Cycles in Language Change (Oxford Studies in Diachronic and Historical Linguistics) 9780198824961, 0198824963

Citation preview

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cycles in Language Change

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

O X F O R D S T U D I E S I N D I A CH R O N I C A ND H I S T O R I CA L L I N G U I S T I CS   Adam Ledgeway and Ian Roberts, University of Cambridge              Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge                         Word Order Change Edited by Ana Maria Martins and Adriana Cardoso  Arabic Historical Dialectology Linguistic and Sociolinguistic Approaches Edited by Clive Holes  Grammaticalization from a Typological Perspective Edited by Heiko Narrog and Bernd Heine  Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou  Indefinites between Latin and Romance Chiara Gianollo  Verb Second in Medieval Romance Sam Wolfe  Referential Null Subjects in Early English Kristian A. Rusten  Word Order and Parameter Change in Romanian A Comparative Romance Perspective Alexandru Nicolae  Cycles in Language Change Edited by Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen For a complete list of titles published and in preparation for the series, see pp. –

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cycles in Language Change Edited by MIRIAM BOUZOUITA, ANNE BREITBARTH, LIEVEN DANCKAERT, AND ELISABETH WITZENHAUSEN

1

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

3

Great Clarendon Street, Oxford,  , 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 Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen  © the chapters their several authors  The moral rights of the authors have been asserted First Edition published in  Impression:  All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press  Madison Avenue, New York, NY , United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number:  ISBN –––– Printed and bound in Great Britain by Clays Ltd, Elcograf S.p.A. 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, 13/8/2019, SPi

Contents Series preface List of abbreviations Notes on contributors . Cycling through diachrony Anne Breitbarth, Lieven Danckaert, Elisabeth Witzenhausen, and Miriam Bouzouita . Cyclical change and problems of projection Elly van Gelderen

vii ix xv 



. When morphological and syntactic change are not in sync: Reassessing diachronic implications of the Rich Agreement Hypothesis Eric Fuß



. The clitic doubling parameter: Development and distribution of a cyclic change Susann Fischer, Mario Navarro, and Jorge Vega Vilanova



. Weak elements in cycles: A case study on dative pronouns in Old Italo-Romance Jacopo Garzonio and Silvia Rossi



. On the emergence of personal articles in the history of Catalan Judy Bernstein, Francisco Ordóñez, and Francesc Roca



. Bare singular nouns in Middle Norwegian Kari Kinn



. What kind of constructions yield what kind of constructions? Andreas Blümel and Marco Coniglio



. Quantificational cycles and shifts Moreno Mitrović



. On the relative cycle: The case of P+che relative clauses from Old to Modern Italian Cecilia Poletto and Emanuela Sanfelici . French negation, the Superset Principle, and Feature Conservation Karen De Clercq

 

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

vi

Contents

. From negative cleft to external negator Elitzur A. Bar-Asher Siegal and Karen De Clercq



. Changes in the argument and event structure of psych verbs in the history of Spanish Montserrat Batllori, Elisabeth Gibert-Sotelo, and Isabel Pujol



References Index

 

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Series preface Modern diachronic linguistics has important contacts with other subdisciplines, notably first-language acquisition, learnability theory, computational linguistics, sociolinguistics, and the traditional philological study of texts. It is now recognized in the wider field that diachronic linguistics can make a novel contribution to linguistic theory, to historical linguistics, and arguably to cognitive science more widely. This series provides a forum for work in both diachronic and historical linguistics, including work on change in grammar, sound, and meaning within and across languages; synchronic studies of languages in the past; and descriptive histories of one or more languages. It is intended to reflect and encourage the links between these subjects and fields such as those mentioned above. The goal of the series is to publish high-quality monographs and collections of papers in diachronic linguistics generally, i.e. studies focusing on change in linguistic structure, and/or change in grammars, which are also intended to make a contribution to linguistic theory, by developing and adopting a current theoretical model, by raising wider questions concerning the nature of language change, or by developing theoretical connections with other areas of linguistics and cognitive science, as listed above. There is no bias towards a particular language or language family, or towards a particular theoretical framework; work in all theoretical frameworks, and work based on the descriptive tradition of language typology, as well as quantitatively based work using theoretical ideas, also feature in the series. Adam Ledgeway and Ian Roberts University of Cambridge

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

List of abbreviations 

first person



second person



third person

ACC

accusative

ADJ

adjective

ADV

adverb

AG

agent

AGR

agreement

Agr(P)

Agreement (Phrase)

AJT

acceptability judgement task

A(P)

Adjective (Phrase)

Appl(P)

Applicative (Phrase)

Arg(P)

Argumenthood (Phrase)

ART

article

ASP

aspect

Asp(P)

Aspect (Phrase)

AUX

auxiliary

BP

Brazilian Portuguese

BSN

bare singular noun

BUF

le bon usage French

C

common gender (Semitic)

Card(P)

Cardinality (Phrase)

Cat

Catalan

CF

colloquial French

C-I

Conceptual-Intentional (interface)

CL

clitic

Class(P)

Classifier (Phrase)

CLD

clitic doubling

CLF

classifier

COCA

Corpus of Contemporary American English

COHA

Corpus of Historical American English

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

x

List of abbreviations

COMP

complementizer

COND

conditional

CONJ

conjunction

COP

copula

C(P)

Complementizer (Phrase)

DAT

dative

DCat

Decadència Catalan (Early Modern Catalan)

DEF

definite

Deix(P)

Deixis (Phrase)

DEM

demonstrative

DEPV

Dative Experiencer Psych Verb

DiGS

Diachronic Generative Syntax

DIR

direct case

DISJ

disjunction

DN

double negation

DN

Diplomatarium Norvegicum

DOE

Dictionary of Old English

DOM

Differential Object Marking

D(P)

Determiner (Phrase)

DUR

durative

EMJ

Early Middle Japanese

EMPH

emphasis

EMSp

Early Modern Spanish

ENHG

Early New High German

EP

European Portuguese

EXH

exhaustification (operator)

EXP

experiencer

F

feminine gender

FC(I)

free choice (item)

Fin(P)

Finiteness (Phrase)

Foc(P)

Focus (Phrase)

FP

functional projection

FSEQ

(universal) functional sequence

FUT

future

GEN

genitive

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

List of abbreviations GN

geographical name

HON

honorific

HPP

Head Preference Principle

i-

interpretable feature

Ident(P)

Identifier (Phrase)

IE

Indo-European

IEC

Institut d’Estudis Catalans

IMP

imperative

IMPERS

impersonal

IND

indicative

INF

infinitive

I(nfl)(P)

Inflection (Phrase)

Init(P)

Initiator (Phrase)

INS

instrumental

IPFV

imperfective

JBA

Jewish Babylonian Aramaic

JP

Japonic

JSp

Judeo-Spanish

K(P)

Kase (Phrase)

L

first language

LA

labelling algorithm

LF

Logical Form

LOC

locative

L(P)

Lexical (Phrase)

M

masculine gender

MHG

Middle High German

N

neuter gender

NC

negative concord

NEG

negation

NOM

nominative

N(P)

noun (Phrase)

n(P)

light noun (Phrase)

NPI

negative polarity item

Num(P)

Number (Phrase)

O

object

xi

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

xii

List of abbreviations

OBJ

object

OBL

oblique

OCat

Old Catalan

OHG

Old High German

OJ

Old Japanese

Op(P)

Operator (Phrase)

OSp

Old Spanish

OT

Optimality Theory

PASS

passive

PCC

Person-Case Constraint

PDG

Present-day German

PL

plural

PN

proper name

PoP

Problems of Projection

P(P)

Preposition(al Phrase)

PPI

positive polarity item

PREP

prepositional case

PRF

perfect

Proc(P)

Process (Phrase)

PROG

progressive

PROX

proximate

PRS

present

PRT

particle

PSI

polarity sensitive item

PST

past

PTCP

participle

Q

question (feature)

Q(P)

(scalar) Quantity (Phrase)

RAH

Rich Agreement Hypothesis

RC

relative clause

REFL

reflexive

REL

relative

Res(P)

Result (Phrase)

RP

Root Phrase

RQM

rhetorical question marker

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

List of abbreviations S

subject

SBJ

subject

SBJV

subjunctive

SG

singular

SI

scalar implicature

Sp

Spanish

SUP

superlative

t

trace

Top(P)

Topic (Phrase)

T(P)

Tense (Phrase)

u-

uninterpretable feature

UG

Universal Grammar

V

verb first

V

verb second

V

verb third

Vfin

finite verb

VOC

vocative

V(P)

(lexical) verb (Phrase)

v(P)

light verb (Phrase)

WALS

World Atlas of Language Structures

WFC

was-für construction

WPG

was + partitive genitive

Σ(P)

Sigma (= polarity) (Phrase)

xiii

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Notes on contributors E A. B-A S is an associate professor at the Language, Logic and Cognition Center at the Hebrew University of Jerusalem. His work covers the two major academic fields of philology and linguistics. Focusing on Semitic languages in general and Hebrew and Aramaic in particular, he seeks to develop new methodologies for finding points of contact between these two distinct but interrelated fields. His recent work focuses on the semantics of negation, causative constructions, and reciprocity, and on the grammar of Jewish Babylonian Aramaic. M B has been a senior lecturer in Spanish linguistics (diachrony) at the University of Girona since . Previously, she was Tenure-track Lecturer at the UdG (–), and Teaching Assistant at the UAB (–). At her home institute she was given multiple awards for her research and lecturing activities. She has also been given recognition for her contribution to university management as vice-dean of the Faculty of Arts and as coordinator of the Degree in Spanish Language. Her research focuses on diachronic syntax, comparative grammar, and language change. J B completed the PhD in Linguistics at the CUNY Graduate Center. She is currently professor of Linguistics at the William Paterson University of New Jersey. Her research interests include synchronic and diachronic cross-linguistic syntactic variation in Romance languages, as well as standard and non-standard varieties of English. A number of her publications focus on the syntax of the noun phrase. A B̈  is a postdoctoral researcher of German linguistics at the University of Göttingen. His research interests include different syntactic phenomena mainly within Germanic, syntactic theorizing, and historical linguistics. He wrote a monograph, Symmetry, Shared Labels and Movement in Syntax (, Studia grammatica ), and a number of papers on topics such as V, long distance dependencies, ATB-movement, and the syntax of adverbial clauses. Moreover, he co-edited a book entitled Labels and Roots. M B works as a professor of Hispanic Linguistics at Ghent University, where she has coordinated the Diachronic and Diatopic Linguistics (ΔiaLing) research group since . Her interests include (Ibero-)Romance historical linguistics and synchronic dialectology, for which she combines traditional fieldwork and new interdisciplinary corpus-building methods. She has published extensively on the grammaticalization of clitic pronouns, the future tense, and the left periphery, as well as various diatopic phenomena. She also leads various research projects on dialectal grammatical variation (e.g. spatial adverbials), the morphosyntactic annotation and parsing of the COSER corpus, and the development of the Dialectos del español app. A B is associate professor of Historical German Linguistics at Ghent University, Belgium. She has published on issues in the historical syntax of, and language change in High

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

xvi

Notes on contributors

and Low German, as well as Dutch and English, and has led projects building parsed corpora for historical Low German and Southern Dutch dialects. Besides a monograph on the development of negation in Old Saxon and Middle Low German (Oxford University Press, ), she is the (co-)editor of several volumes on language change in the domains of negation and polarity, as well as diachronic change and stability in grammar. Currently, her research interests are expanding to include (micro-)diachronic syntactic change in progress and the role of prosody for syntactic change. M C is a junior professor of German linguistics at the University of Göttingen. His research topics are historical linguistics, syntactic theory, and corpus linguistics. He wrote a monograph on the syntax of German modal particles, Die Syntax der deutschen Modalpartikeln: Ihre Distribution und Lizenzierung in Haupt- und Nebensätzen (, Studia grammatica ), and several papers on different topics of diachronic and synchronic syntax, information, and discourse structure (discourse particles, sentence left periphery, historical V declarative clauses, etc.). L D currently works as a CNRS researcher at the University of Lille (France). He was previously employed at Ghent University (Belgium), where he obtained his PhD in . His expertise is in generative grammar and Latin syntax, with special emphasis on the study of word order and the use of quantitative, corpus-based methods. He is the author of two monographs: Latin Embedded Clauses: The Left Periphery (, John Benjamins) and The Development of Latin Clause Structure: A Study of the Extended Verb Phrase (, Oxford University Press). K D C is a postdoctoral researcher funded by the Flemish Fund for Scientific Research and working at Ghent University. She wrote her PhD thesis on the nanosyntax of negative markers. She is currently working on the fine-grained morphosyntax of Quantitywords (many/much; few/little), adjectives, V and V word order, degree comparison, and negation. S F is currently professor of Spanish and Catalan Linguistics at the University of Hamburg. She studied English and Spanish Philology at the FU-Berlin, the Universitat Central de Barcelona, and the University of Manchester. She obtained her PhD in general linguistics from the University of Potsdam and her postdoctoral qualification from the University of Stuttgart. Her main research interests are language change, historical syntax, and the syntax– morphology/phonology interface. She is the author of several books and many articles on different topics in General and Romance linguistics: quirky subjects, stylistic fronting, definiteness-effects, agreement restrictions, contact-induced change, and grammaticalization. E Fß graduated from the Goethe-University Frankfurt and held positions at the Universities of Stuttgart, Frankfurt, Leipzig, and the Institute for the German Language in Mannheim, Germany. He is currently Professor of Historical German Linguistics at the University of Bochum. His primary research interests are language change, linguistic variation, and the interface between syntax and morphology. J G is assistant professor of Linguistics at the University of Padua. His research focuses on morphosyntactic variation in Italo-Romance, with a special interest in

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Notes on contributors

xvii

negation, complementizer systems, wh items, quantifiers, and adpositions, from both a synchronic and a diachronic point of view. He is currently working on a book on the history of Italian negation, and is the principal investigator in a project on the grammar of medieval Venetian varieties. E  G is a syntactician interested in language change. Her work shows how regular syntactic change (grammaticalization and the linguistic cycle) provides insight into the faculty of language. Her  book, The Linguistic Cycle: Language Change and the Language Faculty (Oxford University Press) shows how cyclical change can be accounted for through an economy principle. Her Clause Structure (Cambridge University Press, ) examines a number of current debates in theoretical syntax. Her most recent book, The Diachrony of Meaning (Routledge ), is on the history of argument structure. E G-S is associate lecturer in Spanish Language at the University of Girona and at the Autonomous University of Barcelona. She received her PhD in Diachronic Linguistics and Comparative Grammar from the University of Girona, where she obtained the Extraordinary Doctorate Award within the Doctoral Programme in Humanities, Heritage and Cultural Sciences (–). She has conducted research stays at the University of Cambridge () and the University of Tromsø (). Her work, focused on the study of the syntax–lexicon interface from both a diachronic and a synchronic perspective, has been published in journals including Morphology, and in edited volumes at John Benjamins and Peter Lang, among others. K K is a postdoctoral fellow at the University of Oslo. Her research interests include comparative and diachronic syntax, language contact, and heritage languages. She was awarded a PhD in  at the University of Oslo; before taking up her current post, she held a postdoctoral research position at the University of Cambridge. M M´ holds a PhD in linguistics from Cambridge and has since worked as lecturer and researcher across Europe. He is also the founding director of the Bled Institute, a research and development NGO that promotes social mobility through science. His work focuses on the diachronic and typological syntax/semantics/pragmatics of logical vocabulary (the so-called ‘superparticles’ encoding coordination and quantification). His latest monograph, Microsemantics, is forthcoming with Springer. He is also editing a forthcoming volume on Logical Vocabulary & Logical Change, as well as co-authoring a forthcoming book on the theoretical and typological treatment of adjectives as a lexical category. M N studied Language and Literature in Buenos Aires and Romance Linguistics in Hamburg. He completed his Master’s Degree in Romance Linguistics in  at the University of Hamburg. He is currently a doctoral research assistant and is writing his PhD thesis on parametric emergence and the clitic doubling phenomenon within the biolinguistic framework. His main research interests are related to syntax, language variation and language change, language design, and biolinguistics. He is the author of many articles on these topics. F O´̃  received his PhD in Linguistics from the Graduate Center, City University of New York. He is a professor at Stony Brook University and his specialization has been the comparative study of the syntax of Romance languages. During Autumn  he was

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

xviii

Notes on contributors

a Distinguished Fellow of the Advanced Research Collaborative of the Graduate Center, City University of New York. His present research involves the study of the syntactic differences in the dialects of Spanish spoken in Latin America and Spain as well as studies of syntactic variation in Catalan, Spanish, and Italian dialects. He also co-founded Romania Nova with Mary Kato of the University of Campinas (Brazil). This international research collective promotes comparative research on Romance varieties spoken in the Americas. C P is full professor of general linguistics at the University of Padua and at the Goethe University in Frankfurt am Main. She is an associate member of the CNR research institute of cognitive science and technologies (ISTC). Her main interests are in dialectology, microvariation, and diachronic change of Romance standard and non-standard varieties. She has worked on the morphosyntax of subject clitics, interrogatives, the structure of the left periphery, and sentence typing; and more recently on negation and OV orders and sentence structure in Old Italian. She is part of the ASIt project for a syntactic atlas of Italian dialects. I P P has been a tenured lecturer in Spanish language (diachrony) at the University of Girona (Spain) since . Her research focuses on historical morphology, particularly on complex verbs and prefixation. She explores different aspects of verbal derivatives, such as their interaction with semantics, argument structure, and lexical aspect. She is also interested in the evolution of psychological verbs and in the history of prefixing processes. She is currently principal investigator of the project ‘Microparameters and Linguistic Change in Syntax, Morphology and Discourse’ (FFI--C--P), funded by the Spanish Ministry of Economy and Finance. F R received his PhD in Spanish philology at the Autonomous University of Barcelona in  and he has been professor of Spanish grammar at the University of Girona since . His general research interests are in the fields of theoretical linguistics and comparative grammar, especially among Romance varieties. He has published several studies on the syntax and morphology of Spanish and Catalan. He is a member of the research teams Lexicon and Grammar, and Diachronic Linguistics and Comparative Grammar, both at the University of Girona. S R is currently a postdoctoral researcher at the Goethe University, Frankfurt am Main, working on a DFG-funded project on Old Italian quantifiers under the supervision of Professor Cecilia Poletto. She obtained her PhD in  at the University of Padua with a thesis on the morphosyntax of prepositions and particles in Old English. She has also been working on several aspects of morphosyntactic (micro-)variation in British dialects, Italo-Romance varieties, and Old and Modern Italian, focusing particularly on prepositions, prepositional aspectual constructions, and pronouns. E S is assistant professor at the University of Padua. She received her doctoral degree in linguistics from the Scuola Normale Superiore, Pisa. Subsequently, she was appointed postdoctoral researcher at the Goethe University, Frankfurt am Main. Her research focuses on syntax, especially on word order variation, and relative and complement clauses. Her interests are centred on modelling linguistic variation and syntactic change from both a diachronic and an acquisitional perspective.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Notes on contributors

xix

J V V studied classical music and Spanish Philology at the Catalonia College of Music in Barcelona and the National University of Distance Education (UNED). He then completed his masters degree on Romance Linguistics at the University of Hamburg, where he is currently a doctoral research assistant. His main research interests are the syntax of core grammatical relations, Catalan syntax, language typology, and language change. He has written articles on different syntactic phenomena in Catalan and Spanish, such as verbal agreement, aspect, specificity, and clitic doubling. E W received her PhD in historical linguistics at Ghent University, Belgium, working on the functional change of the preverbal negative marker in Continental West Germanic languages within a generative framework. She is currently a postdoctoral researcher at Ruhr-Universität Bochum. She works quantitatively and corpus-based and is interested in the syntax–semantics interface, clause linking, modality, as well as onomastics.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

1 Cycling through diachrony A N N E B R EI TB A R T H , L I EV E N DA N C K A ER T , E L I S A B E TH W I T Z E N H A U S E N , A N D M I R I A M BO U Z O U I T A

. Cyclical change The fact that linguistic change often proceeds in a cyclical fashion, with new exponents of functional categories developing out of formerly lexical material, is hardly a new discovery; already in the early th century the pioneers of historical linguistics were describing such repeated developments (e.g. Bopp ; von der Gabelentz ). Also in the realm of word order, diachronic connections between different typological possibilities (e.g. VSO out of earlier V) can assume a cyclical character (although this type of cyclical change may be less common: see van Gelderen : – for discussion). Many instances of cyclical change go hand in hand with processes of grammaticalization, though it is not necessarily so. Well-known cyclical developments are, for instance, ‘Jespersen’s Cycle’ (e.g. Dahl  after Jespersen ; in this volume see also De Clercq and Bar-Asher Siegal; De Clercq), which famously describes a set of recurring changes affecting markers of sentential negation, the recurrent development of new agreement morphology from formerly independent pronominal elements (e.g. van Gelderen ; see also Fischer et al. in this volume on the ‘Object Agreement Cycle’; Fuß in this volume on the ‘Subject Agreement Cycle’), or the rise of tense inflection on verbs from earlier auxiliary verbs (e.g. Bybee et al. ; Krug ). In recent years it has been pointed out that cyclic developments of this type are much more pervasive than previously thought, and that they display a remarkably high degree of cross-linguistic regularity. Where a cyclical change is due to processes which are traditionally compiled under the term ‘grammaticalization’, it is worth keeping in mind that ‘grammaticalization’ is an umbrella term which encompasses various processes which interact and that it

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Anne Breitbarth, Lieven Danckaert, Elisabeth Witzenhausen, and Miriam Bouzouita . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

can affect different levels of linguistic signs.1 On the one hand, grammaticalization may involve semantic change, to the effect that the loss of lexical meaning leads to the acquisition of functional meaning (for formal semantic analyses, cf. von Fintel ; Eckardt ). On the other hand, grammaticalizing elements may undergo formal change, typically, the loss of morphological and phonological substance. While in individual instances of grammaticalization both these types of processes may be found (for example, in the case of inflectional morphemes newly created out of former auxiliaries, as in the Romance futures: Company Company ; Bouzouita ; Octavio de Toledo , inter alia), either type of change may constitute an instance of grammaticalization on its own. For instance, there is no formal difference between the auxiliary and the lexical verb haben ‘have’ in German (they have the same inflectional forms, though their syntactic distribution differs), and there is little to no semantic difference between a (non-emphatic) personal pronoun and an agreement marker. In the present volume, for instance, Bar-Asher Siegal and De Clercq analyse a type of change involving both univerbation and reanalysis from a biclausal to a monoclausal structure, without any concomitant change in semantics. At the same time, there are cyclic developments not obviously involving grammaticalization, in the sense that the item undergoing change does not gain or lose any functional content or structure. Examples include the semantic and distributional changes affecting indefinite pronouns and adverbs, which appear to follow a directional path along the semantic map of indefinite functions detailed by Haspelmath (), and which are cyclical in that changing elements are, or can be, replaced in their original functions by new indefinites. For the particular case of indefinites in the scope of negation, it has been proposed that the changes affecting these elements can be described in terms of a ‘Quantifier Cycle’ (Jäger a; Willis ; Willis et al. b; Gianollo ), with formerly non-negative elements gradually becoming ‘more negative’, i.e. restricted first to weak negative polarity item (NPI) contexts, then negative contexts, and ultimately becoming able to express negation on their own. This has, for instance, happened with French rien (‘thing’ > ‘nothing’) and personne (‘person’ > ‘no one’). In their non-polar and non-negative function, these elements have been replaced by indefinites of the quelque-series, originating from a free relative structure quel X que . . . (e.g. Foulet ; Haspelmath ). In the current volume, the contribution by Mitrović discusses one such semantic cycle. 1 In particular, grammaticalization is understood here as the result of different processes of linguistic change, first described as such by Meillet (), by which originally (more) lexical, free and contentful material comes to express increasingly functional, fixed, and abstract categories (e.g. Lehmann ; Traugott ; inter alia). Grammaticalization often occurs in a unidirectional fashion (Hopper and Traugott , ; Bybee et al. ), with (new) lexical material coming to replace the original material in its former function, and may be cyclic (Croft : ), but need not be. While some linguists have sought to establish grammaticalization as a distinct process of language change with unique properties (e.g. Lehmann ; Hopper and Traugott , ; Heine ), it has also been argued that none of the changes leading to grammaticalization are unique to it (as a process of change), but can be found in other, non-directional, or non-grammaticalization changes, too (cf. for example Campbell  and other contributions in the same special issue of Language Sciences).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Cycling through diachrony



Finally, it is not necessary for a particular change to come full circle in order for it to count as a ‘cyclical change’ for the purposes of the present volume. The comparison of attested diachronies across languages makes it possible to see partial cycles as glimpses of larger developments (for further motivation of a ‘comparing diachronies’ method, see Fleischer and Simon ). The contributions to the present volume consider various (partial) cycles in the sense just described, within a broadly generative, mostly feature-based framework, the basic tenets of which are outlined in the following section.

. Cycles in language change: a diachronic generative perspective Because of its directional character, cyclical change at first appeared to be at odds with a generative approach to linguistic change. According to Lightfoot (, ), the primary locus of language change is first language acquisition: during this process, learners can quite radically reanalyse the linguistic data they are exposed to. Under this view, acquisitional parameter setting is in principle not directional, but rather conditioned by cues in the primary linguistic data available to the language acquirer (Lightfoot , ; Fodor ; Fodor and Ferreira ; Yang ; Roberts ). During the last two decades, however, cyclical change has come to be a prolific field of study also within generative diachronic linguistics, providing a fertile testing ground for theories assuming a hierarchical organization of functional clause structure, and the presence of principles not specific to the faculty of language, such as economy principles. This shift was a consequence of a changing conceptualization of the human language faculty in Minimalism, which aims to reduce the amount of apparatus involved in syntactic computation. This led to a reduction of structure-building and -transforming operations to the single operation Merge (e.g. Chomsky , ). Language design is now viewed as the result of interaction between three factors, to wit, (i) genetic endowment (Universal Grammar), (ii) experience (mechanisms of language acquisition), and (iii) general cognitive principles not specific to the language faculty. One particular type of such so-called ‘third factors’ (Chomsky , ) is constituted by economy principles, which, in the form of ‘Fewest Steps’ or ‘Shortest Move’, have played an important role within Minimalism from the start (e.g. Chomsky ; Zwart ). Almost immediately, the particular importance of such principles for diachronic structural change was recognized.2 Roberts and Roussou () and van Gelderen () were among the first to propose formal accounts for grammaticalization within the Minimalist Framework. According to these authors, cyclical change, and in particular grammaticalization, is really a 2 While Lightfoot’s () Transparency Principle (‘derivations should be minimally complex’) was also an economy principle operative in diachronic syntax, it addressed the now obsolete mapping between deep and surface structures (mandating that the relationship between these levels of representation be transparent for the language-acquiring child).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

subcase of structural reanalysis, that is, parameter resetting in language acquisition, taking ‘parameters’ to refer to features of lexical items (Borer ; Chomsky ), driven by economy principles. Assuming that displacement in syntax (Move) is really Merge triggered by the presence of two features on the item affected by the operation—one that causes its merger in its base position, and one that causes the merger in the target position—, Roberts and Roussou argue that it is more economical for items to be merged only once. This leads to direct merger in the former target position and the loss of the earlier movement, as well as the opening of the original base position for the direct merger of new lexical material. Given that movement targets higher positions (originally accounted for in terms of conditions on licensing traces of movement, such as the Empty Category Principle from Chomsky ), this means that the loss of earlier movement leads to what Roberts and Roussou call ‘upward reanalysis’. Repeated occurrence of this process creates cycles of grammaticalization. Similarly, van Gelderen’s work over the past two decades has made heavy use of economy principles. Van Gelderen () proposed two such principles, the Head Preference Principle and the Late Merge Principle. The former mandates that given evidence compatible with either parse, language acquirers will postulate heads instead of phrases. The latter has the effect that elements are merged as high as possible. This proposal in a way combines Roberts and Roussou’s () idea that grammaticalization proceeds by upward reanalysis (= Late Merge), and Cardinaletti and Starke’s (: ) choice principle ‘Minimize Structure (up to crash)’, which mandates the insertion in a given position of the structurally least complex element that still leads to a grammatical outcome, and for which they explicitly suggest a possible diachronic application (Cardinaletti and Starke : , n. ). In more recent work, van Gelderen (e.g. , , ) has reformulated these principles as one single economy principle, namely Feature Economy. ()

Feature Economy Minimize the semantic and interpretable features in the derivation, for example: Adjunct Specifier Head Affix semantic > [iF] > [uF] > [uF] > Ø (after van Gelderen : )

As we will see, many contributions in the current volume analyse (partial) linguistic cycles in terms of upward reanalysis.

. The contributions to this volume The twelve contributions collected in this volume approach processes of (partial) cyclical change from a formal perspective, focusing in particular on how the changing feature composition of lexical and functional categories can affect the syntactic

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Cycling through diachrony



structure of the nominal and verbal domains, and more generally the clause. The chapters form three groups. After two chapters addressing a number of foundational questions about the nature of cyclical change, chapters  through  turn to cyclical changes involving functional categories in the nominal domain, with special reference to pronouns, determiners, and adnominal quantifiers. The final four chapters are devoted to cyclical changes in the clausal domain. .. Theoretical questions In the first contribution, Elly van Gelderen approaches cycles and the well-described reanalysis of a phrase as a head (van Gelderen b) from the perspective of Chomsky’s (, ) Problems of Projection papers (PoP). In the PoP approach, labelling is independent of Merge and happens when a syntactic derivation is transferred to the interfaces. Chomsky argues that labelling paradoxes arise when two phrases or two heads merge. The labelling algorithm is conceived of as a special case of minimal search, looking for syntactic heads in a particular domain. Therefore, when two phrases are merged together, there are two possible heads that could provide a label for the complex syntactic object created through XP–YP merge. Chomsky (: ) argues that there are two solutions to this paradox: either (i) one of the phrases moves so that there is only one syntactic head left (this is for example what happens when a subject moves to T in English), or (ii) both phrases are similar to a certain point, i.e. they share one or more features which can be taken as a label for the syntactic object. In her chapter, van Gelderen argues that there is a third solution to this labelling paradox, namely reanalysing a phrase as a head. She first discusses case studies from French and Athabaskan: in the former, phrasal subject pronouns are reanalysed as T heads, and in the latter, object pronouns become v heads. This type of analysis is extended to the development of articles and complementizers in English, and to copulas in Egyptian, which historically derive from demonstrative pronouns. Additionally, van Gelderen discusses English whether and how. These two elements maintain their phrasal status and are not reanalysed: the author argues that this is due to their being endowed with Q-features, which are relevant to the semantic interface, and therefore prone to being diachronically stable. The last section of the chapter addresses argumenthood, revisiting the question as to what drives the reanalysis of adjuncts as arguments. In van Gelderen (), this process was argued to be the result of the principle, ‘Be a specifier rather than an adjunct’. In PoP terms, it can be explained in a more principled fashion, namely as a reflex of a preference of set-merge (feature-sharing) over pair-merge. Eric Fuß’s chapter is concerned with the relation between morphology and syntax. As mentioned above, one of the driving forces behind cyclical change is often thought to be grammaticalization, i.e. a process primarily affecting individual lexical items. One important question related to this process is whether or not the feature

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

composition of grammaticalized functional categories has a direct impact on the shape of the syntactic structure in which it is merged, and on the impact it can have on other categories in the same configuration. The case study Fuß focuses on is related to the Subject Agreement Cycle, in which independent—often emphatic— subject pronouns grammaticalize to become bound agreement markers attached to the verb (on this development, see Givón , and more recently van Gelderen , this volume, as well as Givón ). An influential idea known as the ‘Rich Agreement Hypothesis’ (RAH) states that there is a non-trivial correlation between the height of verb raising in a given language, and whether this language has rich subject–verb agreement. In its strongest, biconditional form, which was recently revived in Koeneman and Zeijlstra (), the RAH implies that richly agreeing verbs always move to the TP-domain, and conversely that verb movement to T is always indicative of rich agreement. With respect to the Subject Agreement Cycle, the RAH thus predicts that the grammaticalization and subsequent loss of subject agreement markers should systematically correlate with presence and absence of V-to-T movement. Fuß discusses a number of case studies from English, Cimbrian, and Lithuanian which disconfirm the strong RAH, arguing instead that the TenseAspect-Mood morphology overtly marked on a given verb form is a better predictor for verb movement (compare Biberauer and Roberts b; Schifano , ; Haeberli and Ihsane ). This chapter thus helps us to understand why word order changes have proven difficult to analyse in terms of the linguistic cycle (but see van Gelderen : –). .. Cyclical change in the nominal domain Susann Fischer, Mario Navarro, and Jorge Vega Vilanova address in their contribution the emergence and distribution of clitic doubling (CLD) in Romance. The authors draw on a rich array of diachronic and synchronic data, starting with Old Spanish and Old Catalan, and proceeding to include modern Spanish varieties such as those spoken in Buenos Aires, Lima, and the Andes, and Judaeo-Spanish from Bulgaria, among others. The development of CLD is argued to be a cycle with five different stages, which interacts with a verb movement parameter, determining whether or not a language has V-to-C movement. CLD appears to have become fully productive after verb movement to the C-domain in declarative main clauses was lost. Concretely, in Old Spanish and Old Catalan a number of word order patterns were available to indicate that a given object refers to a highly accessible (in the sense of Ariel , among others) referent: once these word orders were lost (as a result of a re-setting of the verb movement parameter), the function accessibility marking was taken over by the CLD strategy, witness the fact that initially CLD appears only with highly accessible objects, i.e. those referring to humans and animates. Only later is CLD extended to less accessible ones, following the

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Cycling through diachrony



accessibility and definiteness scales proposed in e.g. Ariel () and Leonetti (). In view of this, CLD can be said to have partially taken over the function of expressing certain information-structural nuances which were expressed through word order in Old Romance. Also related to the diachronic development of pronouns is the chapter ‘Weak elements in cycles: A case study on dative pronouns in Old Italo-Romance’ by Jacopo Garzonio and Silvia Rossi. Focusing on the dative pronoun loro (them.), Garzonio and Rossi depart from the assumption that the development from strong to weak and eventually clitic pronouns follows a cyclic trajectory, arguing that the structural correlates of this process of progressive reduction need to be rethought. Specifically, they modify the influential proposal from Cardinaletti and Starke (), which says that pronominal elements differ in the amount of functional structure they are endowed with, yielding a tripartite taxonomy of strong pronouns (full XPs), weak pronouns (functionally deficient XPs) and clitics (functionally deficient X°s). Departing from this analysis, Garzonio and Rossi first of all suggest that this three-way classification is too crude, in that more than one type of weak pronoun needs to be distinguished. In addition, they assume that smaller functional categories lack a number of inner (rather than outer) layers, in particular a lexical core (of category nP), and one or more functional projections. One attractive consequence of this approach is that it allows one to maintain that grammaticalization entails the upward reanalysis of functional material (Roberts and Roussou ) considering that, in the case at hand, weak pronouns spell out higher functional heads than strong ones. Judy Bernstein, Francisco Ordóñez, and Francesc Roca are concerned with grammaticalization in the extended noun phrase as well, namely the genesis of personal articles, as exemplified by data from the history of Latin and Ibero-Romance, and in particular Catalan. Many varieties of present-day Catalan have personal articles—en for male referents, and na for female ones—, which occur only with [+human] proper names. The source of these items is the Latin lexical noun dominus / domina ‘lord, master / lady, madam’, which was typically used to refer to people belonging to the upper classes of society. In Old Ibero-Romance, this noun first grammaticalized to become a specialized honorific marker, and in modern Catalan, it further developed into a generic personal article which is used with proper names without any honorific connotation. In a similar way to the Romance development of the definite articles from demonstratives, the evolution of these personal articles is modelled as an instance of a cyclic change, whereby phrasal (adjectival) elements occupying specifier positions first develop into X° heads, and eventually into (clitic) affixes in the form of articles. The early honorific titles of Old Ibero-Romance are thus XPs and not heads. Bernstein, Ordóñez, and Roca further illustrate that this cyclic change encompasses phonological, morphological, syntactic, and also semantic shifts. Kari Kinn’s contribution zooms in on the distribution of bare singular nouns (BSNs) in Middle Norwegian, thereby shedding light on the missing link between

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

Old Norwegian, which is well known to have allowed BSNs (e.g. Rane het maðr ‘there was a man called Rane’, lit. ‘Rane was called [a] man’), and Modern Norwegian, where indefinite articles or definite suffixes are compulsory. Presenting novel data from a corpus of Middle Norwegian charters, Kinn shows that BSNs are found throughout the Middle Norwegian period, and that their loss is a more recent development than hitherto thought, roughly contemporary with, and arguably connected to, the loss of null arguments. Kinn’s data shed light on what has been called the ‘DP Cycle’ (for instance in van Gelderen : ch. ), i.e. the development of the system of definite and indefinite determiners in the history of Norwegian, and provide new evidence for the hypothesis that the development of the indefinite article passed through a stage at which it served as a specificity marker. Besides, BSNs with an indefinite interpretation are significantly more frequent than BSNs with a definite interpretation, lending support to the hypothesis that definite determiners grammaticalized earlier. Kinn proposes to capture the relevant developments in terms of definiteness being encoded in two separate functional projections that did not grammaticalize at the same time. This proposal can account for variation between bare nouns and nouns appearing with (the precursors of) determiners in Middle Norwegian. Andreas Blümel and Marco Coniglio continue with the topic of upwards or cyclical change within the extended functional projection of DP. They argue that the development of the was-für construction (WFC) in German (was für ein Buch ‘what kind of book’) follows a similar cyclic development in quantifying expressions as those discussed in Roehrs and Sapp (). Based on a corpus study, they confirm that the construction emerged in sixteenth-century Early New High German, and replaced an earlier functionally equivalent construction involving was + partitive genitive (WPG), which had already been attested in Old High German (waz worto ‘which of the words’), and which eventually disappeared in the nineteenth century. The authors go on to argue that the WFC not only supplanted the earlier WPGconstruction, it also directly evolved from it by a structural reanalysis of the earlier binominal construction (involving a light / silent noun) as a small-clause-like mononominal one. The D head of the higher (silent) noun was then recategorized as a functional head F, a higher head in the extended projection of the lower DP. Subsequently, F was lexicalized by für, which they argue only superficially resembles the homophonous preposition, but differs from it in not assigning accusative case. The silent noun taking the genitive complement structurally disappears in the process, and was is reanalysed as the specifier of FP. The final chapter on changes in the nominal domain is a contribution by Moreno Mitrović on the unidirectional, and often cyclical, interaction between syntax, semantics, and pragmatics in quantificational meanings, which is part of what Willis () (see also Willis et al. b) has dubbed the Quantifier Cycle. The chapter compares the diachronic developments of quantificational particles, drawing

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Cycling through diachrony



in Indo-European (IE) and Japonic, uncovering unidirectional patterns of change affecting quantificational expressions consisting of a wh-stem and a quantifier particle (‘μ’). Based on a comparison of a range of older IE and Japonic languages, Mitrović argues that all meanings of wh + μ expressions can be traced back to a wh-pronominal or indefinite origin with an existential meaning, and that they follow a universal and unidirectional cycle of grammaticalization, from an original existential use via a universal to a polar existential one. Following Chierchia (), he assumes that phenomena associated with quantifying expressions, such as polarity sensitivity or scalar implicatures, are anchored in syntactic feature specifications, which are licensed under Agree by an exhaustifier at the root of the clause, and which may be affected by syntactic reanalysis. According to Mitrović, diachronic changes in the specification of two features of quantificational expressions—δ (non-scalar alternatives, ‘also’) and σ (scalar alternatives, ‘even’)—can account for the cyclic pattern and the quantificational split into universal and existential expressions recurring across and within language families. .. Cyclical change in the clausal domain The final four chapters of this volume deal with cyclic phenomena affecting the structure of the clause. In the first contribution, the core message of Cecilia Poletto and Emanuela Sanfelici is a cautionary one: the authors argue that it may be worthwhile to explore alternative explanations for changes that are at first sight readily amenable to a ‘cyclical’ treatment. As their starting point they take a recent suggestion by Kayne (), that all overt terminals are necessarily phrasal, and concomitantly that all syntactic heads are always silent. If this proposal is on the right track, it would follow that it is no longer possible to account for the progressive (structural, phonological, semantic) reduction of functional elements undergoing grammaticalization by calling upon the Head Preference Principle (HPP). Acknowledging that the many phenomena that have been accounted for in these terms still call for an explanation once the HPP has been abandoned, Poletto and Sanfelici suggest that some processes of grammaticalization reflect a change in the internal structure of a—by assumption phrasal—function word (see also Garzonio and Rossi, in this volume). In particular, what may change over time is the selectional requirements of a given functional item, such as for example the possibility or otherwise that some category to appear with a silent nominal complement. The concrete case study these ideas are applied to is the history of the Italian relativizer che. In older stages of the language, this element could freely act as the complement of a preposition, whereas this option is highly restricted in the presentday language. In Benincà and Cinque (), this development has been interpreted to mean that che used to be a phrasal relative operator, but that it was later reanalysed as a relative complementizer with X°-status, in compliance with the Relative Cycle (cf. van Gelderen a). As an alternative, Poletto and Sanfelici propose that the

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

observed change affecting the distribution of che reflects the fact that in the modern language, che can no longer select a null nominal classifier (‘THING’). On the one hand, without referring to the notion of grammaticalization cycle, this proposal successfully captures the reduced range of syntactic environments in which che can occur, and on the other hand, it also correctly predicts that che ceases to be incompatible with animate antecedents. The following two chapters turn to issues relating to the Negative Cycle. Karen De Clercq presents a nanosyntactic account for the cyclical change in the expression of sentential negation in French, focusing on the change between the stage represented by le bon usage French (BUF), where sentential negation involves ne together with pas or a negative indefinite (personne, rien, . . . ) and the stage represented by colloquial French (CF), where ne is dropped. With Zeijlstra () and De Swart () (who themselves build on observations made in Haspelmath ), De Clercq assumes that personne and rien enter into concord relations with each other, whereas pas cooccurring with n-words also results in a double negation reading. Furthermore, she proposes a lexical decomposition of personne and rien in both CF and BUF, which she then uses to explain negative concord as well as double negative readings with pas. She proposes that the diachronic loss of ne follows from the interplay between the maximal use of the Superset Principle, which governs lexical insertion in the framework of Nanosyntax (Starke ), and the principle of Feature Conservation. The latter principle entails that the features necessary for the expression of negation remain constant over time, but that they are packaged differently at different diachronic stages. For reasons of economy, lexically stored structures grow over time, enabling them to spell out more features and thus leading to less spelloutdriven movement (on which, see Starke ; Baunaz and Lander ). While both personne and rien in BUF consist of a Foc-feature in addition to a Neg-feature, in CF the lexical structure has grown and spells out a T-feature as well, turning ne, which consists of T- and Mood-related features, into a superfluous element. Furthermore, it is argued that by means of the Superset Principle negative concord relations between negative indefinites can be accounted for, allowing French personne and rien to spell out as NPIs and n-words in BUF, and as NPIs, n-words, and negative quantifiers in CF. Without assuming covert items, the availability of negative concord is thus analysed as a consequence of the interplay between the decomposition of French n-words, the Superset Principle, and the structure of the universal functional sequence. The chapter by Elitzur A. Bar-Asher Siegal and Karen De Clercq continues on the topic of negative cycles. The case study focused on concerns the negative expression lāw in Jewish Babylonian Aramaic, which can be characterized as an external negator taking wide propositional scope, resulting in the meaning, ‘it is not the case’. The authors show that lāw is a univerbation of the regular negator lā and an agreement morpheme (-hu, ). Originally, this expression was a cleft: over time, it

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Cycling through diachrony



loses its biclausal structure, the newly grammaticalized negator being reanalysed as a left-peripheral focus in a single clausal domain. Crucially, the new negator still expresses external negation: in other words, this instance of syntactic change is not accompanied by a semantic change. The authors also consider whether there is a general pathway of change affecting negative clefts that grammaticalize as external negators. The authors provide data for Sicilian neca (Cruschina ; Garzonio and Poletto ) which can be shown to have undergone a similar development: in the latter case, three morphemes undergo a process of univerbation, yielding the negator neca, which can also be analysed as being merged in SpecFocP of the clause it negates. The final chapter of the volume is devoted to what we can call ‘first phase syntax’ (in the sense of Ramchand ), and in particular psych verbs in the history of Spanish. Building on work by van Gelderen (, a), Montserrat Batllori, Elisabeth Gibert-Sotelo, and Isabel Pujol argue that changes in the argument structure of Spanish experiencer predicates are correlated with a change in event structure. In particular, they show that Spanish psych predicates such as gustar ‘savour / like’ undergo a change affecting the realization of their external argument, which appears as an agentive subject in Old Spanish, and as a (dative) object experiencer from the sixteenth century onwards. Relatedly, the same class of verbs also changes from expressing a dynamic event to expressing a stative one, which raises the question as to how tightly argument and event structure are related, synchronically as well as diachronically. The data considered in this chapter stand out because they instantiate a type of cyclical change which is not coupled with, and by this token not driven by, any morphological change. In other words, it seems to be the case that cyclical change does not necessarily have to go hand in hand with a process of grammaticalization. On the other hand, it hardly seems accidental that in the prototypical case grammaticalization and cyclical change do in fact co-occur. The exact extent to which grammaticalization and the linguistic cycle are logically mutually independent no doubt deserves to be investigated more closely in future research. To conclude, the chapters united in this volume provide further evidence that cyclical instances of language change can fruitfully be analysed in a formal framework, although—inevitably—many open questions remain. In particular, in the spirit of Minimalism, where the role of the phonological (Sensory-Motor) and semantic (Conceptual-Intentional) interfaces is heavily stressed, it can be expected that future research in the generative paradigm will pay more attention to the relation between syntactic aspects of the linguistic cycle and mechanisms of semantic change (for a recent programmatic statement, see Deo ), as well as to the exact nature of the correlation between phonological reduction and grammaticalization, and by this token also cyclical change (cf. Roberts and Roussou : – for some speculation on this last topic).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Breitbarth, Danckaert, Witzenhausen, Bouzouita

Acknowledgements This volume unites a thematic selection from the papers presented at the eighteenth Diachronic Generative Syntax conference held at Ghent University in June . We gratefully acknowledge the financial support of the Research Foundation - Flanders (FWO) (conference organization grant K..N), of the research funds of the Faculty of Arts and Philosophy of Ghent University, of the FWO Odysseus project G (Liliane Haegeman), and of the research group ΔiaLing, as well as the organizational assistance given by Tom Parlevliet. We would also like to take the opportunity to thank the series editors of the series ‘Oxford Studies in Diachronic & Historical Linguistics’, Adam Ledgeway and Ian Roberts, for including the volume in the series, Julia Steer and Victoria Sunter for their editorial assistance at Oxford University Press, and thirty anonymous reviewers for their much-appreciated contributions.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

2 Cyclical change and problems of projection E L L Y V AN G E L D E R E N

. Introduction A linguistic cycle describes a regular pattern of language, a round of linguistic changes taking place in a systematic manner and direction. For instance, an independent pronoun may come to be dependent on a verb and be reanalysed as agreement and the independent pronoun may be renewed through a noun or demonstrative. These changes involve phrases (e.g. independent pronouns) reanalysing as heads (e.g. agreement) and adjuncts as renewed specifiers. Heads are also reanalysed as higher heads, but these will not be discussed in this chapter. Because the older phrases and heads are renewed, these changes are seen as cyclical, as argued in, for instance, van Gelderen (; ). The formal explanation for these cyclical changes is that, during acquisition, principles of economy predispose the learner to use simpler structures and features (van Gelderen ). The urge of speakers to be innovative may introduce new, loosely adjoined elements into the structure. In this chapter, I show that it is possible to see these changes as solutions to labelling problems. Chomsky (; ) advocates a system of free merge where labelling (of TP, DP, etc.) is done via a labelling algorithm (LA) because labelling is needed at the interface levels. This framework is known as the Problems of Projection (hence PoP) approach. When a head and a phrase merge, the LA automatically determines the head to be the label. However, in cases where two phrases merge, the LA cannot find the head and this results in a labelling problem. One of the phrases has to either move or share features with the other. In this chapter, I argue that the problematic merge of two phrases can be resolved in another way and this process is evident in language change, namely as a change from phrase to head. The change away from adverbials involves a reanalysis of pair-merge as set-merge, as in Chomsky (). The reanalyses of phrases as heads are varied: (a) subject and object pronouns to agreement, i.e. DPs to T and v heads, respectively, (b) demonstratives to C, D, and Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Elly van Gelderen . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

T heads, (c) wh-elements to C heads, (d) Adverb Phrases to ASP heads. There are additional changes of this type, e.g. PP to C heads and negative adverbs to Neg heads, but I have chosen to focus on changes that are very frequent (those in (a) and (b)) and those that are less frequent ((c) and (d)). As phrases reanalyse as heads, new phrases arise again. I look at one of these, i.e. the renewal of the subject from a topic. I argue that all these changes provide insight into some of the labelling mechanisms, e.g. simple search is preferred over sharing features and sharing features (set-merge) is selected over pair-merge. Each of the cases will be exemplified and discussed in terms of labelling. Section . will outline the basics of Chomsky (; ; ). Section . will provide a few instances of reanalysis of the subject and object DP to a T and v head, from French and Athabaskan languages, respectively. Section . is on the changes affecting demonstrative pronouns. They reanalyse as articles, complementizers, and copulas. I will provide a number of scenarios on what might prompt the reanalysis. Section . will examine further sources for C-heads, namely wh-elements, and will consider Adverb Phrases as they change to ASP. Section . turns to the change from adjunct to specifier. Section . is a conclusion.

.. From projection to the labelling algorithm (LA) Early Phrase Structure Grammar (e.g. Chomsky ) and X’-bar theory (e.g. Jackendoff ) take for granted that a phrase is headed and expands to a maximal projection with a specifier, head, and complement. This X’-schema is seen by many as perhaps one of the greatest insights into syntactic structure. The spirit of the current Minimalist Program (Chomsky  to the present), however, is to attribute as little as possible to the computation, restricting it to simple merge with a labelling algorithm needed for the conceptual–intentional interface. In early Generative Grammar (e.g. Chomsky : ), language-specific phrase structure rules, such as (), are responsible for generating sentence structure. (a) generates the basic sentence and (b) the Verb Phrase. Chomsky () and, especially, Jackendoff (: ) reformulate these rules as a category-independent and language-independent schema, as given in (). ()

a. S b. VP

! !

NP V

VP NP

()

a. XP b. X’

! !

YP X

X’ ZP

In the mid-s, the X’-schema of () is extended to grammatical categories, such as T, C, and D, and the result is the familiar structure in (), again with the head determining the label of the higher phrase.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



CP

()

Spec

C’ C

TP Spec John

T’ T may

vP John eat apples

Taking the Minimalist Program seriously means attributing less and less to Universal Grammar, in particular to rules such as (), and restricting the generative part of a derivation to a computational operation called Merge. External Merge (EM) takes two objects and yields an unordered set {X, Y} without a label (Chomsky : ); Internal Merge (IM) takes an already formed syntactic object and takes part of that and merges it with the original syntactic object. Labelling the set is not part of Merge and should therefore be avoided and left as a requirement of the interface. Not much is said about this requirement. Chomsky (: ) assumes that ‘for interpretation, syntactic objects must be labeled’ where, in an unpublished manuscript, he requires labelling ‘at the CI interface, and for the rules of externalization’. I will briefly come back to labels and interpretability when comparing the phi-features to the Q-features, but do not have much to add otherwise. The labelling algorithm (LA), stated in (a), involves just a minimal search and ‘must take place at the phase level, as part of the Transfer operation’ (Chomsky : ). It is like Agree, not Match, and part of Minimal Computation, i.e. a third-factor effect. Rizzi (: ) formulates it slightly differently, as in (b). ()

a. The Labelling Algorithm is ‘a special case of minimal search’ seeking ‘heads H within its search domain’. (Chomsky : ) b. Labelling Algorithm: The category created by Merge receives the label of the closest head. Labelling must be complete at the interfaces. (Rizzi : )

There are three potential sets in need of labels, namely {X, YP}, {XP, YP}, and {X, Y}. The first case is unproblematic—Chomsky says ‘trivial’—because the LA selects the head X in accordance with (a). The other two are ‘interesting’ because there is no unambiguous label assigned by the LA. Subjects in English exemplify {XP, YP} and the resolution to their labelling, IM, forces movement without having to rely on EPP features, a desired consequence. Thus, in (), a label cannot be found because both X and Y are as accessible to minimal search and therefore appropriate as labels.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen ?

()

v∗P (=YP)

DP (=XP) D (=X) v∗ (=Y)

...

Chomsky (: ) provides two solutions to labelling problems such as these: ‘There are, then, two ways in which [syntactic object] SO can be labeled: (A) modify SO so that there is only one visible head, or (B) X and Y are identical in a relevant respect, providing the same label, which can be taken as the label of the SO. These are the two cases that are prominently found’. Solution (A) applies in (): the DP must move, after which the v*P can be labelled. Other examples where the {XP, YP} set can be modified through movement of one of the maximal projections is the movement of a phrase out of a copula clause. Movement of one of the maximal projections, as in (), would result in a structure that can be labelled. According to Chomsky (: ), ‘[t]he intuitive idea is that the lower XP copy [in ()] is invisible to LA, since it is part of a discontinuous element, so therefore β will receive the label of YP’. ()

(Chomsky : )

XP copula {β XP, YP}

Although β receives a label in (), as does v*P in (), both result in other cases of {XP, YP}. Assuming the next merge will be a T in () and that the copula is in T in (), the result is the well-known issue that subjects in English face: they are drawn to Spec TP. Instead of positing EPP-features, labelling requirements in () and () force DP-movement. The result, given in (), is an unlabelled α because the subject internally merges to the TP resulting in {XP, YP}. ()

α[Tom T [Tom v* read a book]]

(adapted from Chomsky : )

In this case, solution (B) applies since the heads of the DP and TP share phi-features, and the set is successfully labelled as shown in ().

()

DP Tom

TP T

v∗P

Not much is said on the nature of . I assume it to be person and number features but it might as well be just number since subjects that move to the Specifier of TP agree more consistently in number with their verb than subjects that do not (van Gelderen ). This has been known since Greenberg (: ) as Universal

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



 (‘When number agreement between the noun and verb is suspended and the rule is based on order, the case is always one in which the verb precedes and the verb is in the singular’). I leave this issue aside. This second solution to the labelling problem can be exemplified by means of wh-constructions as well. If ‘the most prominent feature of the {XP, YP} set “is shared” ’, labelling is not a problem. It will be labelled using ‘the interrogative feature Q, a feature of C and the head of α’ in (a) (Chomsky : ). Sharing the Q-features between the PP and C in (a) has the result that the PP does not move further, as the ungrammatical (b) shows. ()

a. They wondered [α in which Texas city [ C [JFK was assassinated]]] b. *In which city did they wonder JFK was assassinated.

The fact that the wh-element cannot move further from (a) to (b) is called the ‘halting problem’ or ‘criterial freezing’ in Rizzi (; ), the basic intuition being that the wh-element included in the PP shares contradicting features: y/n for the embedded C and wh for the main clause. Once it has shared features in the embedded CP, it is frozen. Labelling resolutions also provide an account for the that-trace effect in (a): α cannot be labelled by the phase head C if who has moved. When the phase-head C deletes, as in (b), it transfers phasehood to T and who can remain in Spec TP until it is moved in the next phase. () a. * [γ Who do you v* [ε think [δ C that [α t T read the book]]]] b. [γ Who do you v* [ε think [δ C [α t T read the book]]]] (Chomsky : –) Apart from {XP, YP} being challenging, the set {X, Y} is problematic. Here Chomsky (: ) says that this applies when one of the heads is a root and the other a functional element determining its category. If roots do not count as labels, no problem arises. Chomsky (: ) mentions another case of head-movement, namely to T and v* and here ‘T [is] affixed to V. More generally, the conventional theory of head-raising seems to have the story backwards: the host should be affixed to the raised element’ so these are not cases of {X, Y} because ‘the affix is invisible to the labeling algorithm’. See Carstens, Hornstein, and Seely () as well. Labelling paradoxes can be resolved by having one of the XPs move, as in () and (), or by ignoring one label (the root), or by feature-sharing in () and (a). The first two solutions are worked out in Chomsky (), whereas the latter is the focus of Chomsky (). I now turn to some linguistic changes that may be accounted for by the requirements of the labelling algorithm. I will also discuss some changes that do not occur and why these may not be a problem for the labelling algorithm. Other work showing phrase to head reanalysis appears as Jäger (; b), Weiß (), Willis (), Bayer and Brandner (), Bácskai-Atkári and Dekány ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

. Subjects to T and objects to v* In this section, I first sketch a typical subject cycle and provide an explanation from a labelling perspective for the first part of the cycle. I then move to the object cycle and do the same. .. The subject cycle The typical stages of the subject cycle are given in () where English words are used for convenience. ()

a. b. c. d.

They (often) eat tomatoes. They’eat tomatoes. (Them) th’eat tomatoes. Them (often) eat tomatoes.

In (a), the pronoun is fully independent and need not be adjacent to the finite verb whereas, in (b), it is cliticized to the verb. If the pronoun is interpreted as agreement marker, this stage will be one of null subject (or pro-drop). In (c), the earlier independent pronoun is renewed by a new one that is ambiguous between being in topic or in subject position. If them is in topic position, the clitic could still count as the subject; if them is the subject, the clitic is now a marker on the verb. Stage (d) is the same as (a) with a renewed subject pronoun. Diagnostics to decide between topics and subjects include that the former need to be definite, whereas the latter can be quantifiers and indefinites. Once a quantifier appears in (c), it is a subject. Languages can thus be seen as being in different stages of the cycle; they can have just subject pronouns, just agreement, or both. If languages acquire agreement markers from erstwhile pronouns, one expects them to resemble these and that is indeed the case in many languages. This means the forms are available in the lexicon with different sets of features. According to Tauli (: ), the Basque verbal prefixes n-, g-, z- are identical to the pronouns ni ‘I’, gu ‘we’, and zu ‘you’. As early as the nineteenth century, Proto-Indo-European verbal endings -mi, si, -ti are considered to arise from first, second, and third person pronouns (e.g. Bopp ). Hale (: ) argues that in Pama-Nyungan inflectional markers are derived from independent pronouns: ‘the source of pronominal clitics in Walbiri is in fact independent pronouns’. Likewise, Mithun () claims that Iroquoian agreement markers derive from Proto-Iroquoian pronouns and Haugen () argues that Nahuatl agreement markers derive from earlier forms. Fuß () and van Gelderen () cite many additional examples. A language where we have evidence of all the stages in () is French. Old French has optional pronouns that need not be adjacent to the verb, as () shows for the second person singular tu ‘you’.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



() Si con tu meismes le preuves (Old French) If when you self it prove ‘If you prove it yourself ’ (http://romandelarose.org, Selden Supra , v) Foulet (: ) confirms that all personal pronouns can be separated from the verb in Old French. By the time of Modern (colloquial) French, je and tu obligatorily precede the finite verb, as the ungrammaticality of () shows.1 See Kayne (: –) for additional arguments. In addition, a frequent renewal in the form of moi (and toi) appears, as in ().2 () *Je/tu probablement SG/SG probably ‘I’ve probably read that.’

ai/as have

lu read

ça that

(Colloquial French)

() euh moi je trouve ce qui en souffre le plus . . . (Colloquial French) Eh me SG find that who of.it suffers the most ‘I think that the one that suffers the most is . . . ’ (Orléans Corpus). If we look at which pronouns are the first ones to grammaticalize into agreement markers, they are typically the first- and second-person singular ones. For instance, since Lambrecht (), it has been argued that French weak pronouns such as je ‘I’ and tu ‘you.SG’ are agreement markers on the verb and frequently doubled, as in (). What has also been known for a long time is that third-person subject pronouns are slower to gain agreement status. To be an agreement marker, a third-person subject pronoun would have to appear obligatorily and that is not the case in most varieties of French. Most indefinite subjects are not doubled except in a few varieties, as in () from Spoken Swiss French. () Si un: un Russe i va en france . . . (Swiss Spoken French) If a a Russian SG goes to France ‘If a Russian goes to France’ (Fonseca-Greber : ) The reason the third person is ‘slow’ is that there are more features to be shared, e.g. gender. Gender and possibly number are in fact deleted when the pronoun becomes the agreement marker, as in (), where i is marked for only third person (singular or plural) although les tomates are feminine plural. () Les tomates, i sont encore the tomatoes  are still ‘The tomatoes, they are still green.’

vertes green-P

(Spoken French) (Lambrecht : )

Because the status of the Modern French markers is debated, I use SG, SG, etc. to gloss them. In the Corpus d’entretiens spontanés, this doubling occurs in .% with first person ( out of  je/j’ ) and, in the Orléans Corpus, it occurs in % of the first-person singulars ( out of  je/j’ ). This corpus is part of the ELICOP Corpus. 1 2

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

As mentioned, Chomsky argues that DP and TP in () must share phi-features in order to be labelled. These features are not D or T but person and number and this, I will argue, makes a reanalysis possible. Let us look at the stages of the cycle. As (a) shows, a full phrase in subject position and T share person and number and the DP cannot be mistaken by the language learner for an agreement morpheme. Pronouns are ambiguous and, once they have lost definite and gender features, they can be reanalysed as T, either as a T with interpretable features, as in (b), or with uninterpretable ones in (c). Once the phi-features are uninterpretable, a new subject is necessary, as in (c), something that will be discussed in section .. () a.

les tomates [i–3, PL, F, definite]

b.

TP T [u-phi: 3, P]

v∗P

c.

TP T / [i–3]

les tomates ...

v∗P

T [u-phi]

TP v ∗P les tomates ... [i–3, PL, F, definite]

Thus, a very straightforward escape from the labelling paradox in () and () would be to have a subject that has the status of a head. Chomsky (: ) says that (pronoun) subjects cannot be heads because they would label the TP incorrectly, as D-headed, not T-headed. What I will argue is that the features of T in () are in fact agreement features and not T. This explains why pronouns change from phonologically fully independent phrases to agreement markers, as has happened in a number of languages, perhaps the most well-known case being French (see Lambrecht ; Roberts and Roussou ). A fully phrasal pronoun (that can be coordinated and modified) cannot be seen as having the same agreement features as T and can only be labelled as . A head (that has to be adjacent to a verb) can be seen by the child acquiring French (or English) as similar in features to T. When the features of a pronoun overlap with those of the agreeing T, they may disappear and a structure as in () may be the result. This structure can, of course, receive a label.

PhiP

()

Phi

v∗P

This account is very similar to accounts such as those of Roberts (a) and van Gelderen () who suggest the change from pronoun to agreement marker is due to a confusion as to whether the pronoun actually values the features of T or is itself in need of valuation. Taking Chomsky’s idea of feature-sharing, the preference for subjects that are heads with minimal features similarly makes sense. Let us look at how the two scenarios work in the most recent version of PoP.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



In Chomsky (), the T merges with the v*P and the subject moves internally (to Spec TP), after which C is merged. There are, of course, no labels, such as C or T, or branches, just features, but I have added the labels and branches for convenience. C has uninterpretable agreement features (u-phi) which it values with the subject before transferring the features to T. Once this happens, the {DP, TP} sequence can be labelled as after it arrives at the interface. In the scenario I argue for, given in figure ., the DP with its interpretable phi-features is reanalysed as head and valuation and labelling occur without a need to transfer features. a.

b.

CP C u-phi

TP = > DP i-phi

CP C u-phi T i-phi

T’

T vP valuation of u-F on C and transfer to T labelling due to feature transfer

TP vP

just valuation >

labelling is search

FIGURE . Reanalysis of the subject resulting in simpler labelling

After the pronoun is reanalysed as agreement, there is optional renewal in many languages, as () to () show. I come back to this in section .. .. The object cycle I will now turn to the object cycle, which was identified in e.g. Givón (). A typical object cycle is given in (), again a fictitious case for ease of exposition. Let us say that a language has a fully independent object pronoun, as in stage (a). Since this pronoun can be coordinated and modified and need not be close to a verb, it is a full phrase. A possible optional next stage is for speakers to analyse this object pronoun as a head, as in (b). This head cannot be coordinated or modified and is phonologically dependent on the verb. The next stage might be for the object to be reanalysed as an agreement marker. Once it has uninterpretable features, it could be renewed through an emphatic or some other form, as in (c). The last stage, as in (d), is similar to the first with the emphatic counting as the regular argument. ()

a. b. c. d.

I saw yesterday her (and him). I saw ’r (*her). I saw’r HER. I saw her.

French shows an object cycle, as argued in Bahtchevanova and van Gelderen (). Here, I will give some examples from other languages. In the Athabaskan family, there is a change from northern languages to southern ones in going from (b) to

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

(c). A representative of a northern language is Kaska and of a southern one Navajo (see further van Gelderen : –). In Kaska, the incorporated pronoun is in complementary distribution with another noun, as () shows, whereas it is obligatory in Navajo (), indicating it is agreement. () a. meganehtan me-ga-ne--h-tan SG-at-ASP-SG-CLF-look ‘He looks at her.’ b. ayudeni ganehtan girl ga-ne-o-h-tan girl at-ASP-SG-CLF-look ‘He looks at the girl(s).’ () a. ‘atoo’ yí-ní-dlaa’-ísh soup SG-SG-eat-Q ‘Did you eat the soup?’ b. yí-ní-dlaa’-ísh SG-SG-eat-Q, ‘Did you eat it?’

(Kaska)

(Jelinek ) (Navajo)

(Jelinek : )

Other languages are in various intermediate stages, e.g. Persian () has what looks like an affix but is still incompatible with a full object and varieties of Arabic restrict () to certain persons, while Kosrean () shows a doubling that makes the verbal marker into agreement. () pursed-am-ash asked-SG-SG ‘I asked him.’ () ʃuft-ik saw.SG-SG ‘I saw you.’

(Persian)

Ɂinti you

() Nga kihte-l sah I feed-SG him ‘I am feeding him.’

((some) Arabic)

(Kosraen) (Lee : )

How to account for this change in terms of a PoP account? Chomsky () assumes a Root Phrase, which is comparable to a VP in earlier work, whose head R merges with an object, as in (a) of figure .. Unlike the subject in () and (), an object need not move to the Spec of the RP because either R3 can label RP. However, v* can also 3 Chomsky (: ) says, ‘the question turns on whether R is analogous to “weak” T . . . If it is, then object-raising is obligatory’.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



transfer features to R and then the label is . The latter might result in a reanalysis of the DP object as agreement in (b) of figure . because of the ambiguity of the object pronoun. a.

v*P v* [u-phi]

> RP/

R

him [i-phi]

Minimal search or transfer and sharing

b.

v*P v* [u-phi]

RP R [i-phi] ‘m

Minimal search

FIGURE . Reanalysis of the object resulting in simpler labelling

The renewal in stage (d) is, as in the case of subject renewal, not due to the economy of minimal search. I will discuss this in section .. In connection to the subject cycle, I mentioned Greenberg’s Universal . This universal holds for objects as well (as noted by Kayne ). In English, there is no movement of the object to an OV position but in some languages (e.g. French object clitics, Hopi DPs), this movement results in full agreement on the verb. A labelling approach would have to say that transfer does not take place from v* to R. This too remains for further study. In this section, I have discussed two cases of pronouns being reanalysed as agreement markers. These can be seen as a preference for minimal search over feature-sharing.

. Demonstrative pronouns Demonstrative pronouns reduce features in a number of ways. They can lose deictic marking to become articles or complementizers and frequently reanalyse as copulas. Owing to space restrictions, I will not provide a lot of examples but focus on the mechanisms. I will start with the reanalysis of a demonstrative to article, a change that occurred in Romance (Harris , ), Uto-Aztecan, Salish, Egyptian, and many more (van Gelderen ; a). In one chronicle that was written during the eleventh and twelfth centuries around Peterborough in England, the switch is very obvious. In (), from , demonstratives are used regularly (e.g. se is masculine singular nominative) and no articles are but, in (), from  and from a different scribe, articles suddenly appear. ()

Đes feorðe dæges þæræfter wæs se king Heanri on Roueceastre. & se burch forbernde ælmæst. & se ærcebiscop Willelm halgede Sancti Andreas

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen mynstre & ða forsprecon biscop mid him. & se kyng Heanri ferde ouer sæ into Normandi on heruest. ‘On that fourth day thereafter, (that) King Henry was in Rochester, when that town was almost consumed by fire; and (that) Archbishop William consecrated St. Andrew’s monastery, and those aforesaid bishops with him. And (that) King Henry went over sea into Normandy in autumn.’ (Peterborough Chronicle, , Thorpe edition)

()

ðis gære for þe king Stephne ofer sæ to Normandi & ther wes underfangen forþi ðæt hi uuenden ðæt he sculde ben alsuic alse the eom wes. ‘This year, (the) King Stephen crossed the sea to go to Normandy and was received there because they thought he was like the uncle (i.e. his uncle).’ (Peterborough Chronicle, , Thorpe edition)

A possible reanalysis is given in figure ., where the label in (a) is seen as harder to arrive at than the one in (b), which is therefore preferred. Stage (a) is what we call concord and stage (b) involves agreement. a.

>

DP NP that/se book [i-3S] [i-3S] labelling {DP, NP} is due > to feature sharing

b.

DP NP D book the/þe [i-3S] [u-phi] Minimal search

FIGURE . Reanalysis of the demonstrative as article

Another change involving the neuter demonstrative þat is to complementizer, as from (a) to (b). It could be represented as in figure ., very similar to the change to article. () a. mid al þat þe þeron stant with all DEM REL thereon stands ‘with all that stands thereon’ (DOE, Will of Bishop Theodred, ) b. and suggeð feole þinges . . . þat næuere nes iwurðen and say many things REL never NEG.was happened ‘and say many things that never happened.’ (Layamon, Caligula –, Brook and Leslie edition)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection a.

>

DP CP þat [i-3S] C TP [i-loc] [u-phi] labelling {DP, CP} is due to sharing

b.



CP C þat [u-phi] >

TP ...

Minimal search

FIGURE . Reanalysis of the demonstrative as complementizer

Reanalysis of demonstratives as copulas is widely attested in Semitic, Egyptian, various creole languages, Iranian, Slavic, Tibeto-Burman, Swahili, Indonesian, Zoque, Passamaquoddy, Maya, and Chinese (van Gelderen b). An example from Egyptian is given in () where pw is a masculine singular proximal demonstrative in (a) reanalysed as (non-agreeing) copula in (b). () a. rmt pw man MSG.PROX ‘This man’ or ‘this is a man’. b. tmj-t pw jmn-t city-F be west-F ‘The West is a city.’

(Old Egyptian)

(Middle Egyptian) (Loprieno : )

In (), the derivation of a DP with its (copula-lacking) predicate is given. First, the DP and AP merge, as in (a), which results in a labelling paradox. This is resolved after the Pred head4 is merged in (b) and the DP moves internally. This, however, results in another unlabellable phrase. To resolve this, we would have to merge T to (c) and apply internal merge again to the DP, as in (d). () a. b. c. d.

{DP, AP} {Pred, {DP, AP}} {DP, {Pred, {DP, AP}}} {DP, {T, {DP, {Pred, {DP, AP}}}}}

External Merge Merge of copula Internal Merge of DP: unlabellable result Merge of T and Internal Merge of DP

Anti-locality has been defined as ‘movement that cannot be too local’ (Grohmann : ) and a reasonable domain of locality would be the phases PredP (or v*P) and CP. Stage (c) is therefore ruled out, just like movement of sister of V (the object) to specifier of v*P (subject) is. This is why a reanalysis as head takes place, as shown in figure .. Unlike the other cases, features are not involved. 4

Instead of Pred, I could have used little v as well; that choice is irrelevant to the analysis.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen a.

>

?P DP that

b.

Pred’ Pred

Predp Pred that

DP the chief

DP the chief

FIGURE . Reanalysis of demonstrative to copula

In short, in this section, we have seen three instances where a demonstrative is reanalysed in such a way as to enable labelling through simple search.

. Towards C and ASP In .., I discuss two instances where a phrase continues as the specifier of the CP and does not reanalyse as the head of C. These are interesting in that their label is feature-based but very stable, unlike the ones for subject discussed in .. In .., the change from adverb to aspectual affix is discussed. This occurs in many languages and will be exemplified through Modern English which shows an incipient stage. .. Towards the specifier of CP In the PoP-framework, wh-elements in the specifier of the CP share features with the C, as shown in () above, and escape the Labelling Paradox that way. The verb wonder in () requires a CP that has a Q-feature. I have marked this requirement by an uninterpretable feature on the verb in (). ()

I wonder whether he’ll do it.

() I

wonder [u-Q]

α[whether [i-Q]

[C [u-Q: Q]

[ he’ll do it]]].

Whether has interpretable Q features which value the C. The label of α is then . This label seems to be stable, unlike the features of section ., because whether is not being reanalysed as a head. () shows that whether is a specifier because extraction is not possible and, in (), if occupies the head C. () *Who do I wonder whether he saw who. ()

the Congressmen who come in in January and asking whether if one kind of affects the other. (COCA Spoken )

There is another complementizer that checks the Q-features, namely how, as in (). Here too, how remains in specifier position judging from sentences with the impossibility to extract in () and the presence of another C head, as in ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



()

The men will wonder how there’ll ever be enough lobsters around this island for seven more men to . . . (COCA  Fiction)

()

*What will the men wonder how there’ll ever be enough what.

()

by looking on, and watching how that these things might be done as well as others. (COHA )

Whether and how have been complementizers since the Old English period (van Gelderen a, a) and have not changed to heads. A reviewer brings up that embedded contexts, such as () and (), are more resistant to change. This cannot be the reason because both whether and how have been and are currently used as yes/no markers in root clauses (van Gelderen a; a). Even there, they show no evidence of consistently changing to C. The reason why the labelling of is stable could have to do with the relevance to the semantic interface (as the same reviewer points out). Thus, features are not relevant to the C-I interface but features are. .. Towards ASP In many languages, perfective aspect goes through a cycle in which an aspectual prefix weakens and is replaced by an adverb or adposition. For instance, Lehmann (: ) and Diessel (: ) argue that aspectual preverbs derive from relational adverbs and adverbial demonstratives, e.g. hin/her in German hinweisen ‘point out’, hinfahren ‘drive to’, and herbringen ‘bring over’. Miller (: –) provides instances of preposition incorporation in Ancient Greek and Latin and Booij and van Marle () bring together a number of studies on many languages that show a development from adverb to preverb. These cycles occur in Indo-European, but also in the Amazonian language Nadëb, as described by Weir (), in Athabaskan languages such as Dëne Sųłiné/ Chipewyan, as described by Li (), and in the Uto-Aztecan Tohono O’odham. Old and Middle English follow Germanic in having separable and inseparable prefixes on verbs to express aspectual nuances, as in (), as well as particles, as in (). () leofes mannes lic eall forswealg. dear man’s body all up.swallowed ‘He swallowed up the entire body.’

(Beowulf )

() & duste him dun riht to þer eorðe and threw him down right to the ground ‘and threw him right down to the ground’. (Elenbaas : , St. Margarete .) Elenbaas (: ch. ) argues that the particle in () is phrasal in nature and that that situation continues into Middle English because the particles are modified, coordinated, and preposable. There are particles that seem to combine with the verb, as in (), from (late) Old English on.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

()

til he aiauen up here castles ‘till they gave up their castles’

(Peterborough Chronicle , )

The two possibilities, () and (), continue in Modern English, as in (a) and (b). The adverb back is a phrase in (a) because it can be modified by a degree adverb, which it cannot in (b). ()

a. They received the book right back. b. They received (*right) back the book.

Trees for these are provided in (a, b) respectively. ()

a.

vP

b.

v

v

VP/?P VP/V’

vP

ASP back

AP

V received

ASPP

DP back the book

VP V received

DP the book

The VP/?P in (a) is problematic for the Labelling Algorithm because it consists of two merged XPs (by pair-merge). The reanalysis of the AP as a perfective ASP head, as in (b), is therefore expected. The two structures are shown in figure ., as in Chomsky (), namely with R rather than V. vP

a. v [u-phil]

b. ?P

RP

vP v [u-phil]

AP

back R DP received the book ?P is problematic

ASPP ASP back

RP

R received labelling via minimal search

DP the book

FIGURE . Reanalysis of an AP as ASP head

If () and (b) are more economical from a labelling point of view, why do we still have (a)? It may be that, in order to go from adverbial to functional category, the category has to be salient in the language which perfective is not in Modern English. Concluding section ., we have seen two cases that resist reanalysis from phrase to head and that may be due to the kinds of features that are shared in the case of Qfeatures. We have also seen a case where reanalysis is expected and occurs but where the change is not fast in Modern English.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection



. Towards argumenthood In this section, adjuncts are shown to be frequently incorporated as arguments, e.g. in (d) and (d), and this shows a preference of set-merge over pair-merge. I will first briefly discuss pair-merge and then show some examples of the change. Chomsky (: ; ) comes up with the term ‘pair-merge’ to describe adjunction. Merge comes in two kinds: ordered in pair-merge and unordered in setmerge. Chomsky (: –) argues that adjuncts are invisible to normal operations: elements that c-command the pair-merged continue to c-command the set-merged {b,c} but ignore the adjoined element. Pair-merge is invoked for adverbials because they are less integrated into a clause, evidenced by the fact that they are islands for extraction,5 as () shows (and argued in Huang ), and impervious to c-command (as shown in Lebeaux ). ()

*What did he leave the house [because she sang what].

Topicalized DPs are also islands, as () shows, although one might not think they are in adjoined positions since the work by Rizzi (). Rizzi assigns topicalized elements to designated specifier positions in the left periphery, i.e. the TopP, as the tree in () shows. ()

*Whose books do you think that [reviews of whose books] John never reads reviews of whose books? (adapted from Corver )

() CP DP Whose books C do

C’ TP DP you

Tʹ VP

T do DP you

Vʹ V think

ForceP Force that

TopP DP

reviews of whose books

Topʹ Top

TP ...

5 There are some non-finite adverbials that allow extraction (What did John arrive whistling), as Truswell () has shown (William Kruger p.c.).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

So, even if topics have designated positions, like adverbials, they are not as integrated in the sentence structure as subjects and objects are. From a diachronic perspective, adjunct clauses have become incorporated as arguments (Hale ; Kiparsky ) and this may show a preference of set-merge over pair-merge. In what follows, I look at the change of a topic from Spec TopP to a grammatical subject in Spec TP. Van Gelderen () argues that there is a principle that incorporates (innovative) topics and adverbials in the syntactic tree, as in (a), which could be modernized in PoP terms, as in (b). ()

Specifier Incorporation (SIP) a. When possible, be a specifier rather than an adjunct (van Gelderen : ), or b. When possible, set-merge is preferred over pair-merge (even if featuresharing with a sister XP is involved).

Givón () and others have talked about topics that are later reanalysed as subjects and call this a shift from the pragmatic to the syntactic. Earlier, we saw French pronoun subjects reanalysing as agreement affixes and a new topic/subject appearing. () Moi, j’ai me SG.have ‘I am cold.’

froid. cold

(Colloquial French)

Is there evidence of a renewal of the topic moi in () as subject? The answer is yes. Various researchers have commented on colloquial forms where the DP is an indefinite or quantifier, as in () to (), and these cannot be topics and are therefore seen as having been reanalysed, as well as () and (). () si un: un Russe i va en france (Swiss Spoken French) if a a Russian SGM goes to France ‘If a Russian goes to France.’ (Fonseca-Greber : , repeated from ()) () Chaque femme Every woman SGF ‘Every woman talks.’

elle parle talks

() Un Cadien ça travaillait pas. An Acadian  works not ‘An Acadien doesn’t work.’ The change from topic to subject is represented in figure ..

(Pied Noir French) (Roberge : ) (Acadian French) (Girard : )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Cyclical change and problems of projection b.

?P

a. DP



?P

... Top labelling mechanism not understood

DP

T’

T .. labelling after feature-sharing or pair-merge

FIGURE . From topic to subject

The last stage of the object cycle in () involves the renewal of the object through the incorporation of a topic that is right-dislocated. The actual reanalysis depends too much on which model of right-dislocation one follows and I will leave it aside. In this section, I have looked at the renewal of the subject through a topic and suggested a preference of set-merge over pair-merge, which is selected by the language learner once the data are ambiguous, as in ().

. Conclusion In this chapter, I have argued that the Labelling Algorithm motivates the reanalyses involved in the linguistic cycle. Minimalism, PoP included, constitutes a paradigm shift in attributing as little as possible to Universal Grammar and that includes labelling. Merge is forced by interface conditions that require labelling. In the change from specifier to head, we see one solution to the labelling problem: change from feature-sharing and agree to minimal search. The cases of specifier to head change that I have selected range from frequent to less frequent. I have done this on purpose to determine some other factors involved, e.g. the difference between and sharing, the one relevant to the C-I Interface and the other not.. Subject and object cycles are clear examples of the change from phrasal pronoun to agreement head. The changes involving demonstratives to articles and complementizers also involve phrase to head reanalysis that shows a preference for minimal search. In the case of reanalysis to copula, another constraint is at work, namely antilocality. The wh-elements whether and how are specifiers and show no reanalysis to head, which shows that this feature-sharing is stable. Finally, the change from AP to ASP head is frequent in a number of languages but not in English. The changes involving the renewal of the subject pronoun show the preference of set-merge over pair-merge. () summarizes the preferences. ()

minimal search > feature sharing > pair-merge

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elly van Gelderen

Acknowledgements Some of the explanations in this chapter overlap with those in van Gelderen (b) which was accepted and published online in . The current chapter includes much more data and a more updated version of the PoP framework. I would like to thank three anonymous reviewers, William Kruger, Johanna Wood, and the members of ASU’s syntax reading group.

Primary Sources Brook, G. and R. Leslie (). Layamon: Brut. Oxford: Oxford University Press [EETS ]. COCA. The Corpus of Contemporary American English, a -million-word corpus from between  and . https://corpus.byu.edu/coca/. COHA. The Corpus of Historical American English, a -million-word corpus from between  and . https://corpus.byu.edu/coha. Corpus d’entretiens spontanés, https://www.uwe.ac.uk/faculties/CAHE/ELC/Documents/ iclru/corpus.pdf. Dictionary of Old English (DOE) texts. http://www.doe.utoronto.ca. ELICOP Corpus, includes the Orléans, Tours, and Auvergne corpora. The Orléans Corpus (ESLO)contains , words of transcribed spoken French from  to ; https:// www.corpusfinder.ugent.be/corpus/.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

3 When morphological and syntactic change are not in sync Reassessing diachronic implications of the Rich Agreement Hypothesis ERIC FUß

. Introduction It is a well-known fact that morphological change often proceeds in a cyclic fashion. The loss of distinctive/expressive markers is regularly compensated by the grammaticalization of new functional morphology. However, the picture seems to be less clear with regard to syntactic change. While there are many reported instances of the loss of syntactic movement, the rise of displacement operations has rarely been observed, which suggests that the cycles of morphological and syntactic change are not always in sync and differ in their temporal dimensions. This state of affairs does not seem to sit comfortably with recent attempts to reinstate the Rich Agreement Hypothesis (RAH) in its strongest form (Koeneman and Zeijlstra ). This chapter discusses a set of historical developments where morphological change and syntactic change do not seem to go hand in hand, which is at odds with predictions of the strong RAH. In addition to well-known cases such as Danish, English, and French, which display a loss of verbal agreement marking without (or with delayed) loss of verb movement (or vice versa), the chapter presents data from Lithuanian and Cimbrian that suggest that a change from SOV to SVO may result in word order patterns that are not in line with the RAH (despite rich verbal agreement). The latter scenario challenges both strong and weak versions of the RAH, suggesting that approaches that link verb movement to other verb-related inflectional categories such as Tense/Aspect/Mood, possibly in combination with agreement, are more promising (cf. Biberauer and Roberts b on French and English; Holmberg and Roberts ; Haeberli and

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Eric Fuß . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

Ihsane  on English). However, it will be shown that this alternative also cannot account for the full range of relevant phenomena.

. The Rich Agreement Hypothesis: predictions for language change One of the most widely discussed correlations between morphology and syntax is the so-called ‘Rich Agreement Hypothesis’ (RAH), going back to work by Kosmeijer () and Platzack and Holmberg () on the Germanic VO languages. The RAH states that verb movement to INFL is linked to rich subject agreement morphology on the finite verb. The usual diagnostic test for verb movement to a clause-medial inflectional head is based on the linear position of the finite verb relative to elements such as negation and certain types of (low) adverbs, which are commonly taken to mark the left edge of the VP (e.g. manner and frequency adverbs): ()

sbj. Vfin / [VP tVfin . . . (V) . . . obj.]

Like other hypotheses that posit a connection between syntax and morphology, the RAH comes in two basic variants, a ‘strong’, biconditional version (cf. e.g. Vikner ; Rohrbacher ), and a ‘weak’, one-way implication (Platzack and Holmberg ; Roberts ; Koeneman ; Bobaljik ): ()

a. The ‘strong’ RAH: Rich subject agreement morphology ⟷ V-to-INFL b. The ‘weak’ RAH: Rich subject agreement morphology ⟶ V-to-INFL

According to the strong interpretation, a language has V-to-I movement if and only if it has rich verbal agreement morphology. In contrast, the weak RAH states that strong inflection triggers verb movement, but says nothing about the syntax of verbs in the absence of rich agreement. While the original debate on the validity and exact character of the RAH mainly focused on synchronic data, diachronic evidence also played a significant role from early on. Initially, the observation that the erosion of the formerly rich verbal agreement system preceded the loss of verb movement in the Germanic SVO languages was taken to support a strong interpretation of the RAH (cf. Vikner ; Rohrbacher ). However, when it became clear that the loss of agreement morphology and the loss of verb movement do not operate in sync and are often separated by a considerable temporal gap, diachronic evidence came to be used as an argument against the strong RAH (cf. e.g. Bobaljik ). Recently, Koeneman and Zeijlstra () have argued that the RAH should be reinstated in its strongest, biconditional form. They provide a new definition of what counts as rich agreement that links richness of verbal inflection to (alleged) universal properties of subject pronoun inventories (Greenberg’s  Universal ).1 Their basic idea is summarized in () (cf. Koeneman and Zeijlstra : ). 1 According to Greenberg (), the minimal number of distinctions expressed in subject pronoun inventories involves three persons and two numbers (but see Harbour  for a critical evaluation).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH ()



A language exhibits V-to-I movement iff the regular paradigm manifests featural distinctions that are at least as rich as those featural distinctions manifested in the smallest pronoun inventories universally possible.

The presence of rich agreement morphology is taken to signal to the learner the existence of a separate clause-medial functional head Arg (for Argument(hood)) that obligatorily triggers verb movement (while poor agreement is represented in terms of features on v): ()

. . . [ArgP [Arg’ Arg0 . . . A/N . . . [vP [v’ v0 [VP . . . ]]]]]

Koeneman and Zeijlstra then argue that counterexamples against a strong RAH do not stand up to closer scrutiny. For example, they propose that apparent cases of V-to-I in the absence of rich inflection can be explained away by assuming (i) that the finite verb moves to C (Faroese) or (ii) that adverbs occupy an exceptionally low position (Regional Northern Norwegian, ReNN). The opposite case—apparent absence of V-to-I in the face of rich agreement marking—is attributed to an exceptionally high position of adverbs and negation (Övdalian).2 To account for cases where there is a temporal gap between the loss of inflections and its (alleged) syntactic consequences, they suggest that after the erosion of agreement marking, conflicting word order patterns may be preserved as a result of syntactic reanalysis during L acquisition. This proposal is discussed in some more detail in section .. In what follows, I will first take a closer look at strong and weak theories of the morphology/syntax interface and their predictions for language change. .. The strong view Theories that assume a strong causal link between morphology and syntax predict that morphological and syntactic change should proceed more or less simultaneously. The loss of a morphological property M that is linked to a syntactic property S (possibly because both M and S are surface manifestations of an underlying parameter setting +P) should go hand in hand with the loss of S (and vice versa). Likewise, the rise of M is expected to be accompanied by the rise of S (and vice versa):3 ()

a. Loss of morphological property M ⇒ loss of a syntactic property S linked to M b. Rise of morphological property M ⇒ rise of S linked to M

2 As a result, the structural position of adverbs and negation must first be established on independent grounds before the placement of these elements can be used as a diagnostic for verb movement. More precisely, Koeneman and Zeijlstra () accept only vP-adjuncts (i.e. low adverbs and phrasal negation as e.g. in the Germanic family) as a reliable tool for the detection of v/V-to-I. While this move sharpens our understanding of what patterns can be used as diagnostic tests for verb movement, it leaves only very little room for falsifying the RAH. See Tvica () for the application of these criteria to a broad sample of genetically unrelated languages. 3 The strong view further predicts that it should not be possible to innovate a syntactic property S in the absence of a morphological property M to which S is causally linked (likewise, a language cannot lose S as long as M is present).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

The strong position necessarily leads to a conflict with regard to language acquisition. This becomes clear if we take a closer look at scenario (a): at the point when a learner fails to acquire M, M will still be part of the target grammar. As a result, syntactic patterns linked to M will continue to be part of the input the learner receives, leading to a situation where syntactic cues for a given property/parameter (say, verb movement) are in conflict with the morphology acquired by the learner: () Grammar G1 (+M, +S) ⇒ Output 1 (status of M unclear, but synt. cues linked to +S)

Grammar G2 acquired by the learner (–M, –S?) ⇒ Output 2 This raises the question of why the presence of relevant syntactic cues does not inhibit the loss of inflectional morphology linked to S, effectively blocking change. One way to solve this problem is to assume that syntactic opacity is a necessary additional ingredient for change to take place. In other words, the loss of M is only possible if S is ambiguous as well (this seems to be what is suggested by Koeneman and Zeijlstra, see section .. for additional discussion). A similar conflict occurs in cases of grammaticalization where the learner innovates M against the target grammar (e.g. rise of rich agreement via a reanalysis of subject clitics). As in scenario (), the learner will encounter syntactic patterns that do not match the newly acquired morphological property M: since the target grammar lacks both M and S, the innovation of rich inflections (i.e. +M) that are causally linked to a syntactic property S conflicts with the syntactic evidence the learner is presented with, which lacks S. Again, this raises the question(s) of how the learner copes with such a situation and why the presence of clear syntactic evidence does not prevent the rise of conflicting inflectional morphology: () Grammar G1 (–M, –S) ⇒ Output 1 (status of M unclear, but synt. cues linked to –S)

Grammar G2 acquired by the learner (+M, +S?) ⇒ Output 2

.. The weak view Theories assuming a weak causal link between morphology and syntax (such as the weak RAH, Bobaljik ) provide more wiggle room to account for temporal gaps between the loss of inflections and its effects on the syntactic system of a language. The loss of a morphological property M does not necessarily entail a loss of a

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



syntactic property S that is supposedly connected with M, as long as the relevant parameter value can be acquired on the basis of syntactic trigger evidence: ()

Loss of morphological property M ⇒

evidence for S linked to M weakened, but S may remain part of the grammar

Moreover, the loss of inflections does not lead to a conflict during L acquisition, since the absence of the relevant morphological cues does not contradict the syntactic evidence the learner is confronted with; the loss of inflections merely weakens the overall evidence that supports a certain abstract parameter value +P (linked to both M and S). It is also noteworthy that weak approaches in principle allow for the diachronic development of a syntactic property S despite the absence of M, a scenario that is ruled out under the strong view. However, it turns out that the diachronic predictions of weak theories do not differ from those of the strong view when it comes to the rise of M (e.g. via grammaticalization processes). Thus, given the hypothesized nature of the link between morphology and syntax as a one-way implication, the rise of a morphological property M which is causally linked to a syntactic property S is expected to go hand in hand with the development of S:4 ()

Rise of morphological property M ⇒ rise of S linked to M

In the same way, the presence of M blocks the loss of S linked to M. Thus, even under the weak view, the rise of M leads to a conflict: the word order patterns generated by the target grammar (which lacks both M and S) do not match morphological properties posited by the learner (e.g. innovated rich verbal agreement). Again, this raises the question of how learners deal with apparently contradictory pieces of evidence during L acquisition. As with the strong view, a potential solution is to assume that some degree of syntactic opacity/ambiguity is required for the rise of inflections to take place. For example, one might hypothesize that the rise of rich verbal agreement morphology is possible only if the finite verb occupies a surface position that is analysable as a ‘high’ position (i.e. Arg/I; see Alexiadou and Fanselow  and Fuß  for related considerations), possibly involving some form of syntactic reanalysis affecting the underlying structural position of the finite verb. 4 Of potential relevance are various Austroasiatic/Mon-Khmer SVO languages spoken by the Orang Asli, the indigenous people of the Malay Peninsula, where we can observe a(n ongoing) change in which proclitic subject pronouns have turned into bound person markers (see Benjamin  on Temiar; Diffloth  on Jah Hut; Kruspe  on Semelai; Burenhult  on Jahai). At first sight, it appears that despite the apparently rich agreement paradigm, the verb occupies a position to the right of negation, as in (i). However, more research is needed to determine the status of the person/number markers and the structural position of the verb in these languages.

(i) ˀe-loˀ tɔˀ ha-reɲrec sej mɛjmɛj naˀ? why  -eat meat excellent that ‘Why didn’t you eat that excellent meat?’

(Temiar; Benjamin : ff.)

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

. Reanalysis to the rescue? Focusing on scenarios resulting from the loss of inflections, Koeneman and Zeijlstra () propose that conflicts that arise by changes affecting only one part of the M-S link may be resolved via a reanalysis of problematic evidence, which may preserve a certain surface pattern despite contradictory morphological evidence. More precisely, they assume that syntactic patterns in which the finite verb occurs to the left of negation and (low) adverbs, which cannot any longer be parsed in terms of V-to-I/Arg movement (owing to the loss of rich inflections), may be reanalysed (i) as resulting from V-to-C movement (which they posit for Faroese, cf. (a)), or (ii) as involving an exceptionally low position of adverbs (which they assume e.g. for ReNN, cf. (b)):5 ()

[CP [TP [ArgP Vfin+Arg [NegP Neg [vP Adv . . . ]]]]] is reanalysed as either (a) or (b):

() a. [CP Vfin+C [TP [NegP Neg [vP Adv . . . ]]]] b. [CP [TP [vP Vfin+v [VP Adv . . . ]]]] The conflict caused by the loss of verbal agreement marking may also be resolved by grammaticalization processes that restore a rich system of verbal inflections via a reanalysis of subject clitics as agreement markers. This is what Koeneman and Zeijlstra assume for the history of French, in line with the widely held claim that preverbal subject clitics have turned into agreement markers when former clitic left dislocation structures were reanalysed as neutral active clauses (see section .. for details and references): () Moi, je travaille souvent me I work often ‘I often work at night.’

la the

nuit. night

a. [CP moi [TP je [ArgP travaille [vP souvent . . . ]]]] is reanalysed as: b. [CP [TP moi [ArgP je+travaille [vP souvent . . . ]]]] The idea that input data that cannot be parsed after the loss of inflections is subject to syntactic reanalysis provides a potential account for the persistence of V-Adv/Neg patterns after the erosion of rich agreement. However, this approach seems to insulate the strong RAH from problematic diachronic evidence: the assumption that a mismatch between syntactic and morphological cues is always and automatically resolved via reanalysis removes the hypothesis that there is a strong, biconditional relation between morphology and syntax from empirical evaluation by diachronic data. In what follows, I will take a brief look at three of the better-known historical scenarios 5

A related reanalysis may be invoked to account for problematic word order patterns (Subj-Adv/ Neg-V . . . ) that conflict with the status of verb movement after the development of rich verbal agreement morphology, giving rise to a high-attachment site for adverbs and negation.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



and discuss the application of Koeneman and Zeijlstra’s solution in terms of forced reanalysis to these test cases. .. Danish One of the better-known cases where there is a temporal gap between the loss of verbal agreement marking and the loss of verb movement comes from the history of Danish (cf. Vikner ; Sundquist , ; Heycock and Sundquist ). By  (Middle Danish), the erosion of verbal agreement morphology has led to an inflectional paradigm without person distinctions (cf. e.g. Sundquist : ). Still, verb movement to a position to the left of adverbs and negation continues to occur at a rate of over % till the end of the sixteenth century. Koeneman and Zeijlstra suggest that these facts can be accounted for if it is assumed that learners reanalysed the apparently conflicting evidence in a way that is compatible with a strong interpretation of the RAH (see () above). However, Heycock and Sundquist () argue convincingly that this approach does not yield an explanation for the persistence of verb movement in Early Modern Danish. Abstracting away from the possibility that rich agreement was reinstated via a reanalysis of subject pronouns (all historical stages of Danish lack relevant clitics), we are left with two options to capture V-Adv/Neg orders in the absence of rich agreement morphology: (i) a reanalysis of V-to-Arg/I movement as (embedded) V-to-C movement (cf. (a) above), or (ii) a reanalysis leading to a low, vP-internal position for adverbs (cf. (b) above). As is already shown by Sundquist (), (), the first scenario cannot be invoked to account for the Danish data since there is no difference in word order between embedded clauses where embedded V is possible (such as at-clauses embedded by bridge verbs) and embedded clauses where V-to-C/V is generally ruled out (such as relative clauses and indirect questions). Option (ii)—vP-internal attachment of adverbs—predicts an asymmetry between adverbs and negation concerning the position relative to the finite verb (as is found in e.g. ReNN): in a grammar without rich agreement morphology, surface V-Adv orders may result from a low position for adverbs; however, since negation must always occupy a position outside vP (for reasons of scope), a vP-internal finite verb will always appear to the right of negation. Heycock and Sundquist show that this expectation is not borne out and that no such difference between adverbs and negation can be observed in the history of Danish. These findings suggest that the temporal gap is real and that Danish continued to exhibit verb movement in the absence of rich agreement morphology for approximately  years. Thus, it seems that the Danish facts continue to challenge a strong interpretation of the RAH and cannot be circumvented by invoking some form of reanalysis, at least not of the type proposed by Koeneman and Zeijlstra (). In contrast, Danish does not pose a problem for the weak RAH: the loss of rich agreement merely weakens the evidence supporting a

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

grammar with verb movement, which may still be acquired solely based on the word order patterns encountered in the input. .. English At first sight, English seems to be just another instance of morphological change with delayed syntactic change. It is a well-known fact that in addition to have and be, certain frequently occuring lexical verbs such as know resisted do-support and continued to undergo verb movement well into the eighteenth and even nineteenth century, long after the loss of verbal agreement morphology (this fact is not considered by Ellegård ; but see Roberts  and Hale  for discussion): ()

Much that I encountered on the way contributed, I know not how, to heighten the vague sentiments of which I have already spoken. (Edgar Allan Poe, The Fall of the House of Usher, )

()

Now is the dramatic moment of fate, Watson, when you hear a step upon the stair which is walking into your life, and you know not whether for good or ill. (Arthur Conan Doyle, The Hound of the Baskervilles, )

However, recent work by Haeberli and Ihsane (, ) suggests that there might be an additional problem. Haeberli and Ihsane convincingly argue that the relevant developments are more complex than previously thought and that there is diachronic evidence from the history of English that actually challenges even a weak interpretation of the RAH. They show that the loss of verb movement was not a single change, but rather proceeded in two steps. First, English lost verb movement to a ‘high’ position to the left of adverbs, which Haeberli and Ihsane identify as T0. This change started in the middle of the fifteenth century and was completed by the middle of the sixteenth century. Subsequently, English began to lose verb movement to a ‘low’ inflectional head to the left of negation (Asp0, according to Haeberli and Ihsane). This later development started in the middle of the sixteenth century and reached its final stage in the second half of the eighteenth century. Crucially, these findings suggest that the loss of verb movement is not directly linked to the loss of verbal agreement morphology, at least not in Koeneman and Zeijlstra’s sense: owing to the fact that the sg ending ‑(e)st continues to be robustly used in connection with thou, the verbal paradigm counts as ‘rich’ until the seventeenth century. Thus it is fairly clear that the early change—that is, loss of verb movement to the left of adverbs, which began in the fifteenth century—cannot be attributed to the loss of agreement endings. The later change that eventually led to the general loss of verb movement seems to partially overlap with the loss of sg agreement. More precisely, it seems that short verb movement began to erode when the replacement of thou with the polite (originally plural) form you was already well under way (cf. Busse  for details). However, Haeberli and Ihsane (: ) point to the fact that the familiar

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



form thou continued to be part of the child’s input until the eighteenth century, since it ‘is typically used by the dominant member in socially asymmetrical relations’. If the conclusion that the loss of verb movement cannot be attributed to the loss of verbal agreement morphology turns out to be correct, the results of Haeberli and Ihsane are problematic for both the strong and the weak version of the RAH, since we face a situation where the loss of verb movement—a syntactic change—precedes the loss of rich agreement. .. French Another widely discussed test case comes from the history of French. As is well known, French differs from English in that the finite verb moves to a position to the left of low adverbs and negation. Pollock () speculates that this word order difference is linked to properties of verbal inflection, which he considers to be rich in French, but poor in English. However, the relevant markers are distinct in the written language only. In colloquial spoken French, all forms save the pl have fallen together, so that the resulting paradigm is actually as poor as the inflectional system found in English (in both languages, there is only a single distinctive agreement marker, sg ‑s in English, and pl ‑ez [eː] in French). So it appears that spoken Modern (Standard) French exhibits weak verbal agreement morphology, but still requires verb movement to the left of adverbs and negation—similar to Danish, the loss of inflections had no direct influence on the availability of V-to-I.6 From the perspective of the weak RAH, this does not present a problem, since the weak interpretation allows for a scenario in which verb movement may be acquired in the absence of relevant morphological cues. However, no such leeway is available under a strong interpretation of the RAH. As already mentioned above, Koeneman and Zeijlstra propose that in the history of French, the conflict between the loss of verbal agreement suffixes and the persistence of V-Adv/Neg patterns was resolved by another type of reanalysis in which subject clitics turned into prefixal agreement markers, eventually leading to a system that complies with the strong RAH. If this reasoning turned out to be correct, (spoken) French would represent a case where a cyclic morphological change (loss+rise of inflections) is triggered by the need that syntactic and morphological aspects of a given parameter setting be fully in sync. Koeneman and Zeijlstra’s proposal is consistent with the view widely held in the literature that the subject ‘clitics’ of colloquial/non-standard varieties of French have 6 For the relevant historical developments cf. e.g. Wartburg (); Ashby (); Harris (); Roberts (); Vance (); Roberts (b); King et al. (); Simonenko et al. (forthcoming). The loss of verbal inflections began in the Middle French period (fourteenth to sixteenth centuries). However, as pointed out by an anonymous reviewer, the decisive change was (similar to the change from ‘thou’ to ‘you’ in English) the replacement of the pl pronoun nous by on, which triggers sg agreement (see also Koeneman and Zeijlstra : ). This is a rather recent change, which began slowly in the seventeenth century and reached completion in the twentieth century (cf. King et al. ).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

lost their pronominal character and are more appropriately analysed as some form of agreement marking.7 Note that under this account, it would seem that colloquial French ceases to be a problem for the strong RAH, since it exhibits a fully distinctive set of preverbal agreement markers (je, tu, il/elle etc.) and is thus expected to show obligatory verb movement. However, closer inspection reveals that the problem cannot be dismissed that easily. An observation that undermines Koeneman and Zeijlstra’s account is that whereas verb movement is compulsory in all varieties of French, less uniformity is observed with respect to the properties of the preverbal person/number markers. De Cat () argues that at least in some of the relevant varieties, the alleged agreement markers are still pronominal clitics, pointing to the fact that the position of the elements is not fixed (the clitic follows the verb in yes/no questions), cf. (), and that other preverbal clitics (object clitics, elements such as en, y and the negation particle ne) may intervene between the alleged agreement markers and the verb, as shown in (). If the person markers were prefixes, the intervening elements would have to be analysed as affixes as well, which does not seem to be warranted (following Zwicky and Pullum , clitics can attach to hosts+affixes, but affixes cannot attach to hosts+clitics).8 () Peut-il avoir une petite may-he have a little ‘Can he have a little bite?’

bouchée? mouthful (De Cat : )

() a. Je la lui donnerai. I it to.him give: ‘I’ll give it to him.’ b. Je ne t’ en I  to.you of.it ‘I don’t begrudge you.’ c. On y va? we there goes ‘Shall we go?’

veux want

pas. 

(De Cat : )

Moreover, even if we were to accept Koeneman and Zeijlstra’s account for colloquial varieties of French, Standard French continues to be a problem, since it lacks

7 For details cf. Lambrecht (); Roberge (); Auger (, a,b, ); Zribi-Hertz (); Fonseca-Greber (); Gerlach (); Fonseca-Greber and Waugh (); but see Roberts (b) for qualifying remarks and De Cat () for an opposing view. 8 Even if the subject clitics are analysed as clitic agreement markers (as has been suggested by one anonymous reviewer), the data continues to be a problem for Koeneman and Zeijlstra’s analysis: under standard assumptions, clitics do not trigger movement, that is, they move to attach to their host, and not the other way around.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



proclitic/prefixal agreement markers and still exhibits verb movement. So we may conclude that at least those varieties of French in which the preverbal person markers cannot be analysed as agreement prefixes (and thus have not completed the grammaticalization cycle) continue to challenge the strong RAH.

. Verb movement and the change from SOV to SVO In this section I will show that word order patterns that apparently conflict with the RAH may come into existence as a side effect of a change in basic word order from SOV to SVO. More precisely, it appears that after the transition from SOV to SVO order, languages may preserve the preverbal placement of adverbs and negation that is characteristic of SOV languages, giving rise to S‑Adv/Neg‑V‑O orders despite the presence of rich verbal inflections. .. Lithuanian Morphologically, Lithuanian is one of the most conservative (European) IndoEuropean languages. Its system of verbal inflections features three conjugations (marked by thematic vowels ‑a, ‑i, ‑o), distinctive person and number agreement, four different synthetic tenses (present, past, frequentative past (‘used to V’), future), four moods, and a rich system of participles (thirteen different forms) conveying a variety of aspectual differences (cf. e.g. Ambrazas ), cf. table . for the paradigm of dìrbti ‘to work’. TABLE . Tense and agreement marking in Lithuanian (first conjugation, Ambrazas ) dìrbti ‘to work’

Present

Past

Past freq.

Future

sg sg sg pl pl pl

dìrb-u dìrb-i dìrb-a dìrb-a-me dìrb-a-te dìrb-a

dìrb-a-u dìrb-a-i dìrb-o dìrb-o-me dìrb-o-te dìrb-o

dìrb-dav-a-u dìrb-dav-a-i dìrb-dav-o dìrb-dav-o-me dìrb-dav-o-te dìrb-dav-o

dìrb-s-i-u dìrb-s-i dìrb-s dìrb-s-i-me dìrb-s-i-te dìrb-s

Despite its conservative morphology, Lithuanian has been undergoing a major word order change in its recent recorded history (cf. e.g. Reklaitis ; Hock : ). At least until the early twentieth century, Lithuanian was commonly described as a basic SOV language (with a number of additional word order options linked to information-structural distinctions, emphasis, etc.), cf. e.g. Schwentner (: ). In contrast, standard descriptive works on present-day Lithuanian (cf. e.g. Ambrazas

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

: ch. ) unequivocally identify SVO as the basic word order (again with multiple additional orders dependent on the information-structural status of the constituents of the clause): () Vaikai suvalgė [visus obuolius]. the.children: eat:. all apples: ‘The children have eaten all the apples.’ Given the rich verbal inflection of Lithuanian, any theory that links verb movement to properties of verbal agreement—be it agreement, Tense/Aspect/Mood (Biberauer and Roberts b; Haeberli and Ihsane ), or the presence of separate tense and agreement morphemes (Bobaljik and Thráinsson )—should lead us to expect the verb to occur to the left of negation and adverbs after the change from basic SOV to SVO. As will become clear shortly, this expectation is not borne out by the facts (see also Korostenskaja ; Korostienskienė ). At first sight, negation does not seem to be a good indicator of verb position in Lithuanian: the negative clitic ne attaches to the left of the verb and accompanies verb movement, e.g. to clause-initial position in inversion contexts. However, to intensify negation, the particle nẽ/̇ neĩ can be added, which is usually placed before the verb (cf. Ambrazas : ): () Mókytoja nė/nei ne-pàžvelgė į są̃siuvinį. the.teacher  -glanced at the.copybook ‘The teacher did not even glance at the copybook.’ If the additional Neg-element in cases like () signals the position of NegP, then this might be taken to suggest that the verb does not move further than Neg0 in Lithuanian. Note furthermore that negative adverbs such as niekadà ‘never’ also precede the verb in the unmarked order (Ambrazas : ): () Táu niẽkas niekadà ne-dãrė jokių̃ you.. nobody never -make any ‘No one has ever reproached you for anything.’

príekaištų. reproaches

While the evidence from the placement of negators is not fully decisive, it is certainly more in line with an analysis that takes the verb to occupy a lower position, either internal to or directly above VP (cf. e.g. Korostenskaja ). In any case, the placement of reinforcing negative particles and adverbs suggests that subject and finite verb are not hosted by the same functional projection, that is, they are not in a specifier-head relation. This conclusion is further supported by the relative order of the finite verb and certain types of (low) adverbs. Similar to the placement of arguments, the position of adverbs is quite variable in Lithuanian (cf. e.g. Dičpinigaitytė and Korostenskienė ). However, in the unmarked case, the finite verb typically follows low adverbs (specifying manner

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



and frequency) that are commonly used as diagnostics for the structural position of the verb (cf. Mathiassen : ; Ambrazas : ; Korostenskaja ): () onas dažnai bučiuoja Jonas often kiss:. ‘Jonas often kisses Marija.’

Mariją. Marija (Korostenskaja : )

Korostenskaja () concludes from this fact that Lithuanian, like English, lacks systematic V-to-I movement. However, as pointed out by one anonymous referee, the Lithuanian data might also be subsumed under the analysis proposed by Koeneman and Zeijlstra (: –) for Russian. Koeneman and Zeijlstra argue that examples like () should not be used as a diagnostic for V-to-I movement, since manner adverbs are not adjoined to vP, but rather occupy a higher position in Russian. () My často čitali pravila. we often .read rules ‘We read the rules often.’

(Russian) (Koeneman and Zeijlstra : )

In support of their proposal, Koeneman and Zeijlstra (: ) discuss the placement of low adverbs in connection with negation, which is expressed by the clitic ne that attaches to the left of the verb. In clauses that contain both negation and an adverb, the latter must either precede the negated verb, or occur clause-finally; the adverb cannot intervene between the verb and the object:9 () a. ??/*

b.

Ty ne čitala často pravila. you  .read often rules ‘You haven’t read the rules often.’ Ty (často) ne čitala pravila you often  .read rules ‘You haven’t often read the rules.’

(často). often

Koeneman and Zeijlstra assume that the verb must undergo movement to the head of NegP to attach to the negator ne, which for reasons of scope must c-command the verbal projection. If the adverb were adjoined to vP, one would expect it to intervene between the verb and the object. From the fact that this ordering is ruled out, Koeneman and Zeijlstra conclude that the adverb must be adjoined to a higher position (to the left of NegP) and cannot be used as a diagnostic for V-to-I movement.

9

In Koeneman and Zeijlstra’s version of (b), the adverb occurs to the left of the subject. A native speaker of Russian pointed out to me that this order is very marked and that it is much more natural to put the adverb in between the subject and the finite verb. I changed the example accordingly.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

Orders similar to (b) are also found in Lithuanian, as shown in (). Accordingly, we cannot rule out the possibility that Lithuanian might be analysed in a way similar to Russian, featuring high attachment of adverbs and a clitic negator, that both do not tell us much about the exact position of the finite verb. () Jonas visiškai ne-si-tvarko Jonas completely --manage:. ‘Jonas does not clean his room at all.’

kambario (Lithuanian) room (Korostenskaja : )

Still, there are a couple of facts that cast some doubt on whether Lithuanian should be analysed on a par with languages that clearly exhibit V-to-I movement. First of all, notice that in contrast to languages of this latter type, subject and verb often fail to be adjacent in Lithuanian (cf. e.g. (), (), and ()). This suggests that the subject and the finite verb do not occupy specifier and head of one and the same functional projection. In other words, even if the finite verb is taken to move to Asp or Neg, this does not entail movement to Arg/I. This conclusion is supported by the observation that as in English, there is an asymmetry between modals/ auxiliaries and lexical verbs in Lithuanian (cf. e.g. Korostenskaja : ). While finite lexical verbs preferably occur to the right of adverbs, finite modals and auxiliaries may move to a head position further to the left. This contrast suggests that only the latter undergo movement to Arg/I, while finite lexical verbs occupy a lower position: () Marketingo skyriaus vadovas gali visiškai marketing:. department:. head: can:. completely pa-si-titikė-ti savo pavaldiniais. (Lithuanian) --believe- his subordinates:. ‘The head of the marketing department can completely rely on his subordinates.’ (Korostenskaja : ) To be sure, these considerations are not conclusive and do not rule out the possibility that Lithuanian complies with the RAH. But they do show that more work is needed to pinpoint the structural position of the finite verb in Lithuanian. However, a detailed analysis of the clause structure of Lithuanian far exceeds the scope of this chapter, and I will therefore have to leave it for future research. In any case, Lithuanian shows how in a SOV language with rich verbal agreement, word order change may lead to patterns that apparently violate the RAH. More precisely, the change in question merely affected the position of arguments, while adverbs/adjuncts continued to appear in preverbal position, a residue of the former SOV stage. Arguably, this can be traced to an acquisition scenario where S‑Adv‑V patterns continue to be robustly attested in the input even after the change from OV to VO has taken place. So it seems that rich verbal agreement does not necessarily have the expected effects on verb placement if an independent change has affected the verbal

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



syntax of the language. Should it turn out that Lithuanian lacks V-to-I, it would not only be a counterexample to both the strong and the weak version of the RAH, but it would also defy any attempt to establish a systematic link between verb movement and rich verbal inflection. In the remainder of this section, I take a look at a related development in Cimbrian, a Germanic minority language spoken in Northern Italy. .. Cimbrian Cimbrian is a cover term for a set of German dialects spoken in an area between Trento, Verona, and Vicenza. There used to be at least three different major varieties of Cimbrian, which were originally spoken in two groups of municipalities called the seven communities (main village: Roana) and the thirteen communities (main village: Giazza), and the village of Lusern. Today, the active use of Cimbrian is more or less confined to Lusern (around  speakers). The Cimbrian dialects are of particular interest for our purposes, since they exhibit various stages of the transition from OV to VO order, probably due to contact with the surrounding Romance VO varieties (cf. Grewendorf and Poletto  and Kolmer  for details). As a result, it is possible to establish whether the finite verb moves to Arg/I (in contrast to German OV varieties). The basic VO syntax of Cimbrian is illustrated by the following examples. () shows that the object typically follows the verb in both main and embedded clauses; moreover, finite verbs always precede non-finite verbs, and governing non-finite verbs precede governed non-finite verbs in the verbal complex (cf. e.g. Schweizer /; Grewendorf and Poletto ; Bidese ; Kolmer ).10 () a. [ka herbest] han=sa gelest [di in fall have=they harvested the ‘In the fall, they harvested the potatoes.’

patatn]. (Lusern) potatoes (Kolmer : ff.)

b. S Beibe bo da-r-en hat geet [a Liber] (Lusern) the woman  there=he=her has given a book ‘the woman who he has given a book’ (Grewendorf and Poletto : ) As shown in table ., Cimbrian has by and large preserved the German paradigm of verbal agreement markers, which counts as rich according to Koeneman and Zeijlstra’s definition.

10

Note that Cimbrian exhibits some residual SOV effects. For example, certain elements such as object/ reflexive pronouns and indefinites/quantifiers may appear to the left of the non-finite verb, giving rise to a reduced sentence bracket (cf. Grewendorf and Poletto ; Kolmer ).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß TABLE . Verbal person/number markers in Cimbrian (Schweizer /: )

sg sg sg pl pl pl

LUSERN

GIAZZA

ROANA

(-e) -est -et -n -et -n

-e (a) -ast -at -an -at -an

-e (a) -est -et -en -et -ent

Hence we expect Cimbrian to exhibit verb movement to a position to the left of (low) adverbs and negation. However, the predictions of the RAH cannot be tested in main clauses, since it is generally agreed that the finite verb occupies a position in the left periphery in root clauses (cf. e.g. Bidese ; Grewendorf and Poletto ; Kolmer ; Bidese and Tomaselli ; Bidese, Padovan and Tomaselli forthcoming), a residue of the Germanic V constraint. As shown in (a) and (), subject–verb inversion is confined to clitics, while nominal subjects precede the finite verb in root clauses: () [dar mon] [ka herbest] is kent the man in fall is come ‘The man came home in autumn.’

humman. home

(Lusern) (Kolmer : ff.)

Accordingly, the presence/absence of V-to-I movement can be tested only in embedded clauses in which the verb does not undergo movement to the C-domain. In the variety spoken in Lusern, relevant differences between root and embedded clauses include the relative order of the finite verb and negation.11 In root clauses, both finite main verbs and auxiliaries precede the negation net (much as in other Germanic varieties, nèt is taken to occupy a position directly to the left of vP/VP): () a. I gea nèt ka miss. I go not to mass ‘I don’t go to mass.’ b. I hån nèt giböllt gian ka I have not wanted go to ‘I didn’t want to go to Trent.’

Tria. Trent

(Lusern) (Panieri et al. : )

11 Asymmetries between root and embedded clauses can be observed with relative clauses (introduced by bo) and embedded clauses introduced by the complementizers az ‘that, if ’, bal ‘when/if ’, benn ‘if ’, intånto az ‘while’, ånka az ‘even if ’, dopo az ‘after’, fin az ‘until’, and ena az ‘unless’. Other types of embedded clauses, in particular those introduced by the borrowed complementizer ke ‘that’, exhibit the same word order properties as root clauses (cf. Panieri et al. ; Bidese ; Grewendorf and Poletto ; Kolmer ).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



In contrast, the finite verb appears to the right of nèt in az/bo-type embedded clauses:12 () a. às=to nèt geast . . . if=you not go ‘if you don’t go . . . ’

(Lusern) (Panieri et al. : )

b. un bo=bar biar o net han and who=we we too not have ‘ . . . and who we didn’t know, too’

gewisst known

(Lusern) (Kolmer : )

In the literature, these facts are taken to suggest that the finite verb does not move to INFL in embedded clauses of the az/bo type, but rather occupies a vP/VP-internal position (cf. Bidese ; Grewendorf and Poletto ; Bidese and Tomaselli forthcoming; Bidese, Padovan and Tomaselli forthcoming). The same asymmetry is found with a class of (short) adverbs that may occur in preverbal position. Again, the finite verb precedes these adverbs in main clauses, while it follows them in az/botype embedded clauses:13 () a. Dar he ‘He

hat has has

za already already

gerüaft. called called.’

b. az ar za vort is that he already away is ‘that he already went away’

gont gone

(Lusern) (Grewendorf and Poletto : )

So it seems that in certain embedded clauses of Cimbrian, the finite verb occupies a position to the right of negation and certain adverbs, apparently in violation of both the strong and the weak RAH. As in Lithuanian, these word order patterns are probably the outcome of an independent syntactic change—the rise of basic SVO— after which the language has preserved the preverbal position of adverbs and negation typical of an SOV grammar.

. Concluding summary In this chapter, I have taken a look at a couple of diachronic scenarios where morphological change and syntactic change do not go hand in hand, which create various problems for the idea that rich verbal agreement triggers verb movement:

12 Note that finite auxiliaries and modal verbs can optionally precede the negation in embedded clauses (Panieri et al. : ; Grewendorf and Poletto : ). 13 Unfortunately, other adverbs cannot be used as a diagnostic for verb movement since they preferably occur clause-initially or clause-finally (Panieri et al. : ff.).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Eric Fuß

. Morphological change without or with delayed syntactic change (Danish, French) . Syntactic change without or with delayed morphological change (English) . Lack of verb movement after an independent syntactic change from OV to VO (despite rich Agr: Lithuanian, Cimbrian) I have then argued that Koeneman and Zeijlstra’s () attempt to maintain the strong RAH by accommodating the first set of problematic data in terms of syntactic reanalysis runs into difficulties. While a weak interpretation of the RAH provides enough leeway in the diachronic transition from one grammar to another to account for the Danish and French data, it fares no better than the strong RAH when it comes to the second and third scenarios. This has to do with the fact that the diachronic implications of strong and weak theories are identical in cases where the morphology is rich. Therefore, cases like English (section ..) and Lithuanian/Cimbrian (section .) seem to contradict any version of the RAH. As an alternative, the absence/ presence of verb movement in these cases might be linked to other verb-related inflectional categories such as Tense/Aspect/Mood, possibly in combination with agreement (cf. e.g. Biberauer and Roberts b on French and English; Holmberg and Roberts ; Haeberli and Ihsane  on English). For example, an approach that links verb movement to the richness of (synthetic) tense and mood distinctions makes the correct predictions for the differences between English/Mainland Scandinavian and (spoken) French: ()

Spoken French: parle (present indicative/subjunctive), parlerai (future), parlerais (conditional), parlais (imperfect)

()

a. English: speak (present), spoke (past) b. Swedish: snakker (present), snakket (past)

Furthermore, it appears that earlier stages of English/Mainland Scandinavian exhibit a productive indicative/subjunctive distinction, which might have contributed to the richness of verbal inflections. From this perspective, it is perhaps possible to attribute the loss of verb movement to the loss of verbal inflections other than agreement (cf. e.g. Haeberli and Ihsane  on English). While this approach also captures the Cimbrian data (which exhibits the Germanic type of tense system), cases like Lithuanian seem to remain problematic for any attempt to construe a morphological trigger for verb movement, which suggests that there must be triggers of syntactic change independent of morphology (e.g. syntactic opacity leading to reanalysis of both syntax and morphology, cf. Anderson ; Fischer ). This raises the more general question of how grammatical theory should deal with the fact that it still seems to be clear that there is some trade-off relation between syntax and morphology—languages with rich inflectional morphology often exhibit syntactic properties not shared by languages with poor inflectional morphology; over time, a change in one component often leads to changes in the other etc. However, if we

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Diachronic implications of the RAH



want to avoid the pitfalls of earlier theories, it might prove to be necessary to abandon the idea that there is a direct connection between morphology and syntax. As an alternative, we might assume that the observed correlations may also be the reflex of historical developments, that is, recurrent pathways of language change (for related considerations cf. e.g. Alexiadou and Fanselow ; McWhorter : ch. ). For example, morphological change may reduce the evidence for (or practical functionality of) a certain kind of syntactic system (i.e. a combination of parameter settings), which in the long run may lead to a bias against the acquisition of certain syntactic properties—either because adult speakers tend to avoid syntactic strings that in principle express the older setting but give rise to syntactic ambiguity (leading to so-called word order freezing, e.g. the avoidance of scrambling of arguments that lack distinctive case morphology, cf. Lee ; Flack ), or because the loss of inflections opens up the possibility of a new grammar that parses the input more successfully than the older competitor and gradually spreads in a speaker community (cf. e.g. Heycock and Wallenberg  on the loss of verb movement in Scandinavian; Simonenko et al. forthcoming for the connection between verbal agreement marking and pro-drop in the history of French). In addition, the loss of a certain coding option may exert a functional pressure that over time may lead to the emergence of alternative coding options (e.g. fixed SVO word order instead of case marking). Admittedly, historical explanations of this kind are at first sight less interesting from a theoretical point of view; on the other hand, they seem to be in line with recent ideas that aim at minimizing the overall complexity of UG by relegating certain apparent universal properties of language to so-called third factors (cf. e.g. Chomsky ; Hornstein ), which may also include (universal) pathways of language change (possibly determined by ‘developmental constraints . . . including principles of efficient computation’, Chomsky : ).

Acknowledgements I would like to thank audiences in Constance (DGfS ), Hamburg, Ghent (DiGS ), Zurich (Romanistentag ), and Trento for helpful suggestions. The chapter has much benefited from discussions with Ermenegildo Bidese, Patrick Brandt, Thórhallur Eythórsson, Fabian Heck, Achim Stein, Alessandra Tomaselli, and Carola Trips. In addition, I am grateful to three anonymous reviewers for insightful comments, which considerably improved the chapter. All remaining shortcomings are, of course, entirely my own.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

4 The clitic doubling parameter Development and distribution of a cyclic change SUSANN FISCHER, MARIO NAVARRO, A N D JO R G E V E G A V I L A N O V A

. Introduction As its name suggests, clitic doubling (CLD) involves the doubling of a verbal argument by means of a clitic pronoun inside the same propositional structure.1 In Peninsular Spanish and Standard Catalan the clitic obligatorily doubles the full pronoun (), while it optionally doubles the indirect object (). ()

El vam veure him saw ‘We saw him.’

()

(Le) doy el libro a Pedro.2 him give the book to Pedro ‘I give the book to Pedro.’

a ell. to him

(Cat)

(Sp)

1

We assume that CLD is different from clitic left and right dislocation. Left and right dislocation necessarily involves prosodic cues (cf. Jaeggli ) whereas this need not be the case in CLD structures. For a different view on this issue, see Aoun (). Furthermore, clitic left and right dislocation allows wide and narrow scope (Torregrossa ) whereas CLD only allows wide scope. 2 We are aware of the fact that what we call optionality concerning dative CLD is not real optionality, since some indirect objects are more easily doubled than others–depending, for example, on the thematic role (Dufter and Stark ) or on information structure; and with regard to psych-verbs it seems to depend on whether a stative or causative reading is intended (cf. Elvira ). We have piloted a questionnaire for Standard Spanish speakers and bilingual Spanish-Catalan speakers testing dative CLD and there seems to be even more variety than so far acknowledged. Our results are not publishable yet. We noticed some effects that cannot be explained by thematic roles or information structure only, but also seem to depend on issues like bilingualism. Therefore, what we mean by optionality is the fact that we find doubled and undoubled datives depending on different factors, which have yet to be identified. Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Susann Fischer, Mario Navarro, and Jorge Vega Vilanova . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



We examine three types of data in relation to CLD in this chapter: direct object or accusative doubling, indirect object or dative doubling, and dative doubling of the experiencer of psych-verbs. However, we believe that these different types can be seen as one unitary phenomenon, since all three types of CLD obey similar restrictions along the accessibility hierarchy and it is possible to state the following implications cross-linguistically: (i) no language allows doubling of accusative objects without also doubling dative objects; (ii) no language allows doubling of dative nouns without also doubling full pronouns (cf. Fischer and Rinke : ). As is well known (cf. Fischer and Rinke ; Anagnostopoulou  and references therein), CLD is a highly variable phenomenon within the Romance languages. We find enormous differences within one language and across the different Romance languages which cannot be explained only by the different paradigmatic stages the Romance languages are in. CLD has always been a challenge for linguistic theorizing. For one thing, it is hard to account for the variability. In addition, it seems as if the clitic shares one semantic role, one case, and one syntactic function with the full nominal object. Over the years, many factors have been discussed and held responsible for the occurrence and distribution of CLD. We find grammatical factors such as pronominal vs non-pronominal, accusative vs dative, and the occurrence vs non-occurrence of Differential Object Marking (DOM), semantic factors such as animacy, and pragmatic factors such as accessibility, definiteness, and specificity. The interaction of factors immediately suggests that an explanation for CLD needs to refer to several modules of grammar, i.e., CLD is clearly an interface phenomenon. This chapter aims at contributing to the question of why and under what conditions CLD appears in a language. Special attention is given to the grammatical factors that constrain CLD. We want to identify the semantic and pragmatic causes and their interaction with the grammatical ones at different diachronic and paradigmatic stages of Catalan and Spanish varieties. Concerning other Romance languages, e.g. European Portuguese and Romanian, we will only provide some suggestions on how this explanation could also account for these. We will present new diachronic and synchronic data showing that the diachronic development and synchronic distribution can be analysed as a cyclic change (cf. Jespersen ; Sapir ; van Gelderen ). We will argue that the existence of clitics in a language is, of course, a necessary precondition for the emergence of CLD; however, our data allow the assumption that the emergence of CLD requires both a certain categorial status of the clitic pronouns and a certain specification of the verb-movement parameter. On the one hand, it is a well-known fact that CLD is related to the accessibility scale and definiteness scale of objects (cf. Leonetti ; Fischer and Rinke ). On the other hand, it has often been argued that the change from Old Romance to Modern Romance is a change from topic-prominent to subject-prominent languages, from discourse-oriented to syntax-oriented languages

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

(Lehmann : ; Givón : ). In topic-prominent languages, highly accessible constituents occupy higher positions in the syntactic structure (i.e. further to the left), resulting in a seemingly free word order where high placement of the verb ‘opens’ the structure for A’-positions that may host accessible and prominent objects. Thus, we propose that the function of possible word orders in Romance, more specifically, the function of possible A’-positions in front of the verb which during the Old Romance period also hosted easily accessible and prominent objects in addition to subjects, was compensated for by the possibility of CLD. From a diachronic perspective—as we will show—CLD applies first to the most accessible objects, i.e. human and animate (first full pronouns, followed by indirect objects), conveying exhaustive / contrastive focus readings. The loss in the Old Romance languages of the different preverbal A’-positions for accessible objects, due to the loss of verb-movement into the C-domain, meant that something else was needed to mark accessibility and prominence, and this was—among other things—CLD. In other words, we propose that CLD substitutes in part for the information-structural meaning expressed by the ‘freer’ word order of the historical varieties. In section . we will present the historical dimension and synchronic distribution of CLD in Romance. In section . we will argue that CLD is best analysed as a cycle. In section . we will present data showing that there seems to be a correlation between verb movement and CLD, and in section . we will discuss the Clitic Doubling Cycle from a parametric perspective.

. The historical dimension and synchronic distribution of CLD in Romance The data we use in this chapter, unless indicated otherwise, are part of the corpora which have been assembled at the University of Hamburg.3 For Spanish, twenty-four texts covering the thirteenth up to the nineteenth century have been analysed so far. Where possible, juridical texts, personal diaries, letters, and popular language-related dramas have been used in order to reflect the spoken language as closely as possible. A total of  sentences for Spanish have been extracted and coded for different grammatical, semantic, and pragmatic features [+/
def, +/
anim, +/
specific, +/
dat, +/
acc, +/
DOM, person, number, quantifier, full pronoun, etc.] known to be relevant for the occurrence of CLD. Concerning Catalan, we started out by 3 The corpora have been assembled for the DFG research projects FI /- and FI /-. They are called CDAR_HH_Sp, CDAR_HH_Cat, etc. CDAR stands for Clitic Doubling across Romance, HH stands for Hamburg, Sp for Spanish, etc. The year and concrete text names are given in brackets following this information. A list of all the texts used for our corpora can be found under https://www.slm.uni-hamburg. de/romanistik/forschung/forschungsprojekte/downloads/clitic-doubling-corpus.pdf. We would like to thank our student assistants Lisa Figura, Laura Golla, Svenja Gottschick, Sarah Jobus, Clemens Kirsten, and Christina Maruhn for their help in extracting and coding the data.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



using Fischer’s () syntactically coded corpus of object clitics for Catalan. On the basis of these results, we then created the corpus for Catalan as well, which ranges from the thirteenth up to the nineteenth century. We have analysed a total of nineteen prose texts by different authors. As we did for Spanish, we chose—where possible—Catalan texts that reflect the spoken language as closely as possible. Some of the texts were electronically available from the Corpus Informatitzat del Català Antic. A total of , sentences have been extracted and coded for the same features as the Spanish texts. Both corpora have in part already been blind coded. Concerning the synchronic Catalan data, we created an online acceptability judgement task (AJT) with twenty-six test items, distractors, and control items. The participants were provided with a short context and had to rate the acceptability of each sentence on a four-point scale. If they considered a sentence ‘unacceptable’ or ‘rather unacceptable’, they were asked to correct the sentence in the way they could have uttered it. So far,  speakers from Barcelona and the surrounding areas have completed the questionnaire. The data for Judeo-Spanish (a Spanish contact variety) were gathered in Sofia, Bulgaria, also with an acceptability judgement task. However, since there are not many speakers of Judeo-Spanish in Bulgaria any more (most of them are between seventy-five and eighty years old) (see also Studemund-Halévy and Fischer ; Fischer et al. ), only six speakers, all female, took the test.4 Since the preliminary results show that all speakers doubled full pronouns, and allowed doubling of dative nominal objects as well as of accusative nominal objects, we additionally collected data from articles written in Judeo-Spanish in the web journal of the Noticias del mundo Sefaradí, available on eSefarad.com.5 .. From Latin to Peninsular Modern Spanish / Modern Catalan It is a well-known fact that the Romance languages differ with respect to the existence of CLD. Starting at the beginning, we see that Latin did not show any clitics and as a trivial effect of this, it also did not use CLD. In (), (), and () full pronouns or demonstratives are used instead of clitic pronouns. ()

4

Caesar videt eum. Caesar sees him ‘Caesar sees him.’

(Latin)

We would like to thank Elena Kireva, who carried out the AJT in Sofia for us. In the following some data to illustrate basic unmarked sentence structures are used from the JudeoSpanish translation of Antoine de Saint-Exupéry’s ‘Le petit prince’ (El Princhipiko. , translated by Avner Perez and Gladys Pimienta. Neckarstein: Tintenfass). 5

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

()

Credo ego vos. believe I you ‘I believe you.’

()

Quid illum putas? what that. think. ‘What do you think of that one?’

Proto-Romance did not show CLD either, even though from the very first Old Romance texts clitics are attested. However, it might very well be that these clitics are simple clitics or even weak pronouns (cf. Vincent ). In Old Spanish (OSp) and Old Catalan (OCat) up to the fifteenth century we find object clitics, i.e. syntactic special clitics in the sense of Zwicky (), and CLD is attested with full pronouns (b, b), but it is still optional.6 CLD with dative full DPs seems to be sporadically attested.7 Elvira (: ) finds thirty-five examples (.%) of doubling with the psych-verb plazer (to please): Al rey plogole (To the king pleases him). However, he does not clarify how many of the . % are doubled full DPs and how many are doubled full pronouns. In the texts we used, we did not find CLD with full dative DPs, not even in psych-verb constructions (c, c) (pace Rivero , who provides examples of assumedly obligatory CLD with telic psychverbs like olvidarse8 in Old Spanish, and Bouzouita , who shows some doubling cases with preverbal experiencers, in a position which is ambiguous between a clauseinternal and a clause-external one). ()

a. pusieron a ellos a vna part9 put to them to one side ‘They put them to one side.’ b. yo les fiz I them made ‘I let them know.’

saber know

(OSp)

a ellos to them (CDAR_HH_Sp [Fazienda de ultramar_])

6 It seems that coordination with a full DP is a facilitating factor for strong pronouns to be doubled by a clitic (see Rini : ). 7 Elvira (: ) also provides three examples of doubling with postverbal full DPs: E con esta respuesta plogo-l al conde e tornosse pora su tierra (‘The count was pleased by this answer and he returned to his land’). It seems to us that these examples are better analysed as clitic right dislocation. 8 One of the reviewers pointed out that olvidarse is not a psych-verb. We are aware of the fact that it is usually not considered a psych-verb; however, we have taken this designation from Rivero (). 9 For convenience we only highlight those grammatical categories by abbreviations that are not directly translatable from the glosses. Furthermore, since we mostly use diachronic texts, the examples are a direct copy of the texts, i.e. if there are spaces between the clitic and the words preceding or following (b), the examples are given with spaces, if between the clitic and the word preceding or following is an apostrophe (b) or a hyphen (a), the examples are given with these. We are aware of the fact that according to the Leipzig Glossing Rules clitics should be attached by an equals sign. However, since we believe that the categorial status of the clitic (whether it is a phrase, head or agreement marker) changes through time, the equals sign enforces a wrong analysis for most historical stages. The categorial status of the clitics is discussed in the text.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter c. Plazrie a sus parentes de veerla would.like to his parents to see.her ‘His parents like to see her exhausted.’ ()

a. . . . e tan amarg és a mi que . . . and so bitter is to me that ‘ . . . and it is so bitter for me that . . . ’



transida exhausted (Fischer : ) (OCat) (Fischer : )

b. Prec-vos que m’ojats tots ami un poc. ask-you that me-listen all to.me a little ‘I ask you all to listen to me for a while.’ (CDAR_HH_Cat [Meravelles_]) c. Molt plagueren a Fèlix les paraules que dix la pastora much like to Fèlix the words that said the shepherdess ‘Fèlix liked the words very much that the shepherdess said.’ (Fischer : ) The next change with respect to CLD can be perceived during the period from the sixteenth up to the twentieth century. In Early Modern Spanish (EMSp) we see obligatory CLD with full pronouns (a), and in both EMSp and Decadència Catalan (DCat), the first regular instances of doubling with indirect nominal objects start to appear (b), also in psych-verb constructions (b). Since only personal pronouns and nominal indirect objects are doubled, all instances of doubling are [+animate]. ()

a. y por amor de mi agüela me llamaron a mí Aldonza (EMSp) & for love of my grandmother me called to me Aldonza ‘and for the love of my grandmother they called me Aldonza’ (CDAR_HH_Sp [La Lozana Andaluza_]) b. y a media noche les dije a las camaradas & to half night them said to the companions ‘and at midnight I told my companions’ (CDAR_HH_Sp [Vida del capitán Alonso Contreras ])

()

a. tinga per bé de pagar-me a mi lo que m’és degut (DCat) have for good to pay me to me what me’is owed ‘would you please pay me what is owed to me.’ (CDAR_HH_Cat [Epistolaris d’Hipólita ]) b. que li pesa a vostra senyora what her feel bad to your lady ‘what your lady regrets.’ (CDAR_HH_Cat [Epistolaris d’Hipólita ])

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

From the eighteenth century onwards10 CLD with indirect objects in Standard Spanish seems to be the preferred option (b), whereas it becomes obligatory with psychological verbs (c, c). () a. Pedro *(lo) vio Pedro him saw ‘Pedro saw him.’

a él. to him

(Sp)

b. (Le) devolví el coche a Pedro. him gave.back the car to Pedro ‘I gave back the car to Pedro.’ c. *(Le) gusta la música her like the music ‘María likes classical music.’

clásica classical

a María. to María

() a. Ahir no *(el) vaig veure a ell yesterday not him saw to him ‘Yesterday I saw only him (not her).’

(sinó només a ella). (Cat) (but only to her) (CDAR_HH_ Cat [AJT])

b. A la inauguració (li) van regalar flors a l’Ada Colau. at the inauguration her gave flowers to the’Ada Colau ‘At the inauguration, they gave flowers to A.C.’ (CDAR_HH_ Cat [AJT]) c. A en Jordi *(li) agrada to the Jordi him like ‘Jordi likes classical music.’

la música clàssica. the music classical (CDAR_HH_ Cat [AJT])

Usually three types of datives are distinguished: indirect objects in ditransitive constructions, with the thematic role of [recipient], [goal], or [source / location]; the [experiencer] argument in unaccusative transitives (e.g. psych-verbs); and [possessive], [benefactor], and [ethical] datives (Franco and Huidobro ). It is commonly assumed that ditransitive constructions allow optional doubling, however, what can be seen with respect to the data is that thematic roles that are intrinsically marked with the features [+animate, +def, +spec] are more readily doubled than others. This fact cannot be derived by the definiteness scale proposed by Leonetti () alone; it also seems to be dependent on the accessibility hierarchy (Ariel ). Consequently, experiencer objects in psych-verb constructions, non-argumental benefactors (Gutiérrez Ordóñez : –), and possessors, usually carrying the features [+animate, +def, +spec], are obligatorily doubled which (recall examples (c) and (c)) were not yet doubled during the medieval period. Since ethical datives cannot be expressed by full DPs, but only by the clitic, they are not involved in doubling structures. 10 One of the reviewers argues that the many examples of CLD attested in CORDE suggest that it was the preferred option from the seventeenth century onwards. Our data do not allow this assumption. Dative CLD in Peninsular Spanish and Catalan is far from clear-cut; the data nowadays are still very diverse and show differences according to region, register, and bilingualism (see also footnote ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



.. Non-peninsular Spanish varieties When looking at some of the Spanish varieties outside Spain, we see with respect to Buenos Aires Spanish, for example, that full pronouns and indirect objects are obligatorily doubled,11 and that accusative animate nominal objects () are optionally doubled (cf. Di Tullio and Zdrojewski ; Schroten ; Zdrojewski and Sánchez ). () Las saludé a las maestras del jardín. (Buenos Aires Sp) them greeted to the teachers from.the garden ‘I greeted the teachers from the kindergarten.’ (Zdrojewski and Sánchez : ) We also find CLD in the Judeo-Spanish (JSp) variety spoken in Bulgaria. In our data, speakers even use CLD with animate definite full accusative object DPs (), which are not doubled in Standard Spanish. Since the data we gathered in Bulgaria are quite limited, we verified them with articles written in Judeo-Spanish that were available online, which confirmed our findings. In both examples in () a [+animate, +def, +spec] accusative object is doubled. () La vimos a Maria. her saw to Mary ‘We saw Mary.’

(JSp) (CDAR_HH_JSp [AJT])

() a. Yildiz disho ademas ke mi tiya Beki la ayudo (JSp) Yildiz said additionally that my aunt Beki her helped a parir a su madre. to give birth to her mother ‘Yildiz also said that my aunt Beki helped her mother to give birth.’ (CDAR_HH_JSp [eSefarad ]) b. La hija hazina la yamó a la madre, después the daughter sick her called to the mother after di kuarenta días. of forty days ‘The sick daughter called her mother after forty days.’ (CDAR_HH_JSp [eSefarad ]) In other varieties, e.g. in Lima Spanish and Andean Spanish, doubled indefinite full accusative DPs () and even doubled inanimate accusative DPs () are attested.

11

See also Dufter and Stark (), who provide data from C-ORAL-ROM showing that dative CLD is almost obligatory with decir and dar in Modern Peninsular Spanish, and Nishida (), who shows very low rates of dative CLD in formal Mexican varieties.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

() Lo saludé a un estudiante him greeted to a student ‘I greeted a student that I know.’ () a. Lo vendo him.. sell ‘I sell all the cars.

que conozco. (Lima Sp) that know (Zdrojewski and Sánchez : )

toditos all...

b. Eso también lo that too him.. ‘That too kills the plants.’

los carros. the car..

(Andean Sp)

mata las plantas. kill the plant.. (Zdrojewski and Sánchez : )

In (a) the clitic does not agree in number with the nominal object it doubles; in (b) the clitic agrees neither in gender nor in number. This can be seen as evidence that the grammaticalization of the clitic pronoun in this variety is more advanced. Since CLD can be used with all kinds of objects, it seems that lo is used as an agreement marker (cf. Bleam  and further discussion in sections . and .), which would involve the dissolution of the doubling construction (cf. Vega Vilanova  for a proposal on how the change from doubling to agreement takes place).

. CLD as a cyclic process The data collected from the different periods of Spanish and Catalan and from the different Spanish varieties outside Spain allow us to suggest that the development of CLD in the Romance languages follows a cyclic pattern (see also Navarro et al. ). We propose a cycle which, considering the data in section ., seems to consist of five stages (figure .). Stage I

→ no CLD

Stage II

→ optional CLD with full pronouns

Stage III → obligatory CLD with full pronouns,

Latin/ProtoRom OSp/OCat12 EMSp/DCat/Sp/Cat

→ optional CLD with indirect nominal objects [+anim, +def, +spec] Stage IV → obligatory CLD with full pronouns,

Buenos Aires Sp/JSp

→ obligatory CLD with indirect nominal objects → spread of CLD to direct nominal objects [+anim, +def, +spec] Stage V

→ generalized CLD (with all objects, even inanimates)

Lima, Andean Sp

FIGURE . The Clitic Doubling Cycle

12

Other Romance languages were not part of our research project, which focuses on the variation of CLD in Spanish and Catalan varieties. However, it seems that the data cited in the literature from Portuguese and Romanian, and probably also from Italian and French, can also be interpreted as a consequence of changes in verb movement and the means of expressing the accessibility of objects. In European Portuguese (EP), for instance, CLD with full pronouns is still optional (perhaps obligatory in some varieties), as argued by Barrie () and Gonçalves et al. (). Some Brazilian Portuguese (BP) dialects allow CLD with experiencers (Machado-Rocha and Martins Ramos ). This would roughly correspond to stage II: highly accessible objects (i.e. personal pronouns and experiencers) can be optionally

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



The last stage, stage V, is especially interesting. It is represented, for example, by Lima Spanish and Andean Spanish, where CLD is allowed with indefinite and inanimate DPs, and where the Andean Spanish clitic no longer agrees with the object it doubles (). It looks more like some kind of agreement, perhaps parallel to subject agreement. We assume that the next step will be the complete loss of the category ‘object clitic’, which is the starting point of the whole cycle.

. The emergence of CLD One of the most prominent approaches to explain the emergence of CLD has connected the rise of CLD to the grammaticalization path of the clitic itself. Fontana () suggests that the rise of CLD is related to the reanalysis of clitics from being Xmax to Xmin (Fontana : ff.), i.e. from DPs to D-head categories. Combining all proposals concerning the categorial status of Romance clitics over the years, we can identify the following grammaticalization path () (cf. Fischer and Rinke : ): () Grammaticalization path of the clitic Clitic > Clitic > Clitic DPs D-head phi-features However, it seems that the connection is not that simple. At least in OCat, sentences are attested with and without CLD (recall example ()) even though the OCat clitics have to be analysed as heads, i.e., they never appear left of negation and they are never separated from the verb (Fischer ). We agree that part of the variability can be explained by the fact that not all clitics in a specific period of a language are at the same stage of the path, e.g., for Modern Spanish clitics it has been assumed that accusative clitics are D-heads, whereas dative clitics are already agreement markers (phi-features) (Bleam ). But when correlating the five stages of CLD (figure .) with the categorial status of the clitic and the accessibility hierarchy of objects (Leonetti ; Fischer and Rinke ), it becomes obvious that something else is of importance as well. Thus, reviewing the syntax of the languages at the different stages (Old Spanish / Catalan; Early Modern Spanish / Decadència Catalan; different varieties of Spanish) and correlating this with the grammaticalization path of the clitics, it looks as if CLD is connected to the possibility for the verb to move to the front of the sentence and thereby provide a position for A’-movement. .. From Latin to the modern Catalan / Spanish varieties In order to bring to light the interrelation of CLD and verb movement, we will correlate the five stages of CLD with the grammaticalization path of the clitic and the possibility for the verb to move. doubled by a clitic. Of course, the available data are still incomplete, and a series of other properties of EP/ BP, which potentially have an effect on CLD, should be considered. In Standard French and normative Italian, for example, the existence of past participle agreement could have hindered the emergence of CLD (Tsakali and Anagnostopoulou ; Vega Vilanova ). As for Romanian, see footnote .

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

As we have shown in section ., in stage I, represented by Latin and ProtoRomance, no special clitics or CLD are available. Latin and Proto-Romance are discourse-oriented languages, where anything can appear in front of the verb (cf. Menge  for Classical Latin; Devine and Stephens  for Vulgar Latin/ Proto-Romance). In these kinds of languages, where word order is clearly dependent on information structure, the position in front of the verb needs to be analysed as an A’-position. According to Ledgeway (), there is enough robust evidence that Late Latin behaves as a V language, i.e., the verb commonly moves up to C°. In stage II, represented by Old Spanish and Old Catalan, special clitics (cf. Fischer ) and optional CLD of full pronouns (, ) are attested. Concerning Old Spanish, we assume that the clitics should be analysed as DPs, since interpolation is attested (). However, in Old Catalan clitics are clearly heads, D°, since they never appear left of negation (Kayne ), i.e., no interpolation is attested (cf. Fischer ; Batllori et al. ); nevertheless they appear in front of and after the finite verb in main (a) and embedded clauses (b), and the order of the clitic is not yet fixed (a–b) (see Fischer ; Batllori et al.  for further discussion on this issue). () a. assi como les dios auie so how them God had ‘as God had promised them’

prometido promised

(OSp) (Fischer : )

b. como a ty cierto es que lo non hamas as to you true is that him not loves ‘as you know that you do not love him’ (CDAR_HH_Sp [Corbacho_]) () a. e la emperadriu ha-li perdonat francament per ço (OCat) & the empress has him forgiven frankly for that cor Déus e vós li avets feyta tant d’onhor because God & you him have done so much of ’honour ‘And the empress has frankly forgiven him, since God and you have done him such honour’ (CDAR_HH_Cat [Crònica de Bernat Desclot_]) b. e les gens sabien lo rey tan just que desperaven-se de ell & the people knew the king so just that despaired- of him quant lo havien offès . . . when him had offended ‘ . . . and the people knew the king to be so just that they despaired of him when they had offended him . . . ’ (Fischer : ) () a. car lo pare lo li havie tolt because the father it him had removed ‘because his father had removed it from him’ b. e el pages li lo atorga and the peasant him it concede ‘and the peasant admits it to him’

(OCat)

(Fischer : )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



In both languages we find classic V structures (a, a) next to V structures (b, b) and V orders, as well as stylistic fronting (), which has been taken as evidence that the verb can move as high as C° in V and as high as Σ° in V structures (Martins ; Fischer , ). () a. e entonces le dixo Muget . . . and then him said Muget ‘and then Muget said to him . . . ’

(OSp)

b. E entretanto el hermano de Mahomad llego al rrey and meanwhile the brother of Mahomad arrived to.the king ‘and meanwhile the brother of Mahomad came to the king’ (Fontana : ) () a. Tantost e sens triga vengueren Jacob e Curial soon and without haste came Jacob and Curial ‘Jacob and Curial came soon without haste.’

(OCat)

b. E d’aquí avant lo rey féu-li donar tot . . . and from there onwards the king made-him give all ‘And from there onwards the king made him give all . . . ’ (Fischer : ) () que feita aviets la corona del who made had the crown of.the ‘who had made the crown of the empire’

Emperi, empire (Fischer : )

Stage III shows CLD with full pronouns, which becomes obligatory (at the latest) by the end of the seventeenth century, and optional CLD with dative nominals. Neither in Spanish nor in Catalan is interpolation attested during this period, thus the clitics can be analysed as D°, and in some cases perhaps even as phi-features (cf. Bleam ). Word order still has some discourse function. We still find postverbal subjects (b) and postverbal clitics in matrix sentences (a–c), however, for the first time we also find sentence-initial clitics in Catalan (d) (Fischer ). () a. Moríran-hi uns quans capitans died-there some few captains ‘Some captains of Spain died there.’

dʼEspanya. of Spain

(Cat)

b. . . . entraren dintra la ciutat lo dit bisbe entered in the city the said bishop de Malta e lo duch de Gandia, . . . e lo bisbe e lo duc of Malta and the duke of Gandia and the bishop and the duke meteren-lo’s al mig. place-him. into middle ‘and the aforementioned bishop of Malta and the Duke of Gandia came into the city and the bishop and the duke placed him in their middle’ (Fischer : )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova c. Anem, pues, de Lucrècia a veure la bellesa; mostram-li mon go well, from Lucrècia to see the beauty; show-him my ardor, ma flama, ma tendresa. heat, my flame, my tenderness ‘Well, we go to see Lucrècia’s beauty; we show him my heat, my flame, my tenderness.’ (Fischer : ) d. Se contà que en esta nit passada, del  de juny de ,  told that in this night passed, of.the  of June of  alguns moros . . . some Arabs ‘It was told that last night, June , , some Arabs . . . ’ (Fischer : )

Additionally, during this period from the sixteenth to the nineteenth century, we find oblique subjects (so-called quirky subjects) in preverbal position that still pass the syntactic tests identifying them as subjects (Keenan ; Fischer ; Vega Vilanova ). From the nineteenth to the twentieth century, postverbal subjects are attested in many contexts in Standard Spanish (Zagona ) and Standard Catalan (Vallduví ). Oblique subjects pass only a few of the syntactic tests (): for instance, oblique subjects are still able to rise to the subject position of control verbs such as parecer / semblar ‘to seem’. The fact that quirky subjects still pass a few of these tests has been taken as evidence for the verb being in T° and the dual character of SpecTP as an A- / A’-position (e.g. Masullo ), while others explain the specific properties by an additional position within TP (e.g. Gutiérrez-Bravo ; Fischer ).13 () a. A l’autora sembla agradar-li especialment la història to the’author seems to like-her especially the history de Roma. (Cat) of Rome ‘The author seems to especially like the history of Rome.’ b. A Pedro parece gustarle to Pedro seems to like.him ‘Pedro seems to like the meal.’

la comida. the meal

(Sp)

In Stage IV we find obligatory CLD with full pronouns, obligatory doubling with dative nominal objects (see Di Tullio and Zdrojewski ; Schroten  for dative CLD in Buenos Aires Spanish), and optional CLD with accusative nominal objects (recall example ()) for Buenos Aires Spanish and Judeo-Spanish in Bulgaria (recall 13 The additional position in which quirky subjects in Spanish and Catalan end up might be ΣP (cf. Fischer ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



examples (, )).14 Dative clitics are analysed as phi-features (Anagnostopoulou ; Bleam ), and accusative clitics as D°s. It has been shown that Argentinian Spanish has a strong preference for SVO (around ninety per cent of the attested data) in contexts where other Spanish varieties such as Peninsular Spanish would require a non-canonical word order (e.g. VOS) (Gabriel ). The same has been shown for Judeo-Spanish () (cf. Fischer et al. : ). In those varieties where word order is rather strict and does not convey information-structural meaning, we suggest that the verb still moves up to T°; however, A’-movement of the object might only be up to vP. (JSp) () a. El rey no respondio nada. the king not answered nothing ‘The king did not answer anything.’ (CDAR_HH_JSp [El Princhipiko_]) b. Entonses, vos podesh imajinar mi sorpreza. then, you can imagine my surprise ‘Thus, you can imagine / picture my surprise.’ (CDAR_HH_JSp [El Princhipiko_]) Stage V of the cycle is exemplified by Lima and Andean Spanish, which show CLD in all contexts (full pronoun, dative objects, and even inanimate and indefinite objects). As the data in section . showed, clitics in Andean Spanish are ‘bleached’ in so far as they are no longer inflected for number and gender (we suggest that they are mere agreement markers, i.e. phi-features, and no longer clitics), the verb might even stay lower than T° (Alexiadou and Anagnostopoulou ; Ordóñez ) and word order seems to be even stricter: as Muntendam () argues, Andean Spanish often resorts to dislocations in order to express information structure of the object, which can be interpreted as the lack of the relevant object A’-positions within the clause. .. Summarizing When correlating the five stages of CLD with the grammaticalization path, i.e. the categorial status of the clitic and the accessibility hierarchy of objects (Leonetti ; Fischer and Rinke ), it becomes obvious that something additional plays a role. Thus, reviewing the syntax of the languages at the different stages (Old Spanish / Catalan, Early Modern Spanish / Decadència Catalan, Modern Spanish / Catalan, different contemporary varieties of Spanish) and correlating this with verb 14 We take Romanian to also belong to this stage, even though CLD is not yet obligatory with dative nominals (Diaconescu and Rivero ; Cornilescu ). Nevertheless, Romanian also allows doubling with accusative nominals that are animate (Dobrovie-Sorin ; Hill and Tasmowski ; Fischer and Rinke ), i.e., CLD is possible with objects placed at the same high position in the accessibility hierarchy, as predicted in our approach.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

movement, it looks as if CLD took over after verb movement to the C-domain in canonical declarative sentences was lost. The change from Old Romance to Modern Romance has often been called a change from topic-prominent to subject-prominent language (Lehmann : ; Givón : ), which is clearly a change concerning the phrase structure of the languages, i.e. the available A’-positions in front of the verb. These positions, which in Old Romance were used—in addition to subjects— for easily accessible and prominent objects, were replaced by the possibility of CLD.15 Our data clearly confirm that CLD first applies to the most accessible objects, i.e. human and animate (first full pronouns, followed by indirect objects), and only afterwards—paralleling the loss of more and more A’-positions in front of the verb— is CLD extended to less accessible objects following the accessibility scale and definiteness scale (Ariel ; Leonetti ; Fischer and Rinke ). In other words, the categorial status of the clitic and the specification of the verbmovement parameter trigger the emergence of CLD and thus explain the distribution of CLD across the Romance languages (figure .).

Verb Movement v°

T°

Σ°

C° I

Pronouns

II

Datives/Pronouns

III

Accusatives/Datives/Pronouns

IV

Generalized/Accusatives/Datives/Pronouns

V

DP/D°

D°/φ

φ

Clitic Doubling

Grammaticalization

DP

FIGURE . The interaction of verb movement and grammaticalization

Consequently, CLD supplants in part the information-structural function expressed by word order.

15 Other constructions, e.g. past participle agreement, also seem to fulfil the function of marking accessible objects, which would eventually block the emergence of CLD in languages in which all other conditions specified above concerning the grammaticalization status of the clitic and the specification of the verb-movement parameter are satisfied (cf. Tsakali and Anagnostopoulou ; Vega Vilanova ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



. CLD from a parametric perspective Since Government-Binding theory, typological diversity across languages was explained by different parameters which are set according to the positive evidence, i.e. the primary linguistic data (PLD), during language acquisition. Language change was thus explained as the reanalysis of the available PLD by the language learner (Lightfoot ; Kroch ; Roberts and Roussou ). However, the vast number of parameters that have been attributed to universal grammar (UG) have led to an overspecified conception of it. This overspecification consequently increases the structural complexity of UG itself and has been subject to severe criticism in recent years (see e.g. Newmeyer ; Boeckx ). Baker () and Roberts (), among others, have shown that parameters are hierarchically organized. Especially, Roberts () and Holmberg and Roberts (), building on Chomsky (), propose viewing parameters as epiphenomena arising as a consequence of the interaction of a learnability factor with the linguistic experience and the genetic endowment. These factors together thus determine the emergence of parameter hierarchies. In other words, a parameter is a function of this interaction, and the distribution of its different specifications finally gives rise to a hierarchical structure from macro- to microparameters (figure .). grammaticalization path Macroparameter no

yes = Mesoparameter no

unmarked

yes = Microparameter no

yes = Nanoparameter

marked acquisition path

FIGURE . Parameter hierarchies (adapted from Roberts )

The hierarchy thus is organized in a top-down relation: at the top of the hierarchy we find the macroparameter or the unmarked specification and at the bottom of it the nanoparameter, which is equivalent to the most marked specification. From a language-acquisition perspective, the learner goes down the hierarchy, i.e. from an unmarked specification to a marked interpretation of the hierarchy. However, in Roberts’ view, language change—in contrast to language acquisition—always goes up the hierarchy, i.e., in language change, markedness is subject to loss. We agree. However, what seems to be true as well—especially concerning cyclic changes—is the fact that at a certain point, unmarkedness gives way to markedness. Thus, language change can, of course, go up the parametric hierarchy, but at the same

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

time—concerning other parameters—language change must go down the parametric hierarchy, otherwise cyclic change would not be part of language change. According to Roberts’ hierarchies, and taking into account the data presented in section ., we suggest that the distribution of CLD across languages can be explained by a CLD parameter as discussed in Fischer and Rinke (), where one macroparameter gives way to subsequent meso-, micro-, and finally nanoparameters. Along these lines, Fischer and Rinke (: ) suggest that diachronically the emergence of clitic pronouns is the prerequisite for the occurrence of doubling and see this as the macroparameter of CLD. However, while it seems correct that special clitics need to be available in a language in order for CLD to occur, it does not seem correct to assume that this is due to a parametric setting, since the change from a weak pronoun, simple clitic to a special clitic is a gradual process and not categorial. Furthermore, the existence of clitics in a language is, of course, a necessary precondition for the emergence of CLD, but we suggest that it is not part of the parameter hierarchy, since it is not only the status that a clitic reaches during its grammaticalization path that triggers CLD. The resulting hierarchy is shown in (). ()

The CLD parameter hierarchy (adapted from Fischer and Rinke 2013) CLD - full pronouns CLD - dative nominal objects CLD - specific nominal accusative objects CLD - nominal accusative objects

Our data show that verb movement seems to interact with CLD at the same time. Thus, as mentioned before, we suggest that the categorial status of the clitic and the specification of the verb-movement parameter (whether the verb moves to C°, Σ°, T°, v°, or remains in V°) in canonical declarative sentences jointly trigger the emergence of CLD (figure .). Grammaticalization of clitic pronouns

Verb-movement parameter Clitic doubling parameter

FIGURE . Emergence of clitic doubling

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The clitic doubling parameter



Furthermore, we and other linguists understand this phenomenon as a parameter, namely the CLD parameter (among others Tsakali and Anagnostopoulou ; Anagnostopoulou ). However, it should be noted that this kind of parameter, which emerges as the result of interacting factors, seems to be different from, for example, the verb-movement parameter.16 The emergence of the verb-movement parameter depends exclusively on the specification of formal features and their interaction with principles of UG (cf. Holmberg and Roberts ): either languages are V languages, they move the verb up to C° (e.g. German), or they are languages that move the verb up to T° (the Modern Romance languages), or they don’t move the verb at all (e.g. Creole languages). The CLD parameter hierarchy can be compared to an implicational universal, in which higher parameters are contained in the lower ones; for example, to our knowledge there is no language that allows CLD with accusatives without also allowing CLD with datives and full pronouns (cf. Fischer and Rinke ). Also, in addition to formal feature specifications of UG, semantic (e.g. animacy) and pragmatic (e.g. accessibility) features are involved in deriving CLD. So perhaps this explains why language change seems to go from unmarkedness to markedness with regard to this parameter. The question that still awaits an answer is why some varieties freeze at some point in the hierarchy while others change rather rapidly. Looking at the Spanish varieties that extended CLD to nominal accusatives, it becomes obvious that these are all varieties that have been or were in extensive contact situations (e.g. with Basque, Quechua, and Bulgarian). It has been argued that these varieties borrowed an object agreement pattern from the languages with which they were in contact (Urrutia Cárdenas ; Lipski ). However, we found extended CLD in varieties whose contact languages do not show object agreement, e.g. Judeo-Spanish in Bulgaria (, ) or the Spanish varieties spoken in Argentina () and in Patagonia (). In other words, the CLD parameter—differently, for instance, from the verb-movement parameter—seems to be sensitive to language-contact situations. () La toqué a la sonata. .. touched to the sonata ‘I played the sonata.’ () a. Lo tomé .. took ‘I took the bus.’

el colectivo. the bus

b. La pinté la casa. .. painted the house ‘I painted the house.’

(Argentinian Sp) (Bleam : ) (Patagonian Sp)

(CDAR_HH_PaSp)

16 The distinction between the two kinds of parameters is also made in Mario Navarro’s PhD project ‘Language design and parametric interaction. Approaching the emergence of CLD from below’.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Fischer, Navarro, and Vega Vilanova

In contrast to the general assumption that contact Spanish is changed under the direct influence of the contact languages (Quechua ! Spanish, Nahuatl ! Spanish, Basque ! Spanish, etc.), we have proposed elsewhere (Fischer et al. forthcoming) that CLD as an interface phenomenon is especially vulnerable to language variation and change in language-contact situations (Sorace ). Dative CLD has often been claimed to be optional in Peninsular Standard Spanish, but we know that the accessibility hierarchy of objects and their theta-roles as well seem to influence whether a dative object is doubled or not (Leonetti , Dufter and Stark ). Furthermore grammatical (pronominal vs non-pronominal, the occurrence vs nonoccurrence of DOM) and semantic factors (animacy) play a role. Thus, CLD applies at several interfaces, at the internal interface (e.g. syntax–semantics) as well as at the external interface (syntax–pragmatics / discourse). What speakers do in multilingual contexts is to overgeneralize an existing pattern (dative CLD) to other contexts (accusative CLD) (cf. Fischer et al. ) in the sense of Weinreich (), hence moving down the parameter hierarchy in (). However, the exact manner of the extension of CLD in multilingual contexts needs to be investigated further.

. Summary We provided data showing that a cyclic change can be identified with respect to the evolution of CLD in Spanish and Catalan (and is probably applicable to other Romance languages as well). The proposed cycle of five stages explains not only the attested diachronic data at the different stages but also the synchronic variation in Catalan, Spanish, and some of their varieties. In addition, we argued that the emergence of the CLD constructions could not be adequately explained by focusing only on a single factor. Instead, we proposed that the CLD parameter is triggered by the categorial status of the clitic, i.e. its grammaticalization path and the specification of the verb-movement parameter. As a result of this interaction, not only formal features are involved in deriving the CLD parameter but also semantic and pragmatic features.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

5 Weak elements in cycles A case study on dative pronouns in Old Italo-Romance JA C O PO G A R Z O NI O AN D S I L VIA R O S S I

. Introduction Since Cardinaletti and Starke (), it has been standardly assumed that pronominal items, in particular personal pronouns, come universally in three classes, strong, weak, and clitic. Each of these categories is associated with different yet interrelated distributional, morphological, and semantic properties, which are the result of varying degrees of structural deficiency: strong, weak, and clitic share the same underlying structure, but clitic and weak elements have less functional structure than strong pronouns (structures adapted from Cardinaletti and Starke : ). ()

a. strong = [CP [ΣP [IP [LP b. weak = [CP [ΣP [IP [LP c. clitic = [CP [ΣP [IP [LP

For instance, Cardinaletti and Starke (: ff.) argue that strong pronouns—but not deficient pronouns—have to refer to a [+human] entity. This semantic restriction is explained by assuming that the higher layer CP in () contains a ‘referential index’, which forces a [+human] interpretation. Other differences in phonology and syntactic distribution are accounted for in a similar fashion.1 Yet, though being deficient and sharing various properties, weak pronouns and clitic pronouns do not behave alike: weak items can also share some properties with

1 The CP layer is also the place for functional case: ‘Σ° may be taken to be the locus of prosody-related features of L°’ (Cardinaletti and Starke : ). Thus, clitics lacking Σ° must be in a local relation with an item which has Σ° (like verbs). The IP layer contains agreement features, and LP contains the lexical category. We refer the reader to Cardinaletti and Starke’s paper for further details.

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Jacopo Garzonio and Silvia Rossi . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

strong pronouns. For example, they are sometimes homophonous with their strong counterparts (while clitics are usually morphophonologically reduced), and they are not bound morphemes. In order to capture this second divide within the deficient group of pronouns, Cardinaletti and Starke () also propose that weak and strong items are XPs while clitics are heads. In other words, weak pronouns are structurally deficient XPs, while clitics are structurally deficient X°s. The synchronic distribution of these items in a given language is regulated by a choice principle (Cardinaletti and Starke ): ()

Choose the most deficient possible form.

In recent years, Egerland () has proposed that the Choice Principle also holds diachronically, as pronouns have a strong tendency over time to be reanalysed in the order strong > weak > clitic (see also Roberts ). For instance, strong sg person Latin demonstrative pronoun ille ‘that one’ becomes weak egli in Old Italian and clitic e’ in some present-day Italo-Romance varieties (Egerland ).2 Similarly, the same cline could be observed also for the pl genitive/dative pronoun: from the Latin strong illorum ‘of those ones’, (a), Old Florentine has developed a weak dative lor(o) ‘to them’, (b), and Old Sienese has gone a step further and presents a clitic dative lo’ ‘to them’, (c) (see Garzonio and Rossi , and the following sections). () a. ut alii sibi esse illorum similis expetant (Latin) that others themselves. be. them. similar expect. ‘that others would want to have children similar to those’ (Plautus, Mostellaria ) b. da che lor non piacessi . . . (Old Florentine) from that them. not pleased. ‘since they do not like you . . . ’ (; Brunetto Latini, Tesoretto, v. ) c. e mai no lo’ volse dare udienza (Old Sienese) and never not them. wanted give hearing ‘but he never agreed to hear them’ (; Cronaca senese) Hence, the shift from weak pronouns to clitics is structurally a shift from XPs to X°s, which could be interpreted in terms of an incipient cyclical change (van Gelderen , , b). In this chapter, we will show that if one assumes that the Head Preference Principle (van Gelderen c) is sufficient to explain the cline from strong to clitic functional items, the individuation of a strong-weak-clitic cycle for dative loro is problematic (see also Vedovato  on this as a ‘broken cycle’). As we will see, pl 2

Egerland () claims that the trigger behind this diachronic cline is a process of morphophonological weakening. Our approach maintains that these items undergo morphophonological reduction but it also seeks to propose a structural interpretation behind this gradual development.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



dative loro, the prototypical weak pronoun of Modern Italian, does not seem to exhibit the expected properties of well-behaved weak pronouns in Old Tuscan texts. What these texts show is a much more complicated picture, where clearly separated stages are not observable: even in the earlier texts, one can encounter cases of loro displaying clitic properties, while later texts contain instances of loro with (at least some) strong properties. We interpret these unexpected patterns as minimal differences in the morphosyntactic make-up of such elements, which cannot be reduced to a deficient XP vs deficient X° opposition. More precisely, we will argue that a cyclical change from strong to clitic can be maintained but crucially, it has to be reinterpreted in terms of an interaction between the Head Preference Principle and reanalysis as upward movement in minimal steps inside the functional spine of pronouns, à la Roberts and Roussou (). Furthermore, the following discussion also wishes to shed some light on the rise of unfinished cycles, i.e. ‘broken cycles’. A linguistic cycle should also involve renewal, but in the following stages of the language no renewal of these items seems to be found, as weak loro is still present in written Modern Italian (an effect of prescriptive pressure, as discussed by Vedovato ). In what follows, we briefly sketch the distribution of Modern Italian loro (section ..) and then discuss the various cases from some Old Tuscan varieties (section ..), which lend support to the diachronic cline proposed in Egerland (, ). In section . we present and discuss cases of non-well-behaved weak loro, while in section . we outline a proposal for the Old Tuscan situation relating the internal structure of pronouns to more general sentence phenomena of Old Romance varieties like Verb Second (V). Section . concludes.

. Well-behaved strong, weak, and clitic pl datives In this section, we discuss cases of dative loro behaving as expected under an analysis where it is considered a weak pronoun à la Cardinaletti and Starke (). But, before this, we briefly outline loro’s evolution, and its behaviour in Modern Italian. .. From Latin illorum to Modern Italian loro Old and Modern Italian loro ‘to them’ derive etymologically from Latin illorum, the plural genitive form of the demonstrative ille ‘that one’. () pronuntiare . . . illorum esse praedam . . . declare. those. be. booty. ‘ . . . to declare that the booty belonged to those ones . . . ’ (Caesar, De Bello Gallico, .) This entails that from the earliest Italo-Romance attestations, we encounter at least two diachronic shifts, one turning distal demonstrative elements of the ille-series into

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

third-person pronouns (and also articles; see among others Vincent ), and the other involving a development from genitive to dative. This last shift, however, is much more articulated than a simple syncretism inside oblique case, for at least two reasons: i) even in the early Italo-Romance texts, loro is found not only in subject (a) and direct object (b) positions, but also under prepositions (c), that is, it has a ‘direct case’ strong variant; ii) loro can also appear as a possessive, i.e. genitive, inside DPs (). () a. e loro sì nne deo(no) dare noi quello che lloro piace (OFlor.) and they  of.it must give. us that that to.them pleases ‘and they have to give us how much they like’ (; Bene Bencivenni, transaction record) b. et inducendo loro a ssapere le cose utili et oneste (OFlor.) and inducing them to know. the things useful and honest ‘teaching them how to recognize useful and truthful things’ (; Brunetto Latini, La Rettorica) c. ave(n) konperato da loro due peçi di tera (OFlor.) have. bought from them two pieces of land posti ad Aliana situated at A. ‘We bought from them due pieces of land near Agliana.’ (; Ricordi di compere in val di Streda) () Ma loro arme e lor cavalla lassaro dai Pagani (OLucchese) but their weapons and their horses left. at.the heathens ‘But they left their weapons and their horse to the Infidels’ (; Ritmo Lucchese) The analysis proposed by Cardinaletti () and Cardinaletti and Starke () is based on the assumption that there are two distinct entries for loro in Modern Italian: the strong form has direct case and can be used in dative contexts only if introduced by the preposition/dative-case marker a ‘to’, while the weak form is only oblique and receives dative case in a specific position above v/VP. In other words, weak dative loro appears in dedicated sentential positions (usually before direct objects, and low IP adverbs) while the strong a loro occurs inside v/VP. Moreover, these two forms behave differently also with respect to modification, coordination, and topicalization/ focalization. We show this variation in (), adapted from Garzonio and Rossi (), and based on Cardinaletti (): the minimal pair in (a-a’) shows that only weak loro precedes the direct object in the unmarked word order; (b–b’) shows that only weak loro can appear pre-participially; (c–c’) shows that only weak loro precedes low IP-adverbs (Cinque ); (d–d’), (e–e’) and (f–f ’) show that weak loro cannot be coordinated, modified, and focalized/topicalized.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles ()

a. Ho have.

dato given

loro them



il libro. the book

a’. Ho dato il libro a loro. have. given the book to them ‘I gave them the book.’ b. Ho loro detto [CP che . . . ] have. them said that . . . b’. Ho detto a loro [CP che . . . ] have. said to them that . . . ‘I said to them that . . . ’ c. Ho dato loro tutto. have. given them everything c’. Ho dato tutto a loro. have. given everything to them ‘I gave them everything.’ d. *Ho dato il libro loro e loro. have. given the book them and them d’. Ho dato il libro a loro e have. given the book to them and ‘I gave the book to them and to them.’ e. *Ho parlato tutti loro. have. talked all them e’. Ho parlato a tutti loro. have. talked to all them ‘I talked to all of them.’ f. *LORO ho dato il libro. THEM have. given the book f ’. A LORO ho dato il libro. TO THEM have. given the book ‘TO THEM I gave the book.’

a loro. to them

However, Manzini () and Manzini and Savoia () present some data on both direct and oblique loro which contrast with the picture emerging from Cardinaletti () and Cardinaletti and Starke (). In particular, they provide corpus examples and native speakers’ judgements showing that oblique loro can be coordinated and modified by focalizers like anche ‘also’. Furthermore, they present data on direct (i.e. strong) loro with [–animate] interpretation. These facts are accounted for by assuming that there are not two separate lexical items loro, weak vs strong, but only one, which is morphologically decomposed into l-oro, where ‑oro can alternatively encode [plural] and [oblique] or only [plural].

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

As for the clitic vs weak divide, the distributional properties of clitics are even more restrictive, since, unlike weak items, they are bound morphemes. For instance, Modern Italian weak loro cannot appear between preverbal negation non and the finite verb—a position only available to clitics. On the basis of this strong restriction, Cardinaletti () distinguishes homophonous clitic and weak pronouns in Old Italian, which behaves just like Modern Italian in this specific respect. Furthermore, only clitics are subject to the different types of the Person-Case Constraint (PCC) (Bonet ) and can be used as resumptive elements in left dislocation constructions. In the next section we consider cases of pl dative pronouns behaving as expected under Cardinaletti and Starke (); specifically we discuss examples of wellbehaved weak loro in Old Tuscan texts, and of well-behaved clitic lo’ in Old Sienese. .. Loro and lo’ In Old Italian, the pl dative pronoun loro already had a very similar distribution to that it has in Modern Italian. Cardinaletti (: ff.) gives examples of dative loro before the direct object (a), and before the non-finite verb (b), (cf. a–b):3 () a. Et insegnava loro le cose oneste dicendo . . . (OFlor.) and taught. them the things truthful saying ‘And he would teach them the truth, saying . . . ’ (; Brunetto Latini, Rettorica) b. . . . che peggio non potea lor fare Dio che . . . (OFlor.) . . . that worse not could. them do. God that ‘ . . . that God could not do anything worse to them than . . . ’ (; Bono Giamboni, Il Libro de’ Vizi e delle Virtudi, ch. , par. ) Dative loro could also occur in preverbal position, a position marginally allowed in Modern Italian but in very restricted contexts:4

3 Other oblique non-clitic pronouns could appear in these positions, like the dative third singular lui ‘to. him’ and the dative first plural noi ‘to.us’:

a. k’è quella ke noi demo lui la metade that is that which we gave him the half ‘That is the thing of which we gave him half.’

(OFlor.) (; Ricordi di compere, p. )

b. . . . sì nne deono dare noi quello che lloro piacie  of.it must give. us that that to.them pleases ‘and they have to give us how much they like’ 4

(OFlor.) (; Bene Bencivenni)

Modern Italian marginally allows pre-T dative loro under complementizers with some verbs like piacere ‘to like’ or occorrere ‘to be necessary’, where loro encodes the higher argument. These occurrences can be analysed either as dative ‘quirky’ subjects (Manzini ) or as vestigial Stylistic Fronting surviving only in certain embedded contexts (Cardinaletti ; notice that in these cases loro occupies a peripheral TP position).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



() a. . . . e loro ha donato podere delli altri giudicare (OFlor.) . . . and to.them has given power of.the others judge. ‘ . . . and he has given them the power to judge others.’ (; Zucchero Bencivenni, Esposizione del Paternostro, ) b. . . . da che lor non piacessi . . . (OFlor.) . . . from that to.them not please. ‘ . . . since they do not like you . . . ’ (; Brunetto Latini, Tesoretto, v. ) Given that Old Romance languages had a V grammar, cases of pre-T loro—like those in ()—could be analysed as instances of a non-clitic pronoun moved to the left periphery in a sort of V-configuration (for Old Italian in particular see Benincà , Poletto ; for Old Romance see among many others Benincà ; Adams  and Wolfe ). Notice that though (b) is an embedded clause, it cannot be excluded that it exhibits V (see Benincà  for more detailed discussion about V effects in Old Italian). These distributional properties seem to suggest that Old Italian loro was a weak pronoun exactly as its Modern Italian counterpart, and this categorization can be further supported by the fact that pre-T loro usually appears before negation (cf. b).5 As for the exact positions occupied by loro in the Old Italian sentence structure, we assume as in Garzonio and Rossi () that cases like () and () are instances of the same property, that is, cases of V in the higher and the lower periphery (see Poletto

5

Cardinaletti () capitalizes on the relative order of the pronoun with respect to negation to distinguish clitics and weak pronouns: as normally only clitics can intervene between negation and the tensed verb, pronouns appearing before negation should be weak. Notice moreover that the order between negator and clitics is not subject to variation in Old Italian, as object clitics always follow negation (see Cardinaletti :  and the discussion of her example (a)). An anonymous reviewer points out that the position with respect to negation might not be a reliable diagnostic for the weak vs clitic distinction, as there are some (rare) cases of object/oblique me ‘me’ and te ‘you’ after the negator in the OVI database: a. / Ch’ io non me reche a mente quant’ io sono / (OFlor.) that I not me bring to mind how.much I am ‘ . . . / so that I am not reminded how much I am / . . . ’ (; Petrarch, Rime disperse e attribuite ..) b. /ingiuriando sì non te simiglio / cursing so not you resemble. ‘ . . . / cursing this way I do not resemble you / . . . ’

(OTosc.) (; Guittone d’Arezzo, Rime, sonnet )

It should be remembered however, as pointed out in Cardinaletti (: ), that in some cases it is not possible to identify whether me and te are weak or clitic, as there are alternative clitic forms ending in ‑e in Old Italian (me and te are indeed the first and second singular object clitic forms of the vast majority of the Italo-Romance varieties). Thus, the cases above are clitics ending in ‑e in their usual post-negator position, a claim which finds further support in the metrical structure of the hendecasyllable in (a), which requires that me be unstressed. Thus, as the pronouns in the examples above could be analysed as clitics, we maintain that the position with respect to negation is a good indicator of a pronoun’s status: if a pronoun appears after negation it should be viewed more as clitic than as a weak item.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

 on the idea that V and OV are to be analysed as the same property in parallel phases, i.e. movement to an OperatorP position). () a. [TP potea [TopicP [OperatorP lor [Op° fare [vP Dio fare loro]]]]] b. [TopicP [OperatorP loro [Op° ha [TP ha [vP donato loro podere]]]]] In () loro occupies Spec,Op, a position in both the high and low left periphery arguably related to Focus (i.e. FocusP in Rizzi  and Belletti  respectively). We exclude that pre-T loro can appear higher than OperatorP, i.e. in TopicP, as orders like dative loro + XP + Vfinite are virtually absent from the texts in the OVI database.6 Moreover, dative loro is never followed by a finite verb with enclitics: if, following Benincà (), Tobler-Mussafia cases—enclisis on finite verbs in root sentences—are to be analysed as movement of V to a C head in the Topic field, the impossibility of loro’s occurring before them indicates that pre-T loro is restricted to lower CP positions. This is again confirmed by cases like (), in which pre-T loro occurs before the tensed verb but is preceded by another constituent: () a. E rei] operator[loro] disse . . . (Old Pisan) topic[lo and the king to.them said. ‘And the king told them . . . ’ (; Storia di Barlaam e Iosafas, ch. ) b. E topic[‘l nostro singnore] operator[loro] disse: Andate (OFlor.) and the our lord to.them said. go.. ‘And our Lord told them: ‘Go!’’ (; Libro del difenditore della pace, ch. , par. ) c. E topic [così bell’e savie e virtuose parole (Old Aretino) and so nice and wise and virtuous words de conforto] operator[loro] disse of comfort to.them said. ‘And he proffered such nice and wise and virtuous comforting words to them . . . ’ (; Conti di Antichi Cavalieri ) All the cases of dative loro considered so far can be analysed as instances of a wellbehaved weak pronoun, given that loro occupies a dedicated specifier position in both peripheries. As expected under Cardinaletti and Starke’s () analysis of weak items, this entails that dative loro is an XP. However, on a par with most of the present-day Italo-Romance varieties, Modern Italian does not have a clitic form derived etymologically from Latin illorum. Such a form was not attested even in Old Italian, but can be found in some fourteenthcentury Sienese and Aretino texts. In Old Sienese, for instance, the pl dative form lo’ 6

The OVI database is a searchable corpus of Old Italo-Romance texts developed by the Opera del Vocabolario Italiano institute of the CNR (the Italian National Centre for Scientific Research). It can be accessed at: http://gattoweb.ovi.cnr.it/(S(ivevccprkioemxoxba))/CatForm.aspx.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



had a clear clitic distribution, as shown by Egerland (; we refer to this study and to Egerland  for a more detailed description and discussion of lo’): i) Old Sienese lo’ distributes like Modern Italian clitics, appearing proclitically on finite verbs and enclitically on non-finite verbs, (); ii) it is subject to the Tobler-Mussafia Law and follows negation, cf. (a) and (b) respectively; iii) it forms clusters with other clitics in the Modern Italian order dative > accusative, cf. ().7 () a. Allora lo’ donò una pietra di grossezza d’ (OSien.) then to.them gave. a stone of size of una nocella . . . a nut ‘Then he gave them a gem as big as a nut . . . ’ (; Fatti di Cesare (I), book , ch. , p. ) b. Partissi da’ frati senza fare lo’ motto (OSien.) left= from friars without do. to.them word ‘He left the monks without a word.’ (; Giovanni Colombini, Lettere, ) () a. Iddio lo’ dia più conoscimento che non ànno; (OSien.) God to.them give more knowledge that not have. ò lo’ scritto . . . have. them written . . . ‘May God give them more sense than they have; I have written to them . . . ’ (; Giovanni Colombini, Lettere, , p. ) b. Sì no lo’ feci mai nullo ladio, né eglino (OSien.) So not to.them did. never no harm, nor they a me . . . to me ‘I never did any harm to them, nor they to me . . . ’ (, Binduccio dello Scelto, La Storia di Troia, ch. )

7 Further supporting evidence of the clitic nature of Old Sienese lo’ comes from cases where it appears as l’ before tensed verbs and auxiliaries beginning with a vowel, cf. (i), and also from one case in which lo’ seems to double a dative DP, cf. (ii):

(i) Sì l’ andò a la ‘ncontra e combatté contra loro molto vigorosamente (OSien.) so to.them went to the towards and fought against them very vigorously ‘He went towards them and fought them very vigorously.’ (; Binduccio dello Scelto, La Storia di Troia, ch. ) (ii) la matina vegniente lo’ farebe a tutt’ e quattro mozare la testa. (OSien.) the morning following to.them would.make to all and four sever. the head ‘The following morning he would have the heads of the four of them severed.’ (; Cronaca senese, p. )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

() e coloro che mandati vi furo la chiesero a (OSien.) and those that sent there were it asked to lo re Priamo, ed egli lo’ la donò the king P. and he to.them it gave ‘And those who were sent there, asked king Priamus for it, and he gave it to them.’ (; Binduccio dello Scelto, La Storia di Troia, ch. ) The development of a clitic form from Latin illorum is however very exceptional: not only do present-day Italo-Romance varieties not have any such form (with the exception of Friulian, see also Loporcaro ), but also Old Sienese lo’ and Old Aretino ‘ro disappeared quite early from the system. The typical ItaloRomance paradigm for the third dative clitics levels the opposition between plural and singular forms: in Old and Modern Florentine (hence in Modern Italian), the pl dative clitic is gli ‘to him, to them’, while in other varieties the third-person dative clitic is the same as the locative clitic. Similar cases are already found in the earliest attestations of the Italo-Romance vernaculars like Old Venetian, Old Paduan, or Old Veronese, which do not seem to have clear cases of dative loro, and already present the dative clitic ghe/ge with no gender/ number specifications: () Alguni no ge piaxe questo muo’ de bruxa(r)la, . . . (Old Paduan) some not to.them pleases this manner of burn.=it ‘Some do not like this manner of burning it, . . . ’ (; El libro agregà de Serapiom, ch. ) In conclusion, these data lend support to the formulation of the diachronic cline strong > weak > clitic based on the Choice Principle: while Latin illorum had reached the final stage of the development into a clitic in Old Sienese8 (and Old Aretino), in other Old Tuscan varieties the development was still at the weak stage of the cline. Thus, Modern Italian loro is a ‘relic’ precisely in the sense that it has remained in the system, although the system also had the invariable clitic form gli for third plural datives—a form which is now the by far more common pl dative pronoun in colloquial Modern Italian.

8 Notice that Old Sienese texts also present cases of dative non-reduced loro or partially reduced lor. Interestingly however, these forms do not seem to distribute as well-behaved weak pronouns in that they appear after negation and between the sì particle and the tensed verb. As we will see, similar cases are found also in other Old Tuscan varieties, which we discuss in section ..:

e sì che, se non lor tiene danno, sì lor dà paura (OSien.) and so that if not to.them holds damage so to.them gives fear ‘so that, even if it does not harm them, it frightens them’ (; Reggimento de’ principi di Egidio Romano, ..)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



In the next section, we discuss cases which do not fit well into the proposed diachronic cline, suggesting thus that the distinction into three classes has to be readjusted.

. The ‘short-blanket’ problem of pronoun categories .. Strong-like distribution One of the major characteristics distinguishing Modern Italian weak dative loro from its strong counterpart, i.e. a loro, is the fact that weak loro must appear before direct objects and other low-IP adverbs. However, as Cardinaletti () notices, in some Old Tuscan texts loro can indeed be found after direct objects, postverbal subjects, and low-IP adverbs:9 () a. Allora dissi parole] loro . . . (OFlor.) OBJ[queste then said. these words to.them ‘Then I spoke these words to them . . . ’ (; Dante Alighieri, Vita Nuova, ch. , par. –) b. Molte terre donò loro. (OSien.) SBJ[Cesare] many lands gave. Caesar to.them ‘Caesar gave them much land.’ (; Fatti di Cesare, b. , ch. ) c. . . . e dirai loro . . . (OFlor.) ADV[così] . . . and will.say. so to.them ‘ . . . and you will say this to them . . . ’ (; Novellino, ) In these cases, loro seems to be capable of conveying dative case irrespective of its position within the sentence. In other words, its dative interpretation is intrinsic to its morphology and by this token, it does not have to appear in a specific dative-case assigning position, as for instance ApplicativeP or Cardinaletti and Starke’s () AgrP. Hence, in cases like (), loro behaves like a strong item. .. Clitic-like distribution Old Tuscan varieties also present instances of dative loro which cannot be analysed as weak since loro appears in positions where only clitics are allowed. Two clear cases come

9

An anonymous reviewer points out that in () loro could be the weak form in its usual position, while the preceding XPs have been scrambled over it. Although this may be an alternative account for these cases, notice that there also cases like below where loro appears post-participially and after the postverbal subject. ha mostrato Cristo loro [CP come . . . ] has shown Christ to.them how ‘Jesus Christ has shown them how . . . ’

(OFlor.) (; Giovanni dalle Celle, Lettere, )

In such a case the scrambling analysis does not entail that loro has reached the ‘weak’ position above vP.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

from dative loro appearing either between the preverbal negator no(n) and the tensed verb, or between the Focus particle sì ‘so’10 (Benincà ) and the tensed verb: () a. se voi non loro lo date (OFlor.) if you not to.them it. give ‘if you do not concede it to them’ (, Deca prima di Tito Livio Volgarizzata, b. , ch. ) b. però no·lloro avenrà punto dispetto (OFlor.) for this reason not to.them will.happen no contempt né orgholglio . . . nor pride ‘for this reason no contempt or disdain will be directed towards them . . . ’ (; Libro del difenditore della pace, d. , ch. , par. ) () a. Sì loro avviene come per ammonestamento di natura, che . . . so to.them happens how by lesson of nature, that . . . ‘It occurs to them, as nature dictates, that . . . ’ (; Tesoro di Brunetto Latini volgarizzato, b. , ch. ) b. sì lloro rienpie le ciervella. (OFlor.) so to.them fills the brains ‘It fills their brain.’ (; Zucchero Bencivenni, Santà del corpo, pt. , ch. ) Cases like () are particularly interesting, since Cardinaletti () distinguishes homophonous object clitics from weak pronouns (like te ‘you’) precisely on the basis of their relative ordering with the preverbal negator: Old Italian object clitics follow negation, while weak object pronouns precede it (see also Garzonio and Rossi ). Notice, moreover, that there are some examples of pre-T dative loro forming clusters with other clitics as in (a)—a possibility only available to object clitics and the preverbal negation in most (Italo-)Romance varieties, both old and modern. These instances of clitic-like dative loro pose a serious challenge to a classification of pronouns into just three classes: in order to capture these data, it is necessary to postulate that the diachronic cline from weak to clitic envisaged some stages in which pronouns were somewhere in between the two categories. This is tantamount to postulating that there are more than two degrees of structural deficiency, which clearly is not a welcome result, as one could claim that each single distributional,

10 Following Benincà (), we assume that the Focus particle sì ‘so’ occupies a very low specifier position in the Left Periphery (FocusP) and always triggers proclisis on the tensed verb which is assumed to have moved to Foc°. Thus, only clitic elements forming the complex head with the verb, like the preverbal negator and unstressed pronouns, can intervene between sì and the verb.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



morphosyntactic, semantic etc. property of pronominal items can in principle identify a different level of deficiency. If this were true, we should expect a much wider range of variation in both the synchronic distribution and the diachronic development of these forms. More in general, however, the observed cline is systematically from XP to X°, that is, irrespective of the different properties displayed by these items, the diachronic change follows the Head Preference Principle. In the next section, we consider the internal structure of pronominal elements with the intent of determining the factors at play in the observed differences between loro as XP and loro as a head. It will also be shown that the situation emerging in Old Tuscan can be accounted for by postulating an interplay between the functional projections within pronouns and the V syntax of these varieties.

. The internal structure of pronouns: deficient vs non-deficient In the previous sections we have shown that in Old Tuscan texts we find instances of dative loro displaying the properties of strong, weak, and clitic pronouns according to the classification of Cardinaletti and Starke (). One way to interpret this highly unstable behaviour is to assume that the varieties exemplified by those texts have minimally different types of loro. In other words, the typology of deficient pronouns might be richer than traditionally assumed. In particular, there are two types of loro which are problematic when assuming that Old Tuscan dative loro was already a weak pronoun like its Modern Italian counterpart (Cardinaletti ): a) the strong-like loro that can be separated from the verb by a direct object or a postverbal subject (and that can be moved to the left periphery); and b) the clitic-like loro that can appear between negation or the particle sì and the tensed verb. We analyse these two ‘deviant’ loros in different ways. It is however important to stress that the richer typology of forms we encounter in Old Tuscan texts does not invalidate the general idea behind Cardinaletti and Starke (): the account we propose seeks to reconcile their taxonomy with the historically attested gradual cline, in a structurally principled manner. Starting from the strong-like loro, we propose this is indeed to be treated as a strong pronoun, similarly to its Latin equivalent. We assume that, from the point of view of morphosyntax, these items have the complete internal structure represented in (). The only difference between strong pronouns and lexical nouns is the absence of the most deeply embedded NP layer: pronouns do not have an internal restrictor, but are interpreted according to the external context. The projections we represent in () are functional positions where features like Number (IdentificationP) or Person (DeixP) are encoded (in a sense these features are the equivalent of a semantic restriction for pronouns).11 We maintain, however, that strong pronouns 11 Under this view, the obligatory interpretation as [+animate] of strong non-oblique loro in Modern Italian (as discussed by Cardinaletti and Starke ) can be accounted for as evidence for the encoding of

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

still have a lexical layer, represented as nP (see Leu ),12 for reasons we address in the following discussion. () [DP D° [DeixP [IdentificationP [nP illorum [NP Restrictor]]]]]13 (strong illorum) Given () as the basic structure for strong pronouns, the differences between the various types of structurally deficient loro—and consequently between the various degrees of structural deficiency in general—can be captured in terms of upward movement/lexicalization and progressive deletion of the lower projections. In other words, we claim that complex cycles, like the one affecting pronouns in the history of Italo-Romance, are based on two mechanisms: (i) reanalysis as higher items in the functional spine (Roberts and Roussou ); and (ii) progressive loss of the inner layers from the bottom up. In this sense, the difference between strong-like loro and well-behaved weak loro is analysed as the loss of the inner lexical nP layer, after loro has moved (or is directly merged) in a higher functional projection of the internal spine, as represented in (): () [DP D° [DeixP [IdentificationP loro [nP [NP Restrictor]]]]] (well-behaved loro) The main different between our approach and Cardinaletti and Starke’s boils down to the fact that the structure of deficient items is pruned from the bottom up and not from the top down. A straightforward consequence of such an approach is that functional items in general should always be slightly more structurally deficient than lexical items. Both types of pronouns, strong loro and well-behaved weak loro, move as XPs and as such can target different XP positions within the sentence. However, weak loro cannot appear in its base-generated position, unlike strong pronouns, and must move to some dedicated Operator position, either to a ‘Differential Object Marking’-like projection before direct objects (ApplicativeP), or to a projection in the lower portion of the CP (see section ..). Put differently, we claim that only items that lack the nP layer in their internal structure, i.e., structurally deficient elements, can move out of the vP layer and can only target specific dedicated projections, an idea in line with a recent proposal on bare items like the Old Italian universal quantifier tutto ‘every, everything’ which displays a behaviour similar to weak loro

the [animate/human] feature at the nP level. Once the nP layer is lost, as we propose for deficient elements, the [+ / animate] distinction is no longer functionally encoded. Tentatively, this account also establishes a parallel between the verbal and the nominal domain as [+agentive] and [+animate] are encoded at the upper edge of the lexical layer (vP and nP respectively). 12 In the remainder of the chapter we adapt Leu’s () proposal in a more articulated DP structure. In particular, we label Leu’s IPR-R(estrictor) as nP, while the empty category N is taken to be the lexical restrictor in the most deeply embedded lexical layer NP. 13 Notice that in our notation the barred portion of the structure indicates not that it is moved or cancelled, but simply that that portion is not there.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



(Garzonio and Poletto ).14 Hence, structurally deficient elements have no real lexical layer and contain only functional projections in their spine. We take this to be the necessary starting point for further grammaticalization and reanalysis into genuine clitic forms, with consequent morphophonological reductions and additional morphosyntactic restrictions. Closely connected to this is also the occurrence of preverbal loro just described: it is well known that V syntax targets XPs and that weak pronouns are subject to movement to the C-layer (and can satisfy V) in V languages. Thus, we take all the cases of pre-T loro as analysed in Garzonio and Rossi () to be the most favourable context for the potential later development of weak items into clitics. The clitic-like loro cases seem to show this incipient development yet, they also pose a different problem, as we observe a bisyllabic unreduced pronoun that can occupy a position where only clitics and clitic clusters can appear. We take such a distributional possibility to indicate that clitic-like loro was indeed a clitic, i.e. an X°, yet a complex one. In more general terms, we claim that this is the exact step in which the Head Preference Principle applies, turning the basic morphological components of dative loro into distinct heads. () a. [DP D° [DeixP Deix° [IdentificationP Ident° lo-ro [nP [NP Restrictor]]]]]

(HPP)

b. [DP D° [DeixP Deix° lo- [IdentificationP Ident° lo-ro [nP [NP Restrictor]]]]] (reanalysis) More precisely, we propose that reanalysis as upward movement applies, and loro should move from IdentificationP to DeixP, but the Head Preference Principle intervenes, turning the XP loro in IdentificationP into a complex head, lo-ro in Ident°. Clitic-like loro is in a sense interpreted as a sort of cluster, made up of a deictic component and a plural component: the first morphological part, lo‑, is reanalysed as Deix°, as it carries third-person features (and is recognized as part of the paradigm of all other third-person object clitics in Old Tuscan, and Italo-Romance in general), while ‑ro remains in IdentificationP, encoding the Number feature. It can be assumed that from this level of deficiency onwards the functional item can be moved as a head in the sentential structure. We claim that the HPP can apply at this point precisely because the internal structure lacks the lexical layer nP–NP altogether.

14 Garzonio and Poletto () show that the bare quantifier tutto ‘everything’ is obligatorily extracted from the vP and moved to an IP adverbial position, while when tutto is paired with an NP it can remain in its argumental position. Thus, movement outside the vP is linked to the absence of the lexical layer of pronouns. This entails that bare tutto can be analysed as a weak element, as in French. Notice that bare tutto has the same distribution even when it appears under a functional preposition. Thus, the labelling (DP vs PP) is not the trigger of this syntactic behaviour. The case of loro shows that the reverse is also true: different syntactic behaviour obtain even with the same phasal label (DP). Interestingly, tutto, like molto ‘much, many’, is not specified for [animacy], while the negative quantifier niente is specified for [–animacy] and can appear inside the vP (and thus, according to the present analysis, it is a strong pronoun).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Jacopo Garzonio and Silvia Rossi

Some supporting evidence in favour of such a ‘cycle’ comes, we suggest, from the uncommon cases of pl dative clitics derived historically from Lat. illorum. As mentioned earlier, such forms are extremely rare in Italo-Romance and are found only in Old Sienese lo’, Old Aretino ‘ro and present-day Friulian lur. What we see here are different instantiations of the last steps in the grammaticalization path outlined in (): Old Sienese and Old Aretino lexicalize just one of the two heads, and then subsequently move these items to D° as proper clitics,15 while Friulian maintains the complex head and moves it to D° (notice moreover that Friulian lur is an unstressed monosyllable, falling in line with the typical morphophonological constraints of Romance clitics). ()

Old Sienese lo’: a. [DP D° [DeixP Deix° lo’ [IdentificationP Ident° [nP [NP Restrictor]]]]] b. [DP D° lo’ [DeixP Deix° lo’ [IdentificationP Ident° [nP [NP Restrictor]]]]]

()

Old Aretino ‘ro: a. [DP D° [DeixP Deix° [IdentificationP Ident° -ro [nP [NP Restrictor]]]]] b. [DP D° -ro [DeixP Deix° -ro[IdentificationP Ident° -ro [nP [NP Restrictor]]]]]

()

Friulian lur: a. [DP D° [DeixP Deix° [IdentificationP Ident° lur [nP [NP Restrictor]]]]] b. [DP D° lur [DeixP Deix° lur [IdentificationP Ident° lur [nP [NP Restrictor]]]]]

Most Italo-Romance varieties, however, do not maintain such forms and in general, do not present a specialized clitic for pl datives. We interpret such a fact as a further loss of functional structure in the sense that IndentifcationP is no longer projected. In other words, extremely structurally deficient dative pronouns are syncretic: when IdentificationP is no longer in the structure, in Modern Italian the favoured form is gli, which is invariable (not specified) for number (and in colloquial Italian also for gender): () [DP D° gli [DeixP [IdentificationP [nP [NP Restrictor]]]]]

15 In () we maintain that IdentificationP is not dropped from the functional sequence, because Old Sienese lo’ is specialized for pl dative, even though it is homophonous with the sg masculine object clitic. Such an analysis entails that Old Sienese lo’, although exhibiting all the distinctive properties of clitics, should be ‘less’ clitic. We suggest that this is the case as it displayed one striking non-clitic property, namely the possibility of violating the PCC, witness the fact that it can be in a cluster with first/second accusative clitics (Egerland ):

Cristo mai non me lo’ parta dall’ anima. (Old Sienese) Christ never not me to.them divide from.the soul ‘May Christ never take me away from their soul.’ (; Giovanni Colombini, Lettere, ) On the contrary, the invariable dative clitic gli of (colloquial) Modern Italian, which is unspecified for both gender and most importantly number, cannot violate the PCC.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Weak elements in cycles



In more general terms, the account we propose here seeks not only to refine the typology of structural deficiency, but also to establish a relation between the increasing loss of the lower part of the internal morphosyntactic structure of pronouns and syncretism, i.e. levelling of pronominal paradigms.

. Conclusion In the preceding sections we started from Cardinalettti and Starke’s () proposal on the tripartite typology of functional categories as the manifestation of increasing degrees of structural deficiency. The basic tenets are maintained: pronominal forms come synchronically and diachronically in different flavours. Yet, these ‘flavours’ might be more than the canonical strong-weak-clitic. We have claimed in particular that a typology of Italian pl dative pronouns should be characterized in a slightly more fine-grained way, in the sense that there can be more than two levels of deficiency, and that we need a rich internal structure of pronouns in order to capture the crosslinguistic and diachronic variation regarding the clitic/weak divide. Thus, what we observe in the historical development of Italian (and of other Old Tuscan varieties) cannot be captured as a two‑ or three-step cycle, but as a complex one, in which the Head Preference Principle (van Gelderen c) interacts with upward reanalysis. Some of the distributional properties observed in Cardinaletti () and in this chapter, we have argued, are linked to the internal structure of pronouns. Crucially, however, they may vary across languages, as the internal makeup of these items makes them more or less subject to certain sentence-level phenomena. This is the case we argue for Modern Italian, ‘well-behaved’ weak loro, which is extracted from the vP layer into a dedicated projection but cannot reach the left periphery of the sentence, since Italian has long lost V syntax. As a final remark, the account proposed here could possibly be extended to other well-studied linguistic cycles, and in particular, it could also shed light into the inner workings of the cycle turning original demonstratives into third-person deictic pronouns, which in Roberts and Roussou () was analysed as a case of upward reanalysis.

Acknowledgements We would like to thank Anna Cardinaletti, M. Rita Manzini, Diego Pescarini, Cecilia Poletto, Laura Vanelli for comments and suggestions on previous versions of this chapter. Many thanks also to three anonymous reviewers and to the audiences at DiGS  (Ghent,  June– July ) and the th Cambridge Italian Dialect Syntax-Morphology Meeting (CIDSM , Vienna, – July ). Though this chapter is a joint work by both authors, Jacopo Garzonio takes responsibility for sections ., ., and ., Silvia Rossi for sections . and ..

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

6 On the emergence of personal articles in the history of Catalan J U D Y B E R N S T E I N, F R A N C I S C O OR D Ó Ñ E Z , AND FRANCESC ROCA

. Introduction In this chapter we analyse the properties and evolution of personal articles in Catalan. This language has a personal article en/na that is used exclusively to introduce [+human] proper names. From a diachronic point of view, this Catalan article evolved from Latin dominus ‘master’, ‘ruler’, ‘lord’. Other descendants of dominus are displayed across the Romance languages. Among these are honorific titles like don/doña in Spanish (and also similar forms in Catalan and Portuguese), forms we consider to be an intermediate step on the path towards the more grammaticalized personal article. The path we characterize involves a change from an honorific title XP to a nominal classifier X0 that encodes familiarity. The process can be schematized as follows: ()

Title (Latin dominus) XP ! Classifier—honorific (Spanish don/doña) X0 ! Classifier—familiar (modern Catalan en/na) X0

In order to understand how this evolution took place, we compare the values and uses of the personal article en/na with those of honorific don/doña in present-day Spanish and in several stages of Catalan. We will conclude that the personal article is a grammaticalized element that should be analysed as an affixal syntactic head within the functional structure of the DP. This is in accordance with the view of grammaticalization as a process that involves category shifting and semantic, phonological, and morphological weakening, and that can be expressed syntactically as the reanalysis of a lexical head as a functional head (see Lehmann , ; Roberts and Roussou , ). Our analysis is also in line with and draws support from van Gelderen’s () conception of the path of historical change, what she labels the ‘linguistic cycle’. Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Judy Bernstein, Francisco Ordóñez, and Francesc Roca . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



The chapter is organized as follows: in section . we provide an overview of the distribution of personal articles in modern Catalan and the analysis we have put forth in earlier work to account for them, comparing and contrasting these elements with both regular definite articles and honorific titles. In section . we study Latin dominus, the ancestor of Catalan personal articles. We contrast the Latin noun dominus, the descendants of which are still nouns, with the homophonous honorific title, the form that we will claim undergoes grammaticalization in the history of Catalan to yield the personal article. In section . we focus on changes that took place between Old Catalan and the modern variety and in section . we analyse the evolution of the personal article as a classic case of grammaticalization that involved changes in the phonological, morphological, syntactic, and semantic properties of the forms. We view these changes through the lens of van Gelderen’s () ‘linguistic cycle’. In section . we offer some concluding remarks.

. Personal articles in modern Catalan In modern Catalan proper names that denote [+human] referents are introduced by a specialized article en (m.) or na (f.), forms labelled ‘personal article’ in traditional grammars of Catalan:1 () a. En Pere ha arribat  Pere has arrived ‘Pere has arrived late.’

tard. late

b. Na Maria ha arribat  Maria has arrived ‘Maria has arrived late.’

tard. late

(Catalan)

According to Longobardi (), the specialized article en/na is an expletive that occupies the head D and is related to the proper name that heads the lower NP in a way parallel to how expletive there is related to the following NP in existential constructions like there are two books on the table. Thus, the presence of the personal article in D would block the syntactic movement of the noun to the higher determiner

1

The extent of the use of the personal article varies among modern dialects of Catalan. Only Balearic Catalan uses both forms (masculine and feminine) in a generalized way. In Central Catalan dialects there is an alternation between en/na and the definite article el/la (though the use of la is much more general than na). In present-day Valencian Catalan, proper names are not generally introduced by any article. See Badia i Margarit (), Brucart () or Institut d’Estudis Catalans () for a description of the forms and uses in modern Catalan. In this chapter we deal with the use of personal articles found in Balearic Catalan (masculine and feminine forms) and in Central Catalan (masculine form, mainly); in Bernstein, Ordóñez, and Roca () we offer a description and analysis of the use of articles in several varieties of Catalan.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

position. Unlike the typical case of N-to-D movement of bare proper names in Romance (), proper names in Catalan would remain in situ (): () [DP [D Pedroi] [NP [ti] ]] Pedro ‘Pedro arrived late.’ () [DP [D en] [NP Pere]]  Pere ‘Pere arrived late.’

llegó arrived va 

tarde. late

arribar arrive

tard. late

(Spanish)

(Catalan)

In Longobardi’s analysis, the personal article is characterized as a semantically empty expletive article that occupies the same D head as determiners like the true definite article. This definite article can introduce proper names in some varieties of Catalan as well as in colloquial varieties of other Romance languages:2 () a. el the

Pere Pere

(Central Catalan)

b. la the

Maria Maria

(Colloquial Spanish)

c. il the

Gianni Gianni

(Colloquial Italian)

In separate work (Bernstein, Ordóñez, and Roca ), we claim instead that the Catalan personal article en/na patterns with definite articles only in certain respects, and also shares characteristics of titles like Spanish don/doña, to which it is etymologically related. Regular definite articles differ from Catalan personal articles in several ways. First, personal articles cannot be pluralized but regular definite articles can: () a. Personal article en (m. sg.), *ens (m. pl.) na (f. sg.), *nes (f. pl.)

(Catalan)

b. Definite article el (m. sg.), els (m. pl.) la (f. sg.), les (f. pl.)

2 In this work we will not discuss the Central Catalan dialects that have entirely lost the use of the personal article and have kept the definite article (el Joan, la Maria). The definite article with [+human] proper names behaves like a definite article with common nouns in its morphology and syntax. However, it shares with the personal article the familiarity feature when the proper name is used referentially.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



Second, prenominal adjectives like propi ‘same/oneself ’ or mateix ‘same’ cannot intervene between the personal article and the noun, but they can appear between the regular definite article and the noun: () a. *en 

propi same

Pere Pere

(Catalan)

b. el propi professor the same professor ‘the same professor’ Third, personal articles (with null or overt proper names) cannot introduce relative clauses but regular definite articles can, as illustrated in () and (). Example (a) illustrates Pere preceded by en and (b) the deletion of Pere. Neither alternative can introduce a relative clause. In contrast, example (a) illustrates Pere preceded by a definite article and (b) the same definite article with deletion of Pere. Both alternatives in () yield grammatical results and neither article carries the interpretation of a personal article (see footnote ). () a. *en  b. *en 

Pere Pere Ø

que that que that

va  va 

arribar arrive arribar arrive

() a. el Pere que va arribar the Pere that  arrive ‘the Pere that arrived yesterday’

ahir yesterday

(Catalan)

ahir yesterday ahir yesterday

(Catalan)

b. el Ø que va arribar ahir the that  arrive yesterday ‘the one that arrived yesterday’ Although definite articles and personal articles differ in many respects, there is an interesting property that they share: neither definite articles nor personal articles appear with proper names in vocative contexts, as illustrated in (a,b). Instead, in vocative contexts the proper name must be bare, as illustrated in (c). () a. *En Pere! Pots venir? b. *El Pere! Pots venir? c. Pere! Pots venir? Pere can. come ‘Pere! Can you come?’

(Catalan)

We have presented several contrasts between Catalan personal articles and definite articles, which cast doubt on the idea that both elements are aligned as D heads. So rather than pursue that idea, we will show how Catalan personal articles display

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

parallelisms with honorific titles like Spanish don/doña ‘Mr/Ms’.3 This element introduces proper names that are inserted in argument position without the necessity for a preceding article (i.e. the usual pattern with proper names in Spanish): () a. Don Luis ha llegado Mr Luis has arrived ‘Mr Luis has arrived late.’

tarde. late

b. Doña María ha llegado Ms María has arrived ‘Ms María has arrived late.’

(Spanish)

tarde. late

As we saw for en/na in ()–(), the honorific don/doña cannot be pluralized (), cannot be followed by a prenominal adjective (), and cannot introduce a relative clause with or without an overt noun (). So, in these respects, honorific don/doña patterns like personal articles.4 () a. don (m. sg.), *dones (m. pl.) b. doña (f. sg.), *doñas (f. pl.)

(Spanish)

() *don Mr

(Spanish)

mismo same

() a. *don Mr b. *don Mr

Luis Luis Ø

Luis Luis que that que that

llegó arrived llegó arrived

ayer yesterday

(Spanish)

ayer yesterday

The morphological and syntactic patterns summarized above support the idea that Catalan personal articles share characteristics with don/doña. Examples () with don and (a) with en furthermore suggest that both types of elements are structurally

3 In previous stages of Catalan don/dona was used as an honorific. There are still some vestiges of this use in Balearic Catalan:

(i) i don Vicenç Calvet . . . solia aconsellar: . . . (Balearic Catalan) and Mr Vicenç Calvet used to advise ‘ . . . and Mr Vicenç Calvet used to advise: . . . ’ (L. Villalonga ( []), Mort de Dama, p. ) (ii) Des que dona Obdulia . . . declarà oficialment que . . . since that Ms Obdulia announced officially that ‘Since Ms Obdulia officially announced that . . . ’ (L. Villalonga ( []), Mort de Dama, p. ) In this work we analyse the Spanish forms because they are used more generally than their modern Catalan counterparts. 4 The form doña(s) is used as a noun with the meaning of ‘wife’, ‘woman’, or ‘lady’ in Dominican Spanish. This element can be preceded by a definite article or even a possessive pronoun (e.g. la doña, mi doña). The masculine form don can also be used as a noun (el don). These instances of don/doña do not correspond to an honorific title.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



closer to the noun than typical D elements, since intervening material is barred. The structure we propose for the Spanish and Catalan DP is the following: ()

[DP [D el/la] . . . [ClassP [Class en/na, don/doña] [NP N ] ] ]

The regular definite article occupies the highest functional head D, as is commonly assumed, whereas the personal article and the honorific title occupy a lower projection closer to the NP, one that we label Class(ifier)P. Support for the structure above comes from the fact that, at least for some speakers, definite articles and personal articles can co-occur with prenominal adjectives, as illustrated in ().5 () a. el propi en the same  ‘Pere himself ’

Pere Pere

(Catalan)

b. La mateixa Na Trialls manifesta que ‘ella aturàs poc the same  Trialls expresses that she stopped few en la casa . . . ’ in the house ‘The same Trialls expresses that she rarely stayed at home . . . ’ (E. Botinas, J. Cabaleiro, and Mª À. Duran (), Les beguines. La raó il·luminada per amor, p. ) These examples would not receive an obvious explanation under an account that treated definite and personal articles alike. Parallel to the facts in () with co-occurring definite articles and personal articles, and consistent with our proposed structure in (), definite articles and honorific titles may co-occur with prenominal adjectives in Spanish, as shown in (). Recall from () that adjectives cannot intervene between don/doña and the proper name. () el mismo don the same Mr ‘Mr Luis himself ’

Luis Luis

(Spanish)

5 We suspect that the variability in judgements across Catalan speakers and dialects also comes from the fact that in certain dialects the simultaneous use of two determiner-like positions for personal and nonpersonal article is marked. The acceptability is primarily a question of register. The use of the personal article is frequent only in the written register in Central Catalan dialects such as the ones spoken in Barcelona and Tarragona, and indeed () has the flavour of the written register. The important fact is that the occurrence of two instances of either the definite or personal article is prohibited, as shown in (i) and (ii). For speakers that allow only (iii), there might be obligatory deletion of the internal personal article indicating familiarity (thanks to a reviewer for pointing this out to us).

(i) *el petit el Pere (ii) *en petit en Pere (iii) el petit Pere Pons-Moll, Torres-Tamarit, and Mascaró () analyse the phonological conditions on the co-occurrence of the definite and personal articles in Minorcan Catalan.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

So although we find strong evidence for aligning personal articles and honorific titles, expressed in the structure in (), we note two important differences between them: a) personal articles have apparently undergone further phonological reduction to the point of functioning like affixes or clitics; b) personal articles encode familiarity and honorific titles only formality. From an etymological perspective it is relevant to point out that the personal article en/na and the honorific title don/doña, which we analyse as corresponding to the same syntactic head in (), have the same origin: both evolved from Latin dominus (see Corominas and Pascual –; Coromines –). This common origin is at the root of the fact that personal articles appear only with proper names of persons and underlies the [+animate] feature that we postulate for the modern Catalan and Spanish forms.6 In the next section, we focus on this Latin ancestor of the modern forms. Then, in section ., we will study Old Catalan and show that don/doña and en/na have followed the same trajectory and that the differences between them can be attributed to a further degree of grammaticalization in the case of the personal article.

. Origins: Latin dominus Both the honorific title don/doña and the personal article en/na descend from Latin dominus, which was a noun meaning ‘master, owner, lord, ruler’ and was also a term of address (see Caro Reina ). The following examples show several uses of the noun dominus heading a bare NP (a) and in combination with different kinds of noun modifiers in (b–d): possessive pronoun (b), postnominal adjective (c), and relative clause (d). () a. Non ferat si dominus veniat? (Latin) not carry.. if owner come.. ‘If the owner were to come, wouldn’t he (this owner) take it away?’ (Plautus, Rudens , III–II ) 6 The interesting prediction we make is that only names that refer to humans can grammaticalize into personal articles in natural languages. This seems to be true as far as we know. For instance, the word xu (m.) / xa (f.), related to Sardinian tsiu ‘uncle’ / tsia ‘aunt’, has become a personal article in Algherese Catalan (see Institut d’Estudis Catalans ). Similarly, names of anthropomorphized animals (such as famous cartoon characters and even pets) may appear with personal articles (and see Caro Reina :  for some other expanded uses, which lead him to propose that further grammaticalization of the forms has resulted in generalization to [–human] contexts). In this chapter, we have illustrated the tight relationship between the personal article and a proper name, and expressed this relationship via a specialized functional projection, ClassifierP. As one reviewer notes, the Class head is extremely limited, relevant only for [+human] proper names. In ongoing comparative work on Balearic Catalan, Sardinian, and Picard (Bernstein, Ordóñez, and Roca ), which extends the work on personal articles, we have aligned Catalan en/na with other functional elements best treated as definite articles, but definite articles located low in the structure, adjacent to a proper name or a common noun.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan b. Bene me admonet domina well me remind. wife ‘My wife reminds me well.’



mea. mine (Petronius, Satyricon .,  I)

c. ipse dominus dives, operis et laboris expers  master rich toil and work devoid ‘The rich master himself is without work or toil.’ (Plautus, Amphitruo , III–II ) d. et dominus qui nunc est and owner who now is. ‘and I know who the owner is now’

scio know. (Plautus, Rudens , III–II )

Dickey () notices that dominus is often used as a term of address, that is, as a vocative. The following example provides an instance of this use with the feminine form domina: () domina, certe embasicoetan iusseras dari. (Latin) madam. surely lewd.mug had.ordered. give.. ‘Madam, you surely had given orders that a lewd mug be presented to me.’ (Petronius, Satyricon .,  I) Vocative constructions are especially relevant for our purposes because they provide a productive context for the use of dominus as a title. The examples in () provide slightly later examples (from the second century  onwards) of dominus used as a term of address, in each instance displaying vocative case. () a. Sed tibi plane, Luci domine, soli omnium but you. really Lucius. master. only. of.all Chaldaeus ille vera dixerit . . . Chaldean that truth has.told.. ‘But I really hope, master Lucius, that the Chaldean has told the truth to you, if to no one else . . . ’ (Apuleius, Metamorphoses .,  II) b. . . . aut sub dextera tua, domine Constanti, or under right.hand your master. Constantius. Bataviam Britanniam=que . . . Batavia. Britannia.=and ‘ . . . or Batavia and Britannia under your right hand, master Constantius . . . ’ (Eumenius, Pro instaurandis scholis oratio .,  III) c. non unum, inquit, domine Auguste, sed primum. not one. he.said lord. Augustus. but first. ‘Not one , he said, my lord Augustus, but the first.’ (Sidonius Apollinaris, Epistulae .,  V)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

As the examples in () show, the form domine used in these vocative contexts can follow (see (a)) or precede (see (b,c)) the proper name (Luci, Constanti, Auguste) it is combined with. It is this early use of dominus as a title that especially interests us in our pursuit of the development of Catalan personal articles. In Late Latin, we also find dominus preceding other titles like rex ‘king’, regina ‘queen’, imperator ‘emperor’, and soror ‘sister’ (examples from digilibLT): () a. domine rex (Late Latin) master. king. ‘my lord king’ (Anonymous, Historia Apollonii regis Tyri ,  V–VI) b. domina regina madam. queen. ‘my lady queen’ (Anonymous, Historia Apollonii regis Tyri ,  V–VI) c. Domine imperator Valentiniane master. emperor. Valentinianus. ‘my lord emperor Valentianus’ (Symmachus, Relationes ,  IV) d. Domina soror madam. sister. ‘my lady sister’ (Anonymous, Passio Sanctarum Perpetuae et Felicitatis .,  III) We observe that the ordering of these titles in Late Latin does not resemble that of present-day Spanish and Catalan. In these languages the descendants of dominus (don, en) would follow, rather than precede, other titles: () a. el rey don the king  ‘king Carlos’

Carlos Carlos

b. la reina doña the queen  ‘queen Sofia’ () a. el rei en the king  ‘king Jaume’ b. la reina na the queen  ‘queen Maria’

(modern Spanish)

Sofia Sofia

Jaume Jaume

(modern Catalan)

Maria Maria

In sum, Latin dominus was a noun with lexical meaning and the regular distribution of nouns and, in addition, it was used as a term of address for people belonging to the nobility (e.g. domine Auguste). This vocative use explains its combination with

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



proper names. We consider these early titles in vocative contexts a first step towards the more regular use of dominus (and its descendants) with proper names in Late Latin. The word order in () suggests to us that when Latin dominus began to be merged as a title, it was merged higher than other titles such as ‘king’, ‘queen’, and ‘sister’. We represent the merger of dominus as a title within the nominal structure as follows: ()

[DP D [Title  dominus [Title  rex/regina/soror [NP N ] ]]]

With this analysis we are not suggesting that dominus (or rex, regina, etc.) has already been specialized (i.e. grammaticalized) as a particular head that expresses a hypothetical functional projection, TitleP. We rather see it as a sort of modifier of the noun, an XP that displays some phonological, morphological, and syntactic independence. In these respects, the Latin form is very different from Spanish don/doña and Catalan en/na, which cannot stand alone and can only precede proper names, never common nouns.7 Having depicted two instantiations of Latin dominus, one as noun and the other as title, we now want to continue our analysis of the title and dispense with the noun. What is our reasoning? Our goal in this chapter is to trace the development of Catalan en/na (and Spanish don/doña), forms that are almost certainly the descendants of the titles and not the nouns. How do we know this? For one thing, the true honorific titles no longer exist with their original forms or meanings, but vestiges of the earlier items can be found in their descendants in Spanish, Catalan, and other Romance languages (e.g. Spanish don/doña, Catalan en/na). The resulting forms, especially the personal article (en/na), have undergone an incremental and classic process of grammaticalization and now resemble functional elements rather than lexical ones (see Caro Reina ). On the other hand, nouns that descended from dominus can still be found in the contemporary varieties, though they too have undergone changes in form and meaning. Nevertheless, we cannot identify a parallel process of grammaticalization affecting the nouns, since they remain lexical items with full semantic content and greater morphological robustness (they may be pluralized etc.).8 Returning to our focus on Latin titles that were grammaticalized in daughter languages, we emphasize that there are important differences between dominus and 7 In Spanish it is possible to find don/doña preceding common nouns or adjectives: don perfecto ‘Mr perfect’, doña calores ‘Ms hot temper’. These constructions, considered colloquial by the Diccionario de la lengua española (DRAE), have an ironic value and their use and interpretation are similar to those of a proper name: they denote a particular referent. In present-day Catalan a parallel use of en/na does not exist. In this language ironic constructions like the Spanish ones may be constructed with forms like don or madò, which are documented in previous stages of the language: Don perfecte ‘Mr perfect’, Madò remugues ‘Ms grumbler’ (data from Balearic Catalan). 8 For example, present-day Spanish displays the noun dueño/dueña (‘owner’, ‘proprietor’), and Italian and Catalan display donna (‘woman’) and dona (‘woman’), respectively.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

the related forms don/doña in modern Spanish and en/na in modern Catalan. These differences are found in all levels of the grammar. Phonologically, don/doña and en/ na are reduced forms of dominus (a) and, in the case of the Catalan personal article there is, in addition, loss of the first phonological foot and loss of stress, which results in a clitic form: written next to a vowel-initial proper name in (b) and to the form ca ‘home’ or the preposition de in (c) (see Colomina i Castanyer  for the syllabification rules concerning the personal article and Mascaró  for the proclitic nature of this and other prenominal elements in Catalan). The Catalan form ca is reminiscent of French chez ‘home’, ‘place’; the construction with the personal article ca’n resembles the enclitic genitive’s in English. () a. dominus > ne > en/na dominus > don/doña

(Catalan) (Spanish)

b. n’Anna ‘Anna’ c. ca’n Pere, d’en Joan ‘Pere’s (place), of Joan’ Concerning morphology, dominus can be fully inflected for case, gender, and number, whereas, as we have shown above, Spanish don/doña and the Catalan personal article don’t admit number inflection: () domini (m. pl.) (Latin)

vs

*dones, *ens

(Spanish, Catalan)

The syntax of each element is also different. Leaving aside the use as a regular noun (impossible with en/na and restricted to certain uses of don and doña in some dialects of Spanish; see footnote ), and focusing on the forms used as titles, the modern Spanish and Catalan descendants observe strict adjacency to the proper name. In contrast, Latin dominus precedes other titles. The word order differences between the modern varieties and Latin are illustrated in (). () a. title + don/en + proper name b. domine + title + proper name

(Spanish, Catalan) (Latin)

Finally, from the semantic point of view the honorific nature of Latin dominus is lost in its Spanish and Catalan descendants. The Spanish form preserves a meaning related to respect or formality; the Catalan personal article expresses only familiarity. We schematize the nature of the semantic transition from honorific title to personal article as follows: () a. honorific title reserved for the nobility (Latin) ! b. honorific title expressing formality/politeness (Old Catalan, Old/modern Spanish) ! c. personal article expressing familiarity (modern Catalan)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



All of the stages we are proposing capture the evolution of Latin dominus to Spanish don/doña and Catalan en/na (and see Caro Reina :  for a compatible phonological reconstruction from en/na to dominus). There is a process of grammaticalization that spanned several centuries, eventually yielding the present-day personal article. In the following section we study in detail the forms and stages corresponding to this process in the history of Catalan.

. From Old Catalan to modern Catalan Old Catalan texts provide examples of the following sets of forms, all etymologically related to Latin dominus: a) the attested form domino/domina, which functions as a title; b) don/dom/dona/donya, which are close to the Spanish honorific title don/doña; and c) en/na, the personal article generally used in present-day Balearic Catalan. The form domino/domina was probably rare. We have found it in only a couple of texts corresponding to early Old Catalan of the thirteenth century:9 () a. avia escrit per manament de domina Guillelma has written by order of domina Guillelma ‘(s)he has written by order of lady Guillelma’ (Reconeixement de deutes a Pere Canet, XIII) b. Sanctissimo domino nostro Alexandro pape VI most.holy master ours Alexandro pope VI ‘our most holy master pope Alexander VI’ (Cartes triades dels Borja, XIII) The appearance of the possessive nostro between domino and the proper name in (b) contrasts with the patterns found in the descendants of dominus in Catalan and Spanish, where it must be adjacent to the proper name. This difference concerning word order can be attributed to the fact that this form is closer to the Latin one, which, as shown in (), could appear separated from the proper name. This is the only example we found where this element doesn’t immediately precede the proper name. The forms don/dom/dona/donya are attested in later Old Catalan, that is, in texts from the thirteenth century on: () a. la emperadriu dona the empress  ‘empress Constança’

Costança Costança (Desclot, Crònica del Rey en Pere, XIII)

9 Most of the examples we provide in this section were obtained from the Corpus Informatitzat del Català Antic—CICA (http://www.cica.cat). The corpus also provides information about the geographical location of the text, which is useful for tracing the evolution in each dialect. We don’t indicate such locations here because it is not crucial for the purposes of the current work. Here we simply indicate that in Old Catalan the personal article en/na is found in texts across the various regions, even in Valencian texts, where this article has been lost (see footnote ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca b. la reyna dona Johana the queen  Johana ‘queen Johana’ (Desclot, Crònica del Rey en Pere, XIII) c. dom Berenguer d’ Entença  Berenguer of Entença ‘dom Berenguer of Entença’ (Llibre dels fets del rei en Jaume, XIV) d. la senyora reyna donya Maria the madam queen  Maria ‘lady queen Maria’ (Estudi General Lleida del  al , XV)

In these examples the word order is ‘title(s) + dona/dom + proper name’, contrary to the one observed in Late Latin (example () above) and is the same as the one we see in modern Catalan and modern Spanish. This order is also manifested with the more reduced en/na, which reveals the emerging personal article: () a. El rey N’ the king  ‘king Amfos’

Amfos Amfos (Desclot, Crònica del Rey en Pere, XIII)

b. lo dit senyor infant En Pere the said lord prince  Pere ‘the aforementioned prince Pere’ (Crònica de Ramon Muntaner, XIV) c. E la dita na Johana portà=ho a ·n Vives and the said  Joana brought=it to  Vives ‘and Joana brought it to Vives’ (Clams e crims ciutat de València, XIII) These examples illustrate the process of grammaticalization that transformed dominus at several levels of the grammar: phonology (reduction from domino/domina to don/dona and to en/na), morphology (loss of case and number inflection), syntax (strict adjacency to the [+human] proper name) and semantics (no longer used exclusively with members of the nobility—(a,b) vs (c)—but rather as a general honorific). We take these changes to correspond to the syntactic merge of don/dona (and also en/na) in a position closer to the proper name than Latin dominus/domina: such a position is lower in the syntactic structure and would correspond to a functional projection we label Class(ifier). We hypothesize, then, that the descendant of dominus became a Class head that signifies respect and that was originally used to refer to people belonging to the nobility. Thus its distribution is determined by the semantic characterization of the noun that follows, as is typical of classifiers in natural language. As we saw for modern Catalan (recall ()), prenominal adjectives can intervene between definite article and personal article for some speakers, and this is also the case in Old Catalan:

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan () a. La honorable na Damaua the honorable  Damaua ‘the honorable Damaua’ b. La magnífica na the magnificent  ‘the magnificent Yolant’



(Manual del Consells de Gandia, XV)

Yolant Yolant (Manual del Consells de Gandia, XV)

c. La mateixa na Gojona the same  Gojona ‘the same Gojona’ (La llengua dels processos de crims a la Lleida del segle XVI, XVI) Although don/dona will eventually come to be associated with Spanish and en/na with Catalan, both elements are found in Old Catalan, even in the same text (e.g. Bernat Desclot), as in () vs () above. We take these patterns to indicate an intermediate stage of a change in progress, along the path from domino/domina towards en/na. With these considerations in mind, we represent the general structure of the Old Catalan DP as follows (with ClassP the projection corresponding to the grammaticalized forms don/dona, en/na): ()

[DP [D el/la] . . . [AP mateix, honorable . . . ] [TitleP rei/reina [ClassP don/dona, en/na [NP [N] ] ]]]

In fact, this structure strongly resembles the one we saw earlier in () for modern Catalan and Spanish, with the notable shared characteristic that prenominal adjectives may intervene between definite article and personal article (or honorific title in the case of Spanish). Next we consider some data from the thirteenth and fourteenth centuries. Interestingly, these examples involve two consecutive instantiations of descendants of Latin dominus. In the first case, (a), we see madona plus the honorific and in the second case, (b), madona plus the early personal article, na. () a. la dita madona dona Maria de Muntpesller the said madam  Maria de Montpellier ‘the aforementioned lady Maria of Montpellier’ (Crònica de Ramon Muntaner, XIV) b. madona Na Maria de madam  Maria de ‘lady Maria of Montpellier’

Monpeller Montpellier (Desclot, Crònica del Rey en Pere, XIII)

Examples like these suggest upper and lower merge sites for (ma)dona/na, the lower corresponding to the grammaticalized elements and the upper to the Title specifier schematized above in (). These examples again illustrate (recall ()–()) an

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

intermediate stage on the path from full XP status of the Latin term of address, dominus, to X0 clitic status of modern Catalan en/na. If we compare Old Catalan personal articles with the modern Catalan use, we can identify and explore two important semantic changes, one involving the nature of interpretive features and the other involving the predicate–argument distinction. Although Old Catalan don/dona and en/na have lost several characteristics of their Latin ancestor dominus/domina, they continue to be used in formal contexts with persons of the nobility, as the following examples show: () a. lo dit senyor infant En the said lord prince  ‘the aforementioned prince Pere’ b. lo rey En the king  ‘king Jaume’

Pere Pere (Crònica de Ramon Muntaner, XIV)

Jaume Jaume (Desclot, Crònica del Rey en Pere, XIII)

In modern Catalan the honorific/formality interpretation is absent and familiarity (with the referent) is the relevant interpretive feature (see Brucart ; Institut d’Estudis Catalans ). The contrast between the presence and absence of the personal article in varieties of modern Catalan illustrates the semantic role of the familiarity feature. Consider the following examples: () a. Chomsky Chomsky

ha has

publicat published

un an

altre other

b. En Chomsky ha publicat un  Chomsky has published an ‘Chomsky has published another book.’

llibre. book altre other

(modern Catalan) llibre. book

The semantic difference between (a) without a personal article and (b) with one is that the latter example expresses pragmatically the idea that the speaker is familiar with Noam Chomsky. Included is an implicature that the hearer is also familiar with him.10 This meaning is not the default for (a), though it is a possible interpretation. Modern Spanish don/doña, unlike modern Catalan en/na, encodes formality or politeness, and never familiarity. In this sense and several others, the modern Spanish forms are more conservative than the modern Catalan ones. Returning again to the Old Catalan forms, we do find cases of proper names that involve the nobility and yet lack the formal personal articles. In the examples in () from the thirteenth century, no honorific or personal article appears before the proper name Miramoli, even though this person was a king. But as a Moor, he was

10 The familiarity feature is adding to the reference of the proper noun. It is an affective marker on the part of the speaker and presupposes that the hearer is familiar with the referent. This should not be seen in terms of anaphoricity.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



apparently the wrong sort of king to merit the honorific title (compare (b) and the examples in (), all from the same document). () a. E Miramoli trames sos missatgers al rey de Castella and Miramoli sent his messenger to.the king of Castille e als tres reys de Spanya. and to.the three kings of Spain ‘And Miramoli sent his messenger to the king of Castille and to the three kings of Spain.’ (Desclot, Crònica del Rey en Pere, XIII) b. Aquest Miramoli se aparella que passas en Spanya. this Miramoli  decided that cross into Spain ‘Miramoli decided to go to Spain.’ (Desclot, Crònica del Rey en Pere, XIII) c. En la ciutat de Marochs havia hun Serray que havia nom in the city of Marocco had a Moor that had name Miramoli Miramoli ‘In the city of Marocco there was a Moor that was called Miramoli.’ (Desclot, Crònica del Rey en Pere, XIII) Although Miramoli lacks an honorific title in (), (b) illustrates that the proper name may be introduced by a demonstrative adjective. Further note that Miramoli is used as a predicate in (c), a topic we return to shortly. We find further evidence of change. As early as the fourteenth century, we observe en/na appearing with a list of proper names not corresponding to the nobility: () De Sent Lorenç: en Ramon de Soler, en Bernat Piquer, from St. Lorenç:  Ramon de Soler,  Bernat Piquer notari; en Pere Esteve notary;  Pere Esteve ‘from St. Lorenç: Ramon of Soler, Bernat Piquer the notary, Pere Esteve’ (Manual Consells València, XIV) In this example, which comes from official documents of the city council, it is reasonable to suppose that the personal articles convey respect via an honorific form. Although this function is no longer achieved via en/na in modern Catalan, it continues to be the interpretive effect of the present-day honorifics (don/doña) in modern Spanish. In the sixteenth century we observe an interesting example with na. The person described in (), Gojona, is neither a noblewoman or of high social status. Instead, she is a prisoner or criminal, and yet the proper name is introduced by na: () Dita na Gojona no tenie pa. said  Gojona not had bread ‘The aforementioned Gojona had no bread.’ (La llengua dels processos de crims a la Lleida del segle XVI)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

We conjecture that this is an early instance of what has developed into the personal article in present-day Catalan, one that encodes a familiarity feature. The change in interpretive features over time, from the Latin [+formal] honorific title dominus to the modern Catalan [+familiar] personal article en/na, follows the general direction of decreasing formality: honorific term of address (Latin) ! marker of respect (Spanish, Old Catalan) ! marker of familiarity (Catalan). This shift in interpretation correlates with the erosion of phonological and morphological properties. The end result in Catalan is a clitic form with [+familiar] interpretive effects.11 So far we have examined one of the semantic features that undergoes a change from Old Catalan to the contemporary variety. We turn now to a second semantic change, namely one involving the predicate–argument distinction. En/na and don/ dona in Old Catalan can be used with expressions like ‘call someone X’, ‘my name is X’, which involve the use of the proper name as a predicate (see Matushansky ): () a. qui havia nom N ’ who had name  ‘who was called Amfos’

Amfos Amfos (Desclot, Crònica del Rey en Pere, XIII)

b. qui havia nom En Ferrando who had name  Ferrando ‘who was called Ferrando’ (Desclot, Crònica del Rey en Pere, XIII) c. d’altre hac nom don Ferrando of.other has name  Ferrando ‘Another one was called don Ferrando.’ (Desclot, Crònica del Rey en Pere, XIII) And we saw in previous examples such as () that early Catalan en/na can also precede proper names that are arguments. Over the course of development of the modern personal articles, the ability to appear with predicates was lost, and the ability to appear with arguments was retained. The following example shows the impossibility of personal articles with predicates in present-day Catalan: () Em diuen (*en) me call.  ‘They call me Pere.’

Pere. Pere

(modern Catalan)

In contrast, the Spanish honorific title don/doña retains the possibility: () Me llaman don Pedro. me call.  Pedro ‘They call me Mr Pedro.’

11

(modern Spanish)

Although it is not addressed here, we investigate in separate work (Bernstein, Ordóñez, and Roca ) how the personal article has continued to erode. For example, it is optional in some contemporary varieties of Catalan and totally absent in others.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



The last change we discuss concerns vocatives. Recall that the Latin honorific term of address, dominus, appeared in vocative contexts (and displayed vocative case marking). In Old Catalan, which lacks morphological case, the emerging personal article continues to appear in vocative contexts, a property no longer possible in modern Catalan (but possible in modern Spanish with don/doña). Example () from the fifteenth century illustrates na in a vocative context. () Na Caterina, vos dieu  Caterina you say ‘Caterina, you say that . . . ’

que . . . that (Sant Vicent Ferrer, Sermons, XV)

An important observation about Old Catalan na in () is that it already ‘looks’ (phonologically, morphologically) like the modern form. However, it does not yet encode familiarity and it is still licit in vocative contexts.12 Also from the fifteenth century and on the topic of vocatives, we were surprised to find examples of na used with a common noun (in (a)) and with a (nominalized) attributive adjective (in (b)): () a. Na Donzella, aveus ací  maiden have. here ‘Maiden, here you have Alí Baxà’

Alí Alí

Baxà Baxà (Història de Jacob Xalabin, XV)

b. Ay, na beneyta— dix la duquessa— y com oh  silly said. the duchess and how sou tota plena de lealtea are. entire full of loyalty ‘Oh, silly, said the duchess, and how you are so loyal?’ (Tirant lo Blanc, XV) These examples indicate to us that the process of grammaticalization to personal article is not yet complete, because a) personal articles co-occur only with proper names in modern Catalan; and b) personal articles are no longer found in vocative contexts in modern Catalan.

. Evolution of the personal article and the ‘linguistic cycle’ The central aim of this chapter has been to trace the development from Latin (i.e. dominus) of the Catalan personal article en/na. In doing so, we have often contrasted the emerging Catalan form with its Spanish counterpart don/doña. What we have observed from this comparison is that, across linguistic domains, modern Spanish don/doña is consistently more conservative than modern Catalan en/na. We can 12

We don’t have any indication that Caterina was a noblewoman. However, on the path towards sainthood, she was probably a person of respect. Interestingly, within the same text, not every instance of this proper name was preceded with the emerging personal article.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

group the differences between the modern Spanish forms and the Catalan ones as follows: • phonology: the Spanish forms did not lose the first phonological foot; the Catalan forms did. • morphology: the Spanish forms are not affixal; the Catalan forms are. • syntax: the Spanish forms still appear in vocative contexts; the Catalan forms do not. • semantics: a) the Spanish forms are no longer reserved for the nobility and yet retain an honorific or formal interpretation; the Catalan forms contribute nothing more than familiarity; b) the Spanish forms appear with arguments and predicates; the Catalan forms appear only with arguments. This signals that the Catalan forms have surpassed the Spanish ones in terms of degree of grammaticalization. Focusing now on the Catalan forms, we observe some important semantic, syntactic, and phonological differences between Old Catalan and modern Catalan in the evolution of en/na: first, Old Catalan displays the honorific use, in don/donya and the early personal articles, that Spanish don/doña has retained. This indicates that there is a semantic change in which en/na loses its semantic properties as an honorific title and then comes to indicate familiarity in modern Catalan. These semantic changes are linked to the fact that its distribution becomes more restrictive. Thus while Old Catalan en/na could be used with common nouns, as in (a), and predicative contexts, as in (), this is no longer possible in modern Catalan. Also, contrary to modern Catalan, en/na in Old Catalan was used in vocatives. We hypothesize that the early honorific titles of Old Catalan are XPs and not heads. Catalan titles eventually became article-like elements in modern Catalan varieties prompted by the phonological changes that took place, such as the loss of a phonological foot in the original word dominus. Once dominus became an article, its distribution was restricted to argument positions. The historical changes we have traced in this chapter are consistent with those developed in van Gelderen’s () conception of the ‘linguistic cycle’, particularly as it applies to the development of definite articles from demonstratives in Romance. Demonstratives, which are phrasal (adjectival) elements and occupy specifier positions, develop first into X0 heads and eventually into (clitic) affixes in the form of definite articles. Our proposal for the development of Catalan personal articles fits neatly into this model: Old Catalan titles (domino, domina) are phrasal (adjectival) elements and occupy specifier positions. These titles develop first into X0 heads (don/ donya, don/dona) and eventually into pro‑ or enclitic affixes, the personal articles en/na we see in the modern variety. We saw above that along with the phonological, morphological, and syntactic changes, two semantic changes occur: a) the interpretive feature of the titles erodes and changes from one of formality to one of familiarity; b)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

The emergence of personal articles in Catalan



titles that could appear with predicates or arguments can now appear only with arguments. This evolution of Catalan titles from XP phrases to X0 personal articles coincides with van Gelderen’s (: , her ()) ‘Head Preference Principle’: ()

Head Preference Principle (HPP): Be a head rather than a phrase.

This diachronic change also obeys what van Gelderen (: , her ()) calls ‘feature economy’, schematized in (): a phrase in the specifier position changes to a head and then an affix. Under our account, a title is an XP that occupies a DPinternal specifier position and carries interpretable honorific features. It loses its interpretability status (iF) and becomes a head with uninterpretable features (uF). In the final stage, it becomes a (pro‑ or enclitic) affix. ()

Feature Economy Specifier ! Head ! affix ! iF honorific ! uF (marker[+human]) ! uF ([+familiar])

The novelty of the change is that this uninterpretable feature of Catalan en/na only affects a subset of nouns, namely [+human] proper names. We would like to relate this fact to its historical origin as a term of address for the nobility. It seems that the [+human] feature has characterized these forms at all stages of their evolution.

. Conclusion In this chapter we have studied the origins and subsequent evolution of personal articles in Catalan. In Late Latin up until the fifth century, dominus was used as a common noun and could also modify proper names, usually as a term of address. It started to be used primarily as an honorific title in Old Catalan. At that point, dominus had lost phonological material to become the monosyllabic en/na we find today. From Old Catalan to modern Catalan, en/na lost its honorific value and developed into an article preceding proper names and indicated familiarity. Its syntactic distribution became more restricted and it could no longer be used in predicative and vocative contexts. This is consistent with the evolution of definite and indefinite articles across Romance, including Catalan: the definite article developed from a demonstrative adjective and the indefinite article from the numeral ‘one’. Catalan is interesting in that it displays another source for articles, namely, titles. These in turn yielded personal articles, forms not found across Romance as far as we know. Moreover, the evolution of these articles is consistent with van Gelderen’s ‘linguistic cycle’: title of nobility (XP specifier), designation of respect (X0), personal article (clitic X0).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Bernstein, Ordóñez, and Roca

We have shown in our analysis that titles are a source for personal articles in Catalan. Questions raised by this analysis include whether titles are the only source for personal articles cross-linguistically and also whether other titles have undergone grammaticalization to yield personal articles. Some of these issues have been touched upon in Caro Reina’s () work and we plan to pursue them further in our own future work.

Acknowledgements We thank the audiences at the following conferences, where different aspects of this work have been presented: the th Diachronic Generative Syntax Conference (Ghent,  June– July, ), II GETEGRA International Workshop: Nominals (Recife, – March, ), th Colloquium on Generative Grammar (Cáceres, – April, ), rd Formal Ways of Analysing Variation (New York, – May, ), and the th Diachronic Generative Syntax Conference (Stellenbosch, – September, ). We also thank Neus Sànchez for her help with the Balearic Catalan data. Parts of this research have been funded by the following research grants to Francesc Roca:  SGR  (Generalitat de Catalunya), MPCUdG/ (Universitat de Girona), FFI--C--P (MINECO). We are grateful to three anonymous reviewers for their helpful comments.

Sources Corpus Informatitzat del Català Antic. . digilibLT. Digital library of late-antique Latin texts. . Digital Loeb Classical Library. . Muntaner, Ramon. Crònica catalana. [edited by A. Bofarull (), Barcelona]. Perseus under PhiloLogic. . RAE. Diccionario de la Lengua Española. .

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

7 Bare singular nouns in Middle Norwegian KARI KINN

. Introduction Old Norwegian (approx. –) allowed bare singular count nouns (BSNs) in contexts where they are no longer permitted in Modern (i.e. present-day) Norwegian.1 BSNs could be interpreted as indefinite, as illustrated in example (a), or definite, as in (b).2 ()

a. Rane het maðr Rani was.called man ‘There was a man called Rani.’ b. Dæyr nu konongr . . . dies now king ‘Now the king dies . . . ’

(The saga of St. Ólaf, )3

()

In Modern Norwegian, BSNs like those in () are not acceptable; a felicitous translation must include either an indefinite article or a suffixed definite article. Cf. ():

1 Old Norwegian refers to the language spoken in Norway in the period stated above (Bull et al. ). In the literature, Old Norwegian is not always distinguished from Old Icelandic; the two varieties are often collectively referred to as Old Norse (e.g. Nygaard ; Faarlund ). There are well-known phonological differences between Old Norwegian and Old Icelandic (see e.g. Hagland ); the extent to which they also differed syntactically is a topic that invites further research. 2 Mass nouns and plurals, which are regularly bare in Modern Norwegian and other Germanic languages, are not discussed in this chapter; see Carlsson (). 3 All Old Norwegian examples are from The saga of St. Ólaf (in Upps DG  II). The manuscript has been dated to the middle of the thirteenth century (there is some variation among scholars; see Seip , Johnsen  and Mundal ). The examples have been collected via the PROIEL corpus (http://foni. uio.no:/users/sign_in) and are rendered as they appear there. They are identified by their sentence ID in PROIEL.

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Kari Kinn . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

 ()

Kari Kinn a. Det var ein mann som there was a man who ‘There was a man called Rane.’

het was.called

Rane. Rane

b. No døyr kong-en. now dies king- ‘Now the king dies.’ The option of BSNs in Old Norwegian is well known (Falk and Torp ; Nygaard ; Dyvik ; Lander and Haegeman ; Börjars et al. ), and the use of fully grammaticalized articles in Modern Norwegian has also been thoroughly described (e.g. Faarlund et al. ; Julien ; Halmøy ). However, not much is known about BSNs vs nouns with articles in the transitional period between the two language stages, i.e. Middle Norwegian (approx. –). Indrebø (: , my translation) briefly mentions that the indefinite article ‘emerges in late Old Norwegian and Middle Norwegian’, and that the use of the definite suffix ‘increases much’ in Middle Norwegian. More recently, Mørck (: ) notes that the indefinite article seems common in Middle Norwegian, but to my knowledge, there are no more extensive studies of the topic. Investigations of BSNs in the Middle Norwegian period may shed new light on BSNs more generally and the grammaticalization of definiteness. In this chapter, I present novel data from Middle Norwegian and discuss them in the context of previous works on BSNs and the emergence of definiteness in the history of Scandinavian. I will show that although BSNs are not very common in Middle Norwegian, this language stage exhibits some interesting differences compared to Modern Norwegian. For example, the indefinite article is not yet fully grammaticalized in all speakers. Currently, no parsed corpus of Middle Norwegian is available. This chapter is based on manual investigation of a relatively small data set, and although some figures are presented, much of the discussion is qualitative. Following up this study with more large-scale research would be interesting; however, the approach taken here has its virtues, at the very least as a first approximation. As Crisma (: ) points out, if definiteness is not fully grammaticalized in a language, one does not necessarily need a large data set to establish this: ‘ . . . it is sufficient to find a few instances of nominal arguments clearly interpreted as definite even in the absence of any overt marking’.4 Furthermore, to understand the diachronic development of definiteness, one needs to carefully consider the context in which each BSN or noun with an article occurs. Therefore, the qualitative perspective is important. The chapter is structured as follows: in section ., I give a more detailed overview of BSNs in Old and Modern Norwegian. In section ., I discuss diachronic 4

It takes more empirical evidence to establish that definiteness is fully grammaticalized.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



developments in related languages. In section ., I present my observations from Middle Norwegian. In section ., I discuss the extent to which Middle Norwegian differs from Modern Norwegian in terms of the syntactic structure of nominals. Section . concludes the chapter.

. BSNs in Old and Modern Norwegian: a brief overview .. Modern Norwegian Modern Norwegian has a prenominal indefinite article and a suffixed definite article, as illustrated in examples ()–() (the examples follow the Nynorsk standard):5,6 ()

()

a. ein a

katt cat

(m.)

b. ei a

bok book

(f.)

c. eit a

hus house

a. katt-en cat- b. bok-a book- c. hus-et hus-

(n.)

(m.) (f.) (n.)

Articles are generally required when the noun is referential. Thus, nouns in argument positions typically appear with an article, while non-arguments, e.g. predicative nouns, are bare. Cf. (): ()

a. Eg såg ein katt. Katt-en het I saw a cat cat- was.called ‘I saw a cat. The cat was called Buster.’

Buster. Buster

5 In the Bokmål standard, the indefinite article has the forms en (m.), ei (f.) and et (n.); the definite article is identical. 6 Additionally, if a definite noun is modified by an adjective, a pre-adjectival definite determiner is used (this is often referred to as ‘double definiteness’; see e.g. Julien  and Andersen ):

(i) den grøne bok-a the green book- ‘the green book’ For reasons of scope, I do not discuss double definiteness in this chapter.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn b. Per er lærar. Per is teacher ‘Per is a teacher.’

In (a), the indefinite article ein is used to introduce a new referent (ein katt ‘a cat’); katten in the immediately following sentence, with the definite suffix -en, has unique reference (Dyvik , Faarlund et al. ). The noun lærar ‘teacher’ in (b) is bare; this is possible because it is a predicate noun. The correspondence between argumenthood and the use of articles with singular count nouns is not entirely one-to-one. Two systematic exceptions should be mentioned. First, BSNs can be used in argument positions as bare type nouns. These BSNs do not introduce an individuated token as a referent; instead, the type of referent is emphasized (Borthen ; Julien ; Halmøy ; Rosén and Borthen ). Julien (: ) shows that bare type nouns do not easily combine with nonrestrictive relative clauses; this follows from their not introducing individuated token referents (Julien’s example): ()

# Kari fikk ny sykkel, som forresten Kari got new bike  by.the.way ‘Kari got a new bike, which by the way is blue.’

er is

blå. blue

Julien (: ) argues that bare type nouns in Norwegian have a reduced syntactic structure.7 A second exception to the correspondence between argumenthood and the use of articles concerns predicate nouns. In predicate positions, there is a difference between nominals that are objectively classifying and those that are subjectively characterizing (e.g. Dyvik ). While the bare predicate noun in (b) above is objectively classifying (Per is ‘classified’ as a teacher), predicate nouns that convey a subjective characterization typically appear with an indefinite article. Subjectively characterizing predicate nouns are often modified by an adjective (cf. ): ()

Per er en god lærer. Per is a good teacher ‘Per is a good teacher.’

Dyvik (: , my translation) observes that in these contexts, the indefinite article ‘seems to have developed a secondary function’, i.e. that of distinguishing the two types of predicate. A similar distinction is found in other Germanic and Romance languages (de Swart et al. ). 7 In Julien’s analysis, they lack the functional projections DP and CardP. See (among many others) Schmitt and Munn (), Pereltsvaig (), Espinal () and Alexopoulou et al. () on BSNs in other modern languages. Halmøy () and others argue that BSNs in Modern Norwegian are not actually singular; instead, they are underspecified for number. I abstract away from that in this chapter.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



.. Old Norwegian Old Norwegian had a morphologically bound, postnominal element -inn, which is often referred to as a definite article (Faarlund : ).8,9 However, its distribution is not identical to that of the modern article; one often finds BSNs in cases where a suffixed article would be expected in Modern Norwegian (Nygaard : ). Nygaard (: ff.) mentions some contexts in which BSNs are particularly common. These include cases in which there is in practice only one possible referent (e.g. konungr ‘king’, dróttning ‘queen’), nouns denoting the universe or ‘large and unique parts of it’, prominent parts of houses and ships, and body parts. For the latter two cases, Nygaard notes that BSNs are especially favoured after prepositions. Some Old Norwegian examples are given in (): ()

a. Oc tækr nu sol upp koma. and takes now sun up come ‘And now the sun starts rising.’ b. þa var leset guðspiall er þæir then was read gospel when they ‘The gospel was read when they came.’

() kvamo. came ()

BSNs could also have an indefinite interpretation; cf. (): ()

a. Oc vil ec at haugr se gorr mikill and will I that mound is done big ‘And I want a big mound to be made after me . . . ’ b. Kolbæinn het maðr Kolbeinn was.called man ‘There was a man called Kolbeinn.’

æftir after

mic . . . me ()

()

It is a widely held view that einn, the precursor of Modern Norwegian ein, was not an indefinite article in Old Norwegian, see e.g. Nygaard (: ) and Faarlund (: ). Börjars et al. (: e with further references) write (about Old Norse generally) that einn ‘still functioned as a numeral . . . , or had meanings such as “only”, “a single”, “a certain” or “sole” ’. However, Dyvik (), employing a somewhat different perspective, suggests that the transition of einn into an indefinite article began as early as in Old Norwegian, and that the first step was a stage at which einn

8 -inn is the masculine form; the feminine and neuter (nominative) forms are -in and -it. There has been some debate as to whether -inn is a clitic or an affix at the Old Norwegian stage (e.g. Faarlund  and Börjars et al. ). For the purposes of this chapter, the distinction is not crucial. Old Norwegian also has a morphologically free form (h)inn which is used pre-adjectivally with modified nouns. (H)inn will not be taken into consideration here. 9 Lander and Haegeman () argue that -inn should not be analysed as an article in Old Norse.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

was a marker of specificity.10 An example in which einn has specific reference and is presumably not a numeral is given in (): () En þar kom æínn blinndr maðr oc þuo augu sin and there came a blind man and washed eyes his í þui bloðoga vatne in that bloody water ‘And a blind man came there and washed his eyes in the bloody water.’ () Dyvik’s observation that the indefinite article was initially a specificity marker is consistent with Heine’s () cross-linguistic account of how indefinite articles evolve.

. Development in related languages While previous research on the diachronic development of BSNs in Norwegian is scarce, there are a number of studies discussing this topic in related languages.11 The development of fully grammaticalized articles, with the loss of BSNs as a consequence, has taken place across Germanic, although there are important comparative differences, such as the form of the definite article (a suffix in Scandinavian vs a preposed, free morpheme in West Germanic). A recurring observation is that the definite article is older than the indefinite article (see e.g. Leijström  on Swedish, Jensen :  and Heltoft :  on Danish, Crisma :  on English and Börjars et al.  on Faroese). This is consistent with the typological generalization that many languages have a definite article, but no indefinite article: WALS lists ninety-eight languages (out of a sample of ) (Dryer b).12 Present-day Icelandic is a language of this type (see Longobardi  and Thráinsson ); Icelandic is particularly relevant in the present context, given its close genetic relationship with Norwegian. As illustrated in (a), BSNs are used with indefinite interpretation, but BSNs with a definite interpretation are generally not allowed; here, the definite article must be used (see b): () a. rauður hestur red horse ‘a red horse’

(adapted from Thráinsson : )

10 Dyvik (: ) defines specificity in terms of presupposition of existence; it must be possible to locate the referent in the discourse universe. 11 In addition to the works cited in what follows, see e.g. Demske (), Wood (), Stendahl () and Crisma (). 12 The opposite situation, i.e. languages having an indefinite, but no definite article, is rarer; WALS lists forty-five languages out of a total of  (Dryer a).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian b. rauði hestur-inn red horse- ‘the red horse’



(adapted from Thráinsson : )

The research discussed above gives rise to certain predictions about the loss of BSNs in Norwegian: one can reasonably expect BSNs with an indefinite interpretation to be more common than BSNs with a definite interpretation, at least as a general tendency.13

. Observations from Middle Norwegian In this section, I present some empirical observations, focusing on BSNs whose interpretation is (from a modern perspective) definite, indefinite, and ambiguous. First, however, some discussion of the Middle Norwegian data is in order. .. Middle Norwegian data ... Selection of data This study draws on a sample of  charters from the Diplomatarium Norvegicum (DN), dating from the period –.14 Most of the charters are relatively short documents; they typically describe legal acts such as conveyance of property, testimonies, official decrees, and wills. The sample consists of charters that have been investigated for other purposes in previous studies, e.g. those by Ottoson (), Mørck (), Kinn (b), and Wagener (); the selection criteria of these authors are reflected in the sample. The charters are of mixed Norwegian provenance; geographical variation within Norway is not taken into account. There are some challenges associated with charters as linguistic data. One such challenge is influence from the East Nordic varieties Swedish and Danish (Indrebø : –). In the sample, charters whose linguistic features are predominantly Norwegian have been favoured, but in practice it would be very difficult to avoid

13 Stroh-Wollin (: , ), who focuses on Swedish, suggests that while the first instances of indefinite articles appeared significantly later than the first definite articles, ‘ . . . the modern rules regarding indefiniteness marking seem to have been finally settled at about the same time as those concerning definiteness marking . . . ’. Stroh-Wollin () refers to Skrzypek () for observations corroborating this view; however, Skrzypek () is not conclusive about the final stage of indefiniteness marking: in the latest period she has investigated (–), indefinite articles are still not obligatory and she states (: –) that the ‘stabilization period falls outside the scope of this study’. 14 The documents are available in print and at http://www.dokpro.uio.no/dipl_norv/diplom_field_eng. html. Examples are mainly rendered as in the online edition, but some hyphens and editorial insertions have been left out. In the English translations, place names have been standardized, mostly following Oluf Rygh’s Norske Gaardnavne ‘Norwegian farm names’ (online edition at http://www.dokpro.uio.no/rygh_ ng/rygh_felt.html). Names of persons have also been standardized, mostly following Heggstad et al. () and Kruken and Stemshaug ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

East Nordic influence completely. Another issue is posed by formulaic language (Indrebø : ), especially in the introduction (protocol) and conclusion (eschatocol) of the charters (Hamre ). In the sample, long documents with narrative content in addition to the usual formulae have been favoured. Formulae have not been completely excluded, however. Even in formulaic contexts, charters may vary syntactically (Grøtvedt : , Mørck : , Blaxter : ); formulae are thus not without value as linguistic evidence. ... Principles of excerption The study includes nouns in argument positions (typically subjects and objects), but also nouns in non-argument positions (e.g. in PPs) in cases where Middle Norwegian seems to differ from Modern Norwegian. I have looked only at singular count nouns that are not modified by a demonstrative, quantifier, or possessive noun/pronoun. Proper names are excluded, as are names of weekdays and holidays. Certain syntactic environments in which a noun can potentially be bare for independent reasons have been left out. This includes coordinations of two BSNs (Heycock and Zamparelli ), vocatives and predicates (Delsing , Longobardi ). Some more specifically defined environments have been excluded either because they seem to behave like fixed expressions or because a BSN would be allowed in the equivalent context in Modern Norwegian too. This includes, for example, nouns modified by the adjectives fyrirnefndr ‘aforementioned’ and samr ‘same’,15 and the formulations sverja á bók lit. ‘swear on book’, vera á stefnu lit. ‘be at meeting’, til sannenda ‘as proof, to testify’, and af einni/annarri halfu ‘on the one/ other side’. .. Nouns with a definite vs indefinite interpretation A noun with definite interpretation typically has unique and specific reference (see e.g. Julien , citing Enç  and Abbott ). However, I follow Lyons (: ff.), who stresses that definiteness is a grammatical category, and like other grammatical categories, it does not have an entirely straightforward correspondence with any category of meaning (illustratively, Faarlund :  mentions examples from Modern Norwegian where the ‘referential function [of the definite article] is less obvious’). Consistently with this, a basic criterion for deciding whether a Middle Norwegian BSN should be interpreted as definite or indefinite has been which article one would use in a Modern Norwegian translation; this is how I distinguish ‘definite and indefinite interpretations in what follows. Using Modern Norwegian as a measure like

15 Van de Velde () argues that the equivalent adjectives in Dutch should be treated as ‘determinatives’; see also Lyons (: ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



this might seem anachronistic, but as a heuristic tool, it is useful; it helps identify differences between the language stages, which can then be analysed in more detail.16 A quantitative overview of BSNs with definite vs indefinite interpretation is given in table .. Overall, there is a higher proportion of BSNs with indefinite interpretation (.%, vs .% with definite interpretation). The difference is statistically significant (Fisher Exact Test, p = .); note, however, that the figures should be read with caution, as the data are often difficult to interpret. For example, it can be challenging to distinguish a numeral from an indefinite article, since they have the same form, or a bare indefinite noun from a bare type noun (see section ..).17 Also, note that BSNs of the type discussed in section ..., where it is not clear if the interpretation is definite or indefinite, are not included. TABLE . Overview of BSNs vs nouns with -inn (definite interpretation) or einn (indefinite interpretation)

Definite interpretation Indefinite interpretation Total

BSN

Noun with -inn/ein

Total

 (.%)  (.%)  (.%)

 (.%)  (.%)  (.%)

 (%)  (%)  (%)

In what follows, we take a closer look at nouns with a definite, indefinite, and ambiguous interpretation. ... Definite interpretation As is evident from table ., BSNs with a definite interpretation are not very common; .% of nouns with a definite interpretation are bare, while the remaining ones appear with -inn. Table . shows how BSNs vs nouns with -inn are distributed with respect to syntactic function. Börjars et al. (: e) and Skrzypek (: ) observe that BSNs in Old Faroese and Old Swedish have a tendency to occur in the complement position of PPs. BSNs

16 A reviewer points out that there are cases in which the choice of article itself is decisive for the interpretation (Hawkins ). The BSNs discussed in section ... can be seen as examples of this; apart from that I have encountered very few, if any, cases in which the linguistic context does not favour one reading over the other. 17 Each example has been interpreted in with regard to its surrounding context. Some factors considered relevant in cases of doubt are the following: i) Form: the Roman number i is always interpreted as a numeral. If einn (in letters) occurs in a charter where unambiguous numerals are spelt with Roman numbers, I have been inclined to interpret einn as an article. ii) Properties of the noun: when large or valuable items are mentioned (e.g. in charters about inheritance), I am assuming that einn is less likely to be a numeral: cardinality is presumably not central in these cases (one would not expect there to be more than one item). iii) Occurrences of BSNs in very similar environments in other charters: if einn occurs in an environment in which BSNs are used in other (typically, older) charters, I take this to indicate that a nonnumeral reading is plausible.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn TABLE . BSNs vs nouns with -inn, by syntactic function

SUB OBJ In PP OTHER Total

BSN

Noun with -inn

Total

 (.%)  (.%)  (.%)  (%)  (.%)

 (.%)  (.%)  (.%)  (%)  (.%)

 (%)  (%)  (%)  (%)  (%)

Syntactic labels used: SUB = subject, OBJ = direct and indirect objects and complements of adjectives, In PP = complement of preposition. The OTHER category includes other syntactic functions, like adverbials and arguments/ attributes of nouns.

are slightly more frequent in PPs than in other positions in Middle Norwegian too; however, the difference is not statistically significant (Fisher Exact Test, p = .). There is one example of a definite BSN that is a subject; this example is rendered in ():18 () at . . . Signe tager then arff allen . . . oc eigj broder son that Signe takes that inheritance all and not brother son ‘that . . . Signe gets all of the inheritance, . . . and not the nephew.’ (DN.XI , ) The BSN in () refers to Signe’s nephew, who is previously mentioned in the charter; note that it is also a kinship noun. There are three cases of BSNs that are objects. One is from an early charter from : ()

gaf þer þa kirkiu firir legerstad-en . . . atian gave there then church for burial.site- eighteen aura bol jordar j Finnastadum . . . eyrisból land in Finstad ‘ . . . and gave the church there and then eighteen eyrisból of land in Finstad . . . ’ (DN IV., ) ...

ok and

The BSN is an indirect object (kirkja ‘church’); it seems to refer to St. Mary’s church in Oslo, which has previously been mentioned in the document (alternatively, one might perhaps interpret it as referring to the church as an institution, which could be inherently unique).19 18 A modern Norwegian equivalent of this example may optionally be constructed with a possessive pronoun. The data set includes two cases of BSNs where a construction with a possessive pronoun would not only be an option, but instead the most natural modern translation; these cases are not included in table .. 19 The word þa, which precedes kirkiu, could also be an accusative sg. f. determiner, but I assume that any determiner in this context would be dative.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



The two other object BSNs are modified by annarr ‘second’; cf. () for illustration: () i thet samma sloo attherde Gwdelogh atherde in the same hit aforementioned Guðlaug aforementioned Anund i hoffd-it met eth horn oc annith slag han O˛nund in head- with a horn and second blow he sloo raghe i weg-en . . . hit hit in wall- ‘In that moment the aforementioned Guðlaug hit the aforementioned O˛nund in his head with a horn and the second blow hit the wall . . . ’ (DN II., ) The remaining BSNs are, as mentioned, complements of prepositions. Some examples are given in (): æptir þet togo Þiodgæir firnempder ok Helge after that took Þjóðgeir aforementioned and Helgi Kætilsson j stak ok skiorto Haluordz tidnemfdom Ketilsson in jacket and shirt Hallvarð’s oft.mentioned rifwo sunder nider til bæltistad ok wildo draghit han ripped asunder down to waist and would dragged him wr kirkiogardd-enom. out.of churchyard- ‘ . . . after that, the aforementioned Þjóðgeir and Helgi Ketilsson grabbed the oft-mentioned Hallvarð’s jacket and shirt and ripped them down to his waist and wanted to drag him out of the churchyard.’ (DN I., )

() a. . . .

oc vt ad dør gek oc Hiarandher Stæffansson and out of door went too Hjarrand Stefánsson bondh-en i gord-en med honum master- in farm- with him ‘ . . . and the master of the household, Hjarrand Stefánsson, also went out through the door with him.’ (DN II., )

b. . . .

Generally, many of the BSNs with definite interpretation in Middle Norwegian can be characterized as relational or involving inalienable possession; they denote body parts (a), parts of houses (b), and family members () (recall from Section .. that Nygaard  makes similar observations for Old Norse). Some of the nouns are modified by adjectives that imply uniqueness, like annarr ‘second’ in (). With the possible exception of (), which is an early example, I have not found BSNs whose definite interpretation unambiguously stems from previous mentions in the text. One might hypothesize that -inn became more or less obligatory as an anaphoric definiteness marker (Lyons : –) in the first part of the

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

fifteenth century. An illustration of -inn as an anaphoric definiteness marker is given in (): ()

. . . at fader þeire fek einæ skæino that father their got a wound þet inghen hwasse skeina-n kom that no-one how wound- came ‘ . . . that their father got a wound on his foot . . . wound came about . . . ’

a foth-en . . . ok visse on foot- and knew til . . . to and no-one knew how the (DN I., )

The data presented in this section suggest that BSNs with a definite interpretation have a limited distribution in Middle Norwegian, and there are certain recurring patterns with respect to their semantic/pragmatic properties.20 ... Indefinite interpretation BSNs with an indefinite interpretation occur in a higher proportion than BSNs with a definite interpretation (.% vs .%) (see table .). An overview of their distribution with respect to syntactic function is given in table .. TABLE . BSNs vs nouns with einn, by syntactic function

SUB OBJ In PP OTHER Total

BSN

Noun with einn

Total

 (.%)  (.%)  (.%)  (.%)  (.%)

 (.%)  (.%)  (.%)  (.%)  (.%)

 (%)  (%)  (%)  (%)  (%)

There are four subject BSNs, three of which are found in one individual charter (DN XI.), which cites a law. The three nouns are all kinship nouns; one is rendered in ():21 ()

. . . thi att ther stander . . . wtj log-en at broder because that it says in law- that brother werder broder arfuj . . . becomes brother heir ‘ . . . because it says . . . in the law that a brother shall be the heir of his brother . . . ’ (DN XI., )

20 Even in the environments mentioned here, BSNs are not necessarily obligatory; see the form fothen ‘the foot’ in (), where -inn (spelt ) is used with a body-part noun. 21 Kinship nouns have some special properties even in Modern Norwegian; my preference as a native speaker would be to include an indefinite article in a modern equivalent of (), but there may be speaker variation (see Lødrup ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



Many of the BSNs in object position denote parts of properties that have been sold, a common topic in charters. See (): ()

. . . at þaugh hafdu selt Onunde Reidulfssyni merka bool that they had sold O˛nund Reiðulfsson merkrból j Kolstadum . . . in Kolstad ‘ . . . that they had sold O˛nund Reiðulfsson a merkrból in Kolstad . . . ’ (DN III., )

As mentioned, previous research suggests that the indefinite article arose first as a marker of specificity (Dyvik , Heine , see also Brandtler and Delsing  and Skrzypek  on Swedish). If this was the case in Norwegian, one can expect to find a stage at which BSNs were used in non-specific contexts, while einn was required, or at least preferred, in contexts of specificity. Table . shows the distribution of BSNs vs nouns with the indefinite article in specific vs non-specific contexts.22 TABLE . Specificity in BSNs vs nouns with einn

SPEC NON-SPEC Total

BSN

Noun with einn

Total

 (.%)  (.%)  (.%)

 (.%)  (.%)  (.%)

 (%)  (%)  (%)

As is evident from table ., there is overall a higher proportion of BSNs in nonspecific contexts than in specific contexts (.% vs .%; the result is statistically significant at the . level (Fisher Exact Test, p = .)). This is compatible with einn being a specificity marker before it became a generalized indefinite article. An example of a non-specific BSN is given in (): ()

. . . och var ter egh lensman som ed-en skulle taka and was there not sheriff that oath- should take ‘ . . . and there was not a sheriff there to hear the oath.’ (DN II., )

A closer look at BSNs in specific contexts (which one would not expect to occur if einn was required as a marker of specificity) reveals that most of these examples are fairly early; nine out of twelve are from before . This seems consistent with a

Specificity is taken to imply presupposition of existence, following Dyvik (): see footnote . The annotation was also guided by criteria discussed in Karttunen () and Skrzypek (). The figures should be read with some caution as the distinction between specific and non-specific is not always easy to make. 22

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

diachronic path via a stage as a specificity marker; it might also suggest that the requirement for einn in contexts of specificity did not settle until mid/late Middle Norwegian. At the same time, it seems that other factors interact with specificity: both specific and non-specific BSNs to some extent follow patterns that have been observed for earlier stages of Swedish: Leijström (: ) notes that the indefinite article was introduced late in the presence of certain modifiers, e.g. halfr ‘half ’ (see a), and with collective nouns (see b): at þaug hafdo sælt Hermunde oc Þoron halft that they had sold Hermund and Þorúnnr half marka bol jardar j Glænno . . . merkrból land in Glenne ‘ . . . that they had sold Hermund and Þorúnnr half a merkrból land in Glenne . . . ’ (DN IV., )

() a. . . .

b. . . . at han reid heim til os meder that he rode home to us with ‘ . . . that he rode to our homes with a troop . . . ’

flok . . . troop (DN II., )

Some further notes on the use of einn are in order at this point. If the indefinite article started as a specificity marker, one would generally expect its use in non-specific contexts to be more recent, arising through extension. However, a few rather early cases of einn in non-specific contexts can be found. Cf. the request in (), from : ()

...

ok þy bidiom vy yder kere nadhelighen herra . . . and therefore ask we you dear gracious lord at j þaken honom fra os ok skipen os ein annen that you take him from us and give us a different fughut . . . bailiff ‘ . . . and therefore we ask you, dear gracious lord, to take him away from us and give us a different bailiff . . . ’ (DN II., )

Data like these are not necessarily at odds with the idea of the indefinite article starting as a specificity marker; differences in the rates of change across the speech community are to be expected. However, it would be interesting to study the development in more detail. Although predicates are generally excluded from this study, it is also worth mentioning that einn is found with subjectively characterizing predicate nouns quite early on: () a. . . . at han er en skelligia man i alla handa matha . . . that he is a reasonable man in all hands ways ‘ . . . that he is a reasonable man in all sorts of ways . . . ’ (DN XI., )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian før thi jak er en kranker for that I am a sick ‘ . . . because I am a sick man . . . ’

b. . . .



man . . . man (DN II., )

Recall that Dyvik (: ) calls this a ‘secondary function’ of the indefinite article. It is interesting that it is attested already in the first half of the fifteenth century; one might hypothesize that this use of einn is not secondary in terms of time, but instead an independent development that started before the indefinite article was fully grammaticalized in argument contexts. To sum up, BSNs with an indefinite interpretation are overall more common than BSNs with a definite interpretation in Middle Norwegian. The distribution of indefinite BSNs overall seems compatible with the idea that the indefinite article went through a stage of being a specificity marker, although there is some interesting variation in early Middle Norwegian. ... Ambiguous examples The previous subsections discussed BSNs in contexts where Modern Norwegian would clearly require either a definite or an indefinite article. However, Middle Norwegian also exhibits BSNs that are ambiguous with respect to the modern definiteness category. Put differently, Modern Norwegian would require an article, but both articles would be possible, with only very subtle differences in meaning. Some of these cases include a noun modified by a relative clause; cf. () (I found five BSNs of this type): () a.

b.

...

at Þorbiorn Glødersun læidi .ii. vitni . . . er swa that Þorbjo˛rn Glǿðirsson led two witnesses that so sworo . . . at þæir voro neer ok høyrdu a at swore that they were present and heard on that Þorbiorn steifdi Gunlæik j Gausinum til ændamerkis Þorbjo˛rn summoned Gunnleik in Gausen to demarcation ganngu . . . ok vm laxa fiski sæm han walk and about salmon fishing.place that he stikkat hafde þer sæm æingin var fyrir honom. dammed had there where no.one was before him ‘ . . . that Þorbjo˛rn Glǿðirsson led two witnesses . . . that swore . . . that they were present and heard that Þorbjo˛rn summoned Gunnleik in Gausen to stake out the frontiers . . . and regarding a/the salmon fishing place that he had dammed where no one was before him.’ (DN V., ) ...

at Peetar Havordsson kravd landskyl for that Pétr Hávarðsson demanded rent for tredio momat bol j for næmd third mánadarmatarból in aforementioned som hwstrv Gerthrude Jakops dootor modor hans that Mrs Geirþrúðr Jakob’s daughter mother his

halfss half iord land ssællt sold

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn hafd . . . as bref pa lyydar tær vm giort var som had as letter on states there about done was that thæt vart bref ær med fæst. this our letter is with attached ‘ . . . that Pétr Hávarðsson demanded rent worth two and a half mánadarmatarból in he aforementioned land that Mrs Geirþrúðr, Jakob’s daughter, his mother, had sold . . . as a/the letter states that was made about [these affairs], which is attached to this letter.’ (DN IV., )23

The examples in () have in common that they introduce a new referent; this referent is, however, subsequently uniquely identified by means of the following relative clause (see Lyons :  on similar cases in English). In Modern Norwegian, either a definite or an indefinite article would be felicitous, but a BSN would be unacceptable. A similar situation is found with certain appositional structures. A recurring pattern involves one or more prosodically heavy appositions combined in a way that would require an article with the first noun in Modern Norwegian. I found fourteen occurrences of this pattern; for illustration, see ():24 () a.

at han hafde sælt hæiderlighom manne. sira Andrese that he had sold honest man sir Andres Ogmundæsyni korsbrodor j Hamre iii hæfzældæboll j Vikum O˛gmundsson chorister in Hamar  halfsælduból in Viker ‘ . . . that he had sold a/the honest man sir Andres O˛gmundsson, chorister in Hamar, a part of Viker worth three halfsælduból’ (DN I., ) ...

at erligh velbirdig man mester torber olson vor that honest wellborn man master Þorbjo˛rn Ólafsson our ssogne herre talede till ewind hørland ppaa hoffz vangen parish master spoke to Eivind Hørland at Hofsvangen’ ‘ . . . that a/the honest, wellborn man master Þorbjo˛rn Ólafsson, our parish master, spoke to Eivind Hørland at Hofsvangen.’ (DN XXI., )

b. . . .

The examples in () introduce new persons which are then uniquely identified by an apposition within the same sentence. At the Modern Norwegian stage, the first element of these appositional structures would require a definite or indefinite article; 23

The form bref could, in isolation, be interpreted as either singular or plural, but in the given context, there is independent support for a singular reading. There is no plural agreement on any of the verbs that take bref as their subject (lyydar ‘state’, var ‘was’, and ær ‘is’); the verb høyrrum, on the other hand, which takes a first pl. subject, displays agreement in both number and person. Moreover, according to the description of the charter in DN, the charter is physically attached to one (and only one) other charter. 24 Simple titles, such as bisp Monss ‘bishop Mons’ (DN XXI.), are not included. In such cases, the title arguably functions like a part of a proper name, and there are no articles involved even at the Modern Norwegian stage.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



a BSN would not be acceptable. The Middle Norwegian construction is attested in both early and late charters, and with both male and female persons; it seems to be stable throughout the period. One might ask if it would be better to analyse the examples labelled here as ambiguous as indefinite. This would fit in with the observation that indefinite BSNs are more common than definite BSNs. However, examples like those cited in () and () often involve nouns with specific reference. Recall that most unambiguous indefinite BSNs with specific reference occur in early charters, whereas BSNs like those in () and () are found throughout the period. Thus, the ambiguity cannot be straightforwardly resolved by labelling the examples as indefinites. ... Intermediate summary The picture that emerges from the Middle Norwegian data presented in section . can be characterized as follows: BSNs with a definite interpretation are relatively rare. When found, their definite interpretation does not arise solely by anaphoric reference; in such cases the definite article is used (there are a few possible counterexamples of an early date). BSNs with an indefinite interpretation are more common than BSNs with a definite interpretation; this is in accordance with the cross-Germanic observation that the definite article grammaticalized earlier than the indefinite article. BSNs with indefinite interpretation are used at a higher rate in non-specific than in specific contexts; this is in accordance with the idea that the indefinite article was initially a marker of specificity.

. Comparing Middle Norwegian to Modern Norwegian The subject of this section is to what extent, and how, the differences between Middle and Modern Norwegian with regard to BSNs are reflections of differences in syntactic structure. The discussion is only a first approximation to the formal analysis of Middle Norwegian nominals; I limit my attention primarily to aspects that can be discussed on the basis of the findings presented in section .. A full analysis would need to take additional phenomena into account, like the interplay between preadjectival determiners and the suffixed article (see footnote ). This is beyond the scope of the present study (but see observations in Lundeby ). .. The structure of Modern Norwegian nominals A useful point of departure for comparison of Middle and Modern Norwegian is Julien’s () analysis of nominal phrases in Norwegian. A sketch of this analysis is rendered in () (see Julien’s book for further discussion and motivation): ()

[DP . . . [CardP . . . [αP . . . [nP . . . [NumP . . . [NP . . . ]]]]]]

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

The suffixed definite article in Modern Norwegian is first-merged in n, whereas the indefinite article originates in Card. This is illustrated in (); αP hosts adjectives, the remaining labels should be self-explanatory: () a. [DP . . . [CardP [Card ein/eit/ei [αP . . . [nP . . . [NumP . . . [NP . . . ]]]]]]] b. [DP . . . [CardP . . . [αP . . . [nP [n -en/-a/-et [NumP . . . [NP . . . ]]]]]]] Both the definite and the indefinite article are in an Agreement relationship with the higher D head, which must contain overt material (be ‘identified’) (Julien : ff.). The agreeing feature of the definite article is a definiteness feature. In the case of a simple, definite noun, the full nP (including the suffixed article) moves to Spec-DP (cf. the simplified structure in (); see Julien :  for discussion): () a. bok-a book- b. [DP boka [D [nP boka]]] The indefinite article, in the analysis assumed here, does not express a feature []; rather, it is used in the absence of any definiteness feature. The indefinite article Agrees with D in gender and number and moves to D, as illustrated in (): () a. ei a

bok book

b. [DP [D ei [CardP [Card ei [nP bok ]]]]] Importantly, definiteness does not have a single origin in the syntactic structure; there are two syntactic sources, nP and DP. The definite feature in nP is interpreted as specificity, whereas D encodes uniqueness (see also Roehrs , Andersen  and references cited there).25 An illustration is given in (): () [DP Uniqueness . . . [CardP . . . [αP . . . [nP Specificity . . . [NumP . . . [NP . . . ]]]]]] There are cases in which both n and D are spelled out, notably when a definite noun is modified by an adjective (‘double definiteness’, see footnote ). This is schematically shown in (), which includes both a pre-adjectival determiner in D and a suffixed definite article: () a. den the

lille little

bok-a book-

b. [DP [D den [αP lille [nP boka . . . ]]]] An argument supporting the proposition that uniqueness is encoded in D is that when two modified nouns are coordinated, and these nouns can potentially be 25 Julien (: –) prefers the term ‘inclusiveness’; for the purposes of this chapter, ‘uniqueness’ is accurate enough.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



co-referential, the presence of the pre-adjectival determiner in both conjuncts favours a reading whereby there are two unique referents (see Andersen : , who attributes this observation to Anders Holmberg): () den unge professor-en og (den) omsorgsfulle the young professor- and (the) caring ‘the young professor and (the) caring father’

far-en father-

If the second pre-adjectival determiner in () is left out, the professor and father are understood as the same individual. An argument that specificity is encoded in nP is the contrast between (a) and (b) (from Julien : ): () a. Dei oppførar seg som dei verste bøllar they behave  as the worst brutes ‘They behave like the worst brutes [whoever they are]’ b. Dei oppførar seg som dei verste bøll-ane they behave  as the worst brute-. ‘They behave like the worst brutes [and we know who they are]’ In (a), the definite suffix is left out; the interpretation is that no specific set of brutes is referred to. In (b), the suffix is included, and the interpretation is specific. .. Non-specific, indefinite BSNs and the encoding of specificity Section ... argued that the distribution of indefinite BSNs vs nouns with einn in Middle Norwegian is in accordance with the idea that einn went through a phase of being a specificity marker. This raises the question of how einn as a specificity marker would fit into the syntactic structure. It is not straightforward to accommodate it into the structure proposed for Modern Norwegian in (), repeated below: ()

[DP Uniqueness . . . [CardP . . . [αP . . . [nP Specificity . . . [NumP . . . [NP . . . ]]]]]]

Specificity is encoded in nP. However, there are no clear indications that einn as a specificity marker originates that low. Like the modern indefinite article, einn normally precedes adjectives (see ); recall that adjectives are located in α P, which is above nP.26 () eit a

skilrikt trustworthy

vitne witness

(DN II., )

I take it that einn as a specificity marker is not generated as low as the present-day nP; instead, drawing on Lohndal () and Faarlund (), I assume that nP with its 26 More precisely, adjectives are first-merged in the Spec-position of the α-head (Julien ). Mørck (: ) notes some cases in which einn follows the adjective, or in which einn is doubled (see also Vannebo  and Delsing ). This post-adjectival position could be analysed as a Foc-position (see Bennis et al. ; Corver and van Koppen ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

current properties (encoding specificity) is the result of a restructuring of the nominal phrase, and that einn as a specificity marker is only compatible with the old structure. According to Faarlund (: , : ), Old Norse had a functional projection RP (Reference Phrase), which encoded specificity, and which eventually (in late Old Norwegian) became the host of the specificity marker einn.27 RP was situated below demonstratives, considered by Faarlund to be D-elements, but above adjectives (in αP). () [DP [RP [αP [InflP [NP]]]]] Lohndal () and Faarlund () propose that the current nP is a result of downward grammaticalization. Glossing over many details, both accounts imply a reanalysis whereby specificity became a feature of the lower part of the nominal domain. The argumentation is for the most part based on the behaviour of the definite article -inn and on DP-internal word order phenomena; I do not discuss the proposals further here. The important point is that the use of non-specific indefinite BSNs and einn as a specificity marker can be seen as indicative of a nominal structure different from that found in Modern Norwegian. The distribution of indefinite BSNs in Middle Norwegian seems to indicate that the old, underlying structure was retained at least in some speakers until late in the period, or maybe even into early Modern Norwegian. .. Underspecification of definiteness and the role of the DP layer As shown in section ..., Middle Norwegian allows BSNs in certain contexts that are ambiguous with respect to the modern definiteness category. At the Modern Norwegian stage, either a definite or an indefinite article is felicitous, with only very subtle differences in meaning, but the grammar forces an explicitly expressed choice; a BSN would not be acceptable. For convenience, I repeat example (b) in (): ()

. . . at erligh velbirdig man mester Torber olson vor . . . that honest wellborn man master Þorbjo˛rn Ólafsson our sogne herre talede till ewind hørland ppaa hoffz vangen parish lord spoke to Eivind hørland at Hofsvangen ‘ . . . that a/the honest, wellborn man master Þorbjo˛rn Ólafsson, our parson, spoke to EiIvind Hørland at Hofsvangen.’ (DN XXI., )

The observation that Middle Norwegian allows underspecification of definiteness in these cases could be interpreted as a difference concerning the DP-layer at the top of

27

The numeral einn was merged in the same position as adjectives (Faarlund : ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Bare singular nouns in Middle Norwegian



the nominal phrase: While Modern Norwegian has a D-head that must be overtly identified, typically via Agreement and movement of elements from lower down, one might hypothesize that in Middle Norwegian, there is no such need: D does not need to be overtly expressed; alternatively, D is not (obligatorily) present. The second option has been pursued in recent, diachronic works on Scandinavian nominal phrases (Lander and Haegeman ; Stroh-Wollin ; Börjars et al. ). These authors advocate versions of what Abraham et al. () call the ‘Out of nothing-to-D Hypothesis’; they share the view that the present-day functional category DP is an innovation.28 Note that the absence of a D-layer does not exclude the presence of articles or article-like elements; these articles are not necessarily direct manifestations of D (according to Faarlund , , both -inn and the specificity marker einn originate in RP, see example ). The difference from the modern stage is that the presence of articles is not enforced by formal requirements of a higher functional head. This makes the definiteness category less paradigmatic than it is today; under certain conditions, speakers can refrain from making a choice.29,30

. Conclusion This chapter has discussed the distribution of bare singular count nouns in Middle Norwegian. It was shown that BSNs with an indefinite interpretation occur in a 28 The alternative to the ‘Out of nothing-to-D Hypothesis’ is what Abraham et al. () call the ‘Covert-to-overt D Hypothesis’. The ‘Covert-to-overt D Hypothesis’ is consistent with the tradition represented by e.g. Longobardi () et seq., whereby all nominal arguments are DPs. The ‘Out of nothing-to-D hypothesis’ finds resonance in e.g. Bošković (, , ); cf. also Chomsky (: –), who does not assume D to be universally present. 29 BSNs with unambiguous definite or indefinite interpretation can also be derived under the assumption that the topmost D-layer is not obligatory. One could either hypothesize that Middle Norwegian selectively permitted smaller nominal structures without a D-layer in some contexts (see Pereltsvaig ), or, more radically, that there was no DP at all (in the spirit of Bošković , , ). I do not take a stance on this here. 30 According to Bošković (, , ), a common property of languages that do not project a DP is the ability to license null arguments (see also Barbosa , Tomioka , Neeleman and Szendrȍi ). Mørck () and Kinn (a,b) argue that null arguments are retained in Middle Norwegian, even towards the end of the period. In the context of Bošković’s work, retention of null arguments could be taken to corroborate the idea that the present-day DP had not yet emerged in Middle Norwegian. Existential constructions might also be relevant. Existential constructions with an overt expletive arose in the second half of the Middle Norwegian period (Kinn , a). However, it seems that the definiteness effect that applies in Modern Norwegian, preventing the internal argument from being definite (Milsark ), did not categorically hold in Middle Norwegian (see Kinn : –; examples presented in Aasen (: ) suggest that it was still not categorical in dialects spoken in the nineteenth century). Vangsnes (: ) argues that a referential property shared by internal arguments excluded in existentials is uniqueness. Since uniqueness at the Modern Norwegian stage is encoded in D, the apparent lack of a categorical definiteness effect in Middle Norwegian could be interpreted as follows: DP is not obligatorily present, and although markers of definiteness will often coincide with a unique interpretation, the syntactic effects associated with the present-day DP do not necessarily arise.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Kari Kinn

higher proportion than BSNs with a definite interpretation; moreover, although multiple factors seem to be interacting, there is a correlation between indefinite BSNs and non-specific contexts. These observations are in line with findings from diachronic studies of BSNs and the grammaticalization of definiteness in related languages. Although BSNs overall have a limited distribution, I have proposed that Middle Norwegian (or at least some speakers) retained a nominal structure different from that of Modern Norwegian until quite late. The study invites further research into the grammaticalization of definiteness; since the development of the suffixed definite article in particular has already come quite far in Middle Norwegian, a closer look at Old Norwegian would be interesting.

Acknowledgements I would like to thank the editors of this volume and three anonymous reviewers for their constructive comments on an earlier version of this chapter. I am also grateful to the audiences at DiGS , GLAC , SyntaxLab at the University of Cambridge (June ), Tam Blaxter, Pernille Hansen, and Sverre Stausland Johnsen. Any remaining errors or shortcomings are my own.

Primary sources The following charters in the Diplomatarium Norvegicum were investigated (identified by volume and charter number): III., I., I., II., II., V., II., V., V., II., I., I., VII., V., II., I., VI., IV., III., IV., II., V., V., IV., IX., I., III., II., III., III., III., V., V., X., I., II., I., II., V., XI., II., V., VIII., XI., I., I., II., II., V., VI., II., II., I., II., VI., VI., V., VIII., XI., II., IV., IX., I., II., II., X., V., XXI., XI., IV., I., II., IX., I., II., VII., I., I., IV., XI., V., VIII., VII., XI., II., IV., VIII., II., II., I., III., VIII., VI., VI., V., VI., II., IV., X., VIII., I., II., II., II., I., II., IV., XI., XI., II., IV., X., V., II., V., XVIII., II., II., XI., IX., II., XXI., XI., X., VIII., XI., VI., VI., XI., VII., VII., XXI., XXI., XI., IX., VIII..

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

8 What kind of constructions yield what kind of constructions? ANDREA S B LÜME L A ND MARC O C ONIG LIO

. Introduction and background on the was-für construction This chapter gives, to the best of our knowledge, the first formal account of the genesis of the German was-für construction (WFC, cf. Pafel , Leu b).1 We claim that it historically derives from the similar was ‘what’ plus partitive genitive (WPG) construction, which had died out by the nineteenth century. Theoretically, we analyse the latter as a binominal structure that gives rise to a mononominal structure of the WFC, similarly to the cycles described in Roehrs and Sapp () for some quantifying expressions in historical German (cf. Roberts and Roussou  and van Gelderen a, c, ,  for analogous cyclic changes). This chapter is structured as follows: in this section we provide some background on the WFC, highlighting its most salient grammatical properties. Section . then provides some traditional descriptions of this construction and sketches a provisional hypothesis on its origin. Section . presents a pilot corpus study we conducted, which sheds light on some of the issues surrounding the WFC, including the question of the case assignment potential of für, the time of the emergence of the WFC, and the question of contact-induced change. Section . provides a formal syntactic analysis of both the original WPG and the resulting WFC. Section . concludes the chapter.

1 We presented the results in this chapter at the th Diachronic Generative Syntax conference (DiGS ) in Ghent, Belgium, on  June . We would like to point out that, at the time of writing, Hobich () independently arrived at very similar results in her MA thesis. Her data and descriptive analysis (but not her syntactic modelling) are thus very similar to ours.

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Andreas Blümel and Marco Coniglio . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

.. Background: salient properties of the was-für construction This section provides some background on the WFC in Present-Day German (PDG), describes some of its crucial syntactic and semantic characteristics, and briefly highlights some theoretical issues the construction has given rise to. The WFC is a construction type of PDG to form sentence types such as whquestions () and exclamatives (): ()

Was für ein Buch hat Maria what for a book has Mary ‘What kind of book did Mary read?’

()

Was für ein Buch Maria what for a book Mary ‘What books Mary read!’

gelesen? read

gelesen read

hat! has

The construction involves an invariant wh-element was ‘what’, followed by a preposition-like element für ‘for’, as well as an obligatorily indefinite NP: ()

a. was für ein / what for a / ‘what kind of book’

*das the

Buch book

b. was für (*die) Bücher what for the books ‘what kind of books’ As the glosses indicate, the WFC semantically asks for the kinds or sorts denoted by the NP. Formally, it is noteworthy that the preposition-like element für does not assign case to the NP (), differing from the case-assigning properties of the preposition für, used benefactively in (): the preposition für assigns accusative while für in the WFC is case-inactive, and the NP must receive dative from the verbal predicate helfen ‘help’: ()

Was für ein-em / *ein-en what for a- / *a- ‘What kind of boy did you help?’

()

Dieses Geschenk ist für this present is for ‘This present is for a boy.’

Jungen boy

ein-en a-

/ /

hast have *ein-em *a-

du you

geholfen? helped

Jungen. boy

Relatedly, the verb agrees with the nominal of the indefinite NP and not with was, cf. (). If the pronominal was alone is used as a subject, it gives rise to obligatorily singular (default) agreement, cf. (). In contrast, the indefinite NP controls agreement (here: number agreement) on the verb, cf. ():

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions ()

Was für Leute sind / *ist what for people are / is ‘What kind of people arrived?’

()

Was ist / *sind what is / are ‘What happened?’



angekommen? arrived

passiert? happened

A famous property of the construction is the possibility to split was off from the residual part: ()

Was hat Maria für ein Buch what has Mary for a book ‘What kind of book did Mary read?’

gelesen? read

Evidence from scope interaction with quantifiers suggests that the split and the unsplit variant involve different syntactic structures and cannot merely be a phonological effect, say, of pronouncing differential material in an upper and a lower copy of the complex wh-phrase. Thus, Pafel () observes that examples like (a) are ambiguous between a reading in which the request is for a specific kind of book which everyone reads and a distributive reading, in which for everyone there is a request for a book such that he or she reads it. Under the split, shown in (b), only the latter reading survives. ()

a. Was what

für for

ein a

Buch book

b. Was what

hat has

jeder everyone

hat has für for

jeder everyone ein a

Buch book

gelesen? read

wh > ∀; ∀ > wh

gelesen? read

* wh > ∀; ∀ > wh

Previous generative analyses have drawn a relatively detailed map of the properties of the construction across Germanic languages and dialects.2 While a number of these properties have been described, there appears to be a general divide between two approaches: such analyses where the preposition-like element für and the wh-element form a constituent to the exclusion of the indefinite NP and such analyses where für and the indefinite NP form a constituent to the exclusion of the wh-element. These are illustrated by (a) and (b) respectively: ()

a. [was für] ein Buch b. was [für ein Buch]

We will get back to these disagreements and their arguments in section .. 2

Cf. Corver (); Pafel (); Bennis et al. (); Corver and van Koppen (); Leu (a, b); Lohndal (). This body of literature also represents the differences in treatment: the latter two have defended the analysis in (a) while the former four have defended (b).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

.. Was-für construction outside Germanic and its diachrony According to the literature (Zimmermann ; Kwon , ; Hobich ), the WFC exists mainly in Germanic and Balto-Slavic, as the Russian and the Lithuanian examples below show (cited in Kwon : ): () Čto za knigu ty čitaeš? what for book. you read.. ‘What kind of book are you reading?’ () Kas tas par putni? what it for bird. ‘What kind of a bird is this?’

(Russian)

(Lithuanian)

The only diachronic formal works on (the Russian counterpart of) the WFC we are aware of is Kwon (, ).3 He holds that the WFC results from an autochthonous change, stating that ‘the Russian čto za phrase developed from a binominal sentence, which is constituted by a wh-pronoun čto and a lexical noun to the exclusion of the prepositional element za . . . ’ (Kwon : ). An example of such a source sentence is given in () (from Kwon : , citing Danylenko : ): () i čto jazykъ ixъ ... i čto věra and what language their . . . and what faith. ‘[nor] what their language . . . [nor] what their faith is.’

ixъ their

Kwon (: ) identifies the ‘second half of the th century at the latest’ as a rough time of the emergence of the construction, and, once more, the example () is an early specimen from the Novgorod birch-bark documents (letter , quoted based on Zaliznjak : , translation by Kwon : ): () čto jesi dalƄ namъ za kluck-a · za (Novgorod dialect) what are gave us for steward- for nasъ ne stotƄ us not stand ‘What kind of steward did you assign to us?! He is worthless for us.’ The development he suggests is that ‘the prepositional for in what for crosslinguistically arose as a predicator, which scopes over a small clause’ (Kwon : ). According to him, ‘ . . . the construction goes through a stepwise development, expanding its domain, i.e., nominative to direct cases, then to oblique cases, and finally to prepositional cases’ (Kwon : ), which is schematized in (): ()

 >>  >>  >> 

3

But see n..

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



Descriptively speaking, ‘ . . . what for undergoes a categorial change from a clause to a phrase, changing wh-predication to wh-modification in status’ (Kwon : ). A similar development is unlikely for the German counterpart, as we will argue in this chapter.

. Traditional description of was-für constructions and hypothesis on their origin Based on traditional observations in the literature, this section discusses the origin and development of the WFC. In addition, we will argue that the WFC took on the functions of an older genitive construction (a combination of was with the partitive genitive) and very soon replaced it. .. Origins and properties of was-für constructions According to the traditional view, a WFC like () arose during the Early New High German (ENHG) period, no earlier than the sixteenth century (cf. Grimm and Grimm : f.; Behaghel : ):4 () Dieser erzehlete vns / was die Hispanier this told us what the Spaniards für ein-en Wahn hetten . . . for a- madness had. ‘He told us how mad the Spaniards were . . . ’

(Ralegh, )

Paul (: ) and Behaghel (: ) argue that the indefinite noun following für was inflected for accusative case in the beginning, thus seem to imply that für was a preposition proper and lost this function in the course of time. Unfortunately, they do not provide any clear examples with case being unambiguously assigned by für (and not by the predicate or an external preposition). Still in the sixteenth century, Clajus () seems to hint at the possibility that für was not a preposition: pronomen interrogatiuum was / cum particula für omnium generum, numerorum et casuum substantiuis praeponitur, . . . 5 (Clajus , cited in Grimm and Grimm : )

4 If no other sources are indicated for the ENHG examples, they are taken from the Bonner Frühneuhochdeutsch-Korpus, which is presented in section . (in brackets, we indicated the subcorpus from which each token is taken along with the text year). Notice that the WFC comes in different spelling variants for both was and für, e.g. respectively wz, wc, waz, wasz, etc. and für, fúr, vor, etc. 5 ‘The interrogative pronoun was with the particle für is preposed to nouns of all genders, numbers and cases, . . . ’ (our translation).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

Interestingly enough, it is stated that the construction could precede nouns of all genders, numbers and cases. Thus, it is no coincidence that Clajus describes für as a particle and not as a preposition. At the end of the sixteenth century, für could not assign case and is thus case-transparent. With respect to the emergence of the different functions, Grimm and Grimm (: f.) distinguish three stages. After a first period during which the construction was used in accusative functions (for direct objects), a nominative usage emerged in subject functions. It is argued that, only in a final step, the construction occurred in genitive and dative functions.6 As to the possibility of splitting the construction, Gellert indicates in the eighteenth century that the split variant was very common in the spoken language, but he discourages its use: . . . das pronomen von der präposition zu trennen ist zwar im gemeinen leben sehr häufig, aber in der edlern schreibart unerlaubt, weil es die ganze construction zerrüttet: . . . 7 (Gellert, cited in Grimm and Grimm : )

Prescriptive grammarians might have played a role in restricting the use of the split construction during the last centuries (cf. ...). .. Was plus partitive genitive Earlier than the sixteenth century, only constructions with was/waz ‘what’ in combination with a partitive genitive (WPGs), as in (), were attested in the same contexts in which we later find WFCs. ()

was gezüg-s der cirurgicus haben sol. what tool- the surgeon have should ‘What kind of tool the surgeon should have.’ (Brunschwig, ) ...

Notice that the WPG is already attested in Old High German (OHG), mainly in the same functions of the PDG WFC (cf. Grimm and Grimm : f.; Behaghel : ): () waz wort-o ist thiz . . . ? what word-. is this ‘What kind of words are these . . . ?’ (Tatian , , cited in Grimm and Grimm : ) A similar view is held by Ebert et al. (: ), who claim that nominative case marking in WFCs is attested only starting from the beginning of the sixteenth century, while dative and genitive case marking was available from the late sixteenth century. There appears to be an inconsistency in this scenario. Given that WFCs were available from the sixteenth century and that accusative constructions are claimed to have shown up before nominative ones, it is not clear how one can reconcile the fact that nominative constructions should have already been attested at the beginning of the sixteenth century. 7 ‘ . . . it often happens in common life that the pronoun is separated from the preposition, but it is disallowed in elegant writing, since it disrupts the whole construction . . . ’ (our translation). 6

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



In this example, the agreement of the verb ist with the singular waz (and not with the plural worto) illustrates very clearly that this should be analysed as a binominal construction in which waz was visible for agreement. In contrast, the genitive nominal was dependent on the nominal containing waz for case assignment. In some later cases of WPGs, genitive case is not marked, as illustrated by the following example: () waz kriec da ergie what war there took.place ‘what kind of war was taking place there.’ (Berth. II, , , cited in Behaghel : ) In this example, a genitive marking -s on kriec would have been expected. This seems to indicate that this marking was felt to be no longer necessary at that time. Behaghel (: ) relates this loss of genitive in WPGs to the general loss of genitive that is observed in the history of German. Similar examples can rarely be found in certain exclamative clauses in PDG as well: () Was ein Wetter heute what a weather today ‘What a weather today!’

ist! is

Interestingly enough, Behaghel (: ) observes that the WPG later exhibited some contamination of the WFC, as in the following example: () was für Uebel-s what for evil- ‘what kind of evil’

(Amadis  (DW. ), cited in Behaghel : )

Example () displays both genitive marking on the indefinite noun and the particle für. All these facts seem to indicate that the WFC and the WPG could be diachronically related, a hypothesis which is explored in the following sections. A remark is necessary at this point. In PDG, there exists a construction with was in combination with adjectives, which is exemplified in (): () a. was Neu-es what new-| ‘something new’ b. mit was Neu-em with what new- ‘with something new’ The construction is transparent with respect to case assignment, as we see in (b), where the adjective is assigned dative by the preposition mit. However, in the beginning was was part of a structure in which adjectives invariably received genitive

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

case (arguably by a silent nominal). Probably as the consequence of the collapse of the adjectival morphological ending for nominative/accusative ‑ez with the genitive ending ‑es, such constructions were reanalysed as case-transparent (cf. Behaghel : ; Ebert et al. : ). .. Provisional hypothesis and questions to be answered As mentioned in section .., we intend to explore the hypothesis that WFCs developed from WPGs (via a process similar to those described in Roehrs and Sapp  for quantifying expressions), given that their semanto-pragmatic and syntactic properties are very similar and strongly intertwined (also cf. Hobich ): ()

Provisional hypothesis (to be made explicit later): WPG ! WFC

That one structure spread to the detriment of the old structure is a well-known fact, but what we propose is that WFCs are the direct continuation of the earlier WPGs. We think that this development is not only empirically, but also theoretically well motivated. However, before presenting our theoretical analysis (in section .), we will address the following preliminary empirical issues in order to better understand the diachronic scenario: a. Is it true that WFCs arose no earlier than the sixteenth century? And what can be said about WPGs? b. In which dialects was the WFC first attested? Was this construction autochthonous or was it possibly borrowed from other (Germanic or Balto-Slavic) languages? c. Did für really assign accusative at the beginning? This amounts to asking whether it really functioned as a preposition proper at any stage. d. Is it true that the occurrence of WFC-accusatives predated the rise of nominatives, and that of genitives and datives? e. Did the possibility of splitting exist right from the start? This could indicate that was and für did not form a constituent at the beginning. If this turns out to be true, this would also be compatible with the origin of WFCs from binominal constructions like WPGs. In order to answer these questions, we conducted the corpus-based investigation that is presented in the next section.

. A corpus-based investigation of was-für constructions The pilot corpus study conducted is mainly based on the Bonner Frühneuhochdeutsch-Korpus,8 which consists of forty ENHG prose texts from the 8

Das Bonner Frühneuhochdeutsch-Korpus, Korpora.org, www.korpora.org/FnhdC/.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



fourteenth to the seventeenth centuries (from the second half of each century respectively). This time window is considered as the critical period during which WFCs emerged. We extracted and manually checked all WFCs (N = ) and WPGs (N = ) and furthermore enriched the tokens by means of further syntactic information (e.g. case assignment, split/unsplit construction, etc.).9 In order to explore the development of the construction in the following period (especially until the eighteenth century), a similar method was adopted for two further corpora, which are searchable by means of ANNIS:10 . Fürstinnenkorrespondenz .11 (‘Princesses’ Correspondence’): corpus of private mail (from  to ) in Central German. . RIDGES .12 (‘Register in Diachronic German Science’): corpus of scientific texts from the fifteenth to the nineteenth centuries, written in Germany, Austria, and Switzerland. If we consider the three corpora together, we retrieved a total of  WFCs and  WPGs. In the next section, we will present some empirical results based on the three corpora. .. First empirical results In the following subsections, we will consider each of the following aspects separately: a. b. c. d. e.

period of origin of WFCs (...) dialectal distribution of WFCs (...) function of für (...) grammatical functions of WFCs (...) splitting (...).

... Period of origin of was-für constructions As to the claim made in the traditional literature that WFCs arose no earlier than the sixteenth century, the data in the Bonner Frühneuhochdeutsch-Korpus seem to support it. Table . and figure . clearly show that all tokens for WFCs are to be found in the sixteenth and seventeenth centuries in clear contrast to WPGs.

9

We would like to thank Marten Santjer for his considerable help and support. See Krause and Zeldes (). Link: http://corpus-tools.org/annis/. 11 Lühr, Rosemarie; Faßhauer, Vera; Prutscher, Daniela; Seidel, Henry; Fuerstinnenkorrespondenz (Version .), Universität Jena, DFG. http://dwee.eu/Rosemarie_Luehr/?Projekte___DFG-Projekte___ Fruehneuzeitliche_Fuerstinnenkorrespondenz_im_mitteldeutschen_Raum. http://hdl.handle.net// ---A. 12 Lüdeling, Anke; Odebrecht, Carolin; Zeldes, Amir; RIDGES-Herbology (Version .), HumboldtUniversität zu Berlin. http://korpling.org/ridges/. http://hdl.handle.net//---. 10

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio TABLE . Distribution of WFCs and WPGs

WPG WFC

th c.

th c.

 (%)  (%)

 (%)  (%)

th c.

th c.

 (%)  (%)

 (%)  (%)

Source: Bonner Frühneuhochdeutsch-Korpus

100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%

WPG WFC

14th c.

15th c.

16th c.

17th c.

FIGURE . Distribution of WFCs and WPGs in the Bonner Frühneuhochdeutsch-Korpus

It is noticeable that even though we were not able to find a single example of this construction until the end of the fifteenth century (not even in the secondary sources or in other corpora, but see Hobich  for a possible first attestation of this construction in , also cf. Kwon :  n. ), WFCs suddenly appeared in the sixteenth century in about % of cases. The construction rose continuously at the expense of the WPG. WPGs became rarer (and eventually disappeared), as can be observed in the data in table . (represented in figure .) comprising data from all of the three corpora.

TABLE . Distribution of WFCs and WPGs (including eighteenth century)

WPG WFC

th c.

th c.

th c.

th c.

th c.

 (%)  (%)

 (%)  (%)

 (%)  (%)

 (%)  (%)

 (%)  (%)

Sources: Bonner Frühneuhochdeutsch-Korpus, Fürstinnenkorrespondenz ., RIDGES .

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



100% 90% 80% 70% 60%

WPG

50%

WFC

40% 30% 20% 10% 0%

14th c.

15th c.

16th c.

17th c.

18th c.

FIGURE . Distribution of WFCs and WPGs in the three corpora (including eighteenth century)

Since at least at some stage the two constructions were to a large extent functionally interchangeable, these facts seem to lend initial plausibility to our main claim, namely that the WPG was the diachronic source of the WFC. ... Dialectal distribution Given the low frequency of the construction in the corpus, it is not easy to determine in which dialect(s) it was first attested. In the Bonner Frühneuhochdeutsch-Korpus, it is difficult to track the diffusion of the new construction, since the corpus contains only one text for each dialect region for each period considered. Furthermore, the new construction seems to have been rapidly adopted in all dialects right from the start. However, interestingly enough, this corpus does not provide any evidence for WFCs in (Central) Bavarian until the seventeenth century. This could possibly indicate that the new construction spread starting from the Northern or Western dialects, but this point must be left for further research. It is not even possible to determine whether the WFC was an autochthonous construction or whether its distinctive features were borrowed from other languages, given its diffusion in North and West Germanic, and Balto-Slavic languages. There is a long scholarly tradition from Miklosich (–: ), via Sandfeld (), Vasmer (), and Vaillant (: ) to Moser (: ff.) and Podlesskaja (). Some of them hold that the Slavic construction was borrowed from German (with Ertl  and Danylenko  noticeable exceptions), but then they discuss examples which were attested earlier than the sixteenth century.13 One plausible explanation of the origin of WFCs is that their diffusion in High German could have been facilitated by the contact with Low German (or Dutch), in which this construction might be attested earlier, as in the following example:14 13

We want to thank to Uwe Junghanns und Hagen Pitsch for very insightful discussions of the data and for helping us with the relevant literature. 14 Thanks to Melissa Farasyn (p.c.) for providing us with this example. Another interpretation of this example cannot be excluded, though: ‘what they consider to be right’.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

() wat de vor recht vindet what they for right find ‘what kind of right they find’. (Herforder Rechtsbuch, Westphalian, Herford, ) Thus, we could postulate that High German dialects underwent an analogous process to Low German and developed a new construction in which the genitive marking is substituted by the marker für/vor. ... Function of für As to the question whether für really assigned accusative case at the beginning, there is evidence that it is already case-inert in the sixteenth century. In all the three corpora, we found nineteen cases in which the predicate unambiguously assigns case to the NP. There are no clear cases in which it is für that assigns case to the nominal.15 Hence, we can conclude that there is hardly any evidence in the corpus for the idea that für ever functioned as a preposition proper (pace Paul : f.; Behaghel : ; Pafel : ). We will therefore assume that für has never been a preposition or case assigner in WFCs. However, we share the view that the relevant meaning of für in WFCs was probably als ‘as’, an Stelle von ‘in place of ’ (cf. Paul : f.; Behaghel : ; Ebert et al. : ; also cf. Hobich ). In fact, für was very often independently used in predicative constructions as a copular element, as shown in the following example from the Bonner Frühneuhochdeutsch-Korpus: ()

. . . er mercket das sie in für ain gůten man hat. (Neidhart, ) er notices that she him for a good man has ‘ . . . he notices that she considers him a good man.’

This special use of für is well described also in the secondary literature. For example, Grimm and Grimm (: ff.) list a large number of cases in which für is used not only in combination with verbs like haben ‘have’ or halten ‘consider’,16 but also in independent predicative expressions: () wer who

nur only

am at.the

meisten most

gsaufen guzzle

kan can

den him

preist praises

man one

für for

ein helden. a hero ‘who only can drink the most is celebrated as a hero.’ (Uhland volksl. , cited in Grimm and Grimm : ) 15 In one dubious sentence (a translation from Latin), case is not assigned by the predicate. But since this is a late example (from the seventeenth century), we can exclude that it may be prototypical for the first stage of the reanalysis. 16 A vestige of this meaning and use is still observable in PDG small clauses halten für ‘consider as’ (cf. Corver : f.):

(i)

Ich halte ihn für einen I hold him for an ‘I consider him an idiot.’

Idioten. idiot

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



() nichts . . . will ich dir für dein erb-thail nothing want I to.you for your inheritance-portion geben. give ‘I don’t want to give you nothing as your inheritance.’ (Weckherlin  (ps. , ), cited in Grimm and Grimm : ) Such uses of für in predicative constructions could be the origin of für in WFCs. This is in line with the hypothesis put forward by Kwon () on za being a ‘predicator’ in čto za constructions in Russian. The remaining aspect to explain is how für in WFCs becomes case inert in German. In the absence of a larger amount of data, the explanation can only be speculative, but Russian again offers a possible solution to the problem. As pointed out by Kwon (: f.), Russian also marginally displays cases in which für is case-transparent. A similar grammaticalization process could have taken place for für-predications in German as well. In contrast to Russian, that might have happened before für-constructions contaminated with WPGs.17 ... Grammatical functions The next question is whether the occurrence of WFC-accusatives (in object function) predated the other functions, as stated in the traditional literature. As the small data in table .—represented in figure .—show, there do not seem to be clear differences in the distribution of grammatical functions between the first two centuries. TABLE . Distribution of grammatical functions (sixteenth and seventeenth centuries)

Subject Predicative Object, accusative Adverbial Other

th c.

th c.

 (.%)  (.%)  (.%)  (.%)  (.%)

 (.%)  (.%)  (.%)  (.%)  (.%)

Sources: Bonner Frühneuhochdeutsch-Korpus, Fürstinnenkorrespondenz ., RIDGES .

17 An anonymous reviewer pointed out that one could adopt an alternative approach, in which für still assigns accusative case to was. In fact, a similar explanation was already offered by Leu (a, b). In section .., we will show that such an analysis incurs other difficulties. Furthermore, we could not find any historical cues for such an interpretation.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio 100.0 90.0 80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0

Subj

Pred

Obj_acc 16th c.

Adv

Other

17th c.

FIGURE . Distribution of grammatical functions in the three corpora (sixteenth and seventeenth centuries)

The only clear difference could regard the use in adverbial contexts, but given the small amount of data, we must be careful not to overinterpret them. ... Splitting As to the possibility of splitting the construction, the corpora reveal that this option was always available in the history of the language (cf. Hobich ). As a matter of fact, it was the clearly preferred option in the sixteenth century. As shown in table . (and figure .), there seems to be a clear tendency for unsplit construction to become more prominent in the following centuries, but especially for the eighteenth century we should be careful when interpreting the small amount of data. TABLE . Distribution of split and unsplit WFCs in the three corpora

split unsplit

th c.

th c.

th c.

 (%)  (%)

 (%)  (%)

 (%)  (%)

Sources: Bonner Frühneuhochdeutsch-Korpus, Fürstinnenkorrespondenz ., RIDGES .

Example () shows that the possibility of splitting existed right from the start for the WFC. Analogously, the same option was available for WPGs some centuries earlier (), which comes as no surprise given the binominal structure of such constructions (see section ..):

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



100% 80% 60%

split unsplit

40% 20% 0%

16th c.

17th c.

18th c.

FIGURE . Distribution of split and unsplit WFCs in the three corpora

() Was ist=s dan̄ fúr ain hawß what is=it then for a house ‘What kind of house is it then . . . ?’ ()

... (Andreae, )

waz si fundend jung-er edl-er frawen what they found young- noble- women und junkfrawen ... and unmarried.women ‘whatever young noble women and unmarried women they could find.’ (Mair, ) ...

.. Interim summary Before presenting the theoretical analysis, let us resume the discussion so far. The reanalysis scenario that we suggest is the following. To start with, we have constructions like the OHG WPG in (): () waz what

wort-o word-.

(OHG)

In this example, the noun worto is marked for plural genitive. However, owing to independent morphological changes and the ensuing syncretism already present in the OHG period, many words were not unambiguously marked for genitive case. Most unstressed vowels were reduced to schwa, so that clear genitive marking was mostly possible only in singular masculine and neuter forms. Later WPGs were thus characterized by a no longer transparent genitive case marking on the noun following waz ‘what’, as this (invented) MHG example illustrates: () waz what

wort(e) word..?

(MHG)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

The ending ‑e did not unambiguously mark plural genitive forms any longer, since it is a portmanteau morph for different cases in the plural and for dative singular (and it is mostly dropped also in the plural). The ambiguity with respect to the case of the noun following waz probably paved the way to external case marking by means of an external predicate (or preposition). We will see that similar reanalyses, in which case-assigning elements become caseinert, may be observed also with respect to other elements, such as quantifying expressions (cf. Roehrs and Sapp ). The only difference consists in the presence of für, which makes this construction peculiar. With respect to this point, we argued that, in a last step, WFCs like () emerged in the ENHG period via contamination with predicative structures headed by für like in ()–():18 () was für Worte what for words ‘what kind of words’

((E)NHG)

We will now turn to the theoretical analysis of the phenomenon.

. Towards an analysis Based on the findings of the previous sections, this section is devoted to developing a tentative syntactic analysis of the change from the WPG to the WFC, which is thus claimed to be the continuation of the older structure. In effect, our claim is that the complex nominal structure of the former comprises two DPs, and in the process of the change yields the structure of the latter, made up of only a single DP (like cyclic processes described by Roehrs and Sapp  for quantifying expressions in historical German; also cf. Roberts and Roussou  and van Gelderen a, c, ,  for analogous cyclic changes). While doing so, two analytical options of the target structure (WFC) will be taken into consideration and their advantages and drawbacks will be discussed.

18

An anonymous reviewer raises the interesting question of why only the partitive genitive after was was affected by this change, but not for example the genitive DP occurring in combination with other elements in historical German, such as interrogative wer ‘who’ and indefinite pronouns, e.g. nichts ‘nothing’. While it is logical that similar starting conditions should yield the same outputs, we should be careful in thinking of it as an automatic process. Some processes can idiosyncratically affect one input situation, but not another one (mentioning an example from section ..., why did German grammaticalize only the construction halten für, but not haben für?). The wh-phrase was is notoriously very special and this could be one of the reasons for the different grammaticalization process in WFCs. Furthermore, the problem of explaining why only was—and not other wh-elements—occurs in this construction would remain even if we rejected the hypothesis that WFCs are diachronically derived from WPGs. Needless to say, this point deserves a deeper investigation.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



.. Syntactic and semantic change That partitive structures in English involve two nominal phrases has been commonplace throughout the history of generative grammar (cf. e.g. Jackendoff ). Influential work on the construction includes Selkirk () and Jackendoff (). The latter proposed a uniform structure like that shown in () for different types of partitives. The upper NP may comprise PRO or a small set of nominals ( gallon, half, (one) third, etc.) as a head noun. Following Sauerland and Yatsushiro (), we slightly update this rendering as in (), where capital letters indicate the lack of pronunciation: () a. [NP D=a [N’ gallon [PP of wine]]] b. [NP D=some [N’ PRO [PP of the men]]] ()

[DP D=some [NP MEN [PP of the men]]]

We will go into the characterization of this silent nominal head momentarily. For now, let us lay out some of the various arguments which can be adduced to having a) two nominals and b) a null element in the first place. A naïve argument for the binominal nature of the WPG in OHG comes from the fact that a restricted set of partitives in English allow the realization of both nouns as in (a). Furthermore, semantic considerations suggest that we are dealing with two nominals in that the second nominal provides a reference denotation to which the noun in the upper DP must stand in a part-of relation. Thus e.g. (b) is felicitous if and only if PRO is a proper part of the plurality of men and () is felicitous if and only if PRO is a proper part of the plurality denoted by ‘people’:19 () was do lewte erslagen wart, das stehit what there people. slain became. that stands do nicht zu vorzelen, . . . there not to recount ‘what kind of people were slain there is no need to recount, . . . ’ (Rothe, nd half th c.) Example () also provides morphosyntactic evidence for a binominal nature of the WPG: agreement on the verb targets was or the silent nominal, and crucially not the lexical noun of the subordinate DP (lewte ‘people’). Turning to the silent nominal, specific to OHG is the partitive genitive case on DP2, which is plausibly a reflex of this element, WORD in ():

19 Cf. Sauerland and Yatsushiro () for formal details of the semantic analysis of partitives. Notice that it is open which element takes over the relational role of English ‘of ’ in German partitives.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

() [DP1

was D1 [NP1 N1=WORD [DP2 what ‘which (of the) words’

wort-o word-.

]]]

In the following, D1/N1 and D2/N2 are used for exposition to refer respectively to the superordinate and subordinate DP of the WPG. Having laid out the basics for the analysis of the WPG, which we will elaborate on below, two conceivable analytical options for the WFC and its genesis are described. For both options we assume that the silent N1 disappears and DP2 is in effect promoted to become the head nominal. This yields a richer functional structure left of the lexical nominal, for which there is good motivation from e.g. indefinite welch- (‘which’) within the WFC, which can follow the was-für part: () Context: A conversation about books Was für welche hast du gelesen? what for which have you read? ‘What kind (of books) did you read?’

(PDG)

Analytical option  follows a wide tradition of WFC-analyses in different languages indicated in section .. In this option, the transition from the WPG to the WFC preserves the hierarchical relations of all material: () [DP1 was D1 [NP1 N1=WORD [DP2 worto]]] ! [FP was F=für [DP D [NP worte]]] The WPG is on the left-hand side and the resulting WFC follows the arrow on the right. As noted, we assume that, given that the N1-layer remains phonetically unpronounced, it is no longer perceived to be present and thus disappears. This immediately captures the lack of genitive on DP2. Arguably, the change involved a recategorization of D1, which might have turned into a different functional category which we dub F in () and whose specifier hosts was. We might think of F as being part of the extended projection of DP in the sense of Grimshaw (). Analytical variants are conceivable, e.g. that the wh-element invariably occupies a functional head within the left periphery of the DP in both constructions, in which case there might not be a need for a category change. Be that as it may, DP2 becomes promoted to be the head nominal as a consequence of N1’s disappearance. Subsequently, it controls agreement and receives case from noun-external context. This approach straightforwardly explains why the sequence was für cannot split off from the rest as shown in (): () *Was what

für for

...

ein a

Buch book

Unlike the second approach described below, was für is not a constituent to the exclusion of the indefinite NP, hence the prohibition in (). However, it leaves

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



properties and status of the preposition-like element für unclear: why doesn’t it assign case? The idea that für is somehow defective and ‘case inert’ (cf. Pafel ) is but a name for a problem, not its solution. Furthermore, the question is open: what kind of element is it? While we believe that the latter question is not answered by the alternative either, the former puzzle receives a simple solution under analytical option , to which we turn now. The second approach to the syntax of the WFC we would like to consider regards the structure as resulting from rebracketing, i.e. the hierarchical relations of the material are not preserved. As before, the silent nominal N1 disappears with the advantageous consequences described above. However, unlike their behaviour in the previous option, was and D1 rebracket to form a constituent as shown in (): ()

[DP1 was D1 [NP1 N1=WORD [DP2 worto]]] ! [DP [FP was F=für ] D [NP worte]]

That is, was and D1 effectively turn into a specifier of DP2 so that a phrasal determiner hosting was and für obtains (along the lines of Leu a, b). Additionally, it must be assumed that D1 receives a novel category, again labelled F in (). As can be seen, für does not c-command the NP and hence does not case-govern the nominal. Other advantages of this approach include the fact that in other varieties of the construction, additional material can be hosted by the complex determiner, as Leu (a, b) extensively argues. For example, Norwegian perspicuously allows the overt realization of a (genitive bearing) semi-functional noun slags ‘sort’ (). Moreover, () shows that an indefinite article can precede plural nouns in Dutch, sometimes called ‘spurious een’ (cf. Bennis et al. ): () Hva har du lest for slags what have you read for sort ‘What kind of book did you read?’ () Wat voor een jongens zijn what for a boys are ‘What kind of boys are those?’

bok? book

dat? that

(Norwegian)

(Dutch)

Leaving aside details, Leu argues that all this material is uniformly, i.e. crosslinguistically part of the complex determiner structure and the observable variation boils down to parametric choices of pronunciation or non-pronunciation of material (cf. Kayne ), schematized with English words in (): ()

what kind of books a. [what for A SORT] books b. [what for a SORT] books c. [what for A sort] books

! German ! Dutch ! Norwegian

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

Regarding the syntactic or morphological category of this complex determiner, Swiss German suggests that it is adjectival (cf. also Chomsky :  on demonstratives), as Leu shows. Für can be suffixed by the adjectival ending ‑ig in addition to adjectival (i.e. gender) agreement: () Was für-ig-ä wi hesch what for-- wine have.you ‘What kind of wine did you buy?’

kauft? bought

(Swiss German) (based on Leu b: )

These data are suggestive—and yet the analysis might introduce what we consider a disadvantage. It has long been suggested that für+NP can be coordinated to the exclusion of the wh-word (cf. Leu b:  and references therein), suggesting that this sequence must be a constituent, at least at some point in the derivation: () Was [ für Männer ] und [ für Frauen ] hast du gesehen? (PDG) what for men and for women have you seen ‘What kind of men and of women did you see?’ In this case, one is forced to assume the obligatory exit of was from the complex determiner to a specifier position in the left periphery of the DP (whP in Leu b: ), a step for which there is no independent motivation. An anonymous reviewer points out to us that corresponding examples degrade when the NPs are singular (*Was für einen Mann und für eine Frau . . . ). The same reviewer furthermore observes that examples improve with mass nouns (??Was für Fleisch und für Mehl . . . ‘what kind of meat and flour’). Speculating somewhat, this might have something to do with a requirement by the wh-operator was to single out one individual who cannot simultaneously be a man and a woman. In the absence of a better understanding of even the basic facts, we leave the issue open for now, but continue to assume a syntax without a ‘heavy specifier’. Two possible scenarios of structural change from the WPG to the WFC have been sketched and discussed. In the following, we will elaborate on semantic aspects of the WPG and on the feature composition of the D-heads in the development, using analytical option  for illustration. .. Informal semantic aspects of the change and nominal change more generally As noted, we adopt the analysis of partitives in Sauerland and Yatsushiro () and suggest that N1 in the WPG is a deleted head noun which structurally disappears over time in the process from the WPG to the WFC. According to Sauerland and Yatsushiro (), N1 is not restricted to a single element but can be a lexical noun of various kinds, giving rise to different meanings. Thus, for example, N1 in (a) must denote parts and cannot denote book or books. Likewise,

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



N1 is restricted to a plural reading of book in (b), ruling out the corresponding representation of (c). Inserting the N1 parts in (b) arguably predicts an ambiguity between a reading in which parts of individual books are referred to respectively, or a proper part of the plurality of the books (a subset of the books). But only the latter reading is actually available, which justifies the more specific noun books. ()

a. Most parts of the book b. Most books of the books c. One book of the books

Sauerland and Yatsushiro (: –) state that ‘[s]emantically, the deleted noun provides a way of dividing up the plurality the argument of of denotes into countable units’. More needs to be said on semantic characteristics of the construction: Jackendoff (: ) tied partitivity to the definiteness of the embedded DP, based on the following paradigm from English: ()

a. one of these/the/my cats b. *one of all/most cats c. *one of some/three/no cats

He formulated the following condition, which, he explicates, is a semantic constraint on the embedded DP: ()

Partitive Constraint (Jackendoff : ) In an of-NP construction interpreted as a partitive, the NP must have a demonstrative or a genitive specifier.

The constraint correctly rules out the starred examples in (). As a first approximation, we would like to cast the constraint in syntactic terms as follows: D2 is preferentially [+def] as in (a). The issue of partitives is involved and we suggest that the WPG can be [-def] (cf. (b)) only under narrow and, in part, language-specific semantic conditions developed in e.g. de Hoop ().20 ()

a. [DP1 was D1 [NP1 N1=WORD [DP2 D2[+def] NP]]] b. [DP1 was D1 [NP1 N1=WORD [DP2 D2[-def] NP]]]

Given what we have said so far about the path of development, how do we arrive at the WFC from a structure like in (a), which displays an obligatorily indefinite NP?

20

She considers cases like half of a cookie, which she derives on a specific classification of quantifier determiners which must align with specific readings of the type of partitive (entity vs set quantifiers/ partitives).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

First of all, note that D1 in wh-phrases with was is arguably obligatorily [-def].21 Exclamatives in PDG give credence to the claim: () a. welch which b. was what

ein a ein a

/ / / /

*das the *das the

Buch book Buch book

A structure as in () obtains for the WPG: ()

[DP1 was D1[def] [NP1 N1=WORD [DP2 D2[+def] NP]]]

As noted, we assume that in the change to the WFC, N1 disappears. Concomitantly, genitive on DP2 is unavailable. Notice, furthermore, that partitive readings also become unavailable, since the silent nominal cannot semantically partition the denotation of DP2. This, we suggest, is another desirable feature of the analysis. The result of this disappearance is shown in (): () [DP1 was D1[def] [DP2 D2[+def] NP]] What is the underlying cause for recategorizing D1, as we have suggested in ()? One possibility is the following: Heck () proposes that syntax is constrained in such a way that direct recursion is banned, i.e., a configuration like () is ruled out in principle. If the claim can be sustained, it appears natural to adopt it as a force in diachronic change as well. Thus, in a structure (), either of D1 or D2 is forced to change category, and we suggest that it be the higher one to eventually yield () (where F = für): () [FP was F [DP D[+def] NP]] Remember that there was a certain degree of contamination of the WFC by the WPG, as in waz kriec ‘what war’, which simultaneously already lacked the genitive on the noun, and did not yet feature für. It is conceivable that () represents a structure for such contaminated stages. Now if there is a structural condition that forces D to be [-def ] in the local environment of was, as () suggests, we arrive at our final analysis for the WFC: ()

[FP was F = für [DP D[def] NP]]

At this point we would like to briefly situate our proposal for the WFC within what is known about syntactic change of the DP more generally. In fact, the development we

21

German was contrasts with English what, as an anonymous reviewer rightly points out, cf. what the hell. However, there are numerous other fixed expressions one finds after English what, such as -ever/on earth/in God’s name/ . . . So it is not clear what the significance of this apparent counterexample is.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

What kind of constructions



have suggested parallels observations on quantified nominals by Roehrs and Sapp (). They observe that when dependents of quantified nominals are full DPs or pronominals, they cannot integrate in the containing structure but have to remain a separate DP. The contrasts in () show this with PDG to illustrate the diachronically attested pattern: () a. mit with b. *mit with c. mit with d. *mit with

vielen many vielen many vielen many vielen many

der the.gen

Bücher books

den the.dat

DP

Büchern books

ihrer theirs.

DP DP/pronoun

ihren theirs.

DP/pronoun

By contrast, when dependents of quantified nominals are bare nouns or bare nouns modified by adjectives, they obligatorily integrate: () a. mit with b. *mit with

vielen many

(schön-en) beautiful-

Bücher-n books-

Adj+N

vielen many

(schön-er) beautiful-

Bücher books.

Adj+N

In a very elaborate framework, which successfully captures the intricacies of the cyclic development of quantifying elements in the history of German, Roehrs and Sapp () hold that these dependents are reduced nominals and not full DPs. These must obligatorily be integrated into their host nominals (in contrast to full DPs). The current study on the diachrony of a subset of PDG wh-phrases confirms the authors’ findings on independent grounds.

. Conclusions To conclude, our corpus-based study has confirmed the emergence of the WFCs during the sixteenth century to the detriment of the older WPGs. We proposed that the binominal structure of WPGs, in which a partitive genitive DP2 is syntactically embedded in the DP1 containing waz ‘what’, restructures as a mononominal structure and sketched certain explanatory payoffs of such a change. Even if homophonous with a preposition requiring accusative case, the case-inert predicator für in the WFC was shown to have never functioned as a preposition. Finally, we have compared two possible analyses of the WFC and argued for a syntactic analysis in which für forms a constituent with the following nominal.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Andreas Blümel and Marco Coniglio

Acknowledgements We would like to thank the audience of the th Diachronic Generative Syntax Conference (Ghent,  June– July ), of the colloquium Theoretische Linguistik (Göttingen,  January ), of the conference Generative Grammatik des Südens (Göttingen,  June ), of the workshop Sentence Grammar and Discourse Grammar (Nijmegen,  November ), two anonymous reviewers, and especially Melissa Farasyn, Hagen Pitsch, Uwe Junghanns, and Marten Santjer for their helpful comments and support on different parts of the chapter.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

9 Quantificational cycles and shifts M O R E N O M I T RO V IĆ

. Introduction This chapter reports an inter-genetic diachronic study of quantificational particles in Indo-European (IE) and Japonic (JP), making a case for diachronic typology of syntactic–semantic unidirectional patterns of change in quantificational meanings. The quantificational expressions under investigation conform to the bimorphemic expression that comprises a wh-stem and a quantifier particle (which I dub μ), e.g. *kw​e in Proto-IE, and mo in Old Japanese. In Latin and Gothic, for instance, wh þ μ formulas are universal distributive terms while in Indo-Iranian (Sanskrit and Avestan), wh þ μ terms are consistently (polar) existentials (NPIs). The first question we address concerns the historical primacy of the two quantificational meanings (did NPIs get born out of universals or vice versa?). Evidence from JP (and possibly also from Sino-Tibetan; see Mitrović and Hu ) shows a primarily universal function, hence a universal cycle of change is proposed to account for the quantifier shift in IE. The completely cyclical nature of the problem is borne out in light of the morphosyntactic origins of the quantificational *kw​e particle. Philological evidence shows that *kw e has wh-pronominal or indefinite origins with a clearly existential quantificational force. The chapter provides an analysis according to which the pronominal meaning grammaticalized into an existential quantifier in IE (citing evidence from Anatolian), and then further into a universal quantifier particle (Latin, Gothic, Old Japanese, Ancient Chinese, etc.). The cycle is closed once the universal changes into a (polar) existential (Classical Japanese, Indo-Iranian, Slavonic, etc.). One of the core conclusions that this chapter draws is that existential (∃) NPIs, and presumably Polarity-Sensitive Items (PSIs) more generally, diachronically derive from distributive universal (∀) quantificational meanings.1 I will explain this ∀ ! ∃

1 There is an apparent contradiction in stating that there is both a cyclical (∃ ! ∀ ! ∃) and a unidirectional (∃ ! ∀) change. I will contend that the diachronic cyclicity pertains to the indefinite origins

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Moreno Mitrović . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

transition using Chierchia’s () novel syntactic–semantic derivational/interpretational system for explaining polarity-sensitivity phenomena. I aim to demonstrate that the empirical facts we observe derive as unidirectional featural changes and natural parametric switches. In section ., I first outline the fundamental problem that investigations into particle syntax/semantics pose (namely that of particle ‘multifunctionality’). The critical notion of ‘allosemy’, based on Marantz’s () conception, is introduced. Section . then introduces a theory of scalar implicatures, based largely on Chierchia (), where the pivotal technical apparatus is presented which is used to navigate, and eventually explain, the relevant data from Indo-European and Japonic. In Section ., the Indo-European problem from Section . is cross-linguistically contextualized against the background of Japonic. Diachronic facts from Japanese are presented, showing that PSIs historically originate as universal terms. Finally, in Section ., the analysis of Japonic is transplanted and proposed for the diachronic treatment of quantificational expressions in Indo-European with an additional conjecture for the semantic affinity of quantificational meanings to pronominal meanings. Section ., as its title suggests, concludes.

. The problems of particle polysemy and solutions in particle allosemy .. Problems: at least three It has been recognized, at least since Gonda (), that the Proto-IE (PIE) particle *kw​e is problematic with regard to what it means: on the one hand, it perfoms the connective, or rather conjunctive, function and is translatable simply as ‘and’. On the other hand, however, *kw​e also has some non-connective and ‘epic’ functions such as additivity (‘also’), universal distributivity (‘every/each’), as well as negative polarity (‘any’). One of the aims of this chapter is to address this problem with the hope of answering the historical problem: which one of these functions, if any, was first? Or, as Gonda asks: The question may, to begin with, be posed whether we are right in translating Skt. ca, Gr.τε, Lat. que, etc., simply by our modern ‘and’ in regarding the prehistoric *kw​e as a conjunction in the traditional sense of the term. It is a matter of general knowledge that many words which at a later period acted as conjunctions originally, or at the same time, had other functions. (Gonda, : )

of the IE particle like *kw e, while the unidirectionality of change holds for non-pronominal μ particles which form quantifiers (universal quantifiers primarily and existentials secondarily).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



The quantificational meanings that *kw e and *kw e-like particles encode do not constitute a uniform semantic class.2 The quantificational function of *kw e in IE obtains when *kw e is attached to a wh-base. In such expressions, *kw e may have one of the two possible meanings: (a) the universal (∀) distributive meaning (‘each’), or (b) an existential (∃) negative polarity meaning. Another problem lies in the fact that a conjunction particle like *kw e is etymologically related to the interogative/pronominal wh-stem. (Gonda ; Dunkel, , , a,b) Therefore, the three problems of *kw e are in its seeming polysemy: ()

Three problems of *kw e i. The particle *kw e has a conjunctive function. ii. The particle *kw e has a quantificational function. a. The particle *kw e forms universal quantifiers. b. The particle *kw e forms negative existential quantifiers. iii. The particle *kw e has pronominal/interrogative origins.

As the first problem has been addressed elsewhere and since it has been shown how the conjunctive meanings derive from non-conjunctive meanings (Mitrović , forthcoming), I will not address (i) here. Instead, this chapter focuses on the second problem given in (ii).3 In figure ., the relation between these problems associated with the meaning of *kw e is sketched. Its three original functions, in the light of the philological evidence, and the informal relations for a unified analysis are given. Logically, the algebraic meaning of conjunction can be restated in terms of universal quantification (and vice versa). On the other hand, with regard to the allegedly pronominal origins of *kw e (Dunkel a,b, inter alia), the presuppositional meaning of a wh-expression is that of an existential quantifier which may be hypothesized as relating to the existential meaning of the PSIs built from *kw e - and *kw e -like particles. Under these two views, the question of how the semantic obverses derive can be stated.4

2

For instance, in Latin, wh-*kw e (viz. quis-que) encodes a universal quantifier, whereas the same wh-*kw e term (viz. kaś-ca) expresses a downward-monotone existential quantifier or a free-choice implicating existential expression in Indo-Iranian. 3 The second problem is in fact closely tied to the answer to the first problem. While this is explored elsewhere at length (Mitrović forthcoming), the conjunctive function of *kw e and other μ markers results from the abstract junction of two additive expressions. The additivity of *kw e /μ derives, however, in the same vein as *kw e/μ-marked universal quantification (namely, recursive exhaustification, as discussed in the following sections). 4 In stating the semantic multifunctionality of superparticles, I terminologically and conceptually follow Marantz’s () notion of allosemy. See Mitrović (), Mitrovć (forthcoming), and those cited there for details.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović The allosemy of the IE particle like ⟦∗kwe⟧= ii i

iii

conjunction marker

quantification marker

⟦and⟧

⟦quantifier⟧

✓⊨

⟦each⟧ ↓ ∀

⟦any⟧ ↓ ∃

pronoun

⟦wh–⟧

✓⊨

?⊨

FIGURE . A sketch of *kwe meanings: the top left node modelling the first problem, the middle node representing the second problem, and the third (dashed) node showing the third problem.

.. Aims: again, three The main purpose of this chapter is to understand how the quantificational split into universal and existential functions occurred (assuming it did). To answer this question, I resort to investigating similar quantificational splits and shifts beyond IE. Independent evidence suggests that IE languages with the universal function of *kw e, or rather *kw e-type particles, belong to the more retentive subfamily of IE, suggesting that polarity sensitivity, and the existential meaning of *kw e (*kw e-type particles) is the result of change (both morphosyntactic and morphosemantic). To support this view, I show that in Japonic (JP) NPIs are not historically primary but rather that they developed out of universal expressions of the same type which IE shows. While the diachronic patterns of change in JP and IE are not identical, I show how they can be treated using the same technical apparatus that is designed to explain the distribution of PSIs as triggering obligatory Scalar Implicatures (SIs). The conclusion I am led towards suggests that universals are indeed diachronically primary in these two language families, while NPIs are born from them via a particular syntactic (featural) change which has semantic-pragmatic reflexes. Terminologically, and conceptually, I refer to *kw e and *kw e-like particles as μ particles, for reasons given in Section ...

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



.. Superparticles Gonda’s problem, restated in more precise terms in () finds its empirical parallels in languages beyond IE. One such language is Japanese, which clearly shows the same multifunctionality of a single particle—hence in this language termed superparticle. The Japanese particle mo (も) can have both conjunctional and non-conjunctional meanings (i), hence our referring to IE quantifier particles as μ morphemes. Additionally, the combination of a mo superparticle with a wh-pronoun can yield both universal (distributive) or negative existential (NPI) expressions. This suggests that Japanese parallels IE with respect to the distributional semantics of the mo and *kw e particles, respectively. Given in () are the exemplar ‘superparticles’ in Japanese, whose semantics have been analysed by Kratzer and Shimoyama (), Szabolcsi (), Mitrović (), Mitrović and Sauerland (, ), among others. ()

The μ-series (mo/も) a. conjunction ビル (も) メアリー Bill mo Mary B μ M ‘(both) Bill and Mary.’

も mo μ

b. additivity メアリー も Mary mo M μ ‘also Mary’ c. universal quantification i. 誰 も dare mo who μ ‘each-/any-one’ ii. どの 学生 も dono gakusei mo  student μ ‘every/any student’ Old IE languages with respect to the function of *kw e fit the templatic pattern above seamlessly, with one exception: either the wh-μ (c-i) term is universal distributive (= ⟦each one⟧), or else negative polar (existential) (= ⟦anyone⟧). The simultaneous polysemy, or allosemy, of the Japanese type in (c–i) does not obtain in IE. Therefore the question of capturing this parameter is also cross-linguistically relevant.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

Note also that the quantificational expressions require an indefinite host, in both IE and Japanese.5 The semantic analysis is able to explain this universal generalization and why μ superparticles require indefinite hosts to form quantifiers (existential or universal). Indefinites can generally be thought as discrete disjunctions: ⟦who⟧ has the existential meaning of ⟦someone⟧, which in turn can be paraphrased as ⟦a or b or c⟧ for some toy domain of three individuals. The wh-terms are thus indispensable domain-restricting hosts for the μ superparticles. The relevance of the indefinites for the formation of quantifiers will also be made more explicit in the next section, where I introduce the background system and the technical apparatus that is required for the analysis

. The background system: a grammaticized view of scalar phenomena This chapter contributes to the diachronic syntax/semantics of quantification using a rather novel neo-Gricean view that implicatures are a grammatical phenomenon. Using the apparatus of deriving implicatures in the grammar (following essentially Chierchia ), I also demonstrate (in as much detail as space permits, and providing the reader with citations to more detailed discussion) how quantificational expressions can be derived in the same fashion. This explicans alone allows us to model the diachronic semantics of quantification in novel terms. In order to be able to do so, I devote this section to explicating the formal system which underlies the model I am adopting. To assist the reader in understanding the derivational and interpretational mechanics, this section concludes with an example of how a scalar implicature can be derived in the terms I am assuming and adopting.6 The theory I assume is that of grammaticized implicatures (Chierchia et al. ; Chierchia ; Chierchia ; inter alia), which convincingly contends that the locus of some, traditionally considered as pragmatic, phenomena lies in narrow syntax. Departing from the traditional view, according to which implicatures are treated as an entirely pragmatic phenomenon, this view argues that implicatures are derived in the grammar. Following a theory of grammaticized implicatures, I will suggest that polarity sensitivity may diachronically arise as a grammaticalized implicature. For Japonic, I suggest that the SI resulting from the negation of a universal term behaves like a PPI and grammaticalizes into an NPI, owing to a syntactic parameter change. The conception of such a change is the assumption that the relevant parameters are in fact triggers of pragmatic inferences. The Japanese particle-marked polarity system, 5 This requirement seems to be universal: see Haspelmath (), Jayaseelan (), Szabolcsi (, ), Mitrovć (forthcoming), among others for further discussion and cross-linguistic evidence. 6 For some recent evidence in support of the grammatical view implicatures, see, for instance Sauerland () and Schlenker () for an overview.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



which thus arose from an (existential) SI in Old Japanese (eighth c. ), is plotted as an instance of grammaticalization in terms of a Minimalist feature system. Under the assumption that such change is cross-linguistically natural, I will suggest that old Indo-European languages show the same diachronic pattern. Let us now turn to fleshing out the technical apparatus and assumptions, which are based on Chierchia (). Alternative-sensitive inferential processes, such as those associated with implicatures (including polarity sensitivity, freedom of choice implicatures, SIs) or presumably focus (Fox and Katzir, , et seq.), are anchored in feature specifications on syntactic terminals. There are two such relevant features: ½δ encodes the subdomain (non-scalar) ‘axis’ of alternatives and ½σ encodes the scalar dimension of alternatives. Using standard Minimalist assumptions (Chomsky , et seq.), such as Agree and Relativized Minimality (Rizzi, , ), syntactic objects bearing features ½σ and/or ½δ obligatorily undergo exhaustification against a suitable dimension of its alternatives. The dimension is guaranteed by an Agree relation between an exhaustification () operator, which performs contextual enrichment of the proposition, and the ½δ; σ-bearing grammatical formatives. I suggest that μ particles are such grammatical items. A structural template for μ-expressions is roughly, then, the following.

CP

() EXH

[μA : ] μP μo [iδ/iσ]

whP

Note that I do not commit to any specialized structural placement of μ-headed expressions here. Instead, as Mitrović and Sauerland (, ) suggest, the distribution of such expressions is best analysed using semantic type-theoretic and not syntactic category-theoretic criteria. Since μ particles are accordingly restricted to combinations with e-type elements, i.e. roughly, nominals (DPs), the μP is structured as not being directly part of the clausal spine. A root-level exhaustifier  thus probes for one or more goals carrying unvalued [σ;δ] features that provide its Alternative (A) restriction (σA, δA). Scalar terms (or, some, etc.) carry (unvalued) [σ;δ] features which may be targeted by exhaustifiers. Some lexical items, however, such as English any, German irgend-, or indeed the μ particles, obligatorily activate alternatives, i.e., their feature specifications cannot be without a positive setting for the feature bundle, i.e., [σ; δ] is an inadmissible feature setting. Once active, the alternatives must be exhaustified.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

The exhaustification operator combines with a proposition, and denies as many of its alternatives as it can. In more formal terms: ()

ðpÞ ¼ p ∧ ∀q ∈ Að pÞ½½ p ⊬ q ! ¬q (p is true and no (non-entailed) alternatives (q) to p are true)

An example of the system in action: ambiguous disjunction Here, I briefly sketch an example of how Chierchia’s () system works with its narrow-syntactic presence in order to derive the pragmatically enriched meanings. We draw an example from disjunction. A disjunctive sentence in English always carries an implicature: either an ignorance implicature (a) or a scalar implicature (SI) (b). Consider the following example.7 ()

Mary saw John or Bill. a.   i. [/ ] [ Mary saw John or[σ,δ] Bill. ] ii. ◊[ j] ∧ ◊ [b] ∧ ◊ [ j ∨ b] ∧ ◊ [ j∧ b] iii. ‘The speaker doesn’t know whether Mary saw John and the speaker doesn’t know whether Mary saw Bill and the speaker doesn’t know whether Mary saw John and Bill.’ b.   i. σA [ Mary saw John or [+σ,δ] Bill. ] ii. ½ j ∨ b ∧ ¬ ½ j ∧ b iii. ‘Mary saw John or Bill but not both.’

The disjunction in the sentence above makes relevant two possible dimensions of alternatives: either propositions themselves (δ-alternatives), or the conjunction of those propositions (its σ -alternative). There are two means of deriving the exclusive reading of () ()

Two ways of calculating the SI of () and deriving the exclusive component: j∨b assertion [⊓,⊔] { j, m} = j 𝔄 (5) = ℘

b j∧b

δ-alts σ-alts

i. Global calculation of the SI via exhσ exh[σ𝔄]( j ∨ b) = [j ∨ b] ∧ ¬[j ∧ b] ii. Local calculation of the SI via exhδ exh[σ𝔄]( j ∨ b) = exh(j) ∨ exh(b) ⊢ ¬[ j ∧ b]

7 In (a-iii), I imprecisely state the speaker’s not knowing, using an existential modal for convenience and exposition alone.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



The disjunctive example is also relevant for my purposes as I assume that the denotation of wh-pronouns, with which μ superparticles combine to yield universal/existential expressions, are themselves existential quantifiers. The latter are in turn equivalent to (discreet) disjunctions. Thus the bidimensional alternative set of () is on a par with an alternative set of a sentence featuring a bare wh-pronoun. As I suggest, recursively exhaustfying the δ-alternatives will allow us to capture the universal quantifier expressions which are built from wh-pronouns, as presented in Mitrović (); Mitrović (forthcoming). With the technical tools in place, I discuss the diachrony of the Japonic superparticle μ-system in the next section.

. The Japonic quantifier shift The oldest text in Japonic dates back to the eighth century  and allows us to see how the contemporary superparticle system () developed. Unlike this system, the Old Japanese (OJ) superparticle mo did not have the role of performing conjunction, or of encoding negative polarity.8 What is more, OJ μ shows that the stipulated ½σ feature and the general scalar dimension of meaning are empirically motivated. Section .. first shows the data in Old Japanese before the changes that occurred in the Classical Japanese period are addressed in Section ... .. Obligatory scalarity in the Old Japanese period In the earliest OJ corpus (Man’yōshū MYS, eighth century), the [wh+μ] quantificational expressions were confined to inherently scalar (σ) complements, as first noticed by Whitman (). Not only was the polar construction absent from the μ-system in the OJ corpus, but μ0 subcategorized for scalar hosts only. That is, the only eligible hosts of mo were either numeral nominals or inherently scalar wh-terms: how-many and when. The combination of a numeral n and μ, yielded the least likelihood reading along the lines of ‘even n’. In the wh-domain, the μ particle forms universal quantificational expressions, as shown below. As discussed later, this obtains through recursive exhaustification of the indefinite.

8

The details of the system presented in Mitrović (forthcoming) in fact predict the diachronic behaviour according to which the rise of conjunction or additivity is precluded in the absence of the δfeature on μ.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

Chierchia’s () system gives us the descriptive power to label this μ as carrying ½uσ, since non-scalar complements were disallowed:9 () 以都母 々々々 於母加 古比 須々 itu-mo itu-mo omo-ga kwopi susu when-μ when-μ mother- yearning by ‘I always, always think of my mother [i.e. at all times].’ (MYS, .; trans. Vovin : ) () 佐祢斯 [欲能 伊久陀 母] 阿羅祢婆 sa-ne-si [ywo-no ikuda mo] ara-neba -sleep- [night- how many μ] exist-- ‘As there have been few nights in which we slept together . . . ’ (MYS .a, ll. –) To buttress the fact that only scalar wh-terms were allowed, see table ., where counts of μ hosts are given. TABLE . Distribution of [] μ-hosts in OJ # of attestations SCALAR [wh+μ] itu mo ‘when μ’ iku mo ‘how much/many μ’ NON-SCALAR [wh+μ] ado/na/nado mo ‘what/why μ’ ika mo ‘how μ’ ta mo ‘who μ’

total    total    

The OJ mo, given its restriction to scalar complements, carries an uninterpretable σ-feature which sufficiently captures its distribution. ()

OJ: μ½σ

Given the fact that scalar μ-expressions in OJ under negation resist those negative inferences that can be made about NPIs, they may be analysed as Positive Polarity Items (PPIs). However, as Szabolcsi (: ) notes, ‘Some-type PPIs do not occur within the immediate scope of a clausemate antiadditive operator’. As shown in (), ywo-no ikuda mo occurs within the scope of clausemate negation (ne-). More 9 I contend that the ‘few’ reading of this OJ example belongs to the positive scale and does not represent evidence of super-strong negative polarity (in the terminological sense of Zwarts  and Van der Wouden ). I briefly state two arguments: first, there are no μ-marked NPIs in the MYS corpus. Secondly, the meaning of ‘few’ obtains via scale truncation in the sense of Israel () or Abrusán (: ). For discussion, see Mitrović (forthcoming: ch. ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



crucially, the term in non-negative contexts has a clearly universal meaning which weakens under negation but does not leave the positive scale. Thus, such constructions are best analysed as SIs. In (), an informal sketch of positive/negative inferences is given based on example ().
⇝some nights ðscalar readingÞ () ⟦½not ½all nights⟧ ¼ ¬⇝no nights ðpolar readingÞ In the next subsection, we turn to the classical period of Japanese in which the negative polar inferences were licensed. .. Two changes in Classical Japanese Two interlocked changes can be detected in the classical period. I address each of them in turn. ... The loss of obligatorily scalar complementation The first change concerns the loss of restriction on the type of complements that the Classical Japanese μ may associate with. Unlike in the Old Japanese period, Classical Japanese mo can be seen to freely associate with hosts of non-scalar type. One such example is in (), where mo associates with tare ‘who’ which has as a restriction set the non-scalar subdomain of all individuals. () たれ も 見おぼさん事 tare mo mi-obos-an koto who μ see.-think.-/ matter ‘the fact that everybody wanted to see’ (HM II:/; Vovin : ) Given the system we propose, the Classical Japanese μ is parametrized as having the ½δ feature both present and, in the case of (), set to a positive setting. Recall that Chierchia’s () systems of morphologically marked PSI requires that at least one of the ½δ; σ features be positively set. ... The rise of polarity sensitivity Chierchia’s () system, in fact, predicts that if both ½σ and ½δ are available features polarity sensitivity should obtain by definition of the system, ceteris paribus. As it happens, this is exactly what we find in non-archaic Japonic. The rise of μ-marked polarity sensitivity is evident from the data in () where a wh-pronoun nani ‘what’ with a non-scalar domain extension is interpreted under negation.10 10 Another aspect of change concerns the type-complexity of the μ- associate, considering the parallel between (c-i) and (c-ii). As opposed to bare wh-terms in OJ, Classical Japanese allowed [wh NP] complexes (taken to denote sets of type he;ti, as per Shimoyama ) to associate with μ. The same diachronic pattern can be found in the history of Chinese. (Mitrović and Hu )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

() いま は なにの 心 も なし ima Fa nani-no kokoro mo na-si now Top what- idea μ - ‘I do not have any thoughts [but of meeting you] now’ (IM XCVI: .; Vovin : ) Therefore, the Classical (early middle) Japanese μ-system can be analysed as bearing both ½σ and ½δ features, unlike OJ. ()

OJ: μ[σ]

t

σ

CJ: μ

δ This is motivated by the fact that both scalar and non-scalar complements featured in polarity-sensitive and scalar expressions. The novel possibility of non-scalar hosts associating with μ requires us to posit the relevant uninterpretable ½uδ feature on the μ particle, by virtue of which the negative polarity system arises automatically, as per the predictions of Chierchia’s () system. Therefore, the relevant change is that in the type of inference that wh-μ expressions carried. In OJ, such expressions are analysed as (positive) SIs, while the Classical Japanese wh-μ expressions were, or at least could be, NPIs under negation.11 In (a) and (b) I present this view of change in inferential procedure due to featural change, which I analyse as the signature property of the grammaticalization of μ in Japonic. () a. OJ: ¬ > ∀ ⊢ ¬∀ (SI)

i.

exh [uA : σ]

SI+ ¬

μ1P μ01 [uσ]

ii.

exh[σA] ¬ . . . [ μP ∃[+σ]

whP [iσ]

μ]

11 Note that in Chierchia’s () system, which is assumed here, both (positive) SIs and NPIs are derived using the same apparatus.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



b. CJ: ∀ > ¬ ⊢ ¬∃ (NPI)

i.

exh [uA : δ]

SI– ¬

μ1P μ01 [uσ/δ]

ii.

exh[σA] ¬ . . . [ μP ∃[+δ]

whP [iδ]

μ]

I will address in Section .. the technical details of deriving the meanings above. In the next section, I proceed to apply this technical analysis to the IE superparticle system by showing how this application derives the desired natural classes of NPIs and universal (PPI) terms we laid out in Section ..

. Indo-European quantifier shift The quantificational shift of meanings in IE is not identical to that I showed for Japonic, but may be modelled using the same principles. In Section .., I first show the nature of the quantifier split before proceeding in Section .. to providing a synchronic analysis of the two types of meanings underlying the quantifier split. .. Quantifier split and two superparticle meanings Earliest IE languages fall into two classes with respect to the interpretation of [wh-μ] expression. These superparticle meanings are consistent throughout early IndoEuropean (with Homeric -te being an exception) and are given in the table. The relevant classes, or properties, are boxed. TABLE . Semantic distribution of the meanings of μ markers across early Indo-European Language (family)

μ-marker

OC Slavonic (Slav.) Ṛgvedic (Indo-Iranian) Gothic (Germanic) Latin (Italic) Hittite (Anatolian) Tocharian B (Toch.) Old Irish (Celtic) Homeric (Greek)

i -ca -uh -que -(y)a -ra -ch -τε

Distr. (#∀)

NPI (∃)

  + + + + + 

+ +      

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

Assuming the PIE μ is reconstructable, in functional and not necessarily in the single morphophonemic form, table . suggests that a quantificational split took place in early IE with regard to the interpretation of the expression containing an indefinite wh-word and a conjunctive particle μ like *kw e. Note that the categorization of the μ category is based on the multifunctionality of a particle: i.e., if a particle x shows contextual allosemy (Marantz, ; Mitrović ) and differential interpretation, then x is a superparticle. In one group in table ., ⟦wh-term þ μ⟧ shows polar sensitivity and is generally translatable into the NPI ‘any’. This group is generally the satem one. As per traditional taxonomy, it is indeed Indo-Iranian (Ṛgvedic) and Balto-Slavic (Old Church Slavonic), to the exclusion of Tocharian, that shows NPI behaviour. In another group, on the other hand, ⟦wh-term þ μ⟧ is interepreted as a universal distributive, paraphrasable as ‘every’ or rather ‘each’. What is another striking property of the comparative morphosemantics of IE superparticles is that μ was rather universally operative (to the exclusion of Slavonic) in building Free Choice Items (FCIs). Perhaps less surprisingly, it is not coincidental that FCIs show both universal and existential behaviour (Chierchia, ). The analysis of FCIs in (P)IE falls outside the scope of this chapter. Let us now take each of the two groups in turn. For    , take two examples from Ṛgvedic in () and Old Church Slavonic in (). While the former NPI contains an enclitic (Wackernagel) μ clitic, Slavonic had a μ superparticle i that was not a Wackernagel element. The presence of the negative morpheme pricliticizing on to i is a negative concord reflex (Mitrović, ; Gajić, ). ()  &   (-) न

माया

na yasya [kaś-ca] atititarti māyā  . [..-μ] able to overcome illusions. ‘No one [= not anyone] can overcome that (= the Supreme Personality of Godhead’s) illusory energy.’ (Bhāgavatapurāṇ a, ..) ()    () n-i-česo-že -μ--=nothing ‘He answered nothing.’

otŭvěštavaaše answer. .. (Codex Marianus, Mt. :)

On the other hand, the   of IE languages, the same expression containing a wh-indefinite and a μ particle is interpreted as a universal quantifier. Consider the following examples from Latin () and Gothic ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



()  () auent audire quid [quis- que] senserit want.. hear. what. what- μ think.... ‘they wish to hear what each man’s (everyone’s) opinion was’ (Cicero, Philippicae ,) ()  () jah [hvaz- uh] saei hauseiþ waurda meina ‘And every one that heareth these sayings of mine . . . ’ (Codex Argenteus, Mt. :) Such universals expressions are insensitive to the polarity of the (local) environment they appear in. As () shows, the universal term is easily embedded in an antimorphic context in which it gives rise to a SI. ()  () ni 

[hvazwho..

uh] μ

saei ..

qiþiþ say..

imis I

frauja lord.

Frauja inngaleiþiþ in þiudangardja himine Lord. enter.. in kingdom.. heaven.. ‘Not everyone who says to me, “Lord, Lord”, will enter the kingdom of heaven’ (Codex Argenteus, Mt. : ) The question that is central to the concerns of this chapter is: Which of the two meanings, the  or the , is the primitive and reconstructable one? Just as in Japonic, I analyse IE μ as primitively expressing universal quantifications. While the diachronic Japonic data allowed for a clear view that polarity sensitivity and NPI expressibility arose in the classical period, the IE data does not show such a trend and requires reconstruction. Using Chierchia’s () model of grammaticized implicatures, we relegate the relevant semantic change, i.e. from the universal-distributive expression to PSI, to the change in the featural makeup of the μ-hosting wh-indefinite. One crucial aspect of this change and the rise of grammaticized polarity sensitivity is the availability of subdomain δ-exhaustification, the details of which we explicate in the next section. .. The two quantificational meanings of μ This section sketches the meanings μ-marked expressions have with special attention to quantificational meanings (the existential and the universal).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

Lexical items, such as any, -ever, all, also, and and are morphologically marked in many languages with a uniform μ morpheme. I have demonstrated that such meanings obtain in IE and JP as well.12 The semantics we ascribe μ states, informally, that μ superparticles have a dual semantic (or pragmatic) function. The first is to bring into play active alternatives which cannot be pruned by context alone. That is, μ superparticles, such as IE *kw e or (Old) Japanese mo, are presumed to activate obligatorily alternatives of their morphosyntactic associates. The second function is rather independent from the first: the grammatical system then acts on the triggered alternatives by applying the operation of exhaustification by virtue of a grammatically present (covert) formative, conceptually analogous to Rooth’s (, ) ‘’-operator that derives focus meanings. This  operator was introduced and overviewed in Section ., while this section applies its meaning in order to demonstrate, in rather informal terms,13 how such a semantic conception of superparticles successfully explains the range of meanings we find in IE and JP. As noted, the  operator, ‘filters’ activated alternatives by either denying them (in case  applies once) or asserting all of them (in case  applies iteratively; discussed below). What we adopt, then, is a syntactically present focus-sensitive exhaustification operator (itself essentially a silent variant of only). The lexical entry for μ in () (rather imprecisely) states the aforementioned dual function that μ particles have: alternative activation (second line) and exhaustification (third line) against the background of activated alternatives (A). ()

Lexical entry for ⟦μ0⟧

μP μ0

=

⟦μ⟧M, g, w ⟦XP⟧

XP =

⟦XP⟧

A

→ EXH ⟦XP⟧

⟦XP⟧

A

The core building-block of the semantics of μ will be alternative activation and exhaustification as proposed in Chierchia (). Exhaustification is taken to be a 12

See Mitrović (forthcoming) for a detailed empirical motivation of this view as well as details concerning the analysis briefly sketched here. 13 For an extensive treatment, see Mitrović (forthcoming) and those he cites.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



syntactically grounded pragmatic instruction to ‘run the Gricean reasoning’. We also adopt a more detailed instruction, ‘run the Gricean reasoning iteratively’, where we accept an iterative mode of application of the relevant maxims, as noted by Chierchia (: , n. ). The main reason for adopting this ‘extended’ Gricean reasoning and defining exhaustification iteratively (i.e., allowing  to apply iteratively) is that this iterativity characterization grants us a transition between exhaustivity and antiexhaustivity. As Fox () has shown, a double application of  returns ¬ and therefore allows us to see a natural switch between only and also (since not only = also). See Fox’s () for a detailed account and complete proof of this theorem. The range of meanings μ delivers along these lines as a result of applying  at a propositional (p) level, are the following, stated descriptively and informally. 8 () polarity if under ¬ > > > > FC if under ⋄ < ½δA ðpÞ ¼ ∀ if 2 and if μ½u > > > additivity if 2 > : ⊥ otherwise The differential meanings outlined above are amenable to an allosemic analysis, where the locality context for allosemy of μ is taken to be structures containing the relevant anti/licensing properties. For example, the presence of negation will ensure a single application of  takes place and that PSI meanings obtain, while the presence of an existential modal will obligate recursive exhaustification yielding an FC effect. If a language allows for recursive application of  in the absence of the negative of modalized contexts (2 above), then the μ-expressions can be universal quantifiers (if the μ-host is a wh-term) or additives (which are not addressed here; see Mitrović forthcoming). We now take the two relevant types of μ-meanings in turn: the NPI/PSI existential meaning and the universal quantificational meaning. Polarity In an exhaustification-based approach, polarity phenomena are derived in the following way. Our exemplar is based on (), which we repeat in (). () न माया na yasya [kaś-ca] atititarti māyā  whom. [who..-μ] able to overcome illusions. ‘[There is] No one [= not anyone] can overcome that (= the Supreme Personality of Godhead’s) illusory energy.’ (Bhāgavatapurāṇ a, ..; =) I innocently assume that the LF of () is analogous to ‘There is not anyone who can overcome’ (where I ignore the remainder of the sentence and the complexity of the object).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

 ()

Moreno Mitrović [[σA] [There is not [who[σ]-μ] who can overcome ]] =() a. : (¼p) ¬∃x ∈ D ½ðxÞ ∧ ðxÞ 0 0 b. AðpÞ ¼ f¬∃x ∈ D ½ðxÞ ∧ ðxÞ j D ⊂ Dg ½δA ðpÞ ¼ p (∵ all alternatives are entailed under negation)

In order to derive negative polar inferences, a single operation of exhaustification is sufficient to capture the distribution of NPIs which, in our empirical set, are formed by μ-marking the wh-expressions. The superparticle obligates the activation of the δ-alternatives of who (b). Once active, the alternatives require narrow-syntactic pruning which legislates the  -operator. The presence of  exhasutifies the NPI- (or μ-) containing proposition: it asserts the existential proposition and negates all non-entailed alternatives to that proposition. Given the negative polarity of the proposition, the polarity of entailment is reversed and all alternatives to the prejacent existential proposition are entailed, hence the original meaning of the prejacent is returned, as desired, since no alternatives may be be negated. The successful operation of the -operator is guaranteed as long as it targets and eliminates alternatives that are stronger than (and are entailed by) the assertion. What if they are not? In these cases, other principles are at play and the semantics of μ (allosemically) derives universal inferences. Distributive universals. In both negative and non-negative contexts, the μ in the   of IE languages derives universal quantifications. Hence, the presence of negation is not cached in at the point of SI calculation.14 The first ‘parameter’ relevant for the switch between negative-polar (existential) and universal μ-marked wh-constructions lies in the locality: polar constructions are sensitive to structural domains large enough to include negation, while the universal constructions are insensitive to the presence of negation in the relevant domain. I propose to capture this in/sensitivity to the presence of negation using ‘timing’. Before fleshing out the analysis, consider an exemplar in (), repeated in (). ()  () auent audire quid [quis- que] senserit want.. hear. what. what- μ think.... ‘they wish to hear what each man’s (everyone’s) opinion was’ (Cicero, Philippicae ,) I assume the relevant LF to be ‘They wish to hear everyone’ (ignoring the possessive complement for reasons of simplicity of exposition). The denotation of the 14 I assume, generally in line with Chierchia (), that polarity-sensitivity phenomena are special(ized) instances of scalar inferences.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

Quantificational cycles and shifts



wh-pronoun ‘who’, i.e. μ’s host, is taken to be an open existential quantifier, roughly denoting ‘someone’ (for whom a property P holds): ()

⟦who⟧ ¼ ⟦someone⟧ ¼ ðλpÞ∃x½ðxÞ ð∧ pðxÞÞ ¼ a ∨ b ∨ ::: ð∈ PÞ

In the absence of negation, which would always make sure that all alternatives are entailed, exhaustification leads to a contradiction. Assuming δ-exhaustification operates over discrete disjuncts provided by the (presuppositional component of the) denotation of the wh-pronoun quis ‘who’, the meaning we obtain would be analogous to ‘only a’ or ‘only b’, or . . . or ‘only x’ is such that they wish to hear them. ()

½½δA ½They wish to hear ½who½δ -μ ¼ ð22Þ

a. : (= p) ∀x ∈ D½ðxÞ ! -ð; xÞ 0 0 b. AðpÞ ¼ f∃x ∈ D ½ðxÞ ∧ -ð; xÞ jD ⊂ Dg c. ½δA ðpÞ ¼ p ∧ ¬p ¼ ⊥ (∵ exhaustifying the alternatives contradicts the prejacent) Given that this would possibly lead to a contradiction, exhaustification must apply once more. If the exhaustification procedure is considered as an instruction to ‘run the Gricean reasoning’, then the recursive optionality is a minor modification in so far as the speaker, when all else fails, has to ‘run the Gricean reasoning again!’ In the second step, the alternatives are pre-exhaustified (resulting from the first layer of exhaustification). The result of the second exhaustification is anti-exhaustivity, or additivity (informally, from only to not only).15 Now, each alternative is asserted in conjunction with the prejacent: ‘a and b and . . . and x is such that they want to hear them’, a reading fully compatible with the universal meanings these μ particles enforce in the  . () a. first-level exhaustification of A to the wh-pronoun: ½δA ða ∨ bÞ ¼ ðaÞ ∧ ðbÞ ð⊢ ⊥Þ b. second-level exhaustification of A to the wh-pronoun: R½δA ða ∨ bÞ ¼ ¬ ðaÞ ∧ ¬ ðaÞ⊢a ∧ bð⊢ ¬⊥Þ The reapplication of the  is not mandated if the structure in which μ features contains a modal operator—this gives rise to the Freedom of Choice inference, which, as I presented in table .. FCIs thus appear as an economical, or the most economical, means of delivering conjunctive inferences, which may well be the crucial factor for the widespread, if not universal, distribution of μ-marked FCIs in IE.

15

For an implementation and independent arguments for recursive exhaustification yielding antiexhaustive (additive) inferences, see Fox (), Mitrović (), Bowler (), Mitrović and Sauerland (), Szabolcsi (b, a), or indeed Mitrović (forthcoming: –).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

An anonymous reviewer wonders about the borderline between the synchronic and the diachronic roles of iterative exhaustification and the question of a principled restriction on it: why does it not continuously produce changes of the types discussed here? The reason for our not seeing changes resulting from the iterative mode of exhaustification is purely economical. Iterativity of  is a last-resort syntactic/ pragmatic option. (For a synchronic conception of the blocking economy, see Chierchia : ,  and, for a diachronic version of the same principle and further developments of it, see Mitrović forthcoming: ff.) .. A view of μ-development: a wider perspective Before concluding in the next section, I discuss another aspect of the history of μ-markers in IE. Philological literature contends that *kw e and *kw e -like μ superparticles originate as pronominal stems. How can we countenance the view that quantificational meanings diachronically derive from pronouns? In this subsection, I outline a conjecture for a view. The allegedly primary semantic nature of *kw e is puzzling in the light of philological evidence of indefinite/wh-cognates. However, it may be modelled using modern semantics: ⟦wh⟧↦ ⟦∃⟧, where the indefinite core of ⟦wh⟧ with regard to its presuppositional content is grammaticized into a ∃-quantifier. This is represented by ① in figure ..16 In this regard, the existential force of *kw e may be reconstructed and its reflexes in both quantificational and pronominal expressions brought closer. In the former, the existential quantification is part of the presuppositional component of meaning, while in the latter the existential is asserted. The evidence for the reflex of this stage can be found in Anatolian (cf. Hittite particles -ki and -ku), as per ② in the figure. As such, the particle is an ∃-quantifier which is subject to reanalysis along the lines of strengthening, conjecturally via an SI. In combination with wh-terms, an existential quantifier particle would trivially and vacuously assert the meaning that is contributed by the presupposition on the wh-term. As such, I conjecture that there may have been a procedure of the reversal of quantifier scales, i.e. a scalar shift. Presumably analogous, at least programmatically, to the proposed analysis for Japonic, this is seen at the stage which ultimately splits into ④ and ⑤ in the figure.

16 As a reviewer notes, bona-fide presuppositions do not normally become assertions. Since I am unable to submit any direct evidence for my unitary analysis of the pronominal and the quantificational functions of **kw e, I concede this criticism and provide a conjectural view of how the two disparate meanings throughout the unattested stages of IE may be diachronically interlocked. While Heine and Kuteva () note cases of indefinites grammaticalizing into markers of commonality, I leave for further research how the presuppositional content may grammaticalize into assertions within an alternative-based system. For a presupposition-based view of semantic change, see Beck and Gergel () and those they cite.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Quantificational cycles and shifts ⟦**kw–⟧ = ⟦wh⟧ 2

1

⟦*kwe⟧ = ⟦quant⟧

⟦*kwo⟧ = ⟦wh⟧

3

Anatolian Hit. –ki Hit. –ku

4

Gmc., Cel., etc. Goth. –uh OIr. –ch TA. –ra

5

IIr. Skt., Av. –ca CSI. i

FIGURE . A -step semantic-splitting pathway for wh-based semantic origins of superparticles in IE.

. Conclusions and outlook Assuming Chierchia’s () research programme, which rests on the simple observation that the distribution of SIs is a polarity-sensitive phenomenon, this chapter has shown diachronic reflexes of the Polar and Scalar system of pragmatic strengthening. Implicatures strengthen meaning by reducing the logical space of possible meanings, as do PSIs. Thus, treating PSIs as a type of implicature provides us with new ways of understanding their diachronic behaviour. In order to understand the semantic split of polar/universal meanings of early IE wh +μ terms, we have adopted a diachronic-comparative approach and developed an analysis of Japonic. In Old Japanese, the [wh +μ] quantificational expressions were confined to inherently scalar (σ) complements, i.e., either numeral nominals or inherently scalar wh-terms (e.g. how-many/when), as Whitman () first noticed. The combination of a numeral and μ, [μP n μ0 ], yielded ‘even n’. In this chapter, I focused on the latter and ignored the former numeral μ-hosts. The only two kinds of wh-terms which we find in OJ that may serve as μ-hosts are temporal- and quantity-wh-terms (), i.e. those wh-abstracts with only a σ-domain of alternatives. One of the ideas central to the proposal made in the chapter is that the original μ0 associated with scalar hosts, i.e. those elements endowed with the [σ] feature, and that the exhaustification of the scalar space of alternatives, as per Chierchia’s () system, delivered positive inferences. The prediction that scalar exhaustification of

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Moreno Mitrović

existential wh-terms makes is that, under negation, SIs should be borne our – as it is, indeed, the case in OJ. For IE, I have demonstrated that a similar mechanism was in play, arguing for a unidirectional change of quantificational change or shift. In regard to the cyclic view, I discussed in Section .. how the μ superparticles in IE may indeed be understood in light of the presumed origin of a particle such as *kw e. The pronominal/indefinite meaning of **kw e (not being a μ particle) allegedly had a purely existential make-up, through reanalysis of its presuppositional content into a quantifier and, subsequently, through strengthening into *kw e as a μ superparticle. When μ, originally deriving universals, develops into a marker of PSIs, the existential cycle is closed off. What the cyclical implications are beyond this view is something I leave for further research and to be integrated more widely within the general theory of cyclicity (van Gelderen , ). Formally, I have hypothesized a narrow syntactic featural change from ½þσ; ðδÞ to ½þδ; ðσÞ for the development of polar expressions of the same morphosyntactic structure. This has been confirmed by Early Middle Japanese (EMJ) where we encountered the rise of the polarity system. An additional and parallel reflex of this change is the shift in meaning from ‘even’ (as a scalar additive with a presumably intrinsic ½þσ½þδ feature) to ‘also’ (a non-scalar additive with ½þδ specification).17 A crucial aspect of the diachronic analysis reported here, as a reviewer reminds me, is whether the system I have used is sufficiently explanatory. I maintain that the exhaustification-based system predicts not only the rise of PSIs from universal quantifiers, but also the impossibility of the inverse. Since μ-associates in quantificational expressions are inherently existential, or discretely disjunctive, the -operator always returns a conjunction of the prejacent and the negated alternatives. The exhaustification system is correctly unable to associate with a universal and yield an existential, which is a desirable technical result fitting the empirical evidence. I have thus not shown how synchronically explanatory and predictively powerful Chierchia’s () exhaustification-based approach to polarity sensitivity and scalarity systems is. I have demonstrated how using it the semantic changes in Japonic and Indo-European may be modelled. It is also clear that there is (a rather unexplored amount of) room in historical linguistics for cross-linguistic diachronic analyses using such theoretical technology. With novel and theoretically motivated precision of viewing wh-μ terms in IE (= *kw o-*kw e ), I have presented a novel view of a ‘quantificational split’ in IE. Using a cross-diachronic filter from Japonic, I have modelled a view of diachronic evolution of such quantificational terms, and shown how synchronic and formal semantics/ pragmatics can be married with historical linguistics with very desirable results. 17

This independent piece of evidence is also relevant for the nature of development of conjunction marking, the investigation of which fell outside the scope of this chapter. See Mitrović (forthcoming) for details.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

10 On the relative cycle The case of P+che relative clauses from Old to Modern Italian C E C I L I A PO L E T T O A N D E M A N U E L A S A N F E L IC I

. Introduction In this chapter we investigate the relative cycle in the history of Italian with the aim of shedding light on the general principles of language change in the light of certain recent developments in syntactic theory, in particular Kayne’s () proposal that all heads are necessarily silent. In diachronic syntax many changes have been accounted for in terms of reanalysis of an element from a phrase into a head. A case that has usually been analysed as a spec-to-head change is that of relative pronouns, i.e. phrases, reanalysed as complementizers, i.e. heads, in accordance with the Spec-to-head Principle as formulated in van Gelderen (a: , ). Owing to an economy requirement granting that the structure be as minimal as possible, specifiers may be reanalysed as heads. This approach is challenged under recent theoretical proposals according to which heads must be silent (Kayne ; Cinque ). The existence of null elements has been part of formal syntactic analyses from the very beginning: whole theories of traces and ellipsis could not have developed without the postulate that some elements are not spelled out. Various works have argued for the existence of silent elements: there are syntactically represented but unpronounced elements, such as silent prepositions, verbs, nouns etc. (van Riemsdijk , ; Kayne ; Cinque ,  among many others). Building on these findings, Kayne () advances the proposal that all heads are necessarily silent, whereas overt material is merged as a phrase in the specifier position of one such head. He specifies that what was once interpreted as overt head morphemes (complementizers, articles, etc.) should be considered phrasal. In these phrases the single overt morpheme co-occurs with other silent elements. This can be exemplified by the treatment of the English complementizers that and for, which are analysed as

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Cecilia Poletto and Emanuela Sanfelici . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

[that NP] [WHAT for], with silent NP and silent what. If Kayne’s proposal is on the right track, as defended in Cinque (), this approach has major consequences for economy-based theories of language changes that make heavy use of spec-to-head reanalyses. This chapter proposes a detailed analysis of one particular case study which shows how an apparent instance of cyclical change can be interpreted differently, in a way that is consistent with the recent syntactic proposals in Kayne () and Cinque (). The case study investigates the distribution of three relativizers, i.e. che ‘that’, traditionally analysed as a complementizer in Modern Italian, i.e. a head, (il) qual(e) ‘the which’, and cui ‘that:’, both analysed as pronouns, i.e. specifiers, in relative clauses (RCs) throughout the history of Italian. The progressive expansion of the che form in (colloquial) Italian might be seen as a typical case of a cyclic change, which can be explained as the effect of a reanalysis. It is usually claimed that the relativizer che underwent a diachronic change from Old to Modern Italian: whereas it was a wh-pronoun in Old Italian, che underwent a reanalysis to a complementizer in Modern Italian in accordance with the relative cycle proposed in van Gelderen (a,b) (Benincà and Cinque ). This proposal can easily account for the compatibility of che with prepositions in Old Italian and conversely for its incompatibility in the modern stage of the language. In this chapter we intend to explore an alternative account to this economy explanation and connect this change to other independent changes of the language, namely to a change in the distribution of the null element COSA ‘THING’. In doing so, we show not only that our approach is consistent with the recent theoretical proposals, but also that it offers a solid, empirically grounded explanation for why an element like che has changed its distribution in the way we observe. Since the only evidence that che was a specifier in Old Italian and is a head in Modern Italian is the change in distribution with respect to prepositions, we will concentrate on this point. Building on Kayne () and Poletto and Sanfelici (, ), we argue that all the three relativizers mentioned above are of the same structural type, namely determiner-like elements merged as part of the RC-internal head at all diachronic stages of Italian, which never change their categorial status. Focalizing on the relativizer che we propose that the diachronic differences in its distribution can be accounted for by one single property, i.e. the type of NP it can be paired to.¹ Hence, instead of a change in the categorial status of the item, from a phrasal to a head-like element, we rephrase the observed relative cycle in terms of types of restrictors that the wh-item ranges over.

¹ We have no space here to consider Kayne’s () hypothesis that all complementizers are relativizers. This would imply an analysis of complement clauses, which we cannot undertake here. The material in this chapter is actually paving the way to such an account, since before proving that complementizers are relativizers in complement clauses, we have to understand what the exact syntax of elements like che is in RCs through the history of Italian. To solve part of this riddle is precisely the aim of the chapter.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



The work is organized as follows: in section ., we summarize the traditional analysis of the complex system of relativizers in Modern Italian, and on this basis we revise some problems underlying the notion of a linguistic cycle, and more specifically of a relative cycle, applied to the history of Italian. We then summarize the proposal put forward in Poletto and Sanfelici (, ), according to which Modern Italian che is still a determiner-like element, as it was in Old Italian, nothing having changed except for the feature composition of this element. This leaves one major question open, namely: if che has not changed its categorial nature, why was it compatible with prepositions in Old Italian but not in Modern Italian? In order to address this issue, section . presents and discusses the Old Italian data in comparison with Modern Italian with the aim to (i) define the real empirical basis we have to start from, since the generalization presented in the literature, i.e. that relative che in Old Italian can occur with all prepositions, is not clear. On this basis we can then establish what actually changes from Old to Modern Italian in P+che RCs and (ii) determine the distribution of che with respect to the other relativizers in prepositional RCs. We show three main points: a) the change from Old to Modern Italian affects the sequence P+che in lexically headed RCs only. When P+che occurs in RCs resuming VPs/CPs and in both bare as well as lexically restricted interrogatives, the sequence is still grammatical in Modern Italian. Hence, the Modern Italian ungrammaticality of P+che is restricted to contexts where movement of the lexical noun paired to che has applied. This clearly calls for an explanation that puts together interrogative che and relative che as determiner-like elements and not as distinct entities, as already proposed in Poletto and Sanfelici (). Moreover, when we compare che, cui, and (il) qual(e) in lexically headed PP RCs in Old Italian, the data show that they differ with respect to the type of prepositions they can occur with. Whereas che is mainly selected by functional prepositions, cui and (il) qual(e) occur with both functional and lexical prepositions. This asymmetry has gone unnoticed in the literature and should also be captured by a comprehensive analysis of the diachrony of Italian relativizers. In addition, when we compare che, cui, and (il) qual(e) in lexically headed PP RCs, the data show that they differ with respect to the type of antecedents they can resume. Whereas che can only appear with [–animate] antecedents, cui and (il) qual(e) occur with both [–animate] and [+animate] antecedents. Building on these results, our new proposal is put forward in section .. We claim that in Old Italian two relativization strategies were available: Type , where che is a determiner-like element paired with a lexical nominal, giving rise to P cui; Type , where che is a determiner-like element paired with a null nominal THING, as in free RCs, giving rise to P che. The change from Old to Modern Italian consists in the loss of the Type  derivation, namely the one similar to free RC, where che is paired with a null nominal THING. The fall and disappearance of TYPE  as a relativization strategy are due not to a change in the categorial nature of che, but rather to a change in the licensing conditions of the null nominal THING, which was

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

available in Old Italian but is not in the modern stage of the language.² This type of analysis is in line with Kayne’s () idea that some cases of apparent grammaticalization have to do with the null categories an element can be paired to. We further show that a number of predictions arising from this proposal are in fact borne out. In section . we draw some conclusions.

. Preliminary remarks In this section we summarize the state of the art regarding the diachrony of relativizers in Old and Modern Italian. In particular, the focus lies on the analysis put forward in Benincà and Cinque (), according to which the relativizer che underwent a categorial change from pronoun to complementizer, namely from specifier to head, adopting van Gelderen’s model (section ..). This shift is usually treated in the literature as the last step of the grammaticalization cycle of relativizers, more precisely as an instance of the Head Preference Principle and therefore a change which is the application of an economy principle so that the structure projected must be as minimal as possible. In section .. we discuss the advantages and the problems of applying the notion of cycle to relativizers. In section .. we briefly outline the assumptions we adopt for the RC derivation. .. The paradigm of Italian relativizers in diachrony: state of the art Modern Italian, as well as other Romance languages, displays a mixed relativizer system. Three types of elements can introduce a RC: a wh-element, namely (i) (il/la) quale lit. ‘the which’, which is inflected for number, (ii) an invariant wh-element cui ‘that:OBL’, and (iii) an element that also serves to introduce complement clauses, namely che ‘that’. Their distribution is sensitive to two factors: the type of relative clause, that is, restrictive or appositive, and the grammatical function of the relativized phrase. In restrictive relative clauses only che can relativize subjects and complements not selected by prepositions, as in (a,b), while (il) quale and cui are admitted only with PPs, cf. (c,d). On the other hand, in appositives (il) quale can always be used independently of the thematic role and case of the relativized element, cf. (a–d). () a. La ragazza che/*la quale ho incontrato ieri mi the girl that/the which have: met yesterday :. ha parlato di te. has talked of you ‘The girl that I met yesterday talked about you.’

² This is probably related to the fact that also lexical cosa ‘thing’ in OI had different properties with respect to the modern version, since it occurred as a polarity item in negative contexts, as we will see.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle b. Il libro che/*il quale è sul tavolo the book that/the which is on-the table ‘The book that is on the table is about clitics.’

parla talks



di clitici. of clitics

c. La ragazza con la quale/*che/cui ho the girl with the which/that/that: have: ieri si chiama Maria. yesterday : calls Maria ‘The girl with whom I talked yesterday is called Maria.’

parlato talked

d. Il libro del quale/*di che/di cui tutti parlano è interessante. the book of-the which/of that/of that: all talk is interesting ‘The book everybody is talking about is interesting.’ () a. Maria, che/la quale non vedo da oltre tre anni, Maria that/the which not see: from more three years arriva domani. arrives tomorrow ‘Maria, whom I haven’t seen for more than three years, arrives tomorrow.’ b. Roma, che/la quale è la capitale d’ Italia, Rome that/the which is the capital of Italy ‘Rome, which is the capital of Italy, is really beautiful.’

è is

c. Maria, con la quale/*che/cui ho parlato Maria with the which/that/that: have: talked arriva domani. arrives tomorrow ‘Maria, with whom I talked yesterday, arrives tomorrow.’

bellissima. beautiful: ieri yesterday

d. Roma, della quale/*di che/di cui tutti si innamorano, Rome of-the which/of that/of that: all : fall_in_love è bellissima. is beautiful: ‘Rome, which everybody falls in love with, is really beautiful.’ To account for the pattern in () and (), Cinque (, , , ) concludes that whereas (il) quale and cui are wh-pronouns, che is a complementizer, and as such it cannot be combined with prepositions, as (c,d) and (c,d) show. As in Modern Italian, in Old Italian headed relative clauses can be introduced by both che ‘that’, (il) quale ‘the which’, and cui ‘that:OBL’. Benincà and Cinque (: ) claim that whereas in the modern stage of Italian che is simply a subordination marker, in Old Italian che can in some cases be a complementizer but in some others also a relative pronoun. Old Italian che is a complementizer when it introduces NP-RCs, and in fact it does not show any sensitivity to the [
animate] distinction, as shown in (). Therefore, in these contexts Old Italian patterns with the modern language.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

() Andò alli altri giovani che stavano a ricevere l’ acqua Went to-the other: young: that stayed at receive the water ‘He went towards the other young people that were receiving the water.’ (Novellino, , l. –) However, Old Italian che can be combined with prepositions and in these contexts, it is a pronoun sensitive to the [
animate] distinction: when preceded by a preposition it can only appear with a [–animate] antecedent (). () uno bastone con che s’ apogiava perch’ era debole a stick with that : rested because was weak ‘A stick on which he rested because he was weak’ (Fiori e vita di filosafi, , –) As for the other two relativizers, Modern Italian and Old Italian pattern alike: (il) quale and cui are wh-pronouns, as can be shown by their ability to be combined with prepositions, () and () respectively. () Fermezza è virtù d’ animo per la quale l’ uomo sta Firmness is virtue of spirit for the which the man stays fermo in sul buon proponimento. firm in on-the good resolution ‘Firmness is a virtue of the spirit, by which the man remains fixed on a good intention.’ (Bono Giamboni, Vizi e Virtudi, ch. , p. , l. ) () a tutti choloro a chui s’ aparterrà questo libretto to all those to that: =belong: this booklet ‘to all those to whom this booklet will belong’ (Doc. fior., –, p. , l. ) Comparing Modern to Old Italian, Benincà and Cinque () conclude that che was both a head (in NP-relatives) and a specifier (in PP-relatives), i.e. a complementizer and a pronoun in Old Italian, and was subsequently fully reanalysed as a head, namely a complementizer in Modern Italian. Adopting van Gelderen’s (a, a,b) account of grammaticalization, the change in the system of relativizers in the history of Italian that Benincà and Cinque observed could be interpreted as an instance of reanalysis from a specifier into a head, under the Head Preference Principle. .. On the notion of relative cycle Roberts and Roussou () and van Gelderen (a, a) propose the existence of a relative cycle, according to which relative operators stem from either demonstrative elements or wh-indefinite pronouns, become relative operators, hence specifiers, and may subsequently be reanalysed as complementizer heads. The complementizer may be further reinforced by other elements in its specifier, usually

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



labelled as reinforcers (cf. Bácskai-Atkári , Bácskai-Atkári and Dékány ). The cycle is represented in (). () wh-indefinites/demonstratives > relative operators > complementizers > complementizers + reinforcement >… All steps in the cycle obey and are motivated by economy principles, which are formalized in van Gelderen (a) as the Head Preference Principle and the Late Merge Principle. Many diachronic phenomena can be and have been described adopting the notion of cycles. Here, we do not intend to dispute the fact that cycles are a descriptively useful tool, rather we would like to ask whether a theory of language change needs them or whether it can dispense with the notion of cycle. Although cycles seem to be a powerful tool to account for changes from an older grammar to a newer one, the notion of cycle is not as unproblematic as it seems at first sight. The first problem we face when dealing with cycles regards their nature. Although it is true that certain phenomena, such as negation,³ seem to behave properly in certain languages, other cyclical changes may be blocked at an intermediate stage, or under certain conditions an item may interrupt the path and pursue its own idiosyncratic path (Poletto ; Vedovato ). This points to the conclusion that these principles are to be seen as violable constraints rather than laws and that cycles should be stated as descriptive tendencies.⁴ More crucial for the purpose of the chapter are the causes motivating the cycle. At least two economy principles which point towards structure reduction motivate the nature of cycles: the Head Preference Principle (alternatively known as the Spec-to-head Principle) and the Late Merge Principle. However, not all the stages represented in the cycle can be accounted for in terms of economy, in particular the renewal via reinforcement step of the cycle as well as the return to a specifier, both of which are anti-economic: if it is more economic to use a head instead of a specifier, why is then the head deleted forcing the recreation of a new specifier, which will then undergo the cycle again? Finally, there are empirical and conceptual arguments against Late Merge principles (cf. Chomsky ; Castillo, Drury, and Grohmann ; Sportiche ). In conclusion, whereas cycles may be descriptively adequate, the motivations behind them are rather problematic. When applied to relativizers the notion of a cycle and in particular the proposal of a change from Spec to Head is problematic. It is usually assumed that RCs can be introduced by two distinct types of elements, namely (i) relative pronouns and ³ Actually, even negation shows cases in which some steps of the Jespersen’s cycle, i.e. the intermediate doubling stage, are not found (see Garzonio and Poletto ). ⁴ Furthermore, the notion of cycles relies on the assumption that language change proceeds in a single direction. However, critiques of any generalizations concerning the directionality of change have been raised by Lightfoot (e.g. , , ). Newmeyer () discusses various counterexamples to the unidirectionality of language change.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

(ii) complementizers. The reason for this distinction is that complementizers and pronouns display different behaviours with respect to three main properties (e.g. Klima ; Kayne ): (i) Case Marking: complementizers do not inflect for case, pronouns do; (ii) Sensitivity to animacy: complementizers do not carry animacy features, pronouns do; (iii) Compatibility with prepositions: complementizers cannot be combined with prepositions, pronouns can. These three characteristics have been considered in the literature to depend on a single abstract property, namely the fact that complementizers are heads, and as such, cannot be selected by prepositions, and do not possess typical (pro)nominal features such as case and animacy. More recently, however, many proposals have disclaimed that pronouns and complementizers differ in terms of their categorial status. Kayne () has argued that there is no distinction between relative and declarative complementizers, as both are relative pronouns (and not complementizers).⁵ A similar proposal has been put forward by Manzini and Savoia (, ) for Italian che, which, in their view, is not a complementizer, but a wh-item in all the contexts where it appears.⁶ Poletto and Sanfelici () argue that Italian ‘relative pronouns’ (il) qual(e), cui, and ‘relative complementizers’, che, do not differ in categorial terms. They are all determiner-like elements. When these approaches are taken into account, we face the immediate challenge represented by the incompatibility of che with prepositions in Modern Italian. If che has never changed its categorial status through the history of Italian, then the ungrammaticality of P+che in Modern Italian and on the contrary its grammaticality in Old Italian discussed in section .. cannot be due to the reanalysis of che from pronoun to complementizer. Thus, this change must be accounted for in a different way. .. The RC derivation We adopt a derivation of RCs as proposed in Cinque (, ), and refined in Poletto and Sanfelici (, ). Restrictive and free RCs are merged as CPs in the specifier of a prenominal functional projection above the projections which host attributive adjectives and numerals and below the projections hosting determiners and demonstratives (Kayne ; Cinque : , ). In contrast, appositive RCs are merged above the determiner layer (Cinque ). All RCs are double-headed: an external head is base-generated as the complement of the functional projection that hosts the RC in its specifier, dP2 in (); an internal one is base-generated inside the ⁵ Under Kayne’s () view, complementizers do not exist, i.e., there is no such category that is merged in the head of C, since there are no overt heads. This implies that all complementizers, including those of complement clauses, have to be reanalysed as phrasal categories. In this chapter we remain agnostic as to whether che in complement clauses also is a relative pronoun or not. The only crucial point here is that the relative che is syntactically similar to the determiner che in interrogative clauses. ⁶ For a similar proposal see Sportiche () on qui/que in French. The idea that complementizer que and interrogative que in French are the same element was already present in Obenauer ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



RC, dP1 in () (Cinque , ; Hulsey and Sauerland ). These two heads are identical or non-distinct copies (as in Sauerland ). We exemplify the derivation of a restrictive RC in ().

DP

()

FP1

D0 The

FP2 CP

Internal Head

dP1 Wh ragazzo C0

TP

dP2

External Head

ragazzo

Our analysis builds on Kayne’s version of c-command according to which an XP in Spec, FP can c-command out of FP (Kayne : , –). Following Cinque (), the spelled-out head will be the dP that c-commands the other head: it surfaces at PF and controls the PF deletion of the other dP. If the internal head rises higher than the external one, the internal head c-commands the external one, which is then deleted. Hence, dP1 is spelled out. If the external head moves higher and thus c-commands the internal one, the external head, dP2, is spelled out, and the internal one is deleted at PF. Under this approach, both raising and matching operations are in principle available. In a raising derivation, the internal dP must move from being the complement of the wh-item to being its specifier: being the specifier of the phrase which in turn is in the specifier of the CP, the internal dP1 can c-command the lower external dP2 and meet the requirement for deletion proposed in Sauerland (), i.e. a) non-distinctness and b) c-command. Hence, the identification requirement can be satisfied only by stranding the wh-determiner and moving its associate nominal expression, ragazzo in (), to the position where it can locally c-command the external dP2, namely Spec,CP. Stranding the wh-item has the semantic reflex of creating a predicate of the type out of a proposition (Heim and Kratzer ). Under a matching derivation as in Sauerland

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

(), after the raising of the internal head dP1 in (), the external head dP2 moves to a position higher than that of the RC. From this position dP2 c-commands dP1 and regular deletion of the lower copy, i.e. dP1, applies. As for free RCs, we further assume that the external head is a silent nominal expression: THING, PERSON, TIME, PLACE (cf. Kayne ; Cinque , ). Similarly, the nominal element associated with the wh-item, hence the internal head, is a null nominal category, non-distinct from the external one. Under this framework, relativizers come in one single guise, i.e. as determiner-like elements located in a high specifier position of the relative head as in ().⁷ Crucially, in our view, this derivation holds for all the three types of relativizers: che, il quale, and cui. In the remainder of the chapter we will propose an account for the facts outlined in section .., which incorporates these assumptions on RC derivation.

. The distribution of P+che in Old and Modern Italian The questions we need to raise regard the realm of P+che, both quantitatively and qualitatively, namely: (i) what has actually changed from Old to Modern Italian (section ..), and (ii) what is the distribution of P+che, P+cui and P+(il) qual(e) in Old Italian (section ..). .. What changes from Old to Modern Italian In order to address the first question we made a survey of the OVI database via Gattoweb and we searched for the occurrences of che when tagged for the grammatical category ‘relative pronoun’ (rel. in Gattoweb).⁸ The search was conducted on the Old Florentine sub-corpus obtained selecting the specific area as Florentine ( fior. in Gattoweb, which contains  texts from   to ). In the sub-corpus, the total of occurrences of relative che amounted to . Out of these,  occurrences were found of P+che. We further analysed these  occurrences, classifying them into two types, depending on the presence or absence of an antecedent. The former cases are labelled as lexically headed RCs (), and the latter as -headed RCs ().

⁷ As a reviewer pointed out, the form il quale is made up of two parts: the wh-adjective qual(e) and the definite article. According to Cinque’s () proposal the RC-internal head is an indefinite XP, which has been translated in Poletto and Sanfelici () as a dP. It still remains unaddressed how and where the definite article of il quale is generated, but we point out that there are recent accounts where the determiner is not generated in the highest D° position (see Ihsane ). We will address this issue in the near future. For the time being, we refer the reader to the analysis of il quale proposed in Cinque (). ⁸ The databases used for our corpus searches are the following: / OVI Opera del vocabolario italiano. University of Chicago: http://artfl-project.uchicago.edu/ / Gattoweb: http://gattoweb.ovi.cnr.it/(S(lrfgoiinqhprqkvga))/CatForm.aspx / Parsed corpus of Old Italian. Università di Padova (contact: Emanuela Sanfelici).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle ()



e molti altri convenenti v’ àe di tutte le chose and many other: circumstances : are of all the things di che nnoi fumo in concordia of that we were in agreement ‘And many other circumstances are there of all the things upon which we agreed as contained in this paper.’ (Libro giallo, –, p. , l. )

() Questo signore verrà, e tutti i cittadini troverrà this lord come: and all the citizens find: divisi; di che grande scandalo ne seguirà. divided of that great scandal cl follow: ‘This lord will come and find all the citizens divided; from which great scandal will follow.’ (Dino Compagni, Cronica, –, L. , ch. , p. , l. ) In table . we summarize the results of our search showing the number of occurrences of P+che divided per text, date, and type of RC.

TABLE . Overview of the number of P+che occurrences divided per text, date and type of RC Text

Date

Dino Compagni, Cronica Stat. fior. Libro giallo Alberto della Piagentina Boccaccio, Caccia di Diana Valerio Massimo, red. V Libro vermiglio Boccaccio, Filocolo Boccaccio, Teseida Libri astron. Alfonso X Boccaccio, Ameto Valerio Massimo, red. V Boccaccio, Ninfale Ceffi, Dicerie Tavola ritonda Doc. fior. Stat. fior. Rubriche Commedia Jacopo Passavanti, Tratt. Scienza Jacopo Passavanti, Tratt. sogni Jacopo Passavanti, Tratt. superb.

– / – /   – – –  –  / XIV pm. XIV pm. –  –   

Lexically headed RCs

-headed RCs

   



             

    



Total N                      (continued )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

TABLE . Continued Text

Date

Jacopo Passavanti, Tratt. umiltà Jacopo Passavanti, Tratt. vanagl. Diatessaron Matteo Corsini Doc. fior. Petrarca, Trionfi Paolo dell‘Abbaco, Regoluzze Boccaccio, Rime Chiose falso Boccaccio, Purg. A. Pucci, Centiloquio A. Pucci, Noie A. Pucci, Reina Torini, Brieve meditazione Sacchetti, Trecentonovelle Sacchetti, Rime

    – (?)–       / XIV sm XIV sm

Lexically headed RCs        

  

-headed RCs

  

    

Total N               

In table . we quote the texts following the abbreviation system provided by the OVI database and adopted by Gattoweb.

As shown in table ., the overall number of occurrences of P+che is rather small, with a mean of . occurrences per text. Sixty-three occurrences of P+che exhibit a lexical antecedent, whereas thirty-two are instances of P+che as -headed RCs. We further investigated the type of prepositions selecting che. In -headed RCs, che occurs with the following prepositions: con (N = ), di (N = ), di plus the definite article (N = ), in plus the definite article (N = ), per (N = ), secondo (N = ). In all these instances, che resumes the preceding event or proposition, and hence a VP or a CP, as can be shown by the following examples. () ingenerò gran discordia tra i Fiorentini e ‘l generate:sg great contention between the Florentines and the Vescovo e i Ghibellini. Per che ne seguì la terza Bishop and the Ghibellines for that cl follow:sg the third guerra de’ Fiorentini in Toscana, nel . war of-the Florentines in Tuscany in-the  ‘Great contention was generated between the Florentines, the Bishop, and the Ghibellines. For which the third war of the Florentines followed in Tuscany in .’ (Dino Compagni, Cronica, b., ch. , p. , l. )

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle () I suoi occhi, the his eyes purpureo colore purple colour ‘His eyes, red from amazed’



tornati per le lagrime rossi, erano d’ un returned for the tears red were of a intorniati: di che egli si maravigliò molto surrounded of that he cl. amazes a_lot tears, were ringed with a purple colour: at which he was (Boccaccio, Filocolo, b. , ch. , p. , l. )

Notice that Modern Italian still admits this use of P+che.⁹ Hence, the change from Old to Modern Italian did not affect the sequence P+che when che refers to a VP/CP. In lexically headed RCs, che occurs with the following prepositions: a (N=), con (N=), di (N=), in (N=), per (N=), sopra (N=), su (N=). () come gli altri artefici cognoschino quelle cose che sono as the other: authors kwon those things that are da fare e a che sono costretti to do and to that are bound ‘as the other authors would know those things that are to be done and to which they are bound’ (Statuto fiorentino, ch. , p. , l. ) () e altressì per li cieli in che elle stanno che si and also for the heavens in that they stay that : muovono sempre. move always ‘and so much for the heavens in which they stay so that they are always moving.’ (Libri astronomici di Alfonso X, Prologue, p. , l. ) () dando delli sproni a un ronzino su che era giving of-the spurs to a nag on that was ‘giving spurs to a nag on which he was’ (Sacchetti, Trecentonovelle, p. , l. ) All these sixty-three occurrences lead to ungrammaticality when translated into Modern Italian, suggesting that it is only P+che in lexically headed RCs that is ungrammatical in Modern Italian. Hence, the change from Old to Modern Italian does not lie in the sequence P+che per se, which is possible in both CP/VP relatives, but rather in the occurrence of P+che just in lexically headed RCs. In other words, since in Modern Italian che still functions as a relative pronoun resuming VP or CP, it is not the categorial status of che that underwent a change, but the type of referents it can resume. This conclusion seems to be further supported by the fact that Modern Italian, like Old Italian, allows the sequence P+che in other contexts. In Modern Italian che can indeed occur with prepositions, both when it functions as a determiner-like element and when it functions as a pronoun, i.e., when ⁹ The only exception is that of per che in (), which will have the form per cui or perciò in Modern Italian.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

it is bare or paired to a null or lexical classifier cosa ‘thing’ (b). This is the case with interrogative (and exclamative) clauses. () a. Con che ragazzo parli? with that boy talk: ‘With which boy are you talking?’ b. Di che (cosa) parli oggi? of that (thing) talk: today ‘About what are you talking today?’ The examples in () demonstrate that che is indeed compatible with prepositions, but it is so only in two contexts: as a determiner-like element with a following nominal expression, as in (a), or as a bare element paired to a null or lexical ‘thing’ (as in (b)). We capitalize on this observation and assume that che in Modern Italian is a determiner-like specifier that requires a lexical or null N as its domain of quantification. Moreover, the examples in () as well as in (–) show that the ungrammaticality of P+che in Modern Italian cannot be due to case-morphological differences between Old Italian che [+oblique] and Modern Italian che [+direct cases]. In Modern Italian che is also compatible with oblique case morphology in the contexts above. Moreover, these examples also show that the ungrammaticality of Modern Italian P+che cannot simply be due to the impossibility of che occurring with prepositions, but must be related to the movement, in the case of raising, and the deletion procedure, in the case of matching, which the lexical head noun has to undergo, which is not present in interrogative clauses () and in VP/CP RCs (–). .. The relativizers in lexically headed PP-relatives Having clarified what the real change is from Old to Modern Italian, namely the disappearance of P+che only in lexically headed RCs, the second question we raise regards the distribution of P+che, P+cui, and P+(il) qual(e) in lexically headed PPrelatives in the Old Italian texts. In this chapter we consider the distribution of three relativizers, che, cui, and (il) qual(e) only with respect to two factors: (i) the type of prepositions occurring with the relativizers; (ii) the animacy of the antecedents.¹⁰ In order to investigate whether the three relativizers differ with respect to the type of prepositions with which they can occur, we made a survey of the OVI database via Gattoweb and we searched for the occurrences of cui and (il) qual(e) when tagged for the grammatical category ‘relative pronoun’ (rel. in Gattoweb). The search was again conducted on the Old Florentine sub-corpus obtained selecting the specific area as

¹⁰ On other syntactic differences (in terms of extraposition, reconstruction effects, pied-piping) we refer the reader to Sanfelici and Poletto ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



Florentine (fior. in Gattoweb, which contains  texts containing cui and  containing qual(e) from   to  ). The total of occurrences of cui amounted to . Out of these,  occurrences were found of P+cui with P overtly lexicalized.¹¹ We found  occurrences of -headed RCs and  occurrences exhibited P+cui with a lexical antecedent. We further investigated the type of prepositions which select cui in lexically headed RCs. The prepositions in these contexts are the following: davanti (N=), contro/a (N=), con (N=), appresso (N=), di (N=), a (N=), da (N=), in (N=), per (N=), a guisa di (N=). The total number of occurrences of qual(e) was . Out of these,  occurrences were found of P+qual(e). In  instances, (il) qual(e) resumes the preceding event or proposition, hence a VP or a CP, demonstrated by the presence of cosa ‘thing’ following the relativizer. In  instances P+qual(e) occurs in lexically headed RCs. The prepositions in these contexts are the following: in su (N=), in tra/in fra/fra/tra (N=), in/in plus definite article (N=), di/di plus definite article (N=), a/a plus definite article (N=), per/per plus definite article (N=), con/con plus definite article (N=), da/da plus definite article (N=), senza/sanza (N=), dopo (N=), oltre a (N=), sopra (N=), contro (N=), secondo (N=). Comparing the type of prepositions that can occur with both cui and (il) qual(e) with those that select che in lexically headed RCs, we can formulate the first generalization. () Generalization : type of prepositions Whereas the prepositions selecting cui and (il) qual(e) are both functional and lexical ones, the prepositions occurring with che are mainly restricted to the functional type. We then investigated whether there were differences in terms of [animacy] of the referents. The total of PP-relative clauses extracted from the OVI database via Gattoweb is . Considering the animacy of the antecedent, we obtained the distribution outlined in table .. TABLE . Number of PP-RCs divided per type of relativizer and animacy of the antecedents (Gattoweb survey) Type of relativizer (il) qual(e) Che Cui

[+animate] antecedents

[–animate] antecedents

Total N

 (.%)   (.%)

 (.%)  (%)  (.%)

  

¹¹ In the  occurrences, we did not include the occurrences of cui as indirect object without the preposition a, and of cui as genitive without the preposition di. These two types of cui were counted as preposition-less cui.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

Since not all the OVI texts are fully coded according to the grammatical category of the lexemes/lemmas and in many cases the coding was manually done by us, we further made a survey of the syntactically annotated corpus of Old Italian (cf. footnote ), which at the moment contains eleven Old Italian texts from  to , with a total of , words. We queried all the tokens of CP-REL*, which immediately dominates a PP containing a wh-element, i.e. WPP*, namely all relative clauses attached to PPs. We extracted a total of  tokens, of which  instances were excluded because although the trace was a PP, the preposition was not overtly lexicalized. The remaining  tokens were further analysed. The distribution we observed is given in table .:  tokens of P+cui;  P+che;  P+(il) qual(e). TABLE . Number of PP-RCs divided per type of relativizer and animacy of the antecedents (Corpus survey) Type of relativizer (il) qual(e) Che Cui

[+animate] antecedents

[–animate] antecedents

Total N

 (.%)   (.%)

 (.%)  (%)  (.%)

  

On the basis of tables . and ., we can put forward the second generalization.¹² () Generalization : type of antecedent Whereas P+che only occurs with [–animate] antecedents, in the Old Italian texts P+cui and P+(il) qual(e) refer to both [–/+animate] antecedents.

. Our proposal In the previous section we analysed the Old Italian data in comparison with the Modern Italian ones. Three conclusions were reached. First, the change from Old to Modern Italian affects the sequence P+che in lexically headed RCs only. When P+che occurs in RCs resuming VPs/CPs, and in both bare as well as lexically restricted interrogatives, P+che is still grammatical in Modern Italian. We take this result to suggest that the Modern Italian ungrammaticality of P+che is restricted to contexts where movement or deletion of the lexical noun has applied (cf. section ..). Second, when we compare che, cui, and (il) qual(e) in lexically headed PP RCs, the data show that they differ with respect to the type of prepositions they can ¹² Notice that P+cui refers to [+animate] antecedents in the great majority of cases (cf. section .), whereas P+(il) qual(e) is usually combined with [–animate] antecedents. We can speculate that the high proportion of P+(il) qual(e) referring to [–animate] antecedents is a signal of the fact that P+che was already disappearing as a relativization strategy. Since the use of P+che was restricted to certain prepositions, as suggested to us by an anonymous reviewer, we may speculate that P+(il) qual(e) was supplanting it in the [–animate] antecedent contexts.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



occur with. Whereas che is mainly selected by functional prepositions, cui and (il) qual(e) occur with both functional and lexical prepositions. Third, when we compare che, cui, and (il) qual(e) in lexically headed PP RCs, the data show that they differ with respect to the type of antecedents they can resume. Whereas che can only appear with [–animate] antecedents, cui and (il) qual(e) occur with both [–animate] and [+animate] antecedents. In order to account for our results, we propose that two strategies of relativization were available in Old Italian, one of which, i.e. Type , was lost in the change from Old to Modern Italian. () Relativization strategies TYPE : P+cui insensitive to the animacy of the antecedent, which is continued in the modern stage of the language. TYPE : P+cui dedicated to [+animate] antecedent in opposition with P+che reserved for [–animate] antecedent, which is unavailable in Modern Italian. The other context in which the sensitivity of the antecedent regulates the distribution of different types of relativizers in Old Italian is that of free RCs, in which P+cui is the form dedicated to [+animate] antecedents (a) in opposition to P+che, which occurs only with [–animate] antecedents (b). () a. L’ essemplo basti / a cui esperïenza grazia serba. the example suffices to that: experience grace reserves ‘The example may be sufficient for those to whom experience reserves grace.’ (Dante, Commedia, Par. , v. ) b. Messere, tu non ài con che tu atinghi, e Sir you not have with that you draw_water and il pozzo è alto. the well is high ‘Sir, you do not have anything with which you could draw water and the well is high.’ (Diatessaron, ch. , p. , l. ) We therefore formalize our claim by proposing that Type  and Type  result from two different derivations. In (), che is a determiner-like element paired with a lexical N, i.e. the RC-internal head, and after movement or deletion of the N, che spells out the oblique case as P cui. In (), the lexical N bastone is the RC-external head, che is a determiner-like element paired with a null nominal THING, as in free RCs, giving rise to P che. We represent only the crucial points of the derivations.¹³

¹³ It is very tempting to propose that Type  is the older derivation strategy and that it was supplanted by Type , as a reviewer also suggested to us. This intuition is supported by the high proportion of P+cui referring to [+animate] antecedents, as well as by the presence of P+che referring to [–animate]

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

() [DP D [RC [PP a [che bastone]]]…bastone]

! P cui

TYPE 

() [DP D [RC [PP a [che THING]]]…bastone]

! P che

TYPE 

This proposal immediately accounts for the observation that in lexically headed PP relatives the antecedents of P+che can only be [–animate], as they are in free RCs. Since the null element paired to the determiner-like che is THING, which clearly is [–animate], only [–animate] antecedents can be compatible and hence, resumed by THING. Furthermore, it offers a model to formalize the result found in section ., according to which Modern Italian P+che is ungrammatical when resuming a lexical noun but grammatical in RCs resuming VP/CP. The derivation of Type  was lost in the change from Old to Modern Italian. This proposal resolves the tension of having to postulate three categories of che: che complementizer in both Old and Modern Italian as in NP relatives, che pronoun in Old Italian which becomes a complementizer in Modern Italian as in lexically headed PP relatives, and che pronoun in both Old and Modern Italian VP/CP relatives. Moreover, we can account for the high proportion of P+cui with [+animate] antecedents in tables . and .: according to our proposal, the sequence P+cui with [+animate] referents is ambiguous between being the output of the first and that of the second type of derivation. Notice that under this approach we can account for the presence of both P+cui with [+animate] antecedents and P+cui with [–animate] antecedents. The former would be the result of either the Type  or the Type  derivation, whereas the latter would only be the result of Type . If our proposal is on the right track, then Type  relativization should exhibit a distribution similar to that of free RCs. This seems to be confirmed by the fact that the set of prepositions which can select che in lexically headed PP relatives and in PP free RCs is the same (see section ..). In free RCs we observe mainly functional prepositions in the Old Italian texts we analysed.¹⁴

antecedents. Although this might well be the case, this proposal remains at the level of speculation. Whereas it is clear that Type  supplanted Type , their relative chronology (i.e., whether Type  is as old as Type ) still remains open. ¹⁴ As we stated in Generalization , che is selected by functional prepositions, whereas both cui and il quale occur with both functional and lexical prepositions. We believe that the difference has to be spotted in the different structures of the PPs with functional and lexical prepositions (cf. Cinque and Rizzi ) and the anaphoric force of the wh-element. Whereas in a PP with a functional preposition there only is one PP layer, in a PP with a lexical there are at least two PP layers, one hosting the lexical preposition and the lower hosting the functional preposition. The fact that che is almost never found under lexical prepositions is taken as evidence of its weak status (as proposed in Poletto and Sanfelici ). This holds for che in both lexically headed RCs and free RCs.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



() Egli è a’ ricchi con che [la fame ] sazino, con It is to-the riches with that the hunger sate: with che [il freddo e la sete] discaccino. that the cold and the thirst dispel: ‘It is to the rich something with which they can sate hunger, with which they can dispel [cold and thirst].’ (Alberto della Piagentina, L. , ch. , p. ) () E i farisei stavano attenti s’ egli il curasse in and the Pharisees stay careful if he : heal.. in sabato, acciò che avessero di che l’ accusare. Sabbath at-that that have: of that : accuse ‘And the Pharisees paid attention as to whether he would heal him on the Sabbath, so that they could have something to accuse him of.’ (Diatessaron, ch. , p. , l. ) () E quegli sogni che si fanno intorno all’ alba del and those dreams that : make around at_the dawn of_the dì, secondo che dicono, sono i più veri sogni… day according_to that say: are the more true dreams ‘And those dreams which occur at daybreak, according to what they say, are the truest dreams…’ (Jacopo Passavanti, Trattato de’ sogni, p. , l. ) Furthermore, if Type  involves a free RC, then we expect that, since P+che is ungrammatical in Modern Italian lexically headed RCs, it should also be so in free RCs. The prediction is borne out. True free RCs on [–animate] referents are ungrammatical in Modern Italian.¹⁵ () *Oggi vado in libreria e compro che (cosa) mi today go: in bookshop and buy: that thing :. hai consigliato have: suggested ‘Today I shall go to the bookshop and I shall buy what you suggested to me.’ This suggests that the fall of Type  as a relativization strategy is due not to a change in the categorial nature of che, but rather to the conditions licensing the null nominal THING, which was available in Old Italian but not in the modern stage of the language. This is why both P+che and free RCs with [–animate] antecedents are ungrammatical in Modern Italian.

¹⁵ We leave out for the moment the existential modal wh-constructions in both Old and Modern Italian of the type non ho chi si prenda cura di me ‘I don’t have who could take care of me’ (cf. Grosu ; Šimík ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

The third prediction our hypothesis makes regards the type of lexical antecedents in P+che RCs. When a free RC denotes a [–animate] referent, Modern Italian has to resort to light-headed free RCs (): () Oggi vado in libreria e compro quello/ciò che mi today go: in bookshop and buy: that/this that :. hai consigliato. have: suggested ‘Today I shall go to the bookshop and I shall buy what you suggested to me.’ Benincà (, ) and Cinque (, ) analyse free RCs as DPs with a light head of the type THING, PERSON, which in fact is spelled out in Modern Italian as cos(a/e) ‘thing/s’ or with a demonstrative such as quello ‘that’. This pattern is already attested in Old Italian texts, in which the head of a [–animate] free RC is often spelled out, mostly with the same distal demonstrative as exists in Modern Italian. () Forse ancora ciò ch’ io dico non intendi. maybe yet this that I say not understand: ‘Maybe you do not yet understand what I say.’ (Alberto della Piagentina, ch. , p. , l. ) () Reina, che èe quello per che voi tanto vi lamentate Queen what is that for that you much : complain per tale maniera? for such manner ‘Queen, what is it that you are complaining so much about in such a manner?’ (Tavola ritonda, ch. , p. , l. ) If a free RC is involved in P+che RCs, we expect to find the equivalent pattern in lexically headed P+che RCs: in particular, we expect to often find antecedents of the type cos(a/e) ‘thing/s’ and quello ‘that’. Looking at the tokens we extracted from the parsed corpus, we notice that the great majority of the P+che RCs (N=) have quello/cos(a/e) as a [–animate] antecedent. In conclusion, we propose that the change from Old to Modern Italian does not reside in the categorial nature of che, but rather in the licensing conditions of the null nominal THING paired to che. This holds in both lexically headed RCs and free RCs. The observation that the distribution of the null nominal THING changed from Old to Modern Italian can be further supported by the distribution of its lexical counterpart cosa, which was different and broader in the old stage of the language than in Modern Italian. The lexical counterpart of THING, namely cosa, had indeed different properties in Old Italian, since it was found in NPI contexts () and a whole series of cases in which it would be ungrammatical in Modern Italian ():

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

On the relative cycle



() Ragione è quella che contiene la causa, la quale se ne fosse reason is that that contains the cause the which if cl were tolta non rimarrebbe alcuna cosa in contraversia. removed not remain any thing in controversy ‘Reason is the part that contains the cause, which, if it were not present, nothing would remain in the controversy.’ (Brunetto Latini, Rettorica ) () E que’ risponde che ciò era loro agevole cosa d’ imparal-lo. and this replied that this was :. easy thing to know-it ‘And he replied that this would have been easy for them to know.’ (Giamboni, Fiore di rettorica -) Since cosa was a lexical classifier occurring with NPIs and quantifiers like nulla, neun- ‘no’ and tutte ‘all’, we can state that the distinction between Old and Modern Italian che actually depends on the properties of its silent companion, i.e. THING, which had properties in Old Italian which it has lost today.

. Conclusion In this work, we have analysed a specific case of what has standardly been assumed to be a cyclic change from a pronoun into a complementizer head, for reasons of economy. The crucial empirical point to assume that che was a pronoun in Old Italian and has become a complementizer in Modern Italian is that in Modern Italian che is not compatible with prepositions, whereas it was in Old Italian. Therefore, this is the point we have concentrated on in this chapter. We have first shown that the distribution of relativizers is more complex than expected: a) P+che is restricted only to functional prepositions, while this is not true of the other relativizers cui and quale but it is true of free relative clauses; b) in these cases che is only inanimate, while when it has a direct case, animacy is not relevant; c) there has also been a change in the distribution of cui, from being used with [+animate] antecedents (see table .) to being insensitive to the animacy of the head noun in Modern Italian. This change should be tied to the change concerning che. We have argued that the intricate pattern of the development of Italian relativizers can be better explained by assuming that che has never changed its categorial status as a determiner-like element. The distinction between Old and Modern Italian is better captured by assuming that what has changed is the silent companion of che. In Old Italian it was possible to have a null inanimate THING paired to the relativizer che as in free relatives occurring in addition to the lexical N (Type  derivation). This explains why che was inanimate only when occurring with prepositions, and why it could be paired only to functional prepositions (as free relatives did). It also explains why structures like P+che are actually still possible both in interrogatives and in free relatives resuming a whole CP, since what is blocked is not the pairing of che to THING, but the pairing of che to a

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Cecilia Poletto and Emanuela Sanfelici

lexical noun. The change has affected not only che, but also cui, which was able to spell out a [+animate] feature (and presumably be paired to a null animate nominal element like PERSON), which ability it has completely lost in Modern Italian. We think that our analysis is superior to the traditional one, since it captures the data more accurately and accounts for all the changes in the system of relativizers as related to a single abstract property, i.e. the type of null elements which can occur in lexically headed relative clauses, dispensing with unsatisfactory economy explanations. If we are on the right track for the case of relativizers, it might be worth to reconsider other typical cases of cyclic change and see whether they can be analysed in a different perspective as well.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

11 French negation, the Superset Principle, and Feature Conservation KAREN DE CLERCQ

. Introduction French negation has presented researchers with a puzzle for centuries. The puzzle consists of several parts, one of which ties in with the main theme of this volume. This piece of the puzzle is how to model the change from le bon usage French (henceforth BUF) (BUF; Grevisse and Goosse [] , Rooryck ) to colloquial French (henceforth CF) or from formal written French to informal spoken French (De Swart : –). Crucial for BUF is the fact that the expression of sentential negation goes hand in hand with the presence of ne, either in association with pas, as in (), or with a negative indefinite, as in ()–(); in contrast, CF is characterized by the loss of ne and the independent occurrence of pas (as in ()) and negative indefinites (as in ()–()). () Je n’ ai pas I  have  ‘I’m not hungry.’

faim. hunger

() Je n’ ai rien fait. I  have nothing done ‘I haven’t done anything’

() J’ ai pas faim. I have  hunger ‘I’m not hungry.’

() J’ ai vu personne. I have seen no one ‘I haven’t seen anyone.’

() Je n’ ai vu personne. I  have seen no one ‘I haven’t seen anyone.’

() J’ ai rien fait. I have nothing done ‘I haven’t done anything.’

BUF and CF are used as different registers in present-day French, but the data in ()–() reflect two stages of the Jespersen Cycle in the development of negation in French (Jespersen , Dahl ). The Jespersen Cycle is often represented by means of the following three stages in () (De Swart : ). Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Karen De Clercq . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

() a. Stage I: Preverbal expression of sentential negation. b. Stage II: Discontinuous expression of sentential negation. c. Stage III: Postverbal expression of sentential negation. BUF and CF coincide—at least to some extent—with Stage II and Stage III respectively and a theoretical account of these two varieties in present-day French is hence also relevant to capture diachronic change.1 Usually the idea behind this cycle is that the preverbal negator in Stage I gets strengthened at some point by an emphatic element, which leads to the development of bipartite negation in Stage II. In a next step what used to be an empathic element is bleached owing to overuse (Detges and Waltereit , Kiparsky and Condoravdi ) and becomes a regular negator, leading to the loss of the preverbal element in Stage III (Willis et al. a: –). For more details on the Jespersen Cycle, see Meillet (), Horn (), Rowlett (: ), Zeijlstra (: ), van der Auwera (), Breitbarth and Haegeman (), Larrivée and Ingham (), Mosegaard Hansen (), Willis et al. (a), and many others. However, even though the data in ()–() look similar to what we referred to as the Jespersen Cycle, they cannot be equated with it, since neither personne nor rien are verbal negators. They are argumental negative quantifiers or n-words (Laka ) and undergo a related—though slightly different—cycle of change to that of the verbal negators. The cycle of change indefinites are part of is referred to as the quantifier cycle (Ladusaw , Willis ). Whereas it has been claimed in unidirectional theories that indefinites in general change from positive indefinite elements to increasingly more negative elements (Willis ), it has also been argued that change can go in either direction. More in particular, Labelle and Espinal () argue on the basis of a careful study of Old and Middle French that the diachronic evolution of French negative indefinites can go either way. It seems fairly straightforward what a negative quantifier is, i.e. a quantifier that is on its own capable of giving rise to sentential negation. However, it is less clear what an n-word is. Giannakidou (: ) defines n-words informally as in (). An expression α is an n-word iff a. α can be used in structures containing sentential negation or another expression yielding a reading equivalent to one logical negation; and

()

b. α can provide a negative fragment answer. Whereas the data in ()–() mainly require an account for why ne, which is not inherently negative in that it cannot give rise to sentential negation on its own, is nevertheless necessary for the expression of sentential negation in BUF, the data in ()–() force us in addition to also consider issues such as negative concord and double negation. Rien and personne do not only give rise to negative concord 1

See De Swart (: –) for an OT analysis of negation in written and spoken French.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



(NC), but they can also give rise to double negation readings (DN) in particular circumstances, ().2 () a. Personne (n’) a rien no one  has nothing ‘No one said anything.’ (NC) ‘Nobody said nothing.’ (DN)

dit. said

b. Personne (n’) est le fils de Nobody  is the son of ‘No one is the son of anybody.’ (NC) ‘No one is the son of no one.’ (DN)

personne. nobody. (De Swart : –)

Whereas personne and rien most often get into concord relations with each other, pas does not take part in concord relations with n-words and hence when they co-occur this always gives rise to double negation, () (Haspelmath , De Swart , Zeijlstra ). () J’ ai pas rien dit. I have  nothing said ≠ ‘I didn’t say anything.’ (NC) = ‘I didn’t say nothing.’(DN)

(De Swart : )

This chapter will on the one hand discuss and renew ideas put forward in De Clercq (, ) to model the change from Stage II to Stage III or from negation in BUF () to negation in CF (). On the other hand it will probe into the fine-grained structure of negative indefinite words like rien and personne to explain the behaviour of the negative doubling data (den Besten ) in ()–() and the concord and double negation data in (). Whereas it seems obvious based on ()–() that rien and personne have become negative quantifiers in CF in a way similar to how pas has become a full sentential negator in CF, the fact that they can still get into negative concord relations with each other needs to be further explained. In addition, the question will be asked what type of words rien and personne are in BUF, cf. ()–(), being in need of ne on the one hand and being able to get into concord and double negation relations on the other hand. The main idea promoted in this chapter is that the change from BUF to CF shows that the principle underlying language change is Feature Conservation. This principle says that whereas the lexical elements involved in the expression of sentential negation may change and vary across different stages of the language, the features needed for the expression of sentential negation remain stable throughout time. The chapter elaborates on this for negation, but the assumption is that this is valid for all 2 Den Besten () refers to the difference between the data in ()–() and the data in () as the difference between negative doubling and negative spread.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

language change: features are never lost, it is only the packaging that changes. This principle of Feature Conservation ties in directly with the assumption that there is a universal functional sequence (Pollock , Cinque , Rizzi , Haegeman , , Starke , , and many others), henceforth referred to as , which determines, among other things, the internal structure of individual linguistic expressions. Technically, the change observed between the two stages in French n-words is visible in the lexicon: it leads to bigger lexically stored items for the same linguistic expression and hence to less movement when it comes to lexical insertion. The change from BUF to CF thus involves the most efficient operationalization of the Superset Principle, a principle governing lexical insertion, which we will discuss in section ... The change from BUF to CF could thus be viewed as triggered by economy principles (cf. van Gelderen b). However, as is well known from the literature on the Jespersen Cycle, the need to emphasize a marked value like negation will ultimately overrule the economy principle and again lead to redundancy and diversification in the packaging of negative linguistic expressions (Larrivée ). The chapter is structured as follows. In section . I discuss some key concepts related to nanosyntax, I introduce De Clercq (, b) on the nanosyntax of negation in general, and then her account of French negation in particular (De Clercq ). In section .. I propose a decomposition for the negative indefinites rien and personne in the two stages, BUF and CF, that we are interested in. We move on to explain how this decomposition can help us to account for negative concord (in a way similar to what has been proposed by Poletto , ) and then in the last section we take a stab at explaining double negation readings with pas as a consequence of an interplay between focus and the Superset Principle. We end the chapter with a speculation on how strict and non-strict negative concord could be derived in the present system.

. Prerequisites for the analysis .. Nanosyntax The analysis presented in this chapter is couched in the nanosyntactic framework (Starke , , Caha , Baunaz et al. ). Nanosyntax has a postsyntactic lexicon, which contains lexical trees, which are themselves created by syntax. The build-up of these lexical items is determined by the universal hierarchy of features, which we shall refer to as the functional sequence (or ). The order of the  can be observed thanks to a careful study of syncretism patterns, morphological containment, and those syntactic structures or lexical items that are prone to multiple readings (cf. Baunaz and Lander a). Spell-out in nanosyntax is cyclic and phrasal. After each Merge step, the lexicon is checked at the level of the phrase. Whenever the lexicon has a matching lexical item, the lexical item can be inserted. The Superset Principle, (), governs lexical insertion.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle ()



Superset Principle (Starke : ) A lexically stored tree matches a syntactic node iff the lexically stored tree contains the syntactic node.

Whenever there is competition between lexical items, the Elsewhere Principle (Kiparsky ) takes care of that competition, i.e. the closest match gets inserted. If no match can be found in the lexicon for a specific feature merged in syntax, movement is allowed in order to spell out the newly merged feature. I adopt Starke’s () spell-out algorithm, which is presented here in an adaption from Caha et al. (). ()

Spell-out Algorithm Merge F and a. Spell out FP b. If (a) fails, attempt movement of (i) the spec of the complement of F, and retry (a) (ii) the complement of F, and retry (a) c. If (b) fails, spawn a new derivation and retry (a).

As we go along, in section .. the algorithm will be updated slightly, so as to incorporate the interaction between spell-out driven movement and syntactic movement. Another aspect of nanosyntax, which will turn out to be relevant for the more technical issues discussed in this chapter, is the distinction between PRE and POST elements. Starke () argues that PRE elements, i.e. prefixes preceding constituents and roots of lexical items, are structures with a binary grouping at the bottom, (). Suffixes, on the other hand, are structures with a unary bottom, as a consequence of spell-outdriven movement, as in ().

XP ⇒ PRE

()

X

XP

()

YP Y

YP Z

Y

XP ⇒ POST Z

X

In the remainder of the chapter the distinction between PRE and POST elements will become clear. Elements with a binary bottom will be merged either at the bottom of a derivation or as PRE-elements in newly spawned derivations, i.e. in complex specifiers, in the way that most syntactic theories merge subjects or adverbs in the specifier of heads (Cinque ). For more details on how spell-out works and the nanosyntactic model itself, I refer the reader to Starke (), Caha (), Baunaz et al. (), Caha et al. ().

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

.. Nanosyntax of negation De Clercq (, ), De Clercq (b,a), argues for a decomposition of negative markers on the basis of syncretism patterns across different types of negative markers in twenty-two different languages.3 I explain her proposal in what follows.4 Many languages have several negative markers for the expression of negation. English, for instance, has the sentential negative marker n’t, the marker not, and prefixal markers like non-, un-, iN-, dis-. The traditional labels of these markers are negators expressing predicate denial (n’t), predicate negators (not), and predicate term negators (non-, un-, iN-, dis-).5 The traditional terminology reflects one of the crucial properties of these negative markers, i.e. the fact that they take scope in different positions in the clause. In line with work by Zanuttini (), Poletto (, ), Cormack and Smith (), and many others, De Clercq (, ), De Clercq (b), and De Clercq and Vanden Wyngaerd () argue that there are several positions for negation in the clause. De Clercq (, ), De Clercq (b) distinguishes four different categories of negative markers based on their functions, semantics, scope, and differences in stackability. The four different groups are in table . and the labels given to these groups are related to functional projections in the adjectival, nominal, and verbal domain. Q is a feature related to the adjectival functional projection line and De Clercq takes it to stand for gradability or scalar quantity, Class is a feature related to the nominal projection line and stands for Classifier (cf. Cheng and Sybesma , Borer b), Foc is a feature related to the expression of Focus at the edge of vP and in the left periphery (Belletti , , Etxepare and Uribe-Etxebarria ); T is a feature that is related to the extended projection line of the verbal domain (Pollock ). The labels given to these different types of scopal negative markers are labels associated with well-known scope positions in the functional sequence, cf. (). TABLE . Classification of negative markers (from De Clercq a) Tneg -markers Predicate denial scope over stack semantic function

tensed predicate on Foc, Class, Q contradiction denying

Focneg -marker

Classneg -markers

Qneg -markers

Predicate term negation untensed predicate on Class, Q contradiction contrasting/modifying

predicate term on Q contradiction classifying

predicate term — contrariety characterizing

3 De Clercq (, ), De Clercq (b) discusses only nine languages, but recently De Clercq (a) increased the sample to twenty-two languages. 4 The ideas in Poletto (, ) for a decomposed NegP are comparable to De Clercq’s proposal, with the difference that De Clercq’s work includes constituent and affixal negation and focuses on a cross-linguistic varied sample of languages, whilst Poletto’s work is focused on sentence negation and on Italian dialects. 5 See Horn () for discussion of these negators, also in the philosophy of Aristotle and Frege.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



A closer look at twenty-two different languages (De Clercq , De Clercq b,a) in terms of the properties in table . shows that the morphological realizations of these four groups of negative markers can be ordered on the basis of syncretisms, table ., without giving rise to ABA patterns, i.e. a Tneg -marker can only be syncretic with a Qneg -marker if the intermediate types, Classneg - and Focneg -markers, are also syncretic. It is this fact, i.e. the absence of ABA patterns and the fact that morphology follows the natural scope of negation, i.e. from wide to narrow or narrow to wide, which leads to the proposal to decompose negative markers. Thanks to a decomposition it is possible to capture the fact that these negative markers are featurally related, but also different. TABLE . Syncretisms (table from De Clercq a)

Greek English French (formal) Korean French (informal) Swedish Turkish Japanese Khwe Chinese MS Arabic Persian Mayalayam Moroccan Arabic Hungarian Hebrew Dutch Russian Czech Malagasy Hixkaryana Tümpisa Shoshone

Tneg

Focneg

Classneg

Qneg

dhen n’t ne . . . pas (-ci) an(i) (ha-) (-ci) mos (ha-) pas inte degil nai vé bù laa na alla ma (ši) nem lo niet ne netsy -hike(e)

oxi not pas an(i) mos pas inte degil nai ŋya bù laa na alla muši nem lo niet ne ne tsy -hike(e)

mi non non pimolnon ickegayri/olmayan hiófēi ghayrqheyramuši nem lo nietne netsy -hike(e)

auniNpulmoliNo-siz hu(/bu)/mu ó fēi ghayrqheyramuši -tElEn biltionnenetsy -hike(e)

What table . shows is that some languages, like Greek, do not show any syncretisms, whereas other languages, like Czech, are syncretic throughout. These two extremes provide convincing morphological support for the fact that negative markers contain features that relate them and features that make them different. If one wants to do justice to the syncretisms in table ., then one cannot but

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

decompose a negative marker. The underlying featural unity, exemplified by Czech, (), resides (minimally) in the presence of the feature [Neg]. () a. Ja ne- jsem št’astný. I  am happy. ‘I am not happy.’ b. Ja jsem ne- št’astný. I am  happy. ‘I am unhappy.’ c. Je ne- americký. is  American ‘He is un-American.’ ‘He is non-American.’ Each of these four markers has a negative feature, whilst also being featurally different. Consequently, negative markers are always a packaging of Neg with at least one other syntactico-semantic feature, i.e. Neg, and Q, the feature that contributes gradability, yields QNeg -markers like un-. Depending on the scope of the negative markers, more features can be cumulatively added, i.e. Foc or T, so as to capture the *ABA-pattern and the different types of negative markers. The lexical entries for the different negative markers in English are in (). They have binary bottoms, capturing that they are PRE elements, i.e., they precede the predicate they modify. () a. [TP T [FocP Foc [ClassP Class [QP Q Neg]]]] ⇒ not b. [ClassP Class [QP Q Neg ]] ⇒ non c. [QP Q Neg] ⇒ unThe lexical item for not, (a), is big and capable of spelling out not as a sentential and a constituent negator in virtue of the Superset Principle (Caha ). (a) can be inserted both in the syntactic structure in () and in (). It can actually also be inserted in much smaller structures, but in that case, (b) and (c) may be more suitable candidates, thanks to the Elsewhere Principle, which says that in case there is competition between two lexical items the item with least superfluous structure wins the competition (Kiparsky ).

TP

()

T

FocP

()

FocP Foc

Foc ClassP

Class

Class QP

Q

ClassP QP Q

Neg

Neg

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



These complex negative markers take scope in dedicated positions for negation in the clausal spine. The scope positions for TNeg -markers, FocNeg -markers, ClassNeg -markers, and QNeg-markers are depicted in (). Those are the positions that we also used to label the different types of scopal negative markers, i.e. TP, low FocP, ClassP, and QP. As is clear by now, it is thanks to the fact that the same features are active at the nanolevel and at the clausal level that matching can take place and that what happens at the level of the word or morpheme can become relevant at the level of the clause. ()

(NegP) (Neg)

TP T

(NegP)

(Neg)

(FocP) vP

(Foc) v

(NegP)

(Neg)

ClassP

Class

(NegP) (Neg)

QP Q

√

The structure in () shows Neg (and Foc) between brackets, because I adopt the well-known idea that [Neg] and [Foc] are marked values and therefore optional features in the .6 It is due to the optional nature of these features that they do not need to be present in syntax (cf. Starke ). If Neg is merged, the lexicon will be checked. However, [Neg] will not immediately be spelled out, not even after different movement types, cf. (), have been tried, because none of the lexical items for negation has a unary bottom in English.7 In order to satisfy the feature merged in the clausal spine, Starke () proposes that a new derivation will be spawned, which will lead to the projection of the required feature, but then as a complex specifier, a 6

See Horn (), De Swart (), Starke (), and many others for the idea that negation is a marked value and see Reinhart () for the idea that focus is a marked value. 7 But see De Clercq and Vanden Wyngaerd () for an account of -less.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

so-called PRE-element (cf. section ..). One can compare the spawning of such a new derivation to how subjects are merged in SpecvP in most generative theories or how adverbial phrases are merged in the specifier of empty heads in a Cinquean system (Cinque ). The only difference here is that there is no empty head in the main spine, because the idea is that it is the complex new derivation which projects into the main spine and provides the required feature, in line with ideas in Starke (). Starke () argues convincingly that the empty feature that is usually assumed to be present in the main spine with which the complex specifiers enter into SpecHead agreement is redudant, since the Specifier itself provides the feature required by the . Given that a newly spawned derivation is so costly, merge within that derivation will continue until the requirements of the  for sentential negative markers are satisfied, resulting at the level of TP in the structure of a negative marker with the required features up to T. The tree in () provides the structure for a sentence like John is not happy at the point where sentential negation enters the derivation:

NegP

()

TP

not ⇐ TP T

FocP Foc

NP ClassP

John

T’

QP

Class Q

vP

T

QP ⇒ happy

v Neg

Q

√

.. The nanosyntax of French bipartite negation The two pieces used in the expression of sentential negation in BUF can be considered the spell-out of two pieces of the  for negation.8 As such, we can capture the longstanding problem that pas is in need of ne, whilst it feels as if pas is the only real negation (Tesnière ).9 The lexical items for BUF are in () and those for CF in (). 8 As mentioned before, Poletto (, ) also decomposes negation. It was a natural consequence of her account to deal with bipartite negation. The decomposition and the technical implementation is quite different though and I refer the reader to Poletto’s work for a closer comparison. 9 The account sketched in this section is an updated version of De Clercq (, ).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle () a. b. c. d.

< /iN-/, [QP Q Neg] > < /non/ [ClassP Class [QP Q Neg]]> < /pas/ [FocP Foc [ClassP Class [QP Q Neg]]]> < /ne/ [MoodP Mood T] >

() a. b. c. d.

< /iN-/, [QP Q Neg] > < /non/ [ClassP Class [QP Q Neg]]> < /pas/ [TP T [FocP Foc [ClassP Class [QP Q Neg]]]]> < /ne/ [MoodP Mood T ] >



If syntax merges Neg at the level of TP in BUF, there will be no matching lexical item in the lexicon of French that can be inserted. In accordance with the spell-out algorithm different types of movement will be tried, but this will lead to constituents with unary bottoms, and since all negative lexical items have binary bottoms spell-out will be unsuccessful. As a last resort option a new derivation will be spawned, which continues for as long as possible, providing the feature needed by the functional sequence at that point, conforming to the discussion in sections .. and ... Spell-out in this new derivation will be successful until the last merge step. Given the lexical items in (), there will be no way to spell out the syntactic structure once the TP inside the negative marker is merged in the complex specifier in ().

NegP

()

? ⇐ TP T

TP1

pas ⇐ FocP

v

ClassP

Foc

vP

T1

QP

Class Q

QP ⇒ heureuse Q

√

Neg

Consequently, in line with the spell-out algorithm, all movement steps will be tried within the new derivation, i.e. the complex specifier, to spell out T°. Ultimately, because there is only a lexical item with a binary bottom that consists of T, (d), a new derivation will be spawned inside the complex specifier, yielding the structure in () and a spell-out for the complex negative marker ne . . . pas.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq NegP

()

TP ne ⇐ MoodP Mood

TP pas ⇐ FocP

T

Foc

T

vP v

ClassP QP

Class Q

QP ⇒ heureuse Q

√

Neg

The internal structure of ne captures the fact that its structure is related to the  of the low focus negator pas. However, whilst ne is defective and cannot express sentential negation on its own, pas is really negative but needs ne to give rise to sentential negation. The analysis thus captures why pas can give rise to constituent negation and why the combination of ne and pas never gives rise to double negation. Whereas pas is capable of giving rise to real semantic negation, the spine of ne is featurally defective, missing crucial features at the bottom. As such, they need each other to give rise to sentential negation. The account of French provided here shows how both the internal structure and the external syntax of negative markers need to be considered for bipartite negation to be understood.10 In CF the lexical tree for pas has grown, (), making ne unnecessary for the spellout of sentential negation, because if the structure in () is derived, the lexical item in (c) will provide a spell-out, i.e. pas, without any further movements or complications. Ne can from then on either disappear or grow at the top so as to be used as an ‘expletive negative’ marker in Mood-related expressions such as the subjunctive. This intuition is captured in the presence of [Mood] in the lexical item for ne. As such, this account of French bipartite negation endorses push-chain approaches (Breitbarth ) to the Jespersen Cycle, which typically argue for the fact that it is due to changes in the new negator, i.e. pas, that the old negator slowly becomes obsolete or is reused as another marker. What the analysis also shows is that

10 The present proposal also captures why pas, as a focus-negator, in Middle French and Classical French, which in many ways resembles BUF, given the number of references Grevisse uses to Classical French, could be used as a real negative in the context of presuppositional questions (Martineau and Vinet , Bar-Asher Siegal and De Clercq ).

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



sentential negation—whether in BUF or CF—always involves the same features, i.e. there is Feature Conservation with respect to a particular expression. However, the packaging of the lexical items that take care of that linguistic expression changes over time. With respect to the transition from BUF to CF it seems that the trigger for change in this particular stage is economy, resulting in the growth of a lexical item, less spell-out-driven movement and hence the concomitant loss or reuse of other lexical items. In conclusion, we can say that it is thanks to the Superset Principle that one lexical item can be the spell-out of what used to be two different lexical items.

. The structure and distribution of n-words What this chapter wants to contribute to the nanosyntactic work on French negation is an account of the n-words personne and rien in French.11 In section .., we will first decompose negative indefinites in BUF and CF, capturing the diachronic change within polarity-sensitive items and the syncretisms between NPIs and n-words/ negative quantifiers. Next, I will show how the present account fares well in capturing concord and double negation facts in both varieties by drawing largely on the structure of the  and the power of the Superset Principle. Again, as illustrated for the negative marker ne . . . pas in section .., the changing internal structure of n-words from BUF to CF shows that Feature Conservation is at work in linguistic change across time: the same features are involved in the expression of sentential negation across time, but they may get packaged differently, triggered by the economy principle or the need to emphasize. .. The decomposition of polarity indefinites Haspelmath () argues that negative indefinites are often morphologically complex and that they often either contain in their internal make-up a word representing an ontological category or a wh-component. The internal make-up of negative indefinites in French consists of no overt wh-element, but only of a word referring to an ontological category, i.e.  or . Therefore, I propose that the lexical trees of the negative indefinites personne and rien in both varieties of French consist of a root which will be spelled out as an ontological category, , or  (Baunaz and Lander b). This root is turned into a nominal by a categorial feature n (Borer a). At this point the feature [Num] (or #) can be added to the nominal structure and Case Features on top of NumP, which I abbreviate here as K*P (see Caha , , Starke  for a nanosyntactic account of Case.) At this point we

11 With minor modifications I believe the account can be transferred to other n-words. However, it is beyond the scope of this chapter to dive deeper into other n-words.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

are not yet dealing with quantificational structure. The presence of a Q-feature adds scalar quantity (Bresnan , Corver , De Clercq )12 to the nominal, turning it into an scalar quantifier. For English or French we must assume that the lexical item for body/personne, which refers to an ontological category , consists of a Q-feature. Depending on whether Q is absent or present, body/personne has either a regular nominal interpretation or an indefinite intepretation. The Superset Principle will allow the lexical item to be inserted in both quantificational and non-quantificational contexts.

QP ⇒ -body/personne

()

K*P

Q

NumP

K* Num

nP n

√ PERSONNE

However, the structure up until now for rien/personne or -body in English does not give us a negative polarity item yet, which is what we need to capture some of the NPIlike properties of French rien/personne, as illustrated in (). () Il le sait mieux que He it. knows better than ‘He knows it better than anyone.’

personne. anyone (Rooryck : )

Kadmon and Landman () argued that NPIs are endowed with a property which instructs them to consider a wider domain than is usually considered. To make this more concrete, the use of any in a sentence like She did not invite any boy forces us to not only consider the contextually relevant boys (for instance, those in her class), but also boys she knows from other domains (for instance, a youth club, swimming classes, etc). Given that both personne and rien can be used as NPIs, it is instructive to have a look at the internal structure of NPIs in Malayalam. What we see in Malayalam is that NPIs contain a conjunctive particle and a wh-component, (),

12

See also Haegeman and Lohndal () for a decomposition of West Flemish n-words into [uNeg] and [iQ]-features.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



conform to Haspelmath’s observation about indefinites in the world’s languages. Positive indefinites, on the other hand, consist of a wh-component and a disjunctive particle, ().13 () a. *aar-um wannu who- came ‘Anybody came.’ b. aar-um wann-illa who- came- ‘Nobody came.’ c. aar-um war-aam who- come-may ‘Anybody may come.’ () a. aar-oo wannu who- came ‘Somebody came.’ b. aar-oo wann-illa who- come- ‘Somebody did not come.’ c. aar-oo war-aam who- come-may ‘Somebody may come.’

(Jayaseelan )

Jayaseelan () argues that the wh-element is a variable which needs to be bound. Disjunctive operators can bind such a variable; this is what we assume happens with the positive indefinites in (), allowing these lexical items to occur in veridical contexts. However, in (), the wh-variable is not bound by a disjunctive operator. There is a conjunctive operator present though in (a), but this one cannot bind the variable introduced by the wh-feature. The variable hence remains to be bound by a disjunctive operator, presumably higher up in the structure. Elements with a wh-component and/or -um, whose variable has not yet been licensed, can occur only in nonveridical contexts, as in (b–c). We assume that the reason for this is that nonveridical operators (Giannakidou ) contain a disjunctive operator, for instance in the form of a feature [Or], which licenses the wh-variable present in NPIs. The conjunctive morphology in Malayalam, i.e. the presence of -um, can be

13 It is beyond the confines of this chapter to discuss Malayalam in detail. I refer the reader to Jayaseelan () for more discussion of the data.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

looked at as a morphological means to add (=AND) alternatives to the indefinite and thus widen the domain.14 Given that both personne and rien can be used as NPIs, and in the absence of counterevidence, we assume that the morphology in Malayalam is also instructive for the underlying structure of French NPIs, illustrated in ()–().

&P ⇒ personne (NPI)

()

&

WhP wh

&

QP Q K*

&P ⇒ rien (NPI)

()

WhP wh

K*P

QP Q K*

NumP Num n

K*P NumP Num

nP √ PERSON

n

nP √ THING

However, the data we started out with in ()–(), repeated here as ()–(), cannot be accounted for if this were the full structure of rien and personne in BUF, because as such sentential negation could never arise. () Je n’ ai vu personne. I  have seen anybody ‘I haven’t seen anyone.’ () Je n’ ai rien I  have anything ‘I haven’t done anything’

fait. done

We learned in section .. that pas in BUF is a low-scope constituent negator, which consists of all negative layers up until [Foc] and which needs the defective ne to give rise to sentential negation. Consequently, the minimal assumption is that the structure adopted for pas must also be present somehow in the structure of rien and

14 The work done here by the empirically motivated conjunctive feature [&] and the wh-feature is taken care of in Chierchia () and Labelle and Espinal’s () work by a σ.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



personne in BUF. Therefore, I propose that these items also consist of a [Foc] feature in addition to a Neg-feature, capturing not only the parallel with pas, but also the fact that it is due to the emphatic use of the NPI in relation with a negator that the NPI becomes a new negator, leading to the loss of the old negator (Detges and Waltereit , Kiparsky and Condoravdi , Van der Auwera ). The structure for BUF rien and personne can hence be as big as in ()–(), allowing it to give rise to sentential negation together with ne.15 ()

FocP ⇒ personne (BUF)

()

FocP ⇒ rien (BUF) Foc Negp

Foc NegP

Neg &P

Neg &P

& WhP

& WhP

wh QP

wh QP Q K*P K* NumP Num nP n √ person

Q K*P K* NumP Num nP n √ thing

When it comes down to understanding how these elements end up in their scope positions for negation, we need to explain some issues more precisely. Whereas pas in BUF was merged in a complex specifier at the level of NegP above TP, this is not the case for n-words. N-words are merged as arguments within the VP-domain. In line with this, at the moment [Neg] is merged above low FocP in the clausal spine, the derivation that has been built up until then is carefully scrutinized to check whether the feature required by the  has been merged before as part of a larger constituent. I propose that this probing takes place before any spell-out mechanisms are tried, cf. the spell-out algorithm in (). If there is a relevant constituent present, that constituent will be attracted to the specifier of NegP. If there is no such constituent

15

Under Labelle and Espinal ()’s account the nature of rien and personne in this stage would have to be captured by means of a [uNeg]-feature. However, it is unclear how their account would explain the necessary presence of ne in spite of its waning negativity.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

available, the usual spell-out-driven movement steps will apply. This is schematically illustrated in ().

NegP

()

n-word

NegP Neg

vP

Subject

vP v

VP tn-word

The algorithm, as provided in (), hence needs to be updated with a step that involves Move, (), and should thus be considered a Move and Spellout algorithm. ()

Merge F and a. Probe the derivation and attract a constituent to SpecFP that can provide a spell-out for F. If there is no such constituent available, proceed to b b. Spell-out FP c. If (b) fails in the main space, attempt movement of (i) the spec of the complement of F, and retry (b) (ii) the complement of F, and retry (b) d. If (c) fails, spawn a new derivation and retry (b).

Finally, as illustrated in ()–(), and repeated here as ()–(), rien and personne can appear in CF without ne; this suggests that they can occur independently, giving rise to sentential negation on their own. Following the proposal for pas in section .., this means that the structure incorporates a [T]-feature as well as in CF, ()–(), allowing the quantifier to take scope in NegP above TP, in a similar way as proposed for pas. The structures of the quantifier stage in ()–() also indicate how these same structures can also spell out the NPIs and the n-words. () J’ ai vu personne. I have seen nobody ‘I haven’t seen anyone.’ () J’ ai rien fait. I have nothing done ‘I haven’t done anything.’

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle TP ⇒ personne, quantifier

()

T

TP ⇒ rien, quantifier

()

FocP ⇒ n-word

T

Foc NegP



FocP ⇒ n-word Foc NegP

Neg &P ⇒ NPI

Neg &P ⇒ NPI

& WhP

& WhP

wh QP

wh QP

Q KP

Q KP

K NumP

K NumP

Num QP Q

nP n √ person

Num QP Q

nP n √ thing

At first sight it seems counterintuitive to endow a nominal spine with a T-feature. However, this idea is not so far-fetched as it may seem at first sight. In some languages, like Movima for instance, a language spoken in Bolivia (Haude ), there is overt tense marking on nominals. This phenomenon is extensively discussed in Nordlinger and Sadler ().16 Moreover, if we really care about the parallel between the nominal and the clausal domain, then what is T at the clausal level, is presumably what is usually thought of as D at the nominal level. Déprez (), Déprez (), Déprez and Martineau () argue that the internal structure of Standard French n-words like personne is as in (a), whereas n-words in French Creoles, which cannot give rise to sentential negation on their own, have a structure as in (b). In Déprez’ terms indefinites that give rise to sentential negation on their own are quantificational determiners that moved to D, an intuition that is captured in the present proposal by the postulation of a T-feature in their internal syntax, allowing them to have the widest possible negative scope. In Déprez’ account 16 It is tempting at first to think of this [T]-feature and even the [Foc]-feature along the lines of Pesetsky and Torrego (), who argue that structural case is a [uT] feature. This parallel would make [Foc] compatible with accusative-case licensing and [T] on nominals with nominative-case licensing. However, since case checking and negation checking needs to happen independently, this proposal would lead to unwanted results. However, as will be suggested at the end of this chapter, the close relation between Tense and nominative case and Focus and accusative case could be structurally captured to solve many issues related to negation as well.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

NPI-like n-words come with an empty determiner, which turns the indefinites into elements that are in need of proper licensing. We could again say that the same intuition is captured by the present proposal: n-words in BUF cannot give rise to sentential negation, because they lack a T-feature. Only the presence of a T-feature allows them to have full sentential scope. () a. [DP Personne [NP Ø]] (French CF N-words) b. [DP Ø [NP Personne ]] (Creole N-words) For ease of exposition and to allow matching of the internal structure and the external structure, I use the label T even in the case of the nominal domain. With the present decomposition in mind we will now proceed to capture concord with n-words. .. Negative concord In this section we want to account for concord in BUF and CF. However, before we continue doing so, let me first list some problems with one of the more dominant approaches to concord in the literature at the moment. Under a minimalist approach to negative concord (Zeijlstra , Penka and Zeijlstra ), concord crucially depends on the interaction between a semantic (often non-overt) [iNeg] operator and [uNeg] features on n-words/negative markers. However, with such a minimal system it seems hard to do justice to the diachronic change of n-words, on the one hand, whilst at the same time capturing the possible concord and double-negation readings. All goes well as long as one considers the development of one language separately (as for instance in Labelle and Espinal ): in such a case one can say that the presence of [iNeg] is the quantifier stage and [uNeg] is the n-word stage. However, once one wants to account for concord patterns in the quantifier stage (as for instance in French CF) one is confronted with the fact that the n-words (of which there are at least two) each consist of an [iNeg] feature, but they need to get into concord with each other. Hence, in order to get the concord reading one needs to undo capturing the diachronic development and adopt a negative operator at the sentential level, endowed with [iNeg] that does the job of feature checking all n-words which have a [uNeg] feature (Zeijlstra , , Penka , Biberauer and Roberts a). However, the system with a covert [iNeg] becomes even more problematic once one realizes that double negation is a viable option in most concord, i.e. [uNeg], systems. It is unclear how both diachronic worries and synchronic worries can be captured well within one minimalist feature-checking system. The present system, in our opinion, escapes this flaw, since it not only captures diachronic evolution and syncretisms within the domain of polarity items and negation, but can also account for concord, double negation, and bipartite negation—thanks to the rich interplay of decomposition, the Superset Principle, and the structure of the , as I will explain below.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



As we discussed in section ., no matter whether ne is present or not, n-words can get into negative concord relations, as illustrated in (). We will first discuss our proposal for the concord reading in (a) to then move on to our proposal for (b). () a. Personne n’ a rien nobody  has nothing ‘No one said anything.’ b. Personne a rien nobody has nothing ‘No one said anything.’

dit. said (NC BUF)

dit. said (NC CF)

In sections .. and .., we discussed how complex negative markers like pas project into the main clausal spine by spawning a new derivation. With negative arguments, the situation is slightly different, as we also discussed in section ..: they are attracted to the position that is suitable for them, i.e., their internal structure determines which position for negation they will be attracted to. Under the present account negative concord is a natural consequence of the structure of the lexicon and the Superset Principle, without the need to stipulate several lexical items or lexical ambiguity and without having to give up on the idea that a system should be able to capture both the result of diachronic change and synchronic data. Given that the structure of NPIs in ()–(), repeated here as ()–(), is contained in the structure of n-words, we could very well argue that for a sentence like (a) in BUF rien is merged as an NPI in syntax, whilst personne is merged as an n-word and both can be spelled out by the bigger lexical item (i.e. ()–()), thanks to the Superset Principle. ()

FocP ⇒ personne (n-word) Foc NegP

()

FocP ⇒ rien (n-word) Foc NegP

Neg &P ⇒ personne (NPI) & WhP

Neg &P ⇒ rien (NPI) & WhP

wh QP

wh QP

Q K*P K* NumP Num nP n √ person

Q K*P K* NumP Num nP n √ thing

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

If we want to go through the derivation for (a), then the clausal derivation starts by merging the features for the NPI rien. Conforming to the regular spellout principles of nanosyntax, the structure in () will thus be the first syntactic object that gets spelled out. When the verbal head is merged, the verb and its necessary layers are merged. As explained in section .., PRE-elements, like verbs (which in this case precede the object), have a binary bottom and are inserted either at the bottom of a derivation or in a complex specifier (which is a new derivation with a new bottom).

V*P ⇒rien

()

di-(re)⇐ VP* V*

&P √

WhP

&

QP

wh Q

KP K

NumP Num

nP n

√ thing

Then in a next step, (), the subject is added, also as a complex specifier, following well-known traditions in generative syntax. I would like to propose that all subjects need a [T] feature to allow them to scope high. At the end of this chapter I will speculate on how this T-feature interacts with case. For now I would just like to adopt the assumption that all subjects, hence also n-words, need to be merged with a T-feature in their syntax. For the purpose of the present discussion we will assume that there is a feature [v], which requires the merge of a subject. Once the feature [v] is merged in syntax, the lexicon is checked. This does not lead to spell-out, because there is no lexical item available which consists of this structure. After having tried the necessary movements of the spell-out algorithm (see ()) [v] still cannot be spelled out and again a new derivation is

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



spawned, which leads to the derivation of the n-word personne. After syntax has merged the T-feature, the lexicon will be checked, but there is no lexical item in BUF that can spell out personne with this T-feature. Consequently, the different movement steps of the spell-out algorithm will be tried, ultimately leading to yet another complex specifier within the complex specifier, leading to the derivation of ne . . . personne. This derivation is similar to what we discussed for ne . . . pas in BUF in section ... ()

vP V*P ⇒rien

TP ne ⇐ MoodP

FocP ⇒ personne

Mood T Foc

di(re) ⇐ V*P

NegP Neg

V*

√

&P

&

WhP wh

K

QP Q KP K

QP Q

& WhP wh

&P

KP NumP Num

NumP Num

n

nP √ thing

nP

n √ person At the next step in aspectual and verbal layers, the participle ending and the perfective are added (Harwood ). I assume that [AspP] will attract dire and result in the participle dit. I will abstract away now from any details as to how aspectual layers are added in nanosyntax. [vPPerf ] will result in the spell-out of the auxiliary verb avoir, again in a complex specifier, as illustrated in ().

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

()

vperf P avoir ⇐ vperf P AspP

vperf

√ dit ⇐ AspP vP

AspP

V*P

TP ne ⇐ MoodP Mood

T

FocP ⇒ personne Foc

tdire

&P ⇒ rien &

NegP Neg

WhP wh

&P &

QP

wh

KP

Q

WhP

K

QP

Num

KP

Q K

NumP

NumP Num

nP n

√ thing

nP n

√ person

The position for negation above the optional FocP will now be merged. In line with () when Neg is merged in syntax the derivation is checked for the presence of constituents that could be attracted. In this case, the subconstituent of the subject, i.e. personne, can be attracted, since its structure in syntax consists of a Neg-feature. The derivation in () illustrates this.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



NegP

()

FocP ⇒ personne Foc NegP

vperf P

vperf P⇐ avoir

Neg &P vperf √

AspP

dit ⇐ AspP

& WhP

Asp TP

Wh QP

vP V *P

ne ⇐ MoodP tpersonne tdire &P ⇒ rien

Q KP

Mood T

K NumP Num nP n √ person

& WhP Wh QP Q KP K NumP Num nP n √ thing

From this position a subconstituent of personne can be extracted (i.e. KP) and move to SpecAgrSP (see also Caha () on peeling). Rien can also move to SpecAgrOP, a position above focal NegP to license case features.17 As a final step, at the level of TP ne will be attracted as a complex specifier in a tense- or mood-related position above TP (see Cinque () or Zanuttini () for several tense-related positions), yielding sentential scope to an n-word with constituent scope (cf. also Kayne () for the view that ne is a scope marker), just as we described in section .. for ne . . . pas. The derivation of (a) is in ().18 17 Naturally, peeling raises issues with respect to which part gets interpreted. I will assume for now that the highest part is overt and the ‘peel’ or remnant is covert. It is beyond the scope of this chapter to discuss this issue. 18 The fact that n-words target the same position for negation as negative markers is well accepted in the literature on negation. They do so to satisfy the Neg-criterion in the cartographic literature. See Haegeman () for discussion and empirical support.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

() AgrSP KP ⇒ personne K NumP Num QP

TP

MoodP ⇒ ne

TpresP

Mood T v P ⇒ a perf

Q nP n √ person

AgrOP

vperf √ &P ⇒ rien & Whp

NegP vperf P

FocP

Wh QP Foc NegP Q KP

tavoir AspP

Neg &P

di-t ⇐ AspP

K NumP & WhP Num QP Q nP

Wh QP

Asp √ TP

vP V*P

tne tpersonne tdire trien

Q tkp

n √ thing

In CF the structure of the lexical item for personne will have grown to incorporate a [T]-feature, as illustrated in (). Exactly the same happens as in the BUF variety of French with the sole difference that the n-word in subject position can now immediately target the NegP position above TP, since movement inside the complex specifier was not necessary to spell out the [T]-feature. The lowest n-word is still a regular NPI spelled out by the same lexical item as a negative quantifier would be spelled out by, thanks to the Superset Principle.19 As such, without postulating lexical ambiguity or mechanisms like factorization (Haegeman and Zanuttini ) or empty operators (Zeijlstra , Penka ), concord in French BUF and CF can be derived by means of a decomposition of n-words and the Superset Principle. Moreover, the system captures the diachronic change from BUF to CF as a result of growth of lexical trees in accordance with the

19 It is beyond the limits of this chapter to discuss the distinction between rien and personne with respect to how high both arguments can rise. I refer the reader to Rowlett () for a detailed discussion of the differences. On a par with Rowlett () I assume that rien is similar to tout. I propose tentatively that rien can hence still rise to a dedicated Aspectual position in the IP-field.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



, yielding a more economical lexicon. The features of a specific linguistic expression are constant and conserved over time (cf. the Principle of Feature Conservation). Once a lexical item can spell out the entire  on its own, the cycle can be renewed as a consequence of the need to create more emphasis, leading to the loss of features in the original negator. Most importantly, the system is—thanks to the Superset Principle—well equipped to capture the simultaneous co-occurrence of indefinites with different degrees of negativity. In the next section we will briefly turn to how double negation can be derived.

.. A note on double negation In this section I will briefly suggest how the double negation readings in (), repeated here in (), can be derived. () a. Personne (n’) a rien nobody  has nothing ‘Nobody said nothing.’ (DN)

dit. said

b. Personne (n’) est le fils de Nobody  is the son of ‘Nobody is the son of nobody.’ (DN)

personne. nobody. (De Swart : –)

What I want to propose is that double negation also arises thanks to the Superset Principle on the one hand and interactions with focus on the other, a proposal which has been made before for concord, especially in Hungarian (Surányi , Puskás ). Let us have a look at how this works. If we assume that in the case of (a) the structure of the n-word in subject position involves a T-feature and the structure in object position is an n-word, then both will target SpecNegP above low FocP in BUF. However, if low [Foc], which I assume is a marked feature and therefore only merged if required (Starke ), is merged, it comes first in the clausal spine, before [Neg], and one of the two n-words, let’s assume rien, will be attracted to SpecFocP. The other n-word is attracted to NegP right above FocP. From there the subconstituent ne is attracted to a tense- or mood-related position, just as discussed before in section ... In CF the derivation is the same as in BUF, but here the negative quantifier can immediately target the NegP above TP, giving rise to double negation in a maximally simple way without the activation of the clausal FocP. What remains unexplained at the moment is why personne/rien in subject position never gets the NPI-reading, neither in concord nor in double-negation constructions. It is always the subject that has most internal structure and carries the semantic negation. Therefore, what I would like to propose, without going into any detail, is that nominative case sits higher in the  than (optional) negation. As such, the

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Karen De Clercq

structure we proposed for n-words should be as in (), with different case positions at different points of the  of indefinites.20

FocP ⇒ personne (BUF)

()

Foc

KPNomP NegP

KNom

Neg

&P & KAccP KAcc

Whp Wh Q

QP NumP Num n

nP √ person

If this idea is on the right track, it could help us to get closer to why subjects are more often negative than objects and it could help us explain the difference between strict and non-strict negative-concord languages. I postpone the details of this issue to further research.

. Conclusion In this chapter we have discussed how the featural expression of sentential negation remains constant across different stages of a language, in this case in BUF and CF. We called this the principle of Feature Conservation and we have explained how its very existence is immediately tied to the presence of a universal hierarchy of functional features. We showed how the change from bipartite negation to the loss 20

The interaction between genitive case and negation is well known and possibly ties in with the hypothesis as presented here. However, it is beyond the scope of this chapter to work this out in more detail.

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

French negation and the Superset Principle



of bipartite negation across time can be understood as triggered by economy principles, resulting in the growth of lexical items, less spell-out-driven movement and a maximal operationalization of the Superset Principle. Since bipartite negation also arises in combination with negative indefinites, we set out to decompose negative indefinites into their smallest features. The decomposition, in combination with the Superset Principle, turned out to be a powerful tool to account for both concord and double negation in French, in BUF, and in CF. We argued that negative indefinites in BUF are actually constituent negators, which—like constituent pas—are in need of ne to give rise to sentential negation. We discussed how concord can be looked at as a consequence of the Superset Principle, because the lexical item of a negative indefinite can also spell out the syntactic structure of an NPI, leading to a concord interpretation. Negative indefinites in CF have full sentential scope. Again owing to the Superset Principle, the same item can spell out the smaller structure we know from BUF and also NPIs. We moved on to argue that Focus plays a crucial role in giving rise to double negation, both with the negative marker pas and when n-words give rise to focus. Finally, we speculated that a careful investigation of the position of nominative and accusative cases with respect to negation in the  of indefinites could give us a handle on the fact that in concord constructions and double-negation constructions the n-word with semantic negation is always the subject.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

12 From negative cleft to external negator E L I T Z U R A . BA R -A S H E R S I E G A L A N D K AR E N D E C L ER C Q

. Introduction This chapter discusses the syntax and the semantics of the negator lāw in Jewish Babylonian Aramaic (henceforth JBA) through the lens of the diachronic emergence of this negator. The negator lāw is a sentential external negator, whose syntactic and semantic properties will be discussed alongside a diachronic study concerning its origin. Syntactically, we propose that lāw, like negative DPs/PPs in English (Haegeman ) and Sicilian neca (Cruschina ; Garzonio and Poletto ) is merged in SpecFocP in the extended CP-domain from where it takes wide scope. Semantically, lāw takes propositional scope and expresses the meaning of external negation, equivalent to the independent clause: ‘it is not the case’. Diachronically, lāw, as a single-morpheme external negation, developed from a cleft whose matrix clause negates the content of the embedded clause. Following work by Bar-Asher Siegal (b), we argue that the syntactic reanalysis of lāw is triggered by a phonological process of univerbation between the regular negator lā in clefts with the agreement clitic. This syntactic reanalysis involves a morphological univerbation of lāw (Andersen ). The main claim of this chapter is that the syntactic and the semantic characteristics of this negator can be better understood in the light of its historical origin. Moreover, this is an interesting example of how a similar semantic interpretation can be associated with two different syntactic structures, thus allowing a syntactic reanalysis. This type of development is not part of the Jespersen’s Cycle or Croft’s cycle, but constitutes the development of a non-standard negator next to the standard negator. It will be demonstrated that a similar development can be observed for the Sicilian negator neca as well (cf. Garzonio and Poletto ).1 1

The abbreviation to the sources follows the standard abbreviations which appear in The SBL Handbook of Style (Alexander : –). The interlinear glosses are according to the Leipzig Glossing Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Elitzur A. Bar-Asher Siegal and Karen De Clercq . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



In the light of this, the structure of the chapter is as follows: In section . we discuss the historical development of JBA lāw. In section . we review the properties of lāw in the first and second stages of its development, then we move on to an analysis of lāw in both stages. Section . widens the empirical perspective by discussing the emergence of Sicilian neca and the similarity of its properties to JBA lāw. The final section concludes and discusses remaining issues and recommendations for further research.

. JBA lāw: historical development Before embarking upon the evolution of the negative marker in JBA, a few words should be said concerning the history of Aramaic more broadly. Aramaic is a member of the Semitic language family and belongs to the Northwest Semitic subfamily. The history of Aramaic is commonly divided into five phases (Fitzmyer ): . . . . .

Old Aramaic (– ) Official Aramaic (– ) Middle Aramaic ( – ) Late Aramaic ( –) Neo-Aramaic ( –).

Since Late Aramaic, there has ben an opposition between the eastern and western dialects. Our chapter focuses on a development that took place within the eastern dialects of the late periods, a branch which includes three main dialects: JBA, Syriac, and Mandaic. Each of these dialects was spoken by a different ethnic group, Jews, Christians, and Mandaeans respectively. JBA is used to refer to the preserved material that was composed by the Jews during the Late Aramaic period, from the third century onwards, and is the main focus of the present chapter. All examples in this chapter will be drawn from the Babylonian Talmud, the largest corpus written in this dialect. JBA has two negators for sentential negation (Schlesinger : –; Bar-Asher Siegal : –): the unmarked negator lā, common to all branches of the Semitic languages, and the marked negator lāw, which appears in Aramaic and in Hebrew dialects that were heavily under the influence of Aramaic. As will be demonstrated below, diachronically, the evolution of lāw as reflected in the Eastern Aramaic dialects can be characterized as consisting of two stages: Rules, with the addition of the following abbreviations: —geographical name; —proper name; — Rhetorical question marker; d- in JBA is a subordination marker, i.e., it appears at the beginning of all types of embedded clauses. For the sake of simplicity, in this chapter it is always glossed with ‘’. The choice of manuscript for each citation follows Sokoloff ’s (: –) default manuscripts. We wish to thank Silvio Cruschina for informing and helping us with the data from Sicilian Mussomeli and discussing the semantics of the relevant expression with us.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

Stage I: lāw is a contraction of two morphemes. The morphemes lā and the enclitic hu (originally .. personal pronoun) went through a process of synergism to form lāw, owing to the elision of the intervocalic consonant /h/. Thus lā-hu became lāhu, phonetically equivalent to lāw. The two morphemes together constitute a complete clause, with the meaning ‘it is not the case’, always reversing the truth-value of another clause. This sort of contraction manifests a case of phonological univerbation in which only at the phonological level the two independent morphemes are expressed together. Stage II: lāw is a single morpheme, functioning as another type of negator, which is semantically and syntactically marked, i.e. different from the regular sentential negator. From a diachronic point of view, lāw underwent a process of morphological univerbation, as the two morphemes which constitute a clause were reanalysed as a single morpheme, which operates as a sentential external negator. We use the term univerbation as a descriptive term. A theoretical analysis of each stage will be provided in sections .. and ... Furthermore, it must be emphasized that we use the term ‘Stages’ to refer to diachronic developments. This means that Stage II occurred as a consequence of Stage I. These stages are not exclusive, as a language can have two types of lāw, in specific environments. As we will see, in certain environments lāw in JBA reflects stage I, and in others stage II. Looking at this development from a cross-linguistic perspective in terms of van der Auwera (), the type of change can be described as type (c). () a. X ! NEG b. NEG1 X ! NEG1 NEG2 ! NEG2 c. NEG1 X ! [NEG1-X]NEG2 The first type illustrates the derivation of a negator from a non-negative category (see Bar-Asher Siegal : –). The second type portrays the type associated with Jespersen’s Cycle (inter alia Jespersen ; Horn ; van der Auwera and Neuckermans ; van der Auwera , ; Breitbarth and Haegeman : Breitbarth, Lucas, and Willis ; Willis et al. a; De Clercq ) and with Croft’s Cycle (Croft ; Veselinova ). The negator, NEG1, first co-occurs with a new category X, which then becomes NEG2, in that it needs to co-occur with NEG1 to express sentential negation, to finally replace NEG1 and become the new negator. In the third type, NEG1 and another element (X) combine to form a new negator: NEG2. The evolution of JBA lāw is a subtype of the third kind of diachronic process. Namely, it develops a NEG2 on the basis of NEG1+X, but instead of losing NEG1, it retains the old negator while the new negator, [NEG1-X]NEG2, is marked for certain functions. More concretely, the agreement marker (-hu) that often co-occurs with the standard negative marker (here lā), cliticized to that negative marker, with the result

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



of an emergence of the new negative marker, i.e. lāw. Notably, the newly created negator lāw does not lead to the loss of the negator lā, from which it derives. Moreover, lāw also retains its previous functions, i.e. it can still be used either in a cleft or as an independent sentence. Crucially, such a diachronic process adds a new type of negator to the language, whose functions were previously expressed by the use of lā.

. Properties of lāw As claimed in the introduction, we should identify two stages in the development of lāw in the history of Eastern Aramaic. In the following subsections, we will substantiate this claim. We will begin with the properties of lāw in Syriac, which according to our analysis represents Stage I, because Syriac often represents the earlier stage in the diachronic chain of the Eastern Aramaic Dialects (Bar-Asher Siegal : –). This stage is still manifested in certain environments in JBA too. Demonstration of the reanalysis of lāw in JBA will follow this analysis. .. Stage I In all Late Aramaic dialects the standard negator is lā, the common Semitic negator (Walker ). In Syriac, next to lā, we also encounter the form lāw, which has a restricted distribution, as it appears only in negation in the matrix clause of cleft sentences (Joosten ; Pat-El ).2 We argue that this form is a phonological univerbation of two independent morphemes at this stage: lā+hu, the regular negator (lā) merged with the agreement clitic (–hu, rd person singular, cf. Doron ). Thus, lāw on its own is a complete sentence: () lā=w =. ‘It is not the case.’ lit. ‘[it] is not it.’ Support for this hypothesis comes from (). Only when the verb ‘to be’ is absent from the main clause, as in (a), can lā and -hu merge and be pronounced as lāw. If the verb is present in the main clause (in the past tense, for example (see Goldenberg )), as in (b), the contraction cannot take place and the original negator lā remains. () a. lā=w Ḥ īm ʼītaw=y wa =.  exist=. be... ‘It was not  (lit. it is not the case that it was )’ (Ephrem, Genesis , Pat-El , ex. ) 2 Muraoka and Porten (: ) propose that there is one attestation of lāw already in Egyptian Official Aramaic.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq b. lā=wā men ʼūls̩ānā=hu =be... from coercive=. ‘It was not out of coercion (lit. it is not the case that it was out of coercion)’ (Ephrem, Genesis , Pat-El , ex. )

In JBA, lāw also retained its original use as two morphemes, in cleft sentences (cf. ()), as well as in replies to questions: () ʼmar l-eh ʼit l-āk nikse say.... to-. exist to-. property b-qapputqāyā ʼmar l-eh lā=w in- say.... to-.. neg-.. ‘He said to him, “Do you have property in gn?” He replied, “No.” ’ (Ber. b) There are rare examples such as () with a rd feminine singular pronoun (hi: lā+hi [not+it]> lāhi [= lāy]), as the gender of complete statements is interchangeable between masculine and feminine (Bar-Asher Siegal : –). () māy ʼāmart .. dilmā… lā=y what say.... perhaps =.. ‘What would you say, perhaps…it is not so!’

(Tem. b)

The variation between lāw and lāy clearly indicates that in this function lāw consists of two morphemes, as the agreement clitic can be either masculine or feminine. .. Stage II In most of the appearances of lāw in JBA, this conflation can no longer be only phonological in nature. What follows lāw is not necessarily a sentence. This is illustrated by (a), a common phrase in JBA where the adverb hāke ‘such, so’ follows the negator lāw. At this stage lāw, which usually occurs in clause-initial position, can co-occur with a copular verb that has pronominal agreement (b), with the verb ‘to be’ (c), with a feminine copula (d, cf. ()), or with the regular negator lā, as in (e). () a. hā lāw hāke ..  so ‘[In fact] it is not so.’ b. lāw gazlān-e ninhu  thief- ... ‘They are not thieves.’ c. lāw ’isurā hawya  prohibition be.... ‘It was not a prohibition.’

(among others, Menaḥ . b)

(B. Qam. b)

(Yebam. b)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator d. lāw miltā hi  thing .. ‘It is not something (significant)’ e. lāw lā šǝnā   different.. ‘Isn’t it the case that it doesn’t matter?!’



(Sanh. b)

(Šab b)

Since what comes after lāw is not necessarily a clause (a), it can be concluded that it is not a cleft sentence either. Furthermore, Bar-Asher Siegal (b: –) demonstrates that all sentences in () do not have the characteristics of cleft sentences in JBA. Finally, unlike the situation in Syriac (cf. a–c) lāw always appears in this context, and it is never the case that the negator lā is followed by the verb to be. Thus, it is only one main clause, and the agreement features are either expressed by other copular clitics (b, d) or with the verb to be (c). According to this analysis, it must be concluded that at this point in the history of Aramaic, lāw, in these contexts, does not have an agreement feature. Thus, this stage exhibits a morphological univerbation. lāw, accordingly, was reanalysed as consisting of a single morpheme: a negator. ... Distributional properties Bar-Asher Siegal (b) demonstrates the following syntactic difference between the two negators lā and lāw in JBA (Stage II): the standard negator lā always immediately precedes the main predicate, ()–(), whereas lāw in most cases does not appear next to the verb, and tends to appear either in sentence-initial position, (), or following the overt subject (). ()

()

()

ʼnā lā ʼmari I  say... ‘I didn’t tell you.’

l-āk to-..

lā miʽʽrib šappir  mix..... appropriately ‘It is not mixed up appropriately.’ lāw ʽl-eh qā=sāmk-īnan  upon-.. =rely.-. ‘We do not rely upon it.’

() šmuel lāw šappir qā=mǝšanne   appropriately =reply.... ‘pn was not answering him appropriately.’

(Git .̣ b)

(Šab. a)

(Yebam. a) l-eh to-.. (B. Meṣiʿa a)

Furthermore, Bar-Asher Siegal (b) notes that while lā is the unmarked negator, lāw is marked for the following four functions: I) negative rhetorical questions, ()–(); II) antecedents of conditional counterfactual sentences, (); III) to negate a sentence that had been affirmed earlier, () and IV) to reject contextual presuppositions ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

() lāw ʼmari l-āk  say... to-.. ‘Didn’t I tell you that…?!’ () ʼaṭṭu hāhu gabrā lāw yehūdāʼ-e  .. man  Jewish- ‘Is this one [i.e. am I] not a Jewish man?!’

(Moʿed Qat ̣. b) hu ... (ʿAbod. Zar. b)

() ʼi lāw ʼat bahad-an lā hwa   you with-.  be.... sāleq l-an dinā raise.... to-. judgement ‘Had you not been with us, our judgment would not have been conclusive.’ (Sanh. a) () d-mar sābar k-karmelit dāmy-ā -master think.... like-karmelit similar-. w-mar sābar lāw k-karmelit dāmy-ā and-master think....  like-karmelit similar-. ‘As the one person thought it is like a Carmelite; and the other person thought it is not like a Carmelite.’ (Šab. b) () lāw ʽakbrā gnab ʼellā ḥorā gnab  mouse steal.... but hole steal.... ‘It is not the case that the mouse stole, the hole stole.’

(ʿAr. a)

What all these contexts have in common is that the root proposition (what is negated) is presupposed, that is p on which the negator operates is in the common ground. This is clear in the case of negation of sentences that were affirmed before () and in the contexts of rejection of contextual presupposition (). Furthermore, in negative rhetorical questions (–) the root proposition is assumed to be true (‘isn’t it p?!’ presupposes that p is true) and similarly a negative marker in the antecedent of a counterfactual () also presents the root to be true (‘hadn’t it been the case that p’ presupposes that p is true). We would like to add to this list the following observation: while lāw co-occurs with the standard negator lā in one clause (), it is never attested with another lāw in the same clause, which presumably means it is ungrammatical. () lāw lā šnā   different.. ‘Isn’t it the case that it doesn’t matter?!’

(Šab. b)

Moreover, as noted earlier, like lā, lāw can still be used in clefts, () and it may appear in negative answers ().

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



() lāw d-lā qnu minn-eh  - acquire.possession.... from-.. lā d-qnu minn-eh  -acquire.possession.... from-.. ‘Isn’t it the case that it was not purchased from him? No, it was purchased from him.’ (B. Bat. b) () mar l-eh ʼit l-āk nikse b-qapputqāyā, say.... to-.. exist to-.. property in- ʼmar l-eh lā-w say.... to-.. -.. ‘He said to him, “Do you have property in gn?” He replied, “No.” ’ (Ber. b) Although at first it may seem as if lāw serves as the polarity particle ‘no’ () (Holmberg ), this is most likely not the case. As we saw in (), the agreement can vary in this context, and therefore it should be analysed as a full sentence. ... Semantic properties In what follows we will demonstrate that in the environments in which lāw is available, the only available interpretation of the negation in terms of scope, is the widest, i.e. propositional scope. We will turn to support this claim after elaborating first on external negation. In negating the root proposition stated in (a), it is possible to state that (a) is false. This yields the sentence in (b) with a wide scope reading of negation. This is a case in which the external negation is expressed explicitly, in the sense that it is stated with an independent clause expressing that it is not true that a certain number of questions were answered. For the wide-scope reading to be true, it must be the case that the number of questions that were answered is any number which is not three. It is also possible to negate the proposition in (a) by means of a standard sentential negation, as in (c). One salient reading of (c) states how many questions were not answered. This sentence is true if at most seven questions were answered. For (c), especially when ‘three’ is focused, the negation can also be interpreted as external negation: it will then have the same truth conditions as (b). () a. Mike answered three questions (out of ten). (n=) b. It is not the case that Mike answered three questions (out of ten). (n≠) c. Mike did not answer three questions (out of ten). (salient: n or n≠) Crucially, (b) does not have a reading that requires that no more than seven answers were provided by Mike. In the environments where lāw appears in JBA ()–(), the interpretation of the negation is similar to the one that ‘external negation’ has in (b). Whereas the regular negator lā can trigger both the external and internal negation reading, the

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

typical contexts where lāw appears, illustrated by means of the English examples in (b–c), can only give rise to the external-negation reading. Crucially, as expected from external negation, they are about whether it is true that three questions were answered, and not about how many questions were left unanswered, seven or fewer.3 () a. Mike didn’t answer three questions (out of ten). (salient n or n≠) b. If Mike had not answered three questions (out of ten) he would have failed in the exam. (n>) c. Didn’t he answer three questions (out of ten)?! (n=) Accordingly, we may conclude, that the semantics of lāw is similar in Stage I (negative matrix clause of cleft sentences, as in (b)) and in stage II (sentential negation). In both stages lāw takes the widest possible scope, with respect to the clause. .. Overview of the properties of lāw Table . summarizes the properties that we discussed for lāw. TABLE . The characteristics of JBA lāw lāw derived from cleft indicating ~p, when p is presupposed in rhetorical question in antecedent of conditional counterfactual can co-occur with standard negator clause-initial position appears in environments in which negation is always interpreted with wide scope

✓ □ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓

. Analysis In terms of analysis, we would like to propose that lā in JBA expresses regular sentential negation and is base-generated above IP/TP (cf. Belletti ; Laka ; Zanuttini , ; Cormack and Smith ; Holmberg , ; De Clercq , ). Support for this claim comes from the fact that the regular negative marker always takes surface scope over tense and aspect and never follows tense morphology.

3 Bar-Asher Siegal (a) notes about these environments, that besides the fact that they can only take a wide-scope external negation, also PPIs are rescued in these environments (Ladusaw ), and they are also the environments in which German has the so-called ‘light negation’, i.e., the negator nicht is in an unusual position (Schwarz and Bhatt ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator () ʼanā lā qā=āminā I neg dur=say.ptcp.SG ‘I don’t say’



(Pes. a)

More support for the position of sentential negation comes from the data in ()–(). () illustrates existential sentences in JBA. (a) shows an existential sentence in the present tense with the existential expletive ika. In the present tense the copula is not overt or simply not present. In (b) the existential sentence contains a past tense form of to be and the use of ika is optional (cf. Bar-Asher Siegal : –): () a. ika gabra exist man ‘There is a man.’ b. hwa (ika) gabra was. exist man ‘There was a man.’ In the presence of negation, a suppletive form can be used in the present tense (a). This suppletive form seems the univerbation of the negative marker and the existential expletive. However, in the presence of the past tense this univerbation of the negator and ika is not possible (b). The presence of ika is even ungrammatical (c). The only way to negate the sentence is by means of lā preceding the copular verb in the past tense. This suggests two things: ) that lā indeed precedes tense, since it is clearly the tensed copular verb that intervenes in the univerbation process between la and ika and ) that ika is incompatible with the combination of negation and past tense.4 () a. la-yka gabra -exist man ‘There is no man.’ b. la hwa gabra  be.. man ‘There was no man.’ c. *la-yka -exist d. *la 

hwa be..

hwa be..

ika exist

gabra man

4

It is worth noting that with lāw there is no contraction of the negator and the existential marker (Bar-Asher Siegal : ): (i)

lāw mi ika RASHBAG d-qay   exist  -stand.... ‘Isn’t it the case that there is PN who agrees with me?’

kwat-i like-..

(B. Bay b)

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

As proposed in section ., in Stage I, lā could phonologically contract with the agreement marker on the copula, i.e. –hu, in IP/TP. A prerequisite for this contraction to take place is that there is only clitic agreement without the verb to be in these clefts. () lā=w Ḥ īm ʼītaw=y wa =.  exist=. be... ‘It was not PN (lit. it is not the case that it was PN).’ Now, before we represent how phonological contraction could take place in JBA, we need to take a little detour to examining how clefts can be derived in syntax and to the position for focus in syntax. We will illustrate the syntax of clefts with an example from English. Belletti (, , ) derives it-clefts by phrasal movement of the cleft focus to the specifier of the cleft relative, as illustrated in ().5 () ForceP

Force

... ...

T1P T1P

DP it

T1 is

vP v (be)

FocP DP the cat

FocP Foc

Top Top

FinP

Fin

T2P

that

Mary saw the cat

The idea that the cleft focus moves to a position designated to host focus, i.e. FocP, is in line with the cartographic tradition (Rizzi ). Rizzi (: ) argues on the 5 One of our reviewers raises the question of what the cleft focus is in it-clefts without an overt cleft focus, i.e. those of the type: it is (not) [the case] that he left. For this type of clefts we suggest that TP1 selects for a DP ‘the case’ and that the entire embedded FinP is extracted moving to SpecFocP of the embedded CP. It is beyond any doubt that the ramifications of this proposal are beyond the scope of this chapter. We therefore leave a more detailed analysis of this type of cleft for future research.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



basis of the incompatibility of foci and wh-words that they target the same left peripheral focus position, i.e. FocP, in main clauses. The data from Italian illustrate the complementary distribution between foci and wh-words: () a. *A To

GIANNI John

che what

cosa thing

hai have

detto said

(non a Piero)? (not to Peter)

b. *Che cosa a GIANNI hai detto (non a Piero)? Based on these and other data, Rizzi argues that CP, the layer of the clause that anchors the clause in discourse and takes care of clause typing, needs to be split up in several layers that all contribute to the mediation between discourse and the propositional content in IP. He argues that the following features or levels of structure are present at the left periphery of the clause, (). One of these positions is the Focus phrase (FocP), which is targeted by foci and wh-question words alike. ()

ForceP > TopP* > FocP > TopP* > FinP > TP

Rizzi (: ) says the following about Focus in the left periphery: The preposed element, bearing focal stress, introduces new information, whereas the open sentence expresses contextually given information, knowledge that the speaker presupposes to be shared with the hearer.

Belletti (, , ), on the other hand, argues that there are two different positions for focus: the left peripheral FocP is a position dedicated to contrastive and corrective focus, whilst a Focus position above the verb phrase is dedicated to new information focus.6 Abstracting away from this distinction for now, it is argued for a cleft sentence such as () that the cat is the focus and that Mary saw X is the presupposed material (see Haegeman et al. : ). ()

It was THE CAT that Mary saw.

Following arguments put forward by Haegeman et al. (), we thus adopt Belletti’s biclausal analysis for it-clefts, with be projecting its own TP, i.e. TP1, the first clause, and the focus of the cleft, the cat in (), moving to the specifier of a (contrastive) Focus phrase in the left periphery of the second clause, i.e. TP2 (Belletti , , ; Meinunger , ; Frascarelli and Ramaglia ; cf. Spector Shirtz  in the context of Semitic languages).7 The embedded clause is a reduced clause in the sense that ForceP is missing, cf. (). 6 Subject clefts and new information clefts have thus a slightly different derivation from the one depicted in (), with the cleft focus moving to a low FocP in the matrix CP. For the sake of the discussion here, we will abstract away from the distinction between new information clefts and contrastive/corrective clefts and typically treat clefts as contrastive. 7 Haegeman et al. () compare biclausal and monoclausal analyses (Meinunger , ; Frascarelli and Ramaglia ) for it-clefts and argue in favour of Belletti’s biclausal proposal. They do so because it-clefts are compatible with negative inversion and wh-movement. A monoclausal analysis

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

We would like to propose that the original context in which lāw came into being was a biclausal structure, as in (), illustrated for lāw in (). Owing to the absence of an overt subject in SpecTP1 the two heads, i.e. the Neg° head and the agreement -hu on the verbal empty head, could be conflated phonologically.

ForceP

()

Force

NegP T1P1

Neg

T1P

lā 0 law ⇐

T 0-hu

vP v

FocP Foc

TopP FinP

Top Fin

T 2P ...

In Stage II lāw was reanalysed as an external negator. Given its surface appearance below the coordinator () and conditional complementizer (), we deduce that it became a negative element in its own right in a monoclausal structure. () w-lāw hamrā and- wine ‘Isn’t it a wine?!’

hu ...

(Yoma b)

() ʼi lāw ʼat bahad-an lā   you with-.  hwa sāleq l-an dinā be.... raise.... to-. judgment ‘Had you not been with us, our judgment would not have been conclusive.’ (Sanh. a) would involve the postulation of extra left peripheral positions (beyond FocP) to host negative DPs and PPs and wh-constituents at the loss of capturing the complementary distribution between these constituents in the main clause.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



lāw is not (only) contributing emphasis on the polarity expressed by another negative or positive element in the clause (cf. Breitbarth et al. ), as is the case for Hungarian igen (Lipták ), Latin quidem (Danckaert , ), or Flemish en (Breitbarth and Haegeman ),8 but most importantly, it changes the truth conditions of a sentence and contributes real negation. Moreover, as we saw in (e), in combination with the regular negator, lāw gives rise to double negation, i.e., the negations cancel each other out. Following work by De Clercq (, ) the latter fact indicates that the two negators must be merged in different positions and both involve a negative feature. Given that lāw always precedes lā, which—as we argued—is hosted by a NegP above TP—the natural assumption is that it is merged even higher in the structure. Consequently, what we would like to propose is that the newly created negative marker is base-generated in a left peripheral SpecFocP, the projection which was also targeted by the focus of the cleft in the first stage of the development. As such its scope is always the entire p, and accordingly this proposal captures the fact that lāw is on the one hand a reverser of a truth conditions (p is in its scope), providing contrastive information, and on the other hand typically occurs in contexts which are presuppositional, i.e., the root proposition to which it applies is the presupposition, the complement of FocP. The tree structure in () illustrates the two positions for negation, the external negation in SpecFocP on the one hand and the regular sentential negation above TP on the other. ()

ForceP Force

FocP Foc’

law Foc

FinP Fin

NegP Neg la

TP T

vP ...

Further support for this proposal of lāw in SpecFocP comes from the interaction of preposed negative constituents and wh-constituents in English (Haegeman ). 8

But see Breitbarth and Haegeman () for another analysis.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

Haegeman () argues that preposed negative DPs and PPs with sentential negative scope are in a left peripheral SpecFocP. The idea that preposed negative constituents are focal in nature can be traced back to Rochemont (: ), who says that ‘the affective conditioned inversion construction is indicative of a marked focus assignment, much like the cleft construction’. One of the crucial arguments from Rochemont is that these constituents can function as replies to wh-questions, as illustrated by the question–answer pair in (). () a. What job would John be happy with? b. With no job would John be happy.

(Rochemont : –)

Haegeman (: ) adopts this argument and provides more support for the claim that negative constituents are focalized by showing that—like foci (cf. Rizzi )— preposed negative constituents are incompatible with preposed wh-constituents in main clauses, (), again suggesting that they also are in complementary distribution and hence target the same position. () a. *On no account where should I go? b. *Where on no account should I go? Also in the JBA corpus there are no attestations of lāw and a wh-constituent co-occurring, (), but there are attestations of wh- and lā co-occuring (). () (unattested)

amāy why

lāw 

ʼasqu-h bring..-.

() amāy lā ʼasqu-h l-šmuel why  bring..-. -Samuel ‘Why didn’t they bring Samuel with them?’ () kǝma lā miqqṣar how.long  sick.... ‘How long will he not be sick?’

l-šmuel -Samuel

(Ber. a)

(B. Qama a)

It thus seems that lāw in main clauses targets the same position as wh-constituents and negative DPs/PPs. We take this position to be SpecFocP in the left periphery, a position dedicated to focus. Unlike preposed negative constituents, which are merged clause-internally and move to the left peripheral FocP, we propose that SpecFocP is the base-generated position for lāw in Stage II of its development. More support that lāw needs to be in a position outscoping regular sentence negation and regular non-topical subjects, i.e. high in the left periphery, comes from its interaction with universal quantifiers. Preposed negative constituents in English cannot take low scope with respect to the universal quantifier (a), whereas the regular predicate negator allows both scopal patterns (b).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



() a. Under no circumstances would everyone go to the party. ¬ >∀/ ∀ ≯ ¬ b. Everyone hasn’t arrived yet. ¬ >∀/ ∀ > ¬ Also for JBA, one can demonstrate that lāw patterns with the preposed DP and PP and takes widest scope with respect to universal quantifiers, ()–(), while lā is interpreted as being under the scope of the universal quantification, (). () lāw kulle-h ʼālmā ʼbīde  all-m.sg world do.... d-sāyme msan-e -wear... shoe- ‘It is not the case that everyone is apt to wear shoes.’ (Moʿed Qaṭ. a) (¬ >∀) () lāw kulle-h ʼālmā ḥazu lsahdūtā  all-. world see-... to-tsetemony ‘It is not the case that everyone is eligible (to give) testimony.’ (Sanh. a) (¬ >∀) () R. Ḥ anina hu d-ḥakkim kulle-h ʼālmā  . -wise all-. world ‘It’s  that is wise, everyone (else) is not wise.’

lā 

ḥakkim-e wise- (Nid. b) (∀ > ¬)

Given that overt subjects can precede lāw and that, as argued, lāw is base-generated in the left peripheral SpecFocP, overt subjects must target a position in the left periphery above FocP. As illustrated in (), Rizzi () argued on the basis of Italian, a pro-drop language, that there is a position above FocP for topical constituents. Since JBA is a pro-drop language, we assume—in line with other proposals for subjects in pro-drop languages—that overt subjects in cases like () are topics, merged in a high (topic) position (Frascarelli and Hinterhölzl ).9 () šmuel lāw šappir qā=mšanne   appropriately =reply.... ‘ was not answering him appropriately.’

l-eh to-.. (B. Meṣiʿa a)10

In sum, in this section, we have analysed how the JBA negator lāw arose as the consequence of a clitization of the agreement marking –hu with the standard negator lā. Later, in a second stage, this negator became an independent negator used to express external negative scope. We have argued in this section that its

9 See Bar-Asher Siegal (c) for other observations which are related to the analysis of the subjects in JBA as topics. 10 This line comes after a long discussion which begins with a question by Rabbi El’azar and which was responded to by the Babylonian sage Samuel. This discussion assumes that Samuel’s response was valid, but after a long discussion this assumption is rejected by the sentence in (). Thus, this sentence negates what has been already established as part of the common ground, hence presupposed.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

base-generated position is SpecFocP, a left peripheral position of the clause dedicated to contrastive focus. Support for this idea comes from its incompatibility with whquestion words and its wide-scope interpretation over universal quantifiers. The position used for the focus of clefts has become the position for the new negator. At this point, we would like to emphasize that according to the current analysis, both the syntax (position in the clause, non-occurrence with interrogatives, reading above quantifiers, etc.) and the semantic characteristics (widest-scope reading) of the negator lāw can be understood in the light of its historical origin. The main change from Stage I to Stage II is that from a biclausal cleft structure to a monoclausal structure. The regular TP-negator from Stage I gets reanalysed to become a negator that takes the position that the focus of the cleft had in the biclausal cleft structure. The following should be noted about this analysis: . In both structures, the TP(2)/root clause is presupposed, and the negator takes wide scope with respect to the TP(2). It is because of the fact that the same semantic effect of negation, i.e. widest scope, can be obtained in both configurations, that negative clefts are prone to be reanalysed as external negators. . The reanalysis of lāw from a bimorphemic structure to a single morpheme goes hand in hand with the change from a biclausal structure to a monoclausal structure. If two syntactically different constructions express the same semantics, one (or both) of the syntactic environments can easily be prone to a syntactic reanalysis. In the following section, we would like to demonstrate that a similar type of development can be identified in another language, by pointing to similarities between our synchronic and diachronic analysis of the Aramaic lāw and the negator neca in the dialect of Mussomeli (Sicily). This negator also seems to be derived from a cleft (Cruschina ). As Garzonio and Poletto () analyse its syntax, this negator is also analysed to be generated in a left peripheral FocP. Moreover, we will show that the similarities with JBA lāw can be demonstrated at the semantic level as well.

. Cross-linguistic comparison: Sicilian neca The negator neca in the Sicilian dialect of Mussomeli has a similar type of origin to lāw, since as Cruschina (: ) argues, it is derived from a cleft. The path of this derivation is illustrated in (). () Un not

jè it.is

ca that

!

n-è-ca

!

neca

In this section, we will demonstrate other similarities between these two negators. Like the JBA negator, neca cancels presuppositions or ‘characterizes the negated proposition as a wrong expectation made by the interlocutor’ (Garzonio and Poletto : ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



() Neca t’ à scantari! neca .= have. to fear. ‘You don’t have to be afraid!’ (from Cruschina : , glosses Silvio Cruschina p.c.) () Sta lezioni neca si capisci. this lecture neca = understand. ‘This lecture, one does not understand it.’ => ‘One does not understand this lecture.’ (from Cruschina : , glosses p.c. Cruschina) Like lāw neca can also be used in rhetorical questions (), but was perceived as odd in the protasis of counterfactual conditionals, according to our informant: () neca jè vinu? neca is wine ‘Surely this is not wine, is it?’11 The Sicilian negator also has the typical reading of wide-scope, external negators, as shown by the example in (). For the wide-scope reading to be true, as explained in section ..., the number of questions that were answered must be any number which is not three: () neca arrispunnì a tri dumanni ncapu a deci neca answered.SG to three questions on to ten ‘It is not the case that he answered three questions out of ten.’

(n≠)

Garzonio and Poletto (: ) propose an analysis for neca in a similar way to the way the negator lāw in JBA was analysed in section .. According to their proposal, the complex negation-copula-complementizer sequence is re-analysed as a unique functional projection (we assume it is a Focus projection), with respectively the negative marker and the copula in the specifier and the complementizer in the head; then the whole FP is lexicalized as a single functional word (a well-known development in diachronic morphosyntax).12

11 Like questions in JBA with lāw, the questions with neca are rhetorical, in the sense that when asking them, one does not necessarily expect an answer. However, they are different in their content: while asking ‘lāw p?’ one is presupposing that p is true; in contrast, when asking ‘neca p?’ one is asking to confirm that ‘not p’ is true. According to our consultant, it is used in contexts that ask for confirmation, thus the truth value of p is what is being asked for. Therefore the use of external negation is expected. As for the requirement of a presupposition, note that while standard polar questions do not register any commitment (see Farkas and Bruce ) in negative questions with neca there is some commitment that ‘not p’ is what is expected to be true, but there is a lower commitment, and therefore confirmation is asked. In this sense, this use of neca is different in that the presupposition contains the negation. The use of the external negation, is, however, expected. 12 The diachronic derivation proposed by Garzonio and Poletto (: ) is in (i). For them the negation and the copula start off together in SpecFocP and then fuse to one negator in a later stage. The

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

It has been demonstrated that JBA lāw is incompatible with wh-question words and this fact supports an analysis of the negator as a left peripheral focal negator. The same can be shown for neca. Our informant confirmed that neca is incompatible with regular wh-questions ()–(). Only clefted wh-questions are compatible with neca, ()–(): () *A chi to what () *A cu to whom

neca neca neca neca

arrispunnì? answered. arrispunnì? answered.

() A chi jè ca neca arrispunnì? to what is that neca answered ‘What is it that he (surely) didn’t answer to?’ () A cu jè ca neca arrispunnì? to whom is that neca answered. ‘Whom is it that he (surely) didn’t answer to?’ When neca co-occurs with the standard negator, it can also give rise to double negation, just like lāw in JBA and it can license n-words in object position, (), showing that it really contributes negation. () Neca unn’ arrispunnì a tri dumanni ncapu neca not answer.. to three questions on ‘(Surely) he didn’t not answer three questions out of ten.’

a to

deci ten

() Neca fici nenti neca I-did nothing ‘I didn’t do anything.’ In the light of this we propose the following: ) neca reflects a diachronic change from a contraction of three morphemes to a single morpheme (negator), which also involves—in line with what we proposed for JBA—a structural change from a biclausal to a monoclausal construction.

final position is similar to what we propose for lāw, but Garzonio and Poletto () adopt a monoclausal analysis for clefts with the copula and the negator in the left periphery. However, as discussed in footnote  Haegeman, Meinunger, and Vercauteren () point out problems with monoclausal analyses and argue for a biclausal approach. (i)

a. [Spec Focus [un è] [Focus° ca][TP…]] b. [FocusP neca [TP…]]

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

From negative cleft to external negator



) neca can be analysed as an external negator, a negative marker merged in SpecFocP, in line with our analysis of JBA lāw and the proposal made by Garzonio and Poletto (). Table . summarizes the properties of neca alongside those of lāw.13 TABLE . The characteristics of JBA lāw and Sicilian neca

derived from cleft indicating ~p, when p is presupposed in rhetorical question in antecedent of conditional counterfactual can co-occur with standard negator double negation with standard negator clause-initial position not compatible with wh-questions appears in environments in which negation is always interpreted with wide scope

lāw

neca

✓ □ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓ □ ✓

□ ✓ □ ✓ □ ✓ Not enough data □ ✓ □ ✓ □ ✓ □ ✓ □ ✓

neca and lāw are both derived from clefts and seem to express ‘external negation’ semantically, whereas syntactically, they are still within CP, but high up in the left periphery, more specifically in the specifier of a Focus phrase. This captures the fact that these negators are applied to a proposition that was already part of the common ground, and as such they always have a wide-scope reading.

. Conclusion This chapter discussed the development of the negative-polarity marker lāw from a negative cleft in JBA. It was argued that the trigger for the development was a process of phonological univerbation: a cliticization of the agreement clitic –hu with the standard negator lā, and then their subsequent reanalysis as a single morpheme, i.e. as a negator. The semantics of this new negative marker differs from the regular standard negator in that the negation takes wider scope than the scope of a standard negator: it contributes external negation (in fact lāw in stage II retains the semantics of Stage I). On the basis of its incompatibility with wh-constituents and its widescope interpretation with respect to universal quantifiers, it was argued that the 13 There seems to be one difference between lāw and neca: while the former appears in response to questions (–), the latter does not. However, as noted, this may not be a significant difference, as in fact it is more likely that in JBA this is a remnant of Stage  (as in real cleft-sentences, ), and should be interpreted as an independent sentence with the meaning ‘it is not true’. Neca could not appear in such a context, since it contains the complementizer ca, and therefore it does not fit such contexts.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



Elitzur A. Bar-Asher Siegal and Karen De Clercq

grammaticalized negative marker lāw gets base-generated in a left peripheral position FocP of a monoclausal structure. This analysis does not only capture its incompatibility with wh-consituents, but also captures the wide-scope external negation reading and its status as a negative focal operator, which gives rise to a presupposition, i.e., the non-negated p is already part of the common ground. Moreover, it was proposed that this fact was relevant for the reanalysis, since wide-scope negation and the interaction with the presupposed p were already part of what characterized lāw in the earlier stage. Thus, the same sentences, with similar semantics, could be associated with two syntactic structures. In addition, we broadened the empirical scope and showed how the properties of lāw resemble the properties of the Sicilian Mussomeli neca (Cruschina ; Garzonio and Poletto ), for which Cruschina () argued that it is derived from a cleft and for which Garzonio and Poletto () have argued that it is base-generated in SpecFocP. This research predicts that negative clefts may provide the ideal context for the emergence of a new negator. What we do not know at present is how common this pattern is and whether there are languages where this type of negator becomes the standard negator. More research from a cross-linguistic perspective is needed for that. In addition, this chapter presents the syntactic reanalysis as a consequence of the fact that two different syntactic structures have the same semantics. The reanalysis in this particular case involved a shift from a biclausal structure to a monoclausal structure. It would be interesting to examine whether this is always the direction of change, and whether this principle can be derived from some broader principle of reduction in complexity (see Bar-Asher Siegal () for another example of such a phenomenon with expressions that contain negation).

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

13 Changes in the argument and event structure of psych verbs in the history of Spanish M O N TS E R R A T B A T L L O R I , E L I S A B E T H G I B E R T -S OT E L O , A ND I S A B E L P U J OL

. Introduction .. Scope and goals of the chapter This chapter is devoted to the analysis of Modern Spanish class  psych verbs, and in particular to their origin and diachronic development. As is well known, Belletti and Rizzi’s () class  psych verbs select a dative EXPERIENCER (EXP) and a nominative THEME (see section ..). However, these verbs have not always exhibited the same syntactic characteristics. In order to show this, it is instructive to further classify Spanish psych verbs of class  into three subgroups in accordance with the etymological sources of the relevant lexical items, as well as their distinct diachronic trajectories (see table . for a summary). Subgroup A includes verbs inherited from different stages of Latin that display unaccusative configurations with a dative EXP in Latin and in Early, Old, Classical, and Modern Spanish (i.e. verbs that maintain the same argument structure and do not undergo any significant linguistic change). Subgroup B consists of inherited Latin verbs that are transitive from the earliest Latin attestations until sixteenth-century Spanish. From then on, they undergo semantic changes and adopt the unaccusative pattern with a dative EXP that they still present nowadays. Subgroup C embraces verbs of Early Spanish origin that follow the same evolution towards the unaccusative configurations, in spite of being first attested as pronominal verbs. Even if subgroups B and C show an apparently similar syntactic change to accommodate to the unaccusative pattern, the analysis provided in this chapter focuses on capturing the

Cycles in Language Change. First edition. Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen (eds). This chapter © Montserrat Batllori, Elisabeth Gibert-Sotelo, and Isabel Pujol . First published  by Oxford University Press.

TABLE . Origin and development of Belletti and Rizzi’s class  psych verbs in Spanish Structure

Subgroup A Placer ‘like, please’

Doler ‘hurt’ Pesar ‘regret, cause sorrow’ Subgroup B Gustar ‘like, enjoy’

Apetecer ‘fancy’ Importar ‘matter’ Subgroup C Agradar(se) ‘like, please’

Medieval

Classical

Modern

E T

E T

E T

Dative Nominative

Dative Nominative

Dative Nominative

Dative Nominative

Theta Grid A T

A T

A T

E T

E T

Case

Nominative Accusative

Nominative Accusative

Dative Nominative

Dative Nominative

Theta Grid — — —

U T/C

U T/C

U T/C

———

———

U T/C

E T

E T

E T

———

Nominative Oblique (prep.: de, en, con)

Nominative Oblique (prep.: de, en, por)

Nominative Oblique (prep.: de, en, con)

———

———

Nominative Oblique (prep.: de, en, con)

Dative Nominative

Dative Nominative

Dative Nominative

Theta Grid — — —

U T/C

U T/C

U T/C

———

———

U T/C

E T

E T

E T

———

Nominative Oblique (prep.: de, en, por)

Nominative Oblique (prep.: de, en, por)

Nominative Oblique (prep.: de, en, por)

———

———

Nominative Oblique (prep.: de, en, con)

Dative Nominative

Dative Nominative

Dative Nominative

–

 onwards

Theta Grid E T

E T

Case

Case

Antojarse ‘feel like, fancy’

Latin

Case

Dative Nominative

Nominative Accusative

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Spanish Psych verbs: Belletti and Rizzi () Class 

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



exact nature of the change that has taken place in subgroup B, and on introducing the general lines of the change undergone by subgroup C.1 During the evolution of the verbs in subgroups B and C towards their respective unaccusative configurations, attestations of passive constructions of these verbs are commonplace. Note that periphrastic passives, se passives, and the unaccusative pattern under investigation share some basic syntactic similarities: for example, they are all unaccusative structures, whereby external arguments, if present, are ‘demoted’. Furthermore, it has been claimed that Spanish quite generally displays a strong tendency towards unaccusativity2 (in contrast to French, for instance, which seems to be a much more agentive language; see Wolfsgruber ). This tendency brings about a high frequency of periphrastic passive constructions in Medieval Spanish (see Sepúlveda Barrios ; Ricós , ). Later, the periphrastic passive uses became restricted and decreased in frequency from the fifteenth and sixteenth centuries onwards, being replaced by unaccusative psych verbs and se passives. Taking into account the above considerations, we can say that in order to account for the development of class  psych verbs in the history of Spanish, we must capture the following generalizations: (i) their use as transitive verbs (subgroup B) and pronominal verbs (subgroup C) in Old Spanish; (ii) the frequency of periphrastic passive structures of these verbs attested in Old Spanish (see Ricós ); (iii) the consecutive decrease of periphrastic passives in Classical Spanish; and (iv) the assimilation of these verbs to unaccusative configurations with a dative EXP. In addition, as for their eventive evolution, we put forward a change from [+dynamic] to [+stative] configurations that is related to the aforementioned change from transitive to unaccusative structures. Finally, we suggest that these evolutionary paths can be related to the changes in argument structure affecting psych verbs proposed by van Gelderen (a). It must be added, though, that our study focuses only on Belletti and Rizzi’s () class  psych verbs, which represent just a small part of van Gelderen’s cycle. Accordingly, our analysis gives an account of the fact that the trigger of the change lies in the replacement of the passive structures of these verbs, which involve a PP AGENT (i.e. a by-phrase). We consider verbs of subgroup B to have undergone a change whereby the PP AGENT is reinterpreted as a dative EXP.3 Crucially, the PP

1 The evolution of verbs in subgroup C deserves a much more fine-grained account that would lead us beyond the scope of this investigation. We intend to further address this topic in future research. 2 For instance, De Kock (: ) states that in Spanish ‘el sentido pasivo es fundamental . . . [y] el más frecuente’ (‘the passive value is fundamental and the most frequent one’). 3 Note that, according to grammarians, the AGENT by-phrase and the use of a dative EXP are both strategies to express the demotion of the AGENT: see Batllori () and references therein.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

AGENT of passives and the dative EXP of the unaccusative configurations originate in the same position. We also add to the general assumption that dative EXPs are linked to stativity.

.. Belletti and Rizzi’s () typology of psych verbs In their seminal work, Belletti and Rizzi () establish a three-way typology of psych verbs: verbs that take a nominative EXP, like Spanish temer ‘fear’ (); verbs that select an accusative EXP, such as preocupar ‘worry’ (); and verbs with a dative EXP, like gustar ‘like’ (). () Class —Subject = EXP EXPERIENCER (nominative) verb THEME/CAUSE (accusative) JuanEXP teme la oscuridad Juan fear... the darkness ‘Juan is afraid of darkness.’ () Class —Direct Object = EXP THEME/CAUSE (nominative) verb EXPERIENCER (accusative) La delincuencia preocupa a ElenaEXP the crime worry... to Elena ‘Elena is worried about crime.’ () Class —Dative Object = EXP EXPERIENCER (dative) verb THEME/CAUSE (nominative) A María EXP le EXP gusta el chocolate to María =. like... the chocolate ‘María likes chocolate.’ These authors base their classification of psych verbs on the syntax of the EXP argument (i.e. the subject in class , the direct object in class , and the dative object in class ). Concerning the event structure of the three classes of psych verbs, it has generally been assumed that they are all stative—although, as extensively discussed in the literature, the verbs of class  in Belletti and Rizzi () may appear both in dynamic structures and in stative ones: see Fábregas and Marín (); AcedoMatellán and Mateu (), among others. As we mentioned, this chapter focuses on the diachrony of Spanish verbs of class , the examples of which are given in (). In what follows, we will call these predicates Dative Experiencer Psych Verbs (DEPVs). ()

agradar ‘like, please’, desagradar ‘dislike’, apetecer ‘fancy’, placer ‘please’, doler ‘hurt’, gustar ‘like, enjoy’, importar ‘matter’, pesar ‘regret, cause sorrow’, and antojarse ‘feel like, fancy’.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



.. The diachrony of psych verbs: paths and cycles As stated in section .., we propose that Spanish DEPVs that have become psych verbs after changes in their argument structures (namely, DEPVs of subgroups B and C) follow a particular evolutionary path, irrespective of their etymological origins (see figure .). This claim can be related to other studies on the diachrony of psych verbs, which call upon the notion of ‘linguistic cycle’ to capture their changes in argument structure. In particular, van Gelderen (, a) puts forward that there are three main processes of reanalysis that affect psych verbs: (i) a change from object experiencers (OBJ EXP) to subject experiencers (SUBJ EXP), e.g. English fear; (ii) a development from SUBJ EXP to subject agents (SUBJ AG), e.g. like; and (iii) a shift from SUBJ AG to OBJ EXP, e.g. stun. For example, the argument structure of the English psych verb like changed from OBJ EXP in Old English to SUBJ EXP in Modern English, and subsequently from SUBJ EXP to SUBJ AG in current English (van Gelderen : –). OBJ EXP

SUBJ AG

SUBJ EXP

FIGURE . Changes in argument structure affecting psych verbs (van Gelderen a)

Although it is difficult to find cases where the whole cycle is substantiated for the very same verb in a single language, van Gelderen’s proposal can further be illustrated by comparing the evolution of the same kind of verb in different languages. As we will show in section ..., the argument structure of the Spanish psych verbs of subgroup B provides support for the reanalysis from SUBJ AG to (Dative) OBJ EXP. Our study therefore corroborates part of van Gelderen’s cycle. .. Syntactic structures involved in the evolution of psych verbs As we said, the syntactic configurations attested in the evolution of the different subgroups of class  psych verbs include pronominal verbs, transitive verbs, periphrastic passives, se passives, and unaccusative verbs with a dative EXP.4 Before we move on to the diachronic discussion, let us have a brief look at the properties of these different patterns. 4 We assume that successive stages in this proposed development coexist for a certain amount of time. This is indeed what one should expect, since linguistic change is gradual, on the one hand, and since there is inter-speaker and dialectal variation, on the other hand. In current Spanish, though, the most frequent structure for this type of verb is an unaccusative pattern with a dative EXP.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

In pronominal configurations the verb always displays a se pronoun that does not fulfil any specific syntactic function. The example in () illustrates that, with pronominal verbs, the subject is nominative (i.e. los que ‘those who’), and the object is an oblique argument introduced by the preposition de ‘of ’. ()

Pronominal construction y no se ha de poner entre los que por fin & not = have... of put. between the that by end de otras cosas se apetecen. of other things = desire... ‘And he mustn’t interfere between those who desire other things as an aim.’ (CORDE: . Pedro Simón Abril. Traducción de La ética de Aristóteles)

In contrast, in transitive constructions the verb occurs with a nominative subject (i.e. todos ‘everybody’ in example ) and an accusative object (i.e. the accusative pronoun los ‘them’ in the same example): ()

Transitive construction tres géneros de pescados son los mejores que se three types of fish be... the best that = comen . . . porque todos los apetecen como eat... because all =.. desire... as manjar regalado delicacy exquisite ‘Three types of fish are the best to be eaten . . . because everybody desires them as exquisite delicacies.’ (CORDE: . Bernabé Cobo. Historia del Nuevo Mundo)

Passive configurations involve the promotion of the internal argument and the demotion of the external one. In the case of se passives, the se pronoun appears with a third-person singular or plural verb, depending on the number specification of the nominal element that follows the verb, which is in fact the subject of the construction. The passive nature of this type of configuration implies that the relevant construal is possible only with transitive verbs (see RAE and ASALE : § ). ()

Se passive Demuestra asimismo cómo unas cosas se apetecen show... likewise how some things = desire... ‘Likewise, this shows how some things are desired.’ (CORDE: a. . Pedro Simón Abril. Traducción de La ética de Aristóteles)

In the case of periphrastic passives we see the same type of demotion of the external argument. Differently from the se passive, the periphrastic passive features the auxiliary verb ser ‘to be’, which is followed by the past participle of a lexical verb (like perfect-tense forms in the Latin passive conjugation). This pattern always

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



has a PATIENT subject, and the external argument can optionally be expressed by introducing a PP argument (i.e. a by-phrase). As is well known, the preposition that heads this PP in Old Spanish is usually de ‘of ’ (a). From Classical Spanish onwards the preposition por ‘by’ becomes more frequent, replacing de (b). ()

Periphrastic passive a. esto del mandar es apetecido de todos this of.the order be... desire... by all ‘Being in command is desired by everybody.’ (CORDE: –. Jerónimo de Barrionuevo. Avisos. Tomos I, II, III y IV) b. Las partículas de sus vestiduras fueron apetecidas the particles of his clothes be.. desire... por el pueblo by the people ‘The particles of his clothes were desired by the people.’ (CORDE: c.–. Diego Ignacio de Góngora. Historia del Colegio Mayor de Santo Tomás de Sevilla)

Next, on a par with passives, unaccusative psych verbs with a dative EXP involve the promotion of the internal argument, which becomes the nominative subject, and the demotion of the external argument, which appears with dative case. ()

Unaccusative construction with dative EXP a. convocasen . . . a todas las doncellas . . . de las cuales summon... to all the maidens of the which [el Emperador] escogería la que más le the emperor choose.. the that more =. apeteciese desire... ‘They summoned all the maidens . . . from which the emperor chose the one that he desired the most.’ (CORDE: . Antonio de Eslava. Noches de invierno) b. Si al señor le apetece el pan . . . if to.the sir =. desire... the bread ‘if the lord desires bread . . . ’ (CORDE: . Emilia Pardo Bazán. El cisne de Vilamorta)

.. Structure of the chapter The chapter is organized as follows: following this introduction, section . focuses on the diachronic evolution of class  psych verbs in Spanish, and ... includes subsections devoted to each of the subgroups identified above: DEPVs inherited from Latin (subgroup A) are discussed in section ...; transitive Latin verbs that became DEPVs in Romance (subgroup B), in section ...; and DEPVs of Romance

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

origin (subgroup C), in section .... Furthermore, section .. examines the interaction between the changes in argument and event structure that have affected Spanish DEPVs. In section . we elaborate on the cyclic nature of the processes involved in the development of Spanish psych verbs. Section . is devoted to the formal structural analysis. Section . concludes this chapter.

. Diachronic evolution .. Diachronic classification of DEPVs As explained before, we divide Modern Spanish DEPVs into three different types, depending on their etymological origin: subgroup A comprises verbs inherited from Latin that could be used in unaccusative configurations with a dative object (e.g. Lat. placeo ‘please’ or Lat. doleo ‘feel pain’); subgroup B are descendants of Latin verbs that did not develop into unaccusative structures with dative arguments until Golden Age Spanish (e.g. Lat. gusto ‘savour’ or Lat. appeto ‘strive after something, try to get something’); and subgroup C, which contains verbs of Early Spanish origin (e.g. agradar ‘like’, or antojarse ‘fancy, feel like’). In what follows, we discuss each of these three subgroups in turn. ... Subgroup A: DEPVs inherited from Latin The first subgroup includes Spanish placer ‘please’, doler ‘hurt, feel pain’, and pesar ‘upset, cause sorrow’. Their Latin predecessors could be used in unaccusative configurations with a dative object (see Giusti and Iovino ), and were inherited as such in Spanish. An example is Latin placeo ‘please’ (a), which becomes Spanish placer ‘please’ (b). () a. tuisEXP placuit te habere meas litteras yours. please.. you. have my letters. ‘your parents were pleased about your receiving letters from me’ (Cic, Fam., ,,; apud Gaffiot s.v. placeo) b. Si a ustedEXP leEXP place, yo canto if to you.. =. please... I sing... ‘I will sing if it pleases you.’ (CREA: . Francisco Nieva. La señora Tártara) As to Spanish pesar ‘upset, cause sorrow’, although this verb was transitive in Classical Latin ( penso ‘weigh, consider’, as illustrated in (a)), it accommodated to the syntax of DEPVs from the first Spanish attestations, as in (b), which suggests that it may have taken this pattern in Iberian Vulgar Latin. () a. ex factis, non ex dictis, amicos pensare by facts not by words friends. consider. ‘To value friendship by actions, not by words.’ (Liv. ., apud Lewis and Short s.v. penso)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



b. Mucho pesa alos de TecaEXP & alos de TeruelEXP a.lot upset... to.the of Teca & to.the of Teruel non plaze not please... ‘It upsets those from Teca and it disgusts those from Teruel.’ (th c. Cid: v. ; apud Batllori ) ... Subgroup B: Transitive Latin verbs that became DEPVs in Romance Subgroup B contains the Spanish verbs gustar ‘like’, apetecer ‘fancy’, and importar ‘matter’, which come from the Latin transitive verbs gusto ‘savour’, appeto ‘strive after, desire’, and importo ‘bring, imply’, respectively. () is an example of Latin gusto, with a nominative subject and an accusative object: () qui numquam . . . who. never ‘who never savoured love’

gustauit amorem savour.. love. (Lucr. .; apud Batllori )

The transitive nature of these verbs persisted through Old Spanish.5 The examples in () illustrate that gustar ‘savour’ combined with an accusative object in this period (a), and that it even allowed for periphrastic passives (b)6 and passives with se (c).7

5 Note that for some Modern Spanish speakers apetecer can still be used in transitive constructions in the sense of ‘desire’:

(i)

No apetezco estar en not desire... be. in ‘I don’t want to appear in newspapers.’

las the

páginas pages

de los periódicos of the newspapers (CREA: --. Cambio , núm. )

6 A reviewer wonders whether these passives allow demoted agents to surface. The following examples show that this is in fact the case:

(i)

(ii)

7

y que la virtud sea apetecida de varones graves. & that the virtue be... fancy... by men important ‘And that virtue is desired by important men.’ (CORDE: . Alonso López Pinciano. Filosofía antigua poética) Es tan sabrosa que gustada por los forasteros be... so tasty that savour... by the foreigners no quieren salir de aquel pago not want... go.out of that place ‘It is so tasty that when savoured by the foreigners they do not want to leave that place.’ (CORDE: c.. Sebastián de Covarrubias. Suplemento al Tesoro de la lengua española castellana)

We do not use the term middle when talking about se passives to avoid the confusion with anticausative structures such as se hundió el barco ‘the ship sank’, for which the term middle is reserved in the literature on Spanish.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

() a. para dar mayor claridad para que todos lo to give greater clarity so that all =.. gusten savour... ‘so as to shed more light for everybody to savour it’ (CORDE: . Juan de Arfe y Villafañe. Varia Conmensuración para la Escultura y la Arquitectura) b. E si tales cosas como estas . . . se and if such things as these = enpeorassen ante que fuesen gustadas deteriorate... before that be... savour... ‘And if such things deteriorated before they were savoured . . . ’ (CORDE: . Anónimo. Siete Partidas de Alfonso X. BNM I ) c. Calpari: vino nueuo que se saca dela cuba . . . calpari wine new that = extract... of.the barrel ante que se guste before that = savour... ‘Calpari: young wine that is extracted from the barrel before it is savoured.’ (CORDE: . Alfonso de Palencia. Universal vocabulario en latín y en romance) Finally, in Classical Spanish from the sixteenth century these verbs start being attested in unaccusative configurations with a dative EXP, displaying a semantic change that involves a psychological interpretation instead of a physical one. Some of the earliest-known attestations of the new pattern are given in (). () a. No leEXP gusta la nueva a RicardetoEXP not =. like... the news to Ricardeto ‘Ricardeto does not like the news.’ (CORDE: . Jerónimo de Urrea. Traducción de ‘Orlando furioso’ de Ludovico Ariosto) b. sin duda al que mal haceEXP la luz no without doubt to.the.one that evil do... the light not leEXP apetece =. fancy... ‘Without any doubt the one who does evil doesn’t fancy light.’ (CORDE: . Cristóbal de Virués. Historia del Monserrate) c. Que sea de mi tierra o no, esto that be... from my land or not this. meEXP importa poco =. matter... little ‘I don’t mind whether he comes from the same place as I do.’ (CORDE: –. Juan de Valdés. Diálogo de la lengua)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



The above examples show that these Spanish verbs evolve from transitive structures to unaccusative ones with a dative EXP.8 Batllori () has noted that in Spanish the incidence of the periphrastic passive, the frequency of which was higher in the medieval period, decreased over time. She proposes that this might have given rise to the proliferation of unaccusative psychological constructions with a dative EXP in the history of Spanish, a development that did not take place in English.9 We therefore suggest that passives could be considered the structures that trigger the DEPV configuration. As mentioned, periphrastic passives, se passives, and the DEPV constructions share the characteristic of having unaccusative syntax, with, concomitantly, demotion of the external argument. In our view, the fact that unaccusative configurations with a dative EXP came to be preferred over passive construals may, at least in part, be due to language-external reasons (i.e. exogenous triggers). From the very first prescriptive studies of Spanish as a vernacular, grammarians debate on the appropriateness of using passive structures. Periphrastic passives correspond to the learned pattern (since they are inherited from Latin), whereas se passives are related to vernacular, i.e. more colloquial registers. Several prescriptive grammarians condemn periphrastic passives, on the grounds that Spanish lacks passive morphology (e.g. Nebrija, Correas, among others). At the same time, others advise against the use of se passive constructions because they belong to a low register: see Martínez Gavilán (: –) for a detailed account of these aspects. When verbs of subgroup B undergo a semantic change, the speaker adopts the active DEPV pattern characteristic of the verbs of subgroup A (i.e. placer ‘please’, doler ‘hurt, feel pain’, and pesar ‘upset, cause sorrow’), unconsciously conditioned by the prescriptivism that stigmatizes the passive structures. ... Subgroup C: DEPVs of Romance origin In the third subgroup we find verbs of Romance creation that are unattested in Latin, such as agradar ‘please, like’, desagradar ‘dislike’, antojarse ‘fancy, feel like’. The first attestations of agradar exhibit pronominal (a) or passive configurations with the meaning of ‘please’ (b), and it is also occasionally attested as a transitive verb (c). As for antojarse, in the first attested example it is used as a pronominal verb selecting a PP argument introduced by de (d).

8 Note that the evolution of gustar ‘like’ in Spanish follows a different direction from that of like in English: see Melis () and Melis and Flores () for more information about the history of gustar in Spanish. 9 Roberts (: –) suggests that in English, the changes undergone by periphrastic passives with a dative argument had an influence on the evolution of psych verbs, whereby a parametric change affecting vP triggered the reanalysis from dative to nominative subjects.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

() a. El rey Enrrique de Alemaña dáua=las [las mitras] the king Enrrique of Germany give..=.. the mitres a algún su capellán o algún su clérigo de que se to some his priest or some his clergyman of that = agradaua please... ‘King Enrique of Germany gave them [the mitres] to some priest or one of the clergymen that pleased him.’ (DHLE: s.v. agradar: . GConq Ultramar) b. porque . . . los buenos deudos sean agradados . . . in.order.to the good relatives be... please... establezemos . . . como nos debemos mantener . . . establish... how =. must... keep. los unos con los otros the ones with the others ‘We establish the way we must keep up with one another for good relatives to be pleased.’ (CORDE: . Anónimo. Concordia-Acuerdo, Documentación de la cuadrilla de Campezo) c. muchos e nobles . . . lo agradarán many and noble =.. please... ‘Many noble men will please him.’ (CORDE: c.–. Alfonso de Villasandino. Poesías (Cancionero de Baena)) d. antó=s’ dél la mugier de Furtifar, fancy..= of... the wife of Furtifar e dixo: ‘Iaz comygo’ and said sleep.. with.me ‘Furtifar’s wife took a fancy to him, and told him: “Sleep with me”.’ (DHLE: s.v. antojar: c.. Almerich. Fazienda Ultra Mar) The first attestations of the unaccusative construal with a dative EXP date from the second half of the thirteenth century, although this pattern was not generalized until the fifteenth century. The new pattern comes with a change of meaning, as exemplified in (a), where the verb agradar means ‘like’. () a. el mj fillo caro leal muyto meEXP agrada the my son dear loyal very.much =. like... ‘I like my dear and loyal son very much.’ (CORDE: –. Anónimo. Libro del Tesoro. Girona, Catedral a)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



b. cien cosas se les antojan hundred things = =. feel.like... ‘They feel like having hundreds of things.’ (DHLE: s.v. antojar: c. Mexía (Canc. gen. , e)) The chronology of the examples in () and () evidences that linguistic change is gradual; as a result, the synchronic coexistence of structures belonging to different diachronic stages is to be expected (see also footnote ). We find a co-occurrence of pronominal, transitive, periphrastic passives, se passives, and unaccusative constructions with a dative EXP for a long time, and even in some present-day varieties of Spanish there are still some remnants of the passive structures. However, nowadays the most widespread construction is the unaccusative one. .. Interaction between argument and event structure of DEPVs in Spanish This section explores the event structure of Spanish psych verbs of the three subgroups discussed earlier. Crucially, we show that changes at the level of argument structure are correlated with changes in event structure. To establish whether the different subgroups of DEPVs we have identified behave as dynamic events or as states, two tests will be applied: (i) the possibility of appearing in progressive periphrases, which is possible for dynamic predicates but not stative ones; and (ii) the ability to be modified by adverbials like poco a poco ‘little by little’, which measure the event: such modification is impossible with stative predicates, which lack eventivity (on different tests to diagnose dynamicity and stativity, see Fábregas and Marín a). The verbs of subgroup A are unaccusative verbs occurring with a dative EXP. They behave as stative predicates from Early Spanish onwards, which is supported by the fact that they do not appear in progressive periphrases (a’) (unless they are semantically coerced into this pattern), and also because they are not attested with adverbial modifiers such as poco a poco ‘little by little’ (b’), at any stage of the language. Their meaning is always that of a psych predicate.10 () a. Si te place, puedes if =. please... be.able... acompañar=me go.with.=. ‘If you want, you can come with me.’

10

In (i) placer allows for a progressive periphrasis because in this particular context it is used not as a psych verb but as a dynamic verb meaning ‘cause pleasure’, like Modern Spanish complacer. (i) cuando a él placiendo seamos en esos reinos when to him please. be... in those kingdoms ‘when we were giving pleasure to him in those kingdoms’ (CORDE: -. Fray Prudencio Sandoval. Historia de la vida y hechos del Emperador Carlos V)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol a’ *Si te está placiendo, puedes if =. be... please. be.able... acompañar=me go.with.=. b. Me duelen sus prejuicios =. hurt... his prejudices ‘His prejudices hurt me.’ b’ *Me duelen sus prejuicios poco a poco =. hurt... his prejudices little by little

The verbs of subgroup B changed from expressing dynamic events to expressing stative ones. This change in event structure is linked to the development from a transitive structure to an unaccusative construction with a dative EXP. Crucially, this change in event and argument structure also implies a change in meaning. Before the sixteenth century, gustar, for instance, was used in dynamic transitive structures, which is evidenced by the fact that the relevant predicate can be used in progressive periphrases, and that it can be modified by the adverbial poco a poco ‘little by little’. In these dynamic and transitive uses, the meaning of the verb is not that of the psych predicate ‘like’, but rather of the non-psych one ‘savour’ (cf. ()). () a. habiendo quitado la corteza, está have. remove... the peel be... gustando la sustancia savour. the substance ‘After having removed the peel, he is savouring its substance.’ (CORDE: c.–. San Juan de la Cruz (Juan de Yepes). Subida del Monte Carmelo) b. Desseo que poco a poco lo gustéis Desire... that little by little =.. savour... ‘I wish that you savour it little by little.’ (CORDE: –. Fray Bernardino de Sahagún, Historia general de las cosas de Nueva España) From the moment when these verbs are attested as unaccusatives with a dative EXP (i.e. from the sixteenth century onwards, see section ...), they behave as stative predicates. Accordingly, they are hardly attested in the progressive form, and cannot be modified by measure adverbials that highlight the dynamic nature of the event.11 11 It must be noted, however, that we observe another tendency in the evolution of this subgroup of verbs: from the end of the nineteenth century the unaccusative verb gustar (which canonically takes a dative experiencer), for example, is increasingly often used in the progressive form as shown in (i) below: this development points to a new evolution towards dynamic structures.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



Concerning the verbs in subgroup C, their Early Spanish examples do not exhibit the unaccusative psychological pattern with a dative EXP. In fact, they can be pronominal, as in (a) and (d), they can be used in passive configurations, as in (b), and they can sometimes even be transitive (c). In these cases, as exemplified in () for agradar, they are dynamic and admit progressive periphrases. () ¡Dichosas . . . las [mujeres] que . . . conserváis la vida Fortunate the women that preserve... the life en quietud, sin estar agradando un tirano . . . ! in quietness without be. please. a tyrant ‘Lucky [women], that live a quiet life without trying to please a tyrant!’ (CORDE: –. María de Zayas y Sotomayor. Desengaños amorosos) To sum up, this section shows that the verbs of the three subgroups followed two different evolutionary paths with regard to their event and argument structures. The verbs in subgroups B and C changed from dynamic to stative structures at the same time as they changed from transitive or pronominal configurations to unaccusative psych structures with a dative EXP. In contrast, the verbs in subgroup A, which occurred in unaccusative configurations from the earliest attestations, kept their stative nature throughout their history, and they have appeared with a dative argument from the beginning onwards.

. The cyclic evolution of psych verbs On the basis of the above discussion, we can identify two parallel cycles: one related to changes in argument structure, and another one linked to developments affecting event structure. We would like to suggest that the data examined in this chapter provide evidence in favour of the argument structure diachronic pathway given in (): ()

Pronominal / Transitive > Periphrastic passive / Se passive > Unaccusative with dative experiencer

As seen in section .., according to van Gelderen (a), the diachronic changes in the argument structure of psych verbs are related to a cycle, the main stages of which are: (i) a change from object experiencers (OBJ EXP) to subject experiencers (SUBJ EXP); (ii) one from SUBJ EXP to subject agents (SUBJ AG); and (iii) a final change from SUBJ AG to OBJ EXP.

(i) Su mujer le estaba gustando más his wife =. be... like. more ‘He was liking his wife even more.’ (CORDE: –. Benito Pérez Galdós. Fortunata y Jacinta)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

Our work offers further evidence in favour of van Gelderen’s (a) cycle using data from the history of Spanish: as we have shown, verbs of subgroup B support the reanalysis from SUBJ AG to (dative) OBJ EXP. As for the evolution of their event structure, in the previous section we have argued that all the DEPVs inherited directly from Latin (i.e. the verbs of subgroup A, such as placer ‘please’, doler ‘hurt, to feel pain’, and pesar ‘upset, sorrow’) start as stative unaccusative configurations with a dative EXP and are still stative predicates in Modern Spanish. Verbs of subgroup B ( gustar ‘like’, apetecer ‘fancy’, and importar ‘matter’), in contrast, start as dynamic transitive verbs in Latin ( gusto ‘savour’, appeto ‘strive after’, and importo ‘bring’) and develop into DEPVs in Spanish. Finally, the verbs of the subgroup C, which originate in Romance as pronominal and/or transitive predicates, ended up encoding stative eventualities linked to a dative EXP. Thus, the second cycle, which is related to event structure, involves the evolution from [+dynamic] to [+stative] constructions, as shown in (). ()

[+ dynamic] > [+stative]

To sum up, when verbs of different subgroups enter unaccusative configurations with a dative EXP they can be considered stative. As we have seen, the three subgroups show different chronologies in adopting this kind of structure. Furthermore, some of the verbs under study are transitive and dynamic in origin, and become stative as soon as they become unaccusatives and take a dative EXP.12 A question that we would like to leave open is whether the two cycles we have posited are actually part of one integrated system (cf. ()), and if so, how exactly the connection between (changes affecting) argument and event structure is to be captured. () Pronominal / Transitive > Passive > Unaccusative with dative experiencer

[+dynamic]

[+stative]

. A structural account of the changes in argument and event structure of DEPVs This section provides a preliminary account of the syntactic configurations underlying the different steps of the evolutionary paths we have described above. To this purpose, we adopt Ramchand’s () decomposition of the vP into a hierarchy of three subeventive projections: a stative Init(iation) Phrase that encodes causation and licenses the external argument (i.e. the Initiator); a Proc(ess) Phrase that introduces 12

We leave a more detailed account of the evolution from dynamic to stative predicates for future research: this issue deserves more attention, not only as far as the type of affected predicates is concerned, but also regarding the chronology of the relevant development.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



the event variable and is related to dynamicity; and a Res(ult) Phrase that introduces the result of the process and licenses the Resultee. Differently from Ramchand (), but in line with Berro (), we assume that Proc does not license any argument in its specifier (see also Gibert-Sotelo a). The internal argument must, thus, be introduced as the complement of Proc, or by another projection (e.g. as a subject of Res).

InitP

()

Init’

DP1 initiator

ProcP

Init

ResP

Proc

Res’

DP2 resultee Res

…

The first stage of the two evolutionary cycles sketched in () and () corresponds to the transitive and dynamic constructions of the verbs in subgroup B (and probably the ones in subgroup C as well). This is exemplified in () (= (a)) with the verb gustar, which was a transitive dynamic verb meaning ‘savour’ before becoming a DEPV. () para dar mayor claridad para que todos lo to give greater clarity to that all =.. gusten savour... ‘so as to shed more light for everybody to savour it’ (CORDE: . Juan de Arfe y Villafañe. Varia Conmensuración para la Escultura y la Arquitectura) We put forward that the transitive clause todos lo gusten ‘everybody savours it’ in () can be given an analysis like the one represented in (). According to this analysis,

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

this construction involves the presence of an InitP that licenses the nominative external argument todos ‘everybody’ in its specifier, where it receives its interpretation as the agentive Initiator. In the complement position of Init, a ProcP is merged, which encodes the eventive value of an activity. Finally, the accusative object occupies the complement position of Proc, thus further describing the savouring event.

InitP

()

DP

Init’

todos Init

ProcP

gustProc

lo

The next stage is represented by the tendency for the verbs of subgroups B and C to enter into passive configurations in Old and Classical Spanish. If we continue with the case of gustar (subgroup B), the example in () shows that it was attested in passive constructions, and that in these configurations the meaning of the verb was ‘savour’, as in the transitive ones. () la fruta . . . gustada por los forasteros the fruit savour... by the foreigners ‘the fruit . . . savoured by the foreigners’ (CORDE: c.. Sebastián de Covarrubias. Suplemento al Tesoro de la lengua española castellana) In passive structures the formal (nominative) subject is the internal argument. The external argument, if expressed, is encoded by a PP adjunct (i.e. a by-phrase). Following Fábregas (: –), we assume that passive voice involves a weak InitP that does not license a DP in its specifier (see also Gibert-Sotelo b: ), which forces the external argument to be introduced by a preposition and therefore to take the shape of a PP. This PP, merged at Spec,InitP, is interpreted as the agentive Initiator.13 As a

13 This is in accordance with Collins (), who proposes that the external argument is merged in the same position in both active and passive structures.

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



PP cannot receive nominative case, it is the internal argument DP, merged as the complement of Proc, that ends up being the nominative subject.

InitP

()

PP

P por

los forasteros

Init’

Initweak

ProcP

gustProc

la fruta

Recall that along with the use of these verbs in passive configurations, a change in their semantics can be observed. This is exemplified in () for gustar, which adopts a more abstract value that can be translated as ‘appreciate’ or ‘like’, two meanings that lack the agentivity (and, hence, the dynamicity) of the earlier verb, which meant ‘savour’. () a. dio licencia el Cabildo al Sr. Nolasco para que give.. licence the council to.the Sir Nolasco in.order that lleve a los músicos que gustase bring... to the musicians that like... a la fiesta to the party ‘The Council gave permission to Sir Nolasco to bring the musicians that he liked to the party.’ (CORDE: –. Anónimo. Documentos en la catedral de Sigüenza) b. Estos juegos, muy gustados por la nobleza española, these games very like... by the nobility Spanish formaban el encanto de los campamentos constitute... the charm of the camps y de los soldados and of the soldiers ‘These games, very much liked by the Spanish nobility, were the charm of the camps and soldiers.’ (CORDE: . Diego Barros Arana. Historia general de Chile, II)

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

This change in meaning is parallel to the generalization of the use of this verb in unaccusative structures with a dative EXP, in which case gustar cannot mean ‘savour’, and can only convey the abstract psych value of ‘like’ (cf. ()). This corresponds to the last stage in the evolutionary paths that we have traced in section ., in which the agentive external argument, demoted to an adjunct by-phrase in passive structures, is replaced by a dative EXP (). In addition, the dynamicity of the event is lost, giving rise to a more abstract stative meaning (). According to Haspelmath (, apud van Gelderen ), verbs change from a concrete to an abstract meaning and, accordingly, verbs that have an AGENT external argument can develop more abstract psych values that require an EXP rather than an AGENT. () a. Mucho les gusto a todos la propuesta a.lot =. like.. to all the proposal ‘They liked the proposal very much.’ (CORDE: . Ana Francisca Abarca de Bolea. Vigilia y octavario de San Juan Baptista) b. a cualquier hombre le gusta una mujer bien to any man =. like... a woman well ataviada dress.up... ‘Any man likes a beautifully dressed-up woman.’ (CORDE: c.. José Joaquín Fernández de Lizardi. La Quijotita y su prima) Fábregas and Marín (b) explore the links established between dative EXP and stativity and conclude that, in the case of psychological predicates, the presence of a dative EXP necessarily involves stativity.14 Our data provide evidence in favour of this claim, since the development of the new unaccusative pattern with a dative EXP parallels the development of the stative psych value of these verbs. Along the lines of Landau () and Fábregas and Marín (b), we assume that the dative EXP is generated as the complement of a silent P merged in the specifier of Init. Taking into account Cuervo’s () work on dative arguments, we further propose that DEPVs involve the presence of a high Appl(icative) head that is spelled out by the clitic that doubles the dative EXP, which is always present in DEPVs. This high Appl is merged on top of InitP and attracts the dative-marked DP to its specifier, as represented in (). The stative interpretation of DEPVs is expected, given the absence of Proc (the head that introduces dynamicity) in the complement position of Init (see Ramchand : ). If Proc is missing from a given configuration, the 14 The claim that dative EXP are related to stativity has been put forward by many scholars (see, among others, Arad ; Campos ; Marín and McNally ; Acedo-Matellán and Mateu ).

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi

Changes in psych verbs in Spanish



argument in Spec,InitP cannot be interpreted as a dynamic Initiator (i.e. an AGENT), but rather it has to be interpreted as stative (i.e. an EXP). The DP internal argument which acts as the complement of Init identifies the stimulus of the experience (the THEME or CAUSE), and ends up being the nominative subject, given that the EXP argument, which has been assigned dative case by the applicative head, cannot receive nominative case.

ApplP

()

a todos

Appl’

Appl

InitP

les PP

P

Ø

Init’

Init

la propuesta

gust-

As pointed out in the previous sections, we assume that passive configurations and stative unaccusative structures with a dative EXP share a common core. The expression of the AGENT by a PP (i.e. a by-phrase) in passives and the expression of the EXP by a dative argument in unaccusative constructions are two well-known strategies used to demote an external argument. Crucially, the PP AGENT and the dative EXP are both merged in the same position, namely in the Spec,InitP. Additionally, it has been shown that the stative nature of these verbs (when occurring in the unaccusative pattern with a dative EXP) is the result of a semantic change from concrete to abstract that was brought about by the reinterpretation of the AGENT (demoted to an adjunct PP in passive structures) as a dative EXP.

. Conclusion This chapter offers a detailed study of changes affecting the argument structure and the aspectual behaviour of Spanish psych verbs that take a dative EXP (‘Dative Experiencer Psych Verbs’, DEPVs for short). A three-way classification of Spanish

OUP CORRECTED PROOF – FINAL, 14/8/2019, SPi



Batllori, Gibert-Sotelo, and Pujol

DEPVs was proposed, using the etymological origin of the relevant predicates as our main criterion. We further discussed the changes in argument and event structure of DEPVs. In particular, we proposed that they conform to two parallel evolutionary paths that seem to be interconnected: the first involves the development from transitive (or pronominal) to unaccusative constructions with a dative EXP; the second relates to the evolution from dynamic to stative events. We also hypothesized that the trigger of the change is the general use and following decrease of passive configurations in Classical Spanish. Our analysis shows that passive and unaccusative structures with a dative EXP are structurally similar, in that the PP AGENT and the dative EXP both originate in the same position. At a more general level, this chapter provides empirical support for (a part of) the cyclical process of change affecting psych verbs proposed by van Gelderen (a).

Acknowledgements This research was supported by the research projects FFI--C--P and FFI-C--P (Ministerio de Economía y Competitividad), the research groups  SGR  and  SGR  (Generalitat de Catalunya), and the research grant FPU/. We are grateful to two reviewers for their accurate comments and suggestions, and also to the audiences of two congresses where this research was presented, namely the X Congreso Internacional de Historia de la Lengua Española (Universidad de Zaragoza, ) and the th Diachronic Generative Syntax Conference (Ghent University, ).

Sources [CDH]: RAE-Fundación Rafael Lapesa, Corpus del Nuevo diccionario histórico del Español

[Çid]: Poema de Mio Cid. Facsímil de la edición paleográfica, R. Menéndez Pidal, ed. (), Madrid, Dirección General de Archivos y Bibliotecas. [CORDE]: RAE, Corpus diacrónico del español [CREA]: RAE, Corpus de referencia del español actual [DHLE]: RAE (–): Diccionario Histórico de la Lengua Española [Gaffiot]: Gaffiot, Félix (). Dictionnaire latin-français. Paris: Hachette [Lewis and Short]: Lewis, Charlton T., and Charles Short (), A Latin Dictionary. Oxford: Clarendon Press

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References Aasen, Ivar (). Norsk Grammatik. Christiania: P.T. Mallings Forlagsboghandel. Abbott, Barbara (). ‘Support for a Unique Theory of Definiteness’, in T. Matthews and D. Strolovitch (eds), Proceedings of Semantics and Linguistic Theory . Ithaca, NY: Cornell University, –. Available at (accessed  November ). Abraham, Werner, Elisabeth Stark, and Elisabeth Leiss (). ‘Introduction’, in E. Stark, E. Leiss, and W. Abraham (eds), Nominal Determination. Typology, Context Constraints, and Historical Emergence. Amsterdam: John Benjamins, –. Abrusán, Márta (). Weak Island Semantics. Oxford: Oxford University Press. Acedo-Matellán, Víctor and Jaume Mateu (). ‘Los verbos psicológicos: Raíces especiales en estructuras corrientes’, in R. Marín (ed.), Los predicados psicológicos. Madrid: Visor Libros, –. Adams, Marianne (). ‘From Old French to the Theory of Pro-drop’, Natural Language and Linguistic Theory : –. Alexander, Patrick H. (). The SBL Handbook of Style, for Ancient Near Eastern, Biblical, and Early Christian Studies. Peabody, MA: Hendrickson Publishers. Alexiadou, Artemis and Elena Anagnostopoulou (). ‘Parameterizing AGR: Word Order, Verb-movement and EPP Checking’, Natural Language and Linguistic Theory : –. Alexiadou, Artemis and Gisbert Fanselow (). ‘On the Correlation between Morphology and Syntax: The Case of V-to-I’, in J.-W. Zwart and W. Abraham (eds), Studies in Comparative Germanic Syntax. Amsterdam: John Benjamins, –. Alexopoulou, Theodora, Raffaella Folli, and George Tsoulas (). ‘Bare Number’, in R. Folli, C. Sevdali, and R. Truswell (eds), Syntax and its Limits. Oxford: Oxford University Press, –. Ambrazas, Vitaudas (ed.) (). Lithuanian Grammar. Vilnius: Baltos lankos. Anagnostopoulou, Elena (). ‘Clitic Doubling and Object Agreement’, in S. Fischer and M. Navarro (eds), Clitic Doubling and Other Issues of the Syntax/Semantics Interface in Romance DPs. Proceedings of the VII Nereus International Workshop. Arbeitspapier Nr. , University of Konstanz, –. Andersen, Henning A. (). ‘From Auxiliary to Desinence’, in M. B. Harris and P. Ramat (eds), Historical Development of Auxiliaries. Berlin: De Gruyter, –. Andersen, Merete (). ‘A Spanning Approach to the Acquisition of Double Definiteness in Norwegian’, Iberia : –. Anderson, Stephen R. (). ‘On the Development of Morphology from Syntax’, in J. Fisiak (ed.), Historical Morphology. Berlin: De Gruyter, –. Aoun, Joseph (). The Formal Nature of Anaphoric Relations. Doctoral dissertation, MIT. Arad, Maya (). ‘On “Little v” ’, MIT Working Papers in Linguistics : –. Ariel, Mira (). ‘The Function of Accessibility in a Theory of Grammar’, Journal of Pragmatics : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Ashby, William (). Clitic Inflection in French. An Historical Perspective. Amsterdam: Rodopi. Auger, Julie (). ‘More Evidence for Verbal Agreement Marking in Colloquial French’, in W. J. Ashby, M. Mithun, G. Perissinotto, and E. Raposo (eds), Linguistic Perspectives on the Romance Languages. Amsterdam: John Benjamins, –. Auger, Julie (a). Pronominal Clitics in Quebec Colloquial French. A Morphological Analysis. Doctoral dissertation, University of Pennsylvania. Auger, Julie (b). ‘On the Nature of Subject Clitics in Picard’, in M. L. Mazzola (ed.), Issues and Theory in Romance Linguistics. Washington, DC: Georgetown University Press, –. Auger, Julie (). ‘Pronominal Clitics in Picard Revisited’, in R. Núñez-Cedeño, L. López, and R. Cameron (eds), A Romance Perspective on Language Knowledge and Use. Amsterdam: John Benjamins, –. Bácskai-Atkári, Julia (). ‘Cyclical Change in Hungarian Comparatives’, Diachronica : –. Bácskai-Atkári, Júlia, and Éva Dekány (). ‘From Non-Finite to Finite Subordination’, in K. É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax. Oxford: Oxford University Press, –. Bácskai-Atkári, Julia and Éva Dékány (). ‘Cyclic Changes in Hungarian Relative Clauses’, talk given at the th Diachronic Generative Syntax Conference (DiGS ), Reykjavík, University of Iceland,  May . Badia i Margarit, Antoni (). Gramàtica de la llengua catalana. Barcelona: Proa. Bahtchevanova, Mariana and Elly van Gelderen (). ‘The French Subject Cycle and the Role of Objects’, in E. van Gelderen (ed.), Cyclical Change Continued. Amsterdam: John Benjamins, –. Baker, Mark C. (). The Atoms of Language. New York: Basic Books. Bar-Asher Siegal, Elitzur A. (a). ‘Arguments in Favor of an Ambiguist Approach’, talk given at the Göttingen workshop on negation and polarity, September . Bar-Asher Siegal, Elitzur A. (b). ‘The Case for External Sentential Negation: Evidence from Jewish Babylonian Aramaic’, Linguistics : –. Bar-Asher Siegal, Elitzur A. (c). ‘From a Non-Argument-Dative to an Argument-Dative: The Character and Origin of the qṭīl lī Construction in Syriac and Jewish Babylonian Aramaic’, Folia Orientalia : –. Bar-Asher Siegal, Elitzur A. (). Introduction to the Grammar of Jewish Babylonian Aramaic, nd Revised and Extended Edition. Münster: Ugarit-Verlag. Bar-Asher Siegal, Elitzur A. (). ‘Notes Concerning the Emergence of Negation Words in the Semitic Languages’, Leshonenu : – [in Hebrew]. Bar-Asher Siegal, Elitzur A. (). ‘The History of the Forms ‘ilule and ‘ilmale: Linguistic Diachronic and Textual Transmissions’, Leshonenu. Bar-Asher Siegal, Elitzur A. and Karen De Clercq (). ‘From Negative Cleft to External Negator: Eastern Aramaic lāw and Sicilian (Mussomeli) neca.’ lingbuzz/. submitted to DIGS . Barbosa, Pilar (). ‘pro as a Minimal nP: Toward a Unified Approach to Pro-Drop’, Linguistic Inquiry, (early access)  July: –. Barrie, Michael (). ‘Clitic Placement in European Portuguese’, Toronto Working Papers in Linguistics : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Batllori, Montserrat (). ‘La impersonalización en español medieval: Recursos formales y semánticos’, in C. García Turza, F. González Bachiller, and J. Mangado Martínez (eds), Actas del IV Congreso Internacional de Historia de la Lengua Española. La Rioja, – de abril de . Logroño: Universidad de la Rioja, Servicio de Publicaciones, vol. , –. Batllori, Montserrat (). ‘Diacronía de los verbos psicológicos: Una propuesta de entrada léxica’, in G. Clavería (ed.), Historia del léxico. Perspectivas de investigación. Madrid/ Frankfurt: Iberoamericana/Vervuert, –. Batllori, Montserrat, Narcís Iglésias, and Ana Maria Martins (). ‘Sintaxi dels clítics pronominals en català medieval’, Caplletra : –. Baunaz, Lena and Eric Lander (a). ‘Nanosyntax: The Basics’, in L. Baunaz, K. De Clercq, L. Haegeman, and Eric Lander (eds), Exploring Nanosyntax. Oxford: Oxford University Press, –. Baunaz, Lena and Eric Lander (b). ‘Ontological Categories’, in Pavel Caha, Karen De Clercq, and Guido Vanden Wyngaerd. The Unpublished Manuscript. A Collection of Lingbuzz Papers to Celebrate Michal Starke’s th Birthday. –. https://ling.auf.net/ lingbuzz/ Baunaz, Lena, Karen De Clercq, Liliane Haegeman, and Eric Lander (eds) (). Exploring Nanosyntax. Oxford: Oxford University Press. Bayer, Josef and Ellen Brandner (). ‘On Wh-Head-Movement’, in C. Chang and H. Haynie (eds), Proceedings of the th West Coast Conference on Formal Linguistics. Somerville, MA: Cascadilla, –. Beck, Sigrid and Remus Gergel (). ‘The Diachronic Semantics of English again’, Natural Language Semantics : –. Behaghel, Otto (ed.) (). Deutsche Syntax, vol. . Heidelberg: Winter. Belletti, Adriana (). Generalized Verb Movement. Aspects of Verb Syntax. Turin: Rosenberg & Sellier. Belletti, Adriana (). ‘Aspects of the Low IP Area’, in L. Rizzi (ed.), The Structure of CP and IP. The Cartography of Syntactic Structures . New York: Oxford University Press, –. Belletti, Adriana (). Structures and Strategies. New York: Routledge. Belletti, Adriana (). ‘Focus and the Predicate of Clefts’, talk given at GIST, Cartographic Structures and Beyond, Ghent University,  May . Belletti, Adriana and Luigi Rizzi (). ‘Psych-Verbs and Theta-Theory’, Natural Language and Linguistic Theory : –. Benincà, Paola (). ‘Un’ipotesi sulla sintassi delle lingue romanze medievali’, Quaderni Patavini di Linguistica : –. Benincà, Paola (). ‘A Detailed Map of the Left Periphery of Medieval Romance’, in R. Zanuttini, H. Campos, E. Herburger, and P. Portner (eds), Crosslinguistic Research in Syntax and Semantics. Negation, Tense and Clausal Architecture. Washington DC: Georgetown University Press, –. Benincà, Paola (). ‘Headless Relatives in some Old Italian Varieties’, in Roberta D’Alessandro et al. (eds), Syntactic Variation. The Dialects of Italy. Cambridge: Cambridge University Press. –. Benincà, Paola (). ‘Lexical Complementisers and Headless Relatives’, in Laura Brugè et al. (eds), Functional Heads. New York and Oxford: Oxford University Press, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Benincà, Paola and Guglielmo Cinque (). ‘La frase relativa’, in L. Renzi and G. Salvi (eds), Grammatica dell’italiano antico, vol. . Bologna: Il Mulino, –. Benjamin, Geoffrey (). A New Outline of Temiar Grammar, part I. Ms., Nanyang Technological University, Singapore. Bennis, Hans, Norbert Corver, and Marcel den Dikken (). ‘Predication in Nominal Phrases’, Journal of Comparative Germanic Linguistics (): –. Bernstein, Judy, Francisco Ordóñez, and Francesc Roca (). ‘Grammaticalization in Catalan and Spanish Nominals’, talk given at the GETEGRA International Workshop on Nominals, Recife,  March . Bernstein, Judy, Francisco Ordóñez, and Francesc Roca (). ‘Origins and Change: On the Evolution of Articles with Proper Names’, talk given at the th Diachronic Generative Syntax Conference (DiGS ), Stellenbosch University,  September . Bernstein, Judy, Francisco Ordóñez, and Francesc Roca (). ‘Sardinian Descendants of ipse in Comparative Relief ’, in A. De Angelis and A. Chilà (eds), Capitoli di morfosintassi delle varietà romanze d’Italia: Teoria e dati empirici (Supplementi al Bollettino, ), Palermo: Centro di Studi Filologici e Linguistici Siciliani, –. Berro, Ane (). Breaking Verbs. From Event Structure to Syntactic Categories in Basque. Doctoral Dissertation, University of the Basque Country. Biberauer, Theresa and Ian Roberts (a). ‘Comments on Jäger “Anything is Nothing is Something”: on the Diachrony of Polarity Types of Indefinites’, Natural Language and Linguistic Theory : –. Biberauer, Theresa and Ian Roberts (b). ‘Subjects, Tense and Verb-Movement’, in T. Biberauer, A. Holmberg, I. Roberts, and M. Sheehan (eds), Parametric Variation. Null Subjects in Minimalist Theory. Cambridge: Cambridge University Press, –. Bidese, Ermenegildo (). Die diachronische Syntax des Zimbrischen. Tübingen: Narr. Bidese, Ermenegildo and Alessandra Tomaselli (). ‘Developing Pro-Drop: The Case of Cimbrian’, in J. Casalicchio and F. Cognola (eds), Null Subjects in Generative Grammar. A Synchronic and Diachronic Perspective. Oxford: Oxford University Press, –. Bidese, Ermenegildo, Andrea Padovan, and Alessandra Tomaselli (forthcoming). ‘Rethinking V and Nominative Case Assignment: New Insights from a Germanic Variety in Northern Italy’, in T. Biberauer, S. Wolfe, and R. Woods (eds), Rethinking Verb Second. Oxford: Oxford University Press. Blaxter, Tam (). Speech in Space and Time. Contact, Change and Diffusion in Medieval Norway. Doctoral dissertation, University of Cambridge. Bleam, Tonia (). Leísta Spanish and the Syntax of Clitic Doubling. Doctoral dissertation, University of Delaware. Bobaljik, Jonathan (). ‘Realizing Germanic Inflection: Why Morphology does not Drive Syntax’, Journal of Comparative Germanic Linguistics : –. Bobaljik, Jonathan and Höskuldur Thráinsson (). ‘Two Heads aren’t Always Better than One’, Syntax : –. Boeckx, Cedric (). ‘What Principles and Parameters got Wrong’, in C. Picallo (ed.), Linguistic Variation in the Minimalist Framework. Oxford: Oxford University Press, –. Bonet, Eulàlia (). ‘The Person-Case Constraint: A Morphological Approach’, MIT Working Papers in Linguistics : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Booij, Geert and Jaap van Marle (eds) (). Yearbook of Morphology. Dordrecht: Kluwer. Bopp, Franz (). Über das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem der griechischen, lateinischen, persischen und germanischen Sprachen. Frankfurt am Main: Verlag der Andreäischen Buchhandlung. Borer, Hagit (). Parametric Syntax. Case Studies in Semitic and Romance Languages. Dordrecht: Foris. Borer, Hagit (a). In Name Only. Oxford: Oxford University Press. Borer, Hagit (b). The Normal Course of Events. Oxford: Oxford University Press. Börjars, Kersti, Pauline Harries, and Nigel Vincent (). ‘Growing Syntax: The Development of a DP in North Germanic’, Language : e–e. Borthen, Kaja (). Norwegian Bare Singulars. Doctoral dissertation, Norwegian University of Science and Technology. Bošković, Željko (). ‘On the Locality of Left Branch Extraction and the Structure of NP’, Studia Linguistica : –. Bošković, Željko (). ‘What Will You Have, DP or NP?’, in E. Elfner and M. Walkow (eds), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, –. Bošković, Željko (). ‘More on the no-DP Analysis of Articleless Languages’, Studia Linguistica : –. Bouzouita, Miriam (). ‘Future Constructions in Medieval Spanish: “Mesoclisis” Uncovered’, in R. Kempson, E. Gregoromichelaki, and C. Howes (eds), The Dynamics of Lexical Interfaces, Stanford, CA: CSLI, –. Bouzouita, Miriam (). ‘La accesibilidad referencial de las dislocaciones a la izquierda en español medieval’, Boletín de la Real Academia Española : –. Bowler, Margit (). ‘Conjunction and Disjunction in a Language without “and” ’, in T. Snider, S. D’Antonio, and M. Weigand (eds), Proceedings of Semantics And Linguistic Theory , –. Available at (accessed  November ). Brandtler, Johan and Lars-Olof Delsing (). ‘Framväxten av obestämd artikel i fornsvenska: Definithet, specificitet och referens’, in G. Byrman, A. Gustafsson, and H. Rahm (eds), Svensson och svenskan. Med sinnen känslinga för spark. Festskrift til Jan Svensson  januari . Lund University: Department of Scandinavian Languages, –. Breitbarth, Anne (). ‘A Hybrid Approach to Jespersen’s Cycle in West Germanic’, Journal of Comparative Germanic Linguistics : –. Breitbarth, Anne, and Liliane Haegeman (). ‘Continuity is Change: The Long Tail of Jespersen’s Cycle in Flemish’, in A. Breitbarth, C. Lucas, S. Watts, and D. Willis (eds), Continuity and Change in Grammar. Amsterdam: John Benjamins, –. Breitbarth, Anne and Liliane Haegeman (). ‘The Distribution of Preverbal en in (West) Flemish: Syntactic and Interpretive Properties’, Lingua : –. Breitbarth, Anne and Liliane Haegeman (). ‘En en is níet wat we dachten: A Flemish discourse particle’, in V. Kimmelman, N. Korotkova, and I. Yanovich (eds), MIT Working Papers in Linguistics : –. Breitbarth, Anne, Karen De Clercq, and Liliane Haegeman (). ‘The Syntax of Polarity Emphasis’, Lingua : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Breitbarth, Anne, Christopher Lucas, and David Willis (). ‘Incipient Jespersen’s Cycle: The (Non)-Grammaticalisation of New Negative Markers’, in J. Fleischer and H. Simon (eds), Sprachwandelvergleich – Comparing Diachronies. Berlin: De Gruyter, –. Bresnan, Joan (). ‘Syntax of the Comparative Clause Construction in English’, Linguistic Inquiry : –. Brucart, José M. (). ‘Els determinants’, in J. Solà (ed.), Gramàtica del Català Contemporani. Barcelona: Empúries, –. Bull, Tove, Brit Mæhlum, Agnete Nesse, Helge Sandøy, Michael Schulte, Inge Særheim, and Lars Vikør (). ‘Forord til verket’, in H. Sandøy and A. Nesse (eds), Norsk språkhistorie. Mønster, vol. I. Oslo: Novus forlag, –. Burenhult, Niclas (). A Grammar of Jahai. Lund: Department of Linguistics and Phonetics, Lund University. Busse, Ulrich (). Linguistic Variation in the Shakespeare Corpus. Morpho-syntactic Variability of Second Person Pronouns. Amsterdam: John Benjamins. Bybee, Joan, Revere Perkins, and William Pagliuca (). The Evolution of Grammar. Chicago, IL: University of Chicago Press. Caha, Pavel (). The Nanosyntax of Case. Doctoral dissertation, University of Tromsø. Caha, Pavel (). ‘Explaining the Structure of Case Paradigms by the Mechanisms of Nanosyntax’, Natural Language and Linguistic Theory : –. Caha, Pavel, Karen De Clercq, and Guido Vanden Wyngaerd (). ‘The Fine Structure of the Comparative’, Studia Linguistica: –. doi: https://doi.org/./stul.. Available at (accessed  February ). Campbell, Lyle (). ‘What’s Wrong with Grammaticalization?’, Language Sciences /–: –. Campos, Héctor (). ‘Transitividad e intransitividad’, in I. Bosque and V. Demonte (eds), Gramática descriptiva de la lengua española, vol. , Madrid: Espasa Calpe, –. Cardinaletti, Anna (). ‘On Pronoun Movement: The Italian Dative loro’, Probus : –. Cardinaletti, Anna (). ‘Stylistic Fronting in Italian’, in L.-O. Delsing, C. Falk, G. Josefsson, and H. Á. Sigurðsson (eds), Grammar in Focus. Festschrift for Christer Platzack.  November , Volume . Lund: Wallin & Dalholm, –. Cardinaletti, Anna (). ‘Il pronome personale obliquo’, in L. Renzi and G. Salvi (eds), Grammatica dell’italiano antico. Bologna: Il Mulino, –. Cardinaletti, Anna and Michal Starke (). ‘The Typology of Structural Deficiency: A Case Study of Three Classes of Pronouns’, in H. van Riemsdijk (ed.), Clitics in the Languages of Europe. Berlin: De Gruyter, –. Carlsson, Greg N. (). ‘A Unified Analysis of the English Bare Plural’, Linguistics and Philosophy : –. Caro Reina, Javier (). ‘The Grammaticalization of the Terms of Address en and na as Onymic Markers in Catalan’, in F. Debus, R. Heuser, and D. Nübling (eds), Linguistik der Familiennamen. Hildesheim: Georg Olms Verlag, –. Carstens, Vicki, Norbert Hornstein, and Daniel Seely (). ‘Head Movement in Problems of Projection.’ Unpublished manuscript. Available at . Castillo, Juan Carlos, Drury John E., and Kleanthes K. Grohmann (). ‘Merge over Move and the Extended Projection Principle: MOM and the EPP Revisited’, Iberia : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Cheng, Lisa Lai-Shen and Rint Sybesma (). ‘Bare and Not-So-Bare Nouns and the Structure of NP’, Linguistic Inquiry (): –. Chierchia, Gennaro (). ‘Scalar Implicatures, Polarity Phenomena, and the Syntax/Pragmatics Interface’, in A. Belletti (ed.), Structures and Beyond. The Cartography of Syntactic Structures, vol. . Oxford: Oxford University Press, –. Chierchia, Gennaro (). ‘Broaden your Views: Implicatures of Domain Widening and the “Logicality” of Language’, Linguistic Inquiry (): –. Chierchia, Gennaro (). Logic in Grammar. Polarity, Free Choice and Intervention. Oxford: Oxford University Press. Chierchia, Gennaro, Danny Fox, and Benjamin Spector (). ‘Scalar Implicatures as a Grammatical Phenomenon’, in C. Maienborn, K. von Heusinger, and P. Portner (eds), Semantics: An International Handbook of Natural Language Meaning, vol. . Berlin: De Gruyter, –. Chomsky, Noam (). Aspects of the Theory of Syntax. The Hague: Mouton. Chomsky, Noam (). ‘Remarks on Nominalization’, in R. A. Jacobs and P. Rosenbaum (eds), Readings in English Transformational Grammar. Waltham, MA: Ginn, –. Chomsky, Noam (). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam (). The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, Noam (). ‘Minimalist Inquiries: The Framework’, in R. Martin, D. Michaels, and J. Uriagereka (eds), Step by Step. Essays in Minimalist Syntax in Honor of Howard Lasnik, Cambridge, MA: MIT Press, –. Chomsky, Noam (). ‘Beyond Explanatory Adequacy (MIT Occasional Papers in Linguistics ). Cambridge, MA: MIT Press. Chomsky, Noam (). ‘Beyond Explanatory Adequacy’, in A. Belletti (ed.), Structures and Beyond. Oxford: Oxford University Press, –. Chomsky, Noam (). ‘Three Factors in Language Design’, Linguistic Inquiry : –. Chomsky, Noam (). ‘Approaching UG from Below’, in U. Sauerland and H.-M. Gärtner (eds), Interfaces + Recursion = Language? Chomsky’s Minimalism and the View from SyntaxSemantics. Berlin: De Gruyter, –. Chomsky, Noam (). ‘Problems of Projection’, Lingua : –. Chomsky, Noam (). ‘Problems of Projection: Extensions’, in E. Di Domenico, C. Hamann, and S. Matteini (eds), Structures, Strategies, and Beyond. Amsterdam: John Benjamins, –. Chomsky, Noam (). Puzzles about Phases. Ms, MIT. Cinque, Guglielmo (). ‘La sintassi dei pronomi relativi ‘cui’ e ‘quale’ nell’italiano contemporaneo’, Rivista di Grammatica Generativa : –. Cinque, Guglielmo (). ‘On the Theory of Relative Clauses and Markedness’, The Linguistic Review : –. Cinque, Guglielmo (). ‘La frase relativa’, in L. Renzi (ed.), Grande grammatica italiana di consultazione, vol. . Bologna: Il Mulino, –. Cinque, Guglielmo (). Adverbs and Functional Heads. A Cross-linguistic Perspective. Oxford: Oxford University Press. Cinque, Guglielmo (). ‘More on the Indefinite Character of the Head of Restrictive Relatives’, Rivista di Grammatica Generativa : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Cinque, Guglielmo (). Typological Studies. Word Order and Relative Clauses. New York/ London: Routledge. Cinque, Guglielmo (). ‘Three Phenomena Discriminating between “Raising” and “Matching” Relative Clauses’, Semantics-Syntax Interface : –. Cinque, Guglielmo (). ‘On the Status of Functional Categories (Heads and Phrases)’, Language and Linguistics : –. Cinque, Guglielmo and Luigi Rizzi (eds) (). Mapping Spatial PPs. Oxford: Oxford University Press. Collins, Chris (). ‘A Smuggling Approach to the Passive in English’, Syntax : –. Colomina i Castanyer, Jordi (). ‘Paradigmes flectius de les altres classes nominals’, in J. Solà (ed.), Gramàtica del Català Contemporani. Barcelona: Empúries, –. Company Company, Concepción (). ‘Tiempos de formación romance II. Los futuros y condicionales’, in C. Company Company (ed.), Sintaxis histórica de la lengua española: Primera parte, la frase verbal. Mexico: Universidad Nacional Autónoma de México/Fondo de Cultura Económica, –. Cormack, Annabel and Neil Smith (). ‘Modals and Negation in English’, in S. Barbiers, F. Beukema, and W. van der Wurff (eds), Modality and its Interaction with the Verbal System. Amsterdam: John Benjamins, –. Cornilescu, Alexandra (). ‘On the Syntax of Datives in Unaccusative Configurations’, in E. O. Aboh, J. Schaeffer, and P. Sleeman (eds), Romance Languages and Linguistic Theory . Amsterdam: John Benjamins, –. Corominas, Joan and José A. Pascual (–). Diccionario crítico-etimológico castellano e hispánico. Madrid: Gredos. Coromines, Joan (–). Diccionari etimològic i complementari de la llengua catalana. Barcelona: Curial. Corver, Norbert (). ‘The Internal Syntax and Movement Behavior of the Dutch wat voorConstruction’, Linguistische Berichte : –. Corver, Norbert (). ‘Much-Support as a Last Resort’, Linguistic Inquiry : –. Corver, Norbert (). ‘Freezing Effects’, in M. Everaert and H. van Riemsdijk (eds), The Wiley Blackwell Companion to Syntax, nd Edition. Oxford: Wiley-Blackwell, online version at (accessed  November ). Corver, Norbert and Marjo van Koppen (). ‘Microvariation and Ellipsis in the wat voorConstruction’, handout, Sounds of Silence Workshop, Tilburg, – October. Corver, Norbert and van Koppen, Marjo (). ‘Let’s Focus on Noun Phrase Ellipsis’, GAGL: Groninger Arbeiten zur germanistischen Linguistik : –. Crisma, Paola (). ‘The Emergence of the Definite Article in English: A Contact-Induced Change?’ in P. Sleeman and H. Perridon (eds), The Noun Phrase in Romance and Germanic. Structure, Variation, and Change. Amsterdam: John Benjamins, –. Crisma, Paola (). ‘The “Indefinite” Article from Cardinal Operator to Expletive’, in C. Gianollo, A. Jäger, and D. Penka (eds), Language Change at the Syntax-Semantics Interface. Berlin: De Gruyter, –. Croft, William (). Typology and Universals. Cambridge: Cambridge University Press. Croft, William (). ‘The Evolution of Negation’, Journal of Linguistics : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Cruschina, Silvio (). ‘Aspetti morfologici e sintattici degli avverbi in siciliano’, Quaderni di lavoro dell’ASIt : –. Cuervo, María Cristina (). Datives at Large. Doctoral dissertation, MIT. Dahl, Östen (). ‘Typology of Sentence Negation’, Linguistics : –. Danckaert, Lieven (). ‘Quidem as a Marker of Emphatic Polarity’, Transactions of the Philological Society : –. Danckaert, Lieven (). ‘Spelling out the Obvious: Latin quidem and the Expression of Presuppositional Polarity’, Journal of Historical Pragmatics : –. Danylenko, Andrii (). ‘Russian čto za, Ukrainian ščo za, Polish co za “was für ein”: A Case of Contact-Induced or Parallel Change?’ Diachronica : –. De Cat, Cécile (). ‘French Subject Clitics Are Not Agreement Markers’, Lingua : –. De Clercq, Karen (). A Unified Syntax of Negation. Doctoral dissertation, Ghent University. De Clercq, Karen (). ‘The Nanosyntax of French Negation’, in S. Cruschina, K. Hartmann, and E.-M. Remberger (eds), Studies on Negation. Syntax, Semantics and Variation. Vienna: Vienna University Press, –. De Clercq, Karen (a). ‘The Morphosyntax of Negative Markers. A Nanosyntactic Account’. Ms., Ghent University. De Clercq, Karen (b). ‘Syncretisms and the Morphosyntax of Negation’, in L. Baunaz, K. De Clercq, L. Haegeman, and E. Lander (eds), Exploring Nanosyntax. Oxford University Press, –. De Clercq, Karen and Guido Vanden Wyngaerd (). ‘Negation and the Functional Sequence’, Natural Language and Linguistic Theory: –. Available at (accessed  November ). de Hoop, Hellen (). ‘A Semantic Reanalysis of the Partitive Constraint’, Lingua : –. De Kock, Josse (). Gramática Española. Enseñanza e investigación, vol. . Salamanca: Universidad de Salamanca. Delsing, Lars-Olof (). The Internal Structure of Noun Phrases in the Scandinavian Languages. Doctoral dissertation, Lund University. Demske, Ulrike (). Merkmale und Relationen. Diachrone Studien zur Nominalphrase des Deutschen. Berlin: De Gruyter. den Besten, Hans (). ‘Double Negation and the Genesis of Afrikaans’, in P. Muysken and N. Smith (eds), Substrata versus Universals in Creole Genesis. Papers from the Amsterdam Workshop, April . Amsterdam: John Benjamins, –. Deo, Ashwini (). ‘Diachronic Semantics’, Annual Review of Linguistics : –. Depréz, Viviane (). ‘The Roots of Negative Concord in French and French Based Creoles’, in M. DeGraff (ed.), Language Creation and Language Change. Creole, Diachrony and Development. MIT Press, –. Déprez, Viviane (). ‘Parallel (A)symmetries and the Internal Structure of Negative Expressions’, Natural Language and Linguistic Theory : –. Depréz, Viviane and France Martineau (). ‘Micro-Parametric Variation and Negative Concord’, in J. Auger, C. Clements, and B. Vance (eds), Contemporary Approaches to Romance Linguistics. Amsterdam: John Benjamins, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

de Swart, Henriëtte (). Expression and Interpretation of Negation. An OT Typology. Dordrecht: Springer. de Swart, Henriëtte, Yoad Winter, and Joost Zwarts (). ‘Bare Nominals and Reference to Capacities’, Natural Language and Linguistic Theory : –. Detges, Ulrich and Richard Waltereit (). ‘Grammaticalization vs. Reanalysis: A Semantic Pragmatic Account of Functional Change in Grammar’, Zeitschrift für Sprachwissenschaft : –. Devine, Andrew M. and Laurence D. Stephens (). Latin Word Order. Structured Meaning and Information. Oxford: Oxford University Press. Diaconescu, Constanţa Rodica and María Luisa Rivero (). ‘An Applicative Analysis of Double Object Constructions in Romanian’, Probus : –. Dickey, Eleanor (). Latin Forms of Address. Oxford: Oxford University Press. Dičpinigaitytė, Aušra and Julija Korostenskienė (). ‘Manner Adverb Placement in English, Spanish and Lithuanian: A Corpus-Based Approach’, Verbum : –. Diessel, Holger (). Demonstratives. Amsterdam: John Benjamins. Diffloth, Gérard (). ‘Jah-Hut: An Austroasiatic Language of Malaysia’, in N. Dang Liem (ed.), South-East Asian Linguistic Studies, vol. . Canberra: Pacific Linguistics, –. Di Tullio, Ángela and Zdrojewski, Pablo (). ‘Notas sobre el doblado de clíticos en el español rioplatense: asimetrías entre objetos humanos y no humanos’, Filología : –. Dobrovie-Sorin, Carmen (). ‘Clitic Doubling, Wh-Movement, and Quantification in Romanian’, Linguistic Inquiry : –. Doron, Edit (). ‘The Pronominal “Copula” as Agreement Clitic’, in H. Borer (ed.), The Syntax of Pronominal Clitics. Orlando, FL: Academic Press, –. Dryer, Matthew S. (a). ‘Definite Articles’, in M. S. Dryer and M. Haspelmath (eds), The World Atlas of Languages Structures Online, Max Planck Institute for Evolutionary Anthropology. Available at http://wals.info/chapter/ (accessed  November ). Dryer, Matthew S. (b). ‘Indefinite Articles’, in M. S. Dryer and M. Haspelmath (eds), The World Atlas of Languages Structures Online, Max Planck Institute for Evolutionary Anthropology. Available at http://wals.info/chapter/ (accessed  November ). Dufter, Andreas and Elisabeth Stark (). ‘Double Indirect Object Marking in Spanish and Italian’, in E. Seoane and M. J. López-Couso (eds), Theoretical and Empirical Issues in Grammaticalization. Amsterdam: John Benjamins, –. Dunkel, George E. (). ‘IE Conjunctions: Pleonasm, Ablaut, Suppletion’, Zeitschrift für vergleichende Sprachforschung : –. Dunkel, George E. (). ‘The Common Origin of Conjunctive *-kwe and of Adverbial *-s’, in A. Hintze and E. Tichy (eds), Anusantatyai. Festschrift für Johanna Narten zum . Geburtstag (Münchener Studien zur Sprachwissenschaft ). Dettelbach: J. H. Röll, –. Dunkel, George E. (a). Lexikon der Indogermanischen Partikeln und Pronominalstämme, vol. . Heidelberg: Winter. Dunkel, George E. (b). Lexikon der Indogermanischen Partikeln und Pronominalstämme, vol. . Heidelberg: Winter. Dyvik, Helge (). ‘Omkring fremveksten av artiklene i norsk: Språklig markering av referensielle forutsetninger’, Maal og Minne : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Ebert, Robert P., Oskar Reichmann, Hans-Joachim Solms, and Klaus-Peter Wegera (). Frühneuhochdeutsche Grammatik. Tübingen: Niemeyer. Eckardt, Regine (). Meaning Change in Grammaticalization. An Enquiry into Semantic Reanalysis. Oxford: Oxford University Press. Egerland, Verner (). ‘Diachronic Change and Pronoun Status: Italian Dative “loro” ’, Linguistics : –. Egerland, Verner (). ‘I pronomi lo’ e ‘ro nel toscano dei primi secoli’, L’Italia Dialettale : –. Elenbaas, Marion (). The Synchronic and Diachronic Syntax of the English Verb-Particle Combination. Doctoral dissertation, Radboud University Nijmegen. Ellegård, Alvar (). The Auxiliary do. The Establishment and Regulation of its Use in English. Stockholm: Almquist and Wiksell. Elvira, Javier (). ‘Left Forever: Subject Datives and Clitic Doubling in Old Spanish’, in A. Dufter and Á. S. Octavio de Toledo y Huerta (eds), Left Sentence Peripheries in Spanish. Amsterdam: John Benjamins, –. Enç, Mürvet (). ‘The Semantics of Specificity’, Linguistic Inquiry : –. Ertl, Václav (). ‘Co je to za človĕka?’, Naše řeč : –. Espinal, M. Teresa (). ‘Bare Nominals in Catalan and Spanish: Their Structure and Meaning’, Lingua : –. Etxepare, Ricardo and Myriam Uribe-Etxebarria (). ‘On Negation and Focus in Spanish and Basque’, in X. Artiagoitia and J. Lakarra (eds), Gramatika Jaitan Patxi Goenagaren omenez. Bilbao: University of the Basque Country, –. Faarlund, Jan Terje (). The Syntax of Old Norse. With a Survey of the Inflectional Morphology and a Complete Bibliography. Oxford: Oxford University Press. Faarlund, Jan Terje (). ‘On the History of Definiteness Marking in Scandinavian’, Journal of Linguistics : –. Faarlund, Jan Terje, Svein Lie, and Kjell Ivar Vannebo (). Norsk referansegrammatikk. Oslo: Oslo University Press. Fábregas, Antonio (). Las nominalizaciones. Madrid: Visor Libros. Fábregas, Antonio and Rafael Marín (). ‘Deriving Individual-Level and Stage-Level Psych Verbs in Spanish’, The Linguistic Review : –. Fábregas, Antonio and Rafael Marín (a). ‘On Non-Dynamic Eventive Verbs in Spanish’, Linguistics : –. Fábregas, Antonio and Rafael Marín (b). ‘Datives and Stativity in Psych Predicates’, talk given at the Workshop on Dative Structures and Beyond. Autonomous University of Barcelona,  January . Falk, Hjalmar and Alf Torp (). Dansk-norskens syntax i historisk fremstilling. Oslo: H. Aschehoug and Co. Farkas, Donka and Kim Bruce (). ‘On Reacting to Assertions and Polar Questions’, Journal of Semantics : –. Fintel, Kai von (). ‘The Formal Semantics of Grammaticalization’, in J. N. Beckmann (ed.), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Fischer, Susann (). The Catalan Clitic System. A Diachronic Perspective on its Syntax and Phonology. Berlin: Mouton De Gruyter. Fischer, Susann (). ‘Rethinking the Tobler-Mussafia Law’, Diachronica : –. Fischer, Susann (). Word-Order Change as a Source of Grammaticalisation. Amsterdam: John Benjamins. Fischer, Susann (). ‘Some Notes on the Rise of Mismatches’, in N. Pomino and E. Stark (eds), Mismatches in Romance. Arbeitspapier , Konstanz: University of Konstanz, –. Fischer, Susann and Esther Rinke (). ‘Explaining the Variability in Clitic Doubling across Romance: A Diachronic Account’, Linguistische Berichte : –. Fischer, Susann, Christoph Gabriel, and Elena Kireva (). ‘Towards a Typological Classification of Judeo-Espanyol: Analyzing Syntax and Prosody of Bulgarian judezmo’, in K. Braunmüller, S. Höder, and K. Kühl (eds), Stability and Divergence in Language Contact. Factors and Mechanisms. Amsterdam: John Benjamins, –. Fischer, Susann, Mario Navarro, and Jorge Vega Vilanova (forthcoming). ‘Clitic Doubling and Language Contact’, in T. Parodi (ed.), Referential Properties of the Romance DP in the Context of Multilingualism. Proceedings of the VII Nereus International Workshop. Konstanz: University of Konstanz. Fitzmyer, Joseph A. (). ‘The Phases of the Aramaic Language’, in A Wandering Aramean: Collected Aramaic Essays. Missoula, MT: Scholars Press, –. Flack, Kathryn (). ‘Ambiguity Avoidance as Contrast Preservation: Case and Word Order Freezing in Japanese’, in L. Bateman, M. O’Keefe, E. Reilly, and A. Werle (eds), UMass Occasional Papers in Linguistics . Papers in Optimality Theory III. Amherst, MA: Graduate Linguistic Student Association, –. Fleischer, Jürg and Horst Simon (). ‘Vom Wert sprachwandelvergleichenden Arbeitens’, in J. Fleischer and H. Simon (eds), Sprachwandelvergleich-Comparing Diachronies. Berlin: De Gruyter, –. Fodor, Janet Dean (). ‘Unambiguous Triggers’, Linguistic Inquiry : –. Fodor, Janet Dean (). ‘Setting Syntactic Parameters’, in M. Baltin and C. Collins (eds), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell, –. Fodor, Janet Dean and Fernanda Ferreira (eds) (). Reanalysis in Sentence Processing. Dordrecht: Kluwer. Fonseca-Greber, Bonnie (). The Change from Pronoun to Clitic to Prefix and the Rise of Null Subjects in Spoken Swiss French. Doctoral dissertation, University of Arizona, Tucson. Fonseca-Greber, Bonnie and Linda R. Waugh (). ‘The Subject Clitics of Conversational European French: Morphologization, Grammatical Change, Semantic Change, and Change in Progress’, in R. Núñez-Cedeño, L. López, and R. Cameron (eds), A Romance Perspective on Language Knowledge and Use. Amsterdam: John Benjamins, –. Fontana, Josep M. (). Phrase Structure and the Syntax of Clitics in the History of Spanish. Doctoral dissertation, University of Pennsylvania. Foulet, Lucien (). Études de syntaxe française I: Quelque. Romania /, –. Foulet, Lucien  []. Petite syntaxe de l’ancien français (third edition). Paris: Honoré Champion.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Fox, Danny (). ‘Free Choice And Scalar Implicatures’, in U. Sauerland and P. Stateva (eds), Presupposition and Implicature in Compositional Semantics, London: Palgrave Macmillan, –. Fox, Danny and Roni Katzir (). ‘On the Characterization of Alternatives’, Natural Language Semantics : –. Franco, Jon and Susana Huidobro (). ‘Ethical Datives, Clitic Doubling and the Theory of pro’, in J. Bruhn de Garavito and E. Valenzuela (eds), Selected Proceedings of the th Hispanic Linguistics Symposium. Somerville, MA: Cascadilla, –. Frascarelli, Mara and Roland Hinterhölzl (). ‘Types of Topics in German and Italian’, in S. Winkler and K. Schwabe (eds), On Information Structure, Meaning and Form. Amsterdam: John Benjamins, –. Frascarelli, Mara and Francesca Ramaglia (). ‘(Pseudo)clefts at the Syntax-ProsodyDiscourse Interface’, in K. Hartmann and T. Veenstra (eds), Cleft Structures. Amsterdam: John Benjamins, –. Fuß, Eric (). The Rise of Agreement. A Formal Approach to the Syntax and Grammaticalization of Verbal Inflection. Amsterdam: John Benjamins. Gabelentz, Georg von der (). Die Sprachwissenschaft. Ihre Aufgaben, Methoden und bisherigen Ergebnisse. Leipzig: Weigel. (Reprinted Tübingen: Narr ). Gabriel, Christoph (). ‘On Focus, Prosody, and Word Order in Argentinean Spanish: A Minimalist OT Account’, Revista Virtual de Estudos da Linguagem, Special Issue  ‘Optimality-Theoretic Syntax’, –. Gajić, Jovana (). ‘Coordination and Focus Particles (Re?)United’, Proceedings of Sinn und Bedeutung . Preprint available at (accessed  November ). Garzonio, Jacopo and Cecilia Poletto (). ‘The Negative Marker that Escaped the Cycle: Some Notes on manco’, in C. Contemori and L. dal Pozzo (eds), Inquiries into Linguistic Theory and Language Acquisition. Papers offered to Adriana Belletti. Siena: Centro Interdipartimentale di Studi Cognitivi sul Linguaggio, –. Garzonio, Jacopo and Cecilia Poletto (). ‘On Preverbal Negation in Sicilian and Syntactic Parasitism,’ Isogloss Special Issue : –. Garzonio, Jacopo and Cecilia Poletto (). ‘How Bare are Bare Quantifiers? Some Notes from Diachronic and Synchronic Variation in Italian’, Linguistic Variation : –. Garzonio, Jacopo and Silvia Rossi (). ‘Can Structural Deficiency be Parametrized? Oblique Pronouns in Old Tuscan Varieties’, in S. Cruschina, A. Ledgeway, and E. M. Remberger (eds), Italian Dialectology at the Interfaces. Amsterdam: John Benjamins, –. Gelderen, Elly van (). Verbal Agreement and the Grammar behind its ‘Breakdown’. Minimalist Feature Checking. Tübingen: Niemeyer. Gelderen, Elly van (a). Grammaticalization as Economy. Amsterdam: John Benjamins. Gelderen, Elly van (b). ‘Economy, Innovation, and Prescriptivism: From Spec to Head and Head to Head’, The Journal of Comparative Germanic Linguistics (): –. Gelderen, Elly van (). ‘The DP Cycle in Germanic’, Journal of Germanic Linguistics : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Gelderen, Elly van (). ‘Linguistic Cycles and Economy Principles: The Role of Universal Grammar in Language Change’, in T. Eythórsson (ed.), Grammatical Change and Linguistic Theory. The Rosendal Papers. Amsterdam: John Benjamins, –. Gelderen, Elly van (a). ‘Renewal in the Left Periphery: Economy and the Complementiser Layer’, Transactions of the Philological Society : –. Gelderen, Elly van (b). ‘Feature Economy in the Linguistic Cycle’, in P. Crisma and G. Longobardi (eds), Historical Syntax and Linguistic Theory. Oxford: Oxford University Press, –. Gelderen, Elly van (ed.) (c). Cyclical Change. Amsterdam: John Benjamins. Gelderen, Elly van (). The Linguistic Cycle. Language Change and the Language Faculty. Oxford: Oxford University Press. Gelderen, Elly van (a). ‘The Diachrony of Pronouns and Demonstratives’, in T. Lohndal (ed.), In Search of Universal Grammar. From Old Norse to Zoque. Amsterdam: John Benjamins, –. Gelderen, Elly van (b). ‘The Linguistic Cycle and the Language Faculty’, Language and Linguistics Compass : –. Gelderen, Elly van (). ‘Changes in Psych-Verbs: A Reanalysis of Little v’, Catalan Journal of Linguistics : –. Gelderen, Elly van (a). ‘How’, in J. Bayer, R. Hinterhölzl, and A. Trotzke (eds), Discourseoriented Syntax. Amsterdam: John Benjamins, –. Gelderen, Elly van (b). ‘The Copula Cycle’, Lingue e Linguaggio ., Special Issue on Copulas: –. Gelderen, Elly van (ed.) (). Cyclical Change Continued. Amsterdam: John Benjamins. Gelderen, Elly van (a). The Diachrony of Verb Meaning. Aspect and Argument Structure. London: Routledge. Gelderen, Elly van (b). ‘Problems of Projection: The Role of Language Change in Labeling Paradoxes’, Studia Linguistica : –. Gerlach, Birgit (). Clitics between Syntax and Lexicon. Amsterdam: John Benjamins. Giannakidou, Anastasia (). The Landscape of Polarity Items. Doctoral dissertation, University of Groningen. Giannakidou, Anastasia (). ‘Only, Emotive Factive Verbs, and the Dual Nature of Polarity Dependency’, Language (): –. Gianollo, Chiara (). Indefinites between Latin and Romance. Oxford: Oxford University Press. Gibert-Sotelo, Elisabeth (a). ‘On the Encoding of Negation by Source Prefixes and The Satellite-/Verb-Framed Typology: Evidence from Latin and Spanish’, talk given at the Workshop on Morphosyntactic Variation in Adpositions, Cambridge: Queens’ College,  May . Gibert-Sotelo, Elisabeth (b). Source and Negative Prefixes. On the Syntax-Mexicon Interface and the Encoding of Spatial Relations. Doctoral dissertation, University of Girona. Girard, Francine (). ‘Le statut des clitiques sujets cadiens’, in F. Neveu, V. Toke, J. Durand, T. Klingler, L. Mondada, and S. Prévost (eds), in Actes du deuxième Congrès Mondial de Linguistique Française. Paris: Edition Diffusion Presse Sciences, –. Available at

(accessed  November ).

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Giusti, Giuliana and Rossella Iovino (). ‘Psychological Verbs as a Vulnerable Syntactic Domain: A Comparative Study of Latin and Italian’, Lingua : –. Givón, Talmy (). ‘Topic, Pronoun, and Grammatical Agreement’, in C. Li (ed.), Subject and Topic. New York: Academic Press, –. Givón, Talmy (). ‘From Discourse to Syntax: Grammar as a Processing Strategy’, in T. Givón (ed.), Syntax and Semantics . Discourse and Syntax. New York: Academic Press, –. Givón, Talmy (). The Story of Zero. Amsterdam: John Benjamins. Goldenberg, Gideon (). ‘On Syriac Sentence-Structure’, in S. Michael (ed.), Arameans, Aramaic and the Aramaic Literary Tradition. Ramat-Gan: Bar-Ilan University Press, –. Gonçalves, Rita, Duarte, Inês, and Hagemeijer, Tjerk (). ‘Revisiting Dative Constructions: A VP-Shell Based Approach’, talk given at the th Colloquium on Generative Grammar. Cáceres,  April . Gonda, Jan (). ‘The History and Original Function of the Indo-European Particle *kwe, especially in Greek and Latin’, Mnemosyne : –. Greenberg, Joseph (). ‘Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements’, in J. Greenberg (ed.), Universals of Grammar. Cambridge, MA: MIT Press, –. Grevisse, Maurice and André Goosse. [] . Le bon usage. Paris: Duculot. Grewendorf, Günther and Cecilia Poletto (). ‘Von OV zu VO: Ein Vergleich zwischen Zimbrisch und Plodarisch’, in E. Bidese, J. R. Dow, and T. Stolz (eds), Das Zimbrische zwischen Germanisch und Romanisch. Bochum: Brockmeyer, –. Grewendorf, Günther and Cecilia Poletto (). ‘The Hybrid Complementizer System of Cimbrian’, in V. Moscati and E. Servidio (eds), XXXV Incontro di Grammatica Generativa Proceedings. Siena: Centro Interdipartimentale di Studi Cognitivi sul Linguaggio, –. Grimm, Jacob and Wilhelm Grimm (). Deutsches Wörterbuch von Jacob und Wilhelm Grimm, vol. , part . Forschel – Gefolgsmann. Leipzig: Hirzel. Grimm, Jacob and Wilhelm Grimm (). Deutsches Wörterbuch von Jacob und Wilhelm Grimm, vol. . Wenig – Wiking. Leipzig: Hirzel. Grimshaw, Jane (). ‘Extended Projection and Locality’, in P. Coopmans, M. Everaert, and J. Grimshaw (eds), Lexical Specification and Insertion. Amsterdam: John Benjamins, –. Grohmann, Kleanthes (). Prolific Domains. Amsterdam: John Benjamins. Grosu, Alexander (). ‘The Syntax-Semantics of Modal Existential Wh-Constructions’, in O. M. Tomić (ed.), Balkan Syntax and Semantics. Amsterdam: John Benjamins,  . Grøtvedt, Per Nyquist (). ‘Forholdet mellom skrift og talespråk i sørøstnorsk i senmiddelalderen’, Arkiv för nordisk filologi : –. Gutiérrez Ordóñez, Salvador (). ‘Los dativos’, in I. Bosque and V. Demonte (eds), Gramática descriptiva de la lengua española. Madrid: Espasa-Calpe, –. Gutiérrez-Bravo, Rodrigo (). ‘A Reinterpretation of Quirky Subjects and Related Phenomena in Spanish’, in C. Nishida and J.-P. Montreuil (eds), New Perspectives in Romance Linguistics. Amsterdam: John Benjamins, –. Haase, Alfred (). Syntaxe française du XVIIe siècle, (seventh edition), trans. and rev. by M. Obert. Paris: Delagrave.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Haeberli, Eric and Tabea Ihsane (). ‘On the Interaction between Syntax and Morphology: New Evidence from the Loss of Verb Movement in English’, in Travaux du ème CIL/th ICL Papers, University of Geneva, Department of Linguistics. Available at (accessed  September ). Haeberli, Eric and Tabea Ihsane (). ‘Revisiting the Loss of Verb Movement in the History of English’, Natural Language and Linguistic Theory : –. Haegeman, Liliane (). The Syntax of Negation. Cambridge: Cambridge University Press. Haegeman, Liliane (). ‘Negative Preposing, Negative Inversion, and the Split CP’, in L. Horn and Y. Kato (eds), Negation and Polarity. Oxford: Oxford University Press, –. Haegeman, Liliane (). Adverbial Clauses, Main Clause Phenomena, and the Composition of the Left Periphery. Oxford: Oxford University Press. Haegeman, Liliane and Terje Lohndal (). ‘Negative Concord and Multiple Agree: A Case Study of West-Flemish’, Linguistic Inquiry (): –. Haegeman, Liliane and Raffaella Zanuttini (). ‘Negative Heads and the NEG-criterion’, The Linguistic Review : –. Haegeman, Liliane, André Meinunger, and Aleksandra Vercauteren (). ‘The Syntax of It-clefts and the Left Periphery of the Clause’, in U. Shlonsky (ed.), Beyond Functional Sequence. Oxford: Oxford University Press, –. Hagland, Jan Ragnar (). ‘Gammalislandsk og gammalnorsk språk’, in O. E. Haugen (ed.), Handbok i norrøn filologi (second edition). Bergen: Fagbokforlaget, –. Hale, Ken (). ‘Person Marking in Warlbiri’, in S. Anderson and P. Kiparsky (eds), A Festschrift for Morris Halle. New York: Holt, Rinehart and Winston, –. Hale, Ken (). ‘The Adjoined Relative Clause in Indo-European’, in R. Dixon (ed.), Grammatical Categories in Australia. Canberra: Australian Institute of Aboriginal Studies, –. Hale, Mark (). Historical Linguistics. Theory and Method. Oxford: Blackwell. Halmøy, Madeleine (). The Norwegian Nominal System. A Neo-Saussurean Perspective. Berlin: De Gruyter. Hamre, Lars (). Innføring i diplomatikk. Førelesningar. Oslo: Universitetsforlaget. Harbour, Daniel (). ‘Poor Pronoun Systems and What they Teach Us’, Nordlyd : –. Harris, Martin (). ‘ “Demonstratives”, “Articles” and “Third Person Pronouns” in French: Changes in Progress’, Zeitschrift für romanische Philologie : –. Harris, Martin (). The Evolution of French Syntax. A Comparative Approach. London: Longman. Harwood, William (). ‘Rise of the Auxiliaries: A Case for Auxiliary Raising vs. Affix Lowering’, The Linguistic Review : –. Haspelmath, Martin (). Indefinite Pronouns. Oxford: Oxford University Press. Haspelmath, Martin (). ‘Non-canonical Marking of Core Arguments in European Languages’, in A. Aikhenvald, R. M. W. Dixon, and M. Onishi (eds), Non-canonical Marking of Subjects and Objects. Amsterdam: John Benjamins, –. Haude, Katharina (). ‘Tense Marking on Dependent Nominals in Movima’, in R. Musan and M. Rathert (eds), Tense across Languages. Berlin: De Gruyter, –. Haugen, Jason (). Morphology at the Interfaces. Amsterdam: John Benjamins.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Hawkins, John A. (). ‘On (In)Definite Articles: Implicatures and (Un)Grammaticality Prediction’, Journal of Linguistics : –. Heck, Fabian (). ‘Categories, Recursion and Bare Phrase Structure’, talk given at Generative Grammatik des Südens conference, FU Berlin. Ms, University of Leipzig. Heggstad, Leiv, Finn Hødnebø, and Erik Simensen (eds) (). Norrøn ordbok (fifth edition). Oslo: Samlaget. Heim, Irene and Angelika Kratzer (). Semantics in Generative Grammar. Oxford: Blackwell. Heine, Bernd (). Cognitive Foundations of Grammar. Oxford: Oxford University Press. Heine, Bernd and Tania Kuteva (). World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heltoft, Lars (). ‘Paradigmatic Structure and Reanalysis: From NPs to DPs in Scandinavian’, Acta Linguistica Hafniensia : –. Heycock, Caroline and Roberto Zamparelli (). ‘Coordinated Bare Definites’, Linguistic Inquiry : –. Heycock, Caroline and John Sundquist (). ‘Don’t Rush to Rehabilitate: A Remark on Koeneman and Zeijlstra ’, Linguistic Inquiry : –. Heycock, Caroline and Joel Wallenberg (). ‘How Variational Acquisition Drives Syntactic Change: The Loss of Verb Movement in Scandinavian’, Journal of Comparative Germanic Linguistics : –. Hill, Virginia and Liliane Tasmowski (). ‘Romanian Clitic Doubling: A View from Pragmatics-Semantics and Diachrony’, in D. Kallulli and L. Tasmowski (eds), Clitic Doubling in the Balkan Languages. Amsterdam: John Benjamins, –. Hobich, Melanie (). Was für ein interessantes Thema. Zum Ursprung der was für-Konstruktion im Deutschen. MA thesis, Goethe University Frankfurt. Hock, Hans H. (). Principles of Historical Linguistics (second edition). Berlin: Mouton de Gruyter. Holmberg, Anders (). ‘Questions, Answers, Polarity and Head Movement in Germanic and Finnish’, Nordlyd : –. Holmberg, Anders (). ‘The Syntax of Answers to Polar Questions in English and Swedish’, Lingua : –. Holmberg, Anders (). The Syntax of Yes and No. Oxford: Oxford University Press. Holmberg, Anders and Ian Roberts (). ‘The Syntax-Morphology Relation’, Lingua : –. Holmberg, Anders and Ian Roberts (). ‘Parameters and the Three Factors of Language Design’, in C. Picallo (ed.), Linguistic Variation in the Minimalist Framework. Oxford: Oxford University Press, –. Hopper, Paul and Elizabeth Traugott (). Grammaticalization. Cambridge: Cambridge University Press. Horn, Laurence (). A Natural History of Negation. Chicago, IL: The University of Chicago Press. Hornstein, Norbert (). A Theory of Syntax. Cambridge: Cambridge University Press. Huang, James (). ‘Move WH in a Language without WH Movement’, The Linguistic Review : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Hulsey, Sarah, and Uli Sauerland (). ‘Sorting out Relative Clauses’, Natural Language Semantics : –. Ihsane, Tabea (). The Layered DP in French. Form and Meaning of French Indefinites. Amsterdam: John Benjamins. Indrebø, Gustav (). Norsk målsoga (second edition). Bergen: Norsk bokreidingslag. Institut d’Estudis Catalans – IEC (). El català de l’Alguer. Un model d’àmbit restringit. Barcelona: Institut d’Estudis Catalans. Institut d’Estudis Catalans – IEC (). Gramàtica de la llengua catalana. Barcelona: Institut d’Estudis Catalans. Israel, Michael (). ‘Polarity Sensitivity as Lexical Semantics’, Linguistics and Philosophy : –. Jackendoff, Ray (). ‘Quantifiers in English’, Foundations of Language : –. Jackendoff, Ray (). X-bar Syntax. A Study of Phrase Structure. Cambridge, MA: MIT Press. Jaeggli, Osvaldo (). ‘Tres cuestiones en el estudio de los clíticos: El caso, los sintagmas nominales reduplicados y las extracciones’, in O. Fernández Soriano (ed.), Los pronombres átonos. Madrid: Taurus, –. Jäger, Agnes (). ‘Negation in Old High German’, Zeitschrift für Sprachwissenschaft : –. Jäger, Agnes (a). ‘Anything is Nothing is Something: On the Diachrony of Polarity Types of Indefinites’, Natural Language and Linguistic Theory : –. Jäger, Agnes (b). ‘Der Komparativzyklus und die Position der Vergleichspartikeln’, Linguistische Berichte : –. Jayaseelan, Karattuparambil (). ‘Comparative Morphology of Quantifiers’, Lingua : –. Jelinek, Eloise (). ‘Pronouns and Argument Hierarchies’, talk given at Workshop on Agreement and Argument Structure. Utrecht University, June. Jensen, Eva Skafte (). ‘Om udviklingen af den ubestemte artikel i dansk’, in M. Birkelund, S. Fernándes, and H. Nølke Kratschmer (eds), Ny forskning i grammatik . University of Southern Denmark, –. Jespersen, Otto (). Negation in English and Other Languages. Copenhagen: A. F. Høst & Søn. Johnsen, Oscar Albert (). ‘Indledning’, in O. A. Johnsen (ed.), Olafs saga hins helga. Efter pergamenthaandskrift i Uppsala Universitetsbibliotek, Delagardieske samling nr. II. Oslo: Jacob Dybwad, I–XXVII. Joosten, Jan (). ‘The Negation of the Non-Verbal Clause in Early Syriac’, Journal of the American Oriental Society : –. Julien, Marit (). Nominal Phrases from a Scandinavian Perspective. Amsterdam: John Benjamins. Kadmon, Nirit and Fred Landman (). ‘Any’, Linguistics and Philosophy : –. Karttunen, Lauri (). ‘Discourse Referents’, in Proceedings of the  Conference on Computational Linguistics, Oslo: Association for Computational Linguistics. Available at (accessed  November ). Kayne, Richard (). French Syntax. Cambridge, MA: MIT Press. Kayne, Richard (). Connectedness and Binary Branching. Dordrecht: Foris.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Kayne, Richard (). ‘Facets of Romance Past Participle Agreement’, in P. Benincà (ed.), Dialect Variation and the Theory of Grammar. Dordrecht: Foris, –. Kayne, Richard (). The Antisymmetry of Syntax. Cambridge, MA: The MIT Press. Kayne, Richard (). Movement and Silence. Oxford: Oxford University Press. Kayne, Richard (). ‘Why Isn’t this a Complementizer?’ in R. Kayne, Comparisons and Contrasts. Oxford: Oxford University Press, –. Kayne, Richard (). ‘The Silence of Heads’, Studies in Chinese Linguistics : –. Keenan, Edward L. (). ‘Towards a Universal Definition of Subject’, in C. Li (ed.), Subject and Topic. New York: Academic Press, –. King, Ruth, France Martineau, and Raymond Mougeon (). ‘The Interplay of Internal and External Factors in Grammatical Change: First-Person Plural Pronouns in French’, Language : –. Kinn, Kari (). Formelle subjekter i norsk. En diakron undersøkelse. MA thesis, University of Oslo. Kinn, Kari (a). Null Subjects in the History of Norwegian. Doctoral dissertation, University of Oslo. Kinn, Kari (b). ‘Referential vs. Non-Referential Null Subjects in Middle Norwegian’, Nordic Journal of Linguistics : –. Kiparsky, Paul (). ‘ “Elsewhere” in Phonology’, in S. Anderson and P. Kiparsky (eds), A Festschrift for Morris Halle. New York: Holt, Rinehart & Winston, –. Kiparsky, Paul (). ‘Indo-European Origins of Germanic Syntax’, in A. Battye and I. Roberts, Clause Structure and Language Change. Oxford: Oxford University Press, –. Kiparsky, Paul and Cleo Condoravdi (). ‘Tracking Jespersen’s Cycle’, in M. Janse, B. Joseph, and A. Ralli (eds), Proceedings of the nd International Conference of Modern Greek Dialects and Linguistic Theory. Mytilene: Doukas, –. Klima, Edward S. (). Studies in Diachronic Transformational Syntax. Doctoral dissertation, Harvard University. Koeneman, Olaf (). The Flexible Nature of Verb Movement. Utrecht: LOT Publications. Koeneman, Olaf and Hedde Zeijlstra (). ‘The Rich Agreement Hypothesis Rehabilitated’, Linguistic Inquiry : –. Kolmer, Agnes (). Pronomina und Pronominalklitika im Cimbro. Untersuchungen zum grammatischen Wandel einer deutschen Minderheitensprache in romanischer Umgebung. Stuttgart: Steiner. Korostenskaja, Julija (). ‘Setting the Verb Raising Parameter in Lithuanian’, Kalbų Studijos : –. Korostenskienė, Julija (). ‘On Some Properties of the Aspectual Projection and Verbal Movement in Russian and Lithuanian’, Darnioji daugiakalbystė/Sustainable Multilingualism : –. Kosmeijer, Wim (). ‘The Status of the Finite Inflection in Icelandic and Swedish’, Working Papers in Scandinavian Syntax : –. Kratzer, Angelika and Junko Shimoyama (). ‘Indeterminate Phrases: The View from Japanese’, in Y. Otsu (ed.), The Proceedings of the Third Tokyo Conference on Psycholinguistics. Tokyo: Hituzi Syobo, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Krause, Thomas and Amir Zeldes (). ‘ANNIS: A New Architecture for Generic Corpus Query and Visualization’, Digital Scholarship in the Humanities : –. Kroch, Anthony S. (). ‘Syntactic Change’, in M. Baltin and C. Collins (eds), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell, –. Krug, Manfred (). ‘Auxiliaries and Grammaticalization’, in B. Heine and H. Narrog (eds), The Oxford Handbook of Grammaticalization. Oxford: Oxford University Press, –. Kruken, Kristoffer and Ola Stemshaug (eds) (). Norsk personnamnleksikon (third edition). Oslo: Samlaget. Kruspe, Nicole (). A Grammar of Semelai. Cambridge: Cambridge University Press. Kwon, Kyongjoon (). ‘What for Diachronically’. Ms., Sungkyunkwan University. Kwon, Kyongjoon (). ‘What is for for? Reconstructing the Development of what for Construction in Russian’, Transactions of the Philological Society : –. Labelle, Marie and M. Teresa Espinal (). ‘Diachronic Changes in Negative Expressions: The Case of French’, Lingua : –. Ladusaw, William (). Polarity Sensitivity as Inherent Scope Relations. Doctoral dissertation, University of Texas. Ladusaw, William A. (). ‘Negation, Indefinites, and the Jespersen Cycle’, in J. Guenter, B. Kaiser, and C. Zoll (eds), Proceedings of the Nineteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, –. Laka, Itziar (). On the Syntax of Negation. New York: Garland Publishing. Lambrecht, Knud (). Topic, Antitopic and Verb-Agreement in Non-Standard French. Amsterdam: John Benjamins. Landau, Idan (). The Locative Syntax of Experiencers. Cambridge, MA: MIT Press. Lander, Eric T. and Liliane Haegeman (). ‘Old Norse as an NP Language’, Transactions of the Philological Society : –. Larrivée, Pierre (). ‘The Pragmatic of Marked Configurations: Negative Doubling in French’, Journal of Pragmatics : –. Larrivée, Pierre and Richard P. Ingham (eds) (). The Evolution of Negation beyond the Jespersen Cycle. Berlin: De Gruyter. Lebeaux, David (). ‘Relative Clauses, Licensing and the Nature of the Derivation’, in S. Rothstein (ed.), Syntax and Semantics . Perspectives on Phrase Structure. New York: Academic Press, –. Ledgeway, Adam (). ‘Late Latin Verb Second: The Sentential Word Order of the “Itinerarium Egeriae” ’, Catalan Journal of Linguistics : –. Lee, Hanjung (). ‘Markedness and Word Order Freezing’, in P. Sells (ed.), Formal and Empirical Issues in Optimality Theoretic Syntax. Stanford, CA: Stanford Center for the Study of Language and Information, –. Lee, Kee-dong (). Kusaiean Reference Grammar. Honolulu, HA: University Press of Hawaii. Leijström, Gunnar (). Om obestämda artikeln. Et bidrag til nordisk språkhistorie. Stockholm: Hugo Geber. Lehmann, Christian (). ‘Grammaticalization: Synchronic Variation and Diachronic Change’, Lingua e stile : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Lehmann, Christian (). Thoughts on Grammaticalization. Revised and Expanded Version. Munich: Lincom Europa. Lehmann, Christian (). ‘New Reflections on Grammaticalization and Lexicalization’, in I. Wischer and G. Diewald (eds), New Reflections on Grammaticalization. Amsterdam: John Benjamins, –. Lehmann, Winfred P. (). ‘From Topic to Subject in Indo-European’, in C. Li (ed.), Subject and Topic. New York: Academic Press, –. Lehmann, Winfred P. (). Theoretical Bases of Indo-European Linguistics. New York: Routledge. Leonetti, Manuel (). ‘Specificity in Clitic Doubling and in Differential Object Marking’, Probus : –. Leu, Thomas (). ‘Something Invisible in English’, University of Pennsylvania Working Papers in Linguistics : –. Leu, Thomas (a). The Internal Syntax of Determiners. Doctoral dissertation, New York University. Leu, Thomas (b). ‘What for Internally’, Syntax : –. Li, Fang Kuei ( []). Chipewyan. New York: Viking Fund. Lightfoot, David (). Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, David (). How to Set Parameters. Arguments from Language Change. Cambridge, MA: MIT Press. Lightfoot, David (). The Development of Language. Acquisition, Change, and Evolution. Malden: Blackwell. Lightfoot, David (). ‘Grammatical Approaches to Syntactic Change’, in B. Joseph and R. Janda (eds), The Handbook of Historical Linguistics. Oxford: Blackwell, –. Lipski, John M. (). ʻSpanish and Portuguese in Contactʼ, in R. Hickey (ed.), The Handbook of Language Contact. Malden: Wiley-Blackwell, –. Lipták, Anikó (). ‘The Syntax of Emphatic Positive Polarity in Hungarian: Evidence from Ellipsis’, Lingua : –. Lødrup, Helge (). ‘Split Possession and the Syntax of Kinship Nouns in Norwegian’, Journal of Comparative Germanic Linguistics : –. Lohndal, Terje (). ‘On the Structure and Development of Nominal Phrases in Norwegian’, in E. Stark, E. Leiss, and W. Abraham (eds), Nominal Determination. Typology, Context Constraints, and Historical Emergence. Amsterdam: John Benjamins, –. Lohndal, Terje (). ‘Silent Elements and Some Norwegian Exclamatives’, Linguistic Analysis : –. Longobardi, Giuseppe (). ‘Reference and Proper Names: A Theory of N-movement in Syntax and Logical Form’, Linguistic Inquiry : –. Longobardi, Giuseppe (). ‘The Structure of DP: Some Principles, Parameters and Problems’, in M. Baltin and C. Collins (eds), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell, –. Loporcaro, Michele (). ‘Il pronome loro nell’Italia centro-meridionale e la storia del sistema pronominale romanzo’, Vox Romanica : –. Loprieno, Antonio (). Ancient Egyptian. Cambridge: Cambridge University Press.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Lundeby, Einar (). Overbestemt substantiv i norsk og de andre nordiske språk. Oslo: Universitetsforlaget. Lyons, Christopher (). Definiteness. Cambridge: Cambridge University Press. Machado-Rocha, Ricardo and Martins Ramos, Jânia (). ‘Clitic Doubling and Pure Agreement Person Features’, Revista de Estudos da Linguagem : –. Manzini, M. Rita (). ‘Grammatical Categories: Strong and Weak Pronouns in Romance’, Lingua : –. Manzini, M. Rita and Leonardo Savoia (). ‘The Nature of Complementizers’, Rivista di Grammatica Generativa : –. Manzini, M. Rita and Leonardo Savoia (). Grammatical Categories. Variation in Romance Languages. Cambridge: Cambridge University Press. Manzini, M. Rita and Leonardo Savoia (). ‘From Latin to Romance: Case Loss and Preservation in Pronominal Systems’, Probus : –. Marantz, Alec (). ‘Locality Domains for Contextual Allomorphy Across the Interfaces’, in O. Matushansky and A. Marantz (eds), Distributed Morphology Today. Cambridge, MA: MIT Press, –. Marín, Rafael and Louise McNally (). ‘Inchoativity, Change of State, and Telicity: Evidence from Spanish Reflexive Psychological Verbs’, Natural Language and Linguistic Theory : –. Martineau, France and Marie-Thérese Vinet (). ‘Microvariation in French Negation Markers: A Historical Perspective’, in M. Battlori, M.-L. Hernanz, C. Picallo, and F. Roca (eds), Grammaticalization and Parametric Variation. Oxford: Oxford University Press, –. Martínez Gavilán, M. Dolores (). Las ideas lingüísticas en España en el siglo XVII. Los tratados gramaticales. Doctoral dissertation, University of León. Martins, Ana Maria (). Cliticos na história do Português. Doctoral dissertation, University of Lisbon. Mascaró, Joan (). Morfologia. Barcelona: Enciclopèdia Catalana. Masullo, Pascual J. (). ‘Two Types of Quirky Subjects: Spanish Versus Icelandic’, in A. J. Schafer (ed.), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, –. Mathiassen, Terje (). A Short Grammar of Lithuanian. Bloomington, IN: Slavica Publishers. Matushansky, Ora (). ‘On the Linguistic Complexity of Proper Names’, Linguistics and Philosophy : –. McWhorter, John H. (). Defining Creole. Oxford: Oxford University Press. Meillet, Antoine (). ‘L’évolution des formes grammaticales’, Scientia : –. (Reprinted in Meillet, Antoine (). Linguistique historique et linguistique générale. Paris: H. Champion, –.). Meinunger, André (). ‘The Structure of Cleft and Pseudo-Cleft Sentences’, in M. Moosally and R. Blight (eds), Texas Linguistic Forum . The Syntax and Semantics of Predication. Austin, TX: University of Texas Department of Linguistics, –. Meinunger, André (). ‘A Monoclausal Structure for (Pseudo-)Cleft Sentences’, in P. N. Tamanji and K. Kusumoto (eds), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Melis, Chantal (). ‘Sobre la historia sintáctica de gustar’, in C. García Turza, F. González Bachiller, and J. Mangado Martínez (eds), Actas del IV Congreso Internacional de Historia de la Lengua Española. La Rioja, - de abril de . Logroño: University of Rioja, vol. , –. Melis, Chantal and Marcela Flores (). ‘The Dative Experiencer of Spanish gustar’, in M. Bouzouita, I. Sitaridou, and E. Pato (eds), Studies in Historical Ibero-Romance MorphoSyntax, Amsterdam: John Benjamins, –. Menge, Hermann (). Lehrbuch der lateinischen Syntax und Semantik. Darmstadt: Wissenschaftliche Buchgesellschaft. Miklosich, Franz (–). Vergleichende Grammatik der slavischen Sprachen, vol. : Syntax. Vienna: Wilhelm Braumüller. Miller, D. Gary (). Complex Verb Formation. Amsterdam: John Benjamins. Milsark, Gary (). ‘Toward an Explanation of Certain Peculiarities of the Existential Construction in English’, Linguistic Analysis : –. Mithun, Marianne (). ‘The Development of Bound Pronominal Paradigms’, in W. Lehmann and H.-J. Jakusz Hewitt (eds), Language Typology . Amsterdam: John Benjamins, –. Mitrović, Moreno (). Morphosyntactic Atoms of Propositional Logic. A Philo-logical Programme. Doctoral dissertation, University of Cambridge. Mitrović, Moreno (forthcoming). Microsemantics. Series in Natural Language & Linguistic Theory. Dordrecht: Springer. Mitrović, Moreno and Xuhui Hu (). ‘The Grammar of Distributivity in Ancient Chinese’, talk given at the th International Symposium on Ancient Chinese Grammar, Berlin,  July . Mitrović, Moreno and Uli Sauerland (). ‘Decomposing Coordination’, in J. Iyer and L. Kusmer (eds), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, vol. , –. Mitrović, Moreno and Uli Sauerland (). ‘Two Conjunctions are Better than One’, Acta Linguistica Hungarica : –. Mørck, Endre (). ‘Ellipsis of the Subject of Main Clauses in Middle Norwegian’, in E. H. Jahr and O. Lorentz (eds), Tromsø Linguistics in the Eighties. Oslo: Novus Press, –. Mørck, Endre (). Leddstillinga i mellomnorske heilsetninger. Funksjons- og feltanalyse og materialpresentasjon. Oslo: Novus. Mørck, Endre (). ‘Syntaks’, in H. Sandøy (ed.), Norsk språkhistorie. Mønster, vol. . Oslo: Novus, –. Mosegaard Hansen, Maj-Britt (). ‘Negation in the History of French’, in D. Willis, C. Lucas, and A. Breitbarth (eds), The History of Negation in the Languages of Europe and the Mediterranean. Oxford: Oxford University Press, –. Moser, Michael (). Die polnische, ukrainische und weißrussische Interferenzschicht im russischen Satzbau des . und . Jahrhunderts (= Schriften über Sprachen und Texte, ). Frankfurt am Main: Peter Lang. Mundal, Else (). ‘Sagalitteraturen’, in O. E. Haugen (ed.), Handbok i norrøn filologi (second edition). Bergen: Fagbokforlaget, –. Muntendam, Antje (). ‘Crosslinguistic Influence in Andean Spanish: Word Order and Focus’, in M. Bowles, R. Foote, S. Perpiñán, and R. Bhatt (eds), Selected Proceedings of the  Second Language Research Forum. Somerville, MA: Cascadilla, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Muraoka, Takamitsu and Bezalel Porten (). Grammar of Egyptian Aramaic. Leiden: Brill. Navarro, Mario, Susann Fischer, and Jorge Vega Vilanova (). ‘Reconstruyendo un ciclo: Doblado de clíticos y gramaticalización en las lenguas romances’, in S. Gumiel-Molina, M. Leonetti, and I. Pérez-Jiménez (eds), Investigaciones en Lingüística, vol. III: Sintaxis. Alcalá: University of Alcalá, –. Neeleman, Ad and Kriszta Szendrői (). ‘Radical Pro Drop and the Morphology of Pronouns’, Linguistic Inquiry : –. Newmeyer, Frederick J. (). Language Form and Language Function. Cambridge, MA: MIT Press. Newmeyer, Frederick J. (). ‘Against a Parameter-Setting Approach to Language Typology’, Linguistic Variation Yearbook, –. Nishida, Chiyo (). ‘A Corpus Study of Mexican Spanish Three-Participant Constructions with and without Clitic Doubling’, Linguistic Discovery : –. Nordlinger, Rachel and Louisa Sadler (). ‘Nominal Tense in Crosslinguistic Perspective’, Language (): –. Nygaard, Marius (). Norrøn syntax. Oslo: Aschehoug. Obenauer, Hans-Georg (). Études de syntaxe interrogative du français. Quoi, combien et le complémenteur. Tübingen: Niemeyer. Octavio de Toledo y Huerta, Álvaro (). ‘Futuros que se miran el ombligo: Mesoclisis y anteposición de formas no personales en la historia del español’, in M. Castillo Lluch and M. López Izquierdo (eds), El orden de palabras en la historia del español y otras lenguas iberorromances. Madrid: Visor Libros, –. Ordóñez, Francisco (). The Clausal Architecture of Spanish. A Comparative Study. New York: Garland. Ottóson, Kjartan (). ‘Person- og numerusbøying av verb i gammalnorsk og mellomnorsk’, in J. T. Faarlund (ed.), Språk i endring. Indre norsk språkhistorie. Oslo: Novus forlag, –. Pafel, Jürgen (). ‘Die syntaktische und semantische Struktur von was für Phrasen’, Linguistische Berichte : –. Panieri, Luca, Monica Pedrazza, Adelia Nicolussi Baiz, Sabine Hipp, and Cristina Pruner (). Bar lirnen z’schraiba un zo reda az be biar. Grammatica del cimbro di Luserna/ Grammatik der zimbrischen Sprache von Lúsern. Regione Autonoma Trentino-Alto Adige/ Autonome Region Trentino-Südtirol & Istituto Cimbro/Kulturinstitut Lúsern. Pat-El, Na’ama (). ‘Syntactical Aspects of Negation in Syriac’, Journal of Semitic Studies : –. Paul, Hermann (ed.) (). Deutsche Grammatik, vol. III, part IV: Syntax (first half ). Halle: Niemeyer. Penka, Doris (). Negative Indefinites. Doctoral dissertation, University of Tübingen. Penka, Doris (). ‘Negative Features in Negative Indefinites: Evidence from Split Scope’, Journal of Semantics : –. Penka, Doris and Hedde Zeijlstra (). ‘Negation and Polarity: An Introduction’, Natural Language and Linguistic Theory : –. Pereltsvaig, Asya (). ‘Small Nominals’, Natural Language and Linguistic Theory : –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Pesetsky, David and Esther Torrego (). ‘T-to-C Movement: Causes and Consequences’, in M. Kenstowicz (ed.), Ken Hale. A Life in Language. Cambridge, MA: MIT Press, –. Platzack, Christer, and Anders Holmberg (). ‘The Role of AGR and Finiteness in Germanic VO Languages’, Working Papers in Scandinavian Syntax : –. Podlesskaja, Vera I. (). ‘Mnogoznačnost‘ konstrukcii “čto za pljus imennaja gruppa” v svete dannyx Nacional’nogo Korpusa Russkogo Jazyka: Čto že èto za konstrukcija?!’, in Trudy meždunarodnoj konferencii ‘Dialog ’: –. Poletto, Cecilia (). ‘On Negative Doubling’. Ms., University of Venice. Poletto, Cecilia (). ‘Two Instances of a Broken Cycle: Sentential Particles in Old Italian’, in E. van Gelderen (ed.), Cyclical Change. Amsterdam: John Benjamins, –. Poletto, Cecilia (). Word Order in Old Italian. Oxford: Oxford University Press. Poletto, Cecilia (). ‘Negative Doubling: In Favor of a Big NegP Analysis’, in S. Cruschina, K. Hartmann, and E.-M. Remberger (eds), Studies on Negation. Syntax, Semantics and Variation. Vienna: Vienna University Press, –. Poletto, Cecilia and Emanuela Sanfelici (). ‘On the Nature of Complementizers: Insights from Italian Subject Relative Clauses’, talk given at the th Symposium on Romance Linguistics, Going Romance. University of Lisbon,  December . Poletto, Cecilia and Emanuela Sanfelici (). ‘On Relative Complementizers and Relative Pronouns’, Linguistic Variation (): –. [Special issue, ‘Comparative approaches to the Complementizer Phrase’, eds Jacopo Garzonio and Silvia Rossi.] Pollock, Jean-Yves (). ‘Verb Movement, Universal Grammar, and the Structure of IP’, Linguistic Inquiry : –. Pons-Moll, Clàudia, Francesc Torres-Tamarit, and Ignasi Mascaró (). ‘ProsodicallyDriven Morpheme Non-Realization in the Minorcan Catalan DP’, talk given at the th Manchester Phonology Meeting, University of Manchester,  May . Puskás, Genoveva (). ‘Licensing Double Negation in NC and non-NC Languages’, Natural Language and Linguistic Theory . –. RAE and ASALE (). Nueva gramática de la lengua española. Madrid: Espasa Calpe. Ramchand, Gillian (). Verb Meaning and the Lexicon. A First Phase Syntax. Cambridge: Cambridge University Press. Reinhart, Tanya (). ‘Interface Economy: Focus and Markedness’, in C. Wilder, H.-M. Gaertner, and M. Bierwisch (eds), The Role of Economy Principles in Linguistic Theory. Berlin: Akademie Verlag, –. Reklaitis, Janine K. (). ‘Reduction of Case Markers in Lithuanian: Data for Discussion’, in E. Closs Traugott, R. Labrum, and S. C. Shepherd (eds), Papers from the Fourth International Conference on Historical Linguistics. Amsterdam: John Benjamins, –. Ricós, Amparo (). Las construcciones pasivas en español medieval. Valencia: University of València. Ricós, Amparo (). ‘La pasiva con se agentiva en los textos jurídico-administrativos: Su incidencia pragmática’, Hesperia: Anuario de filología hispánica : –. Riemsdijk, Henk van (). ‘Silent Nouns and the Spurious Indefinite Article in Dutch’, in M. Vulchanova and T. Åfarli (eds), Grammar and Beyond. Essays in Honour of Lars Hellan. Oslo: Novus Press, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Riemsdijk, Henk van (). ‘The Absent, the Silent, and the Audible: Some Thoughts on the Morphology of Silent Verbs’, in E. Torrego (ed.), Of Grammar, Words, and Verses. In Honor of Carlos Piera. Amsterdam: John Benjamins, –. Rini, Joel (). Motives for Linguistic Change in the Formation of the Spanish Object Pronouns. Newark, DE: Juan de la Cuesta. Rivero, María Luisa (). ‘Los verbos psicológicos con experimentante dativo en español y el cambio histórico’, in M. Luján and M. Groppi (eds), Cuestiones gramaticales del español, últimos avances. Santiago de Chile: Cuadernos de la ALFAL, –. Rizzi, Luigi (). Relativized Minimality. Cambridge, MA: MIT Press. Rizzi, Luigi (). ‘The Fine Structure of the Left Periphery’, in L. Haegeman (ed.), Elements of Grammar. Handbook in Generative Syntax. Dordrecht: Kluwer, –. Rizzi, Luigi (). ‘Relativized Minimality Effects’, in M. Baltin and C. Collins (eds), The Handbook of Contemporary Syntactic Theory. Oxford: Blackwell, –. Rizzi, Luigi (). ‘On the Form of Chains’, in L. Cheng and N. Corver (eds), Wh-Movement. Moving on. Cambridge, MA: MIT Press, –. Rizzi, Luigi (). ’On the Form of Chains’, in U. Shlonsky (ed.), Beyond Functional Sequence. The Cartography of Syntactic Structures, vol. . Oxford: Oxford University Press, –. Roberge, Yves (). The Syntactic Recoverability of Null Arguments. Kingston and Montreal: McGill–Queen’s University Press. Roberts, Ian (). ‘The Nature of Subject Clitics in Franco-Provençal Valdotain’, in H. van Riemsdijk and L. Rizzi (eds), Clitics and their Hosts. Geneva/Tilburg: ESF-Eurotype, –. Roberts, Ian (). Verbs and Diachronic Syntax. A Comparative History of English and French. Dordrecht: Kluwer. Roberts, Ian (). Diachronic Syntax. Oxford: Oxford University Press. Roberts, Ian (a). ‘A Deletion Analysis of Null Subjects’, in T. Biberauer, A. Holmberg, I. Roberts, and M. Sheehan (eds), Parametric Variation. Null Subjects in Minimalist Theory. Cambridge: Cambridge University Press, –. Roberts, Ian (b). ‘Varieties of French and the Null Subject Parameter’, in T. Biberauer, A. Holmberg, I. Roberts, and M. Sheehan (eds), Parametric Variation. Null Subjects in Minimalist Theory. Cambridge: Cambridge University Press, –. Roberts, Ian (). ‘Macroparameters and Minimalism: A Program of Comparative Research’, in C. Galves, S. Cyrino, R. Lopes, F. Sandalo, and J. Avelar (eds), Parameter Theory and Linguistic Change. Oxford: Oxford University Press, –. Roberts, Ian and Anna Roussou (). ‘A Formal Approach to Grammaticalization’, Linguistics : –. Roberts, Ian and Anna Roussou (). Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rochemont, Michael (). A Theory of Stylistic Rules in English. Doctoral dissertation, University of Massachusetts at Amherst. Roehrs, Dorian (). The Morpho-syntax of the Germanic Noun Phrase. Determiners Move into the Determiner Phrase. Doctoral dissertation, Indiana University. Roehrs, Dorian and Christopher Sapp (). Quantifying Expressions in the History of German. Amsterdam: John Benjamins.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Rohrbacher, Bernhard (). Morphology-driven Syntax. A Theory of V-to-I Raising and Pro-drop. Amsterdam: John Benjamins. Rooryck, Johan (). A Compositional Analysis of French Negation, Ms., Leiden University. Rooth, Mats (). Association with Focus. Doctoral dissertation, University of Massachusetts at Amherst. Rooth, Mats (). ‘A Theory of Focus Interpretation’, Natural Language Semantics : –. Rosén, Victoria and Kaja Borthen (). ‘Norwegian Bare Singulars Revisited’, Bergen Language and Linguistics Studies : –. Rowlett, Paul (). Sentential Negation in French. Oxford: Oxford University Press. Sandfeld Jensen, Kristian (). ‘Notes sur les calques linguistiques’, in Festschrift Vilhelm Thomsen zur Vollendung des siebzigsten Lebensjahres am . Januar , dargebracht von Freunden und Schülern. Leipzig: Otto Harrassowitz, –. Sanfelici, Emanuela and Cecilia Poletto (). ‘Two Relativization Strategies in Old Italian’, Unpublished manuscript, University of Padua. Sapir, Edward (). Language. New York: Harcourt, Brace. Sauerland, Uli (). ‘Unpronounced Heads in Relative Clauses’, in K. Schwabe and S. Winkler (eds), The Interfaces. Deriving and Interpreting Omitted Structures. Amsterdam: John Benjamins, –. Sauerland, Uli (). ‘Disjunction and Implicatures: Some Notes on Recent Developments’, in C. Lee, F. Kiefer, and M. Krifka (eds), Contrastiveness in Information Structure, Alternatives and Scalar Implicatures. Dordrecht: Springer, –. Sauerland, Uli and Kazuko Yatsushiro (). ‘A Silent Noun in Partitives’, in K. Moulton and M. Wolf (eds), Proceedings of the North East Linguistic Society (NELS) . Amherst, MA: Graduate Linguistic Student Association, vol. , –. Schifano, Norma (). ‘The Paradigmatic Instantiation of TAM: A Novel Approach to Romance Verb-Movement’, in E. Aboh, J. Schaeffer, and P. Sleeman (eds), Romance Languages and Linguistic Theory . Amsterdam: John Benjamins, –. Schifano, Norma (). Verb Movement in Romance. A Comparative Perspective. Oxford: Oxford University Press. Schlenker, Philippe (). ‘The Semantics–Pragmatics Interface’, in M. Aloni and P. Dekker (eds), Cambridge Handbook of Formal Semantics. Cambridge: Cambridge University Press, –. Schlesinger, Michel (). Satzlehre der aramäischen Sprache des babylonischen Talmuds. Leipzig: Asia Major. Schmitt, Cristina and Alan Munn (). ‘The Syntax and Semantics of Bare Arguments’, Linguistic Variation Yearbook : –. Schroten, Jan (). ‘Clitic Doubling in Spanish: Agreement of the Third Kind’, in M. Everaert, T. Lentz, H. de Mulder, Ø. Nilsen, and A. Zondervan (eds), The Linguistics Enterprise. From Knowledge of Language to Knowledge in Linguistics. Amsterdam: John Benjamins, –. Schwarz, Bernhard and Rajesh Bhatt (). ‘Light Negation and Polarity’, in R. Zanuttini, H. Campos, E. Herburger, and P. Portner (eds), Crosslinguistic Research in Syntax and Semantics. Negation, Tense and Clausal Architecture. Washington, DC: Georgetown University Press, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Schweizer, Bruno (/). Zimbrische Gesamtgrammatik. Vergleichende Darstellung der zimbrischen Dialekte. Edited by James R. Dow. Stuttgart: Steiner. Schwentner, Hans (). Die Wortfolge im Litauischen. Heidelberg: Carl Winter. Selkirk, Elisabeth (). ‘Some Remarks on Noun Phrase Structure’, in P. W. Culicover, T. Wasow, and A. Akmajian (eds), Formal Syntax. New York: Academic Press, –. Seip, Didrik Arup (). Den legendariske olavssaga og Fagrskinna. Oslo: Dybwad. Sepúlveda Barrios, Félix (). La voz pasiva en el español del siglo XVII. Contribución a su estudio. Madrid: Gredos. Shimoyama, Junko (). ‘Indeterminate Phrase Quantification in Japanese’, Natural Language Semantics : –. Šimík, Radek (). Modal Existential Wh-Constructions. Doctoral dissertation, University of Groningen. Simonenko, Alexandra, Benoît Crabbé, and Sophie Prévost (forthcoming). ‘Agreement Syncretisation and the Loss of Null Subjects. Quantificational Models for Medieval French’, Language Variation and Change. Available at (accessed  September ). Skrzypek, Dominika (). Grammaticalization of (In)definiteness in Swedish. Doctoral dissertation, Adam Mickiewicz University, Poznań. Sokoloff, Michael (). A Dictionary of Jewish Babylonian Aramaic of the Talmudic and Geonic Periods. Ramat-Gan/Baltimore, MD: Bar-Ilan University Press/The Johns Hopkins University Press. Sorace, Antonella (). ‘Pinning down the Concept of “Interface” in Bilingualism’. Linguistic Approaches to Bilingualism : –. Spector Shirtz, Ilona (). The Syntax of Non-verbal Predication in Modern Hebrew. Predicate Nominals, Pseudoclefts and Clefts. Doctoral dissertation, Hebrew University of Jerusalem. Sportiche, Dominique (). ‘French Relative qui’, Linguistic Inquiry : –. Sportiche, Dominique (). ‘Neglect (or Doing away with Late Merger and Countercyclicity)’, Unpublished manuscript, University of California, Los Angeles. Available at (accessed  November ). Starke, Michal (). Move Dissolves into Merge. A Theory of Locality. Doctoral dissertation, University of Geneva. Starke, Michal (). ‘On the Inexistence of Specifiers and the Nature of Heads’, in A. Belletti (ed.), Structures and Beyond. Oxford: Oxford University Press, –. Starke, Michal (). ‘Nanosyntax: A Short Primer to a New Approach to Language’, Nordlyd : –. Starke, Michal (). ‘Towards Elegant Parameters: Language Variation Reduces to the Size of Lexically-Stored Trees’, in C. Picallo (ed.), Linguistic Variation in the Minimalist Framework. Oxford: Oxford University Press, –. Starke, Michal (). ‘Resolving (dat = acc) ≠ gen’, Glossa (): .–. Starke, Michal (). ‘Complex Left Branches, Spellout, and Prefixes’, in L. Baunaz, K. De Clercq, L. Haegeman, and E. Lander (eds), Exploring Nanosyntax. Oxford: Oxford University Press, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Steinki, Johannes (). Die Entwicklung der englischen Relativpronomina in spätmittelenglischer und frühneuenglischer Zeit. Breslau: Hermann Eschenhagen. Stendahl, Anders (). Användningen av ‘en’ som indefinit artikel i fornsvenskan. Doctoral dissertation, Åbo Akademi University. Stroh-Wollin, Ulla (). ‘Understanding the Gradual Development of Definiteness Marking: The Case of Swedish’, Working Papers in Scandinavian Syntax : –. Studemund-Halévy, Michael and Susann Fischer (). ‘What Happens When a Language Ceases to Be Used by its Speakers?’ in M. Studemund-Halévy, C. Liebl, and I. Vučina Simović (eds), Sefarad an der Donau. Lengua y literatura de los Sefardíes en tierras de los Habsburgos. Barcelona: Tocino, –. Sundquist, John D. (). Morphosyntactic Change in the History of the Mainland Scandinavian Languages. Doctoral dissertation, Indiana University. Sundquist, John D. (). ‘The Rich Agreement Hypothesis and Early Modern Danish Embedded-clause Word Order’, Nordic Journal of Linguistics : –. Surányi, Balázs (). ‘Quantification and Focus in Negative Concord’, Lingua (): –. Szabolcsi, Anna (). ‘Positive Polarity-Negative Polarity’, Natural Language and Linguistic Theory : –. Szabolcsi, Anna (). ‘What Do Quantifier Particles Do?’ Linguistics and Philosophy : –. Szabolcsi, Anna (a). ‘Additive Presuppositions are Derived through Activating Focus Alternatives’, in A. Cremers, T. van Gessel, and F. Roelofsen (eds), Proceedings of the st Amsterdam Colloquium. University of Amsterdam, –. Szabolcsi, Anna (b). ‘Disjunctive and Conjunctive Particles Meet their Negative Concord Relatives’, talk given at ICHL Workshop on Logical Vocabulary & Logical Change. San Antonio, TX,  August . Handout available at (accessed  November ). Szabolcsi, Anna (). ‘Two Types of Quantifier Particles: Quantifier Phrase Internal vs. Heads on the Clausal Spine’, Glossa (): , –. http://doi.org/./gjgl.. Takita, Kensuke, Nobu Goto, and Yoshiyuki Shibata (). ‘Labeling through Spell-out’, The Linguistic Review : –. Tauli, Valter (). The Structural Tendencies of Languages. Helsinki: Finnish Academy of Science and Letters. Tesnière, Lucien (). Eléments de syntaxe structurale. Paris: Klincksieck. Thorpe, Benjamin (). Anglo-Saxon Chronicle I and II. London: Longman. Thráinsson, Höskuldur (). The Syntax of Icelandic. Cambridge: Cambridge University Press. Tomioka, Satoshi (). ‘The Semantics of Japanese Null Pronouns and its Cross-Linguistic Implications’, in K. Schwabe and S. Winkler (eds), The Interfaces. Deriving and Interpreting Omitted Structures. Amsterdam: John Benjamins, –. Torregrossa, Jacopo (). Encoding Topic, Focus and Contrast. Informational Notions at the Interfaces. Doctoral dissertation, University of Verona. Truswell, Robert (). Locality of Movement and the Individuation of Events. Doctoral dissertation, University College London.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Tsakali, Vina and Elena Anagnostopoulou (). ‘Rethinking the Clitic Doubling Parameter: The Inverse Correlation between Clitic Doubling and Participle Agreement’, in D. Kallulli and L. Tasmowski (eds), Clitic Doubling in the Balkan Languages. Amsterdam: John Benjamins, –. Tvica, Seid (). Agreement and Verb Movement. The Rich Agreement Hypothesis from a Typological Perspective. Doctoral dissertation, University of Amsterdam. Urrutia Cárdenas, Hernán (). ‘Los clitícos de tercera persona en el País Vasco’, CAUCE, Revista e Filología y su Didáctica : –. Vaillant, André (). Grammaire comparée de langues slaves, tome : la syntaxe. Paris: Klincksieck. Vallduví, Enric (). ‘L’oració com a unitat informativa’, in J. Solà, M. R. Lloret, J. Mascara, and M. Pérez Saldanya (eds), Gramàtica del català contemporaní, vol. : Sintaxi. Barcelona: Empúries. –. Vance, Barbara (). Syntactic Change in Medieval French. Verb Second and Null Subjects. Dordrecht: Kluwer. van der Auwera, Johan (). ‘The Jespersen Cycles’, in E. van Gelderen (ed.), Cyclical Change. Amsterdam: John Benjamins, –. van der Auwera, Johan (). ‘On the Diachrony of Negation’, in L. Horn (ed.), The Expression of Negation. Berlin: De Gruyter, –. van der Auwera, Johan and Annemie Neuckermans (). ‘Jespersen’s Cycle and the Interaction of Predicate and Quantifier Negation in Flemish’, in B. Kortmann (ed.), Typology meets Dialectology. Dialect Grammar from a Cross-linguistic Perspective. Berlin: De Gruyter, –. van der Wouden, Ton (). Negative Contexts. Collocation, Polarity and Multiple Negation. London and New York: Routledge. Van de Velde, Freek (). ‘The Emergence of the Determiner in the Dutch NP’, Linguistics : –. Vangsnes, Øystein Alexander (). ‘On Noun Phrase Architecture, Referentiality, and Article Systems’, Studia Linguistica : –. Vannebo, Kjell Ivar (). ‘Ein sterk ein prylert’, Maal og Minne : –. Vasmer, Max (). Russisches etymologisches Wörterbuch. Bd. . Heidelberg: Winter. Vedovato, Diana (). ‘Weak Pronouns in Italian: Instances of a Broken Cycle?’, in E. van Gelderen (ed.), Cyclical Change. Amsterdam: John Benjamins, –. Vega Vilanova, Jorge (). ‘Towards a Subject Cartography in Catalan: The Example of Oblique Subjects’, in C. Schlaak and A. Hennemann (eds), Korpuslinguistische Untersuchungen. Berlin: Frank and Timme, –. Vega Vilanova, Jorge (). Catalan Participle Agreement. On the Interaction between Syntactic Features and Language Change. Doctoral dissertation, University of Hamburg. Veselinova, Ljuba (). ‘Negative Existentials: A Cross-Linguistic Study’, Rivista di Linguistica : –. Vikner, Sten (). ‘V-to-I Movement and Inflection for Person in All Tenses’, in L. Haegeman (ed.), The New Comparative Syntax. London: Longman, –. Vincent, Nigel (). ‘The Emergence of the D-system in Romance’, in A. van Kemenade and N. Vincent (eds), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, –.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi

References



Vovin, Alexander (). A Reference Grammar of Classical Japanese Prose. London: Routledge. Vovin, Alexander (). Man’yōshū, Book . A New English Translation Containing the Original Text, Kana Transliteration, Romanization, Glossing and Commentary. Leiden: Brill. Wagener, Terje (). The History of Nordic Relative Clauses. Berlin: De Gruyter. Walker, Dean A. (). ‘The Semitic Negative with Special Reference to the Negative in Hebrew’, The American Journal of Semitic Languages and Literatures : –. Wartburg, Walther von (). Einführung in die Problematik und Methodik der Sprachwissenschaft (third edition). Tübingen: Niemeyer. Weinreich, Uriel (). Languages in Contact. Findings and Problems. New York: De Gruyter. Weir, Helen (). ‘Footprints of Yesterday’s Syntax: Diachronic Development of Certain Verb Prefixes in an OSV Language (Nadëb)’, Lingua : –. Weiß, Helmut (). ‘A Question of Relevance: Some Remarks on Standard Languages’, in M. Penke and A. Rosenbach (eds), What Counts as Evidence in Linguistics. The Case of Innateness. Amsterdam: John Benjamins, –. Whitman, John (). ‘否定構造と歴史的変化— 主要部と否定極性表 現を中心に [Hitei kōzō to rekishi teki henka — shuyōbu to hitei kyokusei hyōgen o chūsin ni]’, in Y. Katō (ed.), 否定と言語理論 [Hitei to gengo riron]. Tokyo: Kaitakusha. Ms, –. Willis, David (). ‘Specifier-to-Head Reanalyses in the Complementizer Domain: Evidence from Welsh’, Transactions of the Philological Society : –. Willis, David (). ‘Negative Polarity and the Quantifier Cycle: Comparative Diachronic Perspectives from European Languages’, in P. Larrivée and R. P. Ingham (eds), The Evolution of Negation: Beyond the Jespersen Cycle. Berlin: De Gruyter, –. Willis, David (). ‘A Minimalist Approach to Jespersen’s Cycle in Welsh’, in Dianne Jonas, John Whitman, and Andrew Garrett (eds), Grammatical Change: Origins, Nature, Outcomes. Oxford: Oxford University Press, –. Willis, David, Christopher Lucas, and Anne Breitbarth (eds) (a). The Development of Negation in the Languages of Europe and the Mediterranean. Volume : Case Studies. Oxford: Oxford University Press. Willis, David, Christopher Lucas, and Anne Breitbarth (b). ‘Comparing Diachronies of Negation’, in D. Willis, C. Lucas, and A. Breitbarth (eds), The Development of Negation in the Languages of Europe and the Mediterranean, vol. Case Studies. Oxford: Oxford University Press, –. Wolfe, Sam (). Microvariation in Medieval Romance Syntax. A Comparative Study. Doctoral dissertation, Cambridge University. Wolfsgruber, Anne (). ‘On, se and Related Valency Alternations in Medieval French’, Lingvisticæ Investigationes : –. Wood, Johanna L. (). ‘Is There a DP in Old English?’ in J. Salmons and S. Dubenion-Smith (eds), Selected Papers from the th International Conference on Historical Linguistics, Madison, Wisconsin,  July– August . Amsterdam: John Benjamins, –. Yang, Charles (). ‘Internal and External Forces in Language Change’, Language Variation and Change : –. Zagona, Karen (). The Syntax of Spanish. Cambridge: Cambridge University Press. Zaliznjak, Andrej (). Drevnenovgorodskij dialekt. Moscow: Jazyki slavjanskoj kul’tury.

OUP CORRECTED PROOF – FINAL, 13/8/2019, SPi



References

Zanuttini, Raffaella (). ‘On the Relevance of Tense for Sentential Negation’, in A. Belletti and L. Rizzi (eds), Parameters and Functional Heads. Oxford: Oxford University Press, –. Zanuttini, Raffaella (). Negation and Clausal Structure. Oxford: Oxford University Press. Zdrojewski, Pablo and Liliana Sánchez (). ‘Variation in Accusative Clitic Doubling across Three Spanish Dialects’, Lingua : –. Zeijlstra, Hedde (). Sentential Negation and Negative Concord. Doctoral dissertation, University of Amsterdam. Zeijlstra, Hedde (). ‘Negative Concord is Syntactic Agreement’. Ms., Amsterdam: University of Amsterdam. Zeijlstra, Hedde (). ‘On French Negation’, in I. Kwon, H. Pritchett, and J. Spence (eds), Proceedings of the th Annual Meeting of the Berkeley Linguistics Society. Berkeley, CA: Berkeley Linguistics Society, –. Zimmermann, Ilse (). ‘On the Syntax and Semantics of kakoj and čto za in Russian’, Journal of Slavic Linguistics : –. Zribi-Hertz, Anne (). ‘La syntaxe des clitiques nominatifs en français standard et en français avancé’, Travaux de linguistique et de philologie : –. Zwart, C. Jan-Wouter (). ‘ “Shortest Move” vs. “Fewest Steps” ’, in W. Abraham, S. David Epstein, H. Thráinsson, and C. J.-W. Zwart (eds), Minimal Ideas. Syntactic Studies in the Minimalist Framework. Amsterdam: John Benjamins, –. Zwarts, Frans (). ‘Three Types of Polarity’, in F. Hamm and E. W. Hinrichs (eds), Plurality and Quantification. Dordrecht: Kluwer, –. Zwicky, Arnold M. (). On Clitics. Bloomington, IN: Indiana University Linguistics Club. Zwicky, Arnold and Geoffrey Pullum (). ‘Cliticization vs. Inflection: English n’t’, Language : –.

OUP CORRECTED PROOF – FINAL, 20/8/2019, SPi

Index A’-movement ,  A’-position , , – accessibility –, , , , – acquisition , –,  additivity –,  adjunct , , , , – adverb –, , –, –, , ,  affix , , , , , – agentive initiator  agreement –, –, , , , ,  all  allosemy – also  alternative activation  alternative feature scalar alternative (σ) , , , – subdomain alternative (δ) , –, ,  anti-exhaustivity  anti-licensing  anti-locality  Arabic ,  Aramaic ,  argument , , , ,  argument structure , – demotion of external argument , , , – article –,  definite article –, –, –, , , –, , , , ,  indefinite article , –, –, –,  personal article – Aspect –, , , ,  Athabaskan –,  Balto-Slavic , ,  bare singular noun () – Basque , – biclausal structure –, , ,  binominal , , –, ,  blocking economy  case , –, , – accusative case , –, ,  case inert 

case-transparent  dative case –,  genitive case –, –, , –, – nominative case , ,  Catalan Old Catalan – Chinese  Ancient Chinese  choice principle  Cimbrian – Classical Japanese  Classical Sanskrit  classifier –, , ,  cleft , –, –,  cleft focus – clitic , –, –, –, , , , , , , , ,  clitic agreement/agreement clitic , –, ,  clitic doubling (CLD) – Clitic Doubling Cycle  enclitic –, , –, ,  object clitic , , –, , , , – Person-Case Constraint () ,  proclitic , , , – subject clitic , , – copula – copula clause  counterfactual conditional sentences – cycle –, , , , , , , , –, , , , , , , – broken cycle  Clitic Doubling Cycle  Croft’s cycle  Jespersen Cycle –, , ,  object cycle – quantificational cycle  relative cycle – subject cycle – Czech  Danish –, , – dative experiencer , – argument structure , –, –, , 

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Index

dative experiencer (cont.) diachronic classification of dative experiencer  stativity of dative experiencer , , , – unaccusative constructions with a dative experiencer – definiteness , , , , , , – definiteness scale , ,  deictic  demonstrative –,  Dëne Sųłiné/Chipewyan  determiner phrase () , ,  diachronic typology  diachrony – disjunction ambiguous disjunction  discrete disjunction ,  distributive universal , , – distributivity , – economy  see also minimal search Egyptian ,  English , , –, –, , , , , –, , –, , , , , , –, ,  Middle English  Old English  event structure , –, ,  exhaustification –, – exhaustifier () –,  iterative exhaustification  recursive exhaustification , , –,  existential existential meaning , , ,  existential quantifier , , , ,  existential sentences  extraction ,  Faroese , ,  Old Faroese  feature –, , , , , , , –, , ,  featural change see parametric change feature conservation , –, , – feature economy ,  feature-sharing , , ,  interpretable feature – phi-feature –, –, –, ,  uninterpretable feature –, , 

focus , , , , –,  contrastive  FocP , , , –, – focus particle  focus sensitivity  functional sequence () , , –, , , – Free Choice Item ,  free relative clause – freedom of choice implicature ,  French –, , , , , –, –, , , , , –,  colloquial French () , , , –, –, , –, ,  le bon usage French () –, –, –, –, , – Swiss French ,  gender –, , , , , , , , ,  German –, , , –, –, ,  Early New High German  Low German – Old High German  Swiss German  Germanic –, , , ,  Gothic , ,  grammaticalization , –, , –, –, , –, –, , , –, –, , , , , , , , ,  Greek  Ancient Greek  Homeric Greek  head –, – Head Preference Principle () ,  Hittite  honorific titles , – implicature – indefinite –,  n-word –, –, –, – Indo-European , ,  Indo-Iranian  Ṛgvedic Sanskrit  interface , , ,  Iranian  Iroquoian  island  see also extraction Italian  Modern Italian –

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

Index Old Aretino ,  Old Florentine see Old Italian Old Italian –, , , – Old Paduan  Old Sienese –,  Old Tuscan , – Old Venetian  Old Veronese  Sicilian Mussomeli – Japanese , ,  Classical Japanese – Old Japanese , , – Japonic () , – Jewish Babylonian Aramaic () – Kaska  Korean  *kwe () –,  labelling  labelling algorithm , –, , ,  labelling paradox , , , – Latin , , , , –,  Italic ,  left periphery , –,  low left periphery – Lithuanian –, –,  Man’yōshū ()  Maya  minimal search ,  mo , – monoclausal structure –, , ,  μ –, – additive function of μ , – conjunctive particle –, ,  distributive function of μ , – Mood , , –,  Nadëb  Nahuatl ,  nanosyntax –, , – Navajo  negation constituent negation  double negation –, –, , , , ,  external negation – n-word –, –, –, – negation as predicate denial  negative concord –, – negative polarity , , , 



particle ,  predicate term negation  preposed negative constituent – sentential negation –, , –, –, – nominal –, , , –,  Norwegian Middle Norwegian – Modern Norwegian –, – Old Norwegian –,  Regional Northern Norwegian (ReNN) , – number –, –, , , , , , , –, , , , ,  numeral –, ,  Old Irish  only  operator  Övdalian  pair-merge , – parameter –, , , –, ,  parametric change ,  parameter hierarchy –,  particle , , , ,  quantificational ,  superparticle , –, ,  partitive partitive constraint  partitive genitive –, – Passamaquoddy  passive periphrastic passive , –, , , ,  se passives , –, , ,  perfective –,  Persian  person –, , –, , , –, ,  personal article see article Person-Case Constraint () ,  phi-feature –, –, ,  phrase (as opposed to syntactic head) –, , , , –, –,  polarity sensitivity , , –, –, ,  negative polarity items () , –, , –, –, – polarity-sensitive items () , , , – preposition –, , , , – presupposition –, –,  presuppositional meaning , , 

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi



Index

preverb  problems of projection (PP) –, ,  pronoun –, , –, –, –,  relative pronoun –, , ,  strong, weak, and clitic –, – proper names – Proto-Indo-European () ,  psych verbs – Q-feature ,  quantificational meaning –,  quantificational split, quantifier split –, – quantifier shift , ,  Quechua – quirky subjects  reanalysis –, , –, –, , , , , –, –, , , , –, , –, , , , , ,  relative clause –, , , –, – renewal , , , – reversal of quantifier scales, scalar shift  rhetorical questions –,  Rich Agreement Hypothesis () – Romance , , –, –, –, –, – Italo-Romance – Old Romance –, , , –, –, –, – Root Phrase () –,  Salish  scalar implicature () , –, , ,  calculation of scalar implicature  scalar subcategorization – Semitic , ,  set-merge – Sino-Tibetan  Slavic , , , , ,  Old Church Slavonic – Slavonic see Slavic SOV (word order) –

specificity –, , , , , –, – specifier –, , – spell-out algorithm , , –,  structural deficiency , – Superset Principle , –, , –, –, –,  SVO (word order) –,  Swedish , –, – Old Swedish  syncretism , , , –, ,  Syriac ,  Tense –, , , , , – Tibeto-Burman  Tobler-Mussafia law – Tocharian B  Tohono O’odham  topic –,  topic prominence –,  topicalization  univerbation , , ,  morphological univerbation  phonological univerbation  universal quantifier , , , –,  Uto-Aztecan ,  verb movement –, –, , – loss of verb movement –, –, – V-to-Arg/I movement , – V-to-C movement – V-to-I movement –, , – Verb Second , , , –,  VOS (word order)  Walbiri  was-für construction – wh-pronoun , , –, –,  construction ,  wh-constituent see wh-pronoun wh-element see wh-pronoun wh-term see wh-pronoun Zoque 

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

                                                  Adam Ledgeway and Ian Roberts, University of Cambridge   Cynthia L. Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge   From Latin to Romance Morphosyntactic Typology and Change Adam Ledgeway  Parameter Theory and Linguistic Change Edited by Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar  Case in Semitic Roles, Relations, and Reconstruction Rebecca Hasselbach  The Boundaries of Pure Morphology Diachronic and Synchronic Perspectives Edited by Silvio Cruschina, Martin Maiden, and John Charles Smith  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  Constructionalization and Constructional Changes Elizabeth Traugott and Graeme Trousdale  Word Order in Old Italian Cecilia Poletto  Diachrony and Dialects Grammatical Change in the Dialects of Italy Edited by Paola Benincà, Adam Ledgeway, and Nigel Vincent  Discourse and Pragmatic Markers from Latin to the Romance Languages Edited by Chiara Ghezzi and Piera Molinelli  Vowel Length from Latin to Romance Michele Loporcaro

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

 The Evolution of Functional Left Peripheries in Hungarian Syntax Edited by Katalin É. Kiss  Syntactic Reconstruction and Proto-Germanic George Walkden  The History of Low German Negation Anne Breitbarth  Arabic Indefinites, Interrogatives, and Negators A Linguistic History of Western Dialects David Wilmsen  Syntax over Time Lexical, Morphological, and Information-Structural Interactions Edited by Theresa Biberauer and George Walkden  Syllable and Segment in Latin Ranjan Sen  Participles in Rigvedic Sanskrit The Syntax and Semantics of Adjectival Verb Forms John J. Lowe  Verb Movement and Clause Structure in Old Romanian Virginia Hill and Gabriela Alboiu  The Syntax of Old Romanian Edited by Gabriela Pană Dindelegan  Grammaticalization and the Rise of Configurationality in Indo-Aryan Uta Reinöhl  The Rise and Fall of Ergativity in Aramaic Cycles of Alignment Change Eleanor Coghill  Portuguese Relative Clauses in Synchrony and Diachrony Adriana Cardoso  Micro-change and Macro-change in Diachronic Syntax Edited by Eric Mathieu and Robert Truswell  The Development of Latin Clause Structure A Study of the Extended Verb Phrase Lieven Danckaert

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

 Transitive Nouns and Adjectives Evidence from Early Indo-Aryan John. J. Lowe  Quantitative Historical Linguistics A Corpus Framework Gard B. Jenset and Barbara McGillivray  Gender from Latin to Romance History, Geography, Typology Michele Loporcaro  Clause Structure and Word Order in the History of German Edited by Agnes Jäger, Gisella Ferraresi, and Helmut Weiß  Word Order Change Edited by Ana Maria Martins and Adriana Cardoso  Arabic Historical Dialectology Linguistic and Sociolinguistic Approaches Edited by Clive Holes  Grammaticalization from a Typological Perspective Edited by Heiko Narrog and Bernd Heine  Negation and Nonveridicality in the History of Greek Katerina Chatzopoulou  Indefinites between Latin and Romance Chiara Gianollo  Verb Second in Medieval Romance Sam Wolfe  Referential Null Subjects in Early English Kristian A. Rusten  Word Order and Parameter Change in Romanian A Comparative Romance Perspective Alexandru Nicolae  Cycles in Language Change Edited by Miriam Bouzouita, Anne Breitbarth, Lieven Danckaert, and Elisabeth Witzenhausen

OUP CORRECTED PROOF – FINAL, 12/8/2019, SPi

  Dative External Possessors in Early English Cynthia L. Allen Redevelopment of Case Systems in Indo-Aryan Miriam Butt Classical Portuguese Grammar and History Charlotte Galves, Aroldo de Andrade, Christiane Namiuti, and Maria Clara Paixão de Sousa Morphological Borrowing Francesco Gardani Nominal Expressions and Language Change From Early Latin to Modern Romance Giuliana Giusti Syntactic Features and the Limits of Syntactic Change Edited by Jóhannes Gísli Jónsson and Thórhallur Eythórsson A Study in Grammatical Change The Modern Greek Weak Subject Pronoun τος and its Implications for Language Change and Structure Brian D. Joseph Reconstructing Pre-Islamic Arabic Dialects Alexander Magidow 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 Variation and Change in Gallo-Romance Grammar Edited by Sam Wolfe and Martin Maiden Palatal Sound Change in the Romance Languages Diachronic and Synchronic Perspectives André Zampaulo