yielding theih/theiz (note that both forms also show i-umlaut, cf. Section ...):
23
Examples and numbers (of occurrences in Otfrid) are taken from Somers Wicka (: ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
() a. ni wáne theih thir gélbo not think that.I you deceive ‘Don’t think that I deceive you.’ (Otfrid I , ) b. ál theih iu gibíete all that.I you.PL command ‘All that I command you.’ (Otfrid V , ) c. theiz ein gidróg wari that.it a ghost was ‘That it was a ghost.’ (Otfrid III , ) Note, however, that contracted forms of the demonstrative pronoun thaz and the finite auxiliary ist (> theist) are also attested (Somers Wicka ). The contracted form theiz, which was used by Otfrid much more than the full form thaz iz (cf. Somers Wicka : ), is still attested in the MHG of Wolfram von Eschenbach:24 () a. deiz in dem ermel wider want that.it in the sleeve again wound ‘such that it reached until his sleeve.’ (Parzival , ) b. deiz al daz volc was verswigen that.it all the people was concealed ‘that it was concealed from all of the people.’ (Parzival , ) Apart from thaz, the wh-pronoun waʒ ‘what’ could also be merged with ih to yield the form weih ‘what-I’. This contraction, however, is only attested three times in Otfrid’s Gospel Book: () a. Húgi weih thir ságeti consider what.I you said ‘Consider what I told you.’ (Otfrid I , )
24 See also Somers Wicka (: ). According to her, the ‘entire text [of Parzival] contains twelve tokens [ . . . ] of the clitic group deiz, next to nineteen occurrences of the non-cliticizing daz ez collocation.’ Note, however, that some of the tokens of the contracted form deiz are not found in the manuscripts, but are due to text interpolations made by the first editor, Karl Lachmann. This is, for instance, the case with , or , where the recent edition based on the Lachmann text has deiz, but the notes in the critical apparatus reveal that the manuscripts have the non-contracted form daz ez (see Knecht and Schirok (= Parzival): , ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß b. weih fon thír nam what.I of you took ‘What I took of you.’ (Otfrid II , ) c. Wéist thu weih thir rédinon know you what.I you say ‘Do you know what I told you?’ (Otfrid III , )
However such contractions might be analysed, they are evidence for a cluster formation between the clitic and its host. ... Reanalysed forms There is probably only one step from contractions to reanalysis: subject pronouns in inverted order could get reanalysed and thus became part of the inflection. The most prominent case is the SG, whose original verbal ending consisted only of an ‐s, to which then a ‐t (< thu, du ‘you-SG’) was added—see (). () a. daz thu thes kiscephes that you that scoop ‘That you scoop it.’ (Christus und die Samariterin ) b. biwaz kerost thu why demand you ‘Why do you demand?’ (Christus und die Samariterin ) The same happened later in Bavarian with the PL where an -s (< ös ‘you-PL’) was added to the original ending resulting in the new form -ds (Weiß a). Reanalysed forms led to DA. In the case of the SG, DA disappeared when the new inflectional marker also occurred on verbs in final position—but there are other cases where DA still exists nowadays (see examples given in Section ..). Another prominent case is the PL: the form of DA found in the Swabian dialect spoken in the West Allgäu, as in () (= (), Section ..), is attested since OHG times. () a. mer gloubed we believe
(Swabian)
b. gloube mr believe we The DA with the PL has a long history: it is very characteristic for the MHG period, but according to Braune (: ), the first occurrences can be traced back to the ninth century (see also Axel and Weiß ). The example in () is from Tatian: () (hunc autem nescimus unde sit.) thesan niuuizuuuir uuanan her ist this.ACC NEG.know.we whence he is ‘but as to this man we do not know where he is from’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
In contrast to DA in SG (or PL), a PL pronominal subject in the WP (partially) suppresses overt inflection on the verb (or triggers the deletion of verbal inflection).25 PL subjects preceding the verb still trigger overt verbal inflection (cf. (a)), which indicates that the deletion of verbal inflection linked to PL subjects in WP is indeed a case of DA. In dialects like Central Bavarian (Weiß , a), the pronominal clitic was eventually reanalysed as an inflectional marker.26 All the developments outlined here demonstrate that the WP was a position where subject clitics tended to become reanalysed as (part of the) verbal inflection. .. Summary Although there are no distinct paradigms for full and reduced/clitic pronouns in OHG and the pronouns occurring in the WP are nothing but reduced forms, there is good evidence that at least some kind of WC already existed in OHG times. As in modern dialects, there is a correlation between form and position in that most reduced forms are restricted to the WP, where they cluster with each other and form a prosodic unit with their host. In OHG, the pronouns occurring in the WP seem to have mostly been just reduced pronouns, and not clitics. The interactions between host and reduced pronouns attested in OHG are not confined to phonological processes, but also entail morphological effects (i.e. reanalysis as inflectional markers, and at least in these cases the reduced pronouns must have evolved into clitics first). All these phenomena may be evidence for the special status of the WP already in OHG times (which is NB the third and no longer the second position).
. Emergence of Complementizer Agreement (CA) .. A new explanation DA and CA are the most prominent features of the WC. As we have seen, DA resulted from cliticization of pronominal subjects onto the verb in second position where a special verbal agreement morpheme arose from phonological processes (such as assimilation, fusion, or truncation), which then triggered reanalysis. On the other hand, as noted above, synchronically, CA is in some cases independent of the presence of subject clitics (see example () in Section ..), and in some cases the CA inflection marker did not emerge from a subject clitic, but is identical with the regular verbal inflection (see (b) and (d) in Section ..). So if there is, at least synchronically, no strict correlation between subject clitics and CA, the crucial question for its diachrony is whether subject enclitization at least 25 This form of DA resembles the so-called Northern Subject Rule (NSR) which holds in English dialects spoken in northern England, Scotland, and the north of Ireland: in these dialects, the s-suffix was generalized to all person and number combinations in the present tense. The s-suffix is suppressed when a non-rd person pronoun immediately precedes or follows the verb, yielding contrasts like the birds sings versus they sing and dances (Pietsch ). Although the NSR pattern is assumed to have emerged in inversion contexts in ME (Pietsch ), it was later generalized to all syntactic contexts where a non-rd person pronoun and the verb are adjacent. 26 In very rare cases, the new inflectional ending ‐ma was (nearly) completely generalized: it appears also on auxiliaries and modal verbs in final position, but never on lexical verbs (Weiß , a; Fuß ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
played a role in the emergence of CA—or whether CA is just the morphological expression of the φ-features which C possesses anyway (Chomsky ). As we have seen above, subject clitics were a necessary precondition for the emergence of DA, and if we can make a plausible claim that DA is a necessary precondition for the evolution of CA, then we would have made a good case that subject clitics also stood at the beginning of CA. Such an explanation would be more desirable, because it is in accordance with the extreme infrequence of CA among the languages of the world which would be unexpected if C had φ-features of its own (see also Fuß ). One further aspect in favour of the proposed diachronic scenario is the fact that full CA paradigms are the exception even within the Continental West Germanic dialects, whereas defective paradigms (i.e. the restriction to single persons) are the normal case. Existing hypotheses have always assumed a correlation between the two phenomena. Weise (), for instance, attributed the emergence of complementizer inflection to analogical extension. Recently, Goeman (), Kathol (), and Zwart () also proposed an analogical-extension approach. Thus, Zwart’s () proposal is based on the analogical schema in (): ()
kunt : kunnə :: dat : dattə
This means ‘that complementizer agreement originates as an analogical extension of the morphology of the verb in verb-clitic constructions to the complementizer in complementizer-clitic constructions’ (Zwart : ). However, it remains unclear why speakers should do that, since complementizers do not really belong to the inflectable parts of speech. Weise’s () explanation involves two intermediate steps to account for the transfer of the inflectional marking onto complementizers: at first, new pronominal forms developed in verb-clitic sequences containing the inflectional ending (this is attested for the PL, where we observe forms like der ‘you’ instead of ihr in some dialects); second, when encliticized to a complementizer, the new pronoun gives the impression that the complementizer is inflected (e.g. wenn +der ‘if you’ if-PL+you’). However, Weise’s explanation cannot be the only reason for the emergence of CA, because CA is much more widespread within German dialects than such new pronouns are (cf. Weiß a). Although both analogical-extension approaches cannot convince entirely, I will also assume a kind of analogical extension, but in a different and more detailed form. In those cases where the verb in second position developed a new inflection which differed from the standard inflectional marker (which still occurred in final position), children acquiring the language may have associated the new inflection with the position (i.e. C) rather than with the verb itself. They may then have also produced inflection on the complementizer. So I would like to propose the following hypothesis in () for the emergence of CA: ()
Hypothesis The emergence of CA involved three steps: Subject clitics > DA > CA
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
This means that the analogical extension of inflection to complementizers had its origin in subject clitics which were reanalysed as (part of the) inflection thus producing DA. Children learning a DA dialect could probably get the impression that the special inflectional morphology appearing on verbs in C was connected with the syntactic slot as such, irrespective of whether the lexical item occupying the slot is a verb or a complementizer. They may have got this impression because the new inflection was restricted to the C-position. As we have seen above, DA originated from phonological processes operating on sequences consisting of a verb (in C) and a subject pronoun. Although the DA inflection was always pronominal in origin, in some cases it was replaced by nonpronominal endings: this is attested by several German dialects where in inversion structures, the normal ending in the PL and PL was substituted by a form taken from another tense or mood (Höhle ). This system can be found in Low German dialects since the MLG period—as well as in OE and Dutch (dialects). It is presumably not by chance that there was already a uniform plural (i.e. all three persons in the plural have an identical verbal ending) in OLG (Gallée ), where the PL ending -ð/d also had intruded into the PL and PL. As a consequence, at least in the PL present indicative, the speakers (or language acquirers) could no longer connect the inflectional difference to its phonological motivation. From this point on, DA was a syntactic rule. In other cases, the development of special CA inflection out of subject clitics may have caused the association of the syntactic slot with inflection— and that was presumably the more general way CA emerged. The emergence of CA, i.e. the analogical extension of inflectional marking onto complementizers, may have been boosted by an additional development: in some cases, the phonological interactions between the inflectional endings and the subject pronouns also resulted in new pronominal forms. This happened in the PL (mir ‘we’, which is the common form in modern High German dialects, is attested since MHG times, but may have already existed in spoken language at OHG times, cf. Axel and Weiß ) and in the PL for which forms such as der ‘you’ are attested in many Central German dialects (e.g. in Hessian, cf. Bohn and Weiß ). When these pronominal forms are adjacent to complementizers, the resulting sequences may appear to contain an inflected complementizer, that is wennder ‘if/when-you’ may be (mis-)segmented as wenn-d+der ‘if/when-PL+you’ in analogy to the correct segmentation of gehdder ‘go-PL+you’ as geh-d+der.27 Seen in this way, the analogical extension may have been triggered by more than one cause making it a little bit less mysterious as it seemed to be within previous approaches. Within the explanation proposed here, the emergence of CA is more or less the result of a historical accident (some subject clitics were by chance reanalysed as
27 The development in the PL was slightly different: the clitic form ma (< mir) which suppressed the regular verbal inflection (cf. Section ...) was reanalysed as an inflectional marker in verb clitic sequences so that complementizer clitic sequences like dass-ma ‘that-we’ made the impression of being inflected. Although it is very common within High German dialects that the PL pronoun begins with minstead of w-, the further development into an inflection marker is presumably restricted to Central Bavarian (cf. Weiß a).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
inflection markers) rather than the morphological expression of C’s inherent φ-features. This explanation has several advantages: • It is in accordance with the typological rarity of CA (Weiß a; Corbett )—if C had inherent φ-features in every language, this rarity would be unexpected. • It explains why CA is restricted to single persons (mostly SG) in most dialects, whereas a full CA paradigm is the exception (with a curious areal distribution, cf. Weiß a). • It gives us a natural explanation for the two different types of CA dialects we encounter within Continental West Germanic dialects shown in Table .: dialects with defective paradigms display only those CA markers which are (partly) the result of reanalysis, whereas dialects with full paradigms also have CA markers which do not go back to subject clitics.28 T . Two types of CA dialects Type
CA depends on subject cliticization Dialects with a partial CA paradigm where the CA morphemes are pronominal (in the sense defined above): e.g. Central Bavarian (Weiß ) (i) a. wennsd SG b. wemma PL c. wennds
Type
PL
CA is independent of subject cliticization (and results from analogical extension of verbal inflections) Dialects with a complete CA paradigm where at least some CA morphemes are verbal, i.e. non-pronominal (in the sense defined above): e.g. Sechsämterisch (Rowley ) (ii) a. wálØ i SG d. wáln mer PL b. wálst (du) SG c. wálØ si
SG
e. wálts diets PL f. wáln si
PL
(Zwart ; Weiß a)
.. Diachrony of Complementizer Agreement According to the explanation given here, it was presumably in late OHG or early MHG times that the system of CA developed, because subject enclitization and DA were already present at that time (Axel and Weiß ). However, it is hard, if not impossible
28 Additional evidence for the connection of CA inflection and verbal inflection comes from changes within the verbal paradigm that also occurred in the CA paradigm. For instance, the South Hessian variety spoken in Rheinhessen (which has CA mostly in the PL) has developed a uniform plural, that is all three persons in the plural have an identical verbal ending in ‐ǝn—and this form also appears with complementizers, e.g. dasņ-ɐ ‘that.PL-you’, owǝn-ɐ ‘whether.PL-you’ (cf. Bohn and Weiß ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
to tell exactly when inflected complementizers actually appeared for the first time. Traditional German scholars seem to hold divergent opinions about this issue: Pfalz (: ), for example, claims that complementizer agreement is of younger origin, whereas Weise (: f) suspects that it must be quite old. He argues that the fact that this phenomenon occurs in the language of the Transylvanian Saxons strongly suggests that inflected complementizers must have existed as early as in the twelfth century. As it is very unlikely that complementizer inflection developed spontaneously when the emigrants were already in their new homeland, Transylvania, they must have brought this peculiarity with them from the Luxemburgian-Mosel-Franconian dialect. The majority of people had already emigrated in the twelfth century under the reign of King Geysa II (–). This is why Weise concludes that at that period there was already complementizer inflection in the nd singular and plural as it is still the case in Noesnarian, a variety spoken in Transylvanian Saxony, and in Mosel-Franconian (Axel and Weiß ).
However, the first examples attested in written language showed up only centuries later. The oldest unambiguous example of an inflected complementizer is from and can be found in the Bavarian carnival play Ein kurtzweilig Faßnachtspill, cf. (): wanst ‘if.SG’—which is completely identical to the modern form. () Ja! wanst mir wolst var liegn [vorlügen] yes, if.SG me want.SG to lie ‘Yes, if you want to lie to me.’ (after Axel and Weiß : ) The Nurembergian poet Hans Sachs, however, uses forms of the complementizer— such as dast ‘that.SG’ or weilt ‘because.SG’—which look like inflected complementizers (Volodina and Weiß ) nearly one century earlier. () a. dast ein zygeuner seist that.SG a gipsy are ‘That you are a gipsy.’ (Fastnachtsspiele , ) b. weilt mir warst versprochwn because.SG me were promised ‘Because you were promised me.’ (Fastnachtsspiele , ) Note that the combination of the complementizer dass with the enclitic form of du would rather result in a form like dastu (which is also frequently attested at that time, cf. Volodina and Weiß ). The form weilt ‘because-SG’ resembles the type of CA found in Alemannic and South Bavarian dialects, where the CA inflectional marker consisted only of a ‐t (which was lost in some of these dialects in the meantime (Weiß a)).
. Pronouns in MHG Since special investigations on the syntax of MHG pronouns are largely absent, we know very little about it. However, we can be sure that MHG possessed a WP in the
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß
modern sense, i.e. a special position exclusively reserved for pronouns where associated phenomena like phonological processes or even DA can be observed. Evidence for the existence of a WP also in MHG comes from textual data. In some MHG texts, personal pronouns seem to strongly prefer the WP and not the prefield. Especially in texts of non-professional writers such as Christina Ebner’s Der Nonne von Engelthal Büchlein von der Genaden Uberlast (written between and ) one often finds that the prefield seems to be banned for topical elements, which means for non-pronominal elements that their topicalization is nearly always left dislocation (In den selben zeiten da komen siben ebt ‘at the same time, there came seven abbots’) and for pronouns that they are largely restricted to the position immediately following the finite verb, i.e. to the WP. A count of the first root sentences containing at least one personal pronoun yields that in seventy of them, the pronoun is in the WP; the thirty-six pronouns in the prefield are all subject pronouns. Similar results are found in other texts: Jüngerer Physiologus (WP: versus prefield: ), Mitteldeutsche Predigten (WP: versus prefield: ). This result is somehow unexpected, since the prefield (or SpecCP) hosts phrases which are moved in this position for (more or less) pragmatic reasons.29 An important aspect of the filling of SpecCP seems to be to guarantee coherence between sentences within a series of sentences (Speyer ). Another function of constituents in the prefield is frame-setting, that is constituents which set the (temporal, locative, or modal) frame of the sentence. According to Speyer (), personal pronouns are elements which are highly qualified as prefield fillers, especially in oral and informal registers. However, there are also texts with results almost to the contrary, e.g. the St. Trudperter Hohe Lied (WP: versus prefield: ), Lilie (WP: versus prefield: ), or the Oxforder Benedictineregel (WP: versus prefield: ). Many (but not all) of these texts are written in a rather formulaic style where a few clausal structures are frequently repeated.30 Therefore, due to the lack of larger corpus studies on the pronominal syntax of MHG, we can only be certain that the WP is a highly preferred position for pronouns—besides the prefield. Although one cannot decide in each case whether the pronoun in the WP is just a reduced form or a clitic, there are many clear cases. In spellings like alser ‘when he’ or bater ‘asked he’ (Paul : ), the pronoun is presumably just a reduced pronoun which could also appear in this form in the prefield, whereas a pronoun like siu/sie/sî ‘she, they’ was often reduced to s in the WP (e.g. mohten si > mohtens, daʒ si > daʒs, cf. Mettke : ), but not in the prefield. In this case, it is obvious that s is a clitic pronoun and not just a reduced one. Now let us turn to the other phenomena associated with the WP. According to Paul (: ), all MHG pronouns displayed two forms: a full accentuated form and a reduced form. The reduced form appears in the WP where it triggers phonological changes of the verb which are very similar to Otfrid’s synaloepha (cf. Section ...): the vowel ending of the verb (especially in the SG) was deleted when a vowel initial
29 30
But see Chapter for discussion of the diachrony of prefield filling. In the St. Trudperter Hohe Lied, for instance, there is a series of sentences beginning with ez ist . . . ‘it is . . . ’.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The Wackernagel complex and pronoun raising
pronoun followed, cf. neme ich > nem ich (Herbers ). As was also the case with Otfrid, vowel deletion was much more common in the (often disyllabic) present tense than it was in the (often trisyllabic) preterite tense, and in general in verse texts rather than in prose (Herbers ). Herbers () mentions an interesting development: whereas in the beginning, the deletion of the vocalic verbal ending was triggered only by vowel-initial pronouns, this correlation became looser in the course of the MHG period so that any (subject) pronoun could trigger it.31 All this information may (but need not) be evidence for a development from phonology to DA. At first, vowel deletion seems to have been mainly conditioned by phonological factors such as the number of syllables (linked to present versus preterite) or whether the following pronoun began with a vowel or not. Later on, such phonological requirements obviously became less important and the mere presence of a pronoun was enough to trigger vowel deletion. Vowel deletion in these cases is equivalent to the deletion of the whole inflectional marker on the verb.32 This is reminiscent of the DA case of the PL in Swabian (mentioned in Section ..), where the inflectional marker ‐ed is shortened to ‐e when the pronominal subject follows in the WP. This kind of DA is attested in the PL since late OHG times (Axel and Weiß ) and is especially frequent in MHG times, where the subject pronoun in the WP triggers shortening or even suppression of the verbal inflection (Paul : ). At the same time, the pronoun wir ‘we’ changed its form into mir, which was originally confined to the WP, and was later generalized in other positions as well. Whether the deletion of /ǝ/ in the SG constitutes a further instance of DA has yet to be investigated. Reduced pronouns in the WP also displayed a number of properties which we have identified in Section . as characteristic traits of the WC. They occasionally merged with their host (cf. Section ... with the examples in () from Wolfram), but very often they form clusters as we have already seen in OHG times. If more than one pronoun is present in the WP, the second one is in general segmentally reduced and cliticizes onto the first one (e.g. ërn(e) < ër in(en) ‘he him’, ichn < ich in ‘I him’, cf. Paul : ). This cluster then cliticizes as a whole to its host (e.g. wande ër ëʒ ‘because he it’ > wanders, cf. Mettke ). If we turn now to the order of pronouns within the WP, we see that the relative order of pronominal objects is variable and depends on the specific combination. In his study on the Prosalancelot, Fleischer () showed that st/nd person singular dative pronouns often precede accusative pronouns (cf. (a)), whereas the rd person singular dative pronouns show variation in this respect (cf. (b) and (c)). () a. Nun bringent mirn herre now bring me.him here ‘Now bring him hither to me.’ (Prosalancelot I , ) Herbers (: ): ‘die Tilgung auch vor Konsonant nimmt hingegen deutlich zu’ (‘in contrast, the deletion increases significantly, even before consonants.’). 32 This also holds for the SG in the weak preterite tense: mohte ër > mohter ‘he may’ (Paul : ). 31
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Helmut Weiß b. und er sagten ir and he said.him her ‘and he told it (his name) to her’ (Prosalancelot II , ) c. Ich sagen uch warumb ich ims nit gestattet I say you why I him.it not allowed ‘I tell you why I did not allow it to him.’ (Prosalancelot II , )
This situation is similar to that in modern dialects (Weiß a): whereas subject pronouns nearly always precede object pronouns,33 the surface order of pronominal objects is more variable and seems to depend on non-syntactic factors like the animacy hierarchy as well (cf. Fleischer , ).
. Conclusion In this chapter, I have presented arguments and data in support of the following claims: • The WP in Continental West Germanic dialects differs from the WP in the original sense in that it is not the second position (i.e. the position after the first word or after the first constituent/phrase), but the third position, i.e. the position after the second constituent (i.e. immediately following C/Fin). • There is evidence for the existence of the WP already in OHG times (e.g. phonological interactions between pronouns and verbs in second position or complementizers respectively, reanalyses of subject pronouns as inflection markers). • For MHG, the evidence is less clear, mainly due to lack of detailed empirical work. However, it is obvious that there are phonological changes (such as reduction, deletion, or assimilation) on both the verb and the pronoun when they are both in the WP. • Diachronically, the development of the WC started with subject cliticization to C in root sentences (i.e. V/V sentences), where some of the subject clitics were reanalysed as (part of the) verbal inflection, thus giving rise to DA, which in turn triggered the emergence of CA via analogical extension. This is the reason why synchronically there are two types of CA: a ‘pronominal’ type, where only pronominal inflection was extended to C in general, and a ‘verbal’ type, where any verbal inflection marking person and number is associated with C.
33
There are a very few exceptions to this rule in modern dialects (cf. Weiß ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
9 Serialization of full noun phrases in the history of German AUGUSTIN SPEYER
. Introduction .. Remarks on the German middle field The middle field in German corresponds roughly to the VP and—if assumed—IP architecture without the final V and I heads which form the right sentence bracket (cf. Chapter of this volume and Sabel for the relation between the field model and generative concepts). I am deliberately phrasing this somewhat vaguely, for the details of the analysis are a matter of debate. Suggestions range from a quite articulated structure in the IP domain (e.g. Schmidt , who proposes several Agr-heads at the left periphery of the middle field, forming the Wackernagel position) to an analysis in which the left parts of the middle field are part of the C domain (Coniglio ). Frey (), like others, takes a middle position in suggesting that the middle field contains a Topic Phrase which forms the left edge of the middle field. As the fine structure is far from clear, I use the ‘traditional’ terms IP and CP in this chapter which is not meant to imply that there is no fine structure, but simply is shorthand for the functional projections having to do with finiteness (TP, AspP, AgrXP, etc.) and with sentence mood (ForceP, TopP, etc.), respectively. Leaving the left area of the middle field out of the discussion (to which reference is made in Chapter ), we concentrate in this chapter on the structure of the verbal phrase in the middle field. The problem is that it is not immediately clear what the structure really looks like, for German is a language that makes use of scrambling rather freely (cf. e.g. Rosengren ). Besides the question of which is the base structure of the German verb phrase (see also Chapter ), another question is, ‘what induces scrambling?’ This chapter tries to address the question from a rather descriptive point of view, that is, what surface word orders occur in German, which factors can be identified that determine the surface word order, and, finally, if we can infer some properties of the base structure of the verbal phrase. So, differently from some other chapters in this book, the generative analysis plays only a minor role in this chapter.
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Augustin Speyer . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
.. Serialization of noun phrases in Modern German German is a language with a relatively ‘free’ word order. This does not mean that the relative order of constituents—let alone single words—is completely unconstrained. Rather, it means that constraints that are not part of syntax in the stricter sense have an impact on the serialization. These constraints are partly a matter of the shape of the elements in question, partly belonging to the conceptual-semantic system, and partly of a pragmatic nature (cf. e.g. Müller ; Reis for the interplay of grammatical and pragmatic factors). Among the factors that have been identified in Modern German are the following, which I give in the form of generalizations with respect to serialization (the sign ‘>’ is to be read as ‘tends to precede’). a. noun phrases > prepositional phrases (e.g. Lenerz ) b. subject > dative object > accusative object (e.g. Lenerz ; Fortmann and Frey ; Primus , ) c. pronominal > lexical noun phrase (e.g. Lenerz ) d. shorter constituents > longer constituents (‘law of increasing constituents’ in Behaghel –, Vol. IV: ; see e.g. Lenerz ) e. animate referent > inanimate referent (e.g. Zubin and Köpcke ) f. referent with greater empathy > referent with less empathy (e.g. Lötscher ) g. referent with higher agentivity > referent with lower agentivity (e.g. Zubin and Köpcke ; Fortmann and Frey ) h. definite > indefinite noun phrase (e.g. Lenerz ; Reis ) i. topic > comment (e.g. Lenerz ) j. given > new (e.g. Musan ) The factors (a–d) are clearly ‘shape’ factors. To some extent they are a motley crew, as some of them are syntactic factors (like factor (b), which reflects the ‘case hierarchy’ presented in Primus and is, together with factor (a), the defining factor of serialization in present-day English), some of them probably rather prosodic factors (factors (c, d)). Factors (e–h) represent the conceptual facet of the problem. Depending on the nature of the referent as such (rather than the shape of the expression denoting the referent), expressions can be put in a certain order. A certain egocentric hierarchy seems to be at play (cf. the ‘animacy hierarchy’ of Silverstein as an attempt to formulate the underlying hierarchy) in that expressions denoting referents that are more similar to the speaker/writer (factors (e, f)) or more similar to the conception that he/she has of him/herself (factor (g)) are placed in front of others. Factor (h) is a factor that is on the borderline between ‘shape’, concept, and information structure, as definiteness is clearly correlated with information-structural categories such as newness. Factors (i, j) represent information-structural constraints: given information is placed before non-given information; the entity the sentence is about stands before the information that is said about this referent.1
1 Lenerz’s () theme–rheme condition is a blend between factors (i) and (j), in accordance with the monostratal conception of information structure at the time.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
These tendencies describe the ‘unmarked’ word order in German, which is defined as the word order that can be used in most possible contexts, including wide focus (Höhle ). Some of these tendencies are stronger than others, but note that none of these tendencies is so strong that it can override the others. If middle field serialization is formulated in a coherent theoretical model, competition models suggest themselves (cf. Reis ). The tendencies are either thought of as cumulative—that is, the more tendencies a candidate conforms to, the better it is (Lötscher ; Zubin and Köpcke )2—or they are built into an optimality theoretic model (Müller ). The above generalizations about Modern German can be formulated in standard generative theory. The preference of subjects to stand before objects (factor (b)) reflects the base-generated word order within the VP (or possibly the IP). We can find several adjunction sites for several sorts of adjuncts, including one position above the lowest object position and one below it (cf. Frey and Pittner ). Apart from this basic pattern the structure allows for extensive scrambling. Scrambling is first and foremost any deviation from the word order identified as basic. No further assumptions about the nature of this operation or its landing sites are needed at this point.3 The base generated word order, from which other word orders are derived by scrambling, is however not completely clear. There is evidence for a base order S(ubject) > D(irect)O(bject) > I(ndirect)O(bject), the direct object thus being positioned higher in the tree than the indirect object (especially binding facts, cf. Barss and Lasnik ; Larson ; Müller ; Speyer c), but the evidence is not unambiguous (Rothmayr , but see the argumentation in Speyer c). An argument in favour of a structural order S > IO > DO is the fact that it is clearly the unmarked word order in Modern German (Lenerz ).4 Note that the so-called unmarked word order need not reflect the base-generated word order (see e.g. Höhle ). This is probably most conspicuous with psych verbs like gefallen ‘please’, where the unmarked word order is dative experience before nominative stimulus ( . . . dass dem Mädchen der Film gefällt ‘ . . . that the movie pleases the girl’); the base order is certainly nominative before dative as the noun phrase receiving nominative is base-generated higher in the tree.
2 The numbers given in Zubin and Köpcke () clearly favour an accumulative model, although they are not very explicit about that. See also Lenerz () for a model built on interaction of constraints. 3 There are several conceptions of the nature of scrambling. There are three important questions in that respect: is scrambling an instance of movement, or are word orders deviating from the ‘basic’ word order base-generated, just like basic word order (e.g. Fanselow )? if scrambling is movement, is it movement into pre-fabricated functional projections, or is it movement into ad-hoc generated Chomsky-adjoined positions (e.g. Müller and Sternefeld )? and finally, is scrambling A- or A’-movement (e.g. Fanselow ; Müller and Sternefeld )? I follow here the view that scrambling is movement into ad-hoc generated positions, mostly for the clarity of the exposition. I do not adopt any specific view of the exact nature of scrambling here, mostly because this cannot be decided from the historical data. 4 Additional evidence comes from clitic pronouns in German dialects, which always show the IO > DO order in the WP, see examples (a) and (b) in Chapter (cf. Weiß , ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
Scrambling is done partly to satisfy prosodic requirements (this might be the case with the tendency pronoun > lexical NP, and probably the ‘law of increasing constituents’), and partly to meet information-structural needs (surely this goes for theme > rheme and given > new, but definite > indefinite might be a reflex of given > new in that given elements tend to be definite and new elements tend to be indefinite). It is a matter of debate what the status of the conceptual factors is. They might reflect different base orders, rather than being the outcome of scrambling (e.g. Fortmann and Frey ). If this is so, one would have to assume a multiple-base hypothesis for German, in the sense that the arguments have no predetermined structure reflected by case assignment (such as e.g. [vP DP-nom [v’ [VP DP-acc [V’ DP-dat V]] v0]]), but that the tree is built in accordance with nongrammatical factors such as animacy or givenness. Similar accounts have been given for older stages of English (e.g. Pintzuk ; Petrova and Speyer ), so this idea is not too far-fetched. This is not the place to discuss this matter further; for the purposes of this chapter we treat the conceptual factors (e–g) as factors motivating scrambling. In addition to scrambling, constituents can be moved to positions that are rightadjoined to the clause, leading to the phenomenon of movement to the post field, which seems to be constrained mostly by prosody (cf. the rule that the more complex a phrase is, the more acceptable it is in the post field; Hoberg ; AverintsevaKlisch ) and by conversation structural requirements (Auer ; Uhmann ; Selting ). The clausal structure of Modern German that I adopt can be schematized as in (a), following Suchsland () and Sabel (). In addition to the minimal clause structure given in Chapter , it includes the functional projections vP and IP. The label _P signifies positions, which can be opened up, but in most cases need not be so that material can be moved into these positions. Phrase-internal adjunction positions are ignored. Note that there are two zones in the middle field that can serve as targets for scrambling operations, namely a position to the left of the subject (the target of what I refer to as IP-scrambling, which is movement to a position Chomsky-adjoined to IP; (b)) and a position to the right of the subject (the target of what I refer to as vP-scrambling, which is movement to a position dominated by IP but higher than the vP/VP-complex, Chomsky-adjoined to vP; (c)). The bracketing and indexing in (b) and (c) is simplified. Scrambling to either of these positions can be subject to different constraints so that it is necessary to distinguish between these two types of scrambling. In Modern German, vP-scrambling is much more common. If the relative order of direct object and indirect object (or nominal object and prepositional object etc.) is flipped, relative to the base structure (which is assumed here as accusative > dative, see further below in this section), we can interpret this as an instance of vP-scrambling. This type of scrambling is very susceptible to the cognitive and information-structural factors that have been identified further above. IP-scrambling, on the other hand, is relatively rarely applied (with the exception of scrambling of pronouns to the left edge of the middle field, the exact analysis of which is a matter of debate, see e.g. Lenerz and especially Chapter of this volume). Scrambling of a full noun phrase object over a full noun
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
phrase subject is rarely motivated by cognitive factors if it is applied it is mostly instigated by an information-structural factor such as placing old before new information. ()
a.
CP _P (prefield)
C0 iterative (IP-scrambling)
C’ IP _P
IP IP
iterative (post field)
I’
NPi iterative (vP-scrambling)
_P
Ij0
vP _P
vP NPi
v’ vj0
VP YP
V’
XP V0 b. dass [den wahren Mörder]1 [IP der Inspektor [vP t1 nicht gefunden that [the true killer]ACC [the inspector]NOM not found ‘that the inspector has not found yet the true killer’ c. dass [IP der Zeuge [vP [dem Inspektor]1 den Vorfall t1 gemeldet that [the witness]NOM [the inspector]DAT [the incident]ACC reported ‘that the witness has reported the incident to the inspector’
hat]] has hat]] has
I assume a base order DO > IO in German, although the ‘stylistically unmarked’ word order is IO > DO. This is not a contradiction, as stylistically normal word orders need not reflect the base order (see Höhle ). What is perceived as ‘unmarked’ might be the most common word order, and there may be reasons independent from structural simplicity that can account for a word order variant being the most frequent one. One such factor, as will be outlined in the course of this chapter (Section ..), is the Thematic Hierarchy (following e.g. Primus ). Referents are ordered according to the agentivity. As being animate is a typical feature of an agent, a high degree of agentivity often coincides with the referent being animate. As the IO tends to be more agent-like than the DO, especially with regard to animacy (IOs tend to refer to animate referents, DOs not), the order IO > DO results from this ordering principle. But the main reason for the assumption that the base order is DO > IO comes from certain binding facts. These binding facts presuppose a hierarchically higher
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
position of the direct object as opposed to the indirect object at some stage of the derivation (a). In clauses like (b) and (d), the direct object must bind the indirect object, as can be seen from the fact that if the indirect object is a reciprocal, the clause is interpretable and grammatical (b), whereas it is ungrammatical if the direct object is a reciprocal (c). This judgement is independent from the surface word order (d) and (e). The grammaticality of (b) is a consequence of the higher position of the direct object (a). The ungrammaticality of (c) is a reflex of the fact that the reciprocal in direct object function is not bound by the antecedent which has indirect object function, which shows that the position of the indirect object is hierarchically lower than the position of the direct object (Barss and Lasnik ; Larson ; Müller ). ()
a.
CP dass
IP vP
Laura2 t2
vorstellt1 v’
VP DPacc die Gäste
b.
c.
d.
e.
V’
v0 t1
DPdat V0 einander t1 Ich sehe, dass Laura die Gäste einander vorstellt. I see that Laura [the guests].ACC [each.other].DAT introduces ‘I see that Laura introduces the guests to each other’ *Ich sehe, dass Laura einander den Gästen vorstellt. I see that Laura [each.other].ACC [the guests].DAT introduces ‘I see that Laura introduces each other to the guests’ Ich sehe, dass Laura einander die Gäste vorstellt. I see that Laura [each.other].DAT [the guests].ACC introduces ‘I see that Laura introduces the guests to each other’ *Ich sehe, dass Laura den Gästen einander vorstellt. I see that Laura [the guests].DAT [each.other].ACC introduces ‘I see that Laura introduces each other to the guests’
The evidence from binding is not wholly conclusive, though, as binding dependencies can be produced by scrambling as well (see e.g. the discussion in Larson ). If scrambling—in the sense of adjunction to VP (see Rosengren )—has occurred, however, we would expect ambiguity, as the binding dependencies should hold for all members of a chain. In other words, (e) should be grammatical, even if the underlying word order was DO > IO and the IO had been scrambled on top of the DO, simply because the IO in that case c-commands the DO ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases ()
CP IP
dass Laura2
vP
DPdat den Gästen3
vorstellt1 vP v’
t2
v0 t1
VP DPacc einander DPdat t3
V’ V0 t1
. German free word order in a broader historical context The leading questions that any investigation of the diachrony of serialization has to address are: first, is there a change in the base generated word order, possibly independent of any observable changes in the unmarked word order? Second, the question about the factors that cause scrambling: did the importance of one or other factor increase or diminish during the observable history of German? A question that is even more fundamental, but to which the leading questions might suggest an approach, can be formulated as follows: is the role of scrambling in general gaining or losing importance? That this is an important question can be seen by looking at English in which scrambling was eventually lost (cf. e.g. van Bergen ), which led to a strong superiority of purely grammatical factors for serialization. In order to attempt giving an answer to these questions, we should first try to compare German scrambling to word order phenomena in other, more or less closely related languages. Typologically speaking, German has a position between, say, English, where the word order is almost exclusively governed by the syntactic functions the elements have, and a language such as Latin, where the order of constituents is predominantly due to information-structural requirements (Panhuis ; Pinkster ; Devine and Stephens ; Spevak ). In languages like Latin, there is a grammatically determined constituent order, but it is hard to recognize as it is frequently overridden by information-structural ordering principles (Speyer b). If we look at other early attested Indo-European languages, we see that the ‘Latin’ state of affairs is the prevalent state: Ancient Greek and Sanskrit word order is equally ‘free’, that is, the word order is sensitive to information-structural, conceptual, even stylistic requirements. Other ‘old’ languages such as Hittite are more rigorous in the serialization of constituents, but on the whole, ‘free’ word order seems to have been a hallmark of PIE. If this is so, Proto-Germanic probably had a free word order as well. Given that the nature of the factors influencing the Modern German and the classical
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
Indo-European word order is similar, it is tempting to assume that the German ‘free’ word order is inherited directly from PIE times. Following this assumption, German would be a relatively ‘conservative’ member of the West-Germanic branch, whereas other related languages such as English or Dutch lost this freedom of linearization over time (cf. Speyer : ff). Some of the Modern German constraints on linearization can be observed in a similar fashion in the classical Indo-European languages. One of these is the tendency to put pronouns to the left of lexical noun phrases, an instance of the so-called Wackernagel effect—that is the tendency to move light elements into a position following the leftmost element, ‘element’ being a cover term for any linguistic unit (see Chapter ). This effect is a property of virtually all ancient Indo-European languages. In Ancient Greek, for instance, discourse particles regularly stand after the first word. Typical examples for such discourse particles are gar, marking that the sentence in which it stands is an explanation of the preceding sentence (on the modelling of discourse relations, see e.g. Kamp and Reyle ; Asher and Lascarides ; Asher and Vieu ), de ‘but’, or ge ‘at any rate’ (see any Greek reference grammar on that). On the other side, there are constraints that are clearly of a younger age. The tendency of definite noun phrases to precede indefinite noun phrases might have to do with the morphological marking of such phrases as definite or indefinite; in that case, the constraint can have developed only after definiteness and indefiniteness came to be indicated robustly by the respective article. But, even if some constraints have clearly developed only in the course of OHG and MHG, the system as such might be old—of Indo-European origin. If it is indeed the case that the freedom of Modern German word order is a trait inherited from PIE, we should expect that the same goes for attested historical stages of German. Let us take vP-scrambling as an example: we should expect that the relative order of direct and indirect object should permute in accordance with the constraints identified in Section .. We should find both orders DO > IO and IO > DO; IO > DO probably as a result of ordering according to the Thematic Hierarchy. If this is so, the order DO > IO should occur only if the direct object is more animate, more thematic, or has a higher value of some other relevant property than the indirect object, even though it is the base word order. A few methodological caveats are in order here. In investigating the factors playing a role for serialization of the middle field in earlier stages of German, one has to bear in mind that the modern strict bracket formation—that is, a middle field which hosts all constituents of the clause, with the exception of the prefield constituent and certain sentential constituents that are moved to the right—is a relatively recent phenomenon. Movement to the right used to be much less constrained in earlier stages of the language, well into the ENHG period (cf. e.g. Schildt and Part III of this volume). Example () presents a clause in which several constituents are moved to the right, into the post field.5 Here the following abbreviations are used to indicate the topological fields: PF = prefield; LSB = left sentence bracket; MF = middle field; RSB = right sentence bracket; PostF = post field. For a general overview over the fields and positions assumed within the topological model of the German clause see Chapter . 5
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
() (dv solt daz wissende sin) . . . you shall that knowing be [LSB das ] [MF min herze in mime libe ] [RSB fert ] [PostF spillende that my heart in my body goes playing von rehter froeiden, von diner gegenwertikeit]. of true joy of your presence ‘You shall know that my heart in my body leaps with joy because of your presence’ (Merswin, Mannen , ff) That means that we are dealing with less complex middle fields in OHG, MHG, and ENHG. Consequently, clauses which allow statements about the relative order of constituents—especially of lexical NP arguments—are not very frequent. This fact, and the reluctance to apply a syntactic theory which might guide an investigation of the details of word ordering might have contributed to the fact that hitherto not many studies have focussed on this topic. Some studies address the problem, however. In his seminal study, Behaghel (–, Vol. IV) identifies an unmarked word order subject > indirect object > direct object > adjuncts, from which the word order of clauses can deviate due to two tendencies. The first is more or less coextensive with theme > rheme (‘das Wichtigere später steht als das Unwichtige’, ‘the important [information] follows the unimportant [information]’, Behaghel –, Vol. IV: ); the second is factor (d), the ‘law of increasing constituents’. Subsequent studies of MHG and ENHG word order (e.g. Keienburg ; Erben ; Margetts ) basically repeat Behaghel’s statements without going into more detail. Ebert () notes that the generalization pronoun > lexical NP holds for ENHG, and speculates about the factors influencing the serialization of lexical NPs, citing grammatical factors (subject > objects), possibly pragmatic factors (topicality, focus, givenness), and the ‘law of increasing constituents’. In the following two sections, I give an overview of the order of lexical argument NPs in OHG and ENHG, based on Speyer (, , , c, ). Nonargument phrases are left out, as there is even less research done, but a closer study of non-argument phrases is planned.
. Constituent order in Old High German As shown in Chapter , the left periphery is already well established in OHG. Therefore it makes sense to speak of a ‘middle field’ as the zone following the finite verb in matrix clauses or the complementizer (subordinate conjunction) in subordinate clauses (Axel ; on the application of the topological model to older stages of German, e.g. Lötscher ; Ramers ; Speyer b). The OHG evidence for argument serialization is not really conclusive. This is not surprising, as most of the texts are direct translations from Latin, and some translators stick very closely to the Latin word order (see Chapter for an overview of OHG textual evidence). Some, like the translator of the Isidor, treat the text with more freedom. But even in the texts that are translated most faithfully to the Latin
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
original, such as the Tatian, we find every now and then sentences that differ from the Latin word order (cf. e.g. Lippert ; Robinson ; Dittmer and Dittmer ; Petrova and Solf a; Schlachter on the problem of OHG translations). So the evidence is perhaps clear enough to identify tendencies. I confine myself here to vP-scrambling. As sentences with three-place predicates are rather rare, especially if the condition holds that at least both objects should be realized as lexical NPs, the easiest way is to search for sample three-place verbs that govern a dative and an accusative object. The TITUS database contains virtually all OHG texts and has a built-in search engine, which is string-sensitive.6 The Bonner Frühneuhochdeutschkorpus also has a built-in search engine that allows searching for lemmata (but only stems are listed as lemmata, hence the order of the word list further below). The sample verbs I used for the OHG and the ENHG part are the same. I give here the complete list in the Modern German version, in alphabetical order of the verbal stems. ()
(an-, er-, ge-, ver)bieten, bringen, (anemp-, be-)fehlen, geben, gönnen, erklären, (ver)künd(ig)en, erlauben, (ab)lehnen, ausliefern, (an-, ein-, unter-, zu)ordnen, nehmen, anpassen, raten, rauben, (an-, zu)rechnen, schicken, abschlagen, senden, aussetzen, ausspannen, (gegenüber-, gleich-, vor-)stellen, (aus-, mit-, zu) teilen, anvertrauen, verübeln, verweigern, weisen, (vor-, unter)werfen, überantworten, erzählen, zeigen, (an-, aus-, ent-, unter-, vor)ziehen.
Note that the verbs bringen ‘bring’, geben ‘give’, and nehmen ‘take’ are the only verbs with a relatively high frequency. The numbers in Table . are based on a search in the TITUS database for forms of the sample verbs bieten ‘offer’, bringen ‘bring’, geben ‘give’, künden ‘announce’, nehmen ‘take’, senden ‘send’, teilen (= zuteilen) ‘distribute’, and zeigen ‘show’, leaving out the text of Otfrid. The numbers are too small to draw firm conclusions. The best that can be said at this stage is that there is a preference for the linearization IO > DO. The picture becomes clearer when we add the information about whether the word order is identical to the Latin original, differing from the Latin original, or whether there is no Latin original at all (Table .). We see that all instances of the order DO > IO are found in translated texts (four from Notker, three from Tatian, two from the Rheinfränkische Psalmenübersetzung) T . Variation of relative order of IO and DO in OHG IO > DO DO > IO
6
See also Chapter for an overview of online corpora of historical German.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
T . Variation of relative order of IO and DO in OHG, varying dependency from Latin identical to Latin original IO > DO DO > IO
differing from Latin original –
no Latin original or free translation
Total
–
and copy the word order of the Latin original (). If we look only at authentic OHG texts (a; from Erster Wessobrunner Glauben und Beichte) or translations in which the German text is a non-literal rendering of the content of the original (b) and at the cases in which the word order has been changed from the Latin original (), we get the order IO > DO exclusively. This is highly suggestive, even if we take into account that the data set is very small (due to the fact that such clauses are very infrequent). What it suggests is that the unmarked OHG word order is IO > DO. () diê bríngent kéba demo egelîchen these bring gifts the terrifying ‘they offer gifts to the awe-inspiring one’ (Notker, Psalter , ) Original: offerunt munera terribili offer.PRES.PL presents.ACC terrible.DAT ‘they offer gifts to the awe-inspiring one’ () a. daz danne aller menisclih fure sih selben gote that then all human for him self God.DAT reda geben scol speech.ACC give is.obliged ‘that then every human being is obliged to give account for himself to God’ (Wessobrunner Glauben , ) b. únde nément socrati dás uiuere and take.PRES.PL Socrates.DAT [the to.live]ACC ‘and they take away from Socrates the predicate “live” ’ (Notker, De Interpretatione , –) Original: cum dico non vivit Socrates vitam when say.PRES.SG not lives.PRES.SG Socrates.Nom life.ACC a Socrate disiunxi from Socrates.ABL disjoin.PERF.SG ‘when I say “Socrates does not live” I have disjoined life (= the predicate “to live”) from Socrates
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
() Der allen ménniscon ézen gíbit who all.DAT humans.DAT food.ACC gives ‘who gives food to all human beings’ (Notker, Psalter , ) Original: qui dat escam omni carni who gives food.ACC all.DAT flesh.DAT ‘who gives food to all beings’ Does this mean that the reverse word orders did not occur in OHG? The fact that there are no examples of authentic OHG DO > IO clauses in prose is not conclusive (argumenta ex negativo are not valid), but what is suggestive is that there are no examples in which a Latin order IO > DO was changed into DO > IO. This contrasts with four times in which a Latin order DO > IO was changed into an OHG order IO > DO. The IO > DO order is not as dominant in other texts. In Otfrid’s Gospel Book, roughly one third of double object constructions are in the order DO > IO (Table .). Nevertheless, IO > DO as in () is more common. Note that most cases of DO > IO are part of an idiom antwurti geben ‘give an answer to someone’ and the order DO > IO produces a rhyme, cf. (). So this evidence, while showing that the order is not rigidly IO > DO, does not contradict the impression that the order IO > DO is the unmarked order in OHG. Looking at OLG (Table .), we get similar numbers as in Otfrid (not surprisingly, as the OLG database also consists of bible epics, the Heliand and the Genesis), but here the cases of DO > IO are not influenced by end rhyme (the epics being in alliterative rhyme, but not end rhyme, as Otfrid; alliterative rhyme may have some influence on the word order as well, of course) and they do not cluster around one idiom, but occur with all verbs. These points indicate that, on the whole, there is greater freedom in object order in the OLG of the Heliand than in OHG, even though the quantitative difference is not significant (see Speyer ). T . Order of objects in Otfrid’s Gospel Book
Order IO > DO Order DO > IO Total
Number
Ratio (%)
. .
T . Order of objects in OLG texts
Order IO > DO Order DO > IO Total
Number
Ratio (%)
. .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases ()
bráht er therera uuórolti / diuri árunti brought he [the world].DAT [precious message].ACC ‘he brought a precious message to the world’ (Otfrid I , )
() Er gab tho ántuuurti / then líutin mit givuúrti he gave then answer.ACC [the people].DAT with gladness ‘He gladly gave the people the [following] answer’ (Otfrid III , ) () sagde uuord manag / firiho barnum said [words many].ACC [men’s children].DAT ‘he said many words to the children of men’ (Heliand –) We cannot say at this stage whether the information-structural and cognitive factors that play such an important role for serialization in Modern German are at play in OHG, too. Only considering the ten examples that are not directly following any Latin original, we see that in eight of those cases an animate indirect object precedes an inanimate direct object. In two cases both referents are inanimate, cf. (a) and (b). At any rate, this factor is not relevant enough to cause the scribe to change the Latin word order: there are several examples in which an inanimate object precedes an animate object in the Latin original and in which this word order is kept by the OHG scribe, cf. (a). This shows that the translators were tolerant with respect to this factor. The examples from texts that are not direct translations or are originally written in OHG all exhibit the order animate before inanimate, when there is an animate and an inanimate object involved, but as the animate object happens to be dative in all these cases and the inanimate happens to be accusative, this evidence is not conclusive. It might be the case that it is not an accident but rather a reflex of a general constraint of verb valency in OHG that the bearers of the roles that are associated with dative case are animate (similar constraints hold e.g. for subjecthood in Latin, so such a consideration is not as weird as it seems). We will come back to this issue below. In cases in which both referents are inanimate, the order dative > accusative applies, as in (b). While the evidence against the factor ‘animate before inanimate’ is inconclusive, there is some evidence against the constraint ‘old before new’. In (), a verse from Otfrid, the new information, represented by the dative object, is placed before the given information, represented by the accusative object. This is exactly the case that is ruled out by Lenerz () as infelicitous. This shows that information-structurally sensitive ordering in OHG is outranked by an ordering in accordance with case, or at least potentially subject to poetic licence. () a. An CHRISTO santa er lôsunga sînemo liûte at Christ sent he liberation.ACC his.DAT people.DAT ‘He sent liberation to his people through Christ.’ (Notker, Psalter , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer b. irbíat er ira gúati / mihilo ótmuati offered he her.DAT goodness.DAT great.ACC humility.ACC ‘He offered great humility in response to her goodness.’ (Otfrid I , )
() ther ánderemo thaz líb nam who other.DAT the.ACC life.ACC took ‘who took the life away from the other person’ (Otfrid V , ) In summary we can say: the OHG evidence suggests that in the OHG verb phrase the constituents were ordered on the basis of the grammatical criterion of case, that is dat > acc, and that informational-structural ordering principles such as ‘given before new information’ do not play a role.7 If this is true it shows that OHG does not perpetuate the putative PIE ‘free’ word order, which is thought to be for the most part an information-structurally governed constituent order, but that its word order is somewhat rigidified. This means that the Modern German word order freedom cannot go back to PIE times. A problem with the OHG evidence is that it is relatively sparse—apart from the fact that clauses in which the relevant pattern occurs are rare anyway, most OHG sources cannot be accepted at face value as Latin influence is evident there. The original texts and the clauses that differ from the Latin original are pointing in the same direction, however, towards a strong predominance of the dative before accusative word order as the unmarked option.
. Constituent order in Middle High German The problem with the extant MHG texts is the fact that prose sources are relatively rare.8 Beside several juridical texts, the main sources are two works from the time around , the Prosalancelot, a huge novel which, however, is translated rather faithfully from an Old French original (cf. e.g. Rothstein ), and the sermons attributed to Berthold von Regensburg. As the text of Berthold’s sermons was not available in digital form, it was not possible to use the same method as with the OHG texts. Instead, the traditional method of manual excerption was employed for both texts. It turns out that the vast majority of examples were governed by one of the verbs on the sample list given in (). This shows that the method of extracting the examples by using the sample list and the sample list itself are useful means of investigation. Considering the Berthold text, the impression we got from OHG that the word order is rather rigidly IO > DO is confirmed. The Prosalancelot shows a somewhat
7 See also Chapter on the role of information structure for word order. There it is shown that new information tends to appear after Vfin in embedded sentences (i.e. in the post-field). 8 A number of MHG prose texts have become available very recently in electronic form within the Referenzkorpus Mittelhochdeutsch (see also Chapter ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
greater variability, but notably in most cases in which the text deviates from the order IO > DO neither animacy nor givenness play a role. These cases are most probably translation lapses, as the Old French order of direct and indirect object was variable (cf. e.g. Einhorn ), but the modern option (DO > IO, the IO being a prepositional phrase headed by à) was already employed. On the basis of Berthold and Prosalancelot alone, however, it is hard to make any statements about the diatopical variation. If we assume that the Prosalancelot is, despite the fact that it is a translationary text, a faithful document of the MHG in use in the Rhine Franconian region where the translation was made, no difference to Berthold’s Bavarian is evident, so perhaps the serialization of objects was not subject to dialectal variation. From the examples it is clear that givenness did not play a role for serialization. There are numerous examples in which old-information indirect objects are placed before new-information direct objects (a), but also some examples in which new information precedes old information, following only the condition that indirect objects stand before direct objects (b). () a. daz dû dem almehtigen gote alle tage ein klôster that you [the almighty god].DAT all days [a monastery].ACC stiftest endow ‘that you always endow a monastery for the almighty God’ (Berthold I , –) b. daz man den heiligen engeln ouch eine hochgezît begê that one [the holy angels].DAT also [a feast].ACC keep ‘that one also keeps a feast day for the holy Angels’ (Berthold I , ) With respect to animacy, we see the same effect as in OHG: The order IO > DO does not contradict the animacy condition in any case. As we have more material available than for OHG, we can however say that the purely grammatical factor IO > DO (cf. factor (b) in Section ..) seems to play a role independently from animacy. This is interesting as it does not reflect the base order. There are several clauses in which both objects are either both animate or both inanimate, and in these cases the relative order is always IO > DO (). () a. dû hâst dem almehtigen gote dînen bruoder ermort you have [the almighty god].DAT [your brother].ACC killed ‘you killed your brother taking him away from God Almighty’ (Berthold I , –) b. alsô hat er ouch den sternen kraft gegeben even.so has he also [the stars].DAT power.ACC given ‘even so he gave power to the stars’ (Berthold I , ) In sum, we can say that in both OHG and MHG the usual object order was IO > DO, as it is in Modern German, although Modern German allows for more variation.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
Givenness does not play a role for serialization. Whether animacy is the decisive factor cannot be said for sure; by MHG times it is clear, however, that the grammatical factor ‘IOs stand before DOs’ is a factor in its own right. Definiteness, which is a very relevant factor in Modern German, does not seem to play the same central role in MHG as it does today (Speyer ). A survey of MHG texts from the Mittelhochdeutsche Begriffsdatenbank9 shows that neither morphologically defined definiteness (i.e. the presence/absence of a definite determiner) nor semantically defined definiteness (i.e. whether the uniqueness condition holds or not) has a stronger effect than the grammatical factor ‘IO > DO’. There are examples in which a definite accusative object appears to the left of an indefinite dative object (a)—in accordance with e.g. Lenerz ()—and there are examples in which the dative before accusative condition ousts the ‘definite before indefinite’ condition (b). But there are also examples—which would be ungrammatical according to Lenerz ()—in which an indefinite accusative object appears to the left of a definite dative object (c). () a. daz unser here got sin riche vil mængem that [our Lord God].NOM [his kingdom].ACC [very many sundær gegeben hat sinner].DAT given has ‘that our Lord God gave his kingdom to quite a few sinners’ (Priester Konrad ..) b. wie hie bevor ain heiliger wissage aim andern wissagen as here earlier [a holy prophet].NOM [an other prophet].DAT sin toehter ze wibe gæbe. his daughter].ACC to wife gave ‘as a holy prophet gave his daughter as a wife to another prophet before that’ (Priester Konrad ..) c. unde sagen wir lob und ere dem tot and say we [praise and honor].ACC [the death].DAT ‘and we say praise and honor to Death’ (Tepl, Ackermann , ) Scrambling of objects over subjects is rare. It is possible, however, notably in cases in which a trivalent verb is passivized. In these cases, the word order reflects the animacy principle (a). There is also sometimes scrambling of some focused element over the subject (b). () a. wan dem sünder . . . diu sünde niht vertragen wirt because [the sinner].DAT [the sin].NOM not pardoned becomes ‘because the sinner is not pardoned of his sin’ (Berthold I , )
9
For an overview of online corpora of historical German see also Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
b. daz sît anegenge der werlte ie die selbe sünde that since beginning of.the world ever [the same sin].ACC got griulichen gerochen hât God.NOM gruesomely revenged has ‘that God took gruesome revenge for the same sin since the beginning of the world’ (Berthold I , –)
. Constituent order in Early New High German .. Perpetuation of the rigid word order state For the next period, the ENHG period, there is a substantial number of original prose texts from different genres, from different dialect regions, and from different strata of the population. Thus this is the oldest period in which we can investigate diatopical, and to some extent diastratal, variation. The ENHG texts which were investigated are given in Table .. I used texts from the Bonner Frühneuhochdeutschkorpus. I combined the data from Speyer () and Speyer (), which cover three different time periods and four dialect areas, two Upper German (Bavarian and West Upper German, that is, Alsatian, Swabian, and
T . Texts used (Bonner Frühneuhochdeutschkorpus) –
–
–
Bavarian
Wilhelmus Durandus: Rationale (Vienna )
Helene Kottanerin: Denkwürdigkeiten (Vienna –)
Sigmund von Herberstein: Moscouia (Vienna )
Swabian
Buch von den Heiligen Altvätern (Reute near Wildberg, th century); Rulman Merswin: Buch von den zwei Mannen (Alsace, th century); Hans Mair: Troja (East Swabian, th century)
Hans Neidhart: Eunuchus des Terenz (Ulm ; only commentary)
Leonhart Rauwolf: Reisebeschreibung (Lauingen )
Ripuarian
Dat nuwe boych (Cologne –)
Johann Koelhoff: Cronica (Cologne )
Johann Gropper: Gegenwertigkeit (Cologne )
East Central German
Altdeutsche Predigten (Upper Saxon, th century); Psalter (prob. Erfurt, )
Johannes Rothe: Chronik (Thuringia, nd half of th century)
Johannes Mathesius: Passionale, (Leipzig ); Johannes Bange: Chronik (Mühlhausen )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
T . Relative order of IO and DO in ENHG texts
Bavarian West Upper German Ripuarian East Central German
IO > DO DO > IO IO > DO DO > IO IO > DO DO > IO IO > DO DO > IO
–
–
–
(%) – (%) (%) – (%)
(%) – (%) – (%) – (%) –
(%) (%) – (%) (%)
East Swabian combined), and two Central German (Ripuarian and Eastern Central German, that is Thuringian and Upper Saxon combined).10 We concentrate here on vP-scrambling. IP-scrambling remains rare. The method is the same as used in the OHG part. Although the data set is too small to make statistically valid statements, we can detect clear tendencies. The data are given in Table .. On the whole, IO > DO is more frequent in the periods before . Examples for all dialects in the second time period are given under (). The order appears even more frequent if one bears in mind that the text accounting for most DO > IO orders in West Upper German (Mair’s Troja: three IO > DO, five DO > IO) is written by a humanistic scholar (meaning that much Latin influence is to be expected) and that the text accounting for all DO > IO orders in the East Central German cell is a translation of the Book of Psalms (). () a. do nam man der Jungen KungInn, frawen Elyzabethen then took one [the young queen lady Elizabeth].DAT den swarczen rokch ab [the black garment].ACC away ‘Then they took the black cloak away from the young queen, lady Elizabeth’ (Kottanerin, Denkwürdigkeiten , )
10
The reason for the blend of Thuringian and Upper Saxon is that the representative text for Saxon in the fifteenth century is an edition of a text by Johannes Tauler, an Alsatian living in the fourteenth century and living in Straßburg, Cologne, and Basel, which happened to be printed in Leipzig. Although the printer might have shaped the text in some ways to render it more ‘Saxon’, it is questionable whether any redaction process would change, e.g. the constituent order. Therefore I decided to exclude Tauler. Instead I had to add Thuringian in order to have at least one representative for East Central German for the fifteenth century. The reason for the blend in the West Upper German area is that Speyer () concentrates on texts from this area; therefore, there are more representatives present.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
b. das ich sie ain zyt dem ritter guete wort that I them a while [the knight].DAT [good words].ACC liesse geben let give ‘that I allowed them to give nice words to the knight for a while’ (Neidhart, Eunuchus (Komm), ) c. Vñ Adã gaff allem gedirtzte vnd allen gaenden vnd kriechenden and Adam gave [all creatures and all walking and creeping dieren bequeme namẽ. animals].DAT [handy names].ACC ‘and Adam gave all creatures that walk and creep easy names’ (Koelhoff, Cronica r, ) d. nu gebit her Adam die hirschaft uf dem ertreich now gives he Adam.DAT [the power].ACC on the earth ‘now he gives the power over the earth to Adam’ (Rothe, Chronik , ) () Got gebet tugent sinen luten God gives virtue.ACC [his people].DAT ‘God gives virtue to his people’ (Psalter , ) Original: Dominus fortitudinem populo suo dabit Lord.NOM strength.ACC [people his].DAT give.FUT.SG ‘God will give virtue to his people’ (Vulgata, Ps. , ) In the third period, the picture changes, in that DO > IO occurs relatively more often and also in dialect regions where it did not occur in the earlier periods (still, IO > DO constitutes the main pattern overall). Examples are given under (). What happens in this period is that the condition to place given before new information seems to gain importance. () a. als woellte der selb die Khinder vñ das Land as wanted the same [the children and the land].ACC dem Khünig zu Polln vbergeben [the king to Poland].DAT surrender ‘as if he wanted to surrender the children and the land to the king of Poland’ (Herberstein, Moscouia v D, ) b. gab daß Land zu Hessen vnd Dueringen einem seiner gave [the land to Hessen and Thuringia].ACC [one of.his Freunde friends].DAT ‘he gave the country of Hessen and Thuringia to one of his friends’ (Bange, Chronik r, )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
The findings for OHG, MHG, and ENHG suggest that German underwent a period of relatively fixed word order, looking at the relative order of indirect object and direct object. The trend towards rigidification visible in the history of, for instance, English in the late OE and early ME period has thus a direct parallel in medieval German, without the consequent absolute rigidification that occurred in English, though. Note that it is not a development within the attested history of German, but that already OHG was a language with relatively firm order of the object constituents. The Modern German relative order of DO and IO, being susceptible to cognitive and information-structural factors, is not inherited from PIE, but a later development, pre-dating OHG in the case of animacy, occurring only after OHG and possibly MHG in the case of information structure. The impact of animacy was quite strong throughout the history: this factor even influenced the internal lexical– semantic predicate structure of the verbs, thus giving rise to the IO > DO condition. We will turn to this later. Further findings point in the same direction. Let me mention two: first, there are some verbs that have DO > IO as unmarked word order in Modern German, such as anpassen, ‘adjust’, ausliefern ‘surrender’, aussetzen ‘expose’, (weg)nehmen ‘take away’, unterordnen ‘subordinate’, etc. (Fortmann and Frey ; Hoberg ). Neither the OHG nor the ENHG corpus contained any examples of these verbs. In fact, such verbs are not attested in their three-place form earlier than the sixteenth century (see Speyer for a more detailed discussion). A possible interpretation, suggested by the findings so far, is that verbs with such a subcategorization frame could not be formed before the order DO > IO was available in the language at all. And second, in German, verbal idioms (so-called Funktionsverbgefüge) are quite common, such as einen Rat geben ‘give an advice’, Anteil nehmen ‘be concerned about someone’, auf die Spur kommen ‘get wind of someone’. They occur, however, only in the form direct object + verb or prepositional object + verb, never indirect object + verb. This gap can be easily explained if in earlier periods, it was only direct objects or prepositional objects that could be adjacent to the verb. Verbal idioms of this sort most certainly came into being by a grammaticalization process. In this case it is necessary that at some stage the two words undergoing grammaticalization are adjacent to each other. If indirect objects never show up in verbal idioms, this suggests that they never stood adjacent to the verb. This again is further evidence for a rigid order IO > DO on the surface, which also in this case can be attributed to the Thematic Hierarchy/Animacy constraint, the direct objects in verbal idioms being typically abstract and thus low on the animacy hierarchy. So we can say that German showed a relatively firm order IO > DO from the beginning of its attested history up to the fifteenth century. The PIE word order freedom was not visible in OHG. .. The rise of the ‘new’ word order freedom It is evident from the data in Table . that the order DO > IO started to occur with some frequency in the sixteenth century. It could be concluded from this that nonsyntactic factors started to play a bigger role in serialization at this time. Two kinds of
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
evidence would support this idea. The first kind of evidence would be that the attested deviations from the order IO > DO should be motivated information structurally or by any other weak constraint. The second kind of evidence would be examples that violate information-structural or cognitive serialization constraints from the time before the sixteenth century, i.e. the time in which the word order was rigid. We concentrate on factors (e) ‘animate referent before inanimate referent’ and (j) ‘given before new information’ (see Section ..). Evidence of the first kind is easy to come by. There is not one clause of the thirtyfour clauses from the sixteenth century that violates the information-structural ordering principle ‘given before new information’. This goes especially for the clauses in which the order is DO > IO. An example is given at (). In the context preceding the example, the concept ‘sun’ is salient, whereas the concept of ‘the whole earth’ was not mentioned before. So we have here an example in which a referent anchored to given information (the power of the sun) precedes a new referent. This is thus an instance of the ordering principle ‘given before new’, which has the power to outrank the principle ‘indirect before direct object’. () jedoch theilet sie jre Naturliche krafft vnnd however communicates she (= sun) [her natural power and eigenschafft dem gantzen erdreich mit. feature].ACC [the whole earth].DAT with ‘The sun, however, shows her natural power and substance to the whole earth’ (Gropper, Gegenwertigkeit r, ) There are also clauses in which both object referents are given. In these cases, the referent placed first is more salient than the referent placed second. An example is given in (), taken from a report on the Russian coronation ritual. Both referents, the grand-duke and the Barma, are given, as is evident from the context, the referent of the pronoun Jme ‘him’ being the grand-duke. The appositive relative clause forms an elaborative subdiscourse giving information about the Barma, using it as its topic. The Barma is salient in this subdiscourse, as opposed to the grand-duke, who is not mentioned. Turning to the actual example sentence, here the text returns to the level of the main discourse, as is indicated by the particle also. The Barma, having been salient throughout, is placed first, although it is the direct object. By being salient it has less communicative weight and should be placed first, following the Prague school concept of communicative dynamism (e.g. Firbas ), at any rate before the grand-duke, who has a higher communicative weight because he has to be reintroduced into the discourse. This is exactly what we find. () (context: Als solchs gebett vollendet / beualche der Metropolit zwayen when such prayer fulfilled ordered the metropolite two Abbten / das sy Jme das Barma raichten des sambt dem Huoet mit abbots that they him the Barma handed together.with the hat with ainer seidn (die selb sy SCHIRNIKOYV nennen.) bedegkht waren/) a silk the same they Shirnikov call covered were
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer gibt also das Barma dem Grooßfuersten gives thus [the Barma].ACC [the grand-duke].DAT ‘(When they have finished praying, the metropolite ordered two abbots that they handed him the Barma. This, together with the hat, was covered by a piece of silk which they call Shirnikov.) So he gives the Barma to the grand-duke.’ (Herberstein, Moscouia v D, )
Evidence of the second sort (earlier examples violating the ‘given > new’ condition) is less easy to find, but we have already seen some examples in the OHG section () and (). We also find examples in the ENHG period. In example (), the concept of sick persons (dem sichtigen) has not been introduced into the discourse. We might count this as new information, although the expression does not refer to an individual referent but rather generically to the referents’ set as a whole. The direct object, water (wazzer), does not refer to an individual referent either, being a mass noun, but the concept was salient in the preceding context, so it can count as old information. Yet the given referent, realized as direct object, is placed after the new referent, realized as indirect object. This is in accordance with the ordering principle ‘IO > DO’, it is also in accordance with the animacy ordering principle, but it violates the principle ‘given > new’. In this example, there is potentially a play on words (unsichtig – sichtig) which comes out more clearly by the word order chosen here; that the author can do this shows that the principle ‘given > new’ is sufficiently weak for him. () (context: Und daz wazzer ist auzgegangen von der rechten seitten, and the water is gone.out from the right side wenn von der goettleichen natur Christi get auz der because from the divine nature of.Christ goes out the heilig geist, der ist ein unsichtiges wazzer) Holy Spirit who is an invisible water und geit dem sichtigen wazzer and gives [the sick].DAT [water].ACC ‘(And the water has poured out from the right side. This is because the Holy Spirit emanates from the divine Nature of Christ, and he is an invisible water.) And he gives water to the sick person.’ (Durandus’ Rationale , ) Turning to animacy, it can be noted that the ‘given > new’ condition gains importance not only over the rigid word order but also over the animacy principle. Note that example () constitutes a violation of this principle. In the time before the fifteenth century, the animacy ordering principle plays a bigger role than the ‘given > new’ principle, but can be outranked by the grammatical ‘IO > DO’ factor. This can be illustrated by examples such as (). The referent of the indirect object lanten und lewten ‘country and people’ is on the whole more abstract and less animate than the referent of the direct object ainen Kung vnd erbherren ‘a king and heir’. Yet the less animate object is placed before the more animate object, thus satisfying the ‘IO > DO’ principle.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
() daz got der almoechtig lanten vnd lewten ainen Kung that God the almighty [land and people].DAT [a king vnd erbherren gegeben het. and heir].ACC given has ‘that God Almighty gave a king and heir to the country and its people’ (Kottanerin, Denkwürdigkeiten , )
. Reasons for the changes in word order .. Reasons for the less flexible Old High German word order At some stage between PIE and MHG the word order in German must have become less flexible. The question is when, and how? We can immediately rule out the phonological process of reduction of unstressed vowels to schwa (Nebensilbenabschwächung). As indicated in Chapter , this term denotes the MHG weakening of the vowels of unstressed syllables from a ‘full’ vowel in OHG, that is, a distinguishable vowel with clear highness, backness etc. features in OHG, viz. [i], [e], [ε], [a], [o], [u], to a non-distinct schwa-vowel [ə] in MHG. The vowels could also be reduced totally to zero. As case endings are always part of unstressed syllables, they were subject to this process. Whereas in OHG, information like declension class, case, and number could be read from the ending with its distinctive vowel(s), in MHG, the system of affixal case marking became subject to heavy syncretism. It has been argued with regard to the transition from OE to ME, where a similar process took place, that the loss of distinct case marking forced the development of compensatory strategies to ensure the correct coding of the grammatical relation a constituent has to its verb, and one of the strategies is the rigidification of word order (e.g. Kiparsky ). Rigidification is equivalent to waiving scrambling operations. Thus the rigidified word order reflects the base-generated word order. However, we saw that the OHG evidence already indicated that the word order was relatively fixed there, too. So this explanation cannot work because the case endings were still distinct, then. Another reason which would make this model implausible is that German developed a different strategy to cope with the loss of case endings, namely the development of the determiner as the locus for case marking. This process already began in the OHG period (that is, before the phonological weakening process set in) and is widely developed by the MHG period (e.g. Leiss ; Kraiss ). So the need for rigidification as a compensation for case marking would have been less urgent than in English, where the determiner was rather weakened and was not fit to adopt the function of a case marker. Finally, we would not expect the word order to become freer again, as the vowel weakening process went on in the history of German. After we have eliminated one notoriously promising candidate as a cause for rigidification, we have to dig deeper. The time at which rigidification occurred must lie before the beginning of the OHG period, that is, before the beginning of attestation of German. This means that it must be in the prehistory of the German language. For want of direct evidence, we can therefore only speculate about what happened.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
An interesting fact is that there is evidence that in PIE, the base-generated word order was S > DO > IO rather than IO > DO, and indeed still is in Modern German (see Section ..). In Modern German, we can test this structure directly with the binding test illustrated in Section ...While this test consideration tells us something about the underlying structure, it does not tell us much about the origin of this structure, although one might presume that this ‘hidden’ hierarchical structure must have originated at some stage11 and thus reflect an older stage in the language development. There is, however, evidence that in PIE, the unmarked (and by that potentially the structurally underlying) basic position of the verbal arguments was indeed DO > IO. Hittite is an old Indo-European language, among the earliest ones to be attested (the texts are from the second millennium BC). Hittite is different from the other ‘old’ Indo-European languages such as Greek, Sanskrit, Avestan, or Latin, in that its word order is more rigid than in the other languages. It has been shown (Luraghi : ) that the Hittite word order is DO > IO. Statements about other Indo-European languages are more difficult to make, as they show notoriously mobile word order, but a point could be made if binding data similar to the ones described for German could be found. If this is so, we have to ask how the basic structure changed, and in what way it changed. There are two possibilities: first, the basic structure changed from PIE DO > IO to Pre-German12 IO > DO to Modern German DO > IO. The way to prove or falsify this hypothesis would be to find examples in which the binding dependencies or any other available diagnostics (see e.g. Barss and Lasnik ; Larson ) clearly suggest a given underlying structure.13 Another, more speculative (and not really falsifiable) hypothesis would be that the change in the system is really a change in the importance of one of the ordering principles. The most promising candidate is the animacy principle, which was hinted at throughout the chapter. In PIE, the role of the animacy principle was probably not exorbitantly high, at least not as high as the ‘given before new’ principle (Krisch reports several information-structural reordering principles that are common to classic Indo-European languages). Almost all the OHG and all German examples up to the sixteenth century did obey the animacy principle, which looked like a side effect of the ‘IO before DO’ principle, simply because the available three-place predicates are set up in such a way that the dative is associated with actants that
11 Alternatively, at some stage the speakers of some predecessor language must have chosen the DO before IO strategy in the VP and the light verb construction suitable for it: to give could in theory be ‘to cause that something is with someone’—this is the DO before IO strategy—or ‘to cause that someone has something’—this is the IO before DO strategy (Rappaport and Levin ; Fortmann and Frey ). 12 I deliberately use this pre-theoretic term so as not to commit myself to a statement about whether the standard word order in Proto-West-Germanic or Proto-Germanic was also IO > DO. All this is meant to denote is that it is some stage of the language before the attestation of OHG sets in. 13 I have not found a relevant example so far. A research project on serialization with respect to scope and binding in the history of German and English is planned (together with Remus Gergel) which will hopefully shed light on this issue.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Serialization of full noun phrases
archetypically imply animacy (the recipient of a give-event, the source of a takeevent, the benefactive of a show-event, etc.), and the accusative with actants that are either indifferent with respect to animacy (the theme of a show-event etc.) or that are preferably inanimate (the patient of a give-event or take-event, etc.). Recent psycholinguistic research (Dowty ; Primus , ) identifies a universal principle of thematic hierarchy such that agent-like roles (Proto-Agent in Dowty’s and Primus’s terms) are ordered before patient-like roles (= Proto-Patient). The ProtoRecipient combines properties of Proto-Agent and Proto-Patient and is serialized inbetween them. The assignment of case to these roles follows a slightly different hierarchy: the Proto-Agent and Proto-Patient are assigned the most prominent (and usually structural) cases (in a non-ergative language with more than two cases, this would be the nominative and accusative), the other roles are assigned case following the thematic hierarchy, mapping them to the same case hierarchy. In languages with more than two cases, the next case on the hierarchy would be the dative, which is assigned the Proto-Recipient role as this is, apart from the Agent, still the most agentive role available. Animacy is an important cue for the language learner to identify some actant as agent-like (Primus ). If the animacy principle, as reflex of the thematic hierarchy, gained importance as a serialization strategy, most clauses containing three-place predicates would have to be realized with the dative before the accusative, simply because the referent that is associated with the dative is animate, whereas the referent that is associated with the accusative is not animate. This could cause an overgeneralization in the brains of language learners, namely that it is the dative that stands first and the accusative that follows it. The evidence was not strong enough to cause the language learner to assume a syntactic structure in which the dative was hierarchically higher than the accusative (and thus would be linearized first)—possibly by binding dependencies like the abovementioned—so that the language learner instead assumed a general scrambling operation of the dative to a position higher than the accusative. The assumption of a general scrambling operation is not as outlandish as it looks at first glance, since exactly such generalized movement operations are visible in the process of the clustering of pronouns in the ‘Wackernagel position’ (see Chapter ). So this hypothesis is not unattractive, but it has to remain speculative. If it turned out by compelling evidence that the first hypothesis (two changes in the underlying structure) is wrong, this second hypothesis (no change in underlying structure, but generalized scrambling due to the thematic hierarchy) would be a good candidate for an explanation model. .. Reasons for the ‘liberation’ of word order in Early New High German It remains unclear why the word order became freer in the time after . One possibility is that it has something to do with the emerging prose style, which in turn was heavily influenced by Latin. In Latin, the relative order of direct and indirect object was decidedly not rigid. Moreover, it depended on the givenness principle, which did not play a role for German serialization up to that point, but has become a prominent principle in Modern German (Lenerz ; Musan ). So this would be an example of a change that has external factors. It is also difficult to identify any
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
internal factors that were responsible for this change as the system seems relatively stable up to that point. At this point, we have to ask to which area of grammar the change pertains at all, and how to describe it. It is basically a change in the availability of scrambling in order to produce an information-structurally motivated constituent order. This means that it is definitely not a deep structural change, but rather a change in the frequency of employing a given optional operation, namely scrambling. We may assume that vP-scrambling was possible all the time, otherwise we should not get any variation in the object order at all, but that it was regarded as very marked and therefore usually not used. Such a scenario is modelled best by an optimality-theoretic model (see e.g. Müller ; Heck ). In such a model, German before could be said to have had high ranking for the constraint that selects a candidate in which an animate referent precedes an inanimate referent (ANIM in Heck ) and the constraint that selects a candidate with the order IO > DO (let us call it THEMATICHIERARCHY, following the considerations in Primus ). Both of these constraints outranked the constraint that selects a candidate in which given information precedes new information (let us called it GIVENNESS). After , the ranking value of the constraint GIVENNESS rises and is ultimately higher than ANIM and THEMATICHIERARCHY, so that violations of ANIM and THEMATICHIERARCHY can occur as long as the candidate obeys GIVENNESS. So the change concerns only the surface. This, however, does not mean that such a change in mere frequency cannot induce a change on a deeper level. Given that the order DO > IO is an available option in the language, the animacy condition can free itself, so to speak, from the grammatical ‘IO > DO’ condition, and all of a sudden verbs like ausliefern ‘surrender someone to something’, or aussetzen ‘expose someone to something’ can emerge that obey in their normal order the animacy condition, but construct with the DO before the IO.
. Summary The relative freedom in constituent order in Modern German does not date back to prehistoric times but is indeed quite a recent phenomenon, dating back to the time around . Before that, the relative order of direct and indirect object in German was predominantly IO > DO. Givenness or other information-structural factors played no role for serialization. The order IO > DO obeys the animacy principle as a rule, which indicates that the verb valency was reorganized according to the agentprominent thematic hierarchy, of which animacy is a side-effect, at some stage preceding the attested history of OHG.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
10 History of negation in High and Low German AN N E BREI TB ARTH A N D A GN ES JÄ GE R
Negation as a propositional operator is usually assumed to take scope above the VP but below the C-domain and under many approaches is in some form syntactically represented here (e.g. Pollock ; Ouhalla ; Haegeman ; Zanuttini ; Weiß a; Zeijlstra ).1 The place of negation in terms of the topological model of the German clause is thus in the middle field.2 This is indeed where the Modern German negation particle nicht is found. Despite the diachronic development of the negation particles and the surface-syntactic change in their position, we will argue that the negation particles (as well as any negative indefinites) in fact always originate in the middle field. Thus, negation can be regarded a German middle field phenomenon, although, due to syntactic movement, markers of negation may also occur at the peripheries of the clause—especially historically because of the different syntactic properties of the negation particle, as will be discussed in this chapter. In many languages, especially in (north-)western Europe (Willis, Lucas, and Breitbarth ), the expression of sentential negation underwent a historical development known as Jespersen’s Cycle (Dahl ), by which the original marker of standard sentential negation is first joined and later replaced by a newly grammaticalized marker.3 Both High and Low German have in their histories undergone this development. The first part of the present chapter discusses and compares the
1 Usually between TP and VP/vP; see Weiß (a) for a distinction between what he calls Neg-TP languages (like Italian) and Neg-VP languages (like Bavarian). See also Zeijlstra (: –) for more discussion and references. 2 For an overview of fields and positions assumed within the topological model of the German clause see Chapter . 3 After Jespersen’s () original observation that ‘[t]he history of negative expressions in various languages makes us witness the following curious fluctuation; the original negative adverb is first weakened, then found insufficient and therefore strengthened, generally through some additional word, and in its turn may be felt as the negative proper and may then in course of time be subject to the same development as the original word’ (Jespersen : ).
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Anne Breitbarth and Agnes Jäger . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
evidence for—and the time frame of—these changes in detail. Besides, the way indefinites in the scope of negation interact with the different negative markers (and with each other) changes over time, which is discussed in the second part of this chapter. In particular, the availability and type of Negative Concord, that is, the presence of more than one morphological expression of negation within one sentence while only one semantic negation is present, changes diachronically in High and Low German. Although these developments, Jespersen’s Cycle and the changes in the interaction of indefinites and sentential negation, are largely comparable in High and Low German, they do not always occur at comparable times, at the same speed, or quite in the same way. It is the goal of this chapter to elucidate the details of these developments and changes and their differences between High and Low German, and to offer a formal analysis of them.4
. The negation particle In many languages, including German, the negation particle (neg-particle) used to mark sentential negation changes diachronically according to what has come to be known as Jespersen’s Cycle, given in (). Although presentations differ regarding the number of stages posited (e.g. van Kemenade ; Zeijlstra ; van der Auwera ), most agree that at least three stages can be identified. At a first stage, the negparticle is a verbal clitic. Towards the second stage, some element is grammaticalized into a second, free neg-particle so that negation is expressed by a bipartite negparticle formed by the old and the new particle. Eventually, the original clitic negparticle is dropped so that, at the third stage, the free neg-particle remains as the marker of negation. This neg-particle may over time be reduced to a clitic, returning the language to the beginning of the cycle. ()
Jespersen’s Cycle: • stage I: clitic neg-particle • stage II: clitic and free neg-particle • stage III: free neg-particle
4 This chapter is based on the following corpora: OHG: all negated clauses from Isidor (around AD) and Tatian (before AD), the first negated clauses from Otfrid (– AD), and Notker’s Psalter (before AD), all negated clauses from minor OHG texts (Hildebrandslied, Wessobrunner Predigt, Pariser Gespräche); OLG: all negated clauses from Heliand (c. ), the Genesis fragments (c. ) as well as from a collection of minor OLG texts (verse, ecclesiastic and secular functional prose, glosses, end of ninth– eleventh century); MHG: first negated clauses from Prosalancelot (before ), Nibelungenlied (ms. A: before , with comparison to manuscripts B and C: second quarter of the thirteenth century—dating cf. Schulze ) and the sermons by Berthold von Regensburg (approx. ), for specific research questions Wiener Genesis (approx. – AD); MLG: all negated clauses from charters, legal texts, and similar dated and localized documents from ten cities, dioceses, and convents (Barsinghausen, Börstel, Braunschweig, Lübeck, Mariengarten, Oldenburg, Scharnebeck, Steinfurt, Stralsund, and Uelzen) from to . Note that, despite what the terminology might suggest, MLG is only partly contemporary to MHG, but overlaps partially with the ENHG period (see also Chapter ).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
.. Old High German ... The preverbal clitic negation particle In OHG, sentential negation was virtually always marked by the preverbal clitic neg-particle ni, inherited from Germanic *ni, cf. (). () sí ni mohta inbéran sin she NEG could do.without him ‘She could not do without him’ (Otfrid I , )
V
In late OHG, the vowel was reduced and the neg-particle occurs as ne. Table . gives the ratio of clauses containing the preverbal clitic neg-particle among all OHG negated clauses in the corpus. Ni is clearly the major means of marking negation in OHG, with percentages around ninety and more of all negated clauses. This result falsifies Paardekooper’s () claim that ni never had the monopoly in historical German. In most cases, the preverbal clitic neg-particle is even the only marker of negation in the respective clause. This clearly falsifies Abraham’s (: f) ‘Asymmetrical Neg-Criterion’ according to which ni could supposedly only appear if a phrasal neg-marker, i.e. a negative indefinite or a second neg-particle was present. The fact that ni was the major means of marking negation is also evident from the OHG texts that represent translations; where these deviate from the Latin originals with regard to negation, it is mostly in adding the preverbal neg-particle where the Latin clause did not contain the neg-particle non but for instance only a negative indefinite. Ni and the finite verb are often written as one word, i.e. the scribes perceived of it as a unit, which corroborates the status of ni as a clitic.5 Ni occurs basically always as a proclitic to the finite verb in Isidor, Tatian, Otfrid, and Notker.6 This observation is also supported by the data from various minor OHG texts (all negated clauses in Hildebrandslied, Wessobrunner Predigt, and Pariser Gespräche). In OHG infinitival
T . Ratio of clauses including ni (or ne) in OHG negated clauses
negated clauses including negation particle ni ni as only neg-marker in the clause
Isidor
Tatian
Otfrid
Notker
% () % ()
% () % ()
% () % ()
% () % ()
5 Spelling as one word is attested with various initial vowels and consonants, and not, as according to Wheelock LaBrum (: ), only with initial /i/. This graphematic fact is, however, often obscured in the classic nineteenth-century editions of OHG texts (e.g. Tatian ed. Sievers in contrast to ed. Masser , Otfrid ed. Kelle in contrast to ed. Kleiber ). Therefore, it is imperative to use the modern editions or original manuscripts. 6 There are very few exceptions to this rule, e.g. ni occasionally occurs as a proclitic on a participle in loan-syntactic participle constructions in Tatian.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
constructions, too, it is generally the finite verb that ni attaches to, compare example (). The procliticization of ni is not restricted by any morphosyntactic specifications of the verb—even negated imperatives are attested, as illustrated in (). () Ni fórhti thir bíscof! NEG fear thee bishop ‘Do not be afraid, bishop!’ (Otfrid I , )
V
Furthermore, ni attaches to the finite verb irrespective of verb placement, compare the V clause in (), the V clause in (), and the verb-final (Ve) clause in ().7 () thaz thu irrímen ni máht. that you tell NEG may ‘that you cannot tell’ (Otfrid I , )
Ve
OHG is thus clearly at stage I of Jespersen’s Cycle. However, in late OHG we witness the beginning grammaticalization of a second neg-particle. ... ‘Negation strengtheners’ and the beginning grammaticalization of a verbindependent second neg-particle There are several cross-linguistically relevant types of emphatic ‘neg-strengtheners’ that form the diachronic source of a second, verbindependent negation particle: negative indefinites (e.g. ‘nothing’), non-negative indefinites (e.g. ‘thing’, ‘being’), minimizers (e.g. ‘jot’, ‘drop’, ‘crumb’), and generalizers (e.g. ‘in this world’), cf. Breitbarth, Lucas, and Willis (). Several of these types are evidenced in OHG. The first type, i.e. a negative indefinite being used as an emphatic ‘negstrengtheners’, is attested in the adverbial, non-argument use of late OHG nieht (< niowiht ‘nothing’) meaning ‘in nothing’/‘not at all’.8 This occurs very occasionally in Notker (in four out of negated clauses), for example in (). () Ih nehábo / niêht in geméitun sô uîlo geuuêinot. I NEG.have not at all/NEG in vain so much cried ‘I did not cry that much in vain.’ (Notker, Psalter , ) 7 In OHG (unlike ON cf. Eythórsson , , and OE cf. van Kemenade ), there is no evidence for a greater amount of V word order (ni-Vfin-Subj) in negated clauses as occasionally suggested in the literature (most recently Axel : , but without giving any concrete figures): according to our corpus data, only Otfrid has about twice as much negated V as V. Yet this text generally has more V, also in non-negated clauses. In Isidor and Tatian, there are three to four times as many negated V as negated V clauses (excluding questions and imperatives, which generally show V), in Notker even twenty times as many, i.e. most OHG declaratives are V clauses, whether negated or not. 8 The first attestation of adverbial niowiht has been argued to be one example from Otfrid: ni zawêta imo es niawiht ‘He did not succeed {at any of it/at it at all}.’ (Otfrid II , ) (cf. Behaghel : ; Wheelock LaBrum : ). However, Donhauser (: ) points out that the valency of the verb zâwen, which is only attested in Otfrid, is not entirely clear so it cannot be excluded that niawiht is an argument here, as in all other cases in Otfrid. Furthermore, our data show that Otfrid uses uuiht or drof rather than niowiht as an adverbial ‘neg-strengthener’.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
Indeed, this original negative indefinite meaning ‘nothing’ is eventually grammaticalized as a second neg-particle in German, compare the parallel development of English nought/nawiht > not, Dutch niet > niet, ON eittki > ekki, Danish/Norwegian ikke, Swedish icke, Danish/Swedish inte, Greek oudén > dén. Occasionally, another negative indefinite appears as a ‘neg-strengthener’ in OHG, viz. the negative indefinite adverb nio (in altere), which otherwise means ‘never’ but is rarely used semantically bleached, i.e. without its temporal meaning and thus approaching an adverbial neg-particle.9 This is illustrated in (). () (nequaquam minima és / In principibus Iuda.) nio In altere bist thu minnista / In then heriston iudeno. never in age/NEG are you smallest in the princes Jewish ‘thou art not the least among the princes of Juda’ (Tatian , –) The second type, i.e. a non-negative indefinite being used as an emphasizer is evidenced in OHG in the use of the non-negative equivalent of niowiht/nieht, viz. the (NPI) pronoun wiht ‘(any)thing’ in an adverbial way meaning ‘at all’. This is found in Otfrid in seven out of negated clauses, for instance in the example in ().10 () thaz ér mir hiar ni dérre, ouh uuíht mih ni gimérre. that he me here NEG let.wither also at all/NEG me NEG obstruct ‘that he will not let me wither here and not obstruct me at all’ (Otfrid I , ) While its neg-marked counterpart eventually won out (cf. Modern German nicht), wiht was also partly grammaticalized as a verb-independent neg-particle: The use of a non-neg-marked neg-particle partly continued into MHG and even into certain present-day dialects of German, cf. () below. Finally, there is also evidence for a minimizer or minimal unit expression as a ‘neg-strengthener’ in OHG. In two out of negated clauses, Otfrid uses the noun drof ‘drop’ in this way entirely parallel to his use of wiht, for instance in ().11 () drof ni forahtet ir iu drop/NEG NEG fear you you ‘Do not be afraid at all.’ (Otfrid III , ) It is conceivable that the use of this particular minimizer is due to French influence, although minimizers are generally a common phenomenon in historical
9 This type of adverbial negator from an indefinite adverb meaning ‘never’ is also attested in the history of English, cf. van Kemenade (, ) and Lucas and Willis () and references cited therein. 10 According to Kelle (: –), there are eighteen examples of adverbial wiht in the entire text of Otfrid. However, according to our corpus results, book one contains more occurrences than are listed by Kelle, so the total number is presumably higher. 11 Kelle (: ) notes eighteen occurences of ‘neg-strengthening’ drof in the entire text—the same number that he gives for wiht as a neg-strengthener.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
German, cf. Zingerle (), as well as Modern German dialects.12 Otfrid’s monastery Weißenburg (now Wissembourg) in Lorraine was not far from the linguistic frontier. Minimizers as emphatic ‘neg-strengtheners’ are obviously well known from the history of French. The Modern French pas originated in this way, and there is also evidence for an Old French ‘neg-strengthener’ goutte ‘drop’ (cf. Catalani ). .. Old Low German ... The preverbal clitic negation particle As in OHG, the major means of marking negation in OLG is the preverbal neg-particle ni/ne inherited from Germanic, cf. (). () ‘ni bium ic’, quađ he, ‘that barn godes . . . ’ NEG am I spoke he the child God.GEN ‘I am not the child of God, he said’ (Heliand )
V
Table . illustrates the ratio of clauses containing this neg-particle among all negated OLG clauses in our corpus. More than % of all negated clauses in our OLG corpus contain this negmarker.13 The picture is thus even clearer than in OHG. In OLG, too, preverbal ni/ne is the only marker of negation in most negated clauses, with percentages comparable to those given for OHG in Table .. OLG ni/ne also always procliticizes to the finite verb irrespective of verb placement and thus occurs together with the verb in initial position in V clauses cf. (), in final position in verb-final (Ve) clauses cf. () and in second position in V clauses cf. ().14 T . Ratio of clauses including ni or ne in OLG negated clauses Heliand
Genesis
minor OLG texts
negated clauses including negation particle ni/ne
.% ()
% ()
% ()
ni/ne as only neg-marker in the clause
.% ()
% ()
% ()
12 Cf. also two occurrences of netrophen (‘NEG-drop’, as translations of Latin nihil ‘nothing’ and quid ‘anything’ in the scope of negation) in the heavily Romance-influenced OHG Pariser Gespräche (sentences and ). Grimm (: /), however, implies that the influence went in the opposite direction and that the French ‘neg-strengtheners’ mie as well as goutte were inspired by German equivalents. Already Zingerle () reports numerous minimizers in MHG. 13 The only exception are some negative conjuncts to negative clauses introduced by the disjunction ni, ne ‘and not, nor’, where the negation marker can be omitted before the finite verb. But even in this type of clause, the preverbal negator is still used in the majority of cases: only in five out of seventeen ni/neconjuncts is there no additional preverbal negator ni/ne. 14 While verb-final word order prevails on the whole in our OLG corpus, V word order is the preferred word order in negative main clause declaratives in the Heliand and also in the minor OLG texts, suggesting that the preference for V word order in declaratives reported for ON and OE can be found in OLG, too, in contrast to OHG.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German () ‘ic thes uuirđig ne bium,’ quađ he, / ‘[ . . . ] that thu I the.GEN worthy NEG am said he that you an min hus cumes to my house come ‘I am not worthy . . . , said he, that you come to my house’ (Heliand –) () that nis thi allaro frumono negên that NEG.is you all benefits none ‘that is no accomplishment to your credit’ (Heliand )
Ve
V
Just as in OHG, OLG can therefore safely be classed as a stage I language in terms of Jespersen’s Cycle. ... ‘Negation strengtheners’ In OLG we find evidence for different types of emphatic ‘neg-strengtheners’, too. The most common type of emphasizer are generalizers like ‘in this world’ or ‘in his life’, as in () (twenty-four occurrences in the Heliand and two in the Genesis fragments, that is, .% of all negated clauses in the OLG corpus). () That ni scal an is liƀa gio liđes anbîtan, / uuînes an is uueroldi. that NEG shall to his life ever cider enjoy wine to this world ‘Never in his life will he drink hard cider or wine in this world.’ (Heliand –) The emphasizer type of non-negative indefinite is also attested. In our OLG corpus, there are twenty-two (%) negated clauses containing uuiht with a genitive noun or pronoun, cf. (), eight (.%) containing adverbially used mid uuihti (lit. ‘with anything’) meaning ‘at all, in the least’, cf. (), and another eight (.%) containing pseudo-argumentally (cf. Breitbarth, Lucas, and Willis ), or adverbially used ((n)io)uuiht () and (). () Ne sculun gi ênigumu manne unrehtes uuiht, / derƀie adêlean, NEG shall you any man injustice.GEN anything hostile give ‘you should never pronounce even the slightest detrimental, unjust judgement on any man’ (Heliand –) () That ic an mînumu hugi ni gidar / uuendean mid uuihti, that I in my mind NEG dare change with anything of ic is giuualdan môt. if I it effect could ‘Even if I had the power, I would not dare even to think about changing that in the slightest.’ (Heliand –) In cases like (), uuiht is not originally adverbial in function but represents the head noun with an attributive genitive NP. However, uuiht could at least in some
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
cases be analysed as an adverbial neg-strengthener while the original attributive NP constitutes the argument of the verb. This is particularly likely in cases like (), where the genitive element is a pronoun, and occurs separated from its alleged head noun uuiht. In this case, the pronoun is more likely to be the argument of the verb antkennian ‘to recognize’ than is . . . uuiht ‘anything of him’.15 () sô is thea ni mahtun antkennian uuiht, / so he.GEN those NEG could recognize anything/NEG the thes uuîhes thar uuardon scoldun who the.GEN shrine.GEN there guard should ‘they did not recognize him at all who were supposed to guard the shrine’ (Not: ‘they did not recognize anything of him . . . ’) (Heliand –) Furthermore, this type of construction is the most frequent pattern of emphasizing negation in our OLG corpus (.% of all negative clauses with an emphasizer of some sort), only considering cases like (), where an analysis of uuiht’s complement as internal argument is possible and likely. Emphatic reinforcement with simple (io)uuiht, as was found in OHG, is much less frequent; it only occurs eight times in our OLG corpus (.% of all negative clauses with an emphasizer, .% overall), for instance in ().16 () ni sculun ûs belgan uuiht, NEG shall REFL anger anything ‘We shall not be angry at all’ (Heliand ) Such ‘neg-strengthening’ uses are also attested with niouuiht, cf. (), albeit even less frequently than (io)uuiht, only three times (.%) in our corpus.17 () ‘ne ik thi geth ni deriu neouuiht’18 quađ he, and.not I you also NEG harm (no)thing said he ‘I will also not harm you at all, either’ (Heliand ) 15 It remains an empirical question whether OLG actually had a genitive of negation, that is, a regular or at least optional alternation between accusative and genitive under negation more generally. It is definitely common to find the genitive after nicht in MHG (cf. Dal : ) and MLG, presumably a last remainder of the nominal past of the new neg-particle. For arguments that genitive alternated regularly with accusative in OHG to express indefiniteness of the noun, imperfective aspect of the verb, or partitive, see Donhauser (). This might be relevant for an assessment of the case alternation under negation. 16 Cf. Breitbarth, Lucas, and Willis () on cross-linguistically common ‘bridging contexts’ for negation strengtheners, such as verbs of caring/indifference, or damaging, in which the strengtheners initially appear as pseudo-arguments expressing an extent. 17 Cf. also the extremely frequent use of nichts ‘nothing’ in present-day (High) German with the same verb as in (), schaden ‘damage, harm’ (though with certain person restrictions):
(i) Das wird dir nichts schaden. that will you nothing harm ‘That won't harm you (at all).’ lit. ‘That will not harm you anything.’ 18
The text of the example conforms to manuscript M. In manuscript C, the form is not neouuiht, but uuiht: Ne ik thi geth [ni] deriu uuiht’ quathie.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
In spite of its low frequency, this item does in fact later grammaticalize as the second neg-particle in Low German, as it did in High German, and is firmly established in this function in the form of nicht in our MLG corpus (see Section ...). An attestation gap of about two hundred years at the decisive stage of the grammaticalization process unfortunately makes it impossible to trace the further development of niouuiht in OLG. As the earliest attestations of adverbially used ((n)io)uuiht are found in the context of predicates with an ambiguous argument structure, in which the pronoun could be interpreted as an optional extent argument (Bayer ; Breitbarth, Lucas, and Willis ), we can be fairly sure that such predicates formed the bridging context for the grammaticalization of the MLG neg-particle nicht ‘not’, see also Breitbarth (). .. Middle High German ... The preverbal clitic negation particle Compared to the situation in OHG, the preverbal clitic neg-particle en/ne is much rarer in our corpus during the ensuing period of MHG, as Table . illustrates.19 One can also observe a vast diachronic drop throughout MHG from less than half of all negated clauses containing en/ne in Prosalancelot, to about a quarter in Nibelungenlied and less than a tenth in Berthold’s sermons. Over the course of the MHG period, the preverbal clitic neg-particle is clearly on its way out. Nonetheless, there are very rare remains of en/ne occurring as the only marker of negation in the clause, i.e. of Jespersen stage I.20 One factor that plays a role here is verbal class:21 As observed by Behaghel (: ); Paul (: f) and others, modals as well as lâzen ‘to let/leave’, tun ‘to do’, ruochen ‘to care’, and wizzen ‘to know’22 tend to allow T . Ratio of clauses including en/ne in MHG negated clauses
negated clauses including negation particle en/ne en/ne as only neg-marker in the clause
Prosalancelot
Nibelungenlied (A)
Berthold
% ()
% ()
% ()
% ()
% ()
% ()
19 The numbers given for Nibelungenlied pertain to manuscript A. In manuscripts B and C, the percentages are distinctly higher with % and %, respectively, of all negated clauses containing the neg-particle en/ne. 20 This fact falsifies once more Abraham’s (: f) Asymmetrical Neg-Criterion, which he formulated for OHG and MHG. 21 On the influence of phonological factors see Szczepaniak (). According to her hypothesis, a typological change regarding the isochrony of German from a syllable-timed language to a word-based (stress-timed) language was responsible for the loss of unstressed en/ne. Similarly, Noel () speaks of a conspiracy of syntactic (strengthening of V constraint to the disadvantage of older V-negative declaratives; cf. also Eythórsson and Chapter this volume) and prosodic factors (as a monosyllabic function word preceding in most cases a stressed syllable, ne cannot usually attract rhythmic or secondary stress) leading to its deletion. 22 Jespersen (: ) observes that the clitic negation is kept longer with the verbs ‘to know’ and ‘to want’ in several languages. He mentions Latin nolo and ne scio, French je ne sais, English nill, and also MHG en will.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
negation by simple en/ne for longer, consider the example in () with a form of wizzen. () sô hât er dirz gestoln, dû enweist hiute wie so has he you.it stolen you NEG.know today how ‘He has stolen it from you and you do not know today how.’ (Berthold I , –) Another decisive factor is verb placement.23 Thus, in our MHG corpus, en/ne never suffices to mark negation in verb-final clauses. On the other hand, V clauses of a special type, viz. exceptive clauses with V word order conveying the meaning of ‘unless . . . ’ as in () are typically negated by simple en/ne. () miern zerinne miner frivnde, in wirt arebeit bechant. me.NEG vanish.SBJV my friends them becomes trouble known ‘Unless my friends should vanish, they shall get to know some trouble.’ (Nibelungenlied (A) , ) However, we are arguably not dealing with genuine cases of sentential negation here, as exceptive clauses are distinguished by a number of formal properties from negative conditionals. For instance, they invariably show V order, while regular conditionals are either asyndetic with V order or syndetic with verb-final order, and the finite verb is invariably in subjunctive mood. Also semantically, true (sententialscope) negation seems to be absent from exceptive clauses: They never contain any post-verbal neg-particles (niht) or n-indefinites, even though either would be compatible with the preverbal marker in a regular negative clause.24 The latter two properties cast doubt on en/ne standing for sentential negation in this context. A beginning reanalysis of en/ne as a marker of affective polarity25 rather than strictly of negation (cf. Breitbarth ) would account for the fact that in MHG unlike in OHG, en/ne furthermore occurs in subordinate clauses that depend on an adversative or negated matrix predicate, cf. (). This phenomenon, also known as expletive negation or, in the terminology of Jespersen (), as paratactic negation, is also found in other languages, e.g. French (Je crains qu’il ne vienne ‘I fear he will come’). Yet, this kind of construction is extremely infrequent in our MHG corpus.26
23 This is corroborated by the findings of Schüler () who, for a corpus of Central German charters from the MHG period, reports a higher rate of en/ne in V and verb-final clauses, but a lower rate in V clauses. Furthermore, it is much rarer with prefixed verbs. Both observations support the view that the use of en/ne in MHG is subject also to phonological constraints. 24 Note that the corresponding English exceptives with unless do not license negative polarity items (NPIs) such as any, either (Breitbarth : ). One could argue that the entire construction (dependent subjunctive V clauses with a single preverbal en/ne) encodes the function of an exceptive complementizer, as Wallmeier (: ) suggests. For a proposal regarding this construction in MLG see Breitbarth (: –, a). 25 Affective is used as a cover term (cf. Klima ) for all contexts that license NPIs such as English any, viz. negation but also conditionals, the standard of comparison, questions, etc. 26 In the MHG corpus, there are only two examples (both from Nibelungenlied), i.e. less than one per cent in the entire MHG corpus.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
() die hoh gemvoten degne wolden des niht lan, / the high spirited warriors wanted that NEG omit sin drvngen da si sahen die minnechlichen meit. they.NEG penetrated there they saw the lovely maiden ‘The bold warriors did not want to refrain from forcing their way through to where they saw the lovely maiden.’ (Nibelungenlied (A) , –)27 Note incidentally that, in examples () and (), (e)n encliticizes to the element in front of the finite verb rather than procliticizing to the finite verb itself. This only occasionally occurs in the MHG corpus, notably with (personal) pronouns, adverbs, or particles such as so, do, ia, or nu, and crucially only in V clauses. One can therefore assume that this encliticization is only a secondary phonological process which optionally takes place once the finite verb has moved the clitic neg-particle along with it to the left periphery of the sentence. Syntactically, ne/en is arguably still linked to the finite verb; the decisive generalization is that the clitic neg-particle in MHG is always positioned immediately left-adjacent to the finite verb. ... The bipartite negation particle In MHG, the preverbal neg-particle also cooccurs with the second, verb-independent neg-particle niht, which had started to be grammaticalized in late OHG from the adverbially used negative indefinite niowiht/ nieht ‘nothing’, ‘in nothing’/‘not at all’, as mentioned above.28 There is thus evidence for Jespersen stage II in MHG, cf. () and (). () daz ich drîzic pfunt niht ennaeme that I thirty pound NEG NEG.take ‘that I would not take thirty pound’ (Berthold I , –)
Ve
() ‘Ich enwil es niht erwinden’, sprach aber der chune man. V I NEG.want it NEG omit said but the brave man ‘ “I do not want to omit it”, said the brave man.’ (Nibelungenlied (C) , )29 This bipartite negation particle is generally described as the standard in MHG, consider the following quote from Wolf (: ): ‘Im Mittelhochdeutschen ist die doppelte Negation ne + niht geradezu die Norm’ (‘In MHG, the double neg-particle ne + niht is really the norm’, cf. also the accounts given in Dal ; Schmidt ; Paul ). However, our quantitative corpus analysis reveals that in the investigated MHG corpus, only a minority of negated clauses contains both particles, cf. Table ..30 None of the MHG texts in the corpus predominantly employs this pattern to express sentential negation. 27
B/C: . . . sine drvngen . . . Even in MHG, niht continues to be used also as an n-indefinite meaning ‘nothing’. 29 A/B: ‘Ich ne wils niht erwinden’, sprach der kuone man. 30 Regarding Nibelungenlied, the rate is again significantly higher in the older manuscripts (B: %, C: % of negated clauses containing the bipartite neg-particle), but still lower than in Prosalancelot, thus illustrating further the diachronic decrease in the use of the bipartite neg-particle within MHG. 28
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Bipartite neg-particle en/ne . . . niht in MHG negated clauses
en/ne . . . niht
Prosalancelot
Nibelungenlied (A)
Berthold
% ()
% ()
% ()
T . Ratio of neg-particle niht in MHG negated clauses Prosalancelot
Nibelungenlied (A)
Berthold
negated clauses incl. neg-particle niht
% ()
% ()
% ()
niht as only neg-marker in the clause
% ()
% ()
% ()
From our data, there is no evidence for a stable Jespersen stage II period in the history of German.31 Even the transitional period between OHG and MHG does not constitute a Jespersen stage II system as an additional corpus analysis of the very early MHG Wiener Genesis (approx. – AD) indicates. This text still predominantly displays stage I of Jespersen’s Cycle; the preverbal clitic neg-particle en/ne is used on its own five times as often as either the bipartite neg-particle ne . . . niht or niht on its own. In classical MHG, on the other hand, ne/en is already disappearing as our core MHG corpus shows. In Upper German, ne/en is virtually extinct by the end of the thirteenth century, yet it continues to be very rarely attested in parts of the Central German area until around (cf. Pensel ). ... The verb-independent negation particle The new verb-independent negparticle niht is very widely used in MHG, cf. Table .. Furthermore, it already occurs mostly on its own. Comparing Table . with Tables . and ., it becomes evident that the Jespersen stage III pattern with just the verb-independent neg-particle niht is more common than the stage I pattern, but also—unexpectedly—more common than the stage II pattern. In other words, the investigated MHG corpus texts already predominantly show Jespersen stage III and are therefore much closer to Modern German than expected.32 31 This conclusion is reminiscent of the results in Frisch () for ME. Frisch argues that the bipartite negation particle does not constitute an independent system but rather the temporal overlap of two competing systems. This could also be argued for German: Early MHG was still stage I, classical MHG already mostly stage III, at least in our Upper German corpus texts. Preliminary data from the Central German area suggest, however, that the bipartite neg-particle may indeed have been a majority pattern there in MHG times, cf. also Schüler () on Western Central German. 32 Compared to Frisch’s () data for ME, MHG had progressed much further in the development towards using the verb-independent neg-particle only; for the period from to , which is roughly comparable to our MHG corpus, Frisch (: ) reports that ne as the only neg-marker is almost twice as common as ne + not, which in turn is more than ten times as common as not only. At the same time, only ne/en is generally the minority pattern in MHG and niht as the only neg-marker is used in the majority of cases, a stage that is only reached in English in the second half of the fourteenth century.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
Niht is generally placed in a fixed position in the centre of the topological middle field. It is unaffected by verb movement, that is, in a verb-final clause, the finite verb occurs after niht, whereas it precedes niht in a V clause, cf. () versus ().33 () und da er wúst das er nit dot was and since he knew that he NEG dead was ‘and since he knew that he was not dead’ (Prosalancelot I , )
Ve
() ‘Des ist mir niht ze mvote’, sprach aber Sifrit that is me NEG to mind said but Siegfried ‘ “That is not on my mind”, said Siegfried.’ (Nibelungenlied (A) , )
V
As mentioned above, the morphologically non-negative counterpart of niht is also partly grammaticalized and occasionally occurs as a neg-particle in MHG in the form of iht (< (io)wiht ‘anything’)34 continuing tendencies observed above for OHG uuiht. In our corpus, there is one example, given in (), where iht even constitutes the only marker of negation in the clause, i.e. it occurs in a Jespersen stage III-type construction. () Wir sulen den iungen herren enphahe dester baz, / daz wir iht we shall the young lord receive all.the better that we NEG verdienen des snellen rechen haz. deserve the strong warrior hatred ‘We shall receive the young lord all the better, so that we do not deserve the hatred of the strong warrior.’ (Nibelungenlied (A) , –) .. Middle Low German ... The preverbal clitic negation particle Just as in MHG, en/ne has become very rare by MLG times35, in stark contrast to its omnipresence in OLG, as shown in Table ..36 There is also a drastic decrease over the course of the MLG period. Furthermore, there are much fewer occurrences of MLG en/ne on its own. There is very rare 33 Besides, niht is also occasionally used as a neg-focus particle in negated clauses with narrow/marked focus of negation and then occurs immediately left of the constituent in focus, arguably forming part of that constituent (cf. Jäger : –). 34 MHG iht < OHG iouuiht ‘anything (ever/at all)’ (cf. Schmeller , I: ; Grimm and Grimm –, X: –; Grimm : ; Paul : ), but also < OHG uuiht ‘(any)thing’ (cf. Schmeller , I: ; Grimm : ; Wheelock LaBrum : ). Apart from example (), there is one more possible case in the MHG corpus where iht could, however, alternatively be analysed as argumental Und hútent wol das er myner fert icht wißse ‘And make sure that he does not know of my journey/that he does not know anything of my journey’ (Prosalancelot I , ). 35 Despite the similar term, the MLG period (–) is not directly comparable to the MHG period, as it only slightly overlaps with the latter, and is mostly contemporary rather to the ENHG period (–). 36 Westphalian: Börstel, Steinfurt; Eastphalian: Barsinghausen, Braunschweig, Mariengarten; North Low Saxon (Altland): Oldenburg, Scharnebeck, Uelzen; (East Elbian) Hansa cities (Neuland): Lübeck, Stralsund.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Ratio of clauses including en/ne in MLG negated clauses
Westphalian Eastphalian North Low Saxon Hansa cities
–
–
–
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
–
–
% () % () % () % ()
% () % () % () % ()
evidence, for example in (), for the Jespersen stage I pattern (sentential negation with ne/en only) in MLG, that is, at a time years later than the MHG data considered above.37 () der ik unde myne erven enscholed recht warende wesen of.which I and my heirs NEG.shall law guarding be ‘which I and my heirs shall not guarantee’ (Scharnebeck //) Almost all cases of MLG en/ne being used on its own, viz. of the cases, cf. (), constitute exceptive clauses of the kind already discussed for MHG, cf. example () and ensuing discussion. () vnde dar moste numment yn, he ne gheue V mark and there must nobody in he NEG give.SBJV five marks vp dat minste on the least ‘and nobody shall enter there, unless he give/pay at least five marks’ As argued above, these are unlikely to actually constitute cases of true sentential negation, which is why these clauses were not counted towards the total number of negated clauses containing en/ne in Table .. Especially in MLG, where exceptive constructions are much more frequent than in the MHG corpus, the difference from negative conditionals is clear: while (asyndetic) negative conditionals are one of the first contexts to lose ne/en, and always occur with the verb-independent negparticle or a negative indefinite, exceptive constructions are the only context to maintain the single preverbal marker for an extended period of time. They cannot contain other negative markers or indefinites, nor, in fact, NPI-indefinites. This constitutes further evidence that en/ne in this context does not express sententialscope negation.38 In contrast to MHG, there is no evidence for the use of ne/en as a
37 These are two examples from Scharnebeck (North Low Saxon), one from (.% for that period; of neg. clauses) and one from (.%; of neg. clauses). The former however contains the construction noch . . . noch ‘neither . . . nor’. Opposed to this are clauses containing nicht (of which with nicht alone). 38 See Breitbarth (: –, a) for a proposal of an analysis of the development of MLG en/ne in exceptive clauses in terms of an exaptation of the previous neg-particle as a higher C-related negation scoping over the world-operator quantifying over a possible world (potentialis).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
marker of affective polarity in other weak NPI contexts such as clauses dependent on negated or adversative matrix predicates in our MLG corpus. As textual Low German evidence becomes scarce after , the timing of the loss of the preverbal neg-particle is largely speculative. However, this development is clearly delayed compared to High German and was probably completed over the course of the seventeenth century (cf. also Pensel ). ... The bipartite negation particle The verb-independent middle-field negparticle nicht is firmly established by MLG times as the standard sentential negator. Especially initially, it frequently co-occurs with the preverbal neg-particle as in (a) and (b).39 () a. dar en sculle wii se nicht ane there NEG shall we them NEG from ‘we shall not bar them from it’ (Lübeck //) b. [nymant] de hir mid namen nicht nobody who here with name NEG ‘[Nobody] who is not explicitly named here’ (Steinfurt //)
hinderen bar
V
ensteet NEG.stands
Ve
This Jespersen stage II pattern is fairly common in MLG and, in contrast to MHG, actually occurs as the majority pattern in the fourteenth and partly in the fifteenth century MLG texts in our corpus. With the loss of preverbal ne/en, the bipartite neg-particle becomes less frequent in MLG, too. As Table . shows, this process takes place at different speed in the various dialects, with a longer retention of the Jespersen stage II pattern in the southwestern and southern Low German dialects than in the north and northeast where dialect levelling occurred due to dialect contact during the later colonization of the area as well as to the extensive Hanseatic trade (Breitbarth c). Languageinternal factors also play a role; the preverbal marker is kept significantly longer in V and verb-final contexts than in V clauses, and with lexical verbs significantly longer than with high-frequency modals and auxiliaries.40 T . Bipartite neg-particle en/ne . . . niht in MLG negated clauses
Westphalian Eastphalian North Low Saxon Hansa cities
39
–
–
–
–
–
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
For expository reasons, we will still gloss MLG en/ne as NEG. Recall, however, that it is arguably in the process of being reanalysed as a marker of affective polarity rather than negation. 40 Recall that, in contrast, modals were a context in MHG where stage I was maintained longer.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Neg-particle niht as the only neg-marker in MLG negated clauses
Westphalian Eastphalian North Low Saxon Hansa cities
–
–
–
–
–
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
% () % () % () % ()
... The verb-independent negation particle With the loss of the preverbal clitic neg-particle ne/en, MLG slowly changes from a Jespersen stage II to a stage III language with simple nicht as the standard neg-particle, cf. (). ()
. . . unde dar sole wy en nicht hinderlich an wezen and there shall we them NEG obstructive to be ‘and we shall not bar them from it’ (Steinfurt //)
V
Nicht already occasionally occurs as the only marker of negation in the clause in the early MLG texts in our corpus dating from the beginning of the fourteenth century. In the north and northeast of the Low German area, Jespersen stage III is already becoming more common than stage II over the course of the fourteenth century, cf. Table ., whereas this state-of-affairs is only reached in the southwestern and southern Low German dialects about a century later.41 Syntactically, MLG nicht behaves like MHG ni(c)ht in so far as it is unaffected by verb-movement and generally occurs in a fixed position in the middle field. Accordingly, it occurs after the finite verb in V and V clauses cf. () and (), but before the finite verb if the verb occurs in its base position in verb-final clauses cf. ().42 () alzo se dar nicht by kuonden komen so they there NEG at could come ‘so that they could not reach it’ (Steinfurt //)
Ve
() wil de kleger ohne deß nicht vorlaten wantst he plaintiff him this NEG yield ‘If the plaintiff does not want to yield this to him’ (Braunschweig //)
V
41
Recall that in the MHG corpus, in contrast, this stage was already reached by the thirteenth century. Example () is a case of verb raising (see also Part III of this volume). This phenomenon is very common in MLG compared with MHG and ENHG. We have counted all clauses with a verbal complex in final position as verb-final (Ve), whether or not the finite part of the verbal complex is actually in the absolute final position in the linear order. 42
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
As in MHG, the original indefinite pronoun icht ‘something/anything’, adverbially ‘in anything’/‘at all’, can in rare cases be used as a verb-independent neg-particle besides nicht in MLG, (). () De brutscho scolen ok icht betere wesen wan v sol. the bride.shoes shall also NEG better be than five shilling ‘The bride shoes shall also not be more expensive than five shillings.’ (Braunschweig ) .. Early New High German In ENHG, the period of the High German language that is largely contemporary to MLG, the process of the loss of the preverbal clitic neg-particle, which had already largely taken place during MHG times, was completed. As Pensel (: –) reports for his ENHG data from to , the overwhelming majority, viz. % of the negated clauses including a neg-particle contain nicht as the only marker of negation, that is, observe Jespersen stage III. Note that .% still include the preverbal clitic en (.% even as the only marker of negation, but only en/ne is not attested at all in East Central German and West Upper German). Occasionally, a bipartite neg-particle is used. It is found in .% of the clauses containing a negparticle. En . . . nicht is retained longest in Western Central German (.%). In the Upper German dialects, there are only rare remains (.%), while the bipartite negparticle is no more attested in East Central German around . Two hundred years later, en/ne is extinct in all regional varieties of High German and also in Low German (Pensel : –). .. Modern (High and Low) German With no remains of the former preverbal neg-particle en/ne, Modern German in all its regional varieties constitutes a Jespersen stage III language. The Modern German neg-particle nicht (and its variants in the different dialects, e.g. Bavarian ned, Modern Low German nich, cf. (), and even the phonologically highly reduced Upper Saxon ni) shows no evidence for any difference in syntactic behaviour compared with MHG, MLG, and ENHG ni(c)ht. It is still positioned in the middle field independent of the verbal position and unaffected by verb-movement (cf. also examples () for Bavarian ned and () for Thuringian nech, given in Section ..). () Man pass op, suust mi dor nich daal Modern Low German but watch out scoot me there NEG down ‘But watch out, don’t you scoot down there (on me)’ (Schleswig-Holstein dialect, http://inesbarber.de/ib_vertelln.html) Besides the neg-particle nicht and its dialectal variants, there is still some rare evidence for its morphologically non-negative counterpart being used as a negparticle continuing on from OHG (io)wiht/MHG iht (compare also the use of MLG icht). Some Upper German dialects, viz. certain Bavarian (Schmeller : /), South-East Swabian (Grimm : ), and North-East Swiss varieties use the verb-independent neg-particle it or et (< MHG iht ‘anything’) even today, cf. ().
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
() Des ka it sei. that can NEG be ‘That’s not possible.’
Swabian
. Indefinites in the scope of negation and Negative Concord There are other ways of marking sentential negation besides the neg-particle. Another major type of neg-marker is negative indefinites (n-indefinites, also referred to as n-words after Laka ).43 In a negated clause with an indefinite pronoun, determiner, or adverb in the scope of negation, three basic syntactic patterns of negmarking are possible in so far as both the indefinite as well as a neg-particle, or just one of the two, may act as a marker of negation, cf. (). In the following, we will simply refer to these as patterns A, B, and C. ()
Syntactic patterns with indefinites in the scope of negation: • A: Negative Concord44 (neg-particle + n-indefinite) • B: neg-marking by neg-particle only (non-negative indefinite) • C: neg-marking by n-indefinite only (no neg-particle)
.. Indefinites in the scope of negation and Negative Concord (NC) in OHG In OHG, all three basic syntactic patterns with indefinites in the scope of negation are attested: negation may be marked by an n-indefinite as well as by the standard preverbal neg-particle ni, as illustrated in () and (), by the neg-particle alone while the indefinite does not take the form of an n-indefinite, as in () and (), or by the n-indefinite only without the neg-particle occurring on the finite verb, as in (). A Negative Concord (neg-particle on V + n-indefinite): () (ne cui dicerent) thaz sie niheinigemo nisagatin that they no.one NEG.told ‘that they did not tell anyone’ (Tatian , ) () (& non respondit ei/ ad ullum uerbum) Inti niantligita imo/ zi noheinigemo uuorte and NEG.answered him to no word ‘And did not answer to a single word’ (Tatian , –)
n-indef. before ni+V
n-indef. after ni+V
43 Further types of neg-markers include negative complementizers such as OHG nibu ‘unless’, the negative counterpart of ibu ‘if ’, or special neg-focus particles such as OHG nalles (NEG+FOCUS). For a discussion of these see Jäger () among others. 44 Negative Concord (NC) is the co-occurence of several neg-markers in one clause that is interpreted as containing a single semantic negation. By contrast, the term ‘Double Negation’ is generally used to refer to the opposite phenomenon where two neg-markers in one clause both signal a semantic negation and the semantic negations ‘cancel each other out’ (e.g. Nobody knows nothing = Everybody knows something).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
B neg-marking by neg-particle only (neg-particle on V + (NPI-)indef.): () Er tothes íó ni chóreti, er er then dróst he death ever NEG chose before he the consolation habeti indef. before ni+V had ‘He never chose death before he had this consolation’ (Otfrid I , ) () (In qua sententia nemo dubitet secundam esse personam.) In dhesemu quhide ni bluchisoe eoman, ni dhiz sii In this saying NEG doubt anybody NEG this be chiuuisso . . . indef. after ni+V certainly ‘Nobody shall doubt that in this saying, it is certainly . . . ’ (Isidor , ) C neg-marking by n-indefinite only (n-indefinite, no neg-particle on V): () (In quo nondum quisquam/ positus fuerat.) Inthemo noh nu níoman/ Ingisezzit uuas. in.which still now nobody put was ‘in which nobody had been put yet’ (Tatian , –)45 The distribution of these three patterns in our OHG corpus is given in Table . (percentages of all clauses with indefinites in the scope of negation). NC is the most common syntactic pattern on average that is used in OHG if an indefinite occurs in the scope of negation. It occurs with preverbal as well as with post-verbal indefinites, cf. examples () and (). This fact falsifies Lehmann’s () generalization that indefinites after ni+finite verb are never n-indefinites in OHG. Since OHG uses the preverbal neg-particle also if the n-indefinite precedes the T . Distribution of negation patterns A, B, and C in OHG A (NC) Isidor Tatian Otfrid Notker average
% () % () % () % () %
B (neg-particle only) % () % () % () % () %
C (n-indefinite only) % () % () % () % () %
45 Note that the line break, which constitutes an obstacle for changing the word order according to OHG syntax in the translation in Tatian (see also Chapter ) and which occurs here between the indefinite and the finite verb, does not constitute a general obstacle for the introduction of NC, cf. () where NC is introduced against the Latin original across a line break. For a detailed discussion of this issue see Jäger (: f).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
finite verb, the language can be classed as a so-called strict NC-language (cf. Giannakidou ) in contrast to a non-strict NC language such as Italian where the neg-particle occurs when the n-indefinite follows the finite verb but not when the n-indefinite precedes the finite verb. The fact that NC must have been a genuine OHG syntactic pattern is not only evident from its high frequency but also from the fact that NC was commonly introduced against a Latin original by adding the preverbal neg-particle ni, cf. (), (), and (), when the Latin clause contained only one neg-marker, Latin being a non-NC language. Note that, interestingly, Otfrid himself reflects on this fact in the dedicational letter to Liutbert at the beginning of his gospel book and explicitly states that, in keeping with his own language, he will use NC even though Latin does not show NC. ()
Duo etiam negatiui, dum in latinitate rationis dicta confirmant, in huius linguae usu pene assidue negant, et quamis hoc interdum praecauere ualerem, ob usum tamen cotidianum, ut morum se locutio praebuit, dictare curaui. ‘While two negative markers signify confirmation in Latin, in this language they almost always negate, and even though I could have avoided this at times, I wanted to adhere to the general way of speaking in my writing out of consideration for the everyday usage.’ (Otfrid, Ad Liutbertum, –)
NC can in principle occur in two main subtypes (cf. den Besten b):46 as NegDoubling between an n-indefinite and the neg-particle or, if several indefinites occur in the scope of negation, as Neg-Spread between several n-indefinites (which may then possibly be combined with Neg-Doubling as well). In OHG, NC takes the form of Neg-Doubling between the preverbal clitic neg-particle and an n-indefinite, as illustrated above. Neg-Spread, on the other hand, is not or hardly attested (cf. also Behaghel ; Donhauser ; Jäger , ). In our corpus of over , negated OHG clauses, there is just one case of Neg-Spread, viz. (). () (cui nemo umquam/hominum sedit) in theme neoman neo in aldere/ manno saz in which nobody never in ages of.man sat ‘in which no man ever sat’ (Tatian , –) Generally, if several indefinites occur in the scope of negation, only one is an n-indefinite (mostly, as in Modern German, the first one), as illustrated in (). () (Deum nemo uidit umquam.) got nioman nigisah io in altere god nobody NEG.saw ever in ages ‘Nobody has ever seen god’ (Tatian , ) 46
Other types of NC do not necessarily involve n-indefinites. These types include co-occurrence of the disjunction ‘nor’ (e.g. OHG/OLG/MHG/MLG/ENHG noh) or of a neg-focus particle (e.g. OHG nalles) with another neg-marker. For a further discussion of various OHG and MHG NC types see Jäger ().
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
As is also evident from Table ., the main syntactic pattern competing with NC in OHG is pattern B, i.e. neg-marking by the neg-particle only, while the indefinite in the scope of negation does not appear in the form of an n-indefinite.47 This is also attested with post-verbal as well as with preverbal indefinites cf. () versus (), in contrast to the demands of the so-called Neg-first principle (Jespersen ; Horn ). In Isidor and Otfrid, this pattern is even significantly more common than NC. While in Otfrid, this result is mostly due to his peculiar use of wiht rather than niowiht in the scope of negation, in the case of Isidor it might reflect an earlier stage within the OHG period,48 especially since we find pattern B even in opposition to the Latin original, which generally employs pattern C. Pattern C, on the other hand, is extremely infrequent in OHG with no occurrences at all in our corpus from Isidor or Otfrid. Where it does occur, it may partly be due to Latin influence. Furthermore, it is only attested with preverbal n-indefinites indicating that, at least with post-verbal n-indefinites, the preverbal neg-particle was obligatory in OHG. On the whole, NC is preferred though in no way obligatory in OHG. The optionality is due to a choice of the indefinite type in the scope of negation (as in Modern English or, to a certain degree, in Romance languages).49 There is arguably evidence for a change from a preference for pattern B in early OHG to a preference of pattern A (NC) in classic and late OHG. .. Indefinites in the scope of negation and NC in OLG In OLG, only two of the three basic patterns are attested: Negation is marked by the preverbal neg-particle and an n-indefinite, i.e. NC occurs (pattern A), cf. () and (), or it is only marked by the preverbal neg-particle while the indefinite does not take the form of an n-indefinite (pattern B), cf. () and (). A Negative Concord (neg-particle on V + n-indefinite): () that iu nian scathe ni uuírthid that you no damage NEG be ‘that you suffer no damage’ (Essener Evangeliarglossen , I, –II, ) () Sia ni namon is tho niam [=niaman] they NEG took it.GEN then nobody ‘They did not take (any of) it from anybody’ (Glossen zu Homilien Gregors des Großen , I, )
n-indef. before ni+V
n-indef. after ni+V
47 The indefinite takes the form of a ‘normal’ or notably of an NPI-indefinite: OHG still possessed a largely intact three-set system of indefinite pronouns, determiners, and adverbs comparable to Modern English, cf. for instance eteswaz ‘something’, iowiht ‘anything’, niowiht ‘nothing’. For more details on the OHG indefinite system see Jäger () and for a typology of indefinite systems with regard to polarity Weiß (b) and Jäger (). 48 Axel () has independently argued that Isidor represents an older stage within OHG on the basis of data concerning verb placement. 49 Concerning the choice between patterns A or B, there is no correlation with the syntactic function of the indefinite, in situ, or moved indefinite, or adjacency to Vfin (versus Donhauser ). However, there is some influence of the relative order of indefinite and Vfin; indefinites before ni+V tend to be neg-marked more often than those afterwards. Ni never lacks with Vfin before n-indefinite.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Distribution of negation patterns A, B, and C in OLG A (NC) Heliand Genesis minor texts
% () % () % ()
B (neg-particle only)
C (n-indefinite only)
% () % () % ()
% () % () % ()
B neg-marking by neg-particle only (neg-particle on V + (NPI-)indef.): () sô is io endi ni cumit indef. before ni+V thus it.GEN ever end NEG comes ‘thus the end (of it) will never come’ (Heliand ) () ne dragu ic ênig drugi thing. NEG carry I any deceptive thing ‘I am not bringing any kind of trick/deception’ (Heliand )
indef. after ni+V
Pattern C with neg-marking only by an n-indefinite is not attested in our OLG corpus; the preverbal neg-particle is obligatory in negative clauses as discussed in Section .... Table . summarizes the results from our OLG corpus analysis concerning the distribution of the patterns. While pattern A (NC) is attested in the Heliand, there is a strong preference in this text for pattern B with neg-marking by the neg-particle only, which is used in three quarters of the relevant cases. In the Genesis fragments, B is even the only pattern available. In the minor OLG texts, on the other hand, all four clauses with indefinites in the scope of negation observe pattern A, i.e. show NC. As the minor texts date from the tenth to eleventh century whereas Heliand and Genesis are ninth century texts, and as using neg-marked indefinites in the scope of negation is also the preferred pattern later in MLG (pattern A or C, see below), this may reflect a diachronic development within OLG, despite the very low absolute numbers in the minor texts.50 This development would be parallel to the one discussed for early versus classic and late OHG above from a preference for non-negative indefinites and marking of negation only by the preverbal neg-particle towards using n-indefinites in the scope of negation and thus a preference for NC. As in OHG, NC in OLG takes the form of Neg-Doubling between the preverbal neg-particle and an n-indefinite. Neg-Spread is not attested in our OLG corpus. In fact, there is evidence that, as in OHG, at most one of several indefinites could be negmarked in clauses where several indefinites occurred in the scope of negation, cf. (). 50 The probability that the five instances of n-marked indefinites in the minor texts happen to belong to the % of the Heliand-grammar and that the % share of n-free indefinites is accidentally unattested is . (by Fisher’s exact test in R). In other words, the grammars of the minor texts and of the Heliand being identical in this respect is essentially improbable.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
() Neo that iuuuar ênig ne dua / gumono an thesom gardon never that you.GEN any NEG do men in this garden geldes ettho côpes. money.GEN or buy ‘Never let any of you men, when you are buying or selling here in the middle world, behave in such a way’ (Heliand –) Both patterns A and B occur with preverbal as well as with post-verbal indefinites, thus once again contradicting the Neg-first principle in clauses such as () and (). Where OLG observes pattern A (NC), constructions with preverbal indefinites as in examples () and () indicate that it was a strict NC language, just as OHG. .. Indefinites in the scope of negation and NC in MHG In MHG, the same three basic syntactic patterns are attested as in OHG: (A) Negmarking by the preverbal neg-particle and an n-indefinite, i.e. NC, cf. () and (), (B) neg-marking just by the preverbal neg-particle, cf. (), or finally (C) just by an n-indefinite, cf. () and (). A Negative Concord (neg-particle on V + n-indefinite): () des enchunde im gevolgen nieman, so michel was sin kraft that NEG.could him follow nobody so big was his power ‘Nobody could follow him in that, so great was his power’ (Nibelungenlied (A) , )51 () als er nichts darumb enwúst as he nothing there.about NEG.knew ‘as though he did not know anything about it’ (Prosalancelot I , ) B neg-marking by neg-particle only (neg-particle on V + (NPI-)indef.): () Was die koniginne sprach, die jungfrauw sprach ein wort nicht; what the queen said the maiden said a word NEG ‘Whatever the queen said, the maiden did not say a word’ (Prosalancelot I , ) C neg-marking by n-indefinite only (n-indefinite, no neg-particle): () Und sie hatten nymant miteinander gewunnen dann ein junges and they had nobody with.each.other won than a young knebelin kleyn boy small ‘and they had no children apart from a little boy’ (Prosalancelot I , )
51
Ms. C: no preverbal neg-particle en.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Distribution of negation patterns A, B, and C in MHG
Prosalancelot Nibelungenlied Berthold average
A (NC)
B (neg-part. + (NPI) indef.)
C (only n-indefinite)
% () % () % () %
% () % () % () %
% () % () % () %
() wann nymant kan sich behuten vor verretery because nobody can himself protect from treachery ‘because nobody can protect himself from treachery’ (Prosalancelot I , ) However, the ratio of these patterns differs quite considerably from OHG, as is evident from Table . in comparison to Table .. First of all, one can observe a sharp decrease of pattern A (NC) by % on average or, considering the maxima found in individual texts, of % in OHG (in Tatian) versus only % in MHG (in Prosalancelot), a decrease by even %. Pattern A is clearly not the majority pattern any more. Our corpus data thus prove the accounts in the previous literature wrong that take NC to be even more typical of MHG than of OHG, e.g. Behaghel (: ); Keller (: ); Admoni (: ); de Boor and Wisniewski (: ); Paul (). The decrease of NC in MHG is linked to the decrease of the preverbal neg-particle through Jespersen’s Cycle. Both are already basically complete in Berthold’s sermons.52 In MHG, the loss of ne/en is generally slightly more progressed with n-indefinites than with the neg-particle niht. The observed link between the loss of ne/en and the loss of NC supports the typological implication as proposed by Zeijlstra () according to which, if there is a preverbal neg-particle (that can be analysed as a syntactic head Neg, see Section .), there will be NC. There is also a significant diachronic continuity between OHG and MHG with respect to NC. The main type of NC is still Neg-Doubling between the preverbal negparticle and an n-indefinite. Neg-Doubling with the verb-independent neg-particle niht, on the other hand, is hardly or not attested. The same goes for Neg-Spread between several n-indefinites, as in OHG.53 Consider the example in () where NegSpread would be possible but is not realized since the second indefinite does not occur in the form of an n-indefinite.
52 In ME, this process apparently took slightly longer. According to Ingham (: ), this type of Neg-Doubling was lost in the fourteenth century, whereas co-occurrence of several negative XPs (NegDoubling incl. not, Neg-Spread) is attested for at least a century after that. 53 In ME, in contrast, Neg-Doubling with not is apparently quite common and all indefinites in the scope of negation are strictly neg-marked (Ingham : /f).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
() wir heten ninder einen zagen. we had nowhere a hesitation ‘We did not hesitate at any point’ (Nibelungenlied (A) , ) There is one noteworthy (apparent) exception: the indefinite determiner dehein/ kein, predecessor of Modern German kein (‘no’) co-occurs with niht and also with nindefinites, as illustrated in () and (), respectively. • verb-independent neg-particle niht + (de)kein () wan er des niht enbern wil von dekeinem menschen der ze because he that NEG miss wants of no/any man who to sînen tagen komen ist his days come is ‘because he does not want to miss it of any man who has come to his days’ (Berthold I , –) • n-indefinite + (de)kein () aber sîn freude hât niemer mêr kein ende but his joy has never more no/any end ‘but his joy will never have an end’ (Berthold I , ) A closer inspection of the history and distribution of dehein/kein, however, shows that these patterns do not actually constitute exceptions. In OHG, dehein was predominantly used in non-negative affective contexts (conditionals, standard of comparison, dependent on adversative or negated matrix predicates, after âna ‘without’, ee ‘before’, kûme ‘hardly’, etc.), less often in the scope of negation, but never in non-affective contexts. It was clearly an NPI-indefinite comparable to English ‘any’.54 In MHG, its distribution shifted so that it predominantly occurred in the scope of negation. There are also the first attestations of dehein/kein as the only marker of negation in its clause, cf. ()—a clear indication that it is used as an n-indefinite. At the same time, dehein/kein continues to appear in non-negative affective contexts without giving rise to a negative interpretation, cf. (). () kein as an n-indefinite (only neg-marker in a negated clause): das sol das buch vil wol hernach gesagen, wann that shall the book very well later tell, because wir haben es yczo keyn stadt we have it now no/any place ‘The book shall tell that later, because we do not have any room for it now’ (Prosalancelot I , )
54 In contrast to its n-indefinite counterpart, nehein/enhein, which was diachronically substituted as an n-indefinite by dehein/kein, cf. Jäger ().
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
() kein as an NPI (conditional clause): Nu dunckest du mich als wise, sol kein man radt now seem you me as wise, should any man advice darzu geben, das thust auch du. to.that give that do also you ‘Now you seem to me just as wise. If any man should give advice on this, so should you’ (Prosalancelot I , ) Dehein/kein clearly changed diachronically from an NPI ‘any’ to an n-indefinite ‘no’ (cf. Jäger ). In Modern German, it only occurs as an n-indefinite. However in MHG, it was obviously ambiguous between an NPI and an n-indefinite. The apparent instances of Neg-Spread as well as of Neg-Doubling with the verbindependent neg-particle nicht including dehein/kein presumably involve the NPIvariant of dehein/kein. Thus, there is no compelling evidence in our MHG corpus for Neg-Spread or for Neg-Doubling with nicht.55 The change of dehein/kein with respect to its polarity type is indicative of a number of changes during and at the end of the MHG period which eventually resulted in a change of the entire indefinite system from a largely intact three-set system with regard to polarity in OHG comparable to Modern English including ‘normal’ indefinites (e.g. eteswaz ‘something’), NPI-indefinites (e.g. iowiht ‘anything’) and n-indefinites (e.g. niowiht ‘nothing’), to a two-set system with just ‘normal’ versus n-indefinites (the only exception being the temporal NPI-indefinite je(mals) ‘ever’).56 The original NPI-indefinites either semantically changed towards ‘more negative’ and became n-indefinites as in the case of dehein/kein ‘any’ > ‘no’, or died out as in the case of iht ‘anything’,57 or they changed towards ‘more positive’ and became ‘normal’ indefinites as in the case of ioman, Modern German jemand ‘anybody’ > ‘somebody/anybody’. This change in the indefinite system is arguably related to the changes in the syntax of negation. As the category of NPI-indefinites was virtually lost from the indefinite system, the choice of the type of indefinite in the scope of negation became fixed. Since MHG, n-indefinites were basically obligatory in the scope of negation. Thus, it was increasingly possible to identify negation by the type of indefinite used and the already vanishing preverbal neg-particle became even more dispensable. On the whole, there is a certain degree of optionality of NC in MHG, too. In contrast to OHG, however, this optionality is due to the optional use of the preverbal neg-particle in addition to an n-indefinite, a state-of-affairs reminiscent of Colloquial 55 According to examples in Paul (), however, these patterns are occasionally also found with n-indefinites other than dehein/kein. 56 For a detailed discussion of the changes of the indefinite system and an analysis in terms of Lexical Underspecification Theory see Jäger (). 57 Iht partly also showed tendencies to turn into an n-indefinite in MHG. Adverbially, it had been very occasionally used as a neg-particle since OHG and survives as such in some Upper German dialects (see Sections ..., ..., and ..).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
French or West Flemish.58 Generally, NC is now dispreferred. Instead, pattern C with neg-marking only by the n-indefinite is the majority pattern in our MHG corpus texts. Recall that this syntactic pattern was very rare in OHG. At the same time, the common OHG pattern of neg-marking only by the neg-particle co-occurring with a non-negative indefinite has become extremely rare. In other words, the main pattern competing with (or, in the case of MHG, even outnumbering) NC diachronically changes from pattern B to C. An indefinite in the scope of negation generally takes the form of an n-indefinite in MHG. .. Indefinites in the scope of negation and NC in MLG Of the three basic syntactic patterns discussed so far, MLG predominantly shows optionality between patterns A, cf. (), and C, cf. (), with a diachronic shift from a preference for the former to a strong preference for the latter. Pattern B with the preverbal marker, cf. (), is only attested in four examples out of negative sentences with indefinites.59 A Negative Concord (neg-particle on V + n-indefinite): () Dar enscal se nement to dringhen there NEG.shall them nobody to force ‘Nobody should force them to (do) it’ (Uelzen //) B neg-marking by neg-particle only (neg-particle on V + (NPI-)indef.): () dat ik edder myne erven enwillen noch enschullen dem ergenomeden that I or my heirs NEG.want nor NEG.shall the aforementioned heren dem bischupe . . . to Lubeke yenigherleye wiis schuldigen sir the bishop at Lübeck any way accuse ‘that neither I nor my heirs want nor shall accuse . . . the aforementioned sir the bishop of Lübeck in any way’ (Lübeck //) C neg-marking by n-indefinite only (n-indefinite, no neg-particle): () Nemant mach mit worden welde doen nobody may with words violence do ‘Nobody is allowed to be verbally violent/abusive’ (Oldenburg ) Table . gives the distribution of these syntactic patterns in our MLG corpus. As can be seen from Table ., the preferences shift during the researched period from pattern A to pattern C, which is clearly attributable to the loss of the 58 Again, the choice between patterns A or C does not appear to correlate strictly with the syntactic function of indefinite, whether it is in situ or moved, adjacent to Vfin or not, but with the relative order of indefinite and Vfin. If the verb precedes the n-indefinite, it is more likely to be neg-marked by en/ne than when the n-indefinite precedes the verb, in line with the Neg-first principle. 59 In addition, there is one potential case where an n-indefinite co-occurs with nicht, cf. example (), and one case where an NPI-indefinite co-occurs with nicht.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Distribution of negation patterns A, B, and C in MLG A (NC: preverbal neg-particle B (preverbal neg-particle C (n-indefinite only) + n-indef.) only) – – – – –
% () % () % () % () % ()
– .% () .% () .% () –
% () % () % () % () % ()
preverbal marker in the context of n-indefinites, as pattern B plays virtually no role. The turning point seems to lie around . As in MHG, Neg-Doubling takes the form of co-occurrence of the preverbal negparticle and an n-indefinite but is virtually not attested in the form of co-occurrence of the verb-independent neg-particle nicht (or the bipartite neg-particle) and an n-indefinite. There is only one exception to this generalization in our MLG corpus among sentences with n-marked indefinites, that is, in less than .%. In this case, cf. (), the n-indefinite is extraposed from the negated clause so that the n-marking here presumably serves to indicate that the indefinite is to be interpreted inside the scope of negation. ()
. . . dat ze sik nycht enscholen vorbynden tjegen nyne heren ofte that they REFL NEG NEG.shall unite against no masters or landesheren noch wedder nyne nedderland sovereigns nor against no Netherlands ‘ . . . that they shall not unite themselves against any masters or sovereigns nor (any) Netherlands’ (Oldenburg //)
In contrast to MHG and to OLG, NC frequently occurs in the form of Neg-Spread between several n-indefinites in MLG, regardless of the presence or absence of preverbal en/ne. (Recall, however, that our MLG sources are mainly contemporary to ENHG rather than to MHG so that one cannot directly compare the two, and that High German also partly developed Neg-Spread later, cf. Section ..). Neg-Spread is attested ninety-five times in our MLG corpus, that is in .% of all negative sentences with indefinites and in % of all negative sentences with more than one indefinite. Note that it does not seem to make a difference whether the indefinite is a pronoun or a (more complex) noun phrase: both can co-occur within one sentence, like nemand and nyn helpe in (), unlike in other languages, where Neg-Spread appears to be restricted to identical types of indefinites.60
60
See e.g. Déprez ().
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
() dar en willen wy nemande nyner helpe uop plichtich wezen there NEG shall we nobody no help up obliged be ‘we shall not be obliged to give any help to anyone in this regard’ (Steinfurt //) In MLG, in contrast to MHG, Neg-Spread was nearly obligatory; only in eight out of cases with several indefinites in the scope of negation does only one of the indefinites take the form of an n-indefinite, cf. (). () dat we nemende, dem we schuldich synt to dessem jare . . . ichtes that we nobody who we due are to this year anything gheven konnen give can ‘that we cannot give anything to anyone we are indebted to this year’ (Uelzen //) Comparing the historical stages of LG, we see that a drastic change has taken place since OLG. There, more than three quarters of all cases of indefinites in the scope of negation followed pattern B, that is, negation was marked on the verb, not on the indefinite. Neg-Spread was not an option either. Thus, n-free indefinites were preferred in the scope of negation. In MLG, on the other hand, the preferred pattern is A (NC) in the earlier texts in our corpus (roughly, during the fourteenth century), which is replaced by pattern C, n-indefinites on their own, as the preverbal marker ne/en that differentiates patterns A and C is lost during the fifteenth century. Pattern A including the new neg-particle nicht is only very marginally attested in our MLG corpus, as is pattern B. .. Indefinites in the scope of negation and NC in ENHG and Modern (High and Low) German In ENHG as well as contemporary MLG, we still find occasional occurrences of preverbal en/ne together with an n-indefinite. According to Pensel (: ), this old type of NC is attested in .% of all clauses containing some neg-particle around (but not at all in East Central German and hardly in East and West Upper German). In contrast, the preverbal neg-particle has completely disappeared in all of the German speaking area by early Modern German times (around , cf. Pensel : f). Accordingly, the old NC pattern of Neg-Doubling between the preverbal neg-particle and an n-indefinite disappears as a natural consequence of Jespersen’s Cycle, well before the influence of prescriptive grammar. The fact that Modern Standard German is a non-NC language61 where the first indefinite in the scope of negation has to be neg-marked but no additional neg-marker is possible, unless
61 Weiß (a) argues by contrast that Standard German is a ‘hidden NC-language’ (see Penka for a similar assumption). According to him, NC is necessarily present in all natural languages that have n-indefinites (i.e. there is no distinction of +/– NC in those languages), and may be absent from certain standard languages as those are presumably only ‘second order’ natural languages (Weiß ).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
proper Double Negation is intended, and can therefore be seen as a natural development.62 Yet, as in MLG, we witness the rise of new NC patterns in some regional varieties of German. In contrast to MLG (as well as the previous High German periods), they show Neg-Doubling with the verb-independent neg-particle nicht. Besides, there is partly also Neg-Spread in early Modern German times (Pensel ; Ebert et al. ). According to Pensel (: ), Neg-Doubling with nicht occurs in .% of all clauses containing some neg-particle around . Around , it is at .% (Pensel : ). During both periods, the overwhelming majority of NegDoubling cases (around : thirty-one of thirty-seven, around : six of seven Neg-Doubling cases) still includes the indefinite kein, cf. example (), which arguably was still ambiguous between an NPI and an n-indefinite (cf. also Pensel : f) so that the number of clear cases of Neg-Doubling, cf. (), is very low still in ENHG. The same holds for Neg-Spread, cf. (). () sie haben Keine so wahrhaffte Freude nicht they have no such veritable joy NEG ‘They do not have any such real joy’ (II Br. , – after Pensel : ) () Und geschach sust von min Geschlecht Niemand nütt and happened otherwise of my kin nobody NEG ‘and otherwise happened to nobody of my kin’ (I R. b, – after Pensel : ) () ez ensal auch nieman under den beiden partihen an den it NEG.shall also nobody among the both parties of the Furdern keinen schaden previous no one harm ‘Nobody among both parties shall harm no one of the former’ (I Chr. , – after Pensel : ) In several Modern German dialects, there is also evidence for NC of these more recent types of Neg-Doubling with the verb-independent neg-particle and/or NegSpread. This is for instance the case in Bavarian, cf. (), certain Thuringian, cf. (), and Swiss German dialects, cf. ()—the latter showing NC in the form of NegSpread, only. () Mia hod neamad koa stikl broud ned gschengt me has nobody no piece bread NEG given ‘Nobody gave a piece of bread to me’ (Weiß : )
Bavarian
62 Our data furthermore show that NC cannot be said to have been lost from German due to Latin influence, as widely claimed in the literature, e.g. Paul (: ); Dal (: ); Lockwood (: ); and Qian (), but instead is an independent natural development. At a time when the influence of Latin was overwhelming and most surviving texts are directly or indirectly influenced by Latin originals, viz. in OHG, NC was in fact most frequent.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
() die war aus Berlin un hatte von nischt nech offn Dorfe she was from Berlin and had of nothing NEG at.the village änne Ahnung Thuringian a clue ‘She was from Berlin and didn’t have a clue about anything in the village’ (Blankenhain – Thüringisches Wörterbuch, IV, –) () Es cha niemer nüüt defür. it can nobody nothing there.for ‘It’s nobody’s fault’ (www.medical-info.ch/samwunsch/playlist.php, July )
Swiss German
The Modern German NC dialects, such as Bavarian, which have Neg-Spread, i.e. NC between several neg-XPs, and esp. Neg-Doubling with nicht, thus do not simply preserve the old patterns from OHG and MHG but rather constitute more recent developments (see Jäger , cf. also Weiß b on NC in Modern German dialects in general and especially in Hessian).63 Modern Low German, too, appears to allow for Neg-Doubling, cf. (). However, Breitbarth (: –) argues that such cases should rather be analysed as emphatic multiple negative expressions (Zeijlstra ) in which n-indefinite and neg-particle nich form one constituent: () a. An so wat hest Du [nie nich] dacht to such something have you never NEG thought ‘You never thought about something like that.’ (Reuter, Kein Hüsung, Chapter ) b. ‘Wat hest dor?’ Aewer ick kunn swigen un säd: ‘Oh, what have.SG there but I can be.silent and said oh [nicks nich!’] nothing NEG ‘ “What do you have there?” But I can be silent and said: “Oh, nothing (at all).” ’ (Reuter, Schurr Murr, Chapter ) Such an analysis could perhaps also apply to koa stikl broud ned in () and nischt nech in ().
. Analysis Our analysis of the development of negation in High and Low German essentially follows Jäger (, ). As illustrated in (), we assume that there is consistently 63 These patterns arguably also became possible through the change of dehein/kein: as an NPI, this item could of course co-occur with n-indefinites or nicht. When it then started to change and first occurred as the only neg-marker in a clause, the learner had evidence that dehein/kein is an n-indefinite and at the same time still, that this item can also co-occur with n-indefinites or nicht. If s/he generalized this rule to other indefinites, s/he would end up with general Neg-Spread as well as Neg-Doubling including nicht. This hypothesis is supported by the fact that dialectal NC is still particularly common with kein.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
one functional projection NegP above VP,64 and only the lexical filling of this structure is what changes, not the syntactic structure itself (cf. Longobardi’s Inertial Theory of Syntactic Change). ()
CP C’ C0
...
ni/en/ne/Ø + V
NegP Spec ¬Op/ni(c)ht
Neg’ VP
Neg0
... tV ni/en/ne/Ø +V
(V-movement in V1/V2 clauses)
There are several advantages of an analysis in terms of NegP (cf. Pollock ; Haegeman , and much subsequent literature): the neg-particles in High and Low German and their change throughout the language history can be modelled in a way parallel to the analysis of cross-linguistic variation in accordance with the requirements of Labov’s (: ) uniformitarian principle, viz. as a variation between the head Neg or the specifier or even both positions within NegP being overtly realized (cf. Ouhalla ; Haegeman for a cross-linguistic analysis of neg-particles). Secondly, the different syntactic behaviour of the attested neg-particles is directly derived from the syntactic structure. The preverbal neg-particle that we find in OHG, OLG, and partly still in MHG, MLG, and ENHG constitutes the head of NegP.65 It does not behave like a full phrase (as implied e.g. in Lockwood : ; Admoni : ; Schrodt : ): Clauses with initial ni/ne/en + Vfin syntactically pattern with affirmative V clauses, clauses beginning in XP-ni/ne/en + Vfin pattern with affirmative V clauses.66 According to the Head Movement Constraint, the finite verb has to move through Neg on its way to any higher functional projections in the
64
For an alternative account without NegP cf. Breitbarth (). We assume a head-final NegP for historical High and Low German, in line with standard assumptions on the headedness of German INFL projections (see also Chapter ). This accounts for verb-final negated clauses as well as the order with regard to separable verbal particles, cf. (i): verbal particle (= stranded in V) – ni/en – Vfin. 65
(i) daz er siê fúrder/ ána ne-sêhe. that he her further on NEG-look ‘that he will not look at her any further’ (Notker, Psalter , ) 66 This also holds with respect to word order in clauses introduced by certain complementizers, cf. Jäger (, ) for wanta ‘because’ clauses.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
clause. In Neg, it attaches to the clitic neg-particle and accordingly always appears together with it whether in clause-final position or in C as evidenced in the data discussed in Sections ..., ..., ..., and .... In the tree marker in (), the verb movement is indicated by arrows. On the way to stage II of Jespersen’s Cycle, the phrasal negator nicht (rarely also iht) is grammaticalized. This neg-particle, which generally occurs in a fixed middlefield position, can be analysed as the specifier of NegP.67 This correctly predicts that nicht is unaffected by verb-movement as discussed in Sections ... and ....68 Contrary to proposals in the literature for English (Roberts and Roussou ; van Kemenade ; van Gelderen a), we do not assume that the grammaticalization of the second neg-particle is due to an obligatory movement of n-indefinites to SpecNegP because of the Neg-Criterion (Haegeman and Zanuttini ), with a diachronic loss of the movement (change from Move to Merge in SpecNegP) as illustrated in (). ()
NegP
NegP Neg’
Spec niuuihti
VP ... ti ...
> Neg0 ni/ne
Spec niht
Neg’ VP
Neg0 ni/ne
If the grammaticalization were based on a previous obligatory movement of n-indefinites to SpecNegP, one would expect the most common n-indefinite to turn into a second neg-particle. In OHG, for instance, this would be nioman ‘nobody’ rather than niowiht/nieht. An even more pressing problem for this analysis is the fact that not only n-indefinites may be grammaticalized as a second neg-particle, but exactly the same grammaticalization process may be observed for other items (notably NPIs) such as French pas, or indeed OHG uuiht ‘anything’ > MHG iht > present-day Upper German neg-particle it/et. These items are clearly not subject to neg-movement to SpecNegP. Finally, there is independent evidence against the assumption of an obligatory movement of all n-indefinites into SpecNegP whether in the overt syntax or at LF, cf. Déprez (); Penka and Stechow (). Therefore, we propose instead that the process of grammaticalization started from the adverbial use of niowiht/nieht (cf. Behaghel ; Jäger ; Breitbarth, Lucas, 67 Crucially, SpecNegP precedes VP (versus Abraham ) as VP-internal material occurs intervening between niht/nicht and the verb in final position: swaz ich frivntliche niht [ab in] erbit ‘whatever I cannot ask of him in a friendly manner’ (Nibelungenlied (A) , ). VP constituents such as definite DPs etc. may undergo scrambling and therefore secondarily appear in front of SpecNegP ni(c)ht, cf. for Modern German Webelhuth (); Büring (); Weiß (). 68 As far as the further development of the preverbal neg-marker after the grammaticalization of nicht is concerned, one may hypothesize on the basis of the (weak) evidence for ne/en beginning to evolve into a marker of affective polarity, that it is reanalysed as the spell-out/Alternative Realization (Emonds ) on T (in sentences without V-to-C movement) of the features of a CP-related polarity head, cf. Breitbarth (), or one could assume that it remains in Neg as a historical relic.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
and Willis ).69 The bridging context mentioned above, namely verbs allowing for an optional extent argument, made possible the reanalysis from a VP-internal nominal to a VP-adjoined adverb. The adverbial use of accusative DPs was common in OHG. Ni(o)uuiht was first used as a VP-adverb, that is, in a VP-adjoined adverbial position linearly adjacent to the phonetically empty position SpecNegP. From there, it could easily be reanalysed as occupying this position, cf. (). ()
NegP Spec [¬ Op]
NegP Neg’
VP
DP ni(o)uuiht
> Neg0
VP ni/ne
Spec ni(c)ht
Neg’ VP
Neg0 ni/ne
During MHG and MLG times, respectively, the head Neg is beginning to be no longer filled overtly so that stage III in Jespersen’s Cycle reached with just the Spec position of NegP filled by nicht and its variants, as is still the case in Modern High and Low German. Another advantage of the NegP analysis is thus that Jespersen’s Cycle is very neatly captured as in () (cf. also Rowlett for historical French and van Gelderen a for historical English). ()
Jespersen’s Cycle in terms of NegP • stage I: Neg0 overt • stage II: grammaticalization of SpecNegP > SpecNegP and Neg0 overt • stage III: loss of overt Neg0 > only SpecNegP overt
In order to account for the phenomenon of NC and indefinites in the scope of negation, we essentially follow Zeijlstra’s () agreement-based account, assuming, in spite of the fact that several neg-markers occur in one clause in NC constructions, that actually only one element bears a semantically interpretable negative feature [iNeg] while all other neg-markers carry a merely formal uninterpretable negative feature [uNeg].70 According to Zeijlstra (), n-marked indefinites in NC languages have an uninterpretable negation feature [uNEG] that needs to be licensed under c-command by a carrier of an interpretable negation feature [iNEG]. This can be the standard negator, or a covert negative operator ¬Op. In strict NC languages, he argues, the neg-particle itself also bears an uninterpretable negative feature [uNeg]. As there is evidence that OHG and OLG were strict NC languages, we propose (following Zeijlstra ) that the preverbal neg-particle bears the feature [uNeg].71 69 This adverbial use of the indefinite meaning ‘nothing’ has occurred several times in the history of German, and is also a means of emphasizing negation in present-day German, as Bayer () shows. 70 For a similar agreement-based account see Weiß (a), Jäger (), or Penka () among others. 71 With the preverbal neg-particle being [uNeg], it needs to be licensed by an [iNeg] feature on another element. This is either an empty negative operator ¬Op in SpecNegP or, in the case of the bipartite negparticle, the verb-independent neg-particle nicht in SpecNegP.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
This analysis is corroborated by the fact that the displacement of ni/ne/en due to verb-movement does not result in any semantic scope differences. The verbindependent neg-particle nicht, on the other hand, can be argued to bear the feature [iNeg] after successful grammaticalization as a result of Jespersen’s Cycle. Furthermore, we assume that all n-indefinites throughout the history of High and Low German bear the feature [uNeg]. This holds for the NC stages of the language as well as for the non-NC stages (deviating from Zeijlstra , but following Weiß a; Penka ). Even for Modern Standard German, it can be shown that n-indefinites are not semantically negative, as evidenced by e.g. neg-split data (Penka and von Stechow ) or certain cases of VP-ellipsis (Weiß a, ). In negated clauses in which [iNeg] is not realized in the form of nicht, there is a phonetically empty negative operator ¬Op in SpecNegP72 that bears the feature [iNeg]. NC can now be analysed as an instance of Agree73 between the uninterpretable neg-feature on an n-indefinite [uNeg] and the interpretable neg-feature on the negative operator [iNeg], as illustrated in () for the OHG example (). ()
CP C0 thaz
NegP Neg’
¬ Op [iNeg]
Neg0
VP sie
nisagatin
V’
DP niheinigemo [uNeg]
V0 tsag-
checking under c-command (Agree)
There are different options regarding how to capture the distribution and development of NC in the history of German on this basis. One possibility is to model Neg-Spread in MLG, ENHG, and Modern German NC dialects in terms of Multiple Agree (Zeijlstra ). The lack of Neg-Spread in OHG, OLG, MHG, Modern Standard German, and Modern Low German is then due to a lack of this possibility of Multiple Agree of neg-features at those stages of the language. The fact that nicht generally cannot co-occur with n-indefinites in NC constructions in MHG and MLG (as well as Modern Standard German and Modern Low German) can be captured with Penka’s () extension of Zeijlstra’s () account: for languages such as Standard French, where the sentential negator pas can also not co-occur with n-indefinites 72 Alternatively, one may assume that ¬Op may be freely adjoined to any propositional node, cf. Penka (). 73 That is, there is no general obligatory movement of n-indefinites to SpecNegP in overt syntax or at LF as under the Neg-Criterion approach (Haegeman and Zanuttini ), cf. Penka and von Stechow ().
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Features of German neg-markers: Multiple Agree approach
OHG/OLG MHG/MLG ENHG/Modern Bavarian/Thuringian etc. Modern Standard German/ Modern Low German
ni/ne
nicht
n-indefinites
[uNeg] [uNeg] / – – –
– [iNeg] [iNeg] [iNeg]
[uNegØ] [uNegØ] [uNeg] [uNegØ]
(rien, personne etc.), she proposes that the n-indefinites are not simply [uNeg], but [uNegØ], that is, they require a phonetically empty ¬Op [iNeg] to license them, not an overt one.74 Accordingly, the n-indefinites in the various stages of German are [uNegØ] for the most part. However, in ENHG and those Modern German NC dialects that allow for Neg-Doubling with the verb-independent neg-particle, the n-indefinites would have to be analysed as [uNeg] so that their uninterpretable negative feature may be checked by nicht or its respective dialectal variant. The negfeatures of the various neg-markers throughout the history of German would thus look as given in Table ..75 A second option, to model the distribution and development of NC in the history of German, builds on Haegeman and Lohndal (). They criticize the mechanism of Multiple Agree and its parameterization on conceptual grounds76 and suggest an alternative analysis of NC in terms of pairwise Agree with maximal matching.77 Based on this approach, Breitbarth () suggests the feature analysis given in Table . for historical Low German. She assumes that n-indefinites not only carry an [i/uNeg]-feature but additionally an [i/uQ]-feature. That means that a [uNeg, uQ] indefinite needs to be licensed by an [iNeg, iQ] operator. This operator is inserted in its scope position (under the present analysis: SpecNegP) as a last resort option (cf. Zeijlstra ) in case there is no other bearer of an [iNeg, iQ] feature present in the clause. This is the case in languages where the sentential neg-particle bears a [uNeg] feature, and where there is Neg-Spread (e.g. Czech, or early MLG). According to Zeijlstra (), there may be n-indefinites containing such an operator 74 Zeijlstra () suggests an alternative analysis in which, in addition to the feature [iNeg] and [uNeg], he assumes that certain items (such as the neg-particle pas in French) possess a semantic feature [NEG] and therefore clash with the covert operator ¬Op [iNeg] projected by the [uNeg] n-indefinites. 75 Note that the characterization given for ENHG, Modern Bavarian, and Thuringian obviously does not hold for all NC dialects, for instance not for the Swiss German variety discussed above, which allows Neg-Spread but no Neg-Doubling (see also Weiß b). 76 According to Haegeman and Lohndal (), Multiple Agree violates the Defective Intervention Constraint (Chomsky ). Furthermore, they argue that parameterization should reside in the properties of lexical items and should not affect operations such as Agree, which are part of the core computational system. 77 Two items cannot agree if one of them retains an unmatched uninterpretable feature, but they can agree if not all interpretable features find a match. Agree works bottom-up. Two uninterpretable features of the same type (e.g. x [uNeg]) can match, producing one single occurrence of the feature (Pesetsky and Torrego ; López ), as long as the last occurrence is matched by an interpretable feature.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
T . Features of Low German neg-markers: Pairwise Agree approach NC-type
¬OP
ni/ne
OLG (Genesis, Heliand)
optional (strict) Neg-Doubling
[iNeg, iQ]
[uNeg] – (only argumental n-indefinite)
[iNeg, iQ]
OLG (minor texts)
strict Neg-Doubling
[iNeg, iQ]
[uNeg] [iNeg, iQ]
[iNeg, iQ]> [uNeg, uQ]
MLG
Neg-Spread, Neg-Doubling only with en/ne
[iNeg, iQ]
[uNeg] [iNeg] >–
[uNeg, uQ]
[iNeg, iQ]
–
[uNeg, uQ]
Modern Low Neg-Spread, (probably) German no Neg-Doubling
nicht
[iNeg]
n-indefinites
as part of their internal syntactic structure.78 In case a language has such indefinites ([iNeg, iQ] in the current analysis), the covert ¬OP is not additionally merged in SpecNegP, rather, as proposed by Zeijlstra, the operator comes from the indefinite. In such a language, there can be no Neg-Spread as each n-indefinite contributes a negative operator. The changes from [iNeg] to [uNeg] occurring during this development can be explained on the basis of van Gelderen’s () Feature Economy principle, which defines pathways of change if the respective reanalysis is independently triggered. Thus, according to Breitbarth (), in a pre-attestation stage of LG, ni would have been [iNeg], but would have changed to [uNeg] after it had helped form [iNeg, iQ] n-indefinites by univerbation with NPI-indefinites. Once these [iNeg] n-indefinites had established themselves as the neutral way of expressing indefinite quantification in negative clauses in later OLG, Neg-Spread became virtually necessary in order to enable the use of this ‘neutral’ form of indefinite in clauses with more than one indefinite in the scope of negation, causing the change from [iNeg, iQ] to [uNeg, uQ] in n-indefinites. At the point when nicht lost its emphatic value and turned into the standard expression of negation—a weak (structurally deficient) [iNeg] adverb (cf. also Breitbarth )—en/ne lost its [uNeg] feature and was eventually lost. While 78
Zeijlstra () argues that [iNeg] indefinites are syntactically internally complex items consisting of a neg-operator and a non-negative indefinite. Split-scope readings are derived by spelling the item out low, but interpreting the neg-operator in its QR-scope position, and the indefinite in the base position/lower copy. (i)
. . . dass er im Garten keinen Menschen antraf und im Haus auch nicht => . . . dass er im Garten [¬OP einen Menschen [VP [¬OP einen Menschen] antraf] und im Haus auch nicht [VP einen Menschen antraf] that he in.the garden no person found and in the house also NEG ‘that he didn’t find anybody in the garden nor in the house’ (Zeijlstra : f, after an example from Weiß : ).
According to Weiß (), who first introduced the VP-ellipsis data, sentences such as in (i) primarily show that n-indefinites are semantically non-negative in Standard German.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
Anne Breitbarth and Agnes Jäger
T . Features of High German neg-markers: Pairwise Agree approach NC-type
¬OP
ni/ne
nicht
n-indefinites
OHG
strict Neg-Doubling
[iNeg, iQ] [uNeg] [iNeg, iQ] [iNeg, iQ]
MHG
Neg-Doubling only with [iNeg, iQ] [uNeg] [iNeg] en/ne, (mostly) no Neg-Spread
ENHG
optional Neg-Spread
[iNeg, iQ]
[iNeg, iQ] –
[iNeg]
Modern Bavarian, optional Neg-Doubling, Thuringian etc. Neg-Spread
–
[iNeg]
[iNeg, iQ]/ [uNeg, uQ] [uNeg]
Modern Standard no Neg-Doubling, no German Neg-Spread
–
[iNeg]
[iNeg, iQ]
possible under Feature Economy, the lack of similar triggers seems to have prevented a change from [iNeg] to [uNeg] in nicht since MLG, and no new Neg-Doubling has arisen. Extending this analysis to historical High German, one may assume the features given in Table . for the respective neg-markers.79 Comparing the development of negation in historical High and Low German under this analysis, we see that they evolved largely in parallel, though certain differences become apparent: in ENHG, there is variation in the availability of Neg-Spread (and Neg-Doubling), and those different variants were developed further in different present-day dialects. For Modern Low German, the evidence for Neg-Doubling is not conclusive.
. Conclusion The developments in the expression of sentential negation are to a large extent similar in High and Low German, but not entirely parallel. Both OHG and OLG are at stage I of Jespersen’s Cycle; the verbal clitic neg-particle ni/ne is virtually omnipresent in negative clauses. Modern Standard German and Modern Low German are both stage III languages, with the old clitic negator lost and standard negation expressed by a free adverbial negator nicht. The most important differences in the developments of the two languages manifest themselves in the MHG/ENHG and MLG periods. While early MHG is still predominantly a stage I language, classical MHG is already largely a stage III language. Contrary to the traditional view in the literature, there is no stable stage II period in High German. This is quite different in Low German. Not only is the MLG period later than the MHG period 79 It is likely that the Bavarian situation is somewhat more complex. There is Neg-Spread and (optional) Neg-Doubling with ned, but n-indefinites can also be used on their own and express sentential negation (Weiß ). Possibly, n-indefinites are ambiguous in Bavarian between [uNeg] (licensed by [iNeg] ned) and [uNeg, uQ] (licensed by a covert [iNeg, iQ] operator).
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
History of negation in High and Low German
it also takes much longer for the transition from stage II to stage III. That is, the developments are neither temporally nor internally parallel. With respect to the historical interaction between indefinites and sentential negation, there is large variation within the OHG and OLG periods. However, it can be shown that these differences reflect a diachronic development within these periods. Once this is taken into account, it appears that older OHG, represented by Isidor, preferred pattern B (verbal neg-particle with n-free indefinites), later OHG prefers pattern A (NC). Similarly in OLG, the older text of the Heliand prefers pattern B, the later minor texts pattern A. OHG as well as OLG seems to disprefer Neg-Spread. Both MHG and MLG show variation between patterns A and C, that is, the average negative sentence with an indefinite uses n-marked indefinites, and may optionally use the preverbal neg-particle in addition. In our MLG corpus, we can clearly see a shift from pattern A to pattern C, concomitant with the loss of the preverbal marker. The diachronic development in High German is comparable but takes place earlier. The main difference between MHG and MLG is that the former hardly uses Neg-Spread, while the latter strongly prefers it in negative clauses with more than one indefinite in the scope of negation. As far as the syntactic structure is concerned, we assume that both High and Low German at all their historical stages share the same underlying structure, and that only the different positions of the respective neg-markers within that structure as well as their different lexical features lead to the observed differences in the syntactic behaviour of neg-particles and the (non-)availability of NC, and of different types of NC.
OUP CORRECTED PROOF – FINAL, 16/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Part III The Right Periphery
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
11 Introduction to Part III ERIC FUẞ
This part deals with historical facts and developments concerning the right sentence periphery in German, that is, (morpho-)syntactic properties associated with the right-most part of the middle field and the right sentence bracket as well as the post field (for a brief introduction of the terminology of the topological model and its correspondences in the generative analysis of clause structure see Chapter ). The exposition focuses on the placement of phrasal material (complements, in particular) relative to the verb, and properties of the verbal complex located in the right sentence bracket. Chapter provides an overview of issues related to the apparent ‘mixed’ OV/VO character typical of many early Germanic languages and discusses whether it is possible to identify a basic word order in early German (OHG, in particular). Chapters and examine in some more detail the factors that govern the placement of (phrasal) material relative to the position of the verb. Diachronic aspects of the structure and ordering properties of verb clusters are discussed in Chapters to , which deal with the rise of periphrastic constructions, the so-called IPP (infinitivus pro participio) effect, and the development of the ACI (accusativus cum infinitivo) construction in the history of German.
. Mixed OV/VO patterns and the theoretical analysis of word order variation It is a well-known fact that the early Germanic languages permitted a wider range of word order options that gradually gave way to the more rigid constituent order found in the present-day languages. In Parts I and II, we have already discussed aspects of the ordering of elements at the left periphery and in the middle field of the German clause. In what follows, we will complete the picture with a discussion of properties belonging to the right clausal periphery. As already hinted at above, we will first focus on the ordering of phrasal elements relative to the position of the verb (i.e. the OV/ VO alternation) before we shift our attention to the structure of the verbal complex at the end of the clause.
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Eric Fuß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
As discussed in Chapter , historical syntax was a rather neglected area of research in traditional Germanic linguistics (for early exceptions cf. e.g. Erdmann ; Paul , ; Behaghel –; relevant later work includes Bolli ; Näf ; Borter ; and Lenerz ). It has not been until recently that the study of aspects of word order has moved into the centre of attention, as documented by the number of publications devoted to variation between OV and VO patterns in various historical stages of German in the last twenty years (cf. e.g. Robinson ; Dittmer and Dittmer ; Prell ; Schlachter , , ; Hinterhölzl , , ; Axel ; Petrova ; Schallert ; Petrova and Hinterhölzl ; Sapp ). This shift of attention has been inspired not only by a renewed interest in diachronic syntax but also by the availability of new critical editions (cf. e.g. Masser’s edition of the Tatian translation) and by theoretical work that highlights the connection between syntax and the coding of information-structural distinctions (cf. e.g. Rizzi ; Belletti ). In the following chapters, we aim to clarify a set of empirical issues, reassessing the scope and frequency of extraposition/VO order in earlier stages of German with the help of qualitative and quantitative analyses. In particular, we want to explore whether it is possible to attribute the variation between OV and VO to independent factors (e.g. as a means to mark information-structural distinctions, cf. Petrova and Hinterhölzl ; Schlachter ).1 In addition, we investigate what an adequate theoretical account of the empirical facts should look like. Notice that the analysis of word order variation raises at least two major issues for linguistic theory. As modern syntactic frameworks typically do not leave much room for grammatical optionality, the very fact of linguistic variation is often perceived as a challenge as long as the variants cannot be shown to be related to different functions, meanings, or registers. Thus, it is important to distinguish between ‘free’ linguistic variation, that is, variation which cannot clearly be linked with any independent linguistic or communicative factors, and what might be called ‘principled variation’, where different surface patterns serve to mark certain linguistically relevant distinctions (informationstructure, clause type, etc.). While the former suggests an analysis in terms of competing grammatical options that exist side by side in the mind of the speaker (i.e. grammar competition in the sense of Kroch , or Double Competence in the sense of Weiß b), the latter is perhaps more adequately modelled in terms of a single grammar that generates different surface patterns dependent on the absence/ presence of certain discourse-semantic or prosodic triggers (cf. Hinterhölzl ; Biberauer and Roberts ). Furthermore, the apparently ‘mixed’ OV/VO character of the early Germanic languages is intimately connected to the question of whether it is possible to identify a basic order from which the set of surface orders is derived. Traditional approaches that analyse VO orders as resulting from extraposition presume that the relevant base order is OV. In contrast, analyses invoking competing grammars typically accept the idea that there is more than a single base order, while 1
Related questions that are addressed only in passing concern the historical development of V-XP order in MHG and ENHG (cf. Ebert ; Prell ; Sapp ), and the status and historical development of (close-to-obligatory) extraposition of finite argument clauses.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part III
accounts making use of multiple leftward movement operations triggered by discourse-semantic factors (cf. e.g. Hinterhölzl ) often assume that the notion of an unmarked/underived base order does not constitute a valid theoretical primitive to begin with. The contributions to this volume that address related questions pertaining to variation between VO and OV patterns in early German (focusing on the placement of phrasal constituents) can be summarized as follows. The goals of Chapter by Eric Fuß are twofold. First, it reviews a set of diagnostic tests for basic word order and discusses their application to early German, focusing on OHG. It is shown that OHG exhibits a range of ‘mixed’ OV/VO patterns on the syntactic surface, but still shares a number of characteristics with the present-day Germanic OV languages (e.g. basic OV and V-Aux order, preverbal verbal particles/ derivational affixes, and verb (projection) raising). While acknowledging that postverbal placement of material (leading to surface VO orders) is linked to factors such as prosodic weight and the introduction of new discourse referents (see Chapter for details), it is pointed out that OV order seems to be the unmarked case, which is compatible with a larger set of pragmatic contexts. The conclusion that OHG is more ‘OV-ish’ than other early Germanic languages (OE, in particular) is corroborated by quantitative evidence involving deviations from the Latin source text in OHG translations. In addition, the chapter critically assesses a selection of theoretical approaches to word order variation, highlighting that an adequate theoretical model should capture the fact that there are not only significant differences between OHG and OE, but also between OHG and the present-day OV languages. Observing that the OHG facts alone do not necessarily entail a specific mode of analysis, it is argued that the empirical evidence is in principle compatible with a more traditional account in terms of a basic head-final structure plus extraposition (Axel ; Sapp ), the assumption of multiple base-generated orders (Haider b, ), and an analysis that treats basic SOV as an order derived from a (universal) SVO base (e.g. Petrova and Hinterhölzl ). Arguably, however, a rationale for selecting one theoretical approach over the other(s) becomes available when a diachronic perspective is adopted. A successful account of word order variation in the history of German should not only capture the OHG facts but also shed light on the question of how and why these additional word order options were lost in the transition to Modern German (see Sapp for a relevant line of reasoning). In Chapter , Helmut Weiß and Svetlana Petrova report on a study of the OV/ VO alternation in OHG, focusing on embedded clauses introduced by the complementizer thaz ‘that’. Based on a corpus of sentences collected from the OHG Isidor and a selection of smaller OHG documents, Weiß and Petrova show that the vast majority of orders attested is compatible with a traditional analysis in terms of underlying OV in combination with extraposition of various types of phrasal constituents. Of this set, more than % of all cases are instances of unambiguous OV orders of the type (XP)-V-Aux; another % exhibit V-Aux order in the verbal complex followed by one or more phrasal constituents. The remaining cases (around %) distribute over six additional types, most of which are compatible with both a traditional OV analysis and alternative theoretical approaches that posit leftward movement operations from a VO base. Weiß and Petrova argue that the majority of
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
these orderings can be captured by an analysis that assumes competing settings of the head parameter within VP in combination with a (universally) head-initial functional projection that dominates the VP (Fuß and Trips ). However, they also point out that in contrast to OE, OHG does not fulfil other criteria for a pure VO grammar (i.e. a combination of a head-initial VP embedded by a head-initial vP) such as final placement of prosodically light elements (pronouns and monosyllabic adverbs, in particular). Additional orderings where the finite verb occurs to the left of certain adverbial elements are taken to represent instances where the finite verb has moved to an inflectional head further to the left of the clausal spine (either T, cf. Weiß , or C, cf. Axel ). The question of which factors govern word order variation in the right periphery of OHG is then taken up by Roland Hinterhölzl and Svetlana Petrova in Chapter . Basically following Petrova (), Petrova and Hinterhölzl (), and Schlachter (), it is argued that OHG was a discourse-configurational language, in which the information-structural value of a constituent and its (linear) position in the clause are intimately correlated. In addition, the placement of arguments and adjuncts is influenced by prosodic factors (i.e. weight/heaviness); in OHG, however, the latter are regularly outweighed by information-structural requirements (cf. Hinterhölzl ). Building mainly on evidence from the OHG Tatian translation, Petrova and Hinterhölzl () argue that the position of the finite verb partitions the clause into different domains that are linked to different discourse-structuring functions. Given, presupposed/background information is placed to the left of the finite verb in subordinate clauses, preferably at the left edge of the middle field, while material to the right of the finite verb typically represents novel information (i.e. the asserted part, or the new-information focus in the utterance). Moreover, there seems to be a dedicated position for contrastive foci, which typically occupy a position directly leftadjacent to the finite verb (i.e. in between the background domain and Vfin). Adopting the Universal Base Hypothesis (Kayne ), according to which heads uniformly precede their complements, Petrova and Hinterhölzl provide a formal model of these findings based on the assumption that the finite verb occupies the head of a focus phrase (FocP), which divides the clause into two domains related to background and presentational focus, respectively, and provides a structural position for contrastive foci (which move to SpecFocP).
. The verbal complex Traditionally, the study of phenomena related to the right sentence bracket has focused on properties of the so-called verbal complex, that is, the fact that in OV languages like German, finite and non-finite verbs together form a compact verb cluster that occurs at the end of embedded clauses. While the finite (or hierarchically highest) verb may occur in absolutely final position in the verbal complex (which is often taken to reflect the basic OV character of early Germanic), we can also observe quite a number of alternative ordering patterns in historical and present-day varieties of German (cf. e.g. Maurer ; Behaghel –, Vol. IV; Robinson ; Ebert ; Schmid and Vogel ; Schmid ; Axel ; Sapp a). Patterns where
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part III
the finite verb occurs to the left of the selected non-finite verb (Vfin-V) are usually referred to as ‘verb raising’ orders (originally suggesting a certain mode of analysis, where the non-finite part right-adjoins to the finite verb, cf. e.g. Evers ); cases where a non-verbal constituent intervenes between the finite verb and the non-finite parts of the verbal complex (Vfin-XP-V) are called ‘verb projection raising’ (suggesting rightward movement of a non-finite VP). In what follows, we will take a closer look at a selection of relevant facts, focusing on the rise of periphrastic constructions such as analytic tenses/passives, and certain infinitival constructions, most prominently the so-called IPP (infinitivus pro participio) and the ACI (accusativus cum infinitivo). Both sets of phenomena are typically linked to the re-ordering of elements in the verbal complex. Taking OHG as a starting point, we will trace the relevant historical developments until the NHG period.2 In addition, we will address the question of how the various linearization patterns (and, in the case of IPP, morphological effects apparently connected to certain orderings) can be adequately modelled.3 In Chapter , Augustin Speyer gives an overview of the emergence of periphrastic constructions in the history of German. Based on a discussion of grammaticalization processes giving rise to periphrastic constructions, Speyer identifies a set of morphosyntactic diagnostics (agreement morphology on the adjectival/participial element, free reordering etc.) that can be used to assess the status of a given construction moving along the cline from full verb plus nominal/adjectival complement to verbal complex (i.e. auxiliary plus non-finite verb form). Against this background, he then goes on to review a set of relevant changes affecting verbal constructions in German. He shows that the rise of periphrastic forms came about in two ‘waves’: While relevant passive and perfect constructions seem to be already well established in (early) OHG, the development of an analytic future tense took place in the MHG period. It is shown that the latter change involved a competition between a number of different construction types, in which the combination of werden ‘become’ and an infinitive eventually won out over the other (more ancient) options (i.e. modal plus infinitive, werden ‘become’ plus present participle). As for why exactly the alternative patterns were discarded, Speyer speculates that after the completion of relevant grammaticalization processes, only combinations of genuine auxiliaries (i.e. sein ‘be’, haben ‘have’,
2 The distribution of verb raising and verb projection raising in the present-day dialects (cf. e.g. Patocka ; Schmid and Vogel ; Höhle ; Weiß and Schwalm ) potentially provides important clues concerning the historical development of the individual variants (with verb raising predominantly found in western dialects and non-verb raising orders confined to eastern dialects such as Bavarian, cf. e.g. Kolmer ). 3 A related question which cannot be addressed here in detail concerns the factors that trigger re-ordering in the verbal complex. For example, Diels () and Behaghel (–, Vol. IV) claim that in OHG, verb raising is triggered by rhythmic/prosodic factors (see also Dittmer and Dittmer on the OHG Tatian), while Reis () and Fourquet () argue that verb raising is less likely to take place in cases where fronting of the auxiliary would blur the distinction between main and embedded clauses (i.e. the difference between V and V-final order). An alternative hypothesis is put forward by Schmid and Vogel (), who argue that apart from rhythmic or prosodic factors, the relevant orderings are also linked to information-structural distinctions (e.g. the realization of focus; see also Sapp a on ENHG).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
and werden ‘become’) and the most common non-finite forms (the infinitive and the past participle) could be used to form periphrastic verb constructions. Chapter is concerned with the IPP effect, a phenomenon intimately related to the formation of a verbal cluster in the Germanic OV languages. Based on a thorough review of the historical facts, Agnes Jäger shows that first examples where a pure infinitive replaces a past participle (that selects another infinitival form) are attested as early as the thirteenth century in connection with causatives such as tuon ‘to do’ and lazen ‘to let’ and the perception verb hoeren ‘to hear’. Up to the ENHG period, the class of IPP verbs grows, extending to modals, other perception verbs, benefactives, inchoatives, and a couple of control verbs, before the list shrinks down somewhat to the set of verbs that trigger the IPP effect in Modern German. Jäger argues convincingly that the historical evidence favours the hypothesis that the IPP originated as a true infinitive (Erdmann ) over an alternative scenario according to which the IPP goes back to a ge-less past participle that was later reanalysed as an infinitive (Lachmann ). Besides the historical developments in German, Jäger also examines IPP effects in a range of other Germanic languages and various dialects of German, in particular. By taking into account a broader range of facts, she not only adds a typological dimension to the discussion but also adduces evidence showing that the IPP is actually one repair strategy among others which languages employ to avoid a configuration where a past participle selects an infinitive in three-verb clusters. Finally, Jäger discusses the link between the IPP effect and the order of verbal elements inside the verbal complex, highlighting a set of implications for the theoretical analysis of the IPP effect. In particular, it is shown that in contrast to views widely held in the theoretical literature, reordering of verbal elements is not a necessary ingredient of the IPP phenomenon. Another type of (obligatorily coherent) infinitival construction is the subject matter of Chapter by Augustin Speyer, which focuses on the historical development of ACI patterns in German. Since the term ACI goes back to traditional Latin grammar, describing cases where the subject of an infinitival complement stands in the accusative case, Speyer first gives an overview of the properties and theoretical analysis of the ACI in Latin. He assumes that the infinitival complement of ACI verbs corresponds to a bare VP and discusses three possible structural descriptions of it. In what he labels the ‘biconstituental analysis’, the accusative subject has undergone raising out of the embedded non-finite VP to an object position of the matrix verb; in the so-called ‘uniconstituental analysis’, the accusative NP is part of the non-finite VP and receives case in its base-position. Finally, the ACI may be analysed as a small clause, in which the infinitive is base-generated as a secondary predicate to the accusative object. It seems that this last option represents the historical origin of the ACI in the Indo-European language family, which was later reanalysed as a biconstituental structure. In Latin, early instances of the ACI are confined to causative and perception verbs, before the construction spreads to matrix verbs such as ‘think’ which cannot take a (non-propositional) accusative object. Speyer assumes that this extension became possible when ‘biconstituental’ structures were reanalysed as ‘uniconstituental’ structures.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Introduction to Part III
Speyer then discusses the nature and scope of the ACI in various historical stages of German. He demonstrates that throughout the history of German, the ACI is typically found with causative and perception verbs and concludes that a biconstituental analysis is basically correct for all stages of German. In addition, a set of apparent exceptions to this general pattern is discussed. It is argued that occasional occurrences of the ACI construction with non-perception verbs in OHG most likely represent loan translations from Latin. However, it appears that in MHG, there are at least some cases that cannot be attributed to Latin influence. From this observation Speyer concludes that a reanalysis as a uniconstituental structure was at least an option in MHG. Finally, it is shown that at first sight, ENHG seems to display an even wider array of ACIs. Upon closer inspection, however, it becomes clear that the extension of the range of the ACI construction is a feature of learned written discourse only, whereas texts closer to the vernacular exhibit the familiar restriction to causative and perception verbs.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
12 The OV/VO alternation in early German Diagnostics for basic word order ERIC FUẞ
. Introduction In traditional works on word order in early Germanic, it is often assumed that languages such as OHG and OE more or less faithfully reflect the basic OV character of earlier stages such as Proto-Germanic and PIE (cf. e.g. Behaghel –, Vol. IV). However, it is also well known that in the very same languages, we can observe postverbal placement of all kinds of elements (occupying the post field in both main and embedded clauses),1 giving rise to a degree of word order variation which is not encountered in their present-day descendants.2 The examples in () illustrate this fact for embedded clauses of OHG: () a. (Hic enim post obitum moysi dux effectus principatum obtenuit . . . ) Dher selbo infenc haerduom dhes israhelischin folches, [PP adjunct] that same received dominion of.the Israeli people dhuo ir dhes leididh uuardh [after moysises ablide] . . . when he their leader became after Moses’ death ‘that same one received dominion over the people of Israel, when he became their leader after Moses’ death . . . ’ (Isidor , ) See Chapter for an overview of the terminology of the topological model and its correlations to generative syntactic structure. 2 For OE and ME cf. e.g. Bean (); Mitchell (); Pintzuk (b, ); Kroch and Taylor (); Fischer et al. (); Fuß and Trips (); Biberauer and Roberts (); Taylor and Pintzuk (a, b, ). For variation between OV and VO patterns in various historical stages of German cf. Maurer (); Behaghel (–, Vol. IV); Borter (); Lenerz (); Dittmer and Dittmer (); Robinson (); Prell (); Schlachter (, , ); Hinterhölzl (, b, ); Axel (); Petrova (); Schallert (); Petrova and Hinterhölzl (); Sapp (a, b, ). 1
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Eric Fuß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
b. ( . . . et archana secretorum, ut scias) endi ih uuillu dhazs dhu firstandes [heilac chiruni] [direct object] and I wish that you understand holy secret ‘and I wish that you understand the holy secret’ (Isidor , ) c. (ut esset deo subiectus) dhazs ir chihoric uuari [gote] [indirect object] that he obedient was god.DAT ‘that he was obedient to God’ (Isidor , ) d. (si fuerit oculus tuus simplex) oba thin ouga uuirdit [luttar] [predicative adjective] if your eye becomes light ‘if your eye is good’ (Tatian , ) The examples in () illustrate post-verbal placement of an adjunct PP, a direct object, an indirect object, and an adjective used predicatively. Similar patterns can be observed in late MHG, as shown in () (cf. e.g. Ebert , ; Prell ; Sapp ). ()
VO orders, late MHG (fourteenth century)3 a. daz er sie loest [von irm smertzen] that he her relieves of her pains ‘that he relieves her of her pains’ (CE , – – after Ebert : ) b. da sie in der werlt waz [ein beginn] since she in the world was a beginning ‘since she was a beginning in the world’ (CE , – after Ebert : ) c. daz er in gibt [leiden und süezzikeit] that he in gives suffering and sweetness ‘that he delivers suffering and sweetness’ (AL , – after Ebert : ) d. daz si gwalticlich wider ste [allen untugenden] that she strongly resist all vices ‘that she strongly resists all vices’ (AL , – after Ebert : )
[PP]
[NPNOM]
[NPACC]
[NPDAT]
3 The examples in () all come from works of female mystics from the area around Nuremberg (CE = Christina Ebner (–), Der Nonne von Engelthal Büchlein von der Genaden Uberlast ( pages, ed. by Karl Schröder ); AL = Die Offenbarungen der Adelheid Langmann, Klosterfrau zu Engelthal ( pages, ed. by Philipp Strauch )), but similar examples can be readily found in texts by other authors from other areas.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß e. daz ez mich irret [minez slafens und ezzens] [NPGEN] that it me distract my.GEN sleep.GEN and eating.GEN ‘that it distracts me from sleeping and eating’ (CE , – – after Ebert : )
This chapter discusses a set of theoretical approaches to the OV/VO alternation in early German (with an emphasis on OHG), focusing on the question of whether it is possible to identify a basic serialization pattern (or structural configuration) that underlies the ‘mixed’ word order properties found at the syntactic surface. As in much of the literature on the topic, I will use the label OV/VO not only to refer to the relative position of nominal objects but also to describe orderings involving PPs and predicative phrases (as in (a), (d), and (a)). In traditional accounts, VO orders in the early Germanic languages are often attributed to extraposition (sometimes also referred to as ‘exbraciation’) from an OV base, that is, rightward movement of preverbal material targeting the post field (cf. e.g. Stockwell ; van Kemenade ; Koopman ; Lightfoot ; Stockwell and Minkova on OE; Lenerz and Axel on OHG; see Sapp for a recent update and further discussion of VO patterns in MHG and ENHG).4 However, there are considerable differences between early German (OHG, in particular) and Modern German concerning both the frequency and scope of this operation, which have led some researchers to propose that relevant SVO patterns should not be analysed as derived orders but rather reflect a head-initial VP (cf. e.g. Haider b; Hinterhölzl ; Schallert on OHG; Prell and Haider b on MHG).5 First, it has been repeatedly pointed out that extraposition was much more common in early German than it is in Modern German. For example, Fleischer and Schallert (: ) report that in Notker’s late OHG translations, around % of the main clauses and between % and % of embedded clauses exhibit material in the post field. Sapp (: ), based on an examination of over , clauses, estimates that the rate of extraposition in MHG and ENHG is still around %, while it has dropped to merely % in Modern German (for the latter number, Sapp relies on figures given in Lambert : ).6 Moreover, while extraposition seems to apply freely to a wide range of
4 In generative work, extraposition is traditionally modelled in terms of rightward adjunction to VP or IP. In what follows, I will use the term ‘extraposition’ as a label to refer to the relevant surface word order patterns without committing to a particular theoretical analysis. 5 An alternative approach to deriving additional VO patterns in a basic OV grammar is to assume (optional) leftward verb movement to a clause-medial position, cf. Tomaselli () and Weiß () on OHG; Fuß and Trips () on OE; see also Chapter of this volume. 6 It appears that the rate of extraposition declined rapidly in the sixteenth century. Sapp (: ) observes that in the database he compiled for his study ‘the frequency of extraposition remains stable from the twelfth through to the fifteenth century at around –%, before dropping precipitously in the sixteenth to less than %.’ Note that the rate of extraposition mentioned in this passage (over %) is higher than perhaps expected, as it has been calculated from a corpus for which Sapp selected only clauses with extraposed material, and clauses that contain constituents that can potentially be extraposed. If the complete corpus of over , clauses (containing many clauses without extraposable material) is taken into account, the instances of extraposition found by Sapp amount to only .% of all cases, which seems to be a more realistic estimate for the rate of extraposition in MHG and ENHG (cf. Sapp : ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
elements (including light NPs as in (c)) in OHG (with the possible exception of highly topical material, pronominal elements in particular), it primarily affects PPs and heavy constituents (most prominently clausal categories) in Modern German as shown in () (cf. e.g. Lambert and Zifonun et al. : ff). ()
a. Oskar hat bestätigt, [CP dass Lulu geheiratet hat]. Oskar has confirmed that Lulu married has ‘Oskar confirmed, that Lulu got married.’ b. ?Oskar hat Lulu gesehen [PP in dem Augenblick, als sie ‘Ja’ Oskar has Lulu seen in the instance when she yes sagte]. said ‘Oskar saw Lulu in the instance when she said ‘yes’.’ c. *Oskar hat gesehen [NP Lulu]. Oskar has seen Lulu. ‘Oskar saw Lulu.’ d. *Oskar ist geworden [AP rot]. Oskar is become red ‘Oskar turned red.’
Thus, it appears that in the history of German, there has been a general decrease of VO orders, which affected both the overall frequency of VO patterns and the range of elements that could occur in post-verbal position, and eventually led to a consolidation of basic OV order (a reverse change leading to a consolidation of basic VO order took place in the history of English, cf. e.g. Fischer et al. ; Pintzuk and Taylor ). These facts eventually attracted the attention of a number of theoretical linguists, who pointed out several conceptual and empirical shortcomings of the traditional analysis and developed alternative formal models of the intriguing word order flexibility found in the early Germanic languages and its subsequent loss (mostly focusing on the history of English).7 On the conceptual side, this change of perspective was fuelled by a general drive towards elimination of optional movement processes (cf. e.g. Chomsky ) and a widespread scepticism concerning the availability of rightward movement (cf. Kayne ). In addition, and more importantly, it became clear that there are good reasons to believe that at least in OE, not all instances of surface VO patterns can be attributed to extraposition of underlyingly preverbal material (cf. e.g. Pintzuk b, ; Pintzuk and Taylor ). First, there are VO patterns in OE involving post-verbal prosodically light elements such as object pronouns, monosyllabic adverbs, and verbal particles which resist extraposition across Germanic. An example is given in ().
Compare Pintzuk (b, ); Kiparsky (, , ); Roberts (); Kroch and Taylor (); Hróarsdóttir (a); Fuß and Trips (); Hinterhölzl (, ); Biberauer and Roberts (); Haider (b, ); Schlachter (), to name only a few. 7
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
()
Eric Fuß he wolde adræfan ut anne æþeling he would drive out a prince ‘he would drive out a prince’ (ChronB (T) .– () – after Pintzuk : )
Second, Pintzuk (b, ) argues convincingly that the rise of VO orders cannot be attributed to an increase in the rate of extraposition, which appears to remain constant during the OE period (see also Pintzuk and Taylor ). From these facts, she concludes that basic VO must have been available as a minority pattern in OE. Based on these findings, Pintzuk develops an alternative approach that applies the notion of ‘grammar competition’ (Kroch ) to word order variation in OE. More precisely, she assumes that the mixed OV/VO character of OE reflects a situation where speakers have access to a set of competing internalized grammars that differ with respect to the value assigned to the head parameter for V and INFL (the so-called Double Base Hypothesis (DBH), cf. e.g. Pintzuk b, on OE; Trips on ME; Santorini on Yiddish; Schlachter on OHG). According to the DBH, four different base orders/structures are logically possible. These are schematically illustrated by the (partially made-up) OE examples in (), which are based on () above. In addition to purely head-final/head-initial structures ((a) and (c), respectively) the DBH also predicts the existence of ‘disharmonic’ structures, which result from conflicting settings of the head parameter for IP and VP, as in (b) and (d).8,9 ()
a. I0 right, V0 right: S-O-V-Vfin (þæt) he [I' [VP [VP anne æþeling ut adræfan] ti ] woldei ] that he a prince out drive wanted b. I0 left, V0 right: S-Vfin-O-V (þæt) he [I' woldei [VP [VP anne æþeling ut adræfan] ti ]] c. I0 left, V0 left: S-Vfin-V-O (þæt) he [I' woldei [VP ti [VP adræfan ut anne æþeling]]] d. I0 right, V0 left: S-V-O-Vfin (þæt) he [I' [VP ti [VP adræfan ut anne æþeling]] woldei ]
Following Kroch (, ), Pintzuk assumes that the kind of linguistic variation observable in early Germanic is a characteristic trait of morphosyntactic change involving mutually incompatible parametric choices; over time, the variation eventually levels out when one structural option wins out over its competitors. This approach manages to capture the observed variation in terms of multiple
8 Alternative analyses of the mixed OV/VO character of early Germanic that also invoke a lexical head parameter include Rögnvaldsson () and Haider (b, ), who assume that the head parameter of V is left unspecified in these languages (see also Faarlund and Schallert ). 9 Option (d) is not attested in OE and possibly cross-linguistically absent, cf. e.g. Steele (); Travis (); den Besten (a); Hawkins (); Dryer (); Kiparsky (); Holmberg (a); Fuß and Trips (); Biberauer, Holmberg, and Roberts (); see also the papers in Biberauer and Sheehan ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
base-generated structural options; moreover, it is embedded within a general theory of the connection between linguistic variation and change that integrates insights of William Labov (cf. e.g. Labov ) into a generative setting. Still, Pintzuk’s analysis has been criticized as conceptually problematic and empirically inadequate (cf. e.g. Hinterhölzl and Hale ). The conceptual critique focuses on the notion of grammar competition (see Hale in particular) and the use of the traditional head parameter (which does not fit well with the assumption that parameters are linked to properties of functional categories, cf. Borer ; Ouhalla ; Kayne ; Chomsky ); empirical problems include the (unexpected) absence of the ordering (d), and the fact that Pintzuk must assume additional operations such as scrambling, verb raising, and rightward movement (extraposition) to account for the full range of ordering possibilities found in OE. However, perhaps the most serious shortcoming of Pintzuk’s analysis is that it does not seem to have much to say about the observation that the OV/VO alternation appears to be linked to factors such as weight and information-structural status. In early Germanic, the preverbal position is preferably occupied by prosodically light and/or inferable/discourse-given entities, while heavy constituents and entities newly introduced into the discourse tend to occur post-verbally (cf. e.g. Behaghel –, Vol. IV; Foster and van der Wurff ; Fischer et al. ; Schlachter , , ; Hinterhölzl ; Petrova ; Petrova and Hinterhölzl ; for a recent re-estimation of the OE facts, cf. Taylor and Pintzuk a, b, , ; see also Chapter of this volume). It is not immediately clear how these facts can be accommodated by an account that attributes different orderings to different grammars, or competing values of the head parameter (but see Schlachter for a relevant proposal for OHG); they seem to call for an analysis in terms of movement operations that are triggered by discoursesemantic factors (although it is not so clear how (prosodic) weight as a factor governing object placement can be captured along these lines, but see Hinterhölzl for a relevant proposal). Recent generative attempts to model the impact of information structure on word order in early Germanic assume that OE and OHG were so-called discourseconfigurational languages, in which word order primarily serves pragmatic functions (e.g. to distinguish old versus new information), while Modern English, for instance, uses word order primarily to mark syntactic relations (e.g. subject versus object).10 The observed linguistic variation is then analysed as resulting from leftward movement of objects and/or verbal projections to designated specifier positions linked to case licensing and the realization of topic/focus, usually adopting Kayne’s () Universal Base Hypothesis (i.e. all languages are underlyingly SVO, cf. e.g. Roberts ; Biberauer and Roberts on OE; Hróarsdóttir a on ON; Hinterhölzl , b, ; Petrova and Hinterhölzl on OHG; for details, see Chapter of this volume). Recently, however, Sapp () has argued convincingly that at least for MHG and ENHG, the traditional analysis of surface VO orders
10 Note that with respect to the ordering of NPs within the middle field, however, informationstructural factors seem to be less important in older stages of German than today, see Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
in terms of extraposition from a head-final VP is basically correct (see also fn. (this section) and fn. (Section ..)). Note that the various approaches to the OV/VO alternation in early Germanic differ with regard to the theoretical status they assign to the notion of ‘basic word order’. In generative work on word order variation, the term ‘basic word order’ is sometimes used to refer to an underlying base order/structural configuration from which the attested set of surface orders can be derived in a most economical way. As already pointed out, approaches invoking extraposition to account for VO patterns simply assume that the relevant base order is OV. In contrast, the Double Base Hypothesis in assuming competing base-generated structural options claims that there is more than a single base order. Finally, accounts cast in the Universal Base Hypothesis (Kayne ) largely dispense with the idea that the basic (surface) word order of a language corresponds to an underived underlying base order. Rather, it is assumed that both OV and VO are derived orders that result from the (obligatory) application of different syntactic movement operations that are triggered by various licensing requirements (case, information-structural distinctions, etc.). Another use of the label ‘basic word order’ is widespread in descriptive and typological work, where the term is commonly applied to an unmarked order of elements, which is used e.g. in pragmatically neutral declarative clauses, or is compatible with the largest set of linguistic contexts (cf. Lenerz ; Höhle ). Note that the identification of the basic word order in this sense is largely orthogonal to the theoretical question of whether the relevant order is base-generated or derived by syntactic operations—it is entirely possible that the unmarked order of elements does not reflect any base-generated configuration but rather results from a parameterized set of syntactic operations that apply obligatorily in each and every sentence of a given language. Still, the notion of basic word order might prove to be theoretically relevant. If we can detect a pattern which is—in contrast to other word order options—accepted in a variety of different contexts and apparently compatible with different pragmatic functions, then we might conclude that there exists indeed a discourse-semantically ‘neutral’, unmarked word order in early Germanic and that the OV/VO alternation is not completely determined by discourse-related factors; if that turns out to be true, the hypothesis that the early Germanic languages were ‘discourse-configurational’ perhaps cannot be maintained in its strong form. Furthermore, the notion of ‘basic word order’ might play a role in the theory of parametrization if a number of seemingly unrelated surface properties (e.g. verb raising, adverb placement, scrambling, position of verbal particles, etc.) can be plausibly attributed to a single ‘macro parameter’ (in the sense of Baker , ) that is linked to basic serialization properties (cf. e.g. Saito and Fukui ). We can thus formulate the following research questions concerning the status of word order variation in the early Germanic languages: ()
Is it possible to identify a. a (underlying) base order from which additional serialization patterns can be economically derived? b. a discourse-semantically ‘neutral’, unmarked surface word order?
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
Of course, it might well be that the unmarked surface word order corresponds to the underlying base order (as e.g. in early generative work where SOV was identified as both the underlying and unmarked word order in early Germanic, cf. e.g. van Kemenade on OE, and Lenerz on OHG). However, I would like to stress that a definitive answer to those questions is beyond the scope of this chapter, the primary goal of which is to present and discuss methods than can be used to address (). Of course, the question of which order counts as ‘basic’ (in the sense of either (a) or (b)) should receive an answer based on linguistic arguments and not solely on superficial inspection. This is particularly clear in the case of the early Germanic languages, which exhibit less consistent word order properties than their present-day relatives. In what follows, I will therefore review a set of relevant diagnostics that have been suggested in the literature and discuss their application to historical stages of German (see also Schallert ).
. OV/VO diagnostics This section presents a set of diagnostics that can be used to explore the questions in (). In line with what has been said above, we must distinguish two types of tests that relate to the different interpretations of the term ‘basic word order’. Sections ..–.. explore criteria for identifying underlying orders/structures, while Sections .. and .. focus on tests that can be applied to determine the unmarked order of elements at the syntactic surface. To avoid confusion, I will use the term ‘base order’ to refer to the underlying order, while the term ‘basic word order’ is used refer to the unmarked surface word order. As already hinted at above, the notion ‘basic word order’ does not imply a certain mode of analysis, including the possibility that the ‘basic’ serialization of constituents in a clause is actually not a base-generated sequence but rather derived by the application of syntactic movement operations.11 .. Unambiguous cases In the literature on the OV/VO alternation in early Germanic, there is general agreement that there are a number of clear-cut cases from which the basic OV or VO character of a language can be deduced (cf. e.g. Pintzuk ). These include the relative ordering of verb and auxiliary, and the position of verbal particles. Orders where the finite auxiliary follows the non-finite verb, as in (a), or where verbal particles precede the main verb, as in (b), are commonly linked to an OV base. The rationale behind this reasoning is that finite auxiliaries always precede the lexical verb in the Germanic VO languages, which do not permit reordering of elements in the verbal complex. Likewise, OV languages exhibit only preverbal particles, while in the Germanic VO languages, verbal particles generally follow the lexical verb:12 11 Note that it is sometimes difficult to decide on purely empirical grounds whether a given surface ordering corresponds to a base-generated or a derived syntactic representation. See the following sections for further discussion. 12 But note that English exhibits a residue of quite a number of verbs such as offset, overcome, outrank, undergo, etc. which exhibit a preverbal particle-like element that cannot be separated from the verbal part.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
()
Eric Fuß Indicators of an OV base a. (Comp)–S–(XP)–V–Aux–(XP) b. (Comp)–S–(XP)–Prt–V–(XP)
Under the hypothesis that the orders in (), which are robustly attested in both OE and OHG, signal a basic OV character, material that follows the verb (giving rise to surface VO patterns) must be assumed to have undergone extraposition from an underlyingly preverbal position (cf. e.g. Pintzuk on OE; Axel on OHG; Sapp on MHG and ENHG; but see e.g. Biberauer and Roberts for an alternative analysis of relevant orders). What about properties that unambiguously signal a VO base? Unfortunately, the picture is less clear here. The fact that OV languages generally allow verb raising, that is, reordering of elements in the verbal complex, seems to lessen the value of the order of auxiliary and lexical verb as a diagnostic test for the base order (but see Section .. for discussion). Moreover, as already mentioned above, surface VO orders may be the result of a stylistic operation placing material in clause-final position (extraposition). However, one might argue that a (uniformly) post-verbal placement of particles is a clear indicator of a basic VO character (apart from V effects, reordering of verbal particles giving rise to V-Prt is usually not possible in the Germanic OV languages): () Indicator of a VO base (Comp)–S–(Aux)–V–Prt–(XP) In what follows, I will first discuss a number of additional tests for identifying the base order of elements, including the position of elements that resist extraposition, the properties of periphrastic verbal constructions, the existence of a special class of immobile complex verbs, and the application of quantitative analyses. We are then going to take a closer look at diagnostics for basic word order, focusing on the unmarked serialization of the verb and its complements, and the significance of deviations from word order properties of the source text in translations.13 13 Note that there are a number of further morphosyntactic properties that have been linked to a headfinal VP, including the possibility of scrambling/‘free’ word order (cf. e.g. Saito and Fukui ), the use of case affixes to mark grammatical functions, a strong preference for suffixing inflections (Bybee, Pagliuca, and Perkins ; Julien ), the licensing of wh-in-situ strategies (cf. Kayne ; Julien ), and the possibility of final complementizers (cf. e.g. Bayer ). These potential diagnostics will not be discussed here, either because they do not apply to early Germanic (e.g. final complementizers), or because they can also occur with surface VO order and are therefore only of limited use as diagnostics for an OV base (scrambling, wh-in-situ, inflectional suffixes, case marking). Another potential diagnostic test that we will not discuss in detail concerns the position and linear ordering of event-related adverbs (Haider , ; Hinterhölzl , ). In VO languages such as English, (event-related) adverbs of time, place, and manner typically follow the verb in a specific sequence, cf. (ia). In contrast, OV languages like German exhibit the reverse ordering in preverbal position:
(i) a. VO: V – Manner – Place – Time b. OV: Time – Place – Manner – V (ii) a. that Peter worked [carefully] [in the office] [yesterday] b. dass Peter [gestern] [im Büro] [sorgfältig] gearbeitet hat that Peter yesterday in.the office carefully worked has
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
.. Elements that resist extraposition As already briefly mentioned above, the Germanic OV languages do not permit extraposition of prosodically light elements such as pronouns, (verbal) particles, and monosyllabic adverbs, compare the German examples in ()–().14 ()
a. dass der Student sie im Kino küsste that the student her in.the cinema kissed ‘that the student kissed her in the cinema’ b. *dass der Student im Kino küsste sie
() a. dass der Student sie dort/dann/oft küsste that the student her there/then/often kissed ‘that the student kissed her there/then/often’ b. *dass der Student sie küsste dort/dann/oft () a. dass der Student aufsteht that the student up.stands ‘that the student is standing up’ b. *dass der Student steht auf In contrast, these elements may (or must) occur in post-verbal position in the Germanic VO languages, as shown by the English translations of ()–(). Accordingly, examples like those in () have been used as evidence suggesting that VO base order was a structural option in OE (Pintzuk b, ). () a. . . . swa þæt hy asettan [him] upp on ænne sið so that they set them up in one journey ‘ . . . so that they transported themselves inland in one journey’ (ChronA . () – after Pintzuk : ) b. . . . he wolde adræfan [ut] anne æþeling he would drive out a prince ‘. . . he would drive out a prince.’ (ChronB (T) .– () – after Pintzuk : ) c. þæt martinus come [þa] into þære byrig that Martin came then into the town ‘that Martin then came into the town’ (ÆLS .– – after Pintzuk : ) However, note that the possibility of particle shift in VO languages such as English (Bolinger ; Haider ; Svenonius ; Dehé ) complicates the picture: () a. Peter tore off the cap. b. Peter tore the cap off. 14
Pintzuk’s conclusion that pronouns, short adverbs, and verbal particles do not undergo extraposition in OE is based on the observation that these elements do not show up in post-verbal position in unambiguous OV orders (e.g. orders of the type XP-V-Aux), cf. Pintzuk (b, ) for details.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
On the surface, the result of particle shift is similar to OE examples which Pintzuk () analyses in terms of an OV base + verb movement to a clause-medial INFLnode, where the position of the particle is taken to reflect the base position of the verb: () þæt he wearpi [þæt sweord] onweg ti so.that he threw that sword away ‘so that he threw away the sword’ (Bede . – after Pintzuk : ) Moreover, one must take into account that the Germanic VO languages do not exhibit a uniform behaviour with regard to particle shift. In contrast to English, Swedish requires strict adjacency of verb and particle (cf. e.g. Haider a), while in Danish, the shifted order seems to be the only acceptable option (Herslund ): ()
Swedish a. att han kastade bort mattan that he threw out carpet.the b. *att han kastade mattan bort ‘that he threw out the carpet’
()
Danish a. *Boris skrev under kontrakten. Boris wrote under contract.the b. Boris skrev kontrakten under. ‘Boris signed the contract.’
In addition, it is not entirely clear whether particle verbs can receive the same analysis in VO and OV languages (cf. e.g. Haider , a; Svenonius for discussion). However, if we follow Haider (, a), who argues that particle shift is available only in VO languages, the final position of the particle in examples like (b) or (b) does not create a problem for the use of particle placement as a diagnostic test for basic word order. Thus, we arrive at the following conclusions:15 () a. The order [non-finite verb – verbal particle] is indicative of a VO base. b. The order [verbal particle – non-finite verb] is indicative of an OV base. c. The order [non-finite verb – object – verbal particle] is compatible with both an OV and a VO analysis. d. The availability of particle shift is indicative of a VO base. 15
In many OV varieties, we can observe that verbal particles can shift to the left of a higher (finite) verb in the verbal complex (sometimes called ‘cluster creepers’, Evers ), compare the following examples from Dutch (Neeleman and Weerman : ): (i) a. dat that
Jan het meisje wil Jan the girl want
opbellen on.ring
‘that Jan wants to call the girl’ b. dat Jan het meisje op wil bellen Similar phenomena can be observed in historical stages of German (cf. Behaghel –, Vol. IV: –) and present-day dialects (cf. Schallert and Schwalm ). Since this type of reordering within the verbal complex seems to be confined to (the Germanic) OV languages, it might be used as another diagnostic for a head-final VP.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
Bearing this in mind, let us now turn to the distribution of light elements in early German. At first sight, it appears that examples with post-verbal pronominal elements can also be found in early OHG texts. This is shown in (). () a. (et scies quia dominus exercituum misit me ad te) dhazs uuerodheoda druhtin sendida [mih] zi dhir that the.armies’ Lord sent me to you ‘that the Lord of Hosts sent me to you’ (Isidor , ) b. (ut subiciam ante faciem eius gentes) dhazs ih fora sinemu anthlutte hneige [imu] dheodun that I before his face subdue him nations ‘that I might subdue nations before Him’ (Isidor , ) However, upon closer inspection it becomes apparent that in the above examples, the word order of the OHG translation is very similar to the word order of the Latin source. Therefore, cases such as () do not constitute clear evidence in favour of the existence of a VO base order option in OHG.16 Still, as has been pointed out by Dittmer and Dittmer () and Schallert () among others, there are some cases where post-verbal placement of a pronominal element (in most cases a reflexive pronoun) cannot be attributed to properties of the source text. Relevant examples are given in () and (). () (& qui se humiliat exaltabitur) inti therthar giotmotigot [sih] uuirdit arhában and who.there humbles REFL will.be lifted up ‘and he who humbles himself will be exalted’ (Tatian , –) () a. (si duo ex uobis consenserint super terram de omni re) oba zuuene fon íu gizuftigont [sih] obar erdu if two of you.PL.DAT agree REFL on earth fon iogilicheru rachu of all things ‘if two of you on earth agree about anything’ (Tatian , ) b. (ut diligatis Inuicem) thaz ir minnot [iuuuih] untar zuuisgen that you.PL.NOM love you.PL.ACC under each other ‘that you love each other’ (Tatian , )
16 Dittmer and Dittmer (: ) count seventy-two cases where post-verbal placement of object pronouns mimics the Latin word order in embedded clauses of the OHG Tatian translation.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
In (), the post-posed reflexive pronoun sih corresponds to a preverbal element in the Latin source; in (), the Latin text does not contain elements corresponding to the pronouns occurring in post-verbal position in the OHG translation. Schallert () argues that examples similar to () and () suggest that OHG was characterized by a mixed OV/VO grammar, similar to OE (see also Haider b, and more recently Haider ). However, it should be pointed out that post-verbal placement of pronominal elements is very rare in OHG and seems to be confined to the earliest translations.17 In Dittmer and Dittmer’s (: ) work on the OHG Tatian translation, they mention only three cases lacking a Latin model where an object pronoun occupies a post-verbal position in an embedded clause; in contrast, they list embedded clauses where an object pronoun is inserted in or transferred to a preverbal position. In general, there seems to be a strong tendency for pronouns to occupy a position at the beginning of the middle field, very similar to Modern German (see also Hinterhölzl and Petrova , and Chapters and of this volume). What about the other light elements that are used by Pintzuk () as a diagnostic for a basic VO configuration? The position of verbal particles supports the conclusion that OHG has an OV base. As pointed out by Axel (: ), ‘in contrast to Old English, in OHG there are hardly any cases with post-verbal particles attested in subordinate clauses with particle verbs.’ Similar observations seem to hold with regard to the position of light adverbs such as thô, dhar, or nû. A relevant search conducted in the Isidor and Tatian (using the TITUS and Kali online corpora) did not produce a single example where these light elements follow a non-finite verb, or a finite verb in an embedded clause.18 Based on the same line of
17 In addition, notice that in the majority of cases with post-verbal pronouns, there is additional material in the post field (e.g. two PPs in (a)). This might be taken to suggest that examples like () actually involve extraposition of a larger phrase (e.g. a remnant VP) that contains the reflexive pronoun in addition to other arguments/adjuncts. Note that related phenomena can be found in present-day German in connection with apparent multiple fronting to clause-initial position:
im Spiegel] hat (i) ?[Sich himself in.the mirror has ‘He saw himself in the mirror.’
er gesehen. he seen
In Notker’s OHG translations, we can find at least some examples that exhibit post-verbal placement of particles characteristic of VO languages, compare the following minimal pair (see also Section .. of this volume): 18
(i) a. taz er beiz imo selbemo aba dia zungûn that he bit him self off the tongue ‘that he bit off his tongue’ (Notker, Consolatio , –) b. ter imo selbemo dia zungûn aba / who him self the tongue off ‘the one who bit off his tongue’ (Notker, Consolatio , –)
beiz bit
The order in (ia) seems to suggest that basic VO was an option at least in Notker’s language. However, note that patterns where an object (or other material) intervenes between the finite verb and a verbal particle might also be analysed as resulting from leftward movement of the finite verb (plus extraposition of the direct object in cases like (ia); cf. Pintzuk on OE examples like () in Section ..).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
reasoning, Sapp () concludes from the virtual absence of light elements (both short adverbs and pronouns) in post-verbal position that the MHG and ENHG VP was systematically head-final and that there is therefore no evidence for a parametric change from VO (or, mixed VO/OV) to OV after the OHG period (i.e. since about ). .. Order in the verbal complex The assumption that V-Aux order is indicative of a basic OV character (see above) goes back at least to the work of Greenberg (), who notes that there is a close connection between the basic word order of a language and the relative order of verbal elements (in the verbal complex). In particular, in OV languages, embedded (non-finite) verbs precede the matrix verb, while finite auxiliaries follow the main verb (Universals and ).19 Further support for the assumption that a final placement of finite auxiliaries signals an OV base comes from both diachronic and typological facts suggesting that a VO base requires a clause-medial position for finite auxiliaries (i.e. a clause-medial INFL/T-node): first, it has been observed that the order VO-Aux (i.e. a combination of a head-initial VP embedded under a head-final IP/TP) is cross-linguistically very rare, if not completely absent (cf. fn. in Section .; see Biberauer, Holmberg, and Roberts for an analysis in terms of the so-called Final-over-Final-Constraint (FOFC)). This observation is linked to a generalization concerning possible pathways of word order change, namely the claim that the development of a clause-medial INFL/T-position is a necessary precondition for a change from a head-final VP to a head-initial VP (cf. Kiparsky ; Pintzuk ).20 Thus, it seems that a head-final VP is compatible with both a final and a medial position for auxiliaries, while a head-initial VP requires the auxiliary to occur in clause-medial position (i.e. to the left of the non-finite verb). Examples like () which display a verbal complex with the finite verb in absolutely final position are often taken to manifest the predominant OV character of early OHG. () (QUIA DE TRIBU IUDA SECUNDUM CARNEM CHRISTUS EXPECTANDUS ESSET) . . . BIHUUIU MAN IN JUDASES CHUNNES FLEISCHE CHRISTES why one in Judah.GEN tribe.GEN flesh Christ BIDENDI UUAS expecting was ‘ . . . why one was expecting Christ in the flesh of the tribe of Judah.’ (Isidor , title)
19 However, note that there are some rare exceptions to Universal (finite auxiliaries follow the main verb in OV languages) such as Early Middle English (EME), where we can observe S-Aux-O-V patterns (Pintzuk ). 20 See Biberauer, Newton, and Sheehan () for an attempt to derive this generalization from the FOFC.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
However, it is a well-known fact that alternative serializations of selecting and selected verb are possible in most stages and varieties of German (cf. e.g. Maurer ; Behaghel –, Vol. IV; Bech ; Ebert ; Robinson ; Schmid and Vogel ; Schmid ; Axel ; Sapp a, b; Weiß and Schwalm ; see also Chapters. , , and of this volume). In the generative literature, these are commonly referred to as Verb Raising (VR) and Verb Projection Raising (VPR), reflecting the original analysis of Evers () in terms of right-adjoining a dependent verb or verbal projection to the (finite) matrix verb. The application of VR/VPR is dependent on a number of factors, including tense, type of selecting verb, and number of elements in the verbal complex. For example, in varieties like Zurich German, the auxiliary follows the participle in the perfect tense, while finite modals (in the present tense) precede the dependent infinitive (Lötscher : f): () a. wil mer em Hans es velo geschänkt hand since we the.DAT Hans the bicycle given have ‘since we gave Hans the bicycle as a present’ b. wil mer em Hans es velo wänd schänke since we the.DAT Hans the bicycle want give ‘since we want to give Hans the bicycle as a present’
[VR]
The term VPR is commonly used to refer to cases where the verbal complex contains a non-verbal element, for example one or more nominal objects, as in (a). Note that in the present-day varieties that allow this kind of reordering, VPR is generally ruled out with prosodically light elements such as weak object pronouns as can be seen in (b).21 () a. dass er will [VP em Peter es Gschänk gää] that he wants the.DAT Peter the present give ‘that he wants to give Peter the present’
[VPR]
b. *dass er will [VP em es Gschänk gää] that he wants him the present give ‘that he wants to give him the present’ (Zurich German, Cécile Meier, p.c.) On the one hand, the possibility of VR and VPR might be taken to cast some doubt on the value of the serialization of verbal elements as a diagnostic test for the underlying order, since there are languages like Dutch and certain Swiss German and East Hessian varieties (cf. e.g. Hodler : ; Weiß and Schwalm ) which combine an OV base with a preference for Aux-V order in the verbal complex. On the other hand, it has also been argued that across Germanic, this kind of reordering in the verbal complex (and so-called ‘verb-cluster formation’) is confined to the class
21
However, it has been pointed out that even in Standard German, other relatively light elements such as pronominal adverbs (e.g. darauf ‘there-on’) may occur in VPR orders, cf. e.g. Kefer and Lejeune () for relevant examples.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
of OV languages and should therefore be directly linked to the basic OV character of these languages (Haider ; Vikner ; however see Koopman and Szabolcsi , and Hinterhölzl for analyses that derive verbal complexes in Germanic from a VO base).22 In historical stages of German, VR seems to be more widespread than in Modern Standard German (but cf. e.g. Kolmer ; Weiß and Schwalm on the (re-)ordering possibilities in present-day dialects). For example, Robinson () shows that in the Isidor translation, we not only find the pattern V-Aux exemplified in (), but also quite frequently the VR pattern Aux-V (often against the order found in the Latin source).23 In the examples in (), the Latin source has the finite auxiliary in clause-final position, while the OHG translation exhibits the order Vfin-Vinf (see fn. for three-verb clusters). The fact that the translation deviates from the order found in the original supports the conclusion that VR was a well-established trait of the grammar of early OHG (see also Chapter ). () a. (quod enim homo factus est) dhazs ir man uuardh uuordan . . . that he man was become ‘that he became a man’ (Isidor , )
22 Another important difference between left- and right-branching verbal complexes has to do with the fact that the former are always compact, while in the latter, non-verbal material may intervene between the verbal components of the complex (i.e. V-(*XP)-Vfin versus Vfin-(XP)-V). As it is not immediately clear how this asymmetry can be captured by approaches that derive left-branching orders from a VO base, it seems to provide potential support for alternative modes of analysis (e.g. in terms of the traditional headparameter, cf. Sapp b for discussion). 23 Further ordering possibilities can be observed in clusters consisting of two non-finite verbs and a finite auxiliary. According to Robinson (: ), in three-verb clusters the subordinate non-finite verb must always precede the matrix non-finite verb, while the auxiliary can occupy any position in the verbal complex (the latter is marked by boldface in the following examples), see also Bolli (); Näf (); Sapp (b) (cf. e.g. Schmid and Vogel ; Schmid for the situation in present-day varieties of German).
(i) V-V-V fona huueliihhemu ædhile christ [chiboran uuerdhan scoldi] from which nobility Christ born be should ‘from which noble lineage Christ was to be born’ (Isidor , ) (ii) V-V-V dher dhar [scoldii chiboran uuerdan] who there should born be ‘who was supposed to be born there’ (Isidor , ) (iii) V-V-V dhazs ir in sines edhiles fleische [quhoman scolda uuerdan] that he in his nobility’s flesh come should be ‘that He would come in the flesh of his noble line’ (Isidor , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß b. (Probauimus dominum nostrum iesum christum secundum carnem iam natum fuisse) Chiuuisso chioffanodum uuir nu hear [dhazs unser druhtin certainly prove we now here that our Lord nerrendeo Christ after dheru fleiscliihhun chiburdi iu the.saviour Christ after the fleshly birth already uuardh chiboran] was born ‘Certainly, we proved here now that our Lord, the saviour Jesus Christ was born through fleshly birth.’ (Isidor , )
Still, under the assumption that VR is actually a property linked to a basic OV character, these data do not conflict with the conclusion that OHG was an OV language, at least as long as we also find corresponding ‘unraised’ orders with the finite verb in final position, which can be taken to represent the head-final source of the VR variant. Thus, it seems that the existence of VR can be used as an argument in favour of OV, but not against it. What about the use of VPR orders as a diagnostic test for basic word order? Recall that VPR is subject to a restriction that bans certain prosodically light elements (pronouns in particular) from occurring inside the verbal complex. Interestingly, as pointed out by Pintzuk (), cases of apparent VPR that involve pronominal objects do turn up in OE: () þæt heo wolde hine læran that she would him teach ‘that she would teach him’ (ÆLS . – after Pintzuk : ) Pintzuk analyses examples like () in terms of a head-final VP in combination with leftward movement of the finite verb to a clause-medial INFL/T head and concludes that examples like () provide further empirical support for her hypothesis that OE was characterized by a mixed OV/VO character, where in principle both settings of the head parameter were available for both INFL/T and V. VPR patterns, where a non-verbal constituent intervenes between the elements of the verbal complex, are also attested in (early) OHG:24 () (Dum enim audis deum unctum, intellege christum.) dhazs dhar ist [Christ] chizeichnit that there is Christ described ‘that Christ is described there’ (Isidor , ) 24 According to Robinson (), the presence of a non-verbal element inside the verbal complex always implies verb raising, that is, the order V-XP-Aux is not attested in the OHG Isidor (similar restrictions for VPR can be observed in the modern Germanic languages; see also fn. , and Sapp b; Schallert b). Note that this is reminiscent of the absence of the order V-O-Aux (i.e. the FOFC), see the beginning of this section.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
In addition to nominal arguments, the position between the two verbal elements can be occupied by (relatively light) elements such as quantifiers, demonstratives, and adverbs: () ( . . . quando a patre per illum cuncta creata esse noscuntur) dhazs fona dhemu almahtigin fater dhurah inan ist [al] uuordan, that from the.DAT almighty father through him is all become dhazs chiscaffanes ist that created is ‘that everything that was created came to be from the Almighty Father through him’ (Isidor , ) () (Numquid de illo salomone creditur prophetatum? minime) Neo nist zi chilaubanne dhazs fona dhemu salomone never NEG.is to believe that of the Solomon sii [dhiz] chiforabodot is this prophesied ‘It can never be believed that this was prophesied by Solomon.’ (Isidor , ) () (sanctus sanctorum dominus iesus christus olim uenisse cognoscitur) dhazs dher allero heilegono heilego druhtin nerrendeo Christ that the most of.the.Holy holy Lord saviour Christ iu ist [langhe] quhoman already is long come ‘that the holiest of the holy, Christ the Lord has already come’ (Isidor , ) Note that elements such as demonstratives, quantifiers, and adverbs can also occur in VPR patterns in varieties like Zurich German (Cécile Meier, p.c.). Crucially, however, there do not seem to be any cases of VPR with object pronouns in OHG, unlike OE. Thus, despite the fact that the relevant elements in ()–() are relatively light, the above examples cannot be used as evidence supporting the existence of a clause-medial INFL/T-position in OHG. To summarize, it seems that we can draw the following conclusions concerning the use of serialization patterns in the verbal complex (involving two verbs) as a diagnostic test for underlying word order:25 25 Another characteristic property of the Germanic OV languages that is linked to the possibility of VR is the so-called IPP (infinitivus pro participio) effect (for details see Chapter ). In certain contexts, we can observe that the past participle of a small set of verbs (typically causatives, modals, and perception verbs, see also Schmid for a comprehensive overview) is replaced by an infinitive. As shown in (i), this effect is intimately linked to VR, giving rise to the order V-V-V in the right sentence bracket:
(i) a. dass Peter das Buch hat lesen that Peter the book has read ‘that Peter had to read the book’ V V b. *dass Peter das Buch lesen
müssen/*gemusst must.INF/read.PART V
müssen/gemusst hat
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
() a. V-Vfin is indicative of an OV base. b. Vfin-V is found in both OV and VO languages. However, only OV languages exhibit an alternation between V-Vfin and Vfin-V (i.e. VR). c. VPR is indicative of an OV base (but possibly a first step towards VO, cf. Kiparsky ; Pintzuk ; Kaufmann ; Biberauer, Newton, and Sheehan ). d. VPR patterns that involve object pronouns suggest the presence of a clausemedial INFL/T head (cf. Pintzuk ; see also Chapter of this volume). .. Immobile complex verbs The Germanic OV languages exhibit a class of ‘immobile’ complex verbs that cannot undergo verb fronting in main clauses and show a tendency to occur in non-finite forms (cf. e.g. Grimm : ff; Åsdahl Holmberg ; Wurzel ; H. Koopman ; Eschenlohr ; Fortmann ; Vikner ; Freywald and Simon ; Ahlers ). Relevant examples involving the verb uraufführen ‘to premiere’ are given in () and (). () a. Sollten sie das Stück uraufführen? should they the play original.on.put.INF ‘Should they premiere the stage play?’ b. dass sie das Stück uraufführten that they the play original.on.put.PAST.PTCP c. dass sie das Stück uraufgeführt haben that they the play original.on.put.PAST.PTCP have () a. *Uraufführten sie das Stück? original.on.put.PAST they the play ‘Did they premiere the play?’ b. *Aufführten sie das Stück ur? on.put.PAST they the play original c. *Urführten sie das original.put.PAST they the d. *Führten sie das Stück put.PAST they the play
Stück auf? play on urauf? original.on
However, the IPP cannot be observed in the oldest attested stages of German. According to Paul (: ), the first examples where a pure infinitive replaces a past participle are attested in thirteenth century MHG (see also Chapter of this volume; Behaghel –, Vol. II; Dal ; Ebert et al. ; Fleischer and Schallert : Chapter ; Sapp a). Moreover, it seems that the preference for V-V-V in combination with the IPP effect is a rather recent innovation. Sapp (a: f) shows that in ENHG, the order V-V-V is outnumbered by the order V-V-V (% versus % of all IPP examples). Of course, this raises the question of how and why the new ordering developed and how it spread to virtually all present-day dialects of German. In addition, it is still unclear how the morphological effects of the IPP, that is, the transition from a past participle to an infinitive, developed historically (see Chapter for discussion).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
Verbs that are confined to the right sentence bracket are often analysed as the result of back-formation (Rückbildung) or conversion operating on complex nominal bases. Relevant examples include verbs with a separable and a nonseparable prefix/particle (e.g. German uraufführen ‘to premiere’, auferstehen ‘to resurrect’, wiedereinführen ‘to reintroduce’, Dutch herinvoeren ‘to reintroduce’), verbs with an incorporated (usually nominal) element (German bergsteigen ‘to mountain-climb’, bauchreden ‘to stomach-speak (to ventriloquize)’, bausparen ‘to build-save (to save with a building society)’, preiskegeln ‘to prize-bowl’, bruchlanden ‘to crash-land’, kopfrechnen ‘to head-calculate (to do mental arithmetic)’, sonnenbaden ‘to sunbathe’; Dutch bergklimmen ‘to mountain-climb’, buikspreken ‘to stomachspeak’, bouwsparen ‘to build-save’, prijsschieten ‘to prize-shoot’, etc.), and verbs with two (in principle) separable (voranmelden ‘to preregister‘) or non-separable prefixed elements (strafversetzen ‘to penalty-transfer (to transfer sb. to another position as a punishment)’, or zweckentfremden ‘to purpose-estrange (to use something for a wrong/not originally intended purpose)’). Interestingly, the Germanic VO languages seem to lack a corresponding class of verbs with deviating syntactic properties (cf. Vikner ). Similar back-formations are very rare (compare Danish *bjergbestige ‘to mountain-climb’, *bogliospare ‘to build-save’, *bugtale ‘to stomach-speak’, *solbade ‘to sun-bathe’); in case relevant complex verbs turn up at all, their syntactic distribution does not differ from the distribution of other verbs. This is illustrated with the Danish verb planlægge ‘to planlay (to plan)’ in () (cf. Vikner : ): () a. Hvorfor kunne de ikke planlægge at holde konferencen her? why can they not plan.lay to hold conference.the here ‘Why can’t they plan to hold the conference here?’ b. Planlægger de at holde konferencen i Reykjavík? plan.lay they to hold conference.the in R. ‘Do they plan to hold the conference in Reykjavík?’ c. *Lægger de plan at holde konferencen i Reykjavík? lay they plan to hold conference.the in R. Arguably, the existence of such a class of complex verbs with a peculiar syntactic distribution is bound to the OV property. So if the early Germanic languages can be shown to exhibit a similar class of verbs, this might be considered corroborating evidence supporting the conclusion that these languages were basically OV. Interestingly, it turns out that at least one relevant type of complex verb, namely forms with an incorporated noun (also sometimes referred to as ‘pseudo-compounds’), is robustly attested in OHG and MHG (apparently in contrast to early English). According to Åsdahl Holmberg (), the dictionaries consulted by her (Raven / for OHG; Lexer –, and Jelinek for MHG) list at least seventy entries for OHG and around for MHG, including forms such as OHG halswerfôn ‘to neck-turn’, muotbrehhôn ‘to courage-break’, fuozfallôn ‘to foot-fall’, and MHG luginstrâfen ‘to lies-punish (to prove somebody wrong)’, vederslagen ‘to
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
feather/wing-beat’, nôtzogen ‘to need-tear (to rape)’, or radebrechen ‘to wheel-break (to mangle (a language)’).26 It is not clear, however, whether these OHG and MHG verbs exhibit the restricted syntactic distribution of their present-day counterparts. Åsdahl Holmberg (: ) notes that finite verbal pseudo-compounds commonly appear in main clauses of OHG and MHG, which suggests that these verb forms could undergo verb fronting, in contrast to the ‘immobile’ complex verbs of the present-day language. Unfortunately, Åsdahl Holmberg does not support this statement with OHG or MHG examples, or quantitative evidence. A relevant example from Notker’s Psalms, in which a finite (imperative/adhortative) verbal pseudocompound occurs in clause-initial position, is given in (). () (Omnes gentes, plaudite manibus:) Hantslagot, alle diete, . . . hand.clap.PL all nations ‘all nations, clap your hands’ (Notker, Psalter , ) We may conclude, then, that the status of OHG/MHG ‘immobile’ complex verbs and thus their usability as a diagnostic for a basic OV syntax remains open to question, pending further studies of their syntactic distribution.27 .. Quantitative evidence In recent years, the use of (advanced) quantitative and statistical methods has turned into a fruitful avenue of research in historical linguistics.28 Of particular interest for our present purposes is the application of quantitative evidence to distinguish between derived orders (which might be linked to discourse-semantic distinctions like information structure) and base orders. An instructive example of this approach applied to the OV/VO alternation in early Germanic is Kroch and Taylor (). They observe that EME exhibits variation between surface VO and OV patterns. The latter occur with an average rate of c. % with non-quantified objects. However, at first sight it is unclear whether preverbal placement of objects results from scrambling from a VO base, or reflects an OV base order. To calculate the rate of derived OV orders (resulting from leftward scrambling), Kroch and Taylor measure the rate of scrambling in examples like () which exhibit the order NP-V-object pronoun and are used as a diagnostic test for leftward movement of objects from a VO base.
26
At least in OHG, all these complex verbs belong to the weak declension. Note that more generally, the syntactic distribution of complex verbs in OHG seems to slightly differ from what we find in present-day German; for instance, there seems to be a tendency to move the verb together with its (today separable) prefixes/particles to the left sentence bracket, cf. e.g. Kuroda () on the Tatian. 28 For relevant generative work cf. e.g. Kroch (); Santorini (); Pintzuk (, b, , ); Kroch and Taylor (, ); Haeberli (); Pintzuk and Taylor (); Taylor and Pintzuk (a, b, , ); Ecay (). 27
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
() For alle þeo þe habbeð [ani good]i idon me ti for all those that have any good done me ‘for everyone who has done me any good’ (CMANCRIW, I.. – after Kroch and Taylor : ) Under the assumption that the rate of scrambling in examples like () is representative of the rate of scrambling in general, Kroch and Taylor estimate the average rate of scrambling with non-quantified objects to be about % (while it amounts to % with quantified objects29). From the discrepancy between the frequency of (unambiguous cases of) scrambling and the rate of all surface OV patterns (around %), Kroch and Taylor conclude that not all surface OV patterns can be analysed as derived orders resulting from leftward movement of the object (contra e.g. Roberts ). Thus, it seems that OV did exist as a base order option in EME (in addition to VO).30 A similar line of argument is used by Pintzuk () to decide whether OE examples like () should be analysed in terms of VR as illustrated in (a), or movement of the finite verb to a clause-medial INFL-head as in (b).31 () þæt se eorðlica man sceolde geþeon so.that the earthly man should prosper ‘so that the earthly man should prosper’ (ÆCHom i.. – after Pintzuk : )
29 The fact that the rate of OV with quantified objects is considerably higher is commonly taken to suggest that leftward movement of quantified objects (possibly for reasons of scope taking) was a regular grammatical option in EME, cf. Kroch and Taylor () for details. 30 See Taylor and Pintzuk (a, b, ) for the use of similar quantitative methods to assess the influence of information structure on object placement in OE. Taylor and Pintzuk show that in subordinate clauses with finite main verbs, there is a tendency for new information to appear in post-verbal position, while objects representing given information tend to occur preverbally. In addition, object placement is affected by independent factors such as the prosodic weight of the constituent. Taylor and Pintzuk argue that the loss of OV order cannot be attributed to an overuse of VO order as a focus-marking strategy. Rather, it appears that VO order increases at the same rate for all kinds of objects, both given and new, during the OE period. They conclude that the development of an increasingly fixed object position diminished the role of information structure as a factor determining word order. This suggests that the loss of information structure-related word order variation resulted from an independent change that led to a more rigid word order (the loss of ‘free’ word order is often attributed to the loss of case marking, cf. e.g. Sapir ; see also Roberts on the loss of OV order in the history of English; notable problematic cases include Icelandic with rich case morphology and a fairly rigid word order and Lower German with only two morphological cases and free reordering of material in the middle field (scrambling)). On the role of information-structural factors for OV/VO in historical German see also Chapter . 31 Note that the auxiliary can also occur in clause-final position as shown in (i). That is, VR is merely optional in OE.
(i) hwæt se bisceop don wolde what the bishop do would ‘what the bishop would do’ (ÆLS . – after Pintzuk : )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
() a. þæt se eorðlica man [VP ti sceolde geþeoni] [VR] b. þæt se eorðlica man [IP sceoldei [VP geþeon ti]] [V-to-INFL] To estimate the average rate of VR, Pintzuk measures the frequency of VR in examples like (), which involve at least two constituents to the left of the verb and which she takes to be unambiguous instances of OV. As shown by (b), VR is optional in this context. () a. swa swa sceap from wulfum & wildeorum ti beoð fornumenei just.as sheep by wolves and beasts are destroyed ‘just as sheep are destroyed by wolves and beasts’ (Bede . – after Pintzuk : ) b. þe se ealdormon wiþ hiene gedon hæfde that the alderman against him done had ‘that the alderman had done against him’ (Or .– – after Pintzuk : ) Based on this test case, Pintzuk calculates that the average rate of VR is about %. If all surface VR orders were genuine cases of VR, then the frequency of VR in examples like () should be similar to the frequency calculated by looking at cases like (a). However, it turns out that the pattern Aux-V is much more frequent (.%) in examples like (), which are ambiguous between VR and leftward movement of the auxiliary. From this observation, Pintzuk concludes that many of the surface VR orders are actually the result of leftward movement of the finite verb in OE. These two case studies provide an example of how quantitative analyses can be employed to discover patterns of grammar beneath the surface in cases where the historical evidence at first sight does not seem to be decisive.32 In particular, the analysis of the frequency distribution of word order alternants might tell us something about whether a certain word order pattern represents an unmarked/basic or a marked/derived option (cf. e.g. Dittmer and Dittmer on the OHG Tatian). As shown above, deviations from the average rate with which a certain order/pattern appears might be taken to suggest that the same surface pattern (in the cases above OV, or Vfin-V) results from a different set of operations in a given, specific context. Moreover, if it can be established that a certain word order option is clearly favoured under certain (pragmatic/syntactic) conditions, then we might conclude that this pattern does not represent a base order but rather an order derived by the application of certain syntactic rules/operations (cf. e.g. van der Wurff , ; Kroch and Taylor ; Hróarsdóttir a; Ingham ; Wallenberg on the position of negated and quantified expressions in early Germanic). It is worth mentioning that the feasibility of quantitative studies on historical syntax is greatly
32
Note that the approaches discussed above make use of fairly basic (descriptive) statistical methods like measuring relative frequencies and calculating p-values. For the use of advanced statistical methods in historical syntax cf. e.g. Ecay ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
enhanced by the availability of parsed corpora; for instance, it has often been pointed out that research on the syntax of early English has received a massive boost from the availability of the Penn Corpora of Historical English. Building comparable syntactically parsed electronic corpora for historical stages of German largely remains a desideratum to this day (however, large morphologically annotated and POS-tagged corpora of historical German have recently become available, see Chapter for an overview). Still, quantitative methods have been successfully put to use to investigate the impact of information structure on word order in a number of recent studies on OHG (cf. e.g. Axel ; Petrova ; Petrova and Hinterhölzl ; Schlachter ). Most recently, Sapp () has used quantitative evidence33 to argue convincingly that extraposition in MHG and ENHG does not differ qualitatively from the corresponding construction in Modern German, thereby challenging the view that basic SVO order was an option in MHG (pace Prell and Haider b). So far, I have discussed a set of tests that can be used to determine the underlying base order of elements. We will now shift our focus to the notion of ‘basic word order’ and the question of whether it is possible to identify an unmarked surface word order pattern in OHG. .. Basic order = unmarked order Traditionally, the basic word order of a language has often been identified with the ‘dominant’, that is, the most frequent serialization pattern in a given corpus of utterances (cf. e.g. Greenberg ; see Dryer for critical discussion).34 In more recent approaches, however, the notion of basic word order is often linked to the notion of unmarked order, where ‘unmarked order’ can be defined as () or (): ()
The word order option which is compatible with the largest set of different linguistic/information-structural contexts (cf. e.g. Lenerz ).
()
The word order option which shows up in thetic (‘all-new’) declarative clauses such as presentational and existential sentences, which contain no presupposed/topical information and which typically serve to establish the existence of an entity or a situation (cf. e.g. Sasse ).
33 Based on an investigation of over , embedded clauses selected from the Referenzkorpus Mittelhochdeutsch and the Bonner Frühneuhochdeutschkorpus, Sapp (: ) concludes that Modern German extraposition ‘aside from decreased frequency, is largely similar to the medieval construction,’ affecting the same set of elements (apart from clausal constituents, mostly PPs and heavy or focused NPs are placed in the post-field). 34 A more refined definition is given in Siewierska (: ), who assumes that the basic word order of a given language is to be identified with the serialization pattern found in ‘stylistically neutral, independent, indicative clauses, where the subject is definite, agentive, and human, the object is a definite semantic patient, and the verb represents an action, not a state or an event.’ Note, however, that even this careful formulation does not seem to capture basic differences between V languages like Swedish and German, due to the fact that the respective VO versus OV character of these languages is obscured by general verb fronting in main clauses.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
The definition at () requires a quantitative analysis of the distribution of the relevant word order options across different information-structural environments in a given corpus (cf. e.g. van Kemenade and Los ; Taylor and Pintzuk a, b, , on OE, and the papers in Hinterhölzl and Petrova ; and Ferraresi and Lühr ; see also Chapter of this volume). There seem to be some indications that in OHG, OV qualifies as the unmarked surface order in this sense. For example, in Chapter it is shown that in a corpus of that-clauses, OV is found in over % of all cases (plus % that invite an analysis in terms of OV+ extraposition (XP . . . V-Aux XP )), while unambiguous instances of VO order are pretty rare (less than %). Thus, OV is the most frequent pattern and clearly satisfies the traditional criterion for being classified as the dominant word order option. Moreover, it is unlikely that all instances of OV order involve arguments referring to given or inferable discourse referents, in particular if it is taken into account that there is a cross-linguistic tendency for objects to refer to new information, while subjects typically represent given information (cf. e.g. Birner and Ward ; see Westergaard on possible consequences for word order change). And in fact there are cases where the preverbal position hosts discourse-new material in OHG. A search conducted in the Tatian Corpus of Deviating Examples, which is annotated for information structure, has produced a number of examples like () where the new information focus (or parts of it) is placed in preverbal/prefinite position: () a. (quod uisionem uidiss& In templo) thaz er [gisiht] gisah in temple that he vision saw in temple ‘that he saw a vision in the temple’ (Tatian , ) b. (a terra reducere pussilum) thaz her íz [fon erdu] / arleitti ein luzzil that he it from land push/move a little ‘that he push it [the boat] from the land/shore a little bit’ (Tatian , – ) c. (matheum nomine) ther [matheus] uuas giheizan who Matthew was named ‘who was named Matthew’ (Tatian , ) Similar cases can be found in the OHG Isidor (cf. Schlachter : f). Other early OHG texts such as the Muspilli already show an even more consistent OV syntax where discourse-new regularly occurs in preverbal position. This can be seen in (). () a. daz der man haret ze gote enti imo [hilfa] niquimit that the man hopes to God and him help not.comes ‘that the man hopes for god and help is not coming to him’ (Muspilli )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
b. daz Elias in demo uuige aruuartit uuerde so daz Eliases pluot that E. in the battle wounded will.be so that E.’s blood [in erda] kitriufit in earth drips ‘that Elias will be wounded in that battle so that Elias’s blood drips down to the soil’ (Muspilli –) Thus, it seems that there is no one-to-one relation between information structure and the linear position of arguments; preverbal objects can be linked to different information-structural properties, while VO order is typically used to introduce new referents into the discourse (cf. Petrova and Hinterhölzl ; Schlachter ; and Chapter ). These observations support the conclusion that OV represents the basic word order option in OHG (cf. Sapp for related conclusions concerning MHG and ENHG). However, note that these findings do not necessarily force us to assume that OV represents the base-generated order from which marked alternatives are derived via additional operations such as extraposition/rightward movement. The empirical facts seem to be equally compatible with a Kaynean analysis where ‘basic’ OV orders result from a set of operations that apply in the unmarked case (e.g. raising a vP containing the object to SpecTP, cf. Biberauer and Roberts on OE), while VO orders require some additional machinery linked to the realization of information-structural distinctions (e.g. moving the object to a focus position prior to vP raising, or lack of vP raising/pied-piping, cf. e.g. Biberauer and Roberts ; Hinterhölzl b; Petrova and Hinterhölzl ; and Chapter of this volume). To illustrate criterion (), take a look at the Modern German examples in () and (), which do not contain any material that has already been introduced into the discourse, and are commonly taken to represent the basic order of elements in the German middle field (cf. e.g. Lenerz ):35 () Was ist passiert? ‘What happened?’ Oskar hat [IO einem Studenten] [DO ein Buch] gegeben. Oskar has a.DAT student.DAT a.ACC book given ‘Oskar gave a book to a student.’ () Es hat ein Professor [IO einer Studentin] [DO einen Kuss] gegeben. it has a professor a.DAT student.FEM a.ACC kiss given ‘A professor gave a student a kiss.’ However, it is not entirely clear whether thetic sentences provide a reliable diagnostic test for basic word order in (supposedly) discourse-configurational languages such as OE or OHG. It is a well-known fact that there are languages where thetic sentences exhibit special properties that set them apart from basic declaratives (cf. e.g. Lambrecht ). For example, it has been observed that presentationals and 35
For a discussion of the order of NPs within the middle field in historical German see Chapter .
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
existentials tend to exhibit V order in OHG, as illustrated in (a) and (b) (cf. e.g. Lenerz ; Axel ; Petrova ). In addition, it seems that the subject tends to appear in absolutely clause-final position in thetic sentences (cf. Lambrecht , and Gast and Haas for the claim that the sole argument of presentational sentences must receive stress and preferably occurs late in the clause). () a. (/ . . . & facta est/tranquilitas magna.,/) . . . uuard tho gitan/ mihhil stilnessi.,/ were then made great calm ‘and there was a great calm’ (Tatian , –) b. Stúant tho thár umbiríng / fílu manag édiling stand then there around many noblemen ‘There stood many noblemen around.’ (Otfrid I , ) In a language where word order is primarily used to convey informationstructural distinctions, we would perhaps expect that there is a tendency for objects (representing new information) to occupy a post-verbal position in transitive presentational and existential sentences. Thus it might turn out that thetic sentences cannot tell us much about the base order of elements in early Germanic and that () does not constitute a useful notion of ‘unmarked order’, at least for the languages under investigation. On the other hand, if we could show that objects occupied a preverbal position in thetic sentences in OHG, this could be taken to suggest that word order in early Germanic was not primarily (let alone solely) governed by information-structural properties.36 Unfortunately, relevant examples are not easy to come by, which has to do with the fact that transitive presentationals and existentials are quite rare. Moreover, due to the fact that the finite verb undergoes regular fronting to the left clausal periphery in all stages of German, diagnostics for the basic ordering of the verb and its complement(s) such as () and () can only be applied to (i) embedded clauses or (ii) main clauses with periphrastic verb forms. Together with the fact that historical linguists have only a limited amount of data at their disposal, these restrictions create some practical problems, since thetic clauses are typically main clauses, and periphrastic constructions are only infrequently found in early OHG (see Chapter ). So it might turn out that thetic clauses are simply too rare in the corpus to warrant solid conclusions. .. The significance of translations: deviations from the source text The use of translations as sources of historical linguistic evidence is a double-edged sword. On the one hand, it is a well-known fact that translations (and translations of religious texts, in particular) are often heavily influenced by the linguistic properties
36 Note that, for the order of NPs in the middle field in OHG, information structure apparently did not play much of a role, either, but only became a decisive factor in later stages of German (see Chapter ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
of their sources, which diminishes their overall value as a source of linguistic information. On the other hand, it is generally agreed upon that (systematic) deviations from the source text are particularly informative, since they can be taken to represent instances where the translator/scribe consciously decided to depart from the Latin or Greek original in order to obey core properties of the target language (for OHG cf. Fleischer ). In this section, we will discuss the use of translations as a means to infer (basic) word order properties, primarily drawing on the work of Dittmer and Dittmer () on the OHG Tatian translation (more recent work on OHG that also highlights the significance of translations includes Axel ; Fleischer, Hinterhölzl, and Solf ; Schlachter , ; Petrova ; Petrova and Solf a; Petrova et al. ). Even a cursory look at word order differences between the OHG Tatian translation and the (presumed) Latin source reveals that despite some amount of variation, OV is the dominant pattern in OHG. Dittmer and Dittmer (: ff) observe a strong tendency in embedded clauses to move material that appears post-verbally in the Latin source into preverbal position in the OHG translation. In the parts examined by Dittmer and Dittmer (: ), they find such cases in embedded clauses. Relevant examples are given in ()–(). In contrast, they mention only twelve instances where a constituent appears post-verbally against the Latin original (plus four cases where an object that lacks a corresponding Latin form is inserted in postverbal position). ()
Subject a. (ut circumcider&ur [puer]) thaz [thaz kind] bisnitan uuvrdi that that child circumcized was ‘that the child was circumcized’ (Tatian , ) b. (cum autem dormirent [homines]) thó [thie man] intsliefun when the men fell.asleep ‘when the men fell asleep’ (Tatian , ) c. (sicut constituit [mihi] [dominus]) so [mir] [trohtin] gisazta as me Lord told ‘as the Lord told me’ (Tatian , )
()
Object (qui hab& [sponsam] sponsus est) ther [brut] hab& ther ist brutigomo the.one bride has the.one is bridegroom ‘He who has the bride is the bridegroom’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
() PP (qui est [ex deo] uerba dei audit) ther [fon gote] ist ther horit gotes uuort the.one from God is the.one hears God.GEN word ‘Whoever belongs to God hears God’s word.’ (Tatian , ) Particularly frequent are cases where a post-verbal Latin object pronoun is rendered by a preverbal object pronoun in the OHG version ( times in embedded clauses, according to Dittmer and Dittmer : ; in addition, there are forty-nine instances where an object pronoun that lacks a Latin model is inserted into the middle field). These include examples where additional material remains in postverbal position, as in (). () a. (quem ego mittam [uobis] a patre) then ih [íu] senti fon themo fater who.ACC I you.PL.DAT send from the.DAT father ‘whom I shall send to you from the Father’ (Tatian , ) b. (sicut praecepit [ei] angelus domini) só [imo] gibôt thruhtines engil as him commanded Lord.GEN angel ‘as the angel of the Lord commanded him’ (Tatian , ) Thus, it appears that already in OHG, there is a strong tendency to avoid postverbal placement of object pronouns in embedded clauses, very similar to the present-day Germanic OV languages (however, see () and () in Section .. for some exceptions).37 Table . gives an overview of the differences and similarities between the Tatian and its Latin source observed by Dittmer and Dittmer () for embedded clauses (compare their table on page ). The table (.) shows that differences between the word order of the OHG translation and its Latin source typically lead to OV patterns. The insertion of material that lacks corresponding Latin forms by and large follows the same pattern. Still, it is not entirely clear how to
37 This conclusion is corroborated by cases where light elements that occupy a position in the verbal complex in the Latin original are transferred to a position to the left of the verbal complex in the OHG translation:
(i) (quia possum [hoc] facere uobis) thaz ih íu [thaz] tuon mugi that I you.PL.DAT that do can (Tatian , ) Example (i) is particularly interesting because it features a number of differences between Latin and OHG, all conforming to a basic OV order. In addition to thaz moving from an interverbal to a preverbal position, the dative pronoun íu has been transferred from post-verbal to preverbal position, and a subject pronoun has been inserted. Moreover, note that the order of elements in the verbal complex has undergone a change from V-V (i.e. matrix verb precedes embedded verb) to V-V.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
T . Position of major constituents in embedded clauses of the Tatian: OHG versus Latin preverbal position preverbal position post-verbal post-verbal position (6¼ Latin) (= Latin) position (6¼ Latin) (= Latin) newly inserted subj. pron. subj. NP obj. pron. obj. NP adverb P+pron. P+XP Total
– –
moved
newly inserted
– – – – –
moved – –
–
(Dittmer and Dittmer : )
deal with the small number of cases (all in all sixteen) where the V-XP order found in the OHG translation cannot be attributed to Latin influence. While examples such as () might be amenable to an analysis in terms of PP extraposition (on a par with Modern German), post-verbal placement of (light) NPs (and pronouns, see () and () in Section ..), as illustrated in () (involving an extraposed subject) and (), seems to raise a problem.38 ()
PP a. (qui [in caelis] es) thu thar bist [in himile] you there are in heaven ‘who is in heaven’ (Tatian , ) b. (ut ihesum [dolo] tenerent) thaz sie then heilant fiengin [mit feihan] that they the.ACC saviour caught with guile ‘that they caught the Saviour by trickery’ (Tatian , –)
38 It is perhaps true that we cannot entirely rule out the possibility that VO patterns that correspond to the word order of the Latin source represented a native option in OHG (cf. e.g. the right-most column in Table .). But since it is always possible that the translation simply mimics properties of the Latin source in these cases, most researchers agree that relevant one-to-one correspondences should not be used to draw firm conclusions about the grammar of OHG (cf. e.g. Fleischer ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
()
Subject (per quem [scandalum] uenit) thuruh then quimit [asuuih] through the.ACC comes offence ‘by whom the offence comes’ (Tatian , )
()
Object a. (qui [demonia] habebant) thie thar hab&un [diuual] who.PL there have.PL devil ‘those who were possessed by the devil’ (Tatian , ) b. (qui [deum] non timebat) ther niforhta [got] who NEG.fears God ‘who did not fear God’ (Tatian , )
So, while much of the evidence available points towards the conclusion that OHG was basically an SOV language (or, at least clearly more ‘OV-ish’ than e.g. OE), there remains a (small) set of data that seems to suggest that OHG cannot be analysed on a par with the Modern Germanic OV languages (where argument NPs usually cannot undergo extraposition). It is unclear whether the attested VO patterns represent the residue of an earlier historical stage with a genuine mixed OV/VO character (cf. e.g. Schallert ) or should be attributed to the workings of movement operations that could be put to use to achieve certain communicative/pragmatic effects, but have decreased considerably in subsequent stages of German (cf. Lenerz ; Hinterhölzl ; and more recently Sapp ).
. Conclusion In this chapter, I have reviewed a set of diagnostic tests for underlying and basic word order and discussed their application to the early Germanic languages, OHG, in particular. We have seen that OHG exhibits a number of surface properties that are reminiscent of the present-day Germanic OV languages. Apart from a general predominance of preverbal object placement, these include the existence of preverbal verbal particles/derivational affixes, V-Aux order in the verbal complex (which might be subject to VR and VPR), and a strong tendency to avoid post-verbal placement of light elements that resist extraposition such as pronouns, or light adverbs (in contrast to e.g. OE). These findings are supported by quantitative evidence involving deviations from the Latin source text in the Tatian translation, which exhibit a general preference for preverbal object placement (this preference is particularly strong with pronominal elements). So we arrive at a point where we are confronted with conflicting pieces of evidence: on the one hand, the conclusion seems to be warranted
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The OV/VO alternation in Old High German
that OHG is more ‘OV-ish’ than other early Germanic languages. In particular, there are significant differences between OHG and OE, which suggests that we should not aim at a uniform analysis of the two languages. On the other hand, we have also seen that in comparison with e.g. present-day German, OHG exhibits a larger array of surface VO orders. The majority of VO orders seems to involve constituents that are either heavy, or refer to newly introduced discourse referents (or both); in addition, there is a (very) small number of examples that exhibit post-verbal given/light elements. Still, if we were to apply a label to OHG, then it would be rather SOV than SVO, or ‘mixed’ SOV/SVO. This conclusion is further supported by the impression that the attested word order variation cannot be fully attributed to information-structural factors: While (a subset of) VO orders seem to be linked to length/new information, there are good reasons to believe that OV order is the unmarked case, which is compatible with a larger set of linguistic/pragmatic contexts. Of course, this intuition requires further examination in the form of in-depth quantitative studies that link word order and information-structural status. Furthermore, note that there is evidence suggesting that certain OV orders involving pronominal and (possibly) contrastively focused elements result from additional syntactic movement operations (see Chapter ). As already pointed out repeatedly above, the statement that OHG was basically SOV does not necessarily entail a specific mode of analysis. Thus, while quantitative evidence can be used to distinguish between unmarked (i.e. basic) and marked word order options, it does not seem to be possible to distinguish on purely empirical grounds whether surface SOV orders (apart from those that e.g. involve a ‘high’ position of pronominal elements at the left edge of the middle field) are basegenerated or derived by a set of syntactic operations. In other words, we might say that in principle, the empirical evidence is compatible with both a traditional account in terms of underlying (i.e. base-generated) SOV order plus extraposition (or multiple base-generated orders, cf. Haider b, ; Schallert ), and an analysis that treats basic SOV as an order derived from a (universal) SVO base, with additional mechanisms to account for SVO patterns (but see e.g. Pintzuk ; Fuß for a critical assessment of analyses of OE based on Kayne’s Universal Base Hypothesis). An adequate theoretical model should capture the fact that there are not only significant differences between OHG and OE but also between OHG and the present-day OV languages. In particular, something has to be said about the observation that a good deal of word order variation is linked to factors such as weight and information structure. Last but not least, a theoretical analysis of OHG should provide an answer to the question of how and why these additional word order options got lost in the transition to Modern German. Let me conclude this section with a brief sketch of what such a theory might look like. The impression that OHG had already moved away from a genuine ‘mixed’ OV/VO grammar towards a more consistent OV character, with VO orders representing the residue of this former stage, seems to call for an analysis in terms of grammar competition, where an older grammatical system is gradually replaced by an innovative parametric option. However, for the reasons already mentioned, the Double Base Hypothesis does not seem to be an adequate choice to model the kind of variation we
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Eric Fuß
observe in OHG. A more promising approach is to combine the notion of grammar competition with the idea that different word orders are not base-generated, but correspond to derived structures (which might be linked to discourse-semantic distinctions). Note that this synthesis does not imply a certain mode of analysis, but is in principle compatible with various approaches including leftward object movement, vP/VP raising/pied-piping, rightward movement (possibly at PF, cf. Truckenbrodt ), or principles and parameters governing the phonological realization of syntactic structures (cf. e.g. Richards ; Fuß on a phonological version of the head parameter, Wurmbrand and Bobaljik on extraposition as Spell-Out of lower copies of movement). Under this perspective, VO orders no longer existent in present-day German can be analysed as minority patterns generated by a grammar/parametric option that eventually gave way to a consistent OV grammar. The latter change was possibly driven by the fact that OV orders could serve a number of distinct functions (marking topichood, contrastive focus, etc., see Chapter ), which at some point blurred the original discourse-semantic function of OV order. A change along these lines can be modelled in terms of ‘fossilization’ of movement operations formerly triggered by pragmatic/information structure-related factors as purely syntactic movement (cf. Simpson ; Fuß for the concept of word order change in terms of fossilized movement).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
13 OV versus VO in Old High German The case of thaz-clauses SVE T LAN A PE TROV A AN D H ELM U T W E I ẞ
. Introduction As already discussed in Chapter , word order, and especially the order of the verb and its complements is a basic and much debated issue in the historical treatment of German syntax. As outlined in much literature, this order is variable in that nominal objects can surface both to the left and to the right of their verbal head. Furthermore, in clauses with complex verb forms, non-finite verbs may both precede and follow the finite auxiliary that selects them as complements. In this chapter, we will illustrate this variation in clauses introduced by thaz ‘that’ from OHG and will survey the analyses proposed in the literature. Consider the data in (a) and (b). In (a), the surface order of the constituents in the OHG sentence displays the assumed canonical OV order in that the verb tuome ‘judge’ follows its complement, the object uueralt ‘world’. In contrast, in (b), the governing verb habet ‘have’ precedes its object sibba ‘peace’, giving rise to VO order on the surface. () a. (ut Iudic& mundum) thaz her [uueralt] tuome that he world judge.SBJV ‘that he should judge this world’ (Tatian , ) b. (ut in me pacem habeatis) thaz in mir habet [sibba] that in me have peace ‘that in me you might have peace’ (Tatian , ) Because the word orders in the OHG translations differ from those of the respective Latin originals, it is justified to assume that they follow the rules of genuine Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Svetlana Petrova and Helmut Weiß . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
OHG grammar rather than some non-native patterns (Dittmer and Dittmer ; Fleischer ; see also Chapter of this volume). Note also that the observed variation in the order of the verb and its complements in historical German is by no means exceptional but rather a well-known property of other early Germanic languages like OE and ME (van Kemenade ; Roberts ; Pintzuk , ; Kroch and Taylor ; Fuß and Trips ; Biberauer and Roberts ) and ON (Hróarsdóttir a, : – and the references therein), see also Chapter . In traditional philological work on German, the variation exemplified in () has been accounted for by assuming that the word order in dependent clauses in historical German is not fixed in the way it is in Modern German, in that earlier stages are more hospitable to word orders in which the verb is not in clause-final position but in a position which is generally called ‘later than second’ (Späterstellung), see Sonderegger (: ) or Schrodt (: ). According to this view, the verb-final order in dependent clauses, which is consolidated as the canonical word order pattern towards Modern German, represents merely a tendency rather than a rule in the historical records. Such approaches remain vague concerning the representation of the underlying structure of the clause in historical German as well as concerning the syntactic changes that took place in the later development. These questions have been placed in the centre of attention in the generative literature, in which the explanation of word order variation and change in Germanic has become one of the most intriguing and rich discussions in the formal investigation of Germanic diachronic syntax. The various approaches put forward in the literature are discussed in detail in Chapter . The present chapter focuses on the word order variation found in OHG thaz-clauses from the perspective of the theoretical discussion on the OV/VO alternation in early Germanic. Following corpus work by Weiß (), some very basic empirical issues underlying the theoretical debate on clause structure in historical German will be addressed. In particular, we will present the variety of surface orders that can be found in historical German and will discuss, on the basis of diagnostic evidence, the underlying structure(s) as well as the operations that lead to the derivation of the attested orders. The explanations considered in this chapter will be couched either within the standard account of a basic OV grammar allowing for various movement operations of verbs and XPs both to the left and to the right, or within an approach arguing for grammar competition allowing for variation in the head-complement parameter, plus additional movement operations to various functional projections in the structure of the clause. An alternative explanation couched within the Universal Base Approach proposed by Kayne () will be discussed in Chapter (see also Chapter ).
. Overview of surface patterns and their derivation in OHG The examination of verb-object orders in OHG in this section builds on data originally presented and analysed in Weiß (). This corpus study surveys the word orders in thaz-clauses in the minor OHG documents included in Köbler’s () Sammlung kleinerer althochdeutscher Sprachdenkmäler, an anthology comprising thirty-eight OHG texts of different genres, both translations as well as original
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German
T . Number and percentage of word order patterns in thaz-clauses in the corpus of Weiß () a total n=
.%
b .%
c .%
d .%
e .%
f .%
g .%
h .%
text. In addition, the corpus also contains the thaz-clauses from the OHG Isidor which were already collected and investigated by Robinson (). Factors that might play a role in the interpretation of particular examples, e.g. metrical demands, the order in the original, and other philological issues, are taken into account in each respective case, but in general, the varying character of the source texts helps to guarantee that the identified word orders are representative of the OHG tradition in general, and not of particular texts only. The corpus study includes subordinate clauses, both adverbial and complement ones, introduced by thaz ‘that’ or a combination of thaz with prepositions (like unz thaz ‘until that’). The choice for such a dataset is based on well-known methodological considerations. In this dataset, verb movement to C is excluded because the C domain of the clause is overtly filled. Eight different surface patterns in OHG thaz-clauses can be distinguished. Taking into consideration the position of finite main verbs (Vfin), finite auxiliaries or modal verbs (Aux), and non-finite verbs (V) relative to the remaining constituents in the clause (subjects S, objects O, adverbials Adv, PPs, and predicative nouns/adjectives, generalized as XP), the individual patterns can be represented in the way outlined in ().1 The total number of instances as well as their percentage in the entire corpus is given in Table .: () a. b. c. d. e. f. g. h.
C . . . XP . . . { Vfin/V Aux} C . . . XP . . . { Vfin/V Aux} XP C . . . Aux V C . . . Aux XP V C S Vfin O C O S Vfin C Vfin XP C Aux V XP
Weiß () provides an analysis of these orders and seeks to detect the underlying structure from which they are derived. In resolving structural ambiguities which are ubiquitous in the data, he applies well-known diagnostics, like the position 1 The patterns in () will be described in detail in the following sections. There, it will become clear, for example, that the [C S O Vfin] order falls under pattern (a), which means that XP also comprises both subjects and objects. Note that, in these patterns, subjects and objects will be distinguished only where relevant. In other cases, they are subsumed under the label XP.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
of various immovable elements or evidence from modern Germanic languages displaying properties of OV and VO, discussed in Chapter . Given the derivation of the individual orders in (), three main groups of patterns can be distinguished: () a. orders which can be analysed within a pure OV base order plus additional extrapostition (a) and (b) b. orders which are ambiguous between an analysis within a pure OV base plus additional extraposition, and the assumption of the presence of an additional functional projection vP which is optional with basic OV but obligatory with basic VO (c), (d), and (h) c. orders which can be analysed as involving leftward verb movement to a functional projection higher than vP (e), (f), and (g) In what follows, we will survey the analyses of the three basic groups outlined in (). .. Analysis within a basic OV order ... Evidence for pure OV order A group of examples, or .% of the thaz-clauses found in the data displays a word order in which the verb is in absolute clause-final position, following all remaining constituents in the clause. Some of the examples, like (a), include only the subject and the verb and thus cannot be regarded as conclusive for the examination of verb-complement orders. But other examples are more telling. For example, we find thaz-clauses in which all lexical constituents (b), and even sentential constituents (c), precede the verb in final position: () a. unz dez [iz] blode until that it bleed.SG.PRES.SBJV ‘until it bleeds’ (Basler Rezepte ) b. (ut faciam uoluntatem tuam) dhazs [ih] [dhinan willon] duoe that I your.ACC will.ACC do.PRES.SBJV ‘that I should fulfil your will’ (Isidor , ) c. (prius quam gallus cantet, ter me negabis) daz [aer danne hano chrait] [drim spurtim] [miin] laucnis that before cock cries three times me.GEN.SG deny.SG ‘that [you] will deny me before a cock cries three times’ (Monseer Fragmente XXIII, –) This shows that the verb-final order, which is canonical in dependent clauses in Modern German, is well represented in thaz-clauses in the OHG period. ... Orders derived by extraposition within a basic OV structure In another relatively large group of examples (a total of thirty-four instances or .% of the data), the verb is not in clause-final position because a single non-verbal constituent, e.g. a subject, an object or a PP, follows:
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German
() a. C . . . V Aux S2 (ut consummetur praeuaricatio) dhazs chiendot uuerdhe [dhiu aboha ubarhlaupnissi] that stopped was this false believing ‘that this misbelieving be brought to an end’ (Isidor , ) b. C . . . Vfin O dhazs dhu firstandes [heilac chiruni] that you understand.PRES.SBJV holy secret ‘that you might understand the holy secret’ (Isidor , ) c. C . . . V Aux O daz er kitarnan megi [tato dehheina] that he hide can.SG.PRES.SBJV deed.GEN.PL some ‘that he could hide some deed’ (Muspilli ) d. C . . . Adv . . . Vfin O daz ih ane din gipot nespricho [nohein uuort] that I without your command NEG.say no word ‘that I say no word without your command’ (Psalm CXXXVIII ) e. C . . . Vfin PP daz der man haret [ze gote] that the man shouts to God ‘that the man shouts to God’ (Muspilli ) Note that we can assume that the attested orders are examples of the phenomenon of extraposition, i.e. the constituents that follow the verb are base-generated in clause-internal position but have been moved to the right for independent—e.g. prosodic or information-structural—reasons, while the verb remains in its base position in the end of the clause. In other words, these orders are analysed in analogy
2 In Weiß’s () corpus, the order C . . . V Aux S occurs only once, in the sentence given in (a). The attentive reader will observe that the subject follows the verb in the Latin original as well, thus raising the question as to whether subject extraposition is a genuine OHG feature. Note that conclusive examples are found in texts that were not considered in Weiß (), for example in (i), suggesting that extraposition of lexical subjects was generally possible in native OHG grammar:
(i) (Et ecce angeli accesserunt) giengun thô zuo gotes engila went there to Lord’s angels ‘The angels of the Lord appeared’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
to clauses in which a sentential constituent is moved to the right of the verb because it is syntactically and prosodically complex,3 see (): () thaz ih nigilaupta [thaz ih gilouben scolta] that I NEG.believed what I believe should ‘that I did not believe what I was supposed to believe’ (Lorscher Beichte –) There is additional evidence in support of this analysis. In (a)–(c), the clause contains elements which in OV languages immediately precede the verb in final position, for example uninflected verb forms. Adverbials are also useful in detecting verb movement. They usually occupy designated positions in the VP, including the left edge of the VP (on the position of different classes of adverbials in Modern German, see Frey and Pittner ). An adverbial as a diagnostic element is contained in (d). The fact that the verb follows one of these constituents proves that it has not left its position at the right edge of the clause, rather, the constituent that follows the verb at the surface has undergone movement to the right. This derivation is most straightforward in (e) because the respective post-verbal element is a PP, a category which typically undergoes rightward movement in OV languages including Modern German. In sum, the surface orders found in this group of examples are derivable via rightward movement of XPs within an underlying OV structure as represented in (). ()
thaz . . . ti . . . Vfin [XPi]
For this reason, the patterns in this group of data cannot be taken to provide conclusive evidence of basic VO orders in OHG. .. Ambiguous surface orders In another group of examples, accounting for a total of twenty-five instances or .% of the data, we witness violations of the canonical word order in clauses involving a complex verb form, i.e. the clause contains a finite auxiliary or a modal verb (Aux) and an uninflected main verb (V). In OV languages, the main verb which is a complement of the finite auxiliary precedes it in basic orders (see also Chapter ). This is also attested in OHG, see (): () thaz ig iz cosan muozi that I it choose must.PRES.SBJV ‘that I might choose it’ (De Heinrico ) But in the OHG data, we find dependent clauses in which the finite auxiliary surfaces in clause-medial position, to the left of the main verb. There is some
3 Note that this is a mainstream interpretation of the structural position of dependent clauses, which has been challenged in more recent theoretical approaches. A number of conceptual and empirical arguments have been used to reject the extraposition scenario and support that idea that complement clauses are basegenerated to the right of their governing head, see Haider (a) on a summary of the discussion and Axel-Tober () on clues coming from the diachronic development of subordinate clauses in German.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German
additional variation with respect to the position of the remaining constituents. In (), which represents the order in (c), Aux immediately precedes V, but all additional constituents precede the verbs. In (), an additional nonverbal constituent intervenes between Aux and V, as represented in (d), and finally, in (), there are non-verbal constituents following Aux V, as formulated in (h): ()
a. dáz ih daz godes lóp niuuolda giloson that I the God.GEN admiration NEG.wished hear ‘that I refused to hear the admiration of our Lord’ (Reichenauer Beichte –) b. dhazs ir man uuardh uuordan that he man AUX become ‘that he became human’ (Isidor , )
() a. daz thu maht forasago sin that you may prophet be ‘that you might be a prophet’ (Christus und die Samariterin ) b. (quando a patre per illum cuncta create esse) dhazs fona dhemu almahtigin fater dhurah inan ist that by the.DAT omnipotent father through him AUX al uuordan, dhazs chiscaffanes ist all become that created is ‘that everything that has been created has been made by the omnipotent father and through himself ’ (Isidor , ) () daz er mac ginerian / ze imo dingenten man that he can save to him believing man ‘that he can save the man who believes in him’ (Bittgesang an Sanct Peter ) Variation in the order of the verbal complex is a well-known phenomenon attested in various modern Continental West Germanic OV dialects, for example in Swiss German (Bader and Schmid ), Bavarian (Weiß : –), Hessian (Weiß and Schwalm ), and West Flemish (Haegeman and van Riemsdjik ; Haegeman ; see also Chapter of this volume). The processes linked to the observable variation in the verbal complex are labelled Verb Raising (VR) in () and Verb Projection Raising (VPR) (), respectively, while orders of the type in () have not received special attention in the literature on OHG. VR and VPR in modern Germanic OV dialects are commonly analysed as involving rightward movement of the VP, which may contain some non-verbal material as illustrated in (a) for the orders in (), but may also be emptied of all non-verbal material prior to extraposition, as illustrated in (b) for the orders in ():
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
() a. daz thu tVP maht [forasago sin]VP b. dáz ih [daz godes lóp]i tVP niuuolda [ti giloson]VP
VPR VR
In this sense, V(P)R is another instance of extraposition within an OV base order. In discussing word order variation and change in OE, Pintzuk (, ) presents orders of the type C . . . Aux XP V which force an alternative interpretation (see also Chapter ). These are clauses in which the VP that appears to the right of Aux contains elements excluded from extraposition and VPR, like pronouns and light adverbs (see Chapter ). These patterns are summarized in (). The example in (a) shows that VPR with one of the elements listed in () is ungrammatical in Swiss German varieties that otherwise allow for VPR. Example (b) taken from Pintzuk () shows that patterns of the type in () are attested in OE: ()
C . . . Aux {pronoun, light adverb} V
() a. *das er törft en is Huus bringe that he may him.ACC into house bring.INF ‘that he is allowed to bring him home’ (Santorini : ) b. þæt heo wolde hine læran that she would him.ACC teach.INF ‘that she would teach him’ (Pintzuk : ) In Pintzuk’s interpretation, such patterns, labelled ‘apparent VPR’, are principally different from VPR in modern OV dialects and must receive a different interpretation, accordingly. Pintzuk argues that the pattern in () provides conclusive evidence for a clause-medial, head-initial IP which hosts the finite auxiliary. Evidence in favour of a head-initial IP in OE also comes from patterns in which a pronoun, a verbal particle, or a light adverb is attested to the right of a finite main verb in a dependent clause, see (a). On the assumption that these elements are excluded from rightward movement in OV languages, and the finite main verb (Vfin) moves to I categorically, the pattern in (a) is only derivable via leftward movement of V to a head-initial IP, see (b): () a. C . . . Vfin {pronoun, verbal particle, light adverb} b. þæt [IP he [I0 wearpi [VP þæt sweord onweg ti]]] that he threw that sword away ‘that he threw that sword away’ (Pintzuk : ) A pattern that is equivalent to (b) and might be taken as evidence for apparent VPR in OHG is given in (). Here, a phonologically light element, the demonstrative pronoun thiz ‘this’ intervenes between Aux and V:4
But see the discussion in Chapter for evidence that demonstratives (may) behave differently from personal pronouns. At least in varieties like Zurich German, certain light elements such as demonstratives, quantifiers, and adverbs can occur in VPR patterns (see Chapter ). 4
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German
() dhazs fona dhemu salomone sii dhiz chiforabodot that about the.DAT Salomo.DAT is.PRES.SBJV this prophesied ‘that this was prophesied about Salomo’ (Isidor , ) But the interpretation of this order along the lines of Pintzuk’s proposal is questionable, because the assumption of a head-initial IP is problematic for OHG on empirical grounds. There is no conclusive evidence of the diagnostic orders suggesting leftward movement of Vfin to head-initial I in the corpus. The few examples involving post-verbal pronouns like the one in (a) are explainable as imitations of the word order of the Latin original,5 and the example involving a post-verbal particle given in (b) is a well-known singleton from Notker’s Consolatio (Näf : ): () a. (quia dominus exercituum misit me ad te) dhazs uuerodheoda druhtin sendida mih zi dhir that army.GEN.PL Lord sent me.ACC to you.DAT ‘that the Lord of the armies sent me to you’ (Isidor , ) b. taz er beiz imo selbemo aba dia zungun that he bit REFL.DAT self.DAT PTC the tongue ‘that he bit off himself the tongue’ (Notker, Consolatio , ) Additionally, the model of word order variation in OE proposed by Pintzuk (, ) has been challenged on conceptual grounds. As discussed in Chapter , one of the basic claims of this model is that the surface variation in OE results from synchronic variation in the headedness both of the VP and the IP. Fuß and Trips () propose a modification to the model of grammar competition that gets rid of the problem of overgeneralization. First, they assume that functional projections, in contrast to lexical ones, are uniformly head-initial. Second, they present evidence suggesting that leftward movement of Vfin in dependent clauses in OE targets the head of the functional projection vP that is below TP (corresponding to IP in Pintzuk’s model, but auxiliaries and modal verbs are not base-generated here but in a lower VP above VP). Given this, all orders attested in OE are derivable by assuming variation in the headedness of the VP plus an additional verb (projection) raising which is optional with basic OV but categorically required with basic VO. Weiß () shows that the model proposed by Fuß and Trips () can be straightforwardly extended to OHG. Two types of clause structure must be assumed for OHG. The first one involves pure head-final VP and no additional projection vP, as shown in ()=(). This is the relevant base structure for deriving verb-final dependent clauses, including those with extraposed constituents:6 5 For examples involving pronominal arguments in the OHG Tatian translation and their relation to the Latin source see Petrova (: ). 6 Note that Weiß () assumes that VR is also derivable from this order, because the inversion of Aux and V occurs in PF, on prosodic grounds.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
()
TP Spec
T’ T0
VP1 V1’
S
V10
VP2 V20
O (thaz) ig (Weiß 2006: 10)
iz
cosan
muozi
If the structure in () is enriched by an additional node vP, overt movement of Aux to v results in the order C . . . Aux XP V, which accounts for patterns of the type in (d), incl. those of apparent VPR, as shown in ()=(a): ()
TP Spec
T’ T0
νP S
ν’ ν0
VP1 S
V1’ V10
VP2 O (thaz)
thû maht
t
forasago
V20 sin
t
(Weiß 2006: 11) Finally, overt leftward movement of Aux to v within a head-initial VP7 results in surface orders of the type C . . . Aux V O, as shown in ()=():
7 Note, however, that OHG provides no evidence for the orders considered diagnostic for basic VO by Pintzuk (b, ). According to her, conclusive evidence for basic VO is provided by orders in which the untensed verb V, which is immovable, is followed by a constituent excluded from rightward movement, for example a pronominal argument, a light adverb, or a verbal particle, as in (i). As discussed in Chapter , such orders are attested in OE root and dependent clauses (ii), giving reason to assume that OE has VO in the base.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German ()
TP Spec
T’ T0
νP ν’
Subj ν0
VP1 Subj
V1’ VP2
V10 V20 (daz)
er
mac
t
t
ginerian
DO ze imo dingenten man
Note, however, that we do not find enough conclusive evidence in favour of the existence of basic VO in OHG. In these terms, the structure of orders of the type in (c), (d), and (h) remains ambiguous and the proposed derivations are highly hypothetical, basically a matter of conceptual considerations.8 .. Surface orders involving obligatory verb fronting A group of thaz-clauses unified in (c) requires special attention. As will be shown below, extraposition of XP or larger categories to the right is excluded in this dataset. But the alternative, i.e. leftward movement of the verb, must target a projection that is above the landing site of the verb in the derivation of the dataset in the previous section (v). The orders summarized in this group are of three different types, depending on the type of XP that is moved to the specifier of the projection that attracts the verb. In (a), this is the subject of the clause, in (b) a non-subject, and finally, in (c) no XP fronting applies:
(i) Aux–V–pronoun/light adverb/particle (ii) he wolde adræfan ut anne æþeling he would drive out a prince ‘he would drive out a prince’ (Pintzuk b: ) Note, however, that we find no proper evidence of this type in OHG. Deriving (h) in the way proposed by Fuß and Trips () is motivated merely by conceptual considerations and not by empirical facts. 8 As an anonymous reviewer points out, this order can also be derived within a traditional OV account, by assuming VR plus extraposition of a heavy XP.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
() a. dat du habes heme herron that you have.PRES.SBJV at.home master ‘that you have a good master at home’ (Hildebrandslied ) b. dat Hiltibrant hætti min that Hildebrand was.called.PRET.SBJV my ‘that my father was called Hildebrand’ (Hildebrandslied )
goten good
fater father
c. daz sculi der antichristo mit Eliase pagan that should.PRET.SBJV the pagan with Elias fight ‘that the pagan should fight against Elias’ (Muspilli ) All examples crucially differ from the instances of extraposition discussed in Chapter . Note that (a) contains the local adverbial (heme ‘at home’) marking the left edge of the VP. This order is diagnostic for leftward movement of Vfin across the VP. We find additional evidence suggesting this type of movement is common in OHG. Consider the data in (): () a. dhazs ir selbo Christ ist chiuuisso got ioh druhtin that he himself Christ is truly God and Lord ‘that truly, he is himself God and Lord’ (Isidor , ) b. daz aer fuor danân that he went from.there ‘that he left‘ (Monseer Fragmente X, ) In all of the examples, we find a type of adverbial which occupies a designated position in the VP. Example (a) displays a sentence adverbial (chiuuisso ‘truly’) occupying a high position in the VP. Example (b) involves the adverbial danân ‘from there’, i.e. a representative of the class of adverbials which are fixed in a low position close to the verb in basic order. Additionally, the latter two types of adverbials are excluded from movement to the right. This suggests that the orders in () are due to leftward movement of Vfin. Let us turn to the question regarding the landing site of V in () and (). Because the verb occurs to the left of the subject in (b) and (c), it is plausible to assume that the subject is in SpecvP and the verb targets a higher functional projection in the structure of the clause. Adopting the clause structure proposed by Fuß and Trips () presented above, Weiß (: ) assumes that the landing site of the verb in () and () is T while SpecTP is the target of fronted XPs of any category type, e.g. subjects in (a) and () and the non-subject (b). In this view, OHG differs from OE in the availability of verb fronting to T which is rejected for OE by Fuß and Trips () but obviously present in OHG. Weiß’s () analysis of the orders in () and () is represented in ()=(b):
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OV versus VO in Old High German ()
TP Spec
T’ T0+ν+V
νP S
ν’ ν0+V
VP V’
S
(daz or dat) H. hætti
min fater
O
V0
t
t
t
(Weiß 2006: 14) The interpretation in () proposed in Weiß () is rejected in Axel (: –), who follows Suchsland () in opting for an analysis according to which the orders in () and () are instances of CP recursion, a phenomenon known from other Germanic languages. Basically, Axel () argues that there is no evidence for the existence of any additional functional projections situated between CP and VP in classical OHG9 but rather attributes the respective orders to V-to-C movement in embedded clauses. CP recursion, also known as limited embedded V, is attested in Modern Scandinavian. As demonstrated in (a) from Danish, the verb surfaces in second position, above the negative element ikke in complements of special verbs called ‘bridge verbs’ like know, say etc. but not regret (b), see Vikner (). Additionally, the verb in second position may be preceded by a non-subject: () a. Vi ved at denne bog har Bo ikke læst we know that the book has Bo not read ‘We know that Bo has not read the book’ (Haider a: ) b. *Vi beklager at denne bog har Bo ikke læst we regret that the book has Bo not read ‘We regret that Bo has not read the book’ (Haider a: )
9 On the various empirical and conceptual arguments in favour of such a clause structure in Modern German and the remaining modern Germanic OV languages, see Haider (, b) and Vikner ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Svetlana Petrova and Helmut Weiß
The common interpretation of limited embedded verb second in Scandinavian involves the presence of two subsequent CPs in the complement of certain matrix predicates if they appear in assertive environments, i.e. in non-modalized, non-negative matrix contexts (de Haan and Weerman ). If the conditions for limited embedded verb second are fulfilled, then the head of the higher CP hosts the complementizer of the complement clause while Vfin targets the head of the lower CP. Assuming that the structure of the clause in OHG only provides C as the landing site of the verb whenever it moves to the left, Axel () derives the orders in () and () as V-to-C movement in a complement clause, similar to embedded verb second in Scandinavian. However, it has not been studied if the remaining conditions required for limited embedded verb second in Scandinavian are fulfilled in OHG as well. Therefore, the interpretation of the orders in this group in analogy to the Scandinavian data must be considered an open issue.
. Conclusion In this chapter, we addressed a well-known and intensively studied property of historical German, namely variation in word order, especially in the order of the verb and its complements. We presented a survey of word orders attested in dependent clauses, because such contexts minimize the number of possible movement operations of the verb, and then surveyed different ways in which these orders have been analysed in the formal literature on the issue. The explanation of word order variation is set in the context of the generative discussion in a cross-Germanic comparison and shows how sensitive the models are towards conceptual innovations in the general linguistic discussion. Although there are a variety of ways of analysing the attested patterns, and although some of the questions cannot receive a complete answer because of the lack of conclusive data, we hope to have been able to show that word order in historical German is by no means accidental or irregular but allows for the proper representation within different models of phrase structure.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
14 Prosodic and information-structural factors in word order variation ROLAND HINTERHÖLZL AND SVETLANA PETROVA
. Introduction In Chapter , it was argued that the word order variation found in dependent clauses in OHG can be explained either as extraposition from a head-final base order, or as leftward movement of the finite verb to a designated functional projection in the left periphery of the clause. This chapter presents an account that elaborates on the second option. It outlines an alternative account of the variation in the order of verbs and their complements, which is couched within the Universal Base Hypothesis (Kayne ). According to this approach, all projections, lexical and functional ones, have a universal head-complement order. The different surface orders are derived by adopting a series of cyclic leftward movement operations which affect both NP complements as well as non-finite complements like VPs and small clauses. It has been shown by Zwart () for Dutch and by Hinterhölzl () for German that all attested surface orders in OV languages can be derived from a universal VO base plus movement of arguments and VP-internal predicates to dedicated licensing positions in the middle field, as illustrated in (a). In this approach, movement of NP and PP arguments serves to check the c-selection of the verb, while the movement of VP-internal predicates and particles serves to form complex predicates. In the remainder of the chapter, we will refer to these movement operations as licensing movement. Thus, superficial VO orders are derived by (i) spell-out of the lower copies of licensing movement out of the VP (Hinterhölzl b) or (ii) by remnant movement of the VP (VP intraposition) across the material extracted from the VP before, as illustrated in (b): () a. [CP that [AgrO XPi [PRTj [VP V XPi PRTj]]]] b. [CP that [VP V ti] [AgrO XPi ] tVP] A basic assumption in this approach is that additional movement operations that are triggered by information-structural and prosodic factors play a crucial role in explaining word order variation in early Germanic. This type of approach has been pursued by Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Roland Hinterhölzl and Svetlana Petrova . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova
Roberts () for OE, modified in Biberauer and Roberts (), by Hróarsdóttir (a, b, ) for ON, and by Hinterhölzl (, b) for OHG. In what follows, we will outline the basic arguments of these approaches and will show—on the basis of the corpus study presented by Petrova () and Petrova and Hinterhölzl ()—how these models combine licensing movement and scope-related movement triggered by information-structural requirements with leftward head movement of the finite verb.
. Word order variation in older Germanic It is well known that the early Germanic languages allowed for greater freedom in word order than their modern descendants (see Chapters and ). In this respect, mixed word orders are of special interest since they challenge accounts based on the head complement parameter. Next to mixed word orders in modern Yiddish, as in () (Diesing ), we find mixed OV/VO orders in the older stages of all Germanic languages, as is illustrated for OE in () (Pintzuk ), for ON in () (Hróarsdóttir ), and for OHG in () (taken from the Tatian translation). As the examples in (b) from OE and (b) from OHG clearly show, we do not only find a combination of pure OV word orders, as in (a) and in (a), and pure VO orders, as in (b), in the same text, but also mixed word orders within the same sentence. In the OE example in (b), for example, the infinitive precedes the finite auxiliary, as is typical for an OV language, but the direct object follows the selecting verb (and the auxiliary), as is typical for a VO language (see also Haider b for a summary of the discussion and examples): () a. Maks hot nit gegebn Rifken dos bukh Max has not given Rifken the book ‘Max has not given the book to Rebecca’ (Diesing : ) b. Maks hot Rifken dos bukh nit gegebn Max has Rifken the book not given ‘Max has not given the book to Rebecca’ (Diesing : ) () a. þæt he his stefne up ahof that he his voice up raised ‘that he raised up his voice’ (Pintzuk : ) b. þæt ænig mon atellan mæge ealne þone demm that any man relate can all the misery ‘that no man could tell all this misery’ (Pintzuk : ) () a. efftir þad þeir höffdu eplid eted after that they had apple.the eaten ‘after that they had eaten the apple’ (Hróarsdóttir : )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Prosodic informational-structural factors
b. að hann haði edið kjotið that he had eaten meat.the ‘that he had eaten the meat’ (Hróarsdóttir : ) () a. (quia dictum est antiquis) thaz then alton giqu&an uúas that to.the old.ones said was ‘that it was said to our ancestors’ (Tatian , ) b. (quae uocatur b&hleem) thiu uuas ginemnit b&hleem that was named Bethlehem ‘that was called Bethlehem’ (Tatian , ) While mixed word orders in Yiddish, OE, and ON have been and are subject to thorough investigation and heated debates concerning their correct analysis (Diesing ; Vikner for Yiddish; Roberts ; Pintzuk for OE), the discussion concerning older stages of German has just started in the last years (see Chapters and ). We briefly recapitulate the main facts and points of significance, as a prelude to a more detailed exploration of the information-structural patterns in the data and their derivation within the universal VO base hypothesis. Traditional grammarians have pointed out that word order in older Germanic is less fixed than in the modern varieties, and argued that word order preferences are due to a large degree to stylistic factors. Most notable among these is Behaghel’s law of increasing constituents. Behaghel (–, Vol. IV) notes that pronouns and unmodified nouns tend to precede the verb, while modified nouns, PPs, and other heavy material tend to follow it, which gave rise to the generalization in () (Behaghel –, Vol. IV: ). ()
Light elements precede heavy elements in OE, ON, and OHG
The statement in () raises the question of what ‘light’ means in this context. The first interpretation is that ‘light’ in () is to be understood as prosodically light. In the same passage, Behaghel also talks about information-structural weight and the general rule that constituents with greater informative weight follow informationally light elements. As it turns out, both factors make relevant predictions about the unmarked word order in older Germanic. There is good evidence that prosodic and information-structural factors play a major role in determining word order in ON (Hróarsdóttir a, b, ) and in OE (Taylor and Pintzuk , a, b, c). In particular, Hróarsdóttir () reports that both factors play a role in ON word order but concludes that prosodic weight was the decisive factor in ON. Taylor and Pintzuk (a, b) argue for the relevance of both factors for determining word order in OE. They also show that the two conditions though overlapping are independent of each other and conclude that the information-structural factor weakens at the end of the OE period.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova
An important observation about word order in OHG is that pronouns and verb particles do not appear after the selecting verb, while PP-adjuncts and PP-arguments do. This property can be related to the law of increasing constituents stated in (), or better, to the prosodic condition in (). In (), it is assumed that a syntactic constituent counts as heavy, if it is branching (Hinterhölzl ). Thus, the prosodic condition allows light, non-branching constituents to precede the verb, but requires heavy constituents, that is, phrases containing two words and more to follow the verb. ()
The weight condition (PF-transparency): A specifier that constitutes a heavy syntactic constituent must appear on a right branch with respect to the selecting/modified head (to occupy a more prominent branch than the head in prosodic structure).
In Hinterhölzl (), it is shown that information-structural requirements regularly override the effects of condition () in OHG. Thus, we will concentrate on the information-structural conditions that determine word order in OHG in this chapter.
. Information-structural conditions on word order in OHG Petrova () and Petrova and Hinterhölzl () discuss evidence indicating that mixed word orders are motivated by information structure. They examine corpus evidence from the OHG Tatian translation and show that there is a tight correlation between the information-structural value of a constituent in the clause and its positional realization relative to Vfin. Petrova and Hinterhölzl () compare these data to selected examples from Isidor, which are partially confirmed and slightly extended in Schlachter’s () complete investigation of the role of information structure on word order in early OHG. The information-structural category that is at the centre of attention of these studies is the expression of focus, i.e. of that part of the utterance that signals the presence of alternatives in the linguistic context. Two semantic types of focus are distinguished in the literature. On the one hand, new-information focus (also called ‘presentational focus’) marks that part of the utterance that has been selected out of an implicitly assumed, open set of alternatives. On the other hand, contrastive focus marks constituents that are selected as members of an explicitly mentioned set of alternatives. Below, the main observations on the role of focus on word order variation in OHG will be presented. .. The position of background versus new-information focus As commonly assumed, the new-information focus of the utterance comprises information that is requested, or asserted as new in the sentence. In contrast, background, or presupposed information, is that part that has been explicitly mentioned in the previous context, or that is inferable via world knowledge or anchoring to other given information (Prince ). In OHG dependent clauses, expressions conveying presupposed information regularly appear to the left of Vfin and in the higher middle field. For example, pronominal arguments are firmly associated with the Wackernagel domain, i.e. the
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Prosodic informational-structural factors
left edge of the middle field, immediately below C (see also Chapter ). Consider () showing that the OHG equivalents of pronominal arguments that are post-verbal in the Latin original are systematically realized in the position below C; the same applies to pronouns lacking an equivalent in the Latin: () (usquedum dicam [tibi]) unzan [ih] [thir] quede until I you tell ‘until I tell you’ (Tatian , ) In some isolated cases, pronouns are not located in the Wackernagel position. But there is always an explanation for this distribution. First, there are cases where the OHG word order reproduces the word order of the Latin original. A relevant example is (a) where the post-verbal position of the pronoun can be attributed to Latin influence. Second, fronting of a pronoun to the Wackernagel domain is blocked by the line break, as in (b). Finally, there are cases in which a root-clause interpretation with Vfin in C is plausible,1 see (c); under this interpretation, the pronoun, which follows Vfin immediately, can be analysed as occupying the Wackernagel position: () a. (ut raperent eum) thaz sie fiengin [inan] that they arrest.PRET.SBJV him.ACC ‘that they arrested Him’ (Tatian , ) b. (& in manibus / tollent té) thaz sie mit Iro hanton / thih nehmen that they with their hands you take ‘that they hold you with their hand’ (Tatian , –) c. (quoniam / sicut proph&am eum habebant) uuanta / samasó uuîzagon hab&un Inan because like prophet had him ‘because they regarded Him as a prophet’ (Tatian , –) Note that apart from pronouns, lexical phrases conveying presupposed information also appear in the middle field of OHG dependent clauses. The corpus provides plenty of examples in which NPs referring to given or inferable information are shifted from the post-verbal position in the Latin original to the preverbal one in OHG. Consider the position of the second appearance of the NP uueralt ‘the world’ in (a), which resumes previously mentioned information. The same applies to the constituent gote ‘God’ in (b), which is inferable in the context of the presentation of 1 Jäger (: f) shows that the majority of the wanta clauses in the OHG Tatian display verb second word order, suggesting that they are root clauses.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova
Jesus in the temple. Note also that there is no example in which a lexical constituent conveying given or inferable information is shifted from the preverbal position in Latin to the post-verbal position in OHG. () a. (non enim missit deus filium suum / In mundum ut Iudic& [mundum]) nisanta got sínan sun / In uueralt thaz her NEG.sent God his.ACC.SG son in world that he [uueralt] tuome world condemned.PRES.SBJV ‘God didn’t send His son into the world in order to condemn the world’ (Tatian , –) b. (postquam Impl&i sunt dies / . . . tulerunt illum In hierusalem / ut sisterent eum [domino]) after thiu gifulta uuarun taga / . . . brahtun sie Inan thô In hierusalem / thereafter filled were days brought they him then to Jerusalem thaz sie Inan [gote] giantuuvrtitin that they him God presented.PRET.SBJV ‘after the days [of her purification] they brought him to Jerusalem to present him to God’ (Tatian , –) Let us turn to the informational status of constituents appearing post-verbally in OHG dependent clauses. Special attention will be given to those examples which contain post-verbal material in contrast to the Latin word order. We can distinguish two basic groups of constituents that appear to the right of Vfin in subordinate clauses in OHG: (i) complements of main verbs and (ii) complements of auxiliary or copula verbs. Complements of main verbs appearing preverbally in Latin but post-verbally in the vernacular are rare in the corpus. But it is crucial that in all of them, the postverbal XP represents novel information, as diuual ‘devil’ in (). At the same time, the corpus provides no examples in which novel material positioned after Vfin in the Latin original is shifted to the domain to the left of Vfin in OHG: () (et obtulerunt ei omnes male habentes [ . . . ] & qui demonia habebant) Inti bráhtun imo / alle ubil habante [ . . . ] Inti thie thár hab&un [diuual] and brought him all evil having and those PTC had devil ‘and they brought to him all sick people and those who were possessed by the devil’ (Tatian , –, ) Much more frequently, the constituent that appears after Vfin is the non-finite part of a complex predicate, for example a predicative adjective (a), a nominal complement of a copula verb (b), or the non-finite main verb (c): () a. (ut [ . . . ] caeci fiant) daz sie sin blinte that they are.PRES.SBJV blind.PL ‘that they became blind’ (Tatian , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Prosodic informational-structural factors
b. (de ciuitate nazar&h) fon theru burgi thiu hiez nazar&h from the town which was.called Nazareth ‘from the town which was called Nazareth’ (Tatian , ) c. (nisi conuersi fueritis / & efficiamini sicut parvuli) nibi ir uuerdet giuuentite / inti gifremite soso theser luzilo unless you become converted and formed like this child ‘unless you convert and become like this child’ (Tatian , ) At first sight, the two groups with post-verbal constituents shown in () and () appear to have little in common. However, from the point of view of information structure, these constituents share one important property, namely they constitute the asserted part, or the new-information focus in the utterance. As for the postverbal argument in (), this interpretation results from its novelty in the discourse. By contrast, the post-verbal constituents in () do not display properties of discourse referents and therefore cannot be analysed with respect to the given/ new distinction. However, together with Vfin, they constitute the asserted part of the proposition, as opposed to the domain of background material, which is placed before Vfin. In sum: background information precedes Vfin on the surface, while novel information follows Vfin, with which it constitutes the domain of new-information focus. In other words, the position of Vfin is a device in structuring the utterance according to the well-known iconic given-before-new principle of information packaging. So far, this distribution confirms Schlachter’s () initial observations about the factors that trigger medial position of Vfin in early OHG embedded clauses. This picture, however, will receive an important modification when the role of contrastive focus is considered. .. The position of contrastive information Let us look at the positional realization of constituents justifying an interpretation in terms of contrastive focus. In particular, this applies to elements displaying a relation of correction or semantic opposition to another entity explicitly mentioned in the discourse, as well as to elements bound by a focus-sensitive operator like even, only, or also. Furthermore, there are instances in which a single constituent bears special emphasis due to rhetorical considerations. Petrova () and Petrova and Hinterhölzl () observe that contrastive information is associated with a particular position in the clause, namely the one immediately preceding Vfin in the clause (while background does not need to be adjacent to Vfin). In the corpus of these studies, this position is targeted by thirty-six of forty-three contrastive phrases, i.e. in % of the cases involving contrast or emphasis on an XP. It is important to note that this positional realization applies to all types of phrases, so for example to arguments (a) and (b) but also to complex adjuncts (c) and simple adverbs (d). The phrases in square brackets either form a contrastive pair with another element present in the context (not X . . . but Y), or are in the scope of a silent
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova
focus operator triggering an exhaustive interpretation (for example, the presentation to God is required for the first-born son only, not for others, in (d)): () a. (quid faciat [dextera tua]) niuuizze íz thin uuinistra / uuaz [thin zesuua] tuo NEG.know it your left.hand what your right.hand do ‘your left hand should not know what your right hand is doing’ (Tatian , ) b. (ne uidearis [hominibus] / ieiunans) thaz thu [mannon] nisís gisehán / fastenti úzouh thinemo fater that you people NEG.be seen fasting but your father ‘that you do not appear fasting to the people but to your father’ (Tatian , ) c. (non iurare [omnino]) Ír guhortut thaz then alton giqu&an uúas / nifursuueri thih you heard that to.the predecessors told was NEG.swear you hane ih quidu íu / thaz mán [zi thuruhslahti] nisuuere PTC I tell you that INDEF at all NEG.swear ‘[you heard that it was told to your predecessors: “You should not swear”] But I tell you that you should not swear at all’ (Tatian , ) d. (adaperiens uuluam) íogiuuelih gommanbarn thaz uuamba [êrist] Intuot each male.child which womb for.the.first.time opens ‘each male child which opens the womb for the first time’ (Tatian , ) In seven cases (%), the contrastive phrase is not in the position immediately before Vfin. However, the linear order of the constituents in these examples either corresponds to that of the original Latin sentence, see the bare plural rehte ‘the righteous man’ in (a), or the transposition of the contrastive XP to the left of Vfin is blocked by the line-per-line principle of the translation (cf. Chapter ), see the NP thiu himiliscun ‘the heavenly matters’ in (b): () a. (Non enim ueni uocare [iustos] / sed peccatores in poenitentiam) niquam zi thiu thaz ih giladoti [rehte] / ouh suntige NEG.came for the.reason that I called righteous but sinful zi riuue to repentance ‘I haven’t come to call the righteous people but the sinners for repentance’ (Tatian , ) b. (si dixero uobis / [caelestia]) oba ih íu erdlihhu quad . . . oba ih íu quidu / [thiu himiliscun] when I you earthly told if I you tell the heavenly ‘[when I told to you earthly things] if I tell you the heavenly things’ (Tatian , –)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Prosodic informational-structural factors
.. Theoretical implications To resume, the positional distribution of arguments depending on their informationstructural value in the utterance in the Tatian translation, according to Petrova () and Petrova and Hinterhölzl (), is summarized in (): () a. background information is located preverbally, in the (upper) middle field of the clause b. the domain of new-information focus follows Vfin c. contrastively focused information is placed left-adjacent to Vfin and surfaces between the background domain and Vfin These findings lead to the conclusion that OHG is discourse-configurational, i.e. that there is a tight correlation between the information-structural value of sentence constituents and their positional realization in the clause. In this section, we want to address the question of how these empirical observations can be integrated into a syntactic model. The principles of word order that we have established for OHG bear parallels to those described by Diesing (: –) for Yiddish. According to her, the different syntactic realization of object NPs with respect to the selecting main verb triggers three different types of semantic interpretation. Hinterhölzl (: ) observes that these interpretations correspond to different categories of information structure. Leftward movement of an NP outside the VP is associated with definiteness and specificity, i.e. with background information, while post-verbal placement yields an existential reading of discourse-new indefinite NPs as instances of new-information, or presentational focus. Additionally, both definite and indefinite objects in the position left-adjacent to the verb gain a special, marked status which is only possible when contrastive or corrective emphasis is put on them, i.e. when they are contrastively focused. Hinterhölzl () accounts for this distribution by assuming movement of the verb to the head of a focus phrase (FocP) which is situated at the left edge of the middle field above the case-checking projections for arguments. While contrastive elements move to [SpecFocP], new objects remain in the scope of the focus head where they receive case, as is illustrated in (): ()
[CP Background [FocP ContrFocus Vfin [AgrP NewInformFoc [VP tV XP]]]]
Petrova and Hinterhölzl () retain the idea of a focus phrase which is situated between CP and VP in the structure of the clause, but introduce a significant modification to the previous account. From the viewpoint of the empirical results, they replace V by Vfin, as is illustrated in (). Thus, the position of contrastive focus is left-adjacent not to the main verb but to Vfin in the clause. All patterns attested in OHG subordinate clauses can be derived according to the following operations: () a. Vfin is moved to the head of focus phrase (FocP). b. [SpecFocP] is reserved for contrastively focused information; this explains the left adjacency of contrastive information to Vfin observed in the data. c. constituents which convey new information remain in the scope of the focus phrase, i.e. after Vfin.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova d. given and presupposed constituents leave the scope of the focus phrase by movement to a position outside the FocP, e.g. to the Wackernagel position and to different topic positions in the left periphery.
Petrova () and Petrova and Hinterhölzl () also address the question of why the distinction of two focus positions should be made explicit in the structure of the clause in OHG. One strong reason for doing so comes from the prosodic realization of focus, especially in cases of multiple foci. Consider examples like () which involves two distinct pairs of alternatives, namely the PPs in mir ‘in me’ and In therru uueralti ‘in this world’ on the one hand, and the object expressions sibba ‘peace’ and thrucnessi ‘tribulation’ on the other. In the Latin version, the members of the different contrastive pairs are realized in the same domain relative to Vfin. In the OHG text, however, the members of the distinct focus domains are displaced, being separated by Vfin: () (ut in me pacem habeatis / In mundo presuram habebitis) thisu sprah ih íu/ thaz [in mir] habet [sibba] / [In therru uueralti] this spoke I you that in me have peace in the world habet ir [thrucnessi] have you.PL tribulation ‘these things I have spoken to you that in Me you may have peace; in the world, you will have tribulation’ (Tatian , –) Such examples suggest that OHG seeks to avoid the stacking of two different focus phrases in the same structural domain in the clause. The reason for this might be purely phonological. Focus is associated with stress, and because the two phrases represent two different semantic types of focus, each of them should represent a separate phonological phrase, bearing its own stress. Placing (the unaccented) Vfin in between these two focus constituents helps to avoid the clash of two heavily stressed phrases in one and the same topological domain in the clause. In other words, fronting of Vfin to Foc is motivated by the Clash Avoidance Principle introduced by Speyer () to explain similar phenomena, for example the role of topicalization in the history of English. But this principle still leaves open why the two different semantic types of focus, i.e. new-information versus contrastive focus should be designated special positions in the clause. A possible suggestion is that the different positional realization of focus types in OHG is used to dissolve focus ambiguities (Petrova ). For Modern German, it is well known that main accent on the right-most XP in base order yields either VP- or XP-focus. Scrambling the right-most XP to the left is considered a strategy to minimize the focus domain (Rosengren ), allowing an otherwise unaccented higher XP to receive main stress and narrow-focus interpretation (see Abraham ). Petrova () suggests that a similar strategy is implemented in OHG on syntactic grounds, in that phrases belonging to VP-focus surface in postverbal position while narrow (XP-) focus is only realized in preverbal position, with additional effects of contrast, emphasis, and exclusion of alternatives.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Prosodic informational-structural factors
.. Discussion and supporting evidence Following the above observations, Vfin plays an important role as a discourse structuring device organizing the clause according to information-structural principles. It is important to determine whether this distribution is restricted to Tatian or is also found in other OHG texts, such as the Isidor translation. Recall that for this text, Schlachter () has already described the properties related to (a) and (b). What is still missing in describing the role of information structure in word order variation in Isidor is the condition in (c), i.e. the role of contrast in the organization of the clause. Some facts concerning the main objective of the text are crucial to the interpretation of the examples. It is well known that the purpose of the Isidor treatise was to vindicate the notion of Trinity against the doctrine of Arianism. By references to statements of the Old Testament, the text attempts to verify the divine origin of Christ, who, contrary to the assumptions of the Arianists, was not only adopted, but created by God the Father. Petrova and Hinterhölzl () find examples in Isidor in which the informationstructural categories of background, new-information focus and contrastive focus are exactly as in Tatian. Consider the data in (). Expressions classifying as background information, such as the anaphoric NP christ gotes sunu in (a) or the pronoun ir in (b), appear preverbally, while material conveying the asserted part of the utterance follows Vfin. In turn, expressions bearing a contrastive interpretation, for example those excluding an alternative to the divine origin of Jesus (conveyed by the PP fona fater ‘by the Father’ in (a)), or members of a set of alternatives (iu ‘already’ versus noh ‘still’ in (b)) are placed immediately before Vfin: () a. Mit so mihhiles hęrduomes urchundin ist nu so offenliihho armarit with so great glorious proof is now so evidently said dhazs christ gotes sunu . . . [fona fater] uuard chiboran that Christ God’s son from Father became born ‘By so many facts it is evidently said that Christ, the Son of God, was created by the Father’ (Isidor , ) b. Souhhemes auur uuir nu ziidh dhera christes chiburdi huuedhar ir explore but we now time of.the Christ’s birth whether he [iu] quhami odho uuir [noh] sculim siin quhemandes biidan already came or we still shall his arrival expect ‘But let us explore the time of Christ’s birth, whether He already came or we still have to expect His arrival’ (Isidor , ) Schlachter’s () complete investigation of the position of Vfin in dependent clauses in Isidor confirms these properties, but adds some modifications. She identifies orders in which a contrastive focus follows Vfin, for example cases of so-called emphatic negation (a) or within a larger NP or PP which has been extraposed to
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Roland Hinterhölzl and Svetlana Petrova
the right (b). At the same time, XPs in the pattern Aux – XP – V may also receive contrastive interpretation, see (c): () a. (quia moriens caro eius non uidit corruptionem) dhasz imu arsterbandemu siin fleisc ni chisah enigan unuuillun that he.DAT dying.DAT his body NEG saw any corruption ‘that his body showed no corruption after he died’ (Isidor , ) b. (Quod in persona specialiter christi domini nostri accipitur) dhazs dhiz ist chiquhedan [in unseres druhtines nemin] that that is said in.the our.GEN Lord.GEN name ‘that this is said in our Lord’s name’ (Isidor , ) c. (intellege christum) dhazs dhar ist christ chizeihnit that there is Christ designated ‘that it is Christ who is designated there’ (Isidor , ) Schlachter () concludes that in early OHG texts, the position of Vfin, which is not fixed yet, is used to express focus on an object in various word order patterns to a larger extent than is possible in later texts.
. Conclusion This chapter addressed the derivation of different word order patterns from a universal VO base and leftward movement of constituents related to the expression of information structural categories. It was shown that in OHG, constituents are mapped according to their information-structural contribution into the following three domains in the clause: (i) presupposed, or background material is realized in the (upper) middle field, (ii) material that belongs to the domain of new-information focus appears post-verbally, while (iii) contrastively focused phrases are realized leftadjacent to Vfin. To account for this distribution of constituents in the clause, we assume leftward movement of Vfin to the head of a functional projection FocP which separates the domains of background and presentational focus in the clause, with [SpecFocP] being reserved for phrases with a contrastive interpretation. This explanation takes into account the interaction between prosodic and information-structural conditions in word order variation.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
15 Periphrastic verb forms AUGUSTIN SPEYER
. Periphrastic verb forms as an example of grammaticalization In the context of word order at the right sentence periphery in German, the verbal complex formed by periphrastic verb forms is crucial, cf. Chapters and on the role of the ordering of finite and non-finite verbs in determining the basic word order and on phenomena such as verb raising and verb projection raising. The present chapter discusses the development of periphrastic verb forms and related processes of syntactic change. The rise of periphrastic verb forms is a classic example of the combination of semantic and morphosyntactic change processes that is referred to descriptively as ‘grammaticalization’. Following, among others, van Gelderen (); von Fintel (); Roberts and Roussou (); Öhl (); and Ferraresi (), grammaticalization is understood here as a sequence of language changes that concern different subsystems: • The first step is the semantic change called ‘semantic bleaching’. If there is a prolific use of a combination of one specific lexeme L1 with other lexemes Ln, where all Ln are of the same category, and if the usage is such that L1 shares at least one semantic feature with a grammatically encodable category C (e.g. one of the universal pool of φ-features or other features able to head a functional projection),1 the semantic content of L1 will change such that most, if not all semantic features except the feature(s) conforming to the φ-feature constituting C will be deleted. • Once this is on its way, syntactic changes set in, reflecting the semantic change. Usually these are reanalysis processes that lead to L1 governing Ln and recategorization processes from lexical to functional categories. • If such a sequence of changes happened once, it is possible that this sequence of changes serves as a model for other, similar changes (‘analogical extension’). 1 For instance, if L1 can be understood as a metaphor for C (e.g. ‘going to a place’ as a metaphor for ‘doing in the future’), or is understood as bearing some other logical relation to C (e.g. if an event A causes another event B, A usually temporally precedes B. So lexemes denoting temporal succession can be used for denoting causation, like the German conjunction weil ‘because’, which comes from the noun Weile ‘timespan’).
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Augustin Speyer . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
In the case of periphrastic verb forms, the syntactic changes proceed as follows, illustrated with the example of forms with participle, namely the eventive passive (werden-passive) (cf. Ferraresi ) and the haben-perfect (cf. Öhl ). In OHG and probably even before that period, predecessors of the later eventive passive and the haben-perfect were in use, but with a different meaning. Examples from OHG are given in (). Note that in these examples it is unclear whether the grammaticalization process has already taken place or not as they are ambiguous. () a. inti nioman iro gireinit uuard and no.one of.them cleansed became ‘and no one of them was cleansed / was clean’ (Tatian , ) b. phígboum habeta sum giflanzotan in sinemo uuingarten fig.tree had someone planted in his vineyard ‘somebody had planted a fig tree in his vineyard/somebody owned a fig tree which was planted in his vineyard’ (Tatian , –) Using the examples in () in English wording and simplifying them somewhat, I outline the stages of the grammaticalization process in the following. Stage : L1 is a main verb denoting something that can be used as a metaphor for a given grammatical category, e.g. ‘to move’ for ‘to come into a state’, ‘to exist’ for ‘to be in a state’, etc. This verb governs a [+ N] complement. Ln is either the nominal complement itself (a) or a predicative to the nominal complement (b; the predicative is adjoined below the complement).2 Also deverbal adjectives (i.e. participles) or nouns can be used as [+N] complements. ()
VP
a. NP John
AP | A’ | A0 cleaned (Ln)
V’ V0 turns (L1)
2 I use here a modified small-clause analysis for predicatives, the predicative being a VP headed by the covert verb ‘to be’ (for which I give a simplified lambda calculus above), having an empty element as specifier that is coreferent and controlled by the object of the matrix clause and the predicative in the strict sense as complement (following Rafel ). In order to make the lexical meaning of L at this stage clearer, I use fully lexical English verbs for the concepts involved.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
VP
b.
V’
NP John
NP fig-treei
V’ VP V’
NP PROi
AP | A’ | A0 planted (Ln)
V0 owns (L1) V0 [[∅]] = λyλx.x = y
Stage : In the case of deverbal adjectives/nouns, the deverbal adjective/noun Ln is reinterpreted as verbal category (). If Ln is the complement of L1 by itself, this is all that changes, cf. (a). If Ln is part of a small clause structure, a reanalysis takes place in which the covert verbal head with the meaning ‘to be’ is deleted and instead the reinterpreted deverbal element Ln functions as head of the small clause VP, cf. (b). This is a classical case of simplification due to lacking evidence for the presence of a silent category. The relabelling also induces an interpretation of the object of L1 as the object of Ln. The structure is then that of two recursive VPs, one headed by L1, one headed by Ln. In order to show ‘semantic bleaching’ of the head verb L1, I use at stages and verbs with fewer conceptual meaning components. ()
a.
VP NP John
V’ V0 (turns) → becomes (L1)
VP V’ V0 cleaned (Ln) b.
VP NP Johni
V’ VP
NP PROi NP fig-tree
V’
V0 (owns) → has (L1) V0 planted (Ln)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
Stage : Reanalysis takes place of the recursive VPs as one VP in which the head V(Ln) is adjoined to the head V(L1), cf. (). This is represented here as clause union (cf. e.g. Haegeman and van Riemsdijk : ; Haider ; Schallert b) if one adopts the view that the verbal complex is rather a V-shell structure, the structure would remain as in (). The clause-union analysis, however, has an advantage: if Ln originated as an adjoined small clause structure, this reanalysis leads again to a strong simplification, as the silent subject position of the embedded VP disappears.3 ()
a.
VP NP John
V’ V0
V0 cleaned (Ln) b.
V0 becomes (L1)
VP NP John
V’
NP fig-tree V0 planted (Ln)
V0 V0 has (L1)
There are certain morphosyntactic cues as to which stage the grammaticalization process has proceeded. If the process is at stage , morphological properties of [+N] categories—i.e. nominal inflection4—is visible on Ln.5 Furthermore, the phrase (AP) containing Ln can be moved freely.
3 Whether stage occurs in all cases of periphrastic verbs is a matter of debate. The account given here mainly follows the school represented e.g. by Haider (, a). An argument against the existence of stage would, for instance, be that verb projection raising should not be possible (cf. e.g. Kefer and Lejeune ). However, the possibility of verb projection raising is subject to variation as some varieties of German allow for it while other varieties do not (see also Chapter ). The ‘completion’ of the grammaticalization process outlined here (i.e. stage ) need not have taken place in all varieties of German. One should also bear in mind that it is characteristic of grammaticalization processes that several stages of the process can co-occur. So speakers that allow for verb projection raising may sometimes analyse their periphrastic verb forms as if at stage , sometimes as if at stage . 4 The term ‘nominal inflection’ is meant to denote inflection for categories that are typical of nouns, such as case or gender. It is not meant to denote here case endings identical to substantives. Whereas many languages show similar declension patterns for substantives and adjectives, German adjectives inflect in a way rather different from the nouns from the earliest attestation. 5 For a critical discussion of the link to the presence of inflection on Ln (specifically on the past participle in the case of periphrastic perfect forms) see Gillmann ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
If the process is at stage , the morphological properties indicative of [+N] are missing. Scrambling of the VP headed by Ln is not possible any more as scrambling of non-clausal verbal projections is not possible (see e.g. Grewendorf and Sternefeld : f). Topicalization of the VP headed by Ln is, however, still possible. Furthermore, the order in the verb cluster is fixed (V(Ln)–V(L1) with right-headed VPs; with left-headed VPs, a verb cluster need not originate as material inside the VP(Ln) can intervene, see e.g. Haider ; Schallert a, b). The only way to reorder the verbal cluster would be verb projection raising. At this stage, the semantics of L1 is bleached to such an extent that the constraints originally valid for the parent construction of stage —for instance, that the verb from which the participle Ln is derived must be transitive—can be violated. If the process is at stage , the verbal elements inside the verbal cluster can be permutated freely and verb raising can apply (see e.g. Haegeman and van Riemsdijk ). An indicator that verb raising has taken place can for instance be that negative operators have exclusively narrow scope (see Haegeman and van Riemsdijk : ff). Topicalization of the VP headed by Ln is still possible, as the V(L1) can be moved out of the verb cluster to I. In the following, the grammaticalization model outlined above will be applied to a number of periphrastic verb forms that developed during the history of German.
. Periphrastic verb forms in the history of German .. The stock of periphrastic verb forms In German, numerous periphrastic verb forms have developed over the roughly , years of attested language history. While some are already present in OHG, others can be observed in the textual record as they develop. A list of periphrastic verb forms is given in Table .. The date of first attestation follows either my examination of the TITUS database; Fleischer and Schallert (FS); Harm (H); or Oubouzar (O).6 A dagger before the form indicates that the form is extinct in Modern German, at least in the usage as paradigmatic periphrastic verb form. In this case, the function given is the one the form had after grammaticalization, and the Modern German example is only for illustration. A dagger in brackets indicates that the form is not used in all varieties (in particular not in the standard variety). .. Some forms in detail: the Old High German forms Let us look at some of these forms in more detail. We begin with the passive (cf. for the following sections Fleischer and Schallert : Chapter ). The passive with both auxiliary options, werden ‘to become’ (a) and sein ‘to be’ (b) is attested already in Isidor. The use of the auxiliary sein is not yet confined to 6 As far as periphrastic future forms are concerned, Harm (: ) reports modals such as sollen + infinitive ‘before ’; in fact there are OHG examples, see (a). The earliest examples of the future periphrasis of werden + infinitive, he dates to ‘around ’. However, there are for instance examples already in Berthold (see ()), that is, from the second half of the thirteenth century. See also Luther () on the development of future forms in MHG.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
T . Examples for periphrastic verb forms with first attestation form
function
example (Modern German)
first attested
sein + past participle
stative passive
es ist gebaut ‘it has been built’
th century
werden + past part.
eventive passive
es wird gebaut ‘it is (being) built’
th century
{sein + present part.
progressive
er ist singend ‘he is singing’
th century
sein + past part.
present perfect (intransitive)
er ist gegangen ‘he has gone’
th century
haben + past part.
present perfect (transitive and some intransitive verbs)
sie hat geschrieben ‘she has written’
th century (FS)
({)tun + infinitive
no clear function, sometimes indication of focus on verb
sie tut schreiben ‘she does write’
th century (FS)
{sollen etc. + inf.
future
er soll singen ‘he will sing’
th century
{werden + pres. part.
future
sie wird schreibend ‘she will write’
th century (H)
werden + inf.
future
er wird gehen ‘he will go’
th century
werden (pret. subj.) + inf.
subjunctive (irrealis)
sie würde schreiben ‘she would write’
th century (O)
the stative passive (as in Modern German), but there are already aspectual differences (Oubouzar : f). () a. (quando sit dei filius genitus) in huuelihheru ziti gotes sunu kaboran uurti in which time God’s son born become.PAST.SBJV ‘in which time the son of God should be born’ (Isidor , ) b. (filius dei vocatur) gotes sunu ist ganemnit God’s son is called ‘he is called God’s son’ (Isidor , ) Already in Isidor, most occurrences are without nominal inflection on the participle. There are some exceptions, though, as in ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
() (qui omnipotens deus a patre omnipotente missum se esse testator) ir almahtic got sih chundida uuesan chisendidan he almighty God REFL knew be.INF sent.ACC.SG fona dhemu almahtigin fater from [the almighty father].DAT.SG ‘he, being God almighty, knew that he was sent by the almighty Father’ (Isidor , ) Interestingly, in Isidor all passives with inflected participles are interpretable as stative passives. The status of stative passives as full verb forms (as opposed to stage-like copula constructions) is a matter of debate and has been contested e.g. by Maienborn (), so it is possible that until today, the stative passives simply have not reached the same stage of the grammaticalization process as the eventive passives. This mirrors in some ways the observation of Fleischer (: ) that inflected participles are more closely associated with the sein-passive. Inflection indicates an adjective-like quality, which means that the step towards recategorization of the participle from an adjectival to a verbal category has perhaps not yet taken place. A form that appears as early as Isidor, but did not make it to Modern German is a combination that resembles the English progressive. It can be formed with the auxiliary sein in the present tense (a) or the past tense (b) and the present participle of the main verb. () a. (querunt multi) manage sint sohhenti many are searching ‘many are investigating’ (Isidor , ) b. (quem sit affatus deus in genesi) huuemu got uuari sprehhendi in genesi whom God was telling in genesis ‘to whom God was speaking in the book Genesis’ (Isidor , ) The form ceased to be used during the early ENHG period (Oubouzar finds examples still in early fifteenth century texts, but not in sixteenth century texts). Note that the form of the participle ‘V-enti’ is actually uninflected, as can be seen from the fact that the participle has the same form in (), although it should agree with a singular subject in (b) and a plural subject in (a), if it were at stage of the grammaticalization process. So this form also seems to have reached stage by early OHG. Let us turn to one of the most successful periphrastic forms: the perfect. The seinperfect appears already quite early without nominal inflection on the participle, for instance in the ninth century Otfrid (a), but there are still sporadic examples of participles with nominal inflection as late as Notker (around ; (b)). The earliest text, Isidor, dating from around , has only few examples of a perfect with sein. In (c), a Latin passive is rendered with the active verb uuerdan in its full verb meaning
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
‘to originate’. An example that cannot be understood any more in the ‘original’ predicative sense is (d). () a. Er ist fon héllu iruuúntan / ioh úf fon dóde irstantan he is from hell returned and up from death risen ‘he has returned from hell and has risen from death’ (Otfrid V , ) b. Also der centurio uuas fóne gentibus chómenêr as the centurio was from peoples come.MASC.SG.NOM ‘as the centurio has come from the gentiles’ (Notker, Psalter , ) c. (per illum cuncta creata esse) dhurah inan ist al uuordan through him is all become ‘Everything has originated through him’ (Isidor , ) d. (qui . . . descendere solitus est et ascendere) dher . . . chiuuon ist fona himile nidharquheman endi uphstigan who usually is from heaven down.come and up.climbed ‘who usually has come down from heaven and has gone up’ (Isidor , ) The haben-perfect (formed in OHG with either the verb habên or eigan, both originally meaning ‘to possess’) is not attested in Isidor, but appears in the early ninth century (a) and is quite frequent in Otfrid, i.e. during the second half of the ninth century. The participle is nearly always uninflected, although there are stage examples that render a Latin predicative. Note that quite early on, we find examples in which the interpretation as stage is excluded as there is no accusative object to which the participle could be a predicative (b). At the end of the OHG period we find examples in which the participle is formed from an intransitive verb (c). This indicates that the construction has reached stage in early OHG, and shows extension to other contexts in late OHG. () a. Ir den christianum namun intfangan eigut you [the christian name].ACC received have ‘you have received the Christian name’ (Exhortatio B – – after Fleischer and Schallert : ) b. denne der paldet, der gipuazzit hapet then he becomes.strong who repented has ‘then anyone becomes strong who has repented’ (Muspilli ) c. Nu hábent siê dir / úbelo gedanchot now have they you.DAT badly thanked ‘now they have thanked you badly’ (Notker, Psalter , )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
So it is fair to assume that the grammaticalization process of the haben/seinperfect has already reached stage at the beginning of the OHG period. The fact that earlier stages of the process (see (b)) coexist with later stages is typical of grammaticalization. Note, however, that inflected participles occur only rarely (Fleischer and Schallert ). Summarizing the OHG state-of-affairs, we can say that the occurring periphrastic verb forms—the present perfect, the passive, and the progressive—are already highly grammaticalized in early OHG and probably came into use some time before the textual attestation sets in. This might indicate that these German periphrastic forms are not simply devices to render Latin verb forms (which in the case of the progressive and the haben-perfect would not make sense anyway, as the biblical Latin relevant for the translations does not use these forms and does not make the aspectual difference conveyed by these forms) but are stable means of expression in the ‘normal’, non-translational OHG vernacular. Whether the development is completely indigenous or whether it is influenced by the advent of periphrastic forms in languages like Romance (esp. Old French) is not at issue here, although it is a very interesting question (see e.g. Drinka ). .. The future forms: a new wave of grammaticalization in Middle High German In the MHG period, a future tense develops (cf. Harm ; Luther ). Originally there originated two ways of forming it, either by means of a ‘modal’ wollen/sollen/ müssen (the semantic bleaching path of these verbs is: wollen: ‘to want’ sollen: ‘to owe > to be obliged to do something’, müssen: ‘to have the possibility to do something > to have to do something’) plus infinitive (a) and (b), which goes back to OHG times and corresponds to the English periphrastic future with will and shall, or by means of the auxiliary werden + present participle (c). The forms are sometimes ambiguous between a future and a modal (conditional) reading (e.g. (b)), but this is quite typical for this stage of the development (cf. Breitbarth b). () a. (in domino gaudebo! et exultabo in deo ihesu meo) in drohtino ih sal mendon unde ih sol urouuen in gode in Lord I shall be.glad and I shall be.glad in God.DAT ‘I shall rejoice in the Lord and be glad about God’ (Rheinfränkische Psalmenübersetzung ) b. sol aber ane lon beliben / steter dienest lieben wiben shall but without reward stay enduring service [dear women].DAT we den frouwen woe [the ladies].DAT ‘If the permanent service for the dear women will remain without reward, woe to the women’ (Frauenlob , , ) c. und werdent mir dan alle . . . gebende die schulde and become..PL me.DAT then all giving the blame’ ‘and then they all will blame me’ (Tristan –)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
In both cases, the grammaticalization process is quite straightforward. Let us first look at sollen, wollen, etc. plus infinitive: the verbs that are used for the future periphrasis undergo a process of grammaticalization from full verbs to modal verbs at around the same time; the development towards future auxiliaries is a further grammaticalization process that takes the modal meaning as starting point. The modal meanings of all three verbs entail an action taking place in the future, so a future can develop from that by a simple semantic bleaching process in that the directive semantic components are deleted and the reference to a future action is kept. The construction of werden + present participle follows the same logic as the passive periphrasis a couple of centuries earlier. The verb werden denotes a change of state. In the case of the passive, it was a change of state of the object, expressed by the past participle. Here, it is a change of state of the subject, expressed by the present participle. Sie wird schreibend ‘She becomes writing’ thus means: she comes into a state in which she is engaged in writing. While both constructions are perfectly straightforward, some years later, a third future periphrasis develops, namely the periphrasis with werden + infinitive (). () sie werdent halt got schelten they become just God scold ‘well, they will scold God.’ (Berthold I , ) There are several ways to account for this. Harm (), after having reviewed some options, decides that it is a blend of the two earlier constructions: the auxiliary werden wins over the modal verbs as this auxiliary does not entail any directive meaning. This is to be seen in the context of the still ongoing development of the verbs wollen/sollen/müssen into modal verbs and can be interpreted as a division-oflabour process, in that the modals that used to be ambiguous with respect to future or directive meaning retreated to directive meaning only, whereas the future part was taken over by werden. This explains why the modal plus infinitive periphrasis went out of use, but it does not explain why the werden + present participle periphrasis went out of use too, instead of taking over. A possible way to account for this is to be seen in the simultaneous decrease of the progressive construction. Obviously all periphrastic verb forms that involved a present participle went out of use.7 They are dispensable, as in both cases alternatives exist; in the case of the future werden + participle option there are the infinitival future forms as an alternative, likewise the simple present forms can be used as an alternative to the progressive, under the assumption that the progressive did what all aspectually marked forms in German sooner or later do, namely lose their aspectual marking. So perhaps these forms were no longer seen as central to the paradigm and had a disadvantage in the competition. Perhaps a well-formedness condition on periphrastic forms originated in that the assembly parts of these forms are constrained—only the ‘real’ auxiliaries (sein, haben, 7
In Bavarian, werden + present participle was very common as a progressive form until the middle of the twentieth century, but this seems to be an exception, because many modern dialects have lost the present participle completely (cf. Weiß c).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
werden) and only the most commonly used non-finite forms—namely the infinitive and the past participle, but not the present participle—could be used for periphrastic forms. Or the werden + present participle construction was confused with the progressive sein + present participle; when the latter went out of use, the former followed. Once werden + infinitive as a marker of future meaning has been established, it is possible to form, for instance, a past subjunctive from it. Following Smirnova (), this was originally used to express a hypothetical consequence from a given precondition in conditional sentences. This consequence is to be regarded as future from the viewpoint of the precondition, so the use of a ‘future subjunctive’ is quite straightforward. After this form originated, it became frequent as an alternative expression for the past subjunctive and later for any subjunctive. One reason might be that in weak verbs, the synthetic past subjunctive is formally identical to the past indicative since the MHG period, so this new periphrastic form came in handy to unambiguously express the subjunctive (Fleischer and Schallert ). The ‘second wave’ of grammaticalization in MHG follows a different path compared to the first wave in pre-OHG. Ln is not an adjectival element that turned into a participle, but a nominal element that developed into an infinitive (on the original nominal status of infinitives cf. e.g. Abraham ). The grammaticalization of this nominal element into an infinitive had already happened in pre-OHG times as infinitives governed by the modals (or their full verb predecessors) appear without nominal inflection (see (a)). The use of this infinitival form as part of periphrastic verb forms seems to appear with some time lag; although the use of infinitives depending on verbs like sollen to express future meaning is already attested in OHG, this cannot yet count as a fully grammaticalized form from a semantic perspective as the ‘modal/auxiliary’ part is variable (any of the verbs later developing into modals such as sollen, wollen, müssen can be used) and the meaning of this part has not yet bleached completely. The full grammaticalization of the periphrastic form as a whole (i.e. the transition from stage to stage ) can only be regarded as complete when the werden periphrasis became the only possible future construction (in late ENHG).
. Permutation in the verbal complex From a syntactic viewpoint, the reanalysis towards a verbal complex might have occurred quite early as in MHG we already have a verbal complex with free permutation of auxiliary/modal and infinitive (Sapp a: ). The numbers and percentages are reported in Table .. In contemplating the numbers reported by Sapp (a: ), it is striking that the order ‘full verb form > auxiliary’ is less frequent with periphrastic verb forms containing an infinitive than with the ones containing a participle. This tendency is perpetuated in Modern German dialects (Sapp a: ). According to Sapp’s figures, the future forms even show a clear disinclination towards this order. This might indicate some structural difference between periphrastic forms containing an infinitive and periphrastic forms containing a participle, and the structural difference
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
T . Ordering variants in the MHG verbal complex, depending on type of periphrastic verb form non-finite > finite; tokens
non-finite > finite; percentage
finite > nonfinite; tokens
.
.
.
.
.
.
perfect with sein (sein ‘to be’ + past participle)
.
.
perfect with haben (haben ‘to have’ + past participle)
.
.
modal verb + infinitive
.
.
other verb + infinitive
.
werden future (werden ‘to become’ + infinitive)
Type of verb form
sein passive (sein ‘to be’ + past participle) progressive (sein ‘to be’ + present participle) werden passive (werden ‘to become’ + past participle)
.
finite > nonfinite; percentage
. .
(adapted from Sapp a: )
might well be that they have reached different stages of the grammaticalization process. At stage , an order ‘auxiliary – full verb form’ can be generated by extraposing the lower VP (potentially after emptying it of non-verbal material); at stage , this order must be the result of permutation inside the verbal cluster. As the latter process is a process that might occur post-syntactically, we would expect a higher influence of prosodic factors such as focus or prosodic setup of the full verb form in stage than in stage environments. Sapp (a: ) does not systematically address this question, but he offers one cross-tabulation for syntagm type and prefix type (that is, whether the full verb form starts with a stressed/unstressed prefix or not). The figures given in this table suggest that with verb forms involving a participle, the rate of the order ‘full verb form > auxiliary’ is constant in about three quarters of all cases, regardless of the prefix type, whereas with verb forms involving an infinitive, there is some variation between full verb forms with stressed prefixes (roughly fifty/fifty) and forms with unstressed prefixes (roughly two thirds with the order ‘full verb form > auxiliary’). This shows that there is some difference, but this evidence actually speaks against the hypothesis that forms with participles are at stage , whereas forms with infinitive are at stage . It is not clear, whether the cases
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Periphrastic verb forms
reported by Sapp can all be analysed as permutation inside the verbal complex or as extraposition of the infinitive, but this is perhaps not crucial. This chapter attempted to give an overview of the origin of periphrastic verb forms in the history of German against the background of an articulated theory of grammaticalization as a blend of a semantic change (stripping of semantic features) and syntactic reanalysis. The attested cases of periphrastic forms fit rather naturally with this theory.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
16 On the history of the IPP construction in German A G N ES JÄ G E R
One fascinating phenomenon that is indicative of the formation of a verb cluster and is typically linked to word order permutation, viz. reordering of verbs within the right periphery of the clause, is the occurrence of IPP (infinitivus pro participio). This chapter presents an overview of the history of the IPP construction in German and discusses its development, different hypotheses on its origin, and possible analyses.1 It places the IPP construction as one of several repair strategies in the context of other three-verb constructions attested in the history and in regional varieties of German. The phenomenon of IPP occurs in several West Germanic languages including German. Certain verbs (modals, optionally a number of other verbs selecting an infinitive such as causatives, perception verbs, etc.) that are selected by a perfect or past perfect auxiliary, and themselves select an infinitive, take the form of a bare infinitive rather than appearing in the expected form of a past participle as selected by the auxiliary, cf. (). The verbs involved form a verbal complex, or verb cluster, in which, in Modern German, a reordering V1-V3-V2 can generally be observed. V1 refers to the highest verb, i.e. the perfect auxiliary. V2 to the modal or other IPP verb that would be expected to occur as a past participle but does in fact occur as a bare infinitive. V3 is the infinitive selected by V2. () dass sie ihn [V1 hat] [V3 hören] [V2 können / *gekonnt] Modern German that she him has hear can[=INF] could[=PAST.PTCP] ‘that she could hear him’
. The rise of the IPP construction In the Old Germanic languages, verb clusters with a maximum of three verbs are attested. However, the periphrastic perfect tense forms had not yet fully evolved (see 1
An earlier version of this chapter was presented at the second workshop of the DFG-funded network ‘Historical syntax’, December in Hamburg. The historical examples as well as basic generalizations from my presentation were also published by Hinterhölzl (a). Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Agnes Jäger . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
also Chapter ). Furthermore, to start with, modals with a dependent infinitive— one prominent class of IPP verbs—are only attested in finite forms. In German, we find non-finite modal verb forms beginning only in the twelfth and thirteenth century. This is because these verbs belong to the morphological class of preterite presents, i.e. their present tense forms are reanalysed original past tense forms. Accordingly, there is no infinitive or past participle to go with these reanalysed finite forms. The respective non-finite forms are only secondarily formed later—first in MHG the infinitives,2 then around weak past participles (cf. Ebert et al. ). IPP constructions with modals as well as the competing construction including the actually expected past participle are of course only possible after that. In the case of modals, the IPP construction may have been used originally because the infinitives were formed before the past participles even existed. (For a semantic explanation based on modals being inherently atelic, see Section ..). However, modals are not the first verbs showing the IPP phenomenon. The only comprehensive empirical study of the history of IPP in German so far is the one by Kurrelmeyer (), mainly based on MHG and ENHG chancery documents. The first rare cases of IPP in German occur in the thirteenth century, i.e. in the MHG period.3 According to Kurrelmeyer’s data, the earliest IPP constructions for the most part include the verbs tuon ‘to do, to have something done’ and hoeren ‘to hear’, cf. () and ().4 () han wir disen brief dun besegelen have we this letter do seal ‘have we had this letter sealed’ (Hoefer, Urkunden – Köln ) () ich hân des hoeren jehen I have that hear say ‘I have heard that being said’ (Kudrun , – around /) Occasionally, heizen ‘to let, to tell/bid someone to do something’ also occurs in thirteenth century IPP constructions, cf. (), and very rarely (with only one attestation each) helfen ‘to help’ and müezen ‘must, to have to’, cf. () and () (all other attestations in Kurrelmeyer’s data are from the fifteenth to sixteenth century as are the first examples of other modals).5
2
Birkmann () reports MHG infinitives of müezen, mugen, wellen, and kunnen (not of suln). According to Ebert et al. (: §S ), IPP is found in particular in chancery documents, while Behaghel (–, Vol. II: f) states that it is first attested in lower registers (zunächst in unteren Sprachschichten). For recent IPP data from one MHG and one ENHG text, respectively, see also Müller (), and Maché and Abraham (). 4 The evidence from Kudrun is problematic insofar as the oldest preserved manuscript stems from the ENHG period. 5 Again, the evidence from Biterolf is problematic, since this thirteenth century epos is only attested in a sixteenth century, i.e. ENHG, manuscript, cf. also Fleischer and Schallert (: ). 3
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
() und haete im heizen machen . . . ein wunneclîchez huselîn and had him let make a pretty house ‘and had him make a pretty little house’ (Tristan – – around ) () vnd heten in dann nwer die aller pest . . . helfen in ze fueren and had them then only the very best help in to bring ‘and had helped them to import only the very best’ (Urkunden Heiligenkreuz – Wienerwald ) () sô hete ichz müezen lân belîben so had I.it must let remain ‘so I would have had to leave it’ (Biterolf – around ) However, according to Lachmann (: ); Grimm (: ff); Paul (: ff); Behaghel (–, Vol. II: f); and Ebert et al. (: §S ), one of the earliest verbs to occur in the IPP construction is lazen ‘to let, to have something done’, as in the early thirteenth century example in (). () umbe waz hâstû daz schif lâzen gân for what have.you that ship let go ‘Why did you let this ship go?’ (Tristan f – around ) This is central for their theory on the origin of IPP (see Section ..). In his empirical study, Kurrelmeyer (: ), on the other hand, notes that it cannot be decided for the earliest attestations of lazen whether they are indeed infinitive forms (thus instantiating IPP) or whether they are simply ge-less variants of the past participle. This ambiguity is in fact exploited in Lachmann’s/Grimm’s theory about the origin of IPP (see Section ..). The first safe attestations of lazen as IPP according to Kurrelmeyer stem from Low German and East Central German sources of the fourteenth century in which lazen is attested in the ge-prefixed past participle form in other constructions so that the form lazen in the three-verb clusters in question constitutes a clear instance of IPP, as in (). () und der diz hât lâzen schrîben and he this has let write ‘and he who had this written’ (Hermann von Fritslar , – mid-fourteenth century) In the fourteenth century, the IPP construction is furthermore extended to bitten ‘to ask/beg’, cf. (). () unde han darzu dye schefenin unde dye burgere zu den Gyezin and have for.that the assessors and the citizens at the Gießen ere stat ingesigele an disin briep biddin henkin their town seal on this letter ask hang ‘and have therefore asked the assessors and the citizens of Gießen to hang the seal of their town on this letter’ (Hessisches Urkundenbuch I, , – Gießen )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
A single attestation of türren ‘to dare’ as IPP in Kurrelmeyer’s data also stems from the late fourteenth century, cf. (). All other cases are from the second half of the fifteenth century at the earliest. () er hat ez vrischleich türren wagen he has it freshly dare venture ‘He has lively dared to venture it.’ (Suchenwirt , – c. ) After those very sporadic occurrences of IPP in the thirteenth and fourteenth century, the IPP construction becomes more prominent in the language over the course of the fifteenth century when it also spreads to a number of modals and other verbs, notably mögen ‘may, to be able to’ (more common in IPP constructions than müezen and türren) and wollen ‘to want’, as illustrated in () and (). Very rarely also können ‘can/be able to’ and dürfen ‘may/need to’ (with only two attestations each from the fifteenth and sixteenth century), as well as sehen ‘to see’ and machen ‘to make’ are found, cf. ()–(). These are more common still with the expected past participle construction. Furthermore, there is a single attestation in Kurrelmeyer’s data of IPP including lernen, which actually means ‘to learn’, but in this case, cf. (), occurs in the autoconverse meaning ‘to teach’, a semantic phenomenon that is also widespread in Modern German dialects. The verb leren ‘to teach’ itself is also sporadically found in fifteenth century IPP constructions, cf. (). () des hab aber sein gnade . . . nicht mugen bekomen that has but his honour not may get ‘but His Honour could not get this’ (Nürnberger Chroniken I, – Nürnberg ) () das haben wir nicht wollen verhalten that have we not want keep/hide ‘we did not want to hide this’ (Nürnberger Chroniken I, – Nürnberg ) () aber dieselben . . . haben nicht künnen ermercken but the.same have not can realize ‘but they could not realize’ (Nürnberger Chroniken II, – – Nürnberg ) () er häts vor den heiden nit dürfen thun. he had.it in.front.of the heathens not may do ‘He should not have done it in front of the heathens.’ (Rozmital – c. ) () das man die selbigen gräfin nach kunig Lassle tod that one the same countess after King Lassle death nie hab sehen lachen never have see laugh ‘that one has never seen this countess laugh after King Lassle’s death’ (Rozmital – c. )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
() herre warumb hast vns machen irren Lord why have us make stray ‘Lord, why did you make us stray?’ (Zainersche Bibel, Jes. , – ) () wer hot dich lernen straiten who has you learn fight ‘Who taught you how to fight?’ (Wunderer, ms. D, , – around ) () wyle cristus nit mit sinem exempel yemant hat leren betteln because Christ not with his example anyone has teach beg ‘because, with his example, Christ did not teach anyone to beg’ (Wyle, Translationen , – around ) According to Kurrelmeyer, the modal sollen ‘shall, should’, cf. (), very rarely occurs in IPP constructions and only since the sixteenth century. This is because with modals, especially with sollen, a competing construction in which the modal is promoted to verb one, i.e. occurs as the finite verb (see example () in Section .) is in fact far more common than IPP. () und da man den armen hat sollen geben, da . . . and when one the poor has should give then ‘and when one was supposed to give to the poor, then . . . ’ (Augsburger Chronik VI, – Augsburg ) As Ebert et al. () report, the IPP construction is occasionally also found with wissen ‘to know’ in the sixteenth century, cf. (). () man hat nicht viel vom Aristotele wissen zu sagen one has not much of Aristotle know to say ‘One did not know much to say about Aristotle.’ (Aventin r – ) In the seventeenth century, it is further extended to anfangen ‘to begin’ and pflegen ‘to use to’, as illustrated in () and (). () a. , den . ianiuarii, hab ich in br. Albrechten hauss anfangen year the January have I in brother A.’s house start bawen build ‘On January , I started to build in brother Albrecht’s house.’ (Schad, Reisebuch, – around ) () unbesunnen Urteil hab ich jederzeit mehr zu verachten rash judgement have I always more to despise als zu achten pflegen than to honour use.to ‘I always used to despise rather than honour a rash judgement.’ (Opitz, Poemata – )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
The verbs brauchen ‘to need’, fühlen ‘to feel’, and suchen ‘to seek’ also first occur in IPP constructions in early NHG.6 At the same time, normative grammarians such as Adelung and Rückert began to criticize the IPP construction. Thus Adelung holds that with hören ‘to hear’, i.e. one of the first IIP verbs, allowing IPP already since the early thirteenth century, the actually expected past participle was the more regular and nobler form than the infinitive. In contrast to the constant spread and extension of the IPP construction in ENHG times, the number of IPP verbs has been decreasing again in Modern German. For instance anfangen, pflegen, and wissen are generally construed with the expected past participle construction rather than with IPP, cf. ().7 () Er hat mit zu rauchen {angefangen / *anfangen}8 he has at to smoke started[=PAST.PTCP] start[=INF] ‘He started to smoke at sixteen.’ In the closely related language of Dutch, the number of IPP verbs is much greater today than in Modern German, cf. Ponten (), see also Schmid (: , ). The historical development is comparable, however, insofar as the first attestations of IPP are found in the thirteenth century with causative doen ‘to do/make’, viz. heeft doen + infinitive instead of heeft gedaan + infinitive ‘has done/made . . . ’. During the fifteenth to seventeenth century, it is still mainly causatives but also modals that occur in the IPP construction in Dutch. The historical development in German, and the largely parallel development in related languages such as Dutch, shows that verb class is an important factor for IPP. This is supported by recent cross-linguistic research: based on a comparison of various languages that show the IPP phenomenon, Schmid (: ) has formulated the implicational hierarchy of potential IPP verbs given in ().9 ()
causatives < modals < perception verbs < benefactives < duratives < inchoatives < control verbs
6 According to Kurrelmeyer (), lehren, brauchen, fühlen, wissen, pflegen, and suchen are attested in IPP constructions only as late as NHG because only after NHG do these verbs allow for embedding of a bare infinitive. With zu-infinitives, the past participle construction tends to occur rather than the IPP construction (but see fn. ). Note that, according to Maché (: , ), brauchen only occurs in IPP constructions as well as with bare infinitival complements after the early nineteenth century. 7 A change in verb order within the verb cluster does not make a difference to the (un)grammaticality in Modern German, here. Also, the infinitival marker zu cannot be omitted: Er hat mit angefangen/ *anfangen *(zu) rauchen. This fact is presumably linked to the ungrammaticality of IPP here since, as Kurrelmeyer () observes, zu-infinitives tend not to go with IPP, but compare Modern German brauchen that occurs as IPP even if embedding a zu-infinitive: Du hättest nicht zu kommen {brauchen, */?gebraucht} (‘you did not need to come’) as well as the historical examples of IPP with an embedded zuinfinitive such as () and (). 8 Note, however, that IPP with anfangen is optionally still possible in certain dialects, cf. for Alemannic varieties Schallert (b: ): dass as {aagfanga [=PAST.PTCP]/aafanga [=INF]} hot rägna ‘that it has started to rain’. Here, unlike in the standard language, zu can be omitted with the verb selected by anfangen and a typical reordering of the verbs occurs. 9 On the basis of additional dialectal evidence, Schallert (b: ) suggests a slightly altered version of this hierarchy: causatives < modals < perception verbs < inchoatives < benefactives.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
T . IPP verbs and verb class in the history of German IPP verb
verb class
th century
tuon hoeren heizen (müezen) (helfen) (lazen)
causative perception verb causative modal benefactive causative
th century
lazen bitten (türren)
causative control verb control verb (cf. Schmid : ff ‘dare to’)
th century
türren müezen mögen wollen dürfen können sehen machen helfen lehren/lernen
control verb modal modal modal modal modal perception verb causative benefactive benefactive
th century
sollen wissen
modal control verb (cf. Schmid : f for Dutch/WF weten)
th century and later
anfangen pflegen brauchen fühlen suchen
inchoative inchoative (cf. Schmid : f for Dutch/WF plegen) modal perception verb control verb
The more ‘auxiliary-like’ a verb is, notably causatives and modals, the more likely it is to allow for IPP. The historical development roughly supports this hierarchy. Table . summarizes the development in German as discussed above, giving the verb class for each IPP verb. The verbs in bold face still occur in IPP constructions in Modern German, cf. Schmid (: ).10 The most common verb class among the early IPP verbs which are attested in this construction as early as the thirteenth century is indeed the class of causative verbs, as predicted by the implicational hierarchy. However, causative machen is used in IPP constructions only in the fifteenth century. Most modals are found in IPP constructions before the perception verbs sehen and fühlen, as expected. The perception verb 10
Verbs that do not yet regularly appear in IPP constructions during the given period, and only in single attestations or in morphological forms that cannot unambiguously be identified as infinitives, are written in parentheses in Table ..
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
hoeren, on the other hand, is among the very first IPP verbs. Benefactives generally occur with IPP later than modals; however, there is rare evidence for benefactive helfen even as early as the thirteenth century. As expected, the IPP is extended to inchoatives only several centuries later. With control verbs, on the other hand, the picture is not so clear. While, as expected, there are among the first IPP verbs, some control verbs are attested in IPP constructions relatively early on, along with or even slightly before most modals. Others are among the last verbs to be used with IPP. The verb class of duratives (e.g. ‘stay’, ‘go’, ‘lie’, ‘walk’, ‘stand’, ‘sit’, ‘be’, cf. Schmid : f in Dutch and West Flemish IPP) is surprisingly not attested with IPP in historical German in spite of the fact that several inchoatives and control verbs are. While there is thus some support for the implicational hierarchy from the diachronic data, the extension of the IPP construction clearly does not proceed verb class by verb class. Whether a verb occurs in the IPP construction at a certain point in time, however, also depends on other factors, for instance whether the verb allows for the respective causative, benefactive, etc. reading at that time, and whether the relevant morphological verb forms are available yet. Note that, while the loss of IPP in Modern German times again did not strictly proceed by verb class, this development also corroborates the hierarchy insofar as control verbs and inchoatives disappeared from the IPP construction whereas modals and causatives (and occasionally benefactives and perception verbs) still occur with IPP today.
. Historical origin and theoretical accounts of the IPP construction .. IPP from ge-less past participle Several theories have been suggested on the origin of the IPP construction. The classic theory goes back to Lachmann (: ) and Grimm (: ff). They suggest that the IPP developed on the basis of (strong) past participles that could also occur in a variant without the prefix ge- and contained a stem vowel identical to that of the infinitive. The MHG past participles (ge)lazen, (ge)heizen, and (ge)sehen (as well as the strong past participles of some MHG modals, viz. (ge)mögen, (ge)wollen) in their prefixless variants would be formally identical to the respective infinitives. These forms could accordingly be reanalysed as infinitives and the construction was then extended to other verbs. (Recall, however, that past participles of modals only occur significantly later than the first attestations of IPP in German.) This theory was later endorsed by Paul (: ff) and Behaghel (–, Vol. II: f).11 However, they modified it slightly in that they discarded heizen and sehen,12 and considered only the ge-less participle of lazen as the origin of the IPP construction. According to Behaghel (–, Vol. II: f), the IPP construction originated with lazen where infinitive and past participle were formally identical—not originally, however, as the ordinary past participle was gilazen, but only in the close connection to an embedded infinitive in which ge- supposedly lost its function (infolge von Funktionslosigkeit in der engen Verbindung mit dem Infinitiv). Later, Behaghel proposes, the same syncretism would have occurred with heizen and sehen, which were then reanalysed as infinitives unless they were already simply formed by analogy to the reanalysed lazen. 12 For empirical reasons, thus Behaghel (–, Vol. II: ) states that heizen (originally ‘to be named’) only acquired its causative meaning later. 11
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
Several modern theories of IPP (e.g. Askedal ; Plank ; de Vos ; Wurmbrand ; Zwart ; Hinterhölzl a) also treat the IPP as a kind of allomorph of the past participle rather than as a true infinitive, thus avoiding the morphosyntactic problem of a violation of the selectional requirements of the perfect auxiliary. Some authors assume that the prefix ge- is originally there in the structure but not spelled out, or overridden by some mechanism (e.g. de Vos ). Others propose that the participle bears an equivalent zero morpheme instead (e.g. Hinterhölzl a). It has also been suggested that the participle bears an additional feature in the respective contexts, which leads to the infinitive-like spell-out form (e.g. Zwart : feature [unit]). One question that arises here, however, is how the various constructions competing with IPP can be explained (see Section .). The classic theory of IPP as originating in the reanalysis of ge-less participles formally identical to the respective infinitives faces certain empirical as well as conceptual challenges: IPP occurs also with verbs for which the infinitive and the past participle are clearly morphologically different. According to the data collected by Kurrelmeyer (), it is not so much lâzen, heizen, and sehen,13 but tuon and hoeren (as well as rarely helfen and müezen) that are attested in the early IPP constructions, i.e. verbs for which the past participle is indeed clearly distinct from the infinitive by virtue of the stem vowel and/or the weak past participle suffix ‐t. The same point can be made for Dutch: the first IPP cases are ones including the infinitive doen instead of the clearly morphologically distinct past participle gedaan (cf. Coupé and van Kemenade ). The equivalent laten of MHG lâzen is also frequently attested in Middle Dutch IPP constructions. However, only the prefixed form of the past participle ghelaten is attested otherwise, i.e. there is no ge-less variant of the past participle (cf. Ponten ). Conceptually, the restriction of IPP to certain constructions is not straightforwardly explained under the participle approach: if the use of the infinitive instead of the past participle is simply based on the formal identity of the two forms with certain verbs and an ensuing general reanalysis, one might expect the same phenomenon to occur in perfect or past perfect constructions involving just two verbs. However, there is a syntactic restriction of IPP to clusters of three (or more) verbs, i.e. the embedding of an infinitive under the IPP verb is a necessary precondition for IPP to occur at all.14 Furthermore, IPP is partly syntactically restricted with respect to the ordering of the verbs involved, cf. Section .. Finally, the semantic restriction of IPP by the verb class remains unexplained under the participle approach.15 13 According to Kurrelmeyer’s () data, sehen is first attested in IPP constructions only in the fifteenth century (and only rarely). IPP including heizen and lâzen is found earlier but, at least with heizen, the usual past participle form in MHG is geheizen, whereas the ge-less variant is very rare. 14 Note that a kind of IPP construction with only two verbs is in fact attested in MHG/ENHG: ich het nit fürpaz kunnen—literally: I had not further can ‘I would not have been able to go further’ (Buch von den Wienern , – after Biener : ). However, as is already evident from the translation, cases such as this one in fact constitute elliptical constructions in which the embedded infinitive V3 has been omitted. 15 Eroms (: f) mentions as an additional argument against the participle approach the observation that the use of the actually expected past participle instead of IPP is recently increasing in Colloquial German. This, he contends, would be unexpected if the infinitive in fact already constituted a past
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
.. IPP as a true infinitive In view of the empirical and conceptual problems of the participle theory, it has been suggested that the IPP does not originate in a reanalysed ge-less past participle, but in fact constitutes a true infinitive from the start. Erdmann (: ) proposed that the infinitive was used instead of the past participle as the result of an assimilation process (or in analogy) to the embedded infinitive, crucially building on the syntactic restriction of IPP to three-verb clusters. He offers a vaguely psycholinguistic explanation of interference by the dependent infinitive: the speaker already has the infinitive form of the dependent infinitive in mind and therefore also uses the higher, closely linked verb in the infinitive form.16 Erdmann’s theory on the origin of IPP was basically taken over by Merkes () and Kurrelmeyer () on the basis of the empirical diachronic evidence against the participal theory. As mentioned previously, the verbs tuon and hoeren, the past participles of which differ clearly from the infinitives, are most common among the first IPP constructions according to Kurrelmeyer’s data.17 Kurrelmeyer further stresses the syntactic restriction to threeverb clusters and the dependence on the form of the embedded infinitive of V3, noting that V2 generally does not appear as a bare infinitive if V3 is a zu-infinitive, which supports the assimilation hypothesis (but see fn. in Section .). Dal () and Ponten () offer a different explanation for why the infinitive came to be used instead of the past participle by suggesting an assimilation in the sense of both a formal as well as a semantic syncretism between infinitive and past participle. Formally, there is a more general approximation between the past participle and the infinitive as not only ge-less past participles occur in historical German (see above) but also infinitives prefixed by ge-. According to Dal (), the resulting uncertainty in use, rather than a reanalysis as according to the participle theory, lead to the use of the infinitive instead of the past participle (compare ich hân gesehen ‘I have seen’ versus ich mac gesehen ‘I can see’), but also to the phenomenon of PPI (participium pro infinitivo), see Section .. Furthermore, there is a semantic assimilation according to Dal () and Ponten (). Crucially, the IPP construction is appearing at a time when the periphrastic/ analytic tense forms are beginning to be grammaticalized (see also Chapter ). Through this grammaticalization, the new auxiliaries take over the function of temporal and categorial marking to a large extent so that the past participle is assimilated to an infinitive in grammatical function (denoting a process or action rather than a
participle allomorph. However, it could also be explained by proponents of the Lachmann/Grimm hypothesis as diachronic reduction of past participle allomorphy. 16 The same explanation was independently proposed for Dutch in the seminal work by van Helten (). Note, however, that there is no assimilation to a following infinitive, e.g. in dass er den Film [gesehen/*sehen] haben muss ‘that he must have seen the movie’. 17 Kurrelmeyer () suggests that the IPP construction originated with hoeren on the basis of the common MHG combination of hoeren + infinitive, especially with sagen/jehen (cf. also Dal ). However, according to Schmid’s () implicational hierarchy of IPP verb classes as stated in (), it would be more likely that IPP originated with the other one of the two most common early IPP verbs identified by Kurrelmeyer, viz. causative tuon, rather than the perception verb hoeren.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
state). In those three-verb clusters in which the participle is especially closely linked semantically to an infinitive (according to Ponten, it constitutes a mere ‘connotation’ of the infinitive, i.e. it ‘extends the infinitive by one semantic feature’), the verb that would be expected to occur as a past participle therefore appears in the neutral form of the infinitive, yielding the semantic restriction of the IPP construction. The view that the IPP constitutes a true infinitive has also been taken by several authors in more recent theoretical analyses of IPP in synchronic terms, e.g. by Evers (), Schmid (), and Keine (). Evers assumes that the perfect auxiliary may exceptionally govern a bare infinitive in these constructions. By contrast, Keine assumes an obligatory extraposition of the IPP and embedded infinitive outside of the c-command domain of the perfect auxiliary thus leading to a constellation where the non-finite V2 is canonically spelled out as an infinitive. Schmid, on the other hand, regards the IPP construction as a true violation of the selectional requirements of the auxiliary, which, however, under an optimality-theoretic (OT) perspective, constitutes a last resort or repair strategy in order to fulfil higher-ranked constraints against infinitival complements of past participles as well as against ge-prefixation. Indeed, there is a noteworthy link between the participle prefix ge- and IPP, cf. (): IPP languages/dialects (e.g. German, Swiss German, Dutch, Afrikaans, West Flemish) form past participles by means of ge-prefixation,18 while none of the closely related Germanic languages that do not use the ge-prefix (e.g. English, Frisian, Low German) show the IPP effect. The correlation seems to work only one way, though, insofar as not every language with ge-participles has IPP, one notable exception being Yiddish.19 () Typological correlation: IPP language ! past participles with ge-prefix Of course, the question remains why ge-prefixation is precluded in the respective contexts. This question is in fact independent of whether the IPP constitutes a ge-less participle or a true infinitive. Schmid () assumes that ge- is marked because it is a separate functional head, i.e. it introduces extra structure and is thus more costly than a ge-less form. IJbema () assumes that, like unaccusatives, a ge-prefixed past participle lacks the specifier position that is necessary, in her Minimalist account, for checking the embedded infinitive V3; to allow for this checking of V3, V2 appears as an infinitive, i.e. IPP. Historically, the prefix gi-/ga-/ge- had a telic (or perfectivizing) function in the Germanic languages. Therefore, it did not occur on inherently telic verbs (consider the past participles bracht ‘brought’, funden ‘found’, troffen ‘met/hit’, worden ‘become’) or verbs that could only have an atelic reading (e.g. wissen ‘to
18 A possible exception is West Frisian Dutch, which shows IPP with some verbs but does not form past participles prefixed by ge-. Hoekstra and Taanman (), however, attribute this fact to the bilingualism with Standard Dutch. 19 This fact contradicts Lange (, ) and Hoeksema (), who suggest that the link is a one-toone correlation. Yiddish is not the only exception: as Schallert (a: ) mentions, most Southern Bavarian dialects as well as certain southern Low German dialects (Eastphalian) have ge-prefixed past participles but no IPP.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
know’, heizen in the sense of ‘to be called’). The prefix could originally occur on nonfinite (including past participle and infinitive) as well as finite verb forms. The IPP phenomenon might thus be linked to the historical function of the prefix. Coupé and van Kemenade () analyse ge- as a CAUSE light-verb that is accordingly incompatible with causative verbs, the verb class that is most likely to show the IPP effect cf. () above. Zwart () suggests that three-verb clusters headed by a temporal auxiliary are inherently atelic and therefore ge- is precluded.20
. Related phenomena: IPP as one repair strategy among others Recent theoretical analyses view IPP as one repair strategy among others. This approach is corroborated by diachronic as well as dialectal data, supporting once more the crucial role of dialectal data for diachronic research, cf. Weiß (); Axel and Weiß () among others. According to Schmid’s () optimality-theoretic analysis of IPP, the grammatical well-formedness constraint according to which a perfect auxiliary selects a past participle is violated by IPP. If the opposing markedness constraint against infinitival complements of past participles is ranked higher than the faithfulness constraint on morphological selection, some repair strategy will apply instead of the expected construction of auxiliary + past participle + infinitive. In Schmid’s approach, this constraint comes with a markedness subhierarchy that reflects the implicational hierarchy of verb classes so that the occurrence of substitute forms with one verb class but not others can be accounted for depending on where the well-formedness constraint on morphological selection is inserted in this hierarchy. IPP is but one possibility to avoid the normally expected construction. Whether IPP occurs or not also depends on whether the constraint against ge-prefixation applies in the language in question (because many languages do not use ge-prefixation to form their past participles), on the ranking of the markedness constraint against ge-, and on a further, more specific constraint on morphological selection demanding that the selectional properties of V1 be satisfied. If the markedness constraint against geapplies, IPP results. If it does not apply at all, an alternative repair strategy is predicted, for instance that of PPI (participium pro infinitivo). In PPI constructions, infinitive embedding under a past participle is avoided by the infinitive (V3) also taking the form of a past participle so that the verb cluster consists of a finite perfect auxiliary and two past participles, compare the Frisian example in ().21 20 Compare also Eroms (), who suggests that there is a contradiction between the IPP verb referring to ‘a not yet realized action’ as opposed to the past participle referring to ‘an already realized action’. Eisenberg, Smith, and Teuber () similarly point to an incompatibility of the semantics of modals with the temporal structure of past participles. More recently, Gaeta () argues in a similar fashion that IPP originally occurred because of a semantic incompatibility of causatives and perception verbs with the perfectivizing/telic prefix ge- that was only overcome once the perfect tense was fully grammaticalized in German. Note that, for instance, in two-verb clusters, past participles of modals are of course perfectly grammatical in German, e.g. Ich hab das nicht gewollt ‘I did not want that’. 21 Note that wollen constitutes the past participle form in Frisian, whereas the corresponding infinitive is wolle (cf. den Dikken and Hoekstra : ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
() Hy soe it dien wollen ha he would it done[=PAST.PTCP] wanted[=PAST.PTCP] have[=INF] ‘He would have liked to do it.’ (den Dikken and Hoekstra : ) Schmid sees a shortcoming of her analysis in the fact that PPI is also predicted, if the constraint against ge- applies but is ranked sufficiently low (Schmid : f, fn. , , ). She assumes that there is a clear division between IPP and PPI languages in that the prefix ge- is completely lacking in the lexicon of the latter (Schmid : ). According to Schmid (: , ), a language with ge-participles that also shows PPI is not attested but cannot be excluded by her account. In the words of Schmid, it would be ‘a non-existing language with German vocabulary but PPI’ (Schmid : ). In this respect, historical German represents a particularly interesting case that actually fulfils the predictions of Schmid’s account, as we will see below. The prediction of the possibility of alternative repair strategies is borne out in a diachronic perspective and also synchronically at a dialectal level (compare what Höhle refers to as three-verb phenomena). Several alternative constructions diachronically compete with the IPP construction, compare the data below. The IPP construction itself can be characterized as follows (the order of verbs within the verb cluster is neglected here, for a discussion of the issue of verb order see Section .): V1: finite Aux + V2: infinitive (= IPP) + V3: infinitive Competing with the IPP construction, we first of all find the actually expected construction with a past participle, cf. () and (). (In terms of an optimalitytheoretic approach, it occurs if the faithfulness constraint on morphological selection is ranked higher than the markedness constraint against infinitive embedding under a past participle.) V1: finite Aux + V2: past participle + V3: infinitive () Von wilden getwergen [V1 hân] ich [V2 gehoeret] [V3 sagen] of wild dwarfs have I heard say ‘I have heard things said about wild dwarfs.’ (Nibelungenlied (C) , – first half thirteenth century) () [V1 haben] wir unserr beyder insigel [V2 geheizzen] [V3 henken] have we our both seal let hang ‘have we had both our seals hung [to this letter]’ (Hessisches Urkundenbuch II, , , – – Hanau ) This type is generally older than the IPP construction and for a long time also more common (cf. Kurrelmeyer for heizen etc.). The only exception are modals: their weak participles, which were only formed around , are attested much more rarely than the IPP. In Colloquial German and the Modern German dialects, this type of construction is partly more common than in the standard language. For instance,
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
as illustrated in (), it is also used with modals, which obligatorily occur with IPP in the respective constructions in the standard language: () a [V1 hoots] nee [V3 soorn] [V2 gedorft] he has.it not say might ‘He was not allowed to say it.’ (Silesian: Arnsdorf – after Biener : ) The past participle may also occur in this construction without the prefix ge-, thus obeying the ban on ge-. Nonetheless, the form is generally distinct from the infinitive by virtue of its different stem vowel and/or weak participal suffix ‐t, cf. (). V1: finite Aux + V2: ge-less past participle + V3: infinitive () [V1 hats] [V2 must] [V3 fort gahen] has.it must[=PAST.PTCP] away go ‘it had to go away’ (Luther, Frühe Schriften , – ) Behaghel (–, Vol. II: f) refers to this type as ‘mixing’. This construction with a ge-less participle is reminiscent of a construction including what Höhle () calls the supine. The latter term refers to a non-finite verb form characterized by the weak participle suffix, absence of the ge- prefix, but sometimes a different stem vowel (Ablaut). It is used instead of the past participle or IPP in certain Modern German dialects, e.g. Thuringian cf. (),22 suggesting that this type of construction has survived in regional varieties of German despite being lost from the standard language (cf. Luks and Schwalm ). Comparable data for other dialects is given by Biener (), for example (). () A hads musd mache. he has.it must make ‘He had to do it.’ (Thuringian: Kranichfeld near Weimar – after Höhle : ) () a hett kunt eendr kumma. he had could earlier come ‘He could have come earlier.’ (North Bohemian – after Biener : ) Whether IPP or the ge-less participle/supine is used depends on the particular verb in some dialects (e.g. in Kranichfeld, lassen occurs with supine or IPP, helfen, hören, sehen, and others only with IPP), or on the order of verbs within the verb cluster (IPP with V3-V2 order, otherwise supine, cf. Höhle : ). The historical link to the 22 A variant of this construction is also found in four-verb clusters in Thuringian: Er hat sich must operlasst-riere ‘He had to have an operation’ (literally: He had himself must oper-let-rate). Here, not only the modal occurs in the supine or ge-less participle form but also the verb selected by it, which should normally occur as an infinitive—in this case the typical IPP verb lassen ‘to let’, which interestingly even morphologically splits up the infinitival main verb (V4) operieren ‘to operate’.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
particular function of the ge-prefix and to its use with verb forms other than just the past participle is supported by the dialectal data, too, in so far as the dialect area where the ge-less participle or supine occurs overlaps with the area in which ge-prefixed infinitives are found. Another construction that diachronically competes with the IPP construction is one already briefly mentioned above where V2 is promoted to V1 and thus infinitiveembedding under a past participle is avoided in a way different to the IPP construction. At the same time, the selectional requirements of the auxiliary—otherwise V1, now V2—are fulfilled since V3 occurs in the form of a past participle, cf. (). V1: finite V (= otherwise V2) + V2: infinitive Aux (= otherwise V1) + V3: past participle () von Veldeke der wîse man / der [V1 kunde] se baz [V3 gelobet] [V2 hân]. von Veldeke the wise man the could her better praised have ‘Von Veldeke, the wise man, he could have praised her better.’ (Parzival , – – around –) According to Kurrelmeyer () and Biener (), this alternative repair strategy is the most common one in MHG and into ENHG. Especially with modals, it is much more frequently attested than the IPP construction, whereas it is only occasionally found in Modern German.23 Yet another repair strategy, which could be considered as a variant of the IPP construction, avoids infinitive-embedding under a past participle but at the same time at least partly fulfils the morphological selection requirement demanding ge- on V2 by using an infinitive instead of a past participle (thus comparable to IPP), but an infinitive prefixed by ge-, as in () (Behaghel –, Vol. II: f refers to this as another case of ‘mixing’):24 V1: finite Aux + V2: ge-infinitive25 + V3: infinitive () niemant [V1 hat] es [V2 gedoerffen] [V3 sagen] nobody has it may[=ge-INF] say ‘Nobody was allowed to say it.’ (Hutten, Gespräch-Büchlin r – )
23 Eroms (: –) points out that this construction generally gives rise to an epistemic reading of the modal in contrast to a non-epistemic reading with IPP. However, epistemic readings of modals only appear during MHG/towards ENHG. The mere number of attestations of this construction in MHG/ ENHG suggests that the observed correlation with epistemic/non-epistemic readings is a more recent phenomenon. Indeed, in (), the modal receives a non-epistemic interpretation. 24 In Schmid’s () account, this kind of construction is quite unexpected: V2 violates the constraint on morphological selection and the constraint against ge-. One possibility to account for these data would be to split up the constraint on morphological selection into subconstraints requiring the individual parts of the past participle morphology (suffix + possibly stem vowel, prefix). 25 Note that, as indicated above, ge-prefixed infinitives are also attested in two-verb constructions in historical German (cf. also Section ..) as well as in certain Modern German dialects such as Thuringian and East-Hessian, cf. Schirmunski (). An alternative historical account for these ge-prefixed forms of modals would be to consider them as newly formed strong past participles for these preterite-present verbs rather than as ge-prefixed infinitives.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
The alternative repair strategy of PPI again avoids infinitive-embedding under a past participle, but this time by V3 occurring as a participle instead of an infinitive, as mentioned above. PPI is also attested in the history of German, cf. (), (however only rarely in early chancery documents). V1: finite Aux + V2: past participle + V3: past participle (= PPI) () [V1 hand] wir unser eigen ingesigel [V2 geton] have we our own seal done[=PAST.PTCP] [V3 henket] hung[=PAST.PTCP] ‘we have had our own seal hung (to this letter)’ (Urkundenbuch Basel II, – Basel ) The existence of constructions like these is particularly interesting in terms of the theory of IPP and PPI: it proves contra Schmid () that there is no clearcut division between IPP and PPI languages, but that both phenomena may occur in one and the same language. Furthermore, it shows that the abovementioned ‘non-existing language’ with German vocabulary predicted by Schmid’s account, i.e. a lexicon including the prefix ge-, and PPI did in fact exist a few hundred years ago.26 The typological correlation ‘If ge- then no PPI’ (Schmid : , fn. ) does not hold. PPI constructions are also attested in first language acquisition of German and in fact appear in the respective three-verb clusters before IPP constructions are used; an example is given in (). (At this acquisitional stage, past participles are consistently formed without ge-prefixation and with the weak participle suffix ‐t, but already including Ablaut where appropriate, thus in a form reminiscent of the supine that is found in certain dialects as discussed above.) () [V1 Hat] de Baby [V3 runterfallt] [V2 lasst] has the baby dropped[=PAST.PTCP] let[=PAST.PTCP] ‘The baby has let it drop.’ (Alma, age ;) Incidentally, a kind of PPI construction is also attested in two-verb clusters in MHG: V2 occurs in the form of a past participle despite of the finite verb V1 selecting a bare infinitive cf. () and () (see also Paul : f). () ich [V1 tuon] zehant nâch iu [V2 gesant] I do immediately after you sent[= PAST. PTCP ] ‘I will (let them) send for you, immediately.’ (Tristan – around ) 26 In other words, the fact that Schmid’s () account cannot exclude this possibility is not a shortcoming but is actually empirically supported. Note, however, that of the two participles in (), only one bears a ge-prefix. Nonetheless, the historical German data show that contra Schmid (: ), her constraint ranking *PastP/+Inf-MV >> Morph >> *ge- >> MorphV1 cannot be the same in all PPI languages, but a ranking of *ge- below MorphV1 is necessary to capture these data.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
() waz [V1 sol] lenger hie [V2 gelegen] what should longer here lie[=PAST.PTCP] ‘Why should I lie here any longer?’ (Herbort – ) Paul (: ) and Dal () propose that in turn this construction is due to a reanalysis of infinitives prefixed by ge- as past participles. These two-verb clusters seem to refute, at least for this time period, the syntactic restriction to three (or more)-verb clusters, which otherwise strictly holds for IPP and the alternative repair strategies (thus the term ‘three-verb phenomena’ coined by Höhle ).27 In this context it is noteworthy that, even though this is not a regular case of IPP, the same correlation between two-verb cluster and regular ge-participle on the one hand versus three-verb cluster and lack of ge- on the other hand can be observed with the verb werden in Modern German as illustrated in ().28 () a. Er [V1 ist] alt [V2 geworden] he is old become ‘He has grown old.’ (werden = main verb) b. Er [V1 ist] [V3 gelobt] [V2 worden] he is praised become ‘He was praised.’ (werden = passive auxiliary) Returning to the constructions diachronically competing with IPP, we finally find a repair strategy that so to speak combines IPP and PPI in so far as V2 does not take the selected form of a past participle but occurs as an infinitive as in IPP constructions, while V3 does not take the selected form of an infinitive but occurs as a past participle as in PPI constructions, cf. (). V1: finite Aux + V2: infinitive (~ IPP) + V3: past participle (~ PPI) () dez [V1 han] wir vnser kunichlich Insigel an disen brieff [V2 haissen] that have we our royal seal on this letter let[=INF] [V3 gehenket] hung[=PAST.PTCP] ‘We have therefore had our royal seal hung to this letter.’ (Monumenta Zollerana I, (Urkunde CXXXII) – Stuttgart ) This construction is quite common in late thirteenth and early fourteenth century chancery documents (cf. Ebert et al. : §S ). Again dialectal evidence shows that, while this construction is not found in the standard language, it has survived
See also fn. (Section ..) on historical IPP in (apparent) two-verb clusters. Compare also the correlation between regular prefixed past participle of modals in two-verb constructions (ekont, ewolt, etc.) versus prefix-less past participle forms of the same verbs in three-verb constructions (kont, wolt, etc.) in certain southern Low German dialects that do not show the IPP effect, cf. Schallert (a: ). 27 28
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
until today in regional varieties of German (cf. Höhle ), as in the Thuringian example in ().29 () ich [V1 hou] [V2 half] [V3 gearbet] I have help[=INF] worked[=PAST.PTCP] ‘I helped working.’ (Thuringian: Wasungen – after Höhle : , fn. ) In terms of the optimality-theoretic analysis sketched above, this type of construction violates two faithfulness constraints on morphological selection in order to fulfil the markedness constraint against infinitive-embedding under a past participle. Accordingly, it is predicted that this construction ‘never occurs as optimal’ (Schmid : ), and that IPP should always be preferred.30 Thus, it is particularly noteworthy that it does in fact occur. According to Behaghel (–, Vol. II: f), we are dealing here with wrong standardization (falsche Verhochdeutschungen), i.e. a kind of hypercorrect construction, due to an uncertainty as to which of the verbs should occur as an infinitive and which as a participle. However, this type of construction is probably more adequately described as a transmission or displacement of selectional properties: V3 instead of V2 may fulfil the selectional requirements that V1 actually imposes on V2, while V2 occurs in the default form of the infinitive. (The IPP construction can be described by analogy as a complete suspension rather than a transmission or displacement of selectional requirements.) Further phenomena including a displacement of selectional requirements are attested in Modern German dialects (cf. Höhle ). In the example in (), V1 is a verb that normally selects a ge-infinitive in this variety, but V3 instead appears as a ge-infinitive despite V2 normally selecting a normal infinitive. () E [V1 kon] en ju [V2 las] [V3 gekom] he could him PTC let[=INF] come[=ge-INF] ‘He could let him come.’ (Thuringian: Kleinschmalkalden – after Höhle : ) A similar displacement is also attested with gerund and zu-gerund in Thuringian as well as in Swiss German varieties, cf. Höhle (: ). According to Höhle (: ), immediate satisfaction of V1’s selectional requirements is excluded when the complement V2 selects a verb V3 by itself. Satisfaction of V1’s requirements is mediated to V3, while V2 appears in a substitute form, e.g. the bare infinitive.31
29 Note that, according to Höhle (: , fn. ), the form half constitutes an infinitive rather than a ge-less past participle/supine. Contra Zwart (: f), this pattern is thus not unique to Afrikaans—and accordingly cannot reside in a different status of ge- in Afrikaans versus German, Dutch, etc. 30 In other words, it is harmonically bounded by the IPP-candidate, cf. Schmid (: ). 31 The transmission or displacement of selectional requirements is order sensitive in some dialects, in that the immediate satisfaction by V2 of V1’s formal selectional requirement is excluded only if V2 < V3 (Höhle : ).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
. Order of verbs in IPP constructions and syntactic analysis As mentioned above, the general order of verbs within the verb cluster in Modern Standard German IPP constructions is V1-V3-V2, deviating from the normal order at the right periphery of V3-V2-V1. (Note that V1 may of course secondarily move from the verb cluster and appear in the left sentence periphery in verb-first or verb-second clauses, cf. Chapters –.) Theoretical accounts of IPP differ as to whether they take the reordering within the verb cluster to be accidental (Eroms ; Zwart ; Schallert b), the consequence (Eisenberg, Smith, and Teuber ) or the cause of the IPP effect (Hinterhölzl a; Keine ). The reordering within the verb cluster may be captured syntactically by verb (projection) raising in the sense of Haegeman and van Riemsdijk (),32 by remnant movement of verbal projections, by PF-movement or by base-generating this order. The syntactic analysis of IPP constructions is thus closely linked to the question of verb cluster formation and of underlying word order (see also Chapters , , , and ). Based on the data presented in an earlier version of the present chapter, as well as the relations to the work by Schmid () and Höhle () drawn there, Hinterhölzl (a), for instance, assumes a consistently head-initial underlying word order with remnant movement of a verbal projection of V3 to a specifier position of an aspectual phrase left of V2 for licensing reasons (viz. temporal anchoring). In this specifier position, ge- would normally be merged and is accordingly blocked. Simplifying somewhat, the suggested structure looks as given in (). ()
VP1 AspP
V1
VP2
geVP3
V2
tVP3
V3 By contrast, Keine (), as indicated in Section .., derives the order on the basis of a head-final underlying structure by obligatory extraposition of the verbal projection of the IPP verb. This VP also contains the verbal projection of the embedded infinitive. The assumed movement is due to a stipulated feature on VP2 which must not appear in the c-command domain of a perfect auxiliary. A non-finite verb that is not c-commanded by a perfect auxiliary in turn is regularly spelled out in the form of an infinitive, which explains the IPP effect on V2. The simplified structure looks as in ().33
On the concepts of verb raising and verb projection raising see also Chapters and . What is referred to as VP1 is in fact a mere vP, and VP3 includes its own vP-shell according to Keine (). 32 33
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German ()
VP1 VP1 tVP2
VP2 V1
VP3
V2
V3
However, the occurrence of IPP with verb orders other than V1-V3-V2 that are attested in various languages as well as in dialectal and historical German data constitutes a challenge to theoretical accounts that link the phenomenon of IPP to this particular verb order. Cross-linguistically, all logically possible verb orders, including V2-V1-V3, are attested with IPP constructions, cf. Schmid (). The order V1-V2-V3 is especially common in IPP constructions across languages. This order is for instance used in Standard Dutch, whereas Dutch dialects also have V3-V2-V1 in IPP constructions and West Flemish has V2-V3-V1. In Modern German dialects, we also find orders deviating from the standard German IPP verb order (cf. Schmid : ; Höhle : ), consider (). () a. dass I das immer [V3 mache] [V1 ha] [V2 wöle] that I that always make have want b. dass I das immer [V1 ha] [V2 wöle] [V3 mache] that I that always have want make ‘that I always wanted to do this’ (Swiss German: Sankt Gallen – after Schmid : ) Nonetheless, there is an overall preference for V1-V3-V2 order in IPP constructions in the German dialects as in the standard language (cf. Sapp a). The preferred order in clusters including modals in the Alemannic dialect of the Vorarlberg region, however, is V3-V2-V1 with V1-V2-V3 being the second most common order according to Schallert (b: ). Yet, on the whole, the Alemannic varieties that are investigated by Schallert (b) show no direct correlation between IPP and the ordering of the verbs, either. Schallert (b: –) therefore suggests a much more flexible syntactic analysis with basegeneration of the verb cluster following assumptions in Bader and Schmid (). The IPP effect is thus not derivationally triggered as in the accounts by Hinterhölzl (a) or Keine (). IPP verbs are lexically specified in such a way that, as their left sister, only a simple verb may occur. The various possible syntactic structures including different verb orders base-generated by head adjunction of the respective verbs such as (), as well as structures with intervening non-verbal constituents (cf. nodes optionally selecting an NP), are subsequently evaluated by an optimalitytheoretic component extending the account by Schmid () by syntactic constraints, e.g. demanding that the lowest lexical verb be positioned at the left edge of the verb cluster, that the auxiliary stands at the top of the verb cluster, or
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Agnes Jäger
prohibiting verb-cluster formation etc., which will be ranked differently in different dialectal and diachronic varieties. () a.
VAux: NP ←
VAux: → VMod–[VV]
VMod: NP ←
hat/wird
VMod: V ←
VMain: NP ← lesen b.
wollen
VAux: NP ←
VAux: VMod ←
VMod: NP ←
wird VMain: NP ← lesen
VMod: V ← wollen VMain: NP ←
c.
VMod: → VMain
VMain: NP ← lesen
VMod: →VMain wollen
VAux: VMod ← hat
(VAux = V1, VMod = V2, VMain = V3, → = (optionally) selects the respective category as a complement in the respective direction, e.g. NP ← = selects an NP-complement to its left after Schallert b: –) Diachronically, the order of verbs in IPP constructions differs from that in Modern German. In the earliest German IPP constructions in MHG, the only attested verb order is V1-V2-V3 according to Kurrelmeyer (), compare also examples (), (), and (). Among Kurrelmeyer’s approximately IPP examples including tuon, there is not one with V3 before V2, not even in the sixteenth and seventeenth centuries. This also holds for all cases including helfen and hoeren from the thirteenth to the fifteenth century. The first rare evidence for the modern
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
On the history of the IPP construction in German
V1-V3-V2 order is found in Low German sources from the fourteenth century. Later on, it is also attested in Central German and since the fifteenth century, we find occasional examples in Upper German, suggesting a diachronic spread of this word order preference from North to South. In the sixteenth century and throughout the ENHG period, however, the older order V1-V2-V3 is generally still more common, cf. Kurrelmeyer (); Sapp (a). Note that V2 before V3 is still found, though rarely, in NHG, for instance in idiomatic expressions including IPP such as in (). () Es [V1 hat] nicht [V2 sollen] [V3 sein] it has not should[=INF] be ‘It was not meant to be.’ The preference in the standard language as well as generally in present-day dialects for the order V1-V3-V2, Sapp (a) concludes, is thus a more recent development. The present-day dialects cannot simply be said to continue the older stage of the language (an observation that is reminiscent of the result for negation and the types of Negative Concord in OHG and MHG as opposed to the present-day Negative Concord dialects, see Chapter ). Besides syntactic constraints, pragmatic constraints (notably focus) also play a role for the order within the verb cluster (cf. Schmid ; Sapp a).
. Summary IPP constitutes a particularly striking morphosyntactic phenomenon indicative of verb cluster formation and typically linked to a permutation of word order, viz. verb reordering in the right periphery of the clause. After rare attestations in MHG including the causatives tuon ‘to do’ and lazen ‘to let’ and the perception verb hoeren ‘to hear’, the construction becomes more common and spreads to further verbs in ENHG, roughly in line with Schmid’s () implicational hierarchy of IPP verbs. Regarding the two theories concerning the historical origin of IPP—the infinitive originating in a ge-less strong participle homophonous with and therefore reanalysed as the infinitive, or being a proper infinitive from the start—the morphology of verbs that IPP first occurs with as well as the syntactic restriction to three-verb clusters point towards the latter hypothesis. Recent analyses which view IPP as one possible repair strategy in view of a ban on infinitival complements to past participles are supported by the historical data insofar as a number of competing constructions including PPI and combinations of IPP and PPI also occur, some of which are still found in Modern German dialects. The general preference for a reordering of the verbs within the right periphery of the clause as V1-V3-V2 in IPP constructions, both in Modern German as well as in most dialects, has only evolved more recently; the preferred word order in IPP constructions in historical German was V1-V2-V3.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
17 The ACI construction in the history of German AUGUSTIN SPEYER
. Introduction The right periphery of a German clause is formed by all verbal material that is not moved to C. In verb-second and verb-first clauses, the right periphery consists of all verbal material save for the finite verb which has moved to the left periphery, whereas in verb-final clauses all verbal material is situated in the right periphery (or, in the terminology of the topological field model, right sentence bracket, see also Chapter for a general overview). More traditionally, this clustering of verb forms at the right edge of the clause is called ‘verbal complex’. This verbal complex may consist of periphrastic verb forms, but also of verbs that function as complements to other verbs, as long as they construe obligatorily coherently.1 Examples are modal verbs like sollen ‘shall’, wollen ‘to want’, or raising verbs such as scheinen ‘to seem’. Within verbal complexes formed that way, reordering is possible. Another class of verbs that can enter verbal complexes are the so-called ‘ACI verbs’, in Modern German mainly perception verbs such as sehen ‘to see’ and hören ‘to hear.’ As the ACI construction poses some interesting questions that also go beyond the issue of the right periphery, we decided to devote a special chapter to the history of this construction.
‘Coherence’ in the discussion of German infinitival syntax is a property of infinitives to form a verbal complex with their matrix predicate. The complements and adjuncts of the infinitival clause are not strictly separated from the ones of the matrix clause, but appear in the middle field and can scramble there (Bech ). In this chapter, the term coherence is used together with the term ‘restructuring’, which is more familiar to the grammatical tradition outside Germany. The opposite term is ‘incoherence’, which denotes the property of an infinitival clause not to be integrated in the matrix clause. An incoherent infinitive cannot form a verbal complex with the matrix predicate, and it appears together with its complements in the post field (definition following Bech ). Whether an infinitive clause is coherent or not depends on the matrix verb. Some matrix verbs allow only for coherent infinitives (‘obligatorily coherent’), some allow only for incoherent infinitives (‘obligatorily incoherent’), and a third group allow sometimes for coherent, sometimes for incoherent infinitives (‘optionally (in-)coherent’). 1
Clause Structure and Word Order in the History of German. First edition. Agnes Jäger, Gisella Ferraresi, and Helmut Weiß (eds). This chapter © Augustin Speyer . First published by Oxford University Press.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
.. The ACI in Latin The term ‘ACI’, short for accusativus cum infinitivo (accusative with an infinitive), comes from traditional Latin grammar and denotes a clausal infinitival construction governed by a verb belonging to certain verb classes such as verbs of perception, verbs of saying, verbs of ‘mental perception’ (such as to assume, to know), and a couple of other verb classes. A typical Latin example is given in (). () pro multitudine autem hominum . . . for quantity.ABL however humans.GEN angustos se fines habere arbitrabantur. narrow.ACC REFL.ACC boundaries.ACC have.INF deem.IMPERF.PL ‘But they thought that they have a too small country, relative to the number of inhabitants’ (Caesar I, II ) The verb arbitrari ‘to think that . . . ’ governs a proposition as direct object. In English, this proposition can be realized by means of an infinitival construction or another non-clausal construction (see ()), but is often also realized as a subordinate clause introduced by the complementizer that. In Latin, the proposition is represented by an infinitival construction, the subject of which stands in the accusative (hence the name). The logical subject in () is the reflexive pronoun se, which indicates that, in this example, the subject of the ACI is coreferent with the subject of the matrix clause.2 The use of a reflexive form, which is commonly taken to be subject to Principle A of the binding theory, suggests that the infinitival clause in this configuration cannot be a CP, as a CP would function as a barrier for reflexivization. In a generative framework, it is usually analysed as an IP/TP.3 Note that IP does not function as a barrier for government (see e.g. Grewendorf : ). The status as IP is, however, doubtful, if we assume that subject-verb-agreement is a hallmark of the IP-architecture. The absence of subject-verb-agreement is characteristic for ACI
2 A similar phenomenon is visible for instance with small clauses in English: Hei thought *himi/ √himselfi too unexperienced for the job. 3 It is somewhat questionable whether this view is warranted by clear IP-like properties. If we look at the history of the generative approach, the IP analysis might simply be an echo of the early transformational analysis of an ACI as the outcome of S’-deletion of an underlying that-clause (cf. e.g. Evers ; Rizzi ). If S’ (=CP) is deleted, it follows that what is left must be S (= IP) (cf. e.g. Grewendorf : ff; for an overview on deletion analyses see Sabel : ff; Wurmbrand : ff). This view does not take into account the possibility of VP-internal generation of subjects, so the presence of S/IP is actually only due to the necessity of having some subject position available. Another argument that has been put forward in favour of an IP-analysis was the fact that scrambling is possible within an ACI (Felser ). If scrambling is viewed as movement to Agr-positions (e.g. Schmidt ), the fact that scrambling is possible indicates that there is some IP-structure involving Agr-structures. This argument crucially hinges on the analysis of scrambling as A-movement to Agr-positions. If one subscribes to other analyses of scrambling (A’-movement: e.g. von Stechow and Sternefeld ; Müller and Sternefeld ; or base generation: e.g. Fanselow ), the argument is void.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
constructions, and there are not many good arguments for an IP-status. So let us assume that the infinitival construction is a bare VP.4 Clearly, the Latin ACI is related to Exceptional Case Marking (ECM) constructions in other languages such as English. ECM constructions in English come in two varieties, ECM constructions with bare infinitive, and ECM constructions with toinfinitive (cf. e.g. McCawley : ff).5 The first type is confined to two classes of verbs, namely perception verbs (a) and verbs of causation (b). Note that English control constructions (c) also often correspond to ACI constructions in Latin (d). (c) is an English translation of (d). () a. Betty saw them march towards the station. b. Peter finally made him build the shed. c. If someone believes himself to be something, although he is nothing, he is fooling himself. d. Nam si quis existimat se aliquid esse for if someone.NOM believes himself.ACC something.ACC be.INF cum sit nihil ipse se although be.SBJV.PRES.SG nothing.NOM the.same.NOM himself.ACC seducit. fool.PRES.SG ‘For if someone thinks he is something, although he is nothing, he deceives himself.’ (Vulgata, Epist. ad Galatas , ) The fact that Latin ECM constructions can occur with a greater number of matrix verbs than in English indicates that the analysis might perhaps be a bit different. English and Latin share the exceptional case marking property. The easiest way to think of this is to assume that the logical subject of the ACI proposition is not assigned nominative case (for example, because it has not been raised to IP, if we assume that nominative case is assigned only in SpecIP, or because it is not part of the domain of the infinitival clause. More on these options below). It remains in SpecVP. If we assume that the nominative case cannot be assigned within VP, the 4 Infinitival complements of perception verbs (which is more or less what the Latin ACI is related to, although it functions often as a complement to other verbs too, as outlined in the text) are analysed as CPs e.g. by Rouveret and Vergnaud (); Kayne (). The standard generative IP-analysis is present e.g. in Borer (); Chomsky (); Grewendorf (). Felser () proposes a ‘weakened’ IP-analysis (in her view, infinitival complements correspond to AspP, the lowest projection of the IP-domain). An analysis as bare VP has been suggested by Guéron and Hoekstra () and Bayer (). It was adapted in modified form by Wurmbrand (), see also Enzinger (: ). Formal accounts of the Latin ACI have been brought forward e.g. by Rouveret and Vergnaud (); von Stechow and Sternefeld (); Goldbach (); and Jøhndal (). Goldbach (); focuses on the question of finiteness; since the Latin infinitive lacks mood features but not tense features, there is some functional overhead; the crucial feature to induce subject-verb-agreement is—in her view—Mood, not Tense, therefore it is not applied in the ACI. The relation between ACI and control constructions in Latin has been addressed, e.g. by Cecchetto and Oniga () and Szilágyi (). 5 Leaving aside other similar constructions such as small clauses (They thought their country too small for this great number of inhabitants) or ECM constructions with gerund (I saw him climbing up the wall).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
subject remains caseless. The only possibility for the logical subject to receive case is to receive it from the matrix verb which governs the logical subject and thus can assign case.6 As the logical subject is in the complement position of the matrix verb, it receives accusative case. There is some vagueness in this outline. It is not completely clear in what position the logical subject really is. Is it the ‘real’ direct object of the matrix verb, the remainder of the infinitival clause being another constituent, so that the ACI really consists of two constituents as in ()?7 Or is the whole infinitival clause one constituent, containing the logical subject as in ()?8 A ‘synthesis’ would be an analysis in which the logical subject is the direct object of the matrix verb, and the proposition expressed by the infinitival clause a secondary predicate to the object as in ().9 The ACI would then be something like a small clause (on the analyses and discussion see Speyer a). ()
IP NP pro2 NP t2
I’ I0 arbitrabantur1
VP V’
V’
NP sei
V0 t1
VinfP NP ei
Vinf’ NP ang. fines
Vinf 0 habere
Another analysis has been proposed by Kayne (: ); cf. Felser () for a critical assessment. Crucially, the logical subject is a constituent of the matrix clause independently of the presence or absence of the infinitival clause. This is of course possible only if the valency of the matrix verb would allow for the bare logical subject as only argument. This is for example the case with to see: the valency allows both for something clausal as complement (I see the speaker flee from the daïs) and for a bare NP (I see the speaker). The technical details as to what such a structure really looks like are left open here. The structure represented under () is adapted from Eisenberg (: ), adapted by Speyer (), see also Coleman (: –). It bears resemblance to e.g. Evers (: , ex. ()), or the analysis in Rosenbaum (). What the relationship between the matrix object and the subject position of the ACI is, is left open here. Rosenbaum (); Postal (); and Lasnik () argue for a subject-to-object-raising analysis. A similar analysis in a split-IP-framework was suggested by Grewendorf (). See also the overview in Jøhndal (: –). 8 Similarly e.g. Rouveret and Vergnaud (); Coleman (: ); Evers (); Jøhndal (). 9 See e.g. Hettrich (, ). 6 7
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
()
IP I’
NP pro2
I0 arbitrabantur1
VP V’
NP t2
VinfP
V0 t1
NP se
Vinf ’ Vinf 0 habere
NP ang. fines ()
IP NP pro2 NP t2
I’ I0 arbitrabantur1
VP V’
V0 t1
SC VinfP
NP sei NP ei
Vinf ’ NP ang. fines
Vinf 0 habere
Let us refer to the analysis represented by tree () as biconstituental analysis, to the analysis represented by tree () as uniconstituental analysis, (the ‘constituent’ which gives the name to the terms being bold-faced in the examples () and ()) and to the analysis represented by tree () as small clause analysis (cf. Speyer a). In an earlier paper on this subject (Speyer ), I dubbed the biconstituental analysis ‘video-analysis’ and the uniconstituental analysis ‘dico-analysis’. The analysis represented by the tree in () was not taken into consideration in Speyer (). The reason for the nomenclature in Speyer () is that perception verbs like Latin video ‘I see’ can govern a nominal accusative complement, no matter whether an additional clausal constituent is present or not, whereas other Latin ACI verbs such as dico ‘I say’ cannot govern a nominal accusative complement representing a participant in an event, as will be demonstrated in the next paragraph. So verbs like dico can only receive the analysis under (), whereas for verbs like video, the analysis under () is at least possible. The analysis under () would also only work for verbs like video. For matters of consistency with Speyer () I will sometimes use the terms ‘dico/video’analysis also here.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
In Latin, it is very difficult to find direct evidence for either analysis. It is possible to find indirect evidence, however. It is well known that not all Latin verbs governing an ACI can also govern an accusative object, thus the biconstituental analysis is ruled out for such verbs (Rouveret and Vergnaud ). This can be demonstrated with verbs of mental perception: the Helvetians in () do not think their land, and then something about their land, but they only think a fully fledged proposition which has something to do with their land. This is evident from the fact that a verb like arbitrari in the sense of ‘think’ cannot govern a pure nominal object (*fines arbitrabantur), but verbs of perception like video ‘I see’ can do so (fines videbantur). So this suggests that at least the Latin ACIs governed by verbs like arbitrari construct following the uniconstituental dico-analysis, and neither the biconstituental video-analysis nor the small clause analysis. .. The ACI in Modern German In Modern German, stronger restrictions for ACI constructions hold than in English: the ACI in Modern German is only grammatical if it is governed by verbs of perception (a) or lassen ‘let/make someone do something’, originally a verb of causation (b).10 Verbs of mental perception or other verb classes that are grammatical as governors of ACI constructions in Latin are not so in German (c). () a. Bettina sah den Protestzug zum Bahnhof ziehen. Bettina.NOM saw [the protest.parade]ACC to.the station.DAT march ‘Bettina saw the protesters march to the station.’ b. Peter ließ ihn den Schuppen bauen. Peter.NOM let him.ACC [the shed]ACC build ‘Peter allowed him to build the shed/P. made him build the shed.’ c. *Die Helvetier meinten das Land für sie zu klein sein. [the Helvetii]NOM thought [the land]ACC for them too small be ‘The Helvetii thought that the land was too small for them.’ Perception verbs can construct rather freely with ACI (the sentences under () are spontaneous utterances which I overheard; they are not fabricated examples), although the most frequent matrix verbs are sehen ‘to see’ and hören ‘to hear’. () a. Es sieht echt so aus, als ob der Hund ihr Auto kommen riecht. it looks really so PTC as if [the dog]NOM [her car]ACC come smells ‘It really seems as if the dog smells her car arrive.’
10
In Bavarian, the verb let shows an exceptional behaviour in that it marks the logical subject with dative instead of accusative (cf. (i)), see Haider (: ). Similar constructions are attested for NHG in the seventeenth and eighteenth centuries (cf. Demske-Neumann : ). (i) Loß ia=n do let her.DAT=him.ACC yet ‘Let her marry him.’
heiradn [ . . . ] marry
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer b. Ich spür den Faden durchgehen. I.NOM feel [the thread]ACC through.go ‘I feel that the thread goes through [the needle’s eye].’
The only condition for using an ACI seems to be direct perception (see Clément ). The sentence (a) is acceptable under the presupposition that the roofer is at the moment of speech engaged in climbing to the roof and that the speaker really hears sounds connected with this process, like the scraping of the roofer’s boots or some swearing from above. It would not be acceptable if the speaker did not hear the roofer himself, but only heard from someone that the roofer climbed onto the roof. This is the reason why (b) is unacceptable under all circumstances: As both actants, Ulrich von Hutten and Duke Ulrich, lived in the sixteenth century, a speaker in the twenty-first century simply cannot perceive them directly and therefore the use of an ACI is impossible.11 () a. Ich höre den Dachdecker aufs Dach steigen. I.NOM hear [the roofer]ACC on.the roof.ACC climb ‘I hear the roofer climb onto the roof.’ b. *Ich höre Ulrich von Hutten sich über Herzog Ulrich I.NOM hear [U. v. Hutten]ACC himself.ACC about [Duke U. ]ACC beschwert haben. complained have ‘I hear Ulrich von Hutten to have complained about Duke Ulrich.’ The proposition expressed by the ACI in (a) can also be represented by a finite subordinate clause introduced by wie as illustrated in (a). For the use of this clause the same condition of direct perception holds: (b), the wie-version of (b), is not as unacceptable as the ACI-version, but it sounds deviant. A subordinate clause introduced by dass is possible with or without direct perception (c) and (d). A nonembedded V argument clause (term following Reis ) is only possible if direct perception is excluded. So, (e) is felicitous in the scenario in which the speaker only hears a report that the roofer climbed onto the roof, but not in the scenario in which the speaker hears the roofer himself scratch over the tiles etc. Likewise, (f) in which direct perception is impossible is felicitous with a non-embedded V argument clause. () a. Ich höre, wie der Dachdecker aufs Dach steigt. I.NOM hear how [the roofer]NOM on.the roof.ACC climbs ‘I hear the roofer climbing onto the roof.’ b. ?Ich höre, wie Ulrich von Hutten sich über I.NOM hear how [U. v. Hutten]ACC himself.ACC about Herzog Ulrich beschwert hat. [Duke U.]ACC complained has ‘I hear Ulrich von Hutten to have complained about Duke Ulrich’ 11 As a result of this constraint, perfect forms are not acceptable in the ACI in principle, even if one could construct a scenario in which the event expressed by a perfect infinitive is directly perceived.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
c. Ich höre, dass der Dachdecker aufs Dach steigt. I.NOM hear that [the roofer]NOM on.the roof.ACC climbs ‘I hear that the roofer climbs onto the roof.’ d. ?Ich höre, dass Ulrich von Hutten sich über I.NOM hear that [U. v. Hutten]ACC himself.ACC about Herzog Ulrich beschwert hat. [Duke U.]ACC complained has ‘I hear that Ulrich von Hutten has complained about Duke Ulrich’ e. Ich höre, der Dachdecker sei aufs Dach gestiegen. I.NOM hear [the roofer]NOM be.SBJV.PRES.SG on.the roof climbed ‘I hear that the roofer has climbed onto the roof.’ f. Ich höre, Ulrich von Hutten habe sich I.NOM hear [U. v. Hutten]NOM have.SBJV.PRES.SG himself.ACC über Herzog Ulrich beschwert. about [Duke U.]ACC complained ‘I hear that Ulrich von Hutten has complained about Duke Ulrich.’ The direct perception quality of the German ACI suggests the biconstituental analysis in some ways: the subject perceives somebody or something, and the action in which the object of perception is engaged is a second constituent. So this is a clear difference from the Latin ACI. Another distributional datum that suggests the biconstituental analysis for Modern German is the fact that only verbs that can govern also a nominal accusative object allow for ACIs. If Modern German ACIs had a uniconstituental structure, we should expect other verbs to realize clausal complements as an ACI as well, at least optionally, as they do in Latin. However, this is not the case in Modern German. Eisenberg (: f) brings forward some more arguments for the biconstituental analysis. Thus, it is possible to move the constituents independently from each other to the prefield (a) and (b). This counts as a strong test for constituency. The fact that the ACI as a whole can be moved to the prefield (c) need not speak against this analysis, as remnant movement of VPs containing more than one phrase is possible in German; in this case the (headless) matrix VP with the logical subject and the predication of the ACI as two constituents (on constituent tests see Speyer a with references therein). () a. Den Dachdecker [the roofer]ACC b. Aufs Dach on.the roof.ACC
höre ich hear I.NOM steigen höre climb hear
aufs Dach steigen. on.the roof.ACC climb ich den Dachdecker. I.NOM [the roofer]ACC
c. Den Dachdecker aufs Dach steigen höre ich. [the roofer]ACC on.the roof.ACC climb hear I.NOM ‘I hear the roofer climb onto the roof.’ This would be hard to explain if one applied the uniconstituental analysis. Following the uniconstituental analysis, (c) would be the normal case, and (a)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
and (b) would be split topicalization, that is movement of a partial constituent to the prefield. Consequently, (a) and (b) should be possible only under the conditions that allow for split topicalization (such as double focus), which is clearly not the case. An argument against the small clause analysis is that the infinitive can form a verbal complex together with the matrix predicate; in other words, it is a coherent/restructuring infinitive construction. Evidence for the formation of a verbal complex comes from the fact that the ACI infinitive form is permutable with parts of the matrix verb (); the possibility to change the order of the verb forms is a characteristic of verbal complexes.12 In a small clause, the infinitive is too deeply embedded. Unambiguous small clauses such as in () do not allow verbal complex formation as freely, as can be seen from the fact that permutation of the verbal forms is marked and requires the fulfilment of special information-structural conditions. For these reasons, I assume the video-analysis, i.e. the biconstituental analysis, for Modern German ACI constructions. () a. Laura weiß, dass Jörg den Dachdecker aufs Dach Laura.NOM knows that Jörg.NOM [the roofer]ACC on.the roof.ACC steigen gehört hat. climb heard has ‘Laura knows that Jörg heard the roofer climb onto the roof.’ b. Laura weiß, dass Jörg den Dachdecker aufs Dach Laura.NOM knows that Jörg.NOM [the roofer]ACC on.the roof.ACC hat steigen hören. has climb hear ‘Laura knows that Jörg heard the roofer climb onto the roof.’ () a. Laura weiß, dass Jörg die Haustür gestreift bemalt hat. Laura.NOM knows that Jörg.NOM [the door]ACC striped painted has b. ?Laura weiß, dass Jörg die Haustür hat gestreift bemalt. Laura.NOM knows that Jörg.NOM [the door]ACC has striped painted ‘Laura knows that Jörg painted the door striped.’ Turning to the diachrony of the ACI construction, the observations above raise several questions. (i) What are the properties of the ACI construction in earlier stages of German? (ii) How should we analyse these earlier forms of the ACI? (iii) How does the ACI construction change during the historical development of German (narrowing down of potential uses etc.)? (iv) What is the historical origin of the German ACI construction—is it a loan from Latin or is it inherited?
12
The IPP (Ersatzinfinitiv, or infinitivus pro participio) occuring in (b) constitutes further evidence for the formation of a verbal complex. For details on the history of the IPP construction see Chapter of this volume and Schallert (b).
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
We will see that all of these questions are closely intertwined. Questions for further research that I can only mention here are whether the coherent/restructuring nature of the ACI has developed from an incoherent construction, or whether the incoherent control construction has developed from a coherent construction or if they have nothing to do with each other, which would be puzzling given the fact that in other languages, such as Latin, there is no clear-cut morphosyntactic distinction between both constructions.
. The ACI in Old High German The historical development of the ACI in German bears resemblance to the development in English and Dutch (see Fischer et al. ). We concentrate on German, however. When we look at the prehistory of the ACI, there is evidence that a similar construction must have existed in PIE (contra Behaghel –, Vol. II: ). Hettrich (, ) cites evidence from Hittite and Vedic Sanskrit suggesting that predicate constructions of accusative objects and verbal nouns occur in these languages; as both languages go back to PIE but belong to different subgroups, it can be assumed that in PIE, this option already existed. So the analysis for PIE should clearly be the small clause analysis illustrated in (). After infinitives developed out of verbal nouns, they were used in such predicative contexts, leading to the ACI stricto sensu. Constructions in which a proposition is expressed by means of a complex expression that contains some non-finite form—and the ACI belongs to that group— are most probably an early method of indicating the subordination of one proposition to another, competing with the other method, i.e. using finite clauses with complementizers. Most probably, these constructions started life as event nominalizations, in which case the logical subordination is obvious from the fact that the verbal noun is a prototypical constituent in its matrix clause obtaining a syntactic function in the matrix clause. The method of marking subordination by a verbal noun/non-finite form as such was obviously in use already in PIE, but the details differ in the daughter languages. According to Palmer (: ), the historical origin of the ACI in Latin is a construction governed by verbs of causation. The object of the verb is the patient (the person who is ordered to do something), and the action which this person has to do is expressed by a verbal noun in the dative, indicating the purpose of the matrix action. This purposive usage of verbal nouns is a feature of all daughter languages and partly remains in action even when these verbal nouns have become infinitives. So a sentence like () is originally to be translated as ‘The consul urged the lictores,13 so that they approach him (= some other guy)’ or ‘The consul urged the lictores in order to get them to approach him (= some other guy)’.
13
Lictores in Republican Rome were officers whose duty was to ensure the safety of the high Roman magistrates and to act as an executive force for them. So their duties covered more or less the duties of modern policemen.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
() lictores ad eum accedere consul iussit. policemen.ACC to him.ACC towards.go consul.NOM ordered ‘The consul commanded the lictores to approach him (some person other than the consul).’ (Livius XXIV , ) This, of course, is a classical example of the biconstituental video-analysis; here the infinitival clause is clearly an adjunct. From there, the ACI was extended by analogy to verbs of perception. We may assume that with verbs of perception a participle clause was originally connected but was replaced by an infinitival clause by analogy to the causative verbs. Up to this point, this scenario could hold for German as well. In Latin, the ACIs with biconstituental structure (or small clause analysis, see Hettrich , ) were reanalysed as having a uniconstituental structure (Hettrich : ). It is easy to see how this could have happened, especially since the uniconstituental analysis is structurally much simpler than the biconstituental analysis, and the evidence for language learners in favour of the biconstituental analysis is sparse.14 From there, extension to other verb classes such as verbs of mental perception and verbs of saying can easily happen (Coleman ). Turning to OHG, we see, however, that here, this reanalysis obviously has not taken place. Behaghel (–, Vol. II: ) doubts that the ACI was a feature of OHG language usage: ‘Die selbständigen ahd. Denkmäler . . . zeigen dann, daß die Konstruktion in der echt heimischen Sprache völlig untergegangen ist.’ [‘The original OHG documents reveal that the construction had disappeared completely from the indigenous language.’]. He sees the ACI as an exclusive phenomenon of translations, that is, as an ephemeral example of loan syntax. This view has to be qualified, however. Today, it is widely assumed that ACIs subcategorized by causative verbs and perception verbs reflect indigenous OHG language usage (cf. e.g. Schrodt : ). In Otfrid’s Gospel Book, which counts as the only extensive original OHG text (cf. Chapter ), we find fifteen ACIs, all of which are governed by a perception verb (). () a. Thu sihis sún liaban / zi mártolonne zíahan you.NOM see.SG son.ACC dear.ACC to torture.DAT march ‘You see your dear son marching towards the torture’ (Otfrid I , ) b. tház ir io hortut quédan mih that.ACC you.NOM once heard say me.ACC ‘You once heard me say that’ (Otfrid II , ) c. ih irkánta [ . . . ] / thia kraft hiar fáran fona mir I.NOM noticed the.ACC power.ACC here go from me.DAT ‘I noticed how the power left me’ (Otfrid III , ) 14 Language change is mostly a product of reanalysis on the part of language learners due to the limited evidence they have been exposed to. See e.g. Clark and Roberts (); Lightfoot ().
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
If we look at other extensive texts, namely the Tatian and Notker’s works, the picture is different. Let us begin with Notker, who translates the Latin text more freely than Tatian. Here we find numerous examples of ACI governed by perception verbs (a) and (b), but also governed by other verb classes such as verbs of mental perception (c), and verbs of saying (d). () a. uuánda man dârínne síhet. sámoso éinen brúnnen spríngen because one.NOM within sees likewise a.ACC fountain.ACC rise ‘because one sees inside something just like a fountain rise’ (Notker, Martianus Capella , –) Latin original: ita ut clausam in eo fontaneam putes scaturriginem so as closed in it from.a.fountain you.think whirlpool ‘just as if you think, there was locked in it a whirlpool from a fountain’ b. Uuío ófto nehórta íh tie góta síh how often NEG.heard I.NOM the.ACC gods.ACC themselves.ACC chlágon complain ‘How often did I hear the gods complain!’ (Notker, Martianus Capella , –) Latin original: Quotiens . s. memini . conquestos deos how.often themselves remember.SG complained.ACC gods.ACC ‘How often do I remember that the gods complained!’ c. Núbe gót uueíz ih flégen sînes uuérches but God.ACC know I.NOM care.for his.GEN work.GEN ‘But I know that God cares for his work.’ (Notker, Consolatio , –) Latin original: Verum scio deum conditorem . presidere opera suo but know.SG God.ACC creator.ACC preside work.DAT his.DAT ‘But I know that God, the creator, presides over his work’ d. únde iího íh iz álso uuésen . sô du chîst and say I.NOM it.ACC so be as thou.NOM sayest ‘and I agree that it is as you say’ (Notker, Consolatio , –) Original: et consentio id ita esse . uti tu dicis and agree.SG it.ACC so be as you.SG.NOM say ‘and I agree that it is as you say’ It is obvious that the usage of the ACI is a loan translation from Latin in these cases (cf. also Coleman : ). In free passages of Notker’s work, i.e. passages which do not directly go back to a Latin original or which translate it
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
more freely, we find almost exclusively ACI governed by perception verbs (a). The ACI governed by perception verbs seems to have been a freely available option in OHG, as sometimes it occurs in translation passages, even though the Latin original uses another construction. In (a), the Latin text uses an accusativus cum participio (ACP), that is, a predicative participle dependent on the accusative object. Yet Notker seems to have internalized the Latinate usage of the ACI to such an extent that we find the odd example of an ACI governed by a non-perception verb even in free passages like the prefaces to one of his translations (b). () a. unde siêho sia mih fuôrin ellenden an dero êo and see her me.ACC lead exiled.ACC at the.DAT law.DAT dero súndo the.GEN sin.GEN ‘and I see her lead me as an exiled person in the law of the sin’ (Notker, Psalter , ) Original: et captivum me ducentem in lege peccati and prisoner.ACC me.ACC leading.ACC in law.ABL sin.GEN ‘and [sc. I see] her lead me as prisoner in the law of the sin’ b. Remigius lêret únsih tísen auctorem in álenámen Remigius.NOM teaches us.ACC this.ACC author.ACC in all.name uuésen gehéizenen martianum. be called.ACC Martianus.ACC ‘Remigius teaches us that this author’s real name was Martianus.’ (Notker, Martianus Capella , –) The usage of ACI in the Tatian was investigated at length by Dentschewa (). I will summarize her findings. Usually, the translator follows the usage of ACI in the Latin original, but sometimes he replaces a Latin ACI that is governed by a nonperception verb with another construction. In (), there are two examples in which a Latin ACI governed by dicere ‘to say’ is replaced by a thaz-clause in OHG. () a. thie thar quadun thaz niuuâri urresti who.NOM there said that NEG.was resurrection ‘who said that there was no resurrection’ (Tatian , –) Original: qui dicunt non esse resurrectionem who say.PL not be resurrection.ACC.SG ‘who say that there is no resurrection’ b. quadun thaz thonar gitan uuari say.PL.PRET that thunder.NOM done was ‘they said that thunder was made’ (Tatian , –)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
Original: Dicebant tonitruum factum esse say.PL.IMPERF thunder.ACC done be ‘they said that thunder came into being/was made’ An ACI governed by perception verbs is sometimes translated as an ACP in OHG (a) and (b). But since the Latin infinitives are passive infinitives in these cases, it cannot be excluded that the apparent participles are in reality German passive infinitives (that consist of a participle form and the infinitive of the auxiliary verb wesan or sîn)15 in which the infinitive of the auxiliary is elided. Similar examples can be found in Notker, too: (c) is an example in which the infinitive of the copula is elided as well. () a. zihiu giengut ír úz in uuvostunna sehan to.what.DAT went you.NOM out in desert.ACC see rora fon uuinte giuuegita reed.ACC by wind.DAT moved.ACC ‘What did you expect to see when you went out into the desert? A reed, shaken by the wind?’ (Tatian , –) Original: quid existis in deserto uidere harundinem what.ACC go.out.PL.PERF in desert.ABL see reed.ACC vento moueri wind.ABL be.moved ‘What did you expect to see when you went out into the desert? A reed, shaken by the wind?’ b. mit diu ir gisehet umbigeban fon here with which.DAT you.NOM see surrounded by army.DAT hierusalem Jerusalem.ACC ‘as soon as you see Jerusalem be surrounded by troops’ (Tatian , –) Original: cum autem uideritis circumdari ab exercitu hierusalem when however you.will.see be.surrounded by army.ABL Jerusalem.ACC ‘as soon as you see Jerusalem be surrounded by troops’ c. Uuánda íh tíh óuh sô geréchen sího zefernémenne because I.NOM you.ACC also so ready see to.understand ‘because I also see that you are so ready to understand’ (Notker, Consolatio , )
15 Note that periphrastic passives were already in use in OHG. See Oubouzar () and Chapter of this volume.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer Original: Sed quoniam conspitio te promptissimum esse but because notice.SG you.ACC ready.SUPERL be ad intelligendum at understand.ACC ‘because I also notice that you are very ready to understand’
Summing up, we can say that in indigenous OHG, the usage of the ACI construction seems to correspond to the modern state-of-affairs: the ACI occurs only if governed by perception verbs or verbs of causation. In OHG translationary texts, ACIs can be found that are governed by verbs of other verb classes. In these cases, it is most probably a loan translation from the Latin original. So one might have the impression that no change had occurred at all.
. The ACI in Middle High German There are signs, however, that a change did occur, and in fact was starting in the MHG period. If we look at the MHG evidence (cf. Paul : ; Behaghel –, Vol. II: again doubts the existence of indigenous ACIs), most examples are governed by perception verbs as in (). However, we occasionally find ACIs (and the ‘passive’ variant, the nominativus cum infinitivo) governed by non-perception verbs in texts, cf. (). Note that these texts are original texts (mostly poetry, but for a question such as the present one, we should not expect there to be a difference between poetry and prose). As these are original texts, interference from Latin source texts can be excluded.16 () a. nu gesach ich dich nie mere weinen sô vaste! now saw I.NOM thee.ACC never more weep so strongly ‘I never saw you weep so heavily.’ (Die drei Mönche zu Kolmar –) b. do horent si vil bald lüt an dem there hear they.NOM very soon people.ACC on the.DAT wege gan way.DAT go ‘Then they heard very soon people walking on the street’ (Die Sünderin –)
The data were collected by first examining one entire text (Wigamur) and subsequently searching the Mittelhochdeutsche Begriffsdatenbank subcorpus ‘Kleinere Erzählungen’ (smaller narrative texts) for the sample verbs sehen ‘see’, hören ‘hear’ (perception verbs), meinen ‘think’, denken ‘think’, waenen ‘think’, wizen ‘know’, finden ‘find, regard’ (verbs of mental perception), jehen ‘say’, sagen ‘say’ (verbs of saying), biten ‘beg, ask’, twingen ‘force’, überreden ‘persuade’, überziugen ‘talk someone into doing something’, lêren ‘teach’, bereden ‘persuade’ (other potential ACI verb classes that correspond to object control verbs in Modern German), and hoffen ‘hope’ (another ACI verb that corresponds to a German subject control verb). 16
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
c. daz gehort ich e gesagen nie that.ACC heard I.NOM before say never ‘I never heard this being said before’ (Die beiden Knechte ) () a. das daucht si baide guot sein that.NOM thought them.ACC both.ACC good be ‘Both thought that this was good’ (Die Rache des Ehemannes ) b. die wonte ich mein muter sein her.ACC thought I.NOM my.ACC mother.ACC be ‘I thought she was my mother’ (Wigamur ) c. den vand ich ligen bi dem weib this.ACC found I.NOM lie at the.DAT woman.DAT ‘I found this guy lying with the woman’ (Die Suche nach dem glücklichen Ehepaar ) We see that an extension has taken place from the verbs of direct perception as the only verb class (abstracting away from causative verbs) that subcategorizes an ACI in OHG to verbs of mental perception.17 One has to bear in mind, however, that examples of ACIs governed by a verb of mental perception are less frequent than ACIs governed by verbs of direct perception: Of tokens of sehen ‘to see’ in the subcorpus, fifty-five governed an ACI (.%), whereas of sixty-one tokens of denken ‘to think’, only one (.%) governed an ACI. The fact that this extension happens is natural from a semantic point of view (see Coleman ). It is, however, dramatic from a syntactic point of view as ACIs with verbs of mental perception cannot receive the biconstituental analysis any more as these verbs do not subcategorize a single argument that refers to an entity (cf. also Paul : ). This is excluded for reasons of the semantics of these verbs. So a reanalysis of the biconstituental structure to the uniconstituental analysis (dico-analysis) seems to have taken place in MHG as it did a couple of centuries before in the prehistory of Latin. Let us investigate whether it is possible to trace the way the reanalysis worked. One important observation is that we regularly find sentences which can be analysed as ACIs with ellipsis of the copula ‘to be’ (). () a. do si in in noten sahen when they.NOM him.ACC in need saw ‘when they saw him in need’ (Der ernsthafte König )
17 In that light, it might be the case that OHG examples of ACI governed by verbs of mental perception might be genuine as well and not instances of calque. However, as the extension in MHG seems to go only as far as mental perception, whereas in OHG semantically less related verb classes can govern an ACI, at least these examples are probably loan syntactic phenomena.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer b. die daucht sich fürbaß immer mere this.NOM thought herself.ACC afterwards always more beide hoffertig und here. both proud and glorious ‘Ever after, she thought herself to be proud and glorious.’ (Das Nonnenturnier –) c. wan du mich not durftich wæist der gnaden because you.NOM me.ACC wanting know the.GEN grace.GEN ‘Because you know that I am in need of grace’ (Vaterunser –) d. ir namen ich geschriben vant her.ACC name.ACC I.NOM written found ‘I found her name written’ (Die Klage der Kunst , )
Note that the analysis as ACI with elided copula (a)18 is by no means the only possible analysis of the sentences in (). All instances could equally well be analysed as small clauses, that is, as object NP with a secondary predicate adjoined to the NP (b) and (c). These sentences are thus ambiguous with respect to the underlying structure. From this point of departure, a reanalysis producing a blend of both constructions seems possible in which the secondary predicate is interpreted as a verbal projection in which the verb ‘to be’ is covert (c), especially as the absence of the infinitive of the copula in ACIs seems quite frequent at this stage of the language.19 () a.
IP NP si2
I’ I0 sahen1
VP NP t2
V’ VinfP NP in
V0 t1 Vinf ’
PP in noten
18
Vinf0
I have put angled brackets for material that is elided at spell-out although it is underlyingly present. A reviewer pointed out that one should expect it to happen the other way round, i.e. starting from a small clause and then reanalysing it as an ACI without copula. It is hard to decide in the present case, especially since small clauses of this kind seem not to have been productive in OHG and MHG, but surely this would be a plausible scenario for the origin of the ACI in general out of a small clause construction. This reanalysis would pre-date the OHG times. 19
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
b.
IP I’
NP si2
VP V’
NP t2
I0 sahen1 V0 t1
NP NP in c.
PP in noten
IP NP si2
I’ I0 sahen1
VP NP t2
V’ V0 t1
NP
NP ini
VinfP NP ei
Vinf’ PP in noten
Vinf 0
Predicative constructions (i.e. small clauses) belong to the regular repertoire of MHG, also in other contexts (a). So it is feasible that a language learner having a small clause analysis of sentences like () internalized reanalyses verbless ACIs as small clauses as well on the model of unambiguous small clauses as in (a). An important role as bridging contexts is taken up by verbs like vinden ‘to find’ that are frequent both with ACI and in contexts in which the analysis as small clause is the one that suggests itself (b) and (c). () a. daz man in nante einen zagen that one.NOM him.ACC called a.ACC coward.ACC ‘that one called him a coward’ (Eschenbach, Alexander ) b. want si die uzzern want because she.NOM the.ACC outer.ACC wall.ACC so bitter und so ubel vant so bitter and so bad found ‘because she found the shell [of the nut] so bitter and bad-tasting’ (Die Äffin und die Nuss –)
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer c. so vindet er offen die tur zu vrone so finds he.NOM open the.ACC door.ACC to Lord.GEN himelriche. heavenly.kingdom.DAT ‘So he finds the door to the Lord’s heavenly kingdom open’ (Von drei Freunden –)
If this putative language learner adopted analysis (c) for the cases in (), she is quite close to a uniconstituental analysis, in that the infinitival proposition is no independent constituent stricto sensu any more. A further reanalysis towards the uniconstituental analysis (dico-analysis) in fact would be a simplification of the structure and thus would not be unexpected. It is not clear, however, whether the ACIs governed by verbs of mental perception are ‘real’ ACIs or control constructions. In Modern German, ‘real’ ACIs and control constructions can easily be distinguished by the fact that control constructions are optionally incoherent and take the zu-infinitive, whereas real ACIs are coherent/ restructuring and take the bare infinitive. In MHG, control verbs govern both bare infinitives (a) and (c) and zu-infinitives (b) and (d). Point () offers examples of object control (a) and (b), and subject control (c) and (d). Especially cases of object control are morphologically not distinguishable from ACIs. Coherence is not easy to pinpoint in these texts, although there is one example in which the infinitival clause is probably incoherent as it is set apart from the matrix clause by punctuation (e); note that a bare infinitive is used, in contrast to Modern German. So it is actually not even clear whether the examples for verbs of mental perception like (b) are not control constructions in disguise, especially as verbs of mental perception also govern a zu-infinitive occasionally (a). There are lexical differences: meinen usually governs a zu-infinitive (a), whereas, in my sample, waenen always governs a bare infinitive, even in cases of clear subject control (b). Note that verbs of mental perception can construct as subject control verbs in Modern German ((c) and (d) being modern translations of (a) and (b); cf. also Speyer a). () a. si patt in haimlich mit ir gan she.NOM asked him.ACC secretly with her.DAT go ‘she asked him to go with her secretly’ (Die unschuldige Mörderin ) b. der mentsch sol den priester nit bitten ze singenn the.NOM human.NOM shall the.ACC priest.ACC not ask to sing ‘A man should not ask the priest to sing’ (Predigten und Tischgebete –) c. Und hoffet sîns dienstes warden frô and hopes his.GEN service.GEN become happy ‘and he hopes to enjoy his service’ (Hugo von Trimberg )
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
d. so hoffen ich uwer sůne zu machen so hope I.NOM your.ACC retribution.ACC to make ‘Therefore I hope to counterstrike on your behalf ’ (Prosalancelot II , ) e. unt bittet, im mere buze geben and asks him.DAT more.ACC penitence.ACC give ‘and asks him to give him more penitence’ (Vom heiligen Geist ) () a. und meint den pfaffen abzuschrecken and thinks the.ACC priest.ACC away.to.frighten ‘and he thinks he could frighten away the priest’ (Von einem varnden Schuler ) b. si wânten vinden grôzen schaz they.NOM thought find big.ACC treasure.ACC ‘They thought they would find a big treasure.’ (Der Schlegel ) c. Und er meinte, den Pfarrer abzuschrecken. and he thought [the priest]ACC away.to.frighten ‘and he thought he could frighten away the priest’ d. Sie meinten, einen großen Schatz zu finden. they.NOM thought [a big treasure]ACC to find ‘They thought they would find a big treasure.’ The differentiation of ACI and control construction and the development of the modern infinitival-state-system as described by Bech () is an interesting area of research in itself, but it would go beyond the scope of this chapter to delve into it. Summarizing, we can say for MHG that the OHG state-of-affairs—ACIs being almost exclusively governed by perception verbs (leaving out causative verbs) and thus presumably to be analysed as biconstituental—persists, although an extension might be taking place towards other verb classes that are semantically close to verbs of perception but that cannot govern a true ACI. If the latter cases are not instances of control constructions, it is fair to say that the development towards a uniconstituental analysis is initiated.
. The ACI in Early New High German In ENHG, at least in certain texts, it looks as if the development towards a uniconstituental analysis comes to completion. If we look at texts from the ENHG period, we find numerous examples of ACIs with non-perception verbs. In () there is a small collection of examples: (a) governed by a verb of mental perception, (b) governed by a verb of saying, (c) governed by a verb of sensation, (d) governed by a verb of expectation.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
geloub dasselb lande vnderwürffig sin () a. Ich be I believe [the.same country]ACC subject dem wind borree [the wind Boreas]DAT ‘I think that this land is subject to Boreas (= north wind)’ (Wyle, Translationen , –) b. Aber verender dise din vrtail als du sprichst but change [these your judgements]ACC when you.NOM say min arbait gegen dir vnnütz sin. [my work]ACC against you.DAT useless be ‘But revise your judgement when you say that my work for you was useless.’ (Wyle, Translationen , –) c. darvmb lucrecia laid trůg, ain söliche verlümdete therefore Lucretia.NOM grief.ACC bore [a such of.ill.repute frovwen zů ir gesant sin. woman]ACC to her.DAT sent be ‘Therefore Lucretia was aggrieved that a woman of such ill repute was sent to her.’ (Wyle, Translationen , –) d. vnd forcht betrugnusz dar vnder verborgen ligen and feared deception.ACC there under covered lie ‘and she feared that hidden deception was going on’ (Wyle, Translationen , ) This, however, does not constitute solid evidence that the ACI was really an option in indigenous ENHG in these configurations. This is a well-known fact (cf. e.g. Ebert et al. : ff), and has been characterized as Latin influence in the literature (cf. e.g. Ebert et al. : ; Behaghel –, Vol. II: f). All examples in () are taken from a text by Niklas von Wyle, a humanistic author, and the text itself is a collection of translations of Latin and Old Italian texts, mostly originally written by Boccaccio. In an earlier study (Speyer ), I compared the usage of ACI in this text with a text that was written by a humanistic author as well, but intended for a broader audience, a Gespräch-Büchlin (= humorous dialogue with educational aims) by Ulrich von Hutten. Both texts belong to the same dialect area (Swabian). In Hutten’s text, we find also examples of ACIs governed by non-perception verbs, but less frequently than in Niklas von Wyle. A comparison taking the most common verb classes into account is given in Table .. In (), there are some examples of ACI clauses from Ulrich von Hutten. The Latin influence is obvious in examples like (c) in which the logical subject of the ACI which is coreferent with the matrix subject is rendered by a reflexive pronoun. In Modern German, verbs like glauben ‘believe, think’ can govern a control infinitive in which, characteristically, the logical subject is silent. Example (d) is perhaps not an example of an ACI at all; it could also be a kind of subordinate clause; the subjunctive rd person plural of the copula is sein, as is the infinitive, in Ulrich von Hutten’s writing.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
T . Verb types governing ACI in ENHG Niklas von Wyle verbs of perception verbs of mental perception verbs of saying other verb classes total
Ulrich von Hutten (?) –
Feber hat uns sehen Hühner essen () a. dann das for the.NOM fever.NOM has us.ACC seen chicken.ACC eat ‘for the Fever saw us eat chicken’ (Hutten, Gespräch-Büchlin ) b. Dann itzo dunkt mich . . . sie alle Ding redden und tun then now thinks me.ACC them.ACC [all things]ACC speak and do um der Weiber willen. for [the women]GEN will ‘But now I think that they speak and do everything only because of the women.’ (Hutten, Gespräch-Büchlin ) c. daß er geglaubt, sich von ihr geliebt werden that he.NOM believes himself.ACC by her.DAT loved become ‘that he believes to be loved by her’ (Hutten, Gespräch-Büchlin ) d. sprach: ‘Die Pfaffen ein untreu, argwehnig spoke [the priests]NOM/ACC [a unreliable distrustful Volk sein’ people]NOM/ACC be ‘(sc. he) spoke that the priests are an unreliable, distrustful lot.’ (Hutten, Gespräch-Büchlin ) A brief search in the Bonner Frühneuhochdeutschkorpus showed the same picture: I searched for some ‘Latinate’ ACI verbs, namely glauben ‘to believe’, meinen ‘to think’, wissen ‘to know’. In texts by authors with a probable humanistic or ecclesiastical background (such as Hans Neidhart, the translator of Terence’s Eunuchus in Ulm) we find ACIs governed by verbs such as glauben ‘believe, think’ (a), whereas less influenced authors (such as Sigmund von Herberstein) realize the dependent proposition as a that-clause (b).20
20 The punctuation in (a) indicates that mutual influence of control constructions and ACIs might have been an issue also in ENHG.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
Augustin Speyer
glauben moͤchte, das lauter geredt syn () a. das ich that I.NOM believe would that.ACC a.lot.of gossip.ACC be ‘I believe that this is nothing but gossip’ (Neidhart, Eunuchus , ) b. Derhalben bewegt mich zuglauben / das die Reissen . . . therefore moves me.ACC to.believe that [the Russians]NOM froͤmbder sprach . . . zu Herrschafften genomben / . . . haben. [foreign language]ACC to reign taken have ‘Therefore I am, inclined to believe that the Russians adopted a foreign language when they came into power.’ (Herberstein, Moscouia B, ) We see that the Latinate usage of the ACI was quite widespread among the humanistically educated ‘literati’ of the time around . It is obvious that this usage goes back to the Latin model (cf. e.g. Behaghel –, Vol. II: ; Ebert : ). And it is likewise obvious that the Latinate usage went out of use after humanism ceased to play an important role for the development of a German prose style as the ACI governed by non-perception verbs is ungrammatical in Modern German. Speaking in terms of the putative reanalysis that took place in MHG, it is clear that the reanalysed uniconstituental structure could have spread (and probably the reason that the humanistically educated authors used it at all goes back to the fact that this structure was available in principle), but obviously it did not spread, at least not in vernacular usage. We have to ask at this point why the Latinate uniconstituental ACI governed by verbs of mental perception etc. did not manage to get integrated into the German syntactic system. The answer to this question is less obvious than one might think because there are several Latin-influenced changes observable just in the ENHG era. I mention only one example, namely the higher role of information structure in the serialization of arguments (see Chapter ). A possible explanation is that the two examples of Latin influence do not involve a real change in syntactic structure, but only a change in frequency: in both cases, the ‘Latinate’ syntactic form was an option in the language, yet a dispreferred one. In the case of the ACI, the adoption of the Latin usage would have meant also the adoption of the uniconstituental analysis of ACIs. This would have amounted to a ‘real’ syntactic change, which obviously did not go through, although the beginnings of such a change seem to be visible in MHG in the shape of the extension of ACI to verbs of mental perception. The case shows that stylistic factors such as the Latin-like shaping of prose are not a sufficient trigger to induce a ‘real’ syntactic change. This is what we expect under an acquisition-based model of syntactic change (e.g. Lightfoot ), given that the input for the language learners is typically not confined to some passages from learned translations. Another explanation that is not incompatible with the first one is that the ‘niche’ of propositional complements governed by verbs of mental perception etc. was already filled by the morphologically similar control constructions or came to be filled by control constructions as they developed.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
The ACI construction in the history of German
. Conclusion Now we are in a position to attempt an answer to the questions guiding this chapter. As to the properties and the analysis of the ACI, we can say that the exclusive association with perception (and causative) verbs of Modern German (which suggests a biconstituental structure) was prevalent throughout the history of German. In MHG, a slight extension to other verb classes is noticeable (which suggests that a reanalysis towards a uniconstituental structure has taken place or was at least an option). In ENHG, we find a split between authors who are humanistically educated and write for the cognoscenti, who use the uniconstituental analysis frequently, and authors whose style is more modelled on the vernacular, who use an ACI only after perception verbs. Turning to the question of whether the ACI is inherited or a loan syntactic phenomenon, we can say that the answer is: both. The ACI in general is inherited, although it looks as if it has undergone reanalysis from the small clause structure that seems to have been the original, PIE structure to a biconstituental structure in pre-OHG times. The extension to other verb classes and the uniconstituental structure that goes with it is clearly influenced by Latin. As uniconstituental ACIs are a feature of learned written discourse only, they never made their way into the German vernacular and disappeared even in learned written discourse when the influence of Latin abated over time.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources Altdeutsche Predigten, hrsg. von Anton E. Schönbach, . Band: Texte. Graz: Styria, . Augsburger Chronik = Die Chroniken der schwäbischen Städte. Augsburg. Auf Veranlassung seiner Majestät des Königs von Bayern Maximilian II. Herausgegeben durch die historische Commission bei der Königlichen Akademie der Wissenschaften. Bde. Leipzig: Hirzel, –. (Die Chroniken der deutschen Städte vom . bis . Jahrhundert; ; ; ; ; ; ; ). (Neudruck: Vandenhoeck & Ruprecht, –). Aventin = Johannis Aventini Des Hochgelerten weitberümbten Beyerischen Geschichtschreibers Chronica. Anfenglich durch den Authorem im Latein verfertiget, nachmals aber den Teutschen zu guten, von jhm selber in gut gemein Hochteutsch gebracht . . . durch . . . Nicolaus Cisner . . . in Druck gegeben vnd mit nützlichen Glossen illustriert worden. Gedruckt zu Franckfort am Mayn, durch Johann und Sigmund Feyerabende Gevettern, . [Online: http://reader.digitale-sammlungen.de/de/fs/object/display/bsb_.html]. Bange, Chronik = Thüringische Chronick oder Geschichtbuch ( . . . ) anfenglich auss einem alten geschriebenen ( . . . ) Exemplar colligirt ( . . . ) vnd ferner durch weiland Ern Friderichen Schmidt Pfarherrn zu Grossen Beruingen reuidiert vnd vermehret/dann ( . . . ) von andern ( . . . ) continuirt. Jetzo aber ( . . . ) in Druck geben/durch Johan Bangen. Druck: Andreas Hantzsch, Mühlhausen, . [Online: FnhdC, Text ]. Basler Rezepte, in: Die kleineren althochdeutschen Sprachdenkmäler, VII, –. Berthold = Berthold von Regensburg. Vollständige Ausgabe seiner Predigten mit Anmerkungen von Franz Pfeiffer. Mit einem Vorwort von Kurt Ruh. Bd. . Wien, . (Nachdruck: Berlin: de Gruyter, ). Biterolf = Biterolf und Dietleib. Neu herausgegeben und eingeleitet von André Schnyder. Bern/ Stuttgart: Paul Haupt, . (Sprache und Dichtung; ). Bittgesang an Sanct Peter, in: Die kleineren althochdeutschen Sprachdenkmäler, XXI, –. Braunschweig = Urkundenbuch der Stadt Braunschweig, hrsg. von Ludwig Hänselmann, Heinrich Mack, Manfred R.W. Garzmann, Josef Dolle. Osnabrück: Wenner, . Brunschwig, Cirurgia = Das Buch der Cirurgia des Hieronymus Brunschwig. Straßburg, . Begleittext von Gustav Klein. München, . (Alte Meister der Medizin und Naturkunde in Facsimile-Ausgaben und Neudrucken des .–. Jahrhunderts; ). [Online: FnhdC, Text ]. Buch von den heiligen Altvätern = Hie vahet an das buoch von den hailigen Altvaetern ( . . . ). Cod. theol. et phil. der Württembergischen Landesbibliothek Stuttgart, Kloster Reute, . Jh. [Online: FnhdC, Text ]. Bucholtz, Wunder-Geschichte = Andreas Heinrich Buchholtz, Des Christlichen Teutschen Großfürsten Herkules Und Der Böhmischen Königlichen Fräulein Valiska WunderGeschichte. In acht Büchern und zween Teile abgefasset. Und allen Gott- und Tugendliebenden Seelen zur Christ- und ehrlichen Ergezligkeit ans Licht gestellet. Bd. . Braunschweig: Christoff Friedrich Zilliger, . (Nachdruck: Bern/Frankfurt a. M.: Peter Lang, –). (Nachdrucke deutscher Literatur des . Jahrhunderts; ).
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Busch, Schmetterling = Wilhelm Busch, Der Schmetterling, in: Wilhelm Busch. Werke. Historisch-kritische Gesamtausgabe, Band IV, bearb. und hrsg. von Friedrich Bohne, Hamburg: Standard-Verlag, . Caesar = C. Iulii Caesaris Commentarii de bello gallico, hrsg. von Bernhard Kübler, Leipzig: Teubner, . Christliche Leichpredigt, Bey der Käyserlichen Besing- und Begängnuß, der in Gott ruhenden Römischen Käyserin Mariae, welche den . Februarij dieses lauffenden . Jahrs zu Madrill, seliglich dem Herrn verschieden (Georg Scherer). Ingolstadt: Angermayer, . [Online: http://reader.digitale-sammlungen.de/de/fs/object/display/bsb_. html]. Christus und die Samariterin, in: Die kleineren althochdeutschen Sprachdenkmäler, XVII, –. Das ‚Beichtbüchlein‘ des Thomas Peuntner. Nach den Heidelberger, Melker, Münchner und Wiener Handschriften, hrsg. von Ernst Haberkern. Göppingen: Kümmerle, . Das Leben der heiligen Dorothea von Johannes Marienwerder, in: Scriptores Rerum Prussicarum. Die Geschichtsquellen der preußischen Vorzeit bis zum Untergange der Ordensherrschaft, hrsg. von Theodor Hirsch, Max Töppen, Ernst Strehlke. Zweiter Band, –. Leipzig: Hirzel, (Nachdruck: Frankfurt a. M.: Minerva, ). Das Nonnenturnier, in: Novellistik des Mittelalters. Märendichtung, –. Das St. Trudperter Hohe Lied. Krit. Ausg. von Hermann Menhardt, Bd. I: Einleitung, Bd. II: Text, Wörterverzeichnis und Anmerkungen, –. Halle (Saale): Niemeyer, . (Rhein. Beiträge und Hülfsbücher zur german. Philologie und Volkskunde; ; ). Dat nuwe Boych, in: Quellen zur Geschichte der Stadt Köln, Bd. I, hrsg. von Leonard Ennen und Gottfried Eckertz, –. Köln, . [Online: FnhdC, Text ]. De Heinrico, in: Die kleineren althochdeutschen Sprachdenkmäler, XXIII, –. Der ernsthafte König, in: Kleinere mittelhochdeutsche Erzählungen, Fabeln und Lehrgedichte. Bd. , Nr. , –. Der NONNE VON ENGELTHAL büchlein von der genaden uberlast (Christina Ebner), hrsg. von Karl Schröder. Stuttgart: Litterarischer Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; CVIII). Der Schlegel, in: Novellistik des Mittelalters. Märendichtung, –. Die Äffin und die Nuss, in: Die Kleindichtung des Strickers. Bd. III, Nr. , –. Die beiden Knechte, in: Kleinere mittelhochdeutsche Erzählungen, Fabeln und Lehrgedichte. Bd. , Nr. , –. Die drei Mönche zu Kolmar, in: Novellistik des Mittelalters. Märendichtung, –. Die Klage der Kunst, in: Kleinere Dichtungen Konrads von Würzburg, hrsg. von Edward Schröder. Mit einem Nachwort von Ludwig Wolff. Bd. . Berlin, . (Nachdruck: Dublin/ Zürich: Weidmann,3 ). Die Kleindichtung des Strickers. Gesamtausgabe in fünf Bänden, hrsg. von Wolfgang Wilfried Moelleken, Gayle Agler-Beck, Robert E. Lewis. Bde. Göppingen, –. (Göppinger Arbeiten zur Germanistik; /I–V). [Online: http://www.mhdwb-online.de/Etexte/PDF/ StrKd.pdf]. Die kleineren althochdeutschen Sprachdenkmäler, hrsg. von Elias von Steinmeyer. Berlin: Weidmannsche Buchhandlung, . (Nachdruck: Dublin/Zürich: Weidmann,3 ).
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Die Lilie, in: Die Lilie, eine mittelfränkische Dichtung in Reimprosa und andere geistliche Gedichte, aus der Wiesbadener Handschrift, hrsg. von Paul Wüst, –. Berlin: Weidmannsche Buchhandlung, . Die Rache des Ehemannes, in: Novellistik des Mittelalters. Märendichtung, –. Die Suche nach dem glücklichen Ehepaar, in: Novellistik des Mittelalters. Märendichtung, –. Die Sünderin, in: Eine Schweizer Kleinepiksammlung des . Jahrhunderts, hrsg. von Hanns Fischer, –. Tübingen: Niemeyer, . Die unschuldige Mörderin, in: Novellistik des Mittelalters. Märendichtung, –. Durandus’ Rationale = Durandus’ Rationale in spätmittelhochdeutscher Übersetzung. Das vierte Buch nach der HS. CVP , hrsg. von Gerard H. Buijssen. Assen, . (Studia Theodisca; VI). [Online: FnhdC, Text ]. Eschenbach, Alexander = Alexander von Ulrich von Eschenbach, hrsg. von Wendelin Toischer. Tübingen: H. Laupp für den Litterarischen Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; CLXXXIII). Essener Evangeliarglossen, in: Kleinere altsächsische sprachdenkmäler mit anmerkungen und glossar, –. Fastnachtsspiele = Hans Sachs. Fastnachtsspiele, hrsg. von Adelbert von Keller. Tübingen: H. Laupp für den Litterarischen Verein in Stuttgart, –. (Nachdruck: Hildesheim: Georg Olms Verlagbuchhandlung, ). Frauenlob = Frauenlob (Heinrich von Meißen). Leichs, Sangsprüche, Lieder. . Teil: Einleitungen, Texte, hrsg. von Karl Stackmann und Karl Bertrau. Göttingen: Vandenhoeck & Ruprecht, . Gellert, Betschwester = Christian Fürchtegott Gellert, Die Betschwester: ein Lustspiel in drei Aufzügen, in: Christian Fürchtegott Gellert. Werke, hrsg. von Gottfried Honnefelder, Bd. : Fabeln, Gedichte, Lustspiele. Frankfurt a. M.: Insel, . Glossen zu Homilien Gregors des Großen, in: Kleinere altsächsische sprachdenkmäler mit anmerkungen und glossar, –. Gropper, Gegenwertigkeit = Vonn warer/wesenlicher/vnd pleibeder Gegenwertigkeit des Leybs und Bluots Christi ( . . . ). durch Johannem Gropperum d. Archidiaco der h. Kirchen zu Coellen. Druck: Jaspar Gennep, Köln . [Online: FnhdC, Text ]. Heliand = Heliand und Genesis, hrsg. von Otto Behaghel. . überarb. Aufl. von Burkhard Taeger. Tübingen: Niemeyer, . (Altdeutsche Textbiliothek; ). Herberstein, Moscouia = Moscouia der Hauptstat der Reissen durch Herrn Sigmunden Freyherrn zu Herberstein ( . . . ) zusamen getragen ( . . . ). Wien, . [Online: FnhdC, Text ]. Herbort = Herbort’s von Fritslâr liet von Troye, hrsg. von Karl Frommann. Quedlinburg/ Leipzig: Basse, . (Bibliothek der gesammten deutschen National-Literatur von der ältesten bis auf die neuere Zeit, Abteilung ; ). [Online: http://reader.digitale-sammlungen. de/de/fs/object/display/bsb_.html?zoom=.]. Herder, Briefe = Johann Gottfried Herder. Briefe zur Beförderung der Humanität. Band , hrsg. von Heinz Stolpe in Zusammenarbeit mit Hans-Joachim Kruse und Dietrich Simon. Berlin/ Weimar: Aufbau Verlag, . Herder, Ideen = Johann Gottfried Herder. Ideen zur Philosophie der Geschichte der Menschheit. Band , hrsg. von Heinz Stolpe, Berlin/Weimar: Aufbau Verlag, .
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Hermann von Fritslar = Deutsche Mystiker des vierzehnten Jahrhunderts, hrsg. von Franz Pfeiffer. Erster Band: Hermann von Fritzlar/Nikolaus von Strassburg/David von Augsburg. Leipzig: Göschen, . (Neudruck: Aalen: Scientia, ). Hessisches Urkundenbuch. Abteilung I, hrsg. von Arthur Wyss/Abteilung II, hrsg. von Heinrich Reimer. Leipzig: Hirzel, –. (Publicationen aus den Kgl. Preußischen Staatsarchiven. Veranlasst und unterstützt durch die K. Archiv-Verwaltung; ; ; ; ; ; ; ). [Online: Bd. : http://dbooks.bodleian.ox.ac.uk/books/PDFs/N.pdf; Bd. : https://archive. org/stream/publicationenauarchgoog#page/n/mode/up]. Hildebrandslied, in: Die kleineren althochdeutschen Sprachdenkmäler, I, –. Hoefer, Urkunden = Auswahl der ältesten Urkunden deutscher Sprache im Königl. Geheimen Staats- und Kabinets-Archiv zu Berlin, hrsg. von Ludwig Franz Hoefer. Hamburg: Perthes, . [Online: https://archive.org/stream/auswahlderlteststaagoog#page/n/mode/up/]. Hugo von Trimberg = Der Renner von Hugo von Trimberg, hrsg. von Gustav Ehrismann, Band I–IV. Tübingen: H. Laupp für den Litterarischen Verein, –. (Bibliothek des litterarischen Vereins in Stuttgart; ; ; ; ). Hutten, Gespräch-Büchlin = Gespräch-Büchlin Herr Ulrichs von Hutten. Straßburg: Schott, . [Online: http://daten.digitale-sammlungen.de/~db//bsb/images/]. Imago Flandriae. Diese verborgene weissagung ( . . . ) den gemeinen standt der Königlichen Graeffschafft Flanderen antreffende ( . . . )/Matth. Quadus chalcogr. () (Matthias Quad). [Online: http://sammlungen.ub.uni-frankfurt.de/freytag/content/titleinfo/]. Isidor = Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten, hrsg. von Hans Eggers. Tübingen: Niemeyer, . Iwein = Iwein. Eine Erzählung von Hartmann von Aue, hrsg. von Georg Friedrich Benecke und Karl Lachmann, neu bearb. von Ludwig Wolff, . Ausgabe, Bd. : Text. Berlin: de Gruyter . Jacob Böhme. Morgenröte im Aufgangk (), hrsg. von Ferdinand van Ingen. Frankfurt a. M.: Deutscher Klassiker Verlag, . Johann Michael Moscherosch. Wunderliche und wahrhafftige Gesichte Philanders von Sittewald (), hrsg. von Wolfgang Harms. Stuttgart: Reclam, . (Reclams Universalbibliothek ). Johannes Gropper. Briefwechsel II (–), hrsg. von Reinhard Braunisch. Münster: Aschendorff, . Jüngerer Physiologus, in: Bibliothek des Mittelalters, Bd. , hrsg. von Walter Haug und Benedikt Konrad Vollmann, –. Frankfurt a. M.: Deutscher Klassiker Verlag, . Kafka, Amerika = Franz Kafka, Amerika, in: Franz Kafka. Gesammelte Werke, Bd. , hrsg. von Max Brod. Frankfurt a. M.: S. Fischer, . Kazmair, Denkschrift = Jörg Kazmair’s Denkschrift über die Unruhen zu München in den Jahren –, in: Die Chroniken der baierischen Städte. Regensburg. Landshut. Mühldorf. München. Auf Veranlassung seiner Majestät des Königs von Bayern Maximilian II. Herausgegeben durch die historische Commission bei der Königlichen Akademie der Wissenschaften. Leipzig: Hirzel, . (Die Chroniken der deutschen Städte vom . bis . Jahrhundert; ). Kleinere altsächsische sprachdenkmäler mit anmerkungen und glossar, hrsg. von Elis Wadstein, Norden/Leipzig: D. Soltau’s Verlag, .
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Kleinere mittelhochdeutsche Erzählungen, Fabeln und Lehrgedichte. Bd. . Die Heidelberger Handschrift Cod. Pal. Germ. , hrsg. von Gustav Rosenhagen. Berlin, . (Deutsche Texte des Mittelalters, ). (Nachdruck: Dublin/Zürich: Weidmann, ). [Online: https:// archive.org/details/DieHeidelbergerHandschriftCpg]. Koelhoff, Cronica = Die Cronica van der hilliger Stat va Coelle, Druck: Johann Koelhoff d. J. Köln, . (Nachdruck: Köln ). [Online: FnhdC, Text ]. Kortüm, Untersuchungen = Klaus Kortüm, Aktuelle Untersuchungen in der villa rustica von Hechingen-Stein, in: Archäologische Ausgrabungen in Baden-Württemberg , hrsg. von Dirk Krausse, –. Darmstadt: Theiss, . Kottanerin, Denkwürdigkeiten = Die Denkwürdigkeiten der Helene Kottanerin, hrsg. von Karl Mollay. Wien, . (Wiener Neudrucke; ). [Online: FnhdC, Text ]. Kudrun. Nach der Ausgabe von Karl Bartsch, hrsg. von Karl Stackmann. Tübingen: Niemeyer, . (Altdeutsche Textbibliothek; ). Küng, Moral = Hans Küng, Anständig wirtschaften. Warum Ökonomie Moral braucht. München: Piper, . Leibniz, Gedächtnis = Gotthold Wilhelm Leibniz, Zum Gedächtnis an den . November : Abhandlung über die beste philosophische Ausdrucksweise, Ermanung an die Teutsche, ihren Verstand und Sprache besser zu üben; Unvorgreifliche Gedanken betreffend die Ausübung und Verbesserung der teutschen Sprache, hrsg. und erläutert von Paul Pietsch. Berlin: Verlag des Allgemeinen Deutschen Sprachvereins, . Lessing, Laokoon = Gotthold Ephraim Lessing, Laokoon oder über die Grenzen der Malerei und Poesie, in: Gotthold Ephraim Lessing. Werke. Band IV, hrsg. von Herbert G. Göpfert in Zusammenarbeit mit Karl Eibl, Helmut Göbel, Karl S. Guthke, Gerd Hillen, Albert von Schirmding und Jörg Schönert. München: Hanser, . Lessing, Tod = Gotthold Ephraim Lessing, Wie die Alten den tod gebildet, in: Gotthold Ephraim Lessing. Werke. Band IV, hrsg. von Herbert G. Göpfert in Zusammenarbeit mit Karl Eibl, Helmut Göbel, Karl S. Guthke, Gerd Hillen, Albert von Schirmding und Jörg Schönert. München: Hanser, . Lewald, Clementine = Fanny Lewald, Clementine, in: Fanny Lewald. Gesammelte Werke. Neue revidierte Ausgabe, Bd. . Berlin: Verlag von Otto Janke, –. Livius = Titi Livi ab vrbe condita, ed. by Robert Maxwell Ogilvie. Oxford: Clarendon Press, . Lorscher Beichte, in: Die kleineren althochdeutschen Sprachdenkmäler, XLVI, –. Lübeck = Urkundenbuch der Diözese Lübeck. Bd. , bearb. von Wilhelm Leverkus (), Bde –, bearb. von Wolfgang Prange. Neumünster: Wachholtz, –. Luther, Frühe Schriften = D Martin Luthers Werke. Kritische Gesamtausgabe. . Band. Weimar: H. Böhlau, . (= Weimarer Ausgabe (WA) : Schriften /). (Nachdruck: Stuttgart: Metzler, –; WA Gesamtausgabe; ..). Luther-Bibel = Martin Luther. Die gantze Heilige Schrifft Deudsch. Wittenberg , Letzte zu Luthers Lebzeiten erschienene Ausgabe, hrsg. von Hans Volz unter Mitarbeit von Heinz Blanke, Textredaktion Friedrich Kur. Bd. I–II. München: Rogner & Bernhard, . Luther-Bibel = Die Bibel oder Die ganze Heilige Schrift des Alten und Neuen Testaments, nach der deutschen Übersetzung Martin Luthers, durchgesehen im Auftrag der Deutschen Evangelischen Kirchenkonferenz. Mit erklärenden Anmerkungen. Stuttgart: Privilegierte Württembergische Bibelanstalt, .
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Mair, Troja = Das ‘Buch von Troja’ von Hans Mair. Kritische Textausgabe und Untersuchung, hrsg. von Hans-Josef Dreckmann, München: Fink, . [Online: FnhdC, Text ]. Mathesius, Passionale = Passionale Mathesij, das ist/christliche vnd andechtige Erklerung vnd Außlegung des zwey und zwantzigsten Psalms/vnd drey vnd funffzigsten Capitels des Propheten Esaiæ ( . . . ) geprediget durch ( . . . ) M. Johannem Mathesium/weyland Pfarrer in S. Joachimsthal. Druck: Johann Beyer. Leipzig, . [Online: FnhdC, Text ]. Memento Mori, in: Denkmäler deutscher Reime und Prosa aus dem VIII.–XII. Jahrhundert, hrsg. von Karl Müllenhof und Wilhelm Scherer, dritte Ausgabe von Elias von Steinmeyer. Erster Band: Texte, –. Berlin, . (Nachdruck: Zürich: Weidmann, ). Merswin, Mannen = Des Gottesfreundes im Oberland (= Rulmann Merswin’s) Buch von den zwei Mannen. Nach der ältesten Strassburger Handschrift, hrsg. von Friedrich Lauchert. Bonn, . [Online: FnhdC, Text ]. Mitteldeutsche Predigten, Mitgetheilt von Adalbert Jeitteles, in: Germania. Vierteljahresschrift für Deutsche Alterthumskunde, . Jhg. Neue Reihe . Jhg., –. Wien, . Mitternacht, Soldat = Johann Sebastian Mitternacht, Der unglückselige Soldat, in: Johann Sebastian Mitternacht. Dramen, hrsg. von Marianne Kaiser. Tübingen: Max Niemeyer Verlag, . (Deutsche Neudrucke, Reihe: Barock; ). Monseer Fragmente = The Monsee Fragments. Newly collated text with introduction, notes, grammatical treatise, and exhaustive glossary and a photo-lithographic facsimile, ed. by George Allison Hench. Straßburg: Karl J. Teubner, . Monumenta Zollerana. Quellensammlung zur Geschichte des erlauchten Hauses der Grafen von Zollern und Burggrafen von Nürnberg, hrsg. von Rudolph von Stillfried und Traugott Märcker. Bde. Bd. Halle (Saale): Graeger, Bde. – Berlin: Ernst & Korn, –. [Online: Bd. URL: http://reader.digitale-sammlungen.de/de/fs/object/display/bsb_ .html]. Muspilli, in: Die kleineren althochdeutschen Sprachdenkmäler, XIV, –. Neidhart, Eunuchus = Der Eunuchus des Terenz. Uebersetzt von Hans Neidhart , hrsg. von Hermann Fischer. Tübingen: H. Laupp für den Litterarischen Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; CCLXV). [Online: FnhdC, Text ]. Nibelungenlied = Das Nibelungenlied. Paralleldruck der Handschriften A, B und C nebst Lesarten der übrigen Handschriften, hrsg. von Michael S. Batts. Tübingen: Niemeyer, . Notker, Consolatio = Notker der Deutsche: Boethius, ‘De consolatione Philosophiae’, hrsg. von Petrus W. Tax. Bde. Tübingen: Niemeyer, , , . (Die Werke Notkers des Deutschen, Neue Ausgabe; –). Notker, De Interpretatione = Notker der Deutsche: Boethius’ Bearbeitung von Aristoteles’ Schrift ‘De Interpretatione’, hrsg. von James C. King. Tübingen: Niemeyer, . (Die Werke Notkers des Deutschen, Neue Ausgabe; ). Notker, Martianus Capella = Notker der Deutsche: Martianus Capella, De nuptiis Philologiae et Mercurii’, hrsg. von James C. King. Tübingen: Niemeyer, . (Die Werke Notkers des Deutschen, Neue Ausgabe; ). Notker, Psalter = Notker der Deutsche: Der Psalter, hrsg. von Petrus W. Tax. Bde. Tübingen: Niemeyer , , . (Die Werke Notkers des Deutschen, Neue Ausgabe; –). Novellistik des Mittelalters. Märendichtung, hrsg., übers. und kommentiert von Klaus Grubmüller. Frankfurt a. M.: Deutscher Klassiker Verlag, . (Bibliothek des Mittelalters; ).
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Nürnberg Chroniken = Die Chroniken der fränkischen Städte. Nürnberg. Auf Veranlassung seiner Majestät des Königs von Bayern Maximilian II. Herausgegeben durch die historische Commission bei der Königlichen Akademie der Wissenschaften. Bde. Leipzig: Hirzel, –. (Die Chroniken der deutschen Städte vom . bis . Jahrhundert, Bde. –; ; ). (Neudruck: Göttingen: Vandenhoeck & Ruprecht, –). Oldenburg = Urkundenbuch der Stadt Oldenburg, bearb. von Dietrich Kohl. Münster: Aschendorff, . (Inventare der nichtstaatlichen Archive Westfalens, Vol. ). Opitz, Poemata = Martin Opitz. Teutsche Poemata, hrsg. von Georg Witkowski. Halle (Saale): Niemeyer, . [Online: https://archive.org/stream/martinopitzteutopitgoog#page/n/ mode/up]. Otfrid = Otfrid von Weißenburg. Evangelienbuch, Bd. I: Edition nach dem Wiener Codex , hrsg. u. bearbeitet von Wolfgang Kleiber unter Mitarbeit von Rita Heuser, Teil : Text, Teil : Einleitung und Apparat, mit Beiträgen von Wolfgang Haubrichs, Norbert Kössinger, Otto Mazal, Norbert H. Ott., und Michael Klaper. Tübingen: Niemeyer, . (Otfrid von Weißenburg: Evangelienbuch, hrsg. von Wolfgang Kleiber und Ernst Hellgardt; I/–). Oxforder Benedictinerregel, ed. by Eduard Sievers. Tübingen: Fues’sche Buchdruckerei, . Pariser Gespräche, in: ‘In Francia fui’. Studien zu den romanisch-germanischen Interferenzen und zur Grundsprache der althochdeutschen Pariser (Altdeutschen) Gespräche, nebst einer Edition des Textes, hrsg. von Wolfgang Haubrichs und Max Pfister. Mainz: Akad. der Wiss. und der Literatur, . Parzival = Wolfram von Eschenbach. Parzival, Studienausgabe, . Auflage. Mittelhochdeutscher Text nach der sechsten Ausgabe von Karl Lachmann. Übersetzung von Peter Knecht. Mit Einführungen zum Text der Lachmannschen Ausgabe und in Probleme der ‚Parzival‘-Interpretation von Bernd Schirok. Berlin/New York: de Gruyter, . Pilgerfahrt = Pilgerfahrt des Wilhelm von Thüringen, in: Ostmitteldeutsche Chrestomathie. Proben der frühen Schreib-und Druckersprache des mitteldeutschen Ostens, hrsg. von Johannes Erben. Berlin: Akademie Verlag, . (Veröffentlichungen des Instituts für deutsche Sprache und Literatur. Deutsche Akademie der Wissenschaften zu Berlin; ). Predigt Leipzig, in: Altdeutsche Predigten, hrsg. von Anton E. Schönbach. . Band: Texte, –. Graz: Styria, . [Online: FnhdC, Text ]. Predigten und Tischgebete, in: Altdeutsche Predigten und Gebete aus Handschriften. Gesammelt und zur Herausgabe vorbereitet von Wilhelm Wackernagel. Mit Abhandlungen und einem Anhang, hrsg. mit einem Vorwort von Max Rieger, XLII, –. Basel: Schweighauser, . Priester Konrad = Das Predigtbuch des Priester Konrad, hrsg. von Anton E. Schönbach. Graz: Styria, . (Altdeutsche Predigten; ). Prosalancelot = Prosalancelot Bd. I, II: Lancelot und Ginover. Nach der Heidelberger Handschrift Cod. Pal. germ. , hrsg. von Reinhold Kluge. Ergänzt durch die Handschrift Ms. allem. – der Bibliothèque de l’Arsenal Paris. Übersetzt, kommentiert und herausgegeben von Hans-Hugo Steinhoff. Frankfurt a. M.: Deutscher Klassiker Verlag, . (Bibliothek des Mittelalters; , ; Bibliothek deutscher Klassiker; ). Psalter = Zwei Psalter aus dem . Jahrhundert (Dresden MS. und Hamburg in SCR. ) und drei verwandte Bruchstücke aus Schleiz, Breslau und Düsseldorf, hrsg. von Hans Eggers. Berlin: Akademie Verlag, . (Deutsche Texte des Mittelalters; ). [Online: FnhdC, Text ].
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Psalm CXXXVIII, in: Die kleineren althochdeutschen Sprachdenkmäler, XXII, –. Rauwolf, Reisebeschreibung = Leonharti Rauwolfen/der Artzney Doctorn/vnd bestelten Medici zuuo Augspurg. Aigentliche Beschreibung der Raiß/so er ( . . . ) inn die Morgenlaender ( . . . ) selbs volbracht. Druck: Leonhart Reinmichel. Lauingen, . [Online: FnhdC, Text ]. Reichenauer Beichte, in: Die kleineren althochdeutschen Sprachdenkmäler, LI, –. Reuter, Hüsung = Fritz Reuter, Kein Hüsung. Besonderer Abdruck aus der Volks-Ausgabe der sämmtl. Werke. . Auflage. Wismar: Hinstorff ’sche Hofbuchhandlung, . Reuter, Schurr Murr = Fritz Reuter, Schurr Murr. Wismar: Hinstorff ’sche Hofbuchhandlung, . Rheinfränkische Psalmenübersetzung, in: Die kleineren althochdeutschen Sprachdenkmäler, XXXIX, –. Ritter, Bürger und Bauern, in: Ostmitteldeutsche Chrestomathie. Proben der frühen Schreib-und Druckersprache des mitteldeutschen Ostens, hrsg. von Johannes Erben. Berlin: Akademie Verlag, . (Veröffentlichungen des Instituts für deutsche Sprache und Literatur. Deutsche Akademie der Wissenschaften zu Berlin; ). Rothe, Chronik = Düringische Chronik des Johann Rothe. Namens des Vereines für thüringische Geschichte und Alterthumskunde herausgegeben von R. v. Liliencron. (Thüringische Geschichtsquellen; ). Jena: Friedrich Frommann, . [Online: FnhdC, Text ]. Rozmital = Des Böhmischen Herrn Leos von Rozmital Ritter-, Hof- und Pilgerreise durch die Abendlande, hrsg. von Johann Andreas Schmeller. Stuttgart: Litterarischer Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; VII). [Online: http://reader.digitalesammlungen.de/de/fs/object/display/bsb_.html]. Sammlung kleinerer althochdeutscher Sprachdenkmäler, hrsg. von Gerhard Köbler. Gießen: Arbeiten zur Rechts- und Sprachwissenschaft Verlag, . Schad, Reisebuch = Das Memorial- und Reisebuch des Hans Schad. Ein Beitrag zur Geschichte Ulms im . Jahrhundert, hrsg. v. Hans Greiner, in: Württembergische Vierteljahrshefte für Landesgeschichte, Neue Folge (), –. [Online: http://www.mgh-bibliothek.de/ etc/zeitschriftenmagazin/WVLG_NF__.pdf]. Scharnebeck = Urkundenbuch des Klosters Scharnebeck: –, bearb. von Dieter Brosius. Hildesheim: Lax, . Steinfurt = Inventar des Fürstlichen Archivs zu Burgsteinfurt, bearb. von Alfred Bruns und Wilhelm Kohl. Hrsg. von Alfred Bruns. Münster Westf.: Aschendorff, –. Stralsund = Der Stralsunder Liber memorialis, bearb. von Horst-Diether Schroeder. (Veröffentlichungen des Stadtarchivs Stralsund; V/–V/). Teil : Fol. –, –. Schwerin: Petermänken-Verlag, . Teil : Fol. –, –. Weimar: Hermann Böhlaus Nachfolger, . Teil : Fol. –, –. Weimar: Hermann Böhlaus Nachfolger, . Teil : Fol. –, –. Rostock: Hinstorff Verlag, . Teil : Fol. –, –. Weimar: Hermann Böhlaus Nachfolger, . Teil : Fol. –, –. Weimar: Hermann Böhlaus Nachfolger, . Suchenwirt = Peter Suchenwirt’s Werke aus dem vierzehnten Jahrhunderte. Ein Beytrag zur Zeit- und Sittengeschichte. Zum ersten Mahle in der Ursprache aus Handschriften herausgegeben und mit einer Einleitung, historischen Bemerkungen und einem Wörterbuche begleitet von Alois Primisser. Wien: Wallishausser, . [Online: http://digital.ub.uniduesseldorf.de/ihd/content/pageview/].
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Tatian = Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. , hrsg. von Achim Masser. Göttingen: Vandenhoeck & Ruprecht, . Tauler, Sermon = Sermon des grosz gelarten, in Gnade erlauchte Doctoris Johannis Thauleri Predigerr Ordens ( . . . ). Druck: Leipzig: Konrad Kachelofen, . [Online: FnhdC, Text ]. Tepl, Ackermann = Johann von Tepl. Der Ackermann und der Tod, hrsg. u. übers. von Felix Genzmer. Stuttgart: Reclam, . Thüringisches Wörterbuch, Band IV (L–Q), hrsg. von der Sprachwissenschaftlichen Kommission der Sächsischen Akademie der Wissenschaften zu Leipzig. Berlin: Akademie Verlag, . Tristan = Gottfried von Straßburg. Tristan, hrsg. von Karl Marold. Unveränderter fünfter Abdruck nach dem dritten, mit einem auf Grund von Friedrich Rankes Kollationen verbesserten kritischen Apparat besorgt und mit einem erweiterten Nachwort versehen von Werner Schröder. Berlin/New York: de Gruyter, . Tünger, Facetiae = Augustin Tüngers Facetiae Latinae et Germanicae, hrsg. von Adelbert von Keller. Tübingen: H. Laupp für den Litterarischen Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; CXVIII). Uelzen = Urkundenbuch der Stadt Uelzen, bearb. von Thomas Vogtherr. Hildesheim: Lax, . Ulrich, Syrerinn = Anton Ulrich, Herzog von Braunschweig. Die durchleuchtige Syrerinn Aramena. Teile. Nürnberg: Johann Hofmann, (Band ); (Band ); (Band ); (Band ). (Nachdruck: Bern/Frankfurt a. M.: Herbert Lang & Cie AG, –). (Nachdrucke deutscher Literatur des . Jahrhunderts; /–). Urkunden des Oberlausitzer Hussitenkrieges und der gleichzeitigen die Sechslande angehenden Fehden. Im Auftrage der Oberlausitzischen Gesellschaft der Wissenschaften gesammelt und herausgegeben von Dr. Richard Jecht, I. Band umfassend die Jahre –, Görlitz, . [Online: http://digital.slub-dresden.de/id]. Urkunden Heiligenkreuz = Urkunden des Cistercienser-Stiftes Heiligenkreuz im Wiener Walde. I. Theil. Herausgegeben von Johann Nepomuk Weis. Wien: Kaiserlich-königliche Hof- und Staatsdruckerei, . (Fontes Rerum Austriacarum. Österreichische Geschichts-Quellen. Herausgegeben von der Historischen Commission der Kaiserlichen Akademie der Wissenschaften in Wien. Zweite Abteilung; XI). [Online: https://archive.org/stream/ fontesrerumaustkommgoog#page/n/mode/up]. Urkundenbuch Basel = Urkundenbuch der Landschaft Basel. Bde. Hrsg. von Heinrich Boos. Basel: Detloff, –. Urkundenbuch des Klosters Barsinghausen, bearb. von Achim Bonk. Hannover: Hahn, . Urkundenbuch des Klosters Mariengarten, bearb. von Manfred von Boetticher. Hildesheim: Lax, . Urkundenbuch des Stifts Börstel, bearb. von Roland Rölker und Werner Delbanco. Osnabrück: Selbstverlag des Vereins für Geschichte und Landeskunde von Osnabrück, . Vaterunser, in: Die Kleindichtung des Strickers, Bd. II, Nr. , –. Von drei Freunden, in: Kleinere mittelhochdeutsche Erzählungen, Fabeln und Lehrgedichte. Bd. , Nr. , –. Vom heiligen Geist, in: Die Kleindichtung des Strickers. Bd. II, Nr. , –. Von einem varnden Schuler, in: Novellistik des Mittelalters. Märendichtung, –.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Vulgata = Biblia sacra iuxta Vulgatam versionem, hrsg. von Robert Weber, Stuttgart: Deutsche Bibelgesellschaft, . Wessobrunner Glauben = Bamberger und Erster Wessobrunner Glauben und Beichte, in: Die kleineren althochdeutschen Sprachdenkmäler, XXVIII, –. Wessobrunner Predigt, in: Althochdeutsches Lesebuch, hrsg. von Wilhelm Braune and Ernst Ebbinghaus, –. . Aufl. Tübingen: Niemeyer, . Wickram, Nachbarn = Georg Wickram. Sämtliche Werke. Bd. : Von guten und bösen Nachbarn, hrsg. von Hans-Gert Roloff. Berlin/New York: Walter de Gruyter, . Wigamur. Éd. avec introd. et index par Danielle Buschinger. Göppingen: Kümmerle, . (Göppinger Arbeiten zur Germanistik; ). Wiener Genesis = Die frühmittelhochdeutsche Wiener Genesis. Kritische Ausgabe mit einem einleitenden Kommentar zur Überlieferung, hrsg. von Kathryn Smits. Berlin: Schmidt, . (Philologische Studien und Quellen; ). Wunderer = Der Wunderer, hrsg. von Florian Kragl. Berlin: de Gruyter, . (Texte und Studien zur mittelhochdeutschen Heldenepik; ). Wyle, Translationen = Translationen von Niclas von Wyle, hrsg. von Adelbert Keller. Stuttgart: Litterarischer Verein, . (Bibliothek des litterarischen Vereins in Stuttgart; LVII). [Online: http://reader.digitale-sammlungen.de/de/fs/object/display/bsb_. html]. Zainersche Bibel = Die erste deutsche Bibel, hrsg. von William Kurrelmeyer. Bde. Stuttgart: Litterarischer Verein, –. (Bibliothek des Litterarischen Vereins in Stuttgart). Cited digital corpora Corpus of Historical Low German (CHLG), Anne Breitbarth, George Walkden, Sheila Watts (–), Universiteit Gent, University of Manchester, University of Cambridge, http:// www.chlg.ac.uk/. Das Bonner Frühneuhochdeutschkorpus (FnhdC), Korpora.org, http://www.korpora.org/fnhd/. Das Deutsche Referenzkorpus (DeReKo), Institut für Deutsche Sprache, Mannheim, http:// www.ids-mannheim.de/kl/projekte/korpora/. Deutsche Diachrone Baumbank (DDB), Institut für deutsche Sprache und Linguistik der Humboldt-Universität zu Berlin, http://korpling.german.hu-berlin.de/ddb-doku/. Deutsches Textarchiv (DTA), Berlin-Brandenburgische Akademie der Wissenschaften (–), http://www.deutsches-textarchiv.de/. Freiburger Korpus (Grundstrukturen: Freiburger Korpus), Archiv für Gesprochenes Deutsch, Abteilung Pragmatik am Institut für Deutsche Sprache, Mannheim, http://agd. ids-mannheim.de. German Manchester Corpus (GerManC), Martin Durell et al. (–), The GerManC project: A representative historical corpus of German (–), University of Manchester, http:// www.llc.manchester.ac.uk/research/projects/germanc/. Kali Korpus, Leibniz Universität Hannover, http://www.kali.uni-hannover.de. Mittelhochdeutsche Begriffsdatenbank (MHDBDB), Universität Salzburg, Koordination: Margarete Springeth, Technische Leitung: Nikolaus Morocutti/Daniel Schlager (–), http://www.mhdbdb.sbg.ac.at. Penn Corpora of Historical English, Anthony Kroch et al. (–), Department of Linguistics, University of Pennsylvania, https://www.ling.upenn.edu/hist-corpora/.
OUP CORRECTED PROOF – FINAL, 14/2/2018, SPi
Sources
Referenzkorpus Altdeutsch (DDD), Karin Donhauser, Jost Gippert, Rosemarie Lühr (–), Humboldt-Universität zu Berlin, Universität Frankfurt am Main, Universität Jena, http:// www.deutschdiachrondigital.de/. Referenzkorpus Frühneuhochdeutsch (REF), Hans-Joachim Solms, Ulrike Demske, Stefanie Dipper, Klaus-Peter Wegera, (–), Ruhr Universität Bochum, Martin-Luther-Universität Halle-Wittenberg, Universität Potsdam, http://www.ruhr-uni-bochum.de/wegera/ref/. Referenzkorpus Mittelhochdeutsch (ReM), Thomas Klein, Klaus-Peter Wegera, Stefanie Dipper, Claudia Wich-Reif (), Referenzkorpus Mittelhochdeutsch (–), Version ., https://www.linguistics.ruhr-uni-bochum.de/rem/, ISLRN ----. Referenzkorpus Mittelniederdeutsch/Niederrheinisch (ReN), Ingrid Schröder, Robert Peters (–), Universität Hamburg, Westfälische Wilhelms-Universität Münster, https://vs. corpora.uni-hamburg.de/ren/index.html. Tatian Corpus of Deviating Examples (T-Codex .), Humboldt-Universität Berlin, Universität Potsdam, SFB (Projekt B), https://korpling.org/annis/#_c=YjQudGFaWFu. TITUS (Thesaurus Indogermanischer Text- und Sprachmaterialien). Jost Gippert, Javier Martìnez, Agnes Korn, Universität Frankfurt am Main, http://titus.uni-frankfurt.de.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References Abraham, Werner (). Wortstellung und das Mittelfeld im Deutschen, in Werner Abraham (ed.), Erklärende Syntax des Deutschen. Tübingen: Narr, –. Abraham, Werner (). Autonomous and non-autonomous components of ‘grammatic(al)ization’: Economy criteria in the emergence of German negation, Sprachtypologie und Universalienforschung : –. Abraham, Werner (). The grammaticalization of the infinitival preposition—toward a theory of ‘grammaticalizing reanalysis’, Journal of Comparative Germanic Linguistics : –. Admoni, Wladimir (). Historische Syntax des Deutschen. Tübingen: Niemeyer. Ahlers, Timo (). Komplexe C0-phobe Verben des Deutschen. Diploma thesis, Universität Wien. Altmann, Hans (). Formen der ‘Herausstellung’ im Deutschen: Rechtsversetzung, Linksversetzung, freies Thema und verwandte Konstruktionen. Tübingen: Niemeyer. Anderson, Stephen R. (). Wackernagel’s revenge: Clitics, morphology, and the syntax of second position, Language : –. Åsdahl Holmberg, Märta (). Studien zu den verbalen Pseudokomposita im Deutschen (Göteborger Germanistische Forschungen ). Lund: Acta Universitatis Gothoburgensis. Asher, Nicholas, and Alex Lascarides (). Logics of Conversation. Cambridge: Cambridge University Press. Asher, Nicholas, and Laure Vieu (). Subordinating and coordinating discourse relations, Lingua : –. Askedal, John Ole (). ‘Ersatzinfinitiv/Partizipialersatz’ und Verwandtes, Zeitschrift für germanistische Linguistik : –. Auer, Peter (). Vom Ende deutscher Sätze, Zeitschrift für germanistische Linguistik : –. Auer, Peter (). Zur Verbspitzenstellung im gesprochenen Deutsch, Deutsche Sprache : –. van der Auwera, Johan (). The Jespersen cycles, in Elly van Gelderen (ed), Cyclical Change. Amsterdam: John Benjamins, –. Averintseva-Klisch, Maria (). Rechte Satzperipherie im Diskurs. NP-Rechtsversetzung im Deutschen. Tübingen: Stauffenburg. AWB = Grosse, Rudolf (ed) (–). Althochdeutsches Wörterbuch. Auf Grund der von Elias Steinmeyer hinterlassenen Sammlungen im Auftrag der sächsischen Akademie der Wissenschaften in Leipzig. Begründet von Elisabeth Karg-Gasterstädt und Theodor Frings. Bearbeitet von Siegfried Blum u.a. Berlin: Akademie-Verlag. Axel, Katrin (). Zur diachronen Entwicklung der syntaktischen Integration linksperipherer Adverbialsätze im Deutschen. Ein Beispiel für syntaktischen Wandel?, Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB) : –. Axel, Katrin (). The syntactic integration of preposed adverbial clauses on the German left periphery: A diachronic perspective, in Horst Lohnstein and Susanne Trissler (eds), The Syntax and Semantics of the Left Periphery. Berlin: de Gruyter, –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Axel, Katrin (). Studies in Old High German Syntax: Left Sentence Periphery, Verb Placement and Verb-second. Amsterdam: John Benjamins. Axel, Katrin (a). The verb-second property in Old High German: Different ways of filling the prefield, in Roland Hinterhölzl and Svetlana Petrova (eds), Information Structure and Language Change. New Approaches to Word Order Variation in Germanic. Berlin: Mouton de Gruyter, –. Axel, Katrin (b). The consolidation of verb-second in Old High German: What role did subject pronouns play?, in Paola Crisma and Guiseppe Longobardi (eds), Historical Syntax and Linguistic Theory. Oxford: Oxford University Press, –. Axel, Katrin, and Helmut Weiß (). What changed where? A plea for the re-evaluation of dialectal evidence, in Anne Breitbarth, Christopher Lucas, Sheila Watts, and David Willis (eds), Continuity and Change in Grammar. Amsterdam: John Benjamins, –. Axel, Katrin, and Helmut Weiß (). Pro-drop in the history of German: From Old High German to the modern dialects, in Peter Gallmann and Melanie Wratil (eds), Empty Pronouns. Berlin: Mouton de Gruyter, –. Axel, Katrin, and Angelika Wöllstein (). German verb-first conditionals as unintegrated clauses: A case study in converging synchronic and diachronic evidence, in Susanne Winkler and Sam Featherston (eds), The Fruits of Empirical Linguistics, Vol. II: Product. Berlin: de Gruyter, –. Axel-Tober, Katrin (). (Nicht-)kanonische Nebensätze im Deutschen: synchrone und diachrone Aspekte. Berlin: Walter de Gruyter. Axel-Tober, Katrin (). Historische Syntax des Deutschen aus generativer Perspektive, Zeitschrift für germanistische Linguistik (): –. Bader, Markus, and Tanja Schmid (). Verb clusters in colloquial German, Journal of Comparative Germanic Linguistics : –. Baker, Mark C. (). The Polysynthesis Parameter. Oxford: Oxford University Press. Baker, Mark C. (). The Atoms of Language. The Mind’s Hidden Rules of Grammar. New York: Basic Books. Barðal, Jóhanna (). Case in Icelandic: A Synchronic, Diachronic, and Comparative Approach. Doctoral dissertation, Lund University. Barss, Andrew, and Howard Lasnik (). A note on anaphora and double objects, Linguistic Inquiry : –. Bayer, Josef (). COMP in Bavarian syntax, The Linguistic Review : –. Bayer, Josef (). A note on alleged PRO-government, in Dany Jaspers, Yvan Putseys, Wim Klooster, and Pieter Seuren (eds), Sentential Complementation and the Lexicon. Studies in Honour of Wim de Geest. Dordrecht: Foris, –. Bayer, Josef (). Final complementizers in hybrid languages, Journal of Linguistics : –. Bayer, Josef (). Nominal negative quantifiers as adjuncts, Journal of Comparative Germanic Linguistics : –. Bean, Marian C. (). The Development of Word Order Patterns in Old English. Totowa: Barnes and Noble. Bech, Gunnar (). Studien über das deutsche verbum infinitum. Vol. I. Copenhagen: Munksgaard. Bech, Kristin (). Word Order Patterns in Old and Middle English: A Syntactic and Pragmatic Study. Doctoral dissertation, University of Bergen.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Bech, Kristin, and Christine Meklenborg Salvesen (). Preverbal word order in Old English and Old French, in Kristin Bech and Kristine Gunn Eide (eds), Information Structure and Syntactic Change in Germanic and Romance Languages. Amsterdam: John Benjamins, –. Behaghel, Otto (/). Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern, Indogermanischen Forschungen : –. Behaghel, Otto (). Die Verneinung in den deutschen Sprachen, Wiss. Beihefte zur Zeitschr. des allg. dt. Sprachvereins : –. Behaghel, Otto (–). Deutsche Syntax. Eine geschichtliche Darstellung. Vols. Heidelberg: Carl Winter. Belletti, Adriana (). Aspects of the low IP area, in Luigi Rizzi (ed), The Structure of CP and IP: The Cartography of Syntactic Structure, Vol. II. Oxford: Oxford University Press, –. Benincà, Paola (). A detailed map of the left periphery of medieval Romance, in Raffaella Zanuttini, Héctor Campos, Elena Herburger, and Paul Portner (eds), Negation, Tense and Clausal Architecture: Cross-linguistic Investigations. Washington: Georgetown University Press, –. van Bergen, Linda (). Pronouns and Word Order in Old English. With Particular Reference to the Pronoun ‘man’. London: Routledge. Besch, Werner, Anne Betten, Oskar Reichmann, and Stefan Sonderegger (eds) (–). Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. nd ed. Vols. Berlin: Mouton de Gruyter. den Besten, Hans (). On the interaction of root transformations and lexical deletive rules, in Werner Abraham (ed), On the Formal Syntax of the Westgermania. Amsterdam: John Benjamins, –. den Besten, Hans (a). Decidability in the syntax of verbs of (not necessarily) WestGermanic languages, Groninger Arbeiten zur Germanistischen Linguistik : –. den Besten, Hans (b). Double negation and the genesis of Afrikaans, in Pieter Muysken and Norval Smith (eds), Substrata versus Universals in Creole Genesis. Amsterdam: John Benjamins, –. den Besten, Hans (). Studies in West Germanic Syntax. Doctoral dissertation, University of Tilburg. Betten, Anne (). Grundzüge der Prosasyntax. Stilprägende Entwicklungen vom Althochdeutschen zum Neuhochdeutschen. Tübingen: Niemeyer. Bianchi, Valentina (). Some issues in the syntax of relative determiners, in Artemis Alexiadou, Paul Law, André Meinunger, and Chris Wilder (eds), The Syntax of Relative Clauses. Amsterdam: John Benjamins, –. Biberauer, Theresa, and Ian Roberts (). Changing EPP parameters in the history of English: Accounting for variation and change, English Language and Linguistics : –. Biberauer, Theresa, and Michelle Sheehan (eds) (). Theoretical Approaches to Disharmonic Word Order. Oxford: Oxford University Press. Biberauer, Theresa, Glenda Newton, and Michelle Sheehan (). The final-over-final constraint and predictions for diachronic change, Toronto Working Papers in Linguistics : –. Biberauer, Theresa, Anders Holmberg, and Ian Roberts (). A syntactic universal and its consequences, Linguistic Inquiry : –. Biener, Clemens (). Die Doppelumschreibung der Praeteritopraesentia, Zeitschrift für deutsche Philologie : –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Birkmann, Thomas (). Präteritopräsentia. Morphologische Entwicklung einer Sonderklasse in den altgermanischen Sprachen. Tübingen: Niemeyer. Birner, Betty J., and Gregory Ward (). Information Status and Noncanonical Word Order in English. Amsterdam: John Benjamins. BMZ = Benecke, Friedrich, Wilhelm Müller, and Friedrich Zarncke (). Mittelhochdeutsches Wörterbuch. Mit Benutzung des Nachlasses von Georg Friedrich Benecke ausgearbeitet von Wilhelm Müller und Friedrich Zarncke. Vols. I–III. Stuttgart: Hirzel. [Reprint of the edition Leipzig –]. Boersma, Paul, and Bruce Hayes (). Empirical tests of the gradual learning mechanism, Linguistic Inquiry : –. Bohn, Isabella, and Helmut Weiß (). Komplementiererflexion im Hessischen, in Augustin Speyer and Philipp Rauth (eds), Syntax aus Saarbrücker Sicht —Beiträge der SaRDiSTagung zur Dialektsyntax (Zeitschrift für Dialektologie und Linguistik—Beiheft ). Stuttgart: Franz Steiner, –. Bolinger, Dwight (). The Phrasal Verb in English. Cambridge, MA: Harvard University Press. Bolli, Ernst (). Die verbale Klammer bei Notker. Berlin: de Gruyter. de Boor, Helmut (). Untersuchungen zur Sprachbehandlung Otfrids. Hiatus und Synaloephe (Germanistische Abhandlungen ). Breslau: Marcus. de Boor, Helmut, and Roswitha Wisniewski (). Mittelhochdeutsche Grammatik, th ed. Berlin: Mouton de Gruyter. Borer, Hagit (). Parametric Syntax. Dordrecht: Foris. Borer, Hagit (). I-Subjects, Linguistic Inquiry : –. Borter, Alfred (). Syntaktische Klammerbildung in Notkers Psalter. Berlin: de Gruyter. Bošković, Željko (). On multiple wh-fronting, Linguistic Inquiry : –. Braune, Wilhelm (). Zur Lehre von der deutschen Wortstellung, in Rudolf Kögel (ed), Forschungen zur deutschen Philologie. Festgabe für Rudolf Hildebrand zum . März . Leipzig: Veit & Co, –. Braune, Wilhelm (). Althochdeutsche Grammatik I. Laut- und Formenlehre. Bearbeitet von Ingo Reiffenstein. th ed. Tübingen: Niemeyer. Breindl, Eva (). Die Brigitte nun kann der Hans nicht ausstehen. Gebundene Topiks im Deutschen, Deutsche Sprache (): –. Breitbarth, Anne (). Live Fast, Die Young—the Short Life of Early Modern German Auxiliary Ellipsis (LOT Dissertation Series ). Utrecht: LOT. Breitbarth, Anne (). A hybrid approach to Jespersen’s cycle in West Germanic, Journal of Comparative Germanic Linguistics : –. Breitbarth, Anne (). The History of Low German Negation. Oxford: Oxford University Press. Breitbarth, Anne (a). Exceptive negation in Middle Low German, in Ellen Brandner, Anna Czypionka, Constantin Freitag, and Andreas Trotzke (eds), Webschrift for Josef Bayer— Charting the Landscape of Linguistics. URL: http://ling.uni-konstanz.de/pages/WebschriftBayer/ /title.html [accessed May ]. Breitbarth, Anne (b). The development of conditional should in English, in Chiara Gianollo, Agnes Jäger, and Doris Penka (eds), Language Change at the Syntax–Semantics Interface. Berlin: Mouton de Gruyter, –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Breitbarth, Anne (c). Dialect contact and the speed of Jespersen’s cycle in Middle Low German, Taal en tongval : –. Breitbarth, Anne (). Jespersen’s cycle = minimize structure + feature economy, in Silvio Cruschina, Katharina Hartmann, and Eva-Maria Remberger (eds), Studies on Negation: Syntax, Semantics, and Variation. Göttingen: V&R unipress/Vienna University Press, –. Breitbarth, Anne, Christopher Lucas, and David Willis (). Incipient Jespersen’s cycle: The (non-)grammaticalization of new negative markers, in Jürg Fleischer and Horst Simon (eds), Sprachwandelvergleich—Comparing Diachronies. Berlin: de Gruyter, –. Büring, Daniel (). Mittelfeldreport V, in Brigitta Haftka (ed), Was determiniert Wortstellungsvariation? Studien zu einem Interaktionsfeld von Grammatik, Pragmatik und Sprachtypologie. Opladen: Westdeutscher Verlag, –. Büring, Daniel (). What do definites do that indefinites definitely don’t?, in Caroline Féry and Wolfgang Sternefeld (eds), Audiator vox sapientiae. A Festschrift for Arnim von Stechow. Berlin: Akademie Verlag, –. Büring, Daniel, and Katharina Hartmann (). Semantic coordination without syntactic coordination, in Ida Toivonen, Piroska Csuri, and Emile van der Zee (eds), Structures in the Mind: Essays on Language, Music, and Cognition in Honor of Ray Jackendoff. Cambridge, MA: MIT Press, –. Bybee, Joan L., William Pagliuca, and Revere Perkins (). On the asymmetries in the affixation of grammatical material, in William Croft, Keith Denning, and Suzanne Kemmer (eds), Studies in Typology and Diachrony. Amsterdam: John Benjamins, –. Cardinaletti, Anna, and Michal Starke (). Deficient pronouns: A view from Germanic. A study in the unified description of Germanic and Romance, in Höskuldur Thráinsson, Samuel D. Epstein, and Steve Peter (eds), Studies in Comparative Germanic Syntax, Vol II. Dordrecht: Kluwer, –. Cardinaletti, Anna, and Michal Starke (). The typology of structural deficiency: On the three grammatical classes, in Henk van Riemsdijk (ed), Clitics in the Languages of Europe. Berlin: Mouton de Gruyter, –. Carstens, Vicki (). Rethinking complementizer agreement: Agree with a case-checked goal, Linguistic Inquiry : –. Catalani, Luigi (). Die Negation im Mittelfranzösischen. Frankfurt a. M.: Peter Lang. Catasso, Nicholas (). On post-initial aber and other syntactic transgressions: Some considerations on the nature of V in German, Journal of Germanic Linguistics : –. Cecchetto, Carlo, and Renato Oniga (). Consequences of the analysis of Latin infinitival clauses for the theory of case and control, Lingue e Linguaggio : –. Chafe, Wallace L. (). Givenness, contrastiveness, definiteness, subjects, topics, and point of view. In Charles N. Li (ed), Subject and Topic. New York: Academic Press, –. Chomsky, Noam (). Barriers. Cambridge, MA: MIT Press. Chomsky, Noam (). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam (). Minimalist inquiries: The framework, in Roger Martin, David Michaels, and Juan Uriagereka (eds), Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, MA: MIT Press, –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Chomsky, Noam (). On phases, in Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizarreta (eds), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud. Cambridge, MA: MIT Press, –. Cichosz, Anna (). The Influence of Text Type on Word Order of Old Germanic Languages: A Corpus-based Contrastive Study of Old English and Old High German. Frankfurt a. M.: Peter Lang. Cinque, Guglielmo (). Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: Oxford University Press. Clark, Robin, and Ian Roberts (). A computational model of language learnability and language change, Linguistic Inquiry : –. Clément, Daniele (). Satzeinbettungen nach Verben der Sinneswahrnehmung im Deutschen, in Dieter Wunderlich (ed), Probleme und Fortschritte der Transformationsgrammatik. München: Hueber, –. Coleman, Robert (). The Indo-European origins and Latin development of the accusative with infinitive construction, in Christian Touratier (ed), Syntaxe et Latin. Aix-en-Provence: Université de Provence, –. Coniglio, Marco (). Die Syntax der deutschen Modalpartikel: Ihre Distribution und Lizensierung in Haupt- und Nebensätzen. Berlin: Akademie Verlag. Coniglio, Marco (). On V declarative clauses in Middle High German, Linguistische Berichte : –. Coniglio, Marco, Sonja Linde, and Tom Ruette (). Relative clauses in Old High German: A corpus-based statistical investigation of their syntactic and information-structural properties, Journal of Germanic Linguistics (): –. Corbett, Greville G. (). Agreement. Cambridge: Cambridge University Press. Corver, Norbert, and Denis Delfitto (). On the nature of pronoun movement, in Henk van Riemsdijk (ed), Clitics in the Languages of Europe. Berlin: Mouton de Gruyter, –. Coupé, Griet, and Ans van Kemenade (). The origin and rise of the IPP effect in Dutch dialects. Paper presented at the th Diachronic Generative Syntax (DIGS), University of Trieste. URL: http://repository.ubn.ru.nl/handle// [accessed May ]. Craenenbroeck, Jeroen, and Marjo Koppen (). Pronominal doubling in Dutch dialects: Big DPs and coordinations, in Sjef Barbiers, Olaf Koeneman, and Marika Lekakou (eds), Microvariation in Syntactic Doubling. Bingley: Emerald, –. Dahl, Östen (). Typology of sentence negation, Linguistics : –. Dal, Ingerid (). Kurze deutsche Syntax auf historischer Grundlage. rd ed. Tübingen: Niemeyer. Dal, Ingerid (). Der Infinitiv mit dem syntaktischen Wert eines Participium Praeteriti, in Ingerid Dal, Untersuchungen zur germanischen und deutschen Sprachgeschichte. Oslo: Universitätsverlag, –. [Norwegian version first published ]. Dehé, Nicole (). Particle Verbs in English. Syntax, Information Structure, and Intonation. Amsterdam: John Benjamins. Delbrück, Berthold (). Altindische Syntax. Halle/S.: Verlag der Buchhandlung des Waisenhauses. [Reprint: Darmstadt: Wissenschaftliche Buchgesellschaft, ]. Delbrück, Berthold (). Germanische Syntax. Bd. : Zur Stellung des Verbums (Abhandlungen der Königlich Sächsischen Gesellschaft der Wissenschaften , ). Leipzig: Teubner. Delbrück, Berthold (). Beiträge zur germanischen Syntax IV: Die Stellung des Verbums in Sätzen mit ‘doch’ und ‘ja’, Beiträge zur Geschichte der deutschen Sprache und Literatur : –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Delsing, Lars-Olof (). Diakron syntax i ljuset av generativ grammatik, Nordlyd : –. Demske, Ulrike (). Merkmale und Relationen. Diachrone Studien zur Nominalphrase des Deutschen. Berlin: de Gruyter. Demske-Neumann, Ulrike (). Modales Passiv und Tough Movement. Zur strukturellen Kausalität eines syntaktischen Wandels im Deutschen und Englischen. Tübingen: Niemeyer. Dentschewa, Emilia (). Zur sprachlichen Eigenständigkeit der althochdeutschen TatianÜbersetzung in Bezug auf den Gebrauch des Infinitivs, Berichte zur Erforschung der deutschen Sprache : –. Déprez, Viviane (). The roots of negative concord in French and French-lexicon creoles, in Michel DeGraff (ed), Language Creation and Language Change. Cambridge, MA: MIT Press, –. Déprez, Viviane (). Parallel (a)symmetries and the internal structure of negative expressions, Natural Language and Linguistic Theory : –. Devine, Andrew M., and Laurence D. Stephens (). Latin Word Order: Structured Meaning and Information. Oxford: Oxford University Press. Diels, Paul (). Die Stellung des Verbums in der älteren althochdeutschen Prosa (Palaestra. Untersuchungen und Texte aus der deutschen und englischen Philologie LIX). Berlin: Mayer & Müller. Diercks, Michael (). Indirect agree in Lubukusu complementizer agreements, Natural Language and Linguistic Theory : –. Diesing, Molly (). Yiddish VP order and the typology of object movement in Germanic, Natural Language and Linguistic Theory : –. Diessel, Holger (). Demonstratives: Form, Function, and Grammaticalization. Amsterdam: John Benjamins. den Dikken, Marcel, and Eric Hoekstra (). Parasitic participles, Linguistics : –. Dipper, Stefanie (). Annotierte Korpora für die historische Syntaxforschung: Anwendungsbeispiele anhand des Referenzkorpus Mittelhochdeutsch, Zeitschrift für germanistische Linguistik : –. Dittmer, Arne, and Ernst Dittmer (). Studien zur Wortstellung—Satzgliedstellung in der althochdeutschen Tatianübersetzung (Studien zum Althochdeutschen ). Göttingen: Vanderhoeck & Ruprecht. Donaldson, Bryan (). Initial subordinate clauses in Old French: Syntactic variation and the clausal left periphery, Lingua : –. Donhauser, Karin (). Moderne Kasuskonzeptionen und die Kasussetzung im Althochdeutschen. Überlegungen zur Stellung des Objektsgenitivs im Althochdeutschen, in Anne Betten (ed), Neuere Forschungen zur historischen Syntax des Deutschen. Tübingen: Niemeyer, –. Donhauser, Karin (). Negationssyntax in der deutschen Sprachgeschichte. Grammatikalisierung oder Degrammatikalisierung?, in Ewald Lang and Gisela Zifonun (eds), Deutsch— typologisch. Berlin: Mouton de Gruyter, –. Donhauser, Karin (). Negationssyntax im Althochdeutschen. Ein sprachhistorisches Rätsel und der Weg zu seiner Lösung, in Karin Donhauser and Ludwig M. Eichinger (eds), Deutsche Grammatik. Thema in Variationen. Festschrift für Hans-Werner Eroms. Heidelberg: Carl Winter, –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Donhauser, Karin, and Svetlana Petrova (). Die Rolle des Adverbs tho bei der Generalisierung von Verbzweit im Deutschen, in Monika Dannerer, Peter Mauser, Hannes Scheutz, and Andreas E. Weiss (eds), Gesprochen—geschrieben—gedichtet. Variation und Transformation von Sprache. Berlin: Erich Schmidt Verlag, –. Dowty, David R. (). Thematic proto-roles and argument selection, Language : –. Drach, Erich (). Grundgedanken der deutschen Satzlehre. Frankfurt a. M.: Diesterweg. Drinka, Bridget (). Areal factors in the development of the European periphrastic perfect, Word (): –. Drubig, Hans B. (). Zur Frage der grammatischen Repräsentation thetischer und kategorischer Sätze, in Joachim Jacobs (ed), Informationsstruktur und Grammatik (Linguistische Berichte Sonderheft /–). Opladen: Westdeutscher Verlag, –. Dryer, Matthew (). The Greenbergian word order correlations, Language : –. Dryer, Matthew (). Word order, in Timothy Shopen (ed), Language Typology and Syntactic Description. Vol. I: Clause Structure. nd ed. Cambridge: Cambridge University Press, –. Dryer, Matthew S., and Martin Haspelmath (eds) (). The World Atlas of Language Structures Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. URL: http://wals.info [accessed March ]. Duden = Wöllstein, Angelika/Dudenreaktion (). Duden. Vol. IV: Grammatik. th ed. Berlin: Dudenverlag. Dürscheid, Christa (). Zur Vorfeldbesetzung in deutschen Verbzweit-Strukturen. Trier: Wissenschaftlicher Verlag Trier. Ebert, Robert P. (). Infinitival Complement Constructions in Early New High German. Tübingen: Niemeyer. Ebert, Robert P. (). Historische Syntax des Deutschen. Stuttgart: Metzler. Ebert, Robert P. (). Social and stylistic variation in Early New High German word order: The sentence frame (Satzrahmen), Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB) : –. Ebert, Robert P. (). Social and stylistic variation in the order of auxiliary and non-finite verb in dependent clauses in Early New High German, Beiträge zur Geschichte der deutschen Sprache und Literatur (PBB) : –. Ebert, Robert P. (). Historische Syntax des Deutschen II: –. nd ed. Berlin: Weidler. Ebert, Robert P., Oskar Reichmann, Hans-Joachim Solms, and Klaus-Peter Wegera (). Frühneuhochdeutsche Grammatik. Tübingen: Niemeyer. Ecay, Aaron (). A Multi-step Analysis of the Evolution of English Do-Support. Doctoral dissertation, University of Pennsylvania. Einhorn, E. (). Old French. A Concise Handbook. Cambridge: Cambridge University Press. Eisenberg, Peter (). Grundriss der deutschen Grammatik. Vol. II: Der Satz. th rev. ed. Stuttgart: Metzler Verlag. Eisenberg, Peter, George Smith, and Oliver Teuber (). Ersatzinfinitiv und Oberfeld. Ein großes Rätsel der deutschen Sprache, Deutsche Sprache : –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Emonds, Joseph E. (). Lexicon and Grammar: The English Syntacticon. Berlin: Mouton de Gruyter. Enzinger, Stefan (). Kausative und perzeptive Infinitivkonstruktionen. Berlin: Akademie Verlag. Erben, Johannes (). Grundzüge einer Syntax der Sprache Luthers. Berlin: Akademie Verlag. Erdmann, Oskar (). Grundzüge der deutschen Syntax nach ihrer geschichtlichen Entwicklung. Stuttgart: Verlag der J. G. Cotta’schen Buchhandlung. Erickson, Jon (). Some observations on word order in Old Saxon, in Christa Dürscheid, Karl Heinz Ramers, and Monika Schwarz (eds), Sprache im Fokus. Festschrift für Heinz Vater zum . Geburtstag. Tübingen: Niemeyer, –. Eroms, Hans-Werner (). Die Entwicklung des ‘Ersatzinfinitivs’ im Deutschen, in Michail Kotin, Piotr Krycki, Marek Laskowski, and Tadeusz Zuchewicz (eds), Das Deutsche als Forschungsobjekt und Studienfach. Synchronie—Diachronie—Sprachkontrast— Glottodidaktik. Frankfurt a. M.: Peter Lang, –. Eschenlohr, Stefanie (). Vom Nomen zum Verb: Konversion, Präfigierung und Rückbildung im Deutschen. Hildesheim: Georg Olms. Evers, Arnold E. (). The Transformational Cycle in Dutch and German. Doctoral dissertation, University of Utrecht. Evers, Arnold E. (). Clause union in French and German, Groninger Arbeiten zur Germanistischen Linguistik : –. Evers, Arnold E. (). Verbal clusters and cluster creepers, in Pieter Seuren and Gerard Kempen (eds), Verb Constructions in German and Dutch. Amsterdam: John Benjamins, –. Eythórsson, Thórhallur (). Verbal Syntax in the Early Germanic Languages. Doctoral dissertation, Cornell University. Eythórsson, Thórhallur (). Functional categories, cliticization, and verb movement in the early Germanic languages, in Höskuldur Thráinsson, Samuel David Epstein, and Steve Peter (eds), Studies in Comparative Germanic Syntax II. Dordrecht: Kluwer, –. Eythórsson, Thórhallur (). Negation in C: The syntax of negated verbs in Old Norse, Nordic Journal of Linguistics : –. Faarlund, Jan Terje (). Syntactic Change. Toward a Theory of Historical Syntax. Berlin: Mouton de Gruyter. Fabricius-Hansen, Cathrine (). ‘Wi(e)der’ and ‘again(st)’, in Caroline Féry and Wolfgang Sternefeld (eds), Audiatur vox sapientiae. A Festschrift for Arnim von Stechow. Berlin: Akademie Verlag, –. Fanselow, Gisbert (). Scrambling as NP-movement, in Günter Grewendorf and Wolfgang Sternefeld (eds), Scrambling and Barriers. Amsterdam: John Benjamins, –. Fanselow, Gisbert (). The return of the base generators, Groninger Arbeiten zur Germanistischen Linguistik : –. Fanselow, Gisbert (). Features, θ-roles, and free constituent order, Linguistic Inquiry : –. Fanselow, Gisbert (). Quirky subjects and other specifiers, in Ingrid Kaufmann and Barbara Stiebels (eds), More than Words. Berlin: Akademie Verlag, –.
OUP CORRECTED PROOF – FINAL, 15/2/2018, SPi
References
Fanselow, Gisbert (). Münchhausen-style head movement and the analysis of verbsecond, in Anoop Mahajan (ed.), Syntax at Sunset: Head Movement and Syntactic Theory (UCLA Working Papers in Linguistics ). Los Angeles, CA: University of California, –. Fanselow, Gisbert (). Cyclic phonology–syntax interaction: Movement to first position in German, in Shinichiro Ishihara, Michaela Schmitz, and Anne Schwarz (eds), Interdisciplinary Studies on Information Structure (ISIS) (Working Papers of the SFB , Vol I). Potsdam: Universitätsverlag Potsdam, –. Felser, Claudia (). Verbal Complement Clauses. A Minimalist Study of Direct Perception Constructions. Amsterdam: John Benjamins. Ferraresi, Gisella (). Word Order and Phrase Structure in Gothic. A Comparative Study. Doctoral dissertation, University of Stuttgart. Ferraresi, Gisella (). Word Order and Phrase Structure in Gothic. A comparative Study (Orbis: Supplementa ). Leuven: Peeters. Ferraresi, Gisella (). Wandel in den Kohärenzstrategien des Deutschen: Der Fall der Adverbkonnektoren, Zeitschrift für Dialektologie und Linguistik : –. Ferraresi, Gisella (). Die deiktische Komponente bei Pronominalkonnektoren im Deutschen: Was sie zeigen und was sie nicht mehr zeigen können, in Christiane Maaß and Angela Schrott (eds), Wenn Deiktika nicht zeigen: Deiktische Formen als Satzkonnektoren und Diskursmarker. Deixis, Pragmatik und Grammatikalisierung. Münster: LIT Verlag, –. Ferraresi, Gisella (). Grammatikalisierung. Heidelberg: Carl Winter. Ferraresi, Gisella (). Zur Konjunktion dass in Exklamativsätzen und in Rekursionsstrukturen, in Rosemarie Lühr (ed), Idiosynkrasie—Neue Wege ihrer Beschreibung. Hamburg: Reichert, –. Ferraresi, Gisella (). Adverbkonnektoren und Modalpartikeln. Synchronie—Diachronie— L-Erwerb. Heidelberg: Carl Winter. Ferraresi, Gisella, and Maria Goldbach (). Discourse particles in some ancient IndoEuropean languages, in Maria Kozianka and Rosemarie Lühr (eds), Indogermanistik, Germanistik, Linguistik. Akten der Arbeitstagung der indogermanischen Gesellschaft, Jena – September . Hamburg: Dr Kovacs, –. Ferraresi, Gisella, and Rosemarie Lühr (eds) (). Diachronic Studies on Information Structure: Language Acquisition and Change. Berlin: Mouton de Gruyter. Ferraresi, Gisella, and Helmut Weiß (). >Al die wîle und ich lebe