Syntax over Time: Lexical, Morphological, and Information-Structural Interactions (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated] 9780199687923, 0199687927
This book provides a critical investigation of syntactic change and the factors that influence it. Converging empirical
125 29 7MB
English Pages 432 [439] Year 2015
Cover
Syntax over Time: Lexical, Morphological, andInformation-Structural Interactions
Copyright
Contents
Series Preface
List of Abbreviations
Notes on Contributors
1: Introduction: Changing views of syntactic change
1.1 Introduction
1.2 Syntax and the lexicon
1.3 Syntax and morphology
1.4 Syntax, prosody, and information structure
1.5 Summary
Acknowledgements
Part I: Syntax and the Lexicon
2: Expletive there inWest Germanic
2.1 Introduction
2.2 The behaviour of da in the history of German
2.2.1 Da in Early New High German
2.2.2 Comparing Modern German da
2.2.3 The historical origins of expletive da
2.2.4 Subject gaps and expletives in Germanic
2.2.4.1 Danish der
2.2.4.2 Yiddish subject gaps
2.2.5 Analysing da in historical German
2.3 An expletive in Old English
2.3.1 Old English subject gaps and þær
2.3.2 A change in the distribution of expletive þær
2.3.3 Analysing Old English þær
2.4 Conclusion
3: From passive to active: Stages in the Icelandic New Impersonal
3.1 Introduction
3.2 The New Impersonal: Active or passive?
3.3 Ways in which the New Impersonal differs from the canonical passive
3.3.1 Bound anaphors
3.3.2 Adjuncts
3.3.3 Nonagentive predicates (Unaccusatives)
3.4 Stages in the diachronic development of the New Impersonal
3.5 Conclusion
4: Change in the syntax and semantics of be like quotatives
4.1 Introduction
4.2 Experiment 1: Direct speech and reported thought interpretations of be like quotatives
4.2.1 Method
4.2.2 Results
4.3 Experiment : Be like and verb movement
4.3.1 Method
4.3.2 Results
4.4 The syntax and semantics of be like
4.5 The evolution of be like quotatives
4.6 Conclusion
Acknowledgements
5: The grammaticalization of postpositions in Old Hungarian
5.1 Introduction
5.2 The data: Old Hungarian postpositions
5.2.1 The class of postpositions
5.2.1.1 Possessives and postpositions
5.2.1.2 Suffixes
5.2.2 Variation in Old Hungarian
5.3 AxialParts
5.3.1 The structure of PPs
5.3.2 N > AxialPart > P
5.4 Conclusions
Old Hungarian sources
6: A negative cycle in 12th–15th-century Hungarian
6.1 Introduction
6.2 Background: Negation inModern Hungarian
6.3 The position of the negative particle in Old Hungarian
6.4 Se-expressions in Old Hungarian
6.5 A negative cycle in th–th-century Hungarian
6.6 Summary
Sources
7: Negation and NPI composition inside DP
7.1 Introduction
7.2 ‘Nominal negative inversion’ in European Portuguese
7.2.1 Pronouns vs. full DPs
7.2.2 Negative answers to polar questions
7.2.3 Count vs. mass nouns
7.2.4 Gradable quantifiers
7.3 From Old Portuguese toModern Portuguese
7.4 Contrasting Portuguese with Spanish
7.5 Conclusion, with a brief note on Italian and French
Sources of the data
Part II: Syntax and Morphology
8: Increasing morphological complexity and how syntax drives morphological change
8.1 Introduction
8.2 The syntax–morphology connection in historical grammar change
8.2.1 Erosion and loss of inflectional morphology as a trigger for syntactic change
8.2.2 Morphological complexity arising from fossilized syntax
8.2.3 General principles of exogenous morphological change
8.3 Two diachronic scenarios for the rise of morphological complexity
8.4 Analytic drift and the recovery of verb movement space
8.4.1 Main trends of the analytic drift
8.4.2 Particle movement around a clitic left-dislocated subject DP
8.5 Patterns of morphological innovation and change in Later Egyptian
8.5.1 The Equal Complexity Hypothesis
8.5.2 Case study I:The diachronic rise of relative tenses
8.5.3 Case study II: Emergent modal-evidential categories
8.6 Discourse-configurational syntax as a trigger for morphological renewal
8.7 Concluding remarks
9: Reconstructing complementizer-drop in the dialects of the Salento: A syntactic or phonological phenomenon?
9.1 Introduction
9.2 Cu-drop: traditional approach
9.2.1 Cu-deletion: problems
9.3 Northern Salentino: proposal
9.4 Central-southern Salento: Proposal
9.5 Conclusion
10: On negation, tense, and participles in Finnic and Sámi
10.1 Introduction
10.2 The new pattern
10.3 The old pattern
10.4 The tense of past participles
10.5 The change
10.6 The transitional varieties
10.7 Conclusions
11: On the loss of tense and verb-adjacent clitics in Slavic
11.1 Introduction
11.2 Properties of second position and verb-adjacent clitics
11.2.1 Second position cliticization
11.2.2 Verb-adjacent pronominal cliticization
11.3 Pronominal clitics in Old Slavic
11.4 Tense and aspect in Old Slavic
11.5 Towards an analysis
11.6 Summary
12: The evolution of inherent Case in the diachrony of Greek
12.1 Introduction
12.2 Background assumptions
12.3 Dative arguments in the diachrony of Greek
12.3.1 From Classical Greek to Koine
12.3.2 Medieval CG
12.3.2.1 Competing substitutes and the emergence of dative alternations
12.3.2.2 Coexisting/competing inherent [Case] features
(a) Survival of [iCase]
(b) Reanalysis and emergence of active inherent (u)Case
12.3.2.3 Modern CG
12.4 Conclusions
Part III: Syntax, Prosody, and Information Structure
13: From preposition to topic marker: Old Romanian pe
13.1 Introduction
13.2 Research context
13.3 Pe-DOMis the outcome of grammaticalization
13.4 pe-DOMversus Case marking
13.5 Discourse-driven grammaticalization
13.6 Formalization of pe-DOMas CT
13.7 CD without and with DOM
13.7.1 CD without DOM
13.7.2 CD+DOM
13.8 Conclusions
14: Verb-third in earlyWest Germanic: A comparative perspective
14.1 Introduction
14.2 The structure of the left periphery in Old English
14.3 The structure of the left periphery in Old High German
14.4 The structure of the left periphery in Old Saxon
14.5 Main clauses in Proto-West Germanic: A diachronic scenario
14.6 Conclusion
Primary sources
15: Changes in Friulano subject clitics: Conflation and interactions with the left periphery
15.1 Introduction
15.2 The dialects and corpus
15.3 Properties of Friulano SCLs
15.4 The evolution of Friulano SCLs
15.5 Subject doubling in Friulano
15.5.1 An implicational hierarchy of subject doubling
15.5.2 Conflation as progression along the hierarchy
15.5.3 Extension to wh-doubling and mandatory doubling
15.6 Non-co-occurrence of al and other clitics
15.7 Conclusions
16: The decline of Latin left-peripheral presentational foci: Causes and consequences
16.1 Introduction
16.1.1 ‘Left Edge Fronting’:The phenomenon
16.1.2 The position of subordinating conjunctions
16.2 Interpretation and diachrony of LEF
16.2.1 LEF as left-peripheral presentational focalization
16.2.1.1 The syntax of presentational foci
16.2.1.2 On the interpretation of LEF
16.2.1.3 A diachronic hypothesis
16.3 The derivation of LEF: vP movement and ‘smuggling’
16.3.1 Discourse-neutral word order in Latin
16.3.2 vP movement as a way to satisfy the EPP
16.3.3 vP movement in Latin
16.3.4 Smuggling: a way to avoid intervention
16.4 The loss of vP movement and its consequences
16.4.1 Reanalysis of vP movement
16.4.2 Postverbal presentational foci
16.5 Conclusion
17: Weak focus and polarity: Asymmetries between Spanish and Catalan
17.1 Introduction
17.2 Weak focus in Old Spanish and Old Catalan
17.2.1 Old Spanish
17.2.2 Old Catalan
17.2.3 Further properties of weak focus
17.3 Weak focus inModern Spanish andModern Catalan
17.3.1 A preliminary overview of the data
17.3.2 Properties of weak focus in Modern Spanish and Modern Catalan: A comparative view
17.4 WFF and polarity: Towards an explanation for the asymmetry between Modern Spanish and Modern Catalan
17.5 Conclusion
Sources
Acknowledgements
18: An interface account of word-order variation in Old High German
18.1 Introduction
18.1.1 Word-order variation in older Germanic
18.1.2 Outline of the chapter
18.2 Word-order variation and the Head Complement parameter
18.2.1 OV grammar plus extraposition
18.2.2 An account based on the Universal Base hypothesis
18.2.3 An approach in terms of variable spell-out
18.2.4 Mixed word orders and stylistic preferences
18.3 Prosodic and information-structural constraints in OHG
18.3.1 On the interaction between IS and prosody in OHG
18.3.2 On the nature of the prosodic factor
18.4 On the mapping between syntactic and prosodic structure
18.4.1 Prosodic domain formation in a phase-based approach
18.4.2 Focus, prominence, and rules of accent placement
18.5 Conclusions
19: Verb order, object position, and information status in Old English
19.1 Introduction
19.2 Data and terminology
19.3 The distribution of objects in VAux and AuxV clauses
19.3.1 Factors affecting object position
19.3.1.1 Information status
19.3.1.2 Case
19.3.1.3 Grammatical weight
19.3.1.4 Strength of effect
19.3.2 Non-referential objects
19.3.2.1 Negative objects
19.3.2.2 Semantically incorporated objects
19.3.3 Summary
19.4 Analysis
19.4.1 Change in progress
19.4.2 Two postverbal object positions
19.5 Conclusions
20: Antisymmetry and Heavy NP Shift across Germanic
20.1 Introduction: Towards a unified theory of Heavy NP Shift
20.2 Towards a unified theory of Heavy NP Shift
20.3 The fine-grained information structure of HNPS
20.4 Diachronic predictions
20.5 The problem of vP-internal HNPS
20.6 Conclusions and directions for future research
21: Pronominal object shift in Archaic Chinese
21.1 Introduction
21.2 Previous analyses
21.3 Syntactic movement for case
21.4 Direct evidence: Case and nominalization
21.4.1 Morphological case
21.4.2 Nominalization
21.5 Indirect evidence: Availability of structural case obviates fronting
21.5.1 Aspect and fronting
21.5.2 Complex VPs
21.6 Evidence for object shift
21.7 Loss of pronoun fronting
References
Index of Languages
Index of Subjects
Recommend Papers
![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)
![Micro-change and Macro-change in Diachronic Syntax (Oxford Studies in Diachronic and Historical Linguistics) [1 ed.]
9780198747840, 0198747845](https://ebin.pub/img/200x200/micro-change-and-macro-change-in-diachronic-syntax-oxford-studies-in-diachronic-and-historical-linguistics-1nbsped-9780198747840-0198747845.jpg)
![The Evolution of Functional Left Peripheries in Hungarian Syntax (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198709855, 0198709854](https://ebin.pub/img/200x200/the-evolution-of-functional-left-peripheries-in-hungarian-syntax-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198709855-0198709854.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)
![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)
![Indefinites between Latin and Romance (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780198812661, 0198812663](https://ebin.pub/img/200x200/indefinites-between-latin-and-romance-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780198812661-0198812663.jpg)
![Syntax over Time: Lexical, Morphological, and Information-Structural Interactions (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
9780199687923, 0199687927](https://ebin.pub/img/200x200/syntax-over-time-lexical-morphological-and-information-structural-interactions-oxford-studies-in-diachronic-and-historical-linguistics-illustrated-9780199687923-0199687927.jpg)
- Author / Uploaded
- Theresa Biberauer (editor)
- George Walkden (editor)
File loading please wait...
Citation preview
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Syntax over Time
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OX F O R D S T U D I E S I N D IAC H R O N IC A N D 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; David Willis, University of Cambridge recently published in the series 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent 9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro 11 The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by Katalin É. Kiss 12 Syntactic Reconstruction and Proto-Germanic George Walkden 13 The History of Low German Negation Anne Breitbarth 14 Arabic Indefinites, Interrogatives, and Negators A Linguistic History of Western Dialects David Wilmsen 15 Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden For a complete list of books published and in preparation for the series, see pp.419–20.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Syntax over Time Lexical, Morphological, and Information-Structural Interactions
Edited by T H E R E S A B I B E R AU E R A N D G E O R G E WA L K D E N
1
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
3
Great Clarendon Street, Oxford, ox2 6dp, 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 Theresa Biberauer and George Walkden 2015 © the chapters their several authors 2015 The moral rights of the authors have been asserted First Edition published in 2015 Impression: 1 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 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2014941343 ISBN 978–0–19–968792–3 Printed and bound by CPI Group (UK) Ltd, Croydon, cr0 4yy 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, 7/2/2015, SPi
Contents Series Preface List of Abbreviations Notes on Contributors Introduction: Changing views of syntactic change Theresa Biberauer and George Walkden
vii viii xiv
Part I. Syntax and the Lexicon Expletive there in West Germanic Caitlin Light
From passive to active: Stages in the Icelandic New Impersonal Joan Maling and Sigríður Sigurjónsdóttir
Change in the syntax and semantics of be like quotatives William Haddican, Eytan Zweig, and Daniel Ezra Johnson
The grammaticalization of postpositions in Old Hungarian Veronika Hegedűs
A negative cycle in th–th-century Hungarian Katalin É. Kiss
Negation and NPI composition inside DP Ana Maria Martins
Part II. Syntax and Morphology Increasing morphological complexity and how syntax drives morphological change Chris Reintges
Reconstructing complementizer-drop in the dialects of the Salento: A syntactic or phonological phenomenon? Adam Ledgeway
On negation, tense, and participles in Finnic and Sámi Marit Julien
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
vi
Contents
On the loss of tense and verb-adjacent clitics in Slavic Krzysztof Migdalski
The evolution of inherent Case in the diachrony of Greek Dimitris Michelioudakis
Part III. Syntax, Prosody, and Information Structure From preposition to topic marker: Old Romanian pe Virginia Hill
Verb-third in early West Germanic: A comparative perspective George Walkden
Changes in Friulano subject clitics: Conflation and interactions with the left periphery Ed Cormany
The decline of Latin left-peripheral presentational foci: Causes and consequences Lieven Danckaert
Weak focus and polarity: Asymmetries between Spanish and Catalan Montserrat Batllori and Maria-Lluïsa Hernanz
An interface account of word-order variation in Old High German Roland Hinterhölzl
Verb order, object position, and information status in Old English Ann Taylor and Susan Pintzuk
Antisymmetry and Heavy NP Shift across Germanic Joel C. Wallenberg
Pronominal object shift in Archaic Chinese Edith Aldridge
References Index of Languages Index of Subjects
OUP CORRECTED PROOF – FINAL, 7/2/2015, 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 focussing 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, 7/2/2015, SPi
List of Abbreviations 1
first person
2
second person
3
third person
=
cliticized to
ABL
ablative case
ACC
accusative case
ADESS
adessive case
Agr
agreement
AgrOP
object agreement phrase
AgrP
agreement phrase
AIC
active inherent Case
ANT
anterior
APPL
applicative
Aux
auxiliary
AuxV
auxiliary-verb order
BR
Brindisi
C
complementizer
CAUS
causative
CD
clitic doubling
CE
Common Era
CF
contrastive focus
CG
Cypriot Greek
cl
clitic
CLF
classifier
CLLD
clitic left dislocation
COMP
complementizer
COMPARAT
comparative
COND
conditional
CONJ
conjunction
ContrP
contrastive topic phrase
CP
complementizer phrase
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
List of Abbreviations D
determiner
DAT
dative case
DBH
Double Base Hypothesis
DEF
definite
DEM
demonstrative
df
degrees of freedom
DIE
defective intervention effect
DiGS
Diachronic Generative Syntax
DIM
diminutive
DO
direct object
DOC
double object construction
DOM
Differential Object Marking (marker)
DP
determiner phrase
DU
dual
E
predicate event
EME
Early Modern English
EMPH
emphasis
ENHG
Early New High German
EP
European Portuguese
EPP
Extended Projection Principle
ESS
essive case
EXCL
exclamation
EXPL
expletive
F
focus
f.
feminine
FamP
familiar topic phrase
FF
focus fronting
Fin
finiteness
FinP
finiteness phrase
Foc
focus
FocP
focus phrase (left periphery)
FocvP
focus phrase (low left periphery, presentational focus)
FOFC
Final-over-Final Constraint
ForceP
force phrase
FP
functional projection
ix
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
x
List of Abbreviations
FUT
future
G
given information
GEN
genitive case
H-C
Head-Complement
HAB
habitual
HFF
Head-Final Filter
HNPS
Heavy NP Shift
ILL
illative case
IMP
imperative
IND
indicative
INDEF
indefinite
INE
inessive case
INF
infinitive
INFL
inflection
INS
instrumental case
IO
indirect object
IP
inflection phrase
IPFV
imperfective
IS
information structure
It.
Italian
LAC
Late Archaic Chinese
LE
Lecce
LEF
Left Edge Fronting
LF
Logical Form
LOC
locative
M
mean
m.
masculine
ME
Middle English
MedCG
Medieval Cypriot Greek
MHG
Middle High German
MR
Modern Romanian
Neg
negation
NegP
negation phrase
neut.
neuter
NID
Northern Italian Dialect
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
List of Abbreviations NMLZ
nominalizer
NOM
nominative case
NP
noun phrase
NPI
negative polarity item
Num
number
NumP
number phrase
O
object
ObjF
object focus
OCL
object clitic
OE
Old English
OHG
Old High German
OI
Old Icelandic
OR
Old Romanian
OS
Old Saxon
OT
Optimality Theory
OV
object-verb
OVS
object-verb-subject
P
preposition
PART
partitive
PASS
passive
PAST
past tense
PCL
particle
PF
Phonological Form
PFV
perfective
PFX
prefix
PIC
Phase Impenetrability Condition
PL
plural
PlP
plural phrase
PN
pronoun
Pol
polarity
PolP
polarity phrase
POSS
possessed
PP
prepositional phrase
PPI
positive polarity item
PRES
present tense
xi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
xii
List of Abbreviations
pro-arb
arbitrary pronoun
PROG
progressive
Prop.
proportion
PRT
verbal particle
PTCP
past/passive participle
Q
question particle
QP
quantifier phrase
R
reference time
REFL
reflexive
REL
relative particle
RF
raddoppiamento fonosintattico (‘phonosyntactic doubling’)
Rvík
Reykjavík
S
subject
S
speech event
SAGR
subject agreement
SBJV
subjunctive
SCL
subject clitic
SD
standard deviation
SG
singular
ShiftP
shifting/aboutness topic phrase
SMG
Standard Modern Greek
Spec
specifier
Sub
subordinating conjunction
SubjPron
subject pronoun
SUBL
sublative case
SUP
superessive case
SVO
subject-verb-object
SVOI
subject, non-finite verb, object, finite verb order
T
tense
T
(Old High German) Tatian
TAM
tense/aspect/mood
Top
topic
TopP
topic phrase
TP
tense phrase
TR
Taranto
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
List of Abbreviations UG
Universal Grammar
v
light verb
V
verb
V1
verb-initial
V2
verb-second
V3
verb-third
VAux
verb-auxiliary order
VFF
verum focus fronting
Vfin
finite verb
VO
verb-object
VP
verb phrase
vP
light verb/voice phrase
VSO
verb-subject-object
WFF
weak focus fronting
X∗
any number of complements
XP
phrasal constituent (any)
YCOE
York-Toronto-Helsinki Parsed Corpus of Old English Prose
P
sigma (polarity) phrase
xiii
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Notes on Contributors Edith Aldridge is an Associate Professor in the Department of Linguistics at the University of Washington. Her research focuses on comparative and diachronic syntax of Chinese, Japanese, and Austronesian languages. She has recent publications on topics such as applicative constructions in Archaic Chinese, ergativity in Tagalog, and the pseudo-Chinese writing system used to record Old Japanese known as hentaikanbun. Montserrat Batllori received her PhD in Spanish Philology from the Universitat Autònoma de Barcelona in 1996. She has been a Senior Lecturer at the University of Girona since 1997 and Professor of Spanish diachronic grammar since 1990. She has also supervised Masters theses in Catalan diachronic grammar. Her research interests range from historical, descriptive, and comparative studies to theoretical linguistics. She is currently working on the synchronic and diachronic syntax of n-words and word order in Catalan and Spanish, on the one hand, and parasynthesis in Old Spanish, on the other. Theresa Biberauer is a Principal Research Associate at the University of Cambridge, where she is also a Fellow of Churchill College, and Associate Professor Extraordinary at her South African alma mater, Stellenbosch University. Her research interests are principally in theoretical and comparative (synchronic and diachronic) morphosyntax, with Germanic generally and Afrikaans in particular being areas of specific interest. Her past work has focused on word-order variation, (null) subject phenomena, negation, information structure, and the larger question of the nature of parametric variation. Ed Cormany has recently completed his PhD on the morphosyntax, semantics, and pragmatics of imperative clauses in the Department of Linguistics at Cornell University. His research interests more generally include the universality and variability of clause typing crosslinguistically. He is currently investigating the regularity of ‘exceptional’ imperative phenomena and the diachronic development of clause-marking systems. Lieven Danckaert received his doctoral degree from Ghent University, where he is currently working as a postdoctoral researcher. His main expertise is in Latin syntax, with a special interest in word order. His current research focuses on the structure and diachronic evolution of the Latin verb phrase. He is the author of the monograph Latin Embedded Clauses: the Left Periphery (Benjamins 2012). Bill Haddican is an Assistant Professor in the Department of Linguistics and Communications Disorders at Queens College-CUNY. His work focuses on language variation and change and syntax in dialects of Basque and English. Veronika Hegeds is a Junior Research Fellow at the Research Institute for Linguistics of the Hungarian Academy of Sciences. Her research interests include (secondary) predication, complex predicates, the syntax of particles and adpositions, and diachronic syntax.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Notes on Contributors
xv
Maria-Lluïsa Hernanz is Full Professor of Spanish Linguistics at the Universitat Autònoma de Barcelona. Her main areas of research are Spanish grammar and the comparative syntax of the Romance Languages. She has focused on a number of topics including infinitives, small clauses, adjuncts, exclamatives, and focus. Together with M. Batllori, she has also worked on the syntax of Spanish and Catalan polarity particles from a diachronic perspective. She has further published two books and has contributed to the Gramática Descriptiva de la Lengua Española (edited by Bosque and Demonte), and to the Gramàtica del Català Contemporani (edited by Solà et al.). Virginia Hill is Professor of Linguistics at the University of New Brunswick – Saint John. She has also been an adjunct Research Professor at Concordia University and a Leverhulme Visiting Professor at the University of Kent. Her research interests concern the syntax– pragmatics interface (the syntax of particles of direct address and vocatives), and comparative and diachronic syntax, with a focus on Romanian, Romance, and Balkan languages. Roland Hinterhölzl is Professor of German Linguistics at the University of Venice, Ca’ Foscari. He has done research in comparative syntax and diachronic German syntax. He has published a monograph on scrambling, remnant movement, and restructuring in West Germanic (OUP, 2006) and co-edited (with Svetlana Petrova) a volume on information structure and language change with Mouton de Gruyter (2009). He is currently working on issues of the interfaces between syntax, prosody, and information structure. Daniel Ezra Johnson is a Lecturer in the Department of Linguistics and English Language at Lancaster University. His work focuses on dialectology and language change, the quantitative modelling of variation, and sociolinguistic theory. Marit Julien is Professor of Scandinavian Languages at Lund University, Sweden. Her research interests include the grammar of the Scandinavian and the Sámi languages, in particular syntax and the relation between morphology and syntax, as well as microvariation and dialectal syntax. In Sámi, she has mainly focused on verbal syntax. Katalin É. Kiss is a Research Professor at the Research Institute for Linguistics of the Hungarian Academy of Sciences, and the head of the Doctoral School in Linguistics at Pázmány Péter Catholic University. Her main fields of research are Hungarian syntax, both synchronic and diachronic, and information structure. Adam Ledgeway teaches Romance Linguistics at the University of Cambridge. His research interests include Italian dialectology, comparative history and morphosyntax of the Romance languages, Latin, syntactic theory, and linguistic change. His research is channelled towards bringing together traditional Romance philological scholarship with the insights of syntactic theory. He has published extensively on such topics as alignments, complementation, configurationality, auxiliary selection, grammaticalization, word order, cliticization, clause structure, verb movement, finiteness, and the development of demonstrative systems. Caitlin Light is a Lecturer in English Linguistics at the University of York. Her research interests include variation and change, syntax, and information structure, with a particular focus on the Germanic language family and the history of English. She received her PhD from the University of Pennsylvania in 2012.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
xvi
Notes on Contributors
Joan Maling (PhD MIT 1973) is Professor Emerita of Linguistics at Brandeis University, where she taught from 1972 to 2003. She has published on many aspects of the syntax of Modern Icelandic, especially case, word order, passive, preposition-stranding and long distance reflexives, and on case alternations in Finnish, Korean, and German. In 2009, she received an honorary doctorate from the University of Iceland for her contributions to Icelandic Linguistics. She is currently serving as Director of the Linguistics Program at the National Science Foundation (NSF). Ana Maria Martins is Professor at the Universidade de Lisboa. She has published on different topics concerning Portuguese and Romance syntax, including word order, clitics, negative polarity items, negation, emphatic affirmation, impersonal se, (inflected) infinitives, (hyper-)raising, and deictic locatives. Among her recent publications are ‘Emphatic Polarity in European Portuguese and Beyond’ (published in Lingua) and ‘How Much Syntax is There in Metalinguistic Negation?’ (published in Natural Language and Linguistic Theory). She coordinates the project Syntax-oriented Corpus of Portuguese Dialects (CORDIAL-SIN). Dimitris Michelioudakis holds a PhD from the University of Cambridge and is currently a Postdoctoral Researcher at the University of York. His doctoral research focused on dative arguments in Greek and Romance, from a (micro-)comparative and diachronic perspective, and his research interests also include role-changing constructions, especially passives and impersonals, agreement restrictions, clitic clusters, as well as argument and event structure in nominals and synthetic compounds. Krzysztof Migdalski is an Assistant Professor at the Institute of English Studies, University of Wrocław, Poland. He defended his doctoral dissertation The Syntax of Compound Tenses in Slavic at the Models of Grammar Group, Tilburg University in 2006. He is interested in the comparative syntax of Balkan and Slavic languages and historical linguistics. Susan Pintzuk is Professor of English Language and Linguistics at the University of York. Her research interests include syntactic variation and change, particularly in the history of English and other Germanic languages; statistical models of language change; and the role of information structure in syntactic change. Her work involves the application of statistical methods to historical data and the quantitative investigation of structural variation. Chris Reintges is a Senior Researcher at the Centre National de la Recherche Scientifique (CNRS) at University Paris 7. He received his PhD in Linguistics and Egyptology at the University of Leiden in 1997, with a dissertation entitled Passive Voice in Older Egyptian—A Morphosyntactic Study. Subsequently, he worked extensively on the syntax of Coptic Egyptian, which resulted in an extensive reference grammar. He is currently working on the syntax– morphology interface in a historical syntax of Ancient Egyptian. His research interests are in parametric syntax, minimalism, and morphological theory. Sigríður Sigurjónsdóttir (PhD UCLA 1992) is Professor of Icelandic Linguistics at the University of Iceland in Reykjavík, where she has taught since 1994. Her main research areas are syntax and first language acquisition. She has published on many aspects of Icelandic children’s acquisition of syntax, especially binding, question formation, and verb movement, as well as on long distance reflexives and syntactic change in adult Icelandic.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Notes on Contributors
xvii
Ann Taylor is a Senior Lecturer in English Language and Linguistics at the University of York. She has been involved in the creation of syntactically annotated corpora of Old, Middle, and Early Modern English. Her research focuses on the structural and quantitative analysis of syntactic variation and change in the history of English. George Walkden is a Lecturer in the Department of Linguistics and English Language at the University of Manchester. His research is in historical syntax, and his doctoral dissertation focused on aspects of syntactic reconstruction as applied to the early Germanic languages. He is also Associate Editor of Language, with responsibility for its Historical Syntax section. Joel C. Wallenberg is a Lecturer in the History of English at Newcastle University and one of the Principal Investigators at Newcastle’s Centre for Behaviour and Evolution. His primary research interests are syntactic theory, the empirical study of morphosyntactic change using parsed corpora, and the theory of language change and variation. His most current research centres on mathematical parallels between syntactic change and biological evolution, and on how stable sociolinguistic variation persists over time. Eytan Zweig is a Lecturer in the Department of Language and Linguistic Science at the University of York. His work focuses on the semantics and pragmatics interface, from both theoretical and psycholinguistic perspectives.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change T H E R E S A B I B E R AU E R A N D G E O R G E WA L K D E N
. Introduction There is no such thing as syntactic change, and this is a book about it.1 In some approaches and frameworks, change is seen as integral to the very nature of language: for instance, Bailey’s (1981, 1996) Developmental Linguistics, or Hopper’s (1987) Emergent Grammar. By contrast, generative work on historical linguistics as represented at the Diachronic Generative Syntax (DiGS) conferences has led to a picture in which, in at least three respects, syntactic change does not exist.2 The first of these respects relates to the object of study itself. Within the Chomskyan tradition since Aspects (Chomsky 1965), this object has always been the competence of the individual speaker-hearer, or I-language (Chomsky 1986); proposed largerscale entities, such as Saussure’s (1979 [1916]) langue, or the ‘community grammar’ of Weinreich, Labov, and Herzog (1968), have always been firmly rejected as incoherent or undefinable. But if such entities do not exist, then how can they change? Hence, Lightfoot (1991: 162) takes the position that when we speak about a change in, say, the syntax of French, we are adopting ‘a convenient fiction permitting the statement of certain generalizations’. Hale (1998: 2–3) argues that change should be understood simply as a set of differences between two individual grammars. In the introduction to a previous DiGS proceedings volume, Crisma and Longobardi (2009: 5) attempt 1
The first line of this introduction is a homage to Shapin (: ). Cf. also, from a different perspective, Coseriu (). The Diachronic Generative Syntax (DiGS) conference series, of which the Cambridge instalment marked the th anniversary, shares a set of core principles: an emphasis on rigorous synchronic description, an emphasis on reliable and well-understood data, and scepticism towards independent diachronic processes. A clear introduction to, and manifesto for, the ‘DiGS approach’ can be found in Whitman, Jonas, and Garrett (). In this introduction we focus only on those trends in the field that have made themselves apparent over the last few years and that seem to us to be the most exciting directions of research. 2
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
to understand these relations between grammars and partially formalize them as historically significant relations, or ‘H-relations’.3 However, this relational conception of ‘change’ bears little resemblance to our pretheoretical understanding of the notion. In principle, of course, individual grammars are subject to change, and in this limited sense, syntactic change can be said to exist. However, one of the results of research into language acquisition has been the finding that, after the acquisition period, the possibility for the internalized grammar of an individual to change is severely restricted (see e.g. Clahsen 1991; Meisel 1995, 2011). Examples of syntactic change during the lifespan of individual speakers have been presented: Wagner and Sankoff (2011) present findings indicating that speakers of Montreal French have moved away from the periphrastic aller ‘to go’ + infinitive in favour of the inflectional future tense over a 13-year period, Tagliamonte and D’Arcy (2007) have demonstrated an increase in the frequency of use of be like quotatives in Ontario speakers of English over seven years (see Haddican, Zweig, and Johnson, this volume, on change in the syntactic and semantic components of this construction). This indicates that some plasticity remains; nevertheless, ‘virtually all of the evidence for ongoing linguistic change in adulthood derives from an increase in frequency’ of one of two preexisting synonymous options (Tagliamonte and D’Arcy 2009: 61) rather than from the introduction or loss of a discrete grammatical possibility (see in this connection also the striking findings of Heycock et al. (2013) regarding the ongoing loss of V-to-T movement in modern-day Faroese; as Heycock et al. show, this option is in fact initially produced and accepted by pre-school children, before declining as they get older, with ten-year-olds exhibiting adult production and judgments). So true syntactic change within individual grammars, while a logical possibility, may not exist in practice either. Even assuming that some notion of syntactic change beyond the individual is coherent, another perspective on the non-existence of syntactic change is provided by the Inertial Theory of Longobardi (2001). Under this perspective, it is hypothesized that syntactic change must be ‘a well-motivated consequence of other types of change’ and ‘may only originate as an interface phenomenon’ (2001: 278).4 In other words, when syntactic change occurs, it is not primitive but always parasitic on a change in other linguistic domains, for instance a phonological, morphological, or semantic innovation. This is a specific instantiation of a general principle summarized by Keenan (2002: 149) as ‘Things stay as they are unless acted upon by an outside force or decay’ (see also Keenan 1994, 2009). This idea, though not uncontroversial (see Walkden 2012, and Maling and Sigurjónsdóttir, this volume), has spurred twenty-first-century diachronic 3 Though it is important to recognize that, from an I-language perspective, such relations are just as fictional and inchoate as community grammars, as they depend crucially on the primary corpus available to the learner, itself not easily definable or delimitable. See Walkden (: –) for discussion. 4 Interestingly, Stockwell () makes exactly the same claim a quarter of a century earlier: ‘the notion of “pure syntactic change” . . . implies that there exists a class of formal changes which are not motivated by semantic change or phonetic change. I do not believe that is possible.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
syntacticians to investigate interactions between different linguistic modules, and the study of these interactions is a major focus of this volume. Thirdly and finally, theoretical developments within the Principles and Parameters framework have independently led to the conclusion that syntactic change does not exist. Early Principles and Parameters work hypothesized that there existed a set of syntactic principles that could be parameterized: for instance, Subjacency (stated in Chomsky 1973 as a condition; later reformulated as a principle) states that there exist bounding nodes inhibiting movement, and is parameterized as to which nodes count as bounding nodes in a given language. With the move to Minimalism, Chomsky (1995a: 131) adopted a suggestion originally made by Borer (1984): the syntactic computational component of human language is invariant, with inter-speaker variation existing only within the lexicon, and perhaps only within the functional subset of the lexicon. This proposal, dubbed the Borer-Chomsky Conjecture by Baker (2008: 353), has an immediate conceptual advantage, lucidly stated by Borer (1984: 29): The inventory of inflectional rules and of grammatical formatives in any given language is idiosyncratic and learned on the basis of input data. If all interlanguage variation is attributable to that system, the burden of learning is placed exactly on that component of grammar for which there is strong evidence of learning: the vocabulary and its idiosyncratic properties.
The consolidation of linguistic variation in the lexical component has an obvious consequence for the study of syntax diachronically: if there can be no such thing as lexicon-independent syntactic variation, then there can be no such thing as lexicon-independent syntactic change, and developments previously classed as syntactic changes must be reconceptualized as lexical changes. Interestingly, the same conclusion results if we adopt Chomsky’s more recent argument in favour of the idea that parameters may in fact be exclusively located at PF, the interface leading to the sensory-motor systems. On this view, there is once again no variation internal to the syntactic component, with ‘diversity resulting from complex and highly varied modes of externalization, which, furthermore, are readily susceptible to historical change’ (Berwick and Chomsky 2011). Here, then, ‘syntactic change’ is actually morphophonological. If, as we have suggested here, developments within generative syntactic theory have led to the view that syntactic change does not exist, one might conclude that the theory is worthless for the investigation of diachronic issues, or, at the very least, question the validity of continuing to adopt it. We would argue, however, that the opposite is true, and that the perspectives afforded by these theoretical developments have shed new light on the phenomenon of syntactic change as pretheoretically understood. In the remainder of this introduction we will explain why we think this is the case. Mirroring the structure of the volume, the three following sections outline three main directions in which we feel progress is being made: lexical, morphological, and informationstructural interactions.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
. Syntax and the lexicon Though the adoption of the Borer-Chomsky Conjecture, in which syntactic variation is understood as lexical, may at first glance seem to be a trivial terminological matter, we would argue that it has in fact led to advances in understanding and to a widening of the field’s scope of inquiry. In early DiGS work, the focus was overwhelmingly on macro-level properties of languages; thus the proceedings volume from the first DiGS conference, Battye and Roberts (1995), had clause structure as its theme, and was heavily focused on the verbsecond property, the null subject parameter, the Head Parameter, and related issues. These properties are still the subject of detailed work in the current literature, though the focus has shifted towards the interpretive correlates of these different constructions (see Section 1.4). Complementing these big-picture issues, however, the DiGS literature has increasingly seen investigations of the syntactic properties of individual lexical items and their interactions with the larger system of which they form a part, with authors making existential, rather than universal, claims about the languages under investigation. This development, a natural consequence of the Borer-Chomsky approach to syntactic variation, is paralleled by a renewed focus on microvariation in synchronic syntax along the lines sketched in Kayne (2005a). It also opens the door to a better formal understanding of processes of grammaticalization, which fell largely outside the remit of traditional ‘macrodiachronic’ syntax: foundational works here, arguing that grammaticalization can be understood as change in the featural makeup of lexical items, include Roberts and Roussou (1999, 2003) and van Gelderen (2004, 2011). This focus is a hallmark of twenty-first-century DiGS work, with an entire proceedings volume (Batllori et al. 2005) themed around the relationship between grammaticalization and parametric variation. This ‘microdiachronic syntax’ is also well represented in the present volume by the six chapters that constitute Part 1, Syntax and the Lexicon. Light, for instance, is a treatment of expletives in the West Germanic languages, with particular reference to Early New High German and Old English, and is a fine example of how careful analysis of the properties of individual lexical items can inform bigger-picture issues. The Early New High German da and its Old English cognate þær are both analysed as SpecTP expletives. As a consequence, Light argues that the EPP, as a requirement for subjects to fill SpecTP, was more broadly active across early Germanic than previously thought. The key tool of the DiGS trade has always been careful philological and corpusbased study of historical texts. Fine-grained microdiachronic research, however, necessitates the development of new ways of studying syntactic variation and change. Maling and Sigurjónsdóttir contribute to a lively debate about the Icelandic ‘new’ construction, which has been analysed as both a passive and an impersonal. Their aim is to assess various more nuanced predictions of the two analyses, taking into
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
account how the new construction interacts with different lexical items such as bound anaphors and non-agentive verbs. To this end, they employ the methodology of dialect syntax, basing their conclusions on acceptability judgment questionnaires. The results—demonstrating extensive variability—inform our understanding not only of modern Icelandic, but of syntactic change in general, on the basis that ‘the same mechanisms which operated to produce the large-scale changes of the past may be observed operating in the current changes taking place around us’ (Labov 1972: 161). Haddican, Zweig, and Johnson similarly focus on a present-day construction undergoing change, namely the be like quotative in English. Their chapter supplements earlier corpus-based studies with experimental results, demonstrating that be like, originally stative, is increasingly acceptable for younger speakers with an eventive, quotative function; nevertheless, they argue that the spread of stative and eventive be like is part of a single process of change along the lines of the Constant Rate Hypothesis of Kroch (1989). The analysis of the construction—involving a null lexical item SOMETHING (see Kayne 2007)—furthermore has interesting implications for the interaction of syntax, semantics, and the lexicon. Clearly, both dialect-syntactic and experimental methodology constitute a valuable addition to the diachronic syntactician’s toolbox. Prepositions are of special interest from the point of view of the Borer-Chomsky Conjecture, since they have always occupied an uneasy middle ground between functional and lexical categories. Within Minimalism, the study of the fine structure of prepositions and prepositional phrases has flourished, particularly in work by Svenonius (2008, 2010), applied to the history of English by Waters (2009). Hegedűs extends this approach to the grammaticalization of postpositions from nouns in the history of Hungarian, arguing that here, too, we see a clear case of an ‘adposition cycle’. Using a structure proposed by Svenonius, Hegedűs demonstrates that the grammaticalization process does not proceed directly from noun to postposition but rather through an intermediate category of AxialPart. This chapter, then, offers an excellent illustration of how a theoretical approach that draws fine-grained featural (categorial) distinctions allows us to understand grammaticalization in a more articulated fashion than is usual in more traditional approaches. As pointed out in Roberts (2010), this type of approach may also facilitate novel insight into notoriously challenging synchronic and diachronic questions of gradience and gradualness. Negation has always been the poster-child for formal approaches to cyclical change, and the literature on the subject from the twenty-first century alone is extensive: see Larrivée (2011) and Willis, Breitbarth, and Lucas (2013) for overviews. In the present volume, too, negation is well represented. The chapters by É. Kiss and Martins both seek to understand the emergence of negative indefinites, in Hungarian and Portuguese respectively. The se-indefinites discussed by É. Kiss originally arose through univerbation of a negative particle with a (non-negative) indefinite, but over time their negative semantics were bleached, such that in Modern Hungarian they are
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
required to occur in a Negative Concord configuration with an independent negative particle. É. Kiss argues that word-order changes played a role in this development. The indefinite quantifier algum discussed by Martins, on the other hand, develops into a negative indefinite without overt univerbation, and is argued to be so by virtue of its incorporation (along with the noun) into an abstract DP-internal negative head; the proposal thus has consequences for the debate surrounding the cartography of the nominal domain generally (see Alexiadou, Haegeman, and Stavrou 2009, and Alexiadou 2014 for overview discussion) and the internal structure of negative DPs more specifically (see Déprez 2000, Martins 2000, Poletto 2008, Biberauer and Roberts 2011). It is clear, then, that, far from stifling inquiry into syntactic change, adoption of a lexicocentric perspective on variation has enriched diachronic syntax in terms of both methodology and empirical coverage. We began this section by highlighting the ‘macro’ orientation of the earliest DiGS work, which was very centrally concerned with properties associated with parameters—the verb-second parameter, the null subject parameter, and the head parameter, for example. Interestingly, the term ‘parameter’ is not mentioned in any of the six contributions in this section of the volume. On the one hand, one might interpret this as indicative of the notion having outlived its usefulness (see Newmeyer 2004, Boeckx 2010). Alternatively, this can be seen as a reflection of the way in which the notion has been refined in the context of Chomsky’s (2005) ‘three factors’ approach to language design (see Biberauer 2008, Roberts and Holmberg 2010, Biberauer and Roberts 2014, and the work of the ‘Rethinking Comparative Syntax’ project more generally). On this latter view, all of the phenomena discussed in this section would instantiate lexically specified parameters of different sizes (see again Biberauer and Roberts 2014 on the question of ‘parameter size’).
. Syntax and morphology Exploring the interactions between morphology and syntax is not a new endeavour. Within Principles and Parameters, a series of works have proposed strong (often biconditional) relationships between aspects of morphological paradigms and syntactic parameter settings. For instance, the richness of case morphology has been linked to the possibility of certain word orders or word-order variation, implementing ‘the traditional idea that inflectional morphology and positional constraints are functionally equivalent elements of grammatical structure’ (Kiparsky 1997: 461; see also Weerman 1988, 1997). Similarly, V-to-T movement has been linked to the richness of verbal morphology (Roberts 1985, 1993; Bobaljik and Thráinsson 1998; Rohrbacher 1999; Vikner 1997), as has the availability of null subjects (Taraldsen 1978; Rohrbacher 1999; Müller 2005, 2008; Tamburelli 2006). A full review of this literature can be found in Biberauer and Roberts (2010), who suggest that the properties licensing V-to-T
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
and null subjects should be kept apart, though they retain the connection between morphology and both properties. The diachronic consequences of such deterministic interrelationships between morphology and syntax are obvious: if certain morphological changes occur, then syntactic changes will necessarily follow. Thus, if we assume (following Biberauer and Roberts 2010) that V-to-T movement requires rich tense morphology, then if such morphology is lost then V-to-T movement must also be lost. Since morphological change can often be motivated independently (e.g. through regular sound change, or language contact), this gives us a ready-made causal chain for syntactic developments. A whole DiGS volume (Lightfoot 2002) has previously been dedicated to precisely this issue. This simple picture, in which morphological change triggers syntactic change, is intuitively appealing, but evidence has emerged in recent years that suggests that it is in fact too good to be true. For instance, in modern standard English, indirect objects precede direct objects in neutral declaratives, but in the history of the language this was not always the case; a strong version of the morphology-syntax link would then predict that word-order variability was possible as long as case morphology distinguished accusative and dative. Allen (2001), however, shows that the modern order did not become fixed until the last quarter of the fourteenth century, whereas the dative/accusative distinction had been lost in most dialects by the first quarter of the thirteenth. Unless we want to posit that for a period of 150 years learners entertained morphologically unsupported and thus impossible grammars, then, the deterministic relationship between morphology and syntax must be rethought. Similarly, empirical difficulties with deterministic proposals lead Holmberg (2005: 560) to propose that the correlation between null subjects and rich agreement is ‘due not to a rule or condition of (narrow) syntax . . . but to sentence processing’ (see also Ackema and Neeleman 2007). The five contributions collected in part two of this volume all point towards the conclusion that the traditional hypothesis, in which morphology drives syntax, is at best only partially correct. Reintges, for instance, observes that there exists another logical possibility in a Minimalist framework: syntactic changes may lead to morphological changes, the opposite of the traditional view. He illustrates this ‘exogenous’ morphological change with reference to two developments in Later Egyptian, arguing that the emergence of discourse-configurationality (itself contact-induced) triggers the grammaticalization of new evidential and modal marking and new relative tenses. Another logical possibility is for morphophonologically conditioned alternations to become syntacticized, and this is exactly what Ledgeway argues has happened in the dialects of the Salento region in Italy. In these dialects, the complementizer cu can be deleted under certain conditions. This deletion appears to be ordered after raddoppiamento fonosintattico, a process which lengthens the first consonant of a following word, suggesting that deletion is a purely phonological process. Ledgeway
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
argues that this is correct for southern Salentino, but not for the northern dialects, for which he adduces evidence to show that raddoppiamento fonosintattico has been reanalysed as an irrealis marker. This sort of syntacticization of an originally phonological process is consistent with what has been reported in the literature on the life cycle of phonological processes (see, for example, Bermúdez-Otero and Trousdale 2012). Julien looks at a case in which a morphological form takes on a new function. In some Uralic languages, a past participle can be used in the negated past without giving rise to a perfective reading. Julien argues that the introduction of a new copula was central to this development as it led to surface forms which a new generation of acquirers could analyse as including an overt realization requirement on finiteness. Confronted with conservative, copula-lacking present perfect negative structures featuring just a participle alongside the negation element, the new generation thus concluded that these forms in fact contrast with the likewise copula-lacking, but crucially finite-marked, negative presents. This resulted in the original non-finite participial form being reanalysed as a finite past form. This reanalysis of the verbal inflection marking, in turn, triggered a reordering of the heads in the clausal hierarchy, with Tense and Polarity being inverted in the innovative grammar. Morphological reanalysis, then, drives syntactic reanalysis, with a domain that has been shown to vary cross-linguistically, rather than following a universally fixed hierarchical ordering— negation and polarity (see Cinque 1999, and Laka 1994)—being affected by the change. Migdalski’s chapter provides another illustration of how morphology may drive syntax. He shows that in some Slavic languages, such as Serbo-Croatian, clitics changed from being verb-adjacent to occupying second position. Migdalski relates this to the loss of tense distinctions in verbal morphology, which was contemporaneous with the shift to second-position clitics in various Slavic dialects. These changes he interprets as signalling the loss of T, a category which has been argued not to be universally present (see, for example, Bošković 2012, and also Ritter and Wiltschko in press). Strikingly, T-less languages are predicted to lack expletives, subject-object asymmetries, and, Migdalski argues, sequence of tense effects, i.e. a diverse cluster of properties of the type that one might have expected of a classical parameter. If Migdalski’s analysis is on the right track, it is clear that morphological change may have far-reaching and diverse consequences. Like Allen (2001), Michelioudakis deals with a change in which the supposed syntactic effects of a morphological change seem to lag behind. While morphological dative case is lost during the Hellenistic Greek period, Michelioudakis traces the effects of abstract, inherent, interpretable dative Case into Medieval Cypriot Greek; a later reanalysis as uninterpretable is merely facilitated by, and not deterministically triggered by, the earlier morphological change. What the contributions in this section signal, then, is, that despite the proliferation of work on morphology-syntax interactions over the last thirty years, much more
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
still remains to be done. Diachronic data are of particular interest for the study of these interactions, as they allow the investigation of temporally adjacent varieties that differ only minimally. The contributions in this section represent a small step towards this goal.
. Syntax, prosody, and information structure One of the most striking features of this volume is the number of contributions that go beyond the ‘core functional categories’ C, T, and v to assume a more articulated functional domain. This type of work is typical of the ‘cartographic’ research programme, which has its roots in papers like Pollock (1989) but which crystallized in the late 1990s with the decomposition of the IP-domain by Cinque (1999) and of the CP-domain by Rizzi (1997); see Cinque and Rizzi (2010) for an overview. Most of the chapters in this volume exhibit the influence of the cartographic enterprise: as we have already noted, Hegedűs is inspired by cartographic analysis of the PP, while Julien makes use of an articulated IP. Most influential, however, has been the work of Rizzi (1997) on the clausal left periphery: the chapters by Reintges, Ledgeway, Hill, Walkden, Cormany, Danckaert, Hinterhölzl, and Wallenberg all make explicit reference to a Rizzian left periphery in their analyses. This reflects a broader concern with information structure that is showcased in the third and longest part of this book, and also a recognition that the facts of historically attested languages—within and outside the Indo-European family—are not easily accounted for using only the limited palette of functional categories provided in Chomsky (1995a).5 Information structure as object of study sits uneasily between pragmatics, semantics, and syntax: Vallduví (1992) and Lambrecht (1994) are key early works, and Krifka (2008) provides a readable introduction. That the field is still in its infancy is attested by the fact that there is still substantial disagreement over the crucial theoretical primitives involved. Some notion of topic and focus is assumed by just about everyone, though definitions differ: while the chapters in this section by Hill, Walkden, and Cormany largely focus on topics, focus is the topic of the chapters by Danckaert, Taylor and Pintzuk, Hinterhölzl, and Wallenberg. In addition, contrast deserves some mention here: while Rizzi (1997) makes no reference to it, it has been argued to exist as a primitive information-structural feature by Molnár (2001) and others (e.g. É. Kiss 1998; see also the contributions to the (2010) Lingua special edition on ‘Contrast as an information-structural notion in grammar’), and Neeleman et al. (2009) assume that [contrast] is a privative feature which can be present or absent on topics and foci. 5 Chomsky (: ) explicitly states that CP in his work is shorthand for the clausal region that Rizzi () refers to as the left periphery. Nevertheless, approaches to information structure have been developed within Minimalism that do not rely on an expanded cartographic left periphery: see Neeleman et al. (), Fanselow and Lenertova (), and Abels (), among others. Mensching () is a recent attempt to capture information-structural generalizations in diachrony without additional functional projections.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
The study of information-structural interactions in formal diachronic syntax is a recent development, but one that has proven extremely productive, as the collections of papers in Hinterhölzl and Petrova (2009a), Ferraresi and Lühr (2010), and Meurman-Solin, López-Couso, and Los (2012) illustrate. The crucial notions, however, are not new at all. Work by the Prague School (e.g. Firbas 1966; Halliday 1967) assumes a principle of given-before-new. A similar principle is present in Behaghel’s (1932: 4) Second Law: ‘Es stehen die alten Begriffe vor den neuen’ (old expressions come before new ones). The broadening of generative studies to include notions of information status as possible determinants of syntactic structure thus represents something of a unification of two traditions. If one assumes the existence of purely information-structural functional projections, a further important question is: where are they located? Rizzi’s (1997) original proposal locates them at the left periphery of the clause. Belletti (2001, 2004) has argued that there also exists a ‘low left periphery’ above vP, with Poletto (2006, 2014) motivating a clause-internal left periphery of this kind for Old Italian; Danckaert and Wallenberg further explore this possibility. Hill, in turn, proposes a further left periphery: one that lies at the edge of the DP. Focusing on the development of the Romanian object marker pe, Hill shows that it is no longer a preposition, as it was in Old Romanian, but now serves to mark contrastive topics. If CP, vP, and DP may each have a separate information-structurally active left periphery, it is tempting to suggest that this possibility is definitive of non-defective phasehood. Another indication of the immaturity of studies on syntax-information structure interactions is the fact that Rizzi’s (1997) model of the left periphery has been adapted in two different, and fundamentally incompatible, ways, by Benincà and Poletto (2004) on the one hand and Frascarelli and Hinterhölzl (2007) on the other, based largely on exactly the same language (Italian). The information-structural part of Rizzi’s original split CP is given in (1), that of Benincà and Poletto in (2), and that of Frascarelli and Hinterhölzl in (3). (1) Top∗ > Foc > Top∗
(Rizzi 1997: 297)
(2) Hanging Topic > Scene Setting > Left Dislocated > List Interpretation > Contrastive Focus (adverbs/objects) > Contrastive Focus (circumstantial adverbs) > Information Focus (Benincà and Poletto 2004: 71) (3) Aboutness Topic > Contrastive Topic > Focus > Familiar Topic∗ (Frascarelli and Hinterhölzl 2007: 112–3) As can be seen, Benincà and Poletto (2004) and Frascarelli and Hinterhölzl (2007) decompose the upper Topic field in different ways. Furthermore, while Benincà and Poletto (2004) do away with the lower Topic field entirely, Frascarelli and Hinterhölzl (2007) argue that it is reserved for ‘familiar’ topics, i.e. those that are simply given information, as opposed to those which define ‘what the sentence is about’ (Reinhart
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
1981). If progress is to be made, tensions like these must be resolved. Walkden argues that the preverbal element in early West Germanic V3 constructions meets the description of Frascarelli and Hinterhölzl’s (2007) Familiar Topic, thus providing some small support for the notion of a Topic field in the lowest part of the CP-domain. Walkden argues that the V3 pattern is a retention from Proto-West Germanic, and that generalized V2 was most likely a later development. If a left-peripheral ‘template’ such as that proposed by Rizzi (1997), Benincà and Poletto (2004), and Frascarelli and Hinterhölzl (2007) turns out to be the right way to analyse interactions between syntax and information structure at the left periphery, a further question arises: are all of the functional projections present in all languages, or can they be syncretized in some languages, as proposed by Giorgi and Pianesi (1997) for the IP-domain? Cormany suggests that syncretism is possible through a process of conflation of left-peripheral projections diachronically, drawing his evidence from a corpus study of historical Friulano. Specifically, doubling of various types of subjects becomes increasingly possible in the history of Friulano, and the emergence of the possibility of doubling each kind of subject occurs in just the order that an account based on conflation of left-peripheral projections would suggest. Little has been said about focus so far in this section, though this is probably the information-structural category on which the most has been written. The remaining chapters in part three of this volume all address accounts based on focus. Like topic, focus is likely not an irreducible notion: É. Kiss (1998), for example, distinguishes identificational focus—representing the exhaustive subset of the set of contextually or situationally given elements for which the predicate phrase holds—from information focus (or presentational focus), which simply represents new or non-presupposed information (see also Cruschina 2012 for illuminating overview discussion). Belletti (2001, 2004) then argues that identificational focus is associated with the FocP in the clausal left periphery, while information focus is associated with the ‘low’ leftperipheral FocP above vP. However, as Danckaert observes, Latin seems to counterexemplify this insofar as information foci may occur in the high left periphery of subordinate clauses in the Archaic and early Classical periods of this language. Rather than abandon Belletti’s generalization, Danckaert suggests that the apparent violation in Latin occurs because the lower FocP is made unavailable by independent movement operations; movement to the higher FocP, which is normally a position for identificational foci, then represents the syntax—or, recalling Berwick and Chomsky’s (2011) take on the actual locus of what is typically described as ‘syntactic variation’, PF—making the best out of a bad situation. For Batllori and Hernanz, too, the opposition between two types of focus is crucial. Their empirical starting point is the fact that, while both Old Spanish and Old Catalan allow weak (=information) focus fronting, modern Catalan only allows it with quantifier phrases. Batllori and Hernanz analyse weak focus fronting as movement to a dedicated specifier position in the lower part of the clausal left periphery, following Benincà and Poletto’s (2004) cartography.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Theresa Biberauer and George Walkden
They hypothesize that in Catalan this position was lost, and that weak focus fronting was consequently reanalysed as targeting the specifier of a Polarity Phrase just below it, accounting for the restriction of fronting to quantifier phrases in this language. As well as information structure, Hinterhölzl addresses a further important interaction: the interplay between syntax and prosody. The ambitious goal of his chapter is to reduce all word-order variation to the requirements of prosodic and informationstructural conditions. Hinterhölzl argues that Old High German, like the other early Germanic languages, exhibited mixed OV/VO properties, and analyses this in terms of a Kaynean universal base combined with the possibility to spell out lower copies in movement chains. The choice of which copy to spell out is then resolved with reference to prosodic requirements, e.g. whether the constituent is branching or non-branching, and information-structural requirements, e.g. given vs. new. It is clear that investigating information-structural and prosodic interactions in diachrony is a productive line of inquiry. However, it is only intellectually responsible to do so if reliable, replicable methods for identifying information-structural categories can be developed. Taylor and Pintzuk outline such a method, annotating objects in Old English for their prosodic weight and for whether they represent given or new information. Their findings, they argue, provide evidence that a one-to-one correspondence between VO order and new information status cannot be maintained. Instead two postverbal object positions must be postulated for Old English: one associated exclusively with new information, and another that is more informationstructurally promiscuous. Taylor and Pintzuk do not appeal to a cartographic template, instead representing the position for new objects as right-adjunction to TP; however, the data they present is consistent with an account in which informationstructural movement is to specifiers of dedicated functional projections rather than adjunction, and this is the topic of Wallenberg’s chapter. Heavy NP Shift across Germanic is the empirical domain of this chapter, which lays out a Kaynean cartographic analysis in which ‘rightward’ focus movement is to FocP at the clausal left periphery followed by remnant movement of the rest of the clause to TopP. Using a quantitative approach, Wallenberg compares this account to a traditional analysis in terms of right-adjunction to either TP or vP, and finds that the frequency of Heavy NP Shift is higher in Early Middle English than in either Old or Early Modern English—a result predicted under the traditional analysis, but not under the new one. The closing chapter in this section is something of an outlier, in that it explicitly argues against an information-structural or prosodic analysis. Aldridge looks at pronominal object shift in Archaic Chinese, which previous authors have attempted to account for in terms of prosodically motivated cliticization (Feng 1996) or focus movement (Djamouri 1991, 2000). Aldridge shows that these analyses do not account for the particular distribution of pronominal object shift, and presents an alternative
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Introduction: Changing views of syntactic change
in which the pronoun is fronted in order to value structural accusative Case. The analysis receives support from diachronic facts: pronominal object shift declined at the same time as morphological case was being lost on pronouns. The chapter thus serves as a salutary reminder that appealing to information-structural or prosodic motivations is not the solution to all our problems, and that sometimes a more traditional morphosyntactic account better fits the facts.
. Summary The chapters collected in this volume show that historical work within generative frameworks is branching out. Syntax within a Minimalist framework may remain an autonomous module, but it is ultimately responsible to its interfaces. This volume showcases interactions between the syntactic component and the lexicon, morphology, prosody, and information structure. Working on these interactions is not a simple task, and requires interdisciplinary knowledge as well as new methods for investigating change. The results of such work, however, can shed new light on previously poorly understood syntactic alternations both synchronically and diachronically. Historical work of this kind can also contribute to the ongoing investigation of the Strong Minimalist Thesis: the claim that language is an optimal solution to legibility conditions (Chomsky 2000: 96). Studying lexical, morphological, and informationstructural interactions will help us establish what these legibility conditions are, and how individual languages are configured to meet them in different places and at different times—whether or not syntactic change, in the narrow sense, really exists.
Acknowledgements This volume has its origins in the 12th Diachronic Generative Syntax Conference, held in Cambridge, 14th –16th July 2010. The editors would like to express their gratitude to the staff of Queens’ College, Cambridge, to the conference helper, Tim Bazalgette, and particularly to our co-organizers, Alastair Appleton, Elliott Lash, Dimitris Michelioudakis, and Ioanna Sitaridou. Theresa Biberauer also gratefully acknowledges the financial support of European Research Council Advanced Grant No. 269752 ‘Rethinking Comparative Syntax’ (ReCoS). Thanks, too, to Anna Hollingsworth for her careful and efficient creation of the index to this volume.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Part I Syntax and the Lexicon
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic CAITLIN LIGHT
. Introduction The status of the subject position in German has been the source of some debate.1 For example, some studies (Biberauer 2004; Richards and Biberauer 2005) have argued that Tense-final Germanic languages do not have an EPP requirement in the traditional sense. The absence of an expletive that occurs in the subject position (as opposed to SpecCP) seems to support the argument that SpecTP has no special status in languages like German. This chapter will argue against such analyses. First, I will show that in historical stages of German, we see evidence of a subject expletive licensed specifically to fill SpecTP. This expletive, da in Early New High German (ENHG), is merged when the logical subject does not move to SpecTP, leaving the position empty. This supports a traditional analysis of the EPP in German. Furthermore, the existence of expletive da lends support to the argument that two (non-topic) subject positions are available in the German clause structure (cf. Haeberli 1999b, 2000, 2005). In the second section of the chapter, I will present a more speculative study of Old English (OE), which proposes that þær was available as an expletive before the Middle English period, when the use of presentational/existential there began to rise (Williams 2000). I will propose a link between the facts in ENHG and OE, based on which I argue that the traditional EPP is active in Tense-final Germanic languages.
1 A part of this work previously appeared in the Proceedings of the th Annual Meeting of the Berkeley Linguisitics Society. I would like to thank Anthony Kroch, Julie Legate, David Embick, Joel Wallenberg, the attendees of BLS and DiGS XII, and my two anonymous reviewers for their advice and assistance on this work. I am also deeply grateful to Beatrice Santorini and Florian Schwarz, who provided the Modern German judgments used in this chapter. Of course, any errors are my own.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
. The behaviour of da in the history of German This chapter presents a corpus-based study, providing new data on the behaviour of da across the history of German. The corpus data used for ENHG was taken from a parsed corpus of Martin Luther’s first translation of the New Testament, the Septembertestament, published in 1522. The corpus consists of roughly 100,000 words of fully annotated and hand-corrected text (Light 2011). The data will demonstrate that, while da is certainly not an expletive in Modern German, it was historically available as an expletive. The following subsection will present the basic facts on the phenomenon in ENHG. Section 2.2.2 will compare the ENHG data to Modern German, and show that Modern German da cannot be analysed as an expletive. In Section 2.2.3, I look farther back in the history of German. Section 2.2.4 shows that there are similar phenomena across the Germanic language family, which may contribute to our understanding of the facts about da. Finally, Section 2.2.5 will propose an analysis of the facts throughout the history of German. .. Da in Early New High German As an adverbial pro-form, da generally carries a locative interpretation (1), but given the correct context, it can also be temporal, as in (2). Da can also be used as a complementizer (3). Like other demonstrative pro-forms (such as the personal demonstrative pronouns der, die, and das), locative da may be used as a relative pronoun, alternating with the wh-word wo ‘where’. (1)
Vnd es waren da viel weyber and it was da many women . . . ‘And there were many women there.’
(Septembertestament, Matthew 27:55)
(2) Da berieff Herodes die weysen heymlich da appointed Herod the wise secretly ‘Then Herod secretly appointed the wise men.’ (Septembertestament, Matthew 2:7) (3)
vnd da sie yhn sahen, fielen sie fur yhn nyder and da they him saw fell they before him down ‘And when they saw him, they fell down before him.’ (Septembertestament, Matthew 28:17)
The above facts are essentially constant through ENHG and contemporary German. However, the ENHG element da differs in one significant respect. In subordinate clauses, da frequently occurs in the position of an extracted subject. The use of da
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
in examples like the following sounds archaic, and in fact distinctly biblical, to native speakers of contemporary German. Petrus (4) Simon, der do heyst Simon who da is-called Peter ‘Simon, who is called Peter’
(Septembertestament, Matthew 4:18)
In subordinate clauses, da is almost entirely restricted to clauses with gaps (Table 2.1). Furthermore, only subject gap clauses show this effect; clauses with a gap left by a non-subject behave like clauses without a gap with respect to this phenomenon (Table 2.2). In the subject gap examples where da appears, it is uniformly clause-initial. In contrast, of the six examples of da in a clause with a non-subject gap, five are not clause-initial (5). The behaviour of da in subject gap clauses seems to suggest an entirely different phenomenon. (5)
vnd wie viel korbe hubt yhr da auff? and how many baskets lifted you da up ‘And how many baskets did you collect there?’ (Septembertestament, Matthew 16:9, 16:10)
Although da is available both as a complementizer and a relative pronoun in ENHG, I will argue that the da associated with subject gaps is neither. This da almost always occurs below an overt relative pronoun, making it clear that it cannot itself be a relative pronoun. In addition, when da is used as a relative pronoun, da has a locative interpretation, and corresponds to the extraction of a locative adjunct, not a subject. Table .. Distribution of da in subordinate clauses
Clauses with gaps All other subordinate Total
da
No da
Total
144 11 155
1832 3045 4877
1976 3056 5032
Table .. Distribution of da in clauses with gaps
Subject gap Non-subject gap Total
da
No da
Total
138 6 144
972 860 1832
1110 866 1976
2 In this and subsequent examples from earlier texts, there is some variation in the spelling of da: in ENHG, it may be spelled either do or da.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
We can also see that the subject gap da is not a complementizer because it can occur below the overt complementizer das. (6) Wer sagen die leutt, das da sey des menschen son? who say the people that da is of-the man son ‘Who do the people say is the son of man?’ (Septembertestament, Matthew 16:13) All of these facts suggest that the da which appears in subordinate clauses with a subject gap must be performing a different function. I will argue that it is a subject expletive. .. Comparing Modern German da We have seen that in ENHG, da frequently occurs in the position of an extracted subject. However, as I will demonstrate, da is not an expletive in standard German today. In normal usage, da may carry either a locative or temporal interpretation according to context, much like there in English (cf. Bayer and Suchsland 1997; Koeneman and Neeleman 2001; Kratzer 2004; Richards and Biberauer 2005). This ambiguity of interpretation often makes it difficult to identify whether da may be semantically null, as an expletive would be, in a given context. Despite this, as I will show, da is not an expletive in Modern German. Bayer and Suchsland (1997) and Richards and Biberauer (2005) have proposed that da may in fact be used as a subject expletive in SpecTP in Modern German. They argue that da may be merged into the subject position when the logical subject remains low. (7) a. daß (da) gestern ein Schiff versunken ist that (da) yesterday a ship sunk is ‘. . . that a ship sank yesterday.’ b. ∗ daß da ein Schiff gestern versunken ist c. daß ein Schiff gestern versunken ist
(Richards and Biberauer 2005)
This argument must ultimately be rejected, for contemporary German at least. Although Richards and Biberauer (2005) do not explicitly discuss the intended interpretation of da in (7b), it may be that the given sentence is argued to be ungrammatical only when da does not carry a locative interpretation. In fact, native German speakers easily accept the sentence in (7b), but only with a locative (or temporal) interpretation for da. When paired with another locative, the sentence generally gains a redundant reading, or, where possible, a reading in which two locations are referenced (as in (9)). This suggests that da cannot be semantically null in contemporary German, and thus is not an expletive.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
(8) ?Ich glaube, daß da auf der Party getanzt wurde. I believe that da at the party danced became ‘I believe that there was dancing there, at the party.’ (9) Ich glaube, daß da in China angerufen wurde. I believe that da in China up-called became ‘I believe that someone in China was called (from there).’ Furthermore, the structural position of da does not fit the expletive analysis. In German, weak pronouns are necessarily high in the structure. Subject pronouns are grammatical in sentences with da, but obligatorily precede it. The same is true of object pronouns. (10) a. ∗ Ich glaube, daß da er ihn gegessen hat. I believe that da he it eaten has ‘I believe that he ate it there.’ b. ∗ Ich glaube, daß er da ihn gegessen hat. c. Ich glaube, daß er ihn da gegessen hat. The position of da thus does not match with the claim that it occupies SpecTP. What we see instead is that in Standard German, da is weak, and behaves much like weak pronouns in German. As a result, it prefers a position at the left periphery of the clause, often string-adjacent to the material in C. However, it must appear after all pronominal arguments. Weak elements in German follow an ordering hierarchy: weak adverbs like da, for example, must be lower than a weak object pronoun (cf. Lenerz 1977). The fact that da tends to appear at the left periphery is completely expected under this analysis, although it is not an expletive. An alternative analysis of da has been proposed by Kratzer (2004) in the framework of situational semantics, which captures the ambiguous interpretations associated with the element. She suggests that da may be considered a situation pronoun, an adverbial pro-form which may be used to refer to any salient information in the context situation. (11)
Was riecht denn da so komisch? what smells part da so funny ‘What’s the strange smell here?’
(12)
Da brandelt was. da burns something ‘Something is burning.’
(Kratzer 2004)
As I have noted before, non-subject gap da in ENHG has the same general properties as da in Modern German, including an ambiguous and context-dependent interpretation. I will therefore assume that the general usage of da, both in ENHG
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
and in Modern German, falls under Kratzer’s analysis (excluding, of course, its use as a complementizer or relative pronoun). However, as it stands, this analysis offers no further insight into the particular correlation between da and subject gaps. In the next section, I will show that the necessary evidence can be found farther in the past. .. The historical origins of expletive da Regular sound change led to the collapse of two OHG adverbs, th¯o ‘then’ and th¯ar ‘there’, into a single adverb da. This led to the element’s ambiguous interpretation (cf. Axel 2007), allowing it to develop into a situation pronoun by the modern period. Only the behaviour of OHG th¯ar offers insight into the curious behaviour of subject gap da in ENHG. In fact, th¯ar shows a correlation with subject gaps, just like da in ENHG. It is simpler to diagnose th¯ar as an expletive than da, because it does not have the semantic ambiguity we have seen in ENHG (having not yet merged with the temporal adverb th¯o). If th¯ar does not have a possible locative interpretation in a given clause, it can be assumed to be semantically null. In order to explore the behaviour of th¯ar in these cases, I considered a small sample from the OHG Tatian text. This text, written in the ninth century, is a translation of the Latin translation of a text originally attributed to a second-century author. It is similar in style to the Septembertestament, being a synthesis of the four Gospels. The sample of the Tatian used for this study included 56 relative clauses, 41 of which had a subject gap. As predicted, many of the subject gap examples included an instance of th¯ar, with no possible locative interpretation. (13)
bithiu uuanta mir teta mihhilu thie th¯ar maht¯ıg ist because since to.me does much that da mighty is ‘Because the Mighty One did great things for me’
(14) Thie th¯ar hab¯e o¯ run thie h¯ore. that da have ears that hear ‘He that has ears to hear, let him hear.’
(Luke 1:49)
(Matthew 13:43)
As in ENHG, th¯ar occurs almost exclusively in the position of an extracted subject. Although the sample size is too small to say with complete confidence, this data allows a tentative estimate that th¯ar occurred in subject gaps about 50 of the time in OHG (Table 2.3). I also compared the subject relatives in this sample to 35 parallel clauses from the Septembertestament which were also translated as a subject relative (Table 2.4). The frequency of da/th¯ar has significantly decreased between the two time periods (chi-square = 5.29, p = 0.021). This suggests that the use of the subject expletive has already begun to decline by the ENHG period. According to this preliminary study of OHG, we can establish the following stages: in OHG, th¯ar was available as an expletive, to optionally fill the subject position when
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
Table .. OHG th¯ar in clauses with gaps
Subject gap Non-subject gap Total
th¯ar
No th¯ar
Total
21 1 22
20 14 34
41 15 56
Table .. The subject expletive in OHG and ENHG
OHG ENHG Total
da/th¯ar
No da/th¯ar
Total
21 8 29
20 27 47
41 35 76
the subject is extracted. In ENHG, da shows the same property, although its status as an expletive is obscured by the semantic underspecification of the situation pronoun da. In Standard German, da is no longer available as an expletive. .. Subject gaps and expletives in Germanic This phenomenon, as one might expect, is not unique cross-linguistically. In this section, I will consider two languages in the Germanic family in which subject gaps left by extracted subjects may be filled by expletive elements. This will provide context for the puzzle at hand. A better understanding of the mechanisms underlying these phenomena may help explain the behaviour of da in ENHG. ... Danish der In Danish, the element der appears clause-initially in relative clauses and indirect questions with a subject gap (cf. Jacobsen and Jensen 1982; Erteschik-Shir 1984; Vikner 1991; Mikkelsen 2002). Der is grammatical in an indirect question where the subject has been extracted, as in (15), but not permitted if the object has been extracted instead (16). (15)
Jeg ved ikke hvem der kan li’ ham. I know not who der likes him ‘I don’t know who likes him.’
(16)
∗ Jeg
ved ikke hvem der han kan li’. I know not who der he likes ‘I don’t know who he likes.’
(Erteschik-Shir 1984)
It is important to note that der ‘there’ is independently attested as a subject expletive in Danish, such as in matrix clauses. Two conflicting analyses have been proposed to
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
account for the behaviour of der. Erteschik-Shir (1984) and Mikkelsen (2002) propose that the der which occurs in clauses with a subject gap is indeed the subject expletive der, occupying the gap. Jacobsen and Jensen (1982) and Vikner (1991), in contrast, argue that it is a homophonous element in C. There are two obvious reasons to position der in the subject gap (SpecTP). The first is, of course, that it is independently attested as a subject expletive. The second is that der does not appear in clauses with a nonsubject gap. Mikkelsen (2002) shows that the distribution of der in subordinate clauses with a subject gap supports the subject expletive analysis. It is marked, but acceptable, for der to occur with the complementizer at and the element som, which Mikkelsen analyses as an invariant operator (like OP) in SpecCP. Whenever any of these elements co-occur, they must be in the order som at der, requiring der to appear below the complementizer. The analysis of der as a C-element requires CP-recursion to account for this structure, and idiosyncratic properties of each head must explain the ordering restriction. However, the assumption that der is in SpecTP predicts exactly this distribution. (17)
a. ?Vi kender de lingvister som at der vil læse denne bog. we know the linguists som that der will read this book ‘We know the linguists who will read this book.’ (Vikner 1991) b. ∗ Vi kender de lingvister at der som vil læse denne bog c. ∗ Vi kender de lingvister at som der vil læse denne bog d. ∗ Vi kender de lingvister der som at vil læse denne bog e. ∗ Vi kender de lingvister der at som vil læse denne bog
(Mikkelsen 2002)
Vikner (1991) suggests some problems with this analysis which must be addressed. There is usually a transitivity restriction on the appearance of expletive der: it can generally only occur with intransitive clauses. There is also a restriction on the definiteness of the logical subject, which usually must be indefinite. When der appears in the subject gap position, neither restriction holds. (18)
a. vi kender de lingvister der vil læse denne bog we know the linguists der will read this book ‘We know the linguists who will read this book.’
(Vikner 1991)
Because German (unlike Danish) does have transitive expletive constructions, a discussion of the first issue is not relevant for our current concerns. I will focus instead on the definiteness problem. In response to this point, Mikkelsen (2002) proposes that the gap in relative clauses behaves like an indefinite, even when its antecedent is not. This has been previously argued, particularly for English (cf. Browning 1987; Bianchi 1999), based on examples like the following:
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic (19)
a. ∗ There were the men in the garden b. The men that there were in the garden were all diplomats.3 (Browning 1987)
Mikkelsen relates this to the argument, made in Reinhart (1987), that relative whwords in English are inherently indefinite. She suggests that the evidence provided by der motivates us to extend this analysis to Danish, to explain why the gaps in relative clauses seem to be patterning as indefinites. ... Yiddish subject gaps Yiddish is a symmetric V2 language: verb-second orders are required both in matrix and subordinate clauses. If neither the subject nor any other constituent is raised to the topic position, the expletive element es appears instead. This expletive also appears in subject gap position, when the subject is extracted from a free relative or indirect question (cf. Diesing 1990, 1997; Prince 1989, 1993). However, this does not take place in ordinary relative clauses. (20) a. Ikh veys nit ver es iz gekumen I know not who es is come ‘I don’t know who came.’ b. ∗ Ikh veys nit ver iz gekumen c. Ikh veys nit vos Max hot gegesn I know not what Max has eaten ‘I don’t know what Max ate.’ d. ∗ Ikh veys nit vos Max hot es gegesn (21)
(Diesing 1990)
a. Der melamed vos iz besser far ir iz beser far mir the teacher rel is better for her is better for me ‘The teacher that is better for her, is better for me.’ b. ∗ Der melamed vos es iz beser far ir iz beser far mir
(Prince 1989)
Free relatives and indirect questions have a special status here. Prince (1989) argues that the relevant characteristic of these clause types is that the extracted element in a free relative or indirect question represents brand-new information in the discourse, unlike the extracted element in relative clauses, which has a clear referent. Prince associates this with the pragmatic properties of subject postposition. Using a small corpus of narrative Yiddish, Prince demonstrates that subject postposition in Yiddish is more frequent with brand-new subjects, suggesting that discourse status is one potential motivation for subject postposition.
3 Judgments for this example were mixed, but according to a variety of native speakers of English, the grammaticality of the example dramatically improves if the sentence is placed into a context which places a contrastive or focal accent on the subordinate clause verb.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
In indirect questions and free relatives, the extracted element is pragmatically treated as brand new, because it is not related to any elements already present in the discourse. Prince therefore suggests that, prior to extraction, the subjects of indirect questions and free relatives were postposed. As a result, these subjects have been extracted from the postposed position. The insertion of the expletive es is then required because there is no subject trace in the topic position, and the expletive is licensed to fill the gap. The Yiddish case is somewhat different from ENHG da. Unlike es in Yiddish, da is attested in ordinary relative clauses. In addition, because Yiddish is a symmetric V2 language, we know that there is a requirement to fill the preverbal position in subordinate clauses. In German, subordinate clauses are not V2. However, we may ask whether there is a subject position which must be filled in German subordinate clauses. .. Analysing da in historical German Thus far, we have established the following facts. From OHG until the present day, da may function as a weak adverb, generally occurring in a position in the left periphery of the verbal domain. In addition, while da is never used as a subject expletive in Standard German today, its OHG cognate th¯ar did have this function. ENHG can be taken as the middle stage in between these two systems. Because of the ambiguous interpretation of the situation pronoun, it is often impossible to distinguish semantically null cases of da from cases that are behaving as referential pro-forms for a less obvious situational context. However, the examples of da in subject gap position may occur in non-narrative and even ‘timeless’ contexts, in which there is literally no salient situation for da to refer to, as in (22)–(23). (22) Jhesu Christi, der do ist ein son Dauids des sons Abraham. Jesus Christ who da is a son of-David the son of-Abraham ‘Jesus Christ, who is a son of David, Abraham’s son.’ (Septembertestament, Matthew 1:1) (23) Selig sind, die da geystlich arm sind blessed are who da spiritual poor are ‘Blessed are those who are spiritually poor.’ (Septembertestament, Matthew 5:3) Just as Prince (1989) argued for Yiddish, I propose that da occurs in SpecTP when the subject is extracted from a position lower in the clause. However, unlike Yiddish, the appearance of da in subject gaps is optional (the evidence from OHG, in which the expletive occurs in roughly 50 of the subject gap clauses, supports such an assumption). My analysis of the ENHG expletive must hinge on a mechanism which allows subjects to optionally remain low.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
Haeberli (1999b, 2000, 2005) discusses two subject positions in the Germanic language family. Weak subject pronouns are restricted to the high position, as we have discussed; full DP subjects, meanwhile, may optionally remain low. I follow Wallenberg (2009) in assuming these positions to be SpecTP and the base position of the subject in SpecvP.4 Wallenberg also notes that the low subject position in German is related to the definiteness effect discussed by Diesing (1992). Definite DPs, Diesing observes, prefer a high position (here assumed to be SpecTP); only indefinite subjects may (optionally) remain low (here assumed to be SpecvP). I argue that, in historical Germanic, expletive da is licensed to fill the higher subject position, SpecTP, when it is empty. The expletive occurs in subject gap clauses when the subject is extracted from its base position in SpecvP without first raising through SpecTP. As Mikkelsen (2002) claims for der in Danish (Section 2.2.4.1), there is reason to believe that the gaps in subject relatives are inherently indefinite, regardless of the characteristics of their antecedents. Therefore, I argue that the subject of a relative clause behaves like an indefinite DP in German as well, and may optionally remain low (before extraction). In these cases, da is inserted to fill the empty position in SpecTP. A demonstration can be observed in the comparison of Figures 2.1 and 2.2, which show simplified trees of the relative clause in (23). The subject may have been extracted directly from the low subject position (SpecvP), as in Figure 2.1. In this case, the expletive fills the gap. Figure 2.2 shows a derivation in which the subject passes through SpecTP before being extracted, leaving a trace; in this case, the expletive is not licensed, and da would not appear. However, the subject expletive was not restricted to these contexts. The current study has focused on subordinate clauses with subject gaps entirely because they are the case in which this phenomenon is most easily observed. In matrix clauses, it is more difficult to observe the subject expletive da because it becomes nearly impossible to distinguish between it and the situation pronoun. We do find existentiallike examples with matrix clause da; an example is given in (24). Note that there is no locative or temporal adverb in the corresponding verse in either Greek or Latin. However, it is generally impossible to distinguish between the situation pronoun and the expletive in such cases. (24) a. Vnnd da war eyn weyb, das hatte den bluttgang zwelff iar and da was a woman that had the haemorrhage twelve years gehabt . . . had (Septembertestament) 4 Other accounts offer alternative positions for the two subject positions in Germanic: for example, Haeberli () proposes SpecTP and SpecAgrSP. I settle on an account using SpecTP and SpecvP mainly for simplicity’s sake. What matters for my purposes is the existence of two subject positions in Germanic. My argument can be reworked to fit with an alternative analysis of the two subject positions with only minor adjustments.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light CP
SpecCP
C’
die1 TP
C 0
T’
SpecTP da vP
T sind
SpecvP
v’
t1
geystlich arm
Figure . A relative clause in which the subject is extracted from a low position
CP
SpecCP
C’
die1 TP
C 0
T’
SpecTP t1 vP SpecvP t1
T v’
sind
geystlich arm
Figure . A relative clause in which the subject is extracted from a high position
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
b. Kai gunE tis ousa en rusei haimatos etE dOdeka . . . and woman any being in gushing of-blood years twelve (Greek, Scriveners Textus Receptus, 1894) c. Et mulier, quæ erat in profluvio sanguinis annis duodecim . . . and woman which was in issue blood years twelve (Latin, Clementine Latin Vulgate) ‘And (there was) a woman who had had a haemorrhage for twelve years . . .’ (Mark 5:25) My claim is that, while the majority of these cases are examples with the situation pronoun, some of these examples do indeed involve da behaving as a subject expletive. My analysis has two major consequences. First, the expletive da provides evidence that there is an EPP feature on T in German. When the extracted subject does not leave a trace in SpecTP, the expletive is licensed to satisfy the EPP in its place. In the next section, I will argue that a null expletive was also available to fill SpecTP as early as OHG, and that it eventually replaced da in this function. Second, the presence of da becomes support for the claim that there are two subject positions in German: the expletive may only occupy the high position if the extracted subject can be licensed in a low position. Otherwise, we expect the subject trace to fill SpecTP. We cannot easily motivate an analysis in which da is a pronounced version of the subject trace, because subject gap da is optional. By relating expletive da to the low subject position in German, we account for its optionality: indefinite subjects may be extracted without passing through SpecTP, but the subject may also move to SpecTP, in which case expletive da is not licensed. Crucially, although the element da was available as a subject expletive in historical German, it has lost this function in the modern language. There are two ways of analysing the loss of expletive da: either it was replaced by a null expletive, or German has lost an EPP requirement, and the high subject position may now remain unfilled in these cases. In fact, Abraham (1991) gives evidence that a null expletive was available in OHG and Middle High German (MHG). For example, in a clause with an extraposed subject, an expletive may not occur (25a), although the Modern German equivalent would contain the expletive es (25b). Abraham argues that this null expletive was ultimately overtaken by the expletive es and lost. (25) a. uuár ist, dhaz . . . true is that (Is.(H) 24,5) b. Wahr ist es, dass . . . true is it that ‘It is true, that . . .’
(Abraham 1991)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
It is worth noting that among Abraham’s examples, demonstrating cases with a null expletive in OHG and MHG, are examples containing the historical source of da, th¯ar (or orthographic variants thereof). This may be further evidence of expletive da in OHG. (26) dar dunchet tir rehto da deems you.dat correct ‘You seem to be correct there . . .’
(N(P)I, 42,25), (Abraham 1991)
(27) . . . daz dâ komen waeren ritter vil gemeit that da come were knights very dumb (Nib. 79,2), (Abraham 1991) ‘. . . that dumb knights had come.’5 This suggests that the situation prior to ENHG was complex. In addition to the historical subject expletive, da (or th¯ar), we find evidence for a null expletive in the surviving texts. Some of the functions of this historical null expletive have been overtaken by an additional expletive, es. I suggest that, given the null expletive available in OHG, there was plenty of opportunity for expletive da to be replaced by a null form in the grammar. This, combined with independent changes in the function of the homophonous adverb da, led to a reanalysis of the subject expletive. Recall that, when the OHG temporal and locative adverbs th¯o ‘then’ and th¯ar ‘there’ collapsed into a single adverbial element (da), it resulted in a situation pronoun with the ability to refer to various aspects of the situation according to context. This development alone was sufficient to cause the ultimate loss of da as a subject expletive. Once the adverbial function of da was not unambiguously locative, it became difficult to distinguish between the situation pronoun and the homophonous subject expletive. The two forms would almost always occur in the same surface position: because this adverbial is prosodically weak, it tends to occur at the left edge of a clause, near the high subject position which would be occupied by a subject expletive. By the ENHG period, most examples of da either are unambiguously adverbial, or can be analysed as either the situation pronoun or the subject expletive. It is only in a few cases, and in the overall frequency of clause-initial da in subordinate clauses with a subject gap, that we can see clear evidence of the subject expletive. Learners of German after the OHG period would encounter dramatically less unambiguous evidence for the subject expletive than for the situation pronoun. I argue that this would have led to the gradual loss of the form, as learners eventually reanalysed every instance of da as a situation pronoun. As this change occurred, the subject expletive da was gradually replaced by a null expletive. This assumption is motivated by the fact that a null expletive already existed in the grammar of OHG, and was thus available to overtake the function of expletive 5 Abraham presents this example as an ergative construction (‘purely rhematic, i.e. presentative’). He consequently claims that it involves a null expletive.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
da. Ultimately, da lost its expletive function, and occurs only as a situation pronoun in the modern language.
. An expletive in Old English The second part of this chapter is a speculative exploration of OE. A corpus-based study6 suggests that the ENHG phenomenon of da has a parallel in the history of English in þær ‘there’. I will motivate this claim in the next subsection. Then, in Section 2.3.2, I will briefly outline the behaviour of þær across the OE period, and its loss by the beginning of Middle English. In Section 2.3.3, I will provide an analysis for the existence and loss of ‘subject gap’ þær. .. Old English subject gaps and þær OE þær shows the same pattern as ENHG da in subject gap clauses, but at a much lower frequency. The overall frequency of þær, although low, is significantly higher in clauses with a subject gap than in other subordinate clauses (Table 2.5). Þær also occurs at a higher frequency in subject gap clauses than in clauses with a non-subject gap (Table 2.6); just as in ENHG, clauses with a non-subject gap pattern more like clauses without a gap with respect to the distribution of þær.
Table .. Distribution of þær in subordinate clauses
Þær No þær þær
Subject gap
Other subordinate
515 18014 2.8
1055 79217 1.3
chi-square = 205.7137, p ≈ 0
Table .. Distribution of þær in clauses with gaps
Þær No þær þær
Subject gap
Non-subject gap
515 18014 2.8
186 17074 1.1
chi-square = 133.8805, p ≈ 0
6 Data for OE was taken from the York-Toronto-Helsinki Parsed Corpus of Old English Prose (Taylor et al. ). The data on Middle English is taken from the Penn Parsed Corpus of Middle English (Kroch and Taylor a).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light
In subordinate clauses, we can find examples of þær for which a locative interpretation is unlikely. In (28), the relative clause discusses days of the calendar, and thus entities which are independent of any location. In (29), meanwhile, there is a full locative phrase (to þam scypum), which would make the occurrence of a locative demonstrative redundant. (28) Þa fif dagas þe þær synd betwux VI kalendas and kalendas the five days that there are between VI kalendas and kalendas Martii Martii ‘The five days that are between VI kalendas and kalendas Martii’ (ByrM_1_[Baker-Lapidge]:2.1.77.635) (29) þa ane þe ðær ut ætswymman mihton to þam scypum. the one who there out swim might to the ship ‘The one who could swim out to the ship.’ (ChronC_[Rositzke]:915.1.18.1096) These examples are not enormously convincing on their own. However, we have also seen that the overall frequency of þær demonstrates a different function in subject gap clauses. Its position in the clause further supports the claim that it is behaving as an expletive: þær occurs clause-initially at a much higher frequency in subject gap clauses (Table 2.7). There are two possible interpretations of this fact. Either þær is occupying the subject gap in these clauses, or, alternatively, þær is frequently adjacent to the subject, and as a result it frequently appears to be clause-initial when the subject is extracted. We must therefore ask the question: does þær occur clause-initially in subject gap clauses at the same frequency as it occurs adjacent to an overt subject? If this were the case, we would be forced to conclude that þær is not in the subject gap, but next to it. To answer this question, I compared the frequency of ‘subject gap’ þær to the frequency at which þær appears adjacent to an overt clause-initial subject in subordinate clauses. In fact, the frequency of subject gap þær does not match the ‘subject-adjacent’ pattern (Table 2.8). Other light adverbs do not share this pattern; the frequency at which other light adverbs occur clause-initially in subject gap clauses is essentially the same as the frequency at which they appear adjacent to an overt subject (Table 2.9).
Table .. Position of þær in subordinate clauses
Initial þær Non-initial þær Initial
Subject gap
Non-subj. gap
No gap
472 43 91.7
38 148 44.2
348 521 40.0
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
Table .. Clause-initial and subject-adjacent þær
Initial/adjacent þær Non-initial/adjacent þær initial/adjacent
Subject gap
Other subordinate
472 43 91.7
389 200 66.0
Table .. Clause-initial and subject-adjacent adverbs
Initial/adjacent adv. Non-initial/adjacent adv. initial/adjacent
Subject gap
Other subordinate
1614 1595 50.3
6906 4974 58.1
.. A change in the distribution of expletive þær The behaviour of þær is apparent in both Tense-final and Tense-medial clauses. This can be shown by considering frequencies of initial þær in clauses with a finite auxiliary and a non-finite verb. In clauses with the auxiliary-verb (roughly representing an underlying Tense-medial order, minus some examples of Verb (Projection) Raising), subject gap þær occurs at a much higher rate than subject-adjacent þær: in subject gap clauses with this configuration, þær occurs clause-initially 76.9 of the time, and in other clauses it occurs adjacent to the initial subject 41.1 of the time. The difference is visibly and statistically significant (chi-square = 8.4634, p = 0.0036). In clauses with the verb-auxiliary order (representing an underlying Tense-final order), the rate of subject-adjacent þær rises (80.9), but the rate of subject gap þær is still significantly higher (93.5; chi-square = 5.0505, p = 0.0246). Therefore, the occurrence of ‘subject gap’ þær does not seem to be unique to either the older Tensefinal grammar of OE, or the newer Tense-medial grammar. The frequency at which þær occurs clause-initially in subject gap clauses is relatively stable across the OE period (Table 2.10). However, the overall frequency of ‘subject gap’ þær declines slightly (Table 2.11). By the Middle English period, þær behaves like other light adverbs in clauses with a subject gap (Table 2.12): þær and other light adverbs now occur clause-initially at roughly the same frequency. .. Analysing Old English þær The data suggest, contra Breivik (1991) and Ingham (2001), that þær was not simply a ‘dummy topic’ in OE/early Middle English before it was established as an existential expletive. In subject gap clauses, þær can be used to fill SpecTP. However, the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Caitlin Light Table .. Distribution of þær across the OE period
Initial Non-initial initial
Pre-950
Post-950
82 7 92.1
141 5 96.6
chi-square = 1.427, p = 0.2323
Table .. Frequency of initial þær across the OE period
Initial þær No (initial) þær initial
Pre-950
Post-950
82 3465 2.3
141 7882 1.8
chi-square = 3.711, p = 0.05406
Table .. Þær and other light adverbs in Middle English Þær Initial Non-initial initial
48 95 33.6
Other light adverb 794 1226 39.3
chi-square = 1.6175, p = 0.2
occurrence of þær in subject gaps seems to disappear just before modern expletive there began to establish itself in the language (Williams 2000). English did not replace expletive þær with a null expletive in any context, including in the context of subject extraction. Recall, for example, the sentence from Browning (1987). (30) The men that there were in the garden were all diplomats. In existential/presentational contexts, the subject may still be extracted from a low position, and an overt expletive is licensed to fill the gap. Therefore, unlike German, English has retained the ability to fill a subject gap with overt there. I therefore suggest it is not the expletive which occurs at such a low frequency in these clauses, but the gap. Two facts about OE may contribute to this possibility. First, a great deal of literature has discussed the two subject positions in OE (cf. Haeberli 2001; Speyer 2008; Wallenberg 2009). The high position is obligatory for pronominal subjects, while
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Expletive there in West Germanic
full DP subjects may remain in the lower position. However, Haeberli (2001) notes an overwhelming preference for all subjects (pronominal and full DP) to appear before the tensed verb in subordinate clauses, which may be evidence that subjects predominantly target SpecTP in this context. The York-Toronto-Helsinki corpus of Old English supports this claim. Of the total sample of complement that-clauses in OE, I find that subjects remain postverbal at a rate of roughly 3.8. If this is also the case in clauses with an extracted subject, it may cause the overwhelming majority of subjects to move to SpecTP before extraction. This offers an explanation for the low rate of subject gap þær in OE. In addition, the low subject position is lost as a general option for full DPs (cf. Speyer 2008). The loss of this position further restricted the contexts in which subjects could be extracted from SpecvP. This would lead finally to the current situation, in which SpecTP remains unfilled by the subject only in some existential and presentational contexts, therefore effectively causing the disappearance of ‘subject gap’ there.
. Conclusion I have presented quantitative evidence that both historical German and Old English used an expletive to fill SpecTP in subordinate clauses with a subject gap. This contradicts previous arguments that these languages do not have expletives in SpecTP, and also suggests that the EPP is active in Tense-final Germanic languages. The expletive occurs optionally in subordinate clauses with a subject gap, suggesting that it only occurs when the subject is extracted from a low subject position; however, extraction after the subject has moved through SpecTP is also possible. The loss of this phenomenon occurred in each language for different reasons. In German, the semantically underspecified nature of adverbial da contributed to its reanalysis as an adverb in all contexts, and its expletive function came to be filled by a null expletive already available in OHG. In English, the subject simply lost the option of extracting from a low position in most contexts. This study stands as a demonstration of the importance of parsed corpora to empirical syntactic research. By using tools such as these, we come to understand synchronic facts of a language in ways that are not evident when viewed in isolation. In fact, the phenomenon I have explored in Old English occurs at such a low frequency that, without access to quantitative data of a corpus consisting of 1.5 million words, it would not even be observable. Building and using corpora such as those central to this chapter must be treated as a central concern in the continuing development of the field, both for diachronic and synchronic work.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal J OA N M A L I N G A N D S I G R Í Ð U R S I G U R J Ó N S D Ó T T I R
. Introduction A New Impersonal construction (NI) is developing in Icelandic.1 This is a systeminternal change which is neither the result of borrowing nor the result of any phonological change or morphological weakening, and as such it violates the Principle of Inertia proposed by Keenan (1994, 2002) and developed by Longobardi (2001). The new construction takes the form in (1); it appears to have a passive participle, but differs from the canonical passive illustrated in (2) in that the verbal object remains in situ and gets assigned accusative rather than nominative case if the argument does not bear a lexical case (dative or genitive). (1)
a. Í gær var bólusett mig við svínaflensu. yesterday was vaccinated-neut.sg. me-acc against swineflu ‘Yesterday I was vaccinated against the swine flu’ var hrint stráknum. b. Það itEXPL was pushed-neut.sg. boy.the-dat ‘The boy was pushed’
1 This research was supported by grants from Vísindasjóður Rannsóknarráðs Íslands (RANNÍS), Rannsóknarsjóður Háskóla Íslands, and Lýðveldissjóður; the original pilot study reported in Maling and Sigurjónsdóttir () was supported in part by NSF grant BCS- to Brandeis University. The material in this chapter is based in part on work done while the first author was serving as Director of NSF’s Linguistics Program. Any opinion, findings, and conclusions expressed in this material are those of the authors, and do not necessarily reflect the views of the U.S. National Science Foundation. Note that Sigurjónsdóttir and Maling () and Maling and Sigurjónsdóttir () are essentially the same paper, the former written in Icelandic, the latter in English. Here we will usually refer only to the latter paper.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
(2) a. Í gær var ég bólusettur við svínaflensu. Yesterday was I-nom vaccinated-masc.sg. against swineflu b. Stráknum var hrint. boy.the-dat was pushed-neut.sg. The verbal participle, which agrees with a nominative subject in the canonical passive, as illustrated in (2a), is invariantly neuter singular in the NI (compare (2a) to (1a)). Note that the expletive það in (1b) is not a grammatical subject but merely a place-holder confined to initial position in finite clauses. Three syntactic innovations characterize the NI as compared to the canonical passive (see Maling and Sigurjónsdóttir 2002: 100): (3)
a. accusative rather than nominative case on the underlying object; b. lack of NP-movement to subject position; c. the postverbal NP can be definite as well as indefinite.
An object which is marked accusative in the active voice continues to be marked accusative in the NI, and consequently the auxiliary and the past participle do not agree with it. The postverbal direct object does not undergo NP-movement to subject position, and can be definite as well as indefinite, in apparent violation of the Definiteness Effect (Safir 1985). The subject of a canonical passive can occur in postverbal position only if it is indefinite. When the indefinite postverbal NP is either a neuter singular noun, where nominative and accusative are not distinct, or marked with an oblique case (Dative or Genitive), the clause can be analysed either as a canonical passive, or as an instance of the NI (see Sigurjónsdóttir and Maling 2001: 128, 139; Thráinsson 2007: 276; Eythórsson 2008). The NI was first noted in the linguistic literature in the early 1980s as an innovation characterizing the language of children and adolescents (cf. Bernódusson 1982: 212; Hálfdanarson 1984: 31; Sigurðsson 1989: 355; Kjartansson 1991). Data on the NI has been collected in two nationwide studies. The study conducted in 1999–2000 by Sigurjónsdóttir and Maling was the first study of the syntactic characteristics of the innovative construction. We developed a questionnaire which was distributed to 1,731 tenth graders (age 15–16) in 65 schools throughout Iceland; this number represents 45 of the children born in 1984 in Iceland. Based on the results of our study, we distinguished between a small part of the capital city that we called ‘Inner Reykjavík’ and the rest of the country, which we referred to as ‘Elsewhere’. More than half of 15–16year-old adolescents in most parts of the country (n = 1,475) accepted sentences with an accusative definite postverbal object, like the one in (1a), with a range between 51 and 69, but only 28 of adolescents in Inner Reykjavík (n = 220) accepted these sentences, and very few of the adult controls (n = 200). Dative objects were accepted at a slightly higher rate (Maling and Sigurjónsdóttir 2002).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
Table .. Acceptability rates for the NI construction in (4) by age (Thráinsson et al. 2013) Response
14–15 yrs
20–25 yrs
40–45 yrs
65–70 yrs
Yes, a natural sentence Questionable sentence No, impossible sentence
59 19 23
22 13 65
1 4 95
0 5 95
Total number of subjects
n = 200
n = 179
n = 168
n = 165
A second nationwide study was conducted in 2005–7 (Syntactic Variation in Modern Icelandic, project leader: Höskuldur Thráinsson), and reported in Thráinsson et al. (2013). This study examined many different areas of morphosyntactic change in the language, and included some examples of the NI. No distinction was made between Inner Reykjavík and Elsewhere in that study, but in addition to adolescents, their subjects included three different age groups of adults: 20–25 years, 40–45 years, and 65–70 years. Results for the example of the NI shown in (4) are shown in Table 3.1.2 (4) Loks var fundið stelpuna eftir mikla leit. finally was found girl.the-acc after great search ‘The girl was finally found after a long search’ Both studies showed striking age-related variation. As can be seen in Table 3.1, nearly 60 of adolescents accepted the sentence in (4) as natural, as compared to only 22 of subjects in their early twenties, and almost no one in the two oldest age groups. Survey results show that speakers who use the NI also use the canonical passive. It is not clear whether the discourse function of the NI differs from that of the canonical passive, but younger speakers are observed to use it in contexts where older speakers would use the canonical passive. Before discussing the syntactic properties of the NI in more detail, we would like to note that the innovative construction seems to have established itself in the language more firmly in the decade since our nationwide study of 1999–2000 (Maling and Sigurjónsdóttir 2002). It is gaining ground in Inner Reykjavík, and can now be found not only in the spoken language but also in student papers and in informal written registers of the language, including blogs.
. The New Impersonal: Active or passive? Clearly something in the language is changing. But what exactly is the nature of the change? The analysis of the innovative construction has been the subject of lively 2 Percentages in all the tables have been rounded off, and may not add up to exactly .
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
debate in recent years; scholars differ in their assessment of whether the NI is a passive or an active impersonal construction. A good survey of the empirical facts and theoretical issues can be found in Thráinsson (2007: 273–83). The results of the Sigurjónsdóttir and Maling survey (M&S 2002: 116–18) show clearly that the postverbal NP is an object, so the Definiteness Effect, which only constrains postposed subjects, does not apply. On this there is general agreement. The disagreement lies in what is assumed to occupy the syntactic subject position. There are two basic classes of analyses as sketched in (5a, 5b) (M&S 2002: 100). (5)
Aux [VP VPTCP NP]] a. [IP [e] b. [IP proarb Aux [VP VPTCP NP]]
Passive without NP-movement Active Impersonal
Under the analysis sketched in (5a), the new construction is a variant of the standard passive in which NP-movement to subject position fails to apply, and the NP gets assigned accusative in situ, presumably by the verbal participle. This is not unprecedented: there are other reported cases of accusative-assigning passives, such as Sobin’s (1985) analysis of the –no/to participial construction in Ukrainian. However, the theoretical challenge of this analysis is the appearance of accusative case, which is unexpected under Burzio’s Generalization (Burzio 1986). The alternative sketched in (5b) is that the new construction is a syntactically active construction with a thematic subject that happens to be phonologically null. The null subject is interpreted as [+human], and thus can be seen as the null counterpart to impersonal subject pronouns like French on or German man. Icelandic is not a nullsubject language; personal pronoun subjects are as obligatory as they are in English; however, Icelandic does have null subjects in clauses containing weather verbs and in various impersonal constructions. Under an active analysis, the verbal morphology of the new construction is surprising in so far as it looks identical to the morphology of the canonical passive for neuter singular nouns or when the postverbal NP is marked Dative or Genitive (compare (1b) and (2b)). However, in order to understand the current status of the construction, we have to be willing to look beyond the historical identity of the morphemes involved. In earlier work, we have argued that the NI is a syntactically active construction which has its roots in the reanalysis and gradual extension of the traditional impersonal passive of intransitive verbs (Maling and Sigurjónsdóttir 1997, 2002; Maling 1993, 2006). Eythórsson (2008) and Jónsson (2009) have challenged our hypothesis. (See also Barðdal and Molnár 2003.) Arguing that the NI exhibits the syntactic properties of a ‘true passive’, they interpret the observed degree of variation in the responses of adult speakers as indicating that the syntactic properties of the canonical passive are ‘fuzzy’, and thus that the diagnostic tests we used to distinguish between an active vs. passive analysis of the NI, albeit standard in much of the generative literature, are inconclusive. Sigurðsson (2011) takes an intermediate position, coming to the conclusion that the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
new construction ‘is an unusually “active passive” ’ (p.160) and that it ‘is not an alien but a member of the passive family, albeit a somewhat odd one’ (p.176). As typologists have long pointed out, the distinction between active and passive is not a sharp dichotomy but more of a continuum; see the quote from Shibatani (1985: 822) in (6). (6) ‘the familiar controversy . . . over whether a given construction should be considered a passive is pointless; rather, a description must be offered in terms of how such a construction is similar to or different from the proto-typical passive.’ In trying to understand this ongoing change, we find it useful to compare the properties of the NI with the properties of the canonical passive in Icelandic. The controversy indicates the importance of developing concrete syntactic diagnostics for an active vs. a passive analysis. Based on her study of the Polish and Ukrainian –no/to-constructions and the Irish autonomous construction, Maling (1993) selected the four syntactic properties listed in (7) to use as diagnostics for an active analysis. (7) a. b. c. d.
No agentive by-phrase is possible. Binding of anaphors (reflexive and reciprocal) is possible. Control of subject-oriented adjuncts is possible. Nonagentive (‘unaccusative’) verbs can occur in the construction.
The underlying assumption is that a syntactically present subject argument blocks an agentive by-phrase, but licenses binding of lexical anaphors and control of subjectoriented adjuncts. Furthermore, unaccusative verbs should be able to occur in the construction provided that the verb selects for a human subject. A syntactically active impersonal construction with an overt grammatical subject, e.g. French on or German man, has all four of these properties; in contrast, the standard passive construction lacks all four properties.3 More detailed discussion can be found in Maling and Sigurjónsdóttir (2002) and Maling (2006). Maling and Sigurjónsdóttir (2002: 100–7) compared and contrasted the syntactic properties of the accusative-assigning participial –no/to-construction in Polish and Ukrainian. We showed that, despite their common historical origin and surface similarity (i.e. the shared morphological properties of assigning accusative case and consequent lack of agreement), the Polish and Ukrainian constructions are polar opposites in terms of syntactic behaviour: the Ukrainian –no/to-construction behaves like a true passive, whereas its Polish counterpart does not (for Polish, see also Blevins 3 The dichotomy is not always this clear-cut. For example, in German, impersonal passives allow a byphrase, but also reflexives and reciprocals. Both inherent and noninherent predicates form impersonal passives (see Plank , and especially Schäfer for discussion); moreover, at least some unaccusative verbs can form impersonal passives (Torgrim Solstad, p.c.). A Google search turns up examples like Es wurde gestorben auf beiden Seiten ‘it was died on both sides’. Clearly further investigation of the lexical restrictions on voice is needed. For Icelandic, see Sigurðsson (: , n. ) and Thráinsson (: –).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
Table .. Syntactic properties of various constructions in Polish and Ukrainian Syntactic property agentive by-phrase bound anaphors in object position control of subject-oriented adjuncts nonagentive (‘unaccusative’) verbs
Active
Pol/Ukr Passive
Polish –no/to
Ukrainian –no/to
∗
ok
∗
ok
ok
∗
ok
∗
ok
∗
ok
∗
ok
∗
ok
∗
2003; Kibort 2004). This is summarized in Table 3.2. Polish and Ukrainian both have a canonical passive with the expected properties. The take-home lesson from this comparison is that we cannot tell what the syntactic behaviour of a construction is by looking at superficial morphological properties like case and agreement. The obvious question, then, is this: which of the two polar opposites does the innovative Icelandic construction most resemble? Both possible answers have been offered. (8) a. the NI is parallel to the –no/to-construction in Polish, an active impersonal with a thematic proarb subject (Maling and Sigurjónsdóttir 1997, 2002; Maling 2006) b. the NI is ‘comparable to the –no/to-construction in Ukrainian, a passive preserving structural accusative case’ (Eythórsson 2008) Eythórsson (2008) and Jónsson (2009) claim that the Icelandic NI behaves more like the accusative passive in Ukrainian than like its Polish counterpart. Maling and Sigurjónsdóttir (1997, 2002) hypothesized that it is parallel to the completed development of the –no/to-construction in Polish, as well as the Irish autonomous construction, which also developed from a passive but is syntactically active (Maling 1993; McCloskey 2007). Although the actuation may take several centuries to complete, we predicted the NI in Icelandic will eventually pattern with the Polish –no/to-construction with regard to all four syntactic properties listed in Table 3.2. Because the change is still very much in progress, the dichotomy between the NI and the canonical passive in Icelandic is not as clear-cut as it is in Polish. Nonetheless, we find it useful to think about the ongoing change in terms of the predicted endpoint. In this chapter we expand the argument for our analysis, bringing in two kinds of evidence. Syntactically, we focus on the occurrence of bound anaphors, and show that they provide evidence for the active impersonal analysis of the NI. We also review evidence on control into bound adjuncts, and use of nonagentive/unaccusative verbs.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
Moreover, we examine sociolinguistic evidence for the hypothesis that a reanalysis has occurred and that it is not yet stable; we highlight the significant age-related variation observed in adults in the two nationwide studies mentioned earlier, which is exactly what would be expected during a period of syntactic change.
. Ways in which the New Impersonal differs from the canonical passive .. Bound anaphors It is widely accepted that reflexives and reciprocals require a syntactically realized antecedent, as illustrated by the contrast in English, illustrated in (9). (9) a. They arranged for each other to be on the committee. b. ∗ They talked about it for days. It was finally arranged for each other to be on the committee. (McCloskey 2007: 829, ex. 10a, 10b) In an impersonal passive clause like the one in (9b), neither the implicit agent nor the subject of the preceding sentence can serve as a binder for the reciprocal anaphor each other. As sketched in Table 3.2, Polish and Ukrainian differ in whether bound anaphors are allowed in the –no/to construction. So what about Icelandic? In our earlier work, we suggested that the impersonal passive of an intransitive verb, like the one in (10), is in principle syntactically ambiguous. var dansað. (10) Það itEXPL was danced ‘People danced/There was dancing’. The impersonal passive is open to two different syntactic representations, one as an Impersonal Passive, the other as an Impersonal Active with proarb subject. We agree with Haspelmath (1990: 35), who (in a different theoretical framework using different terminology) asserts that: (11)
‘. . . intransitive desubjectives are indistinguishable from passives of intransitive verbs.’
We posited two kinds of empty subject, one thematic and one not, as sketched in (5a, 5b). The crucial feature of the active representation in (5b) is that the null subject is a thematic [+human] subject which can serve as a syntactic binder, i.e. it is syntactically active. The fact that the understood subject of an impersonal passive can only be interpreted as a volitional agent (typically human) even if the verb allows inanimate subjects, surely contributes to the possibility of the proarb reanalysis. For example, the subject of the verb flauta ‘whistle’ can be many things, including teakettles or trains, but the impersonal passive Það var flautað ‘itEXPL was whistled’ can only be
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
understood as describing human whistlers. However, we will continue to use the traditional term ‘impersonal passive’ to refer to examples like (10) with an intransitive verb, even though we believe that many speakers, including adults, analyse them as syntactically active. We predicted that for those speakers who analyse the traditional impersonal passive in Icelandic as an active construction, the null thematic subject can serve both as a binder for anaphors and also as a controller for subject-oriented adjuncts. The data from Maling and Sigurjónsdóttir (2002) provides evidence that this is in fact the case. Consider first the sentence in (12), which looks like a traditional impersonal passive of an intransitive verb, except that it contains a bound anaphor, the accusative reflexive pronoun sig. The verb drífa sig ‘hurry up’ is inherently reflexive. (12)
Svo var bara drifið sig á ball. then was just hurried refl-acc to dance ‘Then everyone just hurried off to the dance’
(M&S 2002, ex. 30a)
The sentence in (12) was included in both nationwide surveys; the results from the Thráinsson 2005–7 study are shown in Table 3.3. Notice the significant increase in the acceptance of such examples across the agegroups; the younger the speakers, the more likely they are to accept them. Only 17 of the oldest age group accepted this sentence as natural, as compared to 70 of the teenagers. Also, compare the acceptability rates shown in Table 3.3 to those in Table 3.1 for a verb with a nonreflexive object. Significantly more adults accept the NI with simple reflexive sig than with nonreflexive objects. In the three adult age groups there is a much greater difference between the acceptance rate for reflexive vs. nonreflexive objects than in the teenage age group. The age-related variation is striking and shows that sentences with a reflexive object, like the one in (12), are accepted by many more speakers and seem to have a longer history in the language than the NI. We will return to this point in Section 3.4. Eythórsson (2008) and Jónsson (2009) suggest that such occurrences of reflexives can be explained away as an ‘intransitivizing suffix’ which does not require a syntactic binder (see Sells, Zaenen, and Zec 1987). But other bound anaphors, including pos-
Table .. Acceptability rates for the reflexive sentence in (12) (Thráinsson et al. 2013) Response
14–15 yrs
20–25 yrs
40–45 yrs
65–70 yrs
Yes, a natural sentence Questionable sentence No, impossible sentence
70 13 17
64 18 18
30 18 52
17 10 73
Total number of subjects
n = 204
n = 197
n = 191
n = 176
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
sessive reflexives, cannot be analysed that way. The examples in (13) with a possessive reflexive were included in our 1999–2000 survey (M&S 2002, ex. 33a, 33b). (13)
var haldið með sínu liði. Adolescents Adolescents Adults a. Það itEXPL was held with self’s team Elsewhere Inner Rvík ‘People supported their own team’. 63 49 36 b. Á kvöldin var skoðað tölvupóstinn sinn. Elsewhere Inner Rvík Adults in evenings was viewed e-mail-acc self’s 32 10 2 ‘In the evenings people checked their e-mail’.
Such examples look like real syntactic binding, and, indeed, fewer subjects accepted the impersonal passive in (13a) with a possessive reflexive than the sentence in (12), which contains a reflexive verb. The much lower acceptance rates for the example in (13b) are expected, since this is clearly an instance of the NI with an accusative object. Overall, however, our results showed that a significant number of adolescents accept the possessive reflexive, and there is the expected geographical and age-related variation. Eythórsson (2008: 199) agrees with Maling and Sigurjónsdóttir (2002) that this result indicates that ‘reflexive possessives pattern with non-reflexive arguments rather than simple reflexives’ but he dismisses the significance of this pattern. In the face of examples like those in (13), where it looks like real syntactic binding is taking place, Eythórsson (2008: 199) and Jónsson (2009) argue that anaphoric binding is not a reliable diagnostic. They suggest that such examples can be accounted for under an alternative analysis in which speakers allow the implicit agent to serve as an antecedent for the possessive reflexive. Eythórsson does not provide any supporting evidence from Icelandic, but cites English and German as languages in which reflexives and other bound anaphors ‘at least to a limited degree’ may occur in the passive. Jónsson (2009) cites the English example in (14) from Baker, Johnson, and Roberts (1989). (14) Such privileges should be kept to oneself. Jónsson (2009: 296) provides two Icelandic examples of the canonical passive with reflexives, shown in (15), which he judges to be acceptable. (15)
a. Á kvöldin var skoðaður tölvupóstur frá börnunum sínum. in the.evenings was checked email-nom from the.children self’s ‘E-mail from one’s children was checked in the evenings’ b. Sumt er bara gert fyrir sjálfan sig. some is just done for oneself ‘Some things you only do for yourself ’
The exact conditions on anaphoric binding in Icelandic definitely merit further investigation. Lødrup (2007) has collected similar Norwegian examples where the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
anaphor occurs inside the subject NP and has no ‘visible binder’. As Jónsson notes, the best examples of this kind in English involve generic sentences like (14), and (15b) is of the same type. But it does not follow that speakers of Icelandic generally allow the implicit agent to bind anaphors in passives. Note that Jónsson’s examples both involve reflexives inside prepositional phrases, so his examples are not instances of the NI. Neither Eythórsson nor Jónsson provide examples with possessive reflexives modifying the patient NP to test their suggestion that speakers allow the implicit agent to bind anaphors. In May 2009, we checked with four adult speakers of the NI to see whether the implicit agent could bind a reflexive in the canonical passive. The speakers all found a clear contrast between (13a, 13b), which they liked, and the canonical passive sentences shown in (16), which they did not like, regardless of the position of the subject NP. (16) a. Á kvöldin var tölvupósturinn (∗ sinn) skoðaður. in the.evening was e-mail-m.sg-nom (∗ self’s) checked-m.sg b. Tölvupósturinn (∗ sinn) var skoðaður á kvöldin. This contrast between (16) and (13b) indicates that speakers of the NI allow the implicit agent to bind anaphors only in the NI and not in the canonical passive. While further research is clearly needed, this preliminary data supports our view that, contra Eythórsson (2008), the Icelandic NI is more like the Polish –no/to-construction than like the Ukrainian counterpart. .. Adjuncts Further support for our hypothesis that speakers can choose between the two different syntactic representations for impersonal passives sketched in (5a, 5b) comes from another of the syntactic diagnostics, namely subject-oriented adjuncts. One difference between the Polish and Ukrainian –no/to-constructions lies in the ability to control subject-oriented adjuncts; see Table 3.2. What about Icelandic? Our survey included some examples with invariant (nonagreeing) participial adjuncts, like grátandi ‘crying’. Maling and Sigurjónsdóttir (2002, ex. 37a) reported that 52 of adult speakers accepted the subject-oriented adjunct in the impersonal passive shown in (17). (17)
var komið skellihlæjandi í tímann. Það itEXPL was come laughing.out.loud into class-acc ‘People came into class laughing out loud’
We also pointed out that ‘the more subject-oriented participles are accepted, the more simple reflexives are accepted’ (p.126). For adolescents, the correlation was highly significant (r = 0.433, n = 1,693, p = 0.000, 2-tailed); for adults the correlation was also highly significant (r = 0.532, n = 199, p = 0.000, 2-tailed) (M&S 2002: 126, fn. 15). This
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
correlation is evidence that these speakers have a syntactically active representation for so-called ‘impersonal passives’. There are other speakers who in contrast allow neither reflexives nor subject-oriented adjuncts; these judgments reflect a passive analysis. (We take no position on whether the grammar of an individual speaker can have both or only one of the representations sketched in (5); in the aggregate, there is evidence for both grammatical analyses.) No examples of agreeing adjuncts were included in either survey. An example of a canonical impersonal passive with an agreeing adjunct is shown in (18); the adjunct einn ‘alone’ must agree with its antecedent. In this example, there is no overt antecedent, but it takes the masculine singular form which is the default gender. (18)
var farið einn heim. Það itEXPL was gone alone-m.sg. home ‘Someone/people went home alone’
Although sentences like (18) have not been tested yet, we did include in our survey the sentence in (19), which is a reflexive impersonal passive with an agreeing adjunct. The previously unreported results are shown in Table 3.4; the acceptance rate for adolescents in Elsewhere ranged from 52 to 70. (19)
Svo var bara drifið sig einn á ball. so was just hurried-neut. refl alone-m.sg. to dance ‘So people just hurried off alone to the dance’
These results indicate that many speakers accept a reflexive impersonal passive with a subject-oriented agreeing adjunct and as with reflexives there is the expected geographical and age-related variation (see for example the results for the impersonal passive with a possessive reflexive in (13a) which are very similar to the results in Table 3.4). The examples in (20) of the NI and the canonical passive with an agreeing adjunct were not included in the surveys. var skrifað minningargreinina einn. (New Impersonal) (20) a. Það itEXPL was written-neut. obituary-f.sg.acc alone-m.sg. einn. b. ∗ Minningargreinin var skrifuð obituary-f.sg.nom was written-f.sg. alone-m.sg.
(Canonical Passive)
Table .. Acceptability rates for the reflexive impersonal passive with an adjunct in (19) Sentence Svo var bara drifið sig einn á ball
Adolescents Elsewhere
Adolescents Inner Rvík
Adults
60
48
22.5
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
However, the NI speakers we have consulted all agree that there is a sharp contrast between the NI, which allows the adjunct, and the canonical passive, which does not. Just as with reflexives, we conclude that the implicit agent in the canonical passive cannot serve as the controller, and thus to the extent that such adjuncts can be used in impersonal passives, speakers have analysed them as actives with a null proarb subject, as sketched in (5b). .. Nonagentive predicates (Unaccusatives) Another prediction of the active impersonal reanalysis is that the construction will extend from agentive (‘unergative’) predicates to nonagentive (‘unaccusative’) ones, provided that the understood subject is human (or at least animate4 ). In Polish, the –no/to morphology which originated as a neuter passive participle can now attach to unaccusative verbs, as long as the intended subject is human. Such forms do not exist in Ukrainian (see Table 3.2). What about Icelandic? The canonical passive in Icelandic is by and large limited to agentive predicates (Sigurðsson 1989: 320; Thráinsson 2007: 268–70; Sigurðsson and Egerland 2009: 167), although there are well-known exceptions in the standard language as pointed out by Eythórsson (2008) and Jónsson (2009). Four test sentences with nonagentive predicates were included in our 1999–2000 survey; the results varied widely, presumably depending on the lexical semantics of the individual predicates, as discussed by Jónsson (2009: 299), in various ways that remain ill-understood. Thráinsson (2007: 151–2, 268–70) also discusses the lexical constraints on passivization, and provides the examples shown in (21a, 21b) with the nonagentive verb eiga ‘have, own.’ (21) a. Þeir eiga hundinn. they own dog.the-acc b. ∗ Hundurinn er áttur (af þeim). dog.the-nom is owned (by them)
(Thráinsson 2007: 152, ex. 4.15a) (Thráinsson 2007: 152, ex. 4.15b)
var samt alltaf átt marga hesta. c. Það itEXPL was still always had many horses-acc ‘People/they still owned many horses’ This verb was one of several nonagentive predicates tested in the syntactic variation study, which included the Nl sentence shown in (21c). The results across the four age groups are shown in Table 3.5 (Thráinsson et al. 2013). 4 In both the traditional passive and the NI, the understood agent can exceptionally be interpreted as referring to a nonhuman animate, such as a dog, horse, or even a mosquito (Benediktsdóttir : –; Sigurðson & Egerland ). Therefore, the relevant feature is [+animate] rather than [+human], at least for Icelandic.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir Table .. Acceptabililty judgments for the NI in (21c) with a nonagentive verb Response
14–15 yrs
20–25 yrs
40–45 yrs
65–70 yrs
Yes, a natural sentence Questionable sentence No, impossible sentence
23 20 58
5 9 86
1 4 95
2 3 95
Total number of subjects
n = 205
n= 198
n = 191
n = 172
The canonical passive in (21b) is sharply unacceptable for all speakers. On the other hand, 23 of the 14–15-year-olds accepted the sentence in (21c), but only about 3 of those over age 40. These results indicate that the NI is ‘beginning to extend its usage to nonagentive verbs which do not form passives in the standard language’ (Maling and Sigurjónsdóttir 2002: 127). Further research is needed to investigate agentivity and volitional control as factors in forming impersonal passives by age-group, for both transitive and intransitive verbs. Both surveys included several examples of NI sentences with verbs which do not form passives in the standard language, and in both studies the acceptability judgments exhibited a wide range. Overall, the more volitional control on the part of the understood subject, the higher the acceptance for all age groups. So, ‘to be found’ was better than ‘to get lost’. Another example is shown in (22).5 (22) a. Eftir það var forðast óhollan mat í langan tíma after that was avoided unhealthy food-acc in long time ‘After that people avoided unhealthy food for a long time’ (New Impersonal) matur forðast(ur) í langan tíma. b. ∗ Eftir það var óhollur after that was unhealthy food-nom avoided in long time (Canonical Passive) Sentence (22a), an instance of the NI, was tested in the syntactic variation study. The results are shown in Table 3.6 (see Thráinsson et al. 2013). The results show that many speakers accept this sentence and the younger the speakers, the more likely they are to accept it. If the NI is analysed as a ‘true passive’, this change in the lexical restrictions would be mysterious, especially since it applies only to the NI and not to the canonical passive. While this extension to nonagentive 5 Anderson (: –) points out that there are morphological gaps in the paradigms for st-verbs; with two exceptions, the participles lack agreeing forms. So an st-verb which takes an accusative object simply cannot form a canonical passive.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
Table .. Acceptability rates for the NI in (22a) of a verb that does not form a passive Response
14–15 yrs
20–25 yrs
40–45 yrs
65–70 yrs
Yes, a natural sentence Questionable sentence No, impossible sentence
65 12 22
43 16 42
20 16 64
14 14 72
Total number of subjects
n = 205
n = 197
n = 190
n = 176
predicates is clearly a later development than the occurrence of bound anaphors, we see once again that the NI differs from the canonical passive.
. Stages in the diachronic development of the New Impersonal A crucial (and unexpected) observation in the Sigurjónsdóttir and Maling survey was the fact that our adult controls split almost 50–50 in whether or not they accepted impersonal passives of reflexive verbs. Since almost no adults accepted the NI, there is an implicational relation: speakers who accept the NI also accept impersonal passives of reflexive verbs, but not vice versa. In our (2002) paper we suggested that sentences with a reflexive object represent the first step in the reanalysis of the past participle in the NI from passive to syntactically active. We interpreted the age-related variation for reflexive impersonals as reflecting three stages in the diachronic development of the NI, as outlined in (23). (23) Stage 1. Impersonal passives occur only with true intransitive verbs (dansa ‘dance’) Stage 2. Impersonal passives occur with reflexive verbs (leika sér ‘play’, baða sig ‘bathe’) Stage 3. ‘Impersonal passives’ occur with all transitive verbs, with retained ACC case (NI) Eythórsson (2008: 189) observed that reflexive impersonal passive verbs were not found in Old Icelandic, but rather seem to be an innovation of Modern Icelandic which is increasingly gaining ground: (24) ‘I have not been able to find any cases of ImpC [=Impersonal Construction] with reflexive verbs in Old Icelandic; an investigation into its origins is pending. Thus, the reflexive ImpC seems to be an innovation of Modern Icelandic which is increasingly gaining ground and is accepted by many speakers who do not accept the NC [= New Construction] with non-reflexive verbs (cf. Maling and Sigurjónsdóttir, p.122).’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
Further support for this finding comes from corpus studies. A corpus search on an open-access digital library , which hosts digital editions of newspapers and magazines from the seventeenth century to the early twentyfirst century, found only sporadic examples of reflexive impersonal passives in the earlier periods, but after about 1890 or so, the number of examples increases significantly (Árnadóttir, Eythórsson, and Sigurðsson 2011). Passives of reflexive verbs have been considered marginal by some researchers, e.g. Halldór Ármann Sigurðsson (1989). Eythórsson asserts that ‘the ImpC with reflexives (“reflexive passive”) is rather marginal and is hardly robust enough to be a model for the NC transitives’ (Eythórsson 2008: 215, emphasis added). But elsewhere in that same paper, Eythórsson says (p.189) that he himself and most speakers he consulted find them acceptable. Linguists who are now in their fifties tell us that they have noticed an increase in the occurrence of reflexive impersonal passives over the last two decades or so, and believe that their acceptability judgments may have changed (Eiríkur Rögnvaldsson, p.c.). We thought that we might find a difference between various classes of reflexive verbs. Sigurðsson (1989: 355, fn. 60) observed a contrast between an inherent reflexive verb, such as drífa sig ‘hurry’, leika sér ‘play’, etc., and other reflexive verbs, like baða sig ‘bathe’, þurrka sér ‘dry oneself ’, etc., see (25). (25) a. ?Það var leikið sér allan daginn. it was played refl-dat all day b. ??Það var baðað sig á laugardögum it was bathed refl-acc on Saturdays The subtle contrast between one and two question marks could reflect a difference between inherent reflexives (25a) and noninherent reflexives (25b); that is, we might expect the presence of an inherent reflexive to be less salient than other bound anaphors, and thus to be judged more natural than other reflexives. Unfortunately the two surveys only included examples of inherently reflexive verbs. As far as we know, once reflexive impersonal passives come into the language, all types of reflexives are possible. Further research is needed to determine when impersonal passives can occur with other bound anaphors, including possessive reflexives, in PP complements but not NP complements (e.g. horfa á sjálfan sig ‘look at oneself ’, halda með sínu liði ‘support one’s team’). However there is another difference between (25a) and (25b), namely morphological case. A preliminary corpus search on the digital library suggests that the first examples of impersonal passives of reflexive verbs are all with dative sér, and only around the beginning of the twentieth century do verbs with accusative sig appear. This is in line with the speculation of Helgi Skúli Kjartansson (1991: 18) that the new construction was more common with verbs governing dative than with verbs governing accusative. The results of our study (M&S 2002: 112) strongly supported this observation. For adolescents the difference between the NI
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
with an accusative (M = 0.54, SD = 0.33) and dative (M = 0.61, SD = 0.37) object was highly significant, t(1691) = –10.928, p = 0.000 (2-tailed). Even for adults, the difference between accusative (M = 0.03, SD = 0.10) and dative (M = 0.06, SD = 0.15) was significant, t(199) = –2.717, p = 0.007 (2-tailed). This is consistent with observations about the development of the –no/to-construction in Ukrainian and Polish. The change began with those forms where the morphological evidence of nonagreement is least obvious; for centuries, the only examples of the innovative –no/to-construction in Polish contained neuter NPs which could just as well have been canonical passives (Zbigniew Kański, p.c.). Anthony Naro’s (1981) work on Brazilian Portuguese showed that the use of non-standard forms goes down as the surface salience of the deviation from the standard goes up (as pointed out to us by Anthony Kroch, p.c.). Recall that for verbs governing dative objects (as in (2b)), the participle in the canonical passive is default neuter singular, so only the fact that a definite NP occurs in postverbal position marks a sentence as an example of the NI (see (1b)). For verbs governing accusative objects (as in (1a) vs. (2a)), there is in addition the difference in morphological case and the consequent lack of agreement (M&S 2002: 112). Interestingly, only six years later, the second survey showed no difference between the acceptability of accusative and dative objects (Thráinsson et al. 2013). But how can we interpret the less-than-categorical acceptability judgments in both surveys, judgments that are not perfectly consistent with the predictions of either a passive or an impersonal active analysis? Eythórsson (2008) and Jónsson (2009) conclude that the syntactic diagnostics we used, albeit standard in much of the generative literature, must be unreliable. Maling and Sigurjónsdóttir (2002) suggested that the observed age-related variation in acceptability judgments is a reflex of the inherent syntactic ambiguity of impersonal passives. Given the infrequency of impersonal passives in the child’s input, she is unlikely to hear evidence that decides between the two different grammars available to speakers sketched in (5a, 5b). While we agree that the syntactic phenomena are more complex than presented in Maling and Sigurjónsdóttir (2002), we believe that there is more individual variation in acceptability judgments than most syntactic theories lead us to expect, and we suggest that variability in judgments is exactly what one would expect to find during periods of syntactic change. For example, as Old Icelandic gradually changed from OV to VO syntax, the surface strings showed mixed word-order patterns for several centuries (Hróarsdóttir 2000); this variability accounts for the passionate disagreement among linguists about the underlying word order of Old Icelandic during the transitional period.
. Conclusion We noted at the very beginning of this chapter that the development of the NI is a system-internal change which is neither the result of borrowing, nor the result
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joan Maling and Sigríður Sigurjónsdóttir
of any phonological change or morphological weakening. Thus it appears to be a counterexample to the ‘Inertia Theory’ of historical syntax, according to which syntactic change should not arise unless it is triggered by external causes or by a change in the lexicon or in other grammatical subsystems (e.g. phonology, semantics) (Keenan 2002; Longobardi 2001). Where is the variation that leads to syntactic change in this case? There must be some ambiguity in the data that the language learner hears which leads to a change in the grammar(s) that the child constructs. (See also Walkden 2012 who argues against the Principle of Inertia on theoretical grounds.) Keep in mind that this is not a unique linguistic change. We argue that a similar change can be seen in the independent development of the Polish –no/to-construction and the Irish autonomous construction (Maling 1993). The focus of this chapter is the stages in the development of the Icelandic NI. We do not, however, claim that the stages found in Icelandic are universal. Polish shows that there is no single pathway to the development of an active impersonal. Since Slavic languages do not have impersonal passives of intransitive verbs, the Polish –no/to-construction must have undergone a different developmental path from what is happening in Icelandic, although we believe that the end points will be virtually identical. We cannot emphasize too strongly that this syntactic change is still very much in progress. The results of our 1999–2000 study suggest that the crucial first step in the reanalysis of the impersonal passive as an active construction is its extension to reflexive predicates; this then extends to other bound anaphors. As shown in Table 3.1, adults (40 years and older) overwhelmingly reject the NI, but they accept a similar construction with a reflexive pronoun at a markedly higher rate (fully acceptable to approximately 30 of speakers in the 40–45-year-old age group, see Table 3.3). Moreover, many adults also accept control of participial adjuncts, and some accept impersonal passives of presumably unaccusative verbs with unspecified human subjects. All of these factors indicate that, even in the standard language, passive morphology is associated with a human agent reading, which makes possible the reanalysis of a passive construction as an active impersonal with an unspecified human subject. Given this variation among adult speakers, it is not surprising that children have started to extend this analysis to all verb classes, including transitive verbs, which take human subjects. There is a growing recognition that grammatical well-formedness is not a binary opposition/dichotomy but a continuum (see e.g. Featherston 2007). We need a more nuanced approach than a simplistic categorical assignment to Active vs. Passive. Unless we are willing to look beyond the historical identity of the morphemes involved, we cannot understand the nature of this ongoing change. Eythórsson (2008) and Jónsson (2009) argue that the NI is a passive construction, but they fail to provide any account for the variation among the adult controls in our 1999–2000 study, variation which is now even better documented in the recently completed Syntactic
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From passive to active: Stages in the Icelandic New Impersonal
Variation study. We do not share their belief that all native speakers necessarily come to the same grammatical analysis of every construction; on the contrary, we believe that speakers may come to radically different analyses of the same data. The readily observable data underdetermines the analysis; it is only by pushing the speaker to judge more complex, or less common (even ‘vanishingly rare’), sentences that we can see the empirical consequences of choosing one syntactic representation over another.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives W I L L IA M HA D D I C A N , E Y TA N Z W E I G , AND DANIEL EZRA JOHNSON
. Introduction This chapter focuses on change in the syntax and event semantics of be like quotative constructions, as illustrated in (1). (1)
Aaron was like ‘OK, fine.’ a. ‘Aaron thought/felt like saying “OK, fine.” ’ b. ‘Aaron said, “OK, fine.” ’
Be like as an introducer of quoted speech is innovative in many contemporary varieties of English. Recent corpus-based work on be like has suggested that as it has continued to spread, it has undergone syntactic and semantic change: be like predicates, originally used exclusively to describe states of individuals via reported thought as in (1a), have taken on an additional guise as descriptors of saying events as in (1b) (Tagliamonte and Hudson 1999; Buchstaller 2004). This chapter reports on two judgment experiments with speakers of American English intended to further explore claims of syntactic and semantic change in be like quotatives. Our experimental results suggest that the direct speech and reported thought readings of be like are diffusing into American English at a similar rate. From the perspective of Kroch’s (1989) seminal constant rate proposal, this result is consistent with a view of these two guises of be like as different environments in a single abstract process of change. In particular, we relate the ambiguity between direct speech and reported thought be like in (1) to the availability of copula be in active contexts as in (2) and (3) (Partee 1977; Dowty 1979; Parsons 1990; Rothstein 1999). (2) John forced him to be quiet. (3)
Jane is being polite.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
We extend Rothstein’s (1999) proposal for adjectival predicates under copula be to the variation between reported thought and direct speech interpretations of be like quotatives in (1a) and (1b) respectively. In particular, we propose that copula be always selects for an adjectival (stative) argument, and that the availability of eventive readings as in (1b), (2), and (3) is attributable to a semantic coercion mechanism1 akin to operations that make count readings out of mass nouns in the nominal domain. Adapting Haddican and Zweig’s (2012) analysis of be like quotatives, we furthermore argue that a range of exceptional properties of be like as a quote introducer in English reflect the presence of a null something quantifier in such constructions, as originally suggested by Kayne (2007). In describing the relationship between tense and agreement morphology and event semantic interpretation as cues to the phonetically null lexical material involved in this change, our analysis contributes to the theme of this volume, focusing on the interaction of these phenomena in syntactic change. The chapter is organized as follows. In Section 4.2, we report on an experiment to measure the correlation between speaker age and acceptability of eventive and stative interpretations of be like. Section 4.3 describes a second experiment that examines main verb properties of the be in be like constructions. Section 4.4 develops a syntactic and semantic analysis of innovative uses of be like. In Section 4.5, we discuss the evolution of agentive, direct speech interpretations of be like quotatives.
. Experiment : Direct speech and reported thought interpretations of be like quotatives Be like as an introducer of quotes was first described in diachronic and sociolinguistic literature on American English in the 1980s and since has been reported in many other varieties of English worldwide (Blyth, Recktenwald, and Wang 1990; Macaulay 2001; Cukor-Avila 2002; Buchstaller and D’Arcy 2009; Bakht 2010). Early work on be like described it not as an introducer of direct speech, but rather exclusively as an introducer of reported thought (Butters 1982). Much subsequent corpus-based work on be like however has reported that quotes introduced by be like can describe not just states of individuals as in (1a), but also saying eventualities as in (1b). A disadvantage of usage corpora for analysing semantic variation of this kind is that the speaker’s intended reading can be difficult to identify from the speech context. The following discussion therefore describes an experiment, first reported on in Durham et al. (2012), that is intended to examine cross-speaker differences in the availability of speech and non-speech readings of be like using a different technique, a controlled judgment experiment, that compares acceptability scores across conditions contextually biasing these different readings. 1 Rothstein refers to this as a ‘repackaging’ mechanism.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
.. Method Subjects. The participants were 121 self-described native speakers of American English aged 18–73 (M=31.3, SD=11.6)—71 women and 50 men. All had at least some university education. Participants were recruited online through the contacts of the researchers and were not paid for their participation. Materials. The experiment examined acceptability of matched be like and say sentences in six environments. First, we created a baseline context with no stativity/eventivity bias, as in (1). To compare acceptability of be like and say with eventive readings, we used four additional constructions, standardly used to bias such readings—progressives, imperatives, force . . . to complements, and pseudoclefts with do. All are contexts in which eventive predicates are fine, but true states are poor (Dowty 1979). We illustrate this in (4)–(7), which compare stative have $100 with eventive spend $100 in each environment. (4) She was ∗ having $100/spending $100. (5)
Just ∗ have $100/spend $100.
(6) Tim forced him to ∗ have $100/spend $100. (7) What she needs to do is ∗ have $100/spend $100.
(progressives) (imperatives) (force . . . to) (do pseudoclefts)
We used these four environments to compare acceptability of direct-speech and eventive readings of be like and say, as illustrated in (8)–(11). (8) She was being like/saying, ‘They’re coming tomorrow at 11:00.’
(progressives)
(9) Just be like/say, ‘They won’t ever do it.’
(imperatives)
(10) Tim forced him to say/be like, ‘Fine, I’ll do it next week.’ (11) What she needs to do is say/be like, ‘John already quit.’
(force . . . to) (do pseudoclefts)
Finally, to compare acceptability of be like vs. say in contexts biasing non-speech readings, we used for-adverbial phrases. As illustrated in (12), temporal for phrases are fine with atelic predicates in simple tenses but poor with eventives (Dowty 1979). (12)
For an hour, Mark had $100/∗ spent $100.
(for adverbials)
We therefore used such contexts to diagnose the availability of stative, non-speech interpretations of be like and say quotative predicates, as in (13). (13)
For an hour, Mark was like/said, ‘Let’s go to McDonald’s.’
(for adverbials)
Two lexicalizations were created for each of the above six environments, each assigned either to a be like or say condition, yielding two lists. Each participant therefore saw each of the above 12 conditions once. Subjects were randomly assigned to lists, and a
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
unique random order of the 12 test sentences and 18 fillers was created for each subject by the software used. Procedure. The data were gathered in the summer of 2009, through a self-paced, web-based magnitude estimation procedure using WebExp2. In linguistic magnitude estimation experiments, subjects judge stimulus sentences not on an abstract n-point scale but rather in relation to a non-zero score arbitrarily assigned to a benchmark (‘modulus’) sentence (Bard, Robertson, and Sorace 1996). If a participant judges an item to be twice as acceptable as the benchmark sentence, he/she gives it twice the benchmark score; if it is half as acceptable, half the benchmark score, and so on. In the present experiment, the benchmark sentence used was that in (14), which native speakers of English typically find to be of intermediate acceptability. (14) I wouldn’t give to the boy the difficult puzzle. Raw scores were normalized by dividing them by the benchmark score. The base-10 logarithm of these normalized values was then taken in order to make data normally distributed and suitable for parametric tests. In the following results, we report these log-transformed values. .. Results To examine the effect of speaker age on acceptability scores, we fit mixed-effect linear models for each condition using the lme4 package for R (Bates and Maechler 2010). The dependent variables were log-transformed values for each condition, with age and verb as fixed effects and subject and item as random effects. P-values were simulated by Markov chain Monte Carlo (MCMC) sampling (10,000 samples) using the LanguageR package for R (Baayen 2006, 2010). The results are summarized in Figure 4.1, which plots say and be like scores by subject age for each condition. The p-values reported in each plot are for the age∗ verb (be like vs. say) interaction. Zero on the y-axis on these plots corresponds to the modulus score. Scores above zero on each of these plots therefore correspond to a judgment higher than that for (14), and scores below zero reflect lower judgments. The plots in Figure 4.1 show that while the say-be like gap increases with age across these conditions, the age∗ verb interaction reaches significance at α=.05 only for three environments: the baseline context; pseudoclefts, and imperatives; the interaction for force . . . to complements is suggestive at p=.052. For for-adverbials there is no interaction between age and verb, and in fact no main effect for verb. These judgment data therefore align only partially with corpus data suggesting diffusion of be like in direct speech and non-speech contexts. The absence of more consistent age effects in these data may be partially attributable to the fact that our sample is relatively youthful, with a mean age of 31.3. These age effects are in any case orthogonal to the main focus of this study, and we set them aside in the following discussion.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Baseline 0.1 Say
0.0
Be like
–0.5 20
30 40 50 age p=.032
60 70
0.5
Say
0.0
Be like
–0.5 20
Say
0.5 0.0
Be like –0.5 20
30 40 50 age p=.035
60 70
Force...to 0.1 0.5
Say
0.0
Be like
–0.5 20
30 40 50 age p=.052
60 70
0.1 Say
0.5 0.0
Be like
–0.5 20
30 40 50
60 70
60 70
Pseudoclefts
Progressives
0.1
score (logs), 0=modulus
score (logs), 0=modulus
Imperatives
30 40 50 age p=.296
score (logs), 0=modulus
0.5
For adverbials 0.1
score (logs), 0=modulus
William Haddican, Eytan Zweig, and Daniel Ezra Johnson score (logs), 0=modulus
score (logs), 0=modulus
0.1 Say
0.5 0.0
Be like
–0.5 20
age p=.256
30 40 50
60 70
age p=.009
Figure . Say and be like scores by age for six conditions
In addition, the plots in Figure 4.1 show that acceptability of be like is particularly high among younger speakers. For speakers around 25 or younger, acceptability of be like is close to that for say across these conditions. We examine further the data from younger speakers, using models for each of the above six contexts, with data from only under-26-year-olds in the sample (n=50). As in the models just described, the dependent variables were the log-transformed acceptability scores, with fixed effect verb and random effects item and subject. The analyses revealed no significant main effects for verb in any of the six conditions. These results are summarized in Figure 4.2, which shows MCMC-estimated confidence intervals for say and be like sentences; MCMC-estimated p-values for the verb factor appear below the error bars for each condition. Figure 4.2 shows no significant main effect for verb for any of the six conditions, indicating that be like in these environments is on a par with counterpart say sentences for younger speakers. A final important result concerns the similarity in the age slopes for be like in the atelic-biased for condition and eventive-biased conditions in Figure 4.1. Linear mixed effect models revealed no significant interaction between age and condition for any of the four comparisons.2 These findings are in keeping with corpus results, which 2 For for adverbials vs. force . . . to as the eventive/stative comparison, p=.; for adverbials vs. pseudoclefts p=.; for adverbials vs. progressives p=.; and for adverbials vs. imperatives, p=..
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives Say Be like
0.8 score (logs), 0=modulus
0.6 0.4 0.2 0.0 –0.2 p=.553 baseline
p=.646
p=.197
p=.730
p=.117
for-ADV
force..to
imperatives
progressives
p=.374 pseudoclefts
Condition
Figure . Partial effects and 95 HPD intervals for under-26-year-olds’ scores in six conditions
generally converge in suggesting a constancy in the effect of the speech/non-speech contrast in this process of change (Cukor-Avila 2001; Tagliamonte and D’Arcy 2007; Buchstaller and D’Arcy 2009; Durham et al. 2012). The parallel diffusion of the two interpretations of be like in Figure 4.1 is consistent with an approach that treats their diffusion as a single abstract process of change. Work by Kroch and colleagues of the last two decades on patterns of variation in historical corpus data has suggested that for any single abstract process of syntactic change, contextual effects are typically stable over time—an effect that Kroch calls the constant rate effect (Kroch 1989, 1994; see also Pintzuk 1991; Santorini 1992; Fruehwald, Gress-Wright, and Wallenberg 2009). Kroch (1989) explains this effect in terms of individuals’ language-independent faculty for learning frequencies of experienced events. As learners acquire a given set of forms in variation they will also infer propensities of use of variants in different contexts. In the normal course of events, contextual effects will therefore be propagated across generations of speakers, all other things being equal. Sometimes, a given variant may acquire a new kind of social or pragmatic meaning, which may lead to change in contextual effects across time, but the historical corpus-based literature suggests that this is atypical (Kroch 1989). From the perspective of this literature, the present experimental results are therefore consistent with a view of the diffusion these two guises of be like as different contexts in a single abstract process of grammatical change. We spell out the syntactic reanalysis in Section 4.4. To summarize, the experimental results so far support two main conclusions about the syntax and semantics of be like quotatives. First, younger speakers in the sample accept be like readily in environments biasing both eventive direct speech and stative
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
non-speech interpretations. For all six conditions, younger subjects accept be like on a par with say. As eventive, direct-speech be like has entered the grammar, it therefore appears not to have displaced the stative variant of be like; rather, both interpretations are available for the younger speakers who use and accept be like to the greatest extent. Second, the data support corpus evidence suggesting that stative and eventive guises of be like are diffusing at the same rate. These facts are consistent with a view of these two variants as involved in a single underlying change in the grammar.
. Experiment : Be like and verb movement The data in the previous section support evidence from corpus studies indicating that for many younger speakers of English, a speech event interpretation of be like quotatives is readily available. In this section, we consider a second set of experimental data intended to test one formal approach to these facts. In many contemporary approaches to agentivity, change-of-state meaning is associated with a functional head merged low in the functional sequence of the clause, above the main verb (Chomsky 1995a; Kratzer 1996). From the perspective of these approaches, one possible account of the variation between eventive and stative be in be like sentences is in terms of their categorial status and merged position in the functional sequence. The be in stative be like contexts, on this account, will be a gardenvariety copula, merged in a designated copula projection or as a modal or tense head (Schütze 2004) as in (15). (15)
[TP be [PrtP like [ QUOTE ]]]
By contrast, be in eventive contexts will be merged as a main verb low in the functional sequence as in (16). (16) [TP [VP be[[SAY]] [PrtP like [ QUOTE ]]]] This approach predicts that the be of be like in eventive contexts will be un-auxiliarylike on standard diagnostics. One such test involves subject–auxiliary inversion (SAI), which in contemporary English is restricted to modals and auxiliaries. As illustrated in (17), inversion of the subject with the main verb ate is bad; do-support is required instead, as in (18). (17)
∗ Ate you a cucumber?
(18)
Did you eat a cucumber?
If the be of be like is a main verb of the familiar sort, then ceteris paribus, we expect be to be poor in SAI on eventive interpretations on a par with other main verbs. In contrast, if be in stative contexts is an auxiliary we expect it to be acceptable.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
A second standard diagnostic of main-verb-hood concerns placement of VP adverbs like quickly, which never appear to the left of modals/auxiliaries, but can appear to their right (Jackendoff 1972). (19)
∗ George quickly was finishing his dinner.
(20) George was quickly finishing his dinner.
(quickly-Aux) (Aux-quickly)
Main verbs, on the other hand, happily take quickly-type adverbs to their left, but not to their right, as shown in (21) and (22). (21)
Jeremy quickly ate his soup.
(quickly-V)
(22)
∗ Jeremy ate quickly his soup.
(V-quickly)
In the following sections we describe an experiment intended to test these predictions. .. Method Subjects. Participants were 50 volunteer undergraduates and staff at CUNY and NYU, 37 women and 13 men, aged 18–39 (M=20.3, SD=3.19). All were self-described speakers of American English. Materials. The experiment consisted of two subdesigns, one focusing on SAI and the second focusing on adverb placement. The SAI subdesign crossed two factors, each with two levels: verb movement, with levels inversion (a yes/no question) and non-inversion (a declarative); and quotative verb, with levels say and be like. The say sentences are included as a control condition: in order to test whether be like is degraded in inversion vs. non-inversion contexts, we compare the effect of this contrast with that for say as a quotative verb with more familiar main-verb syntax. To bias a speech event interpretation for be like, each test sentence included an adverbial felicitous with a direct speech reading but not non-speech/thought readings, e.g. twice in a row, in five seconds flat, etc. This design is summarized in Table 4.1. Four lexicalizations were created for each cell and assigned to one of four test groups by Latin square. Each subject therefore judged each condition once. The adverb placement subdesign was similar in design, crossing two factors: verbadverb order with levels V-adverb and adverb-V; and quotative verb with levels say and be like. This design is illustrated in Table 4.2.
Table .. SAI subdesign design
Say Be like
SAI
¬ SAI
Did she say, ‘shut up’ twice in a row? Was she like, ‘shut up’ twice in a row?
She said, ‘shut up’ twice in a row. She was like, ‘shut up’ twice in a row.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
Table .. Adverb placement subdesign design
Say Be like
V-Quickly
Quickly-V
She said quickly, ‘shut up’ twice in a row. She was quickly like, ‘shut up’ twice in a row.
She quickly said, ‘shut up’ twice in a row. She quickly was like, ‘shut up’ twice in a row.
Four lexicalizations were created for each cell and assigned to lists by Latin square. Each subject therefore judged each condition once. These sentences together with the four experimental sentences from the SAI subdesign were pseudo-randomized together with 12 fillers. Procedure. The testing procedure was similar to that for experiment 1. The data were gathered through a self-paced online magnitude estimation procedure using WebExp2 in the autumn of 2010. The benchmark sentence used was (14), the same used in experiment 1. After giving consent to participate, subjects were asked to provide some background information, including age, sex, highest level of education completed, and hometown. Subjects were then introduced to the magnitude estimation procedure, and then given two sets of slides providing practice in applying this technique. In the first set, subjects used magnitude estimation to measure lengths of lines; the second set provided sample sentences to judge. The experimental phase followed, which subjects typically completed in between five and ten minutes. As with the data from Section 4.2, we report normalized, log-transformed values following Bard, Robertson, and Sorace’s (1996) procedure. .. Results Subject-Aux inversion. The effects of verb and verb movement interaction were measured by fitting mixed effects linear models, with log-transformed acceptability scores as the dependent variable, verb and verb movement as fixed effects, and item and subject as random effects. P-values were again simulated using MCMC sampling (10,000 samples) using the languageR package. The analysis revealed no significant main effect for verb (p=.085) or verb movement (p=.200), and no significant interaction between these factors (p=.364). We illustrate these results in Figure 4.3, which plots the partial effects of the above model with 95 highest posterior density credible intervals for our four conditions: be like and say sentences in (SAI) yes/no questions and (non-SAI) declaratives. Figure 4.3 shows that aggregate scores for be like and say in non-SAI contexts are near identical. Scores for be like are somewhat degraded on the SAI condition, but this difference is not significant at α=.05. These results, therefore, provide no support for a difference in behaviour between be like and say in terms of SAI. The fact that movement of inflected be to the left periphery is on a par with that for do, in the say
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
score (logs), 0=modulus
0.5
Say Be like
0.4 0.3 0.2 0.1 0.0 Infl-Subj
Subj-Infl Order
Figure . Partial effects and 95 HPD intervals for be like and say in SAI and non-SAI contexts
score (logs), 0=modulus
0.5
Say Be like
0.4 0.3 0.2 0.1 0.0 ADV-V
V-ADV Order
Figure . Partial effects and 95 HPD intervals for be like and say in Adv-V and V-Adv orders
condition suggests that for this sample the be of be like in speech event contexts is not particularly main-verb-like in SAI-triggering contexts. Adverb placement. A mixed-effect linear model revealed no significant main effect for verb-adverb order (p=.245), but a significant main effect for verb (p=.001), and a significant interaction for verb and adverb-verb order (p=.047). These results are illustrated in Figure 4.4, which shows partial effects for the above model with 95 highest posterior density credible intervals.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
The interaction shown in Figure 4.4 is what we expect if the be of be like, unlike say, is not a lexical verb but rather sits higher in the clause. These results, along with the SAI data summarized in Figure 4.2, therefore, suggest little support for a ‘main verb’ be approach. In the discussion below, we develop an approach to the syntax/semantics of be like that takes be in such constructions to be a garden-variety copula. In the previous sections, we have seen experimental evidence for the existence of both eventive and stative readings of be like. We have also seen evidence that, perhaps surprisingly, be acts as a copula under both readings, among younger speakers. In the remaining discussion, we review one approach to ambiguity between stative and eventive interpretations of be like by Haddican and Zweig (2012), before discussing the evolution of be like.
. The syntax and semantics of be like In the previous sections, we have seen experimental evidence for the existence of both eventive and stative readings for be like. We have also seen evidence that, perhaps surprisingly, be acts as a copula under both readings. A natural starting point for an analysis, then, is to relate quotative be like to other cases where copula be can display eventive readings. In this section we summarize one such approach to these facts by Haddican and Zweig (2012), which we use in modelling the syntactic evolution of be like discussed above. Haddican and Zweig’s proposal departs from the well-known fact that copula be, while typically characterized as a stative verb, can take eventive readings in certain contexts, such as (23). (23) John is being silly. The first be in (23) is the auxiliary form that precedes V+-ing forms in progressives. The second, which appears in progressive form, is notable in that while it selects an adjective, the overall meaning imparted is not stative, but rather active. Parsons (1990) called this guise of be ‘be of activity’. Such uses of be are typically discussed in relation to progressive contexts like (23); however, as Rothstein (1999) discusses in detail, be of activity appears in several other contexts as well including object control constructions like (24). This sentence is ambiguous between a stative reading where Mary asked John to adopt a new characteristic, and an eventive reading where she requested that he act in a silly manner. (24) Mary asked John to be silly. Early accounts of the be of activity (Partee 1977; Dowty 1979; Parsons 1990) proposed that it is a case of lexical ambiguity, wherein English has a lexical item be that means something like act. However, Rothstein (1999) argues that this account is untenable, and proposes instead an account that argues that there is a single be in English. The
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
function of be, in Rothstein’s view, is to take the meaning of an adjective phrase, which is a property of states, and repackage it into a Davidsonian eventuality argument. In most contexts, this argument itself will be stative, so the purpose of the repackaging will be to convert a general stative property to a particular one; in (25), it converts the property of silliness to John’s particular silliness. (25) John is silly. However, when there are other cues to the nature of the eventuality—for example, a progressive aspect as in (23) or an embedded be as in (24), the process can result in an eventive reading for the adjective phrase.3 While there are important differences between be like and the data discussed in the majority of the be of activity literature— the main one being that the like-quotative argument of be is not an adjective phrase— Rothstein’s proposal provides an important component to understanding the change involved in the meaning of be like, as it shows that a single be can generate both eventive and stative readings from the same argument. Below we discuss the emergence of these meanings. Our syntax for be like quotatives will need to accommodate the semantic proposal just presented along with several other properties that distinguish be like from other English quote introducers as discussed in Haddican and Zweig (2012). First, as noted above, be like differs from say-type verbs in that it cannot introduce indirect speech, as shown in (26) and (27). (26)
∗ John was like that he was hungry.
(27) John said that he was hungry. Second, as noted by Flagg (2007), be like differs from say in that when a quote introduced by be like is questioned, the question word cannot extract. The question in (28), for example, is fine on an interpretation where the questioner is asking about some salient state of Aaron, but not to ask what Aaron said. Say in quotative contexts shows no such opacity to wh-extraction, as shown in (29). (28) What was Aaron like? a. ∗ ‘What did Aaron say?’ b. OK: ‘What was Aaron’s state?’ (29) What did Aaron say? Third, as Flagg (2007) notes, be like also differs from say, in that a quote introduced by be like cannot raise (Flagg 2007). Examples (30) and (31) illustrate the well-known fact
3 Readers may refer to Rothstein () for detailed argumentation and formal implementations of her theory.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
that quotes can precede say with or without an inverted subject (Collins 1997; Suñer 2000). (30) ‘Shut up,’ Aaron said. (31)
‘Shut up,’ said Aaron.
Be like quotatives on the other hand never allow quote raising with or without inversion, as shown in (32) and (33). (32)
∗ ‘Shut up,’ Aaron was like.
(33)
∗ ‘Shut up,’ was like Aaron.
Fourth and finally, unlike quotative sentences with say, be like quotatives on a direct speech interpretation are most naturally interpreted not as reporting a verbatim quote, but rather a close paraphrase (Buchstaller 2004: 111). We illustrate this difference in (34) and (35). (34a, 34b) show that quotatives with say are felicitously preceded with phrases like word for word and exactly which force verbatim interpretations. As shown in (35), counterpart sentences with be like are odd. (34) a. Word for word, she said, ‘I-didn’t-plagiarize.’ b. She said exactly, ‘I promise to be there.’ (35)
a. #Word for word, she was like, ‘I-didn’t-plagiarize.’ b. #She was exactly like, ‘I promise to be there.’
This ‘mere paraphrase’ component of be like quotatives does not appear to be asserted, but rather shows properties of being an implicature. These include the fact that it can be explicitly cancelled by later discourse, at which point the verbatim interpretation arises, as seen in (36), as well as the fact that it is susceptible to in fact cancellation (37). (36) A: She was like, ‘I-didn’t-plagiarize.’ B: Word for word? A: Yes. (37) She was like ‘I like pomegranates’—in fact, that was exactly what she said. Ignoring the ‘mere paraphrase’ meaning which we return to shortly, we take a view in the spirit of Davidson (1968), wherein the quote has to be the same as the speech event, where ‘same-saying’ allows for contextually agreed upon vectors of variation (for example, if the subject of the sentence spoke with a lisp, the person quoting them does not have to replicate this lisp to count as saying the same). We also adapt Davidson’s proposal in assuming that the quote is introduced by a demonstrative THAT. (See also Partee (1973), Munro (1982), and Etxepare (2010) for likeminded proposals.) In most dialects, this demonstrative is null,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
however in a few other varieties, including Glasgow English, it is optionally overt as in (38). (38) And they were like that ‘How’re you doing, Mary?’
Glasgow English (Macaulay 2001: 13)
Following Partee (1973), Munro (1982), and Collins and Branigan (1997), we assume that the quoted material is not merged as a complement of the saying predicate. Rather, following Etxepare (2010), we assume that the demonstrative is merged in a small clause structure headed by a null relator morpheme (den Dikken 2006). We take the like of be like to be a manner preposition, which takes the Relator-headed small clause as its complement. On these assumptions, the lower portion of a sentence like (1) will have the structure shown in (39). (39) [PP like [RelP that [Rel Rel QUOTE ]]] Something more, however, is required to account for additional properties of be like discussed above, namely (i) its opacity to extraction, and (ii) the ‘mere paraphrase’ implicature. Adapting Kayne’s (2007 fn. 9) brief discussion of be like quotatives, we assume that be like quotatives conceal a null SOMETHING heading a DP complement of be. We illustrate this structure in (40).4 (40) [TP Aaron [T’ was [PP something [PP like [RelP that [Rel Rel QUOTE ]]]]]] On this approach, the unavailability of wh-extraction with direct speech readings will be reminiscent of restrictions on wh-raising out of some-quantified DPs, as in (41) and (42). (41) ??Who did you see some picture of ? (42)
∗ What did you see something like ?
Similarly, the incompatibility of quote raising with be like might now be related to whatever excludes raising in counterpart sentences with say as in (43) and (44). (43)
∗ ‘Shut up,’ Aaron said something like.
(44)
∗ ‘Shut up,’ said Aaron something like.
Finally, the paraphrase meaning for quotes introduced by be like is also explained in terms of (40), which asserts that Aaron said something like the quote. As noted above, such sentences are pragmatically odd in a context in which the quote content is intended as a verbatim quote. From the perspective of the structure in (40), quotative 4 Haddican and Zweig () propose a slightly richer structure motivated in part by a comparison with similar quotative constructions in Dutch. The motivation for these additional assumptions is not directly relevant to the present discussion and we set aside these issues, here.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
be like sentences implicate a ‘mere paraphrase’ understanding of the reported quote in the same way that (45) implicates that cougars are merely similar to mountain lions. (45) A cougar is something like a mountain lion. In addition, the incompatibility of exactly and word for word with be like quotatives might now be related to the presence of the null SOMETHING. That is, the oddness of (35a, 35b), might be understood in the same way that (46) is odd, whereby the speaker at once weakens and strengthens the epistemic commitment to the comparison. (46) #A cougar is exactly something like a mountain lion. That the presence of a null SOMETHING (35a, 35b) and an overt something in (46) is implicated in their oddness is suggested by the fact that the same infelicity does not arise in sentences like (47) without an overt something. (47) A cougar is exactly like a mountain lion. To summarize the proposal so far, we have followed Kayne (2007) in assuming that be like quotative constructions conceal a null SOMETHING indefinite that takes the like-headed PP as its complement. We show that this approach accounts for a set of exceptional properties of be like as a quote introducer in English. It also suggests a fairly simple process of syntactic change: once quotes came to be available as descriptors of states, eventive be like interpretations fall out, with the additional enrichment of a null SOMETHING, which we discuss further below. On this approach, be like quotatives will be a species of manner deictic quotative construction, which have been peripheral to the formal literature on quotatives (Munro 1982; Güldemann 2002; Blain and Déchaine 2007; Etxepare 2010). Particularly reminiscent of be like from the perspective of the above proposal are hebben zoiets van quotative constructions available for some younger speakers with an overt ‘something’, iets, as in (48) (van Craenenbroeck 2002). (48) Dutch Jan had (zo)-iets van, “laat me gerust.”5 Jan had such-something of leave me alone a. ‘Jan thought something like, “Leave me alone.”’ b. ‘Jan said something like, “Leave me alone.”’ As discussed in Haddican and Zweig (2012), Dutch hebben zoiets van constructions share with English be like many of the syntactic properties discussed above. The presence of an overt ‘something’ in these similar Dutch constructions, therefore lends
5
The zo ‘such’ element often appears in such constructions but is not obligatory.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
plausibility to the present proposal that English be like constructions contain a silent SOMETHING, as proposed by Kayne (2007).
. The evolution of be like quotatives As noted, early descriptions of be like characterized it not as an introducer of direct speech but rather of reported thought exclusively; direct speech be like is reported to have emerged subsequently (Butters 1982; Tagliamonte and Hudson 1999; Buchstaller 2004). In this light, one possible analysis of the evolution of be like is that stative and eventive interpretations emerged independently. In particular, one possibility proposed in the diachronic literature is that stative, non-speech uses of be like emerged as a reanalysis of descriptions of states of individuals in sequences of be + focuser/discourse marker like + predicate adjectives or non-lexicalized sounds as in (49) and (50) respectively (Buchstaller 2004: 101–13). Direct speech uses of be like might then have evolved as a reanalysis of these reported thought interpretations. (49) I was like devastated. (50) She was like ‘ugh’. Nevertheless, in the preceding discussion, we have suggested two reasons for viewing diffusion of direct speech and reported thought guises of be like as involving a single abstract process of change. First, from the perspective of Kroch’s constant rate hypothesis, a unified approach explains the similar age slopes in the experimental data reviewed above and much of the corpus-based literature suggesting similar rates of diffusion for these two guises of be like. A second motivation for this approach is that it accounts for eventive interpretations of be like constructions in the absence of any overt material obviously responsible for the change of state interpretation. In particular, we have proposed that the eventive interpretation is produced not by an ambiguity in the meaning of be, but rather by a flexibility in its meaning; as shown by Rothstein (1999), the same be can create both eventive and stative readings. This approach entails that direct speech be like is produced by the grammar from the outset of change, albeit as a disfavoured variant, as it remains in most recent production studies (Tagliamonte and D’Arcy 2007; Buchstaller and D’Arcy 2009). An additional issue unaddressed in the discussion so far is the learner’s cue to posit a null SOMETHING. We suggest that the principal cue for this is in fact the approximate quote meaning; that is, that early reported thought quotes introduced by be like will have been readily interpretable as approximate quotes. We propose that this inferred something like meaning provided the cue for positing the null indefinite illustrated in (40). On this approach, a single innovation is fundamentally responsible for the emergence of contemporary be like quotatives, namely the novel use of quotes to describe states of individuals. No further syntactic changes are required to account for the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
William Haddican, Eytan Zweig, and Daniel Ezra Johnson
emergence of eventive ‘direct speech’ guises of be like on this approach, which come for free as a consequence of a Rothsteinean ‘repackaging’ mechanism independently motivated for ‘agentive’ be contexts. The development of a null SOMETHING motivated by the pragmatics of be like in direct speech contexts accounts for a range of additional syntactic properties of be like quotatives that distinguish it from other English verbs of saying.
. Conclusion Based on results from two judgment experiments and some recent spoken corpus studies, this chapter proposes a diachronic syntax for be like quotatives in English. We propose that evidence suggesting similar rates of diffusion for eventive (direct speech) and stative (non-speech/thought) interpretations of be like quotatives reflect a single abstract process of change. In the spirit of Rothstein’s (1999) proposal for adjectival predicates of copula be, we propose that eventive direct speech interpretations of be like quotatives are derived via a repackaging mechanism akin to those that make count readings out of mass nouns in the nominal domain. Our proposal relates be like to other manner deictic (‘thus’) quotatives cross-linguistically (van Craenenbroeck 2002; Güldemann 2002; Blain and Déchaine 2007). Future formal work might usefully explore the semantics of these under-studied constructions from a comparative perspective. The foregoing discussion also suggests two implications for the study of diachronic syntax more generally. First, we have demonstrated the utility of controlled experiments for understanding syntactic change in apparent time. Over the past decade, advances in experimental methodologies for linguistics have expanded the range of data available to researchers. Of particular interest to the study of linguistic change, techniques for web-based experiments enable researchers to acquire large datasets from a wide range of age groups. The recent emergence of Amazon Mechanical Turk as a web-based source of subjects in linguistic/psycholinguistic research also increasingly facilitates the collection of large amounts of data at relatively low cost (Mason and Suri 2012; Sprouse 2012). The importance of these developments for the study of formal diachronic syntax is that they make possible the study of change in apparent time using controlled techniques that avoid some of the problems inherent in corpus-based studies, including the scarcity of data in crucial environments. This allows for a more direct examination of hypotheses about the nature of language change than is possible from relying solely on limited corpora. For phenomena that allow for interpretation to be contextually biased reliably—as in the case of the ambiguity between reported thought and direct speech readings of be like—controlled techniques potentially provide a better approach to ambiguity than corpus data. In this context, it is worth noting that acceptability judgments, which the present results are based on, are a different kind of linguistic performance from the production
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Change in the syntax and semantics of be like quotatives
data that is recorded in spoken and written corpora, which many diachronic syntax studies are based on. A growing body of literature, however, suggests that acceptability judgments closely mirror relative probabilities of semantically equivalent competing forms in production (Bresnan 2007; Bader and Häussler 2010a, 2010b; Melnick, Jaeger, and Wasow 2011). These results suggest that acceptability judgments are useful for inferring change in apparent time in a way comparable to variation in corpus data. Future formal diachronic work on change in apparent time might therefore avail itself of these new techniques. A second consequence of the above discussion is to highlight the scarcity of formal research on processes of change in the aspectual domain, such as the shift between stative and eventive interpretations of be like. While the studies described in this chapter only focus on a specific change, they show that this is an area of inquiry that can produce profitable results. Also, if we are to take seriously the consensus in the semantic literature (as exemplified by Rothstein’s (1999) coercion approach to agentive be adopted here) that the eventive domain shares structure with the nominal domain, the current research opens up the related issue of understanding change in the semantics of count and mass entities. A rich comparative semantics literature has focused on differences among languages with different ways of marking count/mass and state/event distinctions, but relatively little literature has focused on the diachrony of these issues, and the question of what more general principles might govern such change. Whether the semantic changes involved with be like are representative of this wider field remains an open question, of course, but this chapter shows that this is a domain that may well be usefully addressed by future diachronic syntactic and semantic work.
Acknowledgements We are grateful to the participants in our experiments. Thanks also to two anonymous reviewers, Maryam Bakht, Theresa Biberauer, Jóhannes Gísli Jónsson, Richard Kayne, Tony Kroch, George Tsoulas, George Walkden, and audiences at LingEvid2010 and DiGS XII. This research is supported by ESRC grant number 061-25-0033 and by a grant from the Spanish Ministerio de Ciencia E Innovación (B. Fernández, PI) (FFI2008-00240/FILO).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian VERONIKA HEGEDŰS
. Introduction This chapter provides an analysis of the grammaticalization of a large group of Hungarian postpositions (Ps).1 More specifically, it gives an account for the development of Ps in possessive constructions. I will argue that the syntactic variation found in the structure of Old Hungarian Postpositional Phrases (PPs) is due to the fact that the elements involved are not fully grammaticalized Ps in Old Hungarian (ad 896– 1526). They are at an intermediate stage in the grammaticalization process of nouns becoming Ps. This intermediate step in the grammaticalization is in accordance with what Waters (2009) described as the prepositional cycle, and the elements that are not fully grammaticalized Ps can appear with a possessive marker because some of their original nominal properties are transparent at this intermediate stage. The historical origins of Hungarian Ps can be traced back to several different sources, but the most productive grammaticalization pattern seems to be the one whereby possessive structures change into PPs, with the possessee becoming (part of) the postposition. I will not discuss Ps that have a verbal source for two reasons: firstly, their P-status has been questioned in some places (cf. É. Kiss 1999), and secondly, they do not participate in the variation discussed in this chapter, so their grammaticalization patterns will not be dealt with here. The change under consideration is a clear case of grammaticalization. First of all, there is morphophonemic reduction (e.g. belen > ben/ban ‘in’; belől > ből/ból ‘out of ’). Secondly, the elements undergo semantic bleaching when their original nominal 1 The research presented in this chapter was supported by the project ‘Hungarian Generative Diachronic Syntax’ funded by the Hungarian Scientific Research Fund (OTKA No. ). I would like to thank the audience of the th Diachronic Generative Syntax Conference (Cambridge, UK, June ) and two anonymous reviewers for their useful comments.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
meaning is lost. Finally, they also undergo a categorial change from the lexical category of Ns to the (semi-functional) category of Ps. The chapter is organized as follows. Section 5.2 will introduce the historical data. I will first provide some information on the class of postpositions; then I will turn to the variation in Old Hungarian. Section 5.3 will turn to the structure of PPs and will introduce the category of AxialParts (as defined by Svenonius 2006) in the analysis of Old Hungarian, thus accounting for the remaining nominal properties of otherwise postpositional elements. The grammaticalization process of postpositions from nouns will be taken to go through an intermediate step when the elements are neither fully nominal nor fully postpositional, they are AxialParts. Section 5.4 will conclude.
. The data: Old Hungarian postpositions This section will introduce the relevant historical data. First, the origins of postpositions and their status in Old Hungarian will be discussed; then, I will turn to a variation in Old Hungarian PPs, which will call for the introduction of an intermediate step between noun and postposition on the grammaticalization path of adpositions. .. The class of postpositions Hungarian has postpositions and case suffixes, and while the two groups have been classified as belonging to different categories by traditional grammars, recently it has been shown that they behave in the same way syntactically. A unified syntactic analysis can be assigned to them, and both groups belong to the category of P (cf. É. Kiss 1999, 2002; Asbury 2008; Dékány 2012). Looking at the data from another point of view, we can divide the class of postpositions into two groups: (i) many of them take caseless complements, and (ii) some have complements that bear an oblique case.2 The two groups are illustrated in (1) and (2). (1a) is an example of a suffixal postposition, traditionally called ‘inessive case’, and (1b) is a locative postposition which takes a caseless complement. The two Ps only differ in that the suffix is monosyllabic and has alternating forms corresponding to vowel harmony. Members of the second group of P elements take oblique casemarked complements. The postpositional element át ‘through, across’ in (2) takes a complement that is itself marked for superessive case (which can be regarded as a PP, e.g. Asbury, Gehrke, and Hegedűs 2007). I will not discuss the historical development of this letter group of Ps; only those with a caseless complement, such as the ones in
2 É. Kiss (, ) argues that oblique case suffixes and postpositions with caseless complements belong to the same category: they are Ps; while the other postpositions are adverbs. Recently it has been claimed by various people that they all belong to the category of P (cf. Hegedűs ; Asbury, Gehrke, and Hegedűs ; Asbury ; Dékány ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
(1), will be dealt with since only these provide evidence for an intermediate stage in their grammaticalization in Old Hungarian. (1)
a. a ház-ban the house-ine ‘in the house’ b. a ház mögött the house behind.at ‘behind the house’
(2) a ház-on át the house-sup through ‘through the house’ The Ps in (1) do not show any grammatical agreement with their nominal complement and their complement does not bear a case marker in Modern Hungarian. This is an important point to make because it is Ps such as the one in (1b) that show variation in Old Hungarian: they can show 3rd person agreement and their complement can appear in the dative case. This variation is missing from Modern Hungarian and the pattern in (1) is the only one we find. Semantically, we can distinguish spatial and non-spatial Ps, but even some of the non-spatial ones go back to spatial primary meanings. Similarly to its English equivalent, the Hungarian P helyett ‘instead of ’ (illustrated in (3)) has a spatial origin, meaning ‘at the place of ’. However, it is no longer used as a spatial element; it has developed a new meaning. (3)
János helyett John instead.of ‘instead of John’
The P elements that take caseless complements used to be nouns and grammaticalized in possessive constructions. Historical grammars reconstruct a possessive construction with the order where the (unmarked) possessor is followed by the possessee and the possessee is case marked for locative, lative, or ablative case. These ancient case suffixes on the final element are the origins of the three-way partition we find in the synchronic system: the ancient locative suffix is found on (stative) locative Ps, and the lative and ablative suffixes are found on directional ones. Some of the postpositions have become suffixes: they are monosyllabic and they show vowel harmony with the noun they attach to, but even the ones that are not suffixal always strictly follow their complement. ... Possessives and postpositions Most of the oldest spatial Ps form groups of three, where one is locative, one lative, and one ablative. The examples in (4) illustrate
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
this tripartite system of spatial suffixal Ps in Modern Hungarian, while (5) has examples of non-suffixal Ps, showing the same three-way distinction. (4) a. a ház-ban the house-ine ‘in the house’ b. a ház-ba the house-ill ‘into the house’ c. a ház-ból the house-abl ‘out of the house’ (5)
a. a ház mögött the house behind.at ‘behind the house’ b. a ház mögé the house behind.to ‘(to) behind the house’ c. a ház mögül the house behind.from ‘from behind the house’
The origins of these elements are taken to be in unmarked possessive constructions like the ones in (6), which illustrate the ancient possessive nominal with locative endings preceded by the possessor. The fact that the old possessive construction was unmarked means that neither the possessor nor the possessee had grammatical markers of the possessive relation between the two, only their order was indicative (Zsilinszky 1991). (6) a. ház bele-n house inside-at ‘at the inside of the house’ b. ház möge-tt house back-at ‘at the back of the house’ The assumption in the historical grammars of Hungarian is that Proto-Hungarian had unmarked possessives, but possessive agreement on the possessee and optional dative marking of the possessor developed some time before Old Hungarian. By the time of the first written texts, we only find data with agreeing possessee and optional
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
dative case on the possessor (Zsilinszky 1991). The examples in (7) show the two variants. Either of the two constructions can appear when the possessor forms a constituent with the possessee, but dative marking is obligatory when the possessor is extracted from the phrase. (7) a. kezd-el [ysten aldomas-a- ual] begin-away god blessing-3sg-ins ‘(You) begin, with God’s blessing’ b. De az hews vala [ysten-nek angal-a] but the hero was god-dat angel-3sg ‘But the hero was God’s angel’
(JókK 33)
(JókK 15)
Similarly to the possessive construction, where the possessor often appears in dative right next to the possessee, elements that are generally considered to be postpositions by the Old Hungarian period can also bear an agreement marker and appear with a dative-marked complement. This is illustrated in (8). (8) a. ysten-nek felewl-e god-dat from-3sg ‘from God’
(JókK 29)
One important difference between agreeing Ps and possessive constructions is that agreement marking is obligatory in possessive constructions not only with dativemarked but also with caseless possessors. As for the PPs, agreement seems obligatory with dative-marked complements (although there are a couple of exceptional, nonagreeing examples in the corpus); however, when the complement of the P (that is, the equivalent of the possessor) is caseless, the postpositional element has no agreement marker. According to Zsilinszky (1991), the fact that PPs appear in possessive-like constructions in Old Hungarian is suggestive of their possessive origins; the ancient relationship between the two parts ‘lives on’. The traditional historical grammars describe the change as the possessee becoming a more grammatical element as it loses its original meaning and the possessive relationship becomes oblique. It cannot be completely oblique yet in Old Hungarian, so that the agreement and dative marking are possible. This, however, does not hold for all postpositional elements. Those that are already suffixal, or on the way to becoming suffixes, do not take part in the variation with respect to having or not having agreement marking depending on the form of the complement (dative or caseless, respectively). ... Suffixes There are postpositions which were on the way to becoming suffixes in the beginning of the written period of the language. One example is the old form balól/belől ‘out of (the inside)’, which was becoming -ból/-ből ‘out of (ablative case)’ in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
the Old Hungarian period. As (9) shows, we find both the longer and the shorter form even in the same text, and both show vowel harmony. (9) a. keze-belewl hand-out.of ‘out of/from his hand’
(JókK 60)
b. paris-balol Paris-out.of ‘out of/from Paris’
(JókK 28)
c. az lang-bol the fire-out.of ‘out of/from the fire’
(JókK 43)
The other two items developing from the same noun (bele ‘inside’) were also on the way to becoming suffixes in the oldest texts. (10) shows the locative form, which was already monosyllabic, while (11) illustrates the sometimes still disyllabic illative form, which did not obligatorily show vowel harmony with the stem (today it would be világ-ba ‘world-ill’ with a monosyllabic, harmonic suffix). (10) gimils-ben fruit-ine ‘in fruit’ (11)
vilag-bele world-ill ‘into the world’
(FS)
(FS)
These elements and those that are similarly suffixal or close to suffixal (with some variation in their syllable structure or harmonic properties) do not appear with agreement markers and their complement is never in the dative case. They are not marked for agreement with their complement and their complement is not case marked. I will argue below that the items showing variation are not actually Ps in Old Hungarian; the suffixal elements, however, are already fully grammaticalized Ps. .. Variation in Old Hungarian In Old Hungarian, many postpositional elements exhibited a variation typical of possessive DPs. The Ps could appear with a dative-marked complement and bearing an agreement marker or in the ‘regular’ P construction, that is, with a caseless complement and without an agreement marker. The fact that in the latter variant the P does not bear an agreement marker distinguishes the construction from real possessive DPs.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
Let me illustrate the variation by first using examples from the same text, the Jókai Codex (after 1372/copy from 1448). This is the first text where we find a considerable number of Ps both with respect to types and tokens, so that we can observe the extent of the variation. Sebestyén (2002) cites the following data: the codex contains 21,818 words; there are 39 different postpositions in 351 tokens. The data in (12)–(14) show that the two alternative constructions with Ps were really present at the same time, and since it is not possible to find any rule as to the use of one form or the other, we can assume that they were in free variation. The ‘regular’ use of the P is exemplified in the (a) examples and the ‘possessive’-like use of the P is shown in the (b) examples. (12)
a. keues bezed vtan little talk after ‘after some talk’ b. ez bezedek-nec vtan-a this talks-dat after-3sg ‘after these talks’
(13)
a. az baratok-nak aztal-a elewt the monks-dat table-3sg in.front.of ‘in front of the monks’ table’ b. baratok-nak elewtt-e monks-dat in.front.of-3sg ‘in front of monks’
(14) a. Sokak felet many above ‘above many’ b. menden-nek felett-e everything-dat above-3sg ‘above everything’
(JókK 122)
(JókK 25)
(Jók 84)
(JókK 84)
(JókK 114)
(JókK 79)
Sebestyén (2002) also lists the number of occurrences of the different postpositions, and we can observe that only Ps with a possessive origin alternate. Most of the frequent Ps are locative in meaning (directional ones are on average less frequent in this text), except for után ‘after’, which is the second most frequent one with 40 occurrences. They participate in the alternation to varying degrees, but roughly 19 of the Ps are in the ‘possessive-like’ structure. While the Jókai Codex is useful because it is the oldest codex and thus the oldest long text we have, we can observe the variation in other texts from this period as well. Zsilinszky (1992) gives a list of 50 Ps from late Old Hungarian and their occurrences, where we see the same variation based on several texts, which means that we find
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
many examples with dative-marked complements and agreement-marked Ps; as the examples in (15) also show. (15)
a. een zyvem-nek alatt-a I heart-dat under-3sg ‘under my heart’
(FestK 396)
b. Jordan vyzee-nek elvol-e Jordan water-dat over-3sg ‘over the river Jordan’
(JordK 176)
c. Abel-nek helyett-e Abel-dat place.in-3sg ‘instead of Abel’
(JordK IIIa)
d. a nep-nek közepett-e the people-dat middle.at-3sg ‘in the middle of the crowd’
(BécsiK 21)
e. viadal-nak miatt-a fight-dat because.of-3sg ‘because of the fight’
(BécsiK 19)
Since these Ps developed from nouns in possessive constructions, the variation has often been simply put down to a case of analogy in the Hungarian literature (Benkő 1980; Zsilinszky 1991). It has been claimed that as the ‘doubly’-marked possessive construction is very frequent in these old texts (arguably for stylistic reasons), there is an analogical push to use the same doubly-marked construction in PPs as well.3 However, if these elements are Ps, that is, if they are already grammaticalized elements rid of their nominal properties in this period, then the fact that they can have a dative-marked complement and can agree with that complement is not accounted for. Postpositions in Modern Hungarian do not participate in such variation, so either the properties of Ps have changed diachronically or these elements are not Ps. The fact that non-nominal Ps and suffixal Ps do not alternate in Old Hungarian seems to suggest the latter. The analogical push can only apply in Old Hungarian because the elements still have nominal features. Hypothetically, one could also say that there are two lexical Ps that look very similar; one takes a caseless complement and the other a dative-marked complement, and the second one bears agreement marking. However, this would duplicate things in the lexicon unnecessarily and would require us to assume two different grammaticalization 3 ‘Doubly’-marked refers to the fact that the possessive relation is marked both on the possessor and on the possessee: the possessor bears dative case, and the possessee agrees with it in person and number. The general consensus in descriptive historical grammars is that the genre of the old texts requires as much explicitness as possible, that is why grammatical relations are explicitly marked whenever possible (Benkő ; Zsilinszky ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
times for the two items. The original Ps developed in constructions where the possessor N was not in dative case and they were not agreeing with it. Similarly their complements have no case markers and they do not bear agreement morphemes in Modern Hungarian either, so we would have to assume that the second group of these hypothetical lexical entries developed later but disappeared relatively quickly. This seems to create more problems than it solves. Another possibility is that this could be a case of degrammaticalization. Under such a hypothesis, the seemingly previously grammaticalized P elements become nouns again; they gain nominal features. This does not seem likely, however, since they do not show any other nominal properties; they cannot be pluralized, nor do they have determiners or modifiers. Degrammaticalization is a process that is theoretically problematic, and the data does not support such an analysis, so it is not a path we will take either. One more thing we can say about this kind of variation is that it is not present in Modern Hungarian. While there are some cases where it is possible to have the P agree with a dative-marked complement, in all those cases it is obligatorily extracted from the PP (cf. É. Kiss 2002), and arguably a different base-generated structure should be assigned to the construction from that of regular PPs. The examples in (16) are impossible as constituents, and (17) is only grammatical if the agreement-marked P does not form a phrase with its complement (on the surface). mellett-e (16) a. ∗ a ház-nak the house-dat beside-3sg ‘beside the house’ b. ∗ az autó-nak után-a the car-dat after-3sg ‘after the car’ (17)
János [után-a]i futott [az autó-nak [utána]i ]. John after-3sg ran the car-dat ‘John ran after the car.’
I will argue in the next section that instead of the above-listed problematic explanations, an analysis that attributes a special, intermediate status to the Old Hungarian P-like elements on the grammaticalization path is viable.
. AxialParts .. The structure of PPs As recent studies on adpositions and their possible phrasal extensions have shown, the structure of PPs can be rather complex. We need to minimally distinguish between locative and directional Ps in the structure (cf. van Riemsdijk 1990; Koopman 2000;
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
Svenonius 2010; den Dikken 2010). The grammaticalization discussed here results in Place and Path heads; the minimal structure is illustrated in (18).
pP
(18)
p
PathP Path PlaceP Place
DP
Svenonius (2006) argues that there is an additional projection in the extended PP hosting a group of categorially ambiguous elements that exhibit both nominal and adpositional properties. He names the projection AxialPart since the elements that occur there mostly refer to regions or ‘axial’ parts of objects. Svenonius (2006) also shows that we find such AxialParts in various languages. Their syntactic properties classify them partly with nouns and partly with Ps, and their syntactic projection is in between those two as well. His English example highlights the difference between the properties of front in the two sentences in (19). In the first example front is an AxialPart, while (19b) is a regular possessive phrase with front as a noun. (19)
a. There was a kangaroo in front of the car. b. There was a kangaroo in the front of the car.
(AxPart) (N)
The structure Svenonius (2006) renders to PPs involving an AxialPart is the one in (20), with AxPrtP as an intermediate projection between DP/KP and PlaceP. (20)
PlaceP Place in
AxPrtP AxPrt front
KP K
DP
of
the car
According to Svenonius (2006), English AxialParts cannot be pluralized, modified, replaced by a pro-form, or moved away, while nouns can. AxialPart elements, however, have some nominal features, but these features can be different in various languages. Another observation is that AxialParts can be prepositional (e.g. Persian, Tzeltal) or postpositional (e.g. Korean) in a given language.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
This intermediate category and its properties are relevant here, because the semantic class of elements Svenonius (2006) argues to be AxialParts is exactly the same as the one involved in the observed variation in Old Hungarian. It was ‘axial’ nouns that started to turn into postpositions in Proto-Hungarian, and these grammaticalizing nouns seem to have maintained some nominal properties in the early written period. The original meaning of some of these elements were ‘back’, ‘front’, ‘chest’, ‘bottom’, ‘top’, and similar orientational meanings. Similarly to English, Hungarian AxialParts cannot be pluralized, modified, replaced, or extracted either, but they can bear a nominal agreement marker. This is the property that allows them to appear in constructions similar to simple possessive phrases. Asbury (2008) argued for the presence of AxPrtP in Modern Hungarian PPs, partially because of their nominal origin, although the postpositions are never in AxPrt in her analysis. My analysis claims that AxPrtP is present and filled in Old Hungarian PPs; this is how their marginally nominal nature is accounted for. However, since these elements are Ps in Modern Hungarian, AxPrtP is not filled in Modern Hungarian PPs in the case of the elements under discussion, which prevents them from agreeing with their nominal complements. .. N > AxialPart > P My proposal is that we are dealing with a grammaticalization process which is in a transitional state in the early texts (and in some cases it lasts longer). In Old Hungarian, some of the ‘postpositions’ were not actually Ps yet, they were AxialParts, that is, they belonged to an in-between category between nouns and Ps. AxialParts in Old Hungarian do not have determiners, they cannot be modified, and they have no plural form. The only nominal feature they seem to have is an agreement feature, which allows them to agree with their dative-marked complement. They generally do not agree in number with their complement.4 This is shown in (21) where the complement is plural but the AxialPart still shows singular agreement. (21)
barat-ok-nak elewtt-e monk-pl-dat in.front.of-3sg ‘in front of monks’
(JókK 84)
Historically the first step of the grammaticalization process is when N becomes AxialPart. The Old Hungarian elements that take part in the variation are AxialParts. 4 The lack of number agreement is not surprising if we take into account that lexical possessors and possessees do not agree in number in Modern Hungarian either, contrary to pronominal possessors (cf. den Dikken ; É. Kiss ). Old Hungarian seems to differ to some extent from Modern Hungarian in this respect, since plural agreement with a plural possessor was possible within the possessive DP in Old Hungarian. The exact structure of Old Hungarian possessives and especially their agreement pattern is not fully understood yet (but cf. Egedi for the structure of the left periphery of Hungarian DPs).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
In these constructions the locative/lative/ablative case suffix is the P head. The structure in (22) illustrates such a construction.5 The AxPrt head and the Place head are both lexicalized at this stage of the grammaticalization.
PlaceP
(22)
Place -tte
AxPrtP.3sg AxPrt fele-
DP mendennek
By this stage, the AxialPart elements have lost their nominal reference; they do not refer to body parts or orientations by themselves any more. Together with the suffixal P, they have a locative or directional meaning, in this case ‘above’. In the second step, the morphological border between the case suffix (that is, the P element) and the AxialPart element becomes oblique, and the whole unit gets reanalysed as the P head. Grammaticalization thus proceeds from case-marked possessive nouns to case-marked AxParts to simple P heads. Those elements that were (becoming) suffixal in Old Hungarian did not take part in the variation with agreement marking and dative complements because they were already Ps, generated in the Place/Path head. However, during the Old Hungarian period and later in the Middle Hungarian period, most of the items which showed variation in the earlier texts were reanalysed as P heads as well and the variation slowly disappeared. In Modern Hungarian, the item felett ‘above’ (from the previous examples) is base-generated in Place. There is no alternative with an agreement marking on the P with a nominal complement, and its complement is always caseless. (23) a ház felett the house above ‘above the house’ (24)
PlaceP Place
DP
felett
a ház
This is a grammaticalization process which results in a lexical item losing its nominal properties and becoming a semi-functional element, an adposition. There 5 The structures I assume for Hungarian are head-initial as well, despite the fact that the P item ends up as a postposition. The surface order can be derived by movement or morphological merger.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Veronika Hegedűs
is morphophonological reduction (when becoming suffixal, Ps are monosyllabic), semantic bleaching (the nominal reference is lost) and category change (N > AxPrt > P) involved in the process. The change is similar to other syntactic changes where lexical heads are reanalysed and become functional heads generated under a functional node (cf. Roberts and Roussou 2003; van Gelderen 2004), although the lexical or functional nature of P is not a settled issue yet, and there seem to be lexical Ps. For Proto-Hungarian, descriptive historical grammars suggest that we should reconstruct a possessive structure under PP, where the P elements are the ancient locative, lative, and ablative suffixes and there is a possessee noun, which bears the case marker. In Old Hungarian, we find structures like (22), where the original possessee is already on the way to losing its nominal properties completely. As a next step of the process, AxPrt is reanalysed as part of the Place or Path head, as in (24). In those structures where the AxPrtP is filled, AxialPart can partially agree with its complement, since it has a person feature. The 3rd person agreement percolates up onto P. Those Ps that are becoming suffixal at this point are already Ps, generated in the Place or Path head, hence, they do not take part in the variation. Similarly, in Modern Hungarian, the already grammaticalized P elements do not appear in possessive-like structures. This analysis takes into account both the nominal nature and the lack thereof of Old Hungarian alternating postpositions. The fact that the analogical push of the very frequent ‘doubly’-marked possessive constructions can apply to them at all is due to the fact that they still have some nominal properties. The frequency of the alternation might be due to the frequent use of double marking in possessives as well. Furthermore, the fact that the slightly nominal nature of these grammaticalizing items is present throughout the Old Hungarian period might be explained by the strong presence of the alternation in these formal written texts, where the agreement-marked forms keep the remnants of their nominal origins conserved.
. Conclusions The analysis proposed in this chapter shows that the changes observed in the Hungarian PP fit in with the analyses proposed for other languages, and that the grammaticalization of P elements is parallel to that of other functional material in languages. The variation in the Old Hungarian data between possessive-marked and regular ‘unmarked’ postpositional elements can be explained by assuming a semipostpositional head in the structure as an intermediate step in the grammaticalization from N to P. It was argued that this intermediate step in the grammaticalization process is when the elements are AxialPart heads, a category that has been proposed to exist in various languages hosting exactly the kind of elements that are becoming Ps in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The grammaticalization of postpositions in Old Hungarian
the beginning of the written period of Hungarian. The residual nominal features of the AxialPart account for the possibility of a construction with PPs that is similar to possessive constructions. It can also be added that the change from AxialPart to Place/Path takes place at different times with different items. Some of the postpositional elements are already on the way to becoming suffixal in the first texts from the thirteenth century, which I took as evidence that they are Ps. Those elements which take part in the illustrated variation seem to have some nominal features for much longer, some even after the end of the Old Hungarian period.
Old Hungarian sources BécsiK = Vienna Codex (15th c.) Published as: Mészöly, Gedeon (1916) Bécsi codex. Budapest: MTA. Új Nyelvemléktár I. kötet. FestK = Festetics Codex (1492–1494) Published as: Festetics-kódex, 1494 előtt (1996) Közzéteszi, a bevezetést és a jegyzeteket írta: N. Abaffy Csilla, Budapest: Argumentum, Magyar Nyelvtudományi Társaság. FS = Funeral Speech (around 1195). JókK = Jókai Codex (after 1372/copy from 1448) Published as: P. Balázs, János (1981) Jókai–kódex XIV–XV. század. Budapest: Akadémiai Kiadó. JordK = Jordánszky Codex (1516/1519) Published as: A Jordánszky-kódex bibliafordítása (1888) Sajtó alá rendezte és kinyomtatta: Toldy Ferenc, az eredetivel összevetette, és előszóval ellátta: Volf György. Régi magyar nyelvemlékek 5. Buda.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian K ATA L I N É . K I S S
. Introduction This chapter1 analyses the changes having taken place in the syntax of negation in twelfth–fifteenth-century Hungarian. It points out a change in the position of the negative particle, and shows it to be related to the change of basic word order from ‘SOV’ to ‘TopFocVX∗ ’. The central topic of the chapter is a negative cycle induced by the morphological fusion of the negative particle with different types of indefinites in the scope of negation. The opaqueness of the resulting morphological complexes necessitated the reintroduction of negation into sentences with indefinites, and led to the reinterpretation of negative indefinites as expressions with no negative force, participating in negative concord. The newly introduced negative particle, though morphologically identical with the negative particle that was input to the fusion with indefinites, assumed a different syntactic status in the new ‘TopFocVX∗ ’ sentence structure; it acted as a functional head, eliciting verb movement. The chapter is organized as follows: Section 6.2 provides a background by surveying the syntax of negation in present-day Hungarian. Section 6.3 describes the structural positions of the negative particle in Old Hungarian, and Section 6.4 analyses the syntax of Old Hungarian negative indefinite noun phrases and pronouns. Both sections point out an archaic pattern surviving from Proto-Hungarian, as well as a new variant. Section 6.5 attempts to reconstruct the diachronic process emerging from the declining and novel patterns of negation in twelfth–fifteenth-century Hungarian documents. 1 This chapter was written with the support of grants and of OTKA, the Hungarian National Research Foundation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
. Background: Negation in Modern Hungarian Although this chapter focuses on the history of negation in twelfth–fifteenth-century Hungarian, the directions of changes are clearer if we look at them from the perspective of the present-day language.2 Negation in Modern Hungarian is encoded by the negative particle nem, assumed to head a NegP. NegP has two possible merge-in sites. In the case of predicate negation, it subsumes TP, and elicits verb movement across SpecTP, occupied by a predicative complement, most often a telicizing particle, semantically incorported into the verb.3 (The Hungarian sentence has no distinguished subject position in the left periphery; the subject of (1a, 1b) is in Spec,TopP, a position not available for a non-specific or a universally quantified subject.) Compare the affirmative sentence in (1a), and its negated counterpart in (1b): (1)
a. János meg látogatta Marit. John prt visited Mary-acc ‘John visited Mary.’ b. János nem látogatta meg tV Marit. Mary-acc John not visited prt ‘John did not visit Mary.’
The Hungarian sentence often also includes a focus projection above TP, which also elicits verb movement across the verbal particle in SpecTP (2a). The focus projection can also be negated, i.e. it can also be subsumed by a NegP (2b). (As shown in (2b), V-movement elicited by the presence of negation and/or focus is not cyclic; it stops in the head position immediately preceding TP. It is unclear whether this is the head of a separate functional projection (FP), or is the head of the lowest operator projection (NegP or FocP).) (2) a. János TEGNAP látogatta meg tV Marit. Mary-acc John yesterday visited prt ‘It was yesterday that John visited Mary.’ b. János nem TEGNAP látogatta meg tV Marit. Mary-acc John not yesterday visited prt ‘It wasn’t yesterday that John visited Mary.’ The primary predicate and the focus (an identificational predicate) can also be negated simultaneously:
2
For analyses of Hungarian sentence structure, see É. Kiss (, a).
3 For further details, see Surányi (a, b), Olsvay (), and É. Kiss (, ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(3)
Katalin É. Kiss a. János nem TEGNAP nem látogatta meg Marit. ‘It wasn’t yesterday that John didn’t visit Mary.’
TopP
b.
János
NegP Neg nem
FocP TEGNAP
NegP
Neg nem
FP F látogatta
TP meg
T’ T látogatta
vP … Marit…
Hungarian is a negative concord language. Universal pronouns with scope over negation and existential pronouns in the scope of negation have a negative version beginning with se/so-, which is licensed by an overt negative particle. Indefinite lexical noun phrases in the scope of negation are obligatorily supplied with the minimizer sem. (4) Soha senki nem késett el egy óráról sem. never nobody not was.late prt one class-from not.even ‘Nobody has ever been late for even one class.’
. The position of the negative particle in Old Hungarian In the twelfth–fifteenth-century Old Hungarian texts examined (among them Halotti beszéd és könyörgés ‘Funeral Sermon and Prayer’, a 50-clause sermon from 1193–5; Jókai Codex, a fifteenth-century copy of a fourteenth-century translation of the Legend of St Francis; and the Bécsi ‘Wiener’, Müncheni ‘Münchener’, and Apor Codices, containing fifteenth-century copies of various parts of the so-called Hussite Bible, translated after 1416), the majority of negative sentences represent predicate negation. Focus negation is rare, but so is structural focus itself. Here is an example of
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
focus negation, with the negative particle in pre-focus position as in present-day Hungarian: (5)
nem payzual fegyuerkedet de zent kerestnek yegyuel not shield-with armour-refl-past-3sg but holy cross’s sign-with ‘It wasn’t a shield that he armoured himself with but the sign of the holy cross.’ (Jókai Codex p. 147)
Sentences with predicate negation belong to two word-order types, which co-occur in the same texts. The negative particle may intervene between the verbal particle and the V: i. . . . PRT nem V . . . (6) a. hogy ezt senkynek meg-nem yelentene that this-acc nobody-dat prt-not report-cond-3sg ‘that he would not report this to anybody’
(Jókai 27)
b. ha meg nem kayaltandod kegyetlennek ew kegyetlensegett if prt not shout-fut-2sg cruel his cruelty-acc ‘if you do not declare his cruelty to be cruel’ (Jókai 95) Alternatively, the negated verb precedes the verbal particle. In this case, the verb and the particle are not necessarily adjacent: ii. . . . nem V . . . PRT . . . (7) a. Te nemynemew kewekrel . . . nem fyzettel telyesseguel meg you some stones-subl not paid completely prt ‘You have not paid completely for some stones’ b. hogy en lelkem semegyben nem zegyengett meg engemett that my soul nothing-in not shamed prt me ‘that my soul has not shamed me in anything’
(Jókai 7)
(Jókai 48)
Of the two patterns, pattern (i) is the more archaic variant. It represented the majority pattern in early Old Hungarian, and it has been losing ground to pattern (ii) ever since (cf. Gugán 2008). At present, pattern (i) is productively used only in Csángó, the most archaic dialect of Hungarian, and in two subordinate clause types of Standard Hungarian: in amíg ‘as long as/until’ clauses, and in conditional clauses in combination with hacsak, meaning ‘unless’. It is presumably a relic of the SOV ProtoHungarian period. Jäger (2008) derives a similar pattern in Old High German by the rightward movement of the VP-final V to a right-hand side Neg head. I assume that in sentences displaying the ‘. . . PRT nem V . . .’ order, the negative particle is adjoined to the verb. Pattern (ii), on the other hand, involves a left-
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
peripheral negative head attracting the verb across the verbal particle. Since the basic word order of Hungarian had shifted to TopFocVX∗ by the time of the first surviving coherent Hungarian texts (cf. É. Kiss 2013), it seems likely that Old Hungarian speakers analysed both patterns in the framework of a head-initial verb phrase preceded by left-peripheral functional projections. This hypothesis is supported by the distribution of the two word-order patterns, which is related to the presence versus absence of a negative pronoun or negative indefinite (a se-expression) in the left periphery. In the Jókai Codex, 60 of sentences displaying the ‘. . . PRT nem V . . .’ order contain a seexpression in post-topic position, at the left edge of the comment, but only 13 of sentences displaying the ‘. . . nem V . . . PRT . . .’ order do so. This suggests that in the emerging TopFocVX∗ sentence structure of Old Hungarian, with separate thematic and functional domains, operators were expected to precede and c-command their scope. In sentences with a se-expression in the left periphery, the se-expression acted as the scope marker of negation. In sentences with no se-expression, the scope principle, requiring that the scope of negation be preceded and c-commanded by an overt negative constituent, elicited the preposing of the negated V. First it may have been the negated verb that moved; then the negative particle must have been reanalysed as a head generated in the left periphery, attracting the V. This is the structure I hypothesize for sentences displaying the ‘. . . PRT nem V . . .’ order:
CP
(8)
C hogy
TopP ezt
NegP senkinek
Neg’ Neg Ø
TP meg
T’
T [v nem jelentene]
vP . . . tv ...
that this-acc nobody-to prt not report-cond.3sg ‘that he would not report this to anybody’ If the NegP projection is not lexicalized by a se-pronoun, the negated V is preposed into the Neg head:
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian (9)
TopP Te
TopP nemynemew kewekrel
NegP
Neg [nem fyzettel]
TP
telyesseguel
TP
meg
T’ T tv
vP
you some stones not paid completely prt ‘You have not paid completely for some stones’ In the minority of Old Hungarian sentences that display a ‘. . . PRT nem V. . .’ order but contain no se-expression, I assume a phonologically empty NegP, whose head position is filled by the negated verb in LF. Ürögdi (2009), analysing the presentday relic of this construction occurring in amíg-clauses, e.g. that in (10a), argues for a similar structure, with nem LF-moved into the left periphery. The LF attributed to (10a) reflects the fact that negation must have scope over the adverb hirtelen ‘suddenly’—otherwise the need of the adverb amíg ‘as long as’ for a complement clause denoting a durative eventuality is not satisfied. (10) a. Olvastam, amíg hirtelen ki nem aludt a fény. read-I as.long.as suddenly out not went the light ‘I was reading as long as it wasn’t the case that suddenly the light went out.’ LF: b. Olvastam [CP amíg nem [TP hirtelen [TP ki tnem aludt a fény]]] Verbal particle + V combinations display the same word order as predicative nominal + copula combinations both in Modern Hungarian and in Old Hungarian, with the particle/predicative nominal in SpecTP, and the verb/copula in T. Interestingly, whereas the preposing of the negated verb across the particle still represents a minority pattern in early Old Hungarian, the preposing of the negated copula across the nominal predicate nearly always takes place—even in the presence of se-expressions, e.g.: (11)
sonha nem lez zomoro tV never not be-fut.3sg sad ‘he will never be sad’
(Jókai 55)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
Kádár (2006) argues that the Hungarian copula is not a verb; it is an expletive generated in T, providing lexical support for inflection. Apparently, overt T-to-Neg became general earlier than overt [V+T]-to-Neg in the history of Hungarian.
. Se-expressions in Old Hungarian Though Modern Hungarian is a strict negative concord language, in which negative polarity items, the so-called se-pronouns, require the presence of a negative particle, in early Old Hungarian texts we find negative sentences in which the se-expression is not accompanied by a negative particle. These sentences are so sharply unacceptable for present-day speakers that historical linguists generally regard them as scribes’ mistakes due to Latin interference. However, there is evidence that in ProtoHungarian, and, to some extent, in early Old Hungarian, as well, se-pronouns had negative force. First of all, there are fossilized expressions with a se-expression conveying negation, e.g.: (12)
semmit-tevés, nothing.acc-doing ‘idleness’
semmit-mondó nothing.acc-saying ‘meaningless’
semmire-kellő, nothing.subl-needed ‘good-for-nothing’
semmibe vesz nothing-ill take ‘ignore’
Modern Hungarian also has a productive relic of the pre-negative-concord period of the language; there is a finite negative construction in which a se-expression occurs without a negative particle. The underlying construction from which this pattern derives contains an indefinite in the scope of negation, obligatorily accompanied by the minimizer sem. (13)
a. Nem indult el egy ember sem. not left prt one man minimizer ‘No man left.’
When such an indefinite supplied with the minimizer sem is preposed into focus position, sem lands right in front of the position of the negative particle. In this construction the negative particle is not spelled out. The reason must be that sem appears in the same linear positon where the negative particle is expected, hence present-day speakers reanalyse it as the negative particle, an allomorph of nem. (13)
b. Egy ember sem [Neg Ø] indult el. left prt one man minimizer ‘No man left.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
If the occasional lack of the negative particle in the presence of a se-expression in Old Hungarian were a mistake of the scribe caused by Latin interference, the lack of nem would be random; however, it is systematic to a large extent. Namely, (i) in the non-finite clauses of the Jókai Codex, the negative particle is never spelled out in the presence of a se-expression. Non-finite clauses have archaic features in Old Hungarian; for example, they often retain the strictly SOV order with a morphologically caseless object of Proto-Hungarian. This is never attested in finite clauses any more. The negative pattern they have preserved, in which negation is expressed by a se-phrase without the particle nem, is also likely to be a Proto-Hungarian archaism. Cf.: (14) a. bodog ferencz monda magat alazatost lenny semmy blessed Francis said himself-acc humbly be-inf nothing-Ø tudonak know-participle-dat4 ‘Blessed Francis said himself to be knowing nothing’ (Jókai 95) b. mendenestewlfoguan semegyben meg-haraguuan altogether nothing-in prt-being.angry ‘not being angry for anything at all’
(Jókai 21)
c. ew kerelmenek sem egy haznalattyat aloytuan his request-gen not one use-acc assuming ‘assuming no use of his request’
(Jókai 153)
(ii) In finite clauses, the presence or lack of the negative particle is related to the lexical choice of the se-phrase. Semmi ‘nothing’, semegyben ‘in nothing’, semegyképpen ‘in no way’, semegyik ‘none’, as well as lexical noun phrases modified by sem-egy ‘not one [no]’ can occur either without nem (15) or with nem (16): (15)
a. es azokes semmyre valanak yok and they-too nothing-subl were good-pl ‘and they, too, were good for nothing’
(Jókai 86)
b. Semmy ygazb ezeknel nothing true-comparat these-adess ‘Nothing is more true than these’
(Jókai 93)
c. semegyk mendenestewlfoguan indoltatyk-uala none altogether leave.3sg-past ‘none of them at all was leaving’
4
(Jókai 139)
The dative is a structural case marking a tenseless predicate. See Ürögdi ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
(16) a. ky kewnuek semmyre yok nem leznek which books nothing-subl good-pl not be-fut.3pl ‘which books will not be good for anything’ b. Semegykeppen nem lehett hug . . . not-one-manner-in not was.possible that ‘It was not possible in any way that . . .’
(Jókai 109)
(Jókai 3)
c. hogy mendenestewlfoguan semmy meg nem yelennek that altogether nothing prt not appear-cond-3sg ‘that nothing at all would appear’ (Jókai 66) The se-words senki ‘nobody’ and soha ‘never’, on the other hand, always require the presence of a negative particle: (17)
a. De meg nyttuan az kapput senkett nem lele but prt opening the door nobody-acc not found ‘But opening the door, he did not find anybody’
(Jókai 17)
b. kytt sonha nem latam-uala ez vilagban whom never not see-pfv-1sg-past this world-in ‘whom I had never seen in this world’
(Jókai 47)
(iii) In negative subjunctive, imperative, and optative clauses, the ne allomorph of the negative particle is used. Ne is never omitted in the company of a se-expression: (18)
Hogy semegy frater az zerzetben hust ne ennek that no brother the convent-in meat-acc not eat-cond.3sg ‘That no brother should eat any meat in the convent’
The fact that a ne accompanying a se-expression is always spelled out must be due to the fact that, in addition to the negative feature it shares with the se-expression, it also carries a modal feature. The fact that semegy ‘no’, semegyik ‘[+specific] none’, and semmi ‘nothing’ can occur without the negative particle, whereas senki ‘nobody’ and soha ‘never’ always require the presence of nem/ne in Old Hungarian is obviously related to their morphological make-up. Se-words have a complex morphological structure, involving the particle sem, and the numeral egy ‘one’ or its specific counterpart egyik, or an indefinite pronoun (mi ‘what’, ki ‘who’, ha ‘when’). Sem is also a complex morpheme, the fusion of es, a particle with various (additive, distributive, and emphatic) functions, and the negative particle nem. These ingredients are still transparent in the following example from 1193–5. (The vowel of the negative particle, spelled as u, may have been pronounced as [ü].)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian (19)
isa es num igg ember mulchotia ez vermut surely even not one man avoid-can this pit-acc ‘surely, no [not even one] man can avoid this pit’ (Funeral Sermon and Prayer, 1193–5)
Es has the allomorph s in present-day Hungarian, and it might have had it in Old Hungarian, as well. Old Hungarian did not tolerate word-initial consonant clusters, so a fused snum/snem predictably developed into sum/sem. As a next step, sem fused with the indefinite pronouns. Although the preposing of indefinite pronouns into the left periphery was not obligatory, as shown by the example in (20), it was very general. They may have been preposed via focus movement. (20) de az egyebekrewl nem tudok mytt but the rest-about not know-I what-acc ‘but about the rest, I don’t know anything’
(Jókai 145)
In view of these, the se-expressions of Old Hungarian had the following underlying morphological structure: (21)
semegy: semegyik: semmi: senki: soha:
[es+nem]+egy [es+nem]+egyik [es+nem]+mi [es+nem]+ki [es+nem]+ha
The se-expressions that could convey negation in early Old Hungarian were those in which the particle sem, resulting from the fusion of es+nem, was still transparent. In the case of senki, and, especially, in the case of sonha (Modern Hungarian soha), the fusion of the constituent morphemes was so advanced that sem, let alone the underlying nem, were not recognizable any longer. Senki only preserved the vowel of nem. In the case of sonha, both the vowel of sem was assimilated to the back vowel of ha, and its m was affected by the adjacent h as regards its place of articulation (before disappearing completely). Mary’s Lament from 1300 preserved an earlier form of sonha/soha: (22) qui sumha nym hyul which never not ceases ‘which never ceases’ Apparently, the more opaque a morpheme complex including the negative particle was, the less it could preserve its negative force. The morphologically opaqe senki and soha obligatorily needed the presence of a separate negative particle. For the morphologically more transparent semmi, semegy, semegyik, reinforcement by a preverbal negative particle was still optional in the Old Hungarian period under investigation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
The negative particle also fused with the dual connective es . . . es . . . ‘both . . . and . . .’, yielding sem . . . sem . . . ‘neither . . . nor . . .’. The insertion of an additional negative particle was optional in coordinate clauses introduced by sem . . . sem . . . , as shown by the following example of Jókai Codex, where the second coordinate clause contains an additional nem, and the first one does not. (23) Tehat zent ferenc sem magat valta az heylbelewl sem so Saint Francis neither himself-acc shifted that place-from nor arczayat le nem hayta menbewl face-his-acc down not turned heaven-from ‘So Saint Francis neither moved himself from that place, nor turned his face down from heaven.’ (Jókai 16)
. A negative cycle in th–th-century Hungarian Interestingly, the negative construction that represented the initial stage of the changes having taken place in Old Hungarian has been shown by Gugán (2012) to be the output of a former negative cycle. Negative cycles, beginning with the morphological/phonological and semantic weakening of the negative marker at stage 1, to be followed by its subsequent reinforcement by a negative adverbial element at stage 2, by the degradation of the original negative marker into an optional element at stage 3, and by its eventual disappearance at stage 4, have been observed in a large number of languages from various language families—see, among others, Jespersen (1917), Croft (1991), van Kemenade (2000), Wallage (2008), Biberauer’s, Hoeksema’s, and van der Auwera’s chapters in van Gelderen (2009), Chapter 8 of van Gelderen (2011), and the studies in Larrivée and Ingham (2011). Gugán argues that the Hungarian negative particle nem is also the result of a negative cycle having taken place in ProtoHungarian. Most Uralic languages have a negative auxiliary, which also existed in Proto-Ugric in the form ∗ e ∼ä ∼a. In Proto-Hungarian, however, its negative force underwent weakening, and an indefinite pronominal element reconstructed as nëm was introduced to reinforce it (Sipos 1991: 395). Eventually, the negative auxiliary disappeared (except in yes/no questions, where it has survived as an interrogative particle), and the pronoun assumed the role of negative operator. The negative particle nem is the descendant of nëm , hence it is cognate with the indefinite pronouns and proadverbs né-mi ‘some-what’ (originally meaning ‘something’, today meaning ‘some’), né-hol ‘somewhere’, né-ha ‘somewhen’, né-mikor ‘sometime’, and né-hány ‘some-many’. As Gugán (2012) points out, a similar process has been reported from Old High German and Middle High German, where the indefinite pronouns uuiht and iht, respectively, were introduced to strengthen the negative particle, and came to replace it (Jäger 2008: 118). The negative particle ik of certain Upper-German (Bavarian) dialects is a present-day descendant of this indefinite pronoun.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
In the late Proto-Hungarian period, the cycle began anew. As a first step (resulting in stage 2 of the new cycle), negated indefinites were strengthened by the emphatic/additive/distributive particle es, and the numeral egy, egyik ‘one’.5 (Egy is identical with today’s indefinite article, however, in the Old Hungarian period examined, there was no indefinite article yet in the language.) Recall es num igg ember ‘even not one man’, an example from 1193–5, quoted in (19). Negation was strengthened by es also in the case of indefinite pronouns in the scope of negation. In the third stage of the cycle, the morphological fusion of es+nem, and, especially, the morphological fusion of es+nem+pronoun complexes lead to the semantic weakening of negation, and created a need for further strengthening. This was attained by the adjunction of another negative particle to the verb. The reintroduction of the negative particle was first optional. The se-pronouns soha and senki, whose morphological structure had become completely opaque owing to word-internal phonological processes, lost their negative force and came to require an additional negative particle prior to the Old Hungarian period. In the case of the rest of the se-expressions, the additional, V-adjoined negative particle was still optional in the first Old Hungarian documents. According to the evidence of fourteenth–fifteenth-century codices, the pattern without a reinforcing negative particle was becoming less and less common, and by the end of the fifteenth century it had disappeared completely. In stage 4 of the negative cycle, Hungarian became a strict negative concord language, where negation is conveyed by a negative particle, and se-expressions are negative polarity items. The process of reinforcing negation—first optionally, later obligatorily—by the addition of a negative particle went on parallel with the syntactic restructuring of negative sentences, as a result of which the negative particle assumed head status eliciting verb movement. (The process of the negative particle becoming a high functional head has been identified as a key element in negative cycles by van Kemenade 2000 and van Gelderen 2011.) As was discussed in connection with (6) and (8), in the archaic type of negative sentences, the se-expression occupies the specifier of a left-peripheral NegP. The negative particle, if any, behaves like an adverb; it is leftadjoined to the V, and appears sandwiched between the verbal particle and the verb. In the emerging new pattern, discussed in connection with (7) and (9), Neg attracts the negated verb, which moves forward crossing the verbal particle and the elements adjoined to TP. If the sentence also contains a se-phrase, the negated verb is adjacent to it.
5 The numeral one is frequently employed as a strengthener. In Latin, both the negative particle non derives from the earlier negative marker ne merged with oinum ‘one’, and the negative pronoun nullus derives from ne merged with oinolos ‘one+ diminutive suffix’ (Wackernagel : ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
(24)
TopP én lelkem
NegP
Spec semegyben
Neg’ Neg [nem szégyengett]
TP
Spec meg
my soul nothing-in not shamed prt ‘my soul hasn’t shamed me in anything’
T’ T tv
vP
…engemet… me
Since the Old Hungarian negative cycle reached its final stage, only minor changes have taken place in the syntax of negation. Until the end of the fourteenth century, sentences could only contain a single se-expression, confined to the left periphery.6 From the fifteenth century on, we also find postverbal se-phrases, which is evidence of their analysis as negative polarity items. (25) ninč te bèz˙edidbèn sem eg-megfèdd˙es isn’t your speech-pl-2sg-in not one-scolding ‘there isn’t any scolding in your speech’ (Bécsi Codex (1416/1450), Iudith VIII) In Middle and Modern Hungarian, se-expressions can also be stacked, and can stand either pre- or postverbally. This may be the consequence of the analysis of [+specific] se-expressions as universal quantifiers (cf. É. Kiss 2009, 2010) with scope over negation. As such, they are subject to Q-raising, which is an iterable operation with no fixed direction, realizable as either left- or right-adjunction. Observe an example of the Hungarian National Corpus from 1881. (26) nem lopott el senki semmit not stole prt anybody anything ‘nobody stole anything’ 6
A se-expression could only be extraposed when it was explicitly contrasted, e.g.:
(i) Es nem zeretek egÿebet semmÿt hanem czak tegedet and not love-I else anything but only you ‘I love nothing else but you’
(Jókai )
(ii) Azert nenczen semÿm hanem Czak engalya ruham therefore isn’t anything-1sg but only engalya dress-1sg ‘Therefore I have nothing but only an engalya dress’
(Jókai )
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
The history of negative indefinites involving sem and the numeral egy ‘one’ has been somewhat different from the history of se-pronouns. Both es and sem (es+nem) were premodifiers in the earliest Old Hungarian documents. Later es also came to be used as an enclitic, and its two positions came to be associated with different functions. És, the standard Modern Hungarian version of the proclitic variant, is the connective corresponding to and. Is, the descendant of the enclitic, is an additive/distributive particle today. Sem, incorporating the additive particle, acting as a premodifier in the early Old Hungarian period, has also become a postmodifier. Jókai Codex contains, in addition to the regular archaic structure in (27a) and the regular novel structure in (27d), two patterns (those in (27b) and (27c)) which seem to anticipate the change in the position of sem. (27) a. sem egy N V: ew kerelmenek sem egy haznalattyat aloytuan his request-gen not one use-poss.3sg-acc assuming ‘not assuming any use of his request’
(Jókai 153)
b. sem egy N nem V: kyben semegy nugodalmat nem akaruala ew sebynek what-in not-one rest-acc not want-3sg-past his wound-dat vettny give ‘where he didn’t want to give any rest to his wound’ (Jókai 65) c. sem egy N sem V: Es hogy ottegyel Semegy lakas semuala holot and that there not-one dwelling not-was where feyet le haytana head-poss.3sg-acc down lay-cond-3sg ‘And that there was no dwelling where he could lay his head’
(Jókai 27)
d. egy N sem V: az tonak. . . zygetebe kyben meglen egy ember-sem that lake-gen island-poss.3sg-to where still one man not lakott-uala live-pfv-3sg-past ‘to the island of that lake where still no man had lived’ (Jókai 26) The variants in (27a–d) may corrrespond to subsequent stages of a diachronic process. (27a) contains no negative particle in addition to that incorporated in the particle sem associated with the indefinite. In (27b) the negative particle is reintroduced in a position left-adjoined to the verb. (Since the sentence contains no verbal particle, the preposing of the negated verb from T to Neg is string-vacuous, hence it cannot be verified.) In (27c) we find two sem particles; the second one is between the se-phrase
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Katalin É. Kiss
and the verb, in exactly the same position where the negative particle nem should appear. I hypothesize that in this unique example sem does, in fact, occupy the position of nem; it is a nem phonologically assimilated to the preceding sem. This pattern, not found elsewhere, may represent an intermediate stage in the change to (27d). In (27d), which also occurs only once in Jókai Codex, but has become the winning pattern in the long run, the proclitic sem is missing, but the indefinite is followed by a sem. If the prosody of (27d) was the same as it is today, then its sem is not the stressed negative particle but an unstressed enclitic modifying the indefinite. Its status as an enclitic of a minimizing role is shown in present-day Hungarian by the fact that it can be moved together with the indefinite. (28) a. Nem lakott egy ember sem a szigeten. not lived one man sem the island-on ‘No man lived on the island.’ b. Nem lakott a szigeten egy ember sem. As is clear from these Modern Hungarian examples, and the Old Hungarian example in (25), the enclitic sem could only retain its negative force when cliticized to a focused, hence immediately preverbal, indefinite, where it could be reanalysed as the occupant of the adjacent Neg position. Non-focused, postverbal indefinites in the scope of negation require the presence of both the negative particle nem, and the minimizing enclitic sem.
. Summary This chapter has shown that Hungarian negative constructions of the late ProtoHungarian period, representing the output of a former negative cycle, underwent another cycle in the twelfth–fifteenth century. This more recent cycle was set off by a morphological change. Negated indefinites came to be reinforced by the emphatic/additive/distributive proclitic es, which fused with the negative particle nem, yielding sem. Sem underwent further fusion with indefinite pronouns. Owing to word-internal phonological processes, the sem+indefinite pronoun complexes became morphologically more and more opaque. When the incorporated negative particle ceased to be recognizable, it was reintroduced adjoined to the verb, and negative pronouns were reinterpreted as pronouns participating in negative concord. The sem particle accompanying indefinite noun phrases lost its negative force owing to a change in its position (originally a proclitic, it became an enclitic, and came to be interpreted as a minimizing particle, the negative polarity counterpart of the additive es). It could retain its negative force in a single construction: in the case of focused, i.e. immediately preverbal, negated indefinites, where the enclitic sem could be reanalysed as the negative particle preceding the verb.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
A negative cycle in 12th–15th-century Hungarian
These changes went on parallel with the restructuring of the Hungarian sentence from SOV to TopFocVX∗ , a sentence structure with separate thematic and functional domains. In the new sentence structure, the negative particle is the head of a functional projection, eliciting V-movement.
Sources Bécsi kódex [Wien Codex]. Új Nyelvemléktár 1, ed. by Mészöly, Gedeon. Budapest: MTA. 1916. Der Münchener Kodex, ed. by Décsy, Gyula. Wiesbaden: Otto Harrassowitz. 1966. Halotti beszéd és könyörgés [Funeral Sermon and Prayer]. In Molnár, József & Simon, Györgyi (eds.). Magyar nyelvemlékek, 26–27. Budapest: Tankönyvkiadó, 1977. Jókai-kódex. Codices Hungarici VIII, ed. by P. Balázs, János. Budapest: Akadémiai Kiadó. 1981. Ómagyar Mária-siralom [Old Hungarian Mary’s Lament]. In Molnár, József & Simon, Györgyi (eds.). Magyar nyelvemlékek, 42–43. Budapest: Tankönyvkiadó, 1977.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP A NA M A R IA M A RT I N S
. Introduction In European Portuguese,1 the indefinite quantifier algum (‘some’) is a weak positive polarity item (PPI) that seems to turn into a strong negative polarity item (NPI) when it surfaces in post-nominal position.2 (1)
a. Algum animal vive aqui. some animal lives here ‘Some animal lives here.’ b. Animal algum vive aqui. animal some lives here ‘No animal lives here.’
Old Portuguese does not display such correlation between DP-internal word order and polar interpretation. In fact, in Old Portuguese algum was a bi-polar polarity item (Martins 2000) that would receive a positive or negative reading as a function of being part of a non-negative or negative sentence and independently of being prenominal or post-nominal. ‘Nominal negative inversion’ with algum/alguno is also found in Spanish (which however differs from Portuguese in some respects), but is not a grammatical option in most Romance languages. 1 I am very grateful to Rosario Álvarez Blanco, Montse Batllori, Paola Crisma, Manuel Pérez Saldanya, Victoria Vázquez Rozas, Ernestina Carrilho, and Anthony Kroch for invaluable data and discussion. The author’s research is funded by FCT—Fundação para a Ciência e a Tecnologia. 2 Adopting the typology of polarity items put forth in Martins (), I will be using the term ‘NPI’ to cover both ‘weak NPIs’ and ‘strong NPIs’, the latter corresponding to what many authors strictly designate as ‘n-words’. This is a terminological option with no particular theoretical implications with respect to the matters discussed in the chapter.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
The goal of this chapter is threefold. I will seek to understand how word order brings up the polar contrast illustrated in (1), how the negative interpretation associated with post-nominal algum arose in the course of time, and how exactly Portuguese and Spanish compare to each other with respect to the innovative structure. The three questions are naturally interrelated. The specific contours of the connection will hopefully be made clear throughout the chapter. I will propose that the sequence [N+algum] in contemporary European Portuguese is an NPI built in the syntax through incorporation of the noun and the indefinite quantifier in a DP-internal abstract negative head positioned above NumP, as illustrated in (2). Cyclic head-movement determines that N carries along to the incorporation site the indefinite quantifier (which heads NumP).3 This proposal will be central to developing an integrated account of the cross-linguistic variation attested in the geographic and temporal axes. (2) [DP... [NegP [Neg [ coisai alguma ]k [NumP [Num [ coisai alguma ]k [NP coisai ] ] ] ] ] ] I will be assuming (3a) as the basic structure for the DP (cf. Bernstein 1991, 2001; Zamparelli 1995; Heycock and Zamparelli 2005; Borer 2005, among others), and (3b) as the structure of a DP displaying ‘nominal negative inversion’ with algum. In (3b) the presence of the DP-internal NegP blocks the occurrence of PlP (PluralP).4 That NegP may be part of the functional structure of the DP has been proposed on independent grounds by different authors (see Haegeman 2002; Haegeman and Lohndal 2010; Troseth 2009). (3)
a. [DP [NumP [PlP [NP. . . b. [DP [NegP [NumP [NP. . .
The chapter is organized in five sections. In Section 7.2 ‘nominal negative inversion’ with algum in contemporary European Portuguese and some of its effects is described. In Section 7.3 the path from Old to Modern European Portuguese is considered and partially accounted for. Section 7.4 compares Spanish with Portuguese and shows how the comparative perspective is crucial to clarify the diachronic facts and thoroughly account for the change, which turns out to be a two-step change in European Portuguese. Thus while the structural representation in (2) describes Spanish and seventeenth/eighteenth-century European Portuguese, further Neg-to-D movement 3
NumP (NumberP) is the functional projection also designated as QP (Quantifier P). On the ‘bleeding relation’ between negation and plural, see Roberts and Roussou () and Roberts (). This hypothesis is apparently contradicted by the availability in English of DPs like ‘No animals’ (under the assumption that the negative determiner is first merged in Neg and subsequently moves to D). Note, however, that since there is no plural inflection on the negative determiner (‘no’), the plural marker on the noun (‘animals’) can be thought of as purely post-syntactic (thus not involving the presence of Pl(ural)P in the syntactic structure). Cf. Embick and Noyer (). 4
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
in later stages of European Portuguese is what sets it apart from Spanish. Section 7.5 concludes the chapter. In addition, it contains a brief note on the history of French aucun ‘any/none’ and Italian alcuno ‘any’, suggesting that the proposed analysis of Portuguese and Spanish post-nominal algum allows a novel perspective on the development of the earlier PPIs aucun and alcuno into NPIs.
. ‘Nominal negative inversion’ in European Portuguese The indefinite quantifier algum entails a positive or a negative interpretation depending on whether it surfaces in prenominal or post-nominal position. The examples in (4) and (5) illustrate how word order lies behind the contrast in interpretation and take as term of comparison the regular PPI/NPI pair alguém/ninguém (‘somebody/nobody’). (4) a. Alguém vive aqui. somebody lives here b. Ninguém vive aqui. nobody lives here (5)
a. Algum animal vive aqui. some animal lives here ‘Some animal lives here.’ b. Animal algum vive aqui. animal some lives here ‘No animal lives here.’
As for the interaction with sentential-negation, the inverted sequence [N+algum] displays the preverbal/postverbal asymmetry characteristic of European Portuguese n-words, so it obligatorily co-occurs with the predicative negation marker não (‘not’) when postverbal but excludes the predicative negation marker when preverbal, as illustrated by (6) and (7). (6) a. Não vive aqui ninguém. not lives here nobody b. ∗ Vive aqui ninguém. lives here nobody c. Ninguém vive aqui. nobody lives here d. ∗ Ninguém não vive aqui. nobody not lives here ‘Nobody lives here.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
(7) a. Não vive aqui animal algum. not lives here animal some b. ∗ Vive aqui animal algum. lives here animal some c. Animal algum vive aqui animal some lives here d. ∗ Animal algum não vive aqui. animal some not lives here ‘No animal lives here.’ Like the pronominal n-word ninguém (‘nobody’), but unlike the adjectival n-word nenhum (‘not one’), post-nominal algum blocks plural inflection, as exemplified in (8). Moreover, it must be strictly adjacent to the noun, as shown in (9). (8) a. Alguns animais vivem aqui. some-pl animals live-3pl here ‘Some animals live here.’ b. ∗ Animais alguns vivem aqui. animals some-pl live-3pl here ‘No animal lives here.’ (9) a. ∗ Animal selvagem algum vive aqui. animal wild some lives here ‘No wild animal lives here.’ deserto algum vive aqui. b. ∗ Animal do animal of-the desert some lives here ‘No animal of the desert lives here.’ All the facts can be shown to essentially follow from the structural analysis given in (2). The sequence [N+algum] behaves like strong NPIs such as ninguém (‘nobody’) because it is in fact an NPI built in the syntax with the contribution of the DP-internal Neg-head. Plural inflection is blocked because, by hypothesis, whenever NegP is part of the DP, Pl(ural)P is not projected. The strict adjacency requirement between the noun and post-nominal algum is the regular outcome of cyclic head movement.5 I will now introduce further empirical evidence to support the idea that whenever ‘nominal negative inversion’ takes place, the sequence [N+algum] is the NPI, not the indefinite quantifier by itself. 5 I will not discuss in this chapter the syntax of adjectives, but the simpler assumption would be that adjectives are always maximal projections, not heads—cf. Alexiadou, Haegeman, and Stavrou () for references.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
The availability of the DP internal negative head makes ‘nominal negative inversion’ extensible to the negative indefinite nenhum (‘not one’). In what follows, I will look at the parallel grammatical effects of word-order alternation for algum and nenhum, though only the former exhibits polarity reversal dependent on word order.6 .. Pronouns vs. full DPs Post-nominal algum and post-nominal nenhum are allowed in contexts that require pronominal quantifiers (if available) and exclude full DP quantificational expressions. The fact that the sequences displaying ‘nominal negative inversion’ (i.e. [N+algum]/[N+nenhum]) pattern with pronouns is evidence in favour of their analysis as a NPI unit composed in the syntax. (10) [A] O que é que o João gosta de ler? the what is that the João enjoys of read ‘What does João enjoy reading?’ nada. [B] a. Ele não lê he not reads nothing nenhuma coisa. b. ∗ Ele não lê he not reads not-one thing c. Ele não lê coisa nenhuma. he not reads thing not-one d. Ele não lê coisa alguma. he not reads thing some alguma coisa. e. ∗ Ele não lê he not reads some thing ‘He doesn’t read anything.’ .. Negative answers to polar questions The sequences [N+algum]/[N+nenhum] may constitute a well-formed negative answer to a polar question, while the non-inverted sequences are excluded in the same context. The contrast can be explained under the view that ‘nominal negative inversion’ joins the indefinite quantifier and the noun into a single negative word that may then enter the paradigm of possible polar answers (depending on the degree of referential vagueness of the noun).
6 The NPI nenhum can be post-nominal in a structure that does not involve ‘nominal negative inversion’ (therefore does not include NegP), but solely emphasis on the NPI. In this chapter, I will not pay attention to this other DP-structure displaying what could be called ‘emphatic inversion’. This type of inversion is also available in Spanish, while ‘nominal negative inversion’ with nenhum is not.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP (11)
lá amanhã? [A] Vais go-2sg there tomorrow ‘Are you going there tomorrow?’ [B] a. Não. No. b. De of
maneira nenhuma. manner not-one
c. De of
forma form
alguma some
d. ∗ De nenhuma maneira. of not-one manner e.
∗ De
alguma of some ‘Not at all.’
forma. forma
.. Count vs. mass nouns ‘Nominal negative inversion’ with algum and nenhum interacts with the mass/count distinction on nouns, apparently blocking the count interpretation, as exemplified in (12) and (13).7 The fact that ‘nominal negative inversion’ makes nouns be interpreted as mass can be derived as a consequence of the absence of the head Pl(ural) in the DP structure. According to Borer (2005) nouns denote masses by default. That is to say, in the absence of any grammatical specification contributed by syntactic structure above NP, nouns are unspecified for any properties, including the mass/count property, 7 Judgments vary across speakers with respect to the requirement that ‘nominal negative inversion’ obtains with mass nouns such as ‘fear’, ‘luck’, ‘water’. For speakers that judge sentences (i-b), (i-d), and (ii-b) as ungrammatical, the availability of the NPI unit formed in the syntax seems to have the same type of blocking effect attested in example (-B-b). The judgments given below are my own.
(i)
a. Não temos {medo nenhum/sorte nenhuma}. not have-1pl {fear not-one/luck not-one} b. ∗ Não temos {nenhum medo/nenhuma sorte} not have-1pl {not-one fear/not-one luck}. c. Não temos {medo algum/sorte alguma}. not have-1pl {fear some/luck some} d. ∗ Não temos {algum medo/alguma sorte}. not have-1pl {some fear/some luck} ‘We don’t have any fear/luck (at all).’
(ii) a. As flores não têm água nenhuma the flowers not have water not-any b. ∗ As flores não têm nenhuma água. the flowers not have not-one water ‘There is no water (at all) in the flowers’ vase.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
and are interpreted by default as mass. It is the Pl(ural)/Cl(assifier) head that has the function of portioning-out noun’s denotations making the count interpretation available. Whenever Pl(ural)/Cl(assifier) is absent, nouns are interpreted as mass.8 (12)
a. A chave não entra na fechadura de nenhuma maneira the key not enters in-the lock of not-one way ‘The key doesn’t enter in the lock in any possible way/position.’ b. A chave não entra na fechadura de maneira {nenhuma/alguma}. the key not enters in-the lock of way {not-one/some} ‘The key doesn’t enter in the lock at all.’
(13)
a. Ele não come nenhuma fruta (excepto cerejas). he not eats any fruit (except cherries) ‘He doesn’t eat any kind of fruit (except cherries).’ b. Ele não come fruta {nenhuma/alguma}. (#excepto cerejas). he not eats fruit {not-one/some}. (#except cherries) ‘He doesn’t eat fruit at all (except cherries).’
.. Gradable quantifiers Quantifiers like muitos ‘many’ and poucos ‘few’ admit degree modification. In Portuguese also nada can behave as a gradable quantifier (see example (14)). In the sequence [coisa+alguma], [coisa+nenhuma], the noun coisa (‘thing’) can be modified by the superlative suffix -íssima (‘-est’), deriving coisíssima nenhuma (although ∗ coisíssima is ill-formed by itself). Crucially, the sequence ∗ nenhuma coisíssima, with prenominal nenhum, is sharply unacceptable (see examples (15)–(17)). These data support the idea that ‘nominal negative inversion’ with algum/nenhum gives rise to a NPI unit that changes some of the original properties of its constitutive parts. (14) a. Ainda não fiz nadíssima! yet not did-1sg nothing-est ‘I haven’t done anything at all yet!’ b. Não sabe nada, nadíssima. not knows nothing nothing-est ‘He doesn’t know anything, anything at all.’ (15)
a. Nunca recebi favor do Sr. D. Pedro II nem ele me never received-1sg favour of-the Sir Pedro II nor he me-dat deve coisíssima alguma. owe thing-est some ‘I have never been favoured by the king D. Pedro II, neither does he owe me anything at all.’
8 The Number Phrase (or Quantity Phrase) is responsible for the assignment of quantity to stuff (i.e. masses) or for the counting of portioned-out stuff. ‘Cl(assifier)’ is in Borer’s system what we are calling here ‘Pl(ural)’.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
b. Não preciso dela para coisíssima alguma. not need-1sg her for thing-est some ‘I do not need her for anything at all.’ c. Não têm préstimo para coisíssima nenhuma. not have utility for thing-est not-one ‘They are of no use at all.’ (Corpus do Português, nineteenth/twentieth centuries) (16) a. Não senti dores, não senti nada. Não senti coisíssima not felt-1sg pains not felt-1sg nothing not felt-1sg thing-est nenhuma. not-one ‘I didn’t feel pain, didn’t feel anything. I didn’t feel anything at all.’ b. Não me tem doído coisíssima nenhuma. not me has ached thing-est not-one ‘I haven’t been feeling any aches or pains anywhere.’ (17)
a.
(CORDIAL-SIN)
∗ Não
me tem doído nenhuma coisíssima. not me has ached not-one thing-est ‘I haven’t been feeling any aches or pains anywhere.’
b. ∗ Não têm préstimo para nenhuma coisíssima. not have utility for not-one thing-est ‘They are of no use at all.’
. From Old Portuguese to Modern Portuguese Old Portuguese (i.e. the Portuguese language up to the sixteenth century) does not display the correlation between DP-internal word order and polar interpretation described in the previous sections. In Old Portuguese, algum (‘some’) was a bi-polar polarity item (Martins 2000) that would receive a positive or negative reading as a function of being part of a non-negative or a negative sentence, and independently of being prenominal or post-nominal. Examples (18a–b) show that algum could be prenominal or post-nominal and receive a positive interpretation. Examples (19a–b) illustrate how it could as well have a negative interpretation irrespective of word order. (18)
a. Se aqui ficardes em esta furesta, toste vos poderia v˜ıir if here stay-2pl in this forest soon you-dat could come ende mal alg˜uu˜ from-that harm some ‘If you stay here in this forest any longer, soon some harm may come to you.’ (Demanda do Santo Graal. Nunes 2005: 64)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins b. o coraçom me diz que vos há de contecer alg˜uu˜ mal the heart me-dat tells that you-dat is to happen some harm ‘The heart tells me that some harm is coming to you.’ (Demanda do Santo Graal. Nunes 2005: 227)
(19)
a. E ele nom respondeu a cousa alg˜ua que lhe dissesse, and he not answered to thing some that him would-say-3sg ca era mui sanhudo because was very angry ‘He did not answer to anything that he was asked because he was so angry.’ (Demanda do Santo Graal. Nunes 2005: 82) b. outra vez nom façades tam gram braveza nem tam gram crueza other time not do-2pl so great violence nor so great cruelty como fezestes, ca nom vos pode ende v˜ıir alg˜uu˜ as did-2pl because not you-dat can from-that come some bem, mas todo mal good but all evil ‘Don’t be so wild and cruel next time since that will not bring you any good, but only evil.’ (Demanda do Santo Graal. Nunes 2005: 146)
The emergence of ‘nominal negative inversion’ with algum appears to be a side effect of a series of changes that led to the loss of the Old Portuguese ‘free inversion’ attested in (18)–(19) and had the general effect of tying the polar value of algum to a particular placement with respect to the noun. After the fourteenth century the availability of bare nouns was mostly restricted to singular mass nouns and plurals. This had the effect of excluding (or strongly reducing the occurrence of) sentences like (20a–c). As a result of the change, words with a ‘vague’ referential import like cousa/rem (‘thing’), gente (‘people’), omem (‘man’), that could alternate freely with NPIs like nada (‘anything’) and nenhum (‘anybody’) under the scope of negation (see (21)), either gradually decreased in their use or started to co-occur more frequently with an indefinite quantifier, as shown in (22). The rise in frequency of sentences like (22a)—with post-nominal algum—made them salient enough for ulterior reanalysis. (20) a. E eu vos levarei a lugar u pensarám bem de vós and I you will-take to place where will-treat-3pl good of you ‘I will take you to a place where you will be well cared for.’ (Demanda do Santo Graal. Nunes 2005: 131) b. Nom ia a lugar que nom achasse novas que d˜uus not went-3sg to place that not found-3sg news that of-one-pl que doutros that of-others ‘He wouldn’t go to any place where he wouldn’t have news of (some of) them.’ (Demanda do Santo Graal. Nunes 2005: 147)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
c. Nunca achei cavaleiro, fora el, que me vencesse never met knight except him that me defeated ‘I have never come across a knight that could defeat me, except him.’ (Demanda do Santo Graal. Nunes 2005: 201) (21)
a. E caeu em terra morta que nom fallou mais cousa. and felt-3sg in ground dead that not spoke-3sg more thing ‘And she felt dead so that she wasn’t able to say anything else.’ (Demanda do Santo Graal. Nunes 2005: 95) b. E ainda mais digo que jamais nom tornarei aa corte and still more say-1sg that never not will-come-3sg to-the court por cousa que avenha for thing that happens ‘And I will never come back in court for any reason whatsoever.’ (Demanda do Santo Graal. Nunes 2005: 36) c. Nom comi nem bevi nem achei gente que me not ate-1sg nor drank-1sg nor met-1sg people that me-acc quisesse receber em sua companha. want-3sg take-inf into their company ‘I was never able to eat or drink, neither have I met anyone who would give me any shelter.’ (Demanda do Santo Graal. Nunes 2005: 318–19) d. houve tam gram ledice que o nom poderia homem contar had-1sg so big happiness that it not could man tell-inf ‘He showed such great happiness that it is impossible to anybody to describe it.’ (Demanda do Santo Graal. Nunes 2005: 233)
(22) a. e desy e˜trarão pella casa muyto maravilhados, porque and then entered-3pl across-the house very marvelled because nella não viram cousa allg˜ua in-it not saw-3pl thing some ‘As they entered the house, they were astonished to find out that it was empty.’ (Conde D. Pedro de Meneses. Brocardo 1997: 251) b. primeiramemte o serviço de Deus que outra allg˜ua cousa first the service of God than other some thing ‘The service of God is to be put before any other thing.’ (Conde D. Pedro de Meneses. Brocardo 1997: 204) c. ate’quy nõ temos feita nenh˜ua cousa em que possamos (. . . ) until-here not have-1pl done not-one thing in that can-1pl ser prezados be praised ‘Up until now, we have not done anything that deserves to be praised.’ (Conde D. Pedro de Meneses. Brocardo 1997: 387)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins d. E porque nõ vyram cousa nenhua (. . . ) torno-se and because not saw-3pl thing not-one returned-3sg-himself a galliota to-the boat ‘And because they did not see anything, he went back to the boat.’ (Conde D. Pedro de Meneses. Brocardo 1997: 380–1)
By the end of the sixteenth century, the ‘free’ post-nominal placement of the indefinite quantifier algum was lost, so sentences like (23b) ceased to be a grammatical option. This was maybe the effect of the loss of middle scrambling both at the clausal and the DP level (cf. Martins 2002).9 Later, the earlier bi-polar polarity items like algum evolved to weak PPIs, as part of a more general drift of both positive and negative polarity items (Martins 2000), and were therefore excluded from negativeconcord contexts, so sentences like (23c) disappeared as well. (23) a. Des onte ao serão ouvemos alg˜ua folga since yesterday at night had-1pl some rest ‘Since yesterday night, we had some rest.’ b. Des omte ao serão ouvemos folga alg˜ua since yesterday at night had-1pl rest some ‘Since yesterday night, we had some rest.’ c. Des omte ao serão não ouvemos alg˜ua folga since yesterday at night not had-1pl some rest ‘Since yesterday night, we did not have any rest.’ d. Des omte ao serão não ouvemos folga alg˜ua since yesterday at night not had-1pl rest some ‘Since yesterday night, we did not have any rest.’ (Examples adapted from Conde D. Pedro de Meneses. Brocardo 1997) 9 The change also affected the indefinite outro/outros (‘other/others’), which is rarely attested in postnominal position after the sixteenth century. In the sixteenth century we can still find examples of post-nominal algum with positive meaning (that is to say, sentences like (b)), which disappear when ‘free inversion’ is lost:
(i) Desta gente refresco algum tomámos e do rio fresca água from-this people refreshment some had-1pl and from-the river fresh water ‘This people offered us some refreshment, and we got some fresh water from the river.’ (Corpus do Português: Luís de Camões) (ii) Que chove quando não quero/ e faz um sol das estrelas/ quando chuva that rains when not want-1sg and does a sun of-the stars when rain alguma espero some want-1sg ‘It just rains when I do not want (because it damages the crops) and is sunny and dry when some rain would really be needed.’ (Corpus do Português: Gil Vicente)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
The loss of ‘free inversion’ with algum would have made sentences like (23d) also unavailable had a reanalysis process not taken place.10 Under the analysis put forth in this chapter, the fact that UG makes available a Neg-head as part of the functional structure of the DP allowed the reanalysis of the Old Portuguese structure with NPscrambling represented in (24) as the European Portuguese structure with ‘nominal negative inversion’ represented in (25).11 The change is plausible from an acquisition perspective as it does not imply any backtracking from earlier decisions (cf. Fodor 1998; Dresher 1999; Lightfoot 1991, 1999).12 (24) Nom falou [DP... [NumP [NP cousa]i [Num’ alguma [NP cousa]i ]]] thing not spoke some ‘She didn’t say anything.’ (25) Não disse [DP... [NegP [Neg’ [coisai alguma]k [NumP [Num’ thing some not spoke [coisai alguma]k [NP coisai ] ] ] ] ] ] ‘She didn’t say anything.’
. Contrasting Portuguese with Spanish In Spanish, negative inversion with alguno (‘some’) is available and blocks plural inflection like in Portuguese, as illustrated in (26) and (27).13 10 I am well aware that I am oversimplifying when referring to Old Portuguese inversion with algum as ‘free inversion’. Nonetheless, space considerations preclude me from further developing this point here. 11 As expected under the proposed analysis, the Old Portuguese type of nominal inversion represented in () did not block plural inflection:
(i) a. De meu padre sabedes ou ouvistes novas alg˜uas? of my father know-2pl or heard-2pl news some-f-pl ‘Have you heard of my father?’ (Demanda do Santo Graal. Nunes : ) b. e que roubavam nossos regnos e faziam outras coussas alg˜uas desonestas and that robbed-3pl our kingdoms and did-3pl other things some-pl dishonest ‘They would rob the kingdom and do some other dishonest things.’ (Corpus do Português: fifteenth century) 12 When the structure represented in () ceased to be acquired, there were two logical possibilities. Either it would be reanalysed or would be lost. Portuguese displays the former path, Galician and Catalan the latter. So contemporary Galician and Catalan totally exclude post-nominal algún/algun, although Old Galician and Old Catalan allowed it. As pointed out to me by Manuel Pérez Saldanya, this may well be a consequence of the particular unfavourable sociolinguistic conditions of Galician and Catalan in sixteenthcentury Iberia and afterwards, as the structures we are discussing presumably mostly belonged to high register style. 13 The examples in this section come from Rigau (: ), Sanchéz-Lopez (: –), and Montse Batllori (p.c.). Besides negation proper, also ‘modal’/‘weak negative’ contexts (cf. Bosque ; Giannakidou , ; Milner ; van der Wouden , among others) license post-nominal alguno, as illustrated below. This is not the case in contemporary European Portuguese because n-words, including [N+algum], systematically
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
(26) a. No he visto película alguna esta semana not have-1sg seen movie some this week ‘I haven’t watched any movie this week.’ b. La asemblea no planteó problema alguno a la propuesta. the assembly not raised problem some to the proposal ‘The assembly didn’t raise any objection to the proposal.’ (27) a. No hay solución alguna para ese dilema. not is solution some for that dilemma ‘There is no solution for such dilemma.’ b. ∗ No hay soluciones algunas para ese dilema. not is solutions some-pl for that dilemma ‘There aren’t any solutions for such dilemma.’ Spanish crucially diverges from Portuguese, however, in that ‘nominal negative inversion’ with alguno is only licensed under the scope of negation, typically in postverbal position, as illustrated in (28) and (29).14
behave as strong NPIs (see Martins ). In seventeenth and eighteenth-century Portuguese, however, the Spanish patterns exemplified in (i) are also attested. (i) a. Jamás mi país le ha prohibido a nadie que viaje a lugar alguno que never my country him-dat has forbidden to nobody that travel to place some that desee. wish ‘My country has never forbidden anyone to travel anywhere one may wish.’ b. Durante la peregrinación, constantemente nos sacábamos nuestros zapatos (. . .) during the pilgrimage constantly ourselves took-off-1pl our shoes antes de entrar a lugar alguno before to enter in place some ‘Throughout the pilgrimage, we would always take our shoes off before entering any (sacred) place.’ c. tendrá, por mala que sea, más entradas que otra alguna it-will-have though bad that it-may-be, more entrances than other some ‘Poorly acted as it may be, it will still have more spectators than any other (theatrical performances).’ (Google search, //) 14 Spanish also differs from Portuguese in that it does not impose strict adjacency between post-nominal alguno and the noun. While prepositional modifiers are not allowed to intervene between the noun and the indefinite quantifier (see (i)), evaluative adjectives may and relational adjectives must intervene (see (ii) and (iii)). I will not deal here with the issue of adjectives. A possible way to derive the contrast between Spanish and Portuguese is to take Spanish alguno to merge in Spec,NumP and therefore be left behind when the noun cyclically moves to incorporate in the DP-internal Neg-head.
(i) a. No conozco libro alguno de matemáticas que discuta este teorema. not know-1sg book some of mathematics that discusses this theorem b. ∗ No conosco libro de matemáticas alguno que discuta este teorema. not know-1sg book of mathematics some that discusses this theorem ‘I am not aware of any book of mathematics that might discuss this theorem.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
(28) a. No fue necesaria ayuda alguna. not was necessary help some b. ∗ Ayuda alguna (no) fue necesaria. help some (not) was necessary ‘No help was necessary.’ (29) a. No vive aquí persona alguna. not lives here person some b. ∗ Persona alguna (no) vive aqui. person some (not) lives here ‘Nobody lives here.’ The distribution of [N+alguno] in Spanish, typically occurring in postverbal position, is reminiscent of the distribution of bare nouns discussed by Longobardi (1994). A hypothesis to account for the contrast between Portuguese and Spanish then comes to mind. The restricted distribution of [N+alguno] in Spanish would be a consequence of the need to license the null Determiner in a structure like (2). Since European Portuguese escapes such restriction, that would indicate that in Portuguese Neg-to-D movement can take place to fill in the D position. If this hypothesis can be shown to be on the right track, in European Portuguese the final step of the change will be a case of upward reanalysis along the functional hierarchy in the sense of Roberts and Roussou (2003). Now, if the change in Portuguese in fact proceeds in two steps, we expect to find evidence that at some point in the course of time, Portuguese was like contemporary Spanish. This prediction is born out as seventeenth- and early eighteenthcentury European Portuguese behaves just like Spanish in not allowing the sequence [N+algum] except when it is licensed by negation (or related ‘modal’ contexts) in complement position, namely postverbally or after the preposition sem ‘without’. The Corpus do Português indicates that the second step of the change occurred after the seventeenth century. I could not find any example of post-nominal algum in preverbal subject position or other position outside the scope of negation throughout the seventeenth century (although the corpus provides 470 examples of post-nominal (ii) a. No asistí a conferencia alguna interesante. not attended-1sg to lecture some interesting b. No asistí a conferencia interesante alguna. not attended-1sg to lecture interesting some ‘I did not attend any worthy lecture.’ (iii) a. ∗ No hay avería alguna eléctrica en este barrio. not is failure some electrical in this neighbourhood b. No hay avería eléctrica alguna en este barrio. not is failure electrical some in this neighbourhood ‘There isn’t any electrical failure in this neighbourhood.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
algum for this period). Very few examples of post-nominal algum outside the scope of negation appear in the eighteenth century. One has to wait until the nineteenth century to easily find attestations of the innovation.15 Eighteenth- and nineteenthcentury examples are given in (30) and (31) respectively. Bear in mind that all these sentences are currently excluded in Spanish.16 (30) Coisa alguma há mais deliciosa que a sua alegria, nem mais thing some there-is more delicious than the her joy nor more penetrante que a sua ternura. penetrating than the her tenderness ‘There is nothing more pleasant than her joy nor more moving than her tenderness.’ (Corpus do Português: eighteenth century) (31)
a. Coisa alguma escapou! thing some escaped ‘Nothing was left.’ b. Namorado algum, dos mais ardentes, palpitou com tanta febre lover some of-the more ardent palpitated with such fever no antegozo de uma aventura. in-the anticipation of an adventure ‘No lover was ever so deeply excited with the anticipation of an affair.’ c. Em época alguma tinham os criados conhecido Maurício tão in time some had the servants known Maurício so caseiro. domestic ‘Never before had the servants seen Maurício so domestic.’ (Corpus do Português: nineteenth century)
15 The data found in the diary of Conde da Ericeira, ranging from to , point in the same direction (cf. Lisboa, Miranda, and Olival , , ), showing that in the first decades of the eighteenth century the split between Portuguese and Spanish had not become visible yet. There are occurrences of post-nominal algum in the diary (among the total number of , occurrences of algum) and no single example of post-nominal algum except in complement position under the scope of negation. 16 The fact that at a certain point in its diachronic development Portuguese was like contemporary Spanish has two interesting consequences: (i) it enables us to attain a better understanding of seventeenth- and early eighteenth-century Portuguese by exploring contemporary Spanish; (ii) it comes out as a natural result that the grammar of contemporary European Portuguese that I have described may not be shared by all speakers. In fact, some European Portuguese speakers’ judgments fit better within a Spanish-type grammar. This more conservative European Portuguese grammar seems however to be marginal. The data found in the Corpus do Português show that there are no occurrences of post-nominal algum but adjacent to the noun in the twentieth century, once Brazilian Portuguese texts are excluded. In the eighteenth and nineteenth centuries a few examples appear, as exemplified in (i).
(i) sem nenhuma da solenidade do antigo, nem elegância moderna alguma without none of-the solemnity of-the ancient nor elegance modern some ‘Without any of the ancient solemnity or modern refinement.’ (Corpus do Português: Almeida Garrett, nineteenth century)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
In the turn from the eighteenth to the nineteenth century, just after the second step of the change comes into view and sets Portuguese apart from Spanish, there is a striking rise in frequency of post-nominal nenhum in European Portuguese. From 16 of the total number of examples of adjectival nenhum in the eighteenth century, the frequency of the post-nominal placement rises to 43 in the nineteenth century and approaches 50 in the twentieth century, in Corpus do Português. This rate reaches up to 68 in the corpus FLY, a corpus of personal letters written in the context of war, migration, imprisonment, and exile from the years 1900 to 1974. These data appear to reveal that once Neg-to-D movement is available in European Portuguese grammar, its range extends from algum to nenhum. At this point, inversion with the latter (i.e. [N+nenhum]) becomes an unmarked option, displaying the morphological and semantic effects discussed in Section 7.2. As expected, Spanish does not behave like Portuguese with respect to post-nominal nenhum/ninguno. Not only does it not display the type of word-order effects discussed in Section 7.2 (compare (11) with (32), for example) but it only allows post-nominal ninguno as a marked option (some type of extraposition) with an emphatic import, as illustrated in (33). Much is left to be said with respect to nominal inversion with nenhum in Portuguese, which is here identified as a topic for future research. allá mañana? (32) [A] Vas go-2sg there tomorrow ‘Are you going there tomorrow?’ [B] a. ∗ De of
manera manner
b. De ninguna of not-one ‘Not at all.’ (33)
(Spanish)
ninguna. not-one manera. manner
a. No tenemos ningún miedo. not have-1pl not-one fear
(Spanish)
b. No tenemos miedo ninguno. (marked/emphatic) not have-1pl fear not-one ‘We don’t have any fear (at all).’
. Conclusion, with a brief note on Italian and French This chapter starts from the observation that in contemporary European Portuguese there is a correlation between DP-internal word order and polar interpretation when the polarity item algum (‘some’) is involved—e.g. algum animal (‘some animal’) vs. animal algum (‘no animal’). So while in prenominal position algum is a regular weak
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins
PPI, post-nominal algum can only have a negative interpretation (‘no’). To be more precise, ‘nominal negative inversion’ with algum gives rise to the unit [N+algum] that behaves as a strong NPI. This was not the case in Old Portuguese, and more generally in Old Romance. The investigation pursued in the present chapter addressed the following questions: 1. How is the NPI [N+algum] formed in the syntax? 2. How has the negative interpretation associated with post-nominal algum arisen in the course of time? 3. How exactly do Portuguese and Spanish compare to each other with respect to the innovative structure? The structural representations shown in (34)– (36) summarize the proposed answers. While ‘free inversion’ in Old Portuguese (and presumably Old Romance in general) would be a type of scrambling at the DP-level (see (34)), with no specific effect on polar interpretation, the reanalysis of this former DP-structure (often attested under the scope of sentential negation) as a DP containing a Neg-head gave rise to ‘nominal negative inversion’ (see (35)), hence tying the polar value of algum to a particular placement with respect to the noun. This initial step of the change is shared by Portuguese and Spanish. Later, European Portuguese evolved a step more and diverged from Spanish. This second step of the change is shown in (36) and can be understood as a case of upward reanalysis along the functional hierarchy in the sense of Roberts and Roussou (2003). Old Portuguese (and presumably Old Romance) (34) [DP ... [NumP [NP animal]i [Num’ algum [NP animal]i ] ] ] Spanish and 17th/18th-century European Portuguese (35)
[DP [D’ [e] [NegP [Neg’ [ animali algum]k [NumP [Num’ [animali algum ]k [NP animali ] ] ] ] ] ] ]
European Portuguese (36) [DP [D’ [ animali algum]k [NegP [Neg’ [ animali algum]k [NumP [Num’ [animali algum ]k [NP animali ] ] ] ] ] ] ] In Italian and French, the change progressed further and the correlates of algum were turned into lexical NPIs (cf. Roberts and Roussou 2003; Roberts 2007; Déprez and Martineau 2003; Paola Crisma, p.c.). Still, both French and Italian seem to offer evidence that ‘nominal negative inversion’ was available at a certain point of the diachronic path of aucun/alcuno from PPI to NPI, and played a role in the change. That is to say, Italian and French likely attest how a PPI may develop into a lexical NPI through a stage in which the NPI is syntactically built (through ‘nominal negative inversion’). The data displayed in Tables 7.1 and 7.2 (taken from Déprez and Martineau 2003) are very revealing in two respects. They show that the negative interpretation of aucun in sixteenth-century French is often associated with its post-nominal placement (see Table 7.1). They also show that singular favours and plural disfavours the negative
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
Table .. Aucun in prenominal and post-nominal positions in 16th-century French (Déprez and Martineau 2003) 16th c. aucun N N aucun
Positive context
Polarity context
Negative context
12 1
3 11
27 23
Table .. Aucun as a noun-modifying form in positive, polarity, and negative contexts in 16th-century French (Déprez and Martineau 2003) 16th c. Singular Plural
Positive context
Polarity context
Negative context
0 10.5 (8)
21.1 (16) 5.3 (4)
60.5 (46) 2.6 (2)
interpretation (see Table 7.2). This is precisely what is expected if ‘nominal negative inversion’ was a grammatical option in French at a certain point in the diachronic development of aucun. Recall that in contemporary European Portuguese and Spanish ‘nominal negative inversion’ with algum/alguno blocks plural inflection. Italian is particularly interesting because only singular alcuno turned into an NPI, while plural alcuni is still a PPI. Under the hypothesis that ‘nominal negative inversion’ with alcuno was available at some stage in the history of Italian and played a role in the change, the facts fall into place, because the restriction to singular is precisely an effect of the particular structure involved in ‘nominal negative inversion’, with DP-internal NegP blocking the projection of Pl(ural)P. The Italian data displayed in the following examples illustrate the polarity contrast between alcun(o) (sg. ‘any’) and alcuni (pl. ‘some’). Moreover, the data show that alcuno must be licensed under the scope of negation (like post-nominal alguno in Spanish), and that alcun(o) (‘any’) differently from alcuni (‘some’) can be post-nominal (though it does not display the type of word-order-dependent contrasts discussed in Section 7.2 with respect to European Portuguese). (37) a. Alcuni animali vivono qui. some-pl animals live-3pl here b. Qui vivono alcuni animali. here live-3pl some-pl animals c. Alcuni animali non vivono qui. some-pl animals not live-3pl here
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ana Maria Martins d. ∗ Animali alcuni vivono qui. animals some-pl live-3pl here e. ∗ Qui non vivono animali alcuni. here not live-3pl animals some-pl ‘Some animals {live/don’t live} here.’
(38) a. Qui non vive alcun mammifero. here not lives any-sg mammal ‘No mammal lives here.’ b. Non viveva lì animale alcuno. not lived there animal any-sg ‘No animal lived there.’ (39) a. Non c’è stata alcuna obiezione. not there-is been any-f-sg objection b. Non c’è stata obiezione alcuna. not there-is been objection any-f-sg ‘There wasn’t any objection.’ (40) a. ∗ Alcun mammifero (non) vive qui. alcun mammal (not) lives here b. ∗ Mammifero alcuno (non) vive qui. mammal alcuno (not) lives here ‘{Some/No} animal lives here.’
Sources of the data Corpora CORDIAL-SIN: Syntax-oriented corpus of Portuguese dialects. Available online at . Corpus do Português: Davies, Mark and Michael Ferreira. (2006–) Corpus do Português: 45 million words, 1300s–1900s. Available online at . FLY: Forgotten Letters Years 1900–1974. Available online at . Texts Brocardo, Teresa, ed. (1997). Crónica do Conde D. Pedro de Meneses. Lisboa: Fundação Calouste Gulbenkian / JNICT. Lisboa, João, Tiago C. P. dos Reis Miranda, and Fernanda Olival, eds. (2002). Gazetas manuscritas da Biblioteca pública de Évora, vol. 1 (1729–31). Lisboa: Colibri, CIDEHUS, CHC-UNL.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Negation and NPI composition inside DP
Lisboa, João, Tiago C. P. dos Reis Miranda, and Fernanda Olival, eds. (2005). Gazetas manuscritas da Biblioteca pública de Évora, vol. 2 (1732–4). Lisboa: Colibri, CIDEHUS, CHC-UNL. Lisboa, João, Tiago C. P. dos Reis Miranda, and Fernanda Olival, eds. (2007). Gazetas manuscritas da Biblioteca pública de Évora, vol. 3 (1735–7). Lisboa: CIDEHUS, CHCUNL. Unpublished. Nunes, Irene Freire, ed. (2005). A Demanda do Santo Graal. Imprensa Nacional—Casa da Moeda. 2nd edition.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Part II Syntax and Morphology
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity and how syntax drives morphological change CHRIS REINTGES
. Introduction The nature of the syntax–morphology interface and the complex ways in which syntactic and morphological structure are interrelated is a central question in synchronic and diachronic theories of language. Much comparative work derives interlanguage variation from syntactic parameters, which are restricted to functional categories. Functional categories are instantiated by inflectional morphology, which is part of the lexical inventory. In reducing parametric differences to the ‘narrow category of morphological properties, primarily inflectional’, as Chomsky (2001: 2) suggests, the Minimalist Program would seem to posit a tight connection between syntax and inflectional morphology. Nevertheless, it has become increasingly difficult to formulate this relation in formal terms. For one thing, it is no longer clear whether word-internal syntax is derived in accordance with the Mirror Principle (Baker 1985), with the order of inflectional morphemes reflecting successive-cyclic head movement in the syntax. In Chomsky’s (1995a) checking theory, verbs come out of the lexicon fully inflected, and raise, overtly or at LF, to designated functional heads to ‘check off ’ their inflectional features. It is, however, not further specified how verbal stems are combined with affixes to give rise to a fully inflected lexical array/numeration. Moreover, syntactic computation manipulates formal features on terminal nodes, which may or may not correspond to inflectional material (van Kemenade 2007: 160). Even where the functional superstructure is spelled out in full, the syntax– morphology relation turns out to be far more indirect than would be expected under a checking theory of inflection. The Rich Agreement Hypothesis exemplifies the complexity of the problem. In its earliest presentation, it was interpreted as a causal relation between the presence of highly structured inflectional paradigms in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
Table .. The typology of endogenous and exogenous change Type of change
Directionality of change
Regularity of change
endogenous syntactic change exogenous syntactic change endogenous morphological change exogenous morphological change
syntax → syntax morphology → syntax morphology → morphology syntax → morphology
regular irregular irregular regular
a language and verb raising to Agr(eement) and Tense heads. A diachronic corollary of this hypothesis is that the simplification of morphological properties and paradigm erosion leads to a decrease in verb fronting processes (Roberts 1993). It is nowadays agreed upon by most syntacticians that only a weaker one-way implication of the Rich Agreement Hypothesis is empirically sustainable: the more structured a language’s paradigms are, the more likely it is for that language to have independent verb movement (see i.a. Bobaljik 2001, 2002a; Thráinsson 2003; Biberauer and Roberts 2010). In spite of the complex relation between syntax and morphology, the two components may also change independently. The syntax is not inert, but rather flexible and dynamic and, because of this, liable to change over time. Syntactic variation and change can happen spontaneously, endogenously without any interface pressure from the morphology or the lexicon playing any role (Reintges 2009). The same holds true for endogenous morphological change, which involves paradigm-internal and crossparadigm developments, which are not amenable to sound change (Reintges 2011b). We thus arrive at a more fine-grained typology for syntax–morphology interrelations in historical grammar change, as represented in Table 8.1. Based on a case study on the evolution of the verbal tense system in Later Egyptian, I will present evidence for exogenous morphological change, in which ‘new’ morphology either preserves ‘old’ functional superstructure or activates the topic-focus field in the left periphery of the clause. The systematic way in which morphological innovation and change proceed is reminiscent of the rule-governed behaviour of syntactic change. The increase of morphological complexity that accompanies morphological renewal is not just a module-internal process, but rather driven externally by prior syntactic change.
. The syntax–morphology connection in historical grammar change In generative historical syntax, the global effects of parameter resetting have traditionally been related to the opacity of a morphologically cued parameter, with that opacity following from the erosion or loss of inflectional material. The directionality of grammar change is therefore from the morphology to the syntax. By contrast, functionally oriented approaches to morphosyntactic change view word-internal syntax as a faithful, though fossilized, representation of earlier syntactic structure—a
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
position formulated most forcefully by Givón (1971: 413) in his much-cited aphorism ‘Today’s morphology is yesterday’s syntax’. I will briefly review the issues involved and present my own proposal about how the syntax drives morphological innovation and change. .. Erosion and loss of inflectional morphology as a trigger for syntactic change It is an old idea that morphological change entails erosion and loss of inflectional material. According to Georg van der Gabelentz (1891: 256), old morphology must first go to zero before morphological renewal makes itself felt. The affixes grind down and finally disappear without a trace; yet their functions or similar functions remain and need to be expressible again. The expression is provided by the method of the isolating languages through word order or through clarifying words. In the course of time, the latter are again prone to agglutination, grinding down and loss. [My translation from German]
The idea that syntactic change is linked to prior morphological change appears in one form or the other in much of the theoretical literature. For instance, Lightfoot (2006: 90–102) adduces the well-studied history of the English modals as a prime example of the syntactic consequences of morphological simplification. The loss of inflectional endings had consequences for the category membership of modals, which in turn had repercussions for core syntactic processes. Having been stripped of person-number markings, modal verbs were reanalysed as auxiliary verbs, which are directly merged into the infl node. Fischer et al. (2000: 16) contend, however, that, to the extent that there is an abrupt shift, the history of the English modals is not an instance of grammar change. From a somewhat different perspective, Roberts and Roussou (2003) argue that the locus of parameter change is parameter expression. If the parameter expression is robust enough, it will ultimately lead to the correct parameter setting. Addressing the question of what constitutes a robust parameter, the authors note that many parameters are expressed morphologically. Accordingly, when the parameter expression is obscured by independently operating morphological and phonological changes, the parameter will be set for a value compatible with the input data, but may still differ from that of the target grammar. Due to the computational conservatism of the learning device (Clark and Roberts 1993: 317–19), the lack of a morphological cue will lead to the reanalysis of the input string in terms of a simpler hierarchical representation. There is, however, no a priori reason why morphological change should always lead to the loss of exponence and the associated inflectional category. Rather, it could equally well be possible for morphological systems to gain in complexity over time. In this connection, Lightfoot (2006: 101) makes the interesting observation that:
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
Not many of the world’s languages have a richly recorded history, but many that do have undergone morphological simplification, sometimes with category changes. If our historical records included languages with increasing morphological complexity, we would be in a stronger position to relate morphological and categorial changes. [Emphasis in the text]
In this chapter I intend to show that Ancient Egyptian (Afroasiatic, around 3500 bc to ad 1300) provides the case of a historically well-documented language in which the drift towards greater analyticity leads to increasing morphological complexity in the verbal tense system. .. Morphological complexity arising from fossilized syntax From a functional perspective, Givón (1971, 1979, 2009) takes the strong position that morphological complexity derives from syntactic complexity via grammaticalization processes that involve the combination and the subsequent reduction of syntactic structures. Dahl (2004, 2009) understands complexity as a process of maturation over time, where this approach is largely based on Bybee, Perkins, and Pagliuca’s (1994: 20) conception of grammaticalization as ‘the dynamic co-evolution of form and meaning’: the more grammaticalized a linguistic sign becomes, the more it will tend to lose its autonomy. On this view, increasing morphological complexity is associated with higher positions on grammaticalization clines of the following kind. (1)
Development of grammatical patterns free > periphrastic > affixal > fusional
(Dahl 2004: 106)
Rather than having three developmental stages (parataxis, syntaxis, lexis) with a steady increase of tightness and condensation, as in Givón’s (2009) theory, Dahl (2009: 246–8) proposes that, in grammatical evolution, different kinds of integrative processes may take place simultaneously and to a large extent presuppose each other. After having been derived from two-clause precursors with separate intonation contours, syntactic constructions with single intonation contours may further develop either into inflectionally marked words or, alternatively, into morphologically complex words. However, a caveat is in order here about the use of affix order as a tool for syntactic reconstruction. As Anderson (1992: 348–50) and Dimmendaal (2011: 149–51) cogently point out, it is not feasible to identify all of ‘today’s morphology’ with ‘yesterday’s syntax’, since not all affixes have a potential syntactic source. Even where morphological structure reflects fossilized syntax, the sentence construction in question does not necessarily reflect the basic constituent order of the language at the time when independent words started to form a prosodic unit with the constituent they modified. .. General principles of exogenous morphological change I will now articulate some more specific principles which govern syntax-driven changes in morphological systems. My proposal shares with Givón’s and Dahl’s approach, as just outlined, the basic idea that increasing morphological complexity
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
is contingent on prior syntactic reanalysis. My proposal differs from their approach in various ways. To begin with, new morphology need not always evolve from fossilized syntax, as new patterns may emerge through the recovery of functional superstructure which is no longer operational. I will furthermore assume, with Roberts and Roussou (2003) and Roberts (2010), that grammaticalization clines are the result of successive upward syntactic reanalysis. As lexical and functional items grammaticalize, they move, so to speak, up the tree diachronically, with categorial reanalysis having the effect of eliminating previous movement relations. Against this background, the central hypotheses about the locus and momentum of exogenous morphological change can be stated as follows: (2) Exogenous morphological change and morphological renewal (i) Exogenous morphological change arises as a consequence of previous syntactic change. The regularity and systematicity of this type of morphological change mirrors the regular and constrained pattern typical of endogenous syntax change. (ii) Exogenous morphological change encompasses the recovery of pre-existing movement space. If, after the loss of verb-fronting operations, the functional superstructure involved in such processes is retained and not pruned, it provides the relevant terminal nodes for newly developed morphological material. (iii) Syntax-driven morphological change may lead to increasing complexity in the inflectional system, both in terms of the number of morpho-semantic distinctions and in terms of broader syntactic distribution. (iv) Increasing morphological complexity may, but need not, be entirely economical, with the morphological expression of a parameter bordering on formal ‘over-coding’ and even redundancy. It thus differs systematically from the better-known cases of morphological simplification, in which a parameter expression is ‘under-coded’ and eventually lost beyond recognition. To summarize the argument so far, I have proposed that the locus of morphological complexity is syntactic complexity, defined in terms of the recovery of former verbmovement space. Another possibility is the permanent activation of the topic-focus field in languages with discourse-configurational syntax.
. Two diachronic scenarios for the rise of morphological complexity My point of departure here is that, even though morphological structure and syntactic structure are not isomorphic, syntax must be in a systematic mapping relationship with the morphology, providing the necessary vehicle for it in the form of functional superstructure. Because of this, the morphology varies as a consequence of variation in syntactic structure. Bobaljik and Thráinsson (1998) formulate an attractive proposal
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
pointing in this direction. In languages with a positive setting for their Split-IP parameter, the inflectional domain is subdivided into separate agr(eement) and Tense projections, as shown in diagram (3b). In languages in which this parameter is set negatively, there will only be single, fused infl projection, as shown in diagram (3a). (3)
a. Fused infl-node
b. Split infl-node agr-P
IP Subject
IP I°
Subject VP
agr-P TP
agr° Subject
TP Tense°
VP
From an economy perspective on derivation and representation (Chomsky 1995a), functional superstructure is only projected when needed. When this structure is no longer accessed by grammatical operations, it would normally be pruned, unless—and this is a new claim—it can be ‘recycled’, as it were, to furnish insertion sites for newly developed grammatical markers. As represented in diagram (4), the first scenario relates to the loss of verb movement at one stage of the development, and the recovery of the erstwhile target positions for verb movement by means of auxiliary verbs or verbal particles. (4)
CP Auxiliary → C° tam particle Auxiliary → tam particle Auxiliary → tam particle
Focus-P
Fin(iteness)-P
Focus°
Fin°
IP Subject
Auxiliary → tam particle
IP I°
VP
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
The second scenario for increasing morphological complexity over time concerns the activation of functional superstructure that has previously been accessed by highly marked syntactic reordering processes, involving a departure from the canonical surface word order. This scenario can be observed in the context of a typological shift from a subject-prominent to a topic-prominent language (Li and Thompson 1976). Topic-prominent languages organize their syntax around information structure. Given that a single syntactic head can merge one specifier at most, a variety of ¯ dislocated and A-moved phrases end up as the single specifier of designated topic or focus phrases in the cartography of the clausal left periphery (see i.a. Rizzi 1997, 2001; Frascarelli and Hinterhölzl 2007). The point of interest here is that the terminal nodes of these discourse-related projections may start out as null heads, which seek to be morphologically realized by inserting into them new functional elements. The diagram in (5) further illustrates. (5)
CP C°
Topic-P Topic-DP
Topic-P
Topic particle → Topic°
Focus-P
Focus-DP Focus particle →
Focus-P
Focus°
Topic-P
Preposed Adverb tam particle
→
Fin-P
Fin°
IP Subject
IP I° verb
VP
In the remainder of this chapter, I shall attempt to find ways to distinguish the first scenario, which involves the recovery of functional superstructure that is no longer operational, from the second scenario, in which the activation of the topic-focus field provides an incentive for the development of novel morphological categories and their exponents.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
. Analytic drift and the recovery of verb movement space My focus here will be on evolutionary trends in Later Egyptian. Traditionally, Ancient Egyptian language history has been subdivided into two macro-stages. The first stage is Earlier Egyptian, which comprises Old Egyptian (2650–2150 bc) and Middle Egyptian (2000–1300 bc), and the second is Later Egyptian, which subsumes Late Egyptian (1300–700 bc), Demotic (700 bc–ad 400) and Coptic (ad 250–1300). The most salient change in the language’s history is the Sapirian ‘drift’ from a predominately agglutinative language (Old Egyptian) to a thoroughly analytic language (Coptic). Analytic drift is a cumulative kind of change, which involves at least three independent developments, namely (i) the expansion of the auxiliary verb system, (ii) the proliferation of auxiliary verb-plus-infinitive constructions, and (iii) the spread of ‘do’periphrasis (Reintges 2013). In view of the extensive differences between synthetic and analytic systems, Roberts and Holmberg (2010: 43) conclude ‘analyticization is loss of movement, associated to some degree with loss of morphology’. A hallmark of this longterm morphosyntactic change is, indeed, the exclusion of verb-initial orders derived by long verb movement into the complementizer domain (i.e. ‘Verb Second’ and related verb-fronting processes). Crucially, however, the analyticization process did not lead to morphological simplification, but rather to the formation of a more elaborate verbal tense system. Because of this, the analytic temporal system of Coptic Egyptian is morphologically more complex than its precursors in Late Egyptian and Demotic, which still have a fair amount of synthesis. .. Main trends of the analytic drift Old Egyptian can be classified as a synthetic-agglutinative language, in which concatenative and non-concatenative morphology encodes a range of temporal and aspectual distinctions as well as grammatical voice. Consider in this regard the causative pluractional verb form s-n-fx-fx-n (‘has totally released’), which comprises from left to right the causative prefix s-, the obsolete lexical prefix n-, the verbal stem fx (‘to release’), the reduplicant fx, and the stem-final Perfect tense suffix -n. Accordingly, the maximally inflected verb in this stage expresses up to four distinct categories. (6) Agglutinative stem extensions with four inflectional categories (Old Egyptian) s-n-fx-fx-n Gbb »r-t(j)=f Çr DÇ Çt(j)-nxt caus-pfx-release-plur-perf Geb jaw-f.du=poss.3m.sg for Thoth-nakht tn dem.f.sg ‘(The god) Geb released his two jaws for this Thoth-nakht (here) (the deceased female).’ (Coffin Texts VI 102b/B3Bo)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
Classic Middle Egyptian, the language of the classical literature, displays many novel features, such as frequent recourse to auxiliary verbs within the verbal group. The expansion of the auxiliary verb system has far-reaching consequences for the syntax insofar as verb-initial VSO clauses are gradually replaced by auxiliary-initial AUX-SVO clauses. The minimal sentence pair in (7a–b) illustrates this diachronic trend. (7) Verb-initial vs. auxiliary-initial VSO constructions (Classic Middle Egyptian) xrw=f jw=f mdw=f a. sdî m-n=j hear-perf=1sg voice.m.sg=poss.3m.sg aux=3m.sg speak.pfv=3m.sg ‘I heard his voice while he was speaking.’ (Sinuhe R 25) xrw qrj b. »Ç»-n sdî m-n=j aux-perf hear-perf=1sg voice.m.sg thunder.m.sg ‘And then I heard a thunderclap.’ (Shipwrecked Sailor 56–7) The proliferation of auxiliary verb-plus-infinitive constructions represents another development within the general analytic drift. Infinitival tenses, as they are traditionally called, are based on the locative schema, in which an auxiliary verb is combined with an infinitival clause headed by a locative preposition. From Late Egyptian onwards, these originally biclausal structures undergo syntactic reanalysis and clause fusion, with the concomitant deletion of the prepositional element. This is illustrated by the contrast between examples (8a) and (8b). (8) The biclausal vs. monoclausal structure of infinitival tenses (Late Egyptian) a. wn-jn p≈-+rj [Çr Çms Çr jr-t hrw aux-foc def.m.sg-boy.m.sg at sit.inf at do-inf day.m.sg nfr m p≈y=f-pr] nice.ptcp.m.sg in def.m.sg=poss.3sg.m-house.m.sg ‘The young fellow sat down and spent a holiday in his house.’ (Doomed Prince 7, 14) b. xr tw=j sm≈ p≈-xfty n(y) p≈-R» pcl pres=1sg slaughter.inf def.m.sg-enemy.m.sg link def.m.sg-Re m-mnt in-daily ‘And every day I slaughter the adversary of the (sun-god) Re.’ (Horus and Seth 4, 4–4,5)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
Yet another manifestation of the analyticization process is the rise and spread of the periphrastic ‘do’-construction, in which the light verb jrj (‘to do’) conveys a highly schematic ‘action’ meaning, while the lexical predicate is specified by the infinitival verb. ‘do’-periphrasis is first used with defective verbs that cannot be inflected for tense and aspect themselves. Pluractional verbs such as hb-hb (‘to traverse’) are a case in point (9a). Subsequently, the construction is used to replace increasing unproductive synthetic patterns. In this context, the original action semantics of the light verb is already bleached, as witnessed by its compatibility with bona fide unaccusative verbs such as mwt ‘to die’ (9b). (9) The periphrastic ‘do’-construction (Late Egyptian) d3 w-w Çr ÇÇ a. jry=f hb-hb do.pfv=3m.sg traverse-plur.inf mountain-m.pl at search.inf pÇ sw attack(-ptcp.m.sg) cl.m.sg ‘He (the King) traversed the mountains, while he was searching for the one who attacked him’. (papyrus Anastasi II 4, 1) m s-wt=w b. i-jr=w mwt exph-do.pfv=3pl die.inf in place-f.pl=poss.3pl ‘They died on the spot (lit. in their places).’
(Wenamun 2, 52)
Coptic has gone very far in abandoning its former synthetic features and has developed into a thoroughly analytic SVO language, in which tense-, aspect-, and mood-indicating (tam) particles precede the subject and the lexical verb stem in linear order. (10) TAM-SVO order in the thoroughly analyticized verbal system (Coptic) a tә-sophia ket u-εï na=s pfv def.f.sg-wisdom.f.sg build.nom indef.sg-house for=3f.sg ‘Wisdom has built a house for herself.’ (Proverbs 9, 1) tam particles are located in an uplifted Tense position that corresponds in structural height to the Fin(iteness) head of the Rizzian (1997) cartography. Tense dominates a clause-internal modal projection, which contains in its specifier the clausal subject. The modal projection dominates an aspect projection on top of the vP. The skeleton of functional projections Tense > Mood > Aspect is schematically presented in diagram (11) (see Reintges 2011a: 554–7 for further analysis and explication).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity (11)
TP
Tense° tam particle
Mood-P
Specifier Subject
Mood-P
Mood° verb
Aspect-P
Specifier Object
Aspect-P
Aspect°
VP
……
Illicit long verb movement Licit short verb movement into the C-domain to Aspect° and Mood°
The merge of verbal tam particles in pre-subject position creates a situation reminiscent of that of Welsh, in which long verb movement into the left periphery is excluded (Roberts 2005: 120–4; Rouveret 2010: 260–1; cf. also Borsley et al. 2007: 287–98 on the loss of Verb Second in pre-modern Welsh). Accordingly, Coptic no longer displays the Verb Second effects that are still observable in Earlier Egyptian. The interested reader may refer to Reintges (2005: 76–80, 2012a: 146–52) for discussion of the different scope of verb movement of imperfective and passive verbs in Earlier Egyptian. Even though analyticization severely restricts the movement space of verbs, lower functional projections are still available as landing sites. It was therefore possible for the language to retain residual verb movement for the purpose of aspectual and modal licensing. .. Particle movement around a clitic left-dislocated subject DP In this section, I will present evidence from Coptic particle syntax for the recovery of movement-related functional superstructure. Cinque (1999: 189, note 22)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
distinguishes particles from auxiliary verbs ‘in being less prone to movement (perhaps as a consequence of their being poorer in features)’. Although in essence correct, the picture needs fine-tuning when it comes to the Coptic particle system. In a syntactic variant of clitic left-dislocation (CLLD) of the subject, the selected tam particle surfaces in two different positions. The lower particle copy appears in the canonical pre-subject position, while the higher copy ends up in a functional head on top of the CLLDed subject DP. (12)
Perfect particle a > CLLDed Subject DP > Perfect particle a > resumptive pronoun de әm-pә-ma et әmmau a=u‰i a ne-r:‰mei perf def.pl-man.m pcl link-def.m.sg-place rel there perf=3p weh pә-sc:‰ma әm-pә-makarios Apa Mε‰na epesεt put.nom def.m.sg-body link-def.m.sg-blessed Apa Mena pcl ‘The people of that place put down the body of the blessed Apa Mena (. . .)’ (Mena, Martyrdom 5a, 14–20)
The particle movement-and-copying construction does not exhibit a root/non-root asymmetry, as the higher particle can appear to the right of the finite complementizer t + e ‘that’. The position targeted by particle movement must therefore be lower than C◦ . (13)
COMP > perfect particle a > CLLDed subject DP > perfect particle a > resumptive pronoun a pә-hikanos t+ e a=f әm-pә-dynatos comp perf def.m.sg-sufficient link-def.m.sg-mighty perf=3m.sg ti si+e na=ï emate give.nom grief.m to=1sg much ‘Since the Almighty One has given me a lot of grief ’ (Ruth 1, 20)
In previous work (Reintges 2011a: 562–7) I have proposed an analysis of the particle movement-and-copying construction in terms of verum focus, which is broadly associated with emphasis on the truth of the proposition. Virtually all examples of this construction can be analysed as emphatic assertions, in which verum focus signals the update of a given proposition from the conversational background. In serving similar purposes as affirmative do-insertion in English, particle doubling represents an essentially iconic way to express focus and contrastive emphasis. The syntactic correlate of this focus type is a null operator in the specifier position of a leftperipheral focus projection. The clitic-resumed subject DP, on the other hand, must be defocalized and backgrounded by being marked as a familiar topic; as such it occupies the lowest topic projection in Frascarelli and Hinterhölzl’s (2007) refined cartography of the topic-focus field.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
When formalized in terms of feature valuation (Chomsky 2001), the uplifted Tense head carries an uninterpretable focus feature, which must be matched against the interpretable focus feature on the Focus head; vice versa, the Focus head has an uninterpretable tense feature, which must be valued against the interpretable tense feature on the pre-subject tam particle. The resulting agree-relation between the probing Focus head and the probed-for tam particle is, however, not a sufficient structural relation, but must rather be supplemented by narrow-syntactic head movement, followed by the pronunciation of the displaced particle both in the tail and in the head position of the movement chain. The lower particle copy must be phonologically realized to provide an appropriate verbal host for the enclisis of the resumptive subject clitic, thereby ensuring a convergent derivation at PF. The probe-goal configuration of the particle movement-and-copying construction is represented in diagram (14). (14) Focus-driven particle movement around a CLLDed subject DP CP C° comp
Focus-P OpVerum Focus
Focus-P
Focus [ifoc, uT]
Familiar Topic-P
CLLDed Subject DPi tam Particle
Familiar Topic-P
Familiar-Topic
TP Mood-P
Tense [iT, ufoc]
Resumptive
Mood-P
tam Particle pronouni
Particle movement
Topicalization of the subject DP
Adopting Rizzi’s (1996) ‘Residual Verb Second’ proposal in spirit (though not in the exact execution), we may then say that higher functional heads once targeted by Verb Second and related processes may retain the feature structure relevant to movement, even long after the relevant fronting operation ceases to exist. In other words, the functional heads C◦ , Focus, and Fin◦ may continue to act as probes that establish an agree relation with a lower verbal category—in our case, the verbal tam
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
particle in pre-subject position. The proposed analysis has interesting consequences in that Coptic tam particles behave syntactically more like finitely inflected lexical verbs rather than auxiliaries. This diachronic pattern makes sense if certain particles were directly grammaticalized from main verbs (skipping an intermediate auxiliary stage) or if the transition from auxiliary verb to verbal particle was not long enough for the grammaticalizing item to lose the formal features necessary to entertain agree- and movement-based relations.
. Patterns of morphological innovation and change in Later Egyptian .. The Equal Complexity Hypothesis The received wisdom that all languages are ultimately about equally complex has been subjected to serious criticism in recent years (cf. i.a. Kusters 2003; Dahl 2004; Shosted 2006; Nichols 2009; for a historical overview, see Sampson 2009). The Equal Complexity Hypothesis entails a trading relationship between different parts of the grammar, whereby simplicity in one part is compensated for by complexity in another. As Bane (2008: 69) points out, this hypothesis has far-reaching consequences for the study of historical grammar change, since ‘language change must be subject to a sort of “Newton’s Third Law of Motion”, whereby linguistic complexity remains constant over time’. However, it seems very unlikely that increasing morphological complexity is achieved at the expense of syntactic complexity. In involving phrasal extraction of the subject and the direct object, Coptic has a derivationally more complex syntax for argument licensing and Case than Old and Middle Egyptian, in which verbal arguments are Case-licensed in situ in their vP-internal merge positions (see Reintges 2012a: 145–55 for the detail of the analysis). Theories of complexity in language share the guiding assumption that more structural units, rules, and representations imply more grammatical complexity, and that smaller structural units are more clearly definable than rules and representations (Hawkins 2009: 252). To assess variable degrees of morphological complexity across languages, the number of morphemes in a domain as well as the number of allomorphic and paradigmatic alternatives is counted in Greenberg’s (1960) study of morphological typology. This essentially taxonomic approach has been implemented in Nichols’ (2009) typological approach to grammatical complexity, without, however, further specifying a matrix for quantifying and measuring complexity (but see Shosted 2006 and Bane 2008 for proposals in this direction). I will tackle the question of morphological complexification from a somewhat different angle and look at two concrete examples of morphological innovation. .. Case study I: The diachronic rise of relative tenses The first case study deals with the diachronic pathway of Coptic relative tenses. As I will show, relative tenses register filler-gap dependencies in relative clauses, constituent and polarity questions as well as a range of declarative focus contexts,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
thereby distinguishing these focus-sensitive constructions from simple declaratives. The precursors of this special type of morphosyntactic flagging are the so-called emphatic tenses in Late Egyptian and Demotic. Emphatic tenses are inflectionally marked by the vocalic prefix ỉ- (traditionally called the prothetic yod; cf. Polotsky 1944: 70 §25 footnote 3; Winand 1992: 280 §444). The affirmative–negative pair in (15) provides an illustration. (15)
Selection of emphatic verb forms in contrastive focus contexts (Late Egyptian) ỉ-dî d=n m m≈»t Pr-≈» bn dî d=n emph-speak.pfv=1pl prep truth.f.sg Pharaoh neg speak.pfv=1pl »dÇ ≈ lie.m.sg ‘We are speaking the truth, Pharaoh; we are not lying.’ (Ostracon Nash 2, recto 12–13) (= KRI IV 318: 11–12)
What interests us here is that Late Egyptian and Demotic emphatic tenses are associated with focus and contrastive emphasis, but show an irregular syntactic distribution. This suggests that this focus-sensitive verbal pattern has not yet been paradigmaticized. To illustrate this point, consider the adverbial wh-in-situ questions in (16a–b), in which the emphatic marker surfaces on the light verb of a complex predicate, while the corresponding simplex verb is left unmarked for emphaticity. (16) Adverbial wh-in-situ questions with and without emphatic verb forms (Late Egyptian) jy Çr jx n(y) sÇn ? a. ỉ-jr=k emph-do.pfv=2m.sg come.inf on which link order.m.sg ‘On which authorization did you come (here)?’ (Wenamun 2, 3) nfr ? b. jy=k tnw p≈-+rj come.pfv=sm.sg where def.m.sg-boy.m.sg handsome-ptcp.m.sg ‘Where did you come from, handsome boy?’ (Doomed Prince 5, 10–11) Emphatic tenses demonstrate an increasing frequency in Demotic. Nevertheless, their presence is by no means obligatory for the grammaticality of the construction (Quack 2006: 257–8). This point can conveniently be illustrated with the example of adverbial wh-in-situ. (17)
Wh-in-situ questions with and without emphatic periphrastic tenses (Demotic) w+d nm m-s≈ a. ỉ-jr=n emph-do.pfv=1pl honour.inf who except p≈-mr »-wj def.m.sg-superintendent.m.sg house.m.sg ‘Whom shall we honour except for the superintendent of the house?’ (Saqqara Demotic Papyri I, text 2, x+1, 27)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges jx? b. ỉ-jr=t jy r Nwt r-dÇ b≈ emph-do.pfv=2f.sg come.inf to Thebes because.of what ‘Why are you coming to Thebes?’ (papyrus Ryland IX 9,15) jx? t≈j r–dÇ b≈ c. jw=k dî d aux=2m.sg speak.inf dem.f.sg because.of what ‘Why are you saying this?’ (Inaros, papyrus Krall VI, 5)
The Demotic situation is therefore still very far removed from the Coptic one, in which relative tenses have been fully paradigmaticized. More importantly, however, there is no continuity in morphological exponence, since emphatic inflection has almost completely been replaced by alternating relative particles, which are sensitive to the tense and polarity of the embedded clause (see Reintges 2012b for further details of complementizer allomorphy). In expressing the core relativization function of attribution and subordination, these relative particles can readily be identified with finite relative complementizers, albeit with an interesting twist. Due to their reanalysis as a morphosyntactic flagging device, they are grammatically required and/or permitted in a variety of non-relative environments: wh-in-situ and polarity questions, declarative focus sentences, secondary (depictive) predication, temporal adverb clauses, and asymmetric clause coordination. The distributional behaviour of relative tenses has been subject to intensive study (Reintges 2007, 2011a, 2012b; Reintges and Green 2004; Reintges, LeSourd, and Chung 2006). Examples (19a–c) feature relative clauses as well as matrix and embedded wh-in-situ, where the presence of this special relativization morphology is obligatory. (18)
Selection of relative tams in relative clause and wh-in-situ questions (Coptic) әnhε‰tә=f] a. pә-ma [әnt a=k kj әntә=f def.m.sg-place.m rel perf=2m.sg find=3m.sg inside=33m.sg ‘The place where you found it (the boat)’ (Acts Andrew & Paul 204, 145–6) na=u‰ әnto? b. e +are әr u rel hab.2f.sg do what to=3pl you.f.sg ‘What are you doing with them (the torn clothes)?’ (Shenoute, Amélineau I 1, 108, 9)
As has long been observed in the Egyptological descriptive tradition, Late Egyptian and Demotic emphatic tenses are not morphologically differentiated for present and past tense reference (Erman 1933: 263–5 §§545–50; Polotsky 1944: 71 §26; Quack 2006: 252–5; see also Winand 1992: 284–5 §§451–2 for a less convincing alternative). By contrast, Coptic relative tenses employ different complementizer allomorphs to mark the relevant tense distinctions. Thus, the complementizer particle әnt must be selected in past-tense contexts marked by the Perfect particle a. The relative particle e represents a default form selected in all other affirmative tenses. The 3rd person masculine form of Late Egyptian and Demotic emphatic tenses and the corresponding Coptic relative tenses are represented in Table 8.2. The triconsonantal verb sd Ç m/s:‰tәm ‘to hear’ has been chosen to illustrate a typical paradigm.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
Table .. Tense-neutral emphatic tenses vs. temporally distinctive relative tenses
Present Past Future
Late Egyptian
Demotic
Coptic
ỉ –jr=f sdÇ m ỉ –jr=f sdÇ m ỉ –jr=f sdÇ m
ỉ –jr=f sdÇ m ỉ –jr=f sdÇ m ỉ –jr=f sdÇ m, ỉ –jr jw=f r sdÇ m
e=f s:‰tәm әnt a=f s:‰tәm e=f na s:‰tәm, e=f e s:‰tәm
The grammatical evolution of relative tenses represents a complex case of morphological innovation and change, which, on the one hand, involves the paradigmatization of an already existing pattern and its spread to new domains and, on the other hand, the substitution of an unproductive morphology by functionally versatile complementizer particles, which enter into a paradigmatic opposition with one another. .. Case study II: Emergent modal-evidential categories The second case of morphological renewal concerns the emergence of entirely novel inflectional categories. Two cases in point are the conditional mood and the inferred evidential. The Coptic conditional is a compound tense formed with the relative marker e and the preverbal modal particle ∫an. The conditional has a restricted syntactic distribution and is used to introduce the protasis of a conditional sentence. When the following consequent clause contains a future tense, the conditional assumes a factual interpretation, describing a hypothetical situation that has the potential of becoming reality (Reintges 2004: 322–33 §8.2.3.2). (19)
Realis interpretation of the conditional mood (Coptic) e=f +an eï +aro=n tәn na kj en hεu rel=3m.sg cond come to=1pl 1pl fut gain profit.m tεrә=n hitәn ne=f-+lεl entire.m=poss.1pl through def.pl=poss.1pl-prayer.m ‘(And I believe that) if he comes to us, we would all benefit from his prayers.’ (Hilaria 10, 30–1)
There is only a single attestation of the cognate form ∫»-n≈j in Late Demotic (Johnson 1970). As with the Coptic conditional, ∫»-n≈j co-occurs with an initial focussensitive element—the emphatically inflected light verb ỉ-jr=k ‘you do’. (20) Hapax legomenon of the cognate form ∫»-n≈j (Late Demotic) n≈-s[-w [ntj-jw-jw=k »+=w def.pl-writing-m.pl rel-aux-aux=2m.sg recite.inf=3pl ỉ-jr=k +»-n≈j wd=w r p≈j=w-m≈»] emph-do.pfv=3pl pcl send.inf=3pl to def.m.sg=poss.3pl-place.m.sg ‘The writings that you should recite (them) when you dismiss them (the gods) to their place’ (London/Leiden Papyrus 3, 29–30)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
The inferred evidential is formed with the auxiliary tare. As an isolated example of an introspective evidential, it signals the speaker’s belief in the cogency and robustness of the inference that he or she is making (Reintges 2004: 324–7 §8.2.4). (21)
The speaker-oriented inferred evidential (Coptic) s:‰tәm pa-+ε‰re t+ in te=k-mәnt-kouï def.m.sg.poss.1sg-child.m since def.f.sg=poss.2m.sg-childhood listen.imp әn-te-sßo tare=k he e-u-kharis +a to-def.f.sg-teaching.f infer=2m.sg find prep-indef.sg-grace.f until te=k-mәnt-hllo def.f.sg=poss.2m.sg-old.age ‘My child, listen to the teaching from your childhood onwards and (I assure you) you will find grace until your old age!’ (Sirach 6, 18)
The inferred evidential is historically derived from analytic causative construction with a first person singular subject tw=j ‘I give’ (Polotsky 1944: 12–13 §6). There are but a few attested examples in Late Demotic, in which the causative verb has a transparent evidential meaning. (Late Demotic) (22) 1st sing.causative construction with evidential connotation xm b≈t tw=j »j be.modest.imp character.f.sg caus=1sg increase.pfv rmt+ nb t≈j=k-+f»t n Ç≈tt “ def.f.sg=poss.2m.sg-status.f.sg in heart.f.sg man.m.sg each.m.sg ‘Be modest of character and (I assure you) your reputation will increase in the heart of everyone.’ (‘Onchsheshonqy 17, 26) The developmental path of the inferred evidential is typologically highly exceptional. Causative verbs are completely missing in Aikhenvald’s (2004: chapter 9) survey of diachronic sources for evidential markers. This points more generally in the direction of an accelerated grammaticalization process during which Coptic particles have been created out of a ‘mixed bag’ of auxiliary verbs, relativizing complementizers, focus particles, and other functional elements.
. Discourse-configurational syntax as a trigger for morphological renewal The renewal of the modal system is hardly a historical accident, as modal and evidential categories are connected to higher functional projections in the leftperiphery (Cinque 1999: 53–6, 85–6). Neither is the rise of the relative tenses, which are distributed over several left-peripheral positions (C◦ , Focus, Fin◦ ). I propose here that the language’s move towards a discourse-configurational syntax made available a number of functional nodes to host new grammatical markers.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
Discourse-configurational languages tend to have an articulated left periphery. Just as in Italian (Rizzi 1997), Hausa (Green 2007), and Koine Greek (Kirk 2012), the Coptic left-periphery is lower-bounded by the Fin◦ /Tense◦ head, into which presubject tam particles are merged, and upper-bounded by the C◦ /Force◦ node, which typically hosts clause-typing particles or subordinating complementizers. Evidence for an articulated left periphery has already been discussed earlier in connection with the particle movement-and-copying construction (§4.2). Consider next the embedded topicalization structure in (23), in which the clitic-resumed independent pronoun anon ‘we’ appears to the right of the embedding complementizer t+ekas ‘such that’. In semantic terms, it corresponds to an aboutness-shift topic. This is also confirmed by the modifying Greek loan particle men, which has a topic-shifting function. In Frascarelli and Hinterhölzl’s (2007) cartographic approach, aboutness-shift topics are located in the topic projection on top of the focus phrase, which in this example is instantiated by the relative particle e. (23) comp > aboutness-shift topic > relative particle > subject pronoun > future particle > verb > prepositional object t+ ekas anoni men e=tetәn na nehse әmmo=ni e8ol hәm comp we pcl rel=2pl fut awake prep=1pl pcl from әn-hinε8 pә-nokj def.m.sg-big link-sleep.m
(Coptic)
‘. . .such that, as far as we are concerned, you would wake us up from a deep sleep’ (Zenobius 199, 17–18) Coptic can typologically be classified as a wh-in-situ language, as question words and phrases are generally not moved to the left periphery. However, wh-movement is available as a marked alternative and seems to be the preferred option with questioned cause/reason adverbs. The two interrogative strategies differ from each other morphologically in that wh-in-situ questions are formed with relative particles, whereas the corresponding wh-ex-situ structures lack this special type of morphosyntactic flagging. Reintges, LeSourd, and Chung (2006: 179–84) explain this distributional pattern as a corollary of the ‘Doubly Filled Comp Effect’, whereby either the head or the specifier of the Rizzian (2001) int(errogative) phrase receives substantial specification. (24) Wh-in-situ vs. wh-fronting questions a. eye ere ne=tәn-+ε‰re nu‰t+ e e8ol hәn nim? q rel def.pl=poss.2pl-son.m cast pcl in who ‘In whom are your sons casting out (demons)?’ y:‰rәm әns:‰=n b. eye et8e u tetәn q why (pres)2pl stare at=1pl ‘Why do you stare at us (. . .)?’
(Coptic)
(Luke 11, 19)
(Acts 3, 12)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Chris Reintges
Clitic-resumed personal pronouns can also serve as familiar topics, which are merged low in the left periphery. As expected, such topics appear following fronted wh-adverbials, occupying the specifier of the int◦ node. (25) wh-manner adverb > familiar topic > resumptive pronoun > verb (Coptic) anon tәn s:‰tәm pә-wa әn-he aw:‰ әn-a+ and in-what link-way.m we (pres)1pl hear def.m.sg-one pә-wa hrai hәn te=f-aspe def.m.sg-one pcl in def.f.sg=poss.3m.sg-tongue.f ‘And how can we hear each one in his (own) tongue . . .? (Acts 2, 8) If we assume that the presence of the int-phase excludes the projection of an independent focus phrase, a number of gaps in the Coptic data receive a principled explanation. Thus, neither multiple focus-fronting nor a combination of wh-movement and focus-fronting is attested (Reintges and Green 2004: 171–3). The distributional behaviour of different types of topics and displaced wh-phrases motivates the following structure of the Coptic left periphery. (26)
CP C°/Force° comp Topic
Aboutness-shift Topic-P Aboutness-shift Topic-P
int(errogative)-P Topic° Topic-shifting particles Fronted wh-phrases
int(errogative)-P
int° Relative particle
Familiar Topic-P
Topic DP
Familiar Topic-P
Topic°
TP/Fin-P Tense° tam Particle
Mood-P
. Concluding remarks Based on the grammatical evolution of the Later Egyptian verbal system, I have argued for a hitherto unnoticed process of endogenous morphological change, during
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Increasing morphological complexity
which new morphology either evolves from functional superstructure previously used for verb movement or is newly developed to accommodate discourse-related displacement operations. Morphological innovation and change leads to the overall complexification of the verbal tense system, both in terms of the number of possible morphosemantic distinctions and in terms of the characteristically atypical properties of newly developed functional categories and their exponents. Historical cases of increasing morphological complexity are of considerable theoretical interest as they pose a potential challenge for cue-based approaches to syntactic change. This is so because this type of morphological change yields a surplus of formal cues for a given parameter expression. This would lead one to expect—contrary to fact—that the syntax would be inert to change, since there is an abundance of morphological cues that guide the language learner in constructing the grammatical system of his or her native language. So if we find cases of syntactic change in the context of morphological innovation, we have yet another argument for the autonomy of syntax in diachronic change.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento: A syntactic or phonological phenomenon? A DA M L E D G E WAY
. Introduction The dialects of southern Italy are traditionally reported to display a Greek-style dual complementizer system.1, 2 Accordingly, these dialects distinguish between a complementizer derived from quia (> ca) and a complementizer derived from quid (> che (upper South), chi (extreme South)), quod (> cu), or modo (> mu, ma, mi): while the former heads clauses selected by declarative and epistemic predicates, the latter is employed after predicates such as volitionals that typically characterize the state or events of their complements as unrealized at the time of speaking, as illustrated in Table 9.1. Focusing on the Salento, it can be seen that, in common with the local form of Greek (griko) once extensively spoken across the region but now restricted to the Terra d’Otranto, dialects of this region such as Scorrano in the province of Lecce (LE) limit the use of the infinitive chiefly to complements of a subset of functional (restructuring) predicates (1a–b). As a consequence, dialects such as Scorrano do not exhibit the typical Romance subject obviation rule and Romance-style infinitival clauses are standardly replaced by the dual finite complementizer construction, with
1 Thanks to Federico Damonte, Cecilia Poletto, and, in particular, Christina Tortora for comments on a previous version of this chapter. 2 See Calabrese (); Lombardi (, ); Ledgeway (, , , , , : –); Damonte (, , ); Roberts and Roussou (: –); Manzini and Savoia (, I: –, –); Vecchio ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
Table .. Dual complementizer systems in southern Italy Italian Salento Sicily Sicily: Messina S. Calabria N. Calabria Naples N. Puglia Abruzzo
penso che verrà ‘I think that he’ll come’ crisciu ca vène pensu ca vèni critu ca vèni pensu ca vèni criju ca vèni pènsә ca vènә pènsә ca vènә pènsә ca vènә
voglio che lui mangi ‘I want that he eat’ ogghiu cu mmancia vògghiu chi mmanciassi ògghiu mi mancia vogghiu mu (mi) mangia vuogliu chi mmangia vògliә chU mmangә vògghiә chU mmangә vòjjә chU mmangә
propositional complements introduced by ca (2a–b) and irrealis complements by cu (3a–b).3 (1)
a. se aggiu dire propriu la verità if I.have say.inf really the truth ‘if I really have to tell the truth’ b. putimu ncignare l’ eseciziu pe la voce we.can begin.inf the exercise for the voice ‘we can begin the voice exercise’
(2) a. se ài dittu ca te tolene li peti if you.have said that to.you= ache the feet ‘if you said that your feet hurt’ b. si sicura ca nu sbaj you.are sure that not you.are.mistaken ‘you are sure that you’re not mistaken’ (3)
a. ulia cu me ne piju nu paru! I.wanted that me= of.them= I.take a pair ‘I wanted to buy a pair!’ b. m’ aggiu scurdata cu ni lu cercu me= I’ve forgotten that to.her= it=I.ask ‘I forgot to ask her for it’
Although ca and cu are both complementizers, they differ with respect to their relative positions in conjunction with fronted topicalized and focused constituents 3 With the exception of the rd person of non-st conjugation verbs and some highly irregular verbs in many, but not all, dialects of the Salento, the morphological subjunctive has been lost, with generalization of the indicative to irrealis clauses (Rohlfs : –, : n. ; Mancarella : –; Loporcaro b: –; Damonte ; Bertocci and Damonte ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway
(Calabrese 1993: 36 n. 6, 38; Ledgeway 2004, 2005). As the following examples from Campi Salentina (LE) demonstrate, while ca precedes all such fronted elements (4a–b), cu must follow (5a–b). krai (4) a. aggiu tittu ca la Lia ene I.have said ca the Lia comes tomorrow ‘I said that Lia is coming tomorrow’ b. aggiu tittu ka krai ene la Lia I.have said ca tomorrow comes the Lia ‘I said that Lia is coming tomorrow’ (5)
a. oyyu lu libbru ku lu kkatta lu Maryu I.want the book cu it= buys the Mario ‘I want Mario to buy the book’ b. oyyu krai ku bbene lu Maryu I.want tomorrow cu comes the Mario ‘I want Mario to come tomorrow’
Facts like these find an immediate explanation in terms of Rizzi’s (1997) split CP perspective, with ca lexicalizing the higher Force head and cu the lower Fin head. (6) [ForceP ca [TopP Top [FocP Foc [FinP cu [TP . . . ]]]]] A further distinction between the two complementizers which does not immediately follow from (6) concerns their PF realization (Calabrese 1993; Terzi 1994, 1996). While ca must always be pronounced in all dialects of the region (7a–b), cu may remain unpronounced under certain conditions. For example, in the dialects of northern Salento spoken in the provinces of southern Taranto (TR) and Brindisi (BR), such as Cellino San Marco (BR) in examples (7) and (8), cu may optionally remain unpronounced if subject coreferentiality obtains between matrix and embedded clauses (8a–b). By contrast, in the dialects of central-southern Salento spoken in the province of Lecce, such as in the Leccese examples (7) and (9), cu may optionally remain unpronounced whether matrix and embedded subjects are coreferent (9a) or not (9b). mme stau cu ttie (LE) (7) a. te prumettu ∗ (ca) tornu e you= I.promise that I.return and myself= I.stay with you ‘I promise you (that) I’ll come back and stay with you’ b. te l’ ia tittu ∗ (ca) è nu bravu vagnone (BR) you= it=I.had said that is a good boy ‘I had told you that he is a good boy’ (8) a. Ce bbuei (cu) ddici? (BR) what you.want that you.say ‘What do you want to say?’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
∗ (cu) ddicu? (BR) b. Ce bbuei what you.want that I.say ‘What do you want me to say?’
(9) a. nun bole (cu) ndi sàccia te fatìa (LE) not he.wants that of.it= he.knows of work ‘He’s not interested in work’ b. tocca (cu) mme ndi au te pressa (LE) it.touches that myself= therefrom= I.go of haste ‘I must leave at once’ The PF realization or otherwise of cu is further relevant to the licensing of clitic climbing. In the northern dialects clitic climbing is always possible under what may be informally termed cu-drop (10a), whereas in the dialects of central-southern Salento clitic climbing is invariably excluded regardless of the realization or otherwise of cu (10b). (10) a. lu voli ssapi (Mesagne (BR)) llu sape (Matino (LE)) b. (∗ lu) ole it= he.wants it= he.knows ‘he wants to know it’ The differential behaviour of northern and central-southern dialects raises a number of interrelated questions, which will form the focus of the rest of this discussion. Firstly, is Salentino cu-drop a phonological or a syntactic phenomenon? Whatever the correct answer to that question, this in turn should provide some understanding to the second question regarding why cu-drop proves more constrained in northern Salentino than in the dialects of the centre-south of the region. Furthermore, how are the syntactic effects of cu-drop and, in particular, its interaction with subject coreferentiality and its effects on licensing clitic climbing to be explained? Finally, how are the observed distributional differences in the realization of cu in the dialects of northern and central-southern Salento to be related diachronically? Here it will be demonstrated how the available evidence forces us to conclude that the northern dialects should be seen as an innovative variety, in which the phonologically driven type of cu-drop attested in the dialects of central-southern Salento has been radically reinterpreted as a syntactic phenomenon with concomitant structural effects.
. Cu-drop: traditional approach This section briefly outlines the traditional approach to cu-drop in the dialects of Salento, which, as the representative quotes in (11a–b) highlight, has standardly been assumed to involve a case of complementizer ‘deletion’.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(11)
Adam Ledgeway a. ‘In the Salento the conjunction can be lost entirely’
(Rohlfs 1969: 105)
b. ‘Another difference between /ku/ and /ka/ [. . . ] is that /ka/ can never be deleted [. . . ], whereas /ku/ can be deleted in many cases’ (Calabrese 1993: 81 n. 8) Accordingly, the dialects of the Salento apparently generate parallel structures like those in (12a–b) exemplified from the dialect of Lecce. (12)
a. ogghiu [CP cu [TP pparlu]] [TP pparlu]] b. ogghiu [CP Ø (that) I.speak I.want ‘I want to speak’
However, what evidence is there for postulating that cases of cu-drop involve structures like (12b)? Essentially, within the literature the principal evidence for such deletion analyses comes from the distribution of clitic climbing. For example, on the basis of the dialects of northern Salento, Calabrese (1993: 82 n. 8) argues that ‘[i]f we assume that the presence of /ku/ in COMP blocks movement through this position [. . . ], we have a straightforward account of the fact that the presence of /ku/ impedes clitic climbing and that clitic climbing is possible only when /ku/ is deleted.’ Consequently, according to Calabrese in examples such as (12b) the relevant C◦ position is structurally maintained but the lexical material associated with it (namely, cu) has been deleted, leaving the head radically empty and therefore potentially available as an escape-hatch for climbing clitics. In a similar vein, Terzi (1994, 1996) also interprets cu-drop as a case of lexical deletion of the complementizer. Her analysis is based on a novel approach to clitic climbing, which she interprets as a case of T-raising that takes along any left-adjoined clitics with it to the matrix clause. This is illustrated schematically in (13a) and with the Italian example in (13b). (13)
a. Vmatrix ...
[CP[C
] [TP [cl+T] [v-VP V cl ...]]]
b. Ci+T voglio [CP [C ci+T] [TP ci+T [v-VP andare ci]]] (It.) there+I.want go.INF there ‘I want to go there’ One of the desirable consequences of this analysis is that it allows Terzi to explain two important restrictions on clitic climbing. First, it predicts that if the C◦ position is already lexicalized, it will be unavailable as an escape-hatch for T-raising and clitic climbing will consequently be blocked (14; Kayne 1991), a conclusion which finds an immediate parallel in current theory in terms of the Phase Impenetrability Condition (PIC) which requires movement out of the embedded clause to transit through the left edge (viz. Head or Specifier positions) of the left periphery.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
(14) Non (∗ lo) so [CP [C se] [T [v-VP farlo]]] (It.) not it= I.know if do.inf=it ‘I don’t know whether to do it’ Second, by the same line of reasoning this approach correctly predicts the impossibility of clitic climbing under disjoint reference of matrix and embedded subjects, since T-raising will necessarily involve the coindexing of embedded and matrix T(ense)s under strict identity. Consequently, T-raising resulting in the coindexing of two T heads with different ϕ-feature matrices as in (15) will invariably prove ill-formed. (15)
∗ lo+T
penso [CP [C lo+T] [T lo+T [v-VP lo conoscano]]] (It.) him= I.think they.know.sbjv ‘I think they know him’
If this analysis is now applied to clitic climbing in the dialects of northern Salento, it appears to make the correct predictions. In particular, whenever the C◦ position is overtly lexicalized by cu, T-raising, and hence clitic climbing, is blocked, as illustrated in the Brindisino example (16a). If, however, the complementizer remains unpronounced, which Terzi interprets to mean that the position remains radically empty, then C◦ is available as an escape-hatch for T-raising with concomitant clitic climbing (16b), as long as subject coreferentiality obtains. (16) a. vogghiu [CP [C cu] [T lu+T [v-VP kkattu lu]]] that it= I.buy I.want b. lu+T vogghiu [CP [C lu+T] [T lu+T [v-VP kkattu lu]]] it=T I.want I.buy ‘I want to buy it’ In conclusion, it has been seen that existing approaches to Salentino cu-drop treat it as a case of deletion, in which the C-Fin◦ position must be left both phonologically and lexically empty in order to function as an escape-hatch for climbing clitics. .. Cu-deletion: problems Although the traditional approach to cu-drop in terms of a deletion analysis superficially appears to account for the Salentino data, it also raises some important problems which ultimately force us to abandon the deletion analysis. The principal problem lies in the distribution of raddoppiamento fonosintattico ‘phonosyntactic doubling’ (henceforth RF; Maiden 1995: 72–6; Loporcaro 1997a), a phonological process whereby a restricted number of typically grammatical items (e.g. a ‘to, at’, cchiù ‘more’, nu ‘not’, so/su ‘I am; they are’), synchronically forming a lexically opaque class, cause the lengthening of the initial consonant of the following word in word1 + word2 sequences (e.g. a [m:]ilano ‘in Milan’). Historically, this represents a case of sandhi assimilation of an underlying Latin final consonant (e.g. ad > a, plus > cchiù, non >
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway
nu, sum/sun(t) > so). It is not, however, a simple phonological rule whose application is based on pure linear adjacency, but shows sensitivity to structural configuration (Ledgeway 2009: 46–7), with lengthening licensed only under one of the three core local relations of Head-Comp (17a), Spec-Head (17b), and Head-Head (17c). (17)
a. [PP a [DP [k:]asa]] (LE) ‘at home’ b. [DP [Spec cchiù] [k:]afè] (LE) ‘more coffee’ c. [TP [Neg nu] + [V [k:]apisce]] (LE) not he.understands ‘he doesn’t understand’
Returning now to the Salentino data, it is to be observed that in the dialects of the region cu belongs to those words that trigger RF, since it continues the Latin complementizer quod ‘because; that’, where historical assimilation of the final dental stop has been reinterpreted as the relevant RF trigger. Thus, given an underlying verb form such as partu in (18a), it shows the expected consonantal-initial lengthening when embedded under cu (18b). What is surprising, though, is that this same lengthening obtains even when the apparent RF trigger is ‘deleted’ under cu-drop (18c). Indeed, as Rohlfs (1969: 105) highlights, ‘the conjunction can be lost entirely, leaving behind a trace in the doubling of the initial consonant of the following word’. (18)
a. [p]artu (Cellino San Marco (BR)) I.leave b. vogghiu [CP cu
[p:]artu] (Cellino San Marco (BR))
c. vogghiu [CP Ø [p:]artu] (Cellino San Marco (BR)) (cu) I.leave I.want ‘I want to leave’ The contrast in (18b–c) thus gives rise to an apparent paradox. On the one hand, there is clear evidence from (18b) that RF on the embedded verb is triggered by the irrealis complementizer cu lexicalizing the lower C-Fin head, but (18c), on the other hand, shows that, although the same position is not lexicalized, the embedded verb continues, somewhat puzzlingly, to bear the RF triggered by the missing cu. In short, we are forced into the contradiction of saying that, although an element is absent, it nonetheless has to be assumed to be present in order to account for the observed RF. One obvious solution to this paradox is to treat cu-drop as a purely phonological phenomenon, as indeed is explicit in Vecchio’s analysis of the dialect of Francavilla Fontana spoken in the province of Brindisi.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento (19)
‘This deletion rule is purely phonological, and not syntactic, as witnessed by the fact that an underlying /ku/ continues to trigger lengthening of the initial consonant (so-called raddoppiamento (fono)sintattico) of the following word.’ (Vecchio 2010: 321)
Accordingly, to explain the distribution of RF witnessed in (18b–c), the traditional deletion approach in terms of PF-realization could be reframed along the lines of a Kaynian-style analysis. In essence, Salentino dialects could be argued to present two irrealis complementizers, namely cu and CU, both merged in C-Fin◦ but which differ in that the latter represents a silent version of the former not associated with any PFfeatures (20a–b). On this view, RF in (18b–c) now finds a uniform trigger in the overt (20a) and silent (20b) variants of the irrealis complementizer. (20) a. vogghiu [CP cu [p:]artu] b. vogghiu [CP CU [p:]artu] I.want cu I.leave ‘I want to leave’ Attractive as this analysis might appear, it cannot however be correct, at least for the dialects of northern Salento. Already it was seen in (16a–b) that clitic climbing in northern Salento is only possible if C-Fin◦ is radically empty, since this position needs to be available to act as an escape-hatch through which any clitics may climb to the matrix clause. Yet, under the current phonolexical analysis, the complementizer position is never radically empty, but is invariably lexicalized, irrespective of its PF realization, by either cu or CU, both of which are predicted to block clitic climbing, contrary to fact. At this point, it is instructive to compare the Salentino facts with an apparently similar set of data from Italian. In the latter, the complementizer introducing both propositional and irrealis complements indiscriminately surfaces as che ‘that’. Thus, although Italian appears to formally lack the dual complementizer system typical of the dialects of southern Italy (cf. Table 9.1), an identical distinction can be maintained for Italian at the syntactic level, albeit formally neutralized by the homophony of both complementizers. This is substantiated above all by the observation that, when introducing a propositional complement containing an embedded verb in the indicative, che cannot undergo C-drop, but must always be pronounced (21a). In irrealis clauses, by contrast, which typically contain an embedded verb in the subjunctive or, less frequently, the future or conditional, irrealis che may, especially in formal registers, be ‘deleted’ (21b; Wanner 1981; Giorgi and Pianesi 1997, 2004; Poletto 2001; Scorretti 1994; Giorgi 2010: 43–7). (21)
servirci (It.) a. so ∗ (che) può I.say that it.can.ind serve.inf=us ‘I know that it can be of use to us’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway b. (∗ ci) penso (che) possa servirci (It.) us= I.think that it.can.sbjv serve.inf=us ‘I think (that) it can be of use to us’
Now, although Italian che, both in its propositional and irrealis uses, resembles Salentino cu in triggering RF on account of an underlying final dental stop (viz. < quid), there is a crucial difference between Salentino and Italian in this respect: whereas in the former RF continues to obtain under C-drop (cf. 20b), in the latter RF is only present when the complementizer is pronounced (22a), but not when it is deleted (22b). (22) a. sembra che
[v:]oglia
piovere (It.)
[v(∗ :)]oglia
piovere (It.) b. sembra Ø it.seems (that) it.wants.sbjv rain.inf ‘it seems to want to rain’ If the Italian facts were to receive a phonological analysis, with examples such as (22b) assumed to contain a phonologically null variant of che, namely CHE, then the impossibility of clitic climbing (cf. 21b) follows naturally, but the observed lack of RF, on the other hand, is totally unexpected. Indeed, Poletto (2001) argues (pace Giorgi and Pianesi 1997, 2004) that Italian che-drop involves a process of syntactic deletion that yields an empty C◦ position licensing a case of residual embedded V2, with the modally marked verb raising to the vacant complementizer position. [TP voglia piovere ]] (It.) (23) sembra [CP voglia it.wants.sbjv rain.inf it.seems Based on these observations, it is possible to conclude that whereas the syntactic deletion analysis is appropriate for Italian, in that it correctly predicts the lack of RF and clitic climbing under che-drop, it proves empirically untenable for the dialects of Salento. Rather, the comparison with Italian demonstrates that if Salentino cu-drop genuinely involved syntactic deletion resulting in a radically empty C◦ position, then RF should not obtain. By the same token, the PF-deletion approach (essentially assuming a silent CHE/CU) has also been shown to be inadequate for both Italian and northern Salentino: in the former it would incorrectly predict the presence of RF and in the latter the impossibility of clitic climbing under C-drop. For these reasons, the deletion analysis for northern Salento, whether syntactic or phonological, is rejected here. In its place, a novel and arguably more convincing account of the relevant facts is proposed in the following section.
. Northern Salentino: proposal The basic facts of northern Salentino cu-drop which have to be accounted for are listed in (24).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
(24) a. RF on embedded verb maintained even under cu-drop; b. Clitic climbing only possible under cu-drop; c. Cu-drop only permitted under subject coreference. These facts find an immediate explanation if irrealis predicates in the dialects of northern Salento are assumed to select either for a full clausal CP complement headed by cu (25a), or for a reduced clausal TP complement headed by a finite verb (25b). (25) a. vogghiu [CP cu [TP llu fazzu]] (Mesagne (BR)) cu it= I.do I.want b. lu vogghiu [TP ffazzu] (Mesagne (BR)) I.do it= I.want ‘I want to do it’ Informally, then, the difference between (25a) and (25b) may be thought of in terms of clause-union, reflected in the traditional distinction between a non-restructured biclausal representation and a monoclausal restructured representation, respectively (cf. also the discussion of ‘light’ biclausality in Tortora 2014: ch. 3, §4.3.1). Indeed, this conclusion is essentially forced upon us by general economy principles which would exclude the merger of a radically empty C-Fin head to accommodate a deletion analysis of Salentino cu-drop. Furthermore, the proposed TP analysis in (25b) finds strong parallels with Giorgi and Pianesi’s (2004) approach to Italian C-drop, which takes omission of the complementizer to be a reflex of the syncretic realization of Agr and Mood (what has been called here C-Fin◦ and T◦ ) in a single functional projection. What then are the consequences of the structural proposals in (25a–b)? First, there is no deletion process proper, but simply a selectional alternation between a CP and a TP complement in accordance with whether the relevant Mood/Fin and T features are scattered across two distinct projections (CP. . . TP) or realized syncretically in a single functional projection (TPMood/Fin ). Indeed, this approach to the data provides an original and highly natural interpretation of the distribution of RF under C-drop (cf. 24a). In particular, the structure in (25b) forces a synchronic reinterpretation of RF since, unlike (25a), the proposed representation lacks an appropriate RF trigger. The observed RF on the verb in (25b) is therefore to be interpreted as a PF reflex of an intrinsic inflectional property of the embedded T head, ultimately spelling out the latter’s syncretic irrealis Mood/Fin feature on the raised verb (or pronominal clitic attached to the same).4 What this amounts to saying is that the residue of an 4 As C. Tortora (p.c.) points out, the distribution of RF in these varieties can be profitably employed to test those theories of object clitic placement (Kayne ; Cardinaletti ) which assume (certain) proclitics to reside in separate heads higher than that occupied by the raised verb. If RF is a property of the T head, then only those proclitics adjoined to the verb raised to T◦ should be able to bear RF, whereas proclitics residing in higher heads should fail to display RF in irrealis complement clauses. We leave this topic for future research.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway
erstwhile RF reflex has been reanalysed synchronically as an overt marker of irrealis subordination, in essence to be understood as a novel marker of subjunctive modality on the verb (cf. footnote 3). Consequently, the absence of consonantal-initial lengthening on a verb form like partu in (26a) immediately categorizes it as occurring in a root clause or indicativestyle subordinate clause embedded under ca, whereas its presence in (26b) immediately entails that the verb can only be interpreted as an embedded irrealis form (equivalent to the subjunctive in other Romance varieties). (26) a. (ticu ca) [p]artu (root/embedded indicative) I.say that I.leave ‘(I say that) I’m leaving’ b. ∗ (vogghiu (cu)) [p:]artu (embedded subjunctive) I.want that I.leave ‘I want to leave’ Further compelling evidence for the reanalysis of RF as an innovative marker of irrealis/subjunctive modality comes from the observation that consonantal lengthening has now even spread to those verbs which etymologically were vowel-initial and hence fell outside of the original domain of application of genuine cases of RF (Calabrese 1993: 80 n. 2; Bertocci and Damonte 2007: §3.1). Emblematic in this respect is the case of the verb ‘be’, which frequently presents, alongside traditional vowelinitial forms such as 3sg essa (analogical on the infinitive essere ‘be’) and eggia (analogical on the subjunctive aggia < avere ‘have’), novel reinforced consonantal-initial forms such as bbessa and bbeggia, ddeggia, and ggeggia (cf. also further representative forms such as apre > bbrap(r)e ‘open.3sg’, esse > bbesse ‘go.out.3sg’). Under this novel analysis of RF, the northern Salentino facts find an interesting parallel with many Celtic C-related elements (Roberts 2004a), which frequently remain unpronounced but still interpreted on account of the continuing reflex of a so-called ‘soft’ mutation (historically a case of obstruent lenition) on the immediately adjacent embedded verb. For example, modern Welsh has the indirect interrogative complementizer a ‘if ’ which causes soft mutation of the following verb, as illustrated in (27a) where the underlying voiceless velar of ceith ‘will.get.3sg’ has been voiced to geith. If, however, the complementizer remains unpronounced, an apparent case of phonological deletion, the following mutation is maintained (27b) and serves as the only explicit marker of subordination on a par with the role of RF in examples such as (25b, 26b). (27) a. dw i ddim yn siwr a geith
hi
ddiwrnod rhydd
b. dw i ddim yn siwr Ø geith hi ddiwrnod rhydd am I not in sure if will.get she day free ‘I’m not sure that she’ll have a day off ’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
Turning now to properties (24b–c), under the current proposal the availability of clitic climbing and the correlation between C-drop and subject coreferentiality now also fall out naturally. In the non-restructured variant (25a), clitic climbing, and indeed any form of head movement, is blocked by the intervening C-phase head cu, whereas in the restructured variant (25b) there is no intervening phase head and clitic climbing and subject coreferentiality, both interpreted here, following Terzi (1996), as reflexes of T-raising, are free to occur.5 It must also be observed that the present analysis comes very close to Cinque’s (2004: 141–2) monoclausal interpretation of cu-less clauses, in the sense that both approaches assume that the relevant clausal constituent under cu-drop does not involve a full CP complement, but, rather, a reduced complement (cf. also Tortora 2014: ch. 3, §4.3.1). Indeed, the analysis outlined here can be straightforwardly recast in terms of Cinque’s (2004) theory of restructuring, if cu is assumed to be able to lexicalize distinct functional heads along the clausal spine in each of the three extended projections of the lexical VP, namely a head within the C-domain when subcategorized by lexical control predicates (28a), a head within the T-domain when subcategorized by epistemic/alethic modal, temporal, and higher aspectual functional predicates (28b), and a head within the v-domain when subcategorized by root modal and lower aspectual functional predicates (28c). On this view, cu-drop in the latter case can be taken to indicate the selection of a reduced VP constituent, hence the absence of cu (28d; following examples from the Brindisino dialect of Mesagne). (28) a. m’ è dittu [CP cu [TP [v-VP spariscu]]] cu I.disappear me= he.has told ‘he told me to disappear’ 5 This analysis finds strong parallels with Tortora’s (: ch. , §..) recent proposals about clitic placement in Romance, according to which clitics may potentially adjoin to a functional head Z projected in one or more of the three clausal domains (namely, C-, I-, and V-domains), as long as Z does not inherit a [finite] feature from the matrix T◦ . On this view, observed Romance variation in the functional domains targeted by clitics is convincingly reduced to ‘a function of what “junctures” in the clause act as barriers to feature-spreading’ of the relevant [finite] feature. For example, in simplex tenses in the Piedmontese dialect of Borgomanero the left periphery of the V-domain acts as a barrier to feature-spreading and consequently the lowest Z head qualifies as an appropriate adjunction site for the object clitic; witness its position to the right of the finite verb in (i.a). In Italian, by contrast, the left periphery of the V-domain is not a barrier and the [finite] feature spreads down as far as the V-domain, rendering the lowest Z head an inappropriate clitic site; witness the position of the clitic to the left of the finite verb under T◦ in (i.b).
[vP vônghi-ti [VP vônghi ti]]] (Borgomanero) (i) a. [TP i scl.1sg I.see=you vedo [v-VP vedo ti ]] (It.) b. [TP ti you= I.see ‘I can see you’ Applying this approach to the northern Salentino data, the phase-head status proposed for cu in (a) now immediately translates as a barrier in Tortora’s system. Consequently, it inhibits the spreading of the [finite] feature from the matrix T◦ , thereby allowing the clitic to adjoin to the Z head in the embedded I-domain. By the same token, (b) lacks such a barrier and the [finite] feature is free to percolate down into the embedded clause, thereby forcing the clitic to surface in the I-domain of the matrix clause.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway b. nu spiccia [TP cu [vP [VP parla mali ti mei]]] not he.ceases cu he.speaks bad of me ‘he won’t stop criticizing me’ ti notti]] c. voli [vP cu [VP vveni cu he.comes of night he.wants ti notti] d. voli [VP vveni he.comes of night he.wants ‘he wants to arrive at night’
This analysis reflects the same structural ambiguity of Romance non-finite complementizers such as Italian di ‘of ’ (cf. Cinque 2004: 165; Tortora 2014: ch.3, §4.3.1), which under the same selectional conditions may also variously lexicalize C-related (29a), T-related (29b), and v-related (29c) heads.6 (29) a. mi ha detto [CP di [TP [vP [VP sparire]]]] me= he.has told di disappear.inf male di me]]] b. non finisce [TP di [vP [VP parlare di speak.inf bad of me not he.ceases di notte]] c. desidera [vP di [VP venire di come.inf of night he.wishes Finally, incontrovertible evidence for the reduced clausal analysis of the complement under cu-drop in (25b) is further supported by the unavailability of the topicfocus system under cu-drop, as the following Mesagnese (BR) examples demonstrate. (30) a. vogghiu [TopP lu libbru [FinP the book I.want ‘I want to buy the book’
∗ (cu)
[TP mi lu kkattu]]] cu me= it=I.buy
6 Revealing in this respect is the case of the Italian volitional desiderare ‘to wish’ in (c) which, like northern Salentino (v)ulire ‘to want’ (cf. d), only licenses clitic climbing if the v-related complementizer di, a potential intervening phase head, remains unpronounced, namely not projected (i.a–b). By contrast, when di lexicalizes a T-related head, as in (b) in conjunction with aspectual finire ‘to finish’, it does not block clitic climbing (i.c) since, as predicted, di does not constitute a phase head under this instantiation.
(i) a. (∗ lo) desidero [vP di [VP vederlo]] him= I.wish see.inf=him di b. lo desidero [VP vederlo] see.inf him= I.wish ‘I wish to see him’ c. lo finirò [TP di [v-VP correggerelo]] it= I.will.finish di correct.inf ‘I’ll finish correcting it’ Once again, these findings can be readily recast in terms of Tortora’s () analysis of clitic placement based on potential barriers to feature spreading (cf. footnote ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
b. vogghiu [TopP lu libbru [FinP ∗ (cu) [TP mi kkattu, nu llu I.want the book cu me= I.buy not the giurnali]]] paper ‘I want to buy the book, not the newspaper’ These facts now find a straightforward explanation since, when cu is ‘deleted’, the whole CP layer is also ‘deleted’ along with it (the relevant mood and finiteness features now realized syncretically on TP), including the relevant left-peripheral topic and focus fields.7
. Central-southern Salento: Proposal Turning finally to the dialects of the centre-south of the region, the relevant facts to be accounted for are: (31)
a. RF on embedded verb maintained even under cu-drop; b. Cu-drop with disjoint and coreferent subjects; c. Clitic climbing impossible under cu-drop.
Potentially, there are two possible solutions to explain cu-drop in these dialects. The first is to assume, as with the dialects of the north of the region, a syntactic analysis, according to which irrealis predicates may select, alongside of a full CP complement (32a), for a reduced TP complement (32b). (32) a. ogghiu [CP cu [TP llu fazzu]] (LE) cu it= I.do I.want b. (∗ lu) ogghiu [TP llu fazzu] (LE) it= I.want it= I.do ‘I want to do it’ Under this analysis, the RF on the embedded T head in (32b)—surfacing in this case on the left-adjoined pronominal clitic—would once again have to be interpreted synchronically as an independent PF-inflectional property of the embedded T◦ , spelling out the relevant irrealis subjunctive feature of the embedded clause. However, the unmistakable differences in cu-drop observed above between the dialects of the north and the centre-south of the region make this unified analysis unlikely. In particular, under (32b) clitic climbing is expected to be possible under cu-drop in central-southern Salentino, contrary to fact. Of course, it could be argued that this 7 The same conclusion equally follows under the alternative Cinque-style analysis outlined earlier. More specifically, if the matrix predicate vogghiu ‘I want’ selects for a full vP complement, then the lower leftperipheral topic and focus positions (Belletti , ) are predicted to be freely available. If, however, a reduced-clausal VP complement is selected, then the lower left periphery containing the relevant topic and focus positions is not projected.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway
variation in the availability of clitic climbing between the north and the centre-south of the region is simply a local case of the more general Romance-internal variation observed in this area (e.g. French vs. Italian; Kayne 1991). However, this represents an ad hoc solution and entirely misses the point, since the dialects of the centre-south of the region do indeed license clitic climbing in infinitival contexts (Calabrese 1993: 30–1). (33)
n’ adha cosa] (Matino (LE)) t’ aggiu [VP direte say.inf=you an other thing you= I.have ‘I’ll tell you something else’
Moreover, the availability of cu-drop (and hence T-raising) under disjoint reference is also completely unexpected given the reduced structural analysis in (32b), as is the observed difference between the dialects of the north and the centre-south in this respect. The preferred solution to the central-southern facts is therefore to assume a phonolexical analysis. More specifically, cu-drop in these varieties is to be understood as a process of lexical variation, with irrealis predicates invariably selecting for a fullfledged, non-restructured CP complement headed either by a phonologically overt complementizer cu (34a) or by a silent variant CU (34b). (34) a. ogghiu [CP cu [TP llu fazzu]] (LE) cu it= I.do I.want b. (∗ lu) ogghiu [CP CU [TP llu fazzu]] (LE) it= I.want CU it= I.do Under this analysis, all the properties in (31a–c) now follow straightforwardly. First, consonantal lengthening under cu-drop qualifies as a simple case of RF, since it finds a genuine trigger in the, albeit silent, complementizer CU. Second, clitic climbing is always excluded as C-Fin◦ , the potential escape-hatch for head movement into the matrix clause, is already lexicalized by CU (and, in Tortora’s terms, functions as a barrier to feature spreading). Finally, the possibility of cu-drop in conjunction with both coreferent and disjoint subjects is unsurprising, since C-drop in these dialects involves the selection of a phonologically covert complementizer, and not T-raising, and hence proves totally irrelevant to the coreferentiality or otherwise of matrix and embedded subjects.
. Conclusion Putting the results of the above discussion together, some aspects of the historical development of C-drop in the dialects of the Salento can now be sketched. Given the differential distribution of cu-deletion in both dialectal areas, we conclude that the central-southern dialects, in which cu-drop has emerged as the output of an optional
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Reconstructing complementizer-drop in the dialects of the Salento
rule of PF-deletion—interpreted here as a case of lexical variation between cu and CU (35a–b)—must represent a more conservative variety and, ultimately, the stage which underlies the subsequent development seen in the dialects to the north of the region. In these latter dialects it can be hypothesized that in light of the frequent phonological non-realization of the complementizer—namely the selection of the CU variant in (35b)—subsequent generations of speakers were led to reanalyse this original lexical PF-deletion rule as a case of syntactic deletion (35c), with concomitant pruning of the CP layer and all this entails, namely, clitic climbing, obligatory subject coreference, and loss of the null complementizer CU. Without doubt, the most important consequence of this reanalysis lies in the novel refunctionalization of the original C-determined RF that surfaces on the embedded T head, which has been exapted (Lass 1990, 1997; Vincent 1995; Norde 2002; Traugott 2004) as an independent reflex of irrealis subjunctive marking on the embedded T head.8 (35)
a. [CP cu [TP VRF ]] ([–clitic climbing], [±coreference]) b. [CP CU [TP VRF ]] ([–clitic climbing], [±coreference]) c. [TP VRF=SBJV . . .] ([+clitic climbing], [+coreference])
However, the reanalysis from (35b) to (35c) proposed here for northern Salentino dialects is not quite complete, in that these same dialects still marginally allow, albeit with increasing infrequency, structures such as (36) taken from the dialect of Mesagne (BR), in which cu remains unpronounced and clitic climbing fails to obtain. Undoubtedly, such obsolescent structures represent the last remnant of the now largely surpassed stage (35b) in the reanalysis which took place in the north of the region. Indeed, this view finds support in Vecchio’s (2009: 322) observation that whereas cu-drop in the modern Brindisino dialect of Francavilla Fontana now requires subject coreference, presumably instantiating structure (35c), eighteenthcentury Francavillese texts show that cu-drop was licensed even in the absence of subject coreference, no doubt reflecting the preceding stage (35b). arretu!]] (36) vuè [CP CU [TP mmi mandi me= you.send back you.want ‘you want to send me back’ Finally, the developments outlined here exemplify interesting patterns of diachronic variation in the division of labour in the licensing of irrealis modality and its spell-out, which can be framed in terms of Chomsky’s (2007, 2008) proposals about possible feature inheritance and transfer between the phase head C◦ and its complement T◦ 8 As suggested by C. Poletto (p.c.), one could go further here and view the erstwhile RF in these cases not only as a marker of irrealis subjunctive modality, but also as an overt reflex of EPP-checking of the D-feature on the embedded T head. This might provide an account for the otherwise exceptional absence of a canonical preverbal subject position in such irrealis complement clauses.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Adam Ledgeway
(see also Ouali 2008). In particular, we can construct a descriptive diachronic typology along the lines of (37) with associated examples in (38). (37) a. CMood . . . TSBJV b. CMood . . . TIND c. TRF(=SBJV)
(early Salentino) (northern, central-southern Salentino) (northern Salentino)
(38) a. illu volce [CP cu [TP nuy sappessemu [. . . ]]] he wanted cu we knew.sbjv ‘he wanted us to know [. . . ]’ b. voli/ole [CP cu [TP pparla]] cu he.speaks.ind he.wants c. ole [CP CU [TP pparla]] CU he.speaks.ind he.wants d. voli [TP [p:]arla] he.speaks.sbjv he.wants ‘he wants to speak’ (37a) represents the original situation in early Salentino (Ledgeway 2004: §4.3.1, 2005: §4.2.1), where the irrealis modal feature on the phase head C◦ is inherited by T◦ and irrealis modality is thus marked both in the lexical choice of the complementizer (viz. cu) and in the morphological (viz. subjunctive) form of the verb, as witnessed in (38a) taken from the fifteenth-century prose text Il libro di Sidrac salentino. By contrast, (37b) instantiates the option found in the modern Salentino dialects, where the relevant irrealis modal feature on C◦ is not transferred down, but surfaces purely in the lexical choice of complementizer, be it overt (38b)—an option found in both areas— or covert (38c)—an option now restricted to central-southern Salento—, with the embedded verb under T◦ occurring in the morphologically and modally unmarked indicative form. Finally, (37c) exemplifies the situation in the modern dialects of northern Salento, where we may talk of complete transfer of the relevant irrealis modal feature solely to T◦ (surfacing in the erstwhile RF, a synchronic form of subjunctive marking), with, indeed, realization in this case of all C-related features in a single syncretic TP projection (38d).9 From a diachronic perspective, it is thus possible to recognize a shift from an original equal sharing of features across both functional heads through the mechanism of inheritance and transfer to a unilateral realization of the same, first on the phase head and then on T◦ (cf. Ouali 2008). 9 Unlike in (root) realis clauses where the lexical verb raises very low, Ledgeway and Lombardi () demonstrate that in these dialects (and those of the South more generally) the lexical verb exceptionally undergoes raising to T in irrealis clauses. The consonantal lengthening (viz. RF) exceptionally borne by the verb in irrealis clauses is therefore to be interpreted as a concomitant of V-to-T raising, both ultimately to be understood as morphophonological and syntactic reflexes, respectively, of an irrealis modal featurechecking operation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi MARIT JULIEN
. Introduction In this chapter I will present a syntactic change that took place in some varieties belonging to the Finnic and Sámi branches of Uralic and involved negation, tense, and the past participle. This change is interesting because it can be seen as an illustration of three more general points: • How the order of functional heads can change over time. • How a morphological form that takes on a new function, and thereby seemingly also a new meaning, can be shown to have the same meaning in all its uses, on the ‘right’ analysis. • How change can proceed in a stepwise fashion and hence in a sense be gradual, even though each step involves discrete properties. The first point is relevant in relation to the so-called cartographic approach to syntactic structure, which was initiated by Cinque (1999) and followed up in much subsequent work. Although this approach has focused on establishing invariant ordering patterns involving functional categories, it is interesting that already Cinque (1999: 125) noted that some categories do not have a universally fixed position. For example, the canonical position of the clausal negation varies between languages. The fact that there is variation means that change should also be possible. Changes in the position of the negation have been extensively studied in certain languages where the clausal negation is a particle or adverb. Here we will see how the position of the negation can change in languages where the negation is a verbal element—a negative auxiliary. The second point is not meant to be of relevance for all cases where a morphological form takes on a new function—I am not claiming that there is never any change in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
meaning. However, I think the change that we will look at in some detail can serve as a reminder that when a form appears to take on a new meaning, it is possible that the new meaning was already there—it only becomes more apparent when the form shows up in a new context. The third point is particularly interesting if one assumes, with Lightfoot (1991), that each language learner constructs one and only one grammar on the basis of the positive evidence that is robustly present in the linguistic environment. As we will see, the conclusions that the learner then draws about the linguistic system need not be global, having relevance for the whole system. On the contrary, they can be relatively local, in the sense that they can be restricted to certain contexts. I start the discussion by showing how the negation and the past participle behave today in those Finnic and Sámi languages that have undergone the change. Then I present a sketch of the old pattern, i.e. of the grammar of Finnic and Sámi participles and perfects before the change. After that I give some more support for my claim that the past participles represent tense, before returning to Finnic and Sámi to take a closer look at the change. Finally, I deal with the situation in Lule Sámi, a variety where the change has not yet been completed and where the gradualness of the change is apparent in the present-day language.
. The new pattern The interplay of past tense, negation, and the past participle in modern Northern Sámi is illustrated in the examples in (1). In the affirmative past, shown in (1a), markers of tense and subject agreement are suffixed to the verb. In the negated past, shown in (1b), the negation agrees with the subject and expresses the finiteness of the clause, whereas the lexical verb appears in a non-finite form. If we compare this to the negated present in (1c), we see that the tense distinction is marked on the main verb. Finally, a comparison between the negated past in (1b) and the affirmative present perfect in (1d) shows that the same non-finite form is found in these two constructions. (1)
Northern Sámi a. Moai jeara-i-me Bireh-is. 1du.nom ask-past-1du Biret-loc ‘We (two) asked Biret.’ b. Moai ean jearra-n Bireh-is. 1du.nom neg.1du ask-ptcp Biret-loc ‘We (two) did not ask Biret.’ c. Moai ean jeara Bireh-is. 1du.nom neg.1du ask Biret-loc ‘We (two) do not ask Biret.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi
d. Moai letne jearra-n Bireh-is. 1du.nom be.pres.1du ask-ptcp Biret-loc ‘We (two) have asked Biret.’ It should be noted that in the negated past the negative auxiliary combines with the past participle without giving rise to a perfect reading. The verb form that appears in the affirmative past, on the other hand, bears no likeness to the participle. Although the past participle has a long history in the language, its use in the negated past is an innovation, not only found in today’s Northern Sámi but also in Inari Sámi, Skolt Sámi, Kildin Sámi, Finnish, Karelian, Northern Estonian, Votian, and Ingrian. The development of the new negated past is the focus of this chapter.
. The old pattern According to Korhonen (1973, 1981) the construction that is now the negated past in Finnish, Northern Sámi, and several related languages, was originally the negative of the present perfect construction. We can see this in (2), where I list a selection of verb forms from the reconstructed Proto-Finnic (1500–1000 bc). Since what we are interested in here is the pattern and not the actual form of the markers, I follow Korhonen (1973) and use modern Finnish word forms. (2) Proto-Finnic verb forms affirmative a. Present kuule-n hear-1sg ‘I hear’
(Korhonen 1973: 181) negative e-n kuule neg-1sg hear ‘I don’t hear’
b. Past
kuul-i-n hear-past-1sg ‘I heard’
e-si-n kuule neg-past-1sg hear ‘I didn’t hear’
c. Present perfect
kuul-lut hear-ptcp ‘I have heard’
e-n kuul-lut neg-1sg hear-ptcp ‘I haven’t heard’
We see that in Proto-Finnic the main verb was inflected for tense in affirmative clauses, but in negated clauses the negation carried tense while the main verb appeared in an invariant so-called connegative form. And in the present perfect, there was no auxiliary at all in the affirmative while in the negative a negation with subject agreement (and with null present tense) combined with the participle. The tense-marked negation is still preserved in some varieties at the periphery of the Sámi/Finnic area—in Southern Sámi, Southern Estonian, Livonian, and in the Mordvin, Mari, Permic, and Samoyed languages (Korhonen 1973; Trosterud 1994). The Southern Sámi examples in (3) serve as an illustration.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(3)
Marit Julien Southern Sámi a. Manne guarka-m. I understand-1sg ‘I understand.’ b. I-m guarkah. neg-1sg understand ‘I don’t understand.’ c. Manne guarka-ji-m. I understand-past-1sg ‘I understood.’ d. I-dtji-m guarkah. neg-past-1sg understand ‘I didn’t understand.’
Another interesting fact of Southern Sámi, shown in (4), is that the copula is optional in the perfect, or more precisely, it is not spelled out unless the tense or polarity is emphasized (see Bergsland 1994). In the negative, the overt copula has the invariant form leah in the present tense, but in the past tense it shows the same subject agreement as the negation. (4) Southern Sámi a. Manne (lea-m/li-m) guarke-me. I be.pres-1sg/be.past-1sg understand-ptcp ‘I have/had understood.’ (leah/lim) guarke-me. b. Manne i-m I neg-1sg be/be.past.1sg understand-ptcp ‘I have/had not understood.’ Strikingly, there is a strong positive correlation in Sámi between the presence of a past tense form of the negation and the absence of an obligatory copula. Korhonen (1973) suggests that there is a causal relation between these two properties, since, historically, the introduction of the copula led to the construction that used to be the negated present perfect acquiring the meaning of the negated past. Here I will try to give an explanation of this change. But first we will consider the construction in (3d). If we assume 1) that the order of morphemes in the surface form reflects the underlying order of syntactic elements, and 2) that head movement or an operation equivalent to head movement is possible (see e.g. Matushansky 2006), then the negated clause in (3d) indicates that the clausal functional sequence in Southern Sámi involves the heads shown in (5a). I take the agreement markers to be realizations of a Finite head, which sits above the Tense head, which in its turn is higher than the Polarity head. Heads in the C domain are left out here.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi (5)
a.
FinP Fin
TP T
PolP Pol
b.
VP FinP
Fin T Pol
TP Fin
T
T
PolP Pol
VP
After head movement of the Pol head to Tense and of Pol+Tense to Fin, the result is as in (5b), with polarity, tense, and finiteness/agreement in a complex syntactic head that is separate from the verb root, which stays in VP. Hence, (5b) shows the syntactic structure of (3d). In the affirmative clauses in (3a) and (3c), on the other hand, markers of tense and agreement are suffixed to the verb root. I take this as an indication that in the affirmative, there is successive cyclic head movement of the verb to Pol and then to Tense and Fin. In other words, an affirmative polarity head attracts the verb, whereas a negative polarity head does not. This means that affixhood can be connected to individual features and not to categories. In Finnic and Sámi, only positive polarity heads attract the heads of their complements, which leads to affixation of the polarity element, if it has a phonological realization. Negative polarity heads, by contrast, do not attract the heads of their complements.
. The tense of past participles Turning now to the perfect constructions, I will first point to some data which indicate that the past participle has temporal content. In each of the Northern Sámi constructions given in (6), there is an embedded nonfinite clause. In (6a) the verb of the non-finite clause has the past participial form, whereas in (6b) it appears in the so-called ‘aktio essive’ form, which also could be termed progressive, and in (6c) the embedded verb shows up as an infinitive. (6) Northern Sámi a. Elle lohká/logai Máhte boaht-án. Elle says/said Máhtte.acc come-ptcp ‘Elle says/said that Máhtte has/had come.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien b. Elle lohká/logai Máhte boahti-me. Elle says/said Máhtte.acc come-prog ‘Elle says/said that Máhtte is/was coming.’ c. Elle lohká/logai Máhte boahti-t. Elle says/said Máhtte.acc come-inf ‘Elle says/said that Máhtte will/would come.’
We see that the meanings of these non-finite verb forms can be rendered by finite verbs.1 More precisely, the past participle in (6a) corresponds to a relative past tense, since it places the event of Máhtte coming in the past with respect to the event denoted by the matrix verb, which coincides with the speech time if the matrix verb has present tense but precedes the speech time if the matrix verb is in the past tense. The progressive in (6b) indicates that the coming event is simultaneous with the matrix event, regardless of whether the latter coincides with the speech time or precedes it. Finally, the embedded infinitive in (6c) places the coming event in the future with respect to the matrix event. Although both the past participle and the progressive in Sámi are traditionally taken to represent aspectual categories, the readings that we get in (6) are no different from what we would get from a relative past tense, a relative present tense, and a relative future tense, respectively. This is one reason why I take the past participle to represent a non-finite past tense and the progressive to represent a non-finite present tense, in Sámi as well as in other languages. As for the infinitive, it clearly gives a future reading in (6c). In other cases, it seems to have no temporal semantics at all. One example is given in (7). (7) Northern Sámi Elle liik-o dáns-ut. Elle like-pres.3sg dance-inf ‘Elle likes to dance.’ Hence, I will follow Wurmbrand (2001) and assume that infinitives are sometimes tensed and sometimes not. The idea that participles express tense has consequences for the analysis of perfect constructions. Although the literature on past participles and on perfect constructions has grown enormously in recent years, one can say that in general there are two competing views on the syntax of the perfect. According to the more traditional view, perfect constructions, as in (8a–b), are monoclausal, with the past participle representing an aspectual head situated somewhere below the tense head (see e.g. Ouhalla 1991 and Cinque 1999). 1 The embedded non-finite clauses in (a)–(c) could also be paraphrased by embedded finite clauses in Northern Sámi.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi
(8) a. I have read that book. b. I had read that book. On the competing view, which can be traced back to Ross (1969) and Emonds (1976), perfect constructions are biclausal. This means that the sentences in (8) are made up of a finite matrix clause, with a VP headed by the auxiliary and with a Tense head spelled out by the tense suffix that appears on the auxiliary, and of a non-finite embedded clause, with a VP headed by the main verb and with a Tense head spelled out by the participial suffix (but possibly considerably reduced compared to the finite matrix clause). As for the semantics of the tense elements, I assume that each tense element specifies a relation between two temporal primitives. These temporal primitives could be identified as times (Stowell 1995, 1996; Zagona 1995), as time intervals (Demirdache and Uribe-Etxebarria 2000, 2007), or as events (Giorgi and Pianesi 1997). An advantage of postulating events as the arguments of tense elements is that one then avoids imposing any particular extension in time on these arguments: the events may be punctual or non-punctual, open or closed (see Giorgi and Pianesi 1997). I will therefore follow Giorgi and Pianesi (1997) and assume that a tense element has two events as its arguments, and it specifies a temporal relation between these events. As for the question of how the events are represented in the syntax, I will not go into details here (but some proposals are found in Zagona 1995, Stowell 1995, 1996, 2007, and Julien 2001); I will merely note that VPs headed by full lexical verbs represent predicate events, and that the speech event must be higher than any other event that is represented in a given construction. For each tense element, an event that is higher in the structure will be its first argument while an event that is lower in the structure will be its second argument. If a given tense element is the highest tense element of the construction, it will normally have the speech event S as one of its arguments, since S is higher than any other relevant event. The lowest tense element of a construction must have the predicate event E as one of its arguments. It follows that if a construction contains only one tense element, it must have both S and E as arguments, and, accordingly, it relates S directly to E. In constructions with two tense elements, by contrast, the higher head relates S to a reference time R while the lower head relates R to E, so that the relation between S and E is an indirect one, mediated by R.2 Now the meaning of the past perfect can be shown schematically as in (9). 2 The use of narrative present is an exception to the generalization that the highest tense element has S as one of its arguments. One could perhaps argue, though, that the narrative present relates to a fictitious speech time. Also note that an embedded finite tense does not necessarily have the speech event as one of its arguments. This is shown in the following Northern Sámi example, from Nickel (: ):
(i) De gula-i-met, go unna billo-š gilkkah-a. then hear-past-1pl as little bell-dim sound-pres.3sg ‘Then we heard a/the little bell sounding.’
(cont.)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
(9) I had(PAST1 ) read(PAST2 ) that book.
PAST1: S after R PAST2: R after E E |
R |
S | >
That is, the speech time is after a reference time which in its turn is after the event time.3 The important point here is that on the view I have just sketched out, the difference between the ordinary past tense and the past participle is that the former typically relates R to S, whereas the latter establishes a relation between R and E. In other words, the ordinary past tense has an absolute function while the past participle has a relative function (cf. Comrie 1976; Klein 1995). However, the absolute function of the non-participial past and the relative function of the participial past are not necessarily intrinsic properties of the respective markers. Rather, their different functions follow from their different syntactic positions. A nonparticipial, finite past marker can be the highest tense element of a construction, and accordingly it can have the speech event as its first argument. Hence, finiteness tends to be associated with absolute tense. A participial past marker, on the other hand, will normally be embedded under some finite verbal element. In most cases, it is also embedded under a finite tense element, and it is then the finite tense element which has the speech event as its first argument, while the participial past has some R as its first argument. Because of this, the participial past is associated with relative tense. In other words, the morphological distinction between the ordinary past and the past participle does not necessarily reflect a contrast between tense and aspect. Rather, it seems that the realization of a tense head with the feature [+Past] in many cases depends on whether that tense head appears in a finite or in a non-finite context. Also note that there are quite a few languages where absolute, finite tenses have the same realizations as relative, non-finite tenses. I will present one example here. In (10) we have Northern Sotho, where one and the same marker expresses an absolute past if it is the highest tense marker of the construction, as in (10a), but a relative past elsewhere, as in (10b) and (10c). Here the past tense of the matrix clause relates the speech event to the event denoted by the matrix predicate, i.e. to the event of hearing. The (finite) present tense of the embedded clause relates the event denoted by the embedded predicate, i.e. the event of sounding, not to the speech event but to the matrix predicate event, which serves as an R for the embedded tense. 3 The various readings of the perfect, first discussed in Comrie (), arise from variations in the exact relation between E and R.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi (10) Northern Sotho a. Mosadi o-rêk-ilê nama. woman sagr-buy-past meat ‘The woman bought meat.’
(Louwrens, Kosch, & Kotzé 1995: 51 ff.)
b. Mosadi o-bê a-rêk-ilê nama. woman sagr-aux.past sagr-buy-past meat ‘The woman had bought meat.’ c. Mosadi o-tlô-ba a-rêk-ilê nama. woman sagr-fut-aux sagr-buy-past meat ‘The woman will have bought meat.’ The only morphological difference between the single verb in (10a) and the embedded verb in (10b) and (10c) is the shape of the subject agreement prefix, which is oin the former but a- in the latter.4 Crucially, the past tense marker has the same shape in both environments, and, arguably, the only semantic difference between the past tense in (10a) on the one hand and the past tense in (10b–c) on the other is that the former is directly related to the speech event whereas the latter is not.5 In the model I am proposing here, we would take all these occurrences of past tense to be realizations of a T head with the feature [+Past], and the absolute or relative reading to follow from the position of the head in the construction as a whole. I will conclude that, generally, the past participle represents a past tense which is like any other past tense except that it is non-finite.
. The change Let us now consider Proto-Finnic again. If the present perfect is always the result of combining a matrix present tense with an embedded past tense, then the Proto-Finnic type of present perfect, which is expressed by just the participle (as we saw in (2c)), must include a phonologically null auxiliary. But since the only overt verbal inflection in the present perfect is that of the participle, it is only the participle that shows that we have a perfect and not a simple past. Hence, the past participle suffix in Proto-Finnic was always the realization of a non-finite relative past tense, and, moreover, only the presence of the participle indicated that the construction also contained an auxiliary projection. 4 According to Louwrens, Kosch, and Kotzé () the o- agreement marker is used in the indicative positive, for this class of subjects, and the a- marker elsewhere. They do not comment on the appearance of the a- marker on the embedded verb in (b–c), which also seems to be a positive indicative form. One might guess, though, that the o- only appears on finite verbs, so that a non-finite verb must carry the amarker instead. 5 Note that even in English and other Germanic languages the non-participial past is often similar to the participial past.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
At a later stage, the copula was introduced in many Finnic and Sámi varieties (at different times in different varieties). The reasons for the introduction of the copula into the languages are outside the scope of the present discussion. What will concern us here is how the copula could be the trigger for the changes that eventually led to the constructions that we saw in (1). After the introduction of the copula, the old present perfect constructions were replaced by constructions containing a copula. But as Korhonen (1973) points out, there was a stage where the forms shown in (11) all coexisted. (11)
Verb forms at the intermediate stage affirmative a. Present kuule-n hear-1sg ‘I hear’
(Korhonen 1973: 182) negative e-n kuule neg-1sg hear ‘I don’t hear’
b. Past
kuul-i-n hear-past-1sg ‘I heard’
e-si-n kuule neg-past-1sg hear ‘I didn’t hear’
c. Old present perfect
[kuul-lut] hear-ptcp
e-n neg-1sg
d. New present perfect
ole-n kuul-lut be-1sg hear-ptcp ‘I have heard’
e-n ole kuul-lut neg-1sg be hear-ptcp ‘I haven’t heard’
kuul-lut hear-ptcp
But as indicated, the old affirmative present perfect, consisting of just the past participle, soon fell out of use. The old negated present perfect, by contrast, with negation and participle only, survived alongside the new perfect that contains a copula both in the affirmative and in the negative. But as we will see, the combination of negation and participle became prone to reanalysis. Korhonen (1973) suggests that the reason why the old negative present perfect was retained better than the old affirmative present perfect had to do with finiteness: after the introduction of the copula, every verbal construction includes a finite verbal element. Hence, visible finiteness becomes obligatory. In the new affirmative present perfect, the copula is the finite element and cannot be left out. In negated clauses, on the other hand, the finite element is always the negation, and consequently the copula is not needed in the negative to meet the visible finiteness requirement. Then the question is of course how the combination of negation and participle would be interpreted. Korhonen suggests that since minä en ole kuullut is an indisputable perfect form, minä en kuullut was left competing with minä esin kuule as the realization of the negated past. We can understand this if we see it from the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi
perspective of the language learner: whereas a child exposed to constructions of the form minä kullut (subject + past participle) would have to assume that the past participle suffix represents a non-finite relative past tense, as I have already pointed out, a child hearing the new constructions of the form minä olen kullut (subject + copula + past participle) will conclude differently. In the latter case, the past participle need no longer be specified as the realization of a relative tense, since its relative function is clearly indicated by the fact that it is preceded by a tensed auxiliary. Hence, the past participle can be seen as just a non-finite past tense form that can get a relative tense interpretation from the context. In Northern Sámi, there is however one notable exception to the generalization that the past participle expresses a non-finite past tense. The verb leat ‘be’ has a past participle leamaš, which appears in perfect constructions, as in (12a), and another form lean, which is the past tense connegative form, as in (12b). (12)
Northern Sámi a. Mis lea leamaš gelddolaš jahki. to.us is been exciting year ‘To us it has been an exciting year.’ b. Dat ii lean mu sivva. it neg.3sg be.past my fault ‘It was not my fault.’
Hence, for this verb we have to say that leamaš is the non-finite relative past whereas lean is the non-finite absolute past. But for all other verbs it holds that in the new grammar it suffices to characterize the past participle suffix as the realization of a nonfinite past tense; more precisely, as the realization of a [+Past] tense head that has not undergone head movement to a Fin head. With everything else unchanged, seeing the participial suffix as a marker of nonfinite past tense does not allow it to show up in the negated past, however. The inflected negation in the old negated past (see (11b)) indicates that the Tense head is higher than the polarity head (as in (5a)), and in this structure, the verb cannot move to (or be associated with) the Tense head across the negation. However, Korhonen (1973) also suggested that the use of the participle in the negated past was a consequence of the copula itself having become obligatory. This can now be explained in the following way. For the new generation, the copula is obligatorily spelled out whenever it occurs. It follows that the negated present perfect must be en ole kuullut, with negation, copula, and participle (see (11d)). The older generation, however, still (at least sometimes) produce forms like en kuullut, without a copula. For the younger speaker, there is only one possible analysis for such forms, namely, one where en kuullut is in opposition to en kuule, the negated simple present,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
with the participial past tense marking in en kuullut contrasting minimally with the present tense in en kuule—see (13). In other words, constructions like en kuullut are now analysed as representing the negated simple past.6 (13)
Negated present: Negated past:
en neg.1sg en neg.1sg
kuule V(+pres) kuullut V+past
Importantly, the appearance of the past participle in the negated past does not only mean that the combination of negation and past participle has been reanalysed; it means that a new underlying syntactic structure must be postulated, one where Tense is below the polarity head. I show this in (14a). (14)
a.
FinP Fin
PolP Pol
TP T
b.
VP
FinP Fin Pol
PolP Fin
Pol
TP T
V
VP T
V
In negated clauses, the situation will now be as shown in (14b). The verb moves to Tense, but since a negative polarity head does not attract its complement, as we have seen, the complex head consisting of V and Tense does not move on further. The polarity head itself moves to Fin, as before, and as a consequence the negation ends up with subject agreement marking—but it no longer carries tense marking.
6 It should be noted that the past participle in Sámi is never an exponent of passive. The Sámi passive is a suffix that appears inside tense and agreement marking, as shown in (i).
(i) Johtim-is geavah-uvvu-i veahkki-n helikopter. move-loc use-pass-past.3sg aid-ess helicopter ‘When moving camp a helicopter was used as an aid.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi
In the resulting structure, the Tense head is not included in a complex head with the Fin head, and accordingly, if the Tense head has the feature [+Past], it will be spelled out by the past participial suffix. Recall that in the grammar with an obligatory copula, the past participial suffix is specified as the representation of a non-finite past tense. Hence, it may appear in perfect constructions, where it spells out the non-finite past tense of a clausal projection that is embedded under a finite auxiliary, and it may also appear in the negated past. But note that the feature specification of the past participle has not changed—it is a past tense marker as before. It is the specification of the contexts where it is allowed to appear that is different in the new grammar. What made the reanalysis of the past participle and of the clause structure possible must have been the fact that the present tense marker was (and still is) phonologically null. Hence, a negated present tense construction like en kuule was compatible with the structure in (14) as well as with the one in (5). As I show schematically in (15), the negated present tense could be analysed as having the present tense marker suffixed to the negation, like the old negated past, or suffixed to the main verb, like the new negated past. (15)
Old negated past: Negated present: New negated past:
e-si-n neg-past-1sg e-n neg-(pres?)-1sg e-n neg-1sg
kuule hear kuule hear-(pres?) kuull-ut hear-past
In the new grammar, the latter option is generalized, so that in all negated clauses an overt or covert tense marker is attached to the verb that follows the negation and never to the negation itself. Summing up, we see that analysing the past participle as a realization of past tense paves the way for a rather simple analysis of the appearance of the past participle in the negated simple past in many Finnic and Sámi languages: in these languages, the past participle is always the realization of a tense head that, firstly, carries a [+Past] feature, and, secondly, has not moved to Fin. In the perfect, the participle is combined with a finite and tensed auxiliary. When the past participle is used in the negated past, on the other hand, the participle is the only tense-marked element of the construction, whereas the negation carries the finiteness.
. The transitional varieties At present, there is a transitional zone between the conservative Southern Sámi, shown in (3), and the innovative Northern Sámi, shown in (1). In the varieties spoken in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
between, the copula has been introduced into the perfect construction, as we see in the Lule Sámi examples in (16). (16) Lule Sámi a. Mån la-v tjáll-ám. I be-1sg write-ptcp ‘I have written.’ b. Mån lidji-v tjáll-ám. I be.past-1sg write-ptcp ‘I had written.’ However, the tense marking of the negation in the negated past is still retained, as illustrated in (17). (17)
Lule Sámi a. Mån tjálá-v. I write-1sg ‘I write.’
b. I-v tjále. neg-1sg write ‘I don’t write.’
c. Mån tjáll-i-v. I write-past-1sg ‘I wrote.’
d. I-ttji-v tjále. neg-past-1sg write ‘I didn’t write.’
Interestingly, it is reported that the innovative negated past, with tense marking on the main verb instead of on the negation, as in (18), is often heard in Lule Sámi, although it has not yet replaced the traditional construction completely. Note that the participial ending represents the past tense, as indicated. (18)
Lule Sámi I-v tjáll-ám. neg-1sg write-past ‘I didn’t write.’
This means that the change that has already been completed in Northern Sámi, Lule Sámi’s closest neighbour to the north, is under way in Lule Sámi, which is the northernmost of the transitional languages. The gradualness of the change is seen even more clearly if the paradigms of Lule Sámi verbs are scrutinized. Thus, in (19) I show the negated forms of liehket ‘be’ together with the traditional negated forms of the full verb tjierrot ‘cry’. (19)
Lule Sámi: Negated forms of liehket ‘be’ and tjierrot ‘cry’ Present Past Present Past 1sg iv la iv lim iv tjiero ittjiv tjiero 2sg i la i lim i tjiero ittji tjiero 3sg ij la ij lim ij tjiero ittjij tjiero
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On negation, tense, and participles in Finnic and Sámi 1du 2du 3du 1pl 2pl 3pl
en la ähppe la äbá la ep la ehpit la æl’la
ejma lim eida lim ejga lim ejma lim ejda lim æl’lim
en tjiero ähppe tjiero äbá tjiero ep tjiero ehpit tjiero e tjiero
ejma tjiero ejda tjiero ejga tjiero ejma tjiero ejda tjiero ettjin tjiero
We see that a full verb, like tjierrot, may appear in an invariant connegative form throughout the negative paradigm, while the negation is marked for tense and agreement (with the present tense marker being null). However, in the negative paradigm of liehket tense is obligatorily marked on the verb in the singular and in the third person plural, such that the present tense form la contrasts with the past tense form lim, while the negation is only marked for subject agreement. In other words, the change to the new pattern has been completed in these cells of the paradigm. Then in the negated dual and in the first and second person plural of liehket, tense is marked on the negation as well as on the verb. This is apparently in conflict with my earlier proposals, which are based on the assumption that tense markers correspond to Tense heads in the syntax. In light of this, the double tense marking seen in some of the negated forms of liehket seems to suggest that two heads with a [+Past] feature are present in the construction. Alternatively, the postulated correspondence between tense markers and Tense heads is simply mistaken. However, I think that a third explanation is also available. Note that the past connegative lim is not identical to the past participle of liehket, which is læhkám. (20) Mån lav læhkám moaren. I be.pres.1sg be.ptcp angry ‘I have been angry.’ Hence, we could say that in the negated dual and in the negated first and second person plural of liehket, lim is just a contextually conditioned allomorph of the connegative of liehket, appearing in past tense contexts. But if the tense marking of the negation disappears at a later stage, lim will end up as the realization of non-finite ‘be’ plus an absolute past tense, while the past participle læhkám will be the realization of nonfinite ‘be’ plus a relative past tense. As we have already seen, in example (12), this is the present situation in Northern Sámi. Taken together, the variation that can be found with respect to the interaction between negation, past tense, and finiteness in various Sámi languages shows how gradual the change is that replaces the old pattern with the new pattern. It does not affect the whole language at once; rather, some verbs are affected before others and, moreover, in the paradigm of one single verb some forms can adhere to the new pattern before other forms follow.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Marit Julien
. Conclusions In this chapter, I have given an account of the historical change that has taken place in several Sámi and Finnic languages, whereby the traditional Uralic negated past construction, with tense marked on the negation, has been replaced by the construction that was formerly the negated perfect. I have shown that this replacement, which was triggered by the introduction of the obligatory copula into the perfect constructions, can be explained as a revision of the context for the past participle. In the new grammar, the past participle invariably represents a non-finite past tense, not a relative non-finite past tense. At the same time, the use of the participle in the negated past forces a change in the order of heads in the clausal functional domain: Tense and polarity will have to switch positions. Finally, we looked at Lule Sámi, a variety which is now undergoing the change, and we saw here that the change is gradual in the sense that it does not affect all verbs and all negated forms at once. Instead, it eats its way through the grammar in small steps.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic K R Z YS Z T O F M I G DA L S K I
. Introduction This chapter analyses diachronic changes in the position of pronominal clitics in Slavic.1 The change consists in the switch from verb-adjacent to second position (Wackernagel) cliticization. The change was contemporaneous with the impoverishment of morphological tense distinctions, which I interpret in syntactic terms as the loss of TP. The chapter has the following organization. Section 11.2 presents two types of clitic placement in contemporary Slavic languages, which are exemplified by Serbo-Croatian for languages with Wackernagel clitics and by Bulgarian for languages with verb-adjacent clitics. Section 11.3 shows the way the two patterns evolved by describing diachronic modifications of cliticization in Old Church Slavonic and Old Serbian. Section 11.4 outlines a history of tense marking in Slavic, showing that clitics moved to second position in those languages in which temporal relations are exclusively expressed by aspect. Section 11.5 provides an analysis which postulates nonuniversality of the TP layer across languages. The chapter concludes with a summary in Section 11.6.
. Properties of second position and verb-adjacent clitics The present section discusses second position and verb-adjacent cliticization in contemporary Slavic languages. Cliticization in Slavic has been a matter of extensive 1 I would like to thank Edith Aldridge, Željko Bošković, Paul Kiparsky, Marios Mavrogiorgos, Roumyana Pancheva, Olga Tomić, Julio Villa-García, the audience present at the DiGS- conference, and two anonymous reviewers for their helpful feedback. The initial part of this research was funded by the Netherlands Organization for Scientific Research (NWO) under the Rubicon grant -- completed at the University of Connecticut, and currently it is supported by the Focus grant from the Foundation for Polish Science.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
debates during the last 20 years (see, for example, Tomić 1996, Franks 1998, Franks and King 2000, and Bošković 2001 for a discussion). Therefore, this section is not an overview of various accounts; rather, it serves to juxtapose some relevant properties of the two cliticization types in order to demonstrate that the change described in the chapter did not consist in just a shift in their syntactic position, but rather involved a major modification of the movement operation. Namely, I will demonstrate that the switch to the second position pattern means that pronominal clitics no longer headadjoin to a single syntactic head, but instead each pronominal clitic raises to a different specifier of various projections in the VP domain. Regardless of the cliticization type, clitics in Slavic follow the template exemplified in (1). It shows that operator clitics (which in most contemporary Slavic languages are represented by the interrogative/focus particle li) precede all the other clitics. The auxiliary clitics, which comprise weak forms of the verb BE, have a peculiar distribution, with the 3rd person auxiliary clitic following all the argumental pronominal clitics. There are a few minor differences with respect to the clitic arrangement in some Slavic languages, but they are immaterial for the claims made in this chapter. (1)
operator clitics > AUX (except 3rd SG) > REFL > DAT > ACC > 3rd SG AUX
.. Second position cliticization Second position clitics do not need to be adjacent to elements of any specific grammatical category. What matters is that they follow the first initial element in both main and embedded clauses, as shown in (2) for Serbo-Croatian.2 (2) a. Koliko im ko daje? How-much them-dat who gives ‘Who gives how much to them?’
(Rivero 1991: 335)
b. Stefan kaže da mu je knjigu Ana dala Stefan says that him-dat is-aux book-acc Ana give-ptcp.f.sg ‘Stefan says that Ana gave him a book.’ (cf. Franks 1998) The question of what counts as a valid initial constituent has been a matter of debate. A traditional idea was that clitics are phonologically weak, so they need to be supported by a stressed element in front of them. It seems clear by now that being a stressed element is not enough to be an eligible host and that clitics may occur only after syntactic constituents, that is only after those elements that are able to undergo syntactic movement independently. Consequently, they do not appear after elements that are not syntactically mobile even if they are stressed, such as the first conjunct in (3a’) or the preposition prema in (3b’).
2 Apart from Serbo-Croatian, clitics target second position also in Slovene, Czech, and Slovak.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic (3)
a. Sestra i njen muž će mi ga pokloniti sister and her husband will me it-acc give ‘My sister and her husband will give it to me.’ a’. ∗ Sestra će mi ga i njen muž pokloniti b. Prema Miodragu ga je Ana bacila, a ne toward Miodrag-dat it-acc is-aux Ana throw-ptcp.f.sg and not od njega from him ‘A. threw it towards M., and not away from him.’ b’. ∗ Prema ga je Miodragu Ana bacila, a ne od njega (see Progovac 1996: 419 and Progovac 2005: 137)
As far as the syntactic placement is concerned, it is also evident that there is no designated position for Wackernagel clitics in Serbo-Croatian. The clitics may seem to cluster, as they are inseparable, but still they do not adjoin to a single head. Bošković (1995: 247) shows that in fact they target different positions in different clauses. He concludes this from the potential interpretations of ambiguous adverbs, such as pravilno ‘cleverly’, which receive a sentential or a manner reading. As indicated in (4), when the adverb pravilno occurs with the auxiliary clitic su, both the manner and the sentential readings of the adverb are available; however, when both the auxiliary and the dative clitic joj are present, only the manner interpretation is possible. Since sentential adverbs are standardly assumed to be adjoined to TP, whereas manner adverbs are adjoined to VP, Bošković proposes that the auxiliary clitic su may target different positions in the structure: it raises higher when it appears on its own than when it is accompanied by a pronominal clitic. (4) a. Oni su pravilno odgovorili Mileni they are-aux correctly answer-ptcp Milena-dat ‘They did the right thing in answering Milena.’ ‘They gave Milena a correct answer.’ b. Oni su joj pravilno odgovorili they are-aux her-dat correctly answer-ptcp ‘∗ They did the right thing in answering her.’ ‘They gave her a correct answer.’ The actual landing site of Wackernagel clitics is difficult to determine for yet another reason. What clitics have in common is phonological deficiency, but as the template in (1) indicates, clitics do not form a morphosyntactic natural class, as they do not share morphological features with each other. Therefore, it is hard to postulate a nonad hoc feature triggering their movement to second position. Yet, the movement is subject to strict syntactic locality conditions, so their placement cannot be attributed
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
to PF requirements. For instance, Progovac (1993) observes that clitic climbing from embedded clauses to main clauses in Serbo-Croatian is subject to the same locality restrictions as topicalization and wh-movement. Namely, like a topicalized phrase (cf. (5)) or a wh-element (cf. (6)), clitics may only raise to the main clause out of subjunctive clauses (cf. (7)) but not out of indicative clauses (cf. (8)). (5)
potpisao a. ∗ Pismo ne kažem da sam letter neg say that am-aux sign-ptcp.m.sg ‘??The letter, I don’t say that I have signed.’ b. Pismo ne želim da potpišem letter neg want that sign ‘The letter, I don’t want to sign.’
(6) a. ?∗ Koga ne kažeš da voliš? who neg say that love ‘Whom don’t you say that you love?’ b. Koga ne želiš da voliš? whom neg wish that love ‘Whom don’t you want to love?’ (7) a. Milan želi da ga vidi Milan wishes that him-acc sees ‘Milan wishes to see him.’ b. ?Milan ga želi da vidi (8) a. Milan kaže da ga vidi Milan says that him-acc sees ‘Milan says that he can see him.’ b. ∗ Milan ga kaže da vidi
(S-C, Progovac 1993)
Although it is difficult to pinpoint the landing site of second position clitics, there is some evidence coming from ellipsis suggesting that pronominal clitics do not adjoin to a single head. Namely, Stjepanović (1998, 1999) notices that the lower (accusative) pronominal clitic may be deleted under identity in VP ellipsis, as shown in (9). If the pronominal clitics adjoined to a single head, this operation should be impossible, as it would involve deletion of a non-constituent. Stjepanović proposes that this means that each pronominal clitic targets the specifier of a different projection. (9) a. Mi smo mu ga dali, a i vi ste we are-aux him-dat him-acc give-ptcp.m.sg and also you are-aux mu ga dali him-dat him-acc give ‘We gave it to him, and you did, too.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
b. Mi smo mu ga dali, a i vi ste mu ga dali, (takodje) c. ∗ Mi smo mu ga dali, a i vi ste mu ga dali, (takodje) (S-C, cf. Stjepanović 1998: 530–2) In Migdalski (2006) I provide additional empirical evidence supporting Stjepanović’s claim that comes from the distribution of the Person Case Constraint (PCC) across Slavic. The PCC is a restriction concerning the combination of pronominal clitics in ditransitive constructions: an accusative clitic that appears with a dative clitic must be in the 3rd person, or else the structure is ungrammatical. To explain the constraint, I adopt Anagnostopoulou’s (2003) account, which states that the PCC is a result of an incompatibility of person and number feature checking. It occurs when a dative clitic adjoins to a single head (such as T) to check a person feature and when subsequently an accusative clitic adjoins to the same head, tucking in beneath the dative clitic. When the person feature of T has been checked by the dative clitic, the accusative may only check the remaining number feature of T. Thus, the derivation converges only if the accusative pronoun does not carry a person feature, but just a number feature. On the assumption that the 3rd person is a null person, the accusative clitic may only be marked for the 3rd person for the derivation to converge. As shown in (10), the PCC is not operative in Serbo-Croatian,3 as the dative clitic im may be accompanied by the non-3rd person accusative clitic te. If Anagnostopoulou’s (2003) account of the PCC is on the right track, this means that the pronominal clitics indeed do not adjoin to a single head. (10) Ja im te preporučujem I them-dat you-acc recommend ‘I recommend you to them.’
(Migdalski 2006: 198)
Further evidence for the assumption that pronominal clitics target XP-positions comes from the interaction of clitics and verbs with negation. In many Slavic languages the negative particle n(i)e, which is assumed to be located in Neg0 , attracts and incorporates into other elements. In Serbo-Croatian, the process is restricted to verbs. Example (11) shows that ne incorporates into the verb that is followed by the pronominal clitic me. If the negation and the verb boli were analysed as separate, non-incorporated elements, the clitic me would occur in the third position in the clause, and this would be the only case of a clitic occurring in the third position in Serbo-Croatian. Crucially, negation may not incorporate into the pronominal clitic, but only into the verb. A possible way to account for this restriction is to assume, as is standard, that incorporation adheres to the Chain Uniformity Condition, which says 3 It seems that the PCC does not hold for other Slavic languages with Wackernagel clitics either, as shown for Slovene by Rivero () and for Czech by Lenertová ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
that the only elements of the same syntactic status (that is, only heads, rather than a combination of a head and a phrase) may incorporate. If this account is correct, the negation facts provide more support for the idea that pronominal clitics target specifiers, and that they do not adjoin to a single head. If the latter process were at work, pronominal clitics should be able to incorporate into negation. (11)
a. Ne boli me neg hurts me-acc ‘It doesn’t hurt me.’ a’. ∗ Ne me boli b. Ne cini mi se da . . . neg seems me-dat refl that ‘It doesn’t seem to me that . . .’ b’. ∗ Ne mi se cini da . . .
(Migdalski 2006: 218)
To summarize, Serbo-Croatian facts indicate that pronominal Wackernagel clitics do not target a consistent syntactic position in the structure, although they may only occur after the sentence-initial element that is a syntactic constituent. There is no uniform trigger for their movement to second position. Empirical observations concerning the distribution of the PCC, the ellipsis of pronominal clitics, and the incorporation of verbal heads into negation indicate that pronominal clitics do not adjoin to a single head, but instead they target specifiers of different projections. .. Verb-adjacent pronominal cliticization Two Slavic languages, Bulgarian and Macedonian, have verb-adjacent pronominal clitics. Since by and large they only differ with respect to the direction of cliticization (Macedonian has proclitics whereas pronominal clitics in Bulgarian are phonologically enclitic), which does not bear much relevance for the topic of this chapter, I will focus on Bulgarian. As shown in (12b), the pronominal clitics do not need to appear in second position, but they may not be separated from the verb by any overt material. (12)
a. Vera mi go dade včera Vera me-dat him-acc gave yesterday ‘Vera gave it to me yesterday.’ b. Včera Vera mi go dade c. ∗ Vera mi go včera dade
(Franks and King 2000: 63)
The linear position of pronominal clitics in the structure is not the only property that distinguishes verb-adjacent clitics from Wackernagel clitics. The examples in (7) demonstrate that clitics may climb out of subjunctive clauses in Serbo-Croatian. This operation is never possible in Bulgarian, whatever the type of complement clause that
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
is selected by the main verb, as shown in (13) for a subjunctive complement, and in (14) for an indicative complement. (13)
a. Manol iska da go vidi Manol wish-3sg that him-cl.acc see-3sg ‘Manol wishes to see him.’ b. ∗ Manol go iska da vidi
(14) a. Manol kazva če go vižda Manol say-3sg that him-cl.acc see-3sg ‘Manol says that he can see him.’ b. ∗ Manol go kazva če vižda
(Migdalski 2006: 217)
As far as ellipsis of pronominal clitics is concerned, Bošković (2002: 331) revisits Stjepanović’s (1998) Serbo-Croatian data quoted in (9) and points out that Bulgarian disallows any kind of deletion of pronominal clitics under identity in coordinate structures. (15)
mu go dali, i vie ste mu a. ∗ Nie sme we are-aux him-dat it-acc give-ptcp and you are-aux him-dat go dali (sušto) him-acc give-ptcp too ‘We gave it to him, and you did too.’ b. ∗ Nie sme mu go dali, i vie ste mu go dali (s˘ušto) c. ∗ Nie sme mu go dali, i vie ste go mu dali (s˘ušto)
Furthermore, Hauge (1999) and Rivero (2005) point out that the Person Case Constraint is operative in Bulgarian (I make the same observation for Macedonian in Migdalski 2006) and in fact leads to very strong unacceptability, unlike in languages with second position clitics. te prepor˘učvam (16) a. ∗ Az im I them-dat you-acc recommend ‘I recommend you to them.’ b. Az im ja prepor˘učvam I them-dat her-acc recommend ‘I recommend her to them.’
(Hauge 1999)
Finally, the sentences in (17) show that negation does not attract verbs in Bulgarian, the way it does in Serbo-Croatian, but, rather, it incorporates into pronominal clitics. (17)
a. Ne me boli neg me-acc hurts ‘It doesn’t hurt me.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski a’. ∗ Ne boli me b. Ne mi se struva, če . . . neg me-dat refl seems that ‘It doesn’t seem to me that . . .’ b’. ∗ Ne struva mi se, če . . .
(Migdalski 2006: 218)
It seems that a straightforward way to account for the contrasts between the two language types is to assume that pronominal clitics in Bulgarian move via headmovement and adjoin to a single head such as T, as suggested in Migdalski (2006; a number of previous proposals including Tomić (1996), Rudin (1997), and Franks (1998) postulated that the pronominal clitics land in separate functional heads such as AgrO and AgrIO). (18)
[TP [T i + j + T] . . . [Aux (e) [VP V ti tj ]]]
This proposal explains the impossibility of clitic climbing in (13) and (14), as head movement is restricted to main clauses, as well as the unavailability of clitic ellipsis in (15), which in Bulgarian would involve the deletion of a non-constituent. Furthermore, it captures the validity of the Person Case Constraint as well as the fact that, in compliance with the Uniform Chain Condition, pronominal clitics may be incorporated into Neg◦ in Bulgarian. Summarizing, the current section has shown that the two patterns of cliticization found in contemporary Slavic differ not only in terms of the linear clitic placement but they also involve distinct syntactic operations: head-adjunction in the case of verbadjacent clitics and XP-movement of clitics to separate specifiers in the case of Wackernagel cliticization. The next section will investigate a diachronic shift from verb-adjacency to second position clitic placement that occurred in some Slavic languages. A natural follow-up of the observations made in the current section is that the shift occurred when the option of clitic head-adjunction to a designated head became impossible. I will argue that this happened when T◦ was lost.
. Pronominal clitics in Old Slavic The investigation of the earliest cliticization patterns in Slavic pursued in this chapter is based on Old Church Slavonic (OCS), the first literary Slavic language. The oldest OCS manuscripts come from the tenth century and are ecclesiastical texts translated from Greek. For the purpose of this study I use examples from Codex Marianus and Codex Zographensis that I extracted using USC Parsed Corpora of Old South Slavic and which are referred to as Pancheva et al. (2007a) and (2007b), respectively. I also make use of OCS and Old Serbian data analysed by Radanović-Kocić (1988). Traditional descriptive grammars state that clitics in OCS comply with Wackernagel’s Law and occur after the first word bearing stress (Sławski 1946; Lunt 1974).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
More recent investigations (Radanović-Kocić 1988; Pancheva 2005) challenge this traditional view and point out that obligatory second position placement was confined to just three clitics: the question/focus particle li, the complementizer clitic bo ‘because’, and the focus/interrogative particle že (Radanović-Kocić 1988: 151). All these clitics encode the illocutionary force of a sentence, and for convenience I will term them operator clitics. Incidentally, it has been observed that clitics specifying clause type and other information structure properties target second position in many unrelated languages (see Axel 2007 on Old Germanic and Roberts 2005 on Celtic), so it seems that in this way OCS follows a general cliticization pattern found in ProtoIndo-European. The placement of operator clitics in OCS is exemplified in (19). (19)
a. Ašte li oko tvo˘e lõkavo bõdet]. if q eye your evil is-pfv ‘If your eye should be evil.’ b. Ašte li že ni i novõjõ razderet]. if q foc not also new tear-fut ‘Or else the new one will tear.’
(Radanović-Kocić 1988: 151)
(Pancheva et al. 2007a)
The sentence in (20) shows that pronominal clitics do not need to cluster with operator clitics located in the second position. They are adjacent to the verbs that they are arguments of, as in the case of the reflexive clitic se˛ and the dative clitic ei. (20) Elisaveti že ispl]ni se˛ vremę roditi ei. I rodi Elizabeth foc fulfilled refl time give-birth her-dat and gave-birth sn]. son-acc ‘When it was time for Elizabeth to have her baby, she gave birth to a son.’ (Pancheva et al. 2007a) Radanović-Kocić (1988: 153) points out that in the data she analysed there are instances of pronominal clitics following the initial element and thus occurring in second position, but in the case of accusative clitic, the clitic is also verb-adjacent (see (21a)). Otherwise, pronominal clitics are placed lower in the structure, adjacent to the verb (see (21b, c)). There are cases of dative clitics occurring in second position, especially in the presence of an operator clitic (see (21d)). Željko Bošković (p.c.) suggests that these are not argumental datives but rather ethical datives. Ethical datives are standardly considered to be operator clitics, as they do not refer to actual arguments, but instead they are pragmatic devices of marking closeness between the speaker and the addressee. Thus, they occupy the same position as other operator clitics.4 4 In contemporary Serbo-Croatian ethical datives also target a higher syntactic position than argumental dative clitics. For example, they may precede sentential adverbs, unlike argumental dative clitics (Bošković : –).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(21)
Krzysztof Migdalski a. Sõdii te˛ pr˘edast] slous˘e judge you-acc hands-over guard-dat ‘The judge hands you over to the guard.’
(Radanović-Kocić 1988: 156)
b. Oca moego v] t˘ex] dostoit] mi byti. father my in these be-appropriate me-dat be ‘I had to be in my father’s house?’ (Pancheva et al. 2007a) c. Ašte desna˘e tvo˘e rõka s]blažbaet] te˛. if right your hand sins you-acc ‘If your right hand causes you to sin.’ d. Dobr˘ee bo ti est]. better as you-dat is ‘It is better for you.’
(Radanović-Kocić 1988: 153–4)
Moving to Old Serbian, in her analysis of texts from the period between the twelfth and the fifteenth century, Radanović-Kocić (1988) notices that the distribution of clitics is initially somewhat similar to what is found in OCS: operator clitics follow the initial word, whereas pronominal clitics tend to occur postverbally (see (22a)). However, from the fifteenth century onwards, pronominal clitics gradually start to appear in second position in Serbian, but according to Radanović-Kocić (1988), at first a clear dichotomy can be observed: initially, pronominal clitics are in second position only when they are accompanied by operator clitics (see (22c)). If operator clitics are absent, pronominal clitics are in most cases after the initial word, but they can also be verb-adjacent. At a later stage, pronominal clitics uniformly target second position even when operator clitics are not present (see (22d)). Still, the change is not uniform across dialects. Example (22b) shows that in the dialect of Montenegro pronominal clitics were verb-adjacent and occurred low in the structure as late as the turn of the eighteenth century. Incidentally, in this dialect morphological tense distinctions were retained the longest. (22) a. Koi e privo im˘ı byl˘ı which is first they-dat be-ptcp.m.sg ‘Which was first to them.’ (pre-15th-c. Serbian, Radanović-Kocić 1988: 160) b. Ako iguman sakrivi mi. If prior do-wrong me-dat ‘If the prior does me wrong.’ (18/19th c., Montenegro dialect, Radanović-Kocić 1988: 166) ime taiti. c. Kto li ga who q him-acc has hide-inf ‘Who will be hiding him?’ (Bistrica Manuscript, 15th-c. Serbian, Radanović-Kocić 1988: 158)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
d. Dokle mu se ne ispravi. until him-dat refl neg corrects ‘Until it is corrected (for him).’ (Studenica Chronicle, ca. 1418, Serbian, Radanović-Kocić 1988: 158) A remarkable property of cliticization in the earlier stages of Serbian is that clitics need not stay adjacent to each other. There are many cases in which the auxiliary clitic is separated from a pronominal clitic by some lexical material, as shown in the example from the nineteenth century in (23). The separation is generally not allowed in contemporary Serbo-Croatian, in which clitics are adjacent to each other. The fact that such patterns were possible only a century ago suggests that the generalization of Wackernagel’s law to all clitics was a long process, with dialectal variation and with potential cases of grammar competition. (23) Koji bi barem dodao joj rje˘cnik who would at least hand her-dat dictionary ‘Who would at least hand her the dictionary.’ (19th-c. S-C, Radanović-Kocić 1988: 167) Recall from Section 11.2 that verb-adjacent and second position cliticization differ not only with respect to the position of clitics in the structure, but they also show variation concerning clitic climbing, ellipsis under identity, and the Person Case Constraint. It is to be expected that these contrasts arose along with the diachronic change described in this section. In the next parts of this chapter, I will argue that the change was driven by the impoverishment of morphologically expressed tense distinctions in all Modern Slavic languages except Bulgarian and Macedonian, and I will provide more evidence for this hypothesis from cliticization in other Slavic languages. I will assume that the impoverishment is reflected through the loss of T◦ , the head that pronominal clitics adjoin to in verb-adjacent clitic languages.
. Tense and aspect in Old Slavic This section investigates the way tense is expressed in Slavic and shows that while Old Church Slavonic, Bulgarian, and Macedonian have robust tense distinctions, in other Slavic languages temporal relations are expressed exclusively through aspect. The fact that all Slavic languages make profuse use of aspectual distinctions is well known. Almost all Slavic verbs form aspectual pairs, with one member of the pair being perfective, the other one imperfective. Aspect is marked even on nominalizations. What is less known is how aspect specification originated and how it interacted with tense marking diachronically. In Proto-Slavic, verbs had a typical Proto-Indo-European three-morpheme structure, with a stem that was formed by a root, followed by a thematic suffix and an
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski Table .. The paradigm of the verb ∗ nesti ‘to carry’ in the present tense
1 2 3
Singular
Dual
Plural
nes-¯o-m] nes-e-š] nes-e-t]
nes-e-vě nes-e-ta nes-e-te
nes-e-m] nes-e-te nes-o-nti
(Proto-Slavic, Długosz-Kurczabowa and Dubisz 2001: 265)
Table .. The paradigm of ∗ nesti ‘to carry’ in the present tense (a later version)
1 2 3
Singular
Dual
Plural
nes-õ nes-e-š] nes-e-t]
nes-e-vě nes-e-ta nes-e-te
nes-e-m] nes-e-te nes-õt]
(Late Proto-Slavic, Długosz-Kurczabowa and Dubisz 2001: 265)
inflectional marking. The thematic suffix assigned a stem to a particular inflectional paradigm, but it could also render aspectual information. The structure is exemplified in Table 11.1 for the verb ∗ nesti ‘to carry’ in early Proto-Slavic. In later stages of Proto-Indo-European, verbal morphology was simplified, and in most cases thematic suffixes blended with the inflectional endings. As a result, verbs acquired a two-morpheme structure, as shown in Table 11.2, which presents simplified variants of the 1st person singular and the 3rd person plural forms of the Late-ProtoSlavic paradigm of the verb ∗ nesti, with just two morphemes. In the case at hand, the change was triggered by phonological readjustments caused by the nasalization of the vowel o¯ when it was followed by nasal consonants, as in nes-o-nti → nes- õt]. Because of the fusion of the two morphemes it became increasingly more difficult to mark aspect morphologically and the aspectual system of Late-Proto-Indo-European became less consistent, although some languages developed new aspectual tenses. For instance, imparfait and passé simple developed in French, though an equivalent of imparfait never emerged in Germanic. However, Proto-Slavic was in this respect the most conservative Indo-European language, because it retained the original ways of marking aspect and at the same time inherited the simple past tenses of Proto-IndoEuropean, aorist and imperfect. Consequently, it radically extended the aspect system (see Klemensiewicz, Lehr-Spławiński, and Urbańczyk 1964: 242–53; Młynarczyk 2004).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
In addition, Old Slavic developed a new complex tense constructed with verbal adjectives ending in ∗ -lo reanalysed as l-participles (agreeing with the subject in phifeatures) and the verb BE occurring in different aspectual forms as the exclusive auxiliary: the imperfective form in present perfect, the perfective form in the future, and in the imperfect tense (or in aorist marked for imperfective aspect) in the past perfect. Table 11.3 presents the verb nesti in various tenses in Old Church Slavonic, in perfective and imperfective aspect. This table shows that aspect was doubly marked in OCS: through the tense forms of aorist and imperfect and through the perfective/imperfective aspectual morphemes. In the majority of cases aorist forms were marked for perfective aspect and imperfect forms for imperfective aspect, yet the ‘contradictory’ combinations of perfective forms of the imperfect tense or imperfective forms of the aorist were possible, though less common, as indicated by Dostál’s (1954) calculations in Table 11.4. The fact they were possible indicates the independence of tense and aspect systems. With the exception of Bulgarian and Macedonian, all Slavic languages have lost aorist and imperfect. The only tense that is morphologically marked in these languages is the present tense, and past and future events are characterized exclusively via aspectual morphology or modality. The compound tense formed with the imperfective variant of the auxiliary ‘be’ and the l-participle, which in Old Church Slavonic had a
Table .. Tense and aspect forms in OCS Tense/Aspect
Imperfective
Perfective
3sg present 3sg aorist 3sg imperfect 3sg perfect 3sg future II
neset] nese nesěaše nesl] jest] bõdet] nesl]
poneset] ponese ponesěaše ponesl] jest] bõdet] ponesl]
(the verb nesti ‘to carry’ in different tenses in OCS, van Schooneveld 1951: 97)
Table .. Combinations of the aspectual tenses with imperfective and perfective aspectual markings in OCS texts Tense/Aspect
Imperfective
Perfective
99 40
1 60
Imperfect Aorist (Dostál 1954: 599–600)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
resultative meaning (similar to that of the present perfect in contemporary English) has been adopted as the default past tense. The participle may occur in the perfective or imperfective aspect.5 (24) Ana je (na)pisala pismo Ana is-aux write-pfv/ipfv.ptcp.f.sg letter ‘Ana wrote/was writing a letter.’
(S-C)
With regard to future tenses, in contemporary East and West Slavic languages they are either constructed with the perfective form of the auxiliary BE accompanied by the l-participle or the infinitive (cf. (25a)), or they are formed by using a perfective form of the verb in the present tense (cf. (25b)). (25) a. Będę pisać/pisał list am-pfv write-ipfv.inf/ptcp.m.sg letter ‘I will be writing a letter.’ b. Napiszę list write-pfv letter ‘I will write a letter.’
(Polish)
As far as the timing of the tense decline is concerned, it coincides with the loss of verb-adjacent pronominal clitics. According to Vaillant (1966: 60), aorist was limited to certain verb forms in Old Slovene, and this fact ties in with the occurrence of Wackernagel pronominal clitics already in The Freising Manuscripts, the oldest Slovene text from the tenth to eleventh century. In Czech both aorist and imperfect were lost in the fourteenth century (Stieber 1973: 53), whereas Serbo-Croatian is a curious case, because although aorist is still ‘taught’ at schools, it is restricted to certain dialects, mostly in Montenegro, where it is still used by modern fiction writers (see Lindstedt 1994: 39). Interestingly, this corresponds to the observation made earlier in this chapter that pronominal clitics did not consistently appear in second position as late as in the nineteenth century, in particular in Montenegrin dialects (see example (22b)). Bulgarian, which has retained verb-adjacent clitics, still productively uses aorist and imperfect, and like OCS, it also freely allows the ‘contradictory’ combinations of tense and aspect values (see Lindstedt 1985), which suggests that tense and aspect form two independent systems. In Macedonian, imperfect is the default past tense for verbs specified for imperfective aspect, whereas aorist is the default tense for perfective verbs, but it is also becoming obsolete (Tomić 1989: 366). According to Lunt (1952: 90), aorist may have declined due to its morphological syncretism with imperfect in all but 5 East Slavic languages like Russian or Ukrainian lost the auxiliary in the past tense, so one might claim that the l-participle is specified for the past tense in these languages. This would not be correct, however, as the participle is used in a number of non-past/non-tensed contexts, such as the subjunctive mood and conditional constructions.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
the 2nd and the 3rd person singular forms. Friedman (2002: 267) points out that the decline of the aorist usage is a recent innovation, as imperfective aorist had been found in the literature till as recently as the middle of the twentieth century. Strikingly, these current developments relate to an unstable clitic system of Macedonian. Although it has verb-adjacent proclitics, in non-verbal predicates clitics are in second position, which in Tomić’s (2000) and Bošković’s (2001) views suggests that Macedonian is in an intermediate stage between a language with verb-adjacent and with Wackernagel clitics. e tatko (na deteto) (26) a. ∗ Mu him-dat is father to child-the b. Tatko mu e na deteto father him-dat is to child-the ‘He is the father of this child.’
(Tomić 2000: 295)
. Towards an analysis It has been established that the shift in the position of pronominal clitics was contemporaneous with morphological impoverishment of tense marking. I propose that these two processes are related and that the decline of tense was reflected via the loss of TP, as a result of which pronominal clitics cannot adjoin to T any more. This proposal entails that TP, a core syntactic projection, may be absent in some languages. However, this has been independently suggested for a number of unrelated languages in synchronic and diachronic analysis. For example, the presence of TP in Chinese is a matter of debate. Recently, Lin (2010) has argued contra Sybesma’s (2007) tensed account of Chinese and showed that a TP-less analysis is able to account for a number of empirical facts, such as the ability of a nominal predicate to serve as the main predicate, the absence of a distinction between finite and non-finite clauses, and the lack of case-related movement in Chinese. Likewise, van Gelderen (1993) suggests that the presence of the syntactic category T is a matter of parametric variation and language change. She proposes that T emerges at the end of the Middle English period (ca. 1380) and coincides with the rise of do-support. Kiparsky (1996), who adopts van Gelderen’s proposal, posits that the OV to VO shift in Germanic is related to the rise of T (his I). A TP-less approach has also been postulated for Old English by Osawa (1999), whereas Fukui (1988) and Shon, Hong, and Hong (1996) give arguments for the lack of TP in Japanese and Korean, respectively. Recently Bošković (2012) has suggested that the TP projection is available only in languages that have nominals with the DP layer. He bases the link between DP and TP on Chomsky’s (1986) assumption that SpecDP is the landing site of the counterpart of movement to SpecTP in nominalizations such as John’s destruction of the city. Moreover, he postulates a licensing mechanism holding for functional
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
elements which requires that an (unvalued) interpretable feature be morphologically realized. Following Pesetsky and Torrego (2007), he assumes that there is an Agree relationship between T and V, in which T has an unvalued interpretable tense feature, while T has a valued uninterpretable tense feature. (27) T (unvalued iTense) V (valued uTense) Since the tense feature of T is interpretable and unvalued, it must be morphologically expressed. This is what happens in English, where the opposition between the -ed morpheme versus Ø renders the contrast between the past and the present tense. If tense is not morphologically expressed in a language, the TP projection is missing. Bošković observes that DP/TP-less languages share a lot of properties, and they can be accounted for with the proviso that the DP/TP layers are not universally present. For instance, DP/TP-less languages lack expletives. Since the role of expletives is to satisfy the EPP, which is a property of the TP projection, these elements cannot be present in languages that lack TP. Furthermore, many languages exhibit certain subject–object asymmetries. Thus, in English extraction is possible out of objects, but not out of subjects (see (28)). Correspondingly, English allows extraction of an object, but not of a subject across a clause-mate that. The latter constrain is referred to as the that-trace effect (see (29)). (28) a. ∗ Who did friends of see you? b. Who did you see a friend of? (29) a. Whoi do you think that John saw ti ? b. ∗ Whoi do you think that ti saw John? Interestingly, these restrictions cross-linguistically apply only to the subjects located in SpecTP. Gallego and Uriagereka (2007) demonstrate that Spanish, a language without the that-trace effect, does not permit extraction of the type exemplified in (28), but the restriction applies only to preverbal subjects, that is, those hosted in SpecTP. Extraction out of postverbal subjects, which are standardly assumed not to raise to SpecTP, (see, for instance, Torrego (1984), Suñer (1994), Ordóñez (1998), and VillaGarcía (2012) for a discussion of the status of subjects in Spanish), is possible. Bošković points out that his examination of a large sample of languages indicates that articleless languages do not display the subject–object asymmetries presented in (28) and (29). This observation provides support for the hypothesis that these languages lack the TP projection and that subjects in these languages land in a different projection than SpecTP. Finally, Bošković observes that languages without articles do not exhibit the sequence of tense. If this phenomenon is related to the TP projection, it is expected that it does not occur in languages that lack this projection.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
On the loss of tense and verb-adjacent clitics in Slavic
Moving back to pronominal clitics, I have proposed in this chapter that certain Slavic languages lost verb-adjacent pronominal clitics due to the loss of TP. Since this proposal is based on the idea that pronominal clitics adjoin to T, it is necessary to motivate it and spell out some general assumptions about the way cliticization works. According to Chomsky (1995a: 249), pronominal clitics are ambiguous categories that have both XP and X0 properties. Such ambiguity is possible in Chomsky’s Bare Phrase Structure approach. Clitics raise from their phrasal theta-positions within the VP and move to adjoin to an inflectional head. In the final step of the adjunction they are heads, as otherwise the operation would violate the Chain Uniformity Condition. However, they must raise to their landing position via XP-adjunction, given that they cross a number of heads on the way and the raising would otherwise violate the Head Movement Constraint. As has been pointed out earlier in this chapter, clitics in Slavic do not share any morphosyntactic features. Although pronominal and auxiliary clitics are marked for phi-features, modal and operator clitics are not phi-feature bearers. The only property that all clitics in Slavic have in common is phonological deficiency and a resultant need for a prosodic host for support. Moreover, recall from example (4) that there is no consistent syntactic second position. Auxiliary clitics may move across sentential adverbs, but if they are accompanied by pronominal clitics, they must stay lower. Consequently, finding a uniform trigger for the movement of second posting clitics does not seem to be feasible. A standard assumption made in the literature about verb-adjacent clitics on the basis of Romance languages is that clitics must procliticize by adjoining onto the verb located in T. A number of analyses have been put forward to motivate this requirement. For instance, in Nash and Rouveret (2002: 177), the cliticization on T occurs because clitics must adjoin to a ‘substantive’ (lexical) category endowed with active ϕ-features. Irrespective of the actual motivation, the fact that the PCC holds in Bulgarian and Macedonian shows that the operation indeed takes place. A question that arises, though, is how to explain clitic movement to second position and why pronominal clitics move from their base VP-position even in the absence of T. The movement cannot be attributed to prosodic factors, as clitics can be phonologically supported in their base position. A potential trigger might be semantic requirements, as in Uriagereka (1995), who states that since pronominal clitics are specific and referential, they must move out of VP, given Diesing’s (1990) Mapping Hypothesis. A final issue to be addressed concerns the feasibility of the two cliticization patterns. In this chapter I discuss a switch from verb-adjacent to second position cliticization that becomes available only when T is lost. A question that might be asked is what prevents second position from being a valid option when T is present. Why is the option of adjoining clitics to the verb in T preferred over having them in second position? I suggest that this preference is due to a principle of phonology/syntax matching,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Krzysztof Migdalski
which says that a syntactic constituent should correspond to a prosodic word. This is what happens in languages with verb-adjacent clitics, in which pronominal clitics are adjoined to a single head, but not in languages with second position clitics such as Serbo-Croatian, in which each clitic targets the specifier of a separate functional head.
. Summary In conclusion, this chapter shows that the decline of tense in Slavic has led to a substantial modification of cliticization patterns, which in turn triggered further syntactic repercussions related to negation incorporation, the possibility of ellipsis of pronominal clitics, and the presence of the PCC. Interestingly, the process described in this chapter seems to be the reversal of what happened in Ancient Greek, where, according to Kiparsky and Condoravdi (2002), Wackernagel pronominal clitics became verb-adjacent with the rise of IP. If the analysis presented here is correct, this suggests that the emergence and the decline of TP are both possible diachronic options in language, made available by Universal Grammar.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek D I M I T R I S M I C H E L I O U DA K I S
. Introduction This chapter deals with a series of changes in the syntax of so-called dative arguments in Greek and suggests that they constitute manifestations of a deeper syntactic change affecting the type of inherent Case involved in each period and each class of arguments.1 This change was arguably not directly driven by morphological change, despite being preceded by a major related change in the declensional system of Greek, namely the loss of the morphological dative and its replacement by genitive or accusative in different varieties. Two significant questions arise in relation to the history of datives in Greek: (i) how and why inherent Case, i.e. abstract dative, was not lost in the diachrony of Greek, despite the loss of its morphologically distinct exponent; (ii) how and why the syntactic behaviour of abstract dative Case changed, despite its remaining a rather clear case of inherent Case (i.e. theta-related, not alternating with other Cases depending on the configuration). In relation to (ii), in particular, I will establish that dative constructions in Greek underwent some major changes such as: (a) the emergence of dative alternations and dative-shifted constructions (b) the transition from a grammar allowing for weak pronominal clusters subject to a weak version of the P(erson) C(ase) C(onstraint), or no PCC at all, to a system where only the strong PCC is operative (c) the emergence of minimality/defective intervention effects (DIEs) in DAT-above-NOM configurations (in raising/unaccusative constructions) (d) the rise of dative experiencers with subject-like behaviour. 1 I would like to thank Ian Roberts, Ioanna Sitaridou, and Anna Roussou for helpful comments and suggestions, as well as the audience of DiGS XII. I also thank Adam Gibbins and Georg Hoehn for proofreading the final draft. All remaining errors are my own. The financial aid of the A. Onassis Foundation and the A. Leventis Foundation is also hereby gratefully acknowledged.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
In this chapter I will mainly focus on the diachrony of Cypriot Greek (CG), which is arguably the best-documented Greek variety with respect to the phenomena in question. The diachronic scenario advanced suggests that the abstract dative changed from one type of inherent Case to another, becoming visible to Agree/Move, following a reanalysis in the syntax of ditransitives and the rise of dative-shifted constructions. Thus, the configuration underlying ditransitive constructions with indirect object (IO) DPs changed from ‘DPD(irect)O(bject) > DPI(ndirect)O(bject) ’ (where ‘>’ indicates asymmetric c-command) to ‘DPIO >DPDO ’ (1)–(2). (1)
. . . edeiksen tous agnooumenousi alle:loisi . . . showed.3sg the missing.acc.pl each-other.dat ‘She revealed the missing heroes to each other.’
tisi (2) a. edhiksa tis Mariasi ton eafto showed.1sg the Mary.gen the self.acc her
(Hellenistic Greek)
(Modern Greek)
tu eaftu tisi b. ∗ edhiksa ti Mariai showed.1sg the Mary.acc the self.gen her ‘I showed Mary herself.’ It will be argued that this reanalysis was not triggered but simply facilitated by the morphological loss. This may explain the large discrepancy between the dates of these two changes. Another possible implication of this discussion is the idea that the precise nature of an abstract Case feature, although only inherent Case is discussed here, is largely independent from its morphological exponent, since both accusative and genitive substitutes in different Greek varieties exhibit the same syntactic behaviour and probably underwent the same—or a similar—path of changes; see e.g. (8)–(8’).
. Background assumptions The developments in (a)–(d) in the previous section will be shown to correlate and cluster together in such a way that we may reduce them to a bi-partite syntactic distinction, in which the distinctive feature should be the accessibility/visibility of the ‘dative’ DP to Agree/Move. Following Chomsky (2000, 2001), I assume that it is the value and the ‘timing’ of the valuation of the (abstract) Case feature of a DP that determines whether or not it is an active goal, and consequently that minimality in ϕ-Agree must be relativized to Case features. This implies that Chomsky’s (2000) system should be construed as follows: (i) DPs with unvalued, uninterpretable Case features are active goals (ii) DPs with uninterpretable Case features that have already been valued by a lower ϕ-head H1 when probed by a higher ϕ-head H2 are ‘defective interveners’, in the sense that they cannot value H2 ’s [uϕ] while preventing it from probing further down
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
(iii) DPs with fully interpretable and lexically valued theta-related Case do not induce any minimality effects, as ϕ-heads only look for [uCase], i.e. only [uCase] can make a DP visible to a ϕ-probe. What complicates the picture then, on such an analysis, is the fact that all Greek datives both diatopically and diachronically apparently bear inherent Case which cannot be suppressed in either ECM or passivization.2 This forces us to postulate (like in McGinnis 1998 and Rezac 2008) two types of inherent Case, one that only allows a dative DP to behave as described in (iii) above, and a hybrid type that allows a dative to behave as in (i) or (ii), while retaining its PF- and LF-interpretable part intact. I will assume that the latter type, i.e. ‘active’ inherent Case (AIC), has an uninterpretable/lexically unvalued component which is valued and deactivated via Agree with a ϕ-head, thus meeting Chomsky’s (2000: 127) definition of ‘quirky Case’ (‘(theta-related) inherent Case with an additional structural Case feature’). As such, the computational system recognizes it as an instance of [uCase], while its fully interpretable counterpart is an instance of [iCase]. Configurations such as (3), then, are open to the possibilities described in (i)–(iii) above. (3)
H[uφ] . . . . . . DAT[iφ, uCase/uCase/iCase] . . . . . . DP[iφ, uCase] (i) (ii) (iii)
(i) If DAT bears an unvalued [uCase] feature, then it is an active goal that can match and fully Agree with H, as long as it is its closest potential goal, deactivating its own [uCase], and preventing H from probing further down (ii) If DAT is the most local goal for H but it has already valued its [uCase] via Agree with a closer ϕ-head H’, then DAT is a defective intervener, blocking Agree between H and any lower active goal. This DIE can be obviated if DAT undergoes some (movement-related) process, e.g. cliticization/clitic-doubling (CD) (see Section 12.3.2.3 for analysis), which puts the head of DAT’s chain outside H’s Agree domain, following Chomsky (2000, 2001) (iii) If DAT bears [iCase], it is transparent/invisible for Agree purposes, and H can unproblematically Agree with the next closest DP with [uCase]. The history of dative arguments in Greek provides evidence for a change from dative as [iCase] to dative as [uCase]. Recall that the later stages of Greek feature DPIO >DPDO : evidence that this pattern involves ‘active’ dative Case (i.e. [uCase]) comes from the fact that in passives with promoted, i.e. nominative DOs, the 2 But see Anagnostopoulou and Sevdali () for datives becoming nominative in a restricted number of passive constructions in Classical Greek. Crucially, in all the varieties we examine here, abstract dative (whatever its morphological exponent) is never absorbed.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
configuration [T . . . DPIO . . . DPDO ] is subject to DIEs as described in (3ii) above (see the discussion below on Modern CG and Standard Modern Greek (SMG)). As will become obvious later on, the diachrony of Greek lends support to a strong correlation between the availability of AIC, the availability of dative alternations and the strong PCC. In order to capture this—not otherwise obvious—correlation, I propose that [uCase] on IO DPs is necessary in order to render them active for an Agree relationship which takes place only in the dative-shifted alternant in languages with dative alternations, such as (Modern) Greek, English, etc. As is well known, the most striking difference between prepositional and double-object constructions is the animacy restriction on the latter, cf. (4). (4) I sent the paper to the conference/?? the conference the paper (from Adger and Harbour 2007) I will follow the widely held assumption that applicative heads are universally present in Double-Object Constructions (DOCs), hosting IO DPs in their Spec (see Marantz 1993 and Anagnostopoulou 2003, among others), but not in prepositional ditransitive constructions, and I will further assume that Appl in DOCs bears a probe which can only match animate DPs/pronouns, i.e. an unvalued/uninterpretable feature encoding or entailing animacy. In languages with the strong PCC, this feature could be what Adger and Harbour (2007) refer to as [±Participant], which is valued as follows: (5)
a. [+participant] = a pronoun referring to a discourse participant, either a speaker or a hearer, i.e. a necessarily conscious (at the time of the utterance) participant of the event as well. b. [–participant] = a DP/3rd-person pronoun not referring to a discourse participant, but still able to be assigned a theta-role which requires that its referent be capable of mental experience (cf. Reinhart’s (2000) [± m(ental state)]), e.g. recipient/benefactive/experiencer. c. no value = any other DP/3rd-person pronoun (adapted from Adger and Harbour 2007: 16)
IO DPs, then, need [uCase] to become visible to Appl’s [uParticipant]. I will assume that goal DPs are first merged below DO, like in prepositional ditransitives, in accordance with Baker’s (1997) U(niformity) of T(heta)-A(ssignment) H(ypothesis). Appl probes and Agrees with them, also attracting them to Spec-Appl. There, DPIO can value its [uCase] under full ϕ-Agree with v∗ . The idea that v∗ values DPIO ’s [uCase] rather than Appl itself is empirically motivated by languages such as English, in which the shifted DP bears structural Case, accusative in active ditransitives (‘He gave me the book’), but nominative in goal-passives (‘I was given the book’). Theoretically, the basic background assumption is that Voice is the real phase head, which obligatorily
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
transmits its ϕ-features and hence its Case-assigning capacity to any verbal heads in its domain up to the next phase boundary (cf. Chomsky’s (2008) ‘feature inheritance’). Therefore, in bi-eventive VoicePs such as ditransitives (where a causation and a transfer event are involved), there are two verbal heads, which both inherit a ϕbundle from Voice. Appl, not being verbal, does not inherit any features from Voice. Its merger only creates a configuration such as (7), in which both internal arguments can be matched by a Case assigner (see Michelioudakis (2011) for more details and motivation). In (6), where Appl is absent, the two verbal heads need not and cannot project separately (see also Pylkkänen (2008) on ‘packaged’ heads). When a 1st/2nd-person, i.e. [+Participant], DO intervenes between Appl and DPIO , the latter cannot be matched and attracted by Appl to value its [uCase], which derives the strong PCC. Furthermore, since both IO and DO in DOCs have [uCase], I will assume that DOCs involve two v-heads, each with its own ϕ-set. Therefore, the representations of ditransitives with prepositional IOs/low IO DPs and DOCs are (6) and (7) respectively. (6)
vP EA v
VP DO V
IO
v∗P
(7)
EA v* [uφ]
ApplP IO Appl v2P [uParticipant] VP
v2 [uφ] DO
V
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
. Dative arguments in the diachrony of Greek Although the morphological substitution of the ancient Greek dative started as early as Hellenistic Greek or Koine, i.e. the Greek of the Roman times, and was completed by the end of the thirteenth century as far as CG is concerned (see also Markopoulos 2010), many syntactic properties of the ancient/Hellenistic dative were not lost until long after the morphological loss. In texts of the fifteenth and the sixteenth century we still find goal DPs in configurations befitting DPs with [iCase], i.e. purely interpretable/inactive inherent Case, although they are now realized by the morphological genitive instead of the dative. A similar discrepancy is observed in dialects where the dative was replaced by the accusative, e.g. Pontic Greek (on which see Michelioudakis and Sitaridou 2012). Taking DIEs of datives in T-Agree to be a tell-tale sign of AIC, it is clear that accusative ‘datives’ in the varieties spoken nowadays in Northern Greece, including Pontic, give rise to such intervention effects, making the (morphologically accusative) dative clitic obligatory (see also Dimitriadis 1999: 106). However, there are clear signs that in Medieval Pontic Greek, in which the morphological substitution had already taken place, cliticization/CD of the ‘dative’ argument in DAT-above-NOM configurations is optional, indicating that there is no minimality violation that needs to be obviated. This may be so only if the goal/experiencer argument of the unaccusative construction bears a genuine [iCase] feature, transparent/invisible to Agree, despite being morphologically indistinguishable from, say, (accusative) direct objects (DOs), which are clearly structurally Case-marked. ∗ (ton) arese/irthe i idhea na (8) a. (Ton Kosta) The Kostas.acc him.acc.cl appealed-to/came.3sg the idea sbjv. aniksume maghazi (Northern Greek) open.pfv.1pl store ‘Kostas liked/came up with the idea to open a store.’
to b. (??Tin Anastan) eghraften ∗ (aten) The Anasta.acc was-written her.acc.cl the ghrama letter.nom ‘The letter was written for Anasta.’
(Modern Pontic Greek)3
(8’) a. Sinevi aftin thanatos (Medieval Pontic Greek) Happened.3sg her.acc/dat death.nom ‘She happened to die’ (literally: ‘death happened to her’) (Vazelon, 78,10,1291, 13th–15th cent. ad)
3
Pontic Greek does not allow CD and even CLLD is quite marginal.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
Table .. A typology of dative DPs in the history of Greek Morphological case Abstract Case Koine Medieval Pontic Medieval Cypriot Modern Pontic/Northern Greek Modern Cypriot SMG
mDAT mACC mGEN mACC mGEN mGEN
iCase iCase iCase, emergence of IOs with uCase uCase (uniformly) uCase for goals, iCase for experiencers uCase (uniformly)
b. Eparadothi to milon aftin was-given.3sg the apple.nom her.acc/dat ‘The apple was given to her.’ (Vazelon, 23,5,1260) (data due to Vagiakakos 1964) Table 12.1 presents an overview of the case/Case properties of datives in historical/synchronic varieties of Greek. .. From Classical Greek to Koine The Classical Greek dative was a syncretic Case, doing duty for dative proper, and two lost cases, the locative and the instrumental (cf. Smyth 1920: 337). Thus, the argumental uses of the dative include (a) indirect objects (recipients, though not sources, which were realized as genitives), (b) benefactives/malefactives, (c) unique complements of certain verb classes (e.g. ‘follow’, ‘help’, and their synonyms, as well as P+V compounds, when P normally takes DPdat complements), (d) experiencers selected by piaceretype psych predicates or impersonal/raising predicates, (e) locative arguments (with unaccusative motion verbs), etc. None of these semantic/syntactic roles could be fulfilled by any PPs at this stage. Bare dative DPs could also be used adverbially, e.g. as locative or instrumental modifiers. What is particularly relevant for our discussion here is that dative DPs could be either animate or inanimate, when fulfilling (almost) any of their possible roles. In Hellenistic Greek, the use of morphological dative is in decline: due to a series of phonological changes, especially in the vowel system, dative suffixes are becoming homophonous with genitive suffixes in some dialects, or indistinguishable from the accusative in others—see Horrocks (2010: 114–17) about the emergence of morphological substitutes and its morphophonological triggers. However, as far as syntax is concerned, dative Case, whatever its morphological exponence, retains a significant part of its classical distribution, especially its argumental uses, for both animate and inanimate (9a) arguments. In ditransitives, in particular, we still do not find any dative alternations. Having said that, the locative and other adverbial uses of the dative now
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis Table .. DP/PP competition in Hellenistic Greek
mDAT
Animate/human arguments of dative verbs [–animate]/[–human] arguments of dative verbs Locatives (both argumental and non-argumental)
PP subside considerably, i.e. the lost locative that survived in the dative of Classical Greek is now entirely lost, and PPs headed by appropriate Ps are taking on these syntactic functions (9b) with increasing frequency (see Humbert 1930) (Table 12.2). (9) a. Ou dunatai de o ophthalmos eipein te:i cheiri ‘chreian not can.3sg the eye.nom tell.inf the hand.dat need sou ouk echo:’ e: palin he: kephale: tois posin ‘chreian your.sg not have.1sg or again the head.nom the feet.dat need humo:n ouk echo:’ your.pl not have.1sg ‘And the eye cannot say to the hand: I need not thy help; nor again the head to the feet: I have no need of you’ (Ad Corinthios, 12.21.2, 1st cent. ad) b. dio paredo:ken autous ho theos en tais epithumiais to:n hence abandoned them the god.nom in the desires.dat the kardio:n auto:n hearts.gen their.gen ‘That is why God abandoned them to the shameful desires of their hearts’ (Ad Romanos, 1.24.1, 1st cent. ad) Significantly, at this stage datives do not induce any DIEs in long-distance agreement with nominative themes in raising (10)4 or passive (11) constructions. (10) pro dokoumen men gar auto:i haptesthai ekeino:n pro.1pl seem.1pl therefore him.dat touch.inf those.gen ‘we therefore seem to him to be touching those’ (Themistius, Aristotelis de anima paraphrasis, 5,3.75.8) (11)
he: basileia tou theou [. . .] dothe:setai ethnei the kingdom.fem the God.gen [. . .] will-be-given.3sg nation.dat poiounti tous karpous aute:s making.dat the fruits.acc her.gen ‘God’s kingdom will be given to a nation producing [Kingdom] fruitage.’ (Matthew, 21.43.2)
4 There is indeed evidence from anaphoric binding (omitted here) that dative experiencers with impersonal/raising verbs in ancient Greek could bind into (and therefore had to asymmetrically c-command) nominative themes.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
Furthermore, we can find evidence for DO DPs asymmetrically c-commanding IOs, and no data indicating a reverse underlying configuration. For instance, in (12), the embedded possessor of DO is an R-expression, which is then repeated as the recipient/IO. An underlying representation such as IO>DO (in A-positions) would make (12) look like a Principle C violation, while [DO [Possi ]]>IOi does not. Similarly, (13) is an example of variable binding and, finally, (14) clearly demonstrates that anaphoric binding of IO by DO is possible, as in (6), but not vice versa. (12)
toinun apodote ta Kaisaros Kaisari kai therefore give/pay back the.acc.pl Caesar.gen Caesar.dat and ta tou theou to:i theo:i the.acc.pl the.gen.sg God.gen the.dat.sg God.dat ‘therefore, render unto Caesar the things that are Caesar’s, and unto God the things that are God’s’ (Marc, 12.17.2)
(13)
a. Kai edo:ken autoni te:i me:tri autoui and gave.3sg him the mother.dat his ‘And (Jesus) gave him to his mother.’ b.
(Luke, 7.15.2)
∗ Edo:ken
aute:si aute:ii ton uion gave.3sg her.dat the son.acc her.gen ‘He gave her her son.’
(14) a. Po:s oun he: theos [. . .] tous agnooumenousi edeiksen how so the goddess [. . .] the missing showed.3sg lekso: alle:loisi each-other.dat tell.1sg.fut ‘So now I will tell you how the goddess showed/revealed the two missing heroes to each other.’ (Chariton, Callirhoe, 8.1.5.2) b. ∗ Tois agnooumenoisi alle:lousi edeiksen5 the missing.pl.dat each-other showed.3sg ‘She showed the missing heroes each other.’ Finally, as already implied, this syntactic behaviour coexists with the apparent absence of strong PCC effects: (15) is an example of a cluster of 1st- and 2nd-person weak pronominal forms, which have all the defining properties of clitics (they are accentless, systematically enclitic, morphologically simpler than strong pronouns— see Cardinaletti and Starke 1999), although they might correspond to what Roberts (2010) calls D-clitics. Given the loss of 3rd-person weak/clitic pronouns at this stage of the history of Greek, it is hard to tell whether a weak version of the PCC (see Bonet 1991 on Catalan) or no PCC at all is operative. However, interestingly, Hellenistic Greek 5 The asterisk here indicates that the configuration is always avoided, even when the context would allow/favour it.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
apparently allows 1st/2nd-person accusative clitics in the presence of full lexical IO DPs (16), while the equivalent of this configuration in a strong PCC language like Modern Greek (either Standard or Cypriot) is considerably degraded (see also Adger and Harbour (2007: 7) on the PCC with full IO DPs in Kiowa). (15)
a. Omoson [. . .] te:n Aphrodite:n te:n deiksasan swear.ipfv the Aphrodite.acc the show.past.ptcp.fem.acc me soi me.acc.cl you.dat.cl ‘Swear to Aphrodite, the one who revealed me to you.’ (Chariton, Callirhoe, 3.2.5.4) b. Kai gar aute: me soi diephulaksen and because she me.acc.cl you.dat.cl saved ‘And because she saved me for you.’ (Chariton, Callirhoe, 8.3.2.6)
(16) Thale:s me to:i medeunti Neileo: de:mou Thales me.acc.cl the watching-over (god).dat Nileos.gen state.gen dido:si gives ‘Thales devoted me to the god that protects the people of Nileos.’ (Callim. Th 52) .. Medieval CG ... Competing substitutes and the emergence of dative alternations Locative PPs eventually became the exclusive exponent of all inanimate expressions formerly realized as datives including inanimate goals/IOs. This is arguably due to the fact that the use of locative PPs was probably spread to such an extent that they became the exclusive way to realize locatives (both arguments and adjuncts), replacing all nonargumental uses of oblique cases, coupled with the fact that locative expressions most often have inanimate referents. Meanwhile, the morphological genitive had replaced the dative in argumental uses completely (Table 12.3). In other words, in MedCG genitive argumental DPs can only have [+animate] (or rather [+human]) referents, while goal PPs can only have inanimate/non-human referents. At this stage, then, the animacy restriction on IO DPs may be treated as a
Table .. DP/PP competition in MedCG
mGEN PP
x
x
Animate/human arguments of dative verbs [–animate]/[–human] arguments of dative Vs Locatives (both argumental and non-argumental)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
purely lexical property, perhaps a selectional requirement in the lexical entry of all ditransitive verbs. PP-realization of the IO is only chosen as a last resort for recipients/goals in the following cases: (a) when they are non-human, collective entities (17), (b) when other inanimate nouns are used metonymically in place of [+human] DPs (18) and, interestingly, in several cases where the DO is a 1st/2nd-person clitic, and the IO is 3rd-person (19), i.e. prepositional IOs also seem to be employed as a strategy of avoidance of configurations with PCC-violating combinations of person features, such as (16)—of which no equivalent is attested. (17)
Na to ksighunde is ton kosmon to it.acc.cl narrate.3pl to the people.acc ‘to narrate it to the people’
(Machairas, 2.99.5, 15th cent.)
(18)
oti to dhikon tou na dhothi [. . .] is ta cherja tous that the own his sbjv be-given.sbjv.3sg to the hands.acc the pateres tu San Tomeniku fathers.gen the Saint Dominique ‘that his fortune be given to the hands of the fathers (monks) of St. Dominique’ (Machairas 1.56.1–2)
(19)
kai rikoumantias’ mas is tin afentian tu and recommend/present.ipfv us.cl.acc to the majesty.acc his ‘and recommend/present us to his majesty’ (Machairas, §275.29)
Furthermore, the use of PPs, which were originally restricted to purely locative uses, is often also extended to predicates such as ‘send’, ‘return’, etc. (cf. (20)), which are ambiguous between the ‘caused change of location’ and the ‘caused change of possession’ reading (see Ormazabal and Romero 2010). It then seems reasonable to assume that it was this ambiguity that led to a gradual rise and spread/diffusion of prepositional animate IOs. So, in Cyprus’s Love Poems (21), a collection considered the first sample of (early) Modern Cypriot and compiled between 1560–70, i.e. one century later than the text where the data in (17)–(20) come from, we find uses of animate PP IOs in contexts where the chronicles of the fifteenth century would invariably use the genitive.6 (20) kai esteilan ton eis ton rigan and sent.3pl him.cl.acc to the king.acc ‘and they sent him to the king’ (21)
(Machairas, §360.11)
tote eis afton mou fernoun oi pikres egnais tin thlipsin then to self my bring.3pl the bitter worries the sorrow.acc ‘then the bitter worries bring sorrow to myself ’ (Love Poems, 84b)
6 Note that ferno ‘bring’ in () is one of the ambiguous predicates mentioned earlier, but the collocation ‘bring sorrows’ certainly cannot have a literal locative reading.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis Table .. DP/PP competition in Early Modern CG
animate/human arguments of dative Vs mGEN
[–animate]/[–human] arguments of dative Vs
x x
Locatives (argumental and non-argumental)
PP
Table .. The emergence of dative alternations
Genitive (DPs/pronouns) PPs
Animate IOs Inanimate IOs/locatives
Therefore, the distribution of prepositional and genitive substitutes of the ancient dative at this stage is as illustrated in Table 12.4. More specifically, as far as the syntax of dative/ditransitive verbs is concerned, the emerging pattern is as, shown in Table 12.5. This pattern is, more or less, the standard pattern observed in systems with dative alternations. Now that PP IOs can freely be either animate or inanimate, the animacy requirement concerning IO DPs cannot be regarded as a selectional feature any more. It rather looks like a licensing requirement of inherently Case-marked DPs, i.e. [(abstract) dative Case]→[+Animate/+human] but not vice versa. In other words, the corresponding feature is now optional and may not have to be part of the lexical entry of every dative verb. Instead, it would be reasonable to assume that it is reanalysed as a probe that licenses ‘dative’ DPs and that consequently ‘dative’ DPs need to be active, i.e. to bear an active [uCase] feature, in order to be able to establish an Agree relationship with such a probe. If this is the case, then this might explain why the properties we associated with ‘active’ inherent Case (see the data in the next subsection) emerge as soon as dative alternations become productive. Finally, as I already argued in Section 12.2, the additional probing head that we need to postulate for Greek (both Standard and Cypriot) is an Applicative bearing a [uParticipant] probe, which also derives the strong PCC. ... Coexisting/competing inherent [Case] features (a) Survival of [iCase] The properties [iCase], as in Hellenistic Greek, still survive in MedCG, although as we said the ancient dative has now been replaced by the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
genitive. Firstly, we can still find examples where datives cause no DIEs in raising, unaccusatives, and passives (22)–(24).7 [TP ti efanin-T (22) ekinon, [CP toi tis vulis mu [TP ti ’ which that seemed.3sg the senate/diet.gen my ine kalon]]], fenete mu ki emenan be.inf good seems me.gen.cl and me.gen ‘what seemed to my senators/consultants to be good seems to me (to be good) too’ (Boustr. Chron. A 52.13–15) (23) den areskun tus archondes tus Genuvisus [na ine i not appeal.3pl the masters-acc the Genoans-acc to be.3pl the las mas kai to dikon tus apokato is tin eksusian people-nom our and the fortune-nom their under to the power sas] your ‘The Genoan masters do not like the fact that our people and their fortunes are under your rule.’ (Machairas, 3, 372) ape tus Romeus ke proi (24) pos estrafin to rigatoni that was-returned.3sg the kingdom.nom from the Romans and was edothin tus Latinus ti given.3sg the Latins.acc/gen ‘that the kingdom was returned by the Greeks and was given to the Latins’ (Machairas, 2.99.1–2) Secondly, ‘low’ IO DPs are still available. In her corpus of MedCG, Vassiliou (2002) observes six occurrences of V-DO-IODP but no occurrences of V-IODP -DO in paragraph-initial position, which she considers as the most pragmatically neutral context. This is also consistent with the fact that when both internal arguments are existentially quantified, i.e. when none of them is focused or presupposed, 7 It should be pointed out that in () Spec-T in the relative clause is occupied by the trace of the embedded subject rather than by proexpl , since efanin here is clearly not used as impersonal, taking into consideration that its other occurrence (fenete), in the matrix clause, has an overt referential subject, arguably raised out of an elided complement-TP. Note that in MG, in which fenete appears to have the same usage, the embedded subject of an elided complement clause cannot be (A’)-moved (i.e. topicalized) into the matrix CP, if fenete does not agree with it, i.e. if it is used as impersonal. Consider, for instance, the following MG example, which is as close structurally to () as possible: (i) Ekines i lisis pu su [fenonde.p/??fenete.s] esena na ine kales, [mu fenonde.p/∗ fenete.s] ki emena (na ine kales) ‘Those solutions that seem to you to be good, (they) seem good to me as well’. At any rate, in Med CG there are also quite a few other instances of raising predicates agreeing with embedded nominative DPs (regardless of their surface position) across genitive experiencers unproblematically, without evidence of any intervention effect, e.g. without obligatory cliticization/CD (CD) of the genitive as in SMG (see below). Interestingly, cliticization/CD of the genitive is obligatory in SMG even when the raising predicate is impersonal, i.e. there still appears to be a need to establish some Agree relation with the embedded CP, for which the genitive would act as an intervener.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
V-DO-IODP is preferred (25). DO-IODP order is statistically prevalent anyway (around 70 of all ditransitive constructions, excluding strings with dislocated arguments) but, apart from these observations about linear order, there are also examples in which we need to assume that DO necessarily asymmetrically c-commands IO, see e.g. (26), which presents us with a case of variable binding of a possessor embedded in the IO DP by the DO DP (note that its equivalent in present-day (Cypriot) Greek would sound highly unnatural). (25) Kai afinei kanenan pragman katinos and leaves anything.acc anyone.gen ‘And (if) he leaves anything to anyone’
(Assises f137, 190)
(26) An thelete me to kalon na strepsete [to kastron]i [tou the if want.2pl with the good sbjv return the castle afendi toui ] owner.gen its ‘If you want to willingly return the castle to its owner’ (Machairas 3.472.10–11) Crucially, if ‘low’ IO DPs exist, i.e. if the configuration in (6) is still available, the absence of DIEs in theme passives (24) is simply due to the fact that the IO DP does not structurally intervene between T and DO—apart from the fact that, even if it did intervene, like in raising/psych unaccusatives, it would be transparent due to its [iCase]. (b) Reanalysis and emergence of active inherent (u)Case In the same texts as those containing the above data, especially the later ones, there is also considerable rise of the reverse, IODP -DODP order. This order might have started as the result of A’scrambling of IODP over DO, which began to be extensively used as a disambiguation strategy, given that V-DPACC -DPGEN can be ambiguous between V-DODP -IODP and V-[D [NP NACC POSSGEN ]], especially when the accusative is definite (27). It is important to note that at this stage, NPs in Greek are consistently head initial, with [D [NP N DPGEN ]] being the unmarked configuration. [[DPGENi ] D [NP N ti ]] also exists (but not [D [DPGEN N]], which was the canonical order in Classical Greek). However, this configuration is significantly less frequent, derived from [D [NP N DPGEN ]] (see Panagiotidis 2008), and highly marked. Thus, in principle, V-DPGEN -DPACC can also be ambiguous between V-IODP -DODP and V-[[POSSGENi ] D [NP N ti ]], but certainly not to the same extent as V-DPACC -DPGEN : while in the latter it can be said that the two readings are equally accessible, in the former one of the two readings is very marked and would usually require special intonation. (27) a. etaksen ta rigata tou rigos promised the provinces-of-the kingdom the king.gen ‘he promised the provinces belonging to the kingdom to/of the king’ – genitive ambiguous between ‘recipient’ and ‘possessor’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
b. I avli prepei na pari ta pragmata tous the court.nom must sbjv take.3sg the things.acc the egitades guarantors.acc/gen ‘The court must take the possessions of/from the guarantors.’ (Assises, 58,73) – genitive ambiguous between ‘source’ and ‘possessor’ c. Opou raftei ta rouxa tous ksenous he-who sews.3sg the clothes.acc the foreigners.acc/gen ‘The person who sews up the clothes of/for the foreigners’ (Assises 6,88) – genitive ambiguous between ‘benef.’ and ‘possessor’ Evidence that this permutation of the relative order of the internal arguments must have started as A’-scrambling comes from the fact that several IODP -DODP examples still behave as such, e.g. they are employed for defocusing purposes, cf. (28), where the DP ‘the Genoans’ is old information and as such it has to be vP-external. However, there is also evidence that ‘high’ IOs are already being reanalysed as occupying A-positions in texts such as the Assises (thirteenth–fourteenth century with manuscripts dating from the fifteenth century) and the Chronicle of Machairas (fifteenth century), see e.g. (29), an example of quantifier variable binding of DO by IODP . (28) kai anen kai pepsoun oi Genouvisoi [. . .] tote na and if and send.sbjv.3pl the Genoans [. . .] then sbjv dosoun tous Genouvisous 100 doukata give.sbjv.3pl the Genoans.acc/gen 100 ducats ‘and in case the Genoans send (someone). . . then they (must) give the Genoans 100 ducats’ (Machairas, §353.17) (29) kai edoken pasanou tin douleian tou and gave.3sg everyone.gen the job.acc his ‘and (he) gave everyonei hisi job’
(Machairas, §174.7)
It can be argued then that the trigger of the reanalysis proposed is the synergy of two facts, the rise of V-IODP -DODP and the rise of [+animate] IO PPs, which both gave rise to what is canonically found in languages with dative alternations: prepositional constructions with DO>IOPP and DOCs with IODP >DO. While IODP -DO began as an A’-scrambled order, it soon lost its discourse-related effects, particularly as soon as it did not only serve information-structural (i.e. defocusing) purposes and it started being used extensively purely as a disambiguation strategy. This loss of its interpretive properties meant that the acquirers could no longer attribute to v∗ an optional EPP/edge-feature (cf. the discussion in Roberts 2007: 275–7) and the latter had to be reinterpreted as an obligatory movement-triggering feature, not necessarily linked to v∗ . So, giving the impression that it can be the alternant of DO>IOPP , IODP -DO was reanalysed as such and v∗ ’s optional EF was reinterpreted as the EPP
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
feature associated with Appl’s [uParticipant] probe, in a more articulate structure, such as (7). Therefore, this is how the high position of scrambled IOs was reanalysed as the position in which the genitive/dative DP can check and delete the structural part of an AIC feature. It is clear that, despite this change, the dative did not become a fully structural Case, given for example the fact that it did not become absorbable in passives etc. Valuation of its hybrid [uCase] takes place via Agree with v∗ , i.e. it can only take place in the moved position, so that the intervention of DO is avoided (which is probably why Agree+Move triggered by Appl never changed into Agree only, as in the cases discussed in Roberts 2007: 277). There are two possible scenarios about the steps in which this reanalysis might have taken place. (a) According to the first scenario, the [uParticipant] feature, which probes and attracts dative DPs, was first assigned to v∗ , which often hosted scrambled IOs in its edge. Raising to this position was then reinterpreted as a result of Agree for this [uParticipant] feature, i.e. as A-movement. The stages of the reanalysis, then, are as follows: Step I. A’-scrambling, (30)→(31). (30) [. . . V-v-T [v∗P EA [DO[uCase] IOgen[iCase] ]]] (31)
[. . . V-v-T [v∗P EA [v∗P IOgen[iCase] [DO
]]]]
Step II. A’-movement is reanalysed as Agree-based/A-movement, (31)→(32): [uϕ] Agrees with DO, and [uParticipant] Agrees with (and attracts) IO. (32) [vP EA [vP IO [uCase]
v*
[VP
DO V
]]]
[uφ] [uPrt]
Step III. (32) is reanalysed as a more articulate structure ((33)=(7)), preferable to (27) probably because it involves fewer feature syncretisms (cf. Roberts and Roussou’s 2003 ‘simplicity metric’). (33)
[vP EA v* [ApplP IOgen Appl [v2P v2 [VP DO V ]]]] [uφ]
[uCase]
[uPrt]
[uφ]
The hypothetical structure in (32) may be reflected in the (relatively few) clitic clusters found in the earliest medieval text (Assises), where the order is still unfixed (i.e. both IO-DO and DO-IO clusters are attested, cf. (34)): we may derive the free ordering of the two object clitics, on the assumption that, when DO undergoes cliticmovement to an inner Spec of v, it may land either above or below IO (cf. Richards 1999).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
(34) a. oti to tou afikan ekino to zitai that it.cl.acc him.cl.gen left.3pl that which asks.3sg ‘that they left him what he asks’ (Assises, f134,188) b. Apai ta perpyra κ’ ta sou from/as for the perpers 20 them.neut.cl.acc you.sg.cl.gen eparadoka handed-in.1sg ‘As for the perpers (= local currency) (that I owed to you), I did give you 20.’ (Assises, f74,103) c. oti ekeinos to tou epoulisen ekeinon to alogon that he.nom it.cl.acc him.cl.gen sold that the horse ‘that he sold him that horse’ (Assises, f191.30) d. oti eteros tou to epoulisen that someone-else him.cl.gen it.cl.acc sold.3sg ‘that someone else sold it to him’
(Assises, f191.32)
(b) The alternative scenario would be that (32) never really existed and that acquirers by default prefer (33) over (32), due to Roberts and Roussou’s ‘simplicity metric’. The uncertainty, then, in the ordering of clitic clusters would be attributable to the fact that at this stage there are still datives with both active and non-active Case features, and that there is some tendency to employ a more articulate structure, with two ϕ-probes/v-heads, i.e. two potential targets of clitic movement, even when the IO clitic does not have [uCase], i.e. when no dative-shift is required, thus resulting in clitic-movement of both internal arguments from their first-merged positions, in an order-preserving fashion (35). Note that in Hellenistic Greek, the order of clustering weak pronominal forms/clitics was invariably ACC-DAT, i.e. DOIO. But it might not be quite safe to assume that DO-IO clusters in MedCG are the continuation of this pattern, as clitic clusters in Hellenistic Greek tended to be second-position (see 15b), and their derivation probably involved different targets of clitic-movement. (35)
[vP EA v∗
[uϕ]
[v2P v2
[uϕ]
[VP DO V IO]]]
[iCase]
... Modern CG In Modern CG, all goal arguments bear AIC and they exhibit all the syntactic properties that this was assumed to entail: (i) they always undergo dative-shift (36), yielding IOgen >DO, (ii) they display the strong version of the PCC (37), (iii) in theme passives (38), as well as with motion unaccusatives (39), goal arguments structurally intervene between T and the nominative theme and their presence there is signalled by the blocking effects induced by undoubled genitive/dative DPs in that position.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
ton andran tisi (36) a. Edhiksa kathe jenekasi showed.1sg each woman.gen the man.acc her ‘I showed every woman her husband.’ tui b. ∗ Edhiksa kathe andrani tis jenekas showed.1sg each man.acc the woman.gen his ‘I showed every man to his wife.’ ∗ tu/∗ tis (37) Edhoken me Marias gave.3sg me.cl.acc him.cl.gen/the Mary.gen ‘(S)he gave me to him/to Mary.’
(38) To vivlion en ∗ (tis) epistrafiken tis Marias the book.nom not her.cl.gen was-returned.3sg the Mary.gen ‘The book was not returned to Mary.’ ∗ (-tis) irte (39) [to epidoman]i tis Marias ti the allowance.nom came.3sg her.gen.cl the Maria.gen ‘The allowance came to Mary.’
More specifically, following what we said in (3ii), the intervention of a genitive/dative DP with AIC in passive/unaccusative contexts is an instance of defective intervention: while in active contexts we assumed that v∗ matches the raised IO DP, in DAT-above-NOM contexts it is T that matches the ‘dative’ DP’s ϕ-features and values/deletes its [uCase]. Such Agree blocks further probing of T, leaving the theme’s [uCase] feature unvalued and thus failing the derivation. But if the ‘dative’ cliticizes or undergoes CD, then the clitic will head the chain of the ‘dative’ argument. This saves the derivation in two ways: (a) the clitic, incorporated into T, cannot value T’s bundle of features: T matches all of the clitic’s features, but the clitic, being a ϕP (Cardinaletti and Starke 1999; Déchaine and Wiltschko 2002; Roberts 2010), cannot match and value T’s [uD], forcing T to probe further down; (b) even if there is a full ‘dative’ DP doubling the clitic, it must now be ignored, because the head of the relevant chain is outside T’s complement domain, thus cancelling its DIE (following again Chomsky 2001): so, T is free to Agree with the low nominative theme (40),8 as actually observed. This is why CD of genitive/dative DPs in these contexts is now obligatory, while it was optional in MedCG. (40) [(themenom) ∗ (Cli )-T [vP (v) [ dativeDPi[uCase] ] (Appl) [ V ]]]
8 See Anagnostopoulou () for an alternative account of the repairing effects of cliticization/CD, based on the notion of equidistance of elements in the same minimal domain, which however cannot account for the diachronic variation discussed here.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The evolution of inherent Case in the diachrony of Greek
Interestingly, however, genitive/dative experiencer DPs in raising contexts (41–42), as well as with piacere-type psych unaccusatives (43), do not give rise to any intervention effects. (41) Efanin(-tis) tis Marias (oti eton) eksipnos seemed.3sg-her.gen.cl the Mary.gen (that was.3sg) smart.masc ‘He seemed to Mary (to be) smart.’ (42) Ta mora en (tis) fenonde tis Marias (oti/na en) the kids not her.gen.cl seem.3pl the Mary.gen (that/to be.3pl) kurazmena tired ‘The kids do not seem to Mary to be tired.’ (43) O Janis areski tis Marias polla the John.nom appeals the Mary.gen much ‘Mary likes John a lot.’ The equivalents of these in SMG would require cliticization/CD of the genitive/dative DP, as Anagnostopoulou (1999) observed, again as a means of cancelling its DIE (44–45). tis Marias (na imaste) kurazmeni. (44) Dhen ?∗ (tis) fenomaste her.gen.cl-seem.1pl the Mary.gen (to be.1.pl) tired.masc.pl not ‘We do not seem to Mary (to be) tired.’ (45) O Janis ?∗ (tis) aresi tis Marias poli. the John.nom appeals the Mary.gen much ‘Mary likes John a lot.’ Furthermore, as Anagnostopoulou has also shown, such datives in SMG may also display subject-like behaviour, i.e. they may undergo A-movement as quirky subjects. aresi o Janisj ala proi/∗j sichenete (46) Tis Mariasi tisi detests the Mary.gen her.gen.cl appeal-to.3sg the John but pro tin Katerina the Katerina.acc ‘Mary likes John but hates Katerina.’ (from Anagnostopoulou 1999: 71) This point of divergence between Cypriot and SMG indicates that the emergence of AIC in CG was probably the result of a lexical diffusion that only affected predicates encoding ‘caused change of location’ and ‘caused change of possession’, but not (yet) psych-predicates, while in SMG, which presumably underwent a similar path of changes, the change was diffused to all verbs taking dative arguments. Furthermore, it seems reasonable, precisely because of this selectivity of the spread in terms of
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Dimitris Michelioudakis
theta-roles in CG, to assume that the whole gradual change, and in particular the coexistence of the two patterns in MedCG, was the result of lexical diffusion rather than a change attributable to some competing grammars scenario, e.g. along the lines of Kroch (1989, 1994, and subsequent work), which should have resulted in a homogeneous state of affairs in all possible contexts in all Greek varieties.
. Conclusions To conclude, the evidence discussed in the previous section corroborates the postulation of different types of inherent Case, differing in their visibility to Agree (and Move), as has also been common in synchronic comparative research. More particularly, it was shown how the loss of oblique case morphology did not result in the loss of the corresponding abstract Case feature and how instead the actual syntactic change was triggered by a syntactically motivated reanalysis, merely facilitated by the morphological change. Finally, the diachronic data presented are consistent with, or even point towards, a movement analysis of dative alternations.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Part III Syntax, Prosody, and Information Structure
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe V I R G I N IA H I L L
. Introduction This chapter focuses on Old Romanian (OR) constructions where pe is used for Differential Object Marking, as in (1). The data come from three Moldavian chronicles (Letopiseţe) attesting to the literary language of the seventeenth–eighteenth centuries. The term Differential Object Marking (henceforth, DOM) has been coined by Bossong (1985) to indicate the process of inserting a particle in front of the direct object in a variety of languages, including Romance and Balkan languages.1 (1)
au adus Ştefan vodă pre Mariia din Mangop has brought Stefan king dom Maria from Mangop ‘King Stefan brought Maria of Mangop’
(Ureche 97)
Traditional and formal linguistics consider DOM particles (including pe) as Caseassigning prepositions that may also involve discourse effects (de Hoop and Naransimhan 2005 a.o.). However, current studies on DOM bring arguments to define this operation as a strategy for encoding information structure in syntax rather than for Case marking (Kwon and Zribi-Hertz 2008 for Korean; Leonetti 2004 for Spanish a.o.). In this respect, while still considering pe-DOM as a Case-marking device, von Heusinger and Onea (2008: 104–5) suggest that pe-marked nouns in Romanian are anchored to the speaker, yielding a familiar reading. On the basis of contexts as in (1), I argue that pe-DOM is not a Case marker (cf. Kayne 1975, 2001) but a pragmatic marker, and that its discourse value shifted from contrastive topic to backgrounding topic in the presence of clitic doubling. Theoretically, this chapter supports the exclusive information structure definition of
1 For an exhaustive typological discussion of DOM see Comrie (, ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
DOM, and points out diachronic changes that may affect the information value of the DOM marker. The researched texts are three Letopiseţe ‘chronicles’ of Moldavian history, which provide the first literary texts in Romanian.2 The chronicles follow each other chronologically, in subject matter and in production dates: Grigore Ureche wrote his chronicle between 1642 and 1647; Miron Costin published his in 1675; Ion Neculce began writing his text after 1732. The texts used here are edited by academicians Iordan (1955) and Panaitescu (1958, 1979). I scanned and converted these editions to searchable texts. Data from OR documents, from the same or the preceding centuries, will be used for comparisons and verification as needed.3
. Research context Historical studies of OR (Frâncu 2009; Gheţie and Mareş 1985; Todi 2001) indicate that, at the time of the first written documents (middle of the sixteenth century) and for about a century thereafter, direct objects were treated in various ways, even within the same document. That is, the direct object, morphologically unmarked for Case (except for the pronominal paradigm), occurs either by itself or with two marking devices: clitic doubling (CD) or/and DOM. Examples (2a,c,d) come from Frâncu (2009: 173–4); (2b) from Tasmowski (2008: 15). (2) a. prinserã Pavel caught Pavel ‘they caught Pavel’
–CD; –DOM (Coresi, Praxiu [Br 1563]: 104)
b. se me treacã mine acestu paharu to me pass me this cup ‘let this cup pass me’
+CD; –DOM
c. rogaţi pre Domnedzeu pray.imp.2pl dom God ‘Pray God.’
–CD; +DOM
d. sˇa te=vinzi pre sine sbjv you.acc=sell dom yourself ‘to sell yourself ’
+CD; +DOM
(Evang. [SB] Matei 26:39, 42)
(Prav. I, 7v ms.rom.BAR #5211)
(Varlaam, Leastviţa: 61v )
Based on frequency, Frâncu (2009) concludes that the CD/DOM combination in (2d) is very rare in the first century of written language; however, by the middle of the
2 Written Romanian is first attested through official documents and letters of the fifteenth century, and through translations of religious texts from Slavonic from the fifteenth century onwards. 3 The comparative data include the fragments of Wallachian chronicles, such as discussed in Todi ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
eighteenth century—which is the time of Moldavian chronicles—the patterns in (2a, 2b) become rare, whereas the ones in (2c, 2d) are overwhelming. Only the pattern in (2d) is productive in MR. This situation invites a series of questions: How did CD and DOM function separately? Why did they merge? Are DOM-ed nouns Case marked while the others (e.g. (2a, 2b)) are not? Historical linguistics explains the emergence of DOM as a replacement to the Accusative Case ending from Latin (von Heusinger and Onea 2008). Thus, in languages with flexible word order (such as OR), ‘differentiating’ the object from the subject through DOM made it possible for the hearer to correctly parse the agent and the theme, especially when both nouns were [+human] (Frâncu 2009: 173).4 That is why most nouns displaying DOM belong to semantic classes with a [+human] feature. Along these lines, the research focus fell on the diachronic shift between one or another noun class that undergoes DOM (von Heusinger and Onea 2008). Such approaches justify the co-occurrence of CD with DOM on semantic grounds: this correlation depends on the degree of referential stability in a noun: persons’ names undergo DOM and CD, whereas [+human] bare quantifiers (which have no referential stability) undergo only DOM (Farkas and von Heusinger 2003). Although these studies provide an accurate description of the diachronic changes and spread of DOM, we still need to understand why CD and DOM could operate separately; why, as a preposition, only pe may Case mark the direct object; and, if DOM and CD are so dependent on each other, why did it take over 300 years for them to blend and for this blend to efface the other uses? Formal linguistics is not ready to answer these questions. The most complete account to date comes in Kayne (1975, 2001), and it treats CD and DOM as inseparable. The analysis has been taken for granted with respect to MR (e.g. Cornilescu and Dobrovie-Sorin 2008: 304 a.o.). Basically, the hypothesis is that the doubling clitic in (2d) absorbs the structural Case of the verb, so a preposition is needed (e.g. peDOM) to license the noun in syntax. This view runs into empirical problems not only because of constructions as in (2a, 2b, 2c), which coexisted at the same time, but also because it implies that DOM-ed nouns are interpretively equivalent to unmarked direct objects—which is not the case, as we shall see later. Given this research context, I investigate the behaviour of pe-DOM in the Moldavian chronicles, because these texts display a stable and systematic use of DOM, and this use attests to the midway (i.e. (1)/(2c)) diachronic change to MR. This investigation aims to bring the formal approaches in line with the empirical evidence from OR, and
4 Other languages solve the same problem by marking the subject, or by marking both with different particles (Kornfilt ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
that is why the first task is to show that pe-DOM is dissociated from Case marking. Then I can move the analysis of pe-DOM into the domain of discourse-driven syntax. The data to be investigated is divided according to two patterns: a pattern where a noun is optionally pe-DOM-ed; and a pattern where a noun has obligatory pe-DOM and CD. The main argument is that the same structural configuration underlies the two patterns; that is, a DP with a left periphery for encoding the information structure (TopP>FocP). However, the value of Top in this field is set to contrastive topic when only DOM applies; and to background topic when DOM and CD co-occur. MR has lost the pattern for optional DOM without CD.
. Pe-DOM is the outcome of grammaticalization All the analyses of pe-DOM agree that this is a functional element that lost the lexical semantics of the preposition pe (Mardale 2006 a.o.). In this section, I argue that pe underwent grammaticalization at the same time with semantic bleaching, because of its lexical underspecification; that is why pe, rather than other prepositions, became reanalysed for the purpose of DOM. The analysis capitalizes on the tendency of the OR preposition pe to be reanalysed for the purpose of both derivational and inflectional morphology. Pe is used as a regular preposition in both OR and MR. However, the meaning of this preposition varied considerably in OR; for example, the data in (3) come from the same page in Ureche, Panaitescu (1958: 66). So even in its prepositional use, pe did not have a well-specified meaning in the OR lexicon. (3)
a. chematu-o-au [. . .] pre numele hatmanului called-her-has after name.the governor.the-gen ‘he has named it after the governor’s name’ b. războiu cu şciţii pre acéste locuri war.the with Scythes in these parts ‘the war with the Scythes in these parts’ c. pre carei oi au chiemat-oi Molda dom which her has called=her Molda ‘whom he called Molda’
The vague semantics of pe gave rise to various reanalyses, which may be grouped in two directions. One direction preserved pe as a preposition, but its semantic field has been narrowed to a subset of the meanings possible in OR; for example, the use of pe in (3a) is recognized by the modern speaker but it is considered archaic. The other direction has pe reanalysed as a bound morpheme for the purpose of derivational and inflectional morphology. In derivational morphology, pe becomes a
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
prefix in the derivation of adverbs or adverbial expressions. This can be seen in (4). Adverbial PPs, as in (4a), based on the use of pe as a preposition, have completely disappeared from the language, to the extent that the unadvised reader could not interpret the constituent. In many such contexts, pe becomes a prefix for deriving fullfledged adverbs, as in (4b, 4c). (4) a. să spuie aşa pre cale tot pre rîndu to say so by way everything by turn ‘to say everything rightly and orderly’
(Ureche 64)
b. am socotit pre semne ce arată > pesemne = ‘probably’ have-1sg thought from signs that show ‘I thought, by the sound of it, . . .’ (Ureche 69) c. au rămas pre urma lui la domnie > pe urmă = ‘afterwards’ has stayed after passage.the his at power (Ureche 89) ‘he became king after his (other king’s) death’ These examples from derivational morphology show that pe completely loses its prepositional properties and even its free morpheme status. The reanalysis of pe for the purpose of DOM is seen in the same vein. The main evidence for the non-prepositional status of pe in DOM environments is that it contradicts the expected selectional restrictions in prepositions. More precisely, Mardale (2006) shows that Romanian prepositions (in OR and MR) select only NP, thus contrasting with the P>DP selection in related languages. The use of the preposition pe in the Moldavian chronicles obeys this restriction, as shown in (5), with no exception to be found. (5)
blăstămul săracilor, cum să dzice, nu cade pre copaci(∗ i) curse.the poor.the.gen as refl say not falls on trees(the) ‘As they say, the course of the poor does not fall on the trees.’ (Costin 39)
However, pe-DOM is not bound by this rule: it can be followed by DP, as in (6), and the DP may belong to various semantic classes. (6) a. Schimbasă împărăţiia pre veziriul în scîrbă . . . changed emperor.the dom governor.the in disappointment ‘The emperor changed the governor out of disappointment.’ (Costin 57) b. . . . şi zburdăciunea naşte păcatul şi pre păcatul urmadză and playfulness gives.birth sin.the and dom sin.the follows mînia lui Dumnedzău. wrath.the of God ‘The playfulness gives birth to sin and sin is followed by God’s wrath.’ (Costin 97)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill c. . . . şi pre Radul-vodă avè iară mare nepriietin. and dom Radu.the-king had again big non.friend ‘. . . and he had King Radu again as big enemy.’
(Costin 60)
d. Ştefan vodă au pus muntenilor domnu pre Vladul vodă . . . Stefan king has put Vallahs.the king dom Vlad.the king ‘And king Stefan made Vlad the king of the Vallahs.’ (Ureche 105, 29v) e. Traian întîiu, împăratul, supuindu pre dahii, . . . Trajan first king.the subjugating dom Dacians.the ‘First, Trajan subjugating the Dacians . . .’
(Costin 11)
Therefore, pe in (6) suffered not only semantic bleaching, but it also lost the properties that made it behave as a preposition elsewhere (e.g. (5)). Thus, pe becomes the exclusive candidate for DOM, since other prepositions did not attain and retain this degree of grammaticalization. Indeed, the documents attest other candidates for DOM in the language; that is, spre ‘toward’ and (ca) de ‘as of ’ (Todi 2001: 62).5 However, the grammaticalization of spre was not successful, the preposition preserving the same strength in MR as in OR. On the other hand, de has become a determiner in relation to nouns, so its extension to another nominal function was short-lived. The view provided in this section yields the conclusion that lexical underspecification led to the semantic bleaching of the preposition pe, which ended up with the reanalysis and the recategorization of this element in various directions, one of which is that of a DOM particle. Some studies, however, do not agree with this view (e.g. Mardale 2006 a.o.) by considering that semantic bleaching did not affect the prepositional status of pe. For this reason, the next section must provide further proof that pe-DOM is not a Case-marking device.
. pe-DOM versus Case marking OR texts allow us to do an exercise that is impossible on the basis of MR data alone: that is, we can analyse DOM separately from CD. This separation is paramount for pointing out that pe-DOM occurs independently of Case marking. The Moldavian chronicles display classes of animate or inanimate nouns that undergo DOM on an optional basis and do not involve CD; for example, ethnic names (‘Tartars’, ‘Poles’), nouns denoting collectivities (‘army’, ‘camp’, ‘kingship’), concrete nouns (‘chronicle’), etc. The syntax of these nouns brings clear evidence that direct objects do not need pe-DOM for Case. The arguments are as follows: 5 Frâncu (: ) suspects that these choices have been semantically influenced by the Slavic counterpart na in translated texts. However, this hypothesis works for texts, not for spoken language, where there is no evidence of bilingualism.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
(i) The same noun occurs with or without pe-DOM in the same text, and in the same position (i.e. postverbal), as in (7). (7) a. . . . au gonit pre tătari . . . has chases dom Tartars-the ‘he chased the Tartars’ b. . . . de va putea opri tătarii . . . if will can stop Tartars-the ‘if he’ll be able to stop the Tartars’
(Ureche 68)
(Ureche 183)
c. . . . au adus tătari din Crîm . . . has brought Tartars from Crimea ‘He brought Tartars from Crimea’
(Neculce 105)
d. . . . i=au tăiat pe nişte tătari . . . them=have slain dom some Tartars ‘they have slain some Tartars’
(Neculce 183)
(ii) The eighteenth century, when pe-DOM flourished, is too far away from the loss of the Latin Accusative Case ending to justify the variation in (7) as transitional to an analytical system. (iii) The optionality of pe-DOM in situ is further reflected in the form of the fronted direct object. For example, Ureche uses ‘Tartars’ with or without pe-DOM in situ, as in (7a, 7b), but without pe-DOM in fronting, as in (8a). The same variation applies to the use of ‘Hungarians’, in (8b, 8c, 8d). (8) a. s-au apucat [de tătari a-i bate] refl-have started of Tartars to=them attack.inf ‘he started to attack the Tartars’
(Ureche 68)
b. au bătut pre unguri has beaten dom Hungarians ‘he beat the Hungarians’
(Ureche 100)
şi au fugit c. au lăsat ungurii has left Hungarians.the and has run ‘he left the Hungarians and ran’
(Ureche 118)
d. că ungurii lui Dispot încă=i plecase Tomşa that Hungarians.the of Dispot already=them leaned Tomsa spre sine. towards himself ‘because Tomsa had already corrupted the Hungarians in Despot’s troups’ (Ureche 183) Benincà (2001) points out an important correlation between the Case form of the noun and the type of targeted position under fronting: nouns fronted
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
lower than the head Force in Rizzi (1997) (i.e. within or lower than the CP field) must carry their Case marker, be it an inflectional morpheme or a preposition; nouns fronted higher than Force (i.e. Hanging Topics) meet this requirement on an optional basis. Along these lines, the fronted objects in (8a, 8d) are lower than Force, being preceded by subordinating conjunctions; as such, they are expected to display their Case marker, if there were one. Lack of pe-DOM on these fronted nouns means that pe-DOM does not count for Case. (iv) The first and second person singular in the pronominal paradigm have morphological marking for Accusative. There is no reason to think that these forms were frozen (i.e. Case-less) at the time of pe-DOM emergence, because they occur independently, pe-DOM being optional. For example, in (9) the alternation occurs in the same sentence; in (10), borrowed from Frâncu (2009: 173), the alternation occurs in the same text. (9)
cine va prïimi pre voi, mïne prïmúşte, şi cine pre mïne who will receive dom you me receives and who dom me prïmúşte, . . . receives (from Hill and Tasmowski 2008: 155) ‘who receives you receives me, and he who receives me . . .’
(10) a. . . . puse el domn casei . . . put him lord house.the.gen ‘he made him lord of the house’ b. . . . puse pre el domn casei . . . put dom him lord house.the.gen ‘he made him lord of the house’ Pronouns were identified as the category that underwent the fastest change to obligatory (CD)/DOM (Gheţie and Mareş 1985). For that reason, the examples above come from older documents, the chronicles displaying systematic CD/DOM on pronouns. It is, therefore, puzzling that the category most affected by DOM is also the only category that preserves some inflection for Case. (v) De Jong (1996: 131) shows, on the basis of MR, that pe-DOM does not interfere with the binding of anaphoric sine ‘himself/herself ’, as in (11a), whereas prepositions do, as in (11b). (11)
a. Ion se iubeşte pe sine. Ion refl loves dom himself ‘Ion loves himself.’ construit casa pentru sine. b. ∗ Ion şi-a Ion refl-has built house.the for himself
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
There is no evidence to the contrary in OR texts, so the negative evidence supports the results for MR. (vi) With respect to the semantic classes of nouns that display pe-DOM, Frâncu (2009) notices considerable shifts, even within the classes with the [+human] feature; for example, oaste ‘army’ loses pe-DOM in MR (12). These shifts are possible because the nouns do not depend on pe-DOM to be licensed for Case. (12)
a. să gonească pre oştile lui Vasilie-vodă to chase dom armies.the of Vasilie-king ‘to chase king Vasilie’s armies’ (∗ pre)
b. să gonească dom to chase
(Costin 90)
oştile armies.the
MR
The arguments in (i) to (vi) support the view that pe-DOM is not a preposition, nor a Case-assigning device. Therefore, pe needs categorial redefinition in DOM environments.
. Discourse-driven grammaticalization This section proposes that pe-DOM arises from the grammaticalization of the preposition pe as a marker for the information structure at the left periphery of the DP. The analysis capitalizes on the optional use of pe-DOM in OR. The dissociation between pe-DOM and Case assignment removes the morphosyntactic justification for this operation. Linguists aware of this dissociation in OR suggested that pe-DOM is a device for thematic encoding, distinguishing between agent and theme when both nominal arguments are [+human] (e.g. Frâncu 2009); such distinction would also allow for the fronting of the theme to the left periphery of clauses. However, word order could solve the thematic role distinction, and, as we have seen in (8), the direct object could be fronted without pe-DOM anyway. In fact, Tasmowski (2008) argues convincingly that, if pe-DOM has an effect on thematic interpretation, it is the reverse: it brings the direct object up on the agentivity axis, from theme to beneficiary or experiencer. This effect is confirmed in chronicles, where pe-DOM applies not only to direct objects, enhancing the role the person has played in bringing about the event, but also to indirect objects in need of similar emphasis, as in (13). (13) pre Grigorie-vodăk , de bucurie dom Grigore-king of delight au vinit [. . .] la dînsul, ik = to=him has come to him Muntenească să fie domnu. Muntenească to be king
mare great dat given
ce avè viziriul că that had vizir.the that iară domnia în again kingdom.the in
au has Ţara Ţara
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill ‘And to king Grigore, the vizir granted him again the kingdom of Ţara Muntenească, out of great delight for the fact that this king Grigore chose to show obedience to him.’ (Neculce 139)
Although the involvement of thematic role semantics is an important feature in the interpretation of objects with optional pe-DOM, it is, however, not systematic (e.g. it does not apply to inanimate direct objects). On the other hand, the emphasis or the ‘foregrounding’ intrinsic to such interpretation is pervasive every time the writer uses pe-DOM. Consider (14) and (15): (14) au scris acest letopiseţ . . . has written this chronicle ‘he has written this chronicle’ (15)
(Ureche 64)
fraţilor cetitorilor, cu cît veţi îndemna a ceti pre brothers.the.voc readers.the.voc as much will endeavour to read dom acest letopisaţu mai mult, . . . this chronicle more much ‘readers – my brothers, more you endeavour to read this chronicle over and over, . . .’ (Neculce 104)
In (14) and (15) the noun letopisaţu ‘chronicle’ occurs as a direct object in situ and it has a demonstrative adjective as modifier. The usage difference consists in pe-DOM applying only in (15); this option foregrounds the noun. Thus, the clause in (14) has the interpretation of sentence focus (e.g. as the answer to ‘What happened?’), and the direct object is only a part of the answer. On the other hand, the pe-DOM-ed noun in (15) steals the light, and the utterance is rather an answer to the question ‘What is the point/the news of your speech?’ or ‘What should we read?’ So the difference concerns the discourse pragmatics, and it arises systematically when a noun has the option of +/– pe-DOM. For example, the data presented in (7) also reflect the same switch in readings. Therefore, pe-DOM contributes pragmatic interpretation to the compositional meaning of the utterance, by foregrounding the marked direct object as the carrier of most relevant point(s) of the predication. Formal theories concerned with the encoding of information structure in syntactic configurations propose the term contrastive topic (CT) to grasp the interpretive effects presented for (14) and (15). In line with Lambrecht (1994),6 Molnár (1998) argues that the question ‘What’s the news?’ is a good test to detect CTs. The answer to this question allows for the highlighting of participants in the event, or for listing readings, without involving alternative sets. Examples with pe-DOM fit this description, as in (16) and (17). 6 In Lambrecht (: ) Contrastive Topics provide clarification when several options are possible; for example, ‘I saw MARY yesterday. She says HELLO.’ CTs also allow for listing readings, as in ‘I saw MARY and JOHN yesterday. SHE says HELLO, but HE’s still ANGRY at you.’ This type of topic is different from the contrastive focus, which involves an alternative reading.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
(16) au pusu pe Oprea armaşul logofăt mare şi pre Varticu has made dom Oprea officer.the commander big and dom Vartic vornic mare de Tara de Giosu şi pe Ghiorghie vornic mare governor big of County of Low and dom Gheorghe governor big de Tara de Sus, pe Zota postelnic mare şi pe Alexe of County of Upper dom Zota supervisor big and dom Alexe stolnicu mare şi pre Iane Călugărul visternic mare . . . provisioner big and dom Ian monk.the treasurer big ‘He assigned the officer Oprea as the high commander; Vartic as the governor general of the Low Country; Gheorghe as the governor general of the Upper County; Zota as major administrator; and Alexe as high provisioning officer; and Ian the monk as chief treasurer . . .’ (Ureche 224) (17)
a. Alegeţi=vă un domnu dintre voi chose.2pl.imp=you.dat a king among you ‘Chose a king from your ranks’
(Neculce 126)
trimite împărăţia pe un domnu în Moldova b. păn-a until=will send empire-the dom a king in Moldova ‘until the Emperor will send a king in Moldova’ (Neculce 306) In (16), the news is that a series of promotions took place, and the listing indicates who was assigned what rank. The contrastive topic reading emerges from pe-DOM-ed nouns that stay in situ, with no evidence of a relation with the clausal CT. In fact, if we front the DOM-ed nouns, they become aboutness topics, rather than contrastive; this is theoretically predictable from a classification of topics as in Kiss (2008b). In (17), the direct object is ‘a king’, which has pe-DOM in (17b) but not in (17a). The difference is that in (17a) the verb delivers the new information, with no further information structure, whereas in (17b) the new information situates the phrase ‘a king’ in the spotlight, as the main justification for the event. Again, this reading does not depend on a clausal CT. Accordingly, we can say that pe-DOM marks CT in situ in the relevant constructions. That is, the preposition pe has been reanalysed as a Topic marker at the edge of the nominal phrase. The problem is that syntactic theory provides a clear mapping of topics (as TopP) at the left periphery of clauses, but not at the left periphery of DPs, where pe occurs. Hence, this analysis follows the path of current studies that attempt to extend the articulation of the functional field of the DP.
. Formalization of pe-DOM as CT The analysis proposed in this section represents the left periphery of DPs as articulated over TopP>FocusP, pe being inserted in Top. Two theoretical hypotheses support this
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
analysis: (i) DPs are the syntactic equivalent of CPs; and (ii) adjective movement inside DP is movement to Focus. The first hypothesis has its source in Szabolcsi’s (1994) study arguing for a detailed parallelism between the structures of DPs and CPs, with respect to inflection, possessor extraction, and articles as complementizers. If this parallelism is complete, then DPs must also allow for the same functional articulation of the information structure as CPs do at their left periphery. This theoretical prediction has been validated in Aboh (2004), Aboh et al. (2010), and Haegeman (2004). For example, strong evidence for Topic and Focus phrases within DP comes from specific lexical marking that occurs on Gungbe nouns (Aboh 2004). This analysis found wide cross-linguistic confirmation in European languages (e.g. data in Cardinaletti 1998 for Italian; Siewierska and Uhlířová 1998 for Polish; Villalba 2006 for Romance and Germanic, a.o.), where movement inside DP has been shown to correlate with focus readings. Villalba (2006) includes Romanian in the discussion, showing similar patterns for focus movement within DP in constructions such as in (18), based on doctor idiot ‘idiot doctor’. (18)
a. idiotul de doctor idiot.the-foc of doctor b. los idiotas de médicos the idiots-foc of doctors ‘those idiots of doctors’
Romanian Spanish
Noun phrases in OR show the same possibility: the unmarked word order is N>AP, as in (19a), but a focus reading on the adjective reverses the order, as in (19b). (19)
a. sau întorsu spre Iaşi, la [domnu tînar]. refl have returned to Iassy to king young ‘He returned to Iassy, to the young king.’ b. [Scîrnavă şi nebunească faptă] şi nu fără osîndă . . . disgusting and crazy deed and not without punishment ‘Disgusting and crazy deed that did not go unpunished’. (Costin 123)
On the basis of this theoretical background, I propose to include pe-DOM as another means of encoding the information structure at the left periphery of DP, as in (20). (20) [DPinformation TopP(pe) > FocusP(AP) [DPinflection D. . . ]] In this analysis, the high left periphery of the DP contains a field for information structure, on a par with CP, and within this field Topic and Focus are encoded in the same hierarchical order.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
TopP in DP stands for either backgrounding (familiar) or foregrounding (contrastive) readings. Further research on the information structure in DPs may detail the realization of these topic readings along the same topic hierarchy proposed for clause structure in Frascarelli and Hinterhölzl (2007). In the OR constructions discussed so far (i.e. nouns with optional pe-DOM), the value of Top is [+contrastive] only. Along these lines, the preposition pe has been reanalysed as the highest head of its DP complement. The hierarchy in (20) predicts that pe-DOM precedes adjectives moved to Focus. This is confirmed in (21). (21)
. . . şi au poftit pre singur Jolcovschii la voroavă, and has invited dom alone Jolcovschii to counsel ‘He invited Jolcovschii alone to the counsel.’
(Costin 46)
Since the topic scope of pe is restricted to the DP, (20) also predicts that direct objects with pe-DOM are compatible with fronting to the information structure at clause left periphery. This can be seen in examples as in (22), where the relevant noun is the aboutness topic in a topic-comment structure. lăsa să să odihnească, ci pururea (22) [pre ardelenii ], nu = ii but continuously dom Ardealeans not=them let to refl rest le făcea nevoie. to.them did damage ‘The Ardealeans, they gave them no rest, but continuously caused them damage.’ (Ureche 70) MR lost the configurations that lead to the analysis in (20): that is, in MR, pe-DOM is either absent or obligatory with a given class of nouns, but never optional, as in the OR examples used so far. Importantly, when pe-DOM applies, it co-occurs with CD as a general rule. This change in the use of pe-DOM coincides with the loss of the CT interpretation on the pe-DOM-ed noun in situ. As the path for this change is well represented in the Moldavian chronicles, the next section will present a discussion of these configurations.
. CD without and with DOM This section focuses on the second pattern for pe-DOM in Moldavian chronicles, which occurs in conjunction with CD. Older stages of OR attest to the coexistence of CD and DOM as separate processes, with opposing interpretive effects: CD backgrounds the DP, whereas DOM foregrounds it. By the seventeenth century, the two operations start to merge, with a backgrounding effect. This effect is derived from a local Spec-head relation between the DOM-ed DP and the doubling clitic, which results in the reanalysis of pe-DOM as familiar versus contrastive.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
.. CD without DOM As any other Romance language, OR (and MR) has two pronominal paradigms: weak (clitic) and strong (non-clitic). Weak pronouns are proclitics or enclitics on the verb; strong pronouns can replace any lexical DP. Strong pronouns bring an emphatic reading, whereas clitic pronouns are paired with a familiar reading (Dobrovie-Sorin 1990 a.o.). The first written texts in OR attest to the possibility of having CD independently of DOM, as in (23a, 23b). (23) a. lăsămu=l elu de-a stînga +CD; –DOM leave.1pl=it it to left ‘we leave it on the left’ (CV. Apost., 21:3 apud Tasmowski 2008: 17) b. Şi ok tremeseră muierek lu Candusal şi ficiorii la muma and her sent.3pl woman of Candusal and sons.the to mother.the lu Candusal. of Candusal ‘And they sent Candusal’s woman and the sons to Candusal’s mother.’ (AL 48r apud Tasmowski 2008: 16)7 CD is optional; therefore it is not a condition for grammaticality, but another device for encoding discourse pragmatics. In particular, CD has a backgrounding effect, thus contrasting with DOM, which is foregrounding. Tasmowski (2008) points out that CD occurs mostly with strong pronouns, after they have been introduced in the discourse—that is, after they become familiar, and their referential stability is ensured. Predictably, CD does not occur with indefinite pronouns. Translations from Church Slavonic attest to the backgrounding effect of CD: Church Slavonic lacks CD and uses strong pronouns on a par with nouns, as nonmarked arguments (Assenova 1989/2002 a.o.). In OR, however, strong pronouns are always emphatic. Thus, the translator uses CD on such pronouns to equate the unmarked value of the Slavonic counterpart (Hill and Tasmowski 2008: 152). Tasmowski (2008) presents statistics showing that CD without DOM predominates with strong pronouns (versus other DPs) in the texts of the sixteenth century. Within the pronoun paradigm, the most CD-ed forms are those for the third person. The texts of the seventeenth century show the spread of CD to other pronouns and the overlap with DOM. In the eighteenth century, the overlap is established and increasing, ending up as the only production option in MR.
7 Tasmowski presents this example as problematic, insofar as the CD-ed DP might have an appositive reading in relation to the clitic. The context, however, makes this possibility very unlikely.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
.. CD+DOM Both CD and DOM occur independently in texts, for both strong pronouns and nouns. For strong pronouns, (23) showed CD, whereas (24), from Densuşianu (1997: 707; CC: 372) shows DOM. (24) vrea să slobozească pre el den robie wants to free.sbjv dom him from slavery ‘he want to free him from slavery’ The overlapping between CD and DOM is first attested in texts as a discourse progression, whereby a DP is introduced as DOM-ed (foregrounded), and is subsequently CD-ed, implying that the discourse protagonists share the referential knowledge at this stage (referential stability). Consider (25). (25) Începutul trufiei dosădire arată=se, că ocăraşte pre alalţi şi start.the vanity.gen scolding show=refl that scolds dom others and întru o nemică nu=i socoteaşte pre aceia. in a nothing not=them counts dom those ‘The beginning of vanity shows itself in scolding, for he scolds the others, and discounts them as good for nothing.’ (Coresi EV {5}) The first mention of ‘others’ in (25) is DOM-ed, but not CD-ed. The second mention is CD-ed, and the pronoun form switches from indefinite to demonstrative. As a separate operation, pe-DOM on strong pronouns is shown to have spread quite fast, from third to first and second persons, and became obligatory. The spreading is much slower in nouns. Obligatory pe-DOM may be related to the weakening of markedness and further reanalysis of pe in contexts as in (25) and (26). (26) sˇa mˇa ascultaţi pre mine to me listen.2pl pe me ‘listen to me’
(Ureche 176)
In (26), lack of CD would bring a contrastive focus reading on the DOM-ed pronoun (‘It is me—not someone else—you should listen to.’). CD, on the other hand, allows for an unmarked reading of the pronoun (‘You should listen to me for what I’m going to say.’). That is, in (26), CD applies to the pe-DOM-ed pronoun in order to mitigate the emphasis in reading. The proposed underlying configuration for (25) and (26) is (27). (27)
XP Spec pe-DPpronoun X clitic
X’
YP
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Virginia Hill
The configuration in (27) capitalizes on a local Spec-head agreement between peDOM/DP and the clitic pronoun responsible for CD. The exact point at which this configuration arises is not important. It could arise in situ if V selects XP, where XP is a Clitic Phrase; the clitic then moves to the functional domain of V. It could also arise in the functional domain of V, when both the clitic and the DP move independently through the equivalent of AgrOP in Kayne (1989). What is certain is that the clitic is not directly merged high in the functional domain in OR, but it arrives there through movement. Evidence for this comes from the double spell-out of clitic pronouns in OR (dialectal variation), once in the low position, once in the high position, as in (28). (28) a. Dece a dooadzi îl gătiră de-l porni-l, . . . so the next.day him prepared to him move him ‘so next day they prepared him to be moved’ b. şi i-au inchisu-i,. . . and them has imprisoned them ‘and he imprisoned them’
(Neculce 202)
(Neculce 153)
In (28), the proclitics have their copy spelled out in a lower position, under the past participle forms (28b). This word order situates the location for the direct merge of pronouns in the same area in which direct object strong pronouns/nouns are licensed (i.e. postverbal), so it is reasonable to expect that a local relation as in (27) will emerge. If (27) is correct, then the value of pe is forced by the local agreement with the clitic, which has a non-contrastive Topic value. The internal structure of pe-DOM is the same as in (20), only the value of the Top head in which pe merges has changed, matching the semantics of the clitic. In this respect, I follow current studies that argue for the intrinsic relationship between clitic pronouns and the Topic head at the left periphery of clauses. In particular, Delfitto (2002) proposes a unified treatment of clitic constructions within a semantic/pragmatic approach to syntactic derivation. The impact of discourse semantic features on the syntax of the clitic is justified as follows: semantically, a sentence with a simple pronominal clitic is not a proposition but a predicate, that is, an unsaturated expression (Delfitto 2002: 43). The clitic is, then, a syntactic trigger for a semantic operation consisting in the (re)opening of the corresponding argument position of the verb, which must combine with an antecedent (i.e. an empty Top) to become interpretable. Therefore, every sentence with a clitic (e.g. Marcello lo legge ‘Marcello it reads’) represents a structure of the form [Top e] [λx (Marcello reads x)]. This semantic property, which constrains the derivation of all the clauses with clitic pronouns, is encoded in syntax as a feature associated with Top (e.g. specificity): this feature is checked by the DP to which the clitic relates. The possible variation in the checking configuration results in variation in the type of construction (i.e. CLLD or CD). Along these lines, the clitic triggers a Topic interpretation on the nouns it structurally agrees with, whether it undergoes DOM or not.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
From preposition to topic marker: Old Romanian pe
Since clitics are intrinsically related to Topic, they are compatible only with referential and/or definite nouns. This intrinsic property of clitics explains their initial spread to pronouns and names of persons; that is, nominals with referential stability (Farkas and von Heusinger 2003). Accordingly, (27) applies not only to pe-DOM-ed pronouns, but also to all DPs that meet the semantic requirement.
. Conclusions This chapter proposed a syntactic analysis of pe-DOM in OR, focusing on the use of such constructions in the Moldavian chronicles of the seventeenth–eighteenth centuries. This corpus has the advantage of providing ample data where pe-DOM is dissociated from CD, a situation that is rare (and, therefore, less testable) in MR. The first argument is that pe in pe-DOM has lost its prepositional status, due to grammaticalization. Evidence comes from the various paths of reanalysis pe has undergone, and from the observation that its selectional properties differ from those that apply to the full-fledged preposition pe, concurrent in texts. The consequence of this argument is that pe-DOM must be dissociated from Case marking. This is important for the syntactic theory, since an analysis of pe as a Case marker would have implied that Case checking varies according to semantic classes (i.e. animates need pe for Case, inanimates do not). The main evidence for this point comes from ‘minimal pair’ examples, where pe-DOM is optional on the same noun, in exactly the same environment. The difference pe-DOM makes in these examples is shown to affect the interpretation, not the licensing of the noun. The interpretation triggered by pe-DOM was defined as contrastive topic. Such reading was shown to apply to pe-DOM DPs in situ, disjoint of clausal contrastive topic. A contrastive topic reading whose scope is limited to the DP leads to an analysis where pe is merged as the head of TopP at the left periphery of DP. The diachronic change to MR involves the generalization of CD in configurations with pe-DOM. An analysis of CD is proposed where the doubling clitic triggers the opposite discourse effect to pe-DOM; that is, it backgrounds instead of foregrounding the doubled constituent. For this reason, CD was used as a translating device to mitigate the reading on strong pronouns. Historical and semantic studies on CD/DOM show that CD spread to pe-DOM constructions from the pronominal paradigm to nouns. In light of this chapter, this spread brought about the backgrounding of peDOM-ed nouns, so the initial justification for applying pe-DOM to nouns has been obscured.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective G E O R G E WA L K D E N
. Introduction This chapter focuses on the V2/V3 alternation that has often been observed in Old English (OE) main clauses. Some scholars (e.g. Westergaard 2005; Hinterhölzl and Petrova 2009b) have speculated that the V3 pattern resulted from an innovation in Old English. In this chapter I present this alternation and my analysis of it, then draw on comparative data from the other early Germanic languages, Old High German (OHG) and the little-studied Old Saxon (OS). It will be argued that the possibility of V3 is more likely to be the result of shared retention than of innovation among these languages. The approach has in common with Hinterhölzl and Petrova (2009b) the idea that information structure is key to understanding the V2/V3 alternation, and that this should be captured in terms of a split CP. However, I call into question their hypothesis (2009b: 325–6) that V2 in OHG arose from reanalysis of V1 orders accompanied by a left-dislocated aboutness topic while V3 in OE/OS arose from reanalysis of V2 orders with initial familiar topic accompanied by a left-dislocated aboutness topic. I argue that this hypothesis is not supported by the data, and that it is in any case not well founded from the point of view of diachronic methodological parsimony. Sections 14.2, 14.3, and 14.4 of this chapter deal with the clausal left periphery in OE, OHG, and OS respectively, and Section 14.5 sketches a diachronic scenario. Section 14.6 summarizes and concludes.
. The structure of the left periphery in Old English A first glance at the syntax of OE main clauses ‘suggests a strong parallelism’ between OE and modern Germanic V2 languages such as Dutch and German (van Kemenade 1987: 42). Examples (1)–(3) illustrate this.1 1 References to OE examples are given from the YCOE (Taylor et al. ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective (1)
We habbað hwæðere þa bysene on halgum bocum we have nevertheless the examples in holy books ‘We have, nevertheless, the examples in holy books’ (cocathom1,+ACHom_I,_31:450.315.6332)
(2) On twam þingum hæfde God þæs mannes saule gegodod in two things had God the man’s soul endowed ‘With two things had God endowed man’s soul’ (cocathom1,+ACHom_I,_1:184.161.166) (3)
Hwi wolde God swa lytles þinges him forwyrnan why would God such small thing him deny ‘why would God deny him such a small thing’ (cocathom1,+ACHom_I,_1:181.74.71)
In all of these examples the verb follows the first constituent; in wh-questions such as (3), and where an adverb such as þa or þonne is initial, this pattern is essentially exceptionless (Eythórsson 1995: 293), and it is the majority pattern in main clauses in general. Many scholars have analysed this as uniform leftward verb-movement: to C in the case of van Kemenade (1987) and to Fin in the case of Roberts (1996). However, three alternative patterns exist which cast doubt on the analysis of V2 in OE as parallel to V2 in modern German and Dutch. These are verb-late main clauses (cf. Koopman 1995; Pintzuk and Haeberli 2008), verb-initial main clauses (van Kemenade 1987: 44–5), and verb-third main clauses.2 The latter pattern, in which two constituents precede the finite verb, as in (4), (5), and (6), is the one which concerns us here. (4) æfter his gebede he ahof þæt cild up after his prayer he lifted the child up ‘After his prayer he lifted the child up’ (cocathom2,+ACHom_II,_2:14.70.320) (5)
Fela spella him sægdon þa Beormas many stories him told the Permians ‘The Permians told him many stories’ (coorosiu,Or_1:1.14.27.243)
(6) Nu se rica mann ne mæg her habban . . . now the rich man neg can here have . . . ‘Now the rich man cannot here have . . .’ (coaelive,+ALS[Ash_Wed]:110.2758) 2 I omit (second and subsequent) conjunct clauses from consideration, since it has often been observed (e.g. Mitchell ; Kiparsky ) that these appear to pattern with subordinate clauses in constituent order terms.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
Where the subject is pronominal, as in (4), it almost invariably precedes the verb in main clauses (Haeberli 1999b: 335). Van Kemenade (1987: 138–40) was aware of such examples, in which the second-position constituent is a subject, and argued that an asymmetry between pronominal and non-pronominal subjects arose because the former were clitics. Pintzuk (1999) took a similar line, in the process noting that examples such as (5) existed in which the object appeared to be proclitic. The existence of examples such as (6) was first brought to light in generative research by Allen (1990: 150–1), and their relative prevalence was established by Haeberli (2002), who found that subject–verb non-inversion (i.e. V3) occurred 188 times (28.7 of the time) in a small corpus of 654 clauses with full subjects in second position and a fronted constituent in initial position, taken from ten text samples. The clitic analysis is rendered extremely problematic by the existence of such examples, which indicate that another explanation must be sought for V3 in OE: as Bech (2001: 98) puts it, ‘the fact that one fifth of the subjects in the [XP-Vfin -subject] pattern cannot be clitics, but nevertheless occur in exactly the same position as the clitic elements, can hardly be overlooked’. Koopman (1997) provides further general arguments against a clitic analysis of OE pronouns. Bech (2001), Westergaard (2005), van Kemenade and Los (2006), Walkden (2009), and Hinterhölzl and Petrova (2009b) suggest that the factor unifying examples such as (4)–(6) is that the elements in second position are all discourse-given; Westergaard (2005) and Westergaard and Vangsnes (2005) present a close parallel from a recent synchronic study of Tromsø Norwegian. If this information-structural approach is correct, then analyses such as those of Pintzuk (1999) or Fuß (2003), positing V-to-T movement and variation in whether the subject moves to SpecTP, are unenlightening with regard to V3 unless additions are made. Walkden (2009: 60) and Hinterhölzl and Petrova (2009b: 324) proceed to formalize the information-structural patterns in terms of the cartography of the split CP in the tradition of Rizzi (1997). Here I will base my analysis on the more nuanced split-CP hierarchy presented in (7), from Frascarelli and Hinterhölzl (2007), following Hinterhölzl and Petrova (2009b). (7) ForceP > ShiftP > ContrP > FocP > FamP∗ > FinP (Frascarelli and Hinterhölzl 2007: 22; their (37)) It is assumed that movement of constituents to these left-peripheral positions in OE is relatively unconstrained. Following Aboh (2008) and Cruschina (2009), I take information-structural features to be present in the syntax, added in the numeration; the element bearing these features must then enter into an Agree relation with a left-peripheral head. Where the probing features are associated with a movement-triggering feature, an EPP-feature in the sense of Chomsky (2000, 2001) or a generalized movement-triggering feature ∧ in the sense of Biberauer, Holmberg, and Roberts (2014), the lower element must move into the relevant specifier position.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective
Multiple landing sites for the finite verb are needed; I hypothesize that these are (at a minimum) Force and Fin (cf. also Roberts 1996). I assume that in OE, Force always requires a constituent to be merged in its specifier. This may be satisfied by internal Merge of a high frontable constituent, corresponding to the Stylistic Fronting posited by Fanselow (2003) for modern German. The relevant feature can be conceptualized in the movement typology of Biberauer, Holmberg, and Roberts (2014) as ∧ associated with an Edge Feature; cf. Rizzi (2005) for the suggestion that Force is universally a phase head. Alternatively, a discourse-related adverb (þa, þonne ‘then’) or a null discourse-continuity operator may be first Merged in SpecForceP. In cases where the feature is satisfied by external Merge, the finite verb moves to Force (I remain neutral concerning the exact mechanism of head-movement). This accounts for the consistently V2 nature of clauses introduced by þa, þonne ‘then’, and also for verbinitial main clauses. The latter have been argued to have the specific illocutionary value of expressing/recounting (M. Reis 2000a, 2000b), a property which seems to carry over to the older Germanic languages, where this pattern is referred to as ‘Narrative Inversion’ (cf. e.g. Sigurðsson 1993 on Old Norse). In neutral declarative contexts, on the other hand, the verb moves as far as Fin but no further, with other elements able to move past it into the left periphery of the clause. Crucially, familiar topics, which represent given information, may move to SpecFamP. An analysis of V3 in these terms immediately explains the high prevalence of subject pronouns in SpecFamP, since unstressed subject pronouns are ‘the canonical instance of a given nominal’ (Westergaard and Vangsnes 2005: 137). As FamP may be recursive, as indicated by the Kleene star, sequences where multiple personal pronouns precede the verb such as those discussed by Koopman (1992) can be accounted for unproblematically. (8) and (9) illustrate OE declarative clauses with verb-movement to Fin and given and new subjects respectively, abstracting away from irrelevant levels of structure. [. . .]]]] (8) [ForceP [æfter his gebede] Ø [FamP [he] [FinP ahof after his prayer he lifted . . . (9) [ForceP [On twam þingum] Ø [FinP hæfde [TP [God] . . . ]]] had God . . . in two things One problem with this approach (and that of Hinterhölzl and Petrova 2009b) is that, all other things being equal, one would predict verb-movement to Fin to occur in subordinate clauses, since the classic observation that the complementizer and the finite verb are in complementary distribution (e.g. den Besten 1977; Evers 1981, 1982) can only be used to account for the absence of verb-movement to Force in subordinate clauses under this analysis. Roberts (1996: 160) proposes that complementizers in OE are first Merged in Fin , thus blocking verb-movement. The complementizer then moves to Force because ‘the selected Force of embedded contexts requires
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
PF-realisation’ (1996: 160). While this account solves the technical problem, the PFrealization requirement is no more than a stipulation, and the account also faces learnability problems: how is the acquirer to discern the first-Merge position of the complementizer? In the framework of Roberts and Roussou (2003) and van Gelderen (2004), one might expect it to ‘grammaticalize’ upwards by eliminating the movement and treating Force as its first merged position, but this would give the wrong result for OE. Nevertheless, I will adopt this proposal, since I have no better answer to this problem at present.
. The structure of the left periphery in Old High German In this section I broaden the picture by introducing data from a second early West Germanic language, Old High German. Much of the data is drawn from Tomaselli (1995) and Axel (2007), although the analysis I develop is closer to that of Hinterhölzl and Petrova (2009b). Since many of the relevant facts about OHG are the same as for OE, this section will be briefer than the preceding one. It has long been observed that OHG exhibits a variant of the V2 property that is fairly well established (e.g. H. Reis 1901). Lippert (1974) counts 280 of 380 main declarative clause examples in Isidor as verb-second (73.6), with the rest classifiable ‘into a relatively small number of easily distinguishable and clearly describable types’ (Axel 2007: 63). Examples of subject-initial and non-subject-initial verb-second are in (10) and (11). (10) der antichristo stet pi demo altfiant the antichrist stands with the old.fiend ‘The Antichrist stands with the devil’ (11)
pidiu scal er in deru uuicsteti uunt piuallan thus shall he in the battlefield wounded fall ‘Thus he shall fall, wounded, on the battlefield’
(Muspilli 44)
(Muspilli 46)
However, there are also a number of cases of verb-third main clauses in OHG, first brought to light by Tomaselli (1995): (12)
(13)
erino portun ih firchnussu iron portals I destroy ‘I destroy iron portals’
(Isidor 157)
Dhes martyrunga endi dodh uuir findemes mit urchundin dhes his martyrdom and death we demonstrate with evidence of.the heilegin chiscribes holy writings ‘We demonstrate his martyrdom and his death with evidence from the Holy Writings’ (Isidor 516)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective
Tomaselli argues that subject pronouns are the only elements found in this position (1995: 348). Furthermore, she claims that V3 clauses are only found in the Isidor translation and in the Monsee Fragments; later texts largely do not contain this type of clause. As Axel (2007: 239) points out, these are dated earlier than most OHG prose texts. Tomaselli’s first claim appears to be falsified, at least on the surface, by clauses such as (14) (from Axel 2007: 239): (14) forlazan imo uuirdit forgiven him becomes ‘he will be forgiven’
(Monsee Fragments 6,9)
On the basis of Tomaselli’s second claim, Hinterhölzl and Petrova (2009b) state that V3 in OHG constituted ‘a very rare declining pattern’ (2009: 316), later implying that such clauses are grammatical only in OE and OS: ‘depending on whether the subject is given or represents new information we find V3- or V2-clauses in the former two languages [OE and OS—GW]’ (2009: 324). This has implications for their diachronic analysis, to which I shall return in Section 14.5. Unfortunately, Axel (2007) provides little unambiguous data on non-pronominal elements that may occur in second position, other than a few examples with adverbs: (15)
siu tho giuuanta sih she then turned refl ‘she then turned herself ’
(Tatian 665,19)
She argues that ‘there is no compelling evidence that more than one XP can move to the left periphery in OHG’ (2007: 249–50), but also that the reduction of XP-positions in the left periphery was not yet fully completed in earlier OHG (2007: 202), and that ‘XP-pron-Vfin -sequences were a native and at least partially productive pattern in earlier OHG’ (2007: 248), in contrast to Hinterhölzl and Petrova (2009b). Axel adduces considerable evidence for this, including the fact that pronouns are often inserted in this position counter to the source text in translations (2007: 248). She also demonstrates that such examples cannot be written off as instances of verb-late order (pace Lenerz 1984), since further pronouns may follow the verb, which they may never do in ordinary verb-late clauses: (16) Vnde do iu habeta si leid in-fangen in iro herzen and then you.dat.pl had she sorrow received in her heart ‘and then her heart was filled with sorrow for you’ (N Ps VII 23, 26) She also demonstrates that there is no compelling evidence that these pronouns are X -clitics in OHG as suggested by van Kemenade (1987) for OE and Tomaselli (1995) for OHG; rather, they should be analysed as full XP elements (2007: 277), even if phonologically clitics. This does raise the question of why other XP elements did not
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
move to this position (if indeed they did not), and what the trigger for the movement is; Axel states that it is simply optional (2007: 277). Other than the absence of evidence for full XP movement to the left of the finite verb, the situation in early OHG, at least, seems fully compatible with the hypotheses about clause structure advanced for OE in Section 14.2, namely that second-position pronouns and adverbs are familiar topics in SpecFamP. This similarity of analysis is not unusual in the literature: Eythórsson (1995: 324–30), for example, explicitly assumes syntactic identity between OHG, OS, and OE in verb placement in main clauses, at least to a first approximation. There is no need to assume with Hinterhölzl and Petrova (2009b) that OHG differed substantially from OE in this regard. However, it is true that the movement of subject pronouns to SpecFamP appears to be optional in those OHG texts in which it occurs at all, whereas it is virtually obligatory in OE (Haeberli 1999b: 335). Furthermore, as Axel (2007: 244–5) demonstrates, in early OHG texts pronouns may intervene between wh-elements and the finite verb. While this pattern is usually assumed to be ungrammatical in OE (e.g. van Kemenade 1987: 196), it seems that in OHG wh-questions do not behave differently from other main clauses. These more nuanced differences will undoubtedly reward future research.
. The structure of the left periphery in Old Saxon Along with OE and OHG, OS is one of only three West Germanic languages to have a textual tradition dating back to the first millennium ad. Two main texts exist from this period: the Heliand, a gospel harmony written in alliterative verse of 5,968 lines, and fragments of a version of the Genesis story, also in verse. Both can be dated to the first half of the ninth century. Given the antiquity of these texts, it is surprising that, in comparison to the vast amount of work dealing with the constituent order and clause structure of OE, OS has rarely been given any serious attention, a lack noted elsewhere in the literature (e.g. by Linde 2009: 366). For the most part, traditional philological works on syntax (e.g. Behaghel 1897) and grammars in the philological tradition (e.g. Gallée and Tiefenbach 1993) have had nothing to say about OS clause structure, and the extensive survey of verb position in the early Germanic languages by Eythórsson (1995) only mentions OS in passing. The only book-length study is Ries (1880), which due to its antiquity is useful only as a point of departure for the modern linguist. Rauch (1992), Erickson (1997), and Linde (2009) also discuss constituent order, but without going into particular detail. My own data consist of an exhaustive sample of the finite clauses in the Heliand, using the Behaghel and Taeger (1996) edition. Clauses were manually tagged for clause-type (main, conjunct, subordinate, relative, wh-question, yes/no question, imperative), and for verb position (initial, second, or late), as well as for two other features: the negation morpheme ni/ne proclitic to the finite verb, and absence of overt subject.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective
Of the finite clauses in my corpus, 2,348 were analysed as (non-conjunct) main clauses. As with OE and OHG, a variety of surface orderings are visible: however, these fall into a limited number of types. As observed by Erickson (1997), V2 seems to be the dominant pattern in OS as it is in OE and OHG. A total of 1,597 of the 2,348 main clauses in my corpus (68.0) had the verb in second position, as in (17). (17)
Mattheus uuas he hetan Matthew was he called ‘He was called Matthew’
(Heliand 1192)
A further 481 clauses (20.5) are verb-initial. Only 270 clauses (11.5) have the verb in a position later than second. This is similar to OE and OHG: Koopman (1995), for example, finds for a variety of Old English texts that the frequency of verb-final main clauses is between 0.6 and 6.1, and Cichosz (2010: 73–4) finds that between 10.7 and 16.5 of main clauses in OE, and between 1.2 and 10.7 of main clauses in OHG, have the verb in a position later than second. Some unambiguous examples of verblate order are given in (18)–(20). (18)
(19)
Ic eu an uuatara scal / gidopean diurlico I you in water shall baptize tenderly ‘I shall baptize you tenderly in water’
(Heliand 882–3)
Krist im forH giuuet / an Galileo land Christ refl forth went into Galilee land ‘Christ went forth into the land of Galilee’
(Heliand 1134–5)
(20) Ic is engil bium I his angel am ‘I am his angel’
(Heliand 99)
Crucially, verb-third as found in OE and early OHG, in which the pre-finite element is given information, does not appear to be a productive pattern in OS. I can only find four examples of this order occurring with a personal pronoun subject in my corpus; two of these are given in (21) and (22). (21)
Thanna thu scalt lon nemen / fora godes ogun then you shall reward take before God’s eyes ‘Then you shall be rewarded before God’
(22) Bethiu man sculun / haldan thene holdlico therefore one should hold him favourably ‘Therefore all should keep him in their favour’
(Heliand 1563–4)
(Heliand 1869–70)
Both of these clauses begin with adverbs that may also serve as adverbial subordinators, rendering them potentially ambiguous between main and embedded clauses,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
although they are traditionally read as the former. More tellingly, since no adverbs are present as diagnostics for verb-movement to the left periphery rather than solely to v (cf. Fuß and Trips 2002), neither is an unambiguous example of V3 with verbmovement to Fin as found in OHG and OE. Hinterhölzl and Petrova (2009b: 320) suggest that (23) (their (12)a) is an example of V3 with verb movement as in OE. (23) Thar imu tegegnes quam en idis fan adrom thiodun there him against came a woman from different tribe ‘There, a woman from another tribe approached him’ (Heliand 2984) However, this example is as inconclusive as (21) and (22) with regard to underlying structure. Since extraposition and heavy NP shift must be postulated for OS as for OE, it is possible to argue that the verb in (23) is unmoved and that the postverbal constituent en idis fan adrom thiodun ‘a woman from another tribe’ has in fact been moved rightward over it. Since this constituent represents new information— as acknowledged by Hinterhölzl and Petrova (2009b: 320)—this state of affairs is all the more likely, as rightward movement (at least in OE) appears to be driven by information-structural considerations (Pintzuk 2005: 124, n. 12; Taylor and Pintzuk 2009). Furthermore, as for (21) and (22), in context it is entirely possible to analyse (23) as an embedded clause with the meaning ‘where a woman from another tribe approached him’. The extreme rarity of this order in my corpus must also be taken as an argument against its productivity. For OE, the order XP-SubjPron-Vfin is ‘used consistently’ (Haeberli 1999b: 335) when an element other than þa, þonne ‘then’ or a wh-phrase is fronted. In the Heliand, by contrast, there are 462 examples of V2 declarative main clauses in which the subject pronoun follows the finite verb, e.g. (24) and (25), and 223 examples of V2 declarative main clauses in which the subject pronoun precedes the finite verb. All of these can be seen as ‘missed opportunities’ (Faarlund 1990: 17–18) for V3. (24) mildi uuas he im an is mode mild was he them in his mood ‘He was gentle in spirit to them’
(Heliand 1259)
(25) Thar fundun sea enna godan man there found they a good man ‘There they found a good man’
(Heliand 463)
I therefore conclude that V3 as found in OE and early OHG, with a familiar topic in second position preceding the finite verb, is not a productive feature of OS, or at least of the variety represented by the Heliand. Rather, V2 appears to be generalized in main clauses in OS much as it is in later OHG, modulo the V1 and rare verb-late patterns, which OE and OHG also exhibit.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective
. Main clauses in Proto-West Germanic: A diachronic scenario Hinterhölzl and Petrova (2009b) assume that V3 clauses were the product of a single innovation in pre-OE/OS, and that a different change took place in OHG. The diachronic scenario they posit is as illustrated in (26) (their (28)) for OE and OS, and (27) (their (27)) for OHG. (26) a. Stage I: (topic +non-V1) [Aboutness] [ForceP (familiar topic) [TP . . . Vfin . . .]] b. Stage II: [ForceP [Aboutness] (familiar topic) [TP . . . Vfin . . .]] c. Stage III: [ForceP [Aboutness]i [TP Subject Vfin ti ]. . .] (27) a. Stage I: (topic + V1) [Aboutness] [ForceP Vfin [TP . . .]] b. Stage II: [ForceP [Aboutness] Vfin [TP . . .]] c. Stage III: [ForceP [Aboutness]i Vfin [ti [TP . . .]]] In other words, they posit that OE and OS underwent a process of reanalysis that caused clause-external aboutness topics to be integrated into a clause with a clauseinternal, TP-external familiar topic (26a–b). In OHG, on the other hand, this clauseexternal aboutness topic is integrated instead into a clause in which initial position is occupied by the finite verb (27a–b). These topics are then reanalysed as originating inside the clause (26b–c), (27b–c). V3 as a syntactic possibility in OE and OS thus results from the innovation in (26a–b). Hinterhölzl and Petrova’s (2009b) general approach is appealing in many respects, since they offer a detailed consideration of the interaction between information structure and constituent order which makes nuanced predictions; furthermore, they attempt to account for a wide range of data. However, the specifics of their diachronic proposal are unsatisfactory for a number of reasons, both empirical and theoretical. For a start, Hinterhölzl and Petrova are incorrect in asserting (2009b: 324) that in OS ‘clauses expressing subordinating discourse relations [topic-comment structures—GW] pattern with OE rather than with OHG’ in exhibiting V3; as I have shown in Section 14.4, XP-Vfin -SubjPron rather than XP-SubjPron-Vfin is almost ubiquitous in the Heliand, and there is no clear evidence that clauses in which the verb has moved from its first-Merged position into the left periphery as in OE, but in which a constituent still intervenes between it and an XP in initial position, are possible at all in OS. Of course, our data on the language is limited, and it could be that the non-occurrence of such clauses is coincidental: absence of evidence is not
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
evidence of absence. However, at the very least the prevalence of XP-Vfin -SubjPron is an argument against treating OS and OE as identical in this regard. Hinterhölzl and Petrova’s reanalysis schema in (27) for OHG also cannot account for the fact (mentioned in passing earlier in their paper, 2009b: 316) that V3 orders do exist in this language, as clearly demonstrated by Axel (2007); see Section 14.3. Hinterhölzl and Petrova would either have to argue that the unequivocal examples of this kind (such as (16)) are ungrammatical, which seems unlikely, or that V3 in OHG is in fact the product of a similar innovation to that which took place in OE. There are also a few conceptual problems with this analysis. The schemata in (26) and (27) make numerous assumptions about the syntax of earlier stages of the languages in question. For instance, in order for (26a) to be possible, ProtoWest Germanic (or just Proto-Ingvaeonic) would have had to allow clause-internal preverbal familiar topics, suggesting that a V2 pattern, of a kind, was already possible. But for (27a) to be possible, Proto-West Germanic (or at least prehistoric OHG) would have had to allow verb-initial clauses with verb-movement to Force . Hinterhölzl and Petrova’s analysis thus either requires both V1 and V2 to have been possibilities in Proto-West Germanic, a state of affairs which they do not support with diachronic argumentation, or requires extra changes, which they do not discuss, to have taken place between Proto-West Germanic and the individual prehistoric OHG/OS/OE languages. Furthermore, evidence for stages a) and b) in their schemata is lacking, as they acknowledge in a footnote (2009: n. 7). Finally, Hinterhölzl and Petrova (2009b) motivate neither of the changes that they propose as initiating the reanalysis chains: why would the reanalysis involve a clause with a familiar topic for OE/OS acquirers only, and a verb-initial clause for OHG acquirers only? The alternative I will pursue here is simpler, in that it only involves a single change: the reanalysis of ambiguous SubjPron-Vfin -. . . clauses as involving verb-movement to Force rather than to Fin in early Old Saxon. In terms of the analysis in Section 14.2, I am proposing that the system involving multiple targets of verb-movement, Fin and Force , as found in OE and early OHG, was the original one, and the change that occurred in OS was the generalization of verb-movement to Force . The only plausible alternative is to assume that the change happened in reverse in OE and OHG, which is not as diachronically parsimonious: the well-established family tree structure of West Germanic, in which OE and OS (together with the later-attested Old Frisian and Dutch) are often assumed to form a North Sea Germanic or Ingvaeonic subgroup to the exclusion of OHG,3 prevents one from positing that these two languages shared an innovation, and so one would need to posit two separate (but parallel) 3 This hypothesis is not uncontroversial; however, the debate centres around the affiliation of OS, which shares certain features with OHG that the two do not share with OE (see Nielsen for discussion). For our purposes it is important simply to note that OS can be considered phylogenetically and geographically intermediate between OE and OHG. No one, to my knowledge, has proposed a subgrouping associating OE and OHG with each other to the exclusion of OS.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb-third in early West Germanic: A comparative perspective
identical innovations. Contact is also not a likely explanation, since OE and OHG occupied areas of the West Germanic dialect continuum that were not geographically contiguous, with the OS-speaking area between them. By contrast, the generalization of verb-movement to Force , and hence of V2, in OS and later OHG could plausibly have spread as a single wave of diffusion across the Continental West Germanic area. I can only speculate as to why V2 became generalized in OS and later OHG but not OE. One possibility is to link this generalization to another syntactic difference between these languages, namely the possibility/prevalence of null arguments. In OE the possibility of leaving arguments unexpressed ‘occurs (or survives) only spasmodically’ (Mitchell 1985: 633; cf. also Pogatscher 1901; van Gelderen 2000: ch. 3; Walkden 2014: ch. 5). OHG, on the other hand, seems to allow null arguments more liberally (cf. Axel 2007: ch. 6), and this possibility also exists in OS. Since the second position element in OE and OHG V3 clauses is most often a given subject, and since given subjects are the most likely elements to remain unexpressed if the grammar of the language sanctions null arguments at all, the evidence available to the acquirer for V3 in OS may simply have dropped below a critical level at some point in the language’s prehistory, at which point internal pressures may have intervened to trigger verbmovement to Force and thus generalized V2. Detailed comparative work on null arguments in early Germanic is in its infancy (though see Rosenkvist 2009 for an early summary, and Walkden 2014: ch. 5); if it can be established that null arguments were significantly more common in OHG and OS than in OE, it may be plausible to posit a causal nexus, though such a change would have to have taken place before the onset of the textual record, rendering the claim difficult to assess. Several further, language-specific changes must be posited in order to capture the intricacies of the data. For instance, V2 must have become generalized in OE whquestions, since both in OHG (Axel 2007: 244–5) and Gothic (Eythórsson 1995: 25) pronouns were able to intervene between wh-elements and the finite verb.4 Furthermore, if it is the case that only pronouns and not full XP topics could intervene between the initial XP and the finite verb in OHG, then an explanation for this qualitative difference as compared with OE is required. Detailed consideration of these questions is beyond the scope of this chapter. To summarize: under the scenario sketched here, Proto-West Germanic had generalized V2/V3, i.e. verb-movement to Fin and no further, in ordinary declarative clauses, with the surface occurrence of V2 or V3 depending on the informationstructural status of clausal constituents.
4 Fuß (: ) argues that all such cases in Gothic are word-for-word translations of the Greek Vorlage and ‘do not tell us anything about the syntax of Gothic’. This is problematic in that we must assume that these examples are fully ungrammatical in Gothic if we do not wish to posit this pattern as a native one; furthermore, the existence of an identical pattern in OHG, in which the order does not follow the Latin original, suggests that we are dealing with a shared retention here.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
George Walkden
. Conclusion In Sections 14.2, 14.3, and 14.4 I showed that the syntax of OE, OHG, and OS with respect to verb placement in neutral declarative main clauses was extremely similar, but that differences existed, primarily in the frequency of occurrence of V3 clauses: these are extremely common and apparently the default pattern in certain contexts in OE, rare but still abundantly attested in OHG, and, crucially, non-existent in my corpus of OS. The alternation between V2 and V3 in OE was shown to be information-structurally conditioned. I developed a (partial) feature-based analysis of the relevant movements within a cartographic framework in which informationstructural features are present in the syntax, based on Walkden (2009), Axel (2007), and Hinterhölzl and Petrova (2009b), and argued that OS differed from OE and early OHG in that it had generalized V-to-Force movement. Using the data and analysis from these sections I then outlined a diachronic scenario whereby OS (and later OHG) lost the possibility of V3 and generalized V-to-Force movement. Such a scenario was shown to be preferable to that of Hinterhölzl and Petrova (2009b), an account which faces certain empirical and conceptual problems discussed in detail in Section 14.5. If my account is along the right lines, some light is shed not only on the syntactic properties of the early West Germanic languages but also on those of unattested stages of the Germanic family tree.
Primary sources Behaghel, Otto, and Burkhard Taeger. 1996. Heliand und Genesis, 10th edn. Halle: Max Niemeyer.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics: Conflation and interactions with the left periphery ED CORMANY
. Introduction This diachronic corpus study examines how the varieties that comprise the Friulano dialect cluster of Northern Italy progressed in their use of subject clitic doubling, from none at all in early texts to quite liberal doubling in the present day. In the literature on Northern Italian Dialects (NIDs), Friulano is known for being ‘advanced’ in its ability to double various types of subjects with a subject clitic (SCL). In her broad survey of NIDs, Poletto (2000) outlines a typological hierarchy of SCL doubling according to features of the subject nominal; present-day Friulano occupies one of the highest positions on the hierarchy since it can double wh-phrase subjects, and therefore all types of subjects. Examination of a diachronic corpus reveals how Friulano reached this stage, demonstrating that changes in SCL use proceeded in a stepwise fashion, allowing one additional type of subject doubling at each stage of its historical development. Friulano has two classes of clitics (which I call ‘high’ and ‘low’ SCLs); only the high SCLs, which I analyse as left-peripheral particles (cf. Benincà 1983), participate in the advanced forms of doubling. The distribution of high SCLs cannot be attributed to information structure alone, and must be given a positional, syntactic account. Furthermore, I show that each stage on the doubling hierarchy corresponds to a position in the Rizzi-style left periphery. This provides for a straightforward explanation of the diachronic progression of SCLs in Friulano, as left periphery positions were lost through conflation—a process of syntactic change in which indistinguishable adjacent phrasal positions collapse into a single projection to economize structure.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
. The dialects and corpus Friulano or Friulian is the name given to a cluster of closely related dialects spoken in the Friuli–Venezia Giulia region of north-east Italy. Despite significant debate whether present-day Friulano should be classified with NIDs or with Rhaeto-Romance, early Friulano texts bear significant lexical resemblance to other Italian varieties. Friulano also notably patterns with other NIDs in its use of SCLs to represent agreement information about clausal subjects. SCLs are clearly an innovation when compared to Proto-Romance; Standard Italian lacks them, but French, Gallo-Romance, and some Ladin varieties feature them to varying extents. Compared to neighbouring dialect groups, Friulano is notable for being liberal with respect to subject doubling. In Friulano, an SCL may double any type of subject nominal (pronouns, DPs, QPs, and wh-phrases). Only Trentino has even more extensive doubling, as it mandatorily doubles all subject nominals. Subject clitics are typically internal innovations within a language, and follow a predictable course of development, beginning as non-clitic pronouns. (1)
tonic pronouns → weak pronouns → agreement markers
The latter two stages are the two classes of syntactic elements that are subsumed under the label subject clitic. Weak pronouns are also known as phonological clitics, since they do not satisfy their clitic nature by forming a syntactic unit with any other element in the clause. Phonological subject clitics are found in Standard French, and pattern closely with full DP subjects. As such, they are best treated as maximal projections in ordinary subject position (SpecTP). Since these SCLs compete with DP subjects for a single position, they are in complementary distribution and subject doubling is impossible. Contrast the agreement-marking SCLs found in certain colloquial French varieties and NIDs, which are also known as syntactic clitics, since they necessarily participate in syntactic head-adjunction. Syntactic subject clitics are best analysed as externally merged syntactic heads, to which other Xmin categories (typically the T + v + V complex) adjoin. It is generally the case that syntactic clitics/agreement markers, such as those found in present-day NIDs, have to pass through an intermediate stage of being weak pronouns before becoming agreement markers through a process of Spec-to-head reanalysis. However, the corpus data show that while Friulano has had syntactic clitics for most of its history, there was never a time at which it had phonological clitics. As such, an alternative explanation of the genesis of Friulano SCLs is required. The corpus used in this study is a selection of approximately 4,500 words of Friulano poetry from a historical anthology (Gregor 1975); the texts cover the years 1400–1877.1 The corpus contains 537 tensed clauses, each of which was tagged for 1 The date of birth of the author was used when available. For some earlier texts or those with unknown authorship the estimated date of composition was used.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics
matrix/embedded position, clause type, verb valence, the position and features of the subject, and presence and type of SCL. The earliest texts show no use of SCLs, and from their first attestation in a 1529 text, their frequency increases essentially monotonically over the surveyed period. Statistical analysis of the annotated data shows no strong correlation between SCL frequency and syntactic features other than subject type. I discuss this in more detail in Section 15.5 and conclude that the major factor driving the change is conflation of positions in the articulated left periphery, which hosts certain SCLs. Although no single feature predicts increased SCL use, there are structural contexts that prevent use of SCLs, namely when other types of clitics occur in the same clause. Previous analyses that rely on information structure (Benincà 1983; Poletto 2000) do not predict this fact; in Section 15.6, I explain these non-co-occurrence phenomena in terms of the clausal architecture and the features of SCLs and DPs.
. Properties of Friulano SCLs Drawing upon data from nine dialect groups and over 100 individual varieties, Poletto (2000) subdivides the SCLs found in NIDs into four groups—invariable, deictic, number, and person SCLs—and claims that each occupies a distinct position in the ‘agreement field’, which spans the left periphery and the upper portion of IP. (2) [LDP invariable [CP deictic [FP tinvariable [IP [NegP [NumP number [HearerP person (after Poletto 2000: 36, ex. 63) [SpeakerP T+v+V [ TP ]]]]]]]]] It is important to note that Poletto makes a non-trivial distinction between IP and TP in her cartography. TP contains only tense and argument structure positions, whereas IP dominates polarity (NegP), the two lowest SCL positions, and TP. I will make use of this distinction between left-peripheral and IP-internal SCL positions to explain clitic interaction effects in Section 15.6. Within the Friulano varieties, there is a certain amount of heterogeneity of surface forms of SCLs. However, they do appear to be syntactically homogeneous and cluster into two groups, comparable to deictic and number SCLs in Poletto’s taxonomy. As I do not adopt the same cartography as Poletto, and in light of there being only two classes, I refer to these as ‘high SCLs’ and ‘low SCLs’. The Friulano high SCL al is uninflected, whereas low SCLs are inflected for person and number. While low SCLs are clearly related to their corresponding tonic pronominal forms, al has no such correlate. As a result, the high SCL consistently has the form al throughout the corpus,2 whereas the low SCLs vary depending on the strategy and 2 With the exception of a single text where the high SCL form is a, which may better be classified as belonging to a Padovan variety, or at least affected by contact with Padovan. See Section . for further discussion.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
progress of phonological reduction. A paradigm of these forms (after Gregor 1975), abstracted across several texts and periods, is given in (3). (3)
1sg
2sg
3sg
high low
1pl
2pl
3pl
’o
’o
’a
al ’o
tu
’e
Both types of Friulano SCLs are syntactic clitics, and therefore heads. This is demonstrated by the fact that both high and low SCLs are found in doubling contexts. Also note that in (5) there is no phonological interaction between the vocalic low SCL ’e and the final vowel in the preceding word biciclete. (4) un stradarûl al mene la cariole a street-sweeper scl push.3sg the cart ‘A street-sweeper pushes his cart.’ (5)
la biciclete ’e va the bicycle scl go.3sg ‘The bicycle goes.’
Although at first glance there do not appear to be major structural differences between sentences with high SCLs and those with low SCLs, they do clearly occupy distinct positions. The high SCL al must occupy a left periphery position, as this will account for its variability in terms of what types of subjects it can double (see Section 15.5). The fact that low SCLs do not undergo the changes that high SCLs do, are limited with respect to what types of subjects they can double, and cannot co-occur with any other type of clitic (see Section 15.6) indicates that they occupy a distinct, lower position inside the inflectional field.
. The evolution of Friulano SCLs Recall the canonical path by which syntactic SCLs are taken to develop from strong pronominal forms: (6) tonic pronouns → weak pronouns → agreement markers Poletto (1995) argues that just this type of change has occurred in the Veneto dialect cluster, which geographically neighbours Friulano. It would be reasonable to assume a parallel change occurred in Friulano, except the corpus data contraindicates the existence of a middle stage. In the earliest texts (c.1400), SCLs were not used in any way. After their first attestation (1529), their frequency of use increased consistently until the end of the surveyed period (1877).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics
High SCL frequency by year
(7)
0.3 0.2 0.1 0.0 1400
1500
1600
1700
1800
Low SCL frequency by year 0.5 0.4 0.3 0.2 0.1 0.0 1400
1500
1600
1700
1800
Of all of the clause characteristics coded for in the corpus—including embedding status, verb valence, presence of negation, second conjunct status, and subject position—none proved to be a significant predictor of SCL frequency. However, as the figures in (7) show, the date of a text did correlate with SCL frequency.3 Across the entire corpus, 14.5 of tensed clauses had a high SCL, and 11.4 had a low SCL. High and low SCLs never co-occurred within a single clause. What is more striking, however, is the fact that none of the corpus data exhibits weak pronouns of the French type. The Friulano data confirm a jump from a language with no SCLs to one with syntactic SCLs, with no intermediate stage. Since phonological 3 There was considerably higher variance for low SCLs, although this appears to be due mostly to stylistic variation rather than any grammatical property. Comparatively early texts with a preponderance of low SCLs tend to be heavy in dialogue, which has a high frequency of pronominal and pro subjects. Conversely, later texts that eschew low SCLs entirely are written in a more formal register.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
clitics pattern with DP subjects, they are in complementary distribution, ruling out clitic doubling of an overt subject. Conversely, languages with phonological SCLs tend not to allow pro-drop. In fifteenth-century Friulano, pro-drop was always available for 1st person (8) and 2nd person subjects (9), and for 3rd person in embedded contexts. When pro-drop was unavailable (e.g. for 3rd person, unembedded subjects), tonic subject pronouns were mandatory (10). (8)
Sufriraj peno e torment pluj ch’ogno altri inamorat. suffer.1sg pain and torment more than-any other lover ‘I suffer pain and torment more than any other lover.’
(9)
alegro may no mi vedras happy never not ocl will-see.2sg ‘You will never see me happy again.’
(10) may el sará pur lu to dan but pn.3sg will-be.3sg too the your loss ‘. . . but the loss will be your own.’ SCLs first appear in the sixteenth century. Interestingly, the high SCL al is attested significantly earlier (1529) than any of the low SCLs (1602). This is unsurprising in one sense: since al is indeclinable and does not appear to be derived from a pronominal form, it was certainly never a weak pronoun per se, nor does it appear to have been an ontologically equivalent phonological clitic. This is in accordance with the stance taken by Benincà (1983) that the Padovan clitic a, which has a similar distribution and appears to be cognate with Friulano al, is not a true subject clitic, but rather a clitic left-peripheral particle.4 The major difference between the situations in the two dialect groups is that Benincà claims that Padovan a was once a member of an inflected SCL paradigm, but broke free from the paradigm and was reanalysed as occupying a higher position. This cannot be the case for Friulano, since al is attested before any of the members of the low SCL paradigm. One possible explanation is that Friulano al was introduced by contact with Padovan or other neighbouring dialects in which such a paradigm split did in fact take place. This hypothesis is bolstered by the extraordinarily high rate of high SCL use in a 1722 Friulano text. In that text, the form of the high SCL is a, not al, suggesting that this one text may have been more influenced by Padovan than the rest of the corpus, and also supporting the notion that Padovan a was further along in its course of development.
4 Despite the fact that I agree with Benincà that al is not an SCL, I continue to gloss it as SCL, as this has remained the standard in the NID literature.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics
. Subject doubling in Friulano .. An implicational hierarchy of subject doubling Another prediction to test against the corpus data is the general typology of doubling within NIDs. Dialects vary as to what type of subjects they can double, but these variations form an implicational hierarchy. (11)
wh-phrase > QP > DP > tonic pronoun
According to Poletto (2000), if a dialect can optionally double a particular level on the hierarchy, it mandatorily doubles subjects from all lower levels. At least some presentday Friulano dialects optionally double wh-phrases, as shown in an example of an embedded wh-relative from the variety spoken in San Michele al Tagliamento. (12)
Le fomne che le neta le scale e e ndade via. the women who scl clean the stairs scl have gone away ‘The women who clean the stairs have left.’ (Poletto 2000: 142, ex. 8c)
The historical Friulano corpus does not have any examples of wh-doubling, but does show the expected progression of doubling as far as it went in the surveyed period, in that pronouns are first doubled, followed by DPs, followed by QPs. The table in (13) gives the dates of first attestation of each type of doubling, both with high and low SCLs. (13)
wh
QP
DP
pronoun
high
N/A
1862
1622
1529
low
N/A
1792
1622
1602
However, at no point is there mandatory doubling of any type of subject, despite the fact that DP doubling was possible in the majority of texts and QP doubling was possible in later texts. In the next section I will show how this is possible in a diachronic left-periphery analysis of Friulano SCLs, how further change could easily lead to the mandatory doubling situation found in some present-day Friulano varieties, and how hierarchy-imposed mandatory doubling could be explained by an identical left-peripheral structure plus minimal lexical change. .. Conflation as progression along the hierarchy The cartography of the agreement field, as put forth in Poletto (2000), specifies individual positions for each of the four types of SCLs, but makes little attempt to assimilate these positions to the more traditional cartography of the left periphery, following from Rizzi (1997). Since Friulano uses just two of Poletto’s positions, rather
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
than using a simplified version of her cartography (14), I adopt the Rizzi-style left periphery (15). The one aspect of Poletto’s structural analysis that I agree with is that high SCLs are in the CP layer, while low SCLs are in the TP/IP layer. (14) [CP high SCL [IP [NegP [NumP low SCL [HearerP [SpeakerP T+v+V [ TP ]]]]]]] (15)
ForceP > (TopP) > FocusP > TopP > FinP > TP
(16) Wh-phrase > QP > DP > tonic pronoun One major benefit of adopting the traditional cartography is that it directly mirrors the doubling hierarchy (16), which is reflected in the diachronic progression of doubling possibilities and SCL frequency in the Friulano corpus. This parallelism is not coincidental, but corresponds to the functions that the various left-peripheral positions perform in other contexts and other languages. Spec ForceP is the locus of wh-moved nominals. FocusP is claimed by Rizzi to be inherently quantificational, in contrast to TopP, which is a more generalized information structure position. FinP is the least specified left-peripheral position, and can host any type of subject, including pro. With these facts in mind, we can form a diachronic explanation for Friulano’s progression up the doubling hierarchy. First of all, al is the only overt left-peripheral particle in the language. As such there is no way to directly determine its position by comparison to other left-peripheral heads. This is exactly the learnability problem that was posed to children acquiring a Friulano variety. The potential ambiguity in the position of al eventually led to the loss of distinction between the various possible positions. This is the process of conflation, in which two adjacent syntactic phrases become a single phrase that can bear the features of either, because no positional distinction can be made. In the case of al, its position became conflated with null-headed positions that host nominals in their Specs; those nominals’ features correspond to the projection they are attracted to (ForceP, FocusP, and TopP). Thus conflation allowed al to occupy the same position as these more specialized null heads, and thereby increase its doubling possibilities. The stepwise nature of the change, where each type of doubling was allowed in turn over the span of some 500 years, indicates that these positions were conflated one by one, rather than in one fell swoop. To elaborate on the conflation account, the original position of al must be ascertained. In its first 100 years of attestation in the corpus, al was almost exclusively used in conjunction with pro. (17)
Al pár al mont cu cui cu scrîi in rime al sei scl seems to.the world that to-him who writes in rhyme scl is tignût a fâlu par toscan held to do.it through Tuscan ‘It seems to the world that for those who write in metre, they are required to do it in Tuscan.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics
In both clauses containing al in (17), the subject must be pro. In the first, it is expletive pro used with the raising verb pâr; in the second, the wh-phrase ‘who write in metre’ is contained in a dative relative, and cannot be the structural subject of the clause, although it is coreferential with the pro subject. In later texts in the corpus it becomes very common to use al in sentences with postposed DP subjects. In these sentences too the subject position cannot go empty, and is filled by pro.5 (18)
Al alze i vôi Denêl scl raise.3sg the voice Daniel ‘Daniel raised his voice.’
The impossibility of overt subject doubling in early uses of al indicates that it bears rather impoverished features, since it cannot attract a DP subject. In light of this, it seems best to posit that the first occurrences of al are in Fin . There are early subjectinitial clauses that lack al (19); in such clauses, no dedicated Fin is selected, and the overt subject may remain in SpecTP. (19)
jo cgiantaraj al vuestri honor pn.1sg sing.fut.1sg to-the your honour ‘I will sing in your honour.’
Because al is located in the lowest position in the left periphery, it does not interfere with the function of the higher positions. Certainly there may be clauses where FinP is the highest active projection (according to the Avoid Structure Principle (Rizzi 1997)); in other clauses, ForceP, FinP, and TopP may be selected to be active. This is the case in the embedded clause in (17), where a non-argument relative clause is preposed before al. The possibility of filling the projections above FinP, combined with the fact that Top is null in Friulano, presents a learnability challenge. In a structure that selects for a topic and a null subject, there is no intervening overt element between the topicalized XP and al in Fin ; as such, there is the opportunity to collapse the structure, giving al the ability to license overt topics. The trees in (20) show how the two null positions are eliminated in favour of a structure with a single, conflated Top/Fin projection.
5 A remnant movement analysis, in which the postposed subject does not remain low, is untenable for these constructions. Since al plus its complement (TP) do not comprise a maximal projection, that unit cannot be moved. Thus the DP subject in Spec of al, despite having all its features checked, would have to be moved to a higher Spec position, and then the remnant would have to subsequently cross over it. Given that al heads one of the highest positions (if not the highest) in the clause, there is simply no way to accomplish such movement.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
(20)
TopP XP Top0 Ø
FinP pro Fin0 al
TP
Top/FinP XP Top/Fin0 al
TP
The structure with distinct TopP and FinP is representative of the stage prior to the first attestation of subject doubling in Friulano. Once conflation occurs, subject doubling with al becomes possible.6 As attested in the corpus, at the second stage only DP subjects could be doubled. This is in accordance with the fact that TopP only hosts ‘plain’ DPs in its Spec position, while QPs and wh-phrases must be hosted in SpecFocusP and SpecForceP, respectively. After the conflation of FinP and TopP took place, there was a relatively lengthy, stable period during which optional subject doubling in Top/FinP was present in Friulano. Similar configurations have persisted to the modern day in most other NIDs.7 However, in Friulano, the possibility for further change was realized. One reason that this change did not occur as readily as the conflation between TopP and FinP is because unlike in the prior stage, as shown in (20), the presence of an overt specifier of al prevents the higher projection from conflating with it directly. 6 Note that while TopP could attract other elements while al resided in FinP, it would not be possible to have a subject topic and al in Fin occur in the same clause. If both were selected for, the derivation would necessarily crash, since if the subject could raise directly into Spec TopP, skipping over Spec FinP, the probing features of al would go unchecked. 7 Other varieties that can double both pronouns and DPs may lack high SCLs altogether; in such cases, a conflation has likely occurred within IP. Without drastic modification of the clausal structure—reanalysing material from the inflectional domain as part of the left-peripheral domain—no further conflation can take place, so the grammar remains frozen at the stage of DP doubling. If the presence of a high SCL is requisite for QP or wh-doubling, it could explain the relative infrequency of such doubling across NIDs.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics (21)
FocusP XP Focus0 Ø
Top/FinP DPS Top/Fin0 al
Despite the fact that Focus is null, the structure is completely unambiguous; no structure can be conflated directly. From the configuration shown in (21), additional conflation could be triggered in one of two ways. One possibility is that the conflation of FocusP and TopP proceeds ‘silently’, i.e. only through ambiguity between clauses containing null left-peripheral heads, and the result indirectly affects the featural specification of al. A general lack of co-occurrence of the two projections could lead to such a simplification of structure.8 When FocusP and low TopP are conflated, there is no longer a need for separate null Focus and Top ; a single head can bear the features of each, including the [+Q] feature inherent to Focus. Since al is the only other head merged in that position, by analogy the [+Q] feature is transferred to it as well, allowing for QP doubling, as found in Friulano from 1862 onward. (22) se qualchidùn al cîr di scrivi sclet in tal nestri lengàz if somebody scl tries of write.inf plain in such our language ‘if somebody tries to write clearly in our language’ The other possibility is that the criteria for conflation are simply too strict. On a conservative view, conflation can only take place if a given construction is always ambiguous; if even one grammatical structure disambiguates the positions, a child acquiring the language will have positive evidence for multiple positions and maintain them in the grammar. In the case of subject doubling, while there will be unambiguous structures at the stage represented in (21), al can still host a pro subject, leading to a configuration similar to that in (20), which was conflatable (and indeed conflated). If conflatable structures appear at a significant enough rate in the primary linguistic data, the conflation may take place in spite of the existence of unambiguous structures.
8 Exactly what condition would produce such a scenario needs to be determined by further work on the current corpus, as well as comparison with other cases of Focus/Top conflation. One possibility, however, is the availability of a high TopP, above FocusP. If topics that co-occur with foci overwhelmingly surface as high topics, this could lead language learners to conclude that FocusP and low TopP are essentially the same position.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
What exactly the critical rate for triggering conflation may be is an open matter for the field of diachronic syntax.9 .. Extension to wh-doubling and mandatory doubling The most ‘advanced’ type of doubling found in the corpus is QP doubling, which indicates that for speakers born in the late nineteenth century, conflation of the left periphery had proceeded only as far as FocusP. Present-day data exhibiting wh-doubling, such as (12), show that additional conflation must have taken place. Like the Focus/Top collapse, that change must also have taken place indirectly, again offering an explanation for the existence of a stable period between the changes. The only further possible change is that found in Trentino, in which doubling of all subjects, including wh-phrases, becomes mandatory. Although this has not occurred in Friulano, it is certainly a plausible change. Once a language has wh-doubling, it necessarily has a unitary CP layer. If the only available C elements (ignoring subordinating complementizers) have identical feature specifications, but one is null and the other is overt, lexical loss of the null variant is possible. When that occurs, doubling becomes mandatory for all subjects, since all subjects must move into CP to be licensed, and C must be overtly headed. While the conflation analysis I have presented accounts for the diachronic pattern in Friulano, it only addresses half of the typological claims made by Poletto (2000). What of the claim that if a language optionally doubles subjects at a certain level, it mandatorily doubles subjects of all lower levels? This simply does not seem to be the case in Friulano, as there is no mandatory subject doubling (either by high or low SCLs) at any stage. The conflation analysis I have posited does not necessarily predict mandatory doubling; however, by a lexical loss process similar to that posited above for Trentino, mandatory doubling of lower levels can be accounted for in varieties that exhibit it. After conflation takes place, there are two lexical items that can fill the new position: the null head that occupied the former, higher position, and the high SCL. Since the null alternant for the lower position can no longer be distinguished from any other null left-peripheral head, it is lost. Thus the high SCL becomes the only head that can bear the features previously borne by the lower null head. So, for example, at the optional QP doubling stage, the Focus head can either be filled by a null Focus that is [+Q] or by the high SCL, which is [±Q], [±Topic], and [+phi]. A topicalized element must check its topic feature against the high SCL, whereas a focalized element can check against Ø or the overt clitic. Since this does not occur in Friulano, either the ‘lower’ functional heads are maintained in the lexicon (although they merge in the 9 Preliminary work on a perhaps related issue, namely phonemic merger, has been undertaken by Charles Yang (unpublished ms.) and others. Their findings indicate that even a small percent of merged data can trigger a merger for a particular individual, and then spread rapidly throughout a population.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics
single conflated position), or the ‘lower’ features are transferred to the higher null head, allowing pure optionality between Ø and al. The schemata in (23) and (24) summarize the two paths of change: that which does not predict mandatory doubling, and that which does. (23) Friulano corpus—no mandatory doubling a. [ForceP [FocusP [TopP [FinP pro {al/Ø}[+phi] [ TP ]]]]] b. [ForceP [FocusP [Top/FinP DPS {al/Ø}[±Topic, +phi] [ TP ]]]] c. [ForceP [Focus/Top/FinP QP/DPS {al/Ø}[±Q, ±Topic, +phi] [ TP ]]] (24) Trentino—mandatory doubling a. [ForceP [FocusP [TopP [FinP pro {SCL/Ø}[+phi] [ TP ]]]]] b. [ForceP [FocusP [Top/FinP DPS SCL[±Topic, +phi] / Ø[±Topic, +phi] [ TP ]]]] c. [ForceP [Focus/Top/FinP QP/DPS SCL[±Q, ±Topic, +phi] / Ø[+Q, +phi] [TP]]] d. [CP Wh-/QP/DPS SCL[±Wh, ±Q, ±Topic, +phi] / Ø [+Wh, +phi] [ TP ]] e. [CP Wh-/QP/DPS SCL[±Wh, ±Q, ±Topic, +phi] [ TP ]]
. Non-co-occurrence of al and other clitics Despite being the only overt head in the left periphery, the high SCL al appears to interact with clitic positions elsewhere in the clause. Although it does not behave like a ‘true SCL’ in several respects, one of the reasons for classifying al and Padovan a as SCLs is their inability to co-occur with other SCLs. This is a desirable prediction; since each clause has a single subject, it can only agree with a single clitic element. The prediction is borne out in the Friulano corpus, as there is no clause containing both a high and low SCL. A further pattern which emerged from the corpus, which to my knowledge has not been discussed in the literature for Friulano or any other NID, is that the high SCL cannot co-occur with any type of clitic, including object and dative clitics. Given that 46 clauses in the corpus contain object or dative clitics, with an overall high SCL occurrence rate of 14.5, this pattern cannot be due to chance. One possible explanation would be an appeal to information structure properties of the high SCL. The view that high SCLs ‘indicate that the whole sentence is new information’ (Poletto 2000: 23) was first put forth by Benincà (1983) in her discussion of Padovan a. Naturally, object and dative clitics stand in place of given information, and would be incompatible with such a restriction. However, al appears with other elements that represent old information, including pro, definite DPs, and deictic subjects. (25) L’onor de me’ famèe al ûl cussì the.honour of my family scl wish.3sg thus ‘Thus will be the honour of my family.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
(26) Chest al è pôc this scl is little ‘This is the least of it.’ These data confound an explanation that relies solely on the new versus old information distinction. Fortunately, the data can be explained in terms of syntactic structure rather than information structure. The positions of low SCLs, object clitics, and dative clitics—although not identical—all serve the same role to prevent the selection of al. Object and dative clitics are generally held to occupy a position within an articulated IP. Recall that Poletto (2000) also places low SCLs of the type found in Friulano in a high position within articulated IP, crucially above TP. What these positions all have in common is that they intervene between SpecTP and the position of the high SCL in the left periphery. To be accessible to Agree with al, the subject must be in a high enough position (typically SpecIP), and must have some unchecked feature (to satisfy the Activation Condition of Chomsky (2001)). When there is a clitic in the head of IP, it will affect whether an overt subject DP is available for agreement and movement into the left periphery. If a low SCL is present, it undergoes Agree with the subject (which has moved to SpecTP to satisfy the EPP property of T , but is still active) and moves it to SpecIP, creating a low SCL doubling configuration. However, that Agree relation checks all of the features on the subject, rendering it inactive for further Agree with al and movement into the left periphery (27). In the case of non-subject clitics, they do not bear the features required to Agree with the subject and move it into SpecIP. Furthermore, these clitics bear phi features, and create a defective intervention effect. If al were merged above a non-subject clitic, it would not be able to probe for a suitable goal, and the derivation would crash (28). (27)
CP IP
C
∗al
DPS
Agree
I0 SCL
TP
Agree
…
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Changes in Friulano subject clitics (28)
CP IP
C ∗al
I0 OCL
3
TP
2
Agree
……
Agree
1 Thus the position and featural content of high versus low clitics of all types directly predicts the non-co-occurrence facts exhibited in the corpus data.
. Conclusions I have shown that through examination of a historical corpus of Friulano texts, much can be determined about both the synchrony and diachrony of subject clitics in Friulano and their place in the typology of NIDs and clitic-doubling languages more generally. The corpus findings also reveal some unexpected results about the development of SCLs, compared to previous accounts in the literature. High SCLs appear in the corpus prior to inflected low SCLs, contrary to the case in Padovan. However, this difference actually helps to account for the fact that Friulano seems to have never had phonological clitics, only syntactic clitics, unlike its neighbour to the south, Veneto. In terms of their internal diachrony, Friulano SCLs progress in the expected, stepwise fashion with respect to the types of subjects they can double. I have explained this change as being caused by the conflation of positions in the Rizzi-style left periphery, which correspond directly to the various types of subjects. Subjects with higher final positions are ‘more difficult’ to double, as the high SCL al began in the lowest left-peripheral head, Fin , and over time was able to assume the ‘higher’ roles as the structure of the left periphery collapsed onto it. This model of conflation does not necessarily produce mandatory subject doubling, which is in accordance with the data, but with a single minor addition can produce the hierarchical effect that is argued by Poletto to take place in other NIDs. Finally, the wide range of the historical corpus helped illuminate what is clearly an unchanged synchronic fact about Friulano, namely that high SCLs do not cooccur with any other clitic elements. I have shown that it is not necessary to appeal
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ed Cormany
to information structure considerations to explain these facts. The positions of clitics in the articulated IP, along with the features of DPs and SCLs, predict these outcomes purely in terms of structure and Agree. Thus the cartography of the Friulano clause, as I have outlined it, directly accounts for the major diachronic and synchronic issues involving SCLs.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci: Causes and consequences L I E V E N DA N C K A E RT
. Introduction .. ‘Left Edge Fronting’: The phenomenon In this chapter, I will be concerned with a specific kind of embedded A -movement in Latin whereby one or more constituents are fronted to the left of a conjunction that introduces an embedded clause. The basic pattern is schematically represented in (1), with ‘Sub’ for ‘subordinating conjunction’: (1)
[XPi [Sub [ . . . ti . . .
I will call this particular linear order ‘Left Edge Fronting’ (LEF). In Danckaert (2012), it is argued that Latin LEF comes in two kinds, a topic-like variety LEF1 and a focus-like LEF2.1 In cases of LEF1, the fronted element is always a relative wh-word (2) or a form of the demonstrative pronouns is or hic (eum ‘him, that one’ in (3)). In both cases, the fronted element refers to a discourse-old entity, making it plausible to characterize LEF1 as a type of topicalization. [cum ti scies]], facies ut sciamus. (2) [Quodi which-acc when you-will-know you-will-make that we-know-subj ‘When you know this, you will make sure that we know it as well.’ (= Cic. ad Att. 8.15.1)
1 For more detailed discussion, the reader is referred to Danckaert (: ch. ), where it is shown among other things that LEF is also attested in other old languages (like Vedic Sanskrit and Ancient Greek), as well as in a number of modern languages (such as Modern Greek and Bulgarian).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(3)
Lieven Danckaert [cum ti uidero]], [Eumi Arpinum pergam. him-acc when I-will-have-seen Arpinum-acc I-will-proceed ‘When I have seen him, I’ll move on to Arpinum.’ (= Cic. ad Att. 9.15.1)
Moreover, LEF1 is only attested in embedded clauses which themselves are located in a leftward position in the clause they are embedded in. In Danckaert (2012: chs. 4 and 5), this left–right asymmetry is analysed in terms of clausal pied-piping (see also Bayer 2001). The reader is referred to these chapters for further discussion of LEF1, which, despite surface similarities, can be shown to be very different from LEF2, synchronically as well as diachronically. On the other hand, the second type of LEF is attested in both clause-initial (4) and clause-final embedded clauses (5). (4) [Antoniumi [si ti uidero]], accurate agam de Antonius-acc if I-will-have-seen in-detail I-will-talk about Buthroto. Buthrotus-abl ‘If I see Antonius, I will inform him in detail about Buthrotus.’ (= Cic. ad Att. 14.19.4) (5)
Conloqui uidebamur [[in Tusculano]i [cum ti essem]]. I-was-subj talk-inf we-seemed-impf in Tusculan-abl when ‘It seemed as if we were discussing, when I was in the Tusculan estate.’ (= Cic. ad Att. 13.17–18.2)
LEF1 and LEF2 can co-occur in one and the same clause, as in (6). If so, the LEF1 constituent (quae ‘which’ in (6)) always precedes the LEF2 phrase (the PP in nouam coloniam ‘into the new colony’). (6) [Quae [[in nouam coloniam] [cum introierunt]], which-nom in new-acc colony-acc when have-entered-pf permanent [. . .] libenter [. . .]]. they-stay happily ‘When they have entered the new colony, they stay there happily.’ (= Var. Agr. 3.16.31) In the present chapter, I will only be concerned with this second kind of fronting. In Section 16.2, I will elaborate on the interpretive characteristics of LEF2. After this, I will look at the syntax (Section 16.3) and the diachronic evolution (Section 16.4) of LEF2. First, I will say a couple of words about the ‘cartography’ of the left periphery of embedded clauses.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
.. The position of subordinating conjunctions In Rizzi (1997), it is proposed that clause typing elements are hosted in the highest projection of the split CP, namely ForceP. For instance, the English complementizer that, which marks the embedded clause as having declarative illocutionary force, is said to be base-generated in ForceP. Evidence for this is the observation that the thatcomplementizer systematically precedes embedded topics and foci (7); witness the ungrammaticality of (8). (7) John says [ForceP that [TopP [this book] [TP he likes]]]. (8)
∗ John says [this book [that he likes]].
With Roussou (2000), Rizzi (2001), and Krapova (2010), I will assume that clause typing elements do not systematically coincide with overt subordinating conjunctions. Rather, it seems to be the case that a subordinating conjunction can be merged below ForceP. A point in case is Italian se ‘if, whether’, which can be preceded by a clitic leftdislocated topic. detto la (9) Non so [ForceP OP[TopP [a Gianni] [IntP se [TP gli ha to Gianni if him he-has told the not I-know verità]]]]. truth ‘I don’t know if he told Gianni the truth.’ As indicated, I assume that, in examples like (9), the embedded clause is typed as an interrogative by a null operator in SpecForceP.2 Along the same lines, I would like to propose that, in Latin, subordinating conjunctions like cum ‘when’ and si ‘if ’ are merged in FinP (see also Krapova (2010: 1257) on the Bulgarian that-complementizer deto). Below, I will argue that LEF2 constituents are hosted in the CP-internal focus projection (Section 16.3.4). All this is schematically represented in (10). (10) [ForceP OP [TopP [FocP LEF2 [FinP Sub [TP ]]]]] In the following section, I will first elaborate on the interpretive characteristics of LEF2. Subsequently, I will present the results of a corpus study, which show that LEF2 is only productive in the earliest stages of the Latin language.
. Interpretation and diachrony of LEF .. LEF as left-peripheral presentational focalization ... The syntax of presentational foci In the literature, it has been proposed that broadly two types of foci should be distinguished (see esp. É. Kiss 1998), which I 2 On the clause typing operator in adverbial clauses, see Haegeman (, a, b).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert
will call ‘identificational’ and ‘presentational’. According to É. Kiss, the former are operators that obligatorily move to the clausal left periphery, whereas the latter remain in situ and are marked with a pitch accent. Presentational foci are characterized by the same author as non-contrastive, non-quantificational, and typically conveying new information. With Belletti (2001, 2004), I will assume that these apparently in situ presentational foci actually move to a specialized low focus projection at the edge of the vP phase. I will call this projection FocvP (see also Devine and Stephens 2006: 28). Moreover, in recent work, it has been pointed out that, in some languages, presentational foci do surface in the CP-domain, without being associated with the expected connotations of exhaustivity or contrastivity. Left-peripheral presentational foci have been reported for Russian (Bailyn 2003), Sicilian (Cruschina 2006), Modern Greek (Gryllia 2008), German, and Czech (Fanselow and Lenertová 2011). I will now give a brief overview of the most important characteristics of Latin LEF2. ... On the interpretation of LEF First, in the majority of cases, an LEF2 constituent conveys non-predictable, brand-new information. Consider for instance the sentence in (11). In this example, the city of Capua has not been mentioned before in the letter, and is thus newly introduced into the discourse. (11)
conueni [Capuami [cum ti uenissem [. . .]]], consules Capua-acc when I-had-come-subj consuls-acc I-met-with-pf multosque nostri ordinis. many-acc=and our-gen order-gen ‘When I had come to Capua, I had a meeting with the consuls and many colleagues of the senate.’ (= Cic. ad Att. 7.15.2)
Second, LEF2 is freely available in syntactic domains which are cross-linguistically known to disallow so-called ‘Main Clause Phenomena’, like adverbial clauses (see the examples in (4)–(6) and (11)).3 Third, there seem to be no restrictions on the category of the phrases that can undergo LEF2. Apart from DPs, we find fronted APs, PPs, and even CPs (not illustrated for reasons of space). In general, everything except for inflected verbs and markers of sentential negation can be found to the left of subordinating conjunctions. Fourth and finally, LEF2 can affect constituents which are clearly not referential or D-linked. A first example is given in (12), in which the nominal element of the idiomatic expression castra mouere ‘strike camp (lit. move camp)’ has undergone LEF2. In this example, castra clearly is not used to refer to some specific military camp. (12)
[cum ti mouere uellet]], [. . .]. Itaque [castrai camps-acc when move-inf he-wanted-subj prt ‘And when he then wanted to strike camp, . . .’ (= Anon. Bel. Afr. 6) 3 See Hooper and Thompson () and Haegeman (, a, b).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
Moreover, bare quantifiers like nihil ‘nothing’, omnia ‘everything’, and nemo ‘nobody’ are attested in an LEF2 position. An example is given in (13). (13)
[cum ti proficeret]], ui contra [Nihili nothing-acc when he-brought-about-subj force-abl against uim experiendum putauit. force-acc to-be-tried-acc he-thought-pf ‘When this brought about nothing, he thought that he should try his luck with brute force.’ (= Cic. Phil. 10.23)
The fact that these and other non-referential elements (like non-verbal predicates and secondary predicates) can undergo LEF2 strongly suggests that this phenomenon should not be taken to be a kind of topicalization (Cinque 1986) or scrambling (Diesing 1992): these types of movement typically affect (referential and/or specific) noun phrases. On the basis of these observations, I would like to propose that LEF2 is a type of presentational (i.e. non-quantificational) focalization. In Section 16.3, I will develop an analysis that explains how Latin presentational foci end up in the clausal left periphery rather than in the lower FocvP. Before doing so, I will look at the diachronic development of LEF2. .. Diachronic evolution: Decline of LEF In a large-scale corpus study, I have looked at the frequency of LEF2 in adverbial clauses introduced by the conjunctions cum ‘when, because, although’, si ‘if ’, and ut ‘so that, in order to, . . .’. Looking at adverbial clauses has a number of advantages: apart from the fact that these clauses are ubiquitous and easily retrievable in a corpus, they are also adjuncts and by this token strong islands (see Danckaert 2012: 140–2). It follows that we can be reasonably confident that in a sentence like (11), the fronted constituent Capuam has not been extracted from the cum-clause to a position in the left periphery of the main clause. Rather, it is moved to the C-domain of the embedded clause. The texts of the corpus that I have used (see Table 16.1) are chronologically organized in five different periods.4 (13)
I. II. III. IV. V.
Archaic Latin Classical Latin I Classical Latin II Late Classical Latin I Late Classical Latin II
first half of the second century bc first century bc first century ad first half of the second century ad second half of the second century ad
4 This subdivision into five periods is to some extent motivated by accidental time gaps between two texts in my corpus. Note that data from the early Latin period are unfortunately scarce: Cato’s De agricultura is the only substantial extant prose text.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert Table .. Diachronic evolution of LEF2 Author, work (date)
Frequency of LEF2 n=
I Cato, De agricultura (160 bc) II Cicero, Ad Atticum (68–43 bc) Anon. I, Bellum Afr. (c.40 bc) Anon. II, Bellum Hisp. (c.40 bc) Anon. III, Bellum Alex. (c.40 bc) Varro, Res rustica (36 bc) Vitruvius, De architectura (c.20 bc) III Velleius Pat., Historiae (ad 30) Columella, De agricultura (ad 40–50) IV Frontinus, Aq. + Strat. (c. ad 90) Plinius, Epist. + Paneg. (ad 90–110) Tacitus, Ann. + Hist. (ad 100–110) V Fronto, Epistulae (ad 150–170) Gaius, Institutiones (c. ad 170) Apuleius, Flor. + Mag. (ad 170–180) Total:
49/372 76/2337 11/141 11/125 5/163 88/783 99/1293 2/304 29/2289 5/522 0/1302 15/1281 10/384 13/1161 13/453
13.2 3.3 7.8 8.8 3.1 11.2 7.7 0.7 1.3 1.0 0 1.2 2.6 1.1 2.9
426/12910
Table 16.1 offers an overview of the quantitative data. The figures in the second column show the number of cum-, si-, and ut-clauses exhibiting LEF2, compared to the total number of those clauses. The rightmost column shows the relative frequency of LEF2. Although the limited number of authors per period makes it difficult to make strong diachronic claims (as one cannot exclude that certain tendencies are to be ascribed to stylistic preferences of individual authors rather than to genuine syntactic changes), it seems safe to conclude that LEF2 was mainly productive in the archaic and early classical period. Around the first half of the first century ad, it started to decline, to become almost completely obsolete in the early second century ad. A modest revival is seen in stage V, probably to be ascribed to imitation of the by then classical models of the pre-Augustan era (esp. Plautus, Terentius, Cato, and Cicero).5 5 In the philological literature, the style of Fronto and Apuleius is generally considered to be ‘archaizing’. Thus Goodyear (in Kenney and Clausen (eds.) : ) on Fronto: ‘[. . .] he certainly tried to exploit anew the latent resources of Latin literature, by going back beyond the stylists of the early Empire, and beyond Cicero and his contemporaries, to extract from the archaic writers whatever he might effectively use.’ Similarly, Conte (: ): ‘The limits and variations of the archaizing tendency are neatly symbolized by its triumphant leader, Marcus Cornelius Fronto.’ More or less the same can be said about Apuleius; Conte (: ): ‘Living in a period of enthusiasm for the archaic, Apuleius naturally shares the fondness of his contemporaries for obsolete words [. . .] and for archaic authors [. . .].’ The relatively high frequencies of LEF in these authors can therefore be interpreted as an (indirect) indication that this phenomenon was
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
In the remainder of this chapter, I will try to better understand this diachronic evolution. A necessary condition for this is of course a proper understanding of the initial and the final synchronic stages. .. A diachronic hypothesis Consider first the minimal pair in (15)–(16). Both of these little pieces of discourse have essentially the same structure. They consist of two sentences, the first of which contains a lexical item X (underscored), which is literally repeated in a preposed temporal adverbial clause in the second sentence. The basic scheme is twice something like ‘. . . X . . . . And when I say “X”, I actually mean “Y”.’ (15)
publice non inuitarunt. [Me [cum Mamertini me Mamertines-nom me-acc officially not invited-pf me-acc when dico]], leue est: [. . .]. I-say light-nom it-is ‘The people of Messana did not officially invite me. And when I say “me”, I consider this a light matter.’ (= Cic. Ver. act. sec. 4.25)
(16) Arrianus Maturus Altinatium est princeps. [Cum dico princeps], a-nom m-nom Altinates-gen is chief-nom when I-say chief-nom non de facultatibus loquor, [. . .]. not about means-abl I-speak ‘Arrianus Maturus is the most important man in Altinum. When I say “most important”, I am not referring to his wealth, [. . .].’ (= Pli. Epi. 3.2.2) In the example from Cicero (from 70 bc), the repeated element surfaces in a left-peripheral position, i.e. in an LEF2 position. In contrast, in the example from Pliny the Younger (c. ad 100), the repeated constituent appears postverbally (recall from Table 16.1 that LEF2 is (completely) absent in Pliny’s work). Given the quasi-identical discourse circumstances of the two examples, I will assume that an LEF2 phrase (15) and the postverbal constituent in (16) are functionally equivalent: both are presentational foci (on postverbal presentational foci, see esp. Section 16.4.2). In the literature, it has been observed that the frequency of INFL-final clauses decreases in the course of the first centuries ad (see Linde 1923; Koll 1965; Bauer 1995), which coincides with the period in which LEF2 became obsolete. In the remainder of this chapter, I will pursue the hypothesis that there is a non-trivial correlation a feature of the earlier stages of the Latin language. Finally, it comes as no surprise that Gaius, the third author of stage V, uses LEF less frequently than his two more literary contemporaries: this author wrote a technical treatise on law and did not make any attempt to imitate or emulate the style and language of earlier models.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert
between the loss of LEF2 and the increasing frequency of INFL-medial word order patterns. In Section 16.3, I will first try to explain how presentational foci could end up in the C-domain (where we do not immediately expect them), and in Section 16.4 I will return to the diachronic evolution.
. The derivation of LEF: vP movement and ‘smuggling’ .. Discourse-neutral word order in Latin Some introductory remarks are in order. I assume that in each language, there exists a ‘discourse-neutral’ word order, which is typically found in those sentences which can felicitously be uttered as an answer to the question ‘what happened?’6 Moreover, I will adopt the idea that, in each language, the discourse-neutral (or ‘basic’) word order is derived from the universal base by means of (a series of) movement operations (Kayne 1994; Biberauer and Roberts 2005; Hinterhölzl 2010). Finally, I will assume the Linear Correspondence Axiom (Kayne 1994), which says that in the base, specifiers universally precede heads, which universally precede their complements. There is nowadays some consensus that the discourse-neutral word order in Latin was (S)OV(Aux) (see esp. Devine and Stephens 2006: 79). This basic word order is characterized by two complement-head sequences, namely OV (17–18) and V-INFL (18). The latter can only be diagnosed in clauses with an ‘analytic’ verb form, like secuta est ‘has followed’ in (18), where the lexical root of the verb and the inflectional morphology are not realized on the same word.7 (17)
(18)
Caesar exercitum reduxit. Caesar-nom army-acc led-back-pf ‘Caesar led back his army.’ utilitas amicitiam secuta est. utility-nom friendship-acc followed-nom is ‘Advantage has followed friendship.’
(= Caes. Gal. 3.29)
(= Cic. Lael. 51)
For the purposes of this chapter, I will mainly concentrate on the derivation of the second of the two ‘head final’ orders. Concerning the first, I will just assume that the direct object undergoes ‘short movement’ to a fairly low position inside the verb phrase (see also Hróarsdóttir 2000; Holmberg 2000; Biberauer and Roberts 2005 among many others). In the following sections I will propose that the V-INFL order 6 I refer to Schweikert (: –) for extensive discussion of ways to establish which is the ‘basic’ word order in a given language. 7 Observe that despite the passive morphology and the BE-auxiliary, a predicate like sequor ‘I follow’ (perfect tense secutus sum ‘I followed’) is genuinely transitive (cf. the accusative-marked direct object amicitiam ‘friendship’ in () (see also Embick )).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
is derived through movement of a large verbal chunk to a high TP-internal functional projection. .. vP movement as a way to satisfy the EPP It is often assumed that T is endowed with nominal (N) and verbal (V) features, which need to be ‘checked’ by syntactic objects with matching nominal and verbal features (see for instance Chomsky 2001: 38).8 In the classical case (like in French), T’s verbal feature is checked by means of V-to-T movement (with ‘V’ for the complex V◦ /v◦ ), whereas its nominal feature is checked through movement of the subject DP to SpecTP. Since Chomsky (1981), the latter movement operation has been understood as being triggered by the EPP (‘Extended Projection Principle’), which can loosely be paraphrased as the requirement that every clause have a subject. From Biberauer and Roberts (2005, 2006) and Biberauer and Richards (2006), I will adopt the idea that the EPP requirement can be satisfied by movement of a large verbal projection, namely vP. This can be understood as a case of pied-piping: it is actually the subject DP, base-generated in SpecvP and endowed with the appropriate nominal feature, which is attracted to the middle field. As an accidental by-product, the subject pied-pipes the entire verb phrase, thus giving rise to the surface order V-INFL. Evidence for the claim that in Latin9 it is not a bare subject (typically a DP bearing nominative case morphology) that undergoes A-movement to satisfy the clausal EPP requirement comes from clauses in which a (derived) subject occurs postverbally. Two examples with a passive predicate are given in (19–20), in which the postverbal subject is underscored and the (lexical) verb is marked in boldface.10 (19)
si non siccentur bacae [. . .]. if not are-dried-subj berries-nom ‘if the berries aren’t dried’
(= Plin. NH 15.123)
est, sic [. . .]. (20) quomodo horum [. . .] habita ratio how those-gen held-nom respect-nom is so ‘like respect has been paid to them, so . . .’ (= Sen. Ben. 3.11.1) .. vP movement in Latin We are now in a position to explain how Latin sentences exhibiting the discourseneutral word order SOV(Aux) (as in (17–18)) are derived from the Universal Base. 8
See Alexiadou and Anagnostopoulou () for a slightly different view. That is to say, at least in the earliest stages of the language (see Section .). 10 On examples like (), exhibiting the order ‘passive past participle-subject-auxiliary’, see Biberauer, Holmberg, and Roberts (: , n. ). I will have nothing to say about the special properties of this particular word order pattern: this example was only chosen because it contains a postverbal subject which is clearly not extraposed or right-dislocated. 9
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert
First of all, I will assume that auxiliaries are base-generated in some functional head in the split TP, say T◦ . Furthermore, I will also assume that synthetic verb forms undergo V-to-T movement: given the rich agreement morphology of Latin finite verbs, this seems not an unlikely hypothesis. Second, along the lines of the discussion in the previous section, I assume that the entire vP moves to a position to the left of T◦ : this verbal category then contains (i) the verb’s arguments and (ii) the lexical verb or its trace.11 As outlined in the previous section, I assume that the trigger for this movement operation is the EPP feature with which, by assumption, each clause is endowed. There is reason to believe that the target of vP is not the specifier of the tense phrase, the head of which is occupied by the finite verb. Evidence for assuming a relatively high target position for the moved verb phrase comes from examples like (21), in which the direct object (in this case a CP) and the lexical verb passus ‘permitted, allowed’ are separated from the clause-final auxiliary sum ‘I am’ by the negator non ‘not’. (21)
ego [FP [vP [CP similem ti [meas ruinas] [CP quarumi similar-acc my-acc ruins-acc which-gen i-nom non tj sum]] [. . .]. totam urbem esse] passus]j whole-acc city-acc be-inf permitted-nom not I-am (lit.) ‘my destruction, similar to which I did not allow the entire city to be’ (= Cic. De domo sua 124)
Therefore, I propose that the nominal and the verbal features of the T-domain are not located on the same functional projection:12 rather, they are associated with two different functional heads, which are separated by at least one other functional head, namely NegP. I will not try to be very specific about the identity of the head endowed with the EPP feature: I will just call it FP. Perhaps this projection is to be equated to the dedicated subject position ‘SubjP’ proposed in Cardinaletti (2004) (see also Rizzi 2006), which is also situated in a high position in the split TP. The derivation of a Latin clause with a synthetic verb form would then look like (22), with movement of the V◦ /v◦ -complex to T◦ , and movement of the remnant vP to SpecFP. (22) [fp [vp DPs DPo tj][fº[epp][negp [tp [tº Vj [focvP tvP ]]]]]]
11 See also Haegeman (); Koopman and Szabolcsi (); Pearson (); Hróarsdóttir (, ); Mahajan (); Biberauer and Roberts (, ); Biberauer and Richards (), among others. 12 This complies with a common assumption in the cartographic tradition, which says that there is a one-to-one relation between syntactic features and functional heads (see for instance Shlonsky ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
Crucially, the landing site of vP is not only higher than negation, but also higher than the lower focus phrase FocvP.13 In the following section, I will explain why this is relevant. .. Smuggling: a way to avoid intervention The basic idea is that a presentational focus moves to the CP-internal focus position because it is itself contained in a larger category which by default moves to a position higher than the lower focus projection. This is more formally stated in (23). (23) In Latin, a presentational focus XP moves to FocP if XP is dominated by YP and YP c-commands FocvP. In our Latin case, YP in (23) stands for the A-moved vP. If the analysis developed above is on the right track, this would amount to a ‘smuggling’ derivation, whereby the large verbal projection, containing XP, moves past FocvP, thus avoiding an intervention effect that would arise if XP were to move on its own. The mechanism of smuggling as a way to avoid a minimality violation is defined by Collins (2005a: 97) as follows:14 Suppose a constituent YP contains XP. Furthermore, suppose that XP is inaccessible to Z because of the presence of W (a barrier, phase barrier, or an intervener for the Minimal Link Condition and/or Relativized Minimality), which blocks a syntactic relation between Z and XP (e.g. movement, Case checking, agreement, binding). If YP moves to a position c-commanding W, we say that YP smuggles XP past W.
The ‘smuggling’ derivation of an LEF2 configuration would be as in (24a). I assume that since an independent step in the derivation has made FocvP unavailable for a constituent to move into, the left-peripheral FocP becomes the closest potential Probe for a presentational focus. As a host for presentational foci, FocP is so to speak only a ‘second best’. (24) a. [ForceP [FocP XPi [FinP Sub [FP [vP ti tj][T º Vj [FocvP tvP ]]]]]] Crucially, vP movement is a precondition for LEF2. The ungrammatical derivations (24b, 24c) show that the presentational focus XP cannot reach the higher FocP by 13 A possible example where a postverbal presentational focus (which, as we will see in Section .., only becomes available in a fairly late stadium of the Latin language) is found lower than sentential negation is given in (ia). I assume that this example can be represented as in (ib).
(i) a. Myron [. . .] non inuenit heredem. Myron-nom not found-pf heir-acc ‘Myron [. . .] did not find an heir.’
(= Petr. Sat. )
b. [FP Myronk [NegP non [TP [T◦ inuenitj [FocvP heredemi [vP/VP ti tj tk ]]]]]]. 14 A note of caution is in order: I assume that ‘smuggling’ does not exist as a primitive operation; I consider the avoidance of an intervention effect to be an accidental by-product of two (or possibly more) independently motivated operations.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert
its own force, whether it moves ‘in one fell swoop’ (24b) or with an intermediate movement step to a functional projection in the middle field, say AgrOP (24c)15 . (24) b. * [ForceP [FocP XPi [FinP Sub [FP [T º Vj [FocvP [vP ti tj]]]]]]]
c.
*[ ForceP [FocP XPi [FinP Sub [AgrºP t'i [T º Vj [FocvP
[vP ti tj]]]]]]]
Consider for instance the example in (25a), in which a direct object has been moved to the left periphery of a conditional clause. The schematic derivation of this example is given in (25b). [si ti uidero]], [. . .]. (25) a. [Antoniumi Antonius-acc if I-will-have-seen ‘If I see Antonius, . . .’
(= Cic. ad Att. 14.19.4)
b. [ForceP [FocP Antoniumi [FinP si [FP [vP ti tj ][T◦ uideroj [FocvP [tvP ]]]]]]] Before turning to the diachronic analysis, two remarks are in order. First, the nonquantificational nature of presentational foci (É. Kiss 1998) allows LEF2 constituents to appear in the left periphery of clauses that generally do not allow for ‘Main Clause Phenomena’, like adverbial clauses (see Section 16.2.1.2). I refer to Haegeman (2009, 2010a, 2010b) for an explicit proposal as to why quantificational operators are disallowed in the left periphery of most adverbial clauses. Second, it should be noted that the present proposal is not compatible with a strictly derivational approach to information structure, as e.g. in López (2009). One could assume that ‘peripheral’ discourse-related movement operations take place later than core syntactic operations as A-movement (cf. Zubizarreta 1998: 29–33), so as to avoid a vP-internal phrase with a focus feature moving to FocvP as soon as the latter is merged, but I acknowledge that this is a potential problem. In the remainder of this chapter, I will concentrate on the diachronic development of LEF2. As hinted at in Section 16.2.3, I will propose that the loss of LEF2 goes hand in hand with an increased frequency of postverbal presentational foci, and thus with an increased frequency of non-INFL-final clauses.
. The loss of vP movement and its consequences .. Reanalysis of vP movement I would like to propose that the decline of LEF2 is to be ascribed to the loss of its conditio sine qua non, namely vP movement. Following Biberauer and Roberts (2005: 15 For some more case studies of derivations in which the operation of ‘smuggling’ plays a role, see Cinque (: –); Collins (a, b); Belletti (: ); Belletti and Rizzi ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
25), who discuss a very similar directionality shift that took place during the transition from Old to Middle English, I suppose that movement of a large verbal projection (vP) was reanalysed by the language acquirer as a series of independent movement operations of the verb and its arguments (as was proposed in Zwart (1993) for the OV-order of Dutch). Consider the derivations in (26). (26) a. [ForceP [FinP [FP [vP DPS DPO ti]j[T º Vi [FocvP tj ]]]]]
b. [ForceP [FinP [FP DPS [GP DPO ][T º Vj [FocvP tS tO tV ]]]]]]
In (26a), we see the by now familiar derivation that can give rise to LEF2: vP is piedpiped when the subject DP is attracted to SpecFP. On the other hand, in (26b), the subject moves on its own to the designated subject position, stranding the verb phrase, and the direct object moves to a position in the middle field, possibly a case position (like AgrOP). The crucial observation is that although both derivations yield the same linear output, their underlying structure is fundamentally different. It should be noted that, in the second derivation, the trigger for subject movement, namely the EPP requirement, is arguably stronger than the trigger for object movement. Actually, it is not entirely clear whether the object moves to check a(n abstract) case or agreement feature, or some definiteness feature. In any event, we predict that it is not unlikely that in a grammar without obligatory vP movement, object movement does not take place systematically. As I will show in the final part of the chapter, this is indeed the case. .. Postverbal presentational foci Latin never was a strict INFL-final language: the order Vfin -XP is attested from the earliest records onwards. However, there is an interesting diachronic development concerning the nature of the constituents that could appear in a clause-final position, which seems to support the analysis developed in this chapter.16 As pointed out in Devine and Stephens (2006: 119–23), in authors whose syntax was predominantly INFL-final (like Caesar), the majority of postverbal constituents are PPs, as in (27). (27) Exercitum reducit [PP ad mare]. to sea-acc army-acc leads-back ‘He led the army back to the sea.’
16
(= Caes. Gal. 5.23)
The upcoming discussion is based on Devine and Stephens (). See also Danckaert (: –).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Lieven Danckaert
Assuming that an argumental PP like the directional expression ad mare ‘to the sea’ in (27) is merged in some low position in TP (see Schweikert 2005: 123–9), the present account correctly predicts that it can appear postverbally. As shown in (28), we only have to assume that the base position of the PP, here labelled ‘DirP’ for ‘Directional Phrase’ is not included in the category that undergoes A-movement. (28) [FP [vP exercitumk tj tk ] [TP reducitj [DirP [PP ad mare] tvP ]]]. In a later stage, the possibility for direct objects to appear to the right of the inflected verb becomes productive as well: arguably, this is the stage in which vP movement is (gradually) lost. The postverbal objects typically come with a special pragmatic flavour: they are either destressed ‘tail’ constituents (in the sense of Vallduví 1992) or they are presentational foci. For reasons of space, I will only briefly discuss the latter. According to Devine and Stephens (2006: 128–36), the first type of direct object that appears as a postverbal presentational focus is a non-referential indefinite DP, often denoting an abstract concept, like fugam ‘flight’ in (29).17 (29) plerique [. . .] capessunt fugam. most-nom take flight-acc ‘The majority fled away.’
(= Liv. aUc 33.9.11)
This is exactly the type of DP that is a likely candidate to not undergo movement to the middle field: it is well known that indefinite nouns tend to stay inside the verb phrase, whereas definite or D-linked nouns are typically moved out of it (see esp. Diesing 1992; cf. also Bobaljik and Thráinsson 1998: 54 on the status of object DPs that undergo object shift in Icelandic). Postverbal direct objects are generalized only much later. However, we could assume that the increased frequency of postverbal direct objects has been a considerable factor in the eventual shift from an OV- to a VO-grammar.
. Conclusion In this chapter, I have looked at the syntax and the diachronic evolution of a particular strategy of left-peripheral focalization, which I referred to as LEF2. I have characterized LEF2 constituents as presentational foci, and I have argued that their left-peripheral position is parasitic on vP movement to the middle field. I suggested that the loss of this vP movement is the basis for loss of LEF2. In future research, it would be interesting to widen the empirical domain by looking at main clauses as well. However, it would presumably be very difficult to diagnose this phenomenon in non-embedded environments, as in most main clauses, the linear 17 Other examples with an abstract noun as a postverbal presentational focus include facit amicitiam ‘he makes friendship’ (Nep. Dat. ..) and derigebant cursum ‘they directed their course’ (Liv. aUc ..).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
The decline of Latin left-peripheral presentational foci
string provides us with no clues as to where the boundary between CP (the left periphery) and TP (the core clausal domain) is located: therefore, an LEF2 constituent in many cases would be indistinguishable from, for instance, an XP scrambled to the high TP. The only possibility would be to look at matrix constituent questions (given that XPs can occur to the left of question words). However, assuming that wh-phrases in matrix questions are hosted in FocP (see Rizzi 1997), it is not certain whether a landing site for an LEF2 constituent would be available.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan M O N T S E R R AT BAT L L O R I A N D M A R I A-L L U Ï S A H E R N A N Z
. Introduction As is well known, in Spanish and Catalan contrastive focus obtains whenever a word or constituent is moved to the left, to the Specifier of Contrastive Focus Phrase, bringing about subject-verb inversion. Despite the lack of prosodic clues, the identification of contrastive focus in Old Spanish and Old Catalan, see (1), may still be done on syntactic and semantic grounds.1 (1)
a. Bien sepa el abbat que buen galardón d’ello prendrá well know the abbot who good reward of-it take-fut.3sg ‘Well must the abbot know that he will be richly rewarded for it’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 386)
1 We agree with Cruschina (: , ) that it is difficult to tell when a constituent is topicalized or focused in the absence of prosodic clues. He states that ‘in the old texts there are no clear and unequivocal cases of CFoc-fronting, i.e. fronting of an element which is explicitly contrasted with another element present in the discourse’. Notice, however, that in some cases the discursive syntactic strategies used in old written texts (question-answer structures in dialogues, topic-comment configurations, etc.) can make it possible to identify contrastively focused constituents. Thus, for example, the conditional sentence Si del campo bien salides (‘If you leave the field successfully’), in (b), provides us with the appropriate semantic background to interpret the following sentence as a contrastive focus statement: grand ondra avredes vós (e non grand deshonor)(‘you will have great honour and not great dishonour’). Likewise, the previous question e on és? (‘Where is he?’), in (d), makes it possible to classify the PP a Sexona (‘in Sexona’) as a contrastively focalized constituent. As for the arguments in favour of regarding bien in (a) and poc in (c) as occupying the Specifier of Contrastive Focus, see Batllori and Hernanz (, ), where this subject is explicitly addressed.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
b. Si del campo bien salides, grand ondra avredes vós if from-the field well go-out2-pl great honour will-have you ‘If you leave the battlefield well, you will be honoured to the highest degree’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 3565) c. Mas poc profitara si érem enseyatz e no érem but little will-benefit if were-1pl taught and not were-1pl reemutz redeemed ‘But little would we benefit if we were taught and not redeemed’ (CICA: Vides de Sants Rosselloneses. Segle XIIIb: 14) d. E dix la dita Johana: “e on és?”, e respòs and said the mentioned Joan and where is-3sg and answered é lexat esta, mas nit hic serà . . .” Jacme: “a Sexona l’ James in Sexona him have-1sg left this, but night here (he)will-be ‘And the aforementioned Joan said, “And where is he?”, and James answered, “In Sexona have I left him, but tonight he will be here” ’ (CICA: Llibre de Cort de Justícia d’Alcoi (1263–5). Segle XIIIb: fol. 3r) The data in (2) show that clitic left dislocation differs from contrastive focus in that it exhibits a resumptive pronoun. (2) a. A los de mio Cid ya les tuellen el agua to them of my Cid already them take-away-3pl the water ‘My Cid’s men have already had the water taken away from them’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 661) b. . . . allò no u atorga la comunitat that not it give-3sg.ind the community ‘That is not given by the community’ (CICA: Francesc Eiximenis. Dotzè del Crestià. Segle XVa: Vol. I, 134) Apart from contrastive focus, Old Spanish and Old Catalan display another focalization strategy, which results in a weak or unmarked focus.2 Both contrastive focus and unmarked focus trigger subject–verb inversion, but they can be clearly distinguished by the divergent intonation patterns they exhibit. Benincà (2004) supplies ample evidence for unmarked focus in medieval Romance, as illustrated in (3). In addition, several recent works show that some Modern Romance languages still have an unmarked focalization pattern available, as in Sicilian (see Cruschina 2008,
2 See Martínez Gil () for Old Spanish, Batllori () for Old Spanish and Old Catalan, and Benincà () for Old Romance languages in general.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
2009, 2010: 247–60, and 2011: 95–132) and Sardinian (see Jones 1993; Mensching and Remberger 2010: 261–76; Paoli 2010: 277–8). (3)
a. Autre chose ne pot li roi trouver Old French another thing not can the king find ‘The king cannot find anything else’ (Mort le roi Artu: c.1230, 101; Benincà 2004) b. Mal cosselh donet Pilat Old Provençal bad advice gave Pilate ‘Pilate gave bad advice’ (Venjansa de la mort de Nostre Señor, 106; Benincà 2004) c. Con tanta paceença sofria ela esta with so-much patience suffered her this enfermidade Old Portuguese disease ‘She endured this disease with great patience’ (Diàlogos de Sao Gregório; Benincà 2004) d. Bon vin fa l’ uga negra Old Milanese good wine makes the grape black ‘Black grapes make good wine’ (Bonvesin da la Riva, 1280; Benincà 2004) e. Ciò tenne il re a grande maraviglia Old Florentine this has the king as-a great wonder ‘The king regards this as a great wonder’ (Il Novellino, II, 1300; Benincà 2004)
In addition to prosodic evidence, there are clear semantic differences between the two types of phenomena. Thus, in interpretive terms, weak focus does not necessarily entail a contrastive reading, whereas contrastive focus ‘identifies by contrastive exclusion the complement of the focus within the set of alternatives’ (Cruschina 2009: 24, n. 13). The structure of this chapter is as follows. Section 17.2 is devoted to the study of the nature and properties of weak or unmarked focus and its vitality in Old Spanish and Old Catalan. In Section 17.3, we examine the behaviour of these constructions in Modern Spanish and Modern Catalan, and we conclude that the two languages pattern differently in this respect. On the basis of this evidence, we hypothesize that there has been a grammatical change entailing the deactivation of weak focus in Modern Catalan. Finally, in the last section, we explore the asymmetry between Modern Spanish and Modern Catalan and provide an explanation to account for this asymmetry and also for the grammatical change undergone by Modern Catalan, in which QP-fronting can be related to the properties of the Polarity node.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
. Weak focus in Old Spanish and Old Catalan .. Old Spanish Weak focus involves leftward fronting of constituents which may have different functions, being either arguments or adjuncts. This is shown in (4) and (5), respectively. Object preposing (4) a. Los quinientos marcos dio Minaya al abbat the 500 marks gave Minaya to-the abbot ‘Minaya gave the abbot the 500 marks’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 1422) b. E este logar mostro dios a Abraham and this place showed God to Abraham ‘And God showed Abraham this place’ (CORDE: c.1275. Alfonso X. General Estoria. Primera parte: fol. 62v.) Adjunct preposing (5) a. en mano trae desnuda el espada in hand brings-3sg uncovered the sword ‘He comes with the unsheathed sword in his hand’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 471) b. De Ebron enbio Jacob so fijo Josep a Sychem por veer sos from Hebron sent Jacob his son Joseph to Sechem to see his ermanos brothers ‘Jacob sent his son Joseph from Hebron to Sechem to see his brothers’ (CORDE: c.1200. Almerich. La fazienda de Ultra Mar: 43) Weak focus fronting (WFF) also applies to various categories, such as adjectives (6), past participles (7),3 and quantifiers (8). Adjective preposing (6) a. Alegre es doña Ximena e sus fijas amas happy is lady Ximena and her daughters both ‘Happy is Lady Ximena and both her daughters too’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 1801) b. Non podian bever de las aguas de Mara(z) que amargas eran not could drink of the waters of Maraz that bitter were ‘They could not drink the water of Maraz because it was bitter’ (CORDE: c.1200. Almerich. La fazienda de Ultra Mar: 72) 3 For a detailed description of participle preposing in Old Spanish see Rodríguez Molina (: § ., ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz c. Dixieron: “Vivo es e sano, tu siervo nuestro padre” said-3pl alive is and sound your servant our father ‘They said, “Your servant our father is alive and well” ’ (CORDE: c.1200. Almerich. La fazienda de Ultra Mar: 56)
Past participle preposing (7) a. betatu lo ajat vetoed it-cl have-pres.subj ‘(that he) has forbidden it’ (CORDE: c.950–1000. Anonymous. Glosas Silenses: 295) b. lo que en muchos días recabdado non as it that in many days achieved not have ‘what you have not achieved in many days’ (CORDE: c.1330–43. Juan Ruiz, Archpriest of Hita. Libro de buen amor: 579) c. Benedicto sea Abraam de Dyos blessed be-subj Abraham by God ‘Blessed be Abraham [by God]’ (CORDE: c.1200. Almerich. La fazienda de Ultra Mar: 44) d. las puertas de mi casa aviertas las tenía the doors of my house open them-cl I-had ‘I had the doors of my house open’ (CORDE: 1246–52. Gonzalo de Berceo. Los Milagros de Nuestra Señora: 639c) Quantifier preposing (8) a. Fazerlo he; màs mucho me pesará si os bien do-it have-1sg but much me-cl upset-3sg.fut.ind if you well non fuere not go-3sg.fut.subj ‘I will do it, but it will greatly upset me if you did not succeed’ (CORDE: c.1400–98, Anonymous, El baladro del sabio Merlín con sus profecías) b. & por esto poco se mesclaua con la multitud and for this little refl mixed-3sg with the crowd ‘And for this reason he hardly mixed with the crowd’ (CORDE: 1379–84, Juan Fernández de Heredia, Traducción de Vidas paralelas de Plutarco, III) c. omes & men and algo something
mugieres que algo tienen de lo mio o que women that something have-3pl of the mine or that me deuen en Villa nueva to-me owe-3pl in Villa nueva
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
‘(there are) men and women that have something of mine or that owe me something in Villanueva’ (CORDE: c.1218, a 1300. Anonymous, Carta de nombramiento) .. Old Catalan Old Catalan examples show exactly the same behaviour as their Spanish counterparts.4 Focus-fronted elements can also be arguments, as in (9), or adjuncts, as in (10). Object preposing (9) a. Los chamins e les charreres pupliches fa trencar e the paths and the roads public makes-3sg damage and clodir close ‘He causes the public paths and roads to be damaged and closed’ (1190–1210. Greuges dels Templers de Barberà. Russell-Gebbett 1965: 85; Batllori, Iglésias, and Martins 2005: 158) b. D’aquesta misèria de la comunitat parla la of this misfortune of the community speaks-3sg the Escriptura en molts lochs Holy-Scriptures in many places ‘The Holy Scriptures speak of this community misfortune in many places’ (CICA: Francesc Eiximenis. Dotzè del Crestià. Segle XVa: Vol. I, 180) Adjunct preposing (10) a. E per aquells fou respost encontinent que de bon grat and for these was replied at-once that of good will li·n complaurien to-them-it would-fulfil ‘And they were answered at once that their wish would be happily attended to’ (CICA: Bernat Metge. Lo somni. Segle XVa: 158) b. tu veus bé que totes les oronetes en una manera fan lur manner make their you see well that all the swallows in a niu nest ‘You can see clearly that all the swallows make their nests in a particular way’ (CICA: Bernat Metge. Lo somni. Segle XVa: 116)
4 The only difference we can see between Old Spanish and Old Catalan is that the former displays interpolation, whereas this is attested very rarely in the latter. See Batllori, Iglésias, and Martins () for further discussion.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
Apart from their functional status, focus-fronted elements can belong to different categories, including adjectives, as in (11), past participles, as in (12), or quantifiers, as in (13). Adjective preposing (11) a. Just és e raonable que per aquell sia mantenguda e fair is and reasonable that by that be-subj maintained and ajudada assisted ‘It is fair and reasonable that for him it should be maintained and assisted’ (CICA: Epistolari de la València Medieval. I-4. Segle XIVb: Carta 125, línia 26) b. Foll és aquell qui s’ature e.l camí mad is that who stops-off on-the way ‘Mad is he who stops off on the way’ (CICA: St. Vicent Ferrer. Sermons. Segle XVa: iv-50) Participle preposing (12) a. Dit avem de la pràctica de les regles e mostrada avem spoken have of the practice of the rules and shown have manera amb la qual hom pot soure qüestions ab cascuna de manner with the which one can resolve problems with each of les regles the rules ‘We have spoken of the practice of the rules and shown how one can resolve problems with each of them’ (CICA: Taula general. Segle XIIIb: 420) b. pus vós li donàssets so que promés li avets so you him give this what promised him have ‘So you were to give him what you had promised him’ (CICA: Pergamins, processos i cartes reials. Segle XIVa: Doc. 163, línia 22) Quantifier preposing (13) a. Mon senyor e lo meu bé, molt me enuja la vostra my lord and the my good much to-me upset-3sg the your partida departure ‘My lord and my dear, your departure upsets me greatly.’ (CICA: Curial e Güelfa. Segle XVb)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
b. E ell testis dix: “(. . .) que massa vianda m’ avets and the witness said: that too-much food to-me have-2pl dada, la qual ne volria haver gitada, no que me’ given the which of-it would-like have thrown-up not that to-me n donets més of-it give-2pl more ‘And the witness said, “. . . That you have given me too much food, which I would fain have thrown up than had you give me more.’ (CICA: Procés criminal contra Antònia Marquès. Segle XIVb: fol. 2v) c. Menysprea la glòria d’aq(uest) món, que poch dura scorn the glory of this world which little lasts ‘Scorn the glory of this world, for it lasts but little’ (CICA: Doctrina pueril. 1. Segle XIVb: 13) .. Further properties of weak focus In addition to the properties discussed so far, there is crucial evidence to show that weak focus-fronted constituents pattern as focal elements. Note in this respect that adjacency is required between the fronted element and the finite verb. Consequently, the subject (when lexically realized) appears in postverbal position, as illustrated in (14). (14) a. [XP De Ebron] enbio Jacob so fijo Josep a Sychem =(5b) b. [XP Alegre] es doña Ximena =(6a) c. Mon senyor e lo meu bé, [XP molt] me enuja la vostra partida =(13a) Furthermore, resumptive pronouns never occur in object-fronting constructions, which rules out an analysis along the lines of clitic left dislocation. This is shown in (15). (15)
a. [NP Los quinientos marcos] dio Minaya al abbat =(4a) b. [NP este logar] mostro dios a Abraham =(4b) c. [NP Los chamins e les charreres pupliches] fa trencar e clodir =(9a)
Given the assumption that the constructions under study are focal in nature, we would expect them to only occur in matrix clauses. However, a potential problem arises with the examples in (16), which show that WFF is also attested in embedded sentences. (16) a. Juro ego . . . che . . . treva et paz tenré et a mos swear I that truce and peace will-have-1sg and to my òmens tener la mannaré men have it will-order-1sg
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz ‘I swear that I will keep truce and peace and that I will order my hosts to keep them, too’ (CICA: Jurament de pau i treva del comte Pere Ramon de Pallars Jussà al bisbe d’Urgell. Segle XIIa: 64) b. Dit m’ an que clams volets fer a la said to-me have-3pl that complaints want-2pl make to the cort d’aquest mal fet que ma filla n’ Anthònia vos court of this bad act that my daughter the Antonia you-cl ha fet have-3sg done ‘They told me that you wish to complain to the court of the evil things that my daughter Antonia did to you’ (CICA: Procés criminal contra Antònia Marquès. Segle XIVb: fol. 4r) c. De la reyna Dona Maria, nostra mare, volem aytant dir que, of the queen Lady Mary our mother want-1pl such say that, si bona dona havia e· l món, que ela ho era en if good woman have-3sg.imps in the world that she it was in tembre e en honrar Déu e en altres bones costumes que en fear and in honour God and in other good habits that in ela eren here were-3pl ‘Concerning our mother, Lady Mary, the queen, we wish to say that if there were a perfect woman in the world, it would be her, because she fears and honours God and has other good habits’ (CICA: Llibre dels fets del rei en Jaume. Segle XIVa: fol 4r) d. e dix- li que li retés l’ arcènich que venut li and said to-him that him returned the arsenic that sold him h(avi)a, cor mal li estava que ell, qui son amich era, lo had because bad him were that he who his friend was him aportàs a perill de perdre lo cors e’ l haver brought to danger of losing the body and the assets ‘And he told him to give back the arsenic he had sold him, because he considered it dreadful that he, who was his friend, had put him in danger of losing his body and his assets’ (CICA: Procés criminal contra Antònia Marquès. Segle XIVb: fol. 19r) e. Non podian bever de las aguas de Mara(z) que amargas eran not could drink of the waters of Maraz that bitter were ‘They could not drink that water because it was bitter’ (CORDE: c.1200. Almerich. La fazienda de Ultra Mar: 72)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
Yet, it must be taken into account that the kind of subordinate sentences that host weak focus exhibit properties of root sentences.5 That is, they instantiate cases of embedded clauses depending on verbs that select indicative like jurar (‘to swear’) (16a), or dir (‘to say’) (16b), peripheral adverbial clauses (16c, 16e), and non-restrictive relative clauses (16d). Now, returning to the examples in (14), some authors, noting the OVS order that they exhibit, concluded that medieval Romance displayed V2 properties and that medieval Romance languages were similar to Germanic ones in this respect. (17)
[SpecIP este lugari [[I mostró] [VP Dios [V [V tV ti ] a Abraham]]]] (Fontana 1993: 73)
Consistent with the cartographic approach developed from Rizzi (1997) onwards, which supplied a split CP domain to give an adequate explanation of the relationship between the syntactic representation of the sentence and its pragmatics and information structure, several scholars have regarded the examples under consideration as cases of focalization which can correspond to different types of foci:6 The hypothesis that the Focus Field can host various kinds of Foci is relevant in particular for medieval Romance languages. This area appears to be more easily activated in those languages than in modern Italian, so that we find there not only contrastive Focus or wh elements, but also less ‘marked’ elements (an identificational, informational or ‘unmarked’ focus, an anaphoric operator, or even elements with the pragmatic characteristics of a topic ‘put in relief ’) (Benincà 2004: 251)
This view, applied to Fontana’s example, would give the following representation.7
5 Quer (: –) also says that embedded QP-fronting is possible when the embedding predicate is an epistemic or assertive verb. See Haegeman () and subsequent works on the occurrence of root phenomena in embedded sentences. 6 There has been considerable debate concerning the categorization of medieval Romance as a V language. Cruschina and Sitaridou () and Cruschina (), among others, argue that most of them display V, which does not apply in such canonical V languages as Modern German. This implies that medieval Romance languages have a richer and more complex information structure than Modern Romance, which brings about different distributional orders, emulating V. In fact, Danckaert (: , ) poses a richer hierarchical structure for Latin in which there are Focus and Topic projections both in the high and the low left periphery, apart from a medial Scrambling projection. Note that Poletto (, ) also argues in favour of two peripheries: the high and the low one in vP. If we assume that the change from OV to VO took place within Latin (as in Vincent ) and that medieval Romance languages still have the same structure as Latin, the resulting superficial V order can be readily explained as an effect of information-structural considerations. In fact, Castillo-Lluch () shows that the OV order in charters fluctuates between an archaic Latin flavour and innovative information structure preposing. Extending this to Old Spanish and Old Catalan would cast light on the data and would surely help to distinguish among the different kinds of preposing (nominals, adjectives, past participles, quantifiers, etc.). We leave this for further research. 7 See Benincà (: ).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
(18)
Montserrat Batllori and Maria-Lluïsa Hernanz a. [ForceP [TopicP [FocusP . . . {Unmarked Focus [Spec este lugar i [UFocus’ [UFocus mostró]]]} [FinP . . . . [VP Dios [V’ [V’ tV ti ] a Abraham]]]]]] b. [ForceP [TopicP [FocusP . . .{Unmarked Focus [Spec d’aquesta misèria de la comunitat i [UFocus’ [UFocus parla]]]} [FinP . . . . [VP la Escriptura [V’ tV ti ]]]]]]
The analysis in (18) expresses the fact that this pattern of focus fronting obtains by the leftward movement of unmarked non-focal (or mildly focal) elements that would end up in a left-peripheral functional category hierarchically lower than contrastive focus.
. Weak focus in Modern Spanish and Modern Catalan .. A preliminary overview of the data The general picture sketched in the preceding section exhibits a regularity which sharply contrasts with the one offered by Modern Spanish and Modern Catalan data. A closer examination shows that weak focus is more restrictive in the latter than in the former. The following examples illustrate that, whereas Modern Spanish still allows these configurations (19), most of them are ungrammatical8 in Modern Catalan (20). (19)
a. Mucho me temo que la crisis no ha tocado fondo much me fear-1sg that the crisis not has touched base ‘I am very much afraid that the recession has not bottomed out yet’ b. Sus razones tendrá para actuar de este modo her reasons will-have to act in this way ‘She must have her reasons for acting that way’ c. Eso mismo pienso yo that same think-1sg I ‘I am of exactly the same opinion’ d. Buena tierra es esta good soil is this ‘This is a fertile land’
(Leonetti and Escandell-Vidal 2009)
matinat tu avui (20) a. ∗ Molt has much have-2sg wake.early you today tenir per haver-se enfadat b. ∗ Els seus motius deu the their reasons must-3sg have for have-refl angry 8 The examples in () are ungrammatical with an unmarked declarative intonation. Some speakers might also consider them unacceptable with a contrastive focus intonation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
c. ∗ Això mateix penso jo this same think-1sg I d. ∗ Bona terra és aquesta good soil is this It should be observed, though, that the examples in (21), in which the fronted constituent is a QP, are well-formed in Catalan. (21)
a. Algú hi trobarem, a la Rambla someone there will.find-1pl on the Rambla ‘We are sure to run into someone on the Rambla’
(Quer 2002: 256)
b. Gaires estudiants no deu haver aprovat, aquest professor many students not must have passed this professor ‘This teacher must not have passed many students’ c. Molts diners no han costat, aquestes arracades much money not have cost these earrings ‘They did not cost a lot of money, these earrings’ Interestingly enough, the examples in (21) are grammatical not only in Catalan but also in other Romance varieties in which Weak Focus Fronting (WFF) in general is not possible. Cruschina (2009: 22) notes that ‘existing analyses and new empirical data from other Romance languages show that, contrary to traditional assumptions, noncontrastive F[ocus] F[ronting] is widespread in Romance, especially with quantifiers and quantified expressions (i.e. QP-Fronting)’. Benincà (1988: 141–2) gives the following examples: (22) a. Niente concludi, stando in questo buco nothing conclude-2sg staying in this hole ‘You are not getting anywhere, staying in this hole!’ b. A nessuno nuoce, col suo comportamento to nobody harm-3sg with his behaviour ‘He’s not hurting anyone with his behaviour’ We will examine this issue more extensively in Section 17.4. .. Properties of weak focus in Modern Spanish and Modern Catalan: A comparative view Non-contrastive focus fronting, and particularly QP-fronting, has been regarded as a productive phenomenon in Italian (Cinque 1990), Portuguese (Ambar 1999), and
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
Catalan (Quer 2002).9 However, it does not follow a homogenous pattern within Modern Romance languages, which becomes obvious when the data given in (19) is compared to that in (20). Before taking into consideration the contrast between the examples in (19) and (20), it is worth summarizing the main properties of unmarked focalization in Spanish: I. Unmarked focalizations are incompatible with resumptive clitics (i.e. they diverge from topicalizations): (23)
∗ Sus
tendrá para actuar de este modo razones las in this way her reasons them-cl.acc.fem.pl will-have to act
II. Subject–verb inversion is compulsory: (24)
∗ Algo
estarán tramando estos niños something these children will-be plotting
III. Unmarked focalizations are mutually incompatible (25a), but they can co-occur with topicalized constituents (25b): mucho confía Juan (25) a. ∗ En alguien on somebody much relies John han dado los b. A este enfermo, pocas esperanzas le to this patient, little hope him-cl.acc.masc.sg have given the médicos doctors ‘To this patient, little hope has been given by the doctors’ IV. Contrastive focus and unmarked focus are mutually exclusive: (26) a. ∗ JULIA poco confía en los médicos (y no Pepe) Julia little relies on the doctors and not Pepe b. ∗ Poco JULIA confía en los médicos (y no Pepe) c. ∗ Poco confía JULIA en los médicos (y no Pepe) V. WFF is incompatible with emphatic polarity markers such as sí ‘yes’ or bien ‘well’:10 (27) a. ∗ Tiempo sí / bien habrá para pensar en esto yes / well will-be to think on this time b. Sí / bien habrá tiempo para pensar en eso yes / well will-be time to think on this ‘There will be time indeed to think about this’ 9 10
See Cruschina (: ) for more references concerning quantifier fronting. On emphatic polarity in Spanish, see Hernanz ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
Even though unmarked focus shares several properties with contrastive focus in Modern Spanish, as shown in the preceding examples, it lacks the contrastive value displayed by canonical focalization. As illustrated in (28), unmarked focus poco ‘little’, unlike its contrastive counterpart POCO, precludes the occurrence of a tag with the converse meaning. (28) a. Poco comió María ayer (# y no mucho) little ate Mary yesterday and not a-lot ‘Mary did not eat much yesterday’ b. POCO comió María ayer (y no mucho) little ate Mary yesterday and not a-lot ‘Little did Mary eat yesterday’ Consequently, the above contrast provides crucial evidence that the two kinds of focus must be regarded as separate phenomena. Now, addressing the comparison between Modern Catalan and Modern Spanish, a closer examination of the data suggests that they behave quite differently concerning WFF. Notice, first of all, that in Modern Catalan—as illustrated in (20)–(21)—Weak Focus Fronting is restricted to Quantifier Phrases (QPs). Moreover, Catalan sentences with weak fronted foci convey an added presuppositional value, which establishes a contrast with the preceding context. This implies that the examples in (21) do not count as neutral statements; rather, they cancel a reverse expectation that can be inferred from the previous discourse. Finally, as noted by Quer (2002: 265), Modern Catalan structures require emargination of postverbal elements: (29) a. Algú hi trobarem, a la Rambla =(21a) ‘We are sure to run into someone on the Rambla’ b. ?Algú trobarem a la Rambla
(Quer 2002: 264)
(30) a. Alguns llibres deu haver comprat, l’ Oriol some books must have bought the Oriol ‘Oriol must surely have bought some books’ b. ??Alguns llibres deu haver comprat l’Oriol
(Quer 2002: 264)
According to Quer, emargination prevents postverbal elements from being assigned nuclear stress since, if they were assigned it, there would be a potential conflict with the focal value of the fronted QP. Crucially, in Modern Spanish there is no requirement for the emargination of postverbal elements in WFF and, what is more, (31) and (32) show that it does not affect the grammaticality of the sentence. This poses a problem for Quer’s analysis because it erroneously predicts the ungrammaticality of the examples in (31a) and (32a), which lack emargination.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
(31)
a. Poco se imaginaba María lo sucedido little refl imagined Mary what happened ‘Little did Mary imagine what had happened’ b. Poco se imaginaba, María, lo sucedido
(32) a. Algo estarán tramando estos niños something will-be plotting these children ‘These children must be up to something’ b. Algo estarán tramando, estos niños The contrasts given in (31) and (32) are relevant empirical evidence to sustain the hypothesis that elements submitted to WFF occupy different positions in the Modern Spanish and Modern Catalan left periphery. Our claim is that the requirement for emargination is related to the presuppositional differences between the Catalan and Spanish data. More precisely, emargination is responsible for the added presuppositional import exhibited by Catalan sentences with QP fronting because the right dislocated element which is interpreted as topical is the one that allows the further partition of the information structure of the sentence. Bearing in mind that Old Catalan and Old Spanish, like other medieval Romance languages, displayed an unmarked focus position (Benincà 2004), the next section will examine the asymmetry between Modern Spanish and Modern Catalan to clarify the nature of the grammatical change that took place from Old Catalan to Modern Catalan in relation to the deactivation of the unmarked focus position that yielded WFF structures.
. WFF and polarity: Towards an explanation for the asymmetry between Modern Spanish and Modern Catalan On the basis of the data discussed so far, we are in a position to claim that Modern Spanish still maintains an activated unmarked focus position, which is hierarchically lower than contrastive focus and which serves as a landing site for WF-fronted constituents (see Benincà 2004: 256). (33)
{Topic . . . (CLLD). . . }{Focus . . . (ContrastFocus). . . (UnmFocus). . . }
In order to account for the asymmetry between Modern Catalan and Modern Spanish with regard to WFF, we hypothesize that the syntactic change that has taken place from Old to Modern Catalan can be understood as the deactivation of the unmarked focus position within the left-peripheral focus domain:11 11 We leave the explanation of the diachronic change to future research as it goes beyond the scope of this chapter.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
(34) [ForceP [TopicP [Contrastive FocusP [Unmarked focusP [. . . [FinP ]]]]]] Therefore, information structural mapping constitutes a point of variation between Modern Spanish and Modern Catalan, whereas this was not the case between Old Spanish and Old Catalan. Under the assumption that a syntactic projection for unmarked focus is no longer available in Modern Catalan, the following question arises: what position do the fronted elements subjected to WFF target? In view of the presuppositional interpretation displayed by constructions like (21), we postulate that the projection where these elements move in Modern Catalan is PolP, as (35) illustrates. (35)
[ForceP [TopicP [Contrastive FocusP [Unmarked focusP [PolP [FinP ]]]]]] ←−
This hypothesis offers an explanation for the ungrammaticality of the Modern Catalan examples in (20) and the grammaticality of those in (21). More precisely, the contrast between (20) and (21) clearly suggests that a ‘stronger’ trigger is needed in order to license WFF in Modern Catalan. Notice in this respect that the examples in (21) would be regarded as clearly inappropriate if they were expressed ‘out of the blue’, i.e. in a completely discourse-initial position, being independent of a previous context. The presuppositional import of WFF structures in Modern Catalan cannot be extended to their counterparts in Modern Spanish, which usually do not allow a contrast with the preceding discursive context. This can be shown by examining the minimal pair in (36). Whereas (36a) can be expressed without previous reference to the activity carried out by the subject (the children), (36b), as already noted, refutes or rectifies an affirmation uttered earlier about Lola. (36) a. Algo estarán tramando estos niños = (32a) something will-be plotting these children ‘These children must be up to something’ b. Algun error haurà comès, la Lola some mistake will-have made the Lola ‘Lola must have made some mistake’ Our analysis also predicts the different behaviour of Catalan and Spanish with respect to emargination in WFF constructions. Under the assumption that Modern Spanish exhibits a weak or unmarked focus projection which is no longer available in Modern Catalan, the contrasts illustrated in (31) and (32) can be easily accounted for. Note that WF-fronted constituents targeting this position do not qualify as canonical contrastive elements in Spanish; hence, they are perfectly compatible with non-right dislocated elements receiving focus stress, unlike their Catalan counterparts. It is worth noting that the analysis in (35) connects with the proposal put forward by Leonetti and Escandell-Vidal (2009: 155) to account for WFF structures. These authors claim that this type of focalization, which they label Verum Focus Fronting
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
(VFF), ‘triggers association of focus with sentence polarity’. According to them, ‘in Spanish the most natural paraphrase of a VFF consists in a construction in which the propositional content is embedded under the affirmative particle sí “yes” or under the adjectives seguro “sure” or cierto “true”.’ They illustrate their assumption with the following paraphrases: (37) a. Algo has visto → {Sí / seguro} que has visto algo something have seen yes sure that have seen something ‘You have seen SOMETHING’ → ‘Yes / surely you have seen something’ b. A alguien encontrarás → {Sí / seguro} que encontrarás a alguien to someone will-find yes sure that will-find to someone ‘You will find SOMEONE’ → ‘Surely you will find someone’ c. Miedo me da pensarlo → {Sí / es cierto que me da fear I-cl.dat gives think-it → yes is true that I-cl.dat gives miedo pensarlo fear think-it ‘Afraid as I am to think so’ → ‘Yes / It’s true that I’m afraid to think about it’ Setting aside the issue of the degree of adequacy12 of the paraphrases adduced in (37), they offer an explanation that fits with a picture where WFF properties crucially rely on polarity. The analysis proposed in (35) is also supported by diachronic evidence. Configurational changes undergone by medieval languages correlate with changes in the Polarity node.13 Both Old Spanish and Old Catalan displayed double negation. Spanish lost this option by the fifteenth century, whereas it continues to be present in Catalan. (38) Que los descabecemos nada non ganaremos that them cut-head nothing not win-1pl ‘Even if we cut off their heads, we will gain nothing’ (CORDE: c.1140. Anonymous. Poema de Mio Cid: v. 620) (39) en nula guisa porà, e si fer-ó volrà, res no in no manner will-be-able and if do-it will-want thing no valrà be-worth-3sg.fut ‘He won’t be able to do it anyway, and if he wants to do it, it won’t be worth doing it’ (CICA: Segle XIIIa. Usatges de Barcelona)
12 Although some Modern Spanish WFF structures admit these types of paraphrases, it is not the case in all the Modern Spanish examples of WFF. This may be due to the fact that WFF configurations are not homogeneous in Modern Spanish. We leave this aspect for further research. 13 See Martins (, , ) for an account of the strong/weak nature of åP and its correlation with scrambling, VP-Ellipsis, and clitic placement.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Weak focus and polarity: Asymmetries between Spanish and Catalan
Furthermore, the possibility of attributing the syntactic change we are considering to the properties of the Polarity node in Modern Catalan is supported by an important diachronic fact: Catalan has shown a high tendency to grammaticalize fronted elements as polar markers. Adverbs and adjectives such as poc ‘little’, pla ‘flat’, prou ‘enough’, and bé ‘well’ behave like negative and positive emphatic polarity markers in modern-day usage (see Batllori and Hernanz 2008/2009, 2013). In contrast, in Modern Spanish the only element that has grammaticalized as an emphatic polarity marker is bien. (40) a. En Pere prou que ho deia → sí que ho deia the Peter enough that it said yes that it said ‘Peter said it indeed’ b. Poc ho farà la Maria → no ho farà little it will-do the Mary not it will-do ‘Mary won’t do it’ c. La Maria pla que ho farà → no ho farà14 the Mary flat that it will-do not it will-do ‘Mary won’t do it’ d. Bé hi ha anat a la biblioteca. No l’ has trobat? → sí well there have gone to the library not him have met yes que hi ha anat that there have gone ‘He has gone to the library indeed. Didn’t you run into him?’ Following Batllori and Hernanz (2008, 2009, 2013), we assume that the elements alluded to are base-generated in PolP, as illustrated in (41). (41) [. . . [FocusP béi , proui / poci , plai [PolP ti [FinP . . . ]]]]15 Further evidence that Modern Catalan displays a strong polarity projection is provided by the examples in (42), where two positions for negation are attested. As discussed in Hernanz (2007: 128), this possibility is totally banned in Modern Spanish. (42) No que no ha vingut la Lola not that not has come the Lola ‘But Lola did not come’ These divergent properties of PolP in Catalan and Spanish bring about leftward fronting phenomena which fall under different patterns in each language. While contrastive focus fronting is an available strategy in both languages, Modern Catalan 14
See Rigau (). For a detailed explanation of the analysis instantiated in (), see Batllori and Hernanz (/ and ). 15
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Montserrat Batllori and Maria-Lluïsa Hernanz
displays fronting to PolP (in the case of QP-fronting), whereas Modern Spanish exhibits WFF.
. Conclusion This chapter explores the homogeneous pattern of Old Spanish and Old Catalan with respect to WFF and examines the asymmetry between Modern Spanish and Modern Catalan with respect to this phenomenon. Whereas Modern Spanish still exhibits WFF, Modern Catalan only allows QP Fronting, the licensing of which is directly related to a presuppositional interpretation. Additionally, Modern Catalan displays double negation, has grammaticalized many adverbs as emphatic polarity markers (either negative or affirmative), and allows for two negations (no que no). This leads us to propose that the above-mentioned asymmetry is related to a deeper generalization linked to the behaviour of the two languages. More precisely, we suggest that Modern Catalan QP Fronting is hosted by PolP, which is ‘strong’ enough to attract quantifiers, and also that the syntactic change that took place from Old Catalan to Modern Catalan involves the deactivation of the Unmarked Focus Projection.
Sources (CICA) Corpus Informatitzat del Català Antic, J. Torruella (dir.), with the collaboration of Manuel Pérez Saldanya, Josep Martines and Vicent Martines.
(CORDE) Corpus Diacrónico del Español:
Acknowledgements We acknowledge the following financial support: FFI2011-29440-C03-01 (Ministerio de Educación y Ciencia / FEDER) and 2009SGR1079 (Generalitat de Catalunya) for M. L. Hernanz, and FFI2011-29440-C03-02 (Ministerio de Educación y Ciencia / FEDER) for M. Batllori. Previous versions of this chapter were presented as a poster at the 12th Diachronic Generative Syntax Conference (Cambridge University, Queen’s College. 13–17 July, 2010), and as a talk at the VIII Congreso Internacional de Historia de la Lengua Española (Universidad de Santiago de Compostela. 17 September, 2009) and the 18 Congreso de la Asociación Alemana de Hispanistas. Sección 13. Escorados a la izquierda: dislocaciones y frontalizaciones del español antiguo al moderno (Universität Passau. 23–26 March, 2011). We thank the audiences at these conferences for their comments and suggestions.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German ROLAND HINTERHÖLZL
. Introduction The chapter addresses the issue of whether word-order properties can be reduced to interface properties. Dispensing with the Head Complement parameter (cf. Kayne 1994; Chomsky 1995b), the unmarked order in a language is not a basic property any more and the question arises as to which other properties in the grammar word-order properties can be related to. I will argue that word-order properties follow from the interaction between prosodic and information-structural conditions. In particular, I will look at wordorder variation in Old High German (OHG), and show that the factors that determine this variation are either prosodic or information-structural in nature. .. Word-order variation in older Germanic It is well known that the older Germanic languages showed greater freedom in word order than their respective modern descendents. In this respect, mixed word orders are of special interest since they pose a challenge for accounts based on the Head Complement parameter. Next to mixed word orders in modern Yiddish as in (1) (Diesing 1997), we find mixed OV/VO orders in the older stages of all Germanic languages, as is illustrated for Old English (OE) in (2) (Pintzuk 1999), for Old Icelandic (OI) in (3) (Hróarsdóttir 2000), and for OHG in (4) (taken from the Tatian translation). (1)
a. Maks hot nit gegebn Rifken dos bukh Max has not given Rifken the book b. Maks hot Rifken dos bukh nit gegebn Max has Rifken the book not given ‘Max has not given Rifken the book’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
(2) a. þæt he his stefne up ahof that he his voice up raised ‘. . . that he raised up his voice’ b. þæt ænig mon tellan mæge ealne one demm that any man relate can all the misery ‘. . . that any man can relate all the misery’ (3)
a. eftir það þeir höfðu eplið eteð after that they had apple-the eaten ‘after they had eaten the apple . . .’ b. að hann haði edið kjotið that he had eaten meat-the ‘. . . that he had eaten the meat’
(4) a. thaz then alton giqu&an uúas that to-the old ones said was ‘what was said to the old ones’ b. thaz gibrieuit uuvrdi al these umbiuuerft that listed was-sbjv all this mankind ‘. . . that all this mankind was listed’
(T64, 13a)
(T35,9)
As the examples in (2b) from OE and (4b) from OHG clearly show, we do not only find a combination of pure head-final (OV) word orders, as in (2a) and in (4a), and pure head-initial (VO) orders, as in (3b), in the same text, but also mixed word orders within the same sentence. To take an example from OE, 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. While mixed word orders in Yiddish, Old English, and Old Icelandic have been and are subject to thorough investigation and heated debates concerning their correct analysis (cf. i.a. Diesing 1997 vs. Vikner 2001 for Yiddish, and Roberts 1997 vs. Pintzuk 1999 for OE), the discussion concerning older stages of German has been non-existent, owing to the—as I will show—incorrect assumption that OHG was already an OVlanguage, albeit one permitting a high degree of extraposition. While indisputable VOfeatures in OHG have often been relegated to Latin influence, I will show on the basis of the Tatian translation that these features belong to an independent OHG system. .. Outline of the chapter Focusing on English and German, the above data raise the following questions: A) Were OHG and OE basic OV or basic VO languages? B) Do mixed word orders call for the presence of two grammars? C) How can we characterize OV and VO languages in the absence of the Head Complement Parameter?
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
In the following sections, I will try to provide interesting new answers to these questions. In particular, I will show that word order in OHG was determined by prosodic and information-structural (IS) requirements. Furthermore, I will argue that mixed word orders can be effectively accounted for by assuming a VO-base plus leftward movement triggered by licensing considerations. The difference between OVand VO-orders will be relegated to spell-out options which are taken to be fixed by interface conditions.
. Word-order variation and the Head Complement parameter Traditionally, word-order variation has been accounted for by assuming a basic word order—taken to be fixed by the Head Complement parameter—and by assuming additional rules like extraposition, heavy NP-shift, and the like to derive marked word orders from the unmarked base order. In this section, I will show that such accounts are not able to account for the word-order variation found in OHG. .. OV grammar plus extraposition On the assumption that OHG was an OV language like modern German, the simplest way to account for data like (4b) is to assume that the subject is extraposed from a position preceding the verb cluster. Note that modern German does not allow for the extraposition of non-complex arguments. As is illustrated in (5), DP arguments must be modified or conjoined in order to appear at the right edge of the clause in modern German. Furthermore, note that nominal, adjectival, or prepositional predicates cannot be extraposed in modern German either, as is illustrated in (6a–c) respectively. (5)
a. Auf Gleis 5 fährt ein der Interregio nach Straubing on platform 5 comes in the regional train to Straubing ‘On platform 5, the regional train to Straubing is arriving’ b. ?? Auf Gleis 5 fährt ein der Interregio on platform 5 comes in the regional train c. Es sind eingeladen Peter, Hans und Sabine. it are invited, Peter Hans and Sabine ‘There were invited Peter, Hans and Sabine’ d. ?? Es ist eingeladen der Präsident it is invited the president
(6) a. ∗ Er hat ihn genannt einen Idioten he has him called an idiot b. ∗ Er hat den Hund geschlagen tot he has the dog beaten dead
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl c. ∗ Er hat die Vase gestellt ins Regal he has the vase put into the shelf
In OHG, on the other hand, light arguments can be found in considerable numbers in postverbal position in embedded clauses. Moreover, nominal and adjectival predicates predominantly appear in postverbal position in the OHG Tatian. Since this text constitutes an interlinear translation from Latin, it is important to point out that these features of OHG also appear independently of or, often, in contrast to the Latin original. This is illustrated for arguments in (7) and for predicates in (8). Thus, these features cannot be relegated to Latin influence and must be taken to express genuine OHG properties. (7)
Latin
OHG
a. ut in me pacem habeatis so-that in me peace.acc have-you ‘so that you may have peace in me’
thaz in mir habet sibba so-that in me you-have peace (T 290, 8)
b. & qui demonia habebant and who demons they-have ‘and those that have demons’
inti thie thár hab&un diuual and those that have demon (T 59, 1)
(8) a. cui nomen simeon whose name Simeon ‘whose name was Simeon’
thes namo uuas gihezzan Simeon whose name was called Simeon (T 37)
b. Beati misericordes salige sint thiethar sint miltherze blessed mild-hearted blessed are those-who are mild-hearted ‘Blessed are those who are mild-hearted’ (T 60, 12) In conclusion, if OHG is assumed to have been an OV language, we have to acknowledge the existence of extraposition operations that are quite different from those observed in modern Germanic OV languages like German and Dutch. I argue that, instead, these properties speak in favour of the presence of a VO-grammar in OHG. Similar arguments have been made by Pintzuk (1999) to show that OE must have had a VO-basis. To account for undeniable OV-properties like the preverbal occurrence of verbal particles and the presence of verb clusters of the form V2 V1, with V2 representing the verb selected by V1, Pintzuk proposed the parallel presence of both an OV and a VO base in OE, an approach which has come to be known as the ‘Double Base Hypothesis’ (DBH). .. An account based on the Universal Base hypothesis Given that there is good evidence for the presence of a VO grammar in OHG, the question arises if we also have to assume the presence of an OV grammar to account for OHG’s OV-properties.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
There is a simpler alternative to assuming the presence of two grammars or two divergent parameter settings. Roberts (1997) showed that all the various mixed word orders found in OE can be derived from a universal VO-base plus optional movement operations known to exist in modern German and Dutch. These movements are given in (9). (9) a. licensing movement of arguments into a Case position (cf. Zwart 1993 for Dutch, Hinterhölzl 2006 for German) b. licensing movement of verb particles into the specifier of a low Aspect position (cf. Hinterhölzl 2006) c. licensing movement of predicative elements into a Predicate phrase (cf. Koster 1995 for Dutch, Hinterhölzl 2006 for German) There is one major drawback to Roberts’s (1997) proposal: movement operations that are obligatory in the modern Germanic OV-languages were assumed to be optional in OE to account for mixed word-order patterns. Secondly, Roberts (1997) did not provide any motivation as to when these optional movements did or did not apply. These problems are addressed in Biberauer and Roberts (2005). They propose a complex system of licensing, where the EPP-feature of a higher head can be satisfied by the minimal category targeted by the operation of Agree or by the containing category (a case of pied-piping). .. An approach in terms of variable spell-out The proposal that I am going to make is in line with Roberts (1997), and Biberauer and Roberts (2005). Starting from a universal VO basis, I assume the obligatory licensing movements given in (9). Mixed word orders in this approach are taken to result from spell-out options which are fixed by interface conditions. To provide a simple example, superficial OV-order is derived by obligatory movement of the object into a Caselicensing position in the middle field and spell-out of the higher copy, while superficial VO-order is derived by its obligatory movement into a Case-licensing position and spell-out of the lower copy in vP. This is illustrated in (10a) and (10b) respectively. (10) a. [CP [IP DO [vP V DO]]] b. [CP [IP DO [vP V DO]]] While the derivation in (10a), in which the lower copy is deleted, is uncontroversial, the derivation in (10b) with the spell-out of the lower copy is non-standard and needs additional justification. First, note that the spell-out of the higher copy is in fact based on the stipulation that features are only checked on the remerged copy, but not on the copy in the position prior to movement. Since movement and checking occurs to get rid of uninterpretable features, it is necessarily the lower copy that needs to be deleted at PF, according to
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
the account of Nunes (2004). Chomsky (1993), however, originally proposed that, in a checking operation, a feature is checked (and deleted) in all occurrences of the remerged copy. Second, note that the copy that is interpreted at LF and the copy that is interpreted at PF need not be the same. The simplest account of reconstruction effects and, likewise, the best argument for the existence of copies assumes that a moved constituent can be interpreted in its checking position at PF, but in its base position at LF. Hence it is generally accepted that in (10a)—where the higher copy is interpreted at PF—the lower copy may be interpreted at LF. (10b), on the other hand, should constitute a case in which the lower copy is interpreted at PF and the higher copy should be interpretable at LF. Incidentally, there is empirical evidence from scopal interaction between arguments and adjuncts for the assumption that arguments in English are spelled out in a position that is lower than their LF-position. Note first that arguments in German undergo scrambling in order to bind a pronoun contained, for instance, in a temporal adjunct. This is illustrated in (11). (11)
a. ∗ Hans traf an ihremi Geburtstag jedesi Mädchen birthday every girl Hans met on her b. Hans traf jedesi Mädchen an ihremi Geburtstag on her birthday Hans met every girl ‘Hans met every girl on her birthday’ c. John met everyi girl on heri birthday
In (11a), binding is impossible since in the unmarked order temporal adjuncts precede direct objects. However, when the direct object is scrambled across the adjunct as in (11b), binding of the pronoun in the adjunct by the argument quantifier becomes possible. (11c) displays the parallel state of affairs in English. Note that this is unexpected under the assumption that adjuncts are adjoined to the vP or are introduced in the extended projection of the verb, as proposed in Cinque (1999). Note furthermore that QR of the direct object does not constitute a solution to this problem since it would necessarily lead to a weak cross-over effect. A solution to this problem is provided in Larson (1988), who proposes that event-related adjuncts are contained in vP-shells below the base position of the direct object. This approach, however, fails to account for the comparative dimension and cannot explain the adjunct placement facts in German (cf. Hinterhölzl 2009a). Alternatively, I adopt Cinque’s (1999) approach and follow Barbiers (1995) in assuming that the postverbal position of adjuncts is derived via vP-intraposition. Given that event-related adjuncts introduce predicates on the event-argument of the verb (cf. Davidson 1966; Parsons 1990), I propose that these adjuncts constitute separate phases from the projections of the verb. This will become important when we talk about prosodic domain formation in Section 18.4.1.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
In such an approach, one can assume that the direct object in English undergoes scrambling like the direct object in German does, but that it is the lower copy in the vP that is spelled out.1 .. Mixed word orders and stylistic preferences Traditional grammarians have pointed out that word order in older Germanic is less fixed than in their modern varieties and argued that word-order preferences are due to a large degree to stylistic factors. Most notable among these is Behaghel’s (1932) statement of the Law of Growing Elements. Behaghel observed that pronouns and unmodified nouns tend to precede the verb, while modified nouns, PPs, and other heavy material tend to follow it. This gives rise to the generalization in (12). (12)
Light elements precede heavy elements in OE, OI, and OHG. (Behaghel 1932)
The statement in (12) raises the question of what light means in this context. The first interpretation is that light in (12) 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 factors and information-structural factors play a major role in determining word order in OI and in OE. In particular, Hróarsdóttir (2006) reports that both factors play a role in OI word order but concludes that prosodic weight was the decisive factor in OI. Furthermore, Taylor and Pintzuk (2008) argue for the relevance of both factors for word-order variation in OE and show that the two conditions, though overlapping, are independent of each other. In the following section, I will show that these factors also govern word order in OHG.
. Prosodic and information-structural constraints in OHG An important observation about word order in OHG is that pronouns and verb particles do not appear after the selecting verb, that is, in their presumed base position, abstracting away from the effect of V2, while PP-adjuncts and PP-arguments appear predominantly in postverbal position. This property can be related to the Law of Growing Elements, or better related to a prosodic condition which requires that light, non-branching constituents precede the verb, but heavy constituents—that is, phrases containing three words and more—follow the verb. On the other hand, a careful investigation of the information-structural contribution of arguments and adjuncts in their context gives rise to a different generalization. One basic notion in information structure is the distinction between focus and 1 This analysis is supported by vP-topicalization effects discussed in Hinterhölzl ().
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
background. Interlocutors make assumptions about the shared information in the conversation (also called common ground) and tailor their utterances according to what they believe is already known to the hearer (background) and what provides new relevant information (presentational focus). It turns out that that in mixed word orders in OHG, the verb serves to separate the background domain from the focus domain in the utterance. This is illustrated in (13) (cf. Hinterhölzl 2010; Petrova and Hinterhölzl 2010). The generalization in (13) will be slightly refined below. (13)
C background V focus
If we look at the role of focus in languages, at least three types of focus must be distinguished: wide and narrow presentational focus, as illustrated in (14a, 14b), and emphatic focus or contrastive focus. In general, focus indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions (Krifka 2008). In the case of an information focus, the alternatives are relatively unrestricted by the conversational background. With contrastive focus, a speaker specifically indicates that he considers relevant an alternative distinct from another alternative already under discussion. (14c) illustrates a case in point. With emphatic focus, the speaker indicates that he is positively or negatively surprised by the relevance of a certain alternative that may be already given or new. In intonational languages like German and English, emphatic focus is expressed by an extra high pitch tone. (14) a. What did John do? (broad presentational focus) John [gave a book to Mary] b. What did John give to Mary? (narrow presentational focus) John gave [a book] to Mary c. John gave Mary [a BOOK], not a pen (contrastive focus) In conclusion, similar to what is reported about OI and OE, both prosodic and information-structural constraints are operative in OHG. The important question now is how these conditions can be defined and how they interact with each other. This is the question to which we will turn in the following section. .. On the interaction between IS and prosody in OHG In this section, I report on the results of a small-scale empirical investigation based on the deviations from the Latin original listed in Dittmer and Dittmer (1998). For this investigation, I assume that non-branching constituents count as light and that branching constituents count as heavy. I investigated the information-structural role of constituents violating the Law of Growing Elements. Of particular interest therefore were non-branching constituents in the postverbal field and heavy, branching constituents in the preverbal field.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
A) Cases of preposing: in total, there were 138 cases in which a constituent in the postverbal field in Latin was moved into the preverbal field in OHG. Of these, 102 involved pronominal subjects and objects. Since pronominal constituents are discourse-given and light, their placement follows from Behaghel’s Law as well as from the generalization in (13). Thirty cases involve a nominal subject, among which 23 are branching, and the rest are biblical names like Jesus, Johannes, etc. Six cases involve a nominal object, among which three constituents are branching. Investigating the cases of preposed branching subjects and objects, I found out that all of them are discourse-given. Typically, these noun phrases involve a demonstrative pronoun, which in the absence of a grammaticalized determiner, indicates that the nominal in question is taken to refer back to a referent introduced in the previous discourse. B) Cases of postposing: Dittmer and Dittmer (1998) only list ten cases in which an element from the Latin middlefield appears postverbally in OHG. Obviously, this low number is due to the rare presence of a middlefield in Latin. However, it is interesting to note that seven out of these ten cases involve a light element in our terms. Two examples of this type are given in (15) and (16). When we interpret these sentences in their context, it turns out that both elements are narrowly focused. (15)
thisu sprahih iu thaz in mir habet sibba in therru weralti this I-tell you that in me you-have peace in the world habet ir thrucnessi you-have unrest ‘This I tell you, that in me, you have peace; in the world, you have unrest’ (T 290, 10)
(16) bidiu uuanta iogiuuelih thiedar sih arheuit uuirdit therefore everyone who(ever) that refl lifts-up will-be giotmotigot inti therdar giotmotigot sih wirdit arhaban humiliated and the-one humiliates refl will-be uplifted ‘therefore everyone that exalts himself will be humiliated, but the one who humiliates himself will be uplifted’ (T 195, 16) Let us briefly discuss these cases in turn. (15) involves a contrastive statement, in which the preverbal PPs ‘in me’ and ‘in the world’ function as contrastive topics (cf. Büring 1997; Frascarelli and Hinterhölzl 2007) and the postverbal DPs provide the relevant alternative, both constituting new information in the overall context. The most natural reading of the relevant sentence in (16) is ‘the one that humiliates HIMSELF will be lifted up’. In modern German, the reflexive, incapable of
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
carrying main stress, would be reinforced with the particle selbst, which carries heavy stress. In OHG, it seems to have been sufficient to move the reflexive into the postverbal domain in order to indicate a narrow focus reading. C) contrastive focus: Looking closer at the deviations from Latin in the OHG Tatian text, it turns out that there is another class of elements that regularly appears in a preverbal position in OHG. Even heavy constituents, including modified DPs and PPs, appear left-adjacent to the verb, when they are contrastively focused, as is illustrated in (17). (17)
niuuizze íz thin uuinistra/ uuaz thin zesuua tuo neg-know it your left-(hand) what your right-one does ‘your left hand should not know what your right one is doing’
(T 67, 5)
Thus, we have to adjust the characterization of the interaction between word order and IS given in (13). Taking into account the role of contrastive focus (CF), the generalization in (18) emerges. (18)
C background CF
V presentational focus
The generalization in (18) raises the following questions: A) Why should background information have to precede the verb? B) Why should new information have to follow the verb? C) Why can heavy (branching) constituents that belong to the background or are contrastively focused precede the verb? In Section 18.4, I will argue that (18) follows from the way in which information-structural categories are made visible at the interfaces. .. On the nature of the prosodic factor Above I have simply assumed that a non-branching constituent is prosodically light and therefore I have treated all branching constituents as prosodically heavy. Modern English data show that the overall picture is a bit more complex than this. It is well known that the English middle field—unlike the German middle field, which allows for heavy constituents—is restricted to light adjuncts only. This is illustrated in (19). (19)
a. John (very) carefully read the book b. ∗ John with care read the book
The generalization derivable from this data is that the head of the adjunct may be modified to its left, but may not be extended to its right. This difference between English and German has traditionally been captured by the Head-Final Filter (HFF), first proposed by Williams (1982). A generalized version of the HFF is given in (20). (20) Generalized Head-Final Filter (HFF): A premodifier must be head-final
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
While the HFF covers a great number of empirical facts (cf. i.a. Escribano 2009) and thus constitutes a valid empirical generalization, its status as a genuine syntactic condition is problematic for the following reasons. First, note that HFF-effects do not arise with subjects (the specifier of IP), PPframes, and specifiers of other functional heads in the C-domain. This is illustrated in (21). This raises the question of why the condition should apply to modifiers, but not to specifiers. (21)
a. [Students [of linguistics]] read Chomsky a lot b. [On [Tuesday evening]] I will take out Mary for dinner c. [In [which city]] did John meet Mary?
A second question concerns the issue of its cross-linguistic application: it applies in the I-domain of VO-languages, but fails to apply in this domain in OV-languages. A possible answer to this is that the HFF is somehow linked to the Head-Complement parameter. This line of approach leads to a peculiar conclusion, namely that its application in a VO-language like English has the effect that certain types of phrases must be head-final in an otherwise categorically head-initial language. Third, what is the status of the HFF in the grammar? The HFF is not a likely candidate for being a syntactic condition. In newer treatments of modifiers as specifiers of functional heads in the extended projection of the modified category (Cinque 1999), the HFF can no longer be stated as a genuine syntactic generalization based on the specific configuration of adjunction. Within current minimalist theory, it is best treated as a bare output condition at the PF interface since order and adjacency are taken to be irrelevant to narrow syntax. Note furthermore that the condition, as stated in (20), cannot be a genuine PF-condition either, since the structural difference between specifiers and modifiers is no longer visible at PF. It is generally assumed that prosody has (restricted) access to syntactic structure (cf. Selkirk 1984; Nespor and Vogel 1986). HFF-effects are reminiscent of weight effects in foot-construction systems at the word level (cf. Halle and Vergnaud 1987). In weight-sensitive systems, a heavy syllable must occupy a dominant branch in the metrical structure. In a parallel fashion, I would like to propose that a heavy syntactic phrase has to appear on a dominant branch in the syntactic structure, if the mapping between syntax and prosody happens to be weight-sensitive in a given domain. This raises the following two questions: a) When does a syntactic phrase count as heavy? And b) What counts as the dominant branch in syntactic structure? A syllable counts as heavy if its right branch, the rhyme, is itself branching, the complexity of the onset being immaterial for computing its weight. The parallelism between syllable structure and the X’-schema suggests the definition of syntactic weight given in (22), deriving the classical HFF-effects.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
(22) A syntactic phrase XP counts as heavy if both its head X and the complement of X contain lexical material. HFF-effects can be avoided by placing the adjunct after the modified head, indicating that the dominant branch in prosodic structure should be identified with a right branch in syntactic structure. In anti-symmetric syntax, the right branch constitutes the recursive branch. A standard metrical interpretation of a binary branching tree assigns the metrical value strong (s) to the right-hand branch at each projection level, as is illustrated in (23). (23) Yesterday John visited his mother
w Yesterday
w John
s s w visited
s w his
s mother
The metrical interpretation of the syntactic tree in (23) makes clear why HFFeffects should apply to premodifiers, since a left branch is less prominent than the respective nominal or verbal head. It also makes clear why the postnominal/postverbal placement of a heavy adjunct discards the effect: in this case, the heavy syntactic constituent occupies a more prominent branch than the head with which it is going to form a prosodic constituent, suggesting the relevance of the interface condition in (24). (24) The weight condition (PF-transparency): A heavy specifier in a given domain must occupy a more prominent branch than the selecting/modified head in prosodic structure, if this domain is weightsensitive. The metrical rendition of HFF-effects makes the following prediction in the present approach, in which arguments are taken to be licensed in Specifiers of functional heads in the I-domain. Assuming that the weight condition applies in the English I-domain, but fails to apply in the German I-domain, heavy DP- and PP-arguments are predicted to be spelled out in their base position in English, but may be spelled out in their licensing position in German. This approach raises the following questions: why does the weight condition not apply to subjects in English and why is movement into the C-domain not subject to the weight condition at all, as is indicated by the data in (21b, 21c)? One possible answer is to assume that weight-sensitivity is defined phase by phase.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
Some evidence for this assumption is given in Hinterhölzl (2009a). He argues that while the German I-domain is not weight-sensitive, the German v-domain is weightsensitive; this is to account for restrictions on word order in the verbal complex and leads to a typology of phases and subphases from which these facts follow.2 In the following section, I discuss how the prosodic rendition of the HFF fits into general assumptions about the syntax-prosody interface.
. On the mapping between syntactic and prosodic structure Because of the special role of accents in the focus-background articulation of intonational languages, most researchers favour an accent-first-based approach to the mapping between syntactic and prosodic structure (cf. Gussenhoven 1983; Uhmann 1991; Selkirk 1995; Truckenbrodt 1999). In these accounts, prominence relations in the clauses are adjusted to accent patterns that are derived from syntactic structure with the help of focus projection rules. For instance, in Uhmann (1991), it is assumed that accented syllables are metrically reinforced by receiving an extra beat after accent assignment. As said above, the core of these accounts consists in focus projection rules (Selkirk 1995), which serve to derive the focus domain for a given accented constituent or vice versa, so as to derive the placement of the sentence accent (nuclear accent) for a given focus domain. Büring (2002) proposes that focus projection rules can be dispensed with in a system in which (metrical) prominence relations are taken into account. He also argues that such a prominence-based system (also called stress-first-based accounts) additionally captures the default prosody in prefocal structures. Consequently, I will adopt a stress-first-based approach (cf. Halle and Vergnaud 1987; Ladd 1996), which assumes that accent positions in the clause are (also) determined by prominence relations. .. Prosodic domain formation in a phase-based approach There are two basic approaches to deriving prosodic structure from syntactic structure. End-based approaches (cf. Selkirk 1984) match boundaries of syntactic constituents with prosodic boundaries. These alignment rules are best expressed in an OT-like account (cf. Truckenbrodt 1999). In relation-based approaches (cf. Nespor and Vogel 1986; Wagner 2005), on the other hand, prosodic constituents are built around lexical heads on the basis of the relations they entertain with adjacent constituents. The two approaches differ in the assumption of how much syntactic information is available at the interface: while end-based approaches only assume the visibility 2 The reader is referred to Hinterhölzl (a, ) for an account of why subjects in English and constituents in the C-domain are weight-insensitive.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
of syntactic boundaries, relation-based approaches assume the visibility of syntactic relations expressed within the X’-schema. In this respect, note that it has been argued that prosody must have access to syntactic structure (cf. Gussenhoven 1983; Krifka 1984), since arguments in German and Dutch are phrased with the adjacent verb, while an adjunct and an adjacent verb form two separate phonological phrases. This is illustrated in (25). In the following, I will use round brackets to indicate phonological phrases, square brackets to indicate intonational phrases, and underlining (of the prosodic word) to indicate main stress. (25) a. [(weil Hans) (im Zelt blieb)] since Hans in.the tent remained ‘. . . because Hans stayed in the tent’ b. [(weil Hans) (im Zelt) (rauchte)] since Hans in.the tent smoked ‘. . . since Hans smoked in the tent’ Wagner (2005) proposes that there are two modes of prosodic composition which take into account whether an argument or an adjunct follows or precedes its selecting/modified head. To account for the differences in prosodic phrasing between German and English, illustrated in (26), Wagner proposes that subordination applies to a head and its preceding argument and creates a joint prosodic constituent, while sister matching applies to a head and the argument following it and derives two separate prosodic constituents that may optionally be restructured at a later point in the derivation. (26) a. [(weil Hans) (das Buch las)] since Hans the book read ‘. . . since Hans read the book’ b. [(since John) (read the book)] c. [(since John) (read) (the book)] In analogy to Wagner (2005), Hinterhölzl (2009a) proposes two modes of prosodic composition, which, however, are not directionality-based, but instead take into account the phasal status of two adjacent syntactic constituents. This is illustrated in (27). (27) Modes of prosodic composition a. subordination: (DP) + V → ((DP) V) b. coordination: (PP) & V → (PP) (V)
(Hinterhölzl 2009a)
Subordination applies to constituents that belong to the same phase (the verb and its arguments), irrespective of their relative order, and creates a recursive prosodic constituent, in (27a) a recursive phonological phrase. Coordination, in turn, applies
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
to constituents that belong to separate phases, irrespective of their relative order, creating two separate prosodic constituents of the same type, in (27b) two separate phonological phrases. Remember that adjuncts constitute separate phases in the present approach. Recursive prosodic categories are eliminated at a later level by restructuring operations and the deletion of outer boundaries that take into account global parameters like rate of speech, length, and branchingness of prosodic constituents. In this approach, it is assumed that prosodic composition follows the syntactic composition in a bottom up fashion. That is, parallel to syntactic composition, two prosodic constituents are combined according to the two modes in (27) and a head is determined, according to (28). This head is assigned an extra beat on the higher line, deriving a bracketed grid representation as in Halle and Vergnaud (1987). (28) a. Extrinsic heading (default value): In prosodic composition, the right-hand member is metrically stronger than its sister constituent. b. Intrinsic heading: In the combination of two distinct prosodic constituents, the constituent that is higher on the hierarchical layer counts as metrically stronger than its sister constituent.3 Languages may differ in whether they allow only for extrinsic heading or also for intrinsic heading. Intrinsic heading is necessary to account for the main prominence on the direct object in German. As is shown by the position of manner adverbs in (29), the direct object must be assumed to move out of vP in a Cinque-type approach to modification, and it is spelled out in a position that is structurally higher than the verb, obliterating an account of main-sentence stress in terms of the null theory of Cinque (1991). Without intrinsic heading, main stress would be predicted to fall on the verb in German, contrary to fact. With intrinsic heading, the direct object may receive main stress in a phase-based system for the following reason: at the point of the derivation in which the verb (a prosodic word) is combined with the direct object, the latter has already been mapped onto a phonological phrase (by default), deriving a joint prosodic category whose head is (the prosodic constituent corresponding to) the direct object. (29) a. weil Hans einen Brief sorgfältig las since Hans a letter carefully read ‘. . . since Hans read a letter carefully’
3 I will refer to this effect as strength-sensitivity parallel to the case of weakness-sensitivity triggered by discourse-given constituents to be discussed in the following section.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl b. weil Hans sorgfältig einen Brief las since Hans carefully a letter read ‘. . . since Hans carefully read a letter’ (only subject-oriented reading is possible)
In the following section, I discuss the complex interaction between prominence, accent assignment, and information-structural categories. .. Focus, prominence, and rules of accent placement Adopting a stress-first-based account (cf. Ladd 1996), I assume that accent assignment applies after prosodic domain formation and assigns an accent tone to each prosodic constituent. In particular, I assume that a word-level accent is assigned to each prosodic word, a phrase-level accent is assigned to each phonological phrase, and a sentence accent, usually called the ‘nuclear accent’, is assigned to each intonational phrase of the utterance, according to the principle in (30). (30) The accent must fall on the metrically most prominent syllable in a prosodic domain. If we assume, following Halle and Vergnaud (1987) as outlined above, that the labelled tree is converted into a bracketed grid representation during prosodic evaluation, the relative strengths of the several accents in the clause can be derived. This is illustrated in (31) for a putative German sentence comprising a subject DP, two adjunct XPs, a direct object DP, and the verb.
∗ ∗
(31)
w DP
s w XP
∗ s
w XP
s s DP
w V
∗
∗
∗
∗
∗ ∗
∗
∗ ∗ ∗ ∗ ∗ (DP) (XP) (XP) (DP V)
However, prosodic structure is not only determined by and derived from syntactic structure, as illustrated in (31), but it is also crucially determined by information structure (IS). In particular, it must be ensured that the focused constituent in an utterance is assigned the main stress, independently of its position in the clause. Thus, Jackendoff (1972) argued for the introduction of a syntactic feature F (for focus), which is interpreted both at PF and LF, with the PF-interpretation given as in (32) below. A similar constraint is also proposed in Truckenbrodt (1995), as given in (33).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German (32) Prosodic Effect of F F attracts the strongest stress of the sentence. (33)
Focus Prominence Focus needs to be maximally prominent.
(Jackendoff 1972) (Truckenbrodt 1995)
I assume the following condition for the mapping between accent and focus in intonational languages like German and English. (34) The focused constituent must contain the most prominent accent in the clause. How can focus prominence be achieved in a stress-first-based system? Note that in such a system the relative strength of an accent depends on its metrical value in the clause. If the mapping between syntactic structure and prosodic structure is monotonic, then there must exist a direct mapping relation between metrical values and IS-categories. Accent-first-based systems assume a special interface condition for constituents representing background information, usually marked with the feature G (for given information), as is illustrated in (35). In analogy to the condition in (35), I assume the condition in (36). (35)
Prosodic effect of G G rejects sentence and phrasal stress.
(Fery and Samek-Lodovici 2006)
(36) Background-Transparency A given constituent must occupy a weak position in prosodic structure. Note that there are basically two ways of satisfying a condition like (36). A) A given argument moves out of its postverbal (strong) base position and is spelled out in a preverbal position, which necessarily counts as metrically weak. B) The default value assigned in prosodic composition to a given postverbal constituent is overridden by projecting its intrinsic prosodic value, namely weak, in prosodic composition. We have a case of deaccenting in situ, which, in the present system, represents the third case of sensitivity to inherent metrical properties in prosodic domain formation. Parallel to the condition in (36), I will assume the condition in (37) for constituents pertaining to the focus domain of the utterance. (37) Focus-Transparency A constituent representing new information must occupy a strong position in prosodic structure. How can this condition be satisfied? The simplest option is that an argument undergoes licensing movement, but is spelled out in the vP in a postverbal position where it occupies a strong branch with respect to the verb. Another option in which focus can interact with metrical structure is more indirect and involves the insertion of a functional head in syntactic structure. Such a head, typically called Focus and found
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Roland Hinterhölzl
in the left periphery of the clause (cf. Rizzi 1997) must be assumed to have the property, next to its LF-property of introducing alternative values for an open proposition (cf. Rooth 1992; Krifka 2007), of assigning the metrical value strong to its specifier (the focused constituent) and the metrical value weak to its complement, which represents the presupposition of the utterance. In many languages, this strategy is reserved for emphatic or contrastive focus, while information focus is often unmarked and can be handled most naturally by dictating spell-out options as assumed above. Returning to the generalization in (18), I propose that OHG had such a specialized focus position in the middle field, which was predominantly used for contrastively focused constituents.
. Conclusions With these assumptions about the interface between syntax, prosody, and IS in place, let us now return to the OHG data in Section 18.3. There are in fact two (different) motivations for spelling out an argument in the vP. Weight-sensitivity requires heavy constituents to be spelled out in a postverbal position, and constituents that belong to the domain of new information will also, independently of their prosodic weight, be spelled out in the vP, due to focus-transparency. Preverbal heavy constituents that are contrastively focused do not represent an exception to the weight law since the latter only requires heavy constituents to occupy a strong branch with respect to the verb and, as stated already, the specifier of a specialized focus position counts as metrically strong. Given constituents are spelled out in the preverbal domain. This follows from background transparency. There are two cases that remain problematic at this point. First, it is not clear what forces verb particles to be spelled out preverbally since there is no specific interface condition that requires light elements to be spelled out preverbally. There are two options to resolve this issue. Either we assume a default condition on spell-out, as given in (38), or we assume that there is a specific interface condition on complex predicate formation which requires that particle and verb form a prosodic unit with the default pattern being (s w), as is generally the case in compounding in Germanic. I will leave this issue for further research. (38) Preference for the higher copy: A constituent is spelled out in its checking position, unless interface requirements demand its spell-out in the base position. Second, there remains the issue of why branching given constituents do not violate the weight law in OHG. Also in this case, there are two options for resolving the issue. As we said above, most of these cases involve a noun modified by a demonstrative determiner. Either we assume that the demonstrative element still occupied SpecDP
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
An interface account of word-order variation in Old High German
with D empty, such that this phrase does not count as prosodically heavy (in other words the determiner was not grammaticalized yet in the OHG Tatian translation), or we assume that given constituents were moved into the C-domain in OHG, which is generally not subject to the weight condition, as argued in Hinterhölzl (2013). Both assumptions, however, require more research to explore them in detail, which goes beyond the scope of this chapter.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English A N N TAY L O R A N D SU S A N P I N T Z U K
. Introduction The correct analysis of the Old English (OE) object-verb/verb-object (OV/VO) alternation is a long-standing and highly disputed issue in the diachronic syntax literature (see Taylor and van der Wurff 2005 for some views).1 The situation is complicated by the fact that in addition to the OV/VO variation, there is variation in the order of finite and non-finite verbs. Thus in clauses with a finite auxiliary, a non-finite main verb, and an object, there are two seemingly independent alternations: VAux vs. AuxV order and OV vs. VO order. The result is the four orders illustrated in (1), with the finite auxiliary underlined, the non-finite main verb italicized, and the object in bold.2 (1)
a. OVAux gif heo þæt bysmor forberan wolde if she that disgrace tolerate would ‘if she would tolerate that disgrace’ (coaelive,+ALS_[Eugenia]:185.305) b. VAuxO þæt he friðian wolde þa leasan wudewan that he make-peace-with would the false widow ‘that he would make peace with the false widow’ (coaelive,+ALS_[Eugenia]:209.315)
1 This chapter was presented at the th Diachronic Generative Syntax Conference (DiGS) in Cambridge, July . We thank the audience and four anonymous reviewers for helpful comments and suggestions. All errors and omissions remain our own. 2 We omit two logical possibilities: VOAux and OAuxV. The former do not exist (Biberauer, Holmberg, and Roberts ; Pintzuk ); in the latter, the position of the object may be influenced by factors different from those affecting the OV/VO alternation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
c. Aux O V þurh þa heo sceal hyre scippend understandan through which it must its creator understand ‘through which it must understand its creator’ (coaelive,+ALS_[Christmas]:157.125) d. Aux V O swa þæt heo bið forloren þam ecan life so that it is lost the eternal life ‘so that it is lost to the eternal life’ (coaelive,+ALS_[Christmas]:144.117) Notice that the examples in (1) are from a single text, Ælfric’s Lives of Saints, and this text is not exceptional: all four orders occur in almost all OE texts of any length.3 Thus the variation cannot be attributed to variation within the community, with some OE speakers categorically producing one constituent order, e.g. OV, while other speakers categorically produce the alternate order, e.g. VO. Instead, the variation is at the level of the individual, with all speakers producing all four orders; and therefore all variants must be derivable within the grammar(s) of individual speakers. Syntactically, the most parsimonious approach to analysing object position in OE is to disregard the order of the finite auxiliary verb and the non-finite lexical verb and to derive all postverbal objects in the same way, as has been proposed in, for example, Roberts (1997), van Kemenade (1987), and van der Wurff (1997). This view is based on the reasonable assumption that the two alternations are independent, and thus that verb order is not a factor in determining verb-object order. In analyses of this type, the source of the variation in object position is, rather, attributed to discourse/performance factors such as new information focus. In this chapter we show that this simple view is not supported by the evidence. When we widen our view of the OV/VO data to include the interaction of discourse/performance factors with the order of finite and non-finite verbs, we see a very different picture: there is clear evidence that the distribution of pre- vs. postverbal objects differs systematically in VAux and AuxV clauses. Although this appears to support the non-independence of verb order and verb-object order, we argue instead that the two alternations are indeed syntactically independent, but that each alternation is associated with a change in progress: from head-final to headinitial TP (VAux to AuxV) and head-final to head-initial VP (OV to VO), respectively. Crucially, these two changes are linked in such a way as to create the appearance in the synchronic data of non-independence. We show that, by adopting a model of syntactic
3 Only two texts in our dataset with more than tokens of the types illustrated in () lack any of the variants. The Holy Rood Tree and The West-Saxon Gospels with and relevant tokens, respectively, both lack VAuxO order.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
change in which speakers have access to more than one grammar during the course of change, a conservative ‘outgoing’ grammar and an innovative ‘incoming’ grammar (Kroch 2003), we can account for the empirical facts, while maintaining syntactic independence of the headedness parameter settings in TP and VP.
. Data and terminology The data for this study are taken from the York-Penn-Helsinki Parsed Corpus of Old English Prose (Taylor et al. 2003, henceforth the YCOE). The dataset includes clauses with finite auxiliaries, non-finite main verbs, and nominal objects from all texts (duplicate manuscripts excluded) included in the YCOE with the exception of Ælfric’s Supplemental Homilies, Wulfstan’s Homilies, the Vercelli Homilies, and the Institutes of Polity.4 There are 85 texts in total in our dataset, although some texts are represented by very small numbers of tokens. The sample of VAux clauses is exhaustive; for AuxV clauses, approximately one-third of the total cases (every third token) were collected. The main dataset contains 1,525 tokens in total, 692 VAux clauses and 833 AuxV clauses. The data are further restricted by the exclusion of the following types: • Clauses with movement of the finite auxiliary to C. These clauses are ambiguous between underlying AuxV and underlying VAux structure and order. As discussed in much previous work, e.g. Pintzuk (1999), most so-called V2 clauses in Old English involve verb movement to T; it is only in exceptional clause types (e.g. direct questions, clauses beginning with þa/þonne ‘then’, verb-initial clauses) that the finite verb moves to C. • AuxV clauses with the object before the Aux (OAuxV order); cf. footnote 1. • VAux main clauses with non-overt subjects, or with the object before the subject, to avoid potential cases of topicalization. • Pronominal (personal and demonstrative) objects. Pronominal objects are optionally clitics or weak pronouns in OE (Pintzuk 1996; Wallenberg 2009), and frequently move to a high position in the clause structure. Since their syntax is different from non-pronominal objects, they are not considered here. • Quantified objects. It has been shown (Pintzuk and Taylor 2006) that quantified objects exhibit special syntactic behaviour compared to non-quantified objects. • Non-referential objects (negative and semantically incorporated) are excluded from the main dataset and discussed separately in Section 19.3.2. • A few additional cases where the information status of the object is unclear.
4 These texts were excluded because they lack Modern English translations, a necessity for coding information structure in an accurate and timely manner.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
. The distribution of objects in VAux and AuxV clauses In this section we present data on the distribution of objects in OE VAux and AuxV clauses. After a look at the overall picture, we turn in Section 19.3.1 to three particular factors which have been associated in the literature with surface word order variation: information status, case, and grammatical weight. Under the null hypothesis (that there is no association between verb order and object position), we would expect to find these constraints applying equally in both types of clauses; instead, we find a significant difference in the distribution of objects in the two clause types. In Section 19.3.2, we then consider two types of objects omitted from the main analysis due to their non-referential status, negative objects and incorporated objects, and show that their distribution also differs significantly in VAux vs. AuxV clauses. Table 19.1 presents the overall frequency of VO order in VAux and AuxV clauses; the texts are listed by the frequency of AuxV (vs. VAux) order, which is a rough indication of chronology; see Taylor and Pintzuk (2012a) for discussion. Table 19.1 shows that overall VO order is more frequent in AuxV than in VAux clauses, not only in aggregate, but also for each text. If all postverbal objects are derived by the same discourse/performance constraints, this is an unexpected result. While some variation is inevitable given the nature of the data, if AuxV and VAux clauses do not differ with respect to the derivation of object position, we would expect the frequency of VO to be closer in value overall; in addition, text by text, we would expect to see variation in frequency in both directions, rather than always in the direction of VO. Table .. Frequency of VO order by verb order in texts with more than 100 tokens, VAux
Text
Orosius Bede Boethius Cura Pastoralis Catholic Homilies I Catholic Homilies II Lives of Saints Gregory’s Dialogues (C) total sample total
AuxV
N
VO
N
VO
66 58 74 51 49 42 33 36
4.5 6.9 8.1 21.6 10.2 7.1 45.5 27.8
47 46 49 72 95 80 91 66
31.9 10.9 53.1 55.6 47.4 46.3 62.6 68.2
409 692
13.9 12.6
546 833
49.5 47.1
5 The difference between VAux and AuxV clauses is significant at the 0.001 level for both totals and for all texts, except for Bede and Lives of Saints, where the difference is not significant. 6 Recall that VO order in VAux clauses means . . . V Aux (. . .) O, while VO order in AuxV clauses means . . . Aux V (. . .) O.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
.. Factors affecting object position ... Information status There is a vast literature addressing the assignment of information status to entities within utterances (cf. Bech 2001; Bies 1996; Hinterhölzl 2009b; van Kemenade and Los 2006; Petrova 2009; Sapp 2006; Westergaard 2009, among others, for recent work on discourse factors in the older West Germanic languages). Although details differ, there is general agreement that information status correlates with position in the clause: given elements favour an early position and newer elements a later one (Bech 2001; Pintzuk and Taylor 2006, among many others). We have followed the common line in studies of this sort (e.g. Arnold et al. 2000; Bech 2001; Gries 2003) and used a binary distinction of given vs. new. We divide the data into given and new entities primarily on the basis of insights drawn from the work of Birner (2006) building on Prince (1981) and Gundel, Hedberg, and Zacharski (1993). Details can be found on the website associated with Nevalainen and Traugott (2012): (chapter 64). Our results, given in Table 19.2, show the expected effect in both VAux and AuxV clauses, i.e. new objects favour postverbal position; but the difference between given and new is much greater, and only significant, for VAux clauses. In Tables 19.2–19.5, shaded cells indicate results that are not statistically significant. Table .. Frequency of VO order in VAux vs. AuxV clauses by information status Info Status
given new
VAux
AuxV
N
VO
N
VO
507 185
9.7 20.5
507 326
45.0 50.3
VAux: χ 2 = 14.6, p = 0.0005 AuxV: χ 2 = 2.27, p = 0.132
... Case Dunbar (1979: 175) gives the hierarchy in Figure 19.1 for different constituents according to their likelihood of appearing postverbally in Old High German. As Dunbar notes, this hierarchy is almost the opposite of the accessibility hierarchy of Keenan and Comrie (1977) in which elements higher on the hierarchy are less likely to be topical/themes than those farther down.7
7 Burridge (: –) finds almost the same hierarchy for Middle Dutch. Although her order of DO/IO is reversed from that of Dunbar, she attributes the reversal to her inclusion of pronouns in the counts: almost all of the Middle Dutch indirect objects are pronouns, which are highly topical and therefore show a lower tendency than direct objects (presumably a mixture of pronouns and full DPs) to appear in postverbal position.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
lower postverbal ------------------------------------------------------------higher postverbal subject < pred. nominal < direct object < genitive/indirect object < prep. Phrase higher topicality -------------------------------------------------------------- lower topicality Figure . Hierarchy of postverbal position in OHG
Thus, as indicated in Figure 19.1, subjects are the most likely constituents to be topics, and the least likely to appear postverbally, while complements of prepositions are the least likely to be topics, and the most likely to appear postverbally. The constituents we are interested in, direct and indirect objects, fall in between these extremes, with direct objects expected to be more topical (and less postverbal) than indirect objects. Topicality is of course related to information status: topics represent given information, while non-topics may represent new information. On this basis, as direct objects are primarily accusative and indirect objects are primarily dative or genitive, the case effect may reduce to information status. As shown in Table 19.3, our data do indeed reflect the difference between accusatives and datives as expected, although the difference is not significant. The most extreme effect, however, is with genitives, which appear postverbally more frequently than either accusatives or datives. The reason for this is not clear. Dunbar does not discuss genitives separately from datives, treating them both as ‘indirect objects’, but in our data the difference between datives and genitives is larger than the difference between accusatives and datives. Whatever the source of this difference, the effect is greater, and only significant, in VAux clauses, just as with information status. Table .. Frequency of VO order in VAux vs. AuxV clauses by case Case
accusative dative genitive
VAux
AuxV
N
VO
N
VO
517 133 42
9.9 14.3 40.5
644 140 49
46.4 47.9 53.1
VAux: χ 2 = 33.6, p < 0.0005; acc vs. dat: χ 2 = 2.15, p = 0.142 AuxV: χ 2 = 0.847 , p = 0.655
... Grammatical weight The grammatical weight (length/heaviness) of constituents has frequently been shown to correlate with clause position crosslinguistically, with longer/heavier elements favouring later position (Arnold et al. 2000; Gries 2003; Pintzuk and Taylor 2006; Siewierska 1993; Taylor and Pintzuk 2012b, among many others). Exactly how grammatical weight is measured seems to make little difference to the outcome. The conclusion that can be drawn from studies such
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk Table .. Frequency of VO order in VAux vs. AuxV clauses by length of object (in open-class words) Words
1 2 3 4+
VAux
AuxV
N
VO
N
VO
453 197 28 14
5.3 25.9 35.7 14.38
442 313 53 25
34.8 56.4 74.1 88.5
VAux: p = 3.499e-14 [Fisher’s exact]; AuxV: p = 1.614e-14 [Fisher’s exact]
as Wasow (1997) and Szmrecsányi (2004), as well as our own work, is that while continuous measures are more accurate than categorical ones, all measures are highly correlated. Here we measure weight in the number of open-class words.9 Table 19.4 shows that, for the most part, for each additional open-class word, an object is more likely to appear postverbally in both AuxV and VAux clauses. As with the other factors considered, this factor thus shows an effect in both clause types, but here the effect is significant for both. ... Strength of effect In the previous section we saw that on the basis of frequencies alone, information status, case, and grammatical weight all show an effect in the expected direction in both AuxV and VAux clauses. The former two effects, however, are significant only for VAux clauses, which implies that their effect is less strong in AuxV clauses. We can test this prediction using multivariate analysis to assess the strength of each effect while simultaneously taking all other effects into account.10 The results are given in Table 19.5. The results from the multivariate analysis confirm the picture given by the frequencies, but the log odds and odds ratios also indicate that the effect of grammatical weight is (slightly) weaker in AuxV clauses, a fact not evident from the frequencies alone. We can see this by comparing the log odds for grammatical weight for the two clause types: it is significant for both, but nevertheless slightly lower in AuxV than
8 The low frequency of VO order for objects of + words in VAux clauses is unexpected. We note the relatively small N for this category; nevertheless we have no explanation for the frequency. 9 As definiteness is normally signalled with a demonstrative in OE, while the language lacks a dedicated indefinite determiner, including all words in the count would lead to definite DPs being heavier than indefinite DPs simply due to the inclusion of a determiner. 10 The analysis was carried out using Rbrul , including clause type (main, that-complement, adverbial, relative) as a factor and with text as a random effect, using sum contrasts (the Rbrul default). The coefficients indicate the increased likelihood of a postverbal object given the factor level (e.g. given, new, etc.) against the average case, measured on a logarithmic scale. The numbers, therefore, have no straightforward real-world interpretation, but the scale is linear, so the absolute size of the effects is comparable.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
Table .. Results of multivariate analysis, effects in log odds and odds ratios Variable
grammatical weight info status case
Level
per additional open-class word given new accusative dative genitive
VAux
AuxV
Log odds
Odds ratio
Log odds
Odds ratio
0.99
2.68
0.89
2.43
−0.45 0.45 −1.06 −0.38 1.45
0.64 1.57 0.35 0.68 4.25
−0.04 0.04 −0.22 −0.17 0.38
0.96 1.04 0.81 0.85 1.47
VAux: C = 0.88; AuxV: C = 0.78
in VAux clauses (0.89 vs. 0.99). In terms of odds ratios,11 each additional open-class word added to an object in VAux clauses increases the odds of VO order by somewhat over two and a half times (2.68), while in AuxV clauses it increases the odds by slightly under two and a half times (2.43). The other two factors, information status and case, are significant only in VAux clauses, as we saw previously with the frequencies. As with the frequencies, the effect is in the same (expected) direction in both clause types, but again the log odds show that the effects are stronger in VAux than in AuxV clauses. Thus for VAux clauses, the distance between information status given and new on the log odds scale is 0.90 (0.45+0.45), while in AuxV clauses it is only 0.08 (0.04+0.04). In terms of odds ratios, in comparison to the mean, being new increases the odds of an object appearing in postverbal position in VAux clauses by 57 (1.57), but only by 4 (1.04) in AuxV clauses. The same is true of the case effect. The distance between accusative and genitive on the log odds scale in VAux clauses is 2.51 (1.06+1.45), while in AuxV clauses it is 0.60 (0.22+0.38). In comparison to the mean, being genitive increases the odds of an object appearing postverbally by 4.25 times in VAux clauses, but by only 1.5 times (1.47) in AuxV clauses; in contrast, being accusative decreases the odds by about 65 (0.35) in VAux clauses, but only by about 20 in AuxV clauses (0.81).12 Note finally that the fit for this model is better for VAux clauses than for AuxV clauses: C, an index of concordance between the predicted probability and the observed response,13 is 0.88 for the VAux model and 0.78 for the AuxV model. 11 The odds ratio is the probability of success over the probability of failure (p / ( – p)). Transforming the log odds into odds ratios makes them easier to interpret. The odds ratio is calculated by raising the constant e to the log odds power, e.g. for grammatical weight in VAux clauses, e to the power of . = .. 12 It is unclear whether the relative weakness of the information status effect with respect to case, and particularly heaviness, is a real effect or simply reflects the difficulty of coding this factor in comparison to heaviness and case, which are much more straightforward. 13 C ranges from . (random) to (perfect fit). Values of . or above indicate the model has real predictive value.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
.. Non-referential objects ... Negative objects Negative objects, as in (2), were excluded from the analysis above due to their non-referential status. Their distribution, given in Table 19.6, shows a sharp contrast between VAux and AuxV clauses: negative objects never appear postverbally in VAux clauses (cf. also Pintzuk 2005). In contrast, non-negative objects, as we saw in Table 19.1, occur postverbally at approximately 13 and 47 in VAux and AuxV clauses, respectively. (2) a. heo nan ðing elles don ne mihte she no thing else do neg might ‘she might do nothing else’ (cocathom2,+ACHom_II,_42:316.180.7143) his b. þæt nan cristen mon ne moste habban nænne none (of) his that no Christian man neg might have sunderfolgeþa official-teachings ‘that no Christian man might have any of his official teachings’ (coorosiu,Or_6:31.150.14.3217)
Table .. Frequency of postverbal negative objects and non-negative objects in VAux and AuxV clauses Object type
negative non-negative
VAux
AuxV
N
VO
N
VO
70 692
0.0 12.6
107 833
14.0 47.1
... Semantically incorporated objects Semantically incorporated objects, as in (3), were also excluded from the analysis as they are known not to respond to information status factors in the same way as referential objects; in particular, they do not exhibit continuity of identity with previously or subsequently mentioned noun phrases (du Bois 1980).14 (3)
a. Ac gif ic deað þrowian sceal, leofre me is, þæt . . . but if I death suffer shall dearer me is that . . . ‘But if I suffer death, it is dearer to me that . . . ’ (cobede,Bede_2.9.128.6.1213)
14 As far as we know, there is no completely objective way to determine whether an object is incorporated in OE. The clearest cases are those in which the object is a single bare noun, it is not later referred to, and the verb+object can be paraphrased with a verbal idea alone: e.g. suffer death = die. Another indication of incorporation in this sense is the continued use of the same bare noun (or NP) in subsequent mentions.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
þrowian for Criste b. and wiscton þæt hi moston wite for Christ and wished that they might punishment suffer ‘and wished that they might suffer punishment for Christ’ (coaelive,+ALS_[Chrysanthus].216.7456) We might expect a difference for these objects similar to that for negatives, but in this case the data are less clear. As shown in Table 19.7, in VAux clauses, 2 out of the 32 cases of incorporated objects (6.3) are postverbal where we would expect zero. However, both these cases occur in Ælfric’s Lives of Saints. For reasons which are not entirely clear, but are possibly related to its metrical properties, this text often behaves differently from others (including Ælfric’s own non-metrical Catholic Homilies) with respect to object positioning (cf. Taylor 2006). If we omit this text, then we find the same distribution for incorporated as for negative objects: no postverbal objects in VAux clauses as opposed to 15 in AuxV clauses. This is very close to the difference that we see with negative objects in Table 19.6 (0 vs. 14.0). The difference in the frequency of incorporated vs. non-incorporated objects in AuxV clauses is also instructive. There is no weight distinction here since the data (incorporated and nonincorporated) have been restricted to cases of one open-class word, and yet nonincorporated objects appear postverbally about twice as often as incorporated ones. Table .. Frequency of postverbal position for incorporated and non-incorporated objects (one open-class word only) in VAux and AuxV clauses Object type
incorporated
non-incorporated
Texts
All texts LoS LoS omitted All texts LoS LoS omitted
VAux
AuxV
N
VO
N
VO
32 5 27 463 14 449
6.3 40.0 0.0 5.6 35.7 4.7
45 5 40 451 55 396
15.6 20.0 15.0 34.8 47.3 33.1
.. Summary In summary, we have seen that VAux and AuxV clauses differ on three dimensions: • postverbal objects are less frequent overall in VAux clauses than in AuxV clauses. • Factors associated with end position—weight, information status, and case—have a greater effect, to varying degrees, in VAux clauses than in AuxV clauses.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
• Non-referential objects (negative and semantically incorporated) do not freely appear postverbally in VAux clauses but do so in AuxV clauses.
. Analysis On the face of it, it is hard to reconcile the differences between VAux and AuxV clauses with respect to object position discussed in Section 19.3 with a syntactic analysis in which all postverbal objects are triggered by the same discourse/performance constraints. In this section we suggest a different analysis for the distribution of objects in VAux and AuxV clauses. The analysis retains the independence of the two syntactic alternations (head-final/initial TP and head-final/initial VP), but nevertheless is able to account for the differences in the distribution of objects between the clause types. .. Change in progress The first point to note is that the four word orders under consideration here (illustrated in (1)) are associated not only with different discourse/performance constraints, but also with two independent but linked changes that occur starting in the (pre-)OE period. The first change is the replacement of VAux order by AuxV order, i.e. the change from head-final to head-initial TP. This change is a prerequisite for the second one, the replacement of OV by VO order, i.e. the change from head-final to headinitial VP. As discussed extensively elsewhere (Biberauer, Holmberg, and Roberts 2007; Pintzuk 2005, among others) a head-final TP cannot select a head-initial VP, i.e. the structure [TP [VP V O ] Aux ] is never generated or derived. From this it follows that the change from VAux to AuxV is a necessary precondition for the change from OV to VO. By precondition we do not mean that the first change has to be completed before the second change begins. On the contrary, the time spans of the two changes overlap almost completely: as noted above, all four orders can be found in any OE text of any length. Nevertheless, there is some evidence that the variation in headedness in the TP starts earlier than the variation in headedness in the VP. The earliest extant OE text, the narrative poem Beowulf, was uniformly OV with variation in the headedness of TP (Pintzuk 1999; Pintzuk and Kroch 1989), and the prose Laws of Æthelbert (composition dated to approximately 600 ce) is consistently surface V-final, even in main clauses, according to Oliver (2002: 31).15 , 16 Based on the Germanic ancestors of OE and on the diachronic trends, it is reasonable to assume that the earliest stage of OE was uniformly head-final in both the VP and the TP, with variation in surface object position (OVAux
15 This text is fairly short and in a very particular genre, so it is not clear how seriously to take its consistency. It does differ from later texts in the same genre on the position of the verb, however, indicating that it is not just a genre-specific feature. 16 Both Beowulf and the Laws of Æthelbert are found only in fairly late manuscripts; presumably the OV order of these texts was maintained by scribes who would have had more VO order in their own speech.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English (a)
(b)
(c)
(d)
(e)
pre-OE
Beowulf
first Aux-V-pro17
early ME
late ME
orders
c.1250
16th cent.
(f) PDE18
head-final TP------------------------------------------------------head-initial TP-------------------------------------------------------------------head-final VP------------------------------------------------------head-initial VP--------------------------------------------------Figure . Timeline for head-initial TP and VP in the history of English
vs. VAuxO) triggered by discourse constraints (cf. Pintzuk and Kroch 1989). Moreover, VAux order disappears before OV order. While both survive into Middle English, the frequencies of VAux order are already very low in Early Middle English (Kroch and Taylor 2000b), while OV order continues with some frequency throughout Middle English until finally disappearing from prose in the course of the sixteenth century (van der Wurff 1997, 1999; Pintzuk and Taylor 2006). A rough timeline is given in Figure 19.2: head-final TPs and VPs are used in periods (a)–(d), head-initial TPs are first used in period (b), while head-initial VPs are first used in period (c). This scenario produces a somewhat complicated situation, illustrated in Table 19.8, in which three distinct grammatical systems with their base and derived orders are shown.19
17 Since pronouns do not postpose in OE (Pintzuk ), we can ‘date’ the appearance of the head-initial VP from the first appearance of postverbal pronouns. Six of the texts dated as pre- in our corpus which contain examples of AuxV clauses with pronominal objects have some Aux-V-pro order: Blickling Homilies (/), Boethius’ Consolation of Philosophy (/), Cura Pastoralis (/), Gregory’s Dialogues C (/), Laws of Ine (/), and Augustine’s Soliloquies (/). In the remaining texts—Alexander’s Letter to Aristotle, Bede’s History of the English Church, Life of Saint Chad, early parts of Chronicle A, Documents , Leechdoms, Laws of Alfred, Martyrology II/III, Marvels of the East, and Orosius—all pronominal objects appear before the non-finite verb. 18 G, the grammar incorporating head-initial TP and VP (see Table .), is shown as extending to PDE. But G is actually not quite the PDE grammar, because it allows preposing of negative and quantified objects from postverbal to preverbal position (Pintzuk and Taylor ), as in Modern Icelandic (Rögnvaldsson ; Svenonius ). This option was lost in late ME. 19 Although G/G/G are labelled ‘grammatical systems’, it is not the case that there are major differences between them. If we think of parameters (like the headedness parameter) as features on functional heads, then the differences between the grammars are simply differences in the features on T and V: T in G selects a VP complement to its left, while T in G/ selects a VP complement to its right. Similarly, V in G/ selects a DP complement to its left, while V in G selects a DP complement to its right. Thus the competition between G and G in periods (b)–(d) is simply a competition between two functional heads T that differ in a single feature (directionality), while the competition between G and G in periods (c) and (d) is a competition between two functional heads V. See Kroch () for further discussion.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
Table .. Three grammatical systems
System
Periods used
TP
G1 G2 G3
(a)–(d) (b)–(d) (c)–(f)
head-final head-final OVAux head-initial head-final AuxOV head-initial head-initial AuxVO
VP
Base order
Derived order VAuxO AuxVO AuxVO
Characteristics of postverbal objects (idealized) new and/or heavy new and/or heavy no discourse-related effects
During the period in which both G1 and G2 are used (from the time represented in the language of Beowulf through early ME), the use of G2 increases in frequency at the expense of G1; in other words, the TP changes from head-final to head-initial. During the period in which both G2 and G3 are used (from the OE period to the late ME period), the use of G3 increases in frequency at the expense of G2; in other words, the VP changes from head-final to head-initial. The output of G1 can be distinguished from the output of G2 and G3 by the surface order of the verbs, as all VAux clauses belong to G1. However, G2 and G3 cannot be reliably distinguished by surface order. AuxOV order must belong to G2, but AuxVO can belong to either G2 or G3, depending on how it is derived. If generated by G2, it should be associated with particular discourse constraints; but if generated by G3, it should show no discourse-related effects. In theory, therefore, it ought to be possible to distinguish G2 AuxVO from G3 AuxVO, but of course in practice discourse constraints are neither categorical nor detectable with enough certainty to apply to individual cases. What we can reliably distinguish is cases of G1, in which all VO order is hypothesized to be derived from an OV base, and a mixed bag of G2 and G3 cases, some with VO order derived from an OV base and some from a VO base. In an idealized world, all postverbal objects in G1 are focused (new) and/or heavy; the same is true of postverbal objects in G2. G3 is like PDE, however, where all objects are postverbal regardless of information status or weight; i.e. they do not exhibit discourse-related effects. The expected outcome of this scenario in which AuxVO clauses belong to either G2 or G3 is that, when treated as a group, some of these postverbal objects will exhibit discourse-related effects and some won’t, and thus there will appear to be a weakening of the discourse-related effects in comparison to the uniform VAuxO class. A straightforward way to model the variation described in Table 19.8 is by invoking a headedness parameter for TP and VP. Simplified versions of the relevant derivations, with the minimal amount of structure and movement shown, are given in (4).20 T and 20 These structures are not intended to reflect assumptions about the nature of auxiliary verbs in OE, i.e. whether they appear in monoclausal or biclausal structures.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
V take VP complements. Finite verbs (V1) move obligatorily to T, regardless of the position of T, while non-finite verbs (V2) remain in situ. DP objects may remain in their merged position, or they may postpose to postverbal position (i.e. adjoin to the right periphery of TP), presumably because of focus or heaviness. (4) a. OVAux: head-final TP and VP TP
Subj
T VP1 V1+T
VP2 Obj
tv1
V2
b. VAuxO: head-final TP and VP with postposition of O TP
Subj
Obj
Subj
T
VP1 V1+T tv1
VP2 tObj
V2
c. AuxOV: head-initial TP, head-final VP TP
Subj
T
V1+T VP1 VP2 Obj
tv1 V2
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk d. AuxVO: head-initial TP, head-final VP, with postposition of O TP
TP
Obj
Subj
T
V1+T VP1 tv1
VP2 tObj
V2
The difference between (4a) and (4b) and between (4c) and (4d) is the assignment of focus: focused objects move out of the VP and right-adjoin to TP. However, AuxVO clauses also have a second derivation: all projections can simply be head-initial, as shown in (5). (5)
AuxVO, derivation 2: head-initial TP and VP TP
Subj
T'
V1+T tv1
VP1 VP2 V2
Obj
In summary, an account with two postverbal object positions in AuxV but not VAux clauses predicts differences not only between OV and VO clauses but also between VAuxO and AuxVO clauses, while accounts with a single postverbal object position can capture only the first difference. In the discussion that follows, we will use the head-initial vs. head-final terminology for ease of explanation; but it should be noted that an anti-symmetry analysis such as Biberauer and Roberts (2008) or Wallenberg (2009), incorporating pied-piping and focus movement and providing two postverbal object positions, would work equally well.21
21 Note, however, that the analysis of Biberauer and Roberts () is problematic with respect to data coverage: it predicts constituent orders that do not occur in Old and Middle English, and does not derive all of the attested constituent orders.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
.. Two postverbal object positions In this section we demonstrate how the empirical facts described in Section 19.3 can be accommodated under an analysis in which there are two postverbal object positions which differ with respect to discourse-related effects. As we saw in Table 19.1, postverbal objects are more frequent in AuxV than in VAux clauses, both in aggregate and text by text. Under the reasonable assumption that postverbal objects triggered by discourse/performance constraints occur at approximately the same rate in G1 (VAux) and G2 (AuxV) clauses, the ‘extra’ AuxVO tokens shown in Table 19.1 are the result of the addition of the G3 base-generated AuxVO tokens to the G2 derived ones. In other words, surface AuxVO tokens are derivationally ambiguous, comprising both base-generated postverbal objects and those derived by focus movement, whereas VAuxO tokens are all derived by movement, as illustrated in Table 19.9. Table 19.5 shows that the effects of information status, case, and grammatical weight are less strong in AuxV clauses than in VAux clauses, although in the case of weight, the difference is quite small. As discussed in Section 19.3.1.2 the case effect is likely to indirectly reflect information structure. It is a commonly held view (see, for instance, Burridge 1993: 107 and Niv 1992) that weight is also a reflection of information status, on the assumption that new constituents require more material to identify them. However, in more recent work, Arnold et al. 2000, Gries 2003, and Taylor and Pintzuk 2012b have argued that heaviness has an effect independent of information status, although the relationship between the two is complicated.22 Although it is not entirely clear if/how these three effects are related, the difference in the clause types is the same: all the effects are less strong in AuxV than VAux clauses. Again, we suggest that this follows from the mix of G2 and G3 structures underlying AuxVO order; see Table 19.9. As some (unknown) percentage of AuxV
Table .. Grammars and their outputs
1 2 3 4 5
Grammar
Surface Order
Focus Movement?
Information Status of Object (Idealized)
G1 (VAux, OV) G1 (VAux, OV) G2 (AuxV, OV) G2 (AuxV, OV) G3 (AuxV, VO)
O V Aux V Aux O Aux O V Aux V O Aux V O
no yes no yes no
old new old new old or new
22 In all studies that we are aware of, the effect of length, however measured, is consistently stronger than that of discourse factors. It thus may be that length reflects (partially) a different underlying factor, perhaps processing cost, or it may simply be that length is easier to operationalize than discourse factors such as information status, and thus there is more ‘noise’ masking the effect in the latter case.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Ann Taylor and Susan Pintzuk
clauses (the ones generated by G3) have postverbal objects that are not necessarily triggered by discourse/performance-related effects, they are not affected by these factors, thus diluting the effects. Finally, we saw in Section 19.3.2 that non-referential objects (negative and incorporated) do not appear freely in postverbal position in VAux clauses. In general, incorporated objects are short, and it seems reasonable to assume that they cannot easily be focused.23 The same cannot be said of negatives, however. Rather, negatives appear to be barred from postverbal position for an entirely different reason, perhaps related to the fact that OE is a negative concord language; see Haeberli and Haegeman (1995) for details. Whatever the reason for the inability of negatives to postpose, the difference between VAux and AuxV clauses on this point is clear. If non-referential objects cannot move to postverbal position, as indicated by their distribution in VAux clauses, then all AuxV clauses with postverbal non-referential objects necessarily belong to G3. We also saw in Table 19.7 that in AuxV clauses, the frequency of postverbal non-incorporated objects is approximately twice as high as the frequency of incorporated ones (33 vs. 15 for all texts excluding Lives of Saints). For negatives (Table 19.6) the difference appears greater (47 non-negative to 14 negative), but note that this table includes negatives of any length. If we restrict the negatives to one open-class word to match the incorporated objects in Table 19.7, the difference is 35 non-negative to 15 negative, virtually identical to the frequency for incorporated objects. This distribution can be accounted for under the assumption that for referential objects (non-incorporated and non-negative) our AuxVO sample includes both base-generated postverbal objects and those derived by focus movement, while all the non-referential (incorporated and negative) objects in postverbal position are base-generated by the G3 grammar. This suggests, in addition, that the frequency of underlying VO order in OE overall is around 15.
. Conclusions In this chapter we set out to demonstrate that the distribution of objects differs in VAux and AuxV clauses in OE and to consider the implications for syntactic analyses of object position in OE. Our results show that the distribution of objects differs in VAux and AuxV clauses in three ways that may reasonably be interpreted as related to discourse/performance factors: 1) postverbal objects are less frequent in VAux clauses than in AuxV clauses; 2) factors associated with end position— information status, case, and grammatical weight—have a greater effect in VAux clauses than in AuxV clauses; and 3) non-referential objects do not freely appear postverbally in VAux clauses but do so in AuxV clauses. These results cannot be 23 The apparent exceptions in Lives of Saints require further investigation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Verb order, object position, and information status in Old English
reconciled with a single grammar in which object position is predominantly dependent on discourse/performance factors. Rather, they show that only in VAux clauses (G1) is object position solely dependent on these factors. In AuxV clauses (G2 and G3), object position is partly dependent on discourse factors (G2) and partly fixed syntactically (G3). The primary implication of our results for syntactic analyses of OE is that there need to be two ways to derive postverbal objects, one which involves focus (and/or other discourse/performance factors) and one which does not, and that these two derivations are associated with verb order: in VAux clauses only the G1 (discourse dependent) derivation is available, while in AuxV clauses both G2 (discourse dependent) and G3 (non-discourse dependent) derivations are available. We argue that this is the result of an ongoing change in the syntax of objects during the Old English period: the existence of non-focused objects in postverbal position in AuxV clauses is an innovation associated with the incoming G3 grammar.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic J O E L C . WA L L E N B E R G
. Introduction: Towards a unified theory of Heavy NP Shift This chapter will be somewhat unusual among modern syntax articles in that it very precisely identifies an analytical problem and describes a clear empirical result, but will not attempt to give a full solution to the problem.1 Instead, I would like to make the problem as clear as possible and spell out a number of elements that the correct solution must incorporate, even though the correct solution is difficult to state under current theoretical assumptions. In Section 20.2 I introduce a number of theoretical approaches to ‘Heavy NP Shift’ (HNPS) across Germanic, and argue in favour of a particular antisymmetric approach. Section 20.3 discusses the information structure of HNPS, and the antisymmetric approach correctly predicts HNPS to be a focus construction. In Section 20.4 I show a surprising diachronic result which challenges the antisymmetric account, and then Section 20.5 discusses this conclusion further in the light of synchronic data. Finally, I offer some conclusions concerning the different possible analyses.
. Towards a unified theory of Heavy NP Shift In Modern English ‘heavy’ DPs (i.e. long and/or accented in the prosody of the intonational phrase) can be displaced to the right of the clause, as in the now famous (with minor variations in the literature) example (1).
1 I would like to thank the attendees of DIGS for much helpful discussion of these issues, as well as the attendees of the Comparative Germanic Syntax Workshop in Tromsø, where a portion of this work appeared previously. I would also like to especially thank Anthony Kroch, Caitlin Light, and Liliane Haegeman for important comments and discussion at various stages of this work. I would also like to acknowledge that this work was funded by NSF grant OISE-. All errors are, of course, my own.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic (1)
a. I met my rich uncle from Detroit on the street. b. I met on the street my rich uncle from Detroit.
Going back nearly to Ross (1967), HNPS examples like (1) have been frequently analysed as involving rightward movement of the DP (e.g. ‘my rich uncle from Detroit’) to right-adjoin to either vP/VP or T(ense)P/IP, resulting in the order where the object follows the adjoined modifier (see rightward movement approaches like Baltin 1982, inter alia). I will refer to the rightward movement approach as the ‘classical’ analysis throughout this chapter, and contrast it with an ‘antisymmetric’ analysis to HNPS, under an approach to phrase structure that does not admit rightward movement (Kayne 1994). The analysis of HNPS which I will ultimately argue to be the best of the antisymmetric class of analyses depends on the ‘Split-C’ left periphery of Rizzi (1997 and much subsequent work), and follows the work in Roberts (2005). This antisymmetric account involves movement of the ‘shifted’ DP to the specifier of a FocP projection in an articulated CP domain, followed by remnant movement of the TP/IP to a higher specifier, Spec(TopP), as shown in Figure 20.1.2 I argue that the structure in Figure 20.1 leads to a more productive line of work on both the syntax and information structure of HNPS, and also allows for a unified analysis of object extraposition constructions across Germanic. This is the structure that I argued for in Wallenberg (2009), and its mechanics bear some similarity
TopP
TPj I met ti on the street
Top’
Top
FocP Foc’
DPi my rich uncle from Detroit
Foc
tj
Figure . Sketch of leftward A’-movement account of HNPS, assuming a split-CP left periphery 2 Note that, to avoid confusion, ‘TP’ is only ever an abbreviation for ‘Tense Phrase’ in this chapter. ‘Topic Phrase’, on the other hand, will only ever be abbreviated as ‘TopP’, and never as ‘TP’.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg
to proposals in Kayne (1998), Kayne (2005). Unfortunately, it suffers from certain empirical problems as I will show later in the chapter, but it seems to be the most plausible of the antisymmetric approaches. There are two other main candidates for the type of analysis that must be applied to HNPS under an antisymmetric phrase structure. The first potential analysis of HNPS without rightward movement is the one outlined in Kayne (1994: 74), in which the shifted object is not actually moved at all; rather, it is stranded by the movement of some other constituent (this is the style of analysis that Kayne suggests for all of the constructions that are classically derived with some type of rightward movement). In example (1), the object is stranded at the right of the clause by leftward movement of the PP, as shown in (2). (2) I met [ on the street ]i my rich uncle from Detroit ti . The Kayne (1994) approach suffers from two drawbacks: first, it cannot explain why HNPS licenses parasitic gaps, which indicates the the object itself is moved (by A’-movement), as in the sentence in (3). (3)
I met ti on the street, without recognizing immediately, [ my rich uncle from Detroit ]i .
Secondly, this analysis only applies to HNPS in modern English. It cannot provide a unified analysis for object extraposition constructions across Germanic, particularly in OV and Tense-final varieties. The following are examples of HNPS to the right of the finite verb in Tense-final West Germanic varieties (see also (7)). (4) Old English Ða æfter þam þe hi gewyld hæfdon eall heora feonda Then after that-dat that they controlled have all their enemies’ land land ‘After that time when they conquered all of their enemies’ land’ (Saint Eustace and his Companions, date: c. 11th century, from Taylor et al. 2003) (5)
Early Yiddish in dem kll iz oykh vas an bilngn iz zeyn hndl in the community is also what concern is his trade ‘In public, [it]’s also what concerns one’s business’ (Isaac ben Eliakum’s Preface to Lev Tov, date: 1620, from Santorini 1997/2008)
(6) Pennsylvania German Benjamen Y. Lapp ist Benjamen Y. Lapp is 9 Jahr 1 Mo und 9 years 1 month and
gestorben ten 14 den May 1915 ist alt worden died the 14th the May 1915 is old become 12 tag. 12 days
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
‘Benjamen Y. Lapp died May 14th, 1915. He was nine years, one month, and twelve days old.’ (Amish gravestone in Lancaster County, PA) Only the postposed object appears to the right of the verb in these examples. There are no adjuncts between the last verb and the postposed object in these examples, and there is no constituent that can be moved leftward to strand the object in final position. In this way, the stranding analysis of Kayne (1994) cannot account for this data at the same time as modern English HNPS. And since HNPS has existed throughout the history of English from Old English through Middle English (Pintzuk 1991; Pintzuk and Taylor 2006) and into modern English, with no apparent change even in its characteristic prosody (Pintzuk and Kroch 1989), a unified analysis of these constructions seems to be preferable. In fact, the diachronic data I show later in this chapter could represent a further indirect argument in favour of a unified approach. The second potential type of antisymmetric analysis is the one proposed for various extraposition constructions in Hinterhölzl (2006). This style of analysis generates OV Tense-final structures from underlying left-headed structures by individually ‘evacuating’ the constituents of the TP and vP leftward to the specifiers of various functional projections. HNPS in sentences like the ones in (4)–(6) can easily be generated under this approach, by stranding the object in situ and moving all of the other elements leftward individually. This approach could potentially apply to VO varieties as well, yielding a unified account of HNPS. Unfortunately, the analysis comes at the cost of proliferating as many unmotivated functional projections as there are possible constituents of the vP and TP. It also suffers from the same inability to account for HNPS’s licensing of parasitic gaps, as in the Kayne (1994) stranding account. And finally, the vP-evacuation account would place the adjunct in modern English sentences like (1) in the wrong place: Hinterhölzl proposes that evacuated constituents land to the left of the finite verb’s position, not to the right of v, where the PP actually surfaces in modern English. The classical analysis of HNPS, on the other hand, poses considerable empirical and theoretical problems. First, a system with both left- and right-adjunction is straightforwardly incompatible with the antisymmetric view of phrase structure in Kayne (1994) or Chomsky (1995b), and to the extent that these theories are interesting and useful, it is worth exploring other analyses of HNPS. Secondly, if HNPS is an optional rightward movement of an object to adjoin to vP or TP, then it is hierarchically very similar to scrambling. Under an analysis in which leftward scrambling is adjunction (e.g. Saito 1985; Webelhuth 1989), scrambling and HNPS become distinguishable only in terms of the direction of the movement. The idea that two syntactic operations are distinguished solely on the basis of their directionality rankles with the growing consensus in the literature that the narrow syntax operates on hierarchical structure and that linearization is a product of the syntax-PF interface. However, this is precisely the line taken in Saito and Fukui (1998), who explicitly claim that scrambling in
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg
Japanese and HNPS in English are the same operation (modulo language-specific linearization conditions, which apply at PF). The Saito and Fukui (1998) of HNPS as hierarchically non-distinct from scrambling suffers from (at least) two empirical problems: first, since scrambling is potentially unbounded, as shown by Comp-final languages like Japanese and Korean, there is no explanation for the Right Roof Constraint on HNPS (Ross 1967; Baltin 1982, and references therein). Secondly, they argue that left-headed languages have HNPS and right-headed languages have scrambling, so that the difference between the two operations is derivable from the head-parameter. This very clear hypothesis is falsified by any language that can scramble leftward and HNPS rightward simultaneously in the same clause. This is true of the Tense-final Germanic varieties mentioned above, as well as Early New High German as shown in (7). (7) ir welt Sewald Furzagel von mir ausreichtten III haundert und you will Sewald Furzagel from me pay three hundred and XXX guldn. thirty guilders ‘. . . I ask you to pay Sewald Furzagel three hundred and thirty guilders from me’ (from Bies 1996: 7) In (7), the indirect object is scrambled leftward across the PP while the direct object appears to the right of the non-finite verb by HNPS. The only analysis which does not suffer from the various problems I have discussed is the one in Figure 20.1. Because that analysis does not depend on the internal structure of the TP, it extends straightforwardly to Tense-final, Tense-initial, OV, and VO varieties. HNPS is clearly different from scrambling under this analysis, and since the shifted object moved to an A’-position, the fact that it licenses parasitic gaps is entirely expected. Finally, Roberts (2005) places Germanic complementizers in Force in an articulated CP domain, above TopP, which correctly predicts that HNPS should be available in subordinate clauses in Germanic (even in non-embedded-root contexts, as in (9) and (10)).
. The fine-grained information structure of HNPS The analysis in which the HNPS object moves to Spec(FocP) makes sense in light of the findings of Bies (1996) and Sapp (2011, 2012) that object extraposition in Early New High German and Middle High German is used when there is narrow focus on the shifted object. The following ENHG example from Bies (1996: 30) is particularly clear. (8) Ob er auch das wort Gots predig. Ja, prior hab whether he (the prior) also the word of God preached yes prior has predicirt [F festivus diebus]. preached [F festivus diebus]
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
The example clearly has focus on the shifted object because the rest of the clause is given in context and the shifted object provides the answer to the indirect question; the answer to an explicit question is always the focus of the sentence in which it appears. Similarly, it is clear from naturally occurring examples of modern English HNPS that the shifted object must be accented, as focus typically is in Germanic (Ladd 1996), and it is typically clear that the HNPS object is focused in context. The following two examples illustrate narrow focus on the HNPS object. (9) The letter of the former is somewhat imprudent, upon which I will communicate to him a piece of my mind; (Selections from the dispatches and general orders of Field Marshal the Duke of Wellington, date: 1815, in Kroch, Santorini, and Diertani 2010) (10) Nothing changes tragedy into comedy like gayness. It’s what we call in the entertainment world the GAY EX MACHINA. (from the ‘That’s Gay’ feature on the TV programme infoMania)3 In (9), the previous clause mentions a letter from a particular person, so communication with the person in question is assumed to be common ground between the speaker and hearer for the relative clause containing the HNPS. The purpose of the relative clause is only to convey the information of what the Duke will communicate to his correspondent, namely ‘a piece of his mind’. Sentence (10) has clear focus on the shifted object not only because ‘gay ex machina’ is the punchline to the joke, but also because the verb ‘call’ generally has its third argument focused: ‘call’ renames some known entity with a new (focused) name. The intuition that the preceding naturally occurring sentences represent a general pattern is confirmed by the following quantitative study. Tables 20.1 and 20.2 show data from two sets of personal letters in the York-Helsinki Corpus of Early English Correspondence (PCEEC, Taylor et al. 2006). Since the PCEEC contains samples of several personal letters by a number of individuals, it is possible to compare the individual systems of some of the letter writers. For the purposes of this study, I took the letters of two individuals, Nathaniel Bacon (born c.1546) and Anne Conway (born c.1619), who had letters of a sufficient length and containing enough clauses with the relevant syntactic context to make a quantitative study of HNPS possible. For each author, I extracted all clauses which contained a non-pronominal direct object, an auxiliary (to control for the effect of any V-to-T movement of main verbs), and a two-word (or longer) AdvP or PP.4 . Then for each example, I coded whether the 3
Thank you to Josef Fruehwald and Caitlin Light for this example. The condition on the length of the adverbial or PP was intended to exclude any elements that might be syntactic particles, in the sense of ‘up’ in ‘John picked the book up’. With the exception of Kayne (), there is a general consensus in the literature that the object-particle alternation in English is not the same phenomenon as HNPS (cf. den Dikken and references therein for an overview). 4
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg Table .. Heavy NP Shift and Focus for Nathaniel Bacon HNPS
In Situ
Total
Prop. HNPS
O Focus Other Focus
13 7
33 81
46 88
0.283 0.0796
Total Prop. O F.
20 0.650
114 0.289
134 0.343
0.149
Table .. Heavy NP Shift and Focus for Anne Conway HNPS
In Situ
Total
Prop. HNPS
O Focus Other Focus
6 4
20 56
26 60
0.231 0.067
Total Prop. O F.
10 0.600
76 0.263
86 0.302
0.116
focus was on the DP object or elsewhere in the sentence. As the results in Tables 20.1 and 20.2 show, the HNPS order of PP/AdvP preceding the object is more likely to show focus on the object than the unshifted order, Object > AdvP/PP (‘In Situ’ in the table). Both Nathaniel Bacon and Anne Conway are more likely to use HNPS when the object is focused, with 60 or more of the HNPS cases involving object focus for both speakers. However, HNPS seems to be more strongly associated with focus for Bacon than for Conway. The sample odds ratios show that Bacon is 4.6 times as likely to use the HNPS when there is focus on the object rather than on some other constituent, with chi-square = 8.3 on 1df, p = 0.004. Conway, on the other hand, is 4.2 times as likely to HNPS when there is focus on the object, and chi-square tests are just above the 0.05 Type I Error level (Pearson’s chi-square = 3.3 on 1df, p = 0.069; Fisher exact test p = 0.060). These results make it clear that HNPS is associated with object focus for these speakers, though there are two questions which merit further research: first, how significant are these individual differences in the use of HNPS? And second, why is HNPS used at all in some cases where there is not object focus? I hope that the first question can eventually be addressed with more samples (and larger samples) from a number of different speakers. The second issue may indicate that while object focus is a sufficient condition for HNPS, it may not be the only context in which HNPS is used. Thus, the labelling of the landing site for HNPS as the specifier of FocP may be something of an oversimplification, and I will return to this issue in the final section of the chapter.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
. Diachronic predictions The classical approach to HNPS and the antisymmetric approach described above make two clear predictions about the diachrony of the construction in English. It is well established that English began changing from a head-final TP to a head-initial TP during the Old English period (Pintzuk 1991, 2005), with the change going to completion some time during the Middle English period (Kroch and Taylor 2000b; Biberauer and Roberts 2005). Overlapping with that change was the change in the structure of vP/VP, in which head-initial (VO) vPs entered the language in late OE (Pintzuk and Taylor 2006) and eventually became the dominant pattern by Late Middle English (see references above, inter alia). The antisymmetric FocP/TopP approach makes the simplest prediction: diachronic stability. If HNPS is always movement of the object to Spec(FocP) and movement of the remnant TP to Spec(TopP), then none of the changes that are internal to the TP or the vP should have any effect on the frequency of HNPS. All other things being equal, the frequency of HNPS should be stable over time. The rightward movement account, on the other hand, makes a much more complex prediction: the observed frequency of HNPS should be the same in Tense-final, OV Old English as it is in Tense-initial, VO modern English. However, the frequency of HNPS should rise from the end of the Old English period into the Early Middle English period when OV vPs and Tenseinitial TPs overlap, and then fall again as OV vPs commence to leave the language at the end of the ME period. As we will see below, the surprising result is that this second prediction is borne out, providing strong evidence in favour of the classical account. According to the rightward movement account under a classical phrase structure, HNPS causes objects to right-adjoin freely to either vP or TP, regardless of the headedness of vP and TP. However, in OV Tense-final languages, interruption of the verb-cluster by adjoining any element between final v and T is disallowed (in cases where there is a finite auxiliary). (The unavailability of ∗ SVOI orders in Old English, and generally in Germanic, is discussed in Pintzuk 1991; Kiparsky 1997; Biberauer, Holmberg, and Roberts 2007, and in the subsequent literature on the ‘FinalOver-Final Constraint’.) Under a Kaynian phrase structure, the rightward adjunction analysis for HNPS is simply not available, and so this question does not arise. But for classical phrase structures admitting rightward adjunction, the absence of SVOI orders derived by HNPS cannot be due to the HNPS operation itself, and must rather be some kind of stipulated constraint, which I will take to be a PF-filter (∗ SVOI) for the sake of argument. Thus, while HNPS targets two adjunction sites, only one adjunction site ever actually surfaces in Tense-final OV Old English. All measurable HNPS in Tense-final Old English, in which the object is shifted to the right of the finite verb, is HNPS to TP-adjoined position. Once Tense-final phrase structures are mostly gone from the language by early Middle English (Kroch and Taylor 2000b), the ∗ SVOI filter no longer applies, because
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg
tensed verbs now come initially in the TP. At this point, HNPS is free to right-adjoin objects to either vP or TP. However, there is still a large number of head-final (OV) vPs in the language in Early Middle English (Kroch and Taylor 2000), and every time HNPS applies in a clause with an OV vP, it creates a VO-string from an underlying OV structure regardless of whether the landing site is vP-adjoined or TP-adjoined. Thus, in Tense-initial OV clauses, HNPS has an observable effect every time it applies. If it were possible to measure the frequency of HNPS application in this type of clause, then the frequency of HNPS should appear higher than the measurable frequency for Old English Tense-final clauses, because now both landing sites have measurable effects on the string (rather than just the TP-adjoined one, in OE). Finally, when English vPs become uniformly head-initial in Late Middle English and Early Modern English, HNPS can no longer be observed unless the object has been shifted past a PP or an adverbial at the right edge of the clause. If HNPS to vPadjoined position is present, it could only be observed when an object lands to the right of a specifically vP-adjoined adverb (e.g. a manner adverb) or a PP inside of the vP. Assuming that most clause-final adverbials and PPs will not be of this very restricted class of elements, most examples of HNPS from Early Modern English onward will be of the TP-adjoined type.5 Thus, observed HNPS will overwhelmingly reflect a single landing site, as in the Old English case, and so the measured frequency of HNPS should drop back down to the frequency measured for Tense-final Old English. Table 20.3 shows the frequency of HNPS in Tense-final Old English estimated by Pintzuk (2002) and my estimate for the frequency of HNPS in Early Modern English from the Penn Parsed Corpus of Early Modern English (Kroch, Santorini, and Diertani 2005).6 As in the previous section, my estimate is based on nominal DP objects shifted past two-word (or longer) AdvPs or PPs. In order for the results to be comparable to the Old English sample, I also restricted the count to subordinate clauses.7 The data show that the frequency of HNPS in Old English is nearly identical with the frequency in Early Modern English, a result that is compatible with either the antisymmetric or the classical analysis. Testing the prediction concerning the frequency of HNPS in Early Middle English is more difficult, since every text contains a mixture of OV and VO vPs, and it is not 5 In the future I intend to confirm this assumption by estimating the frequency of HNPS past clause-final vP-adjoined adverbials separately from other adverbials. However, this will take a considerable amount of work, as the available corpora do not make this distinction in their annotation. 6 Elizabeth I’s translation of Boethius was excluded, due to possible translation effects on the syntax from the Latin original (Anthony Kroch, p.c.) 7 Note that while the Old English sample did not include negative or quantified objects, because these are known to not extrapose in Old English (Pintzuk and Taylor ), I have included them in the EME sample because they do extrapose/HNPS in Early Modern English. If I remove them from the sample, there are clauses with HNPS and , in situ, which lowers the HNPS rate to . for EME. I am not sure why negative/quantified objects actually seem to shift more often than positive objects in EME, but it is possible that they are more often focused. In any case, this effect should be explored further, and while it is not predicted under any hypothesis in this section, I do not believe that it necessarily counts against any particular analysis.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
Table .. HNPS in Old English and Early Modern English
Old English EME
HNPS
In Situ
Total
Prop. HNPS
123 354
754 2120
877 2465
0.140 0.139
possible to tell what the underlying structure of the vP is for most clauses that show a surface VO string. It is also not possible to independently estimate the frequency of OV in Early Middle English using the position of pronominal objects in finite clauses, even though they do not participate in HNPS, due to a clitic construction that moves pronominal objects to a Tense-adjacent position (Pintzuk 1996; Kroch and Taylor 2000b) from either a VO or an OV vP (Wallenberg 2009). Instead, I estimated the frequency of OV for a number of Early Middle English texts using purpose infinitives in the Penn Parsed Corpus of Middle English (PPCME2, Kroch and Taylor 2000a), under the following assumptions: that purpose infinitives are adjunct-islands to extraction,8 and that infinitival clauses do not have a Tense node that attracts clitic pronouns.9 The estimates for OV are shown in Table 20.4, for two Early ME Southeastern texts (Vices and Virtues, date: c.1225 and Trinity Homilies, date: c.1225) and for Early ME West Midlands texts (Ancrene Riwle, date: c.1230 and the Katherine group of texts, date: c.1225). I collected a corresponding sample of purpose infinitives containing (positive) nominal objects for these texts in order to measure the surface rates of OV and VO, shown in Table 20.5. I could then use the estimates of underlying OV and VO from the pronominal objects in order to estimate the frequency of HNPS in Early ME, shown in the last column in Table 20.5. A nominal object preceding the infinitive represents an underlying OV clause in this sample, whereas the surface order Infinitive > Object represents a mixture of underlying VO clauses and some underlying OV clauses in which HNPS has applied to move the object to the right of the verb (see Kroch and
8 While purpose clauses are generally weak islands, I have found no examples in the PPCME of an object that was extracted out of a purpose infinitive by clitic movement. 9 This assumption follows the idea from Han and Kroch (), citing a manuscript by Mark Baltin, that English infinitival clauses are headed by MoodP and not TenseP, an idea which Han and Kroch () show makes correct predictions for other aspects of Early Modern English infinitives. While a full discussion of this topic is beyond the scope of this chapter, a comparison between the estimated frequencies of OV below and the frequencies of clitic movement for the same texts in Wallenberg () shows that the OV frequencies are too low to be due to a combination of OV plus clitic movement in these texts, so the assumption that the clitic-attracting head is missing in infinitival clauses seems to be correct. However, I do note that, if this is true, then the OE/ME clitic system is markedly different from the clitic system of the Romance languages, in which some functional head does attract clitics in non-finite clauses (frequently assumed to be non-finite Tense); cf. Kayne (), among many others.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg Table .. OV and VO, pronominal objects, in purpose infinitives in Early Middle English
Vices Trinity Ancrene Riwle Katherine Group
OV
VO
Total
Prop. OV
13 3 5 4
1 9 39 21
14 12 44 25
0.929 0.250 0.114 0.160
Table .. OV and VO, positive nominal objects in purpose infinitives, and estimate of HNPS
Vices Trinity Ancr. Kath.
OV
Total
Prop. OV
Estimate HNPS
8 6 5 2
20 63 84 35
0.400 0.095 0.059 0.057
0.569 0.619 0.476 0.643
Taylor 2000b for a full discussion of this reasoning and methodology). This state of affairs can be represented in the probability equation:10 P(SurfaceVO) = P(UnderVO ∩ (UnderOV ∪ HNPS))
(1)
This equation expands in the following way, according to basic probability theory: P(SurfVO) = P(UnderVO + (UnderOV ∪ HNPS))
(2)
P(SurfVO) = P(UnderVO + (P(UnderOV)P(HNPS))
(3)
P(HNPS) =
P(SurfaceVO) − P(UnderlyingVO) P(UnderlyingOV)
(4)
Applying the equation to the estimates of underlying OV and VO from the pronominal objects and the estimates of surface OV and VO from the nominal objects, it is possible to arrive at an estimate for the rate of HNPS for each text, as shown in Table 20.5. Although there is likely to be some noise due to the small sample sizes, all of the texts show a similar estimated frequency of HNPS, between 45–65, and they all 10 This treatment assumes that the choice of OV or VO vP phrase structure and the application of HNPS are independent events. It is a standard assumption throughout the historical syntax literature cited in this chapter that grammatically independent events are statistically independent, unless proven otherwise. I also assume that the occurrence of OV or VO are mutually exclusive events for a given clause.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
show a much higher frequency of HNPS than the roughly 14 in Old English and Early Modern English. This striking increase in the rate of HNPS specifically in the texts that show the most OV vPs combining with Tense-initial TPs (Kroch and Taylor 2000b) follows naturally from the classical account of HNPS, under which the two adjunction sites for HNPS are only both observable in this combination of phrase structures. Whether or not the rightward movement account is correct in detail, the clear effect of phrase structure change on the frequency of HNPS suggests that there must be two landing sites for HNPS, one at the vP level and another at the clausal level.
. The problem of vP-internal HNPS The quantitative effect in the previous section shows that the HNPS operation has always targeted two positions in the phrase structure throughout the history of English, one in the CP domain and one lower position. In this section, I show that it is possible to construct sentences of modern English which confirm the presence of a low, vP-internal position for HNPS, strengthening the conclusion that HNPS targets two different positions. This causes a theoretical problem for the antisymmetric account of HNPS, and I present one possibility for how to resolve it. As I mentioned in Section 20.2, HNPS licenses parasitic gaps. The examples (11) and (12) show that this is still possible even when the shifted DP originates inside a complex DP, which should be an island to extraction under any analysis of islandhood. Example (11) shows HNPS from within a non-finite clause complement of N, and example (12) shows HNPS from within a mixed nominalization. (11)
John hated his friend’s tendency to borrow, without returning, his new books on syntax.
(12)
John hated Mary’s praising, without reading, the new book on German.
Given the strong islandhood of the complex DPs in which the shifted DP originates, these sentences are not easily amenable to the antisymmetric analysis above, where the object must be moved out of the DP to a high left-periphery position. Rather, it looks like the shifted object has moved to a position within its governing vP, which is itself contained inside the complex DP. Since the shifted object can license a parasitic gap in these sentences, the low vP-internal landing site for HNPS must be an A’-position. So far, these effects are only predicted under the classical rightward movement account. It is also possible to have a reflexive inside the parasitic-gap-PP which is bound by an antecedent within the complex DP. In this case, the parasitic gap must be within the complex DP, not somewhere higher. Example (13) shows that the PP with the parasitic gap may be attached low, at the level of the vP inside of the complex DP, as the reflexive is bound by ‘Mary’. The sentences in (14) and (15) show that the PP containing the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Joel C. Wallenberg
parasitic gap must be attached low, because the reflexive cannot be bound by the matrix subject. (13)
John hated Mary’s praising, without reading herself, the new book on German.
(14)
∗ John
hated Mary’s praising, without reading himself, the new book on German.
(15)
∗ John hated Mary’s praising, before reading himself, the new book on German.
In (13), ‘Mary’ binds the PRO subject of ‘reading’, which in turn binds the reflexive ‘herself ’. In examples (14) and (15), the masculine reflexive inside of the PP with the parasitic gap can only be bound by the matrix subject, ‘John’. The PP must attach higher in these sentences, outside of the complex DP, so that ‘John’ binds the PRO rather than ‘Mary’ intervening and binding PRO, which would prevent the binding of ‘himself ’. This situation forces the landing site of the HNPS to be outside of the complex DP, since it can only attain that position in the string by moving past the high PP. This configuration results in ungrammaticality, presumably because of the island violation. Note that if we construct an analogous example with a non-finite clause complement in place of the complex DP, there is no island violation, and so the sentence is grammatical. (16) Prof. Smith ordered Mary to review, before reading himself, the new book on German syntax. I take this set of constructed examples to be clear evidence that HNPS can target an A’-landing site inside of vP, as well as a higher landing site. Given the data in this section and in the previous section, an antisymmetric account of HNPS is only tenable if it can allow for two targets of the HNPS movement, one low at the vP level and one higher in the CP domain. One approach that could make sense of vP-internal HNPS while still maintaining the theoretical attractiveness of antisymmetry is the idea of ‘low left periphery’ with a FocP and TopP, along the lines of Belletti (2004). If the sequence of movements I suggested for high TopP and FocP could be replicated and motivated within the vP, then the vP-internal HNPS could be derived in much the same way as the CP-level HNPS. However, Belletti’s proposals suggest that the low left periphery is inside the IP but above the vP/VP, which probably places the potential landing sites too high to resolve this problem, even if the sequence of movements that derive HNPS could be independently motivated. The judgment data in this section clearly show that HNPS can take place within the vP, even within mixed nominalizations. If it could be shown that such nominalizations have a TP layer, then the Belletti-style approach could be salvaged. Aside from this problem, as long as the landing site for HNPS is below Tense in the TP (which it is under Belletti’s proposals) the diachronic data in the previous section can be accommodated.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Antisymmetry and Heavy NP Shift across Germanic
The main argument against the Belletti-style low periphery account of HNPS is that HNPS requires two movements, the second of which is unmotivated: the object would move to a low Spec(FocP) and then the remnant vP would move to a low Spec(TopP). The latter movement must take place every time the prior movement takes place, or otherwise modern English would show surface OV orders when objects are focused. One possible solution to this problem is to assume that TP and vP must always move to the next highest Spec(TopP), high and low respectively, whether or not something else has been extracted from them. The proposal would be that every time the TP and vP form part of the topical material (or ‘Ground’ in Vallduví 1992), this must be marked syntactically. Then true topicalization would always be the further movement of a constituent even higher to Spec(ForceP). I leave the details of such a solution for later work.
. Conclusions and directions for future research This chapter has shown both synchronic and quantitative diachronic data which argue strongly that HNPS involves A’-movement of the object, and that it may move either to the edge of vP or to a higher position in the CP domain. Currently, only the classical rightward-movement approach captures all of these facts, though it suffers from a number of theoretical drawbacks. The FocP/TopP antisymmetric account does not suffer from the same concerns, but it cannot straightforwardly accommodate a vPinternal HNPS position. This chapter has also shown that diachronic quantitative data and synchronic judgment data can converge elegantly on a single conclusion. This case study is particularly striking in light of how specific the quantitative diachronic hypotheses were. The particular fact that HNPS is observable at a higher frequency in Tenseinitial OV phrase structures may also help to solve the problem of why there are so many postverbal objects in modern Yiddish, which has been argued to be OV (see references and discussion in Wallenberg 2009). This study has clarified the analytical problem of HNPS, but has not solved it. More research is necessary on both the empirical and theoretical fronts, and hopefully the results here can serve as a good foundation for further work.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese EDITH ALDRIDGE
. Introduction This chapter examines one of the lingering mysteries of Chinese historical syntax, the preverbal positioning of pronominal objects in the context of negation in Late Archaic Chinese1 (LAC). First, note that basic word order in LAC was SVO, with objects following verbs in unmarked order. (1)
a. वृࠌԳवաΖ Zhi zhe shi ren zhi ji wise d make other understand self ‘A wise person makes others understand him.’ b. 胡為而食我？ Hu wei er si wo? what for conj feed 1 ‘Why are you feeding me?’
(Xunzi 29)
(Lü Shi Chunqiu 12.3)
However, when one of several types of negation preceded the VP, pronominal objects raised from their base positions to appear between the negator and the verb. (2a) shows the clausal negator bu ‘not’, (2b) the quantificational negator mo ‘none’, and (2c) the aspectual negator wei ‘not yet’. (2) a. ݺ墋ۖլݺଇΖ Wo ji er bu wo si. 1 starve conj not 1 feed ‘When I was starving, (they) did not feed me.’
(Lü Shi Chunqiu 12.5)
1 I follow Wang () and Zhou () in identifying the Warring States period of the fifth–third centuries bce as a distinct period in the history of Chinese, which they term Late Archaic Chinese.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese b. լ൛๕աवΖ Bu huan mo ji zhi. not worry none self know ‘Do not worry that no one understands you’
(Analects 14)
c. ظհࠃΔآհፊՈΖ (Zuozhuan, Ai 11) Jia bing zhi shi, wei zhi wen ye. armor weapon gen matter not.yet 3.o hear nmlz ‘(Regarding) military matters, I have not heard of such things.’ The preverbal positioning of the object in (2) is often assumed to be a vestige of basic OV order in pre-Archaic Chinese (Li and Thompson 1974; Wang 1980; LaPolla 1994; Feng 1996; Xu 2006, among many others). Others have challenged the claim that pre-Archaic Chinese was an OV language and therefore assume OV to be a derived order (Djamouri 2005; Peyraube 1996; Shen 1992; Whitman and Paul 2005; Djamouri and Paul 2009, among others). This chapter similarly takes examples like those in (2) to be derived from underlying VO order. The question of the correct derivation, however, has remained elusive. In what follows, I offer a novel approach, proposing a syntactic derivation in which the motivation for the movement was the object’s need to value structural case. I further argue that changes in the morphological case system of Chinese provided the trigger for the subsequent loss of this transformation.
. Previous analyses Within the derivational camp, two basic approaches have been put forth to account for pronoun fronting to negation in LAC. Feng (1996) takes the position that the movement was prosodically motivated; pronouns underwent cliticization to the c-commanding negator. In (3), the clitic pronoun raises out of VP and right-adjoins to the negator dominating VP. (3)
NegP Neg Neg
VP Cli
V
ei
Though accounting for the position of the pronoun with respect to negation when fronting occurred, a prosodic approach offers no principled account of the lack of fronting in a wide variety of environments. For example, fronting generally did not take place in the context of perfective aspect marked by the particle yi, as shown in (4a). The complement of certain verbs likewise did not undergo fronting, as in the case of zai ‘be.in’ in (4b).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
(4) a. נԿֲΔլଇհߎΖ Chu san ri, [bu [shi zhi]] yi be.out 3 day not eat 3.o pfv ‘If it has been out for three days, don’t eat it anymore.’ b. ࠫլݺڇΖ Zhi bu zai wo control not be.in 1 ‘The control is not within me.’
(Analects 10)
(Guoyu, Jin 2)
Even more damaging perhaps is the fact that reflexive pronouns sometimes underwent fronting, as in (5a), but sometimes did not, as in (5b). (5)
a. ๕աवՈΖ ye. ej mo jii zhi none self know nmlz ‘No one understands me!’ b. ᆣԳլფաΖ Shengreni bu ai jii saint not love self ‘A saint does not love himself.’
(Analects 14)
(Xunzi 22)
These facts make it extremely doubtful that pronoun fronting to negation was prosodic cliticization. Furthermore, the cliticization analysis offers no explanation for why pronouns were attracted specifically to negation and did not front in other contexts. Addressing the question of why negation triggered object raising, Djamouri (1991) has proposed that pronoun fronting was a type of focus movement. Djamouri (2000) takes this position one step further by arguing that the clausal negator bu ‘not’ was a negative copula in pre-Archaic Chinese. He offers evidence of this in the Shang bone inscriptions (ظ֮) of the early archaic period (fourteenth–eleventh centuries bc). (6a) shows a copula used in an affirmative clause. The constituent following the copula is focused. Djamouri claims that this is a type of cleft construction. In (6b), the negator bu ‘not’ appears in the position for the copula. (6b) is parallel to (6a) in structure and interpretation, the constituent following bu also being focused. (6) a. ഄ׀Ԭࡏഡړ Wei fu yi jiu fu hao. only father Yi overwhelm Lady Hao ‘It is (the ancestral) father Yi who overwhelms Lady Hao.’ (Heji 6032 recto; from Djamouri 2000)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
b. լ׀Ԭࡏഡړ Bu fu yi jiu fu hao. not father Yi overwhelm Lady Hao ‘It is not (the ancestral) father Yi who overwhelms Lady Hao.’ (Heji 6032 recto; from Djamouri 2000) Djamouri suggests that pronoun fronting to negation originated in this type of focus construction. Specifically, the pronoun appears in the position for the focused constituent. This approach addresses the question of the involvement of negation in the raising. However, it remains unclear why pronouns are targeted for movement to the position following bu. Examples like (6b), in which full NPs follow the negator in a focus construction, are extremely rare. Djamouri admits that (6b) is the only such example he has found. He therefore suggests that (6b) is a relic of a historical pattern dating to pre-attested Chinese. But there is no direct evidence connecting (6b) with pronoun fronting to negation, which was already productive in this period, as shown in (7). (7) a. ظ׀୭ݺ Fu Jia hai wo father Jia harm us ‘Father Jia harms us.’ b. ظ׀լݺ୭ Fu Jia bu wo hai. father Jia not us harm ‘Father Jia does not harm us.’
(Heji 2122; from Djamouri 2000)
(Heji 2124; from Djamouri 2000)
There is, admittedly, some later evidence from the LAC period which indirectly suggests a possible connection between pronoun fronting and focus. Fronting generally did not take place in a clause containing a wh-word. Since the wh-word itself is the focused constituent, fronting of the pronoun does not take place (8) a. ֛۶լᢟݺՊ׆Λ Fuzi he bu tan wo yu wang? you why not praise me to king ‘Why don’t you speak in my favour to the king?’ լߠݺՊ׆Λ b. Hu bu jian wo yu wang? why not present me to king ‘Why don’t you present me to the king?’
(Zhuangzi 2.3)
(Mozi 50)
In the following section, I propose an alternative account which, like Djamouri (1991, 2000), is based on syntactic movement but addresses both the question of why only pronouns underwent fronting, as well as the involvement of negation.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
. Syntactic movement for case In this section, I propose an analysis of pronoun fronting to negation as object shift motivated by the need to value structural accusative case. This proposal draws heavily on similarities with genitive of negation in Slavic languages like Russian and Polish. In the following Russian examples, an object can (and often does) receive genitive case in the scope of sentential negation, as in (9a). It is also possible for the object to surface with accusative case, as in (9b). The difference in case-marking correlates with a difference in interpretation. The accusative object receives a definite interpretation, while the genitive object is indefinite. (9) Russian a. Anna ne kupila knig (indefinite) Anna.nom neg bought books.gen ‘Anna did not buy any books.’
(Harves 2002a: 97)
b. Anna ne kupil knigi (definite) Anna.nom neg bought books.acc ‘Anna did not buy the books.’ Many proposals have been made to account for case alternations like the one seen in (9). Most are in agreement that it is the Neg head which is the source of genitive case (Pesetsky 1982; Bailyn 1997; Brown 1999; Harves 2002a, 2002b; Witko 2008). Harves (2002b) implements this idea and accounts for the alternation in (9) in the following way. The Neg head can select a transitive vP with an accusative case feature on v, or it can select a defective vP in which accusative case is unavailable. If accusative case is unavailable within vP, then the object is dependent upon a higher functional head for case licensing and consequently values genitive case with Neg.2 The brackets around the DP in [SpecvP] indicate the trace position of the subject, which has moved to [SpecTP]. (10)
NegP Neg[GEN]
vP v’
v
VP V
DP[GEN]
2 At first blush, this Agree relation seems to violate the Phase Impenetrability Condition of Chomsky (). Harves proposes, however, that the v lacking a case feature is not a strong phase, allowing the case feature on Neg to probe into the VP.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
If Neg selects a transitive vP, then the object is able to value accusative case with v. (11)
NegP vP
Neg
v’
v[ACC]
VP V
DP[ACC]
On this approach, both genitive and accusative objects value their case in situ under Agree. In contrast to this, Brown (1999), Kim (2003, 2004), and Basilico (2008) propose that valuing of structural accusative case is the result of movement of the object. Taking Kim (2003, 2004) as an example, when the Neg head is present, it blocks an Agree relation between the in-situ object and the accusative case valuing head Asp, so the object receives genitive case instead. (12)
AspP AspP Asp[ACC]
NegP
Neg
VP V
DP[GEN]
The object receives accusative case if it moves out of the scope of negation to [SpecAspP]. This movement also accounts for the definite interpretation of the accusative object in examples like (9b). Assuming Diesing’s (1992) Mapping Hypothesis, movement of an object out of VP allows it to receive a presuppositional interpretation at LF. (13)
AspP DP[ACC]
AspP
Asp[ACC]
NegP Neg
VP V
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
There is additional evidence for the correlation between movement of the object and accusative case valuing. Harves (2002b) points out that if the object undergoes wh-movement, it has to have accusative case rather than genitive. (14) Russian a. Kakie fil’my ty ne smotriš’? which films-acc you neg watch ‘What kind of movies don’t you watch?’
(Harves 2002b)
b. ∗ Kakix fil’mov ty nikogda ne smotriš’? which films-gen you never neg watch ‘What kind of movies don’t you ever watch?’ The LAC data also clearly support a movement approach, since the pronoun surfaces to the left of the verb. However, Kim’s (2003, 2004) proposal cannot be adopted directly, because the fronted pronoun in LAC follows the negator in surface order. (15)
ݺ墋ۖլݺଇΖ Wo ji er bu wo si. 1 starve conj not 1 feed ‘When I was starving, (they) did not feed me.’
(Lü Shi Chunqiu 12.5)
For LAC, then, I propose that it is Neg which values accusative case on the fronted object. Similar to Harves, I propose that the head of NegP in LAC typically selected a complement in which structural case was unavailable. Specifically, I propose that Neg selected a nominal complement nP. Since structural case was unavailable, the object was forced to undergo object shift to the edge of nP, where it could exceptionally value accusative case with the head of NegP. I assume that the object could not value accusative case with Neg in situ in NP because n is a strong phase head, making NP an impenetrable domain from outside of the nP phase. (16)
NegP nP
Neg[acc] DP[acc]
n’ n’
n
NP N
Due to the unavailability of case in the domain of n, all DP objects with an unvalued case feature would have been required to move to [Spec, nP] in order to be licensed.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
The fact that only pronouns are spelled out in the landing site is accounted for in the following way. First, I follow Pesetsky (2000), Bobaljik (2002), and others in assuming that either the head or tail of a movement chain can be pronounced. Second, I assume with Halle and Marantz (1993), Harley and Noyer (1999), Embick and Noyer (2007), and many others that insertion of phonetic material takes place post-syntactically in the Morphological Component. Finally, I propose that the pronunciation of the head or tail of a chain depended on whether pronounceable features were inserted at the landing site.3 Case morphology was spelled out on pronouns in LAC but not on nouns, so pronouns were spelled out in the position where they acquired the feature. Nouns, on the other hand, did not display morphological case, so these were spelled out in their base positions. In Section 21.4, I present evidence for morphological case distinctions on pronouns, as well as for the nominalization of the predicate in negated clauses. In Section 21.7, I offer corroborating evidence for my analysis of chain resolution by showing that the loss of pronoun fronting to negation correlates diachronically with the loss of morphological case on pronouns. As in Slavic languages, it was also possible for Neg to select a transitive vP in LAC. I argue in Section 21.5 that this happened when independent factors required the presence of a vP layer. One such case is wh-questions, which Djamouri (2000) used as evidence for a connection between pronoun fronting and focus. (17)
ܩ۶լᜰհΛ Jun he bu ju zhi? you why not promote 3.o ‘Why don’t you promote him?’
(Hanfeizi 32)
The lack of fronting in wh-questions is in fact not a counterexample to the analysis put forth in this chapter but rather is predicted on this approach. As proposed by Aldridge (2010), LAC had short wh-movement to a focus position in the edge of vP. (18) shows movement of an object wh-word. Note that the landing site is above VP but below the surface position of the subject. (18)
ܠᓴུΛ
(Analects 9)
Wu [vP shei [v’ [VP qi ]]]? 1 who deceive ‘Who do I deceive?’ Given that wh-words had to occupy a specifier position in vP, the presence of a wh-word entails the presence of a vP layer and consequently a v to license a direct 3 This restriction applies to PF-interpretable features inserted in the morphological component and not to LF-interpretable features like focus and topic. Focus and topic chains are always resolved by spelling out the head in LAC. Furthermore, unlike modern Chinese varieties, LAC was a wh-movement language, so wh-phrases were likewise spelled out in their landing sites.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
object, allowing the object to remain in situ. This accounts for the lack of pronoun fronting in wh-questions. (19)
NegP Neg
vP v’
XP[WH]
v’
v[ACC]
VP V
DP[ACC]
In the next three sections, I present evidence for the main components of the analysis just sketched. The direct evidence is discussed in Section 21.4 and comes from the fact that morphological case distinctions were spelled out on pronouns but not on nouns in LAC. Additional support is offered from evidence that the predicate in fronting contexts is nominalized. Indirect evidence is presented in Section 21.5 and comes from the lack of fronting if independent factors forced Neg to select a vP rather than a nP. Section 21.6 offers evidence that pronoun fronting is object shift to the phase edge. Finally, I provide additional indirect evidence for the case-based approach by showing a correlation between the loss of pronoun fronting and the loss of morphological case distinctions on pronouns.
. Direct evidence: Case and nominalization In this section, I present evidence for the role of case valuing in accounting for LAC pronoun fronting to negation. Specifically, I show that only accusative pronouns underwent fronting, while objects of prepositions and inherently case-marked pronouns did not front. I interpret this as evidence that movement of the pronoun was motivated by the need to value accusative case. I also provide evidence in this section that the predicate was nominalized in fronting contexts. .. Morphological case Let me begin by presenting morphological evidence for case distinctions on pronouns in LAC. F. Zhou (1959), Yang and He (1992), and Zhang (2001) have identified some distributional restrictions on personal and demonstrative pronouns in the oracle bone inscriptions (fourteenth–eleventh centuries bc). Some pronouns tended to be used more often as possessors, others as subjects, and others as objects. Though the distinctions are no longer completely clear in the language, the tendencies do suggest a historical connection with case. The clearest constraint can be seen in the third person
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
pronoun հ zhi, which was never used in subject position. This restriction on ZHI continues into the Classical period (fifth–third centuries bc). In fact, this pronoun was specifically restricted to accusative case-marked positions in this period. In (20), ZHI functions as a direct object. (20) a. ᖂۖழհ Xue er shi [xi zhi] study conj time practice 3.o ‘To study and periodically practice something . . . ’ b. ֚ඔհߎΖ tian [qi zhi] yi. Heaven aid 3.o pfv ‘Heaven has aided him.’
(Analects 1)
(Zuozhuan, Min 1)
Additional evidence that ZHI is an accusative pronoun comes from ECM constructions. Although the subject of a non-finite clause is not able to value nominative case with embedded T, under certain conditions it can value accusative case ‘exceptionally’ with matrix v Consequently, this case must be analysed as structural accusative case. Following Woolford (2006), I assume that an inherent case is assigned either by a lexical category to its complement or a functional category to its specifier. Consequently, cases valued with a functional head in an Agree relation via c-command are structural cases, i.e. nominative (with finite T) and accusative (with transitive v). (21)
a. ՂᔃࠌհԿֆΖ (Xunzi 12) Shang xian shi [zhi wei sangong]. most able make 3.o be sangong ‘The most capable, make them into sangong (the highest official rank).’ (Analects 17) b. ലࡎृ֪נΔ࠷ᅖۖዚΔࠌհፊհΖ Jiangmingzhe chu hu, qu se er ge, shi [zhi wen zhi]. messenger exit door take zither and sing make 3.o hear 3.o ‘As the messenger was leaving, (Confucius) took up his zither and sang, making him (the messenger) hear it.’
ZHI is also clearly distinguished in this period from the possessive pronoun ࠡ qi ‘3.gen’. (Hanfeizi 31) (22) a. ࠡאଡ଼Ꮨ֜זسعΖ Yu yi qi zi Xiqi dai taizi Shensheng. want appl 3.gen son Xiqi replace heir Shensheng ‘(She) wanted to replace the heir Shensheng with her son Xiqi.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge b. 其子焉往？ qi zi yan wang? 3.gen son where go ‘Where would their sons go?’
(Mencius 7)
Finite embedded clauses are generally not found in LAC. Adjunct, complement, and relative clauses were typically nominalized, and the embedded subject appeared with genitive case. The genitive pronoun QI appears below in subject position of a complement clause in (23a) and a relative clause in (23b). (23) a. ࡌֆवࠡലఊۖࠌհፖΛ (Mencius 4) Zhou gong zhi [qi jiang pan] er shi zhi yu? Zhou duke know 3.gen will rebel conj send 3.o q ‘Did the duke of Zhou send him, knowing [that he would rebel]?’ b. ๕वࢬࠡޣբवृΖ Mo zhi qiu [qi suo yi zhi zhe]. none think seek 3.gen rel pfv know d ‘No one thinks to seek [what he already knows].’
(Zhuangzi 2.3)
The possessive pronoun QI is never attracted to negation. In (24a), QI is the subject of an embedded clause. In (24b), QI is a demonstrative in a DP. (Zhuangzi 1) (24) a. 鵬之背，不知其幾千里也。 Peng zhi bei, bu zhi [qi ji qian li ye]. bird gen back not know 3.gen how.many 1000 li nmlz ‘The back of the great bird, (I) do not know how many thousands of li it is long.’ (Analects 16) b. 吾聞其語矣，未見其人也Ζ Wu wen qi yu yi, wei jian [qi ren] ye. 1 hear 3.gen story pfv not.yet see 3.o person nmlz ‘I have heard such a story, but I have not seen such a person.’ One obvious explanation for the lack of fronting in (24) is locality, if we assume that argumental DPs and nominalized embedded clauses are islands to extraction. But the lack of fronting here is also expected on the case-based analysis. Genitive is an inherent case, so it is not affected by the lack of accusative case in the nominalized predicate. LAC had another pronoun which was used for internal arguments in certain environments and was clearly distinguished from the accusative pronoun ZHI. This pronoun ෫ yan surfaced in complement position of certain verbs, as can be seen in (25a, 25b). Here, the pronoun YAN is used instead of the accusative ZHI. (25c) shows that a full DP is accompanied by a goal preposition in this environment. It is widely recognized that YAN is functionally equivalent to ZHI following the goal preposition yu (He 1989; Pulleyblank 1995, and others). I analyse it as a dative pronoun.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese (25) a. ٣ܩհᐔڇ෫Ζ Xian jun zhi miao zai yan. former lord gen shrine be.in 3.dat ‘The former lord’s shrine is there.’ b. ׆լ៖෫Ζ Wang bu li yan. king not respect 3.dat ‘The king was not respectful toward him.’ c. ֛៖࣍ᇸࡱΖ Fuzi li yu Jia Ji. master respect to Jia Ji ‘The master is respectful toward Jia Ji.’
(Lü Shi Chunqiu 15.4)
(Zuozhuan, Yin 6)
(Zuozhuan, Wen 6)
What is crucial for the discussion at hand is the difference between ZHI and YAN in terms of fronting in the context of negation. The accusative pronoun ZHI does front in this context. (26) a. ܠ٣ٍܩ๕հ۩ՈΖ Wu xian jun yi mo zhi xing ye. 1 former lord too none 3.o do nmlz ‘None of our former lords did this either.’ b. ૨ளհࠃΔآհᖂՈΖ Junlü zhi shi, wei zhi xue ye. army gen matter not.yet 3.o study nmlz ‘Military matters, I have yet to study this.’
(Mencius 5)
(Analects 15)
In contrast, YAN did not front. On the case-based analysis, this is unsurprising, since dative-inherent dative case serves to license the object in the absence of an accusative feature. (27) a. ึߪլ塄෫Ζ zhongshen bu yang yan lifelong not care.for 3.dat ‘(He) was not cared for by them for the rest of his life.’ b. ֚Հ๕ൎ෫Ζ
(Mencius 8)
(Mencius 1)
Tianxia mo qiang yan. world none strong 3.dat ‘No one in the world is stronger than them.’ As shown above, 3rd person pronouns show a morphological distinction for accusative and dative case. The distinction is neutralized in the 1st person pronouns. However, the syntactic asymmetry in the context of negation is still observed. The
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
1st person pronoun fronts in an accusative environment (28a) but not in a dative environment (28b). This contrast is particularly difficult for Feng’s (1996) cliticization analysis to account for, given that the form of the pronoun is the same. (28) a. ݺ墋ۖլݺଇΖ Wo ji er bu wo si. 1 starve conj not 1 feed ‘When I was starving, (they) did not feed me.’ b. ࠫլݺڇΖ Zhi bu zai wo. control not be.in 1 ‘The control is not within me.’
(Lü Shi Chunqiu 12.5)
(Guoyu, Jin 2)
The asymmetry discussed above provides strong evidence for the case-based analysis of pronoun fronting, because it is only pronouns that need to value structural case which move, while those licensed in situ with inherent case do not front. Further support for the case-based approach comes from the fact that the object of a preposition never underwent fronting to negation. Assuming that the object is licensed internal to the PP, it is not affected by the lack of case on n. (29) a. լፖհञ౨Ζ Bu [yu zhi] zheng neng. not with 3.o dispute ability ‘(He) does not dispute ability with them.’ b. اլաشΔլաڽ Min bu [wei ji] yong . . . bu [wei ji] si, people not by self use not for self die ‘If the people cannot be used by you or will not die for you . . . ’
(Xunzi 12)
(Xunzi 12)
Note that the lack of fronting from a PP is not due to a locality restriction. Wh-movement and relativization were both possible from PPs. (30) a. ׆ᓴፖΛ Wang shei [yu e ] wei shan? king who with be good ‘With whom will the king be good?’ b. ႖հࢬ۞ದ luan zhi suo [zi e ] qi unrest gen rel from arise ‘from whence unrest arises’
(Mencius 6)
(Mozi 14)
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
.. Nominalization In this subsection, I discuss evidence that the complement of the Neg head was a nominal projection. Pronoun fronting is frequently found in the presence of the clause-final particle Ո ye, which I treat as a copula or nominalizer. (31)
a. լܠধՈΖ Bu wu pan ye. not 1 betray nmlz ‘(He) will not betray me.’ լܠᐒՈΖ b. Bu wu fei ye. not 1 disown nmlz ‘(He) will not disown me.’
(Zuozhuan, Xiang 31)
(Zuozhuan, Xiang 31)
There is compelling evidence for the analysis of YE as a copula or nominalizer.4 YE occurs with a nominal predicate in (32a), functioning as a copula. In (32b), YE nominalizes its complement so that it may be selected by the determiner ृ zhe (32b). (32c) shows YE following the predicate in a complement clause selected by a perception verb. Complements of perception verbs in LAC were generally required to be nominalized. (32) a. 非吾徒也。
(Analects 11)
Fei wu tu ye. not.be 1 student nmlz ‘(He) is not my student.’ b. ݬݕՈृΔࠡոհءፖΜ (Analects 1) Xiao ti ye zhe, qi wei ren zhi ben yu! pious respectful nmlz d 3.gen be benevolence 3.gen origin excl ‘Filial piety and brotherly respect, these are the foundation of benevolence!’ c. վֲլवࠡՋՈΖ Jin ri bu zhi qi wang ye. now day not know 3.gen gone nmlz ‘Now, you do not know that they have gone.’
(Mengzi 2)
In pronoun fronting contexts like (31), the presence of YE suggests that the predicate is nominalized, as I have proposed in Section 21.3. In this section, I have presented direct evidence for the case-based approach to pronoun fronting in LAC negated clauses introduced in Section 21.3. First, 4 Pulleyblank (, ) analyses Ո ye as an imperfective aspect marker. However, this claim is based to a large extent on the fact that pronoun fronting to negation takes place in the presence of Ո ye but not with the perfective aspect marker ߎ yi ‘Perf ’. Positive evidence for an aspectual analysis of Ո ye is not compelling.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
I established the existence of morphological case distinctions on LAC pronouns and showed that it is only accusative pronouns which undergo fronting to negation. This provides evidence for the proposal that fronting is motivated by the need to value structural accusative case, which is unavailable in the nP projection selected by clausal negation. In contrast to this, pronouns receiving inherent case like dative or genitive are unaffected by the lack of structural licensing and consequently do not front. The existence of morphological case distinctions on pronouns also provides a mechanism for distinguishing pronouns from full DPs in terms of where they are spelled out. Pronouns are spelled out in the landing site, since this is where the case morphology is introduced during vocabulary insertion. Since no morphology is added to full DPs, these are spelled out in their base positions. I have further provided evidence that the complement of the Neg head is a nominalized predicate.
. Indirect evidence: Availability of structural case obviates fronting In Section 21.3, I showed how the case-based analysis accounts for the lack of pronoun fronting in wh-questions. This is because wh-questions require the presence of a vP structure, which also entails the possibility of accusative case licensing from v. This proposal circumvents a potential counterargument posed by Djamouri’s (2000) focus analysis. In this section, I offer additional evidence showing that pronoun fronting is obviated if a vP structure is required for independent reasons. .. Aspect and fronting It is frequently noted that there is a connection between aspect and the availability of structural object case (Tenny 1987, 1994; van Voorst 1988; Bittner 1994; Borer 1994; Benua 1995; Kiparsky 1998; Ritter and Rosen 2000; Spreng 2006; Basilico 2008, and others). For example, Kiparsky (1998: 6) proposes for Finnish that an object has partitive case if it is governed by an unbounded verbal predicate or is itself quantitatively indeterminate. In (33), the appearance of accusative case on the object correlates with a bounded interpretation for the event. If a verb is intrinsically unbounded, it can only license partitive case on its object. (33)
Finnish etsi-n karhu-a/#karhu-n seek-1.sg bear-part/bear-acc ‘I’m looking for the (a) bear.’
(Kiparsky 1998: 3)
Interestingly, LAC pronouns frequently failed to undergo fronting in perfective clauses, as noted by F. Zhou (1959) and G. Zhou (1959).
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
(34) a. נԿֲΔլଇհߎΖ (Analects 10) Chu san ri, [bu [shi zhi]] yi. be.out three day not eat 3.o pfv ‘If it has been out for three days, don’t eat it any more.’ b. ֛वܠലشհΔؘլղߎݺΖ (Guoyu, 10) Fu zhi wu jiang yong zhi, bi [bu [yu wo]] yi. if know 1 will use 3.o then not give 1 pfv ‘If (he) knows that we are going to use him, then (he) won’t give (him) to us any more.’ The lack of pronoun fronting can be accounted for if we assume that LAC was like Finnish and other languages in making accusative case available in perfect aspect contexts. On the analysis presented in Section 21.3, the perfective aspect head selects a vP whose head can structurally license the object. This obviates movement of the object in order to be licensed. Note that the failure of pronoun fronting in perfective contexts presents an empirical problem for Feng’s (1996) prosodic analysis of pronoun fronting discussed in Section 21.2, since Feng treats pronoun fronting as a prosodic phenomenon, which should not be subject to semantic distinctions like aspect. .. Complex VPs There is another structural environment which suggests that Neg was forced to select a verbal complement under certain circumstances. Pronoun fronting did not take place from a VP containing more than one constituent other than the verb. (35) shows examples of ditransitive VPs. (35)
a. վՖլޣհ࣍ءΖ Jin ru bu [qiu zhi yu ben], now 2 not seek 3.o in root ‘Now, you do not seek it in the root.’ լఌհ࣍ܩΖ b. bu [chu zhi yu jun]. not subordinate 3.o to lord ‘(A vassal) will not subordinate this to his lord.’
(Xunzi 15)
(Hanfeizi 32)
Note first that the lack of fronting in (35) would be very mysterious on Feng’s (1996) prosodic analysis. The case-based account does, however, offer an explanation. The word order in (35) suggests that the verb has raised from its base position to v. On the assumption that incorporation of the verb to v is motivated by the need to categorize the root (in the sense of Marantz 1997), it seems safe to conclude that a verbal vP is required in these cases. Given that a verbal v also can carry an accusative case
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
feature, the pronominal object can be licensed in situ, thereby accounting for the lack of fronting in (35). This situation is mirrored in applicative constructions. Aldridge (2012) argues that the LAC functional morpheme yi heads a high applicative phrase (in the sense of Pylkkanen 2002) located above VP within vP. The Appl head YI moves to v, deriving the order YI-DP-VP. As in the case of ditransitive VPs just discussed, I assume that a vP layer is required to support movement of the applicative, making accusative case available for the object in VP, which is the pronoun ZHI in the examples in (36). (Analects 4) (36) a. լࠡאሐհΔլՈΖ Bu [vP yi [ApplP [qi dao] tyi [VP de zhi]]] bu chu ye. not appl 3.gen way get 3.o not stay nmlz ‘If one does not obtain them by the proper means, then they will not remain.’ b. 不以私害հΖ
(Xunzi 3)
Bu [vP yi [ApplP si tyi [VP hai zhi]]] not appl private harm 3.o ‘(He) does not damage it with private concerns.’ Because the applicative construction requires a full vP structure, it is not surprising that the applied object also does not undergo fronting. The pronoun can be licensed by v dominating the ApplP and consequently does not need to front. (Lü Shi Chunqiu 20.6) (37) a. վݺچհሐΔ ۖլݺאԿֆΖ wo wei sangong] Jin wo de di zhi dao, er bu [vP yi appl 1 be official now 1 obtain earth gen way conj not ‘I have achieved the way of the earth and yet you do not make me one of the sangong officials.’ b. ֛Գव׆հլאա݊ՈΖ ji wei du] Furen zhi wang zhi bu [vP yi appl self be jealous wife know king gen not ‘His wife knew the king did not take her to be jealous.’
(Hanfeizi 31)
In this section, I considered environments where pronouns did not front in the context of negation. I have shown that this is not a random occurrence but rather has a principled explanation which is consistent with the case-based analysis of fronting. Objects were required to front when they could not be case licensed in their base positions. But fronting was obviated if a vP structure was projected and the object could be licensed in situ.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
. Evidence for object shift In this section, I present one final argument in favour of analysing pronoun fronting as syntactic movement targeting the edge of nP. The LAC reflexive pronoun ա ji could be bound by the local subject, as in (38a), or long-distance bound, as in (38b). (38) a. ୴ृإաۖ৵࿇Ζ She zhe zheng ji er hou fa. shoot d correct self conj then shoot ‘An archer straightens himself and then shoots.’
(Mencius 3)
b. 壆ঀ༞ࠡ୭աΖ [qij hai jii ]. Zhuhoui wu feudal.lord dislike they harm self ‘The feudal lordsi dislike it that theyj inconvenience themi ’
(Mencius 10)
As a pronoun, ji would undergo object shift in the presence of negation. (39) a. լ൛ԳհլաवΖ huan [renj zhi bu jii zhi] ei bu not worry others gen not self know ‘Do not worry that others do not understand you.’ b. ๕աवՈΖ ye. moi jii zhi none self know nmlz ‘No one understands me!’
(Analects 1)
(Analects 14)
However, only long-distance bound ji could front. When ji was bound by the local subject, it did not front. (40) a. ႉԳۖլ؈աΖ ren er ei bu shi jii . ei shun accommodate person conj not lose self ‘(He) accommodates others and without losing himself.’ b. ᆣԳլფաΖ Shengreni bu ai jii saint not love self ‘A saint does not love himself.’
(Zhuangzi 3.4)
(Xunzi 22)
The asymmetry between (39) and (40) is not predicted by Feng’s (1996) cliticization approach, since prosodic cliticization should not be sensitive to Binding Principles. However, the pattern is accounted for on the present analysis. Since pronoun fronting to negation targets the outer specifier of nP, the pronoun lands in a position which c-commands the local subject. In cases like (40), in which the local subject is the
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
intended antecedent for the anaphor, a Binding Principle C violation will result. Consequently, the pronoun is unable to front when the antecedent is the local subject. The derivation will converge, however, if Neg selects a vP and the object is licensed in situ. To summarize the previous three sections, I have argued that pronoun fronting to negation was syntactic object shift to an outer specifier of nP in order to value structural accusative case. I have shown that only pronouns needing structural accusative case underwent fronting and that this fronting was obviated if case was available for the object in situ.
. Loss of pronoun fronting Pronoun fronting to negation was no longer a productive process by the end of the Han Dynasty (roughly 200 ce). In this section, I show that this change was concomitant with the loss of case distinctions on pronouns, thus providing a morphological trigger for the change in the syntax. The loss of morphological case is evidenced in Han period texts by the distribution of pronominal forms which did not correlate with the case these arguments valued in the syntax. A specific example is the breakdown in the complementarity between the accusative pronoun zhi and the genitive pronoun qi. Recall from Section 21.4 that the accusative pronoun ZHI appeared in the position of an exceptionally case-marked embedded subject in the Classical period. (41) ࠌհፊհΖ shi [zhi wen zhi ] make 3.o hear 3.o ‘making him (the messenger) hear it.’
(Analects 17)
From the Han period, however, we can find examples of the genitive pronoun in this position. (42) a. ٤᎓ַࠡחΖ zhi] Quan Zhao ling [TP qi 3.gen stop protect Zhao make ‘(You) protect Zhao and make them stop.’
(Zhanguoce, Xi Zhou)
(Shishuo Xinyu, Yanyu) b. ࠌࠡ࣍سၸஅۘΖ Yu shi [qi sheng yu jieting] er. want make 3.gen grow in courtyard only ‘(You) only want to make them grow in a courtyard.’
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Pronominal object shift in Archaic Chinese
When used in monoclausal contexts, however, the object following these verbs still had to be the accusative pronoun and was never the genitive pronoun. (Zhanguoce, Qin 5) (43) a. ׆ԯ״ઌΔחհֳΚ Wang nai zhao xiang, ling zhi yue: king then call minister order 3.o say ‘The king then summoned his prime minister and issued him an order, saying . . .’ b. Օױ׆ᇢࠌհΖ Da wang ke shi shi zhi great king can try send 3.o ‘Your majesty could try to send him.’
(Zhanguoce, Zhao 4)
What this asymmetry shows here is that the use of pronominal forms no longer correlates with the abstract cases valued in the syntax, since direct objects and ECM subjects both value accusative case in the syntax. It appears that the difference between accusative and genitive pronouns was reinterpreted in terms of structural position rather than morphological case. In the Han period, there was a tendency to use the genitive pronoun for embedded subjects, regardless of the case which they valued in the syntax. In Middle Chinese Buddhist writings, this generalization was extended to all pronouns occupying specifier positions. Consequently, not only ECM subjects but even the specifier of a ditransitive VP (44a) or the subject of a small clause (44b) is also represented with genitive QI and not accusative ZHI.5 The following texts are fifth-century translations of Buddhist sutras. (44) a. ഄᣋ༇ፖࠡ९ኂΖ Wei yuan Shizun yu qi changshou. only wish Buddha give 3.gen longevity ‘(I) only ask the Buddha to give him long life.’ b. ߠࠡڕਢΔᇢհΖ Jian qi ru shi, yu wang shi zhi. see 3.gen like this want go test 3.o ‘Seeing him thus, (he) want to go and test him.’
(Zabao Zangjing 47)
(Xianyu Jing 1)
I interpret the appearance of the genitive pronoun in positions historically reserved for the accusative pronoun as evidence that, at least in the spoken language, the morphological distinction between these two pronouns had been lost. I posit the following relationship between the loss of morphological case distinctions and the loss of pronoun fronting to negation. First, the lack of morphological case 5 Note that QI in (a) cannot be interpreted as a possessor of the direct object. Possessed DPs in premodern Chinese were always definite. But the referent of the direct object in this example is introduced into the discourse for the first time with this sentence.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Edith Aldridge
meant that both pronouns and full DPs came to be spelled out in their base positions, regardless of whether object shift took place in the syntax. The postverbal position of the pronominal object in the input data subsequently obscured the existence of the transformation to language learners. This also deprived them of evidence for positing the nP complement of Neg and prompted them instead to acquire the default structure in which Neg consistently selects a vP.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References Abels, Klaus (2012). ‘The Italian left periphery: a view from locality’, Linguistic Inquiry 43: 229–54. Aboh, Enoch (2004). ‘Topic and focus within D∗ ’, Linguistics in the Netherlands 21: 1–12. Aboh, Enoch (2008). ‘Information structuring begins with the numeration’, ms., University of Amsterdam. Aboh, Enoch, Corver, Norbert, Dyakonova, Maria, and Koppen, Marjo van (2010). ‘DP-internal information structure: some introductory remarks’, Lingua 782–801. Abraham, Werner (1991). ‘Null subjects: from Gothic, Old High German and Middle High German to Modern German. From pro-drop to semi-pro-drop’, Groninger Arbeiten zur Germanistischen Linguistik 34: 1–28. Ackema, Peter, and Neeleman, Ad (2007). ‘Restricted pro drop in Early Modern Dutch’, Journal of Comparative Germanic Linguistics 10: 81–107. Adger, David, and Harbour, Daniel (2007). ‘The syntax and syncretisms of the Person Case Constraint’, Syntax 10: 2–37. Aikhenvald, Alexandra Y. (2004). Evidentiality. Oxford: Oxford University Press. Aldridge, Edith (2010). ‘Clause-internal Wh-movement in Archaic Chinese’, Journal of East Asian Linguistics 19: 1–36. Aldridge, Edith (2012). ‘PPs and Applicatives in Late Archaic Chinese’, Studies in Chinese Linguistics 33: 139–64. Alexiadou, Artemis (2014). Multiple determiners and the structure of DPs. Amsterdam: John Benjamins. Alexiadou, Artemis, and Anagnostopoulou, Elena (1998). ‘Parametrizing AGR: word order, movement and EPP-checking’, Natural Language and Linguistic Theory 16, 491–539. Alexiadou, Artemis, Haegeman, Liliane, and Stavrou, Melita (2007). Noun Phrase in the Generative Perspective. Berlin: Mouton de Gruyter. Allen, Cynthia (2001). ‘The development of a new passive in English’, in Miriam Butt and Tracy Holloway King (eds.), Time over matter. Stanford, CA: CSLI Publications, 43–72. Allen, Cynthia L. (1990). ‘Review of van Kemenade (1987)’, Language 66: 146–52. Ambar, Manuela (1999). ‘Apects of the syntax of focus in Portuguese’, in Georges Rebuschi and Laurice Tuller (eds.), The grammar of focus. Amsterdam: John Benjamins, 23–53. Anagnostopoulou, Elena (1999). ‘On experiencers’, in Artemis Alexiadou, Geoffrey Horrocks, and Melita Stavrou (eds.), Studies in Greek syntax. Dordrecht: Kluwer, 67–93. Anagnostopoulou, Elena (2003). The syntax of ditransitives: evidence from clitics. Berlin: Mouton de Gruyter. Anagnostopoulou, Elena, and Sevdali, Christina (2012). ‘Case alterations in Ancient Greek passives and the typology of Case’, ms., Universities of Crete and Ulster. Anderson, Stephen R. (1990). ‘The grammar of Icelandic verbs in-st’, in Joan Maling and Annie Zaenen (eds.), Modern Icelandic syntax. San Diego: Academic Press, 235–73.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Anderson, Stephen R. (1992). A-morphous morphology. Cambridge: Cambridge University Press. Árnadóttir, Hlíf, Eythórsson, Þórhallur, and Sigurðsson, Einar Freyr (2011). ‘The passive of reflexive verbs in Icelandic’, in Tania Strahan (ed.), Nordlyd 37: relating to reflexives. Tromsø: CASTL, 39–97. Arnold, Jennifer E., Losongco, Anthony, Wasow, Thomas, and Ginstrom, Ryan (2000). ‘Heaviness vs. newness: the effects of structural complexity and discourse status on constituent ordering’, Language 76: 128–55. Asbury, Anna (2008). The morphosyntax of case and adpositions. Utrecht: LOT. Asbury, Anna, Gehrke, Berit, and Hegedűs, Veronika (2007). ‘One size fits all: prefixes, particles, adpositions and cases as members of the category P’, in Cem Keskin (ed.), UiL OTS yearbook 2006, Utrecht: LOT, 1–17. Assenova, P. (2002). Balkansko ezikoznanie. Osnovni problemi na Balkanskija ezikov săjuz. Sofia: Fabe. Auwera, Johan van der (2009). ‘The Jespersen cycles’, in Elly van Gelderen (ed.), Cyclical change. Amsterdam: John Benjamins, 35–71. Axel, Katrin (2007). Studies on Old High German syntax: left sentence periphery, verb placement and verb-second. Amsterdam: John Benjamins. Baayen, R. Harald (2006). Analyzing linguistic data. Cambridge: Cambridge University Press. Baayen, R. Harald (2010). LanguageR library for R. Bader, Markus, and Häussler, Jana (2010a). ‘The primacy of grammaticality’, ms., University of Konstanz. Bader, Markus, and Häussler, Jana (2010b). ‘Toward a model of grammaticality judgments’, Journal of Linguistics 46, 273–330. Bailey, Charles-James N. (1981). ‘Developmental linguistics’, Folia Linguistica 15: 29–38. Bailey, Charles-James N. (1996). Essays on time-based linguistic analysis. Oxford: Clarendon. Bailyn, John (1997). ‘Genitive of negation is obligatory’, in Wayles Browne, Eva Dornisch, Natalia Kondrashova, and Draga Zec (eds.), Proceedings of the 4th workshop on formal approaches to Slavic linguistics: the Cornell meeting. Ann Arbor, MI: Michigan Slavic Publications, 84–114. Bailyn, John (2003). ‘Does Russian scrambling exist?’, in Simin Karimi (ed.), Word order and scrambling. Oxford: Blackwell, 156–76. Baker, Mark C. (1985). ‘The Mirror Principle and morphosyntactic explanation’, Linguistic Inquiry 16: 373–415. Baker, Mark C. (2008). ‘The macroparameter in a microparametric world’, in Theresa Biberauer (ed.), The limits of syntactic variation. Amsterdam: John Benjamins, 351–73. Baker, Mark C., Johnson, Kyle, and Roberts, Ian G. (1989). ‘Passive arguments raised’, Linguistic Inquiry 20: 219–51. Bakht, Maryam (2010). ‘Lexical variation and the negotiation of linguistic style in a Long Island middle school’, PhD dissertation, NYU. Baltin, Mark R. (1982). ‘A landing site theory of movement rules’, Linguistic Inquiry 13: 1–38. Bane, Max (2008). ‘Quantifying and measuring morphological complexity’, in Charles B. Chang and Hannah J. Haynie (eds.), Proceedings of the 26th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Press, 69–86. Barbiers, Sjef (1995). The syntax of interpretation. The Hague: Holland Academic Graphics.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Bard, Ellen G., Robertson, Dan, and Sorace, Antonella (1996). ‘Magnitude estimation of linguistic acceptability’, Language 72: 32–68. Barðdal, Jóhanna, and Molnár, Valéria (2003). ‘The passive in Icelandic compared to Mainland Scandinavian’, in Jorunn Hetland and Valéria Molnár (eds.), Structures of focus and grammatical relations. Tübingen: Niemeyer, 231–60. Basilico, David (2008). ‘The syntactic representation of perfectivity’, Lingua 118: 1716–39. Bates, Doug, and Maechler, Martin (2010). Lme4 package for R. Batllori, Montserrat (1993). ‘Participle preposing and other related phenomena in Old Spanish and Old Catalan’, paper presented at the 2nd Manchester University Postgraduate Linguistics Conference, University of Manchester. Batllori, Montserrat, and Hernanz, Maria-Lluïsa (2008). ‘Emphatic polarity from Latin to Romance’, poster presented at the 10th Diachronic Generative Syntax Conference, Cornell University. Batllori, Montserrat, and Hernanz, Maria-Lluïsa (2013). ‘Emphatic polarity particles in Spanish and Catalan’, Lingua 128: 9–30. Batllori, Montserrat, Hernanz, Maria-Lluïsa, Picallo, Carme, and Roca, Francesc (eds.) (2005). Grammaticalization and parametric variation. Oxford: Oxford University Press. Batllori, Montserrat, Iglésias, Narcís, and Martins, Ana Maria (2005). ‘Sintaxi dels clítics pronominals en català medieval’, Caplletra 38: 137–77. Battye, Adrian, and Roberts, Ian G. (eds.) (1995). Clause structure and language change. Oxford: Oxford University Press. Bauer, Brigitte (1995). The emergence and development of SVO patterning in Latin and French. Oxford: Oxford University Press. Bayer, Josef (2001). ‘Asymmetry in emphatic topicalization’, in Caroline Féry and Wolfgang Sternefeld (eds.), Audiatur vox sapientiae: a festschrift for Arnim von Stechow. Berlin: Akademie Verlag, 15–47. Bayer, Josef, and Suchsland, Peter (1997). ‘Expletiva und leere Subjekte im Deutschen’, Groninger Arbeiten zur Germanistischen Linguistik 41: 12–38. Bech, Kristin (2001). ‘Word order patterns in Old and Middle English: a syntactic and pragmatic study’, PhD dissertation, University of Bergen. Behaghel, Otto (1897). Die Syntax des Heliand. Vienna: Tempsky. Behaghel, Otto (1932). Deutsche Syntax: eine geschichtliche Darstellung, vol. 4: Wortstellung. Heidelberg: Carl Winter. Belletti, Adriana (2001). ‘Inversion as focalization’, in Aafke Hulk and Jean-Yves Pollock (eds.), Subject inversion in Romance and the theory of Universal Grammar. Oxford: Oxford University Press, 60–90. Belletti, Adriana (2004). ‘Aspects of the low IP area’, in Luigi Rizzi (ed.), The cartography of syntactic structures, vol. 2: The structure of CP and IP. Oxford: Oxford University Press, 16–51. Belletti, Adriana (2005). ‘Extended doubling and the VP periphery’, Probus 17: 1–35. Belletti, Adriana, and Rizzi, Luigi (2012). ‘Moving verbal chunks in the low functional field’, in Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, and Cecilia Poletto (eds.), Functional heads. Oxford: Oxford University Press, 129–37. Benediktsdóttir, Ásbjörg (2008). ‘Nýja þolmyndin: Fyrsta þolmyndun barna?’ [The new passive: children’s first passivizing?], BA dissertation, University of Iceland.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Benincà, Paola (1983). ‘Il clitico a nel dialetto padovano’, in Paola Benincà, Michele Cortelozzo, Aldo Luigi Prosdocimi, Laura Vanelli, and Alberto Zambone (eds.), Scritti linguistici in onore di Giovan Battista Pellegrini. Pisa: Pacini, 25–35. Benincà, Paola (1988). ‘L’ordine degli elementi della frase e le costruzioni marcate’, in Lorenzo Renzi and Giampaolo Salvi (eds.), Grande Grammatica Italiana di Consultazione, vol. 1. Bologna: Il Mulino, 129–94. Benincà, Paola (2001). ‘The position of Topic and Focus in the left periphery’, in Guglielmo Cinque and Giampaolo Salvi (eds.), Current studies in Italian syntax: essays offered to Lorenzo Renzi. Amsterdam: North-Holland, 39–64. Benincà, Paola (2004). ‘The left periphery of medieval Romance’, Studi Linguistici e Filologici Online. Rivista Telematica del Dipartimento di Linguistica dell’Università di Pisa: 243–97. Benincà, Paola, and Poletto, Cecilia (2004). ‘Topic, focus and V2: defining the CP sublayers’, in Luigi Rizzi (ed.), The cartography of syntactic structures, vol. 2: The structure of CP and IP. Oxford: Oxford University Press, 52–75. Benkő, Loránd (1980). Az Árpád-kor magyar nyelvű szövegemlékei. [Hungarian texts from the Árpád-era.] Budapest: Akadémiai Kiadó. Benua, Laura (1995). ‘Yup’ik antipassive’, in Audra Dainora, Rachel Hemphill, Barbara Luka, Barbara Need, and Sheri Pargman (eds.), CLS 31: papers from the 31st Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 28–44. Bergsland, Knut (1994). Sydsamisk grammatikk. Kárášjohka: Davvi Girji. Bermúdez-Otero, Ricardo, and Trousdale, Graeme (2012). ‘Cycles and continua: on unidirectionality and gradualness in language change’, in Terttu Nevalainen and Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English. Oxford: Oxford University Press, 691–720. Bernódusson, Helgi (1982). ‘Ópersónulegar setningar’ [Impersonal Sentences], Cand. mag. dissertation, University of Iceland. Bernstein, Judy (1991). ‘DPs in French and Walloon: evidence for parametric variation in nominal head movement’, Probus 3: 101–26. Bernstein, Judy (2001). ‘The DP hypothesis: identifying clausal properties in the nominal domain’, in Mark R. Baltin and Chris Collins (eds.), The handbook of contemporary syntactic theory. Oxford: Blackwell, 536–61. Bertocci, David, and Damonte, Federico (2007). ‘Distribuzione e morfologia dei congiuntivi in alcune varietà salentine’, in Federico Damonte and Jacopo Garzonio (eds.), Studi sui dialetti della Puglia. Padua: Unipress, 3–28. Berwick, Robert, and Noam Chomsky (2011). ‘Biolinguistics: The current state of its evolution and development’, in Anna Maria Di Sciullo and Cedric Boeckx (eds.), Biolinguistic investigations. Oxford: Oxford University Press, 19–41. Besten, Hans den (1977). ‘On the interaction of root transformations and lexical deletive verbs’, ms., University of Amsterdam. Bianchi, Valentina (1999). Consequences of Antisymmetry: Headed relative clauses. Berlin: Mouton de Gruyter. Biberauer, Theresa (2004). ‘Reconsidering the EPP and Spec-TP in Germanic’, in Lluïsa Astruc and Marc Richards (eds.), Cambridge Occasional Papers in Linguistics (COPiL) 1: 15–40.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Biberauer, Theresa (2008). ‘Introduction’, in Theresa Biberauer (ed.), The limits of syntactic variation. Amsterdam: John Benjamins, 1–72. Biberauer, Theresa (2009). ‘Jespersen off course?’, in Elly van Gelderen (ed.), Cyclical change. Amsterdam: John Benjamins, 91–130. Biberauer, Theresa, Holmberg, Anders, and Roberts, Ian G. (2007). ‘Disharmonic word orders and the Final-over-Final Constraint’, in Antonietta Bisetto and Francesco E. Barbieri (eds.), Proceedings of the XXXIII Incontro di Grammatica Generativa. Bologna: Universita di Bologna, 86–105. Biberauer, Theresa, Holmberg, Anders, and Roberts, Ian G. (2010). ‘A syntactic universal and its consequences’, ms., University of Cambridge. Biberauer, Theresa, Holmberg, Anders, and Roberts, Ian G. (2014). ‘A syntactic universal and its consequences’, Linguistic Inquiry 45: 169–225. Biberauer, Theresa, and Roberts, Ian G. (2006). ‘Loss of residual “head-final” orders and remnant fronting in Late Middle English: causes and consequences’, in Jutta M. Hartmann and Laszlo Molnárfi (eds.), Comparative studies in Germanic syntax. Amsterdam: John Benjamins, 263–97. Biberauer, Theresa, and Roberts, Ian G. (2008). ‘Cascading parameter changes: internally driven change in Middle and Early Modern English’, in Þórhallur Eythórsson (ed.), Grammatical change and linguistic theory: the Rosendal papers. Amsterdam: John Benjamins, 79–113. Biberauer, Theresa, and Roberts, Ian G. (2010). ‘Subjects, Tense and verb-movement’, in Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan (eds.), Parametric variation: null subjects in Minimalist theory. Cambridge: Cambridge University Press, 263–302. Biberauer, Theresa, and Roberts, Ian G. (2011). ‘Negative words and related expressions: a new perspective on some familiar puzzles’, in Richard Ingham and Pierre Larrivée (eds.), The evolution of negation: beyond the Jespersen Cycle. Berlin: Mouton de Gruyter, 23–60. Biberauer, Theresa, and Roberts, Ian G. (2014). ‘Parameter setting’, to appear in Adam Ledgeway and Ian Roberts (eds.), The Cambridge handbook of historical syntax. Cambridge: Cambridge University Press. Bies, Ann (1996). ‘Syntax and discourse factors in Early New High German: evidence for verbfinal word order’, MA dissertation, University of Pennsylvania. Birner, Betty (2006). ‘Inferential relations and noncanonical word order’, in Betty Birner and Gregory Ward (eds.), Drawing the boundaries of meaning: Neo-Gricean studies in pragmatics and semantics in honor of Laurence R. Horn. Amsterdam: John Benjamins, 31–51. Bittner, Maria (1994). Case, scope, and binding. Dordrecht: Kluwer. Blain, Eleanor M., and Déchaine, Rose-Marie (2007). ‘Evidential types: evidence from Cree dialects’, International Journal of American Linguistics 73: 257–91. Blevins, James P. (2003). ‘Passives and impersonals’, Journal of Linguistics 39: 473–520. Blyth, Carl, Recktenwald, Sigrid, and Wang, Jenny (1990). ‘I’m like, “say what?!”: a new quotative in American oral narrative’, American Speech 65: 215–27. Bobaljik, Jonathan D. (2001). ‘The implications of rich agreement: why morphology doesn’t drive syntax’, in Karine Megerdoomian and Leora A. Bar-El (eds.), Proceedings of the 20th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla Press, 82–95. Bobaljik, Jonathan D. (2002a). ‘Realizing Germanic inflection: why morphology does not drive syntax’, Journal of Comparative Germanic Linguistics 6: 129–67.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Bobaljik, Jonathan D. (2002b). ‘A-chains and the PF-interface: copies and “covert” movement’, Natural Language and Linguistic Theory 20: 197–267. Bobaljik, Jonathan D., and Thráinsson, Höskuldur (1998). ‘Two heads aren’t always better than one’, Syntax 1: 37–71. Boeckx, Cedric (2000). ‘Quirky agreement’, Studia Linguistica 54: 354–80. Boeckx, Cedric (2010). ‘What Principles and Parameters got wrong’, ms., ICREA/UAB. Borer, Hagit (1984). Parametric syntax. Dordrecht: Foris. Borer, Hagit (1994). ‘The projection of arguments’, in Elena E. Benedicto and Jeffrey T. Runner (eds.), Functional projections. Amherst, MA: University of Massachusetts, 19–47. Borer, Hagit (2005). In name only. Oxford: Oxford University Press. Borsley, Robert D., Tallerman, Maggie, and Willis, David (2007). The syntax of Welsh (Cambridge Syntax Guides). Cambridge: Cambridge University Press. Bošković, Željko (1995). ‘Participle movement and second position cliticisation in SerboCroatian’, Lingua 96: 245–66. Bošković, Željko (2001). On the nature of the syntax-phonology interface: cliticisation and related phenomena. Amsterdam: Elsevier. Bošković, Željko (2002). ‘Clitics as non-branching elements and the Linear Correspondence Axiom’, Linguistic Inquiry 33: 329–40. Bošković, Željko (2012). ‘On NPs and clauses’, in Günther Grewendorf and Thomas Ede Zimmermann (eds.), Discourse and grammar: from sentence types to lexical categories. Berlin: Mouton de Gruyter, 179–242. Bosque, Ignacio (1996). ‘La polaridad modal’, in Actas del Cuarto Congreso de Hispanistas de Asia. Seoul, Korea: Asociación Asiática de Hispanistas, 7–14. Bossong, Georg (1985). Empirische Universalienforschung. Differentielle Objektmarkierung in den neuiranischen Sprachen. Tübingen: Narr. Breivik, Leiv Egil (1991). ‘On the typological status of Old English’, in Dieter Kastovsky (ed.), Historical English syntax. Berlin: Mouton de Gruyter, 31–50. Bresnan, Joan (2007). ‘Is syntactic knowledge probabilistic? Experiments with the English dative alternation’, in Sam Featherston and Wolfgang Sternefeld (eds.), Roots: linguistics in search of its evidential base. Berlin: Mouton de Gruyter, 77–96. Brown, Sue (1999). The syntax of negation in Russian. Stanford, CA: CSLI Publications. Browning, Marguerite (1987). ‘Null operator constructions’, PhD dissertation, MIT. Buchstaller, Isabelle (2004). ‘The sociolinguistic constraints on the quotative system—US English and British English compared’, PhD dissertation, University of Edinburgh. Buchstaller, Isabelle, and D’Arcy, Alexandra (2009). ‘Localized globalization: a multi-local investigation of quotative like’, Journal of Sociolinguistics 13: 291–331. Büring, Daniel (1997). The meaning of focus and topic: the 59th Street Bridge accent. London: Routledge. Büring, Daniel (2002). ‘Focus projection and default prominence’, ms., University of California, Los Angeles. Burridge, Kate (1993). Syntactic change in Germanic. Amsterdam: John Benjamins. Burzio, Luigi (1986). Italian syntax. Dordrecht: Reidel. Butters, Ronald (1982). ‘Editor’s note [on “be like”]’, American Speech 57: 149.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Bybee, Joan, Perkins, Revere, and Pagliuca, William (1994). The evolution of grammar: tense, aspect, and modality in the languages of the world. Chicago: The University of Chicago Press. Calabrese, Andrea (1993). ‘The sentential complementation of Salentino: a study of a language without infinitival clauses’, in Adriana Belletti (ed.), Syntactic theory and the dialects of Italy. Turin: Rosenberg & Sellier, 28–98. Cardinaletti, Anna (1998). ‘On the deficient/strong opposition in possessive systems’, in Artemis Alexiadou and Chris Wilder (eds.), Possessors, predicates and movement in DPs. Amsterdam: John Benjamins, 17–53. Cardinaletti, Anna (2004). ‘Towards a cartography of subject positions’, in Luigi Rizzi (ed.), The cartography of syntactic structures, vol. 2: The structure of CP and IP. Oxford: Oxford University Press, 115–65. Cardinaletti, Anna (2007). ‘On different types of clitic clusters’, University of Venice Working Papers in Linguistics 17: 27–76. Cardinaletti, Anna, and Repetti, Lori (2008). ‘The phonology and syntax of preverbal and postverbal subject clitics in Northern Italian dialects’, Linguistic Inquiry 39: 523–63. Cardinaletti, Anna, and Starke, Michel (1999). ‘The typology of structural deficiency: a case study of the three classes of pronouns’, in Henk van Riemsdijk (ed.), Clitics in the languages of Europe. Berlin: Mouton de Gruyter, 145–233. Castillo-Lluch, Mónica (2011). ‘Sobre el orden de constituyentes en los fueros castellanos’, in 18 Congreso de la Asociación Alemana de Hispanistas. Sección 13. Escorados a la izquierda: dislocaciones y frontalizaciones del español antiguo al moderno, Passau: Universität Passau. Chomsky, Noam (1965). Aspects of the theory of syntax. Cambridge, MA: MIT Press. Chomsky, Noam (1973). ‘Conditions on transformations’, in Stephen Anderson and Paul Kiparsky (eds.), A festschrift for Morris Halle. New York: Holt, Rinehart, and Winston, 232–86. Chomsky, Noam (1981). Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam (1986). Knowledge of language: its nature, origin, and use. New York: Praeger. Chomsky, Noam (1993). ‘A Minimalist Program for linguistic theory’, in Ken Hale and Samuel Jay Keyser (eds.), The view from Building 20. Cambridge, MA: MIT Press, 1–52. Chomsky, Noam (1995a). The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam (1995b). ‘Bare Phrase Structure’, in Gert Webelhuth (ed.), Government and Binding theory and the Minimalist Program. Oxford: Blackwell, 383–439. Chomsky, Noam (2000). ‘Minimalist inquiries: the framework’, in Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by step: essays on Minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 89–156. Chomsky, Noam (2001). ‘Derivation by phase’, in Michael Kenstowicz (ed.), Ken Hale: a life in language. Cambridge, MA: MIT Press, 1–52. Chomsky, Noam (2005). ‘Three factors in language design’, Linguistic Inquiry 36: 1–22. Chomsky, Noam (2007). ‘Approaching UG from below’, in Uli Sauerland and Hans-Martin Gärtner (eds), Interfaces + Recursion = Language?: Chomsky’s Minimalism and the view from syntax-semantics. Berlin: Mouton de Gruyter, 1–29. Chomsky, Noam (2008). ‘On phases’, in Robert Freidin, Carlos P. Otero, and Maria Luisa Zubizaretta (eds.), Foundational essays in linguistic theory: essays in honor of Jean-Roger Vergnaud. Cambridge, MA: MIT Press, 133–66.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Cichosz, Anna (2010). 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: Peter Lang. Cinque, Guglielmo (1986). ‘Bare quantifiers, quantified NPs and the notion of operator at S-structure’, Rivista di Grammatica Generativa 11: 33–63. Cinque, Guglielmo (1990). Types of A-bar dependencies. Cambridge, MA: MIT Press. Cinque, Guglielmo (1993). ‘A null theory of phrase and compound stress’, Linguistic Inquiry 24: 239–97. Cinque, Guglielmo (1999). Adverbs and functional heads: a cross-linguistic perspective. Oxford: Oxford University Press. Cinque, Guglielmo (2004). ‘ “Restructuring” and functional structure’, in Adriana Belletti (ed.), Structures and beyond. Oxford: Oxford University Press, 132–91. Cinque, Guglielmo, and Rizzi, Luigi (2010). ‘The cartography of syntactic structures’, in Bernd Heine and Heiko Narrog (eds.), The Oxford handbook of linguistic analysis. Oxford: Oxford University Press, 51–65. Clahsen, Harald (1991). ‘Constraints on parameter setting: A grammatical analysis of some acquisition stages in German child language’, Language Acquisition 1: 361–91. Clark, Robin, and Roberts, Ian G. (1993). ‘A computational model of language learnability and language change’. Linguistic Inquiry 24(2): 299–345. Collins, Chris (1997). Local economy. Cambridge, MA: MIT Press. Collins, Chris (2005a). ‘A smuggling approach to raising in English’, Linguistic Inquiry 36: 289–98. Collins, Chris (2005b). ‘A smuggling approach to the passive in English’, Syntax 8: 81–120. Collins, Chris, and Phil Branigan (1997). ‘Quotative Inversion’, Natural Language and Linguistic Theory 15: 1–41. Comrie, Bernard (1976). Aspect. Cambridge: Cambridge University Press. Comrie, Bernard (1978). ‘Ergativity’, in Winfred P. Lehmann (ed.), Syntactic typology. Austin: University of Texas Press, 329–94. Comrie, Bernard (1981). Language universals and language typology. Oxford: Blackwell. Conte, Gian Biagio (1994). Latin literature: a history (translated by Joseph B. Solodow, revised by Don Fowler & Glenn W. Most). Baltimore: The John Hopkins University Press. Cornilescu, Alexandra, and Dobrovie-Sorin, Carmen (2008). ‘Clitic doubling, complex heads and interarboreal operations’, in Dalina Kalluli and Liliane Tasmowski (eds.), Clitic doubling in the Balkan languages. Amsterdam: John Benjamins, 289–319. Coseriu, Eugenio (1985). ‘Linguistic change does not exist’, Linguistica nuova ed antica 1: 51–63. Craenenbroeck, Jeroen van (2002). ‘Van as a marker of dissociation: microvariation in Dutch’, in C. Jan-Wouter Zwart and Werner Abraham (eds.), Studies in comparative Germanic syntax. Amsterdam: John Benjamins, 41–67. Crisma, Paola, and Longobardi, Giuseppe (2009). ‘Introduction: change, relatedness and inertia in historical syntax’, in Paola Crisma and Giuseppe Longobardi (eds.), Historical syntax and linguistic theory. Oxford: Oxford University Press, 1–13. Croft, William (1991). ‘The evolution of negation’, Journal of Linguistics 27: 1–27. Cruschina, Silvio (2006). ‘Informational focus in Sicilian and the left periphery’, in Mara Frascarelli (ed.), Phases of interpretation. Berlin: Mouton de Gruyter, 363–85.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Cruschina, Silvio (2008). ‘Discourse related features and the syntax of peripheral positions. A comparative study of Sicilian and other Romance languages,’ PhD dissertation, University of Cambridge. Cruschina, Silvio (2009). ‘The syntactic role of discourse-related features’, Cambridge Occasional Papers in Linguistics 5: 15–30. Cruschina, Silvio (2010). ‘Fronting as focalization in Sicilian’, in Roberta D’Alessandro, Adam Ledgeway, and Ian G. Roberts (eds.), Syntactic variation: the dialects of Italy. Cambridge: Cambridge University Press, 247–76. Cruschina, Silvio (2011). ‘Focalization and word order in Old Italo-Romance’, Catalan Journal of Linguistics 10: 95–132. Cruschina, Silvio (2012). Discourse-related features and functional projections. Oxford: Oxford University Press. Cruschina, Silvio, and Remberger, Eva-Maria (2009). ‘Focus fronting in Sardinian and Sicilian’, in Vincenzo Moscati and Emilio Servidio (eds.), Proceedings of XXXV Incontro di Grammatica Generativa 2009, Università di Siena. Cambridge, MA: MIT Press, 118–30. Cruschina, Silvio, and Sitaridou, Ioanna (2009). ‘From Modern to Old Romance: the interaction between information structure and word order’, paper presented at the 11th Diachronic Generative Syntax Conference (DiGS), Universidade de Campinas. Cukor-Avila, Patricia (2002). ‘She say, She go, She be like: verbs of quotation over time in African American Vernacular English’, American Speech 77: 3–31. Dahl, Östen (2004). The growth and maintenance of linguistic complexity. Amsterdam: John Benjamins. Dahl, Östen (2009). ‘Two pathways of grammatical evolution’, in Talmy Givón and Masayoshi Shibatani (eds.), Syntactic complexity: diachrony, acquisition, neuro-cognition, evolution. Amsterdam: John Benjamins, 237–48. Damonte, Federico (2005). ‘La diffusione della particella “mi” in alcune varietà messinesi: problemi di metodo’, in G. Marcato (ed.), Dialetti in città: atti del convegno, Sappada/Plodn (Belluno). Padua: Unipress, 237–42. Damonte, Federico (2006). ‘Differenze generazionali nell’uso del congiuntivo presente in salentino’, in G. Marcato (ed.), Giovani, lingue e dialetti: atti del convegno, Sappada/Plodn (Belluno). Padua: Unipress, 87–2. Damonte, Federico (2010). ‘Matching moods: mood concord between CP and IP in Salentino and Southern Calabrian subjunctive complements’, in Paola Benincà and Nicola Munaro (eds), The cartography of syntactic structures, vol. 5: Mapping the left periphery. Oxford: Oxford University Press, 228–56. Danckaert, Lieven (2011). ‘On the left periphery of Latin embedded clauses’, PhD dissertation, Universiteit Gent. Danckaert, Lieven (2012). Latin embedded clauses: the left periphery. Amsterdam: John Benjamins. Davidson, Donald (1966). ‘The logical form of action sentences’, in N. Rescher (ed.), The logic of decision and action. Pittsburgh: University of Pittsburgh Press, 104–12. Davidson, Donald (1968). ‘On saying that’, Synthese 19, 130–46. Davison, Alice (2009). ‘Correlative clause features in Sanskrit and Hindi/Urdu’, in Paola Crisma and Giuseppe Longobardi (eds.), Historical syntax and linguistic theory. Oxford: Oxford University Press, 271–91.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Déchaine, Rosemarie, and Wiltschko, Martina (2002). ‘Decomposing pronouns’, Linguistic Inquiry 33(3): 409–42. Dékány, Éva (2012). ‘A profile of the Hungarian DP: the interaction of lexicalization, agreement, and linearization with the functional sequence’, PhD dissertation, Tromsø University. Delfitto, Denis (2002). ‘On the semantics of pronominal clitics and some of its consequences’, Catalan Journal of Linguistics 1: 29–57. Demirdache, Hamida, and Uribe-Etxebarria, Myriam (2000). ‘The primitives of temporal relations’, in Roger Martin, David Michaels, and Juan Uriagereka (eds.), Step by step: essays on Minimalist syntax in honor of Howard Lasnik. Cambridge, MA: MIT Press, 157–86. Demirdache, Hamida, and Uribe-Etxebarria, Myriam (2007). ‘The syntax of time arguments’, Lingua 117: 330–66. Densuşianu, Ovid (1997/1901). Histoire de la langue roumaine. Bucharest: Editura Grai ţi Suflet– Cultura Naşională. Déprez, Viviane (2000). ‘Parallel (a)symmetries and the internal structure of negative expressions’, Natural Language and Linguistic Theory 18: 253–342. Déprez, Viviane, and Martineau, France (2003). ‘Microparametric variation and Negative Concord’, in Julie Auger, Clancy Clements, and Barbara Vance (eds.), Contemporary approaches to Romance linguistics. Amsterdam: John Benjamins, 139–58. Devine, Andrew, and Stephens, Laurence (2006). Latin word order: structured meaning and information. Oxford: Oxford University Press. Diesing, Molly (1990). ‘Verb movement and the subject position in Yiddish’, Natural Language and Linguistic Theory 8: 41–79. Diesing, Molly (1992). Indefinites. Cambridge, MA: MIT Press. Diesing, Molly (1997). ‘Yiddish VP order and the typology of object movement in Germanic’, Natural Language and Linguistic Theory 15: 369–427. Dikken, Marcel den (1995). Particles: on the syntax of verb-particle, triadic, and causative constructions. Oxford: Oxford University Press. Dikken, Marcel den (1999). ‘On the structural representation of possession and agreement: the case of (anti-)agreement in Hungarian possessed nominal phrases’, in István Kenesei (ed.), Crossing boundaries. Amsterdam: John Benjamins, 137–78. Dikken, Marcel den (2006). Relators and linkers: the syntax of predication, predicate inversion and copulas. Cambridge, MA: MIT Press. Dikken, Marcel den (2010). ‘On the functional structure of locative and directional PPs’, in Guglielmo Cinque and Luigi Rizzi (eds.), The cartography of syntactic structures, vol. 6: Mapping spatial PPs. Oxford: Oxford University Press, 74–126. Dimitriadis, Alexis (1999). ‘On clitics, prepositions and Case licensing in Standard and Macedonian Greek’, in Artemis Alexiadou, Geoffrey Horrocks, and Melita Stavrou (eds.), Studies in Greek syntax. Dordrecht: Foris, 95–112. Dimmendaal, Gerrit J. (2011). Historical linguistics and the comparative study of African languages. Amsterdam: John Benjamins. Dittmer, Arne, and Dittmer, Ernst (1998). Studien zur Wortstellung-Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoeck & Ruprecht. Djamouri, Redouane (1991). ‘Particules de Négation dans les Inscriptions sur Bronze de la Dynastie des Zhou’, Cahiers de Linguistique—Asie Orientale 20: 5–76.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Djamouri, Redouane (2000). ‘Preverbal position of the pronominal object in Archaic Chinese’, paper presented at the 9th International Conference on Chinese Linguistics, The National University of Singapore. Djamouri, Redouane (2005). ‘Typological change and word order patterns in Shang inscriptions’, paper presented at the International Association of Chinese Linguistics (IACL) 13, Leiden. Djamouri, Redouane, and Paul, Waltraud (2009). ‘Verb-to-preposition reanalysis in Chinese’, in Paola Crisma and Giuseppe Longobardi (eds.), Historical syntax and linguistic theory. Oxford: Oxford University Press, 194–211. Długosz-Kurczabowa, Krystyna, and Dubisz, Stanisław (2001). Gramatyka historyczna je˛zyka polskiego. Warsaw: Wydawnictwa Uniwersytetu Warszawskiego. Dobrovie-Sorin, Carmen (1990). ‘Clitic doubling, wh-movement and quantification in Romanian’, Linguistic Inquiry 21: 351–97. Dostál, Antonín (1954). Studie o vidovém systému v staroslovnštině. Prague: Státní pedagogicke nakladatelství. Dowty, David (1979). Word meaning and Montague Grammar. Dordrecht: Reidel. Dresher, B. Elan (1999). ‘Charting the learning path: cues to parameter setting’, Linguistic Inquiry 30: 27–67. Du Bois, John W. (1980). ‘Beyond definiteness: the trace of identity in discourse’, in Wallace L. Chafe (ed.), The Pear Stories: cognitive, cultural, and linguistic aspects of narrative production. Norwood, NJ: Ablex Publishing Corporation, 203–74. Dunbar, Richard (1979). ‘Discourse pragmatics and subordinate clause word order in German: an explanation of related main clause phenomena in German and English clauses’, PhD dissertation, University of Wisconsin, Madison. Durham, Mercedes, Haddican, William, Zweig, Eytan, Johnson, Daniel Ezra, Baker, Zipporah, Cockeram, David, Danks, Esther, and Tyler, Louise. (2011). ‘Constant linguistic effects in the diffusion of be like’, Journal of English Linguistics 40: 316–37. Egedi, Barbara (2012). ‘A story about the left periphery: the functional extension of the Hungarian noun phrase’, paper presented at the 14th Diachronic Generative Syntax Conference (DiGS), Lisbon. Embick, David (2000). ‘Features, syntax and categories in the Latin perfect’, Linguistic Inquiry 31: 185–230. Embick, David, and Noyer, Rolf (2001). ‘Movement operations after syntax’, Linguistic Inquiry 32: 555–95. Embick, David, and Noyer, Rolf (2007). ‘Distributed Morphology and the syntax-morphology interface’, in Gillian Ramchand and Charles Reiss (eds.), The Oxford handbook of linguistic interfaces. Oxford: Oxford University Press, 289–324. Emonds, Joseph (1976). A transformational approach to English syntax. New York: Academic Press. Erickson, Jon (1997). ‘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 65. Geburtstag. Tübingen: Niemeyer, 95–105. Erman, Adolf (1933). Neuägyptische Grammatik. 2nd edition. Hildesheim: Georg Olms Verlag. Erteschik-Shir, Nomi (1984). ‘Der’, Nordic Journal of Linguistics 8: 131–48.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Escribano, José L. G. (2009). ‘Head-final effects and the nature of modification’, PhD dissertation, Universidad de Oviedo. Etxepare, Ricardo (2010). ‘From hearsay evidentiality to samesaying relations’, Lingua 120: 604–27. Evers, Arnold (1981). ‘Verb-second movement rules’, Wiener Linguistische Gazette 26: 15–34. Evers, Arnold (1982). ‘Twee functionele principes voor de regel “Verschuif het Werkwoord” ’, Glot 1: 11–30. Eythórsson, Þórhallur (1995). ‘Verbal syntax in the early Germanic languages’, PhD dissertation, Cornell University. Eythórsson, Þórhallur (2008). ‘The New Passive in Icelandic really is a passive’, in Þórhallur Eythórsson (ed.), Grammatical change and linguistic theory: the Rosendal papers. Amsterdam: John Benjamins, 173–219. Faarlund, Jan Terje (1990). Syntactic change: towards a theory of historical syntax. Berlin: Mouton de Gruyter. Fanselow, Gisbert (2003). ‘Münchhausen-style head movement and the analysis of verb-second’, in Anoop Mahajan (ed.), Syntax at sunset: head movement and syntactic theory. Los Angeles, CA: UCLA Working Papers in Linguistics, 40–76. Fanselow, Gisbert, and Lenertová, Denisa (2011). ‘Left peripheral focus: mismatches between syntax and information structure’, Natural Language and Linguistic Theory 29: 169–209. Farkas, Donka, and Heusinger, Klaus von (2003). ‘Stability of reference and object marking in Romanian’, paper presented at the Workshop on Direct Reference and Specificity, ESSLLI, Vienna. Featherston, Sam (2007). ‘Data in generative grammar: the stick and the carrot’, Theoretical Linguistics 33: 269–318. Feng, Shengli (1996). ‘Prosodically constrained syntactic changes in Early Archaic Chinese’, Journal of East Asian Linguistics 5: 323–71. Ferraresi, Gisella, and Lühr, Rosemarie (eds.) (2010). Diachronic studies on information structure: language acquisition and change. Berlin: Mouton de Gruyter. Ferý, Caroline, and Samek-Lodovici, Vieri (2006). ‘Focus projection and prosodic prominence in nested foci’, Language 82: 131–50. Firbas, Jan (1966). ‘Non-thematic subjects in contemporary English’, Travaux linguistiques de Prague 2: 239–56. Fischer, Olga, Kemenade, Ans van, Koopman, Willem, and Wurff, Wim van der (2000). The syntax of early English. Cambridge: Cambridge University Press. Flagg, Elissa (2007). ‘Questioning innovative quotatives’, Snippets 16: 5–6. Fodor, Janet D. (1998). ‘Unambiguous triggers’, Linguistic Inquiry 29: 1–36. Fontana, Josep M. (1993). ‘Phrase structure and the syntax of clitics in the history of Spanish’, PhD dissertation, University of Pennsylvania. Frâncu, Constantin (2009). Gramatica limbii române vechi (1521–1780). Iaşi: Demiurg. Franks, Steven (1998). ‘Clitics in Slavic’, paper presented at Comparative Slavic Morphosyntax Workshop, Bloomington, IN. Franks, Steven, and King, Tracy Holloway (2000). A handbook of Slavic clitics. Oxford: Oxford University Press.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Frascarelli, Mara, and Hinterhölzl, Roland (2007). ‘Types of topics in German and Italian’, in Kerstin Schwabe and Susanne Winkler (eds.), On information structure, meaning and form: generalizations across languages. Amsterdam: John Benjamins, 87–116. Friedman, Victor A. (2002). ‘Macedonian’, in Bernard Comrie and Greville Corbett (eds.), The Slavonic languages. London: Routledge, 249–305. Fruehwald, Josef, Gress-Wright, Jonathan, and Wallenberg, Joel (2009). ‘Phonological rule change: the Constant Rate Effect’, Proceedings of NELS 40. Amherst: GLSA. Fukui, Naoki (1988). ‘Deriving the differences between English and Japanese’, English Linguistics 5: 249–70. Fuß, Eric (2003). ‘On the historical core of V2 in Germanic’, Nordic Journal of Linguistics 26: 195–231. Fuß, Eric, and Trips, Carola (2002). ‘Variation and change in Old and Middle English: on the validity of the Double Base Hypothesis’, Journal of Comparative Germanic Linguistics 4: 171–224. Gabelentz, Georg van der (1891). Die Sprachwissenschaft, ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Chr. Herm. Tauhhnitz. Gallée, Johan Hendrik, and Tiefenbach, Ingrid (1993). Altsächsische Grammatik. Tübingen: Niemeyer. Gallego, Ángel J. (2007). ‘Defectivitat morfològica i variació sintàctica’, Caplletra 42: 219–49. Gallego, Ángel J., and Uriagereka, Juan (2007). ‘Sub-extraction from subjects: a phase theory account’, in J. Camacho, N. Flores-Ferrán, L. Sánchez, V. Déprez, and M. J. Cabrera (eds.), Romance linguistics 2006. Amsterdam: John Benjamins, 155–68. Gelderen, Elly van (1993). The rise of functional categories. Amsterdam: John Benjamins. Gelderen, Elly van (2000). A history of English reflexive pronouns: person, self, and interpretability. Amsterdam: John Benjamins. Gelderen, Elly van (2004). Grammaticalization as economy. Amsterdam: John Benjamins. Gelderen, Elly van (2011). The linguistic cycle: language change and the language faculty. Oxford: Oxford University Press. Gheţie, Ion, and Mareş, Alexandru (1985). Originile scrisului in limba română. Bucureşti: Editura Ştiinţifică şi Enciclopedică. Giannakidou, Anastasia (1994). ‘The semantic licensing of negative polarity items and the Modern Greek subjunctive’, in Ale de Boer, Helen de Hoop, and Henriette de Swart (eds.), Language and cognition 4: yearbook of the research group for theoretical and experimental linguistics. Groningen: University of Groningen, 55–68. Giannakidou, Anastasia (1997). The landscape of polarity items. Groningen: Groningen Dissertations in Linguistics. Giorgi, Alessandra (2010). About the speaker: towards a syntax of indexicality. Oxford: Oxford University Press. Giorgi, Alessandra, and Pianesi, Fabio (1997). Tense and aspect: from semantics to morphosyntax. Oxford: Oxford University Press. Giorgi, Alessandra, and Pianesi, Fabio (2004). ‘Complementizer deletion in Italian’, in Luigi Rizzi (ed.), The cartography of syntactic structures, vol. 2: The structure of CP and IP. Oxford: Oxford University Press, 190–210.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Givón, Talmy (1971). ‘Historical syntax and synchronic morphology: an archaeologist’s field trip’, in CLS 7: papers from the 7th Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society, 394–415. Givón, Talmy (1979). ‘From discourse to syntax: grammar as a processing strategy’, in Talmy Givón (ed.), Discourse and syntax. New York: Academic Press, 81–112. Givón, Talmy (2009). The genesis of syntactic complexity: diachrony, ontogeny, neuro-cognition, evolution. Amsterdam: John Benjamins. Golato, Andrea (2000). ‘An innovative German quotative for reporting on embodied actions. Und ich so / und er so “and I’m like / and he’s like” ’, Journal of Pragmatics 32: 29–54. Goria, Cecilia (2003). ‘Economy of structure: the case of subject clitics in Piedmontese’, in Josep Quer, Jan Schroten, Mauro Scorretti, Petra Sleeman, and Els Verheugd (eds.), Romance languages and linguistic theory 2001. Amsterdam: John Benjamins, 93–112. Goria, Cecilia (2004). Subject clitics in the Northern Italian dialects: a comparative study based on the Minimalist Program and Optimality Theory. Dordrecht: Kluwer. Green, Melanie (2007). Focus in Hausa. Oxford: Blackwell. Greenberg, Joseph H. (1960). ‘A quantitative approach to the morphological typology of language’, International Journal of American Linguistics 26: 178–94. Gregor, Douglas B. (1975). Friulan: language and literature. New York: The Oleander Press. Gries, Stefan Th. (2003). Multifactorial analysis in corpus linguistics: a study of particle placement. London: Continuum. Gryllia, Stella (2008). ‘On the nature of preverbal focus in Greek: a theoretical and experimental approach’, PhD dissertation, University of Leiden. Gugán, Katalin (2008). ‘Az egyszerű mondat története’, ms., Research Institute for Linguistics of the Hungarian Academy. Gugán, Katalin (2012). ‘Zigzagging in language history: negation and negative concord in Hungarian’, Finno-Ugric Languages and Linguistics 1: 89–97. Güldemann, Tom (2002). ‘When “say” is not say: the functional versatility of the Bantu quotative marker ti with special reference to Shona’, in Tom Güldemann and Manfred von Roncador (eds.), Reported discourse: a meeting ground for different linguistic domains. Amsterdam: John Benjamins, 253–87. Gundel, Jeanette, Hedberg, Nancy, and Zacharski, Ron (1993). ‘Cognitive status and the form of referring expressions in discourse’, Language 69: 274–307. Gussenhoven, Carlos (1984). On the grammar and semantics of sentence accents. Dordrecht: Foris. Haddican, Bill, and Zweig, Eytan (2012). ‘The syntax of manner quotatives in English and Dutch’, Linguistic Variation 12: 1–26. Haeberli, Eric (1999a). ‘On the word order “XP-subject” in the Germanic languages’, Journal of Comparative Germanic Linguistics 3: 1–36. Haeberli, Eric (1999b). ‘Features, categories and the syntax of A-positions’, PhD dissertation, University of Geneva. Haeberli, Eric (2000). ‘Adjuncts and the syntax of subjects in Old and Middle English’, in Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic syntax: models and mechanisms. Oxford: Oxford University Press, 109–31.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Haeberli, Eric (2001). ‘Speculations on the syntax of subordinate clauses in Old English’, Reading Working Papers in Linguistics 5: 201–29. Haeberli, Eric (2002). ‘Inflectional morphology and the loss of V2 in English’, in David W. Lightfoot (ed.), Syntactic effects of morphological change. Oxford: Oxford University Press, 88–106. Haeberli, Eric (2005). ‘Clause type asymmetries in Old English and the syntax of verb movement’, in Montserrat Batllori, Maria-Lluïsa Hernanz, Carme Picallo, and Francesc Roca (eds.), Grammaticalization and parametric change. Oxford: Oxford University Press, 267–83. Haeberli, Eric, and Haegeman, Liliane (1995). ‘Clause structure in Old English: evidence from negative concord’, Journal of Linguistics 31: 81–108. Haegeman, Liliane (2000). ‘Remnant movement and OV order’, in Peter Svenonius (ed.), The derivation of OV and VO. Amsterdam: John Benjamins, 69–96. Haegeman, Liliane (2002). ‘Some notes on DP-internal negative doubling’, in Sjef Barbiers (ed.), Syntactic microvariation. Amsterdam: Meertens Institute, 152–84. Haegeman, Liliane (2004). ‘DP-periphery and clausal periphery: possessor doubling in West Flemish’, in David Adger, Cecile de Cat, and George Tsoulas (eds.), Peripheries: syntactic edges and their effects. Dordrecht: Kluwer, 211–40. Haegeman, Liliane (2007). ‘Operator movement and topicalization in adverbial clauses’, Folia Linguistica 41: 279–325. Haegeman, Liliane (2009). ‘Operator movement and topicalisation in adverbial clauses’, English Language and Linguistics 13: 385–408. Haegeman, Liliane (2010a). ‘The internal syntax of adverbial clauses’, Lingua 120: 628–48. Haegeman, Liliane (2010b). ‘The movement derivation of conditionals’, Linguistic Inquiry 41: 595–621. Haegeman, Liliane, and Lohndal, Terje (2010). ‘Negative concord and (multiple) Agree: a case study in West Flemish’, Linguistic Inquiry 41: 181–211. Haider, Hubert (2000). ‘Adverb placement: convergence of structure and licensing’, Theoretical Linguistics 26: 95–134. Haider, Hubert (2005). ‘How to turn German into Icelandic—and derive the OV-VO contrasts’, Journal of Comparative Germanic Linguistics 8: 1–53. Hale, Mark (1998). ‘Diachronic syntax’, Syntax 1: 1–18. Hálfdanarson, Helgi (ed.) (1984). Gætum tungunnar [Let’s Watch our Language]. Reykjavík: Hið Íslenska Bókmenntafélag. Halle, Morris, and Marantz, Alec (1993). ‘Distributed Morphology and the pieces of inflection’, in Ken Hale and Samuel Jay Keyser (eds.), The view from Building 20. Cambridge, MA: MIT Press, 111–76. Halle, Morris, and Vergnaud, Jean-Roger (1987). An essay on stress. Cambridge, MA: MIT Press. Halliday, Michael Alexander Kirkwood (1967). ‘Notes on transitivity and theme in English, part 2’, Journal of Linguistics 3: 199–244. Han, Chung-Hye, and Kroch, Anthony S. (2000). ‘The rise of do-support in English: implications for clause structure’, in Masako Hirotani, Andries Coetzee, Nancy Hall, and Ji-Yung Kim (eds.), Proceedings of NELS 30. Amherst: GLSA. Harley, Heidi, and Noyer, Rolf (1999). ‘Distributed Morphology’, Glot International 4: 4.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Harley, Heidi, and Ritter, Elizabeth (2002). ‘Person and number in pronouns: a featuregeometric analysis’, Language 78: 482–526. Harves, Stephanie (2002a). ‘Genitive of negation and the syntax of scope’, in Marjo van Koppen, Erica Thrift, Erik J. van der Torre, and Malte Zimmerman (eds.), Proceedings of ConSOLE IX. Utrecht: LOT, 96–110. Harves, Stephanie (2002b). ‘Where have all the phases gone (Non-)defective categories and case alternations in Russian’, in Jindřich Toman (ed.), Formal approaches to Slavic linguistics 10, the second Ann Arbor meeting. Ann Arbor, MI: Michigan Slavic Publications, 97–118. Haspelmath, Martin (1990). ‘The grammaticization of passive morphology’, Studies in Language 14: 25–72. Hauge, Kjetil Rå (1999). ‘The word order of predicate clitics in Bulgarian’, Journal of Slavic Linguistics 7: 91–139. Hawkins, John A. (2009). ‘An efficiency theory of complexity and related phenomena’, in Geoffrey Sampson, David Gil, and Peter Trudgill (eds.), Language complexity as an evolving variable. Oxford: Oxford University Press, 252–68. He, Leshi (1989). Zuozhuan Xuci Yanjiu. [Functional Categories in the Zuozhuan.] 2nd edition, 2004. Beijing: Shangwu Yinshuguan. Hegedűs, Veronika (2006). ‘Hungarian spatial PPs’, in Peter Svenonius (ed.), Nordlyd 33(2): Adpositions. Tromsø: CASTL, 220–33. Hernanz, Maria-Lluïsa (2007). ‘From polarity to modality: some (a)symmetries between bien and sí in Spanish’, in Luis Eguren and Olga Fernández-Soriano (eds.), Coreference, modality, and focus. Amsterdam: John Benjamins, 133–69. Heusinger, Klaus von, and Gáspár, Edgar Onea (2008). ‘Triggering and blocking effects in the diachronic development of DOM in Romanian’, Probus 20: 67–110. Hewson, John, and Bubenik, Vit (1997). Tense and aspect in Indo-European languages: theory, typology, diachrony. Amsterdam: John Benjamins. Heycock, Caroline, and Zamparelli, Roberto (2005). ‘Friends and colleagues: coordination, plurality and the structure of DP’, Natural Language Semantics 13: 201–70. Heycock, Caroline, Sorace, Antonella, Hansen, Zakaris Svabo, and Wilson, Frances (2013). ‘Acquisition in variation (and vice versa): V-to-T in Faroese children’, Language Acquisition 20: 5–22. Hill, Virginia, and Tasmowski, Liliane (2008). ‘Romanian clitic doubling: a view from pragmatics-semantics and diachrony’, in Dalina Kalluli and Liliane Tasmowski (eds.), Clitic doubling in the Balkan languages. Amsterdam: John Benjamins, 135–64. Hinterhölzl, Roland (2004). ‘Language change versus grammar change: what diachronic data reveal about the distinction between core grammar and periphery’, in Eric Fuß and Carola Trips (eds.), Diachronic clues to synchronic grammar. Amsterdam: John Benjamins, 131–60. Hinterhölzl, Roland (2006). Scrambling, remnant movement and restructuring. Oxford: Oxford University Press. Hinterhölzl, Roland (2009a). ‘A phase-based comparative approach to modification and word order in Germanic’, Syntax 12, 242–84. Hinterhölzl, Roland (2009b). ‘The role of information structure in word order variation and word order change’, in Roland Hinterhölzl and Svetlana Petrova (eds.), Information structure
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
and language change: new approaches to word order variation in Germanic. Berlin: Mouton de Gruyter, 45–66. Hinterhölzl, Roland (2010). ‘Information structure and unmarked word order in (Older) Germanic’, in Malte Zimmermann and Caroline Féry (eds.), Information structure: theoretical, typological, and experimental perspectives. Oxford: Oxford University Press, 282–303. Hinterhölzl, Roland (2013). ‘(Dis)Harmonic word order and phase-based restrictions on phrasing and spell-out’, in Theresa Biberauer and Michelle L. Sheehan (eds.), Theoretical approaches to disharmonic word orders. Oxford: Oxford University Press, 16–189. Hinterhölzl, Roland, and Petrova, Svetlana (eds.) (2009a). Information structure and language change: new approaches to word order variation in Germanic. Berlin: Mouton de Gruyter. Hinterhölzl, Roland, and Petrova, Svetlana (2009b). ‘From V1 to V2 in West Germanic’, Lingua 120: 315–28. Hock, Hans Henrich (1982). ‘The Sanskrit quotative: a historical and comparative study’, Studies in the Linguistic Sciences 12: 39–85. Hoeksema, Jack (2009). ‘Jespersen recycled’, in Elly van Gelderen (ed.), Cyclical change. Amsterdam: John Benjamins, 15–34. Holmberg, Anders (2005). ‘Is there a little pro? Evidence from Finnish’, Linguistic Inquiry 36: 533–64. Hoop, Helen de, and Naransimhan, Bhuvana (2005). ‘Differential case-marking in Hindi’, in Mengistu Amberber and Helen de Hoop (eds.), Competition and variation in natural languages: the case for case. London: Elsevier, 295–390. Hooper, Joan, and Thompson, Sandra (1973). ‘On the applicability of root transformations’, Linguistic Inquiry 4: 465–97. Hopper, Paul J. (1987). ‘Emergent grammar’, Berkeley Linguistics Society 13: 139–57. Horrocks, Geoffrey (2010). Greek: a history of the language and its speakers. 2nd edition. Oxford: Wiley-Blackwell. Hróarsdóttir, Þorbjörg (2000). Word order change in Icelandic: from OV to VO. Amsterdam: John Benjamins. Hróarsdóttir, Þorbjörg (2006). ‘Loss of OV and the role of morphology’, ms., University of Tromsø. Hróarsdóttir, Þorbjörg (2010). ‘Information structure and OV order’, in Malte Zimmermann and Caroline Féry (eds.), Information structure: theoretical, typological, and experimental perspectives. Oxford: Oxford University Press, 258–81. Humbert, J. (1930). La disparition du datif en grec (du Ier au Xe siècle). Paris: Champion. Ingham, Richard (2001). ‘The structure and function of expletive there in pre-modern English’. Reading Working Papers in Linguistics 5: 231–49. Iordan, Iorgu (1955). Ion Neculce, Letopiseţul Ţării Moldovei. Bucharest: Editura de Stat. Jackendoff, Ray (1972). Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jacobsen, Bent, and Jensen, Per Anker (1982). ‘Some remarks on Danish weakly stressed der’, ms., Aarhus Business School. Jäger, Agnes (2008). History of German negation. Amsterdam: John Benjamins. Jespersen, Otto (1917). Negation in English and other languages. Copenhagen: A. F. Høst. Johnson, Janet H. (1973). ‘The Coptic conditional particles šan and ene in Demotic’, Journal of Near Eastern Studies 32: 167–9.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Jones, Michael Allan (1993). Sardinian syntax. London: Routledge. Jong, Jelly Julia de (1996). The case of bound pronouns in peripheral Romance. Wageningen: Grafish Bedrijf Ponsen & Loojinen. Jónsson, Jóhannes Gísli (1996). ‘Clausal architecture and case in Icelandic’, PhD dissertation, University of Massachusetts, Amherst. Jónsson, Jóhannes Gísli (2009). ‘The new impersonal as a true passive’, in Artemis Alexiadou, Jorge Hankamer, Thomas McFadden, Justin Nuger, and Florian Schäfer (eds.), Advances in comparative Germanic syntax. Amsterdam: John Benjamins, 281–306. Joppi, Vincenzo (1878). ‘Testi inediti friulani’, Archivio Glottologico Italiano 4: 185–341. Julien, Marit (2001). ‘The syntax of complex tenses’, The Linguistic Review 18: 125–67. Julien, Marit (2003). ‘On the negated past in Finnic and Sámi’, in Satu Manninen and Diane Nelson (eds.), Generative approaches to Finnic and Sámi linguistics: case, features and constraints. Stanford, CA: CSLI Publications, 419–46. Kádár, Edith (2006). ‘A kopula és a nominális mondat a magyarban’, PhD dissertation, University of Cluj. Karttunen, Lauri (1976). ‘Discourse referents’, in James D. McCawley (ed.), Syntax and Semantics 7: notes from the Linguistic Underground. New York: Academic Press, 363–85. Kayne, Richard S. (1975). French syntax: the transformational cycle. Cambridge, MA: MIT Press. Kayne, Richard S. (1985). ‘Principles of particle constructions’, in Jacqueline Guéron, HansGeorg Obenauer, and Jean-Yves Pollock (eds.), Grammatical representation. Dordrecht: Foris, 101–40. Kayne, Richard S. (1989). ‘Facets of Romance past participle agreement’, in Paola Benincà (ed.), Dialect variation and the theory of grammar. Dordrecht: Foris, 85–103. Kayne, Richard S. (1991). ‘Romance clitics, verb movement and PRO’, Linguistic Inquiry 22: 647–86. Kayne, Richard S. (1994). The antisymmetry of syntax. Cambridge, MA: MIT Press. Kayne, Richard S. (1998). ‘Overt vs. covert movement’, Syntax 1: 128–91. Kayne, Richard S. (2001). ‘A note on clitic doubling in French’, in Guglielmo Cinque and Giampaolo Salvi (eds.), Current studies in Italian syntax: essays offered to Lorenzo Renzi. Amsterdam: North-Holland, 40–64. Kayne, Richard S. (2005a). ‘Some notes on comparative syntax, with special reference to English and French’, in Guglielmo Cinque and Richard Kayne (eds.), The Oxford handbook of comparative syntax. Oxford: Oxford University Press, 3–69. Kayne, Richard S. (2005b). Movement and silence. Oxford: Oxford University Press. Kayne, Richard S. (2007). ‘A short note on where vs. place’, in Roberta Maschi, Nicoletta Penello, and Piera Rizzolatti (eds.), Miscellanea di Studi Linguistici offerti a Laura Vanelli da amici e allievi padovani. Udine: Forum, 245–57. Keenan, Edward L. (1994). ‘Creating anaphors: an historical study of the English reflexive pronouns’, ms., University of California at Los Angeles. Keenan, Edward L. (2002). ‘Explaining the creation of reflexive pronouns in English’, in Donka Minkova and Robert Stockwell (eds.), Studies in the history of the English language: a millennial perspective. Berlin: Mouton de Gruyter, 325–54. Keenan, Edward L. (2009). ‘Linguistic theory and the historical creation of English reflexives’, in Paola Crisma and Giuseppe Longobardi (eds.), Historical syntax and linguistic theory. Oxford: Oxford University Press, 17–40.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Keenan, Edward L., and Comrie, Bernard (1977). ‘Noun phrase accessibility and universal grammar’, Linguistic Inquiry 8: 63–99. Kemenade, Ans van (1987). Syntactic case and morphological case in the history of English. Dordrecht: Foris. Kemenade, Ans van (2000). ‘Jespersen’s cycle revisited: formal properties of grammaticalization’, in Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic syntax: models and mechanisms. Oxford: Oxford University Press, 51–74. Kemenade, Ans van (2007). ‘Formal syntax and language change: developments and outlook’, Diachronica 24: 155–69. Kemenade, Ans van, and Los, Bettelou (2006). ‘Discourse adverbs and clausal syntax in Old and Middle English’, in Ans van Kemenade and Bettelou Los (eds.), The handbook of the history of English. Oxford: Blackwell, 224–48. Kenney, Edward, and Clausen, Wendell (eds.) (1982). The Cambridge history of classical literature, vol. 2: Latin literature. Cambridge: Cambridge University Press. Kibort, Anna (2004). ‘Passive and passive-like constructions in English and Polish: a contrastive study with particular reference to impersonal constructions’, PhD dissertation, University of Cambridge. Kim, Min-Joo (2003). ‘The genitive of negation in Russian: a Relativized Minimality account’, in Wayles Browne, Eva Dornisch, Natalia Kondrashova, and Draga Zec (eds.), Proceedings of the 4th Workshop on Formal Approaches to Slavic Linguistics: The Cornell Meeting. Ann Arbor, MI: Michigan Slavic Publications, 295–314. Kim, Min-Joo (2004). ‘The genitive of negation in Russian: blocking of case-licensing by negation’, in Mary Andronis, Anne Pycha, and Keiko Yoshimura (eds.), CLS 38: papers from the 38th Annual Meeting of the Chicago Linguistic Society: Main Session. Chicago: Chicago Linguistic Society, 345–60. Kiparsky, Paul (1995). ‘Indo-European origins of Germanic syntax’, in Adrian Battye and Ian G. Roberts (eds.), Clause structure and language change. Oxford: Oxford University Press, 140–69. Kiparsky, Paul (1996). ‘The shift to head-initial VP in Germanic’, in Höskuldur Thráinsson, Samuel David Epstein, and Steve Peter (eds.), Studies in comparative Germanic syntax, vol. 2. Dordrecht: Kluwer, 149–70. Kiparsky, Paul (1997). ‘The rise of positional licensing’, in Ans van Kemenade and Nigel Vincent (eds.), Parameters of morphosyntactic change. Cambridge: Cambridge University Press, 460–94. Kiparsky, Paul (1998). ‘Partitive case and aspect’, in Miriam Butt and Wilhelm Geuder (eds.), The projection of arguments. Stanford, CA: CSLI Publications, 265–307. Kiparsky, Paul, and Condoravdi, Cleo (2002). ‘Clitics and clause structure’, Journal of Greek Linguistics 2: 1–39. Kirk, Allison (2012). Word order and information structure in New Testament Greek. Utrecht: LOT. Kiss, Katalin É. (1998). ‘Identificational focus vs. information focus’, Language 74: 245–73. Kiss, Katalin É. (1999). ‘Mi tartozik a névutók osztályába?’ [What belongs to the category of postpositions?], Magyar nyelvjárások 37: 167–72. Kiss, Katalin É. (2002). The syntax of Hungarian. Cambridge: Cambridge University Press.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Kiss, Katalin É. (2008a). ‘Free word order, (non-)configurationality and phases’, Linguistic Inquiry 39: 441–74. Kiss, Katalin É. (2008b). ‘Topic and Focus: two structural positions associated with logical functions in the left periphery of the Hungarian sentence’, Acta Linguistica Hungarica 55: 287–96. Kiss, Katalin É. (2009). ‘Negative quantifiers in Hungarian’, in Marcel den Dikken and Robert M. Vago (eds.), Approaches to Hungarian 11. Amsterdam: John Benjamins, 65–94. Kiss, Katalin É. (2010). ‘An adjunction analysis of quantifiers and adverbials in the Hungarian sentence’, Lingua 120: 506–26. Kiss, Katalin É. (2013). ‘From Proto-Hungarian SOV to Old Hungarian TopFocVX∗ ’, Diachronica 30: 202–31. Kjartansson, Helgi Skúli (1991). ‘Nýstárleg þolmynd í barnamáli’ [A New Passive in child language], Skíma 14: 18–22. Klein, Wolfgang (1995). ‘A time relational analysis of Russian aspect’, Language 71: 669–95. Klemensiewicz, Zenon, Lehr-Spławiński, Tadeusz, and Urbańczyk, Stanisław (1964). Gramatyka historyczna je¸zyka polskiego. Warsaw: PWN. Koeneman, Olaf, and Neeleman, Ad (2001). ‘Predication, verb movement and the distribution of expletives’, Lingua 111: 189–233. Koll, Hans-Georg (1965). ‘Zur Stellung des Verbs im spätantiken und frühmittelalterlichen Latein’, Mittellateinisches Jahrbuch 2: 241–72. Koopman, Hilda J. (2000). ‘Prepositions, postpositions, circumpositions, and particles: the structure of Dutch PPs’, in Hilda J. Koopman (ed.), The syntax of specifiers and heads: collected essays of Hilda J. Koopman. London: Routledge, 204–60. Koopman, Hilda, and Szabolcsi, Anna (2000). Verbal complexes. Cambridge, MA: MIT Press. Koopman, Willem (1992). ‘Old English pronouns: some remarks’, in Fran Colman (ed.), Evidence for Old English. Edinburgh: John Donald, 44–87. Koopman, Willem (1995). ‘Verb-final main clauses in Old English prose’, Studia Neophilologica 67: 129–44. Koopman, Willem (1997). ‘Another look at clitics in Old English’, Transactions of the Philological Society 95: 73–93. Korhonen, Mikko (1973). ‘Zur Geschichte des negativen Präteritums und Perfekts im Ostseefinnischen und Lappischen’, in Commentationes Fenno-Ugricae in Honorem Erkki Itkonen. Helsinki: Suomalais-Ugrilainen Seura, 174–84. Korhonen, Mikko (1981). Johdatus lapin kielen historiaan. Helsinki: SKS. Kornfilt, Jaklin (2008). ‘DOM and two types of DSM in Turkish’, in Helen de Hoop and Peter de Swart (eds.), Differential subject marking. Dordrecht: Springer, 79–149. Koster, Jan (1995). ‘Predicate incorporation and word order in Dutch’, in Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi, and Raffaella Zanuttini (eds.), Paths towards Universal Grammar, 255–77. Washington, DC: Georgetown University Press. Krapova, Iliyana (2010). ‘Bulgarian relative and factive clauses with an invariant complementizer’, Lingua 120: 1240–72. Kratzer, Angelika (1996). ‘Severing the external argument from its verb’, in Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon. Dordrecht: Kluwer, 109–37.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Kratzer, Angelika (2004). ‘Covert quantifier restrictions in natural language’, talk given at Palazzo Feltrinelli in Gargnano. Krifka, Manfred (1984). ‘Fokus, Topik, syntaktische Strukur und semantische Interpretation’, ms., Universität Tübingen. Krifka, Manfred (2008). ‘Basic notions of information structure’, Acta Linguistica Hungarica 55: 243–76. Kroch, Anthony S. (1989). ‘Reflexes of grammar in patterns of language change’, Language Variation and Change 1: 199–244. Kroch, Anthony S. (1994). ‘Morphosyntactic variation’, in Katharine Beals, Jeannette Denton, Robert Knippen, Lynette Melnar, Hisami Suzuki, and Erica Zeinfeld (eds.), CLS 30: papers from the 30th Regional Meeting of the Chicago Linguistic Society, vol. 2. Chicago: Chicago Linguistic Society, 180–201. Kroch, Anthony S. (2003). ‘Syntactic change’, in Mark R. Baltin and Chris Collins (eds.), The handbook of contemporary syntactic theory. Oxford: Blackwell, 699–729. Kroch, Anthony S., and Taylor, Ann (2000a). Penn-Helsinki Parsed Corpus of Middle English. 2nd edition. Kroch, Anthony S., and Taylor, Ann (2000b). ‘Verb-complement order in Middle English’, in Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic syntax: models and mechanisms. Oxford: Oxford University Press, 132–63. Kroch, Anthony S., Santorini, Beatrice, and Diertani, Ariel (2005). Penn-Helsinki Parsed Corpus of Early Modern English. Kroch, Anthony S., Santorini, Beatrice, and Diertani, Ariel (2010). Penn Parsed Corpus of Modern British English. Kusters, Wouter (2003). ‘Linguistic complexity: the influence of social change on verbal inflection’, PhD dissertation, Radboud University Nijmegen. Kwon, Song-Nim, and Zribi-Hertz, Anne (2008). ‘Differential function marking, Case and information structure: evidence from Korean’, Language 84: 258–99. Labov, William (1972). Sociolinguistic patterns. Philadelphia, PA: University of Pennsylvania Press. Ladd, D. Robert (1986). ‘Intonational phrasing: the case for recursive prosodic structure’, Phonology 3: 311–40. Ladd, D. Robert (1996). Intonational phonology. Cambridge: Cambridge University Press. Laka, Itziar (1994). Negation in syntax: on the nature of functional categories and projections. New York: Garland. Lambrecht, Knud (1994). Information structure and sentence form: topic, focus, and the mental representations of discourse referents. Cambridge: Cambridge University Press. LaPolla, Randy (1994). ‘On the change to verb-medial order in Proto-Chinese: evidence from Tibeto-Burman’, in Hajime Kitamura, Tatsuo Nishida, and Yasuhiko Nagano (eds.), Current issues in Sino-Tibetan linguistics. Osaka: Organizing Committee of the 26th International Conference on Sino-Tibetan Linguistics, 98–104. Larrivée, Pierre (2011). ‘Is there a Jespersen cycle?’, in Pierre Larrivée and Richard Ingham (eds.), The evolution of negation: beyond the Jespersen cycle. Berlin: Mouton de Gruyter, 1–23. Larrivée, Pierre, and Ingham, Richard (eds.) (2011). The evolution of negation: beyond the Jespersen Cycle. Berlin: Mouton de Gruyter. Larson, Richard K. (1988). ‘On the double object construction’, Linguistic Inquiry 19: 335–91.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Lass, Roger (1990). ‘How to do things with junk: exaptation in language evolution’, Journal of Linguistics 26: 79–102. Lass, Roger (1997). Historical linguistics and language change. Cambridge: Cambridge University Press. Ledgeway, Adam (1998). ‘Variation in the Romance infinitive: the case of the Southern Calabrian inflected infinitive’, Transactions of the Philological Society 96: 1–61. Ledgeway, Adam (2004). ‘Il sistema completivo dei dialetti meridionali: la doppia serie di complementatori’, Rivista Italiana di Dialettologia 27: 89–147. Ledgeway, Adam (2005). ‘Moving through the left periphery: the dual complementiser system in the dialects of southern Italy’, Transactions of the Philological Society 103: 336–96. Ledgeway, Adam (2006). ‘The dual complementiser system in southern Italy: Spirito greco, materia romanza?’, in A. L. Lepschy and A. Tosi (eds), Rethinking languages in contact: the case of Italian. Oxford: Legenda, 112–26. Ledgeway, Adam (2007). ‘Diachrony and finiteness: subordination in the dialects of southern Italy’, in Irina Nikolaeva (ed.), Finiteness: theoretical and empirical foundations. Oxford: Oxford University Press, 335–65. Ledgeway, Adam (2009). Grammatica diacronica del napoletano. Tübingen: Niemeyer. Ledgeway, Adam (2012). From Latin to Romance: morphosyntactic typology and change. Oxford: Oxford University Press. Ledgeway, Adam, and Lombardi, Alessandra (2014). ‘The development of the southern subjunctive: morphological loss and syntactic gain’, in Paola Benincà, Adam Ledgeway, and Nigel Vincent (eds.), Diachrony and dialects: grammatical change in the dialects of Italy. Oxford: Oxford University Press, 25–47. Lenertová, Denisa (2001). ‘Czech pronominal clitics’, paper presented at the Workshop on Slavic pronominal clitics, Berlin, February 2001. Lenerz, Jürgen (1977). Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: TBL-Verlag Narr. Lenerz, Jürgen (1984). Syntaktischer Wandel und Grammatiktheorie: eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Leonetti, Manuel (2004). ‘Specificity and Differential Object Marking in Spanish’, Catalan Journal of Linguistics 3: 75–114. Leonetti, Manuel, and Escandell-Vidal, Victoria (2009). ‘Fronting and verum focus in Spanish’, in Andreas Dufter and Daniel Jacob (eds.), Focus and background in Romance languages. Amsterdam: John Benjamins, 155–204. Li, Charles N., and Thompson, Sandra (1974). ‘An explanation of word order change’, Foundations of Language 12: 201–14. Li, Charles N., and Thompson, Sandra A. (1976). ‘Subject and Topic: a new typology of language’, in Charles N. Li (ed.), Subject and Topic. New York: Academic Press, 457–89. Light, Caitlin (2011). Parsed corpus of Early New High German. Lightfoot, David W. (1991). How to set parameters: arguments from language change. Cambridge, MA: MIT Press. Lightfoot, David W. (1999). The development of language: acquisition, change and evolution. Oxford: Blackwell.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Lightfoot, David W. (ed.) (2002). Syntactic effects of morphological change. Oxford: Oxford University Press. Lightfoot, David W. (2006). How new languages emerge. Cambridge: Cambridge University Press. Lin, Jo-wang (2010). ‘A tenseless analysis of Mandarin Chinese revisited: a response to Sybesma 2007’, Linguistic Inquiry 41: 305–29. Linde, Paul (1923). ‘Die Stellung des Verbs in der lateinischen Prosa’, Glotta 12: 153–78. Linde, Sonja (2009). ‘Aspects of word order and information structure in Old Saxon’, in Roland Hinterhölzl and Svetlana Petrova (eds.), Information structure and language change: new approaches to word order variation in Germanic. Berlin: Mouton de Gruyter, 367–89. Lindstedt, Jouko (1985). ‘On the semantics of tense and aspect in Bulgarian’, PhD dissertation, University of Helsinki. Lindstedt, Jouko (1994). ‘On the development of the South Slavonic perfect’, in Three Papers on the Perfect. EUROTYP Working Papers, Series VI, July 1994 5: 32–53. Lippert, Jörg (1974). Beiträge zur Technik und Syntax althochdeutscher Übersetzungen unter besonderer Berücksichtigung der Isidorgruppe und des althochdeutschen Tatian. Munich: Fink. Lødrup, Helge (2007). ‘Norwegian anaphors without visible binders’, Journal of Germanic Linguistics 19: 1–22. Lombardi, Alessandra (1997). ‘The grammar of complementation in the dialects of Calabria’, unpublished dissertation, University of Manchester. Lombardi, Alessandra (1998). ‘Calabria greca e Calabria latina da Rohlfs ai giorni nostri: la sintassi dei verbi modali-aspettuali’, in P. Ramat and E. Roma (eds.), Sintassi storica. Atti del XXX congresso della società di linguistica italiana, Pavia, 26–28 settembre 1996. Rome: Bulzoni, 613–26. Longobardi, Giuseppe (1994). ‘Reference and proper names’, Linguistic Inquiry 25: 609–65. Longobardi, Giuseppe (2001). ‘Formal syntax, diachronic Minimalism, and etymology: the history of French chez’, Linguistic Inquiry 32: 275–302. López, Luis (2009). A derivational syntax for information structure. Oxford: Oxford University Press. Loporcaro, Michele (1997a). L’origine del raddoppiamento fonosintattico. Saggio di fonologia diacronica romanza. Basel-Tübingen: Francke. Loporcaro, Michele (1997b). ‘Puglia and Salento’, in Martin Maiden and Mair Parry (eds.), The dialects of Italy. London: Routledge, 338–48. Louwrens, Louis J., Kosch, Ingeborg M., and Kotzé, Albert E. (1995). Northern Sotho. Munich: Lincom Europa. Lunt, Horace G. (1952). A grammar of the Macedonian literary language. Skopje: Drzavno knigoizd. na NR Makedonija. Lunt, Horace G. (1974). Old Church Slavonic grammar. Berlin: Mouton de Gruyter. Macaulay, Ronald (2001). ‘You’re like “why not?”. The quotative expression of Glasgow adolescents’, Journal of Sociolinguistics 5: 3–21. McCloskey, James (2007). ‘The grammar of autonomy in Irish’, Natural Language and Linguistic Theory 25: 825–57. McGinnis, Martha (1998). ‘Locality and inert Case’, in Pius N. Tamanji and Kiyomi Kusumoto (eds.), Proceedings of NELS 28. Amherst, MA: GLSA, 267–81.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Mahajan, Anoop (2003). ‘Word order and (remnant) VP movement’, in Simin Karimi (ed.), Word order and scrambling. Oxford: Blackwell, 217–37. Maiden, Martin (1995). A linguistic history of Italian. London: Longman. Maling, Joan (1993). ‘Unpassives of unaccusatives’, ms., Brandeis University. Maling, Joan (2006). ‘From passive to active: syntactic change in progress in Icelandic’, in Benjamin Lyngfelt and Torgrim Solstad (eds.), Demoting the Agent. Passive and other voicerelated phenomena. Amsterdam: John Benjamins, 197–223. Maling, Joan, and Sigurjónsdóttir, Sigríður (1997). ‘The “New Passive” in Icelandic’, in Elizabeth Hughes, Mary Hughes, and Annabel Greenhill (eds.), Proceedings of the 21st Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press, 378–89. Maling, Joan, and Sigurjónsdóttir, Sigríður (2002). ‘The New Impersonal Construction in Icelandic’, Journal of Comparative Germanic Linguistics 5: 97–142. Mancarella, Giovan Battista (ed.) (1998). Salento. Monografia regionale della ‘Carta dei Dialetti Italiani’. Lecce: Edizioni del Grifo. Manolessou, Io, and Beis, Stamatis (2006). ‘Syntactic isoglosses in Modern Greek dialects: the case of the indirect object’, in Angela Ralli, Mark Janse, and Brian Joseph (eds.), Proceedings of the 2nd International Conference on Modern Greek Dialects and Linguistic Theory, Mytilene, September 2004. Patras: University of Patras, 220–35. Manzini, Maria Rita, and Savoia, Leonardo M. (2005). I dialetti italiani e romanci. Morfosintassi generativa, 3 vols. Alessandria: Edizioni dell’Orso. Marantz, Alec (1997). ‘No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon’, in Alexis Dimitriadis, Laura Siegel, Clarissa Surek-Clark, and Alexander Williams (eds.), University of Pennsylvania Working Papers in Linguistics, vol. 4(2): Proceedings of the 21st Annual Penn Linguistics Colloquium. Philadelphia, PA: University of Pennsylvania, 201–25. Mardale, Alexandru (2006). ‘Why on table is on the table? Notes on the definite’s article incorporation with prepositions in Romanian’, in Beata Gyuri (ed.), Proceedings of the First Central Europe Student Conference in Linguistics, 1-12 Markopoulos, Theodore (2010), ‘Case overlap in Medieval Cypriot Greek: a socio-historical perspective’, Folia Linguistica Historica 31(1): 1–46. Martínez-Gil, Fernando (1989). ‘Las inversiones del orden de palabras en el Romancero’, Hispania 72: 895–908. Martins, Ana Maria (2000). ‘Polarity items in Romance: underspecification and lexical change’, in Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic syntax: Models and mechanisms. Oxford: Oxford University Press, 191–219. Martins, Ana Maria (2002). ‘The loss of IP-scrambling in Portuguese: clause structure, word order variation and change’, in David W. Lightfoot (ed.), Syntactic effects of morphological change. Oxford: Oxford University Press, 232–48. Martins, Ana Maria (2005). ‘Clitic placement, VP-ellipsis, and scrambling in Romance’, in Montserrat Batllori, Maria-Lluïsa Hernanz, Carme Picallo, and Francesc Roca (eds.), Grammaticalization and parametric change. Oxford: Oxford University Press, 175–93. Mason, Winter, and Suri, Siddharth (2012). ‘Conducting behavioral research on Amazon’s Mechanical Turk’, Behavioral Research Methods 44: 1–23. Matushansky, Ora (2006). ‘Head movement in linguistic theory’, Linguistic Inquiry 37: 69–109.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Meisel, Jürgen M. (1995). ‘Parameters in acquisition’, in Paul Fletcher and Brian MacWhinney (eds.), The handbook of child language. Oxford: Blackwell, 10–35. Meisel, Jürgen M. (2011). ‘Bilingual acquisition and theories of diachronic change: bilingualism as cause and effect of grammatical change’, Bilingualism: Language and Cognition 14: 121–45. Melnick, Robin, Jaeger, T. Florian, and Wasow, Thomas (2011). ‘Speakers employ fine-grained probabilistic knowledge’, paper presented at the Linguistic Society of America, Pittsburgh. Mensching, Guido (2012). ‘Parameters in Old Romance word order: a comparative Minimalist analysis’, in Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds.), Parameter theory and linguistic change. Oxford: Oxford University Press, 21–42. Mensching, Guido, and Remberger, Eva-Maria (2010). ‘The left periphery in Sardinian’, in Roberta D’Alessandro, Adam Ledgeway, and Ian G. Roberts (eds.), Syntactic variation: the dialects of Italy. Cambridge: Cambridge University Press, 261–76. Meurman-Solin, Anneli, López-Couso, Maria José, and Los, Bettelou (eds.) (2012). Information structure and syntactic change in the history of English. Oxford: Oxford University Press. Michelioudakis, Dimitris (2011). ‘Dative arguments and abstract Case in Greek’, PhD dissertation, University of Cambridge. Michelioudakis, Dimitris, and Sitaridou, Ioanna (2012). ‘Syntactic microvariation: dative constructions in Greek’, in Beatriz Fernandez and Ricardo Etxepare (eds.), Variation in datives: a microcomparative perspective. Oxford: Oxford University Press, 212–55. Migdalski, Krzysztof (2006). ‘The syntax of compound tenses in Slavic’, PhD dissertation, Tilburg University. Mikkelsen, Line H. (2002). ‘Expletive subjects in subject relative clauses’, in C. Jan-Wouter Zwart and Werner Abraham (eds.), Studies in comparative Germanic syntax. Amsterdam: John Benjamins, 71–93. Milner, J. C. (1979). ‘Le système de la négation en français et l’opacité du sujet’, Langue Française 44: 80–106. Mitchell, Bruce (1985). Old English syntax. 2 vols. Oxford: Clarendon. Młynarczyk, Anna (2004). ‘Aspectual pairing in Polish’, PhD dissertation, Utrecht University. Molnár, Valéria (1998). ‘Topic in focus. On the syntax, phonology, semantics and pragmatics of the so-called “contrastive topic” in Hungarian and German’, Acta Linguistica Hungarica 45 (1–2): 89–166. Molnár, Valéria (2001). ‘Contrast from a contrastive perspective’, in Ivana Kruiff-Korbayová and Mark Steedman (eds.), Proceedings of the ESSLLI 2001 workshop on information structure, discourse structure and discourse semantics. Helsinki: Helsinki University, 99–114. Müller, Gereon (2005). ‘Pro-drop and impoverishment’, in Patrick Brandt and Eric Fuß (eds.), Form, structure and grammar: a festschrift presented to Günther Grewendorf on the occasion of his 60th birthday. Tübingen: Narr, 93–115. Müller, Gereon (2008). ‘Some consequences of an impoverishment-based approach to morphological richness and pro-drop’, in Jacek Witkós and Gisbert Fanselow (eds.), Elements of Slavic and Germanic grammars: a comparative view. Frankfurt: Peter Lang, 125–45. Munro, Pamela (1982). ‘On the transitivity of “say” verbs’, in Paul Hopper and Sandra Thompson (eds.), Studies in transitivity: syntax and semantics. New York, San Francisco, London: Academic Press, 301–18.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Naro, Anthony J. (1981). ‘The social and structural dimensions of a syntactic change’, Language 57: 63–98. Nash, Léa, and Rouveret, Alain (2002). ‘Cliticisation as Unselective Attract’, Catalan Journal of Linguistics 1: 157–99. Neeleman, Ad, Titov, Elena, Koot, Hans van de, and Vermeulen, Reiko (2009). ‘A syntactic typology of topic, focus and contrast’, in Jeroen van Craenenbroeck (ed.), Alternatives to cartography. Berlin: Mouton de Gruyter, 15–51. Nespor, Marina, and Vogel, Irene (1986). Prosodic phonology. Dordrecht: Foris. Nevalainen, Terttu, and Traugott, Elizabeth Closs (eds.) (2012). The Oxford handbook of the history of English. Oxford: Oxford University Press. Newmeyer, Frederick J. (2004). ‘Against a parameter-setting approach to typological variation’, Linguistic Variation Yearbook 4: 181–234. Nichols, Johanna (2009). ‘Linguistic complexity: a comprehensive definition and survey’, in Geoffrey Sampson, David Gil, and Peter Trudgill (eds.), Language complexity as an evolving variable. Oxford: Oxford University Press, 110–25. Nickel, Klaus Peter (1990). Samisk grammatikk. Oslo: Universitetsforlaget. Nielsen, Hans Frede (2000). ‘Ingwäonisch’, in Johannes Hoops, Heinrich Beck, Dieter Geuenich, and Heiko Steuer (eds.), Reallexikon der germanischen Altertumskunde, 2nd edition, vol. 15. Berlin: Mouton de Gruyter, 432–8. Niv, Michael (1992). ‘Right association revisited’, paper presented at the Thirtieth Annual Meeting of the Association for Computational Linguistics. Norde, Muriel (2002). ‘The final stages of grammaticalization: affixhood and beyond’, in Ilse Wischer and Gabriele Diewald (eds.), New reflections on grammaticalization. Amsterdam: John Benjamins, 45–81. Nunes, Jairo (2004). Linearization of chains and sideward movement. Cambridge, MA: MIT Press. Oliver, Lisi (2002). The beginnings of English Law. Toronto: University of Toronto Press. Olsvay, Csaba (2006). ‘Negative universal quantifiers in Hungarian’, Lingua 116: 245–70. Ordóñez, Francisco (1998). ‘Post-verbal asymmetries in Spanish’, Natural Language and Linguistic Theory 16: 313–46. Ormazabal, Javier, and Romero, Juan (2010). ‘The derivation of dative alternations’, in Maia Duguine, Susana Huidobro, and Nerea Madariaga (eds.), Argument structure and syntactic relations: a cross-linguistic perspective. Amsterdam: John Benjamins, 203–32. Osawa, Fuyo (1999). ‘The relation between tense and aspect: the emergence of the T system’, UCL Working Papers in Linguistics 11: 521–44. Ouali, Hamid (2008). ‘On C-to-T φ-feature transfer: the nature of agreement and antiagreement in Berber’, in Roberta D’Alessandro, Gunnar H. Hrafnbjargarson, and Susanne Fischer (eds.), Agreement restrictions. Berlin: Mouton de Gruyter, 159–80. Ouhalla, Jamal (1991). Functional categories and parametric variation. London: Routledge. Panagiotidis, E. Phoevos (2008). ‘Diachronic stability and feature interpretability’, in Theresa Biberauer (ed.), The limits of syntactic variation. Amsterdam: John Benjamins, 441–56. Panaitescu, Petre P. (1958). Grigore Ureche, Letopiseţul ޡarii Moldovei. Bucharest: Editura de Stat. Panaitescu, Petre P. (1979). Miron Costin, Letopiseţul ޡarii Moldovei. Bucharest: Editura Minerva.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Pancheva, Roumyana (2005). ‘The rise and fall of second-position clitics’, Natural Language and Linguistic Theory 23: 103–67. Pancheva, Roumyana, Kagle, Janine, and Łazorczyk, Agnieszka (2007). ‘Codex Zographensis’, in USC Parsed Corpus of Old South Slavic. Pancheva, Roumyana, Łazorczyk, Agnieszka, Krivokapić, Jelena, and Minkova, Yulia (2007). ‘Codex Marianus’, in USC Parsed Corpus of Old South Slavic. Paoli, Sandra (2010). ‘In focus: an investigation of information and contrastive constructions’, in Roberta D’Alessandro, Adam Ledgeway, and Ian Roberts (eds.), Syntactic variation. The dialects of Italy. Cambridge: Cambridge University Press: 277–91. Parsons, Terrence (1990). Events in the semantics of English: a study in subatomic semantics. Cambridge, MA: MIT Press. Partee, Barbara (1973). ‘The syntax and semantics of quotation’, in Stephen Anderson and Paul Kiparsky (eds.), A festschrift for Morris Halle. New York: Holt, Rinehart, and Winston, 410–18. Partee, Barbara (1977). ‘John is easy to please’, in Antonio Zampolli (ed.), Linguistic structures processing. Amsterdam: North-Holland, 281–312. Pearson, Matthew (2000). ‘Two types of VO languages’, in Peter Svenonius (ed.), The derivation of OV and VO. Amsterdam: John Benjamins, 327–63. Pesetsky, David (1982). ‘Paths and categories’, PhD dissertation, MIT. Pesetsky, David (2000). Phrasal movement and its kin. Cambridge, MA: MIT Press. Pesetsky, David, and Torrego, Esther (2007). ‘The syntax of valuation and the interpretability of features’, in Simin Karimi, Vida Samiian, and Wendy K. Wilkins (eds.), Phrasal and clausal architecture: syntactic derivation and interpretation, in honor of Joseph E. Emonds. Amsterdam: John Benjamins, 262–94. Petrova, Svetlana (2009). ‘Information structure and word order variation in the Old High German Tatian’, in Roland Hinterhölzl and Svetlana Petrova (eds.), Information structure and language change: new approaches to word order variation in Germanic. Berlin: Mouton de Gruyter, 251–80. Petrova, Svetlana, and Hinterhölzl, Roland (2010). ‘Evidence for two types of focus positions in Old High German’, in Gisella Ferraresi and Rosemarie Lühr (eds.), Diachronic studies on information structure: language acquisition and change. Berlin: Mouton de Gruyter, 189–217. Peyraube, Alain (1996). ‘Recent issues in Chinese historical syntax’, in C.-T. James Huang and Audrey Y.-H. Li (eds.), New horizons in Chinese linguistics. Dordrecht: Kluwer, 161–213. Pintzuk, Susan (1991). ‘Phrase structures in competition: variation and change in Old English word order’, PhD dissertation, University of Pennsylvania. Pintzuk, Susan (1996). ‘Cliticization in Old English’, in Aaron L. Halpern and Arnold Zwicky (eds.), Approaching second: second position clitics and related phenomena. Stanford, CA: CSLI Publications, 366–400. Pintzuk, Susan (1999). Phrase structures in competition: variation and change in Old English word order. New York: Garland. Pintzuk, Susan (2002). ‘The syntax of objects in Old English’, paper presented at the 7th Diachronic Generative Syntax Conference (DiGS), Girona. Pintzuk, Susan (2005). ‘Arguments against a universal base: evidence from Old English’, English Language and Linguistics 9: 115–38. Pintzuk, Susan, and Haeberli, Eric (2008). ‘Structural variation in Old English root clauses’, Language Variation and Change 20: 367–407.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Pintzuk, Susan, and Kroch, Anthony S. (1989). ‘The rightward movement of complements and adjuncts in the Old English of Beowulf ’, Language Variation and Change 1: 115–43. Pintzuk, Susan, and Taylor, Ann (2006). ‘The loss of OV order in the history of English’, in Ans van Kemenade and Bettelou Los (eds.), The handbook of the history of English, 249–78. Oxford: Blackwell. Plank, Frans (1993). ‘Peculiarities of passives of reflexives in German’, Studies in Language 17: 135–67. Pogatscher, Alois (1901). ‘Unausgedrücktes Subject im Altenglischen’, Anglia 23: 261–301. Poletto, Cecilia (1995). ‘The diachronic development of subject clitics in North Eastern Italian dialects’, in Adrian Battye and Ian G. Roberts (eds.), Clause structure and language change. Oxford: Oxford University Press, 295–324. Poletto, Cecilia (2000). The higher functional field: Evidence from Northern Italian dialects. Oxford: Oxford University Press. Poletto, Cecilia (2001). ‘Complementizer deletion and verb movement in Standard Italian’, in Guglielmo Cinque and Giampaolo Salvi (eds.), Current studies in Italian syntax: essays offered to Lorenzo Renzi. Amsterdam: North-Holland, 265–86. Poletto, Cecilia (2006). ‘Parallel phases: a study on the high and low left periphery of Old Italian’, in Mara Frascarelli (ed.), Phases of interpretation. Berlin: Mouton de Gruyter, 261–92. Poletto, Cecilia (2008). ‘On negative doubling’, Quaderni di lavoro dell’ASIt 8: 57–84. Poletto, Cecilia (2014). Word order in Old Italian. Oxford: Oxford University Press. Pollock, Jean-Yves (1989). ‘Verb movement, Universal Grammar and the structure of IP’, Linguistic Inquiry 20: 365–424. Polotsky, Hans Jakob (1944). Études de syntaxe copte. Cairo: Imprimerie de IFAO. Prince, Ellen (1981). ‘Toward a taxonomy of given-new information’, in Peter Cole (ed.), Radical pragmatics. New York: Academic Press, 223–55. Prince, Ellen (1989). ‘Yiddish Wh-clauses, subject-postposing, and topicalization’, in Joyce Powers and Kenneth de Jong (eds.), ESCOL ’88. Columbus, OH: Ohio State University, 403–15. Prince, Ellen (1993). ‘On the discourse functions of syntactic form in Yiddish: expletive es and Subject-Postposing’, in David Goldberg, Marvin I. Herzog, Barbara Kirschenblatt-Gimblett, and Dan Miron (eds.), The field of Yiddish: studies in language, folklore, and literature, vol. 5. Evanston, IL: Northwestern University Press, 59–86. Progovac, Ljiljana (1993). ‘Locality and subjunctive-like complements in Serbo-Croatian’, Journal of Slavic Linguistics 1: 116–44. Progovac, Ljiljana (1996). ‘Clitics in Serbian/Croatian: Comp as the second position’, in Aaron L. Halpern and Arnold M. Zwicky (eds.), Approaching second: second position clitics and related phenomena. Stanford, CA: CSLI Publications, 411–28. Progovac, Ljiljana (2005). Syntax of Serbian: clausal architecture. Bloomington: Slavica. Pulleyblank, Edwin G. (1995). Outline of Classical Chinese grammar. Vancouver: UBC Press. Pylkkänen, Liina (2002). ‘Introducing arguments’, PhD dissertation, MIT. Pylkkänen, Liina (2008). Introducing arguments. Cambridge, MA: MIT Press. Quack, Joachim F. (2006). ‘Zur Morphologie und Syntax der demotischen zweiten Tempora’, Lingua Aegyptia 14: 251–62.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Quer, Josep (2002). ‘Edging quantifiers: on OP-fronting in Western Romance’, in Claire Beyssade, Reineke Bok-Bennema, Frank Drijkoningen, and Paola Monachesi (eds.), Romance Languages and Linguistic Theory 2000. Amsterdam: John Benjamins, 253–70. Radanović-Kocić, Vesna (1988). ‘The grammar of Serbo-Croatian clitics: a synchronic and diachronic perspective’, PhD dissertation, University of Illinois, Urbana-Champaign. Rauch, Irmengard (1992). The Old Saxon language: grammar, epic narrative, linguistic interference. Frankfurt: Peter Lang. Reinhart, Tanya (1981). ‘Pragmatics and linguistics: an analysis of sentence topics’, Philosophica 27: 53–94. Reinhart, Tanya (1987). ‘Specifier and operator binding’, in Eric J. Reuland and Alice G. B. ter Meulen (eds.), The representation of (in)definiteness. Cambridge, MA: MIT Press, 130–67. Reintges, Chris H. (2004). Coptic Egyptian (Sahidic dialect): a learner’s grammar. Cologne: Rüdiger Köppe Verlag. Reintges, Chris H. (2009). ‘Spontaneous syntactic change’, in Paola Crisma and Giuseppe Longobardi (eds.), Historical syntax and linguistic theory. Oxford: Oxford University Press, 41–57. Reintges, Chris H. (2011a). ‘High analyticity and Coptic particle syntax: a phase-based approach’, The Linguistic Review 38: 533–99. Reintges, Chris H. (2011b). ‘The evolution of the Ancient Egyptian stative: diachronic stability despite inflectional change’, Folia Orientalia 48: 7–97. Reintges, Chris H. (2012a). ‘Macroparametric change and the synthetic–analytic dimension: the case of Ancient Egyptian’, in Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds.), Parameter theory and linguistic change. Oxford: Oxford University Press, 133–57. Reintges, Chris H. (2012b). ‘The morphosyntax of alternating complementizers in Coptic relative clause constructions’, Folia Orientalia 49: 395–419. Reintges, Chris H. (2013). ‘Sapirian “drift” towards analyticity and long-term morphosyntactic change in Ancient Egyptian’, in Ritsuko Kikusawa and Laurence A. Reid (eds.), Historical Linguistics 2011: selected papers from the 20th International Conference on Historical Linguistics. Amsterdam: John Benjamins, 289–328. Reintges, Chris H., and Green, Melanie (2004). ‘Coptic second tenses and Hausa relative aspects: a comparative view’, Lingua Aegyptia 12: 157–77. Reintges, Chris H., LeSourd, Philip, and Chung, Sandra (2006). ‘Wh-movement, Wh-agreement, and apparent Wh-in-situ’, in Lisa L.–S. Cheng and Norbert Corver (eds.), Wh-movement moving on. Cambridge, MA: MIT Press, 165–94. Reis, Hans (1901). ‘Über althochdeutsche Wortfolge’, Zeitschrift für deutsche Philologie 33: 212–38, 330–49. Reis, Marga (2000a). ‘Anmerkung zu Verb-erst-Satz-Typen im Deutschen’, in Rolf Thieroff, Matthias Tamrat, Nanna Furhop, and Oliver Teuber (eds.), Deutsche Grammatik in Theorie und Praxis. Tübingen: Niemeyer, 215–27. Reis, Marga (2000b). ‘Review of O. Önnerfors: Verb-erst-Deklarativsätze. Grammatik und Pragmatik’, Studia Linguistica 54: 90–100. Renzi, Lorenzo, and Vanelli, Laura (1983). ‘I pronomi soggetto in alcune varietà romanze’, in Paola Benincà, Michele Cortelozzo, Aldo Luigi Prosdocimi, Laura Vanelli, and Alberto
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Zambone (eds.), Scritti linguistici in onore di Giovan Battista Pellegrini. Pisa: Pacini, 121–45. Rezac, Milan (2003). ‘The syntax of the Person-Case Constraint’, ms., CNRS. Rezac, Milan (2008). ‘The syntax of eccentric agreement: the Person Case Constraint and Absolutive Displacement in Basque.’ Natural Language and Linguistic Theory 26: 61–106. Richards, Marc, and Biberauer, Theresa (2005). ‘Explaining Expl’, in Marcel den Dikken and Christina Tortora (eds.), The function of function words and functional categories. Amsterdam: John Benjamins, 115–54. Richards, Norvin (1999). ‘Featural cyclicity and the ordering of multiple specifiers’, in Sam Epstein and Norbert Hornstein (eds.), Working Minimalism. Cambridge, MA: MIT Press, 127–58. Riemsdijk, Henk van (1990). ‘Functional prepositions’, in Harm Pinkster and Inge Genee (eds.), Unity in diversity: papers presented to Simon C. Dik on his 50th birthday. Dordrecht: Foris, 229–41. Ries, John (1880). Die Stellung von Subject und Prädicatsverbum im Hêliand. Strasbourg/London: Karl J. Trübner. Rigau, Gemma (1999). ‘La estructura del sintagma nominal: los modificadores del nombre’, in Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Madrid: Espasa, 311–93. Rigau, Gemma (2012). ‘Mirative and focusing uses of the Catalan particle pla’, in Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, and Cecilia Poletto (eds.), Functional heads. Oxford: Oxford University Press, 92–102. Ritter, Elizabeth, and Rosen, Sara (2000). ‘Event structure and ergativity’, in Carol Tenny and James Pustejovsky (eds.), Events as grammatical objects. Stanford, CA: CSLI Publications, 187–238. Ritter, Elizabeth, and Wiltschko, Martina (in press). ‘The composition of INFL. An exploration of tense, tenseless languages, and tenseless constructions’, to appear in Natural Language and Linguistic Theory. Rivero, María-Luisa (1991). ‘Long head movement and negation: Serbo-Croatian vs. Slovak and Czech’, The Linguistic Review 8: 319–51. Rivero, María-Luisa (2005). ‘Topics in Bulgarian morphology and syntax: a Minimalist perspective’, Lingua 115: 1083–128. Rizzi, Luigi (1986). ‘On the status of subject clitics in Romance’, in Osvaldo Jaeggli and Carmen Silva-Corvalán (eds.), Studies in Romance linguistics. Dordrecht: Foris, 391–419. Rizzi, Luigi (1996). ‘Residual verb second and the Wh-Criterion’, in Adriana Belletti and Luigi Rizzi (eds.), Parameters and functional heads: essays in comparative syntax. Oxford: Oxford University Press, 63–90. Rizzi, Luigi (1997). ‘The fine structure of the left periphery’, in Liliane Haegeman (ed.), Elements of grammar. Dordrecht: Kluwer, 281–337. Rizzi, Luigi (2001). ‘On the position “Int(errogative)” in the left periphery of the clause’, in Guglielmo Cinque and Giampaolo Salvi (eds.), Current studies in Italian syntax: essays offered to Lorenzo Renzi. Amsterdam: North-Holland, 287–96. Rizzi, Luigi (2005). ‘Grammatically-based target-inconsistencies in child language’, in Kamil Ud Deen, Jun Nomura, Barbara Schulz, and Bonnie D. Schwartz (eds.), The Proceedings of
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
the Inaugural Conference on Generative Approaches to Language Acquisition North America (GALANA). Cambridge, MA: UConn/MIT Working Papers in Linguistics, 19–49. Rizzi, Luigi (2006). ‘On the form of chains: criterial positions and ECP effects’, in Lisa L.–S. Cheng and Norbert Corver (eds.), Wh-movement moving on. Cambridge, MA: MIT Press, 97–133. Roberts, Ian G. (1985). ‘Agreement parameters and the development of English modal auxiliaries’, Natural Language and Linguistic Theory 3: 21–58. Roberts, Ian G. (1993). Verbs and diachronic syntax: a comparative history of English and French. Dordrecht: Kluwer. Roberts, Ian G. (1996). ‘Remarks on the Old English C-system and the diachrony of V2’, in Ellen Brandner and Gisela Ferraresi (eds.), Language change and generative grammar. Opladen: Westdeutscher Verlag, 154–64. Roberts, Ian G. (1997). ‘Directionality and word order change in the history of English’, in Ans van Kemenade and Nigel Vincent (eds.), Parameters of morphosyntactic change. Cambridge: Cambridge University Press, 397–426. Roberts, Ian G. (2004). ‘The C-system in Brythonic Celtic languages, V2, and the EPP’, in Luigi Rizzi (ed.), The cartography of syntactic structures, vol. 2: The structure of CP and IP. Oxford: Oxford University Press, 297–328. Roberts, Ian G. (2005). Principles and Parameters in a VSO language: a case study in Welsh. Oxford: Oxford University Press. Roberts, Ian G. (2007). Diachronic syntax. Oxford: Oxford University Press. Roberts, Ian G. (2010). ‘Grammaticalization, the clausal hierarchy and semantic bleaching’, in Elizabeth Closs Traugott and Graeme Trousdale (eds.), Gradience, gradualness and grammaticalization. Amsterdam: John Benjamins, 45–73. Roberts, Ian G., and Holmberg, Anders (2010). ‘Introduction: parameters in Minimalist theory’, in Theresa Biberauer, Anders Holmberg, Ian Roberts, and Michelle Sheehan (eds.), Parametric variation: null subjects in Minimalist theory. Cambridge: Cambridge University Press, 1–57. Roberts, Ian G., and Roussou, Anna (1999). ‘A formal approach to “grammaticalization” ’, Linguistics 37: 1011–41. Roberts, Ian G., and Roussou, Anna (2003). Syntactic change: a Minimalist approach to grammaticalization. Cambridge: Cambridge University Press. Rodríguez-Molina, Javier (2010). ‘La gramaticalización de los tiempos compuestos en español antiguo: cinco cambios diacrónicos’, PhD dissertation, Universidad Autónoma de Madrid. Rodríguez-Molina, Javier (2011). ‘La anteposición del participio en los tiempos compuestos y la sintaxis V2 del español antiguo’, in 18 Congreso de la Asociación Alemana de Hispanistas. Sección 13. Escorados a la izquierda: dislocaciones y frontalizaciones del español antiguo al moderno. Passau: Universität Passau. Rögnvaldsson, Eiríkur (1987). ‘On word order in Icelandic’, in Robin D. S. Allan and Michael P. Barnes (eds.), Proceedings of the Seventh Biennial Conference of Teachers of Scandinavian Studies in Great Britain and Northern Ireland. London: University College London, 33–49. Rohlfs, Gerhard (1968). Grammatica storica della lingua italiana e dei suoi dialetti. Morfologia. Turin: Einaudi. Rohlfs, Gerhard (1969). Grammatica storica della lingua italiana e dei suoi dialetti. Sintassi e formazione delle parole. Turin: Einaudi.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Rohrbacher, Bernhard (1999). Morphology-driven syntax: a theory of V-to-I-raising and prodrop. Amsterdam: John Benjamins. Romaine, Suzanne, and Lange, Deborah (1991). ‘The use of like as a marker of reported speech and thought: a case of grammaticalization in progress’, American Speech 66: 227–79. Rooth, Mats (1992). ‘A theory of focus interpretation’, Natural Language Semantics 1: 75–116. Rosenkvist, Henrik (2009). ‘Referential null subjects in Germanic: an overview’, Working Papers in Scandinavian Syntax 84: 151–80. Ross, John Robert (1967). ‘Constraints on variables in syntax’, PhD dissertation, MIT. Ross, John Robert (1969). ‘Auxiliaries as main verbs’, in W. Todd (ed.), Studies in philosophical linguistics 1. Evanston, IL: Great Expectations Press, 77–102. Rothstein, Susan (1999). ‘Fine-grained structure in the eventuality domain: the semantics of predicate adjective phrases and be’, Natural Language Semantics 7: 347–420. Roussou, Anna (2000). ‘On the left periphery: modal particles and complementizers’, Journal of Greek Linguistics 1: 65–94. Rouveret, Alain (2010). ‘On verb-subject languages’, Lingua 120: 232–63. Rudin, Catherine (1997). ‘AgrO and Bulgarian pronominal clitics’, in M. Lindseth and S. Franks (eds.), Formal approaches to Slavic linguistics: the Indiana meeting. Ann Arbor, MI: Michigan Slavic Publications, 224–52. Safir, Kenneth (1985). Syntactic chains. Cambridge: Cambridge University Press. Saito, Mamoru (1985). ‘Some asymmetries in Japanese and their theoretical consequences’, PhD dissertation, MIT. Saito, Mamoru, and Fukui, Naoki (1998). ‘Order in phrase structure and movement’, Linguistic Inquiry 29: 439–74. Sampson, Geoffrey (2009). ‘A linguistic axiom challenged’, in Geoffrey Sampson, David Gil, and Peter Trudgill (eds.), Language complexity as an evolving variable. Oxford: Oxford University Press, 1–18. Sánchez-López, Cristina (1999). ‘La negación’, in Ignacio Bosque and Violeta Demonte (eds.), Gramática descriptiva de la lengua española. Madrid: Espasa, 2561–634. Santorini, Beatrice (1992). ‘Variation and change in Yiddish subordinate clause word order’, Natural Language and Linguistic Theory 10, 595–640. Santorini, Beatrice (1997/2008). The Penn Yiddish Corpus. University of Pennsylvania. Sapp, Christopher (2006). ‘Verb order in subordinate clauses from Early New High German to Modern German’, PhD dissertation, Indiana University. Sapp, Christopher (2011). The verbal complex in subordinate clauses from Medieval to Modern German. Amsterdam: John Benjamins. Sapp, Christopher (2012). ‘Microparameters in the verbal complex: Middle High German and some modern varieties’, in Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds.), Parameter theory and linguistic change. Oxford: Oxford University Press, 43–59. Saussure, Ferdinand de (1979) [1916]. Cours de linguistique générale. Paris: Payot. Saxena, Anju (1995). ‘Unidirectional grammaticalization: diachronic and cross-linguistic evidence’, Sprachtypologie und Universalienforschung 48: 350–72. Schäfer, Florian (2009). ‘The passive of reflexive verbs’, GRIMM seminar, University of Lund.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Schooneveld, Cornelis H. van (1951). ‘The aspectual system of the Old Church Slavonic and Old Russian verbum finitum byti’, Word 7: 93–103. Schütze, Carson (2004). ‘Synchronic and diachronic microvariation in English do’, Lingua 114: 495–516. Schweikert, Walter (2005). The order of prepositional phrases in the structure of the clause. Amsterdam: John Benjamins. Scorretti, Mauro (1994). ‘Complementizer deletion’, PhD dissertation, University of Amsterdam. Sebestyén, Árpád (2002). A névutók állománya és rendszere a Jókai-kódexben (1372u.). [The stock and system of postpositions in the Jókai Codex (after 1372).] Debrecen. Selkirk, Elisabeth (1984). Phonology and syntax: the relation between sound and structure. Cambridge, MA: MIT Press. Selkirk, Elisabeth (1995). ‘Sentence prosody: intonation, stress and phrasing’, in John Goldsmith (ed.), The handbook of phonological theory. Oxford: Blackwell, 550–69. Sells, Peter, Zaenen, Annie, and Zec, Draga (1987). ‘Reflexivization variation: relations between syntax, semantics and lexical structure’, in Masayo Iida, Stephen Wechsler, and Draga Zec (eds.), Working papers in grammatical theory and discourse structure: interactions of morphology, syntax, and discourse. Stanford, CA: CSLI Publications, 169–238. Shapin, Stephen (1996). The scientific revolution. Chicago: University of Chicago Press. Shen, Pei (1992). Jiagu Buci Yuxi Yanjiu. [Research on the word order in oracle bone inscriptions.] Taipei: Wenjin. Shibatani, Masayoshi (1985). ‘Passives and related constructions: a prototype analysis’, Language 61: 821–48. Shlonsky, Ur (2010). ‘The cartographic enterprise in syntax’, Language and Linguistics Compass 4: 417–29. Shon, Young-Sook, Hong, Young-Yeah, and Hong, Kyung-Sun (1996). ‘The functional category IP in Korean reconsidered’, Studies in Generative Grammar 6: 351–84. Shosted, R. (2006). ‘Correlating complexity: a typological approach’, Linguistic Typology 10: 1–40. Siewierska, Anna (1993). ‘Syntactic weight vs information structure and word order variation in Polish’, Journal of Linguistics 29: 233–65. Siewierska, Anna, and Uhlířová, Ludmila (1998). ‘An overview of word order in Slavic languages’, in Anna Siewierska (ed.), Constituent order in the languages of Europe. Berlin: Mouton de Gruyter, 321–82. Sigurðsson, Halldór Ármann (1989). ‘Verbal syntax and Case in Icelandic’, PhD dissertation, Lund University. Sigurðsson, Halldór Ármann (1993). ‘Argument drop in Old Icelandic’, Lingua 89: 247–80. Sigurðsson, Halldór Ármann (2011). ‘On the New Passive’, Syntax 14: 148–78. Sigurðsson, Halldór Ármann, and Egerland, Verner (2009). ‘Impersonal null-subjects in Icelandic and elsewhere’, Studia Linguistica 63: 158–85. Sigurjónsdóttir, Sigríður, and Maling, Joan (2001). ‘Það var hrint mér á leiðinni í skólann: Þolmynd eða ekki þolmynd?’ [It was pushed me on the way to school: passive or not passive?], Íslenskt mál 23: 123–80. Sipos, Pál (1991). ‘A névmások’, in Loránd Benkő (ed.), A magyar nyelv történeti nyelvtana I. A korai ómagyar kor és elő zményei. [Historical grammar of Hungarian I. Early Old Hungarian and its ancestors.] Budapest: Akadémiai Kiadó, 353–99.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Sławski, Franciszek (1946). Miejsce enklityki odmiennej w dziejach jêzyka bułgarskiego. Kraków: Polska Akademia UmiejêtnoIJci. Smyth, Herbert W. (1920). A Greek grammar for colleges. New York: American Book Company. Sobin, Nicholas (1985). ‘Case assignment in Ukrainian morphological passive constructions’, Linguistic Inquiry 16: 649–62. Speyer, Augustin (2008). ‘Topicalization and clash avoidance: on the interaction of prosody and syntax in the history of English with a few glimpses at German’, PhD dissertation, University of Pennsylvania. Spiik, N. E. (1989). Lulesamisk grammatik. Lulea: Nils-Erik Spiik and Sameskolstyrelsen. Spreng, Bettina (2006). ‘Antipassive morphology and case assignment in Inuktitut’, in Alana Johns, Diane Massam, and Juvenal Ndayiragije (eds.), Ergativity: emerging issues. Amsterdam: Springer, 247–70. Sprouse, Jon (2011). ‘A validation of Amazon Mechanical Turk for the collection of acceptability judgments in linguistic theory’, Behavioral Research Methods 43: 155–67. Stieber, Zdzisław (1973). Zarys gramatyki porównawczej je¸zyków słowiańskich. Fleksja werbalna. Warsaw: PWN. Stjepanović, Sandra (1998). ‘On the placement of Serbo-Croatian clitics: evidence from VP ellipsis’, Linguistic Inquiry 29: 527–37. Stjepanović, Sandra (1999). ‘What do second position cliticisation, scrambling, and multiple wh-fronting have in common?’, PhD dissertation, University of Connecticut. Stockwell, Robert (1976). ‘Commentary on Lightfoot’, in W. M. Christie, Jr. (ed.), Current progress in historical linguistics: proceedings of the Second International Conference on Historical Linguistics. Amsterdam: North-Holland, 32–3. Stowell, Tim (1995). ‘What do the present and past tenses mean?’, in Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini (eds.), Temporal reference, aspect and actionality, vol. 1: Semantic and syntactic perspectives. Turin: Rosenberg & Sellier, 381–96. Stowell, Tim (1996). ‘The phrase structure of tense’, in Johan Rooryck and Laurie Zaring (eds.), Phrase structure and the lexicon. Dordrecht: Kluwer, 277–91. Stowell, Tim (2007). ‘The syntactic expression of tense’, Lingua 117: 437–63. Suñer, Margarita (1994). ‘V-movement and the licensing of argumental wh-phrases in Spanish’, Natural Language and Linguistic Theory 12: 335–72. Suñer, Margarita (2000). ‘The syntax of direct quotes with special reference to Spanish and English’, Natural Language and Linguistic Theory 18: 525–78. Surányi, Balázs (2006a). ‘Predicates, negative quantifiers and focus: specificity and quantificationality of n-words’, in Katalin É. Kiss (ed.), Event structure and the left periphery. Dordrecht: Springer, 255–86. Surányi, Balázs (2006b). ‘Quantifiers and focus in negative concord’, Lingua 116: 272–313. Svenonius, Peter (2000). ‘Quantifier movement in Icelandic’, in Peter Svenonius (ed.), The derivation of VO and OV. Amsterdam: John Benjamins, 255–92. Svenonius, Peter (2006). ‘The emergence of Axial Parts’, in Peter Svenonius and Marina Pantcheva (eds.), Nordlyd 33(1): Adpositions. Tromsø: CASTL, 49–77. Svenonius, Peter (2008). ‘Projections of P’, in Anna Asbury, Jakub Dotlacil, Berit Gehrke, and Rick Nouwen (eds.), Syntax and semantics of spatial P. Amsterdam: John Benjamins, 63–84.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Svenonius, Peter (2010). ‘Spatial P in English’, in Guglielmo Cinque and Luigi Rizzi (eds.), The cartography of syntactic structures, vol. 6: Mapping spatial PPs. Oxford: Oxford University Press, 127–60. Sybesma, Rint (2007). ‘Whether we Tense-agree overtly or not’, Linguistic Inquiry 38: 580–7. Szabolcsi, Anna (1994). ‘The noun phrase’, in Ferenc Kiefer and Katalin É. Kiss (eds.), The syntactic structure of Hungarian. New York: Academic Press, 179–275. Szabolcsi, Anna (2004). ‘Positive polarity—negative polarity’, Natural Language and Linguistic Theory 22: 409–52. Szmrecsányi, Benedikt (2004). ‘On operationalizing syntactic complexity’, in Gérard Purnelle, Cédrick Fairon, and Anne Dister (eds.), Le poids des mots. Proceedings of the 7th International Conference on Textual Data Statistical Analysis. Louvain-la-Neuve: Presses universitaires de Louvain, 1032–9. Tagliamonte, Sali A., and D’Arcy, Alexandra (2007). ‘Frequency and variation in the community grammar: Tracking a new change through the generations’, Language Variation and Change 19: 199–217. Tagliamonte, Sali A., and D’Arcy, Alexandra (2009). ‘Peaks beyond phonology: adolescence, incrementation, and language change’, Language 85: 58–109. Tagliamonte, Sali A., and Hudson, Rachel (1999). ‘Be like et al. beyond America: the quotative system in British and Canadian youth’, Journal of Sociolinguistics 3: 147–72. Tamburelli, Marco (2006). ‘Remarks on richness’, UCL Working Papers in Linguistics 18: 439–55. Taraldsen, Knut Tarald (1978). ‘On the NIC, vacuous application, and the that-t filter’, ms., MIT. Tasmowski, Liliane (2008). ‘Defilé de MOD en roumain’, in Emmanuelle Danblon, Mikhail Kissine, Fabienne, Martin, Christine Michaux, and Svetlana Vogeleer (eds.), Linguista sum: mélanges offerts à Marc Dominicy à l’occasion de son soixantième anniversaire. Paris: L’Harmattan, 279–96. Taylor, Ann (2006). ‘Rhythmic prose and its effect on verb-object order’, paper presented at the York-Holland Symposium on Historical English Syntax (SHES), York, UK. Taylor, Ann, Nurmi, Arja, Warner, Anthony, Pintzuk, Susan, and Nevalainen, Terttu (2006). York-Helsinki Parsed Corpus of Early English Correspondence. York: University of York and Helsinki: University of Helsinki. Taylor, Ann, and Pintzuk, Susan (2008). ‘The effect of information structure on object position in the history of English’, paper presented at ICEHL 15, Munich. Taylor, Ann, and Pintzuk, Susan (2009). ‘Information structure and the syntax of objects in Old English: a corpus study’, paper presented at the Workshop on Corpus-based Advances in Historical Linguistics, University of York. Taylor, Ann, and Pintzuk, Susan (2012a). ‘The interaction of syntactic change and information status effects in the change from OV to VO in English’, Catalan Journal of Linguistics 10: 1–24. Taylor, Ann, and Pintzuk, Susan (2012b). ‘Rethinking the OV/VO alternation in Old English: the effect of complexity, grammatical weight, and information status’, in Terttu Nevalainen and Elizabeth Closs Traugott (eds.), The Oxford handbook of the history of English. Oxford: Oxford University Press, 1199–213. Taylor, Ann, Warner, Anthony, Pintzuk, Susan, and Beths, Frank (2003). The York-TorontoHelsinki Parsed Corpus of Old English Prose.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Taylor, Ann, and Wurff, Wim van der (2005). ‘Introduction: special issue on aspects of OV and VO order in the history of English’, English Language and Linguistics 9: 1–4. Tenny, Carol (1987). ‘Grammaticalizing aspect and affectedness’, PhD dissertation, MIT. Tenny, Carol (1994). Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer. Terzi, Arhonto (1994). ‘Clitic climbing from finite clauses and long head movement’, Catalan Working Papers in Linguistics 3: 97–122. Terzi, Arhonto (1996). ‘Clitic climbing from finite clauses and tense raising’, Probus 8: 273–95. Thráinsson, Höskuldur (2003). ‘Syntactic variation, historical development, and Minimalism’, in Randall Hendrick (ed.), Minimalist Syntax. Oxford: Blackwell, 152–91. Thráinsson, Höskuldur (2007). The syntax of Icelandic. Cambridge: Cambridge University Press. Thráinsson, Höskuldur, Angantýsson, Ásgrímur, and Sigurðsson, Einar Freyr (eds.) (2013). Tilbrigði í íslenskri setningagerð. [Variation in Icelandic syntax.] Reykjavík: Háskólaútgáfan. Tjihkkom, Kåre (1992). Lulesamisk grammatikk. Guovda-geaidnu: Sáme åhpadusráde. Todi, Aida (2001). Elemente de sintaxˇa româneascˇa veche. Iaşi: Editura Paralela 45. Tomaselli, Alessandra (1995). ‘Cases of verb third in Old High German’, in Adrian Battye and Ian G. Roberts (eds.), Clause structure and language change. Oxford: Oxford University Press, 345–69. Tomić, Olga (1989). ‘Language contact, language system and language code’, in Milorad Radovanović (ed.), Yugoslav General Linguistics. Amsterdam: John Benjamins, 361–80. Tomić, Olga (1996). ‘The Balkan Slavic clausal clitics’, Natural Language and Linguistic Theory 14: 811–72. Tomić, Olga (2000). ‘On clitic sites’, in Frits Beukema and Marcel den Dikken (eds.), Clitic phenomena in European languages. Amsterdam: John Benjamins, 293–316. Torrego, Esther (1984). ‘On inversion in Spanish and some of its effects’, Linguistic Inquiry 15: 103–29. Tortora, Christina (2014). A comparative grammar of Borgomanerese. Oxford: Oxford University Press. Traugott, Elizabeth (2004). ‘Exaptation and grammaticalization’, in M. Akimoto (ed.), Linguistic studies based on corpora. Tokyo: Hituzi Syobo, 133–56. Troseth, Erika (2009). ‘Degree inversion and negative intensifier inversion in the English DP’, The Linguistic Review 26: 67–134. Trosterud, Trond (1994). ‘Auxiliaries, negative verbs and word order in the Sami and Finnic Languages’, in Ago Künnap (ed.), Minor Uralic languages: structure and development. Tartu: University of Tartu, 173–81. Truckenbrodt, Hubert (1995). ‘Phonological phrases: their relation to syntax, focus and prominence’, PhD dissertation, MIT. Truckenbrodt, Hubert (1999). ‘On the relation between syntactic phrases and phonological phrases’, Linguistic Inquiry 30: 219–55. Uhmann, Susanne (1991). Fokusphonologie: eine Analyse deutscher Intonationskonturen im Rahmen der nicht-linearen Phonologie. Tübingen: Niemeyer. Uriagereka, Juan (1995). ‘Aspects of the syntax of clitic placement in Western Romance’, Linguistic Inquiry 26: 79–123. Ürögdi, Barbara (2006). ‘Predicate fronting and dative case in Hungarian’, Acta Linguistica Hungarica 53: 291–332.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Ürögdi, Barbara (2009). ‘Temporal adverbial clauses with or without operator movement’, in Katalin É. Kiss (ed.), Adverbs and adverbial adjuncts at the interfaces. Berlin: Mouton de Gruyter, 133–70. Vagiakakos, Dikaios (1964). ‘Dialektika ek tou Mesaionikou Pontou’, Archeion Pontou 26: 267–89. Vaillant, André (1966). Grammaire comparée des langues slaves, vol. 3: Le verbe. Paris: Éditions Klincksieck. Vallduví, Enric (1992). ‘The informational component’, PhD dissertation, University of Pennsylvania. Vassiliou, Erma (2002). ‘The word order of Medieval Cypriot’, PhD dissertation, La Trobe University. Vecchio, Paola (2010). ‘The distribution of the complementizers /ka/ and /ku/ in the North Salentino dialect of Francavilla Fontana (Brindisi)’, in Roberta D’Alessandro, Adam Ledgeway, and Ian G. Roberts (eds.), Syntactic variation: the dialects of Italy. Cambridge: Cambridge University Press, 312–22. Vikner, Sten (1991). ‘Relative der and other C elements in Danish’, Lingua 84: 109–36. Vikner, Sten (1997). ‘V -to-I -movement and inflection for person in all tenses’, in Liliane Haegeman (ed.), The new comparative syntax. New York: Longman, 189–213. Vikner, Sten (2001). ‘Verb movement variation in Germanic and Optimality Theory’, Habilitation thesis, Universität Tübingen. Villa-García, Julio (2012). ‘Spanish subjects can be subjects: acquisitional and empirical evidence’, Iberia 4: 124–69. Villalba, Xavier (2006). ‘The topic-focus articulation and (non)spurious articles in Germanic and Romance’, paper presented at ‘DP-internal information structure’, Utrecht. Vincent, Nigel (1976). ‘Perceptual factors and word order change in Latin’, in Martin B. Harris (ed.), Romance syntax: synchronic and diachronic perspectives. Salford: University of Salford, 54–69. Vincent, Nigel (1995). ‘Exaptation and grammaticalization’, in Henning Andersen (ed.), Historical Linguistics 1993: selected papers from the 11th International Conference on Historical Linguistics, Los Angeles, 16–20 August 1993. Amsterdam: John Benjamins, 433–48. Voorst, Jan van (1988). Event structure. Amsterdam: John Benjamins. Wackernagel, Jacob (1926). Vorlesungen über Syntax mit besonderer Berücksichtigung von Griechisch, Lateinisch und Deutsch. Zweite Reihe. Basel: Birkhäuser Verlag. Wagner, Michael (2005). ‘Asymmetries in prosodic domain formation’, in Norvin Richards and Martha McGinnis (eds), Perspectives on phases. Cambridge, MA: MIT Press, 329–67. Wagner, Suzanne, and Sankoff, Gillian (2011). ‘Age grading in the Montréal French inflected future’, Language Variation and Change 23: 275–313. Walkden, George (2009). ‘The comparative method in syntactic reconstruction’, MPhil dissertation, University of Cambridge. Walkden, George (2012). ‘Against inertia’, Lingua 122: 891–901. Walkden, George (2014). Syntactic reconstruction and Proto-Germanic. Oxford: Oxford University Press. Wallage, Phillip (2008). ‘Jespersen’s Cycle in Middle English: parametric variation and grammatical competition’, Lingua 118: 643–74.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Wallenberg, Joel C. (2009). ‘Antisymmetry and the conservation of c-command: scrambling and phrase structure in synchronic and diachronic perspective’, PhD dissertation, University of Pennsylvania. Wang, Li (1958). Hanyu Shigao. [Historical Notes on Chinese.] Reprinted 2004. Beijing: Zhonghua Shuju. Wanner, Dieter (1981). ‘Surface complementizer deletion: Italian che ∼ Ø’, Journal of Italian Linguistics 6: 47–82. Wasow, Thomas (1997). ‘Remarks on grammatical weight’, Language Variation and Change 9: 81–105. Waters, Cathleen (2009). ‘The preposition cycle in English’, in Elly van Gelderen (ed.), Cyclical change. Amsterdam: John Benjamins, 285–300. Webelhuth, Gert (1989). ‘Syntactic saturation phenomena and the modern Germanic languages’, PhD dissertation, University of Massachusetts, Amherst. Weerman, Fred (1988). The V2 conspiracy: a synchronic and a diachronic analysis of verbal positions in Germanic languages. Dordrecht: Foris. Weerman, Fred (1997). ‘On the relation between morphological and syntactic case’, in Ans van Kemenade and Nigel Vincent (eds.), Parameters of morphosyntactic change. Cambridge: Cambridge University Press, 427–59. Weinreich, Uriel, Labov, William, and Herzog, Marvin I. (1968). ‘Empirical foundations for a theory of language change’, in Winfred P. Lehmann and Yakov Malkiel (eds.), Directions for historical linguistics: a symposium. Austin: University of Texas Press, 95–195. Westergaard, Marit (2005). ‘Norwegian child language and the history of English: the interaction of syntax and information structure in the development of word order’, in Kevin McCafferty, Tove Bull, and Kristin Killie (eds.), Contexts—historical, social, linguistic. Studies in celebration of Toril Swan. Frankfurt: Peter Lang, 293–310. Westergaard, Marit (2009). ‘The development of word order in Old and Middle English: the role of information structure and first language acquisition’, Diachronica 26: 65–102. Westergaard, Marit, and Vangsnes, Øystein (2005). ‘Wh-questions, V2, and the left periphery of three Norwegian dialect types’, Journal of Comparative Germanic Linguistics 8: 117–58. Whitman, John, Jonas, Dianne, and Garrett, Andrew (2012). ‘Introduction’, in Dianne Jonas, John Whitman, and Andrew Garrett (eds.), Grammatical change: origins, nature, outcomes. Oxford: Oxford University Press, 1–12. Whitman, John, and Paul, Waltraud (2005). ‘Reanalysis and conservancy of structure in Chinese’, in Montserrat Batllori, Maria-Lluïsa Hernanz, Carme Picallo, and Francesc Roca (eds.), Grammaticalization and parametric change. Oxford: Oxford University Press, 82–94. Williams, Alexander (2000). ‘Null subjects in Middle English existentials’, in Susan Pintzuk, George Tsoulas, and Anthony Warner (eds.), Diachronic syntax: models and mechanisms. Oxford: Oxford University Press, 164–87. Williams, Edwin (1982). ‘Another argument that passive is transformational’, Linguistic Inquiry 13: 203–38. Willis, David, Breitbarth, Anne, and Lucas, Christopher (2013). ‘Comparing diachronies of negation’, in David Willis, Anne Breitbarth, and Christopher Lucas (eds.), The history of negation in the languages of Europe and the Mediterranean, vol. 1: Case studies. Oxford: Oxford University Press, 1–50.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
References
Winand, Jean (1992). Études de néo-égyptien, vol. 1: La morphologie verbale. Liège: Centre informatique de philosophie et lettres. Witko, Jacek (2008). ‘Genitive of negation in Polish and single-cycle derivations’, Journal of Slavic Linguistics 16: 247–87. Woolford, Ellen (2006). ‘Lexical case, inherent case, and argument structure’, Linguistic Inquiry 37: 111–30. Wouden, Ton van der (1997). Negative contexts: collocation, polarity and multiple negation. London: Routledge. Wurff, Wim van der (1997). ‘Deriving object-verb order in late Middle English’, Linguistics 33: 485–509. Wurff, Wim van der (1999). ‘Objects and verbs in modern Icelandic and fifteenth-century English: a word order parallel and its causes’, Lingua 109: 237–65. Wurmbrand, Susi (2001). Infinitives: restructuring and clause structure. Berlin: Mouton de Gruyter. Xu, Dan (2006). Typological change in Chinese syntax. Oxford: Oxford University Press. Yang, Bojun, and He, Leshi (1992). Gu Hanyu Yufa ji qi Fazhan. [Archaic Chinese and its development.] Beijing: Yuwen Publishers. Zagona, Karen (1995). ‘Temporal argument structure: configurational elements of construal’, in Pier Marco Bertinetto, Valentina Bianchi, James Higginbotham, and Mario Squartini (eds.), Temporal reference, aspect and actionality, vol. 1: Semantic and syntactic perspectives. Turin: Rosenberg & Sellier, 397–410. Zamparelli, Robert (1995). ‘Layers in the Determiner Phrase’, PhD dissertation, University of Rochester. Zhang, Yujin (2001). Jiaguwen Yufaxue. [Grammar of the oracle bone inscriptions.] Shanghai: Xuelin. Zhou, Fagao (1959). Zhongguo Gudai Yufa [A historical grammar of Ancient Chinese], vol. 3. Taipei: Academia Sinica Institute of History and Philology. Zhou, Fagao (1963). ‘Stages in the development of the Chinese language’, in Zhongguo Yuwen Luncong [Collected Papers on Chinese Language]. Taipei: Zhengzhong Shuju, 432–8. Zhou, Guangwu (1959). ‘Xian Qin Foudingci Binyu Weizhi Wenti’ [On the positions of negators and objects in pre-Qin texts], in Zhongguo Yuwen Zazhi She (ed.), Yufa Lunji. Beijing: Shangwu Yinshuguan, 128–92. Zsilinszky, Éva (1991). ‘A névutók’ [Postpositions], in Loránd Benkő (ed.), A magyar nyelv történeti nyelvtana I. A korai ómagyar kor és elő zményei. [Historical grammar of Hungarian I. Early Old Hungarian and its ancestors.] Budapest: Akadémiai Kiadó, 442–60. Zsilinszky, Éva (1992). ‘A névutók’ [Postpositions], in Loránd Benkő (ed.), A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor. Morfematika. [Historical grammar of Hungarian II/1. Late Old Hungarian. Morphemics.] Budapest: Akadémiai Kiadó, 696–715. Zubizarreta, Maria-Luisa (1998). Prosody, focus and word order. Cambridge, MA: MIT Press. Zwart, C. Jan-Wouter (1993). ‘Dutch syntax: a minimalist approach’, PhD dissertation, University of Groningen.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Languages Bulgarian 179, 184–6, 189, 191–2, 195, 265n. Catalan 113n., 280–2, 290–8 Old Catalan 113n., 280, 282, 285–7, 289n., 294–8 Celtic 156 Chinese 193, 350, 357n., 369n. Archaic Chinese 12 Late Archaic Chinese (LAC) 350–2, 356–69 pre–Archaic Chinese 351–3 Church Slavonic 232 Czech 180n., 183n., 192, 268 Danish 23–5, 27 Dutch 67n., 68, 246, 302–3, 312 Middle Dutch 322n. Egyptian Ancient Egyptian 128, 132 Coptic 132, 134–6, 138–44 Demotic 132, 139–42 Earlier Egyptian 132, 135 Late Egyptian 132–3, 139–41 Later Egyptian 126, 132, 144–5 Middle Egyptian 132–3, 138 Old Egyptian 132, 138 English 35, 44–5, 54–71, 81–2, 103n., 127, 136, 171n., 194, 200, 304–5, 308, 310, 315 Early Modern English 12, 344–5, 347 Middle English 12, 31, 33, 193, 277, 329–30, 332n., 339, 343, 345n. Early Middle English 33, 329–30, 343–6 Late Middle English 329, 330, 343–4 Modern English 2, 329, 330, 336, 338–41, 343, 347 Old English 4, 17, 31–5, 193, 236–40, 242, 244n., 277, 299–300, 302–3, 305–6, 318–30, 332n., 335, 338–9, 343–5, 347 present-day English, see English, Modern English Estonian 165 Faroese 2 Finnic 163–5, 167, 175, 178
Finnish 165, 364–5 French 104, 118–19, 160, 190, 250, 273 Canadian French 2 Frisian 246 Galician 113n. Old Galician 113n. Gallo-Romance 250 German 34–5, 40n., 44, 268, 304–6, 308, 310–15 Early New High German 4, 17–23, 26, 30–1, 340 Middle High German 29–30, 96, 340 Modern German 17–18, 20–3, 27, 29, 31, 289n., 301–3, 307 Old High German 12, 22, 26, 29–30, 35, 89, 96, 240n., 299–302, 305–8, 316–17, 322 Pennsylvania German 338 Upper German (Bavarian) 96 Germanic 171n., 190, 193, 230, 289, 299, 305, 316, 336–8, 340 Gothic 247 Greek 146, 197–200, 206, 208, 209n. 210, 215–16, 265n., 268 Ancient Greek 196, 204n., 265n. Classical Greek 199n., 203–5, 210 Cypriot Greek 198, 200, 202, 206–16 Hellenistic Greek 8, 198, 202–6, 208, 210, 213 Koine Greek 143, 202 Medieval Cypriot Greek 8 Pontic Greek 202 Gungbe 230 Hausa 143 Hungarian 72–5, 82, 83n., 84, 86, 92, 97 Middle Hungarian 83, 98 Modern Hungarian 74–5, 79–80, 82n., 83–4, 87–9, 91–2, 95, 98–100 Old Hungarian 72–80, 82–5, 88–100 Proto-Hungarian 75, 82, 84, 89, 92–3, 96–7, 100 Icelandic Modern Icelandic 4, 36–47, 329n. Old Icelandic 49, 51, 299–300
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Languages
Ingrian 165 Irish 40, 41, 52 Old Irish 305–6 Italian 104, 118–20, 143, 150, 153–5, 157n., 158, 160, 230, 289, 291 Brindisino 151, 161 Cellino San Marco 148 Francavilla Fontana 152 Friulano 11, 249–64 Leccese 148, 150 Mesagnese 158, 161 Padovan 251n., 254, 261, 263 Piedmontese 157n. Salentino 146–56, 157n., 158n., 159–62 Scorrano 146 Trentino 250, 260–1 Veneto 252, 263 Japanese 193, 340 Karelian 165 Kiowa 206 Korean 81, 193, 340 Ladin 250 Latin 151–2, 247, 265–7, 289n., 302, 309 Livonian 165 Macedonian 184, 189, 191–3, 195 Mari 165 Mordvin 165 Northern Sotho 170–1 Norwegian 44, 238
Proto-West Germanic 246 Proto-Ingvaeonic 246 Romance 102, 146, 157n., 158, 160, 195, 219, 230, 281, 289, 291–2, 294 Old Romance 11 Romanian 220n., 230 Middle Romanian 221–4, 226–7, 231–2, 235 Old Romanian 219–20, 222–35 Russian 192n., 268, 354, 356 Old Church Slavonic 179, 186–9, 191–2 Old Slavic 191 Sámi 163–4, 167–8, 174n., 175, 177–8 Inari Sámi 165 Kildin Sámi 165 Lule Sámi 164, 176, 178 Northern Sámi 164–5, 167–8, 169n., 173, 175, 177 Skolt Sámi 165 Southern Sámi 165–6, 175 Samoyed 165 Sardinian 282 Serbian 188–9 Old Serbian 179, 186, 188 Serbo-Croatian 8, 179–85, 187n., 189, 192, 196 Sicilian 268, 281 Slavic 179–80, 183, 186, 189, 191–2, 195–6, 357 Slovak 180n. Slovene 180n., 183n. Old Slovene 192 Spanish 102–4, 106n., 113–19, 194, 230, 280–2, 285n., 290, 292–8 Old Spanish 280, 282–5, 289n., 294–6, 298
Old Saxon 236, 241n. Tzeltal 81 Permic 165 Persian 81 Polish 40–2, 45, 47, 51–2, 230, 354 Portuguese 5, 291 Brazilian 51 European Portuguese 102–9, 113–19 Old Portuguese 102–3, 109–13, 118 Proto-Finnic 165, 171 Proto-Indo-European 187, 189–90 Proto-Slavic 189–90 Proto-Ugric 96
Ukrainian 39, 40–1, 42, 45, 47, 51, 192n. Uralic 96, 163, 178 Vedic Sanskrit 265n. Votian 165 Yiddish 25–6, 299–300, 349 Early Yiddish 338 Welsh 135, 156
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects accent 311, 314–15 accent-first based system 315 acquisition 2, 113, 127, 164, 173, 211, 213, 256–7, 259 Activation Condition 262 adjunct 45–7, 304 adposition 80–1, 83–5 postposition 5, 72–80, 73n., 79n., 82–5, 83n. preposition 5, 67 adverbial clause 269 agent(ive) 40, 40n., 44–5 Agree 137–8, 198–9, 202, 208, 209n., 212, 216, 262, 264, 303, 354n., 355, 359 agreement 72, 74–7, 79–80, 82–4, 164–7, 174, 177 anaphor 5, 40, 41–7, 50, 52 animacy 200, 203–4, 206–8, 211 antisymmetry, see word order, Kaynean approach aspect 168, 170, 179, 189–91, 364–5 auxiliary 96, 127, 129, 132–3, 138, 274 auxiliary verb-plus-infinitive, see tense, infinitival tense negative auxiliary 165 Avoid Structure Principle 257 Bare Phrase Structure 195 barrier 157n., 158n. Behaghel’s Law 307 binding 40, 43–5, 204n., 226, 304, 368–9 Borer-Chomsky Conjecture 3, 4, 5 Burzio’s Generalization 39 cartography 6, 9, 131, 134, 136, 143, 163, 255, 274n., 289 Case/case 8, 13, 36–7, 39, 203, 303, 323–5, 327, 333–4, 355–6, 366 abstract/structural Case 41, 212, 351, 354–6, 359, 362, 364, 368–9 active inherent Case (AIC) 199–200, 202, 208, 212–15 inherent Case 197–202, 208, 216, 359–60, 362, 364 lexical case 36
morphological case/case-marking 6–8, 50–1, 226, 357–8, 361, 364, 368–9 quirky Case 199 chain 357 Chain Uniformity Condition 183, 195 checking theory 125, 273, 303–4 clause-typing 267 clause-union, see restructuring clitic 8, 155n., 157n., 179–87, 193, 195, 199, 202, 205, 212–15, 234–5, 238, 241, 345, 351–2, 362, 367 clitic climbing 149–51, 153–5, 157, 158n., 159–61, 182, 184, 186, 189 clitic doubling 219–22, 224, 226, 231–5 clitic left dislocation (CLLD) 136, 234, 280 clitic movement, see movement second position cliticization, see clitic, Wackernagel cliticization subject clitic (SCL) 249–56, 251n., 253n. 258n., 260–4 verb-adjacent cliticization 179–80, 184–9, 192–3, 195–6 Wackernagel cliticization 179–82, 184–9, 180n., 183n., 192–3, 195–6 competence (I–language) 1 complementizer system, see functional phrase-structure categories C-drop 148–55, 157–61 complementizer deletion 149–51, 153–4, 161 dual complementizer construction 146–8, 153 complexity, see Equal Complexity Hypothesis conflation 256–60, 258n., 259n., 263 Constant Rate Effect/Hypothesis 5, 54, 59, 69 control 40–1, 43, 45–7, 52, 64 copula 54–5, 60, 64, 91–2, 166, 172–3, 175–6, 178 copy 303–5, 316 cycles/cyclic change 5, 72, 86, 96–8, 100 dative alternation 197, 200, 203, 208, 211, 216 dative experiencer 197, 203, 204n., 215 dative shift 197–8, 213
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects
defective intervention effect 197, 199–200, 202, 204, 209–10, 214–15, 262, 275 definiteness 24–5, 29, 278, 354–5 definiteness effect 27, 37, 39 Differential Object Marking (DOM) 219–35 discourse-configurationality 7, 129, 142 ditransitive 198, 200–1, 203, 208, 210 do support 193 Double Base Hypothesis (DBH) 302 Double Object Construction (DOC) 200 Doubly Filled Comp Effect 143 DP/TP-less language 194 drift 132 analytic drift 128, 132–5 economy 129–30, 155, 249 ellipsis 182, 184–6, 189 emphasis contrastive emphasis 139 emphatic tense 139–41 end-based approach 311 EPP 4, 17, 29, 35, 161n., 194, 211, 273–4, 277, 303 Equal Complexity Hypothesis 138 event 169–71, 169–70 eventivity 55–6, 59–60, 64–5, 68, 69–71 expletive 4, 8, 17, 18, 20–7, 29–35, 30n., 37, 92, 194 extraposition 300–2, 337–40 feature 138, 167 Case feature 198–9, 201–2, 208, 210, 212–14, 216 Edge Feature 211, 239 feature inheritance 161–2 201 feature transfer 161 finite feature 157n. focus feature 137, 314, 357n. [±Participant] 200–1, 208, 212 [+Past] feature 170–1, 173–4, 177 tense feature 137 topic feature 357n. finiteness 170, 172, 177 focus 11, 131, 139, 159n., 305, 311, 314–15, 331–2, 334–5, 336, 341–2, 352–3, 357 contrastive focus (emphatic focus) 10, 280–2, 289, 292–3, 297, 306, 308, 316 focus feature, see feature focus movement, see movement identificational focus (exhaustive) 11, 268, 289
information focus (presentational) 10, 11, 268–9, 271–2, 275–6, 278, 289, 306, 308, 316, 319 narrow focus 308, 340–1 unmarked focus 281–90, 292–5 verum focus 136 Verum Focus Fronting (VFF), see movement weak focus, see focus, unmarked focus weak focus fronting (WFF), see movement functional hierarchy 115, 118 functional phrase structure categories 125 Agr 27n., 126, 130, 155 Applicative 200–1, 208, 366 Aspect 134–5, 303, 355 Ax(ial)Part 5, 73, 81–5 Cl(assifier) 108 C(omplementizer) 9, 24, 137, 143, 146–62, 157n., 158n., 230, 260, 279, 320 Contr(ast; head of ContrP) 238 D 10, 102–3, 105, 107, 113, 115, 117–18, 193–4, 222, 229–31 Dir(ectional) 278 Fin(iteness) 137, 142–3, 151–3, 155, 160, 166–7, 173–5, 237, 239, 246, 256–8, 263, 267 Foc 87, 89, 131, 137, 143–4, 222, 229–31, 256, 258–60, 275, 279, 280n., 289n., 315–16, 337, 343, 348–9 FocvP 268–9, 275–6 Force 143, 226, 238–9, 246, 256–8, 267, 340, 349 F(unctional) 274, 277 I 157n., 251, 256, 258n., 262, 264, 310, 337 Infl 127, 130 Int 143–4 Mood 134–5, 155, 345n. n 356, 358, 362, 364, 367–8, 370 Neg(ation) 5, 87, 90–1, 97, 100, 103, 105–6, 113, 114n., 118–19, 354–8, 363–5, 368, 370 Num 103, 108n., 114n. Path 81, 83–5 Place 81, 83–5 Pl(ural) 103, 105, 107–8, 113n., 119 Pol(arity) 8, 12, 166–7, 173–4, 178, 282, 295–8 Predicate 303 Subj(ect) 274 T(ense) 8, 9, 17, 24, 27, 29, 33, 35, 92, 126, 130, 134, 137, 143, 150–1, 155, 157–62, 166–7, 171, 173–5, 177–8, 179, 186, 189,
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects 193–6, 214, 251, 273–4, 279, 319–20, 328–31, 345n., 349, 359 Top 87, 89, 131, 143, 222, 229–31, 234–5, 256–60, 289n., 337, 343, 348–9 v 9–10, 27, 35, 157–8, 159n., 213, 273–7, 289n., 304–5, 311, 313, 315–16, 348–9, 354–5, 357–9, 364–6, 368, 370 V 157n., 319–20, 328–30 v* 157, 200, 211–12, 214 Voice 200–1 functional superstructure 125–6, 129–31, 135, 144–5 gap non-subject gap 19, 21, 23, 31 parasitic gap 338–40, 347–8 subject gap 19–20, 22–35 gradience 5, 52 gradualness 5, 176–8 grammar competition 189, 216 grammatical weight 323–4, 325n., 327, 330–1, 333–4 grammaticalization 4, 72–7, 79–84, 128, 138, 142–3, 222, 224, 227, 235, 297–8 degrammaticalization 80 grammaticalization cline 128–9 Head-Final Filter (HFF) 308–11 Head Movement Constraint 195 hierarchy of postverbal position 322–3 I–language, see competence impersonal construction 36–41, 43–51 active impersonal construction 39–42, 47, 52 incorporation 6 Inertia (Principle) 2, 36, 52 information structure 9, 131, 219, 222, 227, 229–31, 238, 244–5, 249, 251, 261–2, 276, 289, 294–5, 299, 301, 305, 308, 314–16, 322–7, 333–4, 336–7 background 306, 308, 315–16 common ground 306 contrast, see focus; topic discourse structure 25–6, 319, 329–30, 333–5 focus, see focus topic, see topic innovation 236, 246 interface 299, 301, 303, 308, 311, 315–16, 339 island, see locality condition Law of Growing Elements 305–6 left edge 150
Left Edge Fronting (LEF) 265–6, 268–72, 275–9 left periphery 126, 131, 135, 143–4, 150, 159n., 222, 227, 229–31, 234–5, 236n., 249, 251–2, 255–7, 258n., 260–3, 268–9, 276, 289n., 294, 316, 337 low left periphery 348–9 length 333n. Lexicon 3, 5 functional categories, see functional phrase-structure category Linear Correspondence Axiom 272 linearization 339–40 locality condition 181–2, 345, 347–8, 360, 362 Logical Form (LF) 91, 125, 199, 304, 314, 316, 355, 357n. Main Clause Phenomena 268, 276 Mapping Hypothesis 195, 355 mass/count distinction 107–8 Minimal Link Condition 275 Minimalism/Minimalist analysis 5, 7, 9, 13, 125 Minimality, see defective intervention effect Mirror Principle 125 modal 127 modality 146–7, 153–6, 159, 162 morphology 6 agglutination 127, 132 analyticity 132 analyticization, see drift, analytic drift case morphology, see case, morphological case complex morpheme 94 conditional 141–2 erosion, loss, and simplification 126–9, 132, 190, 197–8, 202–3, 216 evidential 7, 141–3 finiteness 8 inflectional morphology 125, 127, 129 isolating 127 morphological complexity 125–9, 131–2, 138, 145 Morphological Component 357 morpho(phono)logy 7 opacity 126–7 paradigmatization 139–41 richness 6 subjunctive 147n. syncretism 11 syntax–morphology interface 125, 129 synthesis 132, 134 tense, see tense Move 198, 212, 216
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects
movement 138, 150 clitic movement 180–1, 184, 186, 195, 212–13, 234, 345n. emphatic inversion 106n. focus movement 12, 95, 144, 230, 332n., 333–4, 352–3 head movement 166–7, 186; see also Head Movement Constraint Heavy NP Shift (HNPS) 244, 336–49 loss of movement 129–30, 132, 135, 137, 145 and negation 87, 89, 90–2, 97, 100, 242 Neg-to-D 103, 115, 117 nominal inversion 113n., 117 nominal negative inversion 102–3, 105–8, 110, 113–14, 118–19 NP-fronting 225–7, 229, 231 NP-movement 37, 39 particle doubling 136–7, 143 particle movement 136–7 post-posing 307 pre-posing 307 pronominal object shift 350–4, 356–8, 360–70 QP-fronting 282, 289n., 291, 293–4, 298 Q-raising 98 rightward movement 12, 244, 337–9, 343–4, 347, 349 scrambling 112, 113, 118, 210–12, 269, 279, 304–5, 339–40 short movement 272 subject–auxiliary inversion (SAI) 60–4 subject postposition 25–6 subject–verb inversion 280–1, 292 T-raising 150–1, 157, 160 verb fronting, see movement, verb movement verb movement 125–6, 129–30, 135 Verb Second (V2), see word order Verum Focus Fronting (VFF) 295–6 vP-intraposition 304 vP-movement 276, 278 V-raising, see movement, verb movement V-to-T movement, see parameter, V-to-T Parameter weak focus fronting (WFF) 283–96, 298 wh-movement 65, 67, 143, 182, 242, 244, 247 Narrow Syntax 339 negated past, see tense, past tense negation 5, 86–101, 104, 163–6, 172–8, 242, 350–4, 356–8, 360–4, 366–9 focus negation 88–9
negative auxiliary, see auxiliary, negative auxiliary Negative Concord 5, 86, 88, 92, 97, 100, 112, 334 negative cycle, see cycles/cyclic change negative force 86, 92, 95–7, 100 negative indefinite 86, 88, 90, 97, 99–100 negative inversion 113 negative particle 86, 87–9, 92–7, 99–100 negative polarity item 92–8, 102–6, 107n., 108, 110, 112, 114n., 118 negative pronouns 88, 90, 100 negative scope 90–1, 98, 100 predicate negation 88 non-finite forms 168n. infinitive 167, 345–6 past participle 164–5, 167–8, 170–5, 177–8 progressive 167 Numeration 125 parameter/parametrization 3, 6, 329n.; see also Borer-Chomsky Conjecture Head Parameter 4, 6, 299–301, 309, 320, 329n., 330, 340 morphological expression of parameters 127, 129, 145 Null Subject Parameter 6 parameter resetting 126–7 presence of T 193 Split-IP parameter 130 Verb Second (V2) Parameter 6; see also word order V-to-T Parameter 6, 162n., 238, 273–4 parsing 221 particle 135–6, 143 particle doubling, see movement particle movement, see movement TAM 134–8, 143 verbal particle 130 passive 4, 36–41, 44–8, 50–2, 174n. impersonal passive 39, 40n., 42–3, 46–52 past participle, see non-finite forms perfect construction 168–9, 170n., 171–2, 175–6, 178 periphrasis do-periphrasis 132, 134 Person Case Constraint (PCC) 183–6, 189, 195–7, 200–1, 205–8, 213 phase 10, 157, 158n., 161–2, 200, 239, 311–13 Phase Impenetrability Condition (PIC) 150, 354n.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects Phonological Form (PF) 3, 137, 148, 153–5, 159–61, 182, 199, 303–4, 309, 314, 339–40, 357n. PF-transparency 310 phonological merger 260n. phonosyntactic doubling, see raddoppiamento fonosintattico (RF) pied-piping 273, 277, 303, 332 polarity 119, 166, 296–7 and word order 102–10, 113–19 positive polarity item 102, 104, 112, 118 possessive 72, 74–80, 82–5 double marking of possessive 79, 84 postposition, see adposition preposition, see adposition processing 7, 333n. pro-drop 254; see also subject, null subject; parameter, Null Subject Parameter prominence 311, 313–15 pronominal object shift, see movement prosody 12, 299, 301, 305–6, 309, 311–16, 351–2, 365, 367 quantifier gradable 108 quotative 54–71 raddoppiamento fonosintattico (RF) 151–6, 159–62 reanalysis 39, 42, 47, 49, 52, 69, 83, 84, 90, 100, 110, 113, 115, 118, 127, 129, 133, 156, 161, 172, 174–5, 198, 208, 211–12, 216, 222–4, 229, 231, 233, 235, 246, 250, 277 reconstruction 128 referential stability 221, 233, 235 reflexive 43–4, 49–50, 52; see also anaphor relation-based approach 311–12 Relativized Minimality 275 relic 353 restructuring 155, 157 retention 236 Rich Agreement Hypothesis 125–6 Right Roof Constraint 340 salience 51 scrambling, see movement semantic bleaching 222, 224 simplicity, see Equal Complexity Hypothesis simplicity metric 213 situation pronoun 21, 22–3, 26–7, 29–31 smuggling 275 soft mutation 156
split CP hypothesis 9, 148, 236, 238, 289, 337 stativity 55–6, 59–60, 64–5, 68–71 stress-first-based account 311, 314–15 stylistic fronting 239 subject 11, 17, 26–7, 29, 34–5, 42, 238 null subject 39, 42–3, 47, 247; see also pro-drop; parameter, Null Subject Parameter subject coreferentiality 148, 151, 155, 157, 159–61 subject extraction 26–9, 34, 35 subject position 17, 26–7, 29, 34, 35, 39, 87 subject-prominence 131 subject doubling 249–502, 52, 254–60, 262–3 suffix 73–6, 83–5 tense 7–8, 126, 132, 140, 165–71, 173, 175, 177–9, 188–93, 196 absolute tense 170, 173 infinitival tense 132–3 past tense 164–6, 170–1, 173–8 relative tense 138–41, 142, 168, 170, 173 sequence of tense 194 that-trace effect 194 thematic encoding 227–8 topic 10, 131, 144, 159n., 323 aboutness topic 10, 143, 229, 231 backgrounding topic 219, 222 contrastive topic 10, 219, 222, 228–9, 231, 235 familiar topic 10, 136, 231–2, 238, 245 topic-prominence 131 topicalization 143, 182, 265, 269, 280n., 292, 305n., 349 topic-focus field 126, 129, 136, 158–9 Universal Grammar (UG) 113, 196 unaccusative (/nonagentive) 40–1, 47–9, 52 Uniform Chain Condition 186 Uniformity of Theta Assignment Hypothesis (UTAH) 200 univerbation 5, 6 Universal Base Hypothesis 302–3 variability 5, 51–3 Wackernagel’s Law 186, 189 wh-in-situ 143 word order 6–7, 12, 86, 128, 133–4, 299 AUX-SVO 133 DP-internal word order 117 free inversion 110, 112n., 113, 118
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
Index of Subjects
Head Parameter based approach 12 Kaynean (antisymmetric) approach 12, 336–9, 343–4, 347–9 mixed word order 299–301, 303, 305–6 narrative inversion 239 and negation 89–90 OV/VO 12, 51, 193, 278, 289n., 300–2, 318–19, 321, 325–35, 338–9, 343–7, 349, 351; see also Head Parameter OVS 289 SVO 134
Tense-final/Tense-initial 318–21, 323–35, 338–9, 343–4, 347, 349 VAux/AuxV, see word order, Tense–final/Tense–initial Verb-Final 237, 272–3, 276–7; see also word order, OV/VO Verb-Initial (Verb First) 32, 133, 236, 237 Verb Second (V2) 25, 26, 132, 135, 137, 154, 239–40, 243n., 289, 305, 320; see also Verb Second Parameter Residual Verb Second 137 Verb Third (V3) 236, 240–1, 243n.
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
OX F O R D S T U D I E S I N D IAC H R O N IC A N D H I S T O R I C A L L I N G U I S T I C S 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; David Willis, University of Cambridge published 1 From Latin to Romance Morphosyntactic Typology and Change Adam Ledgeway 2 Parameter Theory and Linguistic Change Edited by Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar 3 Case in Semitic Roles, Relations, and Reconstruction Rebecca Hasselbach 4 The Boundaries of Pure Morphology Diachronic and Synchronic Perspectives Edited by Silvio Cruschina, Martin Maiden, and John Charles Smith 5 The History of Negation in the Languages of Europe and the Mediterranean Volume I: Case Studies Edited by David Willis, Christopher Lucas, and Anne Breitbarth 6 Constructionalization and Constructional Changes Elizabeth Traugott and Graeme Trousdale 7 Word Order in Old Italian Cecilia Poletto 8 Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent 9 Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli 10 Vowel Length from Latin to Romance Michele Loporcaro
OUP CORRECTED PROOF – FINAL, 7/2/2015, SPi
11 The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by Katalin É. Kiss 12 Syntactic Reconstruction and Proto-Germanic George Walkden 13 The History of Low German Negation Anne Breitbarth 14 Arabic Indefinites, Interrogatives, and Negators A Linguistic History of Western Dialects David Wilmsen 15 Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden in preparation Variation and Change in the Syntax of Portuguese Relative Clauses Adriana Cardoso Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou The Syntax of Old Romanian Edited by Gabriela Pană Dindelegan Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti The Historical Dialectology of Arabic: Linguistic and Sociolinguistic Approaches Edited by Clive Holes A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τ oς and its Implications for Language Change and Structure Brian D. Joseph Gender from Latin to Romance Michele Loporcaro The Syntax and Semantics of Vedic particles John J. Lowe Quantitative Historical Linguistics Barbara McGillivray and Gard Jenset Syllable and Segment in Latin Ranjan Sen Syntactic Change and Stability Joel Wallenberg The History of Negation in the Languages of Europe and the Mediterranean Volume II: Patterns and Processes Edited by David Willis, Christopher Lucas, and Anne Breitbarth