Indefinites between Latin and Romance (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated] 9780198812661, 0198812663

This book investigates the syntactic and semantic development of a selection of indefinite pronouns and determiners (suc

130 31 4MB

English Pages 352 [342] Year 2019

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
Cover
Indefinites between Latin and Romance
Copyright
Contents
Series preface
Preface
Acknowledgments
List of abbreviations
1: The grammar of indefinitesFunctions, variation, and change
1.1 Aim and scope of this work
1.2 Overview of the chapters
1.3 Indefinites: working definitions
1.4 The functional space of indefinites
1.4.1 Haspelmath’s (1997) semantic map
1.4.1.1 Functions
1.4.1.2 Series
1.4.1.3 Diachronic generalizations
1.4.2 Semantic map for Latin indefinites
1.4.3 From Latin to Romance: the case studies
1.5 Dimensions of variation and change
1.5.1 A research program
1.5.2 The classes investigated in this work
1.5.2.1 Epistemic indefinites
1.5.2.2 Indefinites in the scope of negation
1.5.3 Constraints on domains of quantification
1.5.4 Indefinites and the systematicity of semantic change
1.5.4.1 Cycles and directionality
1.5.4.2 Mechanisms of semantic change
1.6 Corpora, methods, and periodization
1.6.1 Corpora and methods for data collection
1.6.2 Periodization for Latin
2: Specific and epistemic indefinites in Latin
2.1 ntroduction
2.1.1 Topics and aims
2.1.2 Roadmap
2.2 The contexts of quidam, aliquis, and quis
2.3 Specificity and Latin indefinites
2.3.1 Notions of specificity
2.3.2 Is aliquis really specific?
2.3.3 Taking stock
2.4 Quidam
2.4.1 Uses of quidam in Classical Latin
2.4.2 Quidam in Late Latin
2.4.2.1 Contexts of use
2.4.2.2 Competition with certus and unus
2.5 Latin aliquis and the class of epistemic indefinites
2.5.1 Epistemic indefinites
2.5.2 The anti-singleton constraint
2.5.3 Dependent licensing
2.5.4 The ignorance implicature
2.5.5 Epistemic indefinites and negation
2.5.6 Crosslinguistic variation with epistemic indefinites
2.6 Classical Latin aliquis
2.6.1 Etymological remarks
2.6.2 Prototypical ‘realis’ contexts
2.6.3 Generic environments
2.6.4 Modalized environments
2.6.5 Negation and other downward-entailing contexts
2.6.6 Protasis of conditional
2.6.7 Scalar uses
2.6.8 Interim summary: aliquis as an epistemic indefinite
2.7 Aliquis in Late Latin
2.7.1 Aliquis and negation in Late Latin
2.7.2 Combination with unus
2.8 Conclusions and outlook
3: Aliquis from Latin to Romance
3.1 Introduction
3.1.1 Topics and aims
3.1.2 Roadmap
3.2 The Quantifier Cycle
3.3 The Romance continuations of aliquis
3.3.1 Latin aliquis and the Quantifier Cycle
3.3.2 The continuations of aliquis: a synchronic overview
3.3.3 Spanish
3.3.4 Portuguese
3.3.5 Catalan
3.3.6 French
3.3.7 Italian
3.3.8 Asymmetry and divergence between singular and plural
3.4 Diachronic developments
3.4.1 Overview
3.4.2 A closer look at Old French
3.4.3 A closer look at Old Italian
3.4.4 Inversion
3.5 Inversion as focalization in the Determiner Phrase
3.5.1 A distributional generalization
3.5.2 The syntax of DP-internal inversion
3.6 The semantic role of focus
3.6.1 Scalar focus and NPIs
3.6.2 Loss of emphasis: a comparison with Jespersen’s Cycle
3.7 Conclusions
4: Indefinites and negation in the history of Latin
4.1 Introduction
4.1.1 Topics and aims
4.1.2 Roadmap
4.2 The Latin negation system in light of Romance
4.2.1 Position of negative markers
4.2.2 Lexical renewal in indefinites interacting with negation
4.2.3 Research questions and hypothesis
4.3 The Classical Latin negation system: a first overview
4.3.1 A Double Negation system
4.3.2 Negative markers
4.3.3 Negative indefinites
4.4 Negation systems: theoretical backgrounds
4.4.1 The interpretation of negation
4.4.1.1 Sentential negation and constituent negation
4.4.1.2 Negation and focus
4.4.1.3 Denials
4.4.3 Types of negation systems
4.4.2.1 A feature-based typology
4.4.2.2 The syntactic status of negative markers
4.4.2.3 Indefinites interacting with negation
4.5 The analysis adopted here
4.5.1 Negative indefinites
4.5.2 n-words
4.5.3 The negative marker
4.6 Latin negation system and Jespersen’s Cycle
4.6.1 Jespersen’s Cycle
4.6.1.1 Typological and diachronic predictions
4.6.1.2 The triggers
4.6.1.3 A generative model of Jespersen’s Cycle
4.6.2 Grammaticalization of the Latin negative marker n¯on
4.6.2.1 Traces of n˘e (Stage I of Jespersen’s Cycle)
4.6.2.2 The rise of noenum (Stage II of Jespersen’s Cycle)
4.6.3 The syntax of n¯on: position in the clause
4.6.4 The phrasal status of Classical Latin n¯on (Jespersen’s Cycle, Stage III)
4.6.4.1 Syntactic autonomy
4.6.4.2 ‘Why no(t)?’ test
4.6.5 Emphatic negation
4.6.6 Interim summary
4.7 The negation system in Late Latin
4.7.1 Late Latin: a ‘concealed non-strict NC language’?
4.7.2 The development of n¯on in Late Latin: a new Stage I of Jespersen’s Cycle?
4.7.3 OV order with negative indefinites
4.7.4 Analysis
4.7.4.1 Object movement with negative indefinites
4.7.4.2 The activation ofNegP and its consequences for indefinites
4.7.4.3 Implications for theoretical models
4.8 Conclusions
5: Developments in negative and polarity-sensitive contexts from Latin to Romance
5.1 Introduction
5.1.1 Topics and aims
5.1.2 Roadmap
5.2 The situation in Early Romance
5.2.1 Diachronic puzzles with Romance n-words
5.2.2 Old French
5.2.3 Old Italian
5.2.4 Conclusions
5.3 The grammaticalization of nec-words
5.3.1 The issue
5.3.2 Etymological remarks
5.3.2.1 The negative particle
5.3.2.2 The inanimate pronoun
5.4 Syntax and semantics of Latin n˘ec
5.4.1 The form
5.4.2 Functions from Classical to Late Latin
5.4.2.1 Discourse-structuring particle
5.4.2.2 Correlative particle
5.4.2.3 Stand-alone focus particle
5.4.3 Semantic analysis
5.4.3.1 Basic ingredients and scope relations
5.4.2.3 Correlative and non-correlative focus-sensitive uses
5.4.3.3 The scalar reading
5.4.4 Syntactic analysis
5.4.4.1 The internal structure of n˘ec
5.4.4.2 Comparison with ne…quidem
5.4.5 Conclusions
5.5 From nec-words to n-words
5.5.1 The internal syntax of nec-words
5.5.2 Focus particles and polarity sensitivity
5.5.3 N˘ec unus
5.5.4 N˘ec ipse unus
5.5.5 Loss of emphasis with nec-words
5.5.6 Summary and outlook
5.6 Redundant uses of n˘ec and the rise of Negative Concord
5.6.1 The redundant uses
5.6.2 The analysis
5.6.2.1 Focus Concord
5.6.2.2 From Focus Concord to Negative Concord
5.6.2.3 A note on redundancy with correlative n˘ec
5.6.2.4 Summary and outlook
5.7 Concord and polarity-sensitive uses: a unification?
5.7.1 NPIs or n-words?
5.7.2 Optional strict Negative Concord in Old Italian
5.8 Conclusions
6: Conclusions
References
Index
Recommend Papers

Indefinites between Latin and Romance (Oxford Studies in Diachronic and Historical Linguistics) [Illustrated]
 9780198812661, 0198812663

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

OUP CORRECTED PROOF – FINAL, //, SPi

Indefinites between Latin and Romance

OUP CORRECTED PROOF – FINAL, //, SPi

OXFORD STUDIES IN DIACHRONIC AND HISTORICAL LINGUISTICS general editors Adam Ledgeway and Ian Roberts, University of Cambridge advisory editors Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge recently published in the series  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 For a complete list of titles published and in preparation for the series, see pp. –

OUP CORRECTED PROOF – FINAL, //, SPi

Indefinites between Latin and Romance C H IAR A G IA NOLLO

1

OUP CORRECTED PROOF – FINAL, //, SPi

3

Great Clarendon Street, Oxford, OX DP, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © Chiara Gianollo  The moral rights of the author have been asserted First Edition published in  Impression:  All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press  Madison Avenue, New York, NY , United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number:  ISBN –––– Printed and bound by CPI Group (UK) Ltd, Croydon, CR YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED PROOF – FINAL, //, SPi

Contents Series preface Preface Acknowledgments List of abbreviations  The grammar of indefinites: Functions, variation, and change . . . .

Aim and scope of this work Overview of the chapters Indefinites: working definitions The functional space of indefinites .. Haspelmath’s () semantic map .. Semantic map for Latin indefinites .. From Latin to Romance: the case studies . Dimensions of variation and change .. A research program .. The classes investigated in this work .. Constraints on domains of quantification .. Indefinites and the systematicity of semantic change . Corpora, methods, and periodization .. Corpora and methods for data collection .. Periodization for Latin  Specific and epistemic indefinites in Latin . Introduction .. Topics and aims .. Roadmap . The contexts of quidam, aliquis, and quis . Specificity and Latin indefinites .. Notions of specificity .. Is aliquis really specific? .. Taking stock . Quidam .. Uses of quidam in Classical Latin .. Quidam in Late Latin . Latin aliquis and the class of epistemic indefinites .. Epistemic indefinites .. The anti-singleton constraint .. Dependent licensing .. The ignorance implicature .. Epistemic indefinites and negation .. Crosslinguistic variation with epistemic indefinites

ix xi xiii xv                                   

OUP CORRECTED PROOF – FINAL, //, SPi

vi

Contents . Classical Latin aliquis .. Etymological remarks .. Prototypical ‘realis’ contexts .. Generic environments .. Modalized environments .. Negation and other downward-entailing contexts .. Protasis of conditional .. Scalar uses .. Interim summary: aliquis as an epistemic indefinite . Aliquis in Late Latin .. Aliquis and negation in Late Latin .. Combination with unus . Conclusions and outlook

 Aliquis from Latin to Romance . Introduction .. Topics and aims .. Roadmap . The Quantifier Cycle . The Romance continuations of aliquis .. Latin aliquis and the Quantifier Cycle .. The continuations of aliquis: a synchronic overview .. Spanish .. Portuguese .. Catalan .. French .. Italian .. Asymmetry and divergence between singular and plural . Diachronic developments .. Overview .. A closer look at Old French .. A closer look at Old Italian .. Inversion . Inversion as focalization in the Determiner Phrase .. A distributional generalization .. The syntax of DP-internal inversion . The semantic role of focus .. Scalar focus and NPIs .. Loss of emphasis: a comparison with Jespersen’s Cycle . Conclusions  Indefinites and negation in the history of Latin . Introduction .. Topics and aims .. Roadmap

                                          

OUP CORRECTED PROOF – FINAL, //, SPi

Contents . The Latin negation system in light of Romance .. Position of negative markers .. Lexical renewal in indefinites interacting with negation .. Research questions and hypothesis . The Classical Latin negation system: a first overview .. A Double Negation system .. Negative markers .. Negative indefinites . Negation systems: theoretical backgrounds .. The interpretation of negation .. Types of negation systems . The analysis adopted here .. Negative indefinites .. n-words .. The negative marker . Latin negation system and Jespersen’s Cycle .. Jespersen’s Cycle .. Grammaticalization of the Latin negative marker n¯on .. The syntax of n¯on: position in the clause .. The phrasal status of Classical Latin n¯on (Jespersen’s Cycle, Stage III) .. Emphatic negation .. Interim summary . The negation system in Late Latin .. Late Latin: a ‘concealed non-strict NC language’? .. The development of n¯on in Late Latin: a new Stage I of Jespersen’s Cycle? .. OV order with negative indefinites .. Analysis . Conclusions  Developments in negative and polarity-sensitive contexts from Latin to Romance . Introduction .. Topics and aims .. Roadmap . The situation in Early Romance .. Diachronic puzzles with Romance n-words .. Old French .. Old Italian .. Conclusions . The grammaticalization of nec-words .. The issue .. Etymological remarks

vii                                        

OUP CORRECTED PROOF – FINAL, //, SPi

viii

Contents . Syntax and semantics of Latin n˘ec .. The form .. Functions from Classical to Late Latin .. Semantic analysis .. Syntactic analysis .. Conclusions . From nec-words to n-words .. The internal syntax of nec-words .. Focus particles and polarity sensitivity .. N˘ec unus .. N˘ec ipse unus .. Loss of emphasis with nec-words .. Summary and outlook . Redundant uses of n˘ec and the rise of Negative Concord .. The redundant uses .. The analysis . Concord and polarity-sensitive uses: a unification? .. NPIs or n-words? .. Optional strict Negative Concord in Old Italian . Conclusions

 Conclusions References Index

                      

OUP CORRECTED PROOF – FINAL, //, SPi

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

OUP CORRECTED PROOF – FINAL, //, SPi

OUP CORRECTED PROOF – FINAL, //, SPi

Preface In this work I follow the development of some indefinite pronouns and determiners between Latin and Romance, with the aim of detecting the mechanisms of semantic and syntactic change leading to the Romance outcomes in this grammatical domain. I survey the history of elements of the functional lexicon such as Latin quidam ‘a certain’, aliquis ‘some’, nullus ‘no’, nemo ‘no one’, nihil ‘nothing’, trying to detect which aspects of their meaning and of their form are responsible for the diachronic success of some of them, and for the disappearance of others, and how they are reanalyzed or replaced in the Romance languages. My work shows that the system of indefinite pronouns and determiners changes profoundly from Latin to Romance, but also that the Romance languages maintain a certain degree of similarity in the way their various systems evolve. I argue that we can account for this similarity of outcomes if we consider the changes happening at the intermediate stage of Late Latin, as witnessed especially by texts of the third and fourth centuries ce. At this stage, the grammar of indefinites already shows a number of changes, which are homogeneously transmitted to the daughter languages, accounting for the parallelism among the various emerging Romance systems. Chapter  will introduce the topic and the aims of this study, by preliminarily discussing the dimensions of variation observed in the realm of indefinite pronouns and determiners, as well as the goals and the methods of the historical investigation conducted in this work. Chapter  and chapter  are dedicated to specific and epistemic indefinites. We will see that the history of Latin aliquis assumes particular relevance in this respect, owing to the large amount of variation observed among its Romance continuations (Italian alcuno, French aucun, Catalan algun, Spanish alguno, Portuguese algum). I will propose an explanation for this variation by looking more closely at the properties of Classical and Late Latin aliquis and by arguing that these properties are responsible for the expansion in further contexts observed in Romance. Chapter  and chapter  deal with indefinites in the scope of negation, respectively in Latin and in some Early Romance varieties (Old French and Old Italian). I propose an explanation for the fundamental process of change leading from the one-to-one correspondence between expression and interpretation of negation seen in Latin (a Double Negation language) to the systems, found in Romance, where multiple expressions of negation correspond to just one negative operator (Negative Concord languages). Some indefinites are involved in this process, since they carry the expression of negation (be it contentful or just formal). I will propose that the trigger to this change is a crucial parametric resetting phenomenon concerning the syntax of negation in Late Latin, and I will show the influence that it has on the syntax and semantics of indefinites in the scope of negation. I will also discuss the importance of the Latin focus-sensitive negation particle nec for the origin of new Romance n-words, such as Old French neuns, Italian nessuno, Spanish ninguno. In view of my conclusions

OUP CORRECTED PROOF – FINAL, //, SPi

xii

Preface

on this point, I will propose a novel interpretation of the Negative Concord systems seen in Old French and Old Italian. The conclusions of this study confirm the fruitfulness of applying methods and models developed within synchronic theoretical linguistics to the study of diachronic phenomena. In turn, they bear witness to the importance of diachronic research for understanding the nature of crosslinguistic variation. In particular, in the case at hand, the apparent heterogeneity of the Romance systems observed in the grammatical domain of indefinites can be reinterpreted in a new light and reduced to a few fundamental determinants. The latter can be shown to have already emerged in Late Latin, thanks to the threefold comparison of Classical Latin, Late Latin, and Early Romance carried out in this work. Moreover, the phenomena observed in the history of Latin and of the Romance languages are shown to follow crosslinguistically recurrent patterns and to proceed in a systematic fashion, shedding light on the nature of the semantic and syntactic categories involved, as well as on the mechanisms of change at the syntax–semantics interface.

OUP CORRECTED PROOF – FINAL, //, SPi

Acknowledgments Many people deserve my gratitude for the support they provided during the time I spent in completing this book and for the ideas they shared with me. First of all I wish to thank Klaus von Heusinger: he had constant faith in me and believed in this project from its beginning. Thanks to his advice and his efforts I was able to enjoy the optimum conditions in carrying out the largest part of the work. I am also very grateful to Martin Becker, who supervised my Habilitation at the University of Cologne and gave me invaluable help, and Aria Adli and Cecilia Poletto, who, as reviewers of my Habilitation, were among the first to read the manuscript and contributed very important insights to this work. The project leading to this book started while I was a postdoctoral fellow at the University of Konstanz: I wish to thank all my colleagues there, and in particular Maribel Romero, who sparked my interest for epistemic indefinites in the first place, Doris Penka, who is, conversely, responsible for making me think about n-words (and for introducing me to LaTeX), Josef Bayer, Ellen Brandner, Miriam Butt, Cleo Condoravdi, Georg Kaiser, Paul Kiparsky, Frans Plank, and Arnim von Stechow. The project continued during my wonderful years at the University of Cologne, where I have more debts of gratitude to friends and colleagues than I could possibly acknowledge here. For their special support in this work, I would like to thank Chris Bongartz, Sofiana Chiriacescu, Marco García García, Stefan Hinterwimmer, Agnes Jäger, Beatrice Primus, Volker Struckmeier, besides the people in my Habilitation committee already mentioned. I finished writing this book at the University of Bologna: I am very grateful to my colleagues there, who supported me in the revision, and especially to Nicola Grandi and Fabio Tamburini. There are many other people who gave me extremely useful insights and inspiration for this work: special thanks go to Ana Maria Martins and Cecilia Poletto, who have been incredibly generous with their time and expertise, Alessandra Bertocchi, Theresa Biberauer, Bernard Bortolussi, Paola Crisma, Lieven Danckaert, Ashwini Deo, Cornelia Ebert, Regine Eckardt, Irene Franco, Michèle Fruyt, Jacopo Garzonio, Remus Gergel, Anastasia Giannakidou, Martin Haspelmath, Katja Jasinskaja, Pierre Larrivée, Manuel Leonetti, Giuseppe Longobardi, Guido Mensching, Anna Orlandini, Svetlana Petrova, Gertjan Postma, Esther Rinke, Ian Roberts, Elisabeth Stark, Lyliane Sznajder, Helmut Weiß, David Willis, and Hedde Zeijlstra. The informants for the questionnaire I used in chapter  also deserve particular gratitude: Livia Assunção Cecilio, Valentina Bianchi, Sonia Cyrino, Victoria Escandell Vidal, Maria Teresa Espinal, Marco García García, Dara Jokilehto, Manuel Leonetti, Ana Maria Martins, Beatriz de Medeiros Silva, Esperanza Torrego, JacopoTorregrossa, and Joaquín Vuoto. I also gratefully acknowledge the help of Kristin Wulfert and Eugenio Mattioni, who assisted me in the annotation and revision of the Latin corpus.

OUP CORRECTED PROOF – FINAL, //, SPi

xiv

Acknowledgments

Extremely helpful insights and suggestions came also from the anonymous reviewers of the OUP proposal and of the manuscript. I also wish to thank the series editors Adam Ledgeway and Ian Roberts, and the editorial team at OUP, especially Vicki Sunter, who provided fundamental support. Finally, I could not have made it without my family, and especially my parents. This book is dedicated to them.

OUP CORRECTED PROOF – FINAL, //, SPi

List of abbreviations Abbreviations for descriptive and theoretical notions are introduced and explained in the text. Abbreviations for corpora are further listed in §.., where the full bibliographical record is provided. Abbreviations for Latin authors and texts are listed below only where they do not follow the standard ones used in the Thesaurus Linguae Latinae (online version of the Index librorum scriptorum inscriptionum: http://www.thesaurus.badw.de/en/user-tools/index). Abbreviations used in the glosses conform as far as possible to the Leipzig Glossing Rules. I list them here for ease of reference. 

pragmatically infelicitous

*

grammatically ill-formed (but before stems: reconstructed form)

ABL

ablative case

ACC

accusative case

BC

Commentarii de bello civili (Caesar)

BG

Commentarii de bello Gallico (Caesar)

BP

Brazilian Portuguese

Cat.

Catalan

CL

Classical Latin

COMP

comparative

CP

Complementizer Phrase

DAT

dative case

DIEC

Diccionari de la llengua catalana. Segona edició

DN

Double Negation

DP

Determiner Phrase

EP

European Portuguese

EPP

Extended Projection Principle

FEW

Französisches Etymologisches Wörterbuch

Foc

Focus

Fr.

French

GEN

genitive case

GER

gerund or gerundive

i

interpretable

IE

Indo-European

IMP

imperative

INF

infinitive

Infl

position of the inflected finite verb

It.

Italian

OUP CORRECTED PROOF – FINAL, //, SPi

xvi

List of abbreviations

Itin. Eg.

Itinerarium Egeriae

JC

Jespersen’s Cycle

LF

Logical Form

LL

Late Latin

LLT-A and LLT-B

Brepols’ Library of Latin Texts

LOC

locative

MCVF

Corpus Modéliser le changement: les voies du français

MI

Modern Italian

NC

Negative Concord

Neg

negative

NI

negative indefinite

NM

negative marker

n-word

element of Negative Concord

NOM

nominative case

NPI

negative-polarity item

NT

New Testament

O

object

OF

Old French

OI

Old Italian

OVI

Corpus dell’Italiano Antico, Opera del Vocabolario Italiano

PASS

passive

PIE

Proto-Indo-European

PF

Phonetic Form

PT

participle

Pt.

Portuguese

Pol

Polarity

REW

Romanisches Etymologisches Wörterbuch

S

subject

Sp.

Spanish

SUP

superlative

TLL

Thesaurus Linguae Latinae

TP

Tense Phrase (corresponds to Infl/IP of earlier models)

u

uninterpretable

val

valued

V

verb

VOC

vocative

OUP CORRECTED PROOF – FINAL, //, SPi

1 The grammar of indefinites Functions, variation, and change . Aim and scope of this work This work is a contribution to the multidimensional study of strategies by which speakers use nominal phrases for introducing and managing discourse entities, by linguistically encoding crucial properties such as (non-)identifiability for speaker and hearer, anaphoricity, degree of saliency in previous and forthcoming discourse, scope preferences and constraints. Taking Haspelmath’s () semantic map as my point of departure, I focus on two core areas in the semantic space of indefinite meanings there defined: specific indefinites, on the one hand, and indefinites occurring in the scope of a negative operator, on the other hand. These two classes have been considered to represent two opposite poles in view of a number of closely related criteria. Specific indefinites score higher than other indefinites on a scale of referentiality, since they can introduce persistent discourse entities that can become part of a referential chain. Indefinites within the scope of negation, in contrast, are particularly short-lived in a discourse perspective: their role consists in contributing to negate the existence of a state of affairs. Connected to this, specific indefinites have been argued to be independent of clausal operators for their licensing. Negative indefinites and n-words, instead, are crucially dependent on negation for their licensing. In this work I subject these criteria to scrutiny in the empirical domain of Latin and Romance, with the aim of accounting for the conspicuous differences between the ancestor and the daughter languages in the grammar of specific indefinites and indefinites in the scope of negation. My research on specific indefinites will focus on the properties of Latin aliquis ‘some’, which is characterized as a ‘specific unknown’ indefinite in Haspelmath’s () map, and on the properties of its Romance continuations (Italian alcuno, French aucun, Catalan algun, Spanish alguno, Portuguese algum). In its history, aliquis appears to expand from one pole to the other in the semantic space considered in Haspelmath (), since its Romance continuations consistently show narrowscope uses under negation, unlike the Latin ancestor. I will show that, in order to understand the diachronic developments and the remarkable variation witnessed by Romance in this area, it is necessary to reconsider the semantic nature of Latin aliquis. In chapter  I compare it with a clearly specific indefinite existing in Latin, quidam ‘a certain’; I conclude that aliquis is not a specific indefinite, and is best Indefinites between Latin and Romance. First edition. Chiara Gianollo. © Chiara Gianollo . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

characterized as an epistemic indefinite. This category is absent from Haspelmath’s system but can subsume some of the distinctions he draws in a theoretically coherent way. Moreover, the fact that epistemic indefinites show a form of licensing dependence on clausal operators can account for the expansion of aliquis into further non-specific contexts, starting in Late Latin. In chapter  we then follow the development of the Romance continuations of aliquis, and uncover a complex and yet principled interplay of conservativeness and innovation. With respect to indefinites in the scope of negation, in chapter  and chapter  I will analyze the interaction of semantic and syntactic factors in their behavior between Latin and Old Romance (in particular Old French and Old Italian, but many of the conclusions reached will have a broader impact). We will have to distinguish different subtypes of indefinites in this area (negative indefinites, n-words, negative polarity items) and to account for their interaction with clausal structure. Given their clear dependence on a negation operator, the study of these indefinites exceeds the boundaries of the nominal domain and can be very interesting in a diachronic perspective, since changes in their syntax are typically a symptom of far-reaching changes involving the whole system of negation and, in the case of Latin, also more general aspects of the syntax of the clause. For this reason, the study of negation systems will assume a relevant role in my work, as a paramount example of the dependence relations that indefinites may be subject to. In this first chapter, I will introduce the general theoretical prerequisites for my research and I will situate it in the broader field of studies on the semantics of nominal phrases. Denotation by means of nominal phrases (DPs) represents a foundational function of language, and an essential domain of linguistic competence. Its multidimensional nature makes it a central testing ground for linguistic theories and methods, since a proper treatment of the function and form of nominal phrases encompasses many levels. Most directly it concerns semantics and pragmatics, but morphology and syntax must also be taken into consideration: the form of nominal phrases obviously constrains interpretation, but crucially not in a univocal way. Even closely related languages show differences as to which interpretations or functions they allow for various morphosyntactic forms of referential and quantificational expressions, and as to which forms are chosen to introduce or reuse entities in discourse depending on various contextual factors. The synchronic facts are well known and amply studied.1 Crucially, however, the constraints governing form and interpretation of nominal phrases may change over time (the grammaticalization of definite and indefinite determiners representing a macroscopic development in this respect). While the relevant diachronic phenomena have been subject to investigation in a number of languages (see Lyons : (–) for an overview, and Haspelmath  for indefinites in particular), the general mechanisms affecting referential and quantificational categories in the diachronic 1 Cf. Abbott (), Szabolcsi (), and the chapters on reference and quantification in von Heusinger et al. (: –) for a survey of the main theoretical stances, as well as Alexiadou et al. () and Stark et al. () for two recent crosslinguistic overviews.

OUP CORRECTED PROOF – FINAL, //, SPi

Aim and scope of this work



development of grammars, and especially their motivations and their dynamics, are far from clear. One aim of my research is to uncover tendencies in the development of strategies for the coding and management of discourse entities by means of indefinite pronouns and determiners. Indefinites display an intriguing amount of variation with respect to their morphosyntactic and semantic–pragmatic properties. They create systems of interrelated series, which can differ profoundly even in closely related languages, like the Romance ones. A diachronic investigation is instrumental for understanding the factors that shape series and systems of indefinite pronouns and determiners in the Romance languages. In particular, and as we will see in the case studies presented here, their contemporary morphosyntactic behavior often retains vestiges of an original semantic-pragmatic motivation that may have been lost meanwhile. In turn, this research has a broader theoretical impact: the domain of indefinite pronouns and determiners, given the numerous dimensions of variation, qualifies as a very promising area in which to observe the evolution of functional items of the lexicon, and to test theoretical models for syntactic and semantic change. Indefinites show complex patterns of interdependencies with various operators in the clause. This turns out to be very interesting in a diachronic perspective as well, since changes in the conditions governing such interdependencies may be a signal for broader structural changes at the syntactic level. Another aim of this work is, thus, to investigate this aspect, by focusing in particular on the interaction between negation and indefinites, which has proved to be subject to recurrent evolutionary patterns, or ‘cycles’. This dependence on the surrounding structural context poses interesting theoretical questions concerning the motor of change (the triggering evidence) and the chain reactions (implicational relations) among mutations. Sometimes the change affects a single lexical item; in the domain of indefinites, however, frequently the change has systemic effects, involving a whole class of items (a series), as well as the paradigmatic relation that the class entertains with others. I address these issues in the empirical domain of Latin and Romance, which, thanks to the uninterrupted written tradition and the rich differentiation, qualifies as an ideal testing ground. I do not aim at a comprehensive overview, but rather focus on phenomena that I singled out as particularly significant in connection with the current theoretical discussion, both in synchronic and diachronic perspective. The synchronic discussion on indefinites centers on their semantic-pragmatic status and on how to model their behavior at the syntax-semantics interface, on the one hand, and in discourse, on the other hand. Diachronic research, in turn, has been interested in the fact that changes affecting indefinites, in terms both of their sources of grammaticalization and of their semantic development, are quite frequent and remarkably similar across languages; they may, thus, disclose important insights into the semantic categories involved, as well as into the general principles of language change. The synchronic and diachronic lines of research are obviously interconnected: restrictions on variation (what can vary across languages; what remains constant; how values for variation points cluster) are restrictions on change as well. That is,

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

conclusions reached by means of the theoretical analysis of crosslinguistic data are expected to allow for the formulation of restrictive theories of change. The general questions I pursue are the following: () a. Functions and Variation: (i) What are the synchronic constraints on possible systems of formmeaning mapping, i.e. on variation in function, in the realm of indefinites, as evidenced by typological and theoretical research? (ii) How do system-internal dynamics (competition, blocking) work, from a synchronic and from a diachronic perspective? b. Change: (iii) How do form and meaning of indefinites evolve in time? To what extent are the changes to which they are subject caused, respectively, by meaning-related and by morphosyntactic factors? (iv) What are the contextual parameters triggering or preventing certain interpretations, and what is their role in diachronic reanalysis? (v) How to deal with historical stages or with individual texts displaying apparent optionality in the distribution of certain indefinites? As mentioned, the classes of indefinites which will be of primary interest in my case studies are those that manifest a semantic and syntactic dependence, i.e. that need licensing by a DP-external operator and, thus, establish formal relations with other elements in the clause: epistemic indefinites, negative polarity items, n-words, and negative indefinites. The reason for this choice is that, exactly because these elements show a dependence, the description of their behavior can rest not only on interpretation judgments, but also on additional structural cues, which allow us to reconstruct their function more safely in the case of historical varieties. In addition to these classes, the operator-independent class of specific indefinites will also be discussed. This is motivated by two main requirements. On the one hand, the expression of specificity in Latin will have to be considered in chapter  in order to reach a more precise description of the point of departure for my investigation. On the other hand, the analysis of specificity that I will choose allows the modeling of the space of variation represented by Haspelmath’s semantic map as a continuum of varying restrictions on domains of quantifications (as we will see in §..).

. Overview of the chapters In this chapter I preliminarily discuss the central theoretical issues introduced in (), which are then addressed by means of case studies in the following chapters, and I describe the methods and the material adopted. I first present Haspelmath’s () semantic map for indefinite meanings, and I then look for correspondences with and differences from the dimensions of variation assumed in current formal treatments. This way, I also provide an overview of the main grammatical categories investigated in this work. I further motivate the diachronic investigation of these issues by

OUP CORRECTED PROOF – FINAL, //, SPi

Indefinites: working definitions



arguing that the study of indefinites is very relevant for our general understanding of systematicity in change at the syntax-semantics interface. In the rest of the work, I consider a number of concrete questions of language transmission that arise when considering differences between Latin and Romance. Why is it, for instance, that some Latin indefinites are historically very successful and are transmitted to the daughter languages, while others, like quidam ‘a certain’, disappear? How could it happen that the continuations of aliquis ‘some(one)’, a ‘positive’ indefinite classed by Haspelmath () as ‘specific unknown’, came to be used as negative polarity items and even n-words or negative indefinites in the Romance descendants? And how can we explain that, while Latin is a so-called Double Negation language, where each negatively marked indefinite receives a semantically negative interpretation, the Early Romance daughters systematically display Negative Concord, where multiple negatively marked expressions co-occur conveying a singlenegation meaning? Chapter  is dedicated to specific and epistemic indefinites in Classical and Late Latin, i.e. to the leftward extreme of Haspelmath’s () map. Early and contemporary Romance data are analyzed in chapter , which is dedicated to the comparative study of the continuations of Latin aliquis ‘some(one)’ in Italian, French, Catalan, Spanish, and Portuguese: these span from the retention of the epistemic uses in some varieties to the development of polarity-sensitive dependencies in all varieties and sometimes to the grammaticalization of a syntactic licensing relation with negation. I then move to the rightward extreme of Haspelmath’s map, represented by indefinites in ‘direct negation’ contexts, in chapter . I discuss at some length the system of Early and Classical Latin negation, since it is a prerequisite for investigating the subsequent stages, first of all Late Latin. The latter is also treated in this chapter: I argue that the conditions triggering the development of Romance Negative Concord already emerge in Late Latin and involve the syntax of negative indefinites. Chapter  deals with the development of Romance indefinites specialized into ‘direct negation’ contexts, the so-called ‘n-words’. I concentrate on those n-words that morphologically contain a negation marker, and I investigate its Latin origin, showing the importance of the interaction between focus and negation for the development of Romance Negative Concord. In Chapter  I summarize the results of my work.

. Indefinites: working definitions In this section I provide some working definitions for the objects of my analysis and the adopted categories. I take Haspelmath’s () groundbreaking typological study as my point of departure for the definition of some synchronic and diachronic problems that will be treated more at length subsequently. In the rest of the book, the working definitions introduced here will be subject to further scrutiny and sometimes substantially modified according to the theoretical analysis adopted. Every comprehensive grammatical description of a language contains a section on indefinites. Following the classical tradition, they are usually treated in the

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

part-of-speech section, under the heading ‘indefinite pronouns’. But what are indefinites exactly? Haspelmath’s discussion in introducing his typological study is illuminating with respect to the difficulties of reaching a proper definition (Haspelmath : –). First of all, both functional and formal criteria are intertwined in the definition. As for the formal criteria, Haspelmath uses the term ‘pronoun’ to indicate that indefinites are ‘grammatical elements’, i.e., part of the functional lexicon. In fact, indefinites may occur both with the value of a full nominal phrase (pronominal use proper) and as part of a complex nominal expression, in the role of determiners (in traditional treatments of Latin and Romance they are considered ‘adjectival’ in view of their agreeing morphology). Often the same lexical items can occur in both uses (with the appropriate morphophonological adjustments), e.g. Italian nessuno ‘nobody / no’. In some cases, however, languages have different items for the pronominal and the determiner use, e.g., English pronominal nobody versus determiner no. In other cases it is not clear at all whether we have to assume one morphosyntactically ‘flexible’ lexical item or rather two different items, cf. French pronominal quelqu’un ‘someone’ versus determiner quelque ‘some’. Indefinite articles are usually treated under a separate heading in traditional grammars, and are left out of Haspelmath’s survey. The reason for this is that articles are considered to be such when they are syntactically necessary to build a nominal phrase in argumental function, i.e., when they realize a functional category (D) that has to be present in the language. Indefinite ‘pronouns’, on the other hand, are considered to express ‘optional’ information, which is strongly context-dependent: e.g., whether the referent is known to the speaker, whether it is going to be picked up again in the following discourse, whether the speaker wants to leave the hearer free to choose any individual from the relevant domain, etc. However, the ‘optionality’ criterion is by no means clear-cut: for instance, the interaction between negation and indefinites shows that sometimes the presence and the form of an indefinite are obligatorily determined by the surrounding morphosyntactic context in a way analogous to what happens with definite and indefinite articles. So, if I want to unambiguously say in English that it is not the case that there exist students who came to my office yesterday, I am forced to use (a) instead of (b), (c), or (d): () a. b. c. d.

No student came to my office yesterday A student did not come to my office yesterday Students did not come to my office yesterday *Any student came to my office yesterday

Moreover, part of the semantic contribution of indefinite pronouns and determiners derives from their paradigmatic relation with the indefinite article, i.e. from the implicit comparison that conversational agents draw between different possible meanings that could be expressed and inferred. In principle, then, it is necessary to consider also the indefinite article as part of the ‘system of indefinites’, and also to address the question of what happens when the language has no (indefinite) articles, as is the case in Latin.

OUP CORRECTED PROOF – FINAL, //, SPi

Indefinites: working definitions



In the course of the discussion, we will see that the process of grammaticalization of the indefinite article from Latin to Romance has important repercussions on the system of indefinites. However, I will not deal with it specifically here: that is, in the title and in the body of this work, I use the term ‘indefinites’ to refer to both pronouns and determiner-like elements expressing an indefinite meaning, to the exclusion of the indefinite article. As ‘determiner-like’ I understand elements realizing functional categories of the nominal phrase related to denotation, without assuming that they necessarily occupy one and the same projection (e.g., DP). I will follow much current literature in assuming a rich functional structure for the DP, where determinerlike elements can occupy different projections (see Alexiadou et al. : part II for an overview of current assumptions). There is an ongoing debate as to whether quantifiers should be considered heads of a different, higher QP projection. Previous work has shown the difficulties of empirically assessing this aspect, especially for a language such as Latin, which is characterized by word-order flexibility (cf. Giusti et al. ). Since deciding on this matter was not immediately relevant for my case studies, and also in view of the controversy surrounding the quantificational status of indefinites (on which see further §.), I will remain agnostic and adopt the label ‘quantificational determiners’ for the subclass of determiner-like elements involved in quantification operations. Concerning the functional criteria, the question naturally arising at this point is: ‘What is an indefinite meaning?’. Roughly, indefinite nominal phrases are expressions with existential quantificational import; they introduce new discourse entities and they are able to convey a series of side-messages that qualify the state of knowledge of the speaker and give instructions on how to update the conversational background (the Common Ground). Under certain conditions the introduced entities may be quite persistent discourse objects, which become part of a referential chain. Other discourse entities are, instead, short-lived and necessarily ‘closed off ’ in the scope of higher operators.2 This definition covers only a subset of the functional items that are treated as ‘indefinites’ in traditional grammars. Mid-scalar or proportional quantifiers (e.g. few, many), universal quantifiers (e.g. each), generic pronouns (e.g. German man), identitives (e.g. same, other) are also often comprised under the traditional label, on the ground that they do not require the conversational agents to precisely identify the individuals denoted by the nominal expressions. However I will follow Haspelmath () in leaving them out of the picture, and focus on the functional space that he considers. Consequently, by ‘system of indefinites’ I will mean the complex of forms that a language displays to cover the functional space considered in Haspelmath’s map. I introduce Haspelmath’s system in the next section.

2 The label ‘discourse entity’ is meant to informally correspond to the notion of ‘discourse referent’ in Discourse Representation Theory. Discourse referents represent indexes connected to all sorts of nominal phrases—not just the referring ones—and subject to different binding procedures depending on their referential or quantificational properties.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

. The functional space of indefinites Haspelmath () represents the overall functional space covered by the indefinites he considers by means of a ‘semantic map’. In Haspelmath (: –) this semantic map format is used to describe the systems of indefinites in forty languages from various families.3 Among them are Latin, Portuguese, Catalan, French, Italian, and Romanian: basing my argument on a comparison among the maps for these languages, I will introduce the main developments that will be dealt with in the next chapters. In §.. I provide an introduction to Haspelmath’s semantic map method and to the categories he uses.4 In §.. I exemplify its application by introducing Haspelmath’s treatment of Latin indefinites, so that we will also have a first overview of the series on which I will focus in the next chapters. In §.. I summarize the historical problems analyzed in the rest of the work. .. Haspelmath’s () semantic map Haspelmath’s () semantic map shows the distribution of indefinites according to functions or uses. The basic structure is represented in (). () Haspelmath (: ): semantic map for indefinites

(1) specific known

(2) specific unknown

(4) question

(6) indirect negation

(5) conditional

(8) comparative

(7) direct negation

(3) irrealis nonspecific

(9) free choice

The map is based on a number of theoretical assumptions (definition of functions, definition of series, diachronic generalizations) that I review below. ... Functions Functions are hybrids of form-context pairs and form-meaning pairs. Their aim is to capture the fact that languages use different indefinite forms depending on a number of semantic and syntactic factors, which are admittedly often difficult to disentangle. In some cases what is crucial is whether the form can express a given meaning (e.g., specific known, specific unknown, irrealis nonspecific, free choice in the map in ()). In other cases the function rather indicates a syntactic context in which languages may employ specialized forms (e.g., question, conditional, 3 The typological study is also based on a broader database of  languages, investigated in less detail. 4 For a broader overview of the issues surrounding the method and other examples of semantic maps, see van der Auwera (); Narrog and van der Auwera ().

OUP CORRECTED PROOF – FINAL, //, SPi

The functional space of indefinites



comparative, indirect negation, direct negation). What is relevant in this latter case is the ability of the indefinite to occur in the scope of a certain operator, and not the meaning contribution of the indefinite per se. For a given form (usually belonging to a series, on which see §...) the map indicates whether it can occur with a certain function. An indefinite series can be found—and actually is typically found—in more than one function: in this case a line circles the multiple functions covered by the series (we will see an example in ()). Haspelmath (: ) encounters ‘massive multifunctionality’ in the languages he examines. An advantage of the notion of ‘function’, according to Haspelmath, is that it makes possible to avoid taking a stance on whether multifunctionality results from polysemy or is rather to be reduced to the combination of a general meaning (Gesamtbedeutung) with contextual effects. Another frequent pattern is represented by the overlap of different indefinite series in a given function. In this case, however, the map does not provide information as to whether there is a meaning difference in a given context when using the one or the other indefinite. Functions are arranged geometrically in the map space. Their disposition is meant to represent implicational relations (hence the label ‘implicational map’ used by Haspelmath): adjacent functions are considered to be related in a systematic and semantically motivated way. The arrangement is, thus, motivated on theoretical and empirical grounds. It is expected that, if an indefinite item is used in more than one function, the involved functions should be adjacent on the map. This expectation is empirically confirmed in Haspelmath’s sample, where the rare exceptions can be explained as due to homonymy, language contact, or loss of intermediate categories. From a theoretical point of view, the geometrical arrangement is intended to represent the semantic distance or closeness, i.e., the degree of similarity, of the meaning categories involved. The two extremes of the map, specific indefinites on the one hand and free choice indefinites and direct negation indefinites on the other hand, represent in many respects the two opposite sides of what can be considered a referentiality scale.5 The referent of a specific known indefinite (e.g., a certain N) is a highly identifiable individual (by the speaker), while in the case of a free-choice indefinite (e.g., whichever N) the identification of the individual is explicitly left entirely open. The function of a free-choice indefinite is, in this respect, exactly the opposite of that of a specific known indefinite: it signals to the hearer that it is not necessary and not possible to identify an individual in the considered domain. Also, indefinites occurring under the direct scope of negation (direct negation function, e.g., nobody) can be considered the opposite of specific indefinites, since they convey that the existence of a referent is negated. The specific unknown function is used by Haspelmath for pronouns and determiners which, like specific known indefinites, convey the presupposition of existence of an entity fulfilling the denotation, but differ from specific known indefinites in that the speaker explicitly indicates that s/he cannot or does not want to identify such an 5 This scale, variously assumed in typological work (cf. Croft :  a.o.), does not refer to the logical type of the nominal expression (referential versus quantificational), but rather to the degree to which the intended denotation is identifiable for the speaker.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

entity. In English the specific unknown function can be expressed by the indefinite some. I will discuss the specific unknown function extensively in chapter . The ‘middle field’ of the map represents a number of environments where polaritysensitive items (e.g., any) are used. We will see more in detail later on that the semantic property these ‘weak polarity’ contexts have in common has been argued to be (Strawson) downward-entailingness; it is a fact that a single indefinite series typically covers several of these environments. Haspelmath considers in particular questions, the antecedents of conditionals, the standard of comparison, but similar effects are observed e.g. in the restriction of universal quantifiers and in the context of before. Another polarity-sensitive context treated by Haspelmath is the indirect negation function: under this heading he considers, on the one hand, the presence of negation in a superordinate clause, and on the other hand, contexts of ‘implicit negation’. The latter comprise cases where the negative contribution comes not from the standard marker of sentential negation (e.g., not), but rather from a preposition such as without, or a verb such as lack or deny. In some languages the indefinites used here are different from those employed in the context of same-clause sentential negation. Under the label ‘irrealis non-specific’ a number of quite heterogeneous modalized contexts are included (e.g., those where the imperative mood or the future tense appear). I will come back to them in chapter . ... Series Series are morphologically and semantically related classes, comprising pronominal, determiner-like, as well as adverbial items (cf. Haspelmath : –). Series comprise dedicated pro-forms for a number of ontological categories (sortal restrictions), e.g., person / animate, thing, place, time, manner, as well as the corresponding determiner: see the example of the English some-series in (): () English some-series (cf. Haspelmath : ) a. ‘person’: somebody, someone b. ‘thing’: something c. ‘place’: somewhere d. ‘time’: sometime e. ‘manner’: somehow f. determiner: some Series can, thus, be understood as a sort of paradigm, in which the same core semantic function (e.g., specific known, free choice, negation) or cluster of semantic functions is expressed by forms derivationally adapted to their syntactic role (argumental pronoun, adverb, determiner). Elements belonging to a series share the same behavior across contexts. They are thus characterized by the same syntactic and semantic properties along the relevant dimensions. Often this homogeneity is reflected by their form: for instance, the same suffix is morphologically combined with various lexical-functional bases, as we saw with English some-. In other cases elements belonging to a series are ‘suppletive’:

OUP CORRECTED PROOF – FINAL, //, SPi

The functional space of indefinites



they do not undergo a parallel morphological derivation, and are related only functionally. For instance, the German direct-negation series (Haspelmath : –) has etymologically negative elements formed with the same negative suffix for the categories ‘person’ (niemand), ‘thing’ (nichts), ‘place’ and ‘direction’ (nirgends, nirgendwo, nirgendher), ‘time’ (nie), but not for the determiner element, which is etymologically distinct (kein). Diachronically, elements of a series may be lost, or new ones may join in. During language change, elements belonging to a series may change at different paces, but they tend ultimately to converge. ... Diachronic generalizations Haspelmath’s map is designed also to make predictions with respect to the directionality of the diachronic developments affecting indefinites. First, the expansion of indefinites into new functions is expected to reflect the semantic similarity of environments adjacent on the map: ‘where markers gradually acquire new functions, they will first be extended to those functions that are adjacent to the original functions on the map, and only later to functions that are further away’ (Haspelmath : ). Secondly, once diachronic data are projected onto the map, one observes that a number of recurrent paths of evolution (‘cycles’) show unidirectionality. Semantic maps are meant to capture these developmental tendencies, by finding a common denominator for the various historical phenomena. For indefinites, for instance, Haspelmath resorts to a notion of bleaching from ‘stronger’ to ‘weaker’ functions to describe the semantic changes leading to shifts or expansions in the map, which tend to have a right-to-left direction. We will come back to these issues in §..., where we will also see that some diachronic conclusions, especially concerning directionality, have not gone unchallenged. Unidirectional developments are relevant for linguistic theory because they point to recurrent mechanisms of change and to implicational relations between meaning ingredients. .. Semantic map for Latin indefinites Haspelmath’s () map for Latin is reproduced in (). As is clear from the examples he uses in the discussion, Haspelmath covers here the grammatical system of Classical and Late Latin (until the fourth century ce). As mentioned in §..., the lines circling multiple functions signal the functional space (in terms of meanings or contexts) where a given indefinite form may be used. () Semantic map for Latin (Classical to Late) in Haspelmath (: )

question specific known –dam

specific unknown

irrealis nonspecific ali–

indirect negation

direct negation –vis/-libet

conditional

comparative –quam

free-choice

n–

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

The map represents a selection of the wealth of indefinite forms that Latin displays, based on frequency of use and functional load.6 The major series considered by Haspelmath (: –) in his map are the following (according to his terminology): ()

(a) the ‘non-emphatic’ ali-series: e.g., aliquis ‘someone’ (b) the free-choice series formed with -vis, and the free-choice series formed with -libet: e.g., quivis ‘whoever you want’, quodlibet ‘whatever you want’ (c) the negative-polarity series marked by -quam (e.g. quisquam, ‘any’; Haspelmath includes also ullus ‘any’ in this series) (d) the negative n- series (e.g. nemo ‘nobody’)

In addition to full-fledged series, the map also shows the specific known indefinite quidam ‘a certain’, which does not form a series, since it occurs only as a determiner and in the pronominal use. The Latin system of indefinites also comprises the pronoun quis, which is identical in form to the interrogative pronoun and is used as indefinite in the scope of a number of operators, i.e. in contexts like ‘irrealis non-specific’, ‘question’, ‘conditional’ in Haspelmath’s map (cf. chapter  and Bortolussi ). In these environments it systematically alternates with aliquis, therefore it is not displayed separately in the map by Haspelmath, who follows the traditional grammatical descriptions in considering it a kind of allomorph of aliquis showing up in a subset of the contexts covered by the latter. Interestingly, except for the negative n- series, all the elements of the Latin system are based on the interrogative stem (qu-). The latter is a very frequent morphological component of indefinites crosslinguistically: in Haspelmath’s -language sample, sixty-three languages show a relation between interrogative and indefinite morphemes. In some languages—including Latin, as just seen—the bare indefinite pronoun is formally identical to the interrogative pronoun, cf. ():7

6 Indefinites not considered in the map are the quite rare negative-polarity series suffixed with -piam and two series of free-choice indefinites, the -cumque series and the reduplicated series, e.g., quisquis. More comprehensive descriptions and analyses of the Latin system of indefinites can be found, a.o., in Orlandini (, ), Mellet (, ), Maraldi (, ), Bortolussi (, , ), Bertocchi et al. (), Fruyt and Spevak (), Devine and Stephens (: ch.  and ), Bortolussi and Sznajder (), Bertocchi and Maraldi (), Pinkster (: ch. ). 7 For Indo-European, it is a matter of debate whether the interrogative and indefinite stems (*kw o/e, *kw i, accented when interrogative; enclitic when indefinite) are synonymous or diachronically derived. Haspelmath (: –; cf. also –) remarks that if there is a morphologically detectable basic-derived relation between interrogatives and indefinites, indefinites are always the morphologically derived form. This etymological matter has gained renewed relevance in the current theoretical debate on the format of indefinites, on the ground of the connection that has been proposed between the meaning of questions and the meaning of indefinites in Hamblin semantics (cf. Kratzer and Shimoyama ). The fact that many genealogically unrelated languages use wh-elements as components of indefinites may be considered evidence in favor of this hypothesis. However, Giannakidou and Quer (: ) caution against jumping too fast to conclusions about synchronic systems on the basis of etymological parallels. Moreover, an adequate analysis should also consider the relation between the interrogative and the relative stem, since free relatives are another common source of indefinites.

OUP CORRECTED PROOF – FINAL, //, SPi

The functional space of indefinites



() Classical Greek tís / tis ‘who’ / ‘some’ Latin quis ‘who’ / ‘some’ German wer ‘who’ / ‘some’ Russian kto ‘who’ / ‘some’ Chinese shéi ‘who’ / ‘some’ Latin series typically comprise pro-forms lexicalizing sortal restrictions such as person (e.g., quisquam ‘anyone’), thing (quidquam ‘anything’), place (usquam ‘anywhere’), time (e.g., umquam ‘ever’), as well as the corresponding determiner (ullus ‘any’); for some series (the interrogative and the -vis free-choice series, as well as the negative series), there is a special determiner for the dual number (respectively, uter, utervis, neuter). To exemplify, in the table below I provide the full series for those indefinites that will be more central to the discussion in the next chapters: the ali-series and the negative n-series. () Example of indefinite series in Latin: Sortal restriction ali-series person thing place time determiner

aliquis ‘someone’ aliquid ‘something’ alicubi ‘somewhere’ aliquando ‘sometime’ aliqui ‘some’

n-series nemo ‘nobody’ nihil ‘nothing’ nusquam ‘nowhere’ numquam ‘never’ nullus / neuter ‘no / neither’

The composition of Haspelmath’s series shows that languages may have different forms for the pronominal indefinites for ‘person’ and ‘thing’, on the one hand, and for the determiner-like indefinites on the other hand. In Latin, we have some such cases: for instance the negative polarity indefinite quisquam ‘anyone’ is used only pronominally, and the corresponding determiner is ullus; the negative indefinites nemo ‘no one’ and nihil ‘nothing’ are pronouns and the corresponding determiner is nullus. But most Latin indefinites can be used either as pronouns or as determiners (‘pronominal adjectives’ in the traditional terminology), i.e., either by themselves or as elements of a complex nominal phrase containing a lexical noun.8 An example is given in (), where (a) has aliquis in the pronominal use, and (b) shows the determiner-like use:

8 See §.. for tendencies in their positioning. In some cases the pronominal and determiner forms follow the same declension class (e.g., quicumque ‘whoever’). Some other items follow partially different inflection classes in their pronominal and determiner uses: e.g., the indefinite pronoun aliquis (m. f.), aliquid (n.) corresponds to the determiner form aliqui (m.), aliqua (f.), aliquod (n.).

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

() a. cum te alicuius improbitas perversitasque when you:acc someone:gen dishonesty:nom wickedness:nom-and commoverit irritate:sg ‘once you are irritated by the dishonesty or wickedness of someone’ (Cic. Quint. ..) b. contra alicuius hominis nobilis voluntatem against some:gen man:gen noble:gen will:acc ‘against the will of some person of high rank’ (Cic. Verr. ..) The classification of an element as pronoun or determiner in its actual textual occurrence is complicated by the fact that even determiner forms such as nullus can be ‘substantivized’ and appear with a null NP complement (cf. (a)); conversely, pronominal forms such as nemo can be accompanied by a nominal phrase in apposition (cf. (b)). () a. Cum constaret istum Syracusis a nullo visum when be.settled:sg this:acc Syracuse:abl by any:abl seen esse archipiratam be:inf captain.pirate:acc ‘When it was established that this pirate captain had not been seen by anyone in Syracuse’ (Cic. Verr. ..) b. invenire neminem Siculum potuit qui pro find:inf nobody:acc Sicilian:acc can:sg who:nom for se cognitor fieret? himself:abl attorney:nom become:sg ‘is it possible that he could find no Sicilian to stand attorney for him?’ (Cic. Verr. ..) We know from Poletto (: ch. –) that sometimes the pronoun / determiner divide is relevant for the diachronic development of a certain item, in terms of its distribution. Therefore, in annotating my data for the studies presented in the following chapters, I always distinguished between determiner and pronominal uses. However, since I was interested in particular in the semantic-pragmatic properties of indefinites, I conflated into the category ‘determiner’ all instances where the item was accompanied by an overt lexical restriction, that is, also cases such as (b) (where, syntactically, we are in fact dealing with an appositional structure). .. From Latin to Romance: the case studies Haspelmath’s survey allows us to draw a first comparison between Latin and the Romance languages, and to single out some diachronically relevant facts. Haspelmath (: –, –) provides maps for a number of Romance languages: Portuguese, Catalan, French, Italian, and Romanian. A comparison between the maps for Latin and for the Romance descendants preliminarily suggests that the following aspects are common to the history of Romance: () a. According to the maps, no variety retains the unambiguous lexical encoding of the specific known versus unknown distinction, i.e., of the epistemic status of the speaker, which Latin had in the quidam / aliquis opposition.

OUP CORRECTED PROOF – FINAL, //, SPi

The functional space of indefinites



b. There is no specialized indefinite unambiguously marking specificity. One single form is used in a cluster of functions comprising the specific indefinite one, as well as other nonspecific contexts, i.e., it is found in irrealisnonspecific, interrogatives, and conditionals, and in some cases also in the indirect negation context. This form is the continuation of Latin aliquis in Portuguese and Catalan. French and Italian have a new indefinite (respectively, quelque and qualche); Romanian also has a new set of forms, the -va series. c. No Romance language has a form exclusively used in direct negation function, as the Latin series of negative indefinites was. Romance languages typically have new forms in this function, which expand to cover at least the indirect negation function (cf. Romanian ni- series, Italian nessuno series), and frequently also further functions in the polarity-sensitive part of the map. These are the phenomena I focus on in this book: they have been chosen exactly because changes that apparently involve individual lexical items are in fact the manifestation of a more global reorganization of the system of indefinites. This reorganization follows a similar path in various Romance languages: we are faced with a scenario where all daughter languages differ from the mother language in a similar way, a scenario that is in principle amenable to three different lines of explanation (cf. Roberts : –, Gianollo , Longobardi  for discussion): (i) language contact; (ii) chance convergence; (iii) (chain-effects of) an inherited change. In many cases of language change, empirically distinguishing between the three is no trivial task; moreover, in the history of the Romance family, contact can practically never be excluded. Nonetheless, my aim for the cases analyzed here is to show that the motor of the change lies in phenomena taking place already in Late Latin and, thus, that the correct line of explanation is (iii): Romance languages may differ profoundly from the Classical Latin stage, but they ‘differentiate in parallel’ because they inherit from Late Latin the crucial seeds for later changes. In order to account for (a) and (b), I will focus on the history of Latin aliquis (chapters  and ). We will have to revise the picture emerging from Haspelmath’s maps in a number of ways. In particular, it will be necessary to distinguish between the singular and plural forms of the Romance continuations of aliquis. Only plural forms can occur in the specific known functions. The singular occurring in the specific unknown function, retained only in some varieties, will be reinterpreted as a nonspecific epistemic indefinite, and this use will be shown to already belong to Classical Latin. The expansion into downward-entailing contexts starts in Late Latin and is continued in different ways in the various daughter languages.9

9 The conclusion in (a) should probably be revised also in view of the fact that Romance languages have developed new means to explicitly indicate the ‘specific known’ interpretation (as a reviewer points out). Important in this respect is the grammaticalization process involving the continuations of Latin certus ‘certain’, on which see §... and Stark (, ); Garassino (). Haspelmath (: –) disregards expressions such as a certain in his map, because he focuses on elements that have a clear pronominal or determiner status. As Zamparelli () observes for Italian, un certo ‘a certain’ shows determiner-like properties, and has definitely a more grammaticalized status than analogous expressions like uno specifico ‘a specific’; in Spanish, where cierto does not cooccur with un, it appears fully grammaticalized as determiner (Eguren and Sánchez ). However, its interpretation is not always straightforwardly

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

In order to explain (c), I will propose (chapters  and ) that the substantial reorganization of indefinites in the scope of negation observed in Romance is ultimately due to far-reaching changes in the syntax of negation that take place in Late Latin. These changes are transmitted to Romance and represent the prerequisite for the development of Negative Concord, crucially involving indefinite pronouns and determiners. Apart from their historical significance, the phenomena I selected are relevant also from a theoretical point of view, and relate to issues that have been extensively investigated in comparative perspective. In the next section, I introduce the main dimensions of variation and change that will be the object of my case studies.

. Dimensions of variation and change This section introduces the main dimensions of variation and change that will be addressed in this work in order to propose an account for the historical puzzles in §... In §.. I formulate the broader theoretical questions that my case studies address. In §.. I motivate my choice of phenomena and I preliminarily introduce the categories and the dimensions of variation that will be involved. In §.. I discuss in particular the importance of conditions on quantificational domains to account for variation. In §.. I argue that my case studies can contribute to an improved understanding of systematic processes of semantic change. .. A research program The crosslinguistic picture and the generalizations emerging from Haspelmath’s map have a clear appeal for formal models of semantic and morphosyntactic variation. First, the map substantiates, for the realm of indefinites, what I call the MatthewsonKratzer conjecture on semantic variation (cf. Matthewson ; Kratzer and Shimoyama ; Kratzer ): denotations expressed by indefinites distribute across a uniform semantic space language after language; crosslinguistic variation resides on the one hand on the selectivity that certain items show for certain contexts in virtue of semantic-pragmatic constraints they encode, and on the other hand in morphosyntactic properties of lexical items; both dimensions of variation are hopefully reducible to a restrictive format. A number of theoretical challenges emerge from this far-reaching research hypothesis: how to define Haspelmath’s functions in terms of formal meanings? how to reach a more fine-grained analysis of the compositional meaning blocks? what is the interplay of truth-conditional and pragmatic, contextually contributed meaning here? how does it reflect in the syntax? how to formally account for ‘adjacency’ of meanings in the map?

analyzable as ‘specific known’ (cf. Jayez and Tovena  for French; Eguren and Sánchez  for Spanish; Garassino  for Italian): the role of this item in the various Romance systems should therefore be ascertained by means of dedicated studies.

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



The way the semantic map is structured overlaps in some important respects with implicational categories proposed in the formal semantic literature, like the hierarchy of negative contexts (Zwarts , van der Wouden ), the continuum of downward-entailing contexts (cf. Hoeksema ), or the continuum of so-called ‘referentially deficient indefinites’ (cf. Giannakidou , Giannakidou and Quer ), which correspond to the central and right-hand functions in Haspelmath’s map. However, the notion of function, owing to its hybrid nature, is insufficient for a rigorous descriptive and explanatory account: we need to more precisely distinguish between context and meaning, by providing lexical entries for indefinites that define the interpretations they may receive and the environments compatible with such interpretations. Secondly, from a diachronic point of view, the map singles out developmental clines and suggests that it is possible to reach principled explanations for them, based on the logical relations among meaning components and on acquisition strategies guiding reanalysis. An answer to the theoretical questions listed above is, thus, inextricably connected to a problem set for historical linguistics: how to turn the map’s connecting lines into arrows? are change phenomena unidirectional? where do cycles start? can they involve the entire space of the map, or do they spread across only certain functions? Many approaches to cyclicity and directionality in the domain of indefinites have been based on a tripartition into positive / nonassertive / negative contexts (cf. e.g., Martins , Weiß a, Jäger , Ingham a). This simplified representation of the space of variation is mainly motivated by the intrinsic difficulty of systematically investigating finer-grained distinctions in historical documents. However, some of the systematic patterns of change observed in the system of indefinites require a more elaborated model in order to capture the regularities underlying the shifts from one function to the other. The case studies which I present in this book are also meant to address these broader questions and to provide some answers based on empirical observation of Latin and Romance facts. While there is a persisting intuition that indefinites form a natural class at some level, the debate on their format is not settled. Many different analyses have been proposed in the literature, with the aim of accounting at the same time for quantificational, scopal, and discourse properties of indefinites, by means of a uniform approach. A line of analysis treats indefinites as existential quantifiers. According to another family of approaches indefinites do not have an intrinsic quantificational force, but receive it from operators in the surrounding structural context; indefinites just introduce variables (or, depending on the theory, choice functions or sets of individual alternatives). Some approaches do not provide a uniform treatment, but sharply distinguish classes of indefinites on the basis of their logical type, differentiating between quantificational and referential (entity-denoting) indefinites.10 10 Cf. Onea (: chapter ) for a recent thorough discussion of the main proposals (as well as an original analysis in the framework of Inquisitive Semantics). Ihsane () comprehensively discusses the syntactic consequences of the various semantic approaches.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

For the aims of my investigation it will be sufficient, for simplicity, to uniformly adopt a quantificational analysis: I treat indefinites as quantificational determiners, forming together with the NP-restriction generalized quantifiers with existential force. Nothing substantial in my conclusions hinges on this choice. The only aspect of the analysis that, in its formulation, depends on the quantificational format is the fact that I describe a fundamental dimension of variation among indefinites as variation in the constraints on their quantificational domains (cf. §..). However, also non-quantificational analyses of indefinites involve conditions that are attached to them in the lexical representation, with analogous effects to constraints on quantification domain: in Discourse Representation Theory, for instance, variables come with constraints on the way they should be bound. I use conditions on domains to model one dimension of Haspelmath’s semantic map as a continuum of domainshifting operations. This way, it will be possible to explain some systematic diachronic processes involving indefinites as due to changes in such conditions. .. The classes investigated in this work As introduced in §., a number of interrelated criteria led to the choice of the case studies presented here. First, the specific (known or unknown) function and the direct negation function represent two opposite poles in the functional space of Haspelmath’s map, yet there exist diachronic processes that connect these two poles, such as the development of Latin aliquis. Second, the chosen phenomena manifest systemic effects: the changes are not limited to a single lexical item, but affect an entire grammatical module (in the cases at hand, mainly the syntax of negation, but also the expression of specificity). Third, the classes on which I will focus display patterns of interdependencies with operators in the surrounding structural context. Epistemic indefinites, negative polarity items, and indefinites with fixed narrow scope with respect to negation are all items that are subject to particular licensing requirements, at the semantic-pragmatic level or at the syntactic level. This provides, from a theoretical perspective, important information on the nature of the selectivity of quantificational determiners, which I will tie to the continuum of varying restrictions on domains of quantification. From the perspective of historical linguistics, we have clearer structural—not just interpretational—evidence for their diachronic analysis and we can better understand how these indefinites are sensitive to changes affecting the grammatical requirements of their licensors. Before discussing these aspects in more detail, it is necessary to refine the terminology by indicating the overlaps and differences between Haspelmath’s functions and the categories that I will adopt from the formal semantic literature. The next section is dedicated to this task, respectively for the left-hand and for the right-hand side of the semantic map. ... Epistemic indefinites As announced in §.. for (a), I will revise the categories on the left-hand side of Haspelmath’s map by adopting a more restricted notion of specificity and by introducing the class of epistemic indefinites. The theoretical status of ‘side-messages’ conveyed by indefinites with respect to the epistemic status of the speaker is a hotly debated topic in current formal semantic

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



and pragmatic research (see Alonso-Ovalle and Menéndez-Benito ,  for an overview). Indefinites conveying the ‘ignorance effect’ that (partly) corresponds to Haspelmath’s ‘specific unknown’ function have been subject to much attention recently under the heading of ‘epistemic indefinites’. European Portuguese algum, a continuation of Latin aliquis, expresses the ignorance component typical of epistemic indefinites, cf. ():11 () Algum aluno me disse isso, mas não sei quem foi ‘Some student told me that, but I don’t know who it was’ In chapter  I will thoroughly introduce this class and propose that also Latin aliquis can be treated as an epistemic indefinite, and that such analysis may account more straightforwardly for its diachronic evolution than one treating it as a specific indefinite. In the framework I adopt, a specific indefinite introduces a discourse entity that is identifiable by the speaker (and is presented as novel for the hearer): in this respect, specific indefinites are high on the referentiality scale, in the sense that their denotation is highly identifiable. Epistemic indefinites, instead, have an intermediate status, since they impose a variation requirement on the operator–variable relation they encode: the speaker conveys that more than one entity in the discourse can correspond to the denotation of the epistemic indefinite. Jayez and Tovena () consider epistemic indefinites as a subclass of free-choice indefinites. The main difference between the two classes consists in the fact that for free-choice indefinites the considered domain is widened so as to comprise marginal cases, which yields a reduced tolerance for exceptions; with epistemic indefinites, instead, the considered domain is widened so as to comprise potential alternative denotations, but not maximally widenened, so that the exclusion of alternatives is possible. As we will see in chapter , the precise meaning contribution of epistemic indefinites is strongly dependent on the type of operators they interact with: in this they are much more similar to polarity-sensitive indefinites (showing ‘dependent reference’ according to Giannakidou ) than to specific indefinites. And indeed, the historical development of aliquis in the Romance languages shows that epistemic indefinites expand into the domain of negative polarity (rather than free-choiceness).12 In order to address (b), in chapter  we follow the history of aliquis. We will see that an originally epistemic indefinite has the semantic potential to expand into a number of functions in the polarity-sensitive landscape: an originally inherited change (the expansion of aliquis into negative-polarity contexts in Late Latin) leads to some

11 Thanks to an anonymous reviewer for providing this example. 12 As most clearly exemplified by English any, free-choice and negative-polarity items are often diachronically related. This has been considered a sign that they share a common meaning component, variously analyzed in the literature. Some have proposed that free-choice indefinites and NPIs share widened or ‘large’ domains (cf. Kadmon and Landman  and Krifka  for this analysis of negative polarity). Others have proposed that they share scalarity: namely, free-choice items involve the operator even on quality scales (e.g. Lee and Horn , Lahiri ), while negative-polarity items are analyzed in terms of even on quantity scales (Lee and Horn ). Chierchia () has tried to combine the two components.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

interesting paths of divergence in the different Romance branches and languages, which sometimes retain the epistemic use and sometimes lose it, but retain the item in negative contexts. In the framework I adopt, the dimension of variation that concerns the degree of epistemic certainty with respect to the identification of the denoted entity is dependent on the constraints imposed by different indefinites on their quantificational domain. As we saw, there are indefinites that express plain ignorance, and others that express indifference with respect to the identification of the referent (cf. in French respectively quelque ‘some or other’ and quelconque ‘any’, Jayez and Tovena ). This has been modeled as variation with respect to the space into which the conversational agents may look for suitable entities, which is represented as maximally widened in the case of free-choice items such as quelconque and only minimally widened for quelque. According to the same line of analysis, specific indefinites represent, instead, quantificational expressions whose domain of quantification is maximally reduced to a singleton set. Pursuing this line of analysis, I will argue that the epistemic component is responsible for the diachronic development of epistemic indefinites into elements that depend on other operators, like negation. In this respect, my analysis provides a principled explanation for the puzzle represented by the fact that an originally ‘positive’ indefinite becomes ‘negative’ in some Romance varieties, in a way which an analysis in terms of specificity would not allow. ... Indefinites in the scope of negation Negative-polarity (and in particular direct negation) contexts will be the topic of chapter  and chapter . The fact that Latin has one series used exclusively in direct negation contexts and that this is not replicated in the Romance languages (c) is linked to the development of Negative Concord. The categories used in the semantic map are not sufficiently fine-grained to address crosslinguistic variation in this domain, especially because of the complex interactions with the syntax. Indefinites taking narrow scope with respect to a sentential negation operator (Haspelmath’s direct negation function) can be of different types, depending on the overall syntactic properties of the system of negation in a given language, as Haspelmath himself discusses at length (Haspelmath : ch. ; cf. also Haspelmath ). Also in this area the debate on how to best model variation is ongoing (for recent overviews see Giannakidou , Penka , Zeijlstra ). These problems will be discussed in detail in chapters  and , but from now on we will already need a terminology that differentiates further. Following much literature on the topic, I distinguish Haspelmath’s direct negation indefinites in three classes: (a) Negative Indefinites (NIs); (b) n-words; (c) negative-polarity items (NPIs). Negative Indefinites (NIs) always contribute a negative meaning and are able to negate the clause with no further negative elements present; if another negative element is present, a double-negation reading with two logical negations arises. A language with NIs is German, cf. ():

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



() German: a. Das interessiert niemanden this interest:sg nobody:acc ‘No one cares about this’ b. Niemand hat nichts gegessen nobody:nom have:sg nothing:acc eaten ‘No one has eaten nothing’ = Everyone has eaten something So-called n-words are indefinites whose negative import depends on the syntactic structure they are found in. In some structural contexts they require the presence of a negative marker (an adverb or particle expressing sentential negation); in other structural contexts they can express negation by themselves. Multiple n-words can cooccur in a single-negation reading. A language with n-words is Italian, cf. (): () Italian: a. Questo non interessa a nessuno this not interest:sg to nobody ‘No one cares about this’ b. Nessuno ha mangiato niente nobody have:sg eaten nothing ‘No one has eaten anything’ (under plain intonation) Negative-polarity items (NPIs) are elements that are never semantically negative per se, but may occur only in a downward-entailing context, i.e., under direct negation, but also in other environments that license similar logical and pragmatic inferences.13 Polarity sensitivity is independent of syntactic category and is not limited to determiners (cf. verbs like Dutch hoeven ‘need’ or entire idioms like lift a finger). An example of a negative-polarity indefinite determiner is English anything in (): () No one has eaten anything The property of a downward-entailing context is to reverse the normal direction of entailments, licensing inferences from supersets to subsets. The licensing of NPIs has been connected to the logical property of downward-entailingness since the work of Fauconnier () and Ladusaw (). This licensing is usually considered to be pragmatic in nature: NPIs will be felicitous only in downward-entailing contexts because only there they will yield an informationally adequate contribution, thanks to the reversal of the relation of strength or informativeness triggered by downwardentailing operators.

13 Contexts where NPIs are felicitous include ‘indirect negation’ contexts (non-negative clauses dependent on negated matrix clauses), questions, the antecedent of conditionals, complement clauses of adversative predicates (like forbid or fear), before-clauses, the restriction of universal quantifiers and of proportional quantifiers like few or most, the standard of comparison, the restriction of superlative noun phrases, some focus-sensitive particles (e.g. only, exceptive negation like French ne . . . que, etc).

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

There is, however, an ongoing debate on whether NPIs in the wrong context give rise to an infelicitous statement (which would be expected under a pragmatic explanation) or to ungrammaticality proper. In some syntactic analyses polarity sensitivity has been modeled as a structural dependence by means of a formal feature ([± aff] or [± pol]), which yields ungrammaticality if not licensed by a proper operator. A further debate concerns the precise definition of the semantic contexts in which NPIs occur. Von Fintel () has put forward the notion of ‘Strawson downward entailingness’ to account for some licensing contexts, like the antecedent of conditionals or the scope of only, which are not strictly downward-entailing: Strawson downward-entailingness also considers presuppositions and conventional implicatures associated with the licensing element, ensuring that they are satisfied, in order to derive the correct licensing conditions (see Gajewski : – for discussion). A broader notion of nonveridicality has been proposed by Giannakidou () (see also Giannakidou  for a comparison of different approaches). Research has also shown that some NPIs (weak NPIs) occur in a broader range of contexts than others (strong NPI). Traditionally, this difference has been attributed to the hierarchy of negative expressions proposed by Zwarts (): strong NPIs would be felicitous in the subset of downward-entailing contexts characterized by anti-additivity (comprising at least sentential negation and negated existentials).14 While NPIs can occur in all types of negation systems, the occurrence of NIs and n-words is dependent on the overall nature of the negation system. In chapter  we will see the determining factors of this variation in detail. For the moment I just introduce the terminology in a purely descriptive fashion. Systems with NIs are called Double Negation systems (DN systems), whereas systems with n-words are called Negative Concord systems (NC systems); the latter are further distinguished into strict and non-strict Negative Concord systems, depending on whether the n-word always cooccurs with the negative marker (strict) or whether an asymmetry between the areas of the clause before and after the inflected verb (which I will indicate as ‘Infl’) is observed (non-strict) (cf. (); see Giannakidou  for these categories). () Strict versus non-strict Negative Concord a. Nimeni nu a cump˘arat cartea Romanian (strict NC) nobody not have:sg bought book-the ‘Nobody bought the book’ b. Nessuno ha comprato il libro Italian (non-strict NC) nobody have:sg bought the book ‘Nobody bought the book’ The history of Romance languages clearly shows that there is a diachronic connection between NPIs and n-words; however its interpretation is not straightforward. One task in chapter  will be to distinguish between an NPI- and an n-word status in the

14 Gajewski (), followed by Chierchia () has proposed a different explanation based on the role of non-truth-conditional aspects of the meaning of the licensors.

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



case of Old Romance negatively marked indefinites (which I will call nec-words, based on the etymology of the negative morpheme they contain). In what follows, I will designate as ‘negation system’ the complex of elements that contributes to express sentential negation in a language, minimally comprising the negative marker and the indefinites scoping below the negative operator. .. Constraints on domains of quantification In investigating the classes introduced in .., I pursue an approach according to which a crucial dimension of variation in the system of indefinites is determined by conditions that different indefinites impose on their quantificational domains and on their alternatives. These conditions change in time in a principled way, through minimal changes in the lexical makeup, obeying functional pressures related to expressivity and informativeness, and ‘incorporating’ originally contextual conditions in the lexical entry of the indefinite. As introduced in §..., some current approaches interpret specific indefinites and free-choice indefinites (the two extremes in Haspelmath’s map) as opposite ends of the same continuum: specificity arises from contextual extreme domain narrowing (i.e., narrowing the restrictor set to a singleton, Schwarzschild ), and free choice results from domain widening (Kadmon and Landman ). Similarly, the interpretation of negative-polarity indefinites (another extreme in Haspelmath’s map) has also been assumed to be connected to domain widening (cf. again Kadmon and Landman ). A diachronic exploitation of this dimension of variation looks extremely promising. Let us see, then, in some more detail how this dimension of variation can be encoded in the lexical representation of quantificational determiners. Conditions on domains emerge as the attempt to formalize the ubiquitous context dependence of quantificational expressions (but also of definite descriptions, cf. Heim : –). In general, the domain, or restriction, of a quantificational determiner is determined by its syntactic complement (NP) and also by contextually salient information. Current approaches propose a formal representation for the role of this contextual information. This can be done by assuming covert domain restrictors.15 Covert restrictors have been modeled in various ways. Here it will be sufficient to assume that a covert restriction is an additional argument of the quantifier, i.e. an additional set, or property, intersecting with the set, or property, expressed by the overt NP restrictor (von Fintel , Stanley and Szabó ). The covert restrictor set is built by assigning a value to a contextual variable, which can be free (and take value from the discourse context) or bound by surrounding operators.16 Alternatively, covert domain restrictions have been represented as functions from sets to subsets, e.g. from the extension of the overt NP restrictor to a contextually relevant subset 15 For an introductory discussion of this issue on the basis of Latin examples see Devine and Stephens (: –). 16 In situation-based semantics this is modeled by assuming that determiners are associated with a free situation variable, which may either take value from the contextual situation or be bound by a structurally superior quantifier. How to precisely model this is a topic of current discussion: cf. Kratzer () for a comparison between covert domain variables and situation arguments, and the question whether they are associated with the determiner or rather with the lexical restriction.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

thereof (such a subset selection function is adopted by Alonso-Ovalle and MenéndezBenito , as we will see in chapter ). Covert domain restrictors can be subject to constraints that establish the characteristics of the domain of quantification. Constraints on covert domain restrictors have figured as a privileged way to model not only contextual, but also structural dependence in current research. Determiners show clear preferences as to the characteristics taken by their covert restrictors. Definite nominals, for instance, take restrictors with the narrowest possible extension, optimally a singleton (Heim : , who attributes this behavior to a pragmatic pressure toward informativeness). As mentioned, a similar restriction to singleton sets has been proposed in order to treat specific readings of indefinite noun phrases: Schwarzschild () proposes that the indefinite article, which is taken to be an existential quantifier, can in certain cases be covertly restricted to apply to the only individual that the speaker has in mind in the context of utterance. A fundamental difference between definites and specific indefinites in this respect is that, with definites, the singleton set is contextually given, i.e. recoverable and known to be a singleton by both the speaker and the hearer, whereas with specific indefinites it has to be newly introduced in the discourse. Therefore the hearer, who is normally able to contextually retrieve the implicit restriction, cannot do so for this type of specific reading (cf. Heim : ). In fact, according to Schwarzschild (), with specific indefinites a ‘Privacy Principle’ is at work, causing an asymmetry between the speaker and the hearer: the information implicitly restricting the set is asymmetrically available to the speaker (or another epistemic agent), but not to the hearer, hence, indefiniteness arises as ‘novelty’.17 Pursuing this line of reasoning further, if the size of the quantificational domain is a dimension of variation (for instance, differentiating between plain and specific indefinites), we expect, in principle, this dimension of variation to be fully exploited in language systems; that is, we expect certain determiners to show a complementary preference for widened domain. These determiners will be felicitous only in contexts where a wide domain is pragmatically motivated. This has been argued to be the case with at least some types of negative-polarity items (Kadmon and Landman ) and free-choice determiners (Kratzer and Shimoyama , Chierchia ). It is well known that these elements are dependent on the surrounding semantic and structural context for their licensing: they need an environment where the inference they license can lead to an informationally significant contribution. In sum, a fundamental dimension of variation among natural language determiners is tackled here by assuming that the lexical entry of determiners contains instructions on how to structure their quantificational domain (cf. Kratzer  for the notion of ‘domain shifters’ applied to indefinites). 17 Note that, since the restriction is a singleton set and the choice of the referent for the indefinite nominal phrase is univocal, scope is neutralized; thus, under this analysis, the exceptional wide scope that is observed with specific indefinites when they scope out of islands would be an ‘illusion’, owing to scope neutralization.

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



As we will see more in detail in chapter , also the epistemic effect of Haspelmath’s ‘specific unknown’ indefinites has been interpreted as (a conversational implicature) resulting from the constraints that epistemic indefinites impose on their quantificational domain. Alonso-Ovalle and Menéndez-Benito (, ), elaborating on Kratzer and Shimoyama’s () treatment of German irgendein ‘some or other / any’, define Spanish algún as an ‘anti-singleton’ indefinite, whose domain of quantification cannot be narrowed down to an individual or singleton set. A sentence like () is uttered when more than one part of the house is an epistemic possibility for the speaker, and would be infelicitous if the part where Juan in fact is had been already established in the Common Ground. It is possible, though, to exclude some possibilities: () is felicitous even if it has been established that some rooms are excluded as epistemic alternatives. () Spanish (Alonso-Ovalle and Menéndez-Benito : –) Juan puede estar en alguna parte de la casa ‘Juan may be in a part of the house’ The fact that some alternatives can already be discarded shows that the domain is not maximally widened. This latter feature represents the difference with respect to the quantificationally stronger free-choice effect, where all members of the domain, even the most marginal ones, are an option (e.g. Spanish cualquiera, Italian qualcunque), cf. Kadmon and Landman (). This is the reason why Alonso-Ovalle and MenéndezBenito () speak of ‘minimal domain widening’.18 Putting these insights together, it is possible to recognize a continuum of operations on quantificational domains, encoded by different classes of indefinites, as schematically represented in () and (): () extreme domain narrowing > minimal domain widening > domain widening () continuum of operations on quantificational domains

≥2

1

extreme domain narrowing (specific) minimal domain widening (epistemic) domain widening (free-choice)

Note that this approach to variation is not necessarily connected to the assumption of a quantificational format for indefinites. In fact, Kadmon and Landman’s () 18 As we will see in chapter , in Alonso-Ovalle and Menéndez-Benito (: –) the ‘anti-singleton’ constraint is modeled as a presupposition constraining the possible values of a subset selection function.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

proposal is formulated in a way that is compatible with either the quantifier- or the variable-based approach. In general, variable-based approaches assume rich lexical specifications for indefinites, including presuppositions constraining their use. For instance, Giannakidou () analyzes free-choice and polarity-sensitive indefinites as introducing variables endowed with a sensitivity feature, specifying the type of dependencies they may undergo. Farkas () develops a whole typology of indefinites based on classes of constraints that indefinites impose on the variables they introduce. The dimension of variation connected to the constraints on domain restriction certainly does not exhaust the expressive possibilities of indefinites. Nonetheless, it suggests a cline along which indefinites are organized, depending on the domain shift they impose, and along which they may change (accounting for directionality of diachronic processes). This cline emerges as a determining dimension of change in my case studies. In the case of epistemic indefinites, constraints on quantificational domains impose dependence relations with surrounding operators, and, thus, select the contexts where these indefinites are found. In the case of indefinites in the scope of negation, on the one hand we see new items being recruited into the negation system on the basis of their semantic potential to act as domain wideners; on the other hand, the bleaching of the domain widening component may lead to the reanalysis of an originally semanticpragmatic licensing dependence as due to a syntactic formal feature. .. Indefinites and the systematicity of semantic change In §.. I introduced a fundamental dimension of variation in the functional space of indefinites, and I argued that it is possible to project this dimension of variation diachronically, as a possible direction for systematic semantic change. In this section, I further support the claim that semantic change involves systematic processes, which can receive a formal analysis, and that that the comparative study of indefinites can substantially contribute to the broader enterprise of understanding the determinants and mechanisms of change at the syntax-semantics interface. Starting with Lightfoot’s () groundbreaking monograph, formal diachronic syntax has been seeking principled, cognitively grounded explanations for syntactic change. This has led to the discovery of generalizations over the format of syntactic change, which made an account of directionality phenomena possible. The main reason for the success of formal diachronic syntax in this respect has been the availability of a formal theory of linguistic variation, the Principles & Parameters Theory, built upon rich (micro)comparative evidence. Grammaticalization research at the syntax-semantics interface has shown that also the semantics of ‘function words’ uncovers important, although still underresearched, diachronic tendencies, whose degree of systematicity appears to be comparable to what is observed in phonology and morphosyntax.19

19 Eckardt () represents a foundational work in this respect. For discussion of the emerging field of formal diachronic semantics see Deo () and Gianollo et al. ().

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



The domain of indefinites, with the rich amount of variation that we just surveyed, represents a privileged area where the study of individual items of the functional lexicon can contribute to our more general understanding of semantic change. One of the main difficulties to overcome when pursuing a systematic investigation of semantic change has to do with the scarcity of truly comparative models of crosslinguistic variation in formal semantics. As discussed, restrictions on variation (what can vary across languages; what remains constant; how values for variation points cluster) are restrictions on change as well, and implicational relations among linguistic features can favor or block a chain reaction, depending on their nature. Thus, a theoretically adequate model of variation is a prerequisite for restrictive theories of change. As discussed, a challenge for formal semantics is represented by the fact that grammaticalization clines have usually been formulated in terms of functions (form– meaning, form–context correspondences), and not in terms of formal meanings interacting with contexts. A more fine-grained analysis of the compositional elements involved would help distinguish between truth-conditional and pragmatic meaning components here, i.e., between meaning that has to be present independent of context, and meaning that results from pragmatic implicatures, accommodation, interplay with the broader clausal meaning, among other factors. An explanatory account of semantic change involves three consecutive steps (cf. also Condoravdi and Deo ): () a. Single out the logical relation between ingredients of meaning that is responsible for the directionality observed in ‘semantic cycles’; in other words, root the potentiality of the (systematic) diachronic connection between categories in the synchronic relation between their meaning parts. b. Identify the mechanism driving the diachronic connection, i.e. the form that the change takes and its cognitive motivation. c. Address the problem of local causes, i.e. of reconstructing the trigger that started the mechanism in the particular diachronic setting. Systematicity emerges as the interplay of (a) and (b), since both are manifestations of the universal architecture of language. Systematicity in syntactic and semantic change has two main general manifestations. On the one hand, a change targeting a specific grammatical domain may repeat itself within languages (cyclicity proper) and across languages (crosslinguistic frequency of attestation); a further dimension in this respect is possibly the (uni)directionality of such changes, along a regular path. On the other hand, different changes across grammatical domains undergo similar stages, and can be shown to be subject to the same mechanisms, i.e., to the same formal constraints and evolutionary tendencies. Below I brief review these aspects of systematicity, connecting them to the study of indefinites and introducing notions that will be useful for the case studies. ... Cycles and directionality As seen in §.., semantic maps encode diachronic observations, since the implicational relations connecting functions in the map are

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

also expected to constrain change across adjacent functions. Cyclical change, in the sense seen above, has two main manifestations in the empirical domain of indefinites: () a. a systematic lexicalization of complex syntactic expressions into simplex morphological units (words): cf. Lat. aliquis ‘some’ < alius quis ‘some other’; Lat. qui-vis ‘whoever’ < ‘who you want’, Cat. qual-se-vol ‘whoever’ < ‘who one wants’. This dimension captures the fact that crosslinguistically recurrent sources of grammaticalization are observed with indefinites; b. a shift along a continuum of meanings / functions: this aspect is represented in semantic maps, and will be the main focus of the case studies: I address the so-called Quantifier Cycle in chapters  and , and the Negative Cycle in chapters  and . As for (a), the fact that some of the recurrent sources of indefinites are composite syntactic expressions explains the observation that indefinites tend to be morphologically complex, some subclasses possibly more than others. Their complex internal structure sometimes is not synchronically transparent to the speaker but is patent in the language history. In other cases it is still transparent in its constituting parts, to the extent that their individual semantic and syntactic contributions can be argued to be clearly detectable in the synchronic grammar (cf. e.g. Martí , Zimmermann ). As Szabolcsi (: ) states, if facts concerning the internal makeup of quantificational elements ‘prove to be sufficiently systematic, we want them to be accounted for by compositional semantics and not to appear as mere etymological curiosities’.20 Haspelmath (: ) also remarks that, crosslinguistically, the origin of indefinite markers is often still transparent, although the meaning of the resulting indefinite pronoun can never be determined purely compositionally. The transparency of the markers is due to their generally recent grammaticalization: ‘indefinite pronouns have a relatively short lifespan and are diachronically quite unstable’. Concerning (b), cycles are intriguing for linguistic theory because they point not only to a deep grammatical and cognitive connection among the linguistic categories involved, but also to factors of language use that trigger systematic renewal in certain areas. The fact that some phenomena are clearly more prone to cyclical renewal than others—think of negation—calls for an explanation. For instance, as we will see in chapters  and , Jespersen’s Cycle has been argued to be due to the pragmatic tendency to strengthen negation, and to ensuing ‘inflationary effects’ that lead to a loss of emphasis on the originally optional reinforcer and to its reanalysis as a plain marker of negation (Schwegler : ch. , Eckardt , Eckardt , Kiparsky and Condoravdi ). Also acquisitional strategies may be argued to play a role in this respect. For example, according to Willis et al. (a: ), a possible motivation for the Negative

20 Szabolcsi’s example concerns the connection between existentials and disjunction, universals and conjunction, cf. further Szabolcsi ().

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



Cycle (the cycle leading from a polarity item to a negative indefinite or n-word) would be children’s failure to acquire some contexts of occurrence of NPIs, owing to insufficient frequency of evidence. Martins () recognizes the existence of a diachronic trajectory from weak to strong polarity items in Romance: she attributes it to a restriction of licensing polar contexts, i.e., to a reduction in polar versatility. Martins proposes a system of syntactic features encoding interface-relevant semantic information on the lexical item, and models the process of reduction of polarity contexts as a reduction in featural underspecification. Such reduction would be motivated in an acquisitional perspective by assuming that less salient contexts of licensing become not robust enough to be successfully acquired by new generations (Martins : ). These previously available contexts become less salient in language use mainly due to ‘competition between negative and positive indefinites in non-negative modal contexts’ (Martins : ). Scenarios like those reconstructed above, where the licensing contexts become more restricted, comply with the Subset Principle (Berwick ). However, there are also cases where the set of licensing contexts is broadened: as we will see in chapters  and , in its development aliquis expands into weak- and strong-polarity contexts. In cases like this acquisition seems to lead to overgeneralization: it is possible that this is due to further cooccurring factors, such as changes affecting paradigmatic complementarity among indefinites (e.g. in Latin between aliquis and quis) and removing morphological blocking (cf. Pereltsvaig  and Penka : , –, and further §...). One issue that cyclical change raises for theoretical linguistics concerns the source of the observed directionality, which, as mentioned, calls for a better understanding of the semantic categories involved. A further question arising in this context concerns the fact that sometimes the observed change is unidirectional: shifts between categories are observed asymmetrically only from Category A to Category B and never the other way round. For indefinites, Haspelmath (: –) proposes that the factor governing unidirectionality is ‘weakening’: right-hand functions on the map are semantically ‘stronger’ than functions to the left; change consists in loss of features, and proceeds unidirectionally from right to left. Weakening can be declined in different fashions in Haspelmath’s system, e.g. as loss of focusing and scalarity, or as loss of nonspecificity or unknownness. Clearly, the interpretation of unidirectionality depends strongly on the adopted system of features and on the assumed mechanism of change. Jäger (), for instance, working from different assumptions, claims that change with indefinites can go either way: in her system, based on the two features [±affective], [±negative], there is either addition of a feature (NPI > NI, PPI > NPI) or loss of a feature (NI > NPI, or NPI > PPI) (where PPI = positive-polarity item; in fact, in some of the cases discussed the indefinite is a polarity-neutral, plain item). () a. PPI > NPI > NI e.g. Latin aliquis > Italian alcuno > French aucun b. NPI > PPI e.g. Old High German io mer ‘ever more’ > German immer ‘always’

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

Also Labelle and Espinal () and Larrivée (a), in their diachronic studies of French indefinites interacting with negation, conclude that change is not unidirectional in this domain. In chapters  and , we will see that Haspelmath’s assumptions on unidirectionality are challenged by the history of aliquis (cf. a), where the development proceeds through extension from the left-hand side to the right-hand side of the map. ... Mechanisms of semantic change Two main mechanisms of semantic change have been invoked in studies of processes targeting functional items of the lexicon: ()

a. bleaching: loss of descriptive content; generalization through loss of contextual constraints; retention, or new association, with logical content b. conventionalization of pragmatic implicatures: lexicalization of previously cancelable inferences

On the one hand, semantic change has been argued to involve an impoverishment of the semantic content in terms of bleaching (cf. e.g. von Fintel ). On the other hand, there are cases of semantic change showing enrichment of meaning through conventionalization of pragmatic implicatures (cf. Traugott and Dasher , Eckardt ). The role of bleaching in the diachronic development of determiners has been investigated more often (e.g., in the development from demonstratives to definite determiners, involving bleaching of the deictic component and, thus, loss of a contextual constraint). The role of conventionalization of pragmatic implicatures, instead, remains to a large extent unexplored in this area, but is widely recognized in other empirical domains and has led to important generalizations on semantic change: cf. the Invited Inferencing Theory of semantic change (e.g., Traugott and König , Traugott and Dasher ), the pragmatic implicature approach of Levinson (), the principle of Avoid Pragmatic Overload in Eckardt (). Aguilar-Guevara et al. () have tried to apply these considerations to the diachronic study of indefinites: their main hypothesis is that ‘different indefinite forms have emerged as a result of a process of conventionalization (or fossilization) of an originally pragmatic inference’ (Aguilar-Guevara et al. : –). For instance, in () the ignorance component, that arises as an implicature with jemand ‘someone’ in (a), is part of the conventional meaning of irgendjemand in (b), a more specified form developed through lexicalization. () German, from Aguilar-Guevara et al. (: ) a. Jemand hat angerufen Conventional meaning: Somebody called Ignorance implicature: The speaker does not know who b. Irgendjemand hat angerufen Conventional meaning: Somebody called and the speaker does not know who The fact that pragmatic inferences arise systematically in language use reveals one dimension of the context-dependence of change: change in word meaning is

OUP CORRECTED PROOF – FINAL, //, SPi

Dimensions of variation and change



inextricably tied to change in sentence meaning, and influenced by discourse meaning. As Eckardt (: ) puts it, during the process of semantic reanalysis ‘the salient overall conveyed information remains the same, but is composed in a different manner. What may have previously been in part assertion, in part implication, turns entirely into a literal assertion after reanalysis’. The conventionalization of pragmatic inferences interacts, thus, with compositional principles, which ensure the adaptability of linguistic means to new communicative needs and inextricably tie word meaning to sentence meaning (cf. also Gergel ). A further factor to be considered in this respect concerns what I will call ‘paradigmatic effects’: if indefinites are really organized into series and share in principled ways a universally established semantic space (represented by the semantic map), they build a paradigm, and are expected to give rise to competition and blocking effects among series (e.g. between any and some, Lat. aliquis and quis).21 These, in turn, may be of diachronic significance. Blocking is explicitly foreseen in systems based on formal features (e.g. Martins , Jäger , Penka ), which obey the Elsewhere Condition (Kiparsky ): featurally underspecified items are blocked by other elements more richly specified for a given context. For instance, the distribution of n-words and NPIs in some Slavic languages has been variously explained as resulting from morphological blocking (Pereltsvaig , Penka : –): although NPIs as such are compatible with the direct negation context, n-words are more specified for it and therefore block NPIs. This causes what has been named a ‘bagel’ distribution (Pereltsvaig , Deni´c ), where the core of the category is blocked for the element that occurs elsewhere in the category, owing to the existence in the lexicon of a more specified candidate for the core. In the case of complementary distribution of NPIs and n-words, the category is represented by the downward-entailing contexts, and its core is represented by direct negation contexts. In (), from Serbian, the NPI i- series (on which see also Progovac ) is blocked under the scope of direct negation by the more specified ni- series: () Nisam upoznala nikoga / *ikoga neg.aux:sg meet anybody / anybody ‘I haven’t met anybody’

Serbian (from Deni´c )

Given paradigmatic effects, changes in the featural specification of a given item (according to the mechanisms of change seen above) may lead to more global reorganizations of the system, owing to removal or creation of a competition scenario. In this respect, the notion of ‘series’ seems to be quite crucial: we can see the interaction among series as the source of part of the meaning of the series themselves. Also in this respect, changes in the system balance, such as the loss or addition of a series, are expected to lead to meaning changes also for the neighboring series. That

21 Very informally, I consider a paradigmatic relation to obtain among members of a certain class competing for a certain basic function. In this sense, also an informativeness scale (on which see further on in this section) is a representation of a paradigmatic relation.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

is, if the pool of competitors varies / changes, the set of admissible inferences may in turn vary / change. Referential and quantificational expressions appear to be ordered along scales of informational strength. Part of their meaning contribution, thus, originates only indirectly from the lexical specification, since it is due to pragmatic reasoning: the expression is considered not in isolation, but rather in comparison to its alternatives. If indefinite series are really organized into a functionally coherent space (the space represented by the semantic map), we expect them to trigger this sort of scalar reasoning. In turn, in view of what we saw above on the role of conventionalization, we may expect inferences arising this way to play a role in the diachronic process of meaning change. As the case studies will show, with indefinites, apparently, we have to deal with two kinds of effects. On the one hand, at least some of these elements participate (also with other non-indefinite determiners, cf. Heim ) in scales based on strength of informativeness. On the other hand, the series into which they are organized compete also with respect to their morphosyntactic features, and may thus trigger blocking in certain contexts. This may lead to a situation where both item A and item B would semantically be compatible with a certain environment, but differ morphosyntactically in a way that makes, say, A featurally more specified than B for the given environment. We will thus have to distinguish at least two types of blocking: (i) morphosyntactic blocking, depending on the specification of formal features; (ii) pragmatic blocking, primarily due to the Maxim of Quantity.22

. Corpora, methods, and periodization Before moving to the case studies, in this last introductory section I describe the main methodological decisions I took and the textual material used as empirical database. .. Corpora and methods for data collection The research presented in this book is based on corpus work that has been conducted on large electronic databases of Latin, Old French, and Old Italian, as well as on more fine-grained annotation work on some individual texts. This twofold strategy was primarily motivated by the need to, first, find out distributional tendencies on material of manageable extension and appropriate thematic cohesion, in order to control for discourse factors, and, subsequently, test such generalizations on a broader set of material.

22 Concerning (ii), I consider also cases related to the ‘Maximize Presupposition’ principle to fall into this class: here what counts is not the strength of the assertion resulting from the use of a certain alternative, but rather the strength of the presuppositions associated with such an assertion. The ‘Maximize Presupposition’ principle had been originally proposed by Heim () as a global principle applying to discourse units, but it is nowadays argued to apply more locally in the presence of presupposition triggers. It forces the choice of the formal alternative that fulfills the greatest number of presuppositions possible in the context, i.e., the choice of the presuppositionally stronger alternative.

OUP CORRECTED PROOF – FINAL, //, SPi

Corpora, methods, and periodization



For Latin I primarily used electronic texts from the LLT-A and B (Brepols’ Library of Latin Texts, Series A and B, http://apps.brepolis.net/Brepolis/Portal/default.aspx), the LASLA Opera Latina database (http://web.philo.ulg.ac.be/lasla/description-operalatina/), and the PROIEL Corpus (Pragmatic Resources in Old Indo-European Languages, http://proiel.github.io). The LLT-A and B are large, non-annotated databases that allow for various types of complex queries on strings of text. The LASLA database is annotated for parts of speech and inflectional categories. PROIEL is annotated for parts of speech, syntactic dependencies, and information structure. In order to counterbalance well-known stylistic tendencies of certain genres, which have a particularly strong impact on strategies of reference and quantification, I avoided texts in verse (with the partial exception of Plautus) and I tried to cover different types of prose (technical treatises, philosophical writings, historiography, public and private letters, biblical and other Christian texts).23 For my data on Old French I worked on selected texts from the MCVF corpus (http://www.voies.uottawa.ca/corpus_pg_en.html) and from the Syntactic Reference Corpus of Medieval French (http://srcmf.org/). Old Italian data come from the Corpus OVI dell’italiano antico (http://www.ovi. cnr.it/index.php/it/?page=banchedati). All these corpora are lemmatized (the Old French corpora are also syntactically annotated) and allow for targeted queries over a large amount of material. For Old French, due to the nature of my research questions, I privileged evidence from the most archaic stage (from the ninth to the twelfth century). However, occasionally, I also discuss data from later stages. For Old Italian I investigated the Old Tuscan variety until the first half of the fourteenth century. For the Modern Romance languages I developed a questionnaire in order to comparatively address the problems presented in chapter . The questionnaire was designed for informants with a solid linguistic background and was administered sometimes in person, if not possible remotely via e-mail and through the web. For Modern Italian I also used the CODIS / CORIS Corpus di italiano scritto (http:// corpora.dslo.unibo.it/coris_ita.html) and the BADIP BAnca Dati dell’Italiano Parlato (http://badip.uni-graz.at/en/description).

23 The English translations of the Latin examples follow as close as possible those provided in the LOEB editions when available (http://www.loebclassics.com), but have been freely modified (especially to be rendered more literally) if necessary. For the Bible I have adopted and adapted the New International Version. The abbreviations for Latin authors and texts follow the standard ones used in the Thesaurus Linguae Latinae (online version: TLL - and for the Index librorum scriptorum inscriptionum also http:// www.thesaurus.badw.de/en/user-tools/index.html). I made an exception for Caesar, where I used the more practical abbreviations BG (= Commentarii de bello Gallico) and BC (= Commentarii de bello civili), and for the Itinerarium Egeriae (Itin. Eg.); lesser-known texts and authors have been cited in full. Vowel quantity has been systematically indicated only on monosyllables, because of its potential morphosyntactic significance. Glosses have been kept to a minimum to ensure legibility for the examples: I only indicate case for nominals, number-person for finite forms of verbs (occasionally mood when relevant for the phenomenon at hand), mood for nonfinite forms. The abbreviations follow the Leipzig Glossing Rules (https://www.eva.mpg. de/lingua/resources/glossing-rules.php). For ease of reference, they have been included in the List of Abbreviations at the start of the volume.

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

In general, I investigated individual indefinites by annotating morphological factors (number, gender); syntactic factors internal to the nominal phrase (pronominal vs determiner use; in the latter case, position with respect to head noun); external syntactic factors, distinguishing grammatical role in the clause (subject, object, use in prepositional phrases); type of clause (matrix versus embedded; in the latter case, type of embedded clause); presence of other operators (negation, modal operators, conditional). For Latin quidam and aliquis also discourse factors were annotated, like the presence of an explicit restricting set in the form of a partitive phrase, the mention of the method of identification of the referent, anaphorical relations in later discourse. A satisfactory semantic annotation in the domain of referential and quantificational categories has to deal with pervasive context-dependence. This raises hard challenges (cf. Aguilar-Guevara et al. , Crisma ), but the broader consideration of contextual factors, in terms both of discourse context and of immediate structural context, cannot be dispensed with in historical research, which has to base its conclusions exclusively on naturalistic data. In naturalistic data optionality and multifunctionality often obscure distributional generalizations, but may be significant in their own right. Optionality may be a sign for language change, i.e., be a temporary variation connected to gradualness of meaning extension into new functions, but it may also reflect principled conditions on use: what might seem optional may in fact turn out to be governed by more or less subtle contextual factors. Multifunctionality involves the presence of many ‘noncore’ readings, which are quite frequent and yet difficult to classify. This was the case, for instance, with some scalar readings that I will discuss for aliquis in chapter . I considered them anyway, even if they did not fit in the ‘bigger picture’, since they are starting to be given more theoretical consideration (cf. e.g. Hinterwimmer and Umbach ) and their relevance may appear clearer in the future. Also given the various sources of indeterminacy in annotation (last but not least the objective difficulty of assessing certain readings in historical texts), I provided quantitative evidence only when it concerned distributional factors that were safe to assess (e.g. word order or co-occurrence with other overt elements) and when it was relevant for the research question. In this respect, the data used in this work are mostly of a qualitative nature: in order to reach my conclusions on a number of phenomena I first of all needed to show that a certain combination (e.g., the occurrence of an indefinite determiner in a certain structural context) was possible, independently of its frequency. .. Periodization for Latin The period referred to as Classical Latin in this work corresponds, roughly, to the ‘Antiquitas’ period in the LLT-A database (until the end of the second century ce). The use found in Cicero (first century bce) has been considered as representative of the Classical Latin norm. Occasionally, I further distinguish the subperiods Early (or Archaic) Latin (until the second century bce) and Post-Classical (or Imperial) Latin (from the first century ce until the end of the second century ce).

OUP CORRECTED PROOF – FINAL, //, SPi

Corpora, methods, and periodization



Given the nature of the documentation and the complex stylistic layering of Latin texts, it is impossible to apply simple chronological criteria in studying diachronic processes from Latin to Romance. It is well known that after the standardization of a classical norm—a process of ‘crystallization’ completed by the first century ce, on which see Rosén ()—Latin literary production, which represents the greatest part of our sources, stops reflecting the variation and the evolution of the spoken language, with very few exceptions. It is, therefore, often impossible to follow the step-by-step development of certain phenomena, since, despite the uninterrupted transmission of written Latin, there are factual gaps in our documentation: many late texts follow the classical norm and are therefore unreliable witnesses of language change. Besides chronological issues, the nature of the documentation causes difficulties also when trying to detect diatopic and diastratic variation in Latin: see Adams (, ), Wright (, ) for painstaking work in this direction. Moreover, the Latin texts I studied represent very different registers, and it is well known that for Latin this dimension may influence the distribution of linguistic phenomena much more incisively than chronology itself: see Dickey and Chahoud () for a recent insightful evaluation of register in Latin. Given the nature of the phenomena investigated in this work, I could largely abstract away from this dimension, but I will frequently make reference to stylistic factors that may have influenced the distribution of certain items or constructions. In order to ensure a reconstruction of the diachronic processes affecting indefinites from Latin to Romance despite the objective difficulties seen above, I carefully selected the documents, singling out texts that exhibit more colloquial features. I made large use of Early Latin documentation, especially Plautus’ texts, in chapters  and , when trying to reach an accurate description of the point of departure for many phenomena I investigate. As for Late Latin, I concentrated my attention in particular on texts dating to the third and fourth centuries ce. When relevant and possible, I extended my consideration to later texts, as e.g. the works of Gregory of Tours (sixth century ce) and Merovingian Lives of the Saints (sixth-eighth century ce). The predominance of Christian writings in this later period is motivated by the necessity to look for genuine representatives of the ongoing linguistic developments. The specific ideological setting in which Christian writings originate legitimates and encourages a decisive detachment from the classical norm. Therefore they are overall reliable witnesses of processes of change (cf. Adams , Clackson and Horrocks : – for an evaluation). Biblical traslations have been included in the corpus, on the basis of the fact that, although they are undoubtedly strongly influenced by the source language, they have been shown to conform to the same grammatical system that governs contemporary texts of comparable register (cf. Gianollo  for discussion). I did not necessarily aim for quantitative representativeness (in terms, e.g., of statistical significance) of my data for a given stage. I rather investigated the grammatical system underlying the individual variety represented by each text. In light of the complex textual tradition of some of these texts, which result from the contributions

OUP CORRECTED PROOF – FINAL, //, SPi



The grammar of indefinites: Functions, variation, and change

of several individuals, this may of course be an oversimplification to some extent (but see Gianollo ,  for the reliability of Latin biblical translations). Of course, phenomena that we observe in Late Latin texts may well have originated earlier in the spoken language, and have been prevented from appearing in the written documentation by prescriptive standardization. Therefore the dating of many changes (especially of those that may plausibly be subject to prescriptive pressure) can only be approximate.

OUP CORRECTED PROOF – FINAL, //, SPi

2 Specific and epistemic indefinites in Latin . Introduction .. Topics and aims In this chapter I discuss the Latin indefinites at the left-hand extreme of Haspelmath’s () map: quidam ‘a certain’ and aliquis ‘some (or other)’. This work will uncover the Latin roots of the principled variation observed in Romance and will be a necessary prerequisite for the investigation of the Romance developments in chapter . In discussing Haspelmath’s notion of specificity and previous work on aliquis I address important theoretical questions concerning the nature of the meanings expressed by indefinite nominal phrases. I will reach a definition of the semantic status of aliquis that departs from Haspelmath’s classification and I will support it on the basis of its empirical adequacy and its ability to make sense of the multifaceted developments observed in Romance. The two indefinites in the left-hand space of Haspelmath’s map for Latin (cf. § ..), quidam and aliquis, have experienced a very different fate in the historical process taking them from Latin to the Romance languages: while quidam is not productively continued in any Romance variety, aliquis has been extremely successful from a diachronic perspective. The different space dedicated to them here mirrors this fact, as well as my research aims: I deal with quidam rather briefly, and mainly in order to empirically support my discussion on specificity. The rest of the chapter is reserved to aliquis in Classical and Late Latin. The Romance developments of the continuations of aliquis (French aucun, Italian alcuno, Spanish algún, Catalan algun, Portuguese algum) will be the topic of chapter . This chapter concentrates on two aspects: Haspelmath’s () specific known versus unknown distinction in Classical Latin and the particular diachronic development of aliquis. Concerning the first aspect, Latin is one of the rare languages in Haspelmath’s () sample with specialized lexical items for the specific known and the specific unknown functions (respectively, quidam and aliquis). This is particularly interesting in view of the fact that Latin has no grammaticalized system of articles: the marking of specificity, thus, assumes particular relevance with respect to the introduction and the management of discourse entities. The unambiguous lexical coding of a specific known form gets lost in Romance, as introduced in §.., since quidam is not Indefinites between Latin and Romance. First edition. Chiara Gianollo. © Chiara Gianollo . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

continued and is not replaced by a functionally equivalent item.1 As for the specific unknown function, one of the tasks of this chapter will be to scrutinize Haspelmath’s notion of specificity and to propose a different characterization for Latin aliquis, in the light of recent work on a class of indefinites that seems to correspond to a large extent to Haspelmath’s specific unknown function: the so-called epistemic indefinites. Concerning the second aspect, the extension into nonspecific functions of specific indefinites is a rare diachronic phenomenon. All the cases considered by Haspelmath (: –) are in fact specific unknown indefinites, which diachronically extend into non-specific functions (questions, conditionals), and which originate from expressions meaning ‘I don’t know who / what’ (e.g. French je ne sais quel- ‘I don’t know which’, i.e. ‘some kind of ’; Romanian ni¸ste ‘some’ < Latin nescio ‘I don’t know’, cf. Farkas : –). Haspelmath notes that the shift from a specific to a nonspecific function is the only kind of development running against his generalization that diachronic shifts affecting indefinites operate, through semantic weakening, unidirectionally from right to left on his semantic map (cf. §...). In the case of aliquis, however, the observed development is not only counterdirectional with respect to Haspelmath’s generalization, but also uncommonly wide-ranging: I show that already in Late Latin we witness the extension of aliquis into right-hand functions, to the detriment of quis in conditional and other nonveridical contexts, and even into negative contexts (the opposite extreme of the semantic map), where it takes the same value as quisquam and ullus ‘any’. The crucial aspect that needs to be accounted for is the original shift from ‘positive’ to polarity-sensitive. It is not surprising to see diachronic fluctuations in the licensing contexts of polarity-sensitive items, as we know from a number of studies (e.g., Ramat  on Italian veruno, Hoeksema  on Dutch enig, Martins  for an overview of Romance, Willis et al. a for a broader crosslinguistic comparison). It is, however, unexpected for a polarity-sensitive item to have a specific indefinite as its source. I connect the shifts undergone by aliquis to its nature of epistemic indefinite, enabling this way an explanation for the further Romance developments. I argue that the diachronic picture becomes much clearer once the ‘unknown’ component in the meaning of Latin aliquis, and its specificity, is re-evaluated in light of recent work on similar indefinites in contemporary languages, which have been analyzed as epistemic indefinites (Spanish alguno, French quelque, Italian qualche). .. Roadmap In §. I describe the functions found in the left-hand part of Haspelmath’s map and I give a preliminary overview of the contexts where the three Latin indefinites quidam, aliquis, and quis are found. In §. I introduce more in detail the debate surrounding

1 Cf. §.. for the status of Romance expressions corresponding to ‘a certain’ in this respect.

OUP CORRECTED PROOF – FINAL, //, SPi

The contexts of quidam, aliquis, and quis



the notion of specificity and discuss the expression of specificity in the system of Latin indefinites. I argue that only quidam is a specific indefinite: its behavior in Classical and Late Latin is the topic of §.. In §. I move to aliquis, summarizing my arguments against its treatment as a specific indefinite and preliminarily proposing an alternative analysis as an epistemic indefinite. I introduce the semantic attributes of the class of epistemic indefinites on the basis of research done on Spanish alguno and of recent crosslinguistic surveys. This way, I single out some distributional criteria that contribute to a more precise characterization of Latin aliquis. Section . investigates in detail to what extent the behavior of Classical Latin aliquis corresponds to what is known about the class of epistemic indefinites. I show that aliquis shares a conspicuous number of features with this class. Section . deals with the distribution of aliquis in Late Latin, showing an extension of contexts for Latin aliquis that preludes the Romance developments.

. The contexts of quidam, aliquis, and quis In this section, I introduce the Latin indefinites that occupy the left-hand side functions in Haspelmath’s () semantic map. I single out some issues both in the definition of the semantic categories involved and in the description of the distribution of Latin aliquis, which I then take up again in the further sections of this chapter. The left-hand side of Haspelmath’s () semantic map of indefinite nominal phrases is occupied by three contiguous functions, as shown in ():2 () left-hand side functions in Haspelmath’s () semantic map: specific known specific unknown irrealis nonspecific According to Haspelmath’s (: , –) analysis, in Latin, the specific known function is expressed by quidam (a), whereas aliquis (b) is reserved to the specific unknown function and to the irrealis nonspecific contexts. Latin aliquis can also appear in questions and conditionals (two functions in the central part of Haspelmath’s map); in these latter environments, however, it is much more usual to find the bare indefinite quis (c), identical in form to the interrogative pronoun (cf. §..), and treated by Haspelmath as an allomorph of aliquis. () a. Modo mihi advenienti nugator quidam occessit Just me:dat come:pt.dat trickster:nom a.certain:nom go.to:sg obviam towards ‘Just now a trickster crossed my way while I was coming’ (Plaut. Trin. )

2 See chapter , §.. for a preliminary discussion of the entire map, and §.. for an introduction to the Latin system of indefinites.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin b. Iam ego illuc praecurram atque inscendam aliquam in now I:nom thither proceed:sg and climb:sg some:acc on arborem tree:acc ‘Now I will run there first and climb up some tree’ (Plaut. Aul. ) c. Tum aquam aufugisse dicito, si quis petet! then water:acc run.away:inf say:sg if anyone:nom ask:sg ‘Say the water has run away if anyone asks for it’ (Plaut. Aul. )

Haspelmath’s classification conforms to traditional descriptions of aliquis and quis, which highlight the seemingly complementary distribution of the two indefinites (cf. a.o. Kühner and Stegmann : , Ernout and Thomas : , Traina and Bertotti : –). For example, Gildersleeve’s grammar says: ‘The common rule is that quis and qu¯ı occur properly only after s¯ı [if], nisi [unless], n¯e [lest], num [interrogative particle], or after a relative; otherwise aliquis, aliqu¯ı.’ (Gildersleeve and Lodge : §).3 This particular distribution also suggested the classification of aliquis as ‘specific’, in the sense that it escapes hypothetical and interrogative contexts. According to the Lewis–Short dictionary, aliquis is ‘more emphatic than quis, denotes that an object really exists, but that nothing depends upon its individuality; no matter of what kind it might be, if it is only one and not none’ (Lewis and Short : s. v.); quis, instead, ‘leaves not merely the object, but even its existence, uncertain’. In order to understand Haspelmath’s classification, we have to define his notion of specificity (to distinguish the two specific functions from the nonspecific ones) and the difference between the specific known and the specific unknown function. Haspelmath adopts a commonly assumed pre-theoretical definition of specificity: when using a nominal phrase specifically, the speaker ‘has a referent in mind’. Specificity expresses the certainty of the speaker about the existence and the identity of the referent. Specific indefinites are, thus, capable of introducing a salient discourse referent, which can be picked up again anaphorically in the subsequent discourse, cf. (). () Epidamniensis quidam ibi mercator fuit. from.Epidamnus:nom a.certain:nom there merchant:nom be:sg Is puerum tollit he:nom boy:acc take:sg ‘And in that place there was a certain merchant from Epidamnus. He took the boy’ (Plaut. Men. –) Specificity, in an articleless language like Latin, is most often left unmarked in nominal phrases and is recovered by means of various contextual factors. For instance, in () the indicative, ‘realis’ modality, the perfective tense, the lexical meaning of the predicate, the apposition further enriching the nominal phrase’s descriptive content contribute a specific reading, despite the absence of explicit morphological marking:

3 The forms qu¯ı, aliqu¯ı are the adjectival counterparts of quis, aliquis; cf. §...

OUP CORRECTED PROOF – FINAL, //, SPi

The contexts of quidam, aliquis, and quis



() non longe ab ostiarii cella canis ingens, catena not far from porter:gen lodge:abl dog:nom huge:nom chain:abl vinctus, in pariete erat pictus tie:pt in wall:abl be:sg paint:pt ‘Not far from the porter’s lodge a huge dog, secured to a chain, was painted on the wall’ (Petr. .) ‘Irrealis’ contexts in Haspelmath’s map are those where an indefinite can in principle be ambiguous between a specific and a nonspecific reading: future and various kinds of nonindicative modality (comprising markers of epistemic modality, e.g. apparently, in otherwise prototypical ‘realis’ environments, such as perfective past and ongoing present sentences). Concerning the specific known / unknown distinction observed in the Latin system, Haspelmath’s () typological survey shows that a number of languages have formal means to encode a distinction in the epistemic status of the speaker with respect to the identifiability of the referent. Specific known indefinites express identifiability on the part of the speaker (e.g. Russian koe- series, German gewiss), whereas a specific unknown indefinite leaves the identity of its referent undetermined, while entailing its existence (e.g. Russian -to series, German irgendein), cf. (–). () German (Haspelmath : ) a. Speaker A: Jemand hat angerufen. – Speaker B: Wer war es? ‘Speaker A: Someone called. – Speaker B: Who was it?’ b. Speaker A: Irgendjemand hat angerufen. – Speaker B:  Wer war es? ‘Speaker A: Someone (I don’t know who) has called.  Speaker B: Who was it?’ () German (Ebert et al. ) Peter sucht schon seit Stunden nach einer gewissen CD –  keine Ahnung, welche genau er sucht ‘Peter has been looking for a certain CD for hours now –  no idea, which one he is looking for’ Moving to Latin examples, we see that in (), with quidam, the context explicitly indicates a specific known reading, since additional information about the method of identification (‘knowing by sight’) is explicitly stated. Compare also (a), where the ‘knowing by sight’ is implicitly contributed by the event description itself, and (), where we know from the broader context of the theater play that the narrator is aware of the merchant’s identity. () Nec diu moratus rusticus quidam and.not long delayed:nom countryman:nom a.certain:nom familiaris oculis meis cum muliercula comite familiar:nom eye:dat my:dat with girl:abl accompanying:abl propius accessit closer come:sg ‘In a little while a countryman, whom I knew by sight, came up with a girl’ (Petr. )

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

In (a), with aliquis, on the other hand, no further information is given about the designated entity and the context is compatible with a reading where the speaker does not know which messenger will show up. However, the existence of a suitable individual is unambiguously entailed in the example, in virtue of the partitive construction, introducing a set to which this individual belongs, i.e., the messengers. In (b), again with aliquis, the identity of the referent is left indeterminate on purpose, since the speaker wants to utter a lawlike general statement. () a. expectabam aliquem meorum wait:sg someone:acc my:gen ‘I was expecting one of my messengers’ (Cic. Att..) b. veritatis argumentum est aliquid omnibus truth:gen argument:nom be:sg something:acc everybody:dat videri seem:inf ‘the fact that all men agree upon something is a proof of its truth’ (Sen. epist. .) Haspelmath (: ) considers an expression specific if ‘the speaker presupposes the existence and the unique identifiability of its referent’. This, of course, raises the question of how it is possible to have a specific reading if the identity of the referent is unknown to the speaker. As we will see more in detail in the next section, Haspelmath (: ) believes that in ‘specific unknown’ readings the specificity comes from the fact that for these nominal expressions ‘the identifiability of their referents is presupposed’, that is, the speaker can in principle identify a referent but s/he decides ‘to withhold the information about its identity from the hearer’ (Haspelmath : ). Concerning the contrast between quidam and aliquis in terms of the ‘knownunknown’ function, Haspelmath states that ‘[s]ome counterexamples to this generalization can apparently be found in Latin texts, but I know of no better description of the distinction between ali- and -dam.’ (Haspelmath : ). He gives the following examples from the Late Latin of the New Testament (I will come back to (b) later on): () from Haspelmath (: ) a. Magister, vidimus quemdam in nomine tuo Master:voc see:pl a.certain:acc in name:abl your:abl eicientem daemonia. cast.out:pt demons:acc ‘Master, we saw someone casting out demons in your name.’ (Vulg. Marc. .) b. Tetigit me aliquis, nam ego novi virtutem touch:sg me:acc some:nom in.fact I:nom know:sg power:acc de me exisse. from me:abl come.out:inf ‘Someone did touch me, for I was aware that power had gone out of me.’ (Vulg. Luc. .)

OUP CORRECTED PROOF – FINAL, //, SPi

The contexts of quidam, aliquis, and quis



While ‘quidam is absolutely never used non-specifically’, aliquis is also used non-specifically ‘in irrealis contexts and in questions and conditionals’ (Haspelmath : ). However, as mentioned, in questions and conditionals the use of the simplex quis is in fact the norm, cf. (). () Si quid petieritis me in nomine meo, hoc if anything:acc ask:pl me:acc in name:abl my:abl this:acc faciam do:sg ‘If you ask Me anything in my name, I will do it’ (Vulg. Ioh. .) Haspelmath attributes the rarer use of aliquis instead of quis in questions and conditionals to the prosodic-syntactic constraints of quis, which has to be enclitic upon another element (more precisely, upon the complementizer, cf. Bortolussi , ). In this, again, he follows the traditional scholarship on Latin, which treats the alternation between aliquis and quis to a large extent as the alternation between a prosodically strong (aliquis) and a prosodically weak (quis) form. However, the presence of aliquis in conditionals seems to be governed by more subtle semantic factors, cf. Bertocchi and Maraldi (), Bertocchi et al. (: –), Bortolussi (), Bortolussi (: ). For instance, the Late Latin example in () shows a Classical use (cf. Bortolussi , Bortolussi and Sznajder ), whereby when two indefinites are found in the same conditional clause, the second one is realized by aliquis: the interpretation of quis co-varies with the interpretation of aliquis.4 () Et si quid aliquem defraudavi, reddo and if something:acc someone:acc defraud:sg give.back:sg quadruplum quadruple:acc ‘And if I have defrauded someone of anything, I will give back four times as much’ (Vulg. Luc. .) In other cases, however, multiple occurrences of quis are found in a comparable context: () si quis quid quaereret if anyone:nom anything:acc ask:sg ‘if anyone had anything to ask’ (Cic. de orat. .) Bortolussi () provides a detailed treatment of the distribution of quis. Semantically, quis simply introduces a variable that has to be bound by a hierarchically superior operator. Typical binders include conditional, modal, question operators. Interestingly, negation does not figure among the possible binders for quis, arguably because of a blocking effect depending on the presence of dedicated negative-polarity series (quispiam, quisquam, ullus), cf. ():

4 On these uses see further Dupraz ().

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

() nec uero ex reliquis fuit quisquam qui iurare and.not in.fact of other:abl be:sg anyone:nom who:nom swear:inf dubitaret hesitate:sg ‘and in fact there was no one among the others who hesitated to swear’ (Caes. BC ..) We may, thus, consider quis as a necessarily bound indefinite that imposes no particular lexically encoded constraints as to the nature of its binder. Syntactically, quis tends to be displaced in a position immediately following the element expressing the binder, which is usually found in the high functional space of the clause. Being prosodically weak, quis cliticizes to its binder. This regular word pattern sometimes yields lexicalizations, as e.g. siquis, nequis (reported by ancient grammarians in their lists of indefinites as separate forms next to aliquis and quis, cf. Bortolussi : ). To summarize, we observe that, in traditional treatments, the most conspicuous feature described for Latin indefinites in the left-hand side of Haspelmath’s map, and especially for aliquis, is, in fact, their complementary distribution with respect to other indefinites: according to grammar descriptions, aliquis is typically found in ‘positive’ contexts, whereas in hypothetical and interrogative contexts the bare form quis ‘any’ is used; quidam contrasts with aliquis in the parameter concerning speaker knowledge. In the remaining sections of this chapter I will be mainly concerned with defining to what extent the contexts in which aliquis is found can really be characterized as ‘positive’. In the literature on Latin indefinites, the classification adopted by Haspelmath is not uncontroversial. If, for instance, Bertocchi et al. (: ) state that ‘aliquis is generally, though not always, specific’, Orlandini (: ), based on pragmatic criteria, classifies aliquis as nonspecific. The difference is due to the notion of specificity adopted in the various works. In order to deal with this issue, I therefore have to introduce the debate surrounding the notion of specificity first. After this discussion, in order to exemplify a clear-cut case of specific indefinite, I will present in more detail the straightforward distribution of quidam. I will then come back to aliquis and, in view of my conclusions on specificity, I will discuss an alternative characterization of this item as an epistemic indefinite.

. Specificity and Latin indefinites In this section I critically review the ‘specific unknown’ category. After discussing which notion of specificity has been adopted in previous analyses of Latin, I argue that the ‘specific unknown’ category is untenable and that an improved categorization can be attained by distinguishing two classes, specific and epistemic indefinites. .. Notions of specificity Specificity is a ubiquitous category in language description, which despite its broad application has proved hard to pin down. The literature on specificity in Latin, which I reviewed in Gianollo (), mirrors this situation. Here I will shortly summarize the main controversial points.

OUP CORRECTED PROOF – FINAL, //, SPi

Specificity and Latin indefinites



As seen in §., Haspelmath (: ) defines specificity as the co-existence of two presuppositions: presupposition of existence and and presupposition of unique identifiability of its referent (cf. Dahl  for a discussion of Haspelmath’s notion of specificity and his use of the term ‘presupposition’). In detecting whether an indefinite in his sample of modern languages is specific, Haspelmath adopts three tests: () a. existential paraphase: ‘There is an x’ b. disambiguation by explicit / unambiguous specific expressions such as Engl. a certain c. ability to introduce a discourse referent, i.e. to be picked up again anaphorically in the subsequent discourse According to Haspelmath, the ‘known / unknown’ distinction is appropriate only in the case of specific indefinites, since only specific indefinites presuppose unique identifiability, whereas with nonspecific ones the referents ‘are not identifiable in principle’ (Haspelmath : , ). With specific unknown indefinites, the (possibility of) referent identifiability is presupposed, but the speaker is not able or does not want to to carry out the identification (Haspelmath : ). Bertocchi et al. (: –), as well as Bortolussi (), take Haspelmath’s definition as their point of departure, and consider presupposition of existence and of unique identifiability as the two basic features of a specific reading. They distinguish between cases where presupposition of existence and of unique identifiability go hand in hand (either they are both present—specific known—or they are both absent— non-specific reading—) and cases where instead the denotation is not uniquely identified, but the existence of a referent is nonetheless presupposed. This latter combination would correspond to the specific unknown function in Haspelmath, but in fact it is a general description for plain wide-scope indefinites, and it is not necessarily connected to specificity. The fourth combination (unique identifiability but no existential entailment) is impossible, since identifiability necessarily leads to presupposition of the existence of a referent. According to Bertocchi et al. (), with aliquis there is always presupposition of existence but there is variation as to whether, according to contexts of occurrence, the indefinite carries or does not carry a commitment to unique identifiability. That is, for Bertocchi et al. (), unlike Haspelmath, identifiability is not always presupposed with aliquis. When the referent is not uniquely identifiable, aliquis is considered nonspecific. This position conforms to Orlandini’s () pragmatic definition of specificity, based on the speaker’s epistemic state: a specific indefinite is used when the speaker has a particular entity in mind. In this approach, the epistemic state of a speaker using a specific indefinite substantially corresponds to the state yielding the referential interpretation of a definite description; the difference consists in the epistemic state of the hearer, who is familiar with the referent in the case of a definite noun phrase (and thus shares the unique identifiability component), but does not know the referent’s identity in the case of a specific indefinite.5

5 Because of this, in some frameworks specific indefinites have been treated as directly referring expressions (Fodor and Sag ). In §.. I have introduced Schwarzschild’s () alternative analysis

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

.. Is aliquis really specific? The criteria seen in §.. for classifying aliquis are primarily meant to distinguish it from quis (on which cf. §.). Unlike quis, which is always found in the scope of structurally superior operators, aliquis has a freer distribution: it does not seem to require licensing by clausal operators (in fact, it shows a largely complementary distribution with quis in conditionals and questions), and it systematically takes wide scope in certain contexts, most notably negation. In this respect, aliquis resembles specific indefinites, which show a non-varying, operator-independent interpretation of the variable introduced by the indefinite noun phrase (Stark ). However, this analysis of aliquis raises some important questions in light of our current understanding of specificity. They concern in particular (i) the semantic status of the existential and of the identifiability commitment (ii) the nature of the scope preferences for aliquis. These problems are not limited to the approach to Latin sketched above: numerous analyses have been proposed for specificity, and none uncontroversially accounts for all aspects that have been connected to it. As for (i), a presuppositional treatment of specific indefinites has been proposed and successively defended in the theoretical literature only for a very restricted class of indefinites, and, crucially, it has been tied to a particular syntactic position (subject of individual level predicates; landing site for scrambled indefinites) or discourse status (topic), cf. von Fintel (), Endriss (: –), von Heusinger (: –), Özge () for discussion. That is, the kind of presupposition involved would not be encoded as a property of the indefinite in the lexicon, but would rather be derived as a contextual effect, i.e., from the informational structure and / or from the interaction with other operators. Now, a first observation on Latin aliquis is that it very rarely appears as subject of an individual-level predicate in a declarative sentence: in all of Caesar I found only one such example, given in (). () dixerat aliquis leniorem sententiam say:sg someone:nom mild:comp.acc opinion:acc ‘someone had expressed a milder opinion’ (Caes. BC .) Besides the rarity of such examples, in the sentence in () what could be interpreted as presupposed (in the sense of being part of the Common Ground) is the contextually established restrictor set (in this case, the members of the assembly on which Caesar is reporting). We would thus have a partitive effect (cf. also the example in (a)), rather than an existential presupposition. It is therefore not clear whether the meaning effects of Latin aliquis that have been analysed as evidence for its presuppositional nature are in fact brought about by its lexical specification or by the interaction with its contexts of use. A similar reasoning applies to the commitment to unique identifiability: if it was brought about by the lexical item itself, it is difficult to find a way of accounting for the

that treats specific indefinites as quantifiers and derives specificity effects from extreme narrowing of the domain of quantification.

OUP CORRECTED PROOF – FINAL, //, SPi

Specificity and Latin indefinites



fact that aliquis can oscillate between contributing and not contributing it. Moreover, the empirical data, as well as the paradigmatic relation with quidam, speak against associating such a commitment with any instance of aliquis in Classical Latin: the primary function of this indefinite is to express that more than one value for the variable it introduces can verify the claim. The definition of the meaning of aliquis given by Bortolussi (: ) seems to me to capture its essence accurately, explicitly excluding a commitment to unique identifiability: aliquis entails the existence of a set of individuals that can freely vary (‘aliquis suppose l’existence d’un ensemble d’individus pouvant alterner librement’). Concerning (ii), the main observation is that aliquis can have wide scope also when it does not carry unique identifiability, i.e., when, according to Bertocchi et al. (), it is nonspecific. That is, the definition of specificity as the conjunction of presupposition of existence and unique identifiability does not account for those cases where aliquis resists scoping under other operators even if the referent is not identifiable. An environment where this can be safely assessed is negation: aliquis systematically scopes above negation, although the identity of the designated entity remains unknown. The example in () can be paraphrased as follows: ‘you use the verb carere [to lack] in the situation in which there is something that you don’t have (whatever it may be) and you are aware that you do not have it’. () dicitur enim alio modo etiam ‘carere’ cum say:sg in.fact other:abl way:abl also lack:inf when aliquid non habeas et non habere te something:acc not have:sg and not have:inf you:acc sentias perceive:sg ‘for “to lack” is being used in a different sense when there is something that you don’t have and you are aware that you don’t have it’ (Cic. Tusc. .) In () we see what is known in the literature as ‘scopal specificity’ (cf. e.g. Farkas , von Heusinger ). Scopal specificity, in its strictest interpretation, refers to the ability of some quantifiers to take exceptionally wide scope over scope islands that usually trap quantifiers in their scopal domain, like e.g. conditional clauses; cf. a classical example in (), cited after von Heusinger (: ): () a. If a friend of mine from Texas had died in the fire, I would have inherited a fortune. (possible: there is a specific friend of mine and if he had died . . . ) b. If each friend of mine from Texas had died in the fire, I would have inherited a fortune. (not possible: for each of my friends, if one of them had died . . . ) The notion has also been applied, more in general, to the cases where an indefinite exhibits ‘regular’ wide scope with respect to a structurally superior operator, like negation in ().

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

The Latin example shows that scopal specificity is in principle independent of the reading implying unique identifiability on the part of the speaker, which falls under the heading of ‘epistemic specificity’ in Farkas () and von Heusinger (), cf. (): () A student in the syntax class cheated on the final exam Specific: ‘she was sitting right in front’ Non-specific: ‘but I don’t know who’ Mismatches between epistemic and scopal specificity are not uncommon crosslinguistically: unique identifiability does not always bring about wide scope (hence the possibility of intermediate scope for specific indefinites, cf. von Heusinger : –), and wide scope indefinites do not always express referent identifiability (Jayez and Tovena ; Alonso-Ovalle and Menéndez-Benito ; Romero ; Ebert et al. ). For the latter type, which has become known as the class of ‘epistemic indefinites’, alternative analyses have been recently proposed that do not resort to the notion of specificity in order to account for wide-scope readings. These approaches focus on the interplay between the semantic contribution of epistemic indefinites and the variable pragmatic effects they trigger according to the context; they capitalize on the semantic source of the ‘lack of speaker’s knowledge’ contribution to account for the context-dependence shown by epistemic indefinites. We will come back to epistemic indefinites in §.., with the aim of showing that the behavior of aliquis becomes much clearer once the theoretically dubious notion ‘specific unknown’ is dispensed with. .. Taking stock Let us take stock of the discussion in this section: we saw that identifiability of the discourse referent on the part of the speaker is necessarily coupled with an existential entailment. The lack of identifiability, on the other hand, is compatible with either an existential entailment or the lack thereof. Haspelmath’s ‘specific unknown’ function corresponds to the case where, despite lack of identifiability on the part of the speaker, an existential entailment is present in the interpretation. At least two problems arise with this category, though. First, we saw that it is not clear that the existential entailment comes about as the result of the presuppositional nature of the indefinite: rather, it may arise from the broader context in which the indefinite nominal phrase is used. Consider furthermore that, in the case of Latin aliquis, which is also found in the irrealis non-specific function, the existential entailment is not always present: this distribution speaks in favor of an analysis that does not attribute to aliquis a presuppositional nature, but rather derives an existential entailment in only some contexts of use. Second, it is debatable that an indefinite item that does not presuppose identifiability on the part of the speaker may count as a specific indefinite at all. Deciding on this point is dependent on the notion of specificity adopted. It seems incompatible, or at least quite difficult to reconcile, with Haspelmath’s own definition of specificity as presupposition of existence and presupposition of unique identifiability (Haspelmath : ). As we saw, Haspelmath defends specificity also in the case of lack of

OUP CORRECTED PROOF – FINAL, //, SPi

Quidam



identifiability on the grounds that the possibility of identifiability (e.g., by someone else than the speaker) would still be presupposed. In this case, however, it seems very hard to test this prediction empirically, at least with Latin aliquis, which is not always accompanied by an existential entailment. If there is an existential entailment, possibly arising through wide scope over a co-occurring operator, then automatically the possibility is granted that some epistemic agent may identify the entity corresponding to the denotation. However, this does not require us to assume a lexical encoding of either presupposition: as discussed, existence may be asserted in a transparent context or in virtue of the wide scope of the indefinite over a co-occurring operator; once this happens, nothing can prevent the further entailment that, in principle, the witness to the existential claim may be identified by some agent. In conclusion, the ‘specific unknown’ function does not qualify as specific if the notion of epistemic specificity is adopted. It may qualify as specific if it corresponds to the ability for an indefinite nominal phrase to be interpreted with wide scope with respect to some structurally superior operator (thus deriving the existential entailment), according to the notion of scopal specificity. This ability, however, does not necessarily have to be lexically encoded in the entry of the indefinite determiner, and may be contextually variable. It is observed also with ‘plain’ indefinites that do not show particular scope preferences and are ambiguous, since they can yield wide-scope readings as well as narrow-scope readings (cf. the English indefinite article in ()). In order to conclude the discussion on specificity, in the next section we will see the case of quidam, which combines scopal specificity (specificity as wide scope) and epistemic specificity (specificity as the speaker’s having one particular individual in mind), qualifying, thus, as a prototypical specific indefinite.

. Quidam According to von Heusinger (: ), ‘[a] prototypical specific indefinite is assumed to have wide scope, a referential reading, an existential presupposition, and to indicate discourse prominence’. In this section we will see that the distribution of Latin quidam conforms to this definition (and thus confirms its classification as ‘specific known’). This unambiguously specific indefinite is not continued in Romance.6 It is nonetheless important to establish its conditions of use, in order to better understand the landscape of specificity marking in Latin. Moreover, an overview of its distribution in Late Latin will give us the opportunity to discuss some aspects concerning the incipient grammaticalization of the indefinite article. Latin cardinal numeral unus ‘one’, the source of Romance indefinite articles, starts, already in Late Latin, to be used to single out specific referents, although sporadically. It qualifies, thus, as a competitor with quidam, and possibly as one of the causes for its loss in Romance.

6 In French, quidam ‘a certain’ was artificially reintroduced in the fourteenth century (according to the dictionaries; cf. also Schnedecker ). The reintroduction is motivated by Latin influence on the French literary standard, but the conditions of use are different from those observed in Latin: in particular, French quidam is compatible with impossibility of identification of the referent, and also develops pejorative readings.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

.. Uses of quidam in Classical Latin As seen in §.., quidam does not form a full-fledged series, but occurs only as pronoun and determiner. It occupies the specific known function in Haspelmath’s semantic map: quidam combines an existential claim with the additional meaning component of identifiability on the part of the speaker in all its uses: its referent is speaker-given but hearer-new and discourse-new. Translations found in the dictionaries for quidam include ‘a certain, a specific’, Germ. ‘ein gewisser’, It. ‘un certo, uno’. It can be used in the singular and in the plural, and as a determiner it is found either before or after the head noun. The form is based on the interrogative / indefinite stem *kw i-, to which a suffixal particle -dam is added. The particle has a demonstrative origin: it derives from the Proto-Indo-European adverb *de, *do ‘here’ (cf. de Vaan  s.v. -dam). This suffixed undeclinable particle is responsible for the internal inflection seen with quidam, where only the pronominal part inflects (e.g. genitive sing. cuius-dam, feminine nominative sing. quae-dam, etc.). As we will see, this type of inflection may have represented a factor in the demise of quidam in the development toward Romance. The main function of quidam is to introduce new referents in the discourse in episodic contexts and existential constructions. The introduction is often followed by further specifications concerning the intended referent. It is very frequent in existential presentative constructions, when a character is introduced at the beginning of the narration: () Mercator quidam fuit Syracusis senex merchant:nom a.certain:nom be:sg Syracuse:abl old.man:nom ‘There was a certain merchant in Syracuse, an old man’ (Plaut. Men. ) We saw in () that the introduced discourse entity can be subsequently picked up again by an anaphoric expression. Contexts of use of quidam like () are unambiguous with respect to the epistemic status of the speaker. Often quidam accompanies proper names (cf. Bertocchi and Maraldi ): its primary function in this case is to indicate the first mention in the discourse of an individual assumed to be unknown to the hearer. () a. ME. Quis is homost? who:nom this:nom man:nom-be:sg b. MA. Menaechmus quidam Menaechmus:nom a.certain:nom ‘ME: Who is this man? MA: A certain Menaechmus’ (Plaut. Men. ) In some cases the context provides explicit indication of the method of identification (cf. (), where the speaker informs us that he knows the intended individual by sight). In other cases an elaboration follows, for instance by means of a relative clause: () Est quidam homo, qui illam ait se be:sg a.certain:nom man:nom who:nom that:acc say:sg he:acc scire ubi sit know:inf where be:sg ‘There is a certain man who says that he knows where she is’ (Plaut. Cist. )

OUP CORRECTED PROOF – FINAL, //, SPi

Quidam



In other cases the speaker does not specify further information about the referent, but often the context gives some indications to the fact that what is evoked is part of some common knowledge that the hearer may also retrieve. This seems to be the case in the following examples, which are set in the dialogic situation of a letter exchange: () a. sed prima duo capita epistulae tuae but first:nom two:nom paragraph:nom letter:gen your:gen tacita mihi quodam modo reliquenda sunt silent:nom I:dat a.certain:abl way:abl leave:ger.nom be:pl ‘but the first two paragraphs of your letter have to remain in a certain way without reply on my part’ (Cic. fam. ..) b. Sed tamen significatur in tuis litteris but nonetheless manifest:sg.pass in your:abl letter:abl suspicio quaedam et dubitatio tua suspicion:nom a.certain:nom and doubt:nom your:nom ‘but nonetheless in your letter a certain suspicion and doubt on your part are apparent’ (Cic. fam. ..) The examples above also show that quidam may be used with abstract nouns (‘manner’, ‘doubt’): in such cases the indefinite indicates that a specific kind of instantiation of the concept is intended. This use is sometimes found also with nouns potentially designating individuals, which however in the context refer to a specific kind of person, and not to a specific individual, as in (). Zamparelli () remarks that this is also possible with a certain in English; he analyzes these cases as involving an implicit kind modifier (‘a certain ruler’ = ‘a certain kind of ruler’). () dux nobis et auctor opus est et leader:nom us:dat and counsellor:nom necessity:nom is:sg and eorum ventorum quos proposui moderator quidam this:gen current:gen which:acc discuss:sg ruler:nom a.certain:nom et quasi gubernator and almost steersman:nom ‘What we need is a leader and a counsellor, a certain kind of ruler and so to say steersman of those currents that I discussed’ (Cic. fam. ..) Some contexts suggest that the attitude / knowledge holder may be different from the speaker, for instance in reported speech and attitude reports: () a. nisi quod addis visum esse quibusdam edictum if.not because add:sg seem:pt be:inf a.certain:dat edict:acc meum quasi consulto ad istas legationes my:acc almost deliberately to this:acc delegation:acc impediendas esse accomodatum prevent:ger.acc be:inf aim:pt ‘were if not for the fact that—as you add—to some my edict has seemed deliberately aimed at not allowing those delegations’ (Cic. fam. ..)

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin b. sin autem quem mea instituta in provincia non if then some:acc my:nom decision:nom in province:abl not delectant et quadam dissimilitudine institutorum meorum please:pl and a.certain:abl difference:abl decision:gen my:gen ac tuorum laedi se putat and your:gen damage:inf.pass himself:acc consider:sg ‘if then someone is not happy with my administration of the province and considers himself injured by a certain difference between my administration and yours’ (Cic. fam. ..)

In some cases quidam and aliquis are explicitly contrasted with respect to identifiability: while quidam always indicates a clearly definable quality pertaining to the denoted element, aliquis is used to express indeterminacy: () a. etsi de tua prolifica beneficaque although from your:abl generous:abl benevolent:abl-and natura limavit aliquid posterior annus nature:abl wear.away:sg something:acc second:nom year:nom propter quandam tristitiam temporum because.of a.certain:acc gloom:acc time:gen ‘although the second year [of administration] wore away something of your generous and benevolent nature on account of a certain gloom of these times’ (Cic. fam. ..) b. non astutia quadam sed aliqua potius sapientia not strategy:abl a.certain:abl but some:abl rather wisdom:abl sum secutus follow:pt be:sg ‘I acted not with a precise strategy but rather with some wisdom’ (Cic. fam. ..) In the case of co-occurrence with sentential negation, quidam always has a specific reading, scoping above negation: () a. Vadimonium mihi non obiit quidam socius recognizance:acc me:dat not go.to:sg a.certain:nom partner:nom et affinis meus and relative:nom my:nom ‘A partner and relative of mine has not answered to his recognizances’ (Cic. Quinct. ) b. Quidam non adiecere numerum certain:nom not add:pl number:acc ‘Certain authors did not add the number’ (Liv. ..) The few cases where quidam is found in the antecedent of a conditional (Bertocchi and Maraldi ) seem to support the conclusion that quidam shares the exceptional wide-scope properties that characterize prototypical specific indefinites. In () Cicero is referring to specific individuals, easily retrievable by the addressee of

OUP CORRECTED PROOF – FINAL, //, SPi

Quidam



his letter (that is, to the members of the senate belonging to Cato’s faction who, in  bce, are trying to deprive Pompey of his authority): if these people will behave more moderately, those who control the republic will be able to allow for some quiet. () otium nobis exoptandum est, quod ii tranquillity:nom us:dat wish:ger be:sg which:acc this:nom qui potiuntur rerum praestaturi videntur, si who:nom obtain:pl power:gen provide:pt.nom seem:pl if quidam homines patientius eorum potentiam certain:nom man:nom more.moderately they:gen force:acc ferre potuerint tolerate:inf can:pl ‘Tranquillity is what we have to pray for, and this the present holders of power seem likely to provide, if certain of them show somewhat more tolerance of their supremacy.’ (Cic. fam. ..) .. Quidam in Late Latin Uses of quidam are still relatively productive in Late Latin. They do not differ from the Classical Latin use in frequency or in breadth of functions. It is also important to remark that, although late texts tend to show an increased presence of determiners, with the main function of improving coherence and explicitly signaling anaphoric relations and new topics (cf. Selig ), quidam does not seem to become more frequent and, especially, is not obligatorily used in contexts where a specific reading is required: as in Classical Latin, specificity can be left unmarked. In the stages I investigated (Christian writers and biblical translations in the third to fourth centuries ce; Merovingian Latin of the sixth century ce) there is no sign that quidam is contributing to the general tendency of adding determination to nominal phrases. ... Contexts of use The privileged context for quidam in biblical texts is, as in Classical Latin, the introduction of new specific referents in the discourse, but also in this functional domain it remains optional, as the comparison between (a) and (b) shows. () a. occurrit illi vir quidam qui come.towards:sg that:dat man:nom a.certain:nom who:nom habebat daemonium have:sg demon:acc ‘a man who was possessed by demons came out to meet him’ (Vulg. Luc. .) b. et ecce venit vir cui nomen Iairus and look! come:sg man:nom who:dat name:nom Iairus:nom ‘Then a man named Jairus came ’ (Vulg. Luc. .) Quidam is typically found in narrative sentences in the ongoing present or perfective past, i.e., the typical environments for specific readings. The most frequent constructions involve (i) the pronominal form followed by a partitive specification; (ii) the adjectival use in the singular to specify descriptions such as homo dives ‘a rich man’,

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

mendicus ‘a beggar’, etc. In presentational sentences it also occurs in the plural (it is difficult to say whether in the plural quidam shows special meaning effects associated with identifiability, cf. (–)). Below we see examples where quidam has the function of introducing new referents in the narration: () a. adulescens autem quidam sequebatur illum young.man:nom then a.certain:nom follow:sg he:acc amictus sindone wrap:pt.nom linen.sheet:abl ‘A young man was following him, wearing a linen sheet’ (Vulg. Marc. .) b. iudex quidam erat in quadam civitate judge:nom a.certain:nom be:sg in a.certain:abl city:abl qui . . . who:nom ‘In a (certain) city there was a (certain) judge, who . . . ’ (Vulg. Luc. .) In a few cases, quidam is found accompanying homo, thereby marking as specific an expression which is otherwise interpreted as generic, or vir, which is typically used specifically, resulting in redundancy in this latter case (cf. Bortolussi and Sznajder , de la Fuente ): () a. Homo quidam habuit duos filios man:nom a.certain:nom have:sg two:acc son:acc ‘A man had two sons’ (Vulg. Luc. .) b. tunc vir quidam de filiis prophetarum dixit then man:nom a.certain:nom from son:abl prophet:gen say:sg ‘Now a (certain) man of the sons of the prophets said’ (Vulg. I Reg. .) As Bortolussi and Sznajder () remark, the behavior of quidam in the Vulgate largely corresponds to what is found in Classical Latin. However, one may suspect that quidam is losing productivity and is used only in very prototypical, almost idiomatic contexts, such as those seen above. In the only case of quidam in the antecedent of a conditional that Bortolussi and Sznajder find in the Vulgate, they remark an interesting difference with respect to Classical Latin: quidam does not take exceptional wide scope outside of the conditional and has a nonspecific interpretation, which could also have been achieved with quis. Note, however, that the indefinite takes wide scope with respect to negation, a fact which may explain why here quidam has been used instead of quis: illorum non crediderunt, numquid incredulitas () si quidam if some:nom that:gen not believe:pl Q unbelief:nom illorum fidem Dei evacuabit? that:gen faith:acc God:gen nullify:sg ‘If some of them did not believe, their unbelief will not nullify the faithfulness of God, will it?’ (Vulg. Rom. .)

OUP CORRECTED PROOF – FINAL, //, SPi

Quidam



In the few cases where quidam is found in negative sentences, it is always interpreted specifically (although in the example in () it seems that identifiability is not relevant for interpretation, as is often the case in the plural): () et quidam credebant his quae dicebantur and certain:nom believe:pl this:dat which:nom say:pl.pass quidam vero non credebant certain:nom in.fact not believe:pl ‘And some people believed what was being said, others instead did not believe’ (Vulg. act. .) Similar conclusions can be reached on the basis of other Late Latin texts, such as the Itinerarium Egeriae (fourth century ce) or Gregory of Tours’ Historia Francorum. All uses conform to the Classical usage and quidam appears in contexts where a new discourse entity is introduced. The referent is always presented as unambiguously known to the speaker. Sometimes, as in () below, the way of identification is provided (in this case, the proper name): () martyrium . . . sancti cuiusdam monachi nomine martyrdom:nom . . . holy:gen a.certain:gen monk:gen name:abl Helpidi Helpidius:gen ‘the martyrdom . . . of a certain holy monk named Helpidius’ (Itin. Eg. .) Interestingly, also in Late Latin we sporadically find cases of reported speech where, arguably, the reporting agent, and not the speaker, can identify the referent: () Et quoniam nescio quando dicitur quidam and since ignore:sg when say:sg.pass someone:nom fixisse morsum et furasse de sancto ligno ... impress:inf bite:acc and steal:inf from saint:abl wood:abl ‘and since it is said that someone—I don’t know when—bit off and stole from the sacred cross . . .’ (Itin. Eg. .) ... Competition with certus and unus The disappearance of quidam in Romance has traditionally been attributed to its morphology, i.e., to the fact that it was an internally inflected element (cf. §..). This kind of inflection pattern was not sustained in Romance in general, cf. the fate of other internally inflected items such as hic ‘this’ and idem ‘the same’.7 7 Wanner (: ) rightly remarks that other reinforced elements, likewise internally inflecting, could be rescued by an early process of reanalysis: ipse ‘he (himself)’ was originally internally inflected, as expected given the etymology (is + -pse) and as shown by Early Latin internally or doubly inflected forms (Plautus: eapsa, eumpse, eumpsum). However, in Classical Latin the pronoun appears with the innovative external inflection (ipsa, ipsum, etc.). According to Wanner, while in principle a similar reanalysis could have happened for other elements at an early stage, this possibility has to be excluded for Late Latin, since declensions would have already been unproductive, i.e. unable to create new inflected elements by means of reanalysis; at the same time, the fixation of a single-case form would have been ‘premature’ in the system of Late Latin, which was still inflectional.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

In addition to the morphological factors, however, competition with various other functional items is plausible, in particular certus ‘certain’ (Väänänen : , Garassino ) and unus ‘one’.8 The employment of certus, an adjective meaning ‘sure, predetermined’, with the function of a specificity marker is seen, albeit rarely, already in Classical Latin: () insolentia certorum hominum insolence:nom certain:gen people:gen ‘the insolence of certain specific people’ (Cic. Marcell. ) According to Garassino (), throughout the history of Latin certus remains an adjective and does not yet develop the determiner-like properties characterizing it in some Romance varieties.9 The fact that it can co-occur with quidam (and aliquis) shows that it still has an adjectival distribution in Classical and Late Latin. The marking of specificity emerges contextually from the adjective’s lexical meaning, which reduces vagueness by pointing to the existence of a unique determined referent. Selig (: ch. ) remarks that in her narrative corpus of Late Latin the explicit marking of indefinite newly introduced entities by means of determiner-like elements is very rare; it is exceptional in the case of definite (world-known) newly introduced entities and anaphoric ones. Like quidam, unus may be employed when characters or places are introduced with some prominence: in the function of character introduction, thus, quidam and unus compete in Late Latin texts. However, at no stage does unus specialize in the marking of specificity, and it can also be found with nonspecific indefinites (Selig : ).10 The examples below show typical contexts where quidam and unus compete in the function of introducing a new referent in the narration. These innovative uses of the cardinal numeral unus are commonly interpreted as preluding its grammaticalization as a determiner. () a. cum venisset autem una vidua pauper misit as come:sg then one:nom widow:nom poor:nom put:sg duo minuta quod est quadrans two:acc mite:acc which:nom be:sg farthing:nom ‘And there came a certain poor widow, and she threw in two mites, which make a farthing’ (Vulg. Marc. .) b. et accessit unus de scribis qui audierat and come:sg one:nom from scribe:abl who:nom hear:sg illos conquirentes that:acc argue:pt.acc ‘and one of the scribes came who had heard them arguing’ (Vulg. Marc. .) 8 In the plural, diversi ‘various’ (lit. ‘different’) starts to be used instead of Classical nonnulli, complures, cf. Väänänen (: ), Garassino (). 9 Also the connection between position in the DP and interpretation is a Romance phenomenon: Old Italian already shows the asymmetry between a determiner-like (‘a certain’) interpretation in pre-nominal position and an adjectival (‘a sure’) interpretation in post-nominal position (although Garassino  remarks that in Old Italian adjectival readings are still possible in pre-nominal position). 10 According to Selig (: ) the use of definite and indefinite determiner-like elements in Late Latin is to be understood as an attention-driving strategy.

OUP CORRECTED PROOF – FINAL, //, SPi

Quidam



Similar cases are found, albeit rarely, in Classical Latin: () Erat unus intus Nervius nomine Vertico be:sg one:nom inside Nervius:nom name:abl Vertico:nom ‘In the camp there was one of the Nervii named Vertico’ (Caes. BG .) The literature on the development of the Romance indefinite article highlights that unus starts its grammaticalization path as a cataphoric, referent-introducing device, especially with highly persistent discourse elements, and that this function is still prevalent in Old Romance texts (Selig , de la Villa  and literature there cited for Late Latin, Carlier ,  for Old French, Stark ,  for Old Italian, Pierluigi  for the oldest stages of French, Italian, Catalan, and Spanish). As we saw from the examples above, unus accompanying specific nominal phrases is found already in texts from the fourth century ce and it is often possible to connect it with the epistemic status of ‘specific known’, but this use is by no means systematic or unambiguous at this stage. Often unus is rather employed, as in Classical Latin, with emphatic strengthening of the cardinal meaning (‘one single’): () lacte colas, cui cruda ova commisces, ut milk:acc strain:sg which:dat raw:acc egg:acc mix:sg so.that unum corpus fiat one:nom body:nom become:sg ‘strain the milk, to which you have mixed raw eggs, so that it becomes one single body’ (Apic. ..) Sometimes in this emphatic use unus is not specific, i.e., its interpretation does not require the identification of a determined witness to the claim; cf. the following example, where it occurs in combination with anima in the scope of a conditional: () Quod si anima una nesciens peccaverit because if soul:nom one:nom not.know:pt.nom sin:sg ‘And if one single person sins unintentionally’ (Vulg. Num. .) In the cardinal meaning unus also frequently shows narrow scope with respect to negation, where unus . . . non, but also non . . . unus, independently of surface position, can be equivalent to nemo, nullus ‘nobody’, as in (). () nonne duo passeres asse veneunt et unus ex not-q two:nom sparrow:nom cent:abl sell:pl and one:nom from illis non cadet super terram sine Patre vestro that:abl not fall:sg on earth:acc without father:abl your:abl ‘Are not two sparrows sold for a cent? And yet not one of them will fall to the ground without your Father’ (Vulg. Matth. .) In negative contexts, unus starts to be used quite frequently in Late Latin as strengthener of negation, often in combination with other elements. In §. we will see some examples of the combination with aliquis. As an end-of-scale element, unus in this use can contribute an emphatic ‘even’ flavor (‘not even one’): in chapter  we will see that this is the reason why unus played a crucial role in the grammaticalization of new Romance narrow-scope indefinites in negative contexts.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

I conclude by observing that in Late Latin quidam and unus overlap in the function of introducing new referents in the discourse, a function which was typical for quidam but innovative for unus. This might have been a factor leading to the loss of quidam in Romance. However, we saw that in the period investigated here instances where unus is used in this specific function are quite rare, and unus essentially remains a cardinal numeral. As argued by Schaden (), all uses of unus in Late Latin may still be explained by assuming the semantics of a cardinal numeral, without positing an early stage of grammaticalization of the indefinite article.

. Latin aliquis and the class of epistemic indefinites This section and the rest of the chapter are dedicated to Latin aliquis. This indefinite pronoun and determiner has a great historical significance for the restructuring of the system of indefinites witnessed from Latin to Romance. It is diachronically very succesful: it is continued by Romance forms such as It. alcuno, Fr. aucun, Sp. algún, Pt. algum, which derive from the univerbation of aliquis with unus ‘one’ > *alicunum. As discussed in the general introduction to this chapter, this item undergoes a typologically rare and theoretically puzzling diachronic development in Romance: in Latin it has been analyzed as a specific indefinite, whereas one of the functions (in some languages the only function) of the Romance continuations in the singular is that of an item dependent upon negation for its licensing. That is, Romance continuations of Latin aliquis are licensed in contexts, like the scope of negation, which excluded Latin aliquis and which, more in general, are known to be incompatible with specific readings of indefinites. My aim here is to lay the basis for a diachronic investigation of this Romance development, by better defining its initial state. I focus on the distribution of aliquis in Classical Latin and try to reach a different characterization of this item than what has been proposed by Haspelmath () and in the scholarship on Latin reviewed in §.. I argue that the distribution and the meaning contribution of aliquis with respect to the parameter of speaker’s knowledge suggest that it belongs to the class of epistemic indefinites (e.g. Alonso-Ovalle and Menéndez-Benito , , Jayez and Tovena ). Epistemic indefinites are characterized by (crosslinguistically diversified) context-sensitivity: they give rise to variable meaning effects according to the syntactic-semantic context they interact with. I further discuss how the (minimal) widening of the domain of quantification brought about by this type of indefinite can represent a first step on a diachronic cline leading to the Romance situation. In this section, I will thus discuss the class of epistemic indefinites, in terms both of their crosslinguistic distribution and of the theoretical analyses that have been offered. In §. we will come back to the behavior of Classical Latin aliquis and in §. we will see some Late Latin developments.11

11 Part of the material considered in the next sections appeared in a preliminary form in Gianollo ().

OUP CORRECTED PROOF – FINAL, //, SPi

Latin aliquis and the class of epistemic indefinites



.. Epistemic indefinites Epistemic indefinites represent a quite heterogeneous class researched under various headings: besides epistemic indefinites (Alonso-Ovalle and Menéndez-Benito , Jayez and Tovena , Aloni and Port ), they are also known as modal indefinites (Alonso-Ovalle and Menéndez-Benito ), or referentially vague indefinites (Giannakidou and Quer ). The table in () lists some cases discussed in the literature. Alonso-Ovalle and Menéndez-Benito (, ) provide an overview of recent work on this class. () Epistemic indefinites discussed in the literature12 Epistemic indefinites English German Romanian French Italian Spanish Greek Russian Finnish

sg. some irgendein vreun quelque (un) qualche algún kapjos -nibud’ series -kin series

Epistemic indefinites owe their name to the fact that they are sensitive to the doxastic alternatives of an epistemic agent: they make an existential claim but also contribute the information that the speaker is ignorant about the identity of the individual that satisfies this claim, cf. ().13 () a. French (Jayez and Tovena ) Le verrou ne coulisse pas; quelque idiot a fermé la porte avec un cadenas ‘The bolt does not slide; some dumb people locked the door’

12 On Spanish algún cf. Alonso-Ovalle and Menéndez-Benito (, ), Giannakidou and Quer (, ); on French quelque Jayez and Tovena (, ); on Italian (un) qualche Zamparelli (); on Romanian vreun Fˇalˇau¸s (, ); on German irgendein Kratzer and Shimoyama (), Lauer (); on Greek kapjos Giannakidou and Quer (); on Russian nibud’ Pereltsvaig (), Onea and Geist (). 13 A terminological note is in order. Jayez and Tovena () used the term ‘epistemic’ to indicate, in a broad sense, all those indefinites that are sensitive to the parameter ‘knowledge of the speaker’, that is, both indefinites requiring the speaker to know the identity of the referent (cf. Farkas’  epistemic specificity; Fr. un certain ‘a certain’) or not to know it (Fr. un N quelconque ‘any’). The correspondence with Haspelmath’s () ‘known / unknown’ classification breaks down when we consider that for Haspelmath speaker’s knowledge (identifiability on the part of the speaker) is a relevant parameter only in the case of specific indefinites, while for the rest of the system it is simply non-applicable. On the contrary, Jayez and Tovena () distinguish between different varieties of free-choiceness and keep identification separate from specificity (which they intend as wide scope). In other work, e.g. Alonso-Ovalle and Menéndez-Benito () and Aloni and Port (), the term ‘epistemic’ has ben restricted to indefinites indicating lack of speaker’s knowledge. This use has become predominant and I follow it here.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin b. European Portuguese (A. M. Martins, p.c.) Algum estudante pagou as propinas. Falei com ele depois. ‘Some student paid the tuition fees. I spoke to him afterwards.’

The ignorance effect is weaker than the free-choice reading: this has been explained by the hypothesis that the two types of indefinite impose different constraints on their domains of quantification. For quelque, algun, etc., the domain of quantification cannot be narrowed down to an individual (‘anti-singleton’ indefinites), but at the same time it is not maximally widened, so that some alternatives may be excluded (an effect variously labeled as ‘partial’ freedom of choice among the members of the restriction set, ‘minimal domain widening’). The ‘side-message’ that the epistemic agent is ignorant or indifferent with respect to the designated individual is absent with the plain indefinite article, which is compatible with a scenario where the speaker elaborates further on the identity of the intended referent. A ‘namely’-continuation is infelicitous with algún in (a), whereas it is compatible with the indefinite article (b). Consequently, a hearer’s reaction asking for more information on the witness of the existential claim is felicitous in (b), but odd in (a). () ‘namely’-continuation (from Alonso-Ovalle and Menéndez-Benito : ) a. María se casó con algún estudiante del departamento de lingüística:  en concreto con Pedro ‘María married algún student of the department of linguistics:  namely Pedro’ b. María se casó con un estudiante del departamento de lingüística: en concreto con Pedro ‘María married a student of the department of linguistics: namely Pedro’ () hearer’s continuation (from Alonso-Ovalle and Menéndez-Benito : –) a. Speaker A: Juan tiene que estar en alguna habitación de la casa Speaker B:  ¿En cuál? Speaker A: ‘Juan must be in algún room of the house’ Speaker B:  ‘Which one?’ b. Speaker A: Juan tiene que estar en una habitación de la casa Speaker B: ¿En cuál? Speaker A: ‘Juan must be in a room of the house’ Speaker B: ‘Which one?’ The two main questions guiding recent research on epistemic indefinites concern the nature of their knowledge-sensitivity and the way to model the contextual variability observed with the ‘epistemic effect’. Differences in the literature arise as to how to treat the nature of the meaning contribution with epistemic determiners (truth-conditional; presupposition; conventional implicature; conversational implicature). In the next section I will focus on Alonso-Ovalle and Menéndez-Benito’s (, ) work on Spanish algún, which elaborates on the treatment of German irgendein in Kratzer and Shimoyama () and derives the epistemic effect as a conversational

OUP CORRECTED PROOF – FINAL, //, SPi

Latin aliquis and the class of epistemic indefinites



implicature resulting from the constraints that epistemic indefinites impose on their quantificational domain. .. The anti-singleton constraint Alonso-Ovalle and Menéndez-Benito () (henceforth, AM) refer to the class of epistemic indefinites as ‘modal indefinites’, because they interpret their meaning contribution as due to an across-worlds (= modal) variation constraint imposed by the lexical entry. This constraint is elsewhere called ‘anti-singleton constraint’ and it amounts to minimal widening of the domain of quantification. The witness of an epistemic indefinite cannot be the same in all of the speaker’s epistemic alternatives. At least two individuals in the domain of quantification must be possible witnesses for the existential claim. If the indefinite is in the scope of an intensional operator, the variation constraint amounts to positing that in all the worlds possible for the speaker there has to be a possible denotation for the indefinite, but this denotation has to be different in each world (Modal Variation). Differently from free-choice items, it is not necessary that all elements in the domain be a potential denotation. Ignorance can concern the individual satisfying the claim, or the kind of which it is a representative (Weir , Alonso-Ovalle and Menéndez-Benito ). () En la pared de mi habitación está creciendo alguna planta ‘There is some plant growing on the wall of my room’ (Alonso-Ovalle and Menéndez-Benito : ) The denoted entity is not required to be unique: uniqueness is an inference that may be contextually triggered—as in the case of the ‘marry’ context seen in (), or in the ‘be in a room’ context in ()—, but does not need to. In other cases the speaker may express ignorance with respect to the total number of witnesses for the existential claim (cf. It. qualche, Zamparelli ): that is, despite singular morphological marking, the algún-phrase is compatible also with a plural interpretation, as in (): () Mi coche tiene algún abollón ‘My car has some dent(s)’

(AM: )

According to AM, the Modal Variation interpretation (ignorance about identity) arises as a quantity-based conversational implicature. Its trigger is provided, as in the case of irgendein in the analysis of Kratzer and Shimoyama (), by the widened domain of quantification (on which see §..). This domain shift is brought about by a presupposition built in the lexical entry, the anti-singleton constraint. The anti-singleton constraint on the domain of quantification is modeled as a constraint on the value of the subset selection function f . Such function selects the contextually relevant quantificational domain which the quantifying determiner takes as an argument. () anti-singleton constraint: | f (P) |>  In () P is the set corresponding to a given property. The cardinality of the set resulting from the application of f cannot be .

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

The lexical entry proposed for algún by Alonso-Ovalle and Menéndez-Benito (: ) is reproduced in (): the string before the colon is the definedness condition; the anti-singleton presupposition f is given immediately after the colon; f is the subset selection function constraining the restriction P of the existential quantifier. () lexical entry for algún in Alonso-Ovalle and Menéndez-Benito (: ): algún = λf λP λQ : anti-singleton(f ).∃x[f (P)(x) ∧ Q(x)] The impossibility for algún to quantify over a singleton set is shown by its infelicitous behavior in examples where a restrictive relative clause further speficies the uniqueness of the denotation: () anti-singleton behavior (examples from AM: , their –) a. Pedro contrató a un candidato que era el más incompetente de los que se presentaron ‘Pedro hired a candidate who was the most incompetent of those who applied’ b.  Pedro contrató a algún candidato que era el más incompetente de los que se presentaron ‘Pedro hired algún candidate who was the most incompetent of those who applied’ The ignorance implicature arises as a scalar implicature triggered by the Maxim of Quantity, because of the competition between algún and the stronger (more informative) alternatives represented by those determiners which require the restriction of their quantificational domain to be a singleton. AM follow Schwarzschild () here and assume that such stronger alternative is represented in Spanish by the specific reading of the indefinite article un (for specific indefinites as singleton indefinites, cf. §..). .. Dependent licensing In AM’s model, the conditions on the domain of quantification observed with epistemic indefinites cause the indefinite to interact with other operators in the clause (e.g. modal verbs, negation), and to ultimately become dependent on (some of) them for its licensing. The minimal domain widening originating from the antisingleton constraint is responsible for the observed dependence: epistemic indefinites are felicitous only in some contexts (i.e. in the interaction with some operators) because in other contexts they yield a statement which is too weak with respect to a general informativeness requirement. The derivation of the implicature proceeds as in Kratzer and Shimoyama’s () Hamblin semantics for indefinites. Domain-widened indefinites induce a set of alternatives, which originate as individual alternatives and then grow into propositional alternatives; this way they are ‘visible’ to modals and can be distributed over the set of accessible worlds introduced by the modals. As in Kadmon and Landman’s () original proposal, domain widening must happen for a reason. In the case of downward-entailing contexts, the motivation is strengthening: domain-widened expressions (such as NPIs) are maximally informative in scale-reversal contexts.

OUP CORRECTED PROOF – FINAL, //, SPi

Latin aliquis and the class of epistemic indefinites



But which function does (minimal) domain widening have in non-downwardentailing contexts, where the direction of entailments leads to a weaker reading? According to Kratzer and Shimoyama (), depending on the surrounding semantic conditions, minimal domain widening in non-downward-entailing contexts can in principle be driven by either of two pragmatic motivations, which the hearer may assume in order to cooperatively interpret the fact that the speaker did not make a stronger claim: (i) avoidance of a false claim or (ii) avoidance of an exhaustivity inference (i.e. unwarranted exclusion of alternatives). Note that in both cases the resulting assertion is informationally weaker than the one which would arise with the indefinite article; however, since there is a pragmatic motivation for such a weaker statement, it is felicitous. In this respect, thus, minimal widening of the quantificational domain happens for the opposite reason to extreme widening with free-choice indefinites and NPIs, which strengthen the assertion. Also with epistemic indefinites, the felicity of their use depends on the surrounding semantic context: pragmatic reasoning aiming to avoid a false claim or an unwarranted exhaustivity inference will need, thus, the right context, i.e. one where the reasoning leads to a meaningful conversational contribution. A crucial aspect of AM’s proposal, which has important consequences for my diachronic analysis, concerns the role of the context-sensitivity observed with epistemic indefinites. Epistemic indefinites are selective, in ways that vary crosslinguistically, as to the structural and semantic contexts in which they may occur. In this respect, epistemic indefinites are not simple ‘positive’ or ‘plain’ indefinites: they carry a condition that constrains their use. Unlike specific indefinites, they need to appear in the scope of a licensing operator. AM model this licensing relation pragmatically. In other approaches (e.g. Chierchia , Giannakidou and Quer ) the licensing also involves a syntactic dependency. In §.. we will further see that the inventory of licensing operators, and consequently of compatible contexts, is subject to crosslinguistic variation. In a diachronic perspective, and especially with respect to the analysis of aliquis, these facts have great significance. They point to the fact that epistemic indefinites, unlike specific indefinites, have a fundamental meaning component in common with neighboring indefinites on Haspelmath’s () semantic map, like indefinites appearing in conditionals, questions, and, further to the right, free-choice and negativepolarity items: the dependence upon a structurally superior operator. Consequently, an epistemic indefinite is a more plausible starting point for the change leading the Romance continuations of aliquis to expand into right-hand functions in the semantic map (cf. §..). .. The ignorance implicature In order to exemplify the reasoning that derives the ignorance implicature in AM’s model, let us see how Spanish algún behaves under necessity modals, as in (). () Spanish (AM: ) Juan tiene que estar en alguna habitación de la casa ‘Juan has to be in some room of the house’

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

Under necessity modals the weaker claim made by the speaker by using an antisingleton indefinite (as opposed to one quantifying over a singleton domain) is made sense of by the hearer by assuming that the speaker wanted to avoid a false claim. This yields the ignorance implicature: the hearer will conclude that assuming singletonbased alternatives would be unwarranted according to the speaker’s epistemic state. () Derivation of the ignorance implicature with necessity modal (adapted from AM : –) a. Juan tiene que estar en alguna habitación de la casa ‘Juan has to be in algún room of the house’ b. assertion: 2 [∃x[x ∈ f (room) ∧ Juan is in x]] c. anti-singleton constraint: | f (room) |>  d. assertion in ‘context’: 2(J. is in the bedroom ∨ J. is in the living-room ∨ J. is in the bathroom) e. set of alternatives: {2(J. is in the bedroom), 2(J. is in the living-room), 2(J. is in the bathroom)} f. implicature: ¬2(J. is in the bedroom) ∧ ¬2(J. is in the living-room) ∧ ¬2(J. is in the bathroom) With possibility modals, as in (), the pragmatic reasoning based on avoidance of a false claim does not work and would lead to contradiction, since the speaker holds at least one competing alternative to be true. () Spanish (Alonso-Ovalle and Menéndez-Benito : ) Juan puede estar en alguna habitación de la casa ‘Juan may be in some room of the house’ AM propose that in this case anti-exhaustivity drives the inference: the hearer motivates the anti-singleton component as a signal to not prematurely exclude viable alternatives. The assertion is interpreted as meaning that there is at least one accessible world where Juan is in a room of the house, and is pragmatically enriched to mean that the speaker does not know which (otherwise he would have used an exhaustivity-triggering expression): this results again in the epistemic effect of ignorance. () Derivation of the ignorance implicature with possibility modal (adapted from AM : –) a. Juan puede estar en alguna habitación de la casa ‘Juan may be in algún room of the house’ b. assertion: ♦ [∃x[x ∈ f (room) ∧ Juan is in x]] c. anti-singleton constraint: | f (room) |>  d. assertion in ‘context’: ♦(J. is in the bedroom ∨ J. is in the living-room ∨ J. is in the bathroom)

OUP CORRECTED PROOF – FINAL, //, SPi

Latin aliquis and the class of epistemic indefinites



e. set of alternatives: {♦(J. is in the bedroom), ♦(J. is in the living-room), ♦(J. is in the bathroom)} f. implicature: ♦(J. is in the bedroom) → ♦(J. is in the living-room ∨ J. is in the bathroom) ♦(J. is in the living-room) → ♦(J. is in the bedroom ∨ J. is in the bathroom) ♦(J. is in the bathroom) → ♦(J. is in the living-room ∨ J. is in the bedroom) Concerning non-overtly modalized assertions, AM follow Kratzer and Shimoyama () in assuming that assertions are implicitly modalized, i.e that they contain a covert assertoric operator. The covert assertoric operator they assume is a kind of necessity operator (i.e. it has universal quantifying force), takes widest scope and ‘ranges over the speaker’s epistemic alternatives’ (AM: , cf. also Alonso-Ovalle and Menéndez-Benito :  and Chierchia ). The result is ignorance, and the implicature is derived as with overt necessity modals, i.e. is motivated as avoidance of a false claim. .. Epistemic indefinites and negation Alonso-Ovalle and Menéndez-Benito () mention negation only in passing, since they consider only pre-nominal algún, which cannot scope under negation (cf. chapter ). They discuss negation when arguing against a treatment of the epistemic effect as a presupposition: use of algún under no es verdad que ‘it is not true that’ (cf. ()) shows, according to AM, that the ignorance effect does not project, as a presupposition would instead do. () Spanish (AM: ) No es verdad que Juan salga con alguna chica del departamento de lingüística. ‘Juan is not dating any girl in the linguistics department’ However, since () is a denial context, and not a plain sentential negation, this aspect of the argumentation is not convincing. Moreover, it disregards the fact that, as we will see in detail in chapter , the behavior of algún with respect to negation differs substantially depending on its position in the DP: when pre-nominal, it never scopes under negation; when post-nominal, algún behaves similarly to an NPI, and it is in fact treated as such by Aloni and Port (). Of course, a unified treatment of the licensing conditions of epistemic indefinites under negation and other operators is desirable, and it is actually called for by the observation that elements inducing (maximal) domain widening function as NPIs in downward-entailing contexts and as free-choice items otherwise, English any being the textbook example (Kadmon and Landman , cf. also Haspelmath :  for the diachronic connection). German irgendein is also a case of epistemic indefinite with special meaning effects in negative contexts, although its behavior with negation is complex and cannot be

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

fully described here (cf. Kratzer and Shimoyama  and Penka : – for discussion). A clear-cut case of NPI behavior is (a), where irgendein co-occurs with a negative indefinite: () German (Penka : –) a. Niemand hat irgendein Auto gesehen ‘Nobody has seen any car’ b. Niemand hat ein Auto gesehen ‘Nobody saw a car’ In (a) no epistemic effect arises because computing the implicature under negation would be vacuous: the pragmatic reasoning yields only statements that are already implied by the assertion (cf. Kratzer and Shimoyama , Condoravdi , Lauer ). The use of irgendein in this context has no special epistemic effects, but yields a more emphatic reading than the indefinite article in (b), especially when irgendein is stressed, and forces narrow scope (whereas wide scope of the existential is possible with ein). Penka (: –) proposes that, in fact, irgendein can be used in the scope of negation only if the global meaning is richer (contributing emphasis or denial) than what would arise with a negative indefinite. This kind of paradigmatic competition with negative indefinites, which Penka treats as morphological blocking, would explain the fact that irgendein cannot occur with the sentential negator nicht in the intended NPI meaning, whereas its NPI use is possible when negation is expressed by negative indefinites or emphatic adverbials like keinesfalls ‘in no case’.14 This interestingly connects with what we will observe in chapter  about DPinternal inversion with the Romance continuations of aliquis: the emphasis added by the operation of inversion between the determiner and the rest of the DP allows for the use of epistemic indefinites under negation, despite their epistemic component being vacuous, and despite there being negative indefinites (e.g. Spanish ningún, Italian nessuno) as paradigmatic competitors. .. Crosslinguistic variation with epistemic indefinites In this section we take a quick look at crosslinguistic variation in the domain of epistemic indefinites, which has been systematically investigated in work by Aloni and Port (, ). They identify four main uses or functions of epistemic indefinites and survey their crosslinguistic distribution. Their treatment of the meaning of epistemic indefinites differs in some important respects from the analysis by AlonsoOvalle and Menéndez-Benito (). Aloni and Port consider the epistemic effect (which they call ‘modal variation’) a conventionalized implicature, and account for it, within a dynamic semantics framework, in terms of shifts in the conceptual covers (naming, ostension, description) used to identify the referent. In what follows, I will not go into the details of their analysis, but I will capitalize on their overview of some

14 Kratzer and Shimoyama (: ) note that irgendein under negation is rescued by stress, as in (i): ich hab’ nicht irgendwas gelesen = ‘I didn’t read anything’, ‘I didn’t read just ANYthing’. Penka (: ) interprets this reading as metalinguistic negation of the implicature.

OUP CORRECTED PROOF – FINAL, //, SPi

Latin aliquis and the class of epistemic indefinites



important dimensions of variation, with the aim of more precisely classifying Latin aliquis. Aloni and Port () designate with ‘use / function’ a meaning-context pair, and define the four basic functions of epistemic indefinites as shown in the table in () and exemplified with German irgendein in (). Note that the ‘specific modal variation’ function corresponds to the ‘specific unknown’ function in Haspelmath ().15 () Functions (meaning-context pairs) of epistemic indefinites in Aloni and Port () abbreviation function

context

meaning

(i) (ii) (iii) (iv)

episodic sentences under epistemic modal under deontic modal downward-entailing

ignorance ignorance indifference plain existential

spMV epiMV deoFC NPI

specific modal variation epistemic modal variation deontic free choice negative polarity

() German (from Aloni and Port ) a. Irgendein Student hat angerufen ( nämlich Peter) ‘Some student called (namely Peter)’ b. Maria muss irgendeinen Arzt geheiratet haben ‘Maria must have married some doctor’ c. Maria muss irgendeinen Arzt heiraten ‘Maria must marry some doctor or other’ d. Niemand hat irgendeine Frage beantwortet ‘Nobody answered any question’

(spMV) (epiMV) (deoFC) (NPI)

As discussed by Kratzer and Shimoyama (), German irgendein brings about an ignorance effect in positive declarative contexts (a), as well as under the scope of epistemic modals (b); under deontic modals, instead, it brings about a free choice flavor (c); when occurring in negative contexts (d) no special effect is recognizable, and the indefinite behaves like a plain existential (but see above the discussion on emphasis based on the contrast between (a) and (b)). In Aloni and Port’s system, an indefinite in a given language qualifies for a function if both of the following conditions are fulfilled (cf. Aloni and Port : ): (i) the use of the indefinite is grammatical in the context and (ii) it gives rise to the meaning specified by the function. That is, besides the case where an indefinite is simply ungrammatical in a given context, it is also possible that a given indefinite is indeed grammatical, but does not have the meaning specified by the function. Aloni and Port (: ) provide the example of English some: some is grammatical in the scope of a 15 Although Aloni and Port (: ) say that they follow Haspelmath () in their definition of ‘function’, their classification is in fact more fine-grained, in that they consider a particular meaning in a particular context. As discussed in chapter , in some cases Haspelmath looks instead for the availability of a form in a certain context (e.g. ‘irrealis’, ‘comparative’), independently of the meaning it takes.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

root modal such as deontic may, but does not have the free-choice reading expected for the deoFC function.16 The table in (), adapted from Aloni and Port (: ), shows the distribution of some epistemic indefinites crosslinguistically. Aloni and Port arrange the functions in a specific order to mirror the expectations arising in their theoretical account. Note that this way they preserve contiguity among same-polarity values for any given lexical item. () Crosslinguistic distribution from Aloni and Port (: ) yes: the indefinite is grammatical in the context and has the meaning indicated in the function no: the indefinite is ungrammatical in the context or it might be grammatical but not have the meaning indicated in the function spMV epiMV deoFC NPI German irgendein Spanish algún Italian un qualche Czech si Romanian vreun English any

yes yes yes yes no no

yes yes yes no yes no

yes yes no no yes yes

yes no no no no yes

We see that the ordering of the functions in the table might in principle correspond to the degree of widening of the domain (partial versus total), which I have interpreted as a fundamental determinant of variation in indefinite systems (cf. §..). However, Aloni and Port (), as said, choose a different analysis. They criticize approaches based on domain widening such as the one seen above, since the mechanism of domain widening does not account for the sensitivity of epistemic indefinites to the (deontic or epistemic) modal base, whereby they convey ignorance with epistemic modals, but indifference (a free-choice effect) with deontic modals. For our purposes, independently of the different solutions chosen by the various authors, it is crucial to observe that the class of epistemic indefinites shows crosslinguistic variation along the lines seen in (). It is, thus, not necessarily to be expected that we will find Latin aliquis in all of the contexts considered in the table. At the same time, Aloni and Port’s system provides a number of criteria in order to check to what extent Latin aliquis qualifies as an epistemic indefinite. In the next section I will proceed to add Latin aliquis to the picture seen in ().

16 Note, moreover, that the functions singled out by Aloni and Port () do not exhaust the possible readings of an indefinite in the modal contexts, since they are defined as scope under the modal. However, at least some epistemic indefinites systematically allow also for a wide-scope reading: (c) is in fact ambiguous and allows for wide scope of the indefinite over the deontic modal. Under the wide-scope interpretation of epistemic indefinites, the conveyed meaning is ignorance anchored to the speaker, and not to the attitude holder or the subject of the modal verb (cf. Alonso-Ovalle and Menéndez-Benito : –, Lauer ).

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



. Classical Latin aliquis We saw in §. that two meaning components have been particularly relevant for previous analyses of aliquis as a specific indefinite, which I rejected: (i) its wide scope: aliquis is able to have wide scope over other operators (modals, conditional, intensional verbs) and sometimes must have wide scope (as is most clearly the case with negation). (ii) its epistemic component, which has been related to an existential commitment: the identity of the witness to the existential claim is presented as unknown to the speaker. In this section I present data from Latin that support, instead, the classification of aliquis as an epistemic indefinite. I show that Classical Latin aliquis is not so specific: it does take wide scope with respect to many operators, but its epistemic contribution makes it ‘referentially vague’ (Giannakidou and Quer ), i.e., ‘less referential’ than a standard specific indefinite. After providing some remarks on the etymology of aliquis (§..), I proceed to discuss the distribution of data from modalized and negative contexts, according to the criteria for crosslinguistic comparison seen in §... We will see that aliquis behaves similarly to Spanish algún and Italian (un) qualche with respect to modals. In negative contexts, instead, it systematically scopes above the negative operator. In §., however, we will see that evidence for a change in this respect can be already found in Late Latin, especially since the fourth century ce. .. Etymological remarks Latin aliquis ‘some (or other)’ is an interrogative-based indefinite, from the ProtoIndo-European interrogative/indefinite stem *kw i-. The first morpheme originates from (the same stem of) Lat. alius ‘other’, yielding the original meaning ‘some or other’. The origin of aliquis must have been a syntactic combination of alius and quis, which was subject to univerbation at some point. The original additive component can be observed when aliquis is used in lists, meaning ‘something else’: () semper petunt aquam hinc aut ignem aut vascula always seek:pl water:acc from.here or fire:acc or vessel:acc aut cultrum aut veru aut aulam extarem aut or knife:acc or spit:acc or pot:acc for.entrails:acc or aliquid something:acc ‘they always look for water from here or fire or vessels or a knife or a spit or a pot for the entrails or something else’ (Plaut. Rud. –) As remarked in the dictionaries, early examples where it co-occurs with alius ‘other’ show us that the item’s etymology was not completely transparent for the speakers already in Early Latin: () dum aliud aliquid flagitii conficiat while other:acc some:acc roguery:gen plan:sg ‘while he plans some other piece of roguery’ (Ter. Phorm. )

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

Bortolussi (: –) remarks a meaning difference between aliquis and the combination alius quis (also quis alius) found in Latin texts: aliquis expresses free variation within a set of possible values, whereas alius quis picks up a unique individual, denoted by means of a contrast with a previously explicitly mentioned individual. In the example below, provided by Bortolussi (: ), the individual denoted by quis alius has to be chosen in the set of the Roman kings (expressed in the partitive genitive) and is contrasted with the previously mentioned Tarquinius Priscus: () sedem eam acceperat a Tarquinio Prisco, district:acc that:acc receive:sg from Tarquinius:abl Priscus:abl seu quis alius regum dedit or some:nom other:nom king:gen give:sg ‘he had received the district as a settlement from Tarquinius Priscus, or another of the kings gave it to him’ (Tac. ann. ..) Given the general impossibility for quis to occur in episodic contexts by itself (cf. §.), we have to conclude that alius is here the denoting term, and quis adds indeterminacy with respect to the identification of the denotation. The resulting meaning is very similar to the contribution of aliquis. In addition, however, alius quis has to establish an anaphoric relation with an explicitly given set. Such a meaning ingredient is absent with aliquis: this may be explained as the result of a grammaticalization process, where meaning generalization (‘bleaching’, cf. §...) consists in the loss of a contextual condition (namely, anaphoricity).17 The alternative denotations relevant for the interpretation of aliquis will then have to be retrieved non-anaphorically. This can be argued to be the ingredient triggering minimal domain widening with aliquis, i.e. the variation effect observed by Bortolussi (: ) when he describes the meaning of aliquis as entailing the existence of a set of individuals that can freely vary as a value for the indefinite. .. Prototypical ‘realis’ contexts One of the arguments found in the literature for analyzing aliquis as a specific indefinite is that it can occur in prototypical ‘realis’ contexts, as subject of affirmative declarative sentences with a perfective past or ongoing present predicate. I have already mentioned, however, that these instances are very rare: cf. example (), repeated below in an extended form as (). () dixerat aliquis leniorem sententiam . . . ut primo say:sg someone:nom mild:comp.acc opinion:acc . . . as first Calidius . . . ut M. M. Marcellus . . . ut M. M.:nom Marcellus:nom . . . as M.:nom Calidius:nom . . . as M.:nom Rufus Rufus:nom ‘someone had expressed a milder opinion . . . like first of all M. Marcellus . . . like M. Calidius . . . like M. Rufus’ (Caes. BC .) 17 It is possible that anaphoricity, as reference to a contextually established set, bleaches into an existential entailment, which according to Bertocchi et al. () is what distinguishes aliquis from quis.

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



In this case aliquis is used to avoid an exhaustivity inference with respect to the subsequent list of the people who spoke during the assembly: not everyone is mentioned in Caesar’s report. Examples are found slightly more frequently in the plural, where in general it is difficult to detect epistemic effects: () et sunt aliqua epistulis eorum inserta, ex and are some:nom letter:dat their:gen inserted:nom from quibus mutua malignitas detegitur which:abl mutual:nom ill.will:nom expose:sg.pass ‘and certainly there are some passages in their letters which show mutual ill-will’ (Tac. dial. .) In the plural, the Romance continuations of aliquis behave as plain indefinites (but see discussion in §..). .. Generic environments Much more frequent are uses in generic contexts (cf. the adverbs semper ‘always’, persaepe ‘very often’ in ()), where the absence of indication of a specific witness to the claim is felicitous: () a. aut aliqua mala crux semper est, quae or some:nom bad:nom torment:nom always be:sg who:nom aliquid petat. something:acc ask:sg ‘or there is always some tormentor demanding something.’ (Plaut. Aul. ) b. Unum genus excipio sermonis in quo persaepe one:acc type:acc exclude:sg speech:gen in which:abl very.often aliquid dicitur quod te putem something:nom say:sg.pass which:acc you:acc think:sg nolle dici not.want:inf say:inf.pass ‘I make an exception only for one type of speech in which very often something is said which I think you would not want to be said’ (Cic. fam. ..) In some cases aliquis is accompanied by unus ‘one’ (on which we will say more in §.), cf. (). In this and similar examples unus precedes aliquis, differently from the structure from which the Romance continuations of aliquis (e.g. Sp. alg-ún, It. alc-uno, etc.) must have originated. Also here, as in the combination of aliquis with alius seen in (), the role of aliquis is to add a variation condition. () Haec vitia unus aliquis inducit, ... this:acc fault:acc one:nom someone:nom introduce:sg . . . ceteri imitantur other:nom imitate:pl ‘There is someone who introduces these faults, . . . the others go on imitating’ (Sen. epist. .)

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

Similar are also the examples where aliquis is found with the indicative in an iterative sense: () cum ad aliquod oppidum venerat, eadem lectica usque in cubiculum deferebatur when to some:acc town:acc come:sg same:abl litter:abl until in chamber:acc bring:sg.pass ‘when he would come to some town he would be carried in the same litter up to his chamber’ (Cic. Verr. ..) I consider generic readings evidence for the fact that aliquis may be embedded under (covert) operators (in this case, the covert operator GEN), similarly to what happens in modalized environments. .. Modalized environments Recall the Late Latin example that Haspelmath provides for the ‘specific unknown’ function of aliquis, seen in (b) and repeated below: () Tetigit me aliquis, nam ego novi virtutem touch:sg me:acc some:nom in.fact I:nom know:sg power:acc de me exisse from me:abl come.out:inf ‘Someone did touch me, for I was aware that power had gone out of me’ (Vulg. Luc. .) Here in fact the assertion has to be interpreted as implicitly modalized: ‘I believe / I am certain that someone touched me’, since Jesus (who is the speaker in the context) infers that someone touched him from indirect evidence (expressed in the second part of the clause, introduced by nam). That means that this context cannot count as a ‘prototypical realis’ one. In fact, the use of aliquis in (overtly or covertly) modalized contexts is very frequent. Traina and Bertotti (: ) note that aliquis is often used to talk about future states of affairs. This is amply confirmed by my corpus study; some examples with future tense are given below for Early Latin (–) and Post-Classical Latin (). For the modal use of the future tense in Latin see Magni (: –). () deus respiciet nos aliquis god:nom protect:sg us:acc some:nom ‘some god will protect us’,  namely Jupiter (Plaut. Bacch. a) () ostium pultabo atque intus evocabo aliquem foras door:acc knock.at:sg and inside call:sg someone:acc outside ‘I will knock at the door and I will call someone outside from inside the house’ (Plaut. Pseud. ) () nam aut incendium oportet fiat, aut for either fire:nom be.necessary:sg become:sg or in vicinia animam abiciet aliquis someone:nom in neighborhood:abl soul:acc give.up:sg ‘Either there must be a fire, or someone close by is just going to give up the ghost’ (Petr. )

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



In () Trimalchio is commenting, upon hearing a rooster crow, that it must mean bad luck, and makes a prediction on possible outcomes: clearly he is not thinking of a specific referent here. Epistemic possibility, compatible with the meaning of aliquis, can also be conveyed by indicative present tense, as for instance in (), where the adverbs forsitan ‘perhaps’ and utique ‘be it as it may’ express the speaker’s epistemic attitude:18 () hunc forsitan, proclamo, in aliqua parte terrarum this:acc perhaps say:sg in some:abl part:abl land:gen secura expectat uxor, forsitan ignarus unsuspecting:nom wait:sg wife:nom perhaps unaware:nom tempestatis filius, aut pater; utique reliquit aliquem, storm:gen son:nom or father:nom anyway leave:sg someone:acc cui proficiscens osculum dedit who:dat leave:pt.nom kiss:acc give:sg (context: the speaker is looking at the body of a man dead at sea) ‘I say: perhaps in some part of the world an unsuspecting wife is waiting for him, perhaps a son who is not aware of the storm, or a father; at any rate, he left someone, whom he kissed in leaving’ (Petr. ) The use of aliquis in implicitly modalized contexts is very richly exemplified in Latin texts, and seems to be one of its primary functions. Turning to overt modals, Latin can express epistemic and deontic modality by means of subjunctive forms or by means of a set of modal verbs (cf. Magni : –). Modal verbs for possibility are the verb posse ‘can’ or the impersonal licet ‘it is permitted’ / ‘it is possible’. However, in the Classical texts I analyzed, these verbs most often express dynamic (ability) rather than epistemic or deontic modality: () ut, dum tempus anni esset idoneum, so.that until time:nom year:gen be:sg convenient:nom aliquid negoti gerere possem something:acc duty:gen accomplish:inf can:sg ‘so that I am enabled to perform some tasks, while the season is still convenient’ (Cic. fam. ..) While aliquis is frequently found in combination with these modals, I could not find many clear examples of epistemic or deontic readings. One instance where the deontic possibility reading is plausible is given in (): here the modal possum ‘can’ is embedded in the complement clause of the verb concedo ‘permit’:

18 Note that forsitan ‘perhaps’ is usually found with the subjunctive in Classical prose, at least until Cicero; the use of the indicative in () belongs to the colloquial style.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

() vos quaeso date hoc et concedite pudori you:voc please give:pl this:acc and allow:pl modesty:dat meo, ut aliquam partem de istius impudentia my:dat that some:acc part:acc of this:gen shamelessness:abl reticere possim keep.silent:inf can:sg ‘Please, do permit this, and grant to my modesty that it may be allowed to refrain from describing some of his shamelessness’ (Cic. Verr. ..) The part of Verres’ misdeeds that Cicero wants to pass over in silence is not further specified, although clearly some alternatives can already be discarded (Cicero’s speech is in fact devoted to enumerating and condemning Verres’ misdeeds). The effect is one of modal variation, dependent on the possibility modal. Epistemic possibility can be expressed by the potential subjunctive: aliquis combines with the present or the perfect potential subjunctive in expressions such as (). () aliquis dicat someone:nom say:sg ‘someone might say’ (e.g. Ter. And. ) The evidence with overt modals is more abundant in the case of necessity, both for epistemic and deontic modality. Overt modal verbs expressing such notions in Latin are debeo ‘must’, the impersonal oportet ‘it is necessary’, and the phrase necesse est ‘it is necessary’. The examples in () show that, both with epistemic (a) and with deontic (b–c) modal base, the meaning contribution of aliquis is ignorance, and the modal variation effect observed with Spanish algún obtains in Latin as well. In (c) the indefinite aliquid has narrow scope also with respect to the universal quantifier omnia. () a. necesse est aliquem dixisse municipem aut necessary be:sg some:acc say:inf fellow.citizen:acc or vicinum neighbor:acc (context: when something like that happens, judges, you usually say the following) ‘some fellow citizen or neighbor must have told him’ (Cic. S. Rosc. ) necesse est b. sed si vasta sunt oliveta, but if big:nom be:pl olive.groves:nom necessary be:sg aliqua pars eorum maturo fructui some:nom part:nom they:gen ripe:dat harvest:dat reservetur reserve:sg.pass ‘but if the olive-groves are very big, it is necessary that some part be reserved to the harvest’ (Colum. .)

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



c. omnia enim ista referri ad aliquid everything:acc in.fact this:acc refer:inf.pass to something:acc necesse est necessary be:sg (context: Quintilian discusses the rhetorical organization of a juridical cause, which at a certain stage has to discuss questions like ‘if someone has the right to accuse’, ‘if he should do it now’, ‘if he should do it this way’) ‘because it is necessary that all of these questions be referred to something’ (Quint. inst. ..) Deontic modality can be expressed in Latin also by means of a periphrastic construction combining the gerundive with a form of the verb ‘to be’: in () it is not the case that there is something specific that has to be contributed to the community; aliquid is embedded under the modal and also in this case contributes ignorance with respect to the identification of the referent. () quare et his utendum est et semper therefore and this:abl adopt:ger be:sg and always aliquid ad communem utilitatem afferendum something:nom to common:acc interest:acc contribute:ger ‘Therefore, one should adopt these principles and always contribute something to the common interest’ (Cic. off. .) We can conclude that aliquis under deontic modals does not have the free-choice import that has been observed e.g. for German irgendein. This supports also for Classical Latin aliquis an analysis in terms of minimal domain widening: this indefinite signals, similarly to Spanish algún, that its domain cannot be reduced to a singleton set, but at the same time it does not trigger maximal domain widening. .. Negation and other downward-entailing contexts Classical Latin aliquis can and must scope out of negation: cf. () and the example in (): () Inter haec tamen aliquis non gemuit. among this:acc though someone:nom not lament:sg ‘Among all this someone, though, did not lament’ (Sen. epist. .) This is a very robust generalization, for which no exceptions are observed during the Classical Latin stage, independent of register and text type. Also with some examples that at first would suggest a narrow-scope reading, the broader context shows that in fact the indefinite takes scope above negation: in () Varro is talking about the loss of noun endings and uses a metaphor comparing this phenomenon to statues that do not have the head or ‘some other parts’ (and not ‘any other part’, as is clear from the continuation ‘in the remaining ones’):

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

() signa quae non habent caput aut aliquam statue:nom which:nom not have:pl head:acc or some:acc aliam partem, nihilo minus in reliquis membris other:acc part:acc nothing less in remaining:abl part:abl eorum esse possunt analogiae, sic in vocabulis casuum they:gen be:inf can:pl regularity:nom so in words:abl case:gen possunt item fieri iacturae can:pl similarly happen:inf loss:nom ‘For as some statues lack the head or some other part without destroying the regularities in their other limbs, so in words certain losses of cases can take place, with as little result.’ (Varro ling. ..) For expressing existential quantification in the scope of negation, Classical Latin has a special negative-polarity series, formed by the pronoun quisquam and the adjective ullus: () a. neque quisquam omnino consisteret and.not anyone:nom at.all stop:sg ‘and no one at all stopped’ (Caes. BC .) b. sine timore ullo without fear:abl any:abl ‘without any fear’ (Caes. BG .) c. quae non modo numquam nocet cuiquam, sed which:nom not only never harm:sg anyone:dat but contra semper addit aliquid instead always add:sg something:acc ‘not only does [Justice] never cause anyone harm, but on the contrary it always adds some benefit’ (Cic. fin. .) Moreover, as we will discuss at length in chapter , Latin has a series of negative indefinites, that may not co-occur with other negative elements in the single-negation reading: () interit tamen nemo die:sg nevertheless nobody:nom ‘nevertheless nobody died’ (Caes. BC .) If aliquis co-occurs with negation, either it is interpreted outside its scope or the negation is interpreted as metalinguistic (cf. §...): cf. the dialogue in (), where negation is brought about by the negative verb nolo ‘not-want’. () a. PH. Iam edes aliquid. now eat:sg something:acc ‘Now you shall eat something’ b. CV. Nolo hercle aliquid, certum quam not.want:sg for.Hercules something:acc certain:acc than aliquid mavolo. something:acc prefer:sg ‘for Hercules, I do not want “something”: I prefer “certain” to “something”’ (Plaut. Curc. –)

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



In case of double negation, aliquis is possible, as expected given the overall positive polarity of the context, cf. (). In these uses, aliquis typically expresses a relevant quantity, similarly to the examples we will see in §... () non sine aliquo divino numine not without some:abl divine:abl providence:abl ‘not without some divine providence’ (Cic. prov. ) More rarely aliquis can co-occur with the preposition sine ‘without’ in a downwardentailing context, as in (a). The interpretational difference with respect to the NPI ullus in (b) consists in the fact that aliquis, also in this context, expresses the existence of a relevant quantity, especially in combination with abstract mass nouns, while the NPI is a maximal domain widener (cf. Traina and Bertotti : ). () a. sine aliquo vulnere without some:abl wound:abl ‘without relevant losses’ (Caes. BC ) b. sine ulla spe without any:abl hope:abl ‘with no hope’ (Cic. Att. ..) In other downward-entailing contexts, such as e.g. the restriction of a universal quantifier, most often the negative polarity quisquam / ullus-series is found: () omnes . . . qui ullam agri glebam possiderent everyone:nom . . . who:nom any:acc land:gen clod:acc possess:pl ‘All those who owned any clod of land’ (Cic. Verr. ..) Aliquis as simple existential is also possible: () omnes qui aliquid scire videntur everyone:nom who:nom something:acc know:inf seem:pl ‘Everyone who seems to know something’ (Cic. parad. ) Also in this case it is possible to recognize a meaning difference (Devine and Stephens : ): aliquis does not refer to a scalar end point. This fact is quite important in diachronic perspective, and we will come back to it: it shows that also in this case no exhaustivity effect is connected with aliquis. In the standard of comparison the NPI-series and free-choice indefinites, e.g. quivis, are licensed, cf. Devine and Stephens (: ), but aliquis is—more rarely— possible: pluris esse necesse est quam () Mundum universum world:acc universe:acc more:gen be:inf necessary be:sg than partem aliquam universi part:acc some:acc universe:gen ‘The whole universe is necessarily worth more than some part of the universe’ (Cic. nat. .) An interesting aspect of the interaction of aliquis with negation concerns its cooccurrence with the negative complementizer n¯e, which functions as negation in prohibitions and a number of nonveridical environments (cf. Ernout and Thomas : ).

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

() ne relinquatur aliquid erroris in verbo lest leave:sg.pass some:nom mistake:gen in word:abl ‘lest some misunderstanding be left in the interpretation of the word’ (Cic. Tusc. .) Orlandini (a: –), Bortolussi (: –), Devine and Stephens (: –) all agree that the negation introduced by n¯e has special semantic and pragmatic properties that can account for the appearance of aliquis with ‘narrow scope’ in this context. With different implementations, they take n¯e to be able to introduce, in some of its contexts of use, a particularly high negation, similar to what is found in denials, in that it targets the assertability of the entire clause, without interacting with its parts (cf. further §...). .. Protasis of conditional As seen in §., in the antecedent of a conditional, aliquis is found rarely. Its substitute here and in other irrealis contexts (especially interrogatives and comparatives) is chiefly quis (a) and more rarely quisquam (b). Quis is always in the scope of an operator, which licenses it. As mentioned in §., one additional constraint with quis concerns its prosodic-syntactic status: it has to be enclitic upon its licensor (Bortolussi ). () a. de quo etiamsi quis dubitasset antea about which:abl even.if someone:nom doubt:sg before ‘even if anyone had been in doubt about this before’ (Cic. Sull. ) b. Si quemquam nactus eris qui perferat, litteras if anyone:acc find:pt be:sg who:nom bring:sg letter:acc des antequam discedimus give:sg before leave:pl ‘If you find anyone to deliver it, let me have a letter before I leave’ (Cic. Att. ..) In the cases where aliquis is indeed found in the protasis, it does not show the exceptional wide-scope behavior observed with quidam in () and is interpreted in the scope of the conditional, triggering the ignorance interpretation. istorum videret . . . () Si L. Memmius aliquem if L.:nom Memmius:nom someone:acc this:gen see:sg . . . utrum illum civem excellentem an atriensem whether that:acc citizen:acc outstanding:acc or butler:acc diligentem putaret? diligent:acc consider:sg ‘If L. Memmius saw one of your lot . . ., would he think him an outstanding citizen or a diligent butler?’ (Cic. parad. ) The alternation with quis is remindful of the alternation between English some and any in conditional clauses. In the few examples in which it occurs in the antecedent of a conditional, aliquis is frequently accompanied by a partitive specification, as

OUP CORRECTED PROOF – FINAL, //, SPi

Classical Latin aliquis



in () (cf. Bertocchi and Maraldi  for further examples), which reinforces the impression that the existence of a possible witness to the claim is not at issue (cf. Bertocchi and Maraldi : ; Devine and Stephens : ). Another possible factor determining the presence of aliquis in conditionals could be the avoidance of quis in contexts where also other operators are present and quis would be trapped in their scope. For instance, in () aliquis takes scope above negation and yields the desired scope relations conditional > aliquis > negation: () si sustinere aliquis aceti vim non potest, if sustain:inf someone:nom vinegar:gen strength:acc not can:sg vino utendum est wine:abl use:ger be:sg ‘if some patient cannot bear the strength of vinegar, wine must be used’ (Cels. ..E) This would connect to the already examined cases of co-occurrence of aliquis and quis in the antecedent of a conditional, where the use of aliquis disambiguates the scope relation and the value for quis co-varies depending on the value of aliquis, cf. (). .. Scalar uses Very frequently in my corpus, aliquis receives a scalar interpretation, expressing the existence of an undetermined degree, with variable meaning effects according to the context, which cannot be reconciled with a classification as specific indefinite. These uses have not been discussed in connection to epistemic indefinites in the literature, but Bertocchi et al. (: –, –, ) have linked them to factors such as scale orientation and scale type (quality versus quantity). In some cases, according to Bertocchi et al. (), aliquis behaves similarly to a free-choice pronoun. More research would be needed to assess whether also these uses can be accounted for by assuming an anti-singleton constraint (rather than a requirement of maximal widening, as in free-choice indefinites) on the quantificational domain of aliquis. Although these uses have not been discussed very often, they are in fact robustly attested for well-known epistemic indefinites (cf. Spanish algo, Amaral ). In §.. we saw that ignorance with epistemic indefinites can concern the individual satisfying the claim, or the kind of which it is a representative. We might expect that ignorance may be applied to a scale of degrees as well. In this section I will provide some data from Classical Latin, because these scalar uses may represent bridging contexts in a diachronic perspective. In Early and Classical Latin we find aliquis with numerals or measure words, to express approximation (Bertocchi et al. : –): () Elleborum potabis faxo aliquos viginti dies hellebore:acc drink:sg make:sg some:acc twenty:acc days:acc ‘I will see to it that you will drink hellebore for some twenty days’ (Plaut. Men. )

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

The use is present also in Late Latin (cf. Löfstedt : –): () Ac sic ergo aliquo biduo ibi tenuit nos and so therefore some:abl two.day:abl there keep:sg we:acc sanctus episcopus holy:nom bishop:nom ‘And so the holy bishop kept us there for some two days’ (Itin. Eg. .) Aliquis may apply to a quantity scale, or to a quality scale. We find aliquis also to express a relevant quantity: () a. vicitque dolorem aliquem domesticum patriae caritate win:sg pain:acc some:acc private:acc country:gen love:abl ‘and has overcome some private pain by his love of his country’ (Cic. Phil. .) b. ut . . . aliquam Caesar ad insequendum facultatem so.that . . . some:acc Caesar:nom to pursue:ger possibility:acc haberet have:sg ‘so that Caesar might have some means of pursuing him’ (Caes. BC .) With mass nouns, the use of aliquis triggers quality alternatives: it is oriented towards a low point on a scale, yielding a flavor of indifference (these are the uses that Bertocchi et al. :  ff. consider ‘next to a free-choice pronoun’): () sic te ipse abicies atque prosternes, ut so you:acc yourself:nom lower:sg and bend:sg so.that nihil inter te atque inter quadripedem nothing:nom between you:acc and between four.feeted:acc aliquam putes interesse? some:acc think:sg differ:inf ‘will you make yourself so abject and so low an outcast as to deem that there is no difference between you and any four-feeted beast?’ (Cic. parad. .) These scalar uses could be relevant in a diachronic perspective since they show aliquis operating on ordered alternatives, i.e., on scales. As we will see in the next section, starting in Late Latin aliquis, also thanks to its combination with unus ‘one’, will start to be used as an NPI, expressing a scalar endpoint and functioning, this way, as negation strengthener. In the scalar uses of aliquis discussed here we see how the alternatives of the minimally widened domain can be ordered and organized into a scale, possibly representing an intermediate step of the development. .. Interim summary: aliquis as an epistemic indefinite Classical Latin aliquis seems to entertain an intermediate status with respect to its denotational properties. It has a preference for wide scope with respect to co-occurring operators, which it shares with specific indefinites like Engl. a, some, German ein bestimmter; however, with respect to its ignorance component, it

OUP CORRECTED PROOF – FINAL, //, SPi

Aliquis in Late Latin



resembles the class of epistemic indefinites. The ignorance effect may be treated as an implicature along the lines of the analysis discussed in §.. With respect to the four functions discussed in a comparative perspective by Aloni and Port (), aliquis behaves similarly to the Romance epistemic indefinites included in the table in (), in that it does not allow for free-choice uses under deontic modals, unlike German irgendein or English any: the meaning contribution is always ignorance. This suggests that an anti-singleton constraint, rather than maximal domain widening, is imposed on the quantificational domain of aliquis. An analysis of aliquis in terms of specificity cannot account for the cases where it takes narrow scope with respect to the modal operators. We further saw that aliquis can also appear in unembedded positive contexts: here aliquis patterns with algún and un qualche in the table in (), and unlike Romanian vreun. Many of these contexts seem to be implicitly modalized, thus suggesting the presence of a covert assertoric operator, as in the treatments of epistemic indefinites discussed in §... Finally, with respect to negation, Classical Latin aliquis patterns like Italian un qualche in not having narrow readings. In the next section we will see that this changes already in Late Latin, preluding the Romance developments.

. Aliquis in Late Latin In the previous sections we have seen how modeling the constraint that epistemic indefinites impose on their quantificational domain in terms of minimal domain widening can derive the ignorance component as a pragmatic effect. The wide-scope uses, as well as the apparently unembedded ones, are in fact quantified over by a covert assertoric operator, and the epistemic component is anchored to the speaker. In intensional contexts narrow-scope readings can arise, where epistemic indefinites impose the anti-singleton constraint on the worlds in the quantification domain of the modal. This analysis accounts for contextual variability (in fact, predicts it) better than an analysis based on specificity. My aim is to exploit this analysis in a diachronic perspective, as anticipated in §..: by introducing (minimal) domain widening, epistemic indefinites represent a bridging context towards other functions, like the free choice or the NPI function. At the same time, the crosslinguistic survey in Aloni and Port () shows that often epistemic indefinites display only a subset of the uses seen in the table in (), hinting at possible diachronic processes (loss / gain of semantic features) affecting the inventory of functions. .. Aliquis and negation in Late Latin In this section we examine some important developments happening in Late Latin. The clearest context where we can observe them is in the presence of sentential negation, on which I will focus here. We will then see in chapter  that aliquis undergoes further interesting developments in the process of differentiation happening in the Romance languages.

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

The Classical Latin subsystem on the left-hand side of Haspelmath’s () map is largely preserved in Late Latin. The Vulgate translation of the New Testament (fourth century ce) does not show any substantial difference with respect to the Classical norm (cf. also Bortolussi and Sznajder ). The only exception, as we saw in §..., is the innovative use of unus as a (completely optional) means of marking specificity. Deviations from the norm in the Vulgate are coherent with phenomena observed in non-translated contemporary varieties of Latin. Despite the general conservativeness of the system, we see important signs of an expansion of aliquis into previously banned contexts: the most significant phenomenon in this respect is represented by narrow readings under the scope of negation or in other downward-entailing contexts. Particularly interesting data can be found in texts between the late third and the early fifth centuries ce, an age that also witnesses the first steps in the development of Negative Concord (cf. Molinelli , , and chapter ). Some examples for the expansion of aliquis into negative contexts within the scope of negation are already found in the Vulgate: () si ergo corpus tuum totum lucidum fuerit non if then body:nom your:nom all:nom shining:nom be:sg not habens aliquam partem tenebrarum erit lucidum totum having any:acc part:acc darkness:gen be:sg shining:nom all:nom ‘If therefore your whole body will be full of light, with no dark part in it, it will be wholly illumined’ (Vulg. Luc. .) Similar examples where aliquis scopes under negation and functions like a negativepolarity item, bringing about extreme domain widening, are encountered in the Itinerarium Egeriae: () Esca autem eorum quadragesimarum diebus haec food:nom then they:gen Lent:gen day:abl this:nom est, ut nec panem . . . nec oleum gustent, nec be:sg that and.not bread:acc . . . and.not oil:acc taste:pl and.not aliquid, quod de arboribus est, sed tantum anything:acc which:nom from tree:abl be:sg but only aqua et sorbitione modica de farina water:acc and gruel:acc small:acc from flour:abl ‘For their food during the days of Lent is as follows: they eat neither bread nor oil nor anything that grows on trees, but only water and a little gruel made of flour’ (Itin. Eg. .) This use becomes more frequent in later texts. Bertocchi et al. (: –, ) note that Gregory of Tours (sixth century ce) has many cases of co-occurrence of aliquis and quisquam (an NPI), where no semantic difference between the two indefinites can be detected.

OUP CORRECTED PROOF – FINAL, //, SPi

Aliquis in Late Latin



() ut nec ibi quidem aut in via aliquem expoliarent that and.not there indeed or in street:abl anyone:acc spoil:pl aut res cuiusquam diriperent or thing:acc anyone:gen rob:pl ‘that either there or in the street they must not spoil anyone or steal anyone’s things’ (Greg. Tur. Franc. .) In particular the combination with the negative particle n˘ec (on which see chapter ) seems to trigger a widened-domain interpretation: () sed nec pulmentum aliquod utebatur but and.not food:acc any:acc use:sg ‘but he would not even take any food’ (in addition to not taking any drink) (Greg. Tur. Franc. ..) In later Medieval Latin a new formation aliqualis appears, which in some contexts corresponds to quicumque / ullus and in other contexts to aliquis, confirming the ongoing confusion between a specific unknown and a free-choice reading. Another interesting example from the sixth century ce, where the context makes clear that all the previously listed alternatives have to be excluded, is presented in (): () si quis crudos lactes vult bibere, mel if someone:nom raw:acc milk:acc want:sg drink:inf honey:acc habeant admixtum vel vinum aut medus; et si non have:pl mixed:acc or wine:acc or mead:acc and if not istis poculis, sale fuerit aliquid de be:sg something:nom from this:abl beverage:abl salt:nom/acc mittatur modicum put:sg.pass little:nom/acc ‘if one wants to drink raw milk, they should mix honey, or wine, or mead; and if there won’t be any of these beverages, add a little salt’ (Anthim. ) These observations suggest that the development of aliquis into an NPI must have been completed, at least in some geographical areas, by the fourth to fifth centuries ce. As mentioned, during this time we also start seeing signs of disruption in the Latin system of negation. Signs of the development of Negative Concord are already observed and acknowledged by Latin grammarians from the fourth century ce on, as discussed by Molinelli (). The example in () shows a case from an early biblical translation where a negative indefinite exceptionally co-occurs with the negative marker non in a single-negation reading, something excluded by the Classical norm. The Vulgate translation ‘revises’ this slip and uses the Classical quisquam: () non respondes nihil? not answer:sg nothing:acc ‘You don’t answer anything?’ (Vetus Latina, Marc. .) cf. Vulgata: non respondes quidquam?

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

Molinelli (: –), in discussing the development of Negative Concord, cites interesting examples from Augustine (fourth century ce), where he comments on Bible translations and indicates what would be the correct Latin usage. Interestingly for us, Augustine reacts to Negative Concord structures, which he deems unacceptable, with alternative formulations that are correct in his opinion. In fact, however, in () his suggested formulation contains a non-Classical use of aliquis as NPI, which must have been fully grammatical in his variety. () ‘Non est relictum viride nihil in lignis’ not be:sg left:pt green:nom nothing:nom in trees:abl dicendum fuit more locutionis nostrae: ‘non est say:ger be:sg use:abl language:gen our:gen not be:sg relictum viride aliquid in lignis’ left:pt green:nom anything:nom in trees:abl ‘ “there isn’t nothing (nihil) green left on the trees” should rather be according to our way of speaking: “there isn’t anything (aliquid) green left on the trees” ’ (Aug. Loc. Hept. , de Ex. .) My corpus confirms that what we see in () is not just an occasional slip: many instances are documented where Augustine uses aliquis in negative contexts with a narrow-scope reading. Some clear examples are given in (). Significantly, they all come from the Sermones, a collection of homilies written in a colloquial language, to be understood by the common people. In these texts negative contexts containing aliquis are very frequent, and in many cases the reading for the indefinite is unambiguously that of an NPI. I did not find such frequent and unambiguous uses in the other works by Augustine that I examined, the Confessiones and the De doctrina christiana, which are written in a more formal style. () a. res ipsa, quam vis dicere, quae thing:nom itself:nom which:acc want:sg say:inf which:nom corde concepta est, non est alicuius linguae, heart:abl conceive:pt be:sg not be:sg any:gen language:gen graecae, nec latinae, nec punicae, nec nec and.not Greek:gen and.not Latin:gen and.not Punic:gen and.not hebraeae, nec cuiusquam gentis. Hebrew:gen and.not any:gen people:gen ‘the thing itself that you want to say, that has been conceived in your heart, is not of any language, neither Greek, nor Latin, nor Punic, nor Hebrew, nor of any people’ (Aug. serm. ) b. dicitur tibi: non moechaberis, id est, non tell:sg.pass you:dat not fornicate:sg that:nom be:sg not eas ad aliquam aliam praeter uxorem tuam go:sg to any:acc other:acc except wife:acc your:acc ‘it is told to you: you shall not commit adultery, that is, you shall not go to any other woman except your wife’ (Aug. serm. )

OUP CORRECTED PROOF – FINAL, //, SPi

Aliquis in Late Latin



In many different documents of Late Latin there is, thus, evidence that aliquis has expanded into the right-hand side of Haspelmath’s () semantic map and starts to be used in the direct negation context. At the same time, it maintains the functions that it had in Classical Latin and further expands into polarity contexts to the detriment of quis (cf. Bortolussi and Sznajder ). Note that these innovative NPI uses coexist with epistemic uses, cf. (): () si non potest, aliqua necessitate impeditus . . . if not can:sg some:abl necessity:abl hinder:pt.nom ‘if he cannot, becase he is hindered by some necessity or other . . .’ (Aug. serm. ) In the antecedent of conditionals, no extreme domain widening is observed, and aliquis patterns as in Classical Latin (although it is found more often in this context): () si enim aliquem sanctum non audes invitare in if namely some:acc holy:acc not dare:sg invite:inf in domum tuam, nisi prius mundaveris domum home:acc your:acc if.not previously clean:sg home:acc tuam, ne oculi eius patiantur iniuriam . . . your:acc lest eyes:nom he:gen suffer:pl offense:acc ‘for if you do not dare to invite some holy man to your home, if you have not previously cleaned your home, lest his eyes suffer an offense . . .’ (Aug. serm. a) Also the plural behaves as in Classical Latin, and outscopes negation: () sed aliqui non fecerunt but some:nom not do:pl ‘But some people did not do so’ (Aug. serm. a) Augustine seems to have a very Romance grammar with respect to aliquis, corresponding to a large extent to what is observed in the Early Romance varieties (cf. chapter ). A relevant observation emerging from the corpus I examined is that negation is the most innovative context for the behavior of aliquis in Late Latin. This will connect in interesting ways to what I will conclude in chapter  concerning the interaction of negation and focus, and the use of the Romance continuations of aliquis as strengtheners of negation. .. Combination with unus The last issue I would like to briefly address concerns the form of aliquis: its Romance continuations show univerbation with unus ‘one’. Although the form *alicunus is never attested in Late Latin documents, it can be safely reconstructed as the source of the various Romance continuations. Given what we saw in §... about the incipient use of unus as a potential specificity marker, the following question arises: which ‘one’ combines with aliquis? the ‘old’ numeral or the ‘new’ specificity marker?

OUP CORRECTED PROOF – FINAL, //, SPi



Specific and epistemic indefinites in Latin

The traditional explanation for the univerbation of aliquis and unus has interpreted unus as a reinforcer of the specificity value of aliquis (e.g., Manoliu-Manea ). But the situation that we just saw for Late Latin suggests an alternative interpretation of the process, namely that unus contributes to the salience of scalarity in the meaning of aliquis. Haspelmath (: , ) remarks that ‘[l]ike generic nouns, ‘one’ has a tendency to become restricted to negative-polarity and negative functions’. Being an expression of minimal quantity, it qualifies as a ‘scalar-endpoint indefinite’ (cf. Lat. ullus < *un-elos ‘one’ + diminutive, cf. Engl. any; n¯on < * ne-oenom ‘not’ + ‘one’, It. nessuno < nec ipsum unum ‘not even one’). Some uses of aliquis co-occurring with unus are found, albeit not frequently, in Classical Latin: they occur in generic statements (cf. ()) but in other cases they may also contribute to a scalar reading implying indifference, as in the following example:19 () quare unius alicuius esse similem satis habebit thus one:gen someone:gen be:inf similar:acc enough have:sg ‘for that reason he will be content with being similar to just anyone’ (Rhet. Her. .) But also the meaning ‘one single’, ‘even one’ is already found, albeit rarely, in Classical Latin: () ut . . . numquam tam frequens senatus fuerit cum so.that . . . never so crowded:nom senate:nom be:sg as unus aliquis sententiam tuam secutus sit? one:nom someone:nom opinion:acc your:acc follow:pt be:sg ‘How is it that the senate has never yet been so full as to enable you to find one single person to agree with your motion?’ (Cic. Phil. .) A question which I leave open for the moment is which role is played by focus in examples like the latter. I will come back to these issues in chapter  and , where we will see the relevance of focus for the grammaticalization of Romance n-words. In conclusion of this section, let me observe in passing a fact that may corroborate my hypothesis on scalarity. A further Latin indefinite formed with ali-, aliquantus ‘some (quantity)’, used at all stages of Latin to express a sufficient or considerable degree, survives in Italian alquanto (Old Italian also aliquanto) ‘rather, quite’. This could possibly be attributed to the nature of the wh-element quantus ‘how great, how much’, which does not lend itself to be reanalyzed as a scalar endpoint, differently from unus. This indirectly points to the fact that scalarity is a determining factor for the further development of the Romance continuations of aliquis into negative-polarity contexts.

19 Hofmann-Szantyr (: ), followed by Bertocchi et al. (: ), state that in Classical Latin the meaning of unus aliquis is one of approximation, which can become indifference (cf. also () above).

OUP CORRECTED PROOF – FINAL, //, SPi

Conclusions and outlook



. Conclusions and outlook I have discussed the synchronic and diachronic problems that arise once Latin aliquis is analyzed as a specific indefinite, and I have proposed to consider it an epistemic indefinite instead. I have concluded that only quidam is a genuine specific indefinite in Classical Latin, and discussed possible reasons for its disappearance in Romance. Aliquis manifests context-sensitivity effects similar to those observed in the literature for Spanish algún, Italian qualche, German irgendein among others. I proposed that epistemic interpretations may constitute a bridging context, an intermediate step in the extension of aliquis to further environments, comprising negation. The key change is the development of domain widening, starting with epistemic use (minimal domain widening), which in turn, in the framework I adopted, is connected with the establishment of a licensing dependence with a structurally superior operator. Extension into negative contexts, which were incompatible with a narrow-scope reading of aliquis in Classical Latin, can be observed starting from the fourth century ce. As shown by Haspelmath’s (: –) semantic maps for Latin and some Romance languages, the Latin subsystem of specific indefinites has not survived into the daughter languages. In fact, we witness a split in Western Romance: French and Italian develop a new element (quelque, qualche), which expands into the expression of the ‘specific unknown’ function (i.e. what I treated as the epistemic indefinite function) from an original free-choice function (cf. Foulet  and §..); Portuguese, Spanish, and Catalan continue Latin aliquis respectively with alguém, algún, algun, whose origin is to be traced to the Late Latin combination aliquis unus. Two aspects are common to all these languages. First, there is no dedicated form for an unambiguous specific use: forms that can function as specific indefinites also cover other nonspecific contexts, i.e. they are found in irrealis–nonspecific (future or modalized contexts), interrogatives, and conditionals, and in some cases may also function as negative-polarity items. Secondly, none of these languages retains the unambiguous lexical encoding of Haspelmath’s ‘specific known’ distinction.

OUP CORRECTED PROOF – FINAL, //, SPi

3 Aliquis from Latin to Romance . Introduction .. Topics and aims In chapter  I discussed the distribution of Latin aliquis ‘some (or other)’ and proposed to analyze it as an epistemic indefinite. In so doing, I drew a parallel with the behavior of algún in Spanish, following the account offered by Alonso-Ovalle and MenéndezBenito (, ). We saw that aliquis and algún pattern similarly in modalized contexts; however, in contrast to its descendant algún, Classical Latin aliquis is not used as a narrow-scope indefinite under negation. The extension into this context happens first in Late Latin (§.). In this chapter I follow more closely the diachronic development of the continuations of aliquis in Romance. I start with a synchronic comparative overview of the functions that these descendants have in the Standard Romance languages: the main divide turns out to be the availability of epistemic uses, which are possible in Spanish, Catalan, and Portuguese but excluded in French and Italian. I then move to the diachronic dimension and show that the epistemic variants were available at early stages of French and Italian and subsequently went missing in the history of the two languages. I argue, thus, that the epistemic uses are a common inheritance from Latin, which is conservatively maintained in Ibero-Romance but lost in French and Italian.1 Concerning the uses in the scope of negation, they are common to all continuations. In some varieties the dependence relation with respect to sentential negation becomes privileged, and the continuations of aliquis further evolve from NPIs into n-words (and negative indefinites in Colloquial French), i.e., become an integral part of the negation system.2 DP-internal word order, and more precisely the phenomenon of DP-internal inversion between the indefinite and the rest of the DP, assumes particular importance in my discussion. I observe that the languages preserving the epistemic use are innovative in conventionalizing DP-internal inversion when the indefinite expression is in the scope of a negative operator. I motivate syntactic inversion as a way to

1 Here and in what follows, I include Catalan within Ibero-Romance, because it shares with Spanish and Portuguese the retention of the epistemic uses, which I interpret as an inherited linguistic feature. 2 For the difference between NPIs, n-words, and negative indefinites, cf. §... (and further chapter ).

Indefinites between Latin and Romance. First edition. Chiara Gianollo. © Chiara Gianollo . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi

Introduction



convey an emphatic meaning, resulting in the strengthening of negation, and I intepret it as due to the movement of the determiner to a DP-internal Focus position. I further argue that, once inversion becomes fully conventionalized, i.e. obligatory, as in European Portuguese, the emphatic value bleaches, and the indefinite becomes a plain n-word (characterized by a formal uninterpretable negative feature [uNeg]). Finally, I discuss the case of Catalan, which provides evidence that the development can take a different turn: in this variety polarity uses are those that are lost, while epistemic ones are preserved. I conclude, then, that the Romance continuations of aliquis were highly versatile at an early stage, and specialize in different domains in the course of history. So much for the empirical coverage of this chapter. But what do we learn exactly from this case study? I would like to show that the development seen with aliquis and its continuations not only is relevant for the comparative history of the Romance languages; it also allows us to draw some more general conclusions on the dynamics of semantic change in general. In §. we will see that the development we are investigating is an instantiation of a cyclical pattern of change at the syntax-semantics interface, which is known from the recent literature as the ‘Quantifier Cycle’. This cycle often intertwines with diachronic phenomena affecting negation, which are comprehended under a further, much better studied cyclical pattern, Jespersen’s Cycle. I propose an analysis according to which the Quantifier Cycle shares with Jespersen’s Cycle not only part of its developmental path, but also its fundamental pragmatic motivation, namely emphatic strengthening. I understand emphatic strenthening as a form of focus, and I discuss how this additional meaning component interacts with the procedures that establish the domain of quantification of the indefinite. I also argue that focus is, at the outset, only one possible interacting factor, but it develops into a necessary condition for the use of the continuations of aliquis under negation. However, as frequently happens, the emphatic value of the construction tends to bleach with time. Also in this case we can draw a parallel with the pragmatics of Jespersen’s Cycle, where bleaching of the negative reinforcer has been attributed to ‘inflationary effects’ which, in diachronic perspective, seem to be a natural counterpart of emphasizing procedures. .. Roadmap The discussion will proceed as follows. I start by presenting the Quantifier Cycle (§.). Then, after a short summary of the analysis proposed for Classical Latin aliquis, I offer a synchronic overview of its continuations in the Standard Romance varieties (§.). Here I also discuss the value of the plural forms and how current hypotheses account for the unity of the lexical item despite its contextually variable manifestations. In §. we will then see how the meaning development of the continuations of aliquis unfolded diachronically in Romance. The next step will be to discuss the nature of DP-internal syntactic inversion, and to propose a motivation for its origin in terms of the semantic value of the indefinite on the one hand, and of focus on the other hand (§.–§.). In §. I will summarize my conclusions.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

. The Quantifier Cycle Willis () and Willis et al. (a) label ‘Quantifier Cycle’ the crosslinguistically frequent, systematic semantic change that affects certain indefinite items, leading to a progressive restriction of their contexts of use, up to the point where the negative operator becomes the only possible licensor. French personne and rien (Eckardt , ), Dutch enig (Hoeksema ), German kein (Jäger ), the Welsh r(h)ywseries (Willis ) are examples studied in the literature. Similar developments have been discussed under a different heading by Haspelmath (), Ladusaw (), Martins (), Jäger (). At the beginning of the cycle, the element is a plain indefinite or a generic nominal expression. Intermediate stages of the cycle involve the restriction to weak-polarity environments. The (pre-)final stage often witnesses the employment of the focused indefinite as a reinforcer of negation in emphatic constructions, thus potentially intertwining with Jespersen’s Cycle. Emphasis may then bleach at the terminal stage, especially when the indefinite becomes a ‘plain’ element of Negative Concord structures or a negative indefinite, depending on the overall type of negation system. The cycle is schematically presented in (a). Not all elements undergoing this cyclic development go through all steps: in some cases they stop at intermediate stages, as shown by the Dutch and Welsh examples in (b). Even in these cases, however, as with ‘full’ cycles, the development follows a systematic cline, with no skipping of intermediate steps. () The Quantifier Cycle a. (i) plain indefinite / generic nominal expression > (ii) weak polarity item > (iii) strong polarity item (reinforcer of negation) > (iv) element of Negative Concord or Negative Indefinite b. French personne (i) ‘person’ > (iv) ‘nobody’ French rien (i) ‘thing’ > (iv) ‘nothing’ Spanish nada (i) ‘born thing’ > (iv) ‘nothing’ Dutch enig (i) ‘some’ > (ii) ‘any’ German kein (ii) Middle High German dehein ‘any’ > (iv) ‘nothing’ Welsh r(h)yw-series (i) ‘some’ > (ii) ‘any’ According to Willis (), Willis et al. (a: –; –), the cycle involves a specialization of the indefinite element first into weak-polarity contexts (comprising questions and antecedents of conditionals) and direct negation, and subsequently only into strong-polarity ones (besides direct negation, also indirect negation and standard of comparison). The least understood part of the cycle involves the change from stage (i) to stage (ii). In particular, elements entering the cycle seem to have quite heterogeneous origins: sometimes they already belong to the inventory of functional items and have a determiner-like status; sometimes they are open-class lexical elements, whose meaning makes them prone to being used as generalizers. In any case, at stage (i) they seem to be quite unconstrained in their distribution and scopal possibilities: crucially,

OUP CORRECTED PROOF – FINAL, //, SPi

The Quantifier Cycle



they are already compatible with all the relevant environments involved in the cycle, and the change affecting them leads to a restriction of possible contexts. This step of the change is quite difficult to pinpoint in historical documentation because it is not categorical and, at the outset, manifests itself in terms of frequency: although the element is still compatible with ‘positive’ contexts, it tends to appear more often in downward-entailing ones (Hoeksema () refers to these intermediate elements as ‘semi-NPIs’). In some cases, as e.g. with Middle High German dehein, the indefinite is a polarity item from the start (Jäger ). Further comparative research is needed here to assess whether there exists some particular meaning component which qualifies a ‘positive’ indefinite as a potential candidate for the cycle. In §.. I argue that this is indeed the case with the Romance continuations of aliquis. The cyclic nature of the change resides in two aspects. On the one hand, it is systematic in nature, both in terms of its crosslinguistic frequency of attestation and in terms of the directionality observed in the language-internal development (cf. §...). On the other hand, it typically co-occurs with the creation of a new ‘positive’ indefinite, replacing the element undergoing the cycle by undertaking its functions. In speaking of co-occurrence, I remain deliberately agnostic as to whether the creation of a new ‘positive’ indefinite is a case of pull-chain or push-chain development. In principle, either scenario is a possibility: the innovative element may be pulled into use by the meaning shift of the older item, or it may cause the meaning shift itself, by taking over (parts of) the functional domain of the older item and thus pushing it aside. I believe that this has to be assessed on a case-by-case basis, since there seems to be no a priori factor leading in either direction. A further dimension of potential differentiation has to do with the pervasiveness of the cycle: while, in some cases, we see that only one element of a series undergoes the cycle (e.g. the determiner, or the animate pronoun), also other scenarios are attested, where the change involves an entire series. This seems to be the case in the history of Welsh (Willis ) and of Romance, as we will discuss. Clearly, the more morphologically coherent a series is, the more likely it is that it will undergo the shift as a whole. But the range of action of the cycle is presumably determined also by the nature of the triggering factor: as we will see in detail in chapters  and , the cycle involving Romance narrow-scope indefinites under negation is the long-term consequence of the restructuring affecting the entire negation system, i.e., the shift from the Latin Double Negation to the Romance Negative Concord system. This last observation takes us to a topic which will assume particular relevance in our discussion: the interplay between the Quantifier Cycle and the cyclic developments in the syntax of clausal negation, which go under the name of Jespersen’s Cycle. In observing the cyclical nature of the development just seen for indefinites, Ladusaw () draws a clear parallel with the systematic changes affecting negation and speaks of a ‘Jespersen argument cycle’.3

3 Ladusaw () uses the label ‘argument cycle’; however, the same pathway also applies to indefinite adverbials.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

Plain indefinites and generalizers, which represent typical stage (i) elements for the Quantifier Cycle, can also be recruited as reinforcers of negation, together with minimizers; as a consequence, they constitute typical sources for the renewed negative markers originating from Jespersen’s Cycle (cf. §..).4 A very important difference between the two cycles consists in the fact that, in the renewal of the negative marker, the reinforcing element (e.g. French pas) loses its original lexical category and argumental status and becomes an adverbial (and later a functional head). In the Quantifier Cycle, instead, the affected element remains a (pronominal) determiner (e.g. Dutch enig) or becomes one (e.g. French personne), that is, a component of DPs, and not of NegPs. Kiparsky and Condoravdi () remark that, when used as strengtheners of negation, indefinite pronouns have an emphatic flavor. This observation has farreaching consequences, if we connect it with the long-standing intuition that emphasis represents a fundamental functional motivation for polarity sensitivity in general. I take it, thus, as a starting point to explore the role that emphasis, which I understand as a form of focus, has in the diachronic process leading to the restriction of licensing contexts with the Romance continuations of aliquis.

. The Romance continuations of aliquis In this section I shortly recapitulate the analysis for Latin aliquis reached in chapter  and I discuss its role within the Quantifier Cycle (§..). I then move to its continuations in the Standard Romance varieties, offering first an overview of the developments (§..) and subsequently discussing in detail each Standard Romance language involved in the change (§..–..). I also introduce the debate on the analysis of plural forms (§..). .. Latin aliquis and the Quantifier Cycle At this point, readers who have gone through chapter  will have noticed that the Quantifier Cycle presents some similarities with the development undergone by aliquis from Classical to Late Latin, especially with respect to the cline leading from weak- to strong-polarity contexts. However, a fundamental difference also emerges quite clearly: the history of aliquis during the Latin stages is a history of extension of use into previously excluded semantic environments and scopal configurations. The Quantifier Cycle represents, instead, a history of restriction in the range of possible contexts. Classical Latin aliquis does not qualify as a typical stage (i) element. It never behaves as a plain indefinite, as shown particularly well by negative contexts, where it systematically resists narrow scope below the negative operator. That is, although 4 Minimizers have never been discussed in the context of the Quantifier Cycle: although we have examples of indefinites formed from minimizers (e.g. Latin ullus ‘any’ > *unulus diminutive of unus ‘one’), these indefinites do not seem to go from ‘positive’ to ‘negative’ as the typical representatives of the Quantifier Cycle. Rather, they are negative-polarity items from the start. For the different case of adverbial strengtheners of negation derived from minimizers see Batllori ().

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



Classical Latin aliquis is found in negative contexts, it obviously does not behave as expected from a typical starting point for the Quantifier Cycle, which is a ‘neutral’ element with respect to scope preferences. In chapter  we saw that this changes in time, and that the Late Latin documentation shows signs of uses under negation. Given the advanced situation we find in Old Romance, these uses arguably had broader application in some spoken varieties. In what follows I present data on further developments in Romance. I argue that the latter are genuine instantiations of the Quantifier Cycle. At their earliest stages, the Romance varieties continuing aliquis feature an item that fully qualifies as a stage (i) element in the Quantifier Cycle, since it has a broad distribution in all relevant contexts. The considerable homogeneity found in Old Romance in this respect, as well as the Late Latin situation, lead us to assume that this kind of distribution is a common inheritance. In the history of the individual Romance varieties, we then see the Quantifier Cycle deploying to different extents and with partially different outcomes on the overall system of indefinites. Before delving into the Romance evidence, let me recapitulate the main points concerning the analysis of Classical Latin aliquis, which will be useful for further discussion. In chapter  I concluded that Classical Latin aliquis can be characterized as an epistemic indefinite and I adopted for it the analysis given for Spanish algún by Alonso-Ovalle and Menéndez-Benito (, ). The main feature of epistemic indefinites is to signal that the speaker or another agent is not committed to a particular referent, i.e., that no specific witness for the existential claim can be singled out. The ignorance effect is explained pragmatically as a conversational implicature, triggered by a lexically encoded presupposition carried by the indefinite, the ‘antisingleton’ constraint on the (covert) domain restriction. The anti-singleton presupposition imposes the condition that the domain of quantification cannot be reduced to a singleton set: more than one element of the quantificational domain must be an epistemic possibility for the agent. At the same time, with algún and, we argued, with aliquis, the domain need not be maximally widenened, as is the case instead with freechoice items. The ignorance implicature arises through comparison of the proposition containing the epistemic indefinite with its informationally stronger alternatives (singleton indefinites): it is a scalar implicature triggered by the Maxim of Quantity. It disappears in downward-entailing contexts since, as the direction of entailments is reversed, no competition arises with indefinites expressing smaller domains: it would be useless from the point of view of informational strength. A fundamental characteristic of epistemic indefinites is, thus, that they trigger inferences that are subject to a pragmatic felicity condition: they need the right environment that makes them useful for communication, where they achieve an informationally significant contribution. This yields constraints on their distribution and their scopal possibilities: epistemic indefinites are ‘dependent’ indefinites. They depend on operators that create the pragmatically right environment for the implicatures they license, or that make them innocuous (as in downward-entailing contexts).

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

I would like to argue that this is indeed what qualifies them, and Latin aliquis as an instantiation, as potential candidates for the Quantifier Cycle: epistemic indefinites are a special type of ‘positive’ indefinites. They are referentially not self-sufficient and are bound to appear in the scope of a licensing operator. The nature of the licensing relation may change over time, owing to the interplay of frequency effects and acquisitional strategies, but also to the interaction with other factors. One prominent factor, I will propose, is focus. One big question I have to address is why these epistemic indefinites, more than other elements, would be targeted by focus to such an extent as to trigger a grammaticalization process, like the one we will see in this chapter. In accounting for the path seen with the Romance continuations of aliquis, I capitalize on the dimension of variation in the system of indefinites that is represented by the variable constraints that indefinites impose on their quantificational domains (cf. §..). I will show that these conditions shift over time in a principled way, through minimal changes in the determiners’ lexical makeup which are triggered by functional pressures, like the need for expressivity or informativeness. A recurrent motivation, in this respect, is represented by emphatic focus: focus interacts with the lexically activated domain alternatives, leading, as I will argue, to a further widening of the domain of quantification. The association with focus, once conventionalized, may lead to the grammaticalization of originally optional inferences. This, in turn, results in bleaching of emphasis. Interestingly, the Ibero-Romance continuations of aliquis suggest the hypothesis that grammaticalized association with focus may have morphosyntactic consequences, in the form of DP-internal inversion. In the next section I will present the relevant data. .. The continuations of aliquis: a synchronic overview In this section I outline the distribution of the continuations of aliquis in a number of Standard Romance languages: Spanish, Portuguese, Catalan, French, and Italian. The observed variability will motivate the subsequent diachronic investigation. All the languages listed above continue aliquis in a modified form that is never attested in our Latin documentation (cf. §.) but has to be reconstructed for Proto-Romance: *alicunum, arising from the univerbation of aliquis and unus, the cardinal numeral for ‘one’. This form is used as a determiner and as a pronoun: Italian alcuno, French aucun, Catalan algun, Spanish algún, Portuguese algum. Some languages have also continued aliquis as a distinct form reserved for pronominal uses: in these cases, the form is not derived with the addition of unus, but continues Latin accusative forms; Ibero-Romance continues the masculine and neuter accusative of aliquis with e.g. Sp. alguien, Pt. alguém < aliquem, Sp. and Pt. algo < aliquod. In what follows, I will mostly concentrate on the determiner form. The table in () provides a first overview of the observed distribution. It distinguishes between plural and singular uses, and indicates the various functions for the latter: Ep(istemic), N(egative) P(olarity) I(tem), n-word, N(egative) I(ndefinite). For the uses under negation, the status of inversion is omitted for the moment: it will be defined more precisely in the following discussion (cf. especially §..). Also the meaning in the plural will be further discussed (§..): for the moment, I indicate

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



with ‘plain’ the availability of a plural form and the absence of the ignorance effect. More will have to be said concerning the possibility of accounting for the plural uses with a unique lexical entry for the various continuations. () Overview of functions: Ep: epistemic; NPI: negative polarity item; n-word: element of negative concord; NI: negative indefinite

Pl Sg

It

Fr

Cat

Sp

Pt

alcuno plain NPI

aucun (n-word with pluralia tantum) n-word / NI

algun plain Ep

algún plain Ep & NPI

algum plain Ep & n-word

We can see from the table that only the plural can be interpreted specifically (a). With the apparent exception of Catalan (but see §.. for its diachrony), the singular shows, in one form or another (NPI, n-word, NI), uses where it is dependent on negation for its licensing, in all (Italian, French) or in some (Spanish, Portuguese) contexts of occurrence. Under conditions that will have to be specified in the next sections, in the scope of negation DP-internal inversion between the determiner and the rest of the nominal phrase may take place, as in (b). According to the analysis of Negative Concord that I follow in this work (cf. §... and further chapter ), Standard French aucun (nowadays found only in formal registers) is an n-word, i.e., its predominant licensing domain is the scope of clausal negation (c) and it can occur as a negative elliptical answer (d). In Colloquial French (the variety which is gradually losing Negative Concord) aucun can occur without the negative marker ne in sentences with a negative meaning (e), thus behaving like a negative indefinite of Double Negation languages.5 Only Ibero-Romance shows the epistemic indefinite use (f vs. g). In this function, French and Italian have developed a new item, quelque / qualche (g) from an original free-choice function (Foulet , Zamparelli , Jayez and Tovena ). () a. Italian: Franco vuole comprare alcuni libri nel negozio qui vicino ‘ Franco wants to buy some books in the nearby shop ’ (wide scope possible) b. Spanish: no toma precaución alguna ‘ He doesn’t take any precaution ’ c. French: Aucun de mes amis n’est venu ‘ None of my friends came ’ d. French: A: Avez-vous une objection? B: Aucune ‘ A: Do you have an objection? B: None ’ e. Colloquial French: Il y en a aucun qui mange ton gâteau ‘ There’s none of them eating your cake ’

5 I am grateful to Fabienne Martin for the judgments.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance f. Spanish: María se casó con algún estudiante del departamento,  en concreto con Pedro ‘ Mary married a student of the department,  namely Pedro ’ g. Italian: *Maria si è sposata con alcuno studente del dipartimento h. Italian: Qualche idiota ha dimenticato di spegnere la luce ‘ Some idiot has forgotten to switch off the light ’

This first overview already gives us a good impression of the remarkable amount of variation witnessed in Romance with the continuations of aliquis. From the observations above it also emerges quite clearly that the distributional properties of the various continuations are tightly linked to the type of negation system the particular language has. In this respect, we will need a more fine-grained analysis of syntactic contexts in order to precisely determine differences among languages. In particular, the conditions governing inversion between the head noun and the determiner turn out to be particularly relevant. We have seen an example of such inversion for Spanish in (b). The pre- or post-nominal position of the indefinite correlates with its epistemic or ‘negative’ interpretation. Under some syntactic conditions, which we will examine in the following sections, inversion can be the only marking for the ‘negative’ function of the indefinite, as in the European Portuguese and Spanish cases below: () Portuguese (Martins a) a. Algum animal vive aqui ‘Some animal lives here’ b. Animal algum vive aqui ‘No animal lives here’ () Spanish (my questionnaire, see end of this section) a. No ha pagado la cuota ningún estudiante ‘No student paid the tuition fees’ b. No ha pagado la cuota estudiante alguno ‘No student paid the tuition fees’ c. *No ha pagado la cuota algún estudiante intended reading: ‘No student paid the tuition fees’ d. Algún estudiante no ha pagado la cuota. ‘Some student did not pay the tuition fees’ Uses of algún in the scope of sentential negation are attested throughout the documented history of Spanish and Portuguese, and are possible—at least at certain historical stages—in all Romance varieties continuing aliquis. Syntactic inversion is similarly attested in all varieties, although to differing extents of conventionalization. There is a remarkable amount of variation in this domain, caused by a complex interplay of factors. Inter-speaker variation is also worthy of attention, since it might give insights on diachronic processes that are still in the making. Specifically, French shows signs of a change from a strict N(egative) C(oncord) system to a D(ouble) N(egation) system in those varieties that drop ne both with pas

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



and with personne, rien, etc., leading to a system where an indefinite can be the only marker of sentential negation. Catalan displays a kind of sociolinguistic variation in the negation system that has led some scholars to assume a change in progress, from a more conservative strict NC variety to an innovative non-strict NC variety. The latter is used in colloquial speech, especially by younger generations, and has been attributed to influence from Spanish (cf. Zeijlstra ). Déprez et al. () provide recent experimental data on the contemporary variation in Catalan. I analyze the contrast in (–) in §.. There I formulate a syntactic proposal concerning the inversion operation: I analyze it syntactically as a DP-internal focalization movement, drawing a parallel with analogous DP-external and DP-internal phenomena in Romance. Before I do that, I present more in detail the Modern Romance data in the remaining parts of this section, and in §. I discuss the diachronic developments in Old French and Old Italian. For the purpose of better evaluating the variation in the Romance continuations of Latin aliquis, I developed a questionnaire (cf. §..). The questionnaire focused in particular on two aspects: (i) the contextual conditions licensing the appearance of the indefinite form and constraining its interpretation; (ii) the syntactic structure of the nominal phrase containing it, in particular the presence of inversion with respect to the head noun. The questionnaire was distributed among professional linguists who are native speakers of one or more Standard Romance varieties.6 Possible interference due to bilingualism was controlled for. In the case of languages, such as French and Catalan, showing sociolinguistically determined variation in the behavior of negation and negation-related indefinites (a sign of ongoing change), the informants were asked to pay attention and be consistent with respect to the type of variant employed. The general aim was to check under which conditions the functions for the indefinite item that we saw in () appear: (i) (ii) (iii) (iv) (v)

plain indefinite / partitive meaning (for the plural) epistemic indefinite (nonspecific, alternative-evoking indefinite) polarity-sensitive indefinite (NPI) n-word (element of Negative Concord) negative indefinite (negative element in a Double Negation system)

I systematically distinguished between canonical subject and object position, that is, more generally, between the positions preceding (pre-Infl) and following (postInfl) the inflected verb (Infl). This distinction is very relevant for Romance varieties displaying non-strict Negative Concord: as we will see in chapter , constituents in the pre-Infl position precede the functional projection for negation NegP-; in non-strict

6 I am very grateful to my informants: Livia Assunção Cecilio, Valentina Bianchi, Sonia Cyrino, Victoria Escandell Vidal, Maria Teresa Espinal, Marco García García, Dara Jokilehto, Manuel Leonetti, Ana Maria Martins, Beatriz de Medeiros Silva, Esperanza Torrego, Jacopo Torregrossa, Joaquín Vuoto. I am also grateful to Klaus von Heusinger and Stefan Hinterwimmer for commenting on a preliminary version of the questionnaire, helping me improve it.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

Negative Concord varieties they cannot co-occur with the overt negative marker under a single-negation reading, i.e. they do not exhibit Concord, unlike elements in the post-Infl field. .. Spanish I start my discussion of the Modern Romance data with Spanish since we already know some facts about the distribution of algún from chapter . Alonso-Ovalle and Menéndez-Benito (: , n. ) say, as we discussed in §.., that algún cannot be in the scope of sentential negation. However, this is not empirically accurate. Although relatively rare and typical of a high stylistic register, sentences like (a) are perfectly acceptable in Modern Spanish, and truthconditionally equivalent to the correspondent structure (b) with the n-word ningún: both indefinites take narrow scope with respect to the negative operator realized by sentential negation: () a. No vi a estudiante alguno ‘I didn’t see any student’ b. No vi a ningún estudiante ‘I didn’t see any student’ Are (a) and (b), then, equivalent in all respects? The answer seems to be negative: speakers perceive (a) as more ‘emphatic’ than the corresponding structure with the n-word. Moreover, for the speakers of European Spanish that I consulted, syntactic inversion between the noun and the determiner is obligatory in a negative context. This is a further difference with respect to ‘positive’, epistemic algún, together with the absence of the ignorance effect (which, as discussed in §.., can be derived by considering it an implicature, which gets neutralized in downward-entailing contexts). For the speakers of European Spanish the following generalizations emerge from the questionnaire: inversion is only possible in the singular and it happens exclusively in connection with negation. That is, every time there is inversion, a negative meaning is conveyed and, conversely, if algún is in the scope of a negative operator, inversion is obligatory. Importantly (and in contrast with Portuguese, as we will see next), nominal phrases with inverted algún can only be found in the post-Infl area of the clause: inversion is impossible in preverbal subject position (). This means that inverted algún always has to occur with an overt manifestation of the negative operator, expressed either by the negative marker no or by a pre-Infl n-word (Spanish being a non-strict Negative Concord language). I thus analyze inverted algún as an NPI in European Spanish: its NPI status is shown by the fact that it always has to be hierarchically (and linearly) preceded by an overt marker of sentential negation. () a. *Persona alguna (no) vive aquí ‘Nobody lives here’ b. No vive aquí persona alguna ‘Nobody lives here’

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



A further confirmation of the NPI-status of inverted algún comes from its behavior in negative short answers, where it is judged ungrammatical by my informants: () Speaker A: ¿Tiene objeciones a eso? ‘Do you have objections to this?’ Speaker B: *objeción alguna ‘no objection’ Speaker B: ninguna objeción ‘no objection’ As expected, a short answer with the non-inverted order is possible and has an affirmative meaning. Interestingly, the same applies if the answer consists of algún in the pronominal use, with no overt nominal restriction: () Speaker A: ¿Tiene objeciones a eso? ‘Do you have objections to this?’ Speaker B: alguna objeción ‘yes, some objection(s)’ Speaker B: alguna ‘yes, some’ Apparent counterexamples to my claim that inverted algún always has to cooccur with a c-commanding negation are examples like the one in (), from Herburger (: ), which is however judged marginal by my European Spanish informants: () En modo alguno se puede tolerar tal actitud! ‘Under no circumstances can one tolerate such an attitude!’ (Herburger : ) Possibly here the focus involved in the displacement of the adverbial PP can play a licensing role for some speakers: similar phenomena are observed in Italian and seem to involve in particular adverbial modifiers. Speakers producing these structures possibly show a more advanced, Portuguese-like grammar. Apparently, thus, Spanish conventionalized inversion, but this conventionalization did not lead yet to the grammaticalization of an n-word: the element keeps having an NPI distribution (excluding the pre-Infl position). In the post-Infl position, inverted algún competes with ningún as narrow-scope indefinite under negation. As mentioned, the use of inverted algún is perceived as more emphatic: in the words of one of my informants, ‘with the canonical order (ningún estudiante) you make a ‘normal’ negation, while with the inverted order (estudiante alguno) you rather make a more expressive negation with a negative polarity flavor focusing that there was not a single student that paid the tuition fees.’ (Marco García García, p.c.). These generalizations are corroborated by a speaker of Argentinian Spanish, who confirms the presence of inversion under negation with algún, and also remarks that the structure sounds more emphatic than the plain one with ningún. Inversion with ningún is judged impossible by my informants:

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

() a. Ningún estudiante pagó la matrícula No student (sg.) paid the tuition fees b. *Estudiante ninguno pagó la matrícula c. *No pagó la matrícula estudiante ninguno However, as a reviewer remarks, there might be intra-speaker variation in this respect, which plausibly reflects a diachronic development, since inversion with ningún was possible at earlier stages. As Martins (a: ) discusses, some speakers accept inversion with ningún as a marked emphatic option: () No tenemos miedo ninguno ‘We don’t have any fear (at all)’

(from Martins a: )

.. Portuguese DP-internal inversion with Portuguese algum has been extensively investigated by Martins (, a,b), who was the first to draw attention to the important syntactic and semantic consequences of this phenomenon. The discussion in this section will rely on her observations, and will be complemented by the data collected through the questionnaire. Martins analyzes inversion as a form of negative inversion targeting a DP-internal NegP projection in the left periphery of the nominal phrase. I will, instead, propose that the relevant DP-internal peripheral position is a Focus one (§.). The Portuguese situation is similar, to a certain extent, to what is seen in Spanish: both epistemic and ‘negative’ uses of the continuation of aliquis are attested, and they differ in terms of word order, since inversion has to be present when the indefinite takes narrow scope with respect to negation (this happens only in the singular form). But there is a crucial difference between Spanish and Portuguese: while structures with inverted algún are banned from the pre-Infl position in Spanish and must always co-occur with the overt manifestation of the negative operator, in Portuguese inverted algum can be found pre-Infl with a negative meaning. See the contrast in (), where the examples come from Martins (a): () E(uropean) P(ortuguese) versus Sp(anish) (from Martins a) a. Algum animal vive aqui ‘Some animal lives here’ b. Animal algum vive aqui ‘No animal lives here’ c. *Persona alguna (no) vive aquí ‘Nobody lives here’ d. No vive aquí persona alguna ‘Nobody lives here’

(EP) (EP) (Sp) (Sp)

We see, therefore, that in Portuguese inversion may be the only overt sign of sentential negation in certain contexts. More concretely, inverted algún behaves like an n-word in European Portuguese and conforms to the non-strict Negative Concord nature of the negation system (cf. Martins a: , Martins b: ): the nominal

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



phrase containing algún co-occurs with the negative marker não only if post-Infl; cooccurrence under a single-negation reading is not possible in the pre-Infl position, where the indefinite itself functions as the overt manifestation of sentential negation. Martins (, a,b) remarks that the first examples in pre-Infl subject position appear in nineteenth-century European Portuguese (EP); she also states that some speakers still have the more ‘Spanish-like’ grammar and do not accept pre-Infl structures with inverted algum. We can thus assume a diachronic cline according to which European Portuguese (or, more precisely, those speakers of European Portuguese accepting inverted algum in pre-Infl position) is more advanced than Spanish in the Quantifier Cycle: the direction of reanalysis goes from an NPI use to an n-word use. Interestingly, my informants do not notice any emphasis in the construction with inverted algum, unlike what we have seen with Spanish. It seems, thus, that the shift from NPI to n-word also involves loss of emphasis: this will be important for my theoretical account. For Brazilian Portuguese (BP), my data indicate that this variety also is more advanced than Spanish and similar, in this respect, to European Portuguese: inversion is possible only in the singular and is unambiguously connected to a negative context. () Brazilian Portuguese a. Maria não se casou com estudante algum em seu grupo de tênis ‘Mary hasn’t married any student in her tennis group’ b. Nunca vi animal algum aqui ‘I have never seen any animal here’ Nominal phrases with inverted algum can occur in pre-Infl position in Brazilian Portuguese, e.g. as subjects, without a further negative marker: they show, as in European Spanish, an n-word behavior. () Estudante algum pagou as taxas ‘No student paid the fees’

(BP)

The n-word behavior is also confirmed by the distribution in negative short answers: differently from the way Spanish works, inverted algum is possible:7 () a. Speaker A: Vais vender a tua casa? (EP, from Martins b: ) ‘Are you going to sell your house?’ Speaker B: Em caso algum. ‘Under no circumstance’ b. Speaker A: Tem objeções a isso? (BP) ‘Do you have objections to this?’ 7 This possibility is limited to algum in its determiner use: algum with no overt nominal restriction (i.e., in its pronominal use) is ungrammatical as short negative answer. This could be attributed to the fact that, as argued by Martins (b: ), algum has the status of a syntactic head, which would exclude it from elliptical structures (for short answers as elliptical structures cf. further §...). Note that in fact the epistemic version also cannot occur in its pronominal use in a (positive) short answer: the structure is judged ungrammatical by my informants, unlike what happens in Spanish, where my informants accept pronominal alguno as positive short answer.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance Speaker B: Objeção alguma ‘No objection’

Interestingly, in both European and Brazilian Portuguese, inversion is possible also with nenhum: the examples in () show that this is the case, for instance, in short answers, parallel to ():8 () a. A: Vais vender a tua casa? ‘Are you going to sell your house?’ B: Em caso nenhum. ‘Under no circumstance’ b. A: Tem objeções a isso? ‘Do you have objections to this?’ B: Objeção nenhuma ‘No objection’

(EP, from Martins b: )

(BP)

The widespread inversion observed with Portuguese nenhum is a peculiar phenomenon in the landscape of contemporary Romance languages: Martins’ (a; b) corpus study reveals that the rise in frequency of this pattern happens in the nineteenth century, and follows the conventionalization of inversion with algum in negative contexts. She attributes this development to the activation of a DPinternal Neg projection, which would be targeted by both algum and nenhum (or, more precisely, by the complex unit formed by the determiner and the head noun). Since some of my informants notice emphasis in the use of inverted nenhum, and also given the correlation with emphatic negation (não . . . nada) noted by Martins (b: –), the role of focus should be further investigated in this respect (I make a very tentative suggestion in §..). .. Catalan Catalan also continues aliquis with algun. However the modern language does not have the narrow-scope use in negative contexts (Vallduví , Montserrat Batllori p.c., Maria Teresa Espinal p.c.), but only the epistemic one. The epistemic use contrasts with the specific reading of the indefinite determiner un in the appropriate contexts and is shown in (). Some Catalan varieties (e.g. Balearic) also have qualque, which Wheeler et al. (: ) describe as synonymous with algun. () a. Algun estudiant va pagar la matrícula, però no sé qui és ‘Some student (sg.) paid the tuition fees, but I don’t know who s/he is.’ b. La Maria busca un estudiant /  algun estudiant. Es diu Joan ‘Mary is looking for some student. His name is John.’ Given the absence of the ‘negative’ variant, no asymmetries between the pre- and postInfl position are observed.

8 Martins (b: –) discusses in detail the conditions governing inversion with nenhum.

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



The example in () shows that it is impossible to use algun under negation. It either has wide scope, or the sentence is understood as a denial of an indifference meaning (‘I did not greet one random student, I greeted the best student in the class’). () No he saludat algun estudiant ‘There is some student whom I didn’t meet’ Interestingly, however, Par (: –) shows that in the language of Bernat Metge (fourteenth century) alcu / alcun was regularly used as an NPI indefinite under negation. Moreover, Par remarks that this use is still completely grammatical in his contemporary language if algun is postponed to the noun. Thus, the disappearance of algun from the context of direct negation in Catalan seems a relatively recent phenomenon, and this language as well witnesses a stage where nominal inversion becomes obligatory. Further diachronic research is needed to safely assess this point.9 Catalan also has the plural form of algun. Interestingly, for my informant it is possible to detect an epistemic effect also in the plural, where the indefinite is incompatible with a disambiguation indicating that the speaker can identify the witnesses to the claim: () Alguns estudiants es deuen haver oblidat de pagar la matrícula.  Els he vist al despatx ‘Some students must have forgotten to pay the tuition fees. I saw them in the office.’ As a reviewer notes, however, this could simply be due to the modal context in which the indefinite is used (similar effects are observed in Portuguese in analogous contexts). .. French Various diachronic studies have shown that the continuations of aliquis had a much broader distribution in Early French and Early Italian than in the contemporary languages, occurring also—as we will see more precisely in §.—in nonveridical environments such as the protasis of conditional and modalized environments (Stark ,  for Old Italian; Prévost and Schnedecker ; Déprez and Martineau ; Ingham a; Ingham and Kallel  for Old and Middle French). In contemporary French aucun is overwhelmingly found in negative contexts. Only formal registers still witness some residual polarity uses (cf. Milner : –, Ingham and Kallel ), cf. () for the use in embedded questions and (), where aucun is licensed by the negative verb ‘refuse’ plausibly across a clause boundary. () je me demande si j’ai jamais lu aucun livre de cet auteur? ‘I wonder whether I ever read any book by this author’

(Milner : )

() Soulignant qu’il refuse d’entrer dans aucune polémique au moment où la justice doit faire son travail, . . . ‘stressing that he refuses to start any discussion at the moment when justice has to do its job’ (Le Figaro May , , p. ) 9 I am very grateful to Montserrat Batllori for pointing me to Par ().

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

The epistemic use has completely disappeared: quelque / quelqu’un is used in that function () and cannot scope below negation (). The pronoun quelqu’un can be used for specific known and for unknown referents alike. While very widespread in the plural, in the singular quelque as a determiner is not very frequent in contemporary French. My informants tend to prefer the indefinite article to quelque in the determiner function, especially in episodic contexts, cf. the marginality of (): () Quelqu’un a sonné, va voir qui c’est ‘Someone has rung, go see who it is’

(Muller : )

() * Je n’ai pas vu quelqu’un ‘I did not see anybody’

(Muller : )

() ?? Quelque étudiant a demandé à vous voir ‘Some student has asked to see you’

(Muller : )

Déprez and Martineau (: ) remark that in Quebec French NPI uses of aucun are still possible in contexts such as questions, conditionals, indirect negation, comparatives: () from Déprez and Martineau (: ): a. T’as-tu vu aucun chien dans les parages? ‘Did you see any dog in the neighborhood?’ b. Si tu vois aucun étudiant, appelle-nous ‘If you see any student, call us’ c. Elle refuse de dire aucun mot à la police ‘She refuses to say anything to the police’ () Yannick Noah, lui, le sait mieux qu’aucun autre joueur de tennis français ‘Yannick Noah, he knows this better than any other French tennis player’ (Le Monde March , ) The indefinite determiner aucun is nowadays always used in the singular, although it has a plural form which appears with the rare ‘pluralia tantum’ (e.g. vacances ‘holidays’ or lunettes ‘eyeglasses’), behaving like the singular with respect to its scope properties, i.e., always scoping below negation (). Otherwise, the plural survives only in the archaic plain or specific pronominal use d’aucuns ‘some, certain’. () Aucunes lunettes ne lui conviennent ‘No glasses suit him well’ As mentioned, in negative environments aucun behaves as an n-word in the conservative Negative Concord variety, as a negative indefinite in the developing Double Negation variety. Since Standard French has strict Negative Concord, the nominal phrase with aucun always co-occurs with the negative marker ne or with another n-word, independently of its position in the clause: () a. Je n’ai vu aucun chat ‘I haven’t seen any cat’

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



b. Aucun étudiant n’est venu ‘no student came’ c. Dans  ans, aucun médecin, quel qu’il soit, ne gagnera bien sa vie dans ce village ‘In  years no doctor, whoever he may be, will earn a good living in this village’ (Muller : ) The n-word status of aucun is further confirmed by its use in negative short answers, as we saw in (d). Inversion is nowadays very rare and stylistically constrained: it is found especially in combination with the preposition sans, with abstract nouns, and in formulaic expressions, e.g. sans émotion aucune ‘without any emotion’, sans objection aucune ‘without any objection’, typical of solemn style, as is sometimes found in sport reports () or poetry (). () Le PSG est donc sorti du tournoi sans regret aucun ‘The PSG exited the competition round with no regrets at all’ () Il n’est resté de trace aucune / Aucun souvenir n’est resté ‘No trace, no memory remained’ (Arthur Rimbaud (authorship disputed), , Poison perdu) Dhoukar () reports diachronic data on inversion when aucun combines with chose ‘thing’: in a total of , occurrences in her corpus, aucun is inverted in only  cases, and the first case is found at the end of the fourteenth century. Interestingly, in  of the  cases inverted aucun is found in negative contexts. In the other two cases, it is found in the protasis of a conditional (an NPI-environment). () s’elle a fait ou dit chose aucune qui lui soit a desplaisance ‘if she has done or said anything that may displease him’ (De Pizan, , Le Livre des trois Vertus, , cited in Dhoukar : ) It seems, thus, that inversion never gained a grammaticalized status in French. In Modern French, some emphasis may be associated with aucun even if no inversion is present, in cases where it can be contrasted with un ‘one’ (in predicative units like ‘order a pizza’, where ‘one’ has no cardinal value): () from Muller (: ) a. Attendez, moi, je n’ai pas commandé une pizza ‘Wait, I haven’t ordered a pizza’ b. Mais je n’ai commandé aucune pizza! ‘But I haven’t ordered any pizza at all!’ But the opposite is true in cases where un is used as a cardinal numeral: its negation pas un (equivalent to pas un seul ‘not a single’) is in this case more emphatic than the use of (non-inverted) aucun, by virtue of the scalar semantics associated with the negative marker, cf. Muller ():

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

() a. Pas un homme n’échappe à la mort ‘No man can escape death’ (Muller : ) b. ne dites pas un mot! ‘Don’t say a word!’ c. il n’apporterait pas un seul fait décisif, il n’oserait affirmer aucun détail précis sur l’assassin ‘He could not adduce a single deciding fact, he could not dare provide any precise detail about the murderer’ (E. Zola, , La bête humaine, p.  Frantext, from Muller : ) .. Italian In Italian, as in French, epistemic uses of alcuno have been lost. The indefinite is used in negative contexts, mostly in formal or literary registers. It behaves semantically and syntactically like a (strong) NPI: it obeys the typical conditions for NPIs, which have to be in the scope of the licensing operator and are sensitive to the position of overt realization of such an operator, having to follow it in the linear order. The use of alcuno in pre-Infl position is excluded (independently of whether the negative marker is realized or not). () a. b. c. d. e. f.

Non è venuto alcuno studente ?non è venuto studente alcuno *alcuno studente non è venuto *alcuno studente è venuto *studente alcuno non è venuto *studente alcuno è venuto ‘no student came’

Marginal exceptions are represented by cases where alcuno is contained in a pre-Infl PP and contributes sentential negation, without a NM being present: Bernini and Ramat (: ) report the example in (), which can be compared to Spanish (): () In alcun modo si intende agire contro il trattato ‘In no way it is meant to act against the treaty’ (TV news April , ) In my competence this example is ungrammatical, but admittedly less severely so than the examples with pre-Infl subjects in (). Apparently, adverbial modifiers, possibly because of their focus-related displacement, have reached a more advanced grammaticalization stage in the competence of some speakers. The privileged context for alcuno is the scope of the negative operator in direct negation contexts. Polarity uses in contexts other than clause-mate sentential negation are found, although rarely, in the results of the query that I ran on the CORIS corpus (cf. §..). In both (a) (where the broader context reveals that the writer is satirically using an archaizing register) and (b), we have cases of long-distance licensing by a superordinate negative operator; () a. ma non ci piace che alcuno se ne accorga ‘but we don’t like that anyone notices it’

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



b. non si può in alcun modo conciliare con l’impegno per consentire ad alcuno la violazione dei trattati ‘it cannot be reconciled in any way with the effort to agree to anyone’s violating the treaties’ Inversion is possible, but not necessary and felt by my informants as archaizing, though more natural with abstract nouns and as a complement of the preposition senza ‘without’. It is perceived as emphatic. () a. senza alcuno sforzo b. senza sforzo alcuno ‘with no effort’ Some speakers tend to reserve it for negative clefts (suggesting a possible interaction with clause-level information structure): () Non c’è studente alcuno che possa superare questo test ‘There doesn’t exist any student who can pass this test’ For some speakers, inversion may have a presuppositional flavor, e.g., it may be appropriate when reacting to an expectation: () Mi aspettavo di trovare i miei studenti, ma non ho incontrato studente alcuno ‘I was expecting to find my students there, but I haven’t met any student at all’ In order to complement the data on this aspect collected through the questionnaire, I performed a search over the CORIS corpus of Modern Italian. As the table in () shows, inversion with alcuno is rather frequent in certain stylistic varieties. A query for the word ‘alcuno’ in the subcorpus ‘Stampa’ (press: newspapers and periodicals) retrieves  examples; of these, only  are relevant for an evaluation of inversion, since the remaining  have alcuno in a pronominal use. Table () presents the incidence of inverted structures in the sample: () Distribution of alcuno (word form) in CORIS, Press subcorpus alcuno

Tot

Licensor

D-NP

NP-D

non





senza













Inversion is quite frequent, and even predominates when the licensor is senza ‘without’, as preposition or in connection with a finite (senza che + subjunctive) or nonfinite (senza + infinitive) subordinate clause. Some of the senza-PPs occur multiple times and have the status of idioms (e.g. senso alcuno ‘any sense’; dubbio alcuno ‘any doubt’), which may be a factor explaining the predominance of inversion with senza. However, the fact that inversion is a genuinely productive option is shown by the fact that it is found also when the NP restrictor does not form an idiom with the determiner, and

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

is not typical of a particularly high register (a), and also in minimal pairs such as (b) and (c). () from the CORIS corpus a. . . . il Parma non ha incassato gol alcuno. ‘Parma did not concede any goal’ b. senza mostrare sforzo alcuno ‘without showing any effort’ c. senza alcuno sforzo ‘without any effort’ A comparison with the data from spoken Italian available through the LIP corpus of the BADIP (cf. §..) shows that the adjectival use is more frequent ( instances) than the pronominal one ( instances). Only five of the occurrences in the corpus are singular (four in a negative context, one as an NPI), confirming that singular alcuno nowadays belongs to a non-colloquial register.10 In no case is inversion observed. In Modern Italian, inversion with nessuno ‘no’ is very marginal, if accepted at all (in a very literary register). The data for nessuno in the CORIS corpus are very different from those observed with alcuno. For purely practical reasons I searched for the feminine form nessuna in a subset of the CORIS Press subcorpus (periodicals).11 The table in () presents the results: of the  occurrences,  yield a possible context of inversion (i.e., have an NP complement). In no case is inversion observed with the n-word. () Distribution of nessuna (word form) in CORIS, Press subcorpus nessuna

D-NP

NP-D









.. Asymmetry and divergence between singular and plural An important point we have not discussed yet concerns the split between the singular and the plural with the continuations of aliquis. This type of split seems to be a recurrent diachronic phenomenon, cf. Hoeksema () for the history of Dutch enig and, in general, the development of this item in Germanic. The fact that the form used in the scope of negation is always singular is not particularly surprising. It is well known that negation shows interactions with the category of number on nominal elements, motivated by the contribution of number

10 An additional case of alcuno used in a negative environment comes from a plural form which is clearly a case of hypercorrection: non c’erano alcuni problemi proprio di sorta ‘there were really no problems at all’ (R.A...C). 11 This limitation is due to the fact that in the case of nessuno we are dealing with a high-frequency item, especially in the morphologically masculine form, which is used as the default pronominal one. The feminine form is much more often found with an overt NP restrictor, which is the configuration relevant for us.

OUP CORRECTED PROOF – FINAL, //, SPi

The Romance continuations of aliquis



to the referential status of the nominal. Jacobs (: ) cites the case of Hindi and Urdu, where plural number is a cue for interpreting an indefinite outside the scope of negation. According to Davison (), the reason behind this is pragmatic, and has to do with the Maxim of Relevance: the plural would be superfluous if the indefinite were to be interpreted narrowly below negation. Languages, however, show fine-grained morphosyntactic distinctions in this domain, which are not yet well understood. For example, plural forms of the n-words are out in French and Italian (It. *Non ho visto nessuni bambini ‘I haven’t seen any children’), but are acceptable, under certain pragmatic conditions, in Spanish and Portuguese (Sp. Ya no somos ningunos ninõs ‘We are certainly no children’, from the RAE dictionary).12 More surprising is the fact that, when the plural of alcuno, algún, etc. is used in positive contexts, the epistemic ignorance contribution is absent. Alonso-Ovalle and Menéndez-Benito () test this by showing that the ‘namely’ contribution, which— as we saw—is infelicitous with singular epistemic algún, is felicitous with the plural (); the result is replicated in Italian (). () María vive con algunos estudiantes, en concreto con Pedro y con Juan ‘María lives with some students, namely Pedro and Juan’ (Alonso-Ovalle and Menéndez-Benito : ) () Maria ha parlato con alcuni studenti, più precisamente con Pietro e con Giovanni ‘Maria spoke with some students, more precisely with Pietro and with Giovanni’ This raises the question whether it is possible to assume the same lexical entry for the epistemic singular and the plural uses.13 Also in this case Alonso-Ovalle and Menéndez-Benito () propose a pragmatic explanation, based on the interaction of the lexically encoded domain constraint on algún with the meaning of the plural: the fundamental intuition is that this interaction blocks the implicature arising with singular epistemic algún. They follow Link () in assuming that plural individuals are sums of atomic individuals, and Martí () in considering the morphologically plural determiner algunos responsible for introducing semantic plurality. The denotation of a nominal phrase headed by algunos will thus comprise all the atomic individuals in the extension of the nominal predicate and their sums. When the plural meaning combines with the semantics of algún, the effects due to the competition with singleton alternatives that we saw for the singular form (§..) do not arise anymore: alternative singleton domains containing an atomic individual would yield a contradiction (given the prerequisite of plurality); alternative singleton domains containing a plural individual (a sum), on the other hand, would

12 For some tricky data on number with German kein cf. Kratzer () and discussion in Penka (:  n. ). 13 The other natural question, whether it is possible to assume a unique lexical entry for the epistemic and polarity-sensitive singular uses, will be the topic of the next sections.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

end up being equivalent to the asserted proposition. This derives a ‘plain’ reading for the plural. One problem with this analysis is, however, that there is evidence that in fact the reading of plural algunos is not so plain: as Martí (, ) has shown, the Spanish plural contrasts with the ‘plural indefinite article’ unos in some contexts. Martí’s conclusions on the behavior of the plural are apparently at odds with the analyses that have been given for the singular: while the singular emerges as an antispecific form, the plural is apparently necessarily specific, in the sense that it has to be D-linked to the previous context, by being anaphoric or, better, implicitly partitive (referring to previously introduced sets, cf. Enç ). Martí (: ) provides the following example: in contrast to (b), in (a) the children have to be part of the contextually salient group of children that the teachers are supervising (‘some of the children’). In the example in (b), instead, the teacher could be talking about a random news s/he heard on the radio (‘some children’). () Martí (: ); context: Teachers A and B are on an excursion with a group of children, of whom they are in charge. Teacher A comes running to teacher B: a. Teacher A: ‘¿ Te has enterado? Algunos niños se han perdido en el bosque’ ‘Have you heard? algunos children got lost in the forest’ b. Teacher A: ‘¿ Te has enterado? Unos niños se han perdido en el bosque’ ‘Have you heard? unos children got lost in the forest’ Martí accounts for this as a partitivity implicature introduced by the component alg-, which in her analysis contains a contextual variable that needs to receive its value from the discourse context (not such a variable being present with unos). At this point, however, as Etxeberria and Giannakidou () remark, the description of the alg- morpheme given for the plural is incompatible with the description given for it in antispecificity (epistemic) accounts, and forces the assumption of two different lexical entries, with alg- taken to be ambiguous. Etxeberria and Giannakidou (, ) have proposed a unified analysis, arguing that also the plural is ‘antispecific’, in the sense that the anti-singleton constraint also holds in plural uses. In the different versions of their proposals, they maintain that the component of referential vagueness, which they take to be responsible for the epistemic use in the singular, is present also in the plural; the anaphoric link of the plural to the discourse context is considered not to be a unitary phenomenon, but to depend on various contextual conditions. In my diachronic corpus work, I did not control for possible fine-grained differences in the meaning of the plural forms of Latin aliquis, and I disregarded the plural forms in my study of the Romance developments, since they seem to display substantial continuity with respect to the Latin use, and do not participate in the Quantifier Cycle. In a way, thus, the split between the singular and the plural (however it is to be analyzed) is already in place in Classical Latin, and is not a diachronic phenomenon. I will therefore leave plural forms aside in what follows. I should remark, however, that the anaphoric behavior of the Spanish plural form is remindful of what was observed in §. with respect to the role of alius ‘other’ in the etymological origin of aliquis: if the diachronic continuity is not an illusion, as further research

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



should ascertain, the same ingredient that ensures the variation component for the epistemic reading ‘some or other’ could also be responsible for the dependence on the discourse context manifested by anaphoric readings of the Spanish plural.

. Diachronic developments .. Overview In the previous sections, I already gave some indications of the fact that historically we witness both a reduction and an extension of contexts with the Romance continuations of aliquis. We also frequently saw signs that we are dealing with a still ongoing process of change (cf. the situation of Portuguese according to Martins , a,b, as well as the fluid situation in French). Epistemic uses for the continuations of aliquis are found in all Old Romance varieties, cf. in () an example from Old Italian (Stark : ) where alcuno contrasts in its epistemic function with specific uno: () domandarono alcuno santo e savio uomo che udisse la confessione d’un lombardo che in casa loro era infermo; e fu lor dato un frate antico di santa e di buona vita ‘[The two brothers] asked for a holy and wise man who could hear the confession of a Lombardian who was in their house, sick, and they were given an old monk of holy and good life’ (Boccaccio, Decameron ) The reduction of contexts consists mainly in the loss of epistemic uses in some languages (Italian and French), as well as in the sensible reduction of polarity contexts also in languages that continue aliquis as narrow-scope indefinite under negation. A complementary form of reduction of contexts is found in Catalan, where the epistemic function is maintained and the NPI one is lost. As for the extension of contexts, once we consider that the extension into downward-entailing contexts is already a Late Latin phenomenon and is, thus, inherited, we conclude that phenomena of extension proper to Romance are, instead, of a syntactic nature, and of quite recent attestation: in Portuguese, where algum is reanalyzed as an n-word, it expands to the pre-Infl area. Martins (Martins a: –) finds the first examples of pre-Infl orders in her corpus starting from the eighteenth to nineteenth centuries. Semantically, thus, we see only processes of restriction of licensing contexts, that is, processes that are compatible with the Subset Principle (cf. Martins  and §...). The extension, instead, happens ‘mechanically’ in the syntax once the contexts for the licensing of the NPI are so restricted as to reduce to direct negation; at this point a reanalysis happens interpreting the originally semantic-pragmatic licensing of an NPI as a syntactic licensing relation between an n-word and the negative operator (see further the discussion in §.). Once the lexical entry of the indefinite is reanalyzed as containing a formal feature driving the syntactic licensing, as in Portuguese, the syntactic contexts extend automatically to all compatible environments in a non-strict Negative Concord language,

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

i.e., also to the pre-Infl position. The syntactic reanalysis of the lexical item leads to a lexical split between the epistemic and the ‘negative’ variant, as I will discuss in §.. The schematic representation in () summarizes the trajectories discussed: () Schematic overview of the development Latin

Romance Plural plain/specific (Dlinked) indefinite

aliquis epistemic indefinite

Singular weak-polarity item

strong-polarity item negative-concord item

The cline implied by the representation in () is well documented in the history of Romance. It has been investigated especially for Old French, where the extension of aucun into negative-polarity contexts has been related to the demise of the Old French negatively marked indefinite nul ‘none’, which directly continued Latin nullus. I examine the Old French situation more in detail in §.., and I compare it to the Old Italian evidence in §... The limits of this work prevent me from undertaking a diachronic analysis of the Ibero-Romance developments, which must be left for future research. However, I will include some data from Martins’ (, a, b) diachronic survey of Old Spanish and Old Portuguese, as well as data from Old Catalan from Bergareche and Saldanya (), when, in concluding this section, I will single out the phenomenon of DP-internal inversion, which plays a crucial role in my account (§..). .. A closer look at Old French In Old and Middle French aucun could be used both as a pronoun and as a determiner, and both in the plural and in the singular. It could be preceded by the definite article in the plural. As a determiner, it could appear both pre- and post-nominally (Déprez and Martineau , Labelle and Espinal : –). In Old French aucun typically has a ‘positive’ meaning, whereas nul is used in negative environments as the typical Concord element.14 This element, which had both determiner and pronominal uses, has been lost in Modern French, where it survives only in fixed expressions such as nulle part ‘nowhere’ and, in very high stylistic registers, in subject function (Foulet : §, Muller , Ingham a). In the adjectival function it has been replaced by aucun (and pas un), in the pronominal function by personne and rien (for OF nule chose). I will come back to the behavior of nul in chapter . For now, let us concentrate on aucun. Old French aucun had the same value as Modern French quelque, both in its singular and in its plural () form.15

14 Also in the form neül. An archaic form nun is very rarely attested in the thirteenth cent., cf. Foulet (: §). See further §... 15 The spread of quelque is a later phenomenon, cf. later in this section.

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



() a. Aucun se sont aati ‘Some have boasted..’ (Feuillée , Foulet : §) = Mod. Fr. ‘Certains se sont vantés’ b. Aucunes gens dient . . . ‘Some people say . . .’ (Roman de la Rose , Buridant : ) It was in general quite rare in the oldest texts (Foulet (: §) counts only seven examples in his corpus). It retains its ‘positive’ value until the fifteenth century and occurs very infrequently in negative contexts (–), cf. Foulet (); Buridant (); Déprez and Martineau (); Ingham (a); Ingham and Kallel (); Labelle and Espinal (). () Dame, or ne faites tel despit / k’il n’aient de vous aucun bien ‘Milady, do not be upset to such a point that they will not receive from you any good’ (Feuillée –, Foulet : §) () Car li roys ne fait jugement / D’ aucun chevalier nullement ‘because the king cannot judge any knight’ (La prise d’Alexandrie –) It is, however, already found in irrealis contexts, especially in the singular: () E si aucuns vescunte u provost mesfait as humes de sa baillie . . . ‘and if any viscount or provost mistreats people under his authority . . . ’ (Leis Willelme art. .) () Et s’il venoit aucun noble homme / De France, d’Espaingne ou de Rome, / De Lombardie ou d’Alemaingne, / Ou d’Angleterre ou de Sardeingne, / Ou de quelque part qu’il venist . . . ‘and if a(ny) noble man from France, from Spain or from Rome, from Lombardy or from Germany, or from England or from Sardinia, or from whichever part he may come, arrives . . . ’ (La prise d’Alexandrie, –) () Et si vous prie que ad ce tenez la main que je n’aye cause d’en trouver aucune matere de jalousie ‘and I beg of you to guard her so that I may have no reason to find any question of jealousy’ (Middle French: CNN,.) Dhoukar (: ) reports some later cases where aucun occurs in the plural under the scope of negation, with the expression aucunes choses ‘some things / anything’: () eux qui ne connoissent véritablement aucunes choses de la nature ‘those who actually do not know anything about nature’ (Pascal, , Entretien avec M. de Sacy, –, Dhoukar : ) Ingham (a) has studied in detail how the replacement of nul by aucun happened in continental French (–) and insular Anglo-Norman (–), focusing on administrative documents. He finds a substantially similar pattern of evolution in Anglo-Norman and in the rest of Medieval French. In both linguistic areas, the indefinite aucun replaces the polarity indefinite nul first in polarity-sensitive environments (‘nonassertive’ in Ingham’s terminology) and then

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

in negative clauses. This happens in insular Anglo-Norman slightly later (second half of the fourteenth century) than in (northern) Continental French (first half of the fourteenth century), but according to the same overall pattern.16 Ingham’s (a) study is complemented by the diachronically more extensive analysis of a French corpus of private correspondence in Ingham and Kallel (), covering the time span –. Both studies are notable in that they provide previously disregarded evidence from non-literary texts. Ingham’s (a) study focuses on two main contexts: the prototypical negative one, represented by clauses with the negative marker ne, and the antecedent of conditionals, taken to be the most representative polarity-sensitive context, due to its frequency and to the fact that polarity uses of nul and aucun tend to concentrate there. In Ingham and Kallel () also further polarity contexts are taken into consideration: interrogative clauses, comparatives, cases of indirect (superordinate) negation. Ingham (a) also controls for the pre- versus postverbal position of the indefinite, reaching interesting conclusions in this respect. Unfortunately, differences in morphological number are disregarded in both studies, so that we have no information as to which kind of aucun is in fact present or more frequent at the onset of the change, and survives during the diachronic process, in positive contexts.17 A further aspect of this reconstruction which should be investigated in greater depth concerns the original value of Old French aucun. Ingham (a: ) analyzes twelfth-century Old French as featuring a two-way split between a ‘positive’ indefinite aucun and another form serving as indefinite in both nonassertive and negative contexts, nul. However, as can be deduced from Foulet (), Buridant (), and from Ingham himself, there is no documented stage of French where aucun cannot occur in the antecedent of conditionals. That is, a polarity use of aucun had always been possible, and, given the Late Latin situation, where aliquis had already expanded into some polarity contexts, this is indeed what we expect in Old French. Ingham’s (a: ) data on Old French administrative documents from the northern cities show that already in the time span – there are  instances of aucun in antecedents of conditionals against four instances of nul. The latter is not attested in this context in the subsequent period (–).18 What seems to clearly emerge from Ingham’s analysis is that aucun follows a precise cline in its development, 16 Ingham interprets the delay as a typical feature of a change process spreading from a center (apparently Paris in this, as in many other cases. cf. Ingham a: ) to a periphery. 17 We know from later sources that the plural form survives the longest in the ‘positive’ use: all positive examples cited by Haase (: ) for the normal use of aucun in the seventeenth century are plural. Interestingly, Haase (: ) remarks that in the seventeenth century it was possible to have the plural form of aucun (and nul) in negative contexts. 18 In this respect, the situation found in Anglo-Norman administrative texts by Ingham seems to be really different: here nul is indeed much more frequent than aucun in conditional contexts in the first period ( instances vs. ), and still sufficiently well attested in the second one, cf. Ingham (a: ): the question is here whether we are witnessing a previous change, involving the spread of nul to conditionals, or whether Anglo-Norman rather displays a conservative grammar in this respect (the latter being Ingham’s interpretation of the situation).

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



first raising in frequency in the antecedent of conditionals, and only later significantly spreading into negative contexts (cf. also Buridant : , ).19 In the latter, the postverbal occurrences are substantially more frequent than those in preverbal position: this is in accordance with the well-known path leading from an NPI behavior to proper Concord-word status. For Continental French, Ingham’s corpus for the Old French period (until ) displays only  instances of aucun in negative contexts, all of them postverbal. The situation changes in the second period (early Middle French until ), where aucun is in  of the cases the indefinite used in preverbal position in negative contexts ( instances); the cases raise to a remarkable  when the indefinite is in postverbal position ( instances). As mentioned, in insular Anglo-Norman the spread is slower: the period – still witnesses nul in the majority of the negative cases, with just three attestations of aucun in preverbal position (vs.  for nul). In postverbal position aucun represents  of the cases. Although aucun surpasses nul in negative contexts, and is the most frequent of the three indefinites considered by Ingham and Kallel () during the first period, we see that it in turn recedes in subsequent periods under the pressure from quelque, which takes its place in nonassertive and positive contexts. During the last period considered by by Ingham and Kallel () (–), there are no instances of aucun in positive contexts: the environments previously covered by aucun have been taken over by quelque. Negative contexts become the preferred environment for aucun, accounting for  of its total attestations. The quantitative data collected in Ingham and Kallel () agree with the findings in Déprez and Martineau (), who show for their corpus that during the sixteenth to eighteenth centuries the plural uses, as well as the pronominal ones, decay. A strong increase is observed in the use as postnominal determiner, as well as in polarity and negative contexts, in the sixteenth and seventeenth centuries. These points are further confirmed by Labelle and Espinal’s () corpus study, where they reach similar conclusions concerning the phases and the chronology of the change. I conclude the discussion of Old French with a very short overview of what is known from the history of the competitor of French aucun in the epistemic use, i.e., quelque. The competition with quelque is relevant for our discussion because it may have been a relevant factor in the specialization of French aucun in direct-negation contexts. The diachronic development of French quelque has been investigated by Foulet (), Combettes () and Jayez and Tovena (). The origin of quelque is a concessive relative clause: ‘quel N que CP’ (or ‘quelque N que’), cf. Buridant (: ). In this structure, quel is a wh- correlative element (introducing a variable ranging over a domain of individuals); que (qui, où) is a relative pronoun. The verb in the relative 19 A similar path can be observed for English any, which is originally excluded from negative contexts and spreads there only in the sixteenth century (Ingham , a: –). The late chronology of this phenomenon excludes the possibility that it might have influenced the Anglo-Norman developments.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

clause is typically in the subjunctive form: according to Jayez and Tovena () the use of the subjunctive adds a domain widening / scalar interpretation, which results in a concessive interpretation. () a. et en quel lieu ou il soit, en avroilt il molt grand duel ‘and, in whichever place he would be, he would have much pain’ (Lanc. Graal, Jayez and Tovena : ) b. vos poez aler quel part que vos volez, et je sui ci anprisonee ‘You can go wherever you want, and I am imprisoned here’ (Yvain ) At a further stage of development, a univerbation has happened (with elision of the material in the relative clause) and quelque has the distribution of a determiner. Its function, according to Jayez and Tovena (), is that of an antispecific determiner with a free-choice flavor; the antispecific meaning results from the grammaticalization of implicatures arising in the original concessive construction. Note that this stage is only logically subsequent, since many texts show examples of both the sentential and the word-like structure. Foulet () sees the origin of the univerbation (a) in fixed expressions that are found quite frequently in Old French texts, such as à quelque pain(n)e (cf. b): the interpretation of this item is equivalent to the free-choice expression à quelle peine que ce soit, i.e. ‘at whatever pain, independently of what it may cost’. () a. Tourjorz a chascuns quelque teche (Roman de la Rose , Jayez and Tovena : ) ‘everytime everybody has some stain’ b. a quelqu’enui, a quelque painne, ting cele voie (Yvain ) ‘At the cost of whatever trouble, whatever pain I followed that road’ At the last stage of development, starting in Middle French, quelque becomes an epistemic determiner: the content changes from indifference to ignorance. This corresponds to the modern usage. () Si s’en va et fait mauvese chiere, dont sa femme cognoist bien qu’il y a quelque chose ‘However he goes and makes a face, hence his wife realizes that something is wrong’ (Quinze Joies de Mariage , Jayez and Tovena : ) .. A closer look at Old Italian Since the earliest Old Italian texts, singular alcuno, as a determiner and as a pronoun, is attested with two functions: it can be an epistemic indefinite or an NPI; in the plural, it only has the plain indefinite function (Stark , ). Differently from Modern Italian, it does not entertain a privileged relation with negation. It is predominantly used as an epistemic indefinite or, as NPI, in non-negative polarity contexts. Table () shows the distribution of alcuno in Brunetto Latini’s Rettorica:

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



() Distribution of alcuno in Brunetto Latini, Rettorica Total forms



Singular



Plural

 (all in non-negative contexts)

Under negation

 (all singular)

In non-negative NPI contexts / with epistemic value



Inversion

 (under negation)

An example of alcuno with epistemic value in the singular is shown below: () Publiche questioni son quelle nelle quali si tratta il convenentre d’alcuna cittade o comunanza di genti ‘Public cases are those in which the situations of some city or community of people are treated’ (Brunetto Latini, Rett., p.  l. –) The epistemic use is possible in pre-Infl position: () a. Alcuno à furato d’una chiesa uno cavallo ‘Someone has stolen a horse from a church’ (Brunetto Latini, Rett. p.  l. ) b. et alcuna cosa nuova v’agiunse ‘and he added something new’ (Brunetto Latini, Rett. p.  l. ) Compare with uno in the specific known function (cf. Stark , Stark : ch.  for detailed discussion of the conditions favoring the specific interpretation of uno): () e là trovò uno suo amico della sua cittade e della sua parte ‘and there he found a friend of his from his own town and his own political party’ (Brunetto Latini, Rett., p.  l. ) The epistemic use of alcuno is most typically found in generic contexts, but episodic contexts are also possible (cf. ). It is frequent in the antecedent of conditionals, where it takes narrow scope; in these cases it is often compatible with either an ignorance or an indifference reading, as in (): () se alcuna di queste cinque parti falla nella diceria . . . ‘if one / any of these five parts is missing in the oration . . .(Brunetto Latini, Rett. p. . l. –) ‘Positive’ and ‘negative’ uses can coexist in the same clause, showing that alcuno at this stage is still unmarked with respect to polarity preferences. In the example in (), the Latin source passage for the Rettorica has aliquis only for the positive context, whereas a negative indefinite nihil is used in the negative context. In the Old Italian translation, alcuno is found in both contexts:

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

() a. Assuntiva è quella che per sé non dà alcuna ferma cosa a difendere, ma di fuori prende alcuna difensione; ‘The assumptive issue is that which of itself does not provide anything substantial for rebuttal, but takes some reason of defense from the outside’ (Brunetto Latini, Rett. p. . l. –) b. assumptiva, quae ipsa ex se nihil dat firmi ad recusationem, foris autem aliquid defensionis assumit. ‘The assumptive issue is that which of itself provides no solid basis for a counter plea, but seeks some defence from external circumstances.’ (Cic. de inv. ) In the scope of negation, where it is still quite rare in the stage I have analyzed (until the first half of the fourteenth century), alcuno can be used interchangeably with nessuno; however, in conformity with its NPI nature, it must be c-commanded by the overt expression of the negative operator, i.e., it occurs mainly post-Infl or, when pre-Infl, preceded by the high correlative negation né (cf. also Kellert ): () o diremo che studiano in cose che non sono da neuno uso né d’alcuna utilitade ‘or we will say that they occupy themselves with matters that are of no application and no utility’ (Brunetto Latini, Rett. p. . l. ) Stark’s () broader corpus study, which also considers the fifteenth and sixteenth centuries, reveals that alcuno increases in frequency much faster than French aucun in negative structures, which become its privileged context (cf. Stark : ). In negative contexts, alcuno invariably takes narrow scope. For instance, in (), Brunetto Latini discusses the conditions required for a controversy: at least one of four issues (‘a question about a fact, or about a definition, or about the nature of an act, or about legal processes’, cf. Cic. de inv. .) has to be applicable; if none is present, there cannot be any legal case. () E così conviene che ssia l’una di queste inn ogne maniera di cause, perciò che in qual causa no’nde fosse alcuna, certo in quella non porrebbe avere contraversia ‘And thus it is appropriate that one of these be in any kind of case, because, in a case in which there was not any, for sure there could not be an argument’ (Brunetto Latini, Rett. p. , l. –) () Ragione è quella che contiene la causa, la quale se ne fosse tolta non rimarrebbe alcuna cosa in contraversia. ‘Reason is what holds the case together: if this were taken away there would be nothing left in the controversy.’ (Brunetto Latini, Rett. p. , l. ) Often an emphatic reading is observed in negative contexts: () tu non dirai alcuna cosa dell’aversarii, né questo né quello ‘you will not say anything about the enemies, neither this nor that’ (Brunetto Latini, Rett. p.  l. )

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



But under negation also a non-widened reading is possible, which in Modern Italian would translate as una qualche ‘some’: () a. i parlieri si sforzano di provare una cosa essere onesta o disonesta, non nominando alcuna certa persona ‘The orators try to prove that something is right or wrong, without mentioning a specific person’ (Brunetto Latini, Rett. p.  l. –) b. Quella è contraversia in ragionamento nella quale non si considera alcuna cosa che ssia per scrittura, ma prendesi argumento . . . ‘A case of general reasoning is one in which one does not consider something written, but some argument . . . ’ (Brunetto Latini, Rett. p. , l. –) The only example with inversion in the Rettorica is reported below; it occurs under negation: () nelle quali non à tencione alcuna intra llui e la donna ‘in which there isn’t any conflict between him and the lady’ (Brunetto Latini, Rett. p.  l. ) The replacement of alcuno in the epistemic use in Modern Italian, qualche, is very rare at the stage that I considered, and does not yet qualify as a competitor. Dante and Brunetto use it only a couple of times (cf. ),  instances are found in Boccaccio’s work. () se tu tronchi qualche fraschetta d’una d’este piante ‘if you cut some little branch from one of these plants’ (Dante, Inf. .) Stark (: ch. ) observes that qualche suddenly increases in frequency in the fifteenth and sixteenth centuries, and that it occurs in a broader set of negative-polarity and modalized contexts than in Modern Italian. Rohlfs (: §) considers a correlative construction of the type already seen for Old French to be its etymological source, cf. (): () la riviera del sangue in la qual bolle qual che per violenza in altrui noccia ‘the bloody stream within which those boil who harmed others with violence’ (Dante, Inf. ., Rohlfs : §) .. Inversion Let us now add some diachronic data on DP-internal inversion, for which I will provide a theoretical analysis in in §.. At the early stages of the Romance languages inversion between certain determiners and the rest of the nominal phrase is freer than at later periods. Here I point out some distributional facts that show that, at least with certain determiners, inversion starts to become more frequent when the indefinite is interpreted under the scope of negation. This will be important for my subsequent proposal.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

First of all, let me remark that inversion between the head noun and determiner-like elements was already possible in Latin, with all categories of determiners. In the table below are some data on the distribution of Latin indefinites from Marouzeau (). Marouzeau remarks that the normal position for these elements is the first in the nominal constituent, independently of information status. He attributes the inverted order to the special information status of the head noun; the determiner occurs postposed if the head noun is highlighted (‘mis en relief ’, Marouzeau : ), for instance in contrastive (disjunctive) constructions and enumerations.20 () Linear order within the DP after Marouzeau (: –): Item pre-N post-N aliquis Cicero speeches aliquis Caesar quidam Cicero speeches quidam Caesar quidam Sallust ullus Caesar ullus Sallust nonnullus Caesar nonnullus Sallust nonnullus Cicero letters (Fam) nullus, nemo Cicero speeches & letters nullus, nemo Caesar nullus, nemo Sallust

            

            

Interestingly, Marouzeau also notes some special cases with the negative indefinites, which point to a special information status of the determiner rather than of the nominal restriction: if nullus is the antecedent of an exceptive phrase introduced by nisi or praeter, it may be postposed: () consolatio nulla nisi . . . consolation:nom no:nom if.not ‘no consolation at all, were it not for . . . ’ (Cic. Fam. ..) We will see in §.. that exceptive negation is a focusing construction. It is natural to propose, thus, that some cases of Latin DP-internal inversion were due to a focalization of the determiner. This fits well also with recent investigations of Latin syntax that have highlighted the particular productivity of movement to the left edge of various constituents (crucially comprising elements of the nominal phrase) in Latin, cf. in particular Devine and Stephens (: –), Giusti and Iovino (), Ledgeway (: –), Danckaert (). Concerning Romance, optional inversion of the continuations of aliquis was possible in all functions in Old Romance, and apparently, as in Latin, was due to a number of different factors. Martins (a) attributes it to a more widespread possibility of 20 A phrase-initial position is the unmarked one also for demonstratives; yet they could also be postnominal in some constructions; for an account cf. Gianollo (), Iovino ().

OUP CORRECTED PROOF – FINAL, //, SPi

Diachronic developments



DP-internal middle scrambling in Medieval Romance varieties. For instance, in Old Italian, while the unmarked position of quantifiers in the nominal phrase is DP-initial, there are also marked DP-final orders, with both universal and existential quantifiers (Giusti : –). With respect to inversion, the situation emerging can be summarized as follows: (i) Italian and French: no substantial diachronic changes are observed; inversion remains optional and recedes in the modern varieties; (ii) European Portuguese: inversion was optional in all contexts in Old Portuguese, and possible also in weak-polarity environments (where algum later disappeared); (iii) Spanish: inversion was optional in all contexts in Old Spanish, and possible also in weak-polarity environments; (iv) Catalan: an NPI use was possible in earlier stages (fourteenth century) and was recently lost (twentieth century): Par () witnesses a stage of obligatory inversion. For Old Italian, Stark (: –) finds that all narrow-scope indefinites (veruno, nessuno, niuno, nullo, alcuno, punto) are found in the inverted order, although at different rates. In Stark’s () corpus, the majority of postposed orders of alcuno and niuno (the only two elements where inversion is statistically relevant) are found in direct negation contexts. Also Giusti (: ) notes that the marked order N(P) - D is more frequent when quantifying elements such as ne(ss)uno ‘no one’, alcuno ‘anyone’, veruno ‘anyone’ are in the scope of morphological markers of negation like non, né, sanza = ‘senza’.21 () from Giusti (: –) a. sanza numero altro alcuno (Dante, Vita Nuova .) ‘with no other number’ (cf. in alcuno altro numero, Dante Vita Nuova .) b. né guarnigione alcuna né fortezza ‘with no garrison nor defension’ (Rinuccino, Sonetti, m.–) Here, however, we should note that the rates of inversion seem to be different with alcuno and niuno / nessuno: namely, inversion is found much more often with the former than with the latter in the Old Italian texts of the OVI corpus. With nessuno, postposition is found only once in Stark’s () corpus. Examples of postposition with niuno are given in (a–b), respectively in a negative context and in the antecedent of a conditional, where niuno has the function of an NPI:

21 See Ramat () on Italian veruno, which has a non-negative origin (< Lat. vere ‘really’ + unum ‘one’) but has negative and NPI uses since the first Old Italian documents. Ramat () concludes that the origin of the negative value is to be found in comparative structures, where it becomes frequent in the standard of comparison, an NPI-environment. For the discussion on the possible parallel etymology of Romanian vreun see Fˇalˇau¸s (), Gergel ().

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

() a. e non ebbi paura niuna ‘and I didn’t feel any fear’ (Boccaccio, Decameron .) b. se bisognerà far cosa niuna, noi la faremo ‘and if anything will be needed, we will do it’ (Boccaccio, Decameron .) Inversion with nessuno is not productive anymore in Modern Italian. Archaic examples (and only with the ‘light noun’ cosa ‘thing’) are still found in authors of the nineteenth century (e.g. non val cosa nessuna, Leopardi, A sé stesso). I provided quantitative data on this phenomenon in §... In French inversion was productively used (apparently with an emphatic contribution) up to the language of the seventeenth century, and typically after the preposition sans ‘without’. It is attested in later writers such as Chateaubriand and is given as a possible rule of the language in the Grammaire nationale (Bescherelle and Bescherelle ), from which the following examples are taken: () a. allons sans crainte aucune (Molière) ‘without any fear’ b. sans réserve aucune (Molière) ‘without any reservation’ c. sans peine aucune (Voltaire) ‘without any effort’ In this stylistically conventionalized use it may still belong to the competence of some contemporary speakers, as we saw in §... Martins (, a) has examples for Old Spanish and Old Portuguese where there is no DP-internal inversion under negation: () a. No consientes algun consejo ni tienes reposo Old Spanish ‘You don’t take any advice, nor do you have any rest’ (Elicia ., Martins :  from Keniston : ) b. ca nom vos pode ende v˜ıir alg˜uu˜ bem Old Portuguese ‘because this will not bring you any good’ (Demanda do Santo Graal, Martins a: ) Finally, in Old Catalan we find again the situation seen in the other Medieval varieties. Catalan is particularly interesting, however, in that inversion seems to have been grammaticalized, i.e., made obligatory in the negative polarity uses, quite early; moreover, we have seen above that the polarity use of algun has disappeared in later stages of Catalan. Bergareche and Saldanya () observe that algun in Old Catalan has a clear determiner-like distribution: it precedes prenominal adjectives and other indefinite items (e.g., altro ‘other’). The only exception they find to this behavior is the ‘inverted’ postnominal use, which they consider a residual use of the Latin-like syntax. Crucially, however, they remark that this order is only found in special contexts and correlates with a special interpretation: ‘Postnominal algun is a polarity item, so it can only be found in modal or negative contexts’ (Bergareche and Saldanya : ).

OUP CORRECTED PROOF – FINAL, //, SPi

Inversion as focalization in the Determiner Phrase



() from Bergareche and Saldanya (: ) Amich, no m’ajusté ne m’acosté a neguna fembre, ne fiu mal ne desplaer a home algun (Exemples .) ‘My friend, I did not join or approach any woman, nor did I cause any evil or upset to any man’ I believe that this use need not be considered an exception to the determiner use of algun, but rather results from an additional syntactic operation of inversion. As Bergareche and Saldanya () correctly observe, this use develops separately from the other functions of algun.

. Inversion as focalization in the Determiner Phrase The material examined in the previous sections shows quite clearly that what is observed with the Romance continuations of aliquis is an instance of the crosslinguistically attested Quantifier Cycle. The only difference with respect to the classical model is that the point of departure is not a plain indefinite, but rather an element already incorporating constraints on its use. I modeled this by analyzing Latin aliquis as an epistemic indefinite in chapter . Now that we have obtained a picture of the synchronic and diachronic variation in the uses and in the syntactic distribution of the Romance continuations of aliquis, I will proceed to address the question whether it is possible to assume a unique lexical entry for the epistemic and polarity-sensitive singular uses. My answer will be largely positive: I propose that, in those ancient and modern Romance varieties that display the two uses, the polarity-sensitive one, and in particular the use in direct-negation contexts, emerges through a semantically triggered operation of focalization of the determiner inside the Determiner Phrase. In my account, a lexical split obtains only in Modern Portuguese, which witnesses a particularly advanced stage of grammaticalization (cf. §..), whereby the continuation of aliquis behaves like an n-word (i.e., an item endowed with a formal negative feature) in direct-negation contexts. The grammaticalization of an n-word status happens earlier in the history of French (cf. §..), but in this case no lexical split obtains, since the process of grammaticalization in negative contexts correlates with the disappearance of the ‘positive’ epistemic variant. In my account, this lack of need for disambiguation is the reason why inversion never becomes obligatory in the history of French. I will start from the premise (cf. §..) that variation in the system of indefinites is due to presuppositions that constrain the way in which indefinites choose their quantificational domains. The functional need for expressivity may drive diachronic change phenomena affecting these constraints and lead to shifts in the distribution of indefinites. Original contextual effects may then result in grammaticalization, modifying the lexical entry of the indefinite. I propose that emphasis, intended as a form of focus, is a fundamental contextual condition in the case of the Ibero-Romance continuations of aliquis, and that

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

it interacts with the anti-singleton presupposition built into the lexical entry of (epistemic) algún / algum (cf. chapter ). I analyze DP-internal inversion with the continuations of aliquis as a form of (emphatic, i.e., scalar) focus (Krifka  for even): alternatives ordered along a scale are evoked; the focus denotation is then the low extreme of the scale. Focus alternatives interact with the alternatives already evoked by the lexical entry of the determiner in virtue of the anti-singleton constraint, triggering maximal widening of the domain of quantification. This makes the nominal expression pragmatically suitable for a polarity-sensitive environment. Focus has a syntactic effect within the DP: the quantificational determiner surfaces in a DP-internal Focus position, and the remnant raises past it to a Topic phrase hosting backgrounded material. Emphasis, intended as focus, may trigger the grammaticalization of originally optional entailments: in Spanish and Portuguese this is manifested in the obligatoriness of syntactic inversion in negative contexts. Grammaticalization, in turn, results in bleaching of emphasis in the course of time: once inversion is conventionalized, the construction loses its emphatic flavor, and in principle the item may turn into an n-word, as we saw for Portuguese. At this point a lexical split happens, since the n-word, as an element of Negative Concord, carries an uninterpretable formal feature [uNeg]. The non-emphatic status of n-words is confirmed also by the very constrained application of inversion in French (cf. §..). In what follows, I will first propose a syntactic analysis for inversion, and then turn to the semantic account. In dealing with the syntax of inversion, I will capitalize on previous work on DP-internal inversion in Romance (Bernstein ) and on the parallelism with similar phenomena happening in the clause. .. A distributional generalization Let me summarize what we saw above concerning the crosslinguistic distribution of DP-internal inversion with the polarity-sensitive continuations of aliquis. We have observed that in the modern varieties inversion appears only under the scope of the negative operator and of anti-additive elements such as ‘without’.22 It is optional and subject to register constraints in Italian, very rare in French; it yields an emphatic assertion and has the function of strengthening the negation: () Non esiste correlazione alcuna tra questi fenomeni ‘There is no correlation at all among these phenomena’

(It)

() sans attention aucune ‘With no attention at all’

(Fr)

In European Portuguese it is obligatory when the indefinite has an n-word use. No emphasis is conveyed. It is also obligatory in Spanish in the NPI use when c-commanded by an overt expression of negation, whereas it is optional in weak-polarity contexts. The speakers perceive this construction as emphatic. 22 As I will discuss in chapter , in the analysis I adopt for Negative Concord also pre-Infl n-words are in the scope of an abstract negative operator.

OUP CORRECTED PROOF – FINAL, //, SPi

Inversion as focalization in the Determiner Phrase



The crosslinguistic picture is summarized in (): () Behavior with respect to DP-internal inversion = Inv (only singular): It Fr Cat Sp Pt alcuno aucun algun algún algum Sg NPI (optional Inv) n-word Ep (no Inv) Ep (no Inv) & Ep (no Inv) or NI NPI (under & n-word (optional negation: (obligatory Inv) obligatory Inv) Inv) A crucial generalization emerges at this point: inversion becomes obligatory in the course of history only in those varieties where the singular form continuing aliquis also maintains another use outside the scope of negation (= the epistemic variant), i.e., Spanish and Portuguese. In varieties where only polarity-sensitive uses are preserved, inversion is only optional and it has an emphatic value. Note that obligatory inversion is not necessarily correlated to n-word-status (cf. the NPI status of the Spanish continuation), but is conditioned by the dependence on a negative operator. .. The syntax of DP-internal inversion Belletti (, ) analyzed cases of ‘subject inversion’ (VS orders) in Romance as a way to focalize the subject. For Italian, she proposes a low, clause-internal focalization of the subject in a vP-peripheral position in case of information focus on the subject. Examples are given in () and the proposed structure in (). () Belletti (: ) a. Speaker A: Chi è partito / ha parlato? ‘Who has left / has spoken?’ b. Speaker B: È partito / ha parlato Gianni is left / has spoken Gianni c. Speaker B: * Gianni è partito / ha parlato () . . . [TP . . .[AspP [TopicP [FocusP [TopicP [vP [VP . . . ]]]]]]] Given the rules governing nuclear stress, the last (= most embedded) constituent ends up being prosodically prominent. Similarly, I propose the syntactic analysis in () for structures containing Sp. algún and Pt. algum in the epistemic and polarity-sensitive version. I will gloss over a possible finer structure of the functional projections responsible for quantification in the nominal phrase. () a. no inversion: (epistemic indefinite): [DP [D alguno . . . [NP estudiante ]]] b. inversion: (NPI): [DP [TopicP [ estudiante ]NP [FocusP [Focus alguno] . . .]]] Inversion is due to movement of the quantificational determiner to a DP-internal Focus position. This movement is followed by the displacement of the material from

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

the restrictor of the quantificational determiner in focus: I assume a raising operation to the specifier of a DP-internal Topic Phrase. As in Belletti’s case, inversion is derived by (leftward) movement of the ‘inverted’ element to a functional position for information-structural reasons (Focus), and by subsequent movement of the lexical head or of the remnant above this position. The theoretical observation is that this and similar phenomena in Romance constitute important evidence for a DP-internal Focus position and, in general, for the presence of complex left peripheries in each syntactic phase. Differently from the general idea in Belletti (), DP-internal inversion is due to the displacement of the NP-remnant to a higher DP-internal Topic position. This is not the general derivation assumed for VS orders in Italian, since the verb raises for independent requirements. However, as we will see below, Belletti () proposes a similar derivation, with an additional low Topic position, for some VOS orders in Italian. Another difference has to do with the functional motivation for the inversion: while subject movement to Focus in the clause is due to information focus, I will assume that in the case of DP-internal inversion we have emphatic (scalar) focus, licensed by DP-external elements and with DP-external consequences. Now that I have given the general gist of the syntactic analysis, let me discuss its ingredients more in detail. First of all, the existence of a DP-internal Focus projection is not uncontroversial.23 While e.g. for Giusti () there is no FocusP in the DP, Poletto (: ch. ), in dealing a.o. with historical data, assumes that an Operator position is available in the DP as well (): () [DP [TopicP [OpP [dP [AgrP . . . [NP ] ] ] ] ] ] adapted from Poletto (: ) The most convincing evidence for the existence of a Focus-related displacement in the Romance nominal phrase comes from Bernstein’s () work. Bernstein shows that throughout Romance the DP-final position of determiner-like elements such as demonstratives, possessive pronouns, indefinites, yields a focus interpretation: () DP-internal inversion in Bernstein () a. el libro interesante este ‘THIS interesting book’ b. cette femme intelligente ci ‘THIS intellingent woman’ c. uno studente di lettere qualsiasi ‘ANY student of Letters’ d. il libro mio ‘MY book’

(Sp) (Fr) (It) (It)

Some further arguments in favor of a DP-internal Focus position come from the behavior of DP-final only, discussed by Brennan () on the basis of the original 23 For a thorough discussion of the left periphery of nominal phrases, see Aboh (), Giusti (), Ihsane (: –), Alexiadou et al. (: –), Aboh et al. ().

OUP CORRECTED PROOF – FINAL, //, SPi

Inversion as focalization in the Determiner Phrase



observations in Kayne (). When only precedes its focus associate it can take scope over a variety of constituents, giving rise to potential ambiguity; when appearing finally, instead, it must associate with a preceding DP (or a subconstituent of it) and cannot take larger units as its focus associate. This speaks for the fact that only in this case is DP-internal: () Brennan () a. John spoke to [one linguist]F , too b. JohnF spoke to one linguist, too c. John spoke to [one linguist]F , only d. ??JohnF spoke to one linguist, only I assume that inversion takes place because the remnant NP raises across the Focus position to a DP-internal Topic Phrase, hosting frame-setting material. We know that restrictions of quantificational expressions are systematically backgrounded (cf. Herburger : ). Once the focused element takes a marked position in the structure, i.e., is displaced because of its informational status, we may argue that a further movement of the backgrounded material is necessary in order to achieve the right focus–background partition.24 Also, thanks to the movement of the de-emphasized material to the left, the focused element in final position may receive nuclear stress. This reminds us of prosodically grounded interpretations of displacement phenomena: Zubizarreta () accounts for Spanish VOS orders as resolution of the conflict between the Nuclear Stress Rule and the Focus Prominence Principle. Depending on the model of the interfaces that one assumes, a prosodic reordering may be a viable alternative analysis also in the case of DP-internal inversion. Leaving these considerations aside for further research, the structure I propose for examples such as (b) is shown in (). The assumption of a pair of focusrelated formal features [iFoc]-[uFoc], respectively on the Focus operator and on the indefinite, will be better motivated in chapter . It is meant to account for the obligatoriness of inversion in Spanish (and in Portuguese, although more will have to be said concerning its further grammaticalization: cf. §..). Adopting the split DP structure proposed by Giusti (), I represent the original Merge projection of the indefinite as dP, i.e., the projection closing off the inflectional domain of the DP.

24 As mentioned, the backgrounded restriction has a frame-setting function in this case (a subtype of Topic, cf. Jacobs : ). Given the parallelism with the clausal domain, it is tempting to label this position GroundP, like the left-peripheral landing site for remnant IP-movement identified by Poletto and Pollock (), and also exploited by Bianchi and Zamparelli () in their analysis of correlatives. In the split CP, GroundP hosts backgrounded, presupposed material.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

() DP-internal inversion: DP TopP

FocP NP estudiantej

Foc

[iFoc] Foc0

dP

algunoi [uFoc]

d

d0 NP algunoi [uFoc]

estudiantej

The analysis in terms of syntactic movement that I put forward here is similar to how Belletti (: –) accounts for some acceptable VOS orders in Italian. She proposes that, in structures like (), pre-S material is given information in a (possibly low, clause-internal) Topic position: () Belletti (: –) a. Speaker A: Chi ha capito il problema? ‘Who has understood the problem?’ Speaker B: Ha capito il problema Gianni. ‘Gianni has understood the problem’ b. Protegge l’uscita di Marchegiani Nesta ‘Nesta protects Marchegiani’s coming out’. Concerning our DP-internal case, there are differences across Romance concerning the size of the material that can be moved across FocusP. The nature of the restrictions is poorly understood: Martins (a) attributes those found with the European Portuguese n-word to the syntactic NPI-building operation that she assumes in her framework, with cyclic head movement and an ensuing requisite of adjacency: () Martins (a: ) a. *Animal selvagem algum vive aqui animal wild algum lives here

(EP)

OUP CORRECTED PROOF – FINAL, //, SPi

The semantic role of focus b. *Animal do deserto algum vive aqui animal of.the desert algum lives here

 (EP)

Spanish and Italian seem different. For at least some speakers, preferences seem to be conditioned also by phonological phrasing, cf. (a) and (b).25 () Martins (a: –) a. No conozco libro alguno de matemáticas que discuta este teorema (Sp) ‘I am not aware of any book of mathematics that might discuss this theorem’ b. * No conozco libro de matemáticas alguno que discuta este teorema (Sp) c. No asistí a conferencia alguna interesante (Sp) ‘I did not attend any interesting lecture’ d. No asistí a conferencia interesante alguna (Sp) () a. Non conosco libro di storia alcuno che tratti di questo tema (It) ‘I am not aware of any book of history that might discuss this topic’ b. ?*Non conosco libro di matematica alcuno che tratti di questo tema (It) ‘I am not aware of any book of mathematics that might discuss this topic’

. The semantic role of focus In the previous section I argued that the fact that the epistemic indefinite assumes the semantics of a negative-polarity element is connected to syntactic movement to a DP-internal Focus position. The reason behind this proposal is that NPI-uses have been convincingly argued to be semantically and pragmatically correlated with focus: summarizing a long tradition of research on polarity sensitivity (see Tovena , Israel  for comprehensive discussion), Chierchia (: ) states: ‘The functional basis of it all is emphasis, potential or actual’. Exactly which form of focus is involved? Two interpretations are in principle possible: one is to consider it as a form of verum (polarity) focus, which focuses on the truth value of a sentence; the other consists in considering it a form of scalar focus. Willis et al. (a: ), in discussing the origin of reinforcers of negation, choose the first option: they treat emphasis as emphasizing of polarity. As I preliminarily said, I interpret the focus involved in DP-internal inversion as a kind of scalar focus. In this section I will elaborate on my proposal, justifying this assumption and proposing a mechanism by which the lexical entry of the epistemic indefinite interacts with the focus operator. .. Scalar focus and NPIs Recall that the semantic account for the ignorance effect that we follow here derives it as an implicature arising through competition with smaller domain alternatives. Owing to the anti-singleton presupposition, the domain of quantification of an epistemic indefinite cannot be narrowed down to an individual, giving rise to a ‘partial’ 25 I am grateful to Jacopo Torregrossa for discussing these examples with me.

OUP CORRECTED PROOF – FINAL, //, SPi



Aliquis from Latin to Romance

freedom of choice among the members of the restriction set (‘minimal domain widening’). According to Alonso-Ovalle and Menéndez-Benito (, ), who base their work on the original idea of Kratzer and Shimoyama (), the presence of domain alternatives is exploited pragmatically for a number of reasons related to the weakening of the assertion: avoid making a false claim or avoid an exhaustivity inference, depending on the interacting operators. Now, also the felicity of an NPI in a given environment depends on the interplay between the lexical meaning of the NPI and the pragmatic implicatures it triggers (Fauconnier , Horn , Kadmon and Landman , Krifka , Chierchia , Chierchia  a.o). The underlying intuition is that the wider the quantificational domain, the stronger the assertion in a downward-entailing context will be. NPIs encode in their lexical entry the necessity of considering a maximally widenened domain. My proposal for the case at hand is that, although the lexical entry of the epistemic indefinite evokes only singleton alternatives in a minimally widened domain, the application of focus evokes further domain alternatives that undergo union with those in the restrictor of algún / algum: thanks to this operation, the domain from minimally widened becomes maximally widened. In current theorizing, focus is represented as an operator, which acts as a quantifier and receives its quantificational domain from the focused constituent through a process of generation of alternatives. The focus operator quantifies over a set of alternatives, comprising its associate and at least one alternative of the same type (Rooth ). Under the most general understanding, ‘Focus indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions’ (Krifka : ). That is, the presence of focus leads the conversation participants to assume a set of alternative meanings, which may (or may not) be constrained by the contextual situation. As alternatives may be exploited in various ways, and come in various guises, different subtypes of focus have to be distinguished (see Krifka  for an overview). Moreover, the set of alternatives may have different sizes (plausibly with a minimum of two alternatives): sometimes the set is limited and sometimes it is unconstrained, comprising all possible alternatives. Some linguistic elements (e.g. particles such as scalar even, additive also, exclusive only) are dependent on focus for their interpretation. A widespread way of treating focus-sensitive elements is assuming that association with focus is specified in their lexical entry as the condition that the restrictor be the set of alternatives to the ordinary semantic value of the focus associate.26 A natural question arising at this point concerns whether focus alternatives can also have a role for the restrictor of quantifying determiners (for this and other issues concerning the interaction of focus with quantifiers, see Hinterwimmer : –). In principle, it is possible to imagine that focus may interact, together with other contextual factors, in the determination of the contextual restriction for quantification.

26 See Hinterwimmer (: ) for alternative analyses.

OUP CORRECTED PROOF – FINAL, //, SPi

The semantic role of focus



Focus may then pragmatically exclude elements from the contextual information considered salient, cf. Hinterwimmer (: ). Another possibility that can be proposed to explain the effect of focus in the case at hand is that a certain type of focus forces the restrictor, i.e. the domain of quantification, to be extended beyond expected, normally salient boundaries to include peripheral cases, resulting thus in maximal domain widening. As mentioned above, I interpret the focus interacting with the indefinite in DPinternal inversion as scalar (Krifka ): with scalar focus alternatives are ordered along a scale; the focus denotation (‘even-some’) is the extreme of the scale; this is the kind of focus connected to the ‘even’ covert operator E in Chierchia .27 Emphatic NPIs correspond to Chierchia’s () ‘even-some’ type: their alternatives (σ (= scalar)-ALT in ()) represent a hierarchy of quantificational domains D ranked in terms of a probability measure μ (measuring ‘how likely it is to find something in D’, Chierchia : ): () from Chierchia (: ) even − some = some = λPλQ ∃x ∈ D [P(x) ∧ Q(x)] even − someσ −ALT = {someD : D  μ D} In Chierchia’s framework, all NPIs activate subdomain alternatives. Emphatic NPIs in addition involve a scalar component: the domain of the NPI must be the most likely (and a greater width of the domain correlates with a greater probability of finding something in it). The operator interacting with ‘even-some’ NPIs is the E operator, corresponding to the meaning of English even. One fundamental assumption underlying this treatment of NPI-licensing consists in the assumption of a covert counterpart of even. Chierchia’s treatment of even-exhaustification is based on Krifka’s () analysis, where Krifka proposes the existence of an Emph.Assert operator. Chierchia (: ) provides the following formalization for exhaustification of the alternatives by (covert) even:28 () even-exhaustification (Chierchia : ): EALT (p) = p ∧ ∀q ∈ ALT [p aliquis unus > *alicunus) and of its Romance continuations. Some other indefinites are first grammaticalized in the different varieties. In the course of time, these elements become organized into series, i.e., despite their heterogeneous etymological

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

sources, they assume a largely homogeneous syntactic behavior, at least with respect to their relation with the negative operator (e.g., Spanish nada ‘nothing’, originally positive res nata ‘born thing’, and ningun, originally negative nec unus ‘not (even) one’, cf. Willis et al. a: ).

. The Classical Latin negation system: a first overview .. A Double Negation system In this section I introduce the essential background on the Latin negation system, which will then be analyzed in depth in the further sections. I understand as ‘negation system’ the set of functional elements used in the language to encode the presence of a negative operator in the logical representation. Here I will mostly restrict myself to the analysis of the Latin negative marker of plain sentential negation (n¯on) and of the indefinites dependent on the negative operator, that is, negative indefinites such as nemo ‘no one’ and negative polarity items such as ullus ‘anyone’. In order to provide a comprehensive picture of the distribution of indefinites interacting with negation, it would also be necessary to consider other elements of the system, especially the coordinative negation n˘ec / n˘eque and the negative marker found in prohibitives, n¯e. However, for reasons of space, I will not treat n¯e here.6 I will instead extensively discuss the coordinative particle n˘ec / n˘eque in Chapter  since it is an important morphological component of the ‘new’ Romance n-words. Classical Latin displays the typical behavior of a Double Negation language: there is a one-to-one correspondence between overt expression of negation and presence of a semantic negation operator. Negation is marked either by the negative marker (NM) (cf. a) or by a negative indefinite (NI) (cf. b and c). () a. interiores plerique frumenta non serunt inlander:nom most:nom corn:acc not grow:pl ‘most of those living in the inland do not grow corn’ (Caes. BG ..) b. mentionem fecerat nemo mention:acc make:sg nobody:nom ‘Nobody had made mention of it’ (Cic. Verr. .) c. neminem reperies qui neget nobody:acc find:sg who:nom deny:sg ‘you will not find anyone who would deny it’ (Cic. Verr. ..) Only a double-negation reading is possible when a negative indefinite co-occurs with the negative marker (a and b); that is, there is no Negative Doubling (i.e., single-negation reading of the combination NM + indefinite), as instead is the case in Romance. 6 See Orlandini (a) for a thorough treatment of its semantics, and Moscati () for its syntax. Latin has further means to mophosyntactically realize negation in the left periphery, e.g. the complementizers nisi, ni ‘if not’, qu¯ominus ‘that not’, qu¯ın (originally ‘why not’, that is, an interrogative particle in independent clauses; later it also develops uses as a coordinator and as a subordinator after verba impediendi, cf. Fleck ).

OUP CORRECTED PROOF – FINAL, //, SPi

The Classical Latin negation system: a first overview



() a. aperte enim adulantem nemo non videt blatantly in.fact flattering:acc nobody:nom not see:sg ‘no one does not recognize someone who is blatantly flattering’ (Cic. Lael. ) b. Nemo ergo non miser nobody:nom then not wretched:nom ‘There is no one then who is not wretched’ (Cic. Tusc. .) Indefinites co-occurring with a negative marker or a negative indefinite must be NPIs (a and b): that is, no Negative Spread (Concord among negatively marked indefinites) is possible. () a. non ante tibi ullus placebit locus not before you:dat any:nom please:sg place:nom ‘Before that (otherwise) no place will please you’ (Sen. epist. .) b. quae non modo numquam nocet cuiquam, sed which not only never harm:sg anyone:dat, but contra semper addit aliquid on.the.contrary always add:sg something:acc ‘not only does [Justice] never cause anyone harm, but on the contrary it always adds some benefit’ (Cic. fin. .) Negative indefinites are also banned from the complement position of ‘without’ (), unlike, e.g., Romance n-words. () sine timore ullo without fear:abl any:abl ‘without any fear’ (Caes. BG .) They can occur as negative answers in elliptical structures: () a. Speaker A: num tu pudicae quoipiam insidias locas Q you:nom chaste:dat anyone:dat trap:acc set:sg aut quam pudicam esse oportet? Speaker B: Nemini. or who:acc chaste:acc be:inf ought:sg nobody: dat Speaker A: ‘Are you setting a trap for any chaste woman, or one who ought to be chaste?’ Speaker B: ‘Certainly not’ (lit.: ‘To no one’) (Plaut. Curc. –) b. quis spopondisse me dicit? Nemo. who:nom promise:inf me:acc say:sg nobody:nom ‘Who says that I promised? Nobody.’ (Cic. Q. Rosc. ) Latin negative indefinites, thus, always result in the insertion of a negative operator. In Classical Latin the ability for a NI to realize sentential negation is independent of the position of the NI before or after the finite verb: () a. Infl > NI Ratione utuntur: ludis poscunt neminem reason:abl use:pl game:abl ask:pl nobody:acc ‘They are reasonable: during the games they don’t demand from anyone’ (Plaut. Cas. )

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin b. NI > Infl iam neminem antepones Catoni now nobody:acc prefer:sg Cato:dat ‘you won’t prefer anyone to Cato’ (Cic. Brut. )

Examples of Double Negation such as (a) are quite rare and stylistically marked. They typically occur in the order NI > NM. The inverse order NM > NI is also attested: there are cases in which the NM is nonadjacent to the NI (a), but the typical examples occur in the form of adjacent combinations such as non nullus (further lexicalized as nonnullus), non nihil: in these latter cases n¯on seems to have the function of a constituent negation, cf. (b).7 () a. non tamen ideo neminem in provinciam mitti not yet therefore nobody:acc in province:acc send:inf.pass ‘and nonetheless governors still went out to governorships’ (Tac. Ann. .) b. quod itinerum meorum ratio te non nullam in because trip:gen my:gen reason:nom you:acc not no:acc in dubitationem videtur adducere (Cic. fam. ..) doubt:acc seem:sg lead:inf ‘The reason for my trips seems to cause you some doubts’ The fact that the co-occurrence of the negative marker and a negative indefinite yields a double-negation reading is exploited here rhetorically as a form of litotes (cf. further Orlandini : – for the semantic-pragmatic value of these constructions). .. Negative markers Latin has a number of negative particles. The marker used to express plain sentential negation is n¯on, which occurs with finite (indicative and some subjunctive) and nonfinite verb forms. It is also used as constituent negation and as a short answer particle. I dedicate to n¯on a separate section (§..), where I assess the status of the Latin negation system with respect to the phases of Jespersen’s Cycle. The distribution in Classical times sees n¯on as the unmarked negation with finite and nonfinite verb form. Some types of subjunctives are negated by n¯e (or its coordinating counterparts n¯eve, neu) (cf. Ernout and Thomas : §). Imperative forms cannot be negated in the Classical language, and a special periphrasis with the verb nolo ‘not-want’ is used (although cases of n¯e + imperative are documented in archaic texts). Frequently, sentential negation is expressed by the coordinative negative particles n˘eque, n˘ec ‘and not, nor’, positioned at the beginning of the coordinated clause or constituent; the preceding conjunct is often positive. Since this particle plays an important role in the formation of Romance indefinites, I will explore its distribution and semantics more thoroughly in chapter .

7 But see Orlandini (: –) for an alternative analysis.

OUP CORRECTED PROOF – FINAL, //, SPi

The Classical Latin negation system: a first overview



The form n˘eque originates quite transparently from the inherited negation n˘e < IE *ne—which is also a formative of the standard negative marker n¯on—and the enclitic coordinative particle -que. The particle n˘ec is a reduced form of n˘eque.8 The two markers n˘eque and n¯on may co-occur, in this order, in a sort of lexicalized litotes (‘and it is not true that not. . .’, Fruyt a: –, Fruyt : ) functionalized as anaphoric discourse marker ‘and moreover’, ‘and in addition to that’. This combination unambiguously show that the two markers do not compete for the same syntactic position. A further negative marker, haud, appears to lose productivity from the archaic to the Classical times, probably also owing to weakening of its phonological material (Kühner and Stegmann : II..–, b, Fruyt : –). In the Classical texts it is confined to an use as constituent negator and to fixed expressions like haud scio ‘I don’t know’ (for its interesting semantic-pragmatic properties see Magni et al. ). It is never found as an answer particle (Rijksbaron : , n. ). .. Negative indefinites The etymology of Latin negative indefinites is quite transparent: in archaic texts we still find independent use of their lexical bases.9 This fact suggests that the grammaticalization process involving Latin negative indefinites is a recent one; in §.. we will see that the same can be said for the plain sentential negative marker n¯on. The elements constituting the Latin negation system apparently evolved during a stage shortly preceding the first written attestations. The table in () shows that the first element of Latin negative indefinite pronouns is represented by the inherited negation n˘e < IE *ne. The grammaticalization path whereby a negative marker merges with a lexical base yielding an indefinite meaning is crosslinguistically frequent. Traditionally, the starting point of this grammaticalization process is analyzed as a type of constituent negation with scope over a (pro)nominal element that can be represented by an indefinite base, a function word (e.g. the cardinal ‘one’), or nouns semantically functioning as minimizers (e.g. ‘crumb’) or generalizers (e.g. ‘person’).10

8 Early Latin, however, also shows another, non-coordinative n˘ec, apparently acting as plain sentential negation, for instance in the XII Tables (and also appearing as affix in adjectival formations like necopinans ‘unaware’). The historical relation between the two forms is complex: while it is possible that the two particles share, in fact, the same etymology, it could also be that the -c in the second, archaic n˘ec is ´ and not the a reinforcer, possibly reflecting the same formative of demonstrative hic ‘this’ (that is, *ke coordinative *kw e, see de Vaan : s.v., Fruyt b: –). 9 For evidence that the etymology was transparent for Classical Latin scholars see Fruyt (b: –). 10 All Indo-European languages develop negatively marked indefinites independently of each other (cf. Lehmann : ch. .). Lehmann also remarks that the negative element of negatively marked indefinites in the various Indo-European languages is always a reflex of the sentential negative marker (typically *ne, but e.g., in Greek ou), and not of the negative derivational prefix *n; the (pro)nominal base differs from language to language. The (very tentative) hypothesis that seems to me most plausible in view of these observations is that proto-Indo-European may have lacked negatively marked indefinites. Instead, the sentential negative marker *ne would have combined with simple indefinites (possibly already specialized as negative-polarity items). Many Germanic languages show exactly this state of affairs at their first stage ◦

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() Latin negative indefinite pronouns nemo

nihil

neuter

nullus

‘no one’

‘nothing’

‘neither’

‘no’ (adjective)

< n˘e+*h˘emo = homo ‘man’

< n˘e+h¯ılum ‘minimal quantity’

< n˘e+˘uter ‘which of the two?’

< n˘e+ullus < *oinolos / unulus, dim. of ‘one’

Latin’s very rich system of indefinites sharply distinguishes between negative indefinites and negative polarity items (Molinelli , ; Orlandini a, Bertocchi et al. ). The main NPIs in direct-negation contexts are the pronoun quisquam and the determiner ullus: () Latin negative-polarity pronouns (selection) quisquam

ullus

‘anyone’

‘any’

< quis + adv. ‘how’ (?)

< *oinolos / unulus, dim. of ‘one’

The Latin negative pronoun for animates, n¯emo, features the merger of the negative morpheme n˘e and the noun h˘om¯o < *h˘em¯o ‘human being’, thus a generalizer acting as reinforcer of the negation.11 The forms of n¯emo typically occur pronominally; however they can also, more rarely, be used in apposition with a noun. The already very abstract meaning of the nominal component bleaches to a mere [+animate] grammatical feature. Fruyt (a: –) shows how this desematicization, and the ensuing loss of trasparency for the speaker, must have happened early, earlier than with nihil, since archaic texts alread show the appositive use of n¯emo with the noun homo, as in (). This use is attested also in Classical texts, cf. (). () Tune id dicere audes, quod nemo you:nom-ne this:acc say:inf dare:sg which:acc nobody:nom umquam homo antehac / vidit nec potest fieri ever man:nom before see:sg and.not can:sg become:inf ‘Do you dare to tell me a thing which no one has ever seen before and which is impossible?’ (Plaut. Amph. –) () Hominem esse arbitror neminem, qui nomen man:acc be:inf believe:sg nobody:acc who:nom name:acc istius audierit this:gen hear:sg ‘I believe that there exists no human being who has heard his name’ (Cic. Verr. .) of attestation: cf. e.g. Ingham (: –) for Gothic, van Kemenade () for Beowulf ’s English, Breitbarth () on Old Low and High German. 11 This base yields the nom., acc. (neminem), and dat. (nemini) forms, whereas in the gen. and abl. a pronominalized form of the adjective nullus is used. Similarly, the neuter pronoun n˘ıh¯˘ıl has only the direct nom.-acc. form; in the oblique cases an analytic form consisting of the noun res ‘thing’ modified by adjectival nullus occurs (e.g. abl. nulla re).

OUP CORRECTED PROOF – FINAL, //, SPi

The Classical Latin negation system: a first overview



In the Latin neuter pronominal form n˘ıh¯˘ıl ‘nothing’, the negative morpheme merges with a nominal element h¯ılum.12 The etymology of h¯ılum itself remains unclear (de Vaan : s.v.); according to Festus, it designated quod grano fabae adhaeret, i.e. ‘sprout of a bean’ (Fr. ‘hile du haricot’, Fruyt a: , Fruyt : ).13 It is still attested as autonomous noun in archaic texts, albeit already with an abstract meaning: ‘a minimal quantity’, ‘a smallest element’, cf. (). () a. aut aliquid prorsum de summa detrahere or something:acc absolutely from total:abl subtract:inf hilum minimum:acc ‘or detract at least a small something from the total’ (Lucr. .) b. filum, quod minimum est hilum: id thread:nom, because minimal:nom be:sg element:nom: it:nom enim minimum est in vestimento. namely minimum:nom be:sg in garment:abl ‘(one says) ‘filum’ (= thread) because it is the minimal element; and in fact the thread is the minimal part of a garment’ (Varro ling. ..)14 The archaic or archaizing texts (especially Lucretius) also show that, at an intermediate stage of grammaticalization, it frequently co-occurs with another element as reinforcer, e.g. with the indefinite aliquid in (a) or as argumental and later adverbial element with a negative marker, especially with the correlative negation n˘eque, n˘ec. Of the  examples of acc. h¯ılum that I retrieved in the LLT-A for the time span ‘Antiquitas’ (cf. §..),  show h¯ılum in a negative context. In some of these cases, h¯ılum behaves like an adverb after intransitive verbs, meaning ‘at all’, ‘not a bit’. Desemanticization in the course of the grammaticalization process led the lexical element to completely lose its concrete meaning and to just retain the grammatical feature [+ inanimate]. In turn, the adjectival determiner n¯ullus is formed by n˘e- and the diminutive (*-lo-) of the cardinal numeral u¯ nus ‘one’ (*oino-lo-s); the latter is also the source for the corresponding negative-polarity adjective ullus. The numeral ‘one’, being a scalar minimum, is a natural candidate for such grammaticalization process, as we will also see in the case of n¯on (§..) and in the grammaticalization of Romance n-words (chapter ). In addition to the nominal elements, there are a number of indefinite adverbial elements belonging to this class and formed in similar ways (e.g. nusquam ‘in no place’,

12 Also in this form, as in n¯on, we witness an early loss of final -um. The vowel shortening n˘ıh˘ıl attested in poetry, as well as the monosyllabic pronunciation that left a trace in the frequent graphic form nil must have originated quite early. See Fruyt (a: –) for a detailed discussion of the morphophonological properties of nihil and related grammaticalization outputs such as nihilum, nihilominus etc. 13 Isidore reports another etymology similarly coming from the agricultural lexicon, and attributes it to Varro (he notes in passing its minimizer value: Hilum autem Varro ait significare medullam eius ferulae quam Graeci α σφδε ον uocant; et sic dici apud nos nihilum quomodo apud Graecos ο υ δ γρ ‘Varro says that ‘hilum’ indicates the pith of that plant that the Greeks call asphodelon (asphodel); and we say ‘nihilum’ in the same way as the Greek say oude gru ‘not even a grunt’.’ (Isid. or. .). 14 Cf. the translation of nihil as ‘nicht ein Fäschen’ < Faser ‘fiber’ given by Kühner and Stegmann ().

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

‘in no way, not at all’; numquam ‘never’, with their corresponding negative-polarity versions usquam, umquam). Interestingly, in cases of correspondence between negative and negative-polarity pronominals or adverbials, the process of word formation shows that the polarity element is historically prior.15

. Negation systems: theoretical backgrounds Now that we have seen the inventory of Latin indefinites belonging to the negation system, we have to address the problem of how to analyze them, and to pave the way toward an explanation for why they largely were lost in the transition toward Romance. This section is dedicated to detailing the theoretical framework on which the categories preliminarily introduced in §... are based. I introduce the main analytical tools for the semantics (§..) and syntax (§..) of negation: inevitably, my discussion will include not only indefinites in direct-negation contexts, but also the negative marker, since the two categories are inextricably connected. In §. I will present my model to account for the fundamental difference between Double Negation and Negative Concord systems, on which I capitalize for my diachronic hypothesis. .. The interpretation of negation In what follows I focus on those interpretive aspects that are particularly relevant when annotating historical data, in order to distinguish standard direct-negation contexts from other cases. ... Sentential negation and constituent negation Sentential negation is commonly represented as the Boolean operator ¬ applying to a proposition, and reverting its truth value.16 The semantic criterion to define sentential negation is that the negative operator must take scope over the matrix predicate, that is, (at least) above the existential quantification on the event argument of the verb (Acquaviva ). This means that, in order to qualify as sentential, negation has to apply at least above v/VP, i.e., above the projection where the predicate and all its arguments have been merged. Since the existential quantifier binding the event argument is usually considered to be introduced by aspectual operators, negation has in fact to apply above the latter.17 Note that semantic considerations do not immediately bear on the locus of realization of the sentential negative marker, since the latter does not necessarily correspond to the locus of interpretation. That is, the classical distinction between 15 This suggests again, purely speculatively, that the polarity item might be the rest of a pre-historic system of negation, where indefinites are not able to negate by themselves. 16 In some frameworks it is thought of as a negative existential quantifier binding the event variable of the verb. This finds a parallel in the analysis of negative indefinites in terms of negative quantifiers, on which see §.... Here I will stick to a decompositional analysis of negative indefinites, which reaches the same result in the syntax, instead of assuming a complex negative entry. 17 The relationship with temporal operators is less well understood, and much depends on the theoretical model of tense adopted. See Penka (: –) for a survey of these complex issues.

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



so-called VP-negation languages and TP-negation languages does not concern the interpretation of negation, only its morphosyntactic ‘signaling’. Constituent negation, on the other hand, involves the negation of a constituent that does not have an impact on the polarity of the entire clause. Constituent negation has scope on constituents other than the predicative nucleus, and thus does not impact on clause polarity. Willis et al. (a: ) show that cases of constituent negation (a) are different from cases where sentential negation co-occurs with (contrastive) focus on some element of the clause (cf. (b), where the focus of negation can be placed variously on ‘burglar’, ‘window’, ‘cricket’, ‘bat’, verb phrase). () from Willis et al. (a: ) a. Not long after, I decided to go home (constituent neg.) b. The burglar did not break the window with the cricket bat (narrow focus) Narrow-focus negation is a form of sentential negation, since the event predicated in the clause ends up being negated: this negation still takes the entire proposition in its scope, but its focus is narrower than in the standard case (i.e., narrower than its scope). Constituent negation, instead, has no effect on the truth conditions of the (main) predication: () Willis (: ) a. Mary isn’t going to PARIS this weekend b. There are some pretty villages not far from here

(narrow focus) (constituent neg.)

Usually languages employ the same negative marker for both sentential and constituent negation; but there are some languages with special means for constituent negation, cf. nalles in Old High German, haud in Latin (cf. Willis et al. a: ). A further morphosyntactic difference has to do with Negative Concord: Jäger () for Old High German and Willis () for Old English show that, while narrowfocus negation exhibits NC with the sentential negative marker, constituent negation doesn’t (attesting to the fact that, owing to its semantic scope, it does not establish a syntactic relation with the negative projection related to the main verb). In Latin, n¯on marks both sentential and constituent negation. While examples like () are unambiguous cases of constituent negation, in other contexts it is difficult to decide. In () it is not clear if just the adverb valde ‘energically’ or also the predicate are negated. () a. cogor non numquam homines non optime de me force:sg.pass not never man:acc not highly from me meritos . . . defendere deserving:acc defend:inf ‘Quite often (lit. not never) I am forced to defend men not deserving highly from me’ (Cic. fam. ..) b. in illo sumptu non necessario in that:abl expense:abl not necessary:abl ‘in that unnecessary expense’ (Cic. fam. ..)

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() hos ego sermones . . . lacessivi numquam, sed non this:acc I:nom discussion:acc . . . allow:sg never but not valde repressi energically repress:sg ‘As for discussions of this kind, I never allowed them, but I also did not repress them energetically’ (Cic. fam. ..) We have seen that sentential negation must scope at least above the quantifier binding the event variable. Is there also an upper limit for its scope? It seems that this is the case: namely, sentential negation can never outscope operators hosted in the Force projection in the left periphery of the clause (cf. Han , Zeijlstra ). Also evidentials (probably, hopefully) are normally placed outside the scope of sentential negation. Quantificational expressions often exhibit clear preferences with respect to their scopal relations with negation. ... Negation and focus In the previous section we have already seen some examples of the interaction of negation with focus. It is well known that sentential negation is sensitive to the background–foreground partition of the sentence (Jacobs , , Partee , Herburger ). In its plain use, sentential negation does not apply to backgrounded or presupposed material, and targets only the informational focus. In this respect, negation functions, in a way, like a focus particle. Jacobs (, ) has highlighted the similarities between the German negative marker nicht and focus particles. Iatridou and Sichel (: ) have proposed that English negation can sometimes function as a focus particle (overriding in this way otherwise active constraints on scope-taking).18 Zanuttini () discusses raising to a focus projection FocP of NMs in some Northern Italian varieties. The relation with focus is very interesting for us in a diachronic perspective, especially in consideration of the systematic origin of negative markers as reinforcers of a pre-existing plain sentential negator in Jespersen’s Cycle (cf. §..). Emphatic reinforcers are obviously connected to forms of contrastive and emphatic (scalar) focus, as for example we saw with the Romance continuations of aliquis in §.. This has led some researchers to assume that the syntactic projection for negation may be connected to a Focus one. For instance, Biberauer and Roberts () interpret bleaching in Jespersen’s Cycle as the loss of a Focus shell narrowly associating with a constituent and precluding Concord. The loss of semantic reinforcement would be accompanied by a reanalysis of the original reinforcer as an element bearing formal features and entering an Agree relation with a negative operator. My proposal for the rise of Romance Negative Concord will capitalize on these insights: in chapter  I will investigate in detail the behavior of some focus-sensitive negative particles of Latin (n˘ec and ne . . . quidem), arguing that they have an important role in the development of the new Romance n-words. I will propose that

18 The case they discuss concerns Engl. must, which typically has to outscope negation, but is interpreted with narrow scope when not functions as a focus particle.

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



Romance n-words formed with the negative morpheme ne- / ni- < Lat. n˘ec originate as focus-sensitive items and derive their [uNeg] feature from the reanalysis of a focusinduced syntactic relation. ... Denials Negation can have a special use in denials, where the negation operator does not target the truth value of a proposition, but rather elements of meaning ‘above’ the propositional content, like presuppositions, implicatures, or even formal aspects of sentences. In (a) we have a plain sentential negation; in (b) negation targets, instead, the presupposition of existence encoded by the definite nominal phrase: () a. The present king of France is not bald. He has a mane of red hair. b. The present king of France is not bald, because there is no present king of France. In () negation targets scalar implicatures: it reacts to an (explicit or implicit) assertion, which in virtue of its scalar implicatures would yield a truth-conditionally correct, but pragmatically too weak statement. () from Geurts () a. The room wasn’t warm (implies that the room is less than warm) b. The room wasn’t warm; it was sweltering (overrides this inference) This kind of negation is, thus, denial of assertability. Horn () calls the plain sentential negation descriptive, and the negation targeting non-propositional aspects metalinguistic. Other authors (e.g. van der Sandt , Geurts , Repp ) use the term denial as a more general category comprising metalinguistic negation proper, as I will do here. In scholarship on Latin, uses of negation as denials go under the heading of négation polémique or dénégation (Orlandini a). More specifically, the type indicated in Orlandini’s work as négation polémique corresponds to Geurts’ () proposition denial. Although usually languages have the same negative marker for both types of negation, subtle morphosyntactic differences accompany the semantic-pragmatic distinction. According to Horn (: ; ), since negation here is operating on a different level, it cannot trigger certain phenomena correlated to ‘plain’ negation, such as, e.g., the licensing of NPIs. As Devine and Stephens (: ) put it, owing to the pragmatic difference in the import of denial with respect to plain negation, with denials ‘the minimizer component of the negative polarity indefinite is absent’. Distinguishing denials in the annotation was, therefore, necessary in order to better understand the behavior of indefinites (and the syntactic position of negation itself). An interesting case is shown in (): here aliquis occurs, despite the fact that it generally avoids negative contexts and it does not scope below negation (cf. §..):19

19 Bortolussi (: ), who discusses the example in (), remarks that this type is characterized by ‘le placement initial de la négation, ce qui peut s’analyser comme un positionnement syntaxique extérieur au domaine propositionnel’: ‘it is not true that’.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() Non aliquis socios rursus ad arma vocat non someone:nom ally:acc back to arm:acc call:sg ‘it is not true that there is someone calling his allies back to arms’ (Ov. rem. ) In addition to this type, a denial targeting an implicature may also be expressed by n¯on, positioned preferably sentence-initially, and allowing the occurrence of indefinites that otherwise mostly escape the scope of negation, such as aliquis, as in the following example from Bortolussi (: ): () non enim declamatorem aliquem de ludo, . . . sed not indeed declaimer:acc some:acc from school:abl . . . but doctissimum et perfectissimum quaerimus skilled:sup-acc and accomplished:sup-acc seek:pl ‘it is not a mere declaimer in a school that we seek, . . . but a scholarly and finished speaker’ (Cic. orat. ) Here a contrast is established between aliquis declamator, with aliquis bringing about a scale and denoting an indeterminate intermediate value, and the superlative doctissimus ‘an accomplished scholar’. We encounter here a rhetorical device that is very frequent in Latin literary texts: replacive negation (cf. Jacobs ), used to correct a previous statement or a contextually salient assumption: the element to be corrected and the element to be substituted are focused: () Peter is not [in Leipzig]Foc but [in Berlin]Foc

(Repp : )

Repp (: ch. ) interprets replacive negations in corrections as denials with a substitution part. The substituted part is typically introduced by corrective ‘but’, e.g. German sondern, Spanish sino, as opposed to contrastive ‘but’ like German aber, Spanish pero. Latin belongs to those languages, like English, that do not lexicalize the difference between corrective and contrastive coordination: sed and verum are used in both senses. The corrective use of ‘but’ is a classical test for metalinguistic negation (Horn ). In denials negation is ‘high’ in a semantic sense, in that it targets units above the propositional content. It has been argued that this may have syntactic reflexes (Alonso-Ovalle and Guerzoni , Repp , Martins ). Repp (: ch. ) analyzes the negation involved in denials as a ‘widest-scope’ negation on the speechact level. This negation, which she calls ‘illocutionary negation’, operates beyond the propositional level and also corresponds to a different, higher syntactic position for the negation, as she shows for English and German with tests related to stripping and gapping. In Repp’s framework, speech acts have a syntactic realization in form of illocutionary operators scoping on propositions. ASSERT and DENIAL, for instance, are two such operators. A negative sentence like () can thus be ambiguous in the following way: () Max isn’t tall a. ASSERT (¬ Max is tall) b. DENIAL (Max is tall)

(Repp : )

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



.. Types of negation systems Let us move to the syntactic foundation of my analysis: after a presentation of the feature-based typology that I follow (§...), I discuss in turn the negative marker (§...) and the indefinites in direct-negation contexts (§...). ... A feature-based typology Ample crosslinguistic research on negation systems has shown that the syntactic parameters of negation display a complex pattern of interdependencies, restricting the number of possible ‘types’ and, consequently, of possible ‘changes’. This has been represented in the parametric hierarchy format by Biberauer et al. (a), and in a parameter schemata format by Longobardi (). I base my proposal on Zeijlstra’s (, ) analysis and adopt a macroparametric approach, which sees the main source of variation and change in the structural properties of the negative marker, which are in turn related to the presence or absence of a functional projection for negation. I show how this has an effect on the variation observed among indefinite pronouns and determiners interacting with the negative marker. I follow the recent theoretical literature in assuming three basic types of negation systems (cf. Giannakidou  and §...):20 () a. Double Negation (DN): each negatively marked element conveys a semantic negation (Latin, German) b. Strict Negative Concord (strict NC): the negatively marked indefinites always co-occur with a sentential negative marker and convey only one semantic negation (Russian, Modern Greek, Romanian) c. Non-strict Negative Concord (non-strict NC): the negatively marked indefinites co-occur with a negative marker conveying one semantic negation if they linearly follow the inflected verb (Infl); however, if pre-Infl they are able to express sentential negation by themselves; in fact, if pre-Infl, they cannot co-occur with a negative marker in a single-negation reading (Italian, Spanish) Romance varieties exhibiting non-strict NC are particularly relevant for my discussion since they show an interesting asymmetry between two areas of the clause, separated by the Infl(ection) / T(ense) Phrase (see Ledgeway : –, Danckaert : ch.  for this category in Latin). The finite verb in Romance is realized in this position, and the standard preverbal negative markers derived from Latin n¯on occupy a position above it (NegP- in Zanuttini ). As mentioned, recent studies on Early Romance (Zanuttini , Parry , Poletto ) have confirmed that this has been the case since the beginning of attestation (cf. §..). Note that in non-strict NC systems the possibility of co-occurrence of the indefinite with a negative marker is dependent on the hierarchical relations between elements in their final landing sites, and not on the original position of Merge: both subject and 20 As will become clearer, the notion ‘negatively marked’ below should be understood as ‘endowed with a negative feature’, independently of the actual morphological makeup of the word (which sometimes contains no negative morphemes, cf. the French direct-negation series).

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

object indefinites cannot co-occur with the negative marker under a single-negation reading if pre-Infl (), and must co-occur with it if post-Infl (). () a. Nessuno studente ha consegnato il compito no student has handed.in the homework ‘No student handed in the homework’ b. NIENTE ha mangiato! nothing has eaten ‘S/he ate NOTHING!’ (capitals indicate emphatic focus)

(Italian)

() a. Non ha consegnato il compito nessuno studente not has handed.in the homework no student ‘No student handed in the homework’ b. Non ha mangiato niente not has eaten nothing ‘S/he did not eat anything’ Negatively marked indefinites belonging to NC systems, the so-called ‘n-words’, have, therefore, an ambiguous status, since the ability to negate by themselves depends on their position.21 This fact has been captured in various ways. In the table in () I present the system proposed by Zeijlstra (, ), which assumes a varying featural composition for the different elements in negation systems. Zeijlstra’s model is based on a threefold system of features: a semantic negative feature [Neg], an interpretable formal feature [iNeg] and an uninterpretable formal feature [uNeg].22 () Features in Zeijlstra’s () system: Type

Negative marker

Indefinites

Double Negation

[Neg]

[Neg] (Neg. Indef.)

Non-strict Negative Concord

[iNeg]

[uNeg] (n-word)

Strict Negative Concord

[uNeg]

[uNeg] (n-word)

Two assumptions—that we will more thoroughly review in the following sections— are crucial to this system: (i) the LF-scope position of the semantic negative operator Op¬ does not necessarily coincide with the surface position of the morphosyntactic sentential negation marker, and (ii) Concord is real syntactic agreement (cf. a.o. Watanabe , Zeijlstra , , , Penka ); the negation operator has to 21 This applies also to strict NC systems, where n-words negate by themselves in negative short answers (an option precluded to NPIs). 22 A further, rarer system, which is expected given Zeijlstra’s typology, is attested in the variety of Afrikaans investigated by Biberauer and Zeijlstra (), where indefinites are [iNeg] (and thus do not undergo Negative Spread) and the NM is [uNeg] (and can thus be stacked). Also Old Low German as analyzed by Breitbarth () has [iNeg] indefinites and a [uNeg] negative marker. In Zeijlstra’s system, the further expected type, where all elements are [iNeg], would be undistinguishable from the Double Negation system, where all elements are [Neg]. In the absence of redundant marking, the speaker has no evidence for a formal (interpretable) feature and opts for a semantic feature.

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



c-command the indefinite. Indefinite n-words can undergo (Multiple) Agree, unlike negative indefinites of DN languages, which are endowed with a semantic feature [Neg] that does not trigger the establishment of a syntactic relation. In this system, ‘real’ negative indefinites of DN systems are distinguished from n-words of NC systems in that the former are intrinsically semantically negative. In DN systems the impossibility of co-occurrence of negatively marked elements in a single-negation reading is captured by assigning to all elements a self-licensed semantic feature [Neg] that does not trigger (licensing-related) syntactic phenomena after its insertion. In NC systems, instead, n-words may be marked for negation only formally, but their negative feature is uninterpretable ([uNeg]) and needs to enter a syntactic relation Agree with an interpretable [iNeg] feature. The latter may be overtly realized by the negative marker or come from a covert operator.23 The first option is observed in non-strict NC systems when the n-word is in post-Infl position. In all the other cases (with pre-Infl n-words in non-strict NC systems, and unambiguously in strict ones) the licensor is a covert operator. That is, the asymmetry between the pre- and the post-Infl positions in nonstrict NC systems is accounted for by assuming that, if the n-word is below (i.e., ccommanded by) the position NegP where the [iNeg] feature sits (on which see further §...), the [iNeg] feature can and must be realized by the negative marker (adverb or particle), entering an Agree relation with the [uNeg] on the n-word. If, instead, for some independent reasons (raising to subject position, information structural requirements) the n-word is above the position where [iNeg] sits, i.e. in the pre-Infl area, it suffices by itself to convey sentential negation. To explain this behavior, Zeijlstra (, ) postulates a ‘last resort’ insertion of a covert negative operator carrying the [iNeg] feature above the landing site of the n-word (to delete its [uNeg] feature). I find explanations in terms of (phase-based) locality considerations (Biberauer and Roberts , Longobardi : ) more attractive and elaborate on them in §., capitalizing on the intuition that ‘being in the same (CP) phase’, i.e., pre-Infl, must be a sufficient condition for the n-word to be licensed in non-strict systems. Redundancy in the morphosyntactic marking of negation is a crosslinguistically relevant feature. As Weiß (b: ) remarks, ‘there seems to be a strong necessity to ‘close off ’ or ‘encapsulate’ weak indefinites when occurring in the scope of clausal negation’, i.e., to overtly signal the narrow scope of the existential quantification they introduce. ... The syntactic status of negative markers Zanuttini () proposes the existence of four different clausal positions where languages may realize a negative marker.24 The highest of these is the NegP- projection, where the negative marker 23 The presence of an abstract negative operator had already been proposed by Laka (), Ladusaw (). 24 Zanuttini (: ) remarks that her cartography of negative positions must not necessarily lead to to the assumption of four different Neg projections: she suggests that NegP- is possibly a general polarity head (cf. SigmaP in Laka ), and that the lower, post-Infl projections may be independent contentful

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

for plain sentential negation is realized in the Standard Romance languages.25 The position of the NegP- projection is shown in (): () The NegP- projection (adapted from Zanuttini : , ): CP NegP-1 AgrP TP1 ... NegP- is structurally higher than the position for the inflected verb (Infl, corresponding to TP in ()); it is also structurally higher than the position for complement clitics and for agreement subject clitics (AgrP). The other three positions are lower than Infl. Zanuttini’s cartography covers up to the TP domain. But languages expressing sentential negation at the higher, CP-level are well known: Celtic is a particularly well-studied branch in this respect (McCloskey ,  on Irish). As mentioned, Latin shows a modality-sensitive CP-level negation with n¯e and other negative complementizers. It has been proposed that not all languages activate a functional projection NegP: for instance, according to Zeijlstra (, , ) a.o., since DN languages lack formal features for negation [i/uNeg], they do not grammaticalize a Neg projection (as there would be no sufficient acquisitional evidence to acquire it). In Zeijlstra the presence or absence of NegP is a consequence of Bare Phrase Structure, according to which structure is a reflex of syntactic dependencies; thus, a functional projection is postulated during language learning only if there is evidence for the presence of formal uninterpretable features in the derivation (in the form of agreement or dislocation).26 In fact, it has been frequently proposed to analyze at least the negative markers of some languages (e.g. German) as syncategorematic expressions, that can adjoin to any verbal projection (Jacobs , Jacobs , Bayer ), or, with nonsentential

projections that serve as adjunction sites for the (clitic / non-clitic) negative markers. A further elaboration of this model is represented by Poletto (), who proposes to derive the various positions for the negative markers (as well as doubling phenomena) from the movement of subconstituents of a layered NegP to various landing sites in the clause. 25 The dialects show rich variation in this respect (see again Zanuttini  for Italo-Romance). 26 For a recent Neg-less approach see also Breitbarth (). For Romance Negative Concord varieties a notable Neg-less approach is proposed by Manzini and Savoia (), who analyze markers in NegP- as nominal clitics, licensed in the argument structure of the verb. Their analysis extends to negative markers in Zanuttini’s lower NegP projections, which are considered to be quantificational elements with a nominal nature.

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



scope, to any targeted category. Also in this they would be similar to focus particles (cf. §...). Negative markers differ crosslinguistically not only in their featural specification, but also in their phrase-structural status: a long-standing observation is that the ‘smaller’ a NM is, the more prone it is to be ‘redundantly doubled’ by other negatively marked elements in the clause. This has been variously understood: as the need to reinforce a phonologically weak element (Jespersen’s  original motivation for the renewal cycle affecting negative markers), as reflecting the level of morphosyntactic integration in the verbal projections (Jacobs ), more recently as an indication of the amount of internal structure of the negative marker (Breitbarth ). An important dimension distinguishing negative systems crosslinguistically has to do with whether a negative marker has the status of a head (X ) or of a phrase (XP). If phrasal, the negative marker may either occupy the Specifier of NegP, or, as adjunct, the outer Specifier of another category. If, instead, the NM is a head, the only possibility is that it projects its own functional projection NegP. Within the generative syntactic tradition, the head status of the negative marker is explicitly related to the occurrence of NC in Haegeman and Zanuttini (), Déprez (), Rowlett () a.o. Concerning the featural specification of negative markers, a number of authors follow the assumption that there is a threefold distinction between [Neg], [iNeg], and [uNeg] NMs. In order to derive the difference between strict and non-strict NC, both Zeijlstra () and Penka () have to stipulate that the sentential negative marker carries [iNeg] in non-strict systems and [uNeg] in strict ones. That is, all negatively marked elements of a strict NC language are, in fact, only formally negative, and they just signal the need for the insertion of a covert negative operator. Negative markers carrying [uNeg] of strict NC languages behave like n-words, appearing in Concord with a covert negative operator. ... Indefinites interacting with negation My aim in this section is to spell out the criteria according to which we distinguish between negative indefinites and n-words, since this will be relevant for the discussion of the Classical and Late Latin data. In chapter  I will come back to this topic and further discuss the problem of the demarcation between n-words and NPIs. The hybrid status of n-words (e.g. It. nessuno, Sp. nadie, Fr. personne, Rom. nimeni) is a perduring puzzle: the label itself under which they are known since Laka () attests to our limited understanding, since it is motivated just by the initial sound appearing in many (but not all!) of them in Romance. The main issue posed by n-words concerns their variable negative import. The clearest case in this respect is represented by non-strict NC systems, such as Italian or Spanish. While the (post-Infl) n-word nessuno is unable to negate by itself in (a) and (b) and must co-occur with the negative marker without yielding an additional semantic negation, it has negative import in (c), where it is pre-Infl, and in (d), where it is found in isolation as negative short answer. Also in Negative Spread structures, as (e), n-words appear to negate a sentence by themselves, provided at least one n-word is in the pre-Infl field.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() Italian a. Non è venuto nessuno ‘Nobody came’ b. * È venuto nessuno c. Nessuno è venuto ‘Nobody came’ d. Speaker A: Chi è venuto? Speaker B: Nessuno ‘Speaker A: Who came? Speaker B: Nobody’ e. Nessuno ha mangiato niente ‘Nobody ate anything’ In strict NC systems there is no pre-/post-Infl asymmetry and the n-word always co-occurs with the negative marker; still, also in these system n-words can have negative import when they are used as negative short answers. Zeijlstra (, ) proposes a solution to this puzzle in terms of formal features; he proposes that in NC languages, n-words do not introduce a negative operator: they carry an [uNeg] feature, which has to be eliminated during the derivation by entering a syntactic relation with an [iNeg] negative operator inserted elsewhere in the clause. In this system, thus, Concord involves Agree. As mentioned, pre-/post-Infl asymmetries are explained by Zeijlstra () by assuming that in non-strict systems, when the [uNeg] indefinite is syntactically above NegP, the insertion of a covert negative operator licenses it. Similarly, in Negative Spread structures, all n-words (pre- and post-Infl) are licensed by a covert negative operator, whose insertion is triggered by the n-word in pre-Infl position. A covert operator is responsible for the licensing of all negative elements (also the [uNeg] negative marker) in strict systems. In non-strict systems, the overt [iNeg] negative marker licenses the post-Infl n-words. Zeijlstra’s system crucially involves a form of Reverse Agree: the lower element in the Agree configuration is the one triggering the relation, since it carries uninterpretable features that have to be eliminated before LF (downward valuation). Independently of the precise formulation, the fundamental intuition behind this system is that the lower element, and not the higher one, is ‘deficient’ in some sense. This system is in line with our semantic understanding of certain pronominal elements as deficient expressions, in need of a suitable licensor. Moreover, it allows for the establishment of a complex relation between one licensor and multiple dependents (Multiple Agree).27 With respect to negation, the (potentially abstract) negative operator is the only interpretable element of the dependence, and it is always merged in a syntactic position from which it c-commands its dependent elements. Also the status of negative indefinites in DN languages is particularly debated. Two main groups of analyses have emerged: according to one, negative indefinites are negative quantifiers, i.e. a negative operator is ‘built in’ in their lexical entry (a.o. Haegeman and Zanuttini , Haegeman ); according to another line of research, negative indefinites are elements bringing about existential quantification (as quantifiers or

27 For an alternative proposal see Haegeman and Lohndal ().

OUP CORRECTED PROOF – FINAL, //, SPi

Negation systems: theoretical backgrounds



as variables under existential closure) and associating with a sentence-level negative operator (Zeijlstra , Penka ). A very important piece of evidence in the debate is represented by so-called split-scope readings: as in NC-languages, the negation and the indefinite meaning component of negative indefinites can take scope independently of each other (cf. Jacobs , Penka and von Stechow , Abels and Martí , Penka , Zeijlstra ). () Du musst keine Krawatte anziehen you must no tie wear a. It is not required that you wear a tie b. There is no tie that you are required to wear c. It is required that you do not wear a tie

(German) ¬>2>∃ ¬>∃>2 2>¬>∃

Since Jacobs’ seminal analysis (Jacobs , ), in order to account for these data, negative indefinites have been argued to decompose into two elements: a negative operator and an indefinite, spelled out as a single unit but able to take scope independently . Zeijlstra () proposes that the lexical entry of negative indefinites is syntactically complex and can be decomposed into two elements: a negative operator and an indefinite, spelled out as a single unit but able to take scope independently of each other. This complex indefinite can undergo Quantifier Raising, but the indefinite part may be reconstructed in a structurally lower position, giving rise to split-scope readings. Crucially, as we saw, according to Zeijlstra in DN languages all elements are [Neg]: in these languages Neg is not a formal, only a semantic feature, which does not trigger further Agree processes in the syntax.28 The ‘intrinsic negativity’ of negative indefinites, which allows them to be ‘selflicensed’,29 in contrast to n-words, has often be traced back to their etymological origin. For instance, Haspelmath () concludes, on the basis of the WALS survey, that ‘in the great majority of languages where negative indefinites preclude predicate negation, the indefinites themselves contain a negation marker’. He finds exceptions to this generalization only in Western Europe: ‘Exceptions to this generalization are found exclusively in western European languages (e.g. German kein ‘no’, Swedish ingen ‘no(body)’, and several Romance expressions which preclude verbal negation only when occurring preverbally, such as Spanish nada ‘nothing’, Italian mai ‘never’). The probable diachronic explanation for this, according to Haspelmath, is that the pattern 28 Penka () makes a different proposal: she argues that negative indefinites of DN languages are similar to n-words of NC languages in being [uNeg]; however, differently from n-words of non-strict NC languages, they can be licensed only by a covert negative operator, i.e., they are [uNeg ∅]. This system, while having the advantage of providing a more homogeneous format for all indefinites involved in negation systems, does not allow us to derive in a straightforward way the presence of NC from the presence of uninterpretable formal features on the indefinites, which is a fundamental aspect of my diachronic analysis. However, see §.. for an important connection emerging between my analysis of NC and Penka’s analysis of DN systems. 29 The notion of self-licensing originates in Ladusaw ().

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

where negative indefinites preclude predicate negation typically arises by ‘attraction of the negator to the negative indefinite pronoun’ (Haspelmath : section ). We will see in chapter  that it is difficult to make sense of this generalization without considering the entire system of negation, i.e. also the syntax of the negative marker. Romance languages are very instructive in this respect: on the one hand, some n-words do incorporate a negation (e.g. It. nessuno < nec ipsum unum) and yet they do not always suffice to express sentential negation (and they also have non-negative readings); on the other hand, some n-words that do not etymologically incorporate a negative element can negate in some contexts (cf. Haspelmath’s examples above) or have even become negative indefinites in changing systems like contemporary French. In chapter  we will deal with some of these phenomena again.

. The analysis adopted here In this section I further elaborate on the distinction between negative indefinites (NIs in what follows) and n-words: I propose a system to derive their differential behavior, in a principled way, from the absence of formal syntactic features in Double Negation systems. I argue that this is connected to the timing of the operations involved at the syntax-semantics interface, more precisely to the fact that the existential quantification that a negative indefinite introduces has to be closed off immediately by the negative operator. I model this as a type of selection, which derives the strict locality of the operation, since selectional properties require immediate satisfaction.30 .. Negative indefinites I assume that both NIs and n-words are indefinites that lexically encode the requirement to be interpreted within the scope of negation; NIs impose the requirement that the licensing relation be established as soon as possible by the negative operator, whereas the licensing required by an n-word can wait. This is a way of implementing the notion of ‘semantically negative indefinites’ needed for DN languages, together with their well-known requirement of PF-adjacency between the negative operator and the existentially quantified variable (Jacobs , Penka : ch. , Zeijlstra ). With NIs, the structure building operation of merging a negative operator with the indefinite has to happen immediately, as soon as the existential quantification is introduced. This also derives the systematic absence of ambiguity between negative and non-negative readings observed with NIs: unlike n-words, NIs of DN languages never show NPI uses (where they are not associated with a negative interpretation). They combine exclusively with an immediately local negative operator and, thus, cannot be captured by other semantically compatible licensors.

30 A preliminary version of this analysis has appeared in Gianollo (a).

OUP CORRECTED PROOF – FINAL, //, SPi

The analysis adopted here



The earliness of the combination of the indefinite and the negative operator is due to the absence of (uninterpretable) formal features in NI, in the spirit of Zeijlstra () and (), and follows without further stipulations from the nature of syntactic computation: since, differently than with n-words, there is no need to wait for the insertion of the appropriate [iNeg] counterpart, the syntactic operation happens as soon as possible (cf. Collins  for earliness ‘ASAP’ economy principles in syntax). The syntactic relation between the negative operator and the indefinite part in NIs may be understood as a form of selection. In Wurmbrand’s () system, selection is modeled in terms of valued / unvalued features, and thus reduces to the general conditions for Merge. The selector has a valued feature, while the complement is unvalued for that feature: by means of (Reverse) Agree the feature on the selected complement becomes valued. The interpretable / uninterpretable distinction represents an independent dimension. We could thus say that the interpretable / uninterpretable dimension is extraneous to the features of NIs: in this sense NI do not have ‘formal’ features. They do, however, involve a pair, ‘unvalued / valued’, that determines selection and has to be created immediately. More specifically, for indefinites we may assume that the DP containing the indefinite carries an unvalued (semantic) negative feature [Neg: ], and the Neg element selecting it carries the valued counterpart [Neg: val ].31 Note that we do not have a NegP in the usual sense: we do not have a clausal projection entering a longer-distance dependency, possibly with multiple elements; in other words, we do not have an [i/uNeg-uNeg] pair. Rather, we have an unmediated selection dependency. The negative operator is adjoined to the nominal projection (possibly to the completed DP phase, also containing operator / quantificational elements), as in the cases of constituent negation. The projection resulting from the Merge operation with the negative operator is a nominal projection for all the ensuing operations required in the derivation.32 Further LF-movement of negation takes place if negation is to take sentential scope, in principle independently of the existential quantification. The creation of the structure in the syntax (and not in the lexicon) ensures, as in Zeijlstra (), that this latter operation does not run against Lexical Integrity, as a (corresponding) negative-quantifier approach would do in the cases of split scope. In (), the absence of formal [i/uNeg] features is represented by the semantic feature [Neg] next to the nominal DP projection.

31 Semantically, the selectional requirement of NIs for a negative operator could be modeled as a presupposition or a conventional implicature (the latter solution is proposed by Alonso-Ovalle and Guerzoni  for Romance n-words). 32 Cf. Biberauer et al (b: ) for the general fact that negative markers are invisible for c-selection / subcategorization. This seems to be a valid conclusion also for focus and topic markers.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() Negative indefinite DP [Neg]

Op¬ [Neg: val]

DP ∃ [Neg: —]

NP x

.. n-words With n-words, the presence of [uNeg] in the lexical entry represents the constraint that forces the computation to wait for the appropriate licensor, i.e., the [iNeg] operator in the CP–TP phase. This captures the idea that n-words are ‘dependent indefinites’ (Giannakidou ). Possibly a more complex feature structure than the one I provide here accounts for the possibility to license n-words by means of other operators creating (Strawson) downward-entailing contexts, in case they c-command the indefinite, i.e., for the NPI uses. The presence of formal uninterpretable features on the n-word, forcing to a syntactic dependence, would thus be indirectly responsible for the fact that n-words of the Romance languages—unlike NIs of, e.g., Germanic languages—often show non-negative NPI uses.33 In chapter  I will formulate a precise proposal in this respect. For now, let us consider just the [uNeg] feature involved in direct-negation contexts. The proposed structure for n-words is shown in (): () n-word DP[uNeg]

∃ [uNeg]

NP x

In this system the difference in featural content between negative indefinites and n-words is reflected in the timing of syntactic operations affecting the two classes. With respect to NIs, consequently, n-words lack the superordinate structure resulting from the adjunction of the semantic negative operator in (), since the appropriate [iNeg] licensor will necessarily be located in a NegP projection.

33 Martins () shows that there is variation within Modern Romance with respect to the possibility of n-words in non-negative NPI uses: these uses are excluded in strict NC varieties such as Romanian and Venetian dialects (Martins : –) and in Portuguese, a non-strict NC language (Martins : –). In Slavic languages, n-words are restricted to the direct-negation function and do not occur in non-negative NPI uses, a phenomenon that has come to be known as the ‘bagel problem’ (Pereltsvaig ). Various proposals have tried to explain it in terms of morphosyntactic blocking, cf. §....

OUP CORRECTED PROOF – FINAL, //, SPi

The analysis adopted here



For well-known cases of Negative Concord, such as those seen in Romance and Slavic, there is ample evidence that the licensing of the indefinite may take place across phases, the indefinite being possibly in the vP phase (e.g., when it is an object) and the NegP projection being connected to Infl in the CP–TP phase. In Germanic, instead, negative markers are inserted lower, adjoining the predicative layer vP. In Zeijlstra’s system, no functional projection NegP is projected in this case, since the negative marker is [Neg]. A [iNeg] counterpart of the ‘low’ [Neg] negation of Germanic DN languages seems to correspond to the Negative Concord system of Bavarian (), as described in Weiß (c). () Gesdan hod neamd ned angrufa yesterday has nobody not called ‘no one called yesterday’

Bavarian (Weiß c)

The parametrization of the Spell-Out position of the sentential negative marker is anyway in principle independent of its locus of interpretation, which I take to be universally invariant. Once a NegP projection is activated, it has to be identified, i.e., overtly realized by an element in its head (NM) or in its specifier (NM or n-word). Now the question arises of how the object remains visible for Agree after completion of the vP phase, if we assume conditions such as the Phase Impenetrability Condition and the intervention effect of phase heads to affect Agree. The issue is in fact common to all n-words remaining in the vP phase, i.e., also low subjects. The valuation requirement of [uNeg] features per se does not seem to lead to obligatory displacement of argument n-words in Romance: if they move to the CP–TP phase, they do it based on independent requirements.34 This follows from the fact that two strategies are available for realizing NegP: either the n-word (a) or the NM (b): () Italian a. Nessuno ha telefonato nobody has called b. Non ha telefonato nessuno not has called nobody ‘No one called’ The [uNeg] feature of the n-word inserted in the argumental domain must remain visible at the vP phase edge. One way to implement this is to take inspiration from Poletto () in assuming that there is a parallelism between the mechanism seen with clitic doubling and the mechanism of Negative Concord. Poletto proposes that, in the same way as clitics originate as part of a ‘big DP’ structure and then raise (Belletti , Cecchetto ), the negative markers of Romance languages originate in a complex ‘big NegP’, whose internal structure is semantically motivated, and then rise to various clausal positions. My proposal is that, similarly, we can think of n-words as ‘big DPs’ where the ‘doubling’ NM is generated in an outer specifier, in

34 For Bavarian, instead, Weiß (c) has proposed obligatory movement to Spec,NegP motivated by checking requirements. I will propose a similar mechanism for Late Latin.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

case uninterpretable formal features need to be passed up to the next phase.35 The NM carries the [uNeg] feature to the edge of the vP phase and from there to the NegP in the CP–TP phase, where the [iNeg] Operator is hosted, yielding Concord. () n-word as ‘big DP’ DP [uNeg]

DP

XP [uNeg] non

∃ nessuno [uNeg]

NP

If the n-word itself is at the edge of vP and then raises for independent (EPP, information-structural) requirements, landing in a position which c-commands NegP, no ‘doubling’ is generated. This is accounted for by Zeijlstra (, ) by assuming that, when the [uNeg] indefinite is syntactically above NegP, the insertion of a covert negative operator c-commanding it performs licensing as a Last Resort measure. I follow instead Longobardi (: ) for the idea that, under certain conditions, being in the CP phase for independent checking reasons suffices for valuation of the negative feature on the indefinite as well (through intermediate movement steps, in my implementation). The n-word carries the [uNeg] feature to the CP–TP phase, where it enters an Agree relation with the [iNeg] Operator in NegP, with no need for doubling. I assume, thus, that the abstract negative operator is inserted in NegP, i.e. in the same position as the overt negative marker (cf. Laka , Ladusaw ). The n-word can rise further to the split CP, extending the scope of the negative operator, and the existential component reconstructs in the scope of the negative operator. The analysis is exemplified below with the examples seen in (a) and (b), repeated in (): () Italian a. Nessuno ha telefonato nobody has called [CP [NegP Op¬[iNeg] nessunoi, [uNeg] [InflP ha [vP telefonato ti ]]]] b. Non ha telefonato nessuno not has called nobody [CP [NegP Op¬[iNeg] noni, [uNeg] [InflP ha [vP telefonato ti nessuno]]]] Note that according to this formulation the NegP position in Romance languages would not induce ‘criterial freezing’ (Rizzi ): the element establishing Agree with the operator is not ‘frozen in place’ after the operation and can move further. 35 In the structure I assume, the doubling clitic is generated as a maximal projection in the specifier of the doubled DP; cf. Cardinaletti (: –) for arguments in favor of this structure.

OUP CORRECTED PROOF – FINAL, //, SPi

The analysis adopted here



Support for the conclusion that [uNeg] elements can land higher than NegP comes from cases like (), discussed by Zanuttini (), where mai ‘never’, the adverbial n-word conveying negation, precedes the subject and is thus uncontroversially higher than NegP: () Italian (Zanuttini : , n. ) Mai sua madre gli avrebbe permesso di comportarsi così never his mother to.him would.have allowed to behave so ‘Never would his mother have allowed him to behave that way!’ A solution reconciling these facts with Rizzi’s criterial approach could come from considering the NegP projection syncretic with other, higher CP-related projections which do induce criterial freezing, as, e.g., the Focus projection. I will come back to this point in §.. and further in chapter . As for the negative marker inserted (as non-doubled marker of sentential negation) if no n-word is present, I remain agnostic as to whether a Move approach like Poletto’s () is preferable to a more traditional external Merge solution: for the time being I will assume that the negative marker is externally merged as the head of NegP in languages such as Italian, Spanish, Portuguese. We must anyway assume that the negative marker undergoes a cliticization process, since it moves together with the verb (and the pronominal clitics that may attach to the latter) when required. Strict Negative Concord, where the negative marker is present independently of the pre-/post-Infl position of the n-word, would represent a type where the negative marker in NegP is an obligatory expletive, constituting part of the inflectional complex of the verb (cf. Zeijlstra  for evidence concerning verb movement). As such, it may be argued to originate within the verbal projections and to be dependent on the verb for all syntactic operations. Agree with the [uNeg] feature of n-words would be parasitic on the remaining Agree operations that the verb entertains. The proposed system maintains a fundamental assumption of Zeijlstra’s analysis, namely that functional projections are activated only if there is acquisitional evidence for formal features participating in syntactic processes (agreement or displacement). The main difference between DN and NC systems remains the fact that this evidence is absent in the former type and present in the latter. .. The negative marker In concluding this section let me point out an issue that arises under my analysis, and sketch a possible way out, which may disclose a new perspective for looking at the difference between strict and non-strict NC systems. Owing to the limitations of this research, my discussion will remain highly abstract at this point. However, in chapter  we will come back to some of the themes discussed here in analyzing concrete examples from Latin and Romance. My account of non-strict NC systems forces an ambiguous representation for the NM: if the NM is the only marker of sentential negation it is analyzed as [iNeg] as in Zeijlstra’s account; but if it co-occurs with a lower n-word it is the carrier of [uNeg] features (being a clitic element that originates in the n-word itself). An ambiguity

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

account is obviously unwelcome on the grounds of learnability issues and purely theoretical considerations, so let us see if a way out is possible. Assuming that the NM originating in the ‘big DP’ structure in () is a carrier of [iNeg] features like the externally merged version is an option that would require too many adhoc assumptions. Its movement to NegP in the CP–TP phase would have to be motivated as a purely semantic scope requirement. Moreover, we would expect that in some cases, where no sentential scope for negation is required and the n-word remains in the vP phase, we should actually see the string [NM n-word] (e.g., It. non nessuno) in the vP phase, which is never the case.36 The other possibility is to assume that the NM is always a carrier of [uNeg], also in the externally merged version. This would amount to positing that the interpretable operator is always abstract, and that all the elements of Negative Concord languages are invariably [uNeg]. I find this option more viable and I outline here the main consequences this theoretical move would have for the system. In Zeijlstra’s (, ) model the sole purpose of assuming a difference between [iNeg] and [uNeg] NMs is to account for the different behavior of strict and non-strict NC languages in the pre-Infl (= pre-NegP) area. In non-strict NC languages Concord is blocked in the pre-Infl area because the Op¬ [iNeg] inserted in NegP is realized by the NM. If an n-word appears in pre-Infl position it will necessarily be higher than NegP and the only way to license it will be through the Last Resort insertion of an abstract negative operator c-commanding it. This is the reason why the co-occurrence of an n-word and the NM in pre-Infl position will necessarily yield a double-negation reading: two [iNeg] operators will be present: () Nessuno non ha telefonato nobody not has called ‘Nobody did not call’ = everyone called

Italian

In strict NC languages, according to Zeijlstra’s analysis, since also the NM is [uNeg], only the Last Resort option is possible: an abstract Op¬ [iNeg] is inserted above the highest negatively marked element and licenses all instances of the [uNeg] feature via Multiple Agree. Now, if it were possible to derive the strict versus non-strict nature of Negative Concord from some independent characteristic of the language, the difference between [iNeg] and [uNeg] NMs could be dispensed with. Let us assume that, in both types of languages, having an active NegP means having to realize it overtly: the [uNeg] marker would be an expletive, morphosyntactically signaling the presence of a semantically active element (the abstract Op¬ [iNeg] with sentential scope) and entering an Agree relation with it. Despite being uninterpretable, the [uNeg] marker is obligatory because it (i) establishes a relation with a lower n-word, marking the sentential scope of negation, or (ii) is the only overt marker of negation (if no n-words co-occur).

36 For the marginal cases of n-words taking scope at the vP level see Herburger (), Zeijlstra (), Biberauer and Zeijlstra (). I discuss some examples in §...

OUP CORRECTED PROOF – FINAL, //, SPi

The analysis adopted here



The difference between strict and non-strict NC languages in the pre-NegP area would then arise from the fact that n-words of non-strict systems can be means of overt realization for NegP (i.e., they represent a further strategy besides Merge of an NM, either externally or internally), while n-words of strict systems cannot. Why would it be so? It cannot depend on the feature content of n-words, which I assumed to uniformly be [uNeg]. The tentative answer I put forward is that this could be due to the independent fact that in non-strict NC languages there can be syncretism between the NegP position and the FocP position. Syncretic syntactic projections spell out contiguous functional categories (Giorgi and Pianesi : ch. ). Consequently, n-words in the CP–TP phase of, e.g., Italian and Spanish realize at the same time and in a single configuration negative and focus features. This way, they remain local enough to the NegP projection to count as its overt realization (with no need for a further expletive). In strict NC languages, instead, n-words would move further to a higher, separate Focus position, leaving behind an expletive NM to realize the NegP head. Note that the syncretism between the NegP and the FocP position would not necessarily lead to a focused reading for the negation. I assume that the only semantic effect of this syncretism is to guarantee that the scope of sentential negation coincides with the informational focus of the clause, the default case (cf. §...). For Italian and Spanish, some influential proposals have argued for a syntactic connection between negation and focus in the CP–TP layer. Frascarelli () locates Italian pre-Infl n-words (and the NM as well) in a general Operator position FP, where Focus and wh- features are also located. AlonsoOvalle and Guerzoni () propose that the abstract negation that licenses pre-Infl n-words in non-strict NC Romance systems is higher than the position of the NM (corresponding to Zeijlstra’s Last Resort operator) and argue as well that it may be hosted in a CP Focus position. Isac’s () proposal, although assuming a different format for n-words, is very similar to mine: in Isac’s analysis, in non-strict NC systems the focus and negation features are hosted in the same projection (FocP). In the strict NC language Romanian, instead, they are different heads. Evidence for an analysis in terms of Focus movement is represented, e.g., by the impossibility of clitic doubling with n-words, shown in (): () a. *Nientei (non) li ’ha mangiato ‘S/he hasn’t eaten anything’ b. *A nadiei loi vi ‘I didn’t see anybody’

Italian (Frascarelli : , n.) Spanish (Isac : )

More generally, ellipsis-based accounts of negative short answers posit that the NM or n-word is located in the specifier of a Focus projection, whose complement is elided (cf. Giannakidou , Merchant , Holmberg , Holmberg , Brunetti , Watanabe , Poletto , Repp  a.o.). It is also clear, at least empirically, that double-negation readings in Italian are dependent on focus: namely, the element not entering Concord, and contributing, thus, a further semantic negation, is focused (resulting in a pragmatically marked construction, a.o. in contrastive denials, cf. Haegeman and Zanuttini ):

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() NESSUNO non ha pagato nobody not has paid (Differently from what has been said / from what one may think) ‘No one has not paid’ = everyone has paid (capitals indicate focus accent) As Biberauer and Roberts (: ) say, focus ‘seals off ’ the additional negatively marked item: the Focus operator locally associated with it disrupts the chain of Agree, resulting in independent negation operators.37 Double-negation readings in non-strict NC languages would represent a pragmatically driven, marked dissociation of the Focus projection and NegP. Examples like the one we saw in () are strong hints at the fact that, in a non-strict language like Italian, FocP and NegP can be dissociated: in () the subject intervenes between the NM and a focused adverb in sentence-initial position, therefore no syncretism can be assumed here (cf. Panizza and Romoli : – for the conclusion that It. mai ‘never’ is in a Focus position when preverbal). If we assume that, instead, in ‘plain’ Concord structures the scope of the Focus operator and the scope of sentential negation coincide, we can think of the preInfl elements in non-strict NC systems as simultaneously and obligatorily realizing both the NegP and FocP projection. Only in pragmatically marked cases the FocP projection will be realized disjointly (and narrowly associate with one element). In strict NC systems, in turn, the FocP and the NegP projections would be systematically distinct. This does not seem to have effects on the semantic interpretation, but it has a very important syntactic consequence, namely that two positions are available for negatively marked items (NM and n-words) in the pre-Infl field. Since the NM marker in NegP is not focused, no intervention effect arises when the n-word is above it. Still, double-negation readings can arise in Romanian as well; similarly to Italian, the ‘special’ interpretation is signaled by means of prosody: () Nimeni nu iubeste pe nimeni ‘Everybody loves somebody’

Romanian (Isac : )

In Romanian, the ‘positive’ (double-negation) reading is possible only if both n-words are stressed. According to Isac, they are both focused. Similar judgments are known from the literature on French, another strict NC system. Also in French, Concord can be suspended under special pragmatic conditions: the following example can be ambiguous, depending on focus accent: () Personne ne commet aucune erreur (Déprez and Martineau ) NC reading: no one makes any error Double-negation reading: no one does not make any error

37 In fact, Biberauer and Roberts () argue that this mechanism related to Focus is responsible for Double Negation languages in general, not just double-negation readings in NC languages. According to their proposal, ‘the difference between Standard and NC varieties of English lies in the obligatory Agreecompromising focus requirement associated with multiple-negation-containing structures in the former (in the latter, it is merely an option).’ (Biberauer and Roberts : ).

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



Of course, establishing the validity of my proposal would require a thorough crosslinguistic survey, which I am not able to provide here. In the rest of the discussion, I will therefore remain conservative: I will keep assuming, with Zeijlstra, that the NM of non-strict NC languages is [iNeg] and I will base my diachronic analysis on this assumption. Nonetheless, I believe that the hypothesis of a distinction between strict and non-strict NC systems based on factors independent of the featural content of the NM merits further investigation. We can interpret [iNeg] on the NM as a mnemonic for the fact that in a non-strict NC system the Spell-Out phase for negation is fixed and certain locality requirements may allow for only one element bearing a formal negative feature to appear in that phase.

. Latin negation system and Jespersen’s Cycle One fact emerged quite clearly from the discussion of the theoretical framework in the previous sections: since the nature of indefinites in negation systems is inextricably tied to the nature of the negative marker, their evolution in Latin can be understood only in the context of the diachronic processes globally affecting the expression of negation, and in particular the negative marker. These go under the name of Jespersen’s Cycle, a crosslinguistically frequent and systematic process leading to the formal renewal of negative markers. In this section, I therefore provide a background for various syntactic and semantic aspects concerning the Classical Latin negative marker n¯on; I base this on my own corpus work and on existing extensive studies (especially Orlandini a, Fruyt a, Bertocchi et al. , Danckaert : ch. , Danckaert : chs. , , Pinkster : ch. ). In §. I proceed to examine the evolution of the system in Late Latin, with the aim of tracing back to this stage developments affecting the Early Romance varieties. I start by presenting some core aspects of Jespersen’s Cycle (§..). In order to to evaluate the status of Early and Classical Latin with respect to it, I discuss the prehistoric grammaticalization path leading to Latin n¯on (§..). I subsequently analyze the synchronic syntax of the Early and Classical Latin plain sentential marker (§..–..): this will imply reviewing some basic aspects of Latin sentence structure as well. I also shortly discuss cases of ‘redundant’ negation noted in the literature for Early and Classical Latin (§..). The section closes with a short summary of my conclusions (§..). .. Jespersen’s Cycle ... Typological and diachronic predictions The theoretical analysis introduced in §. leads to important typological and diachronic predictions. Since DN languages lack a formal, syntactically active feature pair for negation [i/uNeg], they do not grammaticalize a Neg projection, as there is no evidence for the establishment of Agree relations during acquisition. From this the following prediction arises: () Negative heads are predicted not to be available in non-Negative-Concord languages. There is no language without Negative Concord that exhibits a negative marker that is a syntactic head. (Zeijlstra , : )

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

In other words, if the NM of a given language has the syntactic status of a head, then the language will have Negative Concord. A negative head projects its own category, NegP, which requires establishing visible relations with other negatively marked elements in the clause: this yields Negative Concord. In languages without Negative Concord each negative element contributes a logical operator for negation, potentially in its base position: the negative marker attaches as adjunct phrase to a verbal projection, and no NegP is projected in the clausal spine. The fact that the structural status of the NM is related to the presence of Negative Concord connects in interesting ways with the generative interpretation of Jespersen’s Cycle. It is well known that the exponent of sentential negation is subject to a cyclical renewal process, resulting in its reinforcement. During the cycle the original negative marker gets to be reinforced—first optionally, and then obligatorily—by an emphatic additional element; at the end of the cycle the reinforcer has assumed ‘negative force’ and replaced the original element as marker of plain sentential negation. Elements acting as reinforcers are typically minimizers and generalizers, triggering pragmatic strengthening owing to their activation of scalar reasoning (see especially Eckardt, , , ). Minimizers evoke a pragmatically relevant scale and convey that the proposition does not hold even for the lowest end point of such scale. Generalizers are domain wideners, and carry the message that even considering the most marginal, irrelevant peripheries of the domain, the proposition still does not hold. Typically, the use of such devices goes hand in hand with focalization of the reinforcers.38 Also indefinites may act as reinforcers of negation. They do so both as adverbs and as potentially emphatic markers of arguments in the scope of negation. Adverbial uses of indefinites, in turn, very often have their origin in an argumental usage (cf. §...). The cycle is triggered either by morphophonological erosion or by semantic bleaching of an originally emphatic negator (and most probably by both factors; see §... for discussion). Syntactically, this process may be seen as the reinforcement of an X negative marker by means of an XP element, which may subsequently become the ‘real’ negator (Rowlett , van Kemenade , van Gelderen , Jäger ). In the course of time, the cycle may proceed further, leading to the reanalysis of the XP element as an X . At this point, cyclically, a further reinforcement process may become productive. The classical model of Jespersen’s Cycle (formulated after the original observations in Jespersen ; from now on abbreviated JC) comprises three basic stages.39

38 Although this reinforcing strategy is frequent also outside Indo-European, non-Indo-European languages show further renewal strategies of the negative marker, as for instance the development from a negative verb. See van Gelderen (: ch. ) for a survey of case studies. 39 Recent useful overviews of current research on Jespersen’s Cycle are van der Auwera (), Hoeksema (), van Gelderen (: ch. ) (with a focus on non-Indo-European families), Willis et al. (a), various contributions in Larrivée and Ingham (), Hansen and Visconti (), Meisner et al. ().

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



These stages are schematically summarized in (), but from the discussion above it is already clear that e.g. Stage III may be decomposed into two substages, one retaining the XP status and the further one reducing the negator to a X and thus leading to a new Stage I.40 () Correlation between stage in Jespersen’s Cycle and phrase-structural status of negative marker a. Stage I: simple negative marker—X (Old French ne) b. Stage II: reinforced negative marker—X + XP complex (French ne . . . pas) c. Stage III: renewed simple negative marker—XP or X status (Colloquial French pas) So far, JC has been described as a phenomenon involving the NM. However, according to the global perspective on negation systems that I presented, it is to be expected that changes affecting the negative marker have an effect also on the indefinites interacting with it. In particular, a negative marker changing to head status, i.e. becoming [iNeg] or [uNeg], will cease to be compatible with [Neg] negative indefinites in a singlenegation reading, since both elements always result in the insertion of a logical negative operator and they cannot undergo Agree. Thus, reaching Stage III of JC will trigger a chain reaction in the indefinites belonging to the negation system. In my reconstruction, this is exactly what happens in Late Latin, and constitutes the starting point for Romance Negative Concord systems. Detailed investigation of individual language histories has uncovered that, behind the neat generalizations that we can undoubtedly formulate about this cyclical process, a number of variables are hidden, that have to do with system-internal factors (clause structure, DP structure, phonological processes), as well as with sociolinguistic factors (language contact, conservativism / prescriptivism). The process is not completely deterministic: languages may persist in a given stage throughout very different time spans, and may even ‘opt out’ of the cycle (Biberauer ). In fact, as remarked by Willis et al. (a), the stages in the cycles should be understood as not necessarily exclusive, not necessarily temporally discrete ‘construction types’. Recent research has also uncovered crosslinguistic differences in the duration and in the coexistence of the various stages (cf. Wallage  for English, Jäger  for High German, Breitbarth  for Low German). Also, despite formal renewal in the domain of NMs (and indefinites interacting with them), some properties of the system may be very pertinacious. This is, for instance the conclusion reached by Kiparsky and Condoravdi () with respect to the history of Greek, where they find ‘a paradoxical combination of structural stability and constant lexical innovation’. The role of prescriptivism in inhibiting the dismissal of a stage, or in imposing a reduction of variation on the basis of a ‘high’ cultural model, has been traditionally acknowledged: it is most obvious for those contemporary situations where standard languages remain ‘behind’ in the cycle in comparison to dialectal varieties

40 See van der Auwera () for a more complex model, involving more fine-grained stages.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

(cf. Standard Italian versus many Northern Italian dialects, Standard European French versus Québec French), or where differences between registers are so deep as to allow two different grammatical systems of negation (Standard versus Colloquial French).41 Of course, the influence of sociolinguistic factors is much more difficult to evaluate, and thus probably easier to overlook, for historical varieties: it may to some extent explain the variation that is often met in our textual sources, and the speed at which change proceeds. Current diachronic research, however, has shown that for the majority of sources a clear text-internal grammatical coherence can be disclosed, once the analysis is based on sound synchronic models. This applies also to the Latin sources I examined for my research, which appear to be valid witnesses of coherent systems. ... The triggers A much-debated aspect of JC concerns its triggers. The factors triggering the transition from one stage to the next may vary. In general, two competing explanations have been proposed for entering Stage II: a phonological weakening (pull chain) explanation and a pragmatic strengthening (push chain) explanation. Current research generally acknowledges that the two need not be complementary, since both processes seem to go hand in hand in language history. The weakening explanation emerges in Jespersen’s () foundational description of the cycle. The strengthening explanation has a long history as well, since it already figures in Meillet’s discussion of grammaticalization processes involving negation (Meillet ). Its import for research at the syntax-semantics interface has been highlighted recently in a particularly convincing way by Kiparsky and Condoravdi () (but see also Eckardt , : –). They argue that the cycle is pragmatically driven, and is caused by the formal and semantic instability of emphatic negation, which is however a necessary function (expresses a necessary contrast) in every linguistic system and has to be renewed. As we saw in chapter , emphatic negation is subject to semantic bleaching due to an ‘inflationary effect’ (Dahl ) coming about with overuse. First, it loses its original lexical meaning (‘step’, ‘crumble’, ‘drop’, etc.) and is thus able to occur in a broader number of contexts; subsequently, it loses emphasis and changes into a plain (possibly discontinuous) negation. However, despite this pragmatic pressure, it seems that reinforcers tend to not become obligatory unless the original marker reaches a certain degree of phonological weakness. Thus, for instance, Italian has been argued to be a case of ‘blocked’ JC (van Gelderen , Poletto ): reinforcers of negation exist from the beginning of written tradition (Zanuttini ), and many are available in nonstandard varieties and in colloquial regional Italian; however, Gallo-Italic dialectal varieties mostly underwent a full change to Stage II and III, whereas in the remaining varieties and in the standard language the reinforcer remained optional.

41 German is also a language for which both the imposition of a prestige model, represented by Latin, and prescriptive pressure have been similarly argued to have contributed to preserve the Double Negation system artificially, at least to a certain extent (cf. Haspelmath : §., Weiß c, Haspelmath ).

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



Poletto and Garzonio () argue that JC is blocked in Italian and other ItaloRomance dialects, unlike e.g. French, by the strength of the negative marker (which they analyze in morphological terms).42 The presence of sufficient phonological material to allow for focal stress seems to be a factor protecting old negators from being systematically doubled and replaced: focus on the original negative marker may still serve the purpose of pragmatically strengthening the negation.43 ... A generative model of Jespersen’s Cycle Under a generative interpretation of JC, the stage which sees the loss of the old negative marker in favor of the new one (Stage III in the classical model) must be decomposed in two substages: one in which the new negator retains its original phrasal status, and one in which it is reanalyzed as a head. This latter step ensures the cyclical nature of the change, since it creates the potential conditions for a new cycle, i.e., it yields an element that may be reinforced by means of additional lexical material. It is also very important with respect to the relation that the NM entertains with co-occurring narrow-scope indefinites, since, as we have seen, there is a strong (if not universal) correlation between having a head in NegP and displaying Negative Concord (Rowlett , Zeijlstra ). Van Gelderen (, , : ch. ) motivates the phrase-structural reanalysis involved in the conclusion of JC in terms of her ‘Head Preference Principle’ (), to be understood as the manifestation of a general economy strategy leading to minimization of projected structure (). As soon as a reanalysis as a smaller element becomes possible, it will be preferred in first-language acquisition (see Roberts and Roussou , van Gelderen  for a number of other cases that may receive an explanation along this line). () Head Preference Principle (van Gelderen , van Gelderen : ) Be a head, rather than a phrase () i.

NegP Neg

Spec NM

Neg0

NegP

ii. Neg0 XP

XP

NM

0

42 Poletto and Garzonio () argue that alternation among different forms of the negative marker may suggest to the language learner that negation is strong enough, and thus block JC. They argue that ItaloRomance non is morphologically complex (bimorphemic: [no]-[n]), a reflex of ‘frozen syntax’ from the Latin bimorphemic *ne-oinom, on which see §... Complexity would be suggested to the contemporary learner by the phonologically and syntactically conditioned alternation between a full and a reduced form observable in many varieties (Florentine, Sicilian). 43 Other phonological factors interacting with JC have been recently discussed by Meisner and Pomino () for French and by Hoeksema () for Middle Dutch.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

The process of structure reduction goes hand in hand with featural reanalysis. In van Gelderen (: ) elements occupying the specifier of NegP are [iNeg] and heads tend to become [uNeg], however she allows for newly reanalyzed heads, e.g. in creole languages (pp. –) to keep their [iNeg] features. In particular, she follows DeGraff () in assuming that in French-lexified creoles such as Haitian Creole, pas (Haitian Creole pa) has become a head, preceding the inflected verb.44 Zeijlstra (, ) also distinguishes between [iNeg] and [uNeg] heads. According to this interpretation, JC would be a manifestation of a more general directionality in feature change, whereby an originally semantic ([Neg]) feature becomes an interpretable formal feature (i.e., a syntactically active element) ([iNeg]), and eventually an uninterpretable formal feature ([uNeg]).45 The motivation for this kind of directionality is seen as a form of economy guiding language acquisition, cf. van Gelderen () and van Gelderen () for a recent formulation: () Feature Economy (as formulated in van Gelderen : ) a. Utilize semantic features: use them as functional categories, i.e. as formal features. b. If a specific feature appears more than once, one of these is interpretable and the others are uninterpretable. In principle, under this analysis of JC, long-range diachronic processes are foreseen in which a language may first acquire a NegP projection (becoming a Negative Concord system) and later lose it (shifting to a Double Negation system), think e.g. of the cline from Latin to Colloquial French. The assumption of de- and re-grammaticalization of NegP involved in this account may be criticized for being uneconomical. But, in fact, in a I-language perspective it is not, since the reanalysis, during acquisition, in the individual cognitive systems does not involve the ‘loss’, but rather the lack of activation of a NegP: the primary data simply do not provide the trigger(s) for it. The real problem is to account for how and why the primary data stop providing certain cues at a given stage in time. .. Grammaticalization of the Latin negative marker n¯on I now move to assess at which stage of Jespersen’s Cycle Classical Latin is situated. There is widespread consensus concerning the etymology of the plain sentential negative marker n¯on: it derives from the inherited Proto-Indo-European (PIE) negative morpheme n˘e < PIE *ne and *oinom > oenum (= u¯ num), nom-acc. sing. of

44 Other varieties of French show a different development: e.g., in Quebéc French pas, which has also become the only NM, remains post-Infl and becomes a Concord element: Je juge pas personne ‘I don’t judge anybody’; Personne est pas capable de parler français à Montréal? ‘Is nobody able to speak French in Montreal?’ (from Zeijlstra : ). 45 Interestingly, Breitbarth (: ) notes that, against van Gelderen’s Feature Economy, NPIs becoming n-words never develop a [iNeg] feature, so they go from semantic directly to uninterpretable. We will come back to this point in §.. and in chapter .

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



numeral ‘one’, ‘not (even) one’.46 The intermediate form n¯oenum is still found in archaich texts, as we will see in §.... A parallel etymological source is witnessed by German nein, the reinforced negation used for negative answers, < *n˘e + ein ‘one’, neuter form of the indefinite article. The resulting form, as well as some phenomena observable in archaic Latin texts, clearly show that the Latin plain negative marker is the product of a pre-historic but plausibly relatively recent Jespersen’s Cycle, involving reinforcement of the original negation n˘e by means of a scale-evoking minimizer. The latter acts as a semantic reinforcer, which later becomes bleached and is incorporated into the negative morpheme through an agglutination process (cf. Fruyt a, ). In pre-historical Latin, the PIE negative morpheme n˘e was targeted by the general tendency to avoid tonic monosyllables ending in a short vowel (Fruyt a: ). In this case a phonetic trigger for the further grammaticalization of the reinforced noenum seems plausible. The fact that there is no attestation of n¯on in the XII Tables (Adams : ) confirms a relatively late origin of what would become the standard marker of negation in historical Latin. ... Traces of n˘e (Stage I of Jespersen’s Cycle) As Kühner and Stegmann (: II., ) put it, ‘[a]ls ursprüngliche Negation muß ohne Zweifel n˘e angesehen werden’: the original NM was doubtlessly n˘e. As we saw above, this is the continuation of PIE *ne, functioning as default negation in many Indo-European languages (e.g. Sanskrit, Avestic, Germanic, Balto-Slavic) and leaving a trace in many others as affixal element in word-formation processes (Fortson : ). The use of simple, non-reinforced n˘e is still (rarely) found in Plautus, especially with modal verbs, such as volo. Inherited short monosyllables in Latin are prosodically weak and generally attach to the preceding word: n˘e could be enclitic (cf. e.g., quin ‘why not’) or, more frequently, proclitic on the following word. In §.. we saw the role of this element in the formation of negative indefinites. It can also be recognized in negative particles such as nisi or correlative n˘eque, n˘ec (on which see chapter ), and in products of grammaticalization such as negative verbs (n˘escio ‘ignore’, lit. ‘not know’, n˘ego ‘deny’, n˘equeo ‘not be able’, n¯olo < n˘e-v˘olo ‘not want’, etc.), but also nominal and adjectival lexemes such as n˘efas, n˘efarius, etc.

46 This etymology is mentioned by Meillet’s influential article on grammaticalization (Meillet : ), and accepted by all major sources. The most thorough discussion of this grammaticalization process is given by Fruyt in several works (a, b, ), on which my discussion in the main text is based. Note that the word’s phonetic evolution is not completely clear. The vocalic part has been attributed either to elision of the short vowel in n˘e in front of the following diphthong o¯i (> o¯ in open syllables, cf. Sihler : ) or to contraction of the two vocalic parts e˘ and oi in o¯i; in this latter case, however, one would expect the resulting diphthong to retain the vocalic quality of the first element, i.e., e¯i instead of o¯i, cf. Fruyt (a: ). Moreover, de Vaan (: s.v.) notes that the apocope of final -um (which we also observed in the etymological source for n˘ıh¯˘ıl) is irregular, that is, cannot be attributed to an otherwise active general morphophonological process. He advances the hypothesis that this may be due to unstressed use of the word. However, as we will see, there are reasons to assume that n¯on can behave as a prosodically strong element in Classical Latin. Cases of ‘irregular’ phonetic evolution with elements of the functional lexicon are not at all rare: phonetic erosion is indeed a hallmark of grammaticalization processes and has been treated as an effect of high frequency.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

According to a recent comprehensive study by Fruyt (b), the marker n˘e is also ‘hidden’ in a number of adverbial elements, which synchronically show a long-vowel n¯e but do not belong to the functional domain of the modality-sensitive negation: the discontinuous focalizing adverb n¯e . . . quidem ‘neither, not even’,47 n¯equiquam ‘in vain’, n¯equaquam ‘in no way, not at all’. Fruyt’s arguments in this respect are relevant for a general evaluation of the triggers for the pre-historical Latin Jespersen’s Cycle: she explains the vowel lengthening by proposing that the old negation n˘e occurred, in fact, in two forms, a tonic one and an atonic one.48 In Latin, the systematic lengthening of the short monosyllabic n˘e as a strategy to strengthen the negative morpheme would then have been blocked by the existence of the modality-sensitive negation n¯e (Fruyt : , n. ). In light of the evidence reviewed above, it appears that n˘e underwent the typical development of negative particles at Stage I of JC: an original tonic, independent negative marker becomes prosodically weaker, behaving like a clitic and being reanalyzed, in some cases, as an affix. ... The rise of noenum (Stage II of Jespersen’s Cycle) Early Latin shows traces of various ways of reinforcing n˘e. The form noenum, which will evolve into the standard negation, is, however, the only one where the reinforcement involves agglutination of originally separate lexemes (cf. Fruyt a: ), similarly to what happens with negative indefinites (cf. §..), and not just addition of phonetic material (in terms of lengthening). Archaic authors such as Plautus, Ennius, Lucilius, as well as Lucretius, still witness noenu(m). As Fruyt (a: , n. ) notes, the emphatic contribution of the reinforced negation is still appreciable in many passages: in (a), for instance, the negated form of the predicate is contrastively focused to the positive form in the main clause. In other cases (cf. b, c), the reinforced negation occupies the first position in the clause (a pragmatically marked configuration), and may be separated from the verb by many intervening constituents, as in (b): here noenum negates the verb queo ‘I can’ and is separated from it by the emphatic adverb mecastor ‘for God’s sake’ and a complex subordinate clause. () a. si hodie noenum venis, cras quidem sis veneris if today not.at.all come:sg tomorrow at.least please come:sg ‘if you do not come (at all) today, at least please do come tomorrow’ (Varro apud Non. .), from Fruyt (: )

47 On the semantic-pragmatic function of simple quidem see Danckaert (), and on ne . . . quidem see further §.... In later Latin (the first attestation is in Petronius, .), quidem in n¯e . . . quidem may be dropped, resulting in a ‘new’ ne meaning ‘not even’, cf. Fruyt (b: ). 48 Such a situation is not rare in contemporary languages, and signs that it must have pertained at earlier diachronic stages have been recently brought to attention for French (Ingham ) and Low German (Breitbarth ). In general, it is expected from a theoretical perspective, since it is known that monosyllabic function words may have different phonological mappings, and appear as an autonomous prosodic unit only in certain contexts (under focus, in isolation), alternating with prosodically weak variants, forming a prosodic constituent with another element in the clause (cf. Selkirk , and a recent diachronic application in Hinterhölzl and van Kemenade ).

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



b. Noenum mecastor, quid ego ero dicam meo / not.at.all by.Castor what:acc I:nom master:dat say:sg my:dat / malae rei evenisse quamve insaniam, / bad:gen thing:gen happen:inf which:acc-or madness:acc / queo comminisci can:sg invent:inf ‘For God’s sake, I absolutely cannot fancy what accident or madness I should say has happened to my master’ (Plaut. Aul. –), from Fruyt (a: ) c. Noenu decet mussare bonos qui not.at.all be.becoming:sg mumble:inf brave:acc who:nom facta labore / nixi militiae peperere deeds:acc toil:abl / strained:pt.nom war:gen give.birth:pl ‘No, it is not meet that good warriors should mumble; warriors who, straining in the toil of battle-fields, have given birth to deeds.’ (Enn. Ann.  V = W ), from Fruyt (a: ) In order to account for univerbation, the configuration resulting in the grammaticalization of noenum to n¯on must have involved a pre-Infl neuter indefinite object NP generalizing into an adverbial use with intransitive verbs, as in (a). For Latin an immediate parallel is represented by the adverbial use of neuter nihil / nil in examples such as () (cf. Pinkster : –).49 () a. nihil amas nothing love:sg ‘you don’t love me at all’ (Plaut. Mil. )50 b. ego complexum huius nihil moror I:nom hug:acc this:gen nothing wait:sg ‘but I care nothing about his hug’ (Plaut. Asin. ) c. quia ecastor mulier recte olet, ubi nil because by.Castor woman:nom right smell:sg when nothing olet smell:sg ‘Because, by Castor, a woman smells right when she doesn’t smell of anything’ (Plaut. Most. ) The process by which a quantificational argument becomes an adverb and further proceeds to grammaticalize into a NM is amply attested crosslinguistically.51 Bayer () has discussed it for German nichts, van Kemenade () and Willis () 49 In some cases it is in fact difficult to ascertain whether the pronoun is used adverbially or is inserted as a complement of the verb: in (c), for instance, the pronoun could be an ‘internal object’, since olere ‘to smell of something’ could be construed with the accusative. I am grateful to Eugenio Mattioni for pointing this out to me. 50 Cf. the parallel use of Old Italian niente: . . . una gentile pulzella, la quale non amava neente lui . . . ‘a gentle lady, who did not love him at all’ (Novellino ..). 51 For an alternative interpretation of these facts, according to which in fact NMs are uniformly nominal arguments, see Manzini and Savoia (: ch. ).

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

for English not, Garzonio and Poletto (), Poletto (: ch. ) for Italian niente; cf. also Gugán () and Kiss () for the Hungarian NM nem from indefinite ‘anything’. In Piedmontese nen, originally ‘nothing’, has become a NM (Zanuttini ). In Greek, the Modern NM dhen originates from the adverbially used neuter form of the indefinite oudén ‘nothing’, cf. Kiparsky and Condoravdi (), Chatzopoulou (). Bridging contexts for this development are represented by verbs whose internal argument may be optional or ambiguous (cf. c), such as optionally transitive verbs (e.g. eat, read) or predicates taking a measure phrase (extent or degree, e.g. motion verbs, psychological predicates) (cf. Willis et al. a, Willis ). Unfortunately, the textual evidence we have for noenu(m) does not allow us to reach safe conclusions as to how the process may have developed in detail. At the first documented stage, noenu(m) is an adverb (‘not at all’) which has no link to the argumental structure of the predicate. The most plausible reconstruction seems to me one where, as in the case of negative indefinites, n˘e merges as constituent negator with a DP, and the complex formed this way raises to a pre-Infl argumental position. The reinforcement of negation would not be a morphosyntactic ‘doubling’ similar to French ne . . . pas, in this case, but rather arise pragmatically through the choice of a different, inherently emphatic (because of its scalar semantics) lexical element. This is indeed what productively happens in historical Latin with adverbial nihil / nil (see Pinkster :  for its emphatic value). Once the originally argumental position is reinterpreted as an adverbial one, a formally renewed NM with the status of a phrasal constituent completes its grammaticalization in Early Latin. .. The syntax of n¯on: position in the clause There is a very clear generalization concerning the position of the negative marker in Classical Latin: n¯on regularly precedes the finite verb, i.e., in analytical forms () it appears immediately before the auxiliary, not before the participle: cf. Kühner and Stegmann (: II.,): ‘wenn die Negation zu dem ganzen Satze gehört (Satznegation), so tritt sie regelrecht vor das Verbum finitum, bei zusammengesetzen Formen vor das Hilfsverb’ (if the negation belongs to the whole clause—sentential negation—then it regularly appears before the finite verb, in analytical forms before the auxiliary). () Romanus equitatus ipsum quidem regem Elatiae Roman:nom cavalry:nom himself:acc then king:acc Elatea:gen adsecutus non est reach:pt not be:sg ‘but the Roman cavalry did not reach the king of Elatea himself ’ (Liv. ..) () tunc is moratus non est then he:nom wait:pt not be:sg ‘Then he did not delay action’ (Vitr. .) The data on periphrastic forms in Adams (), as well as the extensive corpus work by Danckaert (: –), empirically confirm this generalization. Recent studies

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



offer a syntactic analysis: Devine and Stephens (: ), Danckaert (: ), Danckaert (: –) identify the position of n¯on before Inflection.52 Pre-participle cases are most typically cases of constituent negation or metalinguistic negation (denial): () non profectus est sed profugit not departed:pt be:sg but bolt:sg ‘he did not depart, but rather he bolted’ (Cic. Phil. .) Classical Latin is an Infl-final language, cf. Devine and Stephens (: ), Danckaert (: ); cf. Ledgeway (: –) for discussion of variation in Classical Latin and asymmetry between main and embedded clauses. The unmarked order of a main clause with an analytical verb form, with and without sentential negation, is given in (): () a. unmarked linear order: S O Participle(V)—Aux(Infl) b. with negation: S O Participle(V)—non—Aux(Infl) The fact that, in Latin as in other languages, surface syntactic scope of negation does not necessarily correspond to its semantic scope (in fact, it does not in most cases) appears very clearly when considering the behavior of objects. In Early and Classical Latin, the pre-V positioning of objects is the unmarked order. Owing to the wordorder properties of the NM, preverbal objects end up preceding the negation; linear precedence, however, clearly does not preclude them from scoping narrower than negation, as in the following example with a preverbal nonspecific indefinite (from Devine and Stephens : ): () Silius culcitas non habet (Cic. Att. ..) Silius:nom cushions:acc not have:sg ‘Silius does not have cushions’ It has to be assumed, thus, that at some level of representation the object is c-commanded by the negative operator.

52 There has been some discussion in the typological and formal syntactic literature concerning the role of the position of the NM with respect to a broader typology of word-order types (in terms of headfinality versus head-initiality). However, it is doubtful that negation is subject to head-order harmonization principles, owing to its acategorial status, cf. Cinque (: –), Biberauer et al. (b: , ). Lehmann (: ch. ) attempts to reconstruct at least some patterns with postverbal negation in IndoEuropean, which however are very marginal and due to phonological factors. Lehmann himself concludes that evidence is too scarce in this respect. Delbrück (: –) remarks the preferential position of negation before the verb and before indefinites in the most anciently attested Indo-European varieties. Cyclic alternation of pre- and post-V negation (as e.g. in French) suggested to linguists (e.g., Vennemann , Harris ) ‘the idea that the cycle allowed a language to realign its word order along more harmonic lines’ (Willis et al. a: ). However, change in word order is not the defining property of JC: rather, ‘The defining property of Jespersen’s cycle is . . . the cyclic strengthening and weakening of negative markers, rather than the concomitant changes in word order that are often present.’ (Willis et al. a: ). For similar cases of JC not involving changes in word order cf. Greek (Chatzopoulou ) and, within Romance, manco in the variety from Basilicata discussed by Garzonio and Poletto ().

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

In what follows I discuss how the linear order of Classical Latin n¯on preceding Infl may be structurally analyzed as c-command relation with respect to Infl. I follow Danckaert’s (: –) analysis for the Infl-final surface order of Latin, based on the typology of EPP satisfaction in Biberauer and Roberts (); my account is, however, different in what concerns some crucial aspects of the syntax of the negative marker. Danckaert (: –), (: –) highlights how the placement of sentential negation constitutes a clear example of the structural constraints regulating Latin syntax: despite a certain amount of flexibility in word order, motivated by pragmatic factors, the marker of sentential negation n¯on always precedes the inflected verb, i.e., the V-T complex, as already noted in the classical reference grammars. More precisely, although the NM does not have to be strictly adjacent to the verb, but can be separated from it, e.g., by the core verbal arguments, it always precedes the hierarchically highest verb in a clause (Danckaert : ). Danckaert () analyzes the unmarked SOV / Infl-final order of Classical Latin as the effect of a phrasal movement targeting the whole verbal complex below Infl, i.e., vP. More specifically: (i) Classical Latin satisfies the EPP requirement of TP by moving the (remnant) v/VP to the specifier of a projection in the split-TP that has to be higher than NegP. (ii) In turn, NegP is argued to be higher than the Infl part of TP. This yields Infl-final word orders, assuming independent V-to-Infl movement in synthetic forms, and derives the position of the NM between the lexical verb and the auxiliary in analytic forms. The analysis is summarized in (): () (Danckaert : ): [SubjP [EPP] [VP S O V ] [Subj [NegP Neg [TP T t VP ] ] ] ] According to Biberauer and Roberts (), EPP satisfaction can be performed either by the minimal category representing the goal for Agree or by a bigger projection containing it (pied-piping). In the choice between these two options, languages display a certain degree of optionality. In Latin, the whole projection containing the (trace of the) predicate and the arguments move (VP in Danckaert’s structure). The finite verb (either a synthetic form of the lexical predicate or, more rarely, an auxiliary) moves separately to a projection in the split-TP. The landing site for the finite verb is lower than the attachment level of the NM. The landing site responsible for EPP valuation is, instead, higher: Danckaert (: ) leaves the nature of this F(unctional) P(rojection) undetermined; it corresponds to the standard position for Latin subjects labeled SubjP in Devine and Stephens (: –). Danckaert’s hypothesis on Latin negation differs from mine in a number of aspects. He assumes a more advanced stage of Jespersen’s Cycle for Classical Latin, where n¯on has already been reanalyzed as an X . He also assumes that, starting from later Classical Latin, there is a further reduction process in which the head

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



becomes proclitic on the hierarchically highest verb. This would be connected to the fact that in later Classical Latin (starting in the first centuries CE) the arguments start to move separately; the vP remains in situ, resulting in the decline of Infl-final orders. Danckaert (, ) treats n¯on as the head of a NegP. If this were the case, according to the generalization in () we would expect Classical Latin to be a Negative Concord language, contrary to facts. My alternative proposal, safeguarding (), is that n¯on is a phrasal category sitting in a specifier attached to a projection in the TP-area, above the landing site for the inflected verb. The structure I assume is given in (): () XP-negator as TP-adjunct: [FP [EPP] [VP S O V ] [F [TP Op¬P [ T tVP ] ] ] ] The example in () is analyzed accordingly below: () Structure for example () CP FP

TP

[S O V]V Pi equitatus regem adsecutus

T

Op¬P non T0

VP

est

ti

.. The phrasal status of Classical Latin n¯on (Jespersen’s Cycle, Stage III) In this section I provide evidence to support my hypothesis that Classical Latin n¯on is a phrasal element. A specifier status is diachronically plausible for the product of a recent Jespersen’s Cycle, and we have seen that Early Latin shows signs that the process must have been concluded shortly before the beginning of attestation. There are also a number of distributional arguments speaking in favor of this hypothesis, which I will review in what follows. Note that the phrasal status of NMs is usually assessed on the basis of classical tests related to verb movement (cf. e.g., Zanuttini ). These tests are based on the assumption that, if an NM is a (phonologically independent) head, it should block movement of the verbal head to a landing site above it (Head Movement Constraint). Unfortunately, such tests are difficult to apply in Latin, a.o. since in prohibitions, the clearest potential case of V-to-C (cf. Lohnstein ), a different modality-sensitive negator, n¯e, with clause-typing ability appears:

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

() hoc facito, hoc ne feceris this:acc do:sg.imp this:acc not do:sg.subj ‘Do this, do not do this’ (Cic. div. .) I will therefore base my argument on other sources of evidence. ... Syntactic autonomy It has been observed that in archaic Indo-European varieties the negative marker enjoys a substantial flexibility and may appear separated from the main predicate, often in sentence-initial position (Fortson : ). An explicit interpretation of this phenomenon as a sign of the lack of a NegP projection in early Indo-European has been proposed by Ingham (). The marker n¯on is not a clitic in Classical Latin: it counts as ‘full word’ for secondposition phenomena (Wanner ) and can itself host prosodically weak elements, for instance forms of esse ‘to be’ (Adams ; Spevak : ) or special clitics such as -ne in nonne (Wanner : ). Wanner (: ) allows for the existence of an unstressed variant of both n¯on and n˘ec, but at the same time remarks that these negative elements can host a weak pronoun in Wackernagel position, as also complementizer and relative pronouns could do, qualifying thus for filling the first position in the clause.53 Adams (: ) investigates the factors underlying the frequent second-position patterning of the copula esse ‘to be’, according to Wackernagel’s Law (whose formulation had in fact been based also on this kind of Latin evidence). He shows that in Early and Classical texts, both in its existential and auxiliary use, ‘esse is regularly attracted to one particular host, namely the negative non and some other negatives’ (that is, negative indefinites in first position, which appear to be quite frequent). () non sum ego propter nimiam fortasse constantiae not be:sg I:nom because.of excessive:acc perhaps consistency:gen cupiditatem adductus ad causam desire:acc brought:pt to cause:acc ‘I did not go over to his side on account of a perhaps unreasonable regard for consistency’ (Cic. Pis. ) Adams attributes this attraction to the enclitic status of the forms of esse, and to their need for a strong prosodic host. This tendency, however, does not apply blindly, since the infinitive form of the predicate that esse attaches to would in principle suffice as host; interestingly for us, Adams proposes that it is subject to information-structural constraints. Besides negative elements, other possible hosts for esse are antithetical terms, adjectives of quantity or size, superlative adjectives, intensifiers (e.g. multo, valde), demonstratives. According to Adams, what these hosts have in common is that they are focused or contrastive in some way, his more general point being that the second-position constraint is not applied blindly in Latin, but it is sensitive to the

53 In this respect, n˘ec behaves differently from its positive counterpart et, which is usually extraclausal and does not count as first position for cliticization phenomena, cf. Wanner (: ).

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



informational status of the host. Adams’ general conclusion is that ‘esse is often placed after the focus of a clause’ (Adams : ). This would be a manifestation of a general phenomenon in Latin, according to which ‘clitics were becoming increasingly dependent on “prominent constituents” ’; as a consequence, ‘esse might be drawn away from its usual host, the predicate, to a constituent which was more markedly the focus.’ (Adams : ). The forms of esse prosodically attach to their hosts and are dependent on them for their positioning, a fact which indirectly attests to the freedom of positioning of non: ‘non may move around in the colon, from first position to second last, with esse always behind it’ (Adams : ). As said, flexible positioning of n¯on is motivated with its ability of being focused.54, 55 In general (that is, with all verbs and not only with esse), n¯on always precedes, but is not necessarily adjacent to, the finite verb. Many discontinuous instances seem to be cases where n¯on undergoes Operator movement to a C-peripheral Focus position (a–c) (cf. Adams , Bernini and Ramat : , Pinkster : – for the ‘emphasis’ of clause-initial n¯on). A flexible, autonomous positioning of n¯on is also possible with replacive negation (d) (on the focus-sensitive nature of replacive negation cf. Jacobs : –, Repp ). () a. non edepol nunc ubi terrarum sim scio not by.Pollux now where lands:gen be:sg know:sg ‘I absolutely do not know where of all places I am’ (Plaut. Amph. ) b. non esse servos peior hoc quisquam potest not be:inf servant:nom worst:nom this:abl any:nom can:sg ‘There cannot be any servant worse than this one’ (Plaut. Asin. ) c. non hic Philolaches adulescens habitat hisce not this:nom Philolaches:nom young:nom live:sg this:abl-very in aedibus? in house:abl ‘Isn’t it true that the young Philolaches lives in this house?’ (Plaut. Most. ).56 54 Also in Greek negative elements serve as host for enclitics. Adams remarks how his explanation for the role of negative elements as hosts connects with a similar hypothesis formulated for Greek by Horrocks (: ). 55 Interestingly, Adams (: ch. ) discusses cases in which esse itself can be in first position, or anyway precede the subject: according to Adams, these uses are ‘veridical’ and ‘assertive’, like some Greek cases studied by Kahn: ‘the verb expresses the truth of statements and cognition of the being-so of facts and states-of-affairs’ (Kahn : ): cf. Cic. Lael.  sunt ista, Laeli translated by Adams (: ) as ‘What you say, Laelius is true, IS the case’ (emphasis in original). I agree with Adams that also in this case focus, and more precisely verum (polarity) focus, is at play. This shows us two important things. First, Latin has the option of displacing elements to express verum (polarity) focus. Secondly, as argued by Adams, when in first position esse cannot be enclitic; this means that the copula must have been subject to two distinct accentuation patterns (exactly like personal pronouns, which if emphatic can be placed at the beginning of a colon). This reminds us of what has been observed in §... concerning the coexistence of multiple ‘versions’ of some functional elements, for which the prosodic status corresponds to the amount of projected structure, and thus, to the types of functions it may assume. 56 This is a reaction to a previous negative statement of another character (: Puere, nemo hic habitat ‘My boy, no one lives here’).

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin d. non mihi iam furtum, sed mostrum ac prodigium not me:dat then trick:nom but wonder:nom and prodigy:nom videbatur seem:sg ‘in fact it did not look to me like a trick, but rather as a wonder and a prodigy’ (Cic. Verr. .)

A further case in which a semantic motivation may lead to a structure where the negative marker is separated from the main predicate is represented by denials, where it is crosslinguistically frequent for negation to occupy a higher syntactic position (cf. §...). In Latin there are cases where the NM n¯on scopes over two sentential conjuncts, with the reading ‘it is not the case that A and B’, where A and B are full-fledged clauses. Cf. the following example from Devine and Stephens (: ): () Non et legatum argentum est et non est not and bequeathed:pt silver:nom be:sg and not be:sg legata numerata pecunia bequeathed:pt cash:nom coin:nom ‘it is not possible to say that silver was bequeathed, but coin was not’ (Cic. top. ) In this case the negation semantically targets the conjuction: ‘it is not the case that at the same time / in the same situation . . . ’. We thus see that particular information-structural factors could condition the positioning of Latin n¯on. A change in this domain must be considered responsible for the loss of autonomy of the NM in Romance and for its stricter verb-adjacent positioning. Despite the observed syntactic flexibility in Latin, it would be wrong to treat such flexibility as a sign of ‘non-configurationality’ or ‘loosened configurationality’. Wanner (: ) states that the position of Latin n¯on ‘was only regulated by the semantic and prosodic constraints on Latin period construction’ and that the particle was ‘preverbal in a loose sense’; but we saw that especially data concerning the placement of the NM with respect to the elements of the verbal complex point to a fixed syntactic position for n¯on in unmarked orders. The observations that it can also function as constituent negation (thus placed immediately before the negated constituent) and that it is still syntactically independent and can undergo displacement do not undermine this conclusion, since these phenomena are found also in well-studied ‘configurational’ contemporary languages. ... ‘Why no(t)?’ test The ‘why no(t)?’ test was originally proposed by Merchant () in order to detect the phrase-structural status of negative markers: he argues that only phrasal negators can be found in elliptical constructions like ‘why not?’: cf. Germ. warum nicht?, Fr. pouquoi pas? versus It. *perché non?. The reason he proposes for this is that in elliptical constructions, only XP elements can adjoin other XP elements, whereas heads are excluded. Given that the word for ‘why’ is an XP constituent, only XP-negators will be able to adjoin it.57 57 An alternative analysis of this crosslinguistic generalization consists in assuming, together with much literature on the topic, that fragments of this kind are derived by ellipsis following the movement of the NM

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



Latin n¯on can adjoin to ‘why’ and to other phrasal elements in elliptical constructions. Latin has various ways of expressing ‘why not?’: the most important are the interrogative adverb quidn¯ı?, composed of the neuter form of the interrogative pronoun quid and the enclitic negative particle -n¯ı (a strengthened form of n˘e, cf. Fruyt b: ), and the combination of words meaning ‘why’ (cur, quare) and n¯on. A further element, which however does not occur in elliptical constructions, is qu¯ın, from quid and enclitic negative n˘e, on which see Fleck (). In the Early Latin of comedy, we find various instances of quor non ‘why not?’ in elliptical questions (in these texts cur occurs in its archaic form quor) (a and b); in (c), adverbial qu¯ı ‘how’ is combined with n¯on and followed by a ‘because’ answer. () a. Speaker A: tibi ego credam? Speaker B: quor non? you:dat I:nom believe:sg why not? Speaker A: quia . . . because . . . Speaker A: ‘Should I believe you?’ Speaker B: ‘why not?’ A: ‘because . . . ’ (Plaut. Pseud. ) b. Speaker A: Quor non? Speaker B: Sanum te credis why not sane:acc you:acc believe:sg esse? be:inf Speaker A: ‘Why not?’ Speaker B: ‘Are you sure you are in your right mind?’ (Ter. Ad. ) c. Speaker A: Qui non? Speaker B: Quia flagitium est why not because disgrace:nom be:sg Speaker A: ‘Why not?’ Speaker B: ‘Because it is a disgrace’ (Plaut. Bacch. ) There are also other elliptical structures in which n¯on is found adjoining plausibly phrasal elements: () a. Vel adest vel non. either be.present:sg or not ‘Either he is present or he is not’ (Plaut. Mil. ) b. Eam si nunc sequor, quanam? cum illo non; ad it:acc if now follow:sg by.what.way with that:abl not to quem cum essem profectus . . . which:acc when be:sg go:pt If I now follow it (sc. the impulse of fleeing), which is going to be my way? with him (sc. Pompey) I won’t go. When I set out to join him . . . ’ (Cic. Att. ..) to the specifier of a polarity projection, similarly to what happens in short ‘yes / no’ answers (cf. Holmberg ). If we assume that this movement can apply only to phrasal constituents, we derive the generalization. The Latin Library corpus powered by CQPweb (https://cqpweb.lancs.ac.uk/latinlib/index.php) allows one to search automatically for those instances where the NM n¯on occurs in clause-final position. It retrieves  matches for Plautus, which include short answers and elliptical structures: the two contexts would be unified by the line of analysis sketched here.

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

Merchant () notes a limitation to the application of this test in languages where the NM is homophonous with the particle used in negative answers: e.g., in Spanish, where the sentential NM no clearly has the distribution of a head, ¿porqué no? is nonetheless possible. This, however, does not represent a counterexample to the generalization, as Merchant remarks: the negative element used in ¿porqué no? is not the NM, but the homophonous XP particle used for negative short answers. In other words, in Spanish, as in other Romance languages, a lexical split occurred separating the continuation of n¯on used as sentential NM, which has an X status, and the XP particle used in short answers; however, while in other languages the split is clearly mirrored by the morphophonological forms (cf. It. NM non versus answer particle no; Fr. NM ne versus answer particle non), in Spanish the two items are homophonous. Therefore, the test does not apply straightforwardly. With Latin n¯on we have, in principle, a similar situation. The word n¯on is also used as negative short answer (alone or with repetition of main predicate), cf. (). () A: ‘venitne homo ad te?’ B: ‘Non!’ come:sg-ne man:nom to you:acc no ‘Is the man not coming to you? No!’ (Plaut. Pseud. .) Merchant’s test may thus be inconclusive for Latin, since in principle a lexical split with homophony between an X negative marker and an XP answer particle cannot be excluded. However, given the fact that n¯on, as we saw, is a recent product of grammaticalization, I find the hypothesis that it might have undergone a lexical split already at the stage of Early Latin unplausible, and I consider its distribution in elliptical structures as evidence of its XP status. .. Emphatic negation To conclude the discussion on Classical Latin, I discuss cases of ‘redundant’ marking of negation, which are frequently cited in the literature as possible signs of an incipient development of Negative Concord. I discuss them in the context of emphatic negation, since formal redundancy seems to correlate with pragmatic strengthening. Emphatic negation in Latin had a number of possible realizations. It could be expressed by means of adverbs such as minime: () nisi mi L. Clodius noster Corcyrae dixisset if.not I:dat L.:nom Clodius:nom our:nom Corcyra:loc say:sg minime id esse faciendum absolutely.not this:acc be:inf do:ger ‘if our friend L. Clodius had not told me at Corcyra that this should not be done at all’ (Cic. fam. ..) But also numquam, lit. ‘never’ could assume the meaning ‘absolutely not’: () Numquam nummum never coin:acc ‘not even a red cent’ (Plaut. Most. ) A number of minimizers and idiomatic expressions are found in the texts, especially in the more colloquial varieties:

OUP CORRECTED PROOF – FINAL, //, SPi

Latin negation system and Jespersen’s Cycle



() from Väänänen (: ): a. quoi neque paratast gutta certi consili who:dat and.not provided:pt-be:sg drop:nom safe:gen advice:gen ‘Not a drop of certain counsel is ready for you’ (Plaut. Pseud. ) b. ab hac mihi non licet transversum, ut from this:abl I:dat not be.allowed:sg lying.across:acc as aiunt, digitum discedere say:pl finger:acc diverge:inf ‘from this standard I cannot diverge a finger’s breadth, as the saying is’ (Cic. ac. .) As commonly happens, some of these idiomatic expressions have lost transparency for the speakers, as the following example with nauci ‘trifle’ (an original genitive form of unclear origin) testifies: () Qui homo timidus erit in rebus dubiis, who:nom man:nom timid:nom be:sg in situation:abl uncertain:abl nauci non erit; atque equidem quid id esse trifle:gen not be:sg and indeed what:acc it:acc be:inf dicam verbum nauci nescio say:sg word:acc trifle:gen ignore:sg ‘Someone who is timid in emergencies won’t be worth a farthing (nauci). And I don’t know what I should say the word ‘farthing’ (nauci) means.’ (Plaut. Most. –) Also negative indefinites and NPIs could be used as negation strengtheners. In principle, two basic strategies could be used in Latin (cf. also Chatzopoulou :  for the similar situation of Ancient Greek): () a. (i) NM—Verb—NPI indefinite b. (ii) Negative indefinite—Verb The predominant OV order with negative indefinites that we will see for Late Latin, as well as the focused use of nihil ‘nothing’ in adverbial function (cf. §...) may be examples of strategy (ii); examples of strategy (i) comprise combinations with NPIs, but also early, ‘exceptional’ cases of reinforcement by means of the negative indefinites co-occurring with the NM. These are the cases that I will call ‘redundant’. Colloquial varieties sporadically show emphatic single-negation readings of two co-occurring negative elements, cf. Molinelli (), who notes how these examples remain very rare until a quite late age. In () I present examples from Early, later Classical, and Late Latin: () from Ernout and Thomas (: –) a. Iura te non nociturum esse homini de swear:sg you:acc not hurt:pt be:inf man:dat because hac re nemini this:abl matter:abl no.one:dat (Plaut. Mil. ) ‘swear that you won’t be obnoxious to anyone because of this matter’

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin b. neminem nihil boni facere oportet no.one:acc nothing:acc good:gen do:inf be.appropriate:sg ‘nobody should never do no good’ (Petr. .) c. neminem tamen nihil satis est nobody:acc though nothing:nom enough be:sg ‘No one is ever content with anything’ (Petr. .)58 d. non respondes nihil not answer:sg nothing:acc ‘you don’t answer anything’ (Vetus Latina, Marc. .)

These uses have of course suggested to many scholars that they might be early examples of Romance-like Negative Concord structures. However, as mentioned, their very low frequency in our witnesses does not allow us to reach safe conclusions here. Moreover, the extensive lexical replacement in the domain of indefinites under negation during the transition to Romance seems to indicate that this kind of emphatic strengthening is not the decisive factor for the change during and after Late Latin, since, in general, Negative Concord structures first arise with newly grammaticalized indefinites (cf. chapter ). Note also that previous discussions of these examples have often failed to distinguish between these rare cases involving n¯on and the negative indefinites, on the one hand, and the more frequent cases involving, instead, the negative particle n˘eque / n˘ec. As we will see in chapter  (cf. especially §.), redundancy with n˘eque / n˘ec is systematic in some contexts already in Classical Latin and—I will argue—observes precise grammatical constraints, which are very revealing with respect to the possible bridging contexts toward a Negative Concord grammar. Other Double Negation languages are known that show a form of redundant negation in emphatic contexts, i.e., multiple negatively marked elements yielding no double-negation reading: Zeijlstra (: –) discusses data from Dutch (): () a. Hij gaat nooit niet naar school He goes never not to school ‘He never ever goes to school’ (Zeijlstra : ) b. Hij gaat niet nooit naar school he goes not never to school ‘He sometimes (not never) goes to school’ (Zeijlstra : ) Similar constructions are available in English, but Zeijlstra warns of the possibility that they may be cases of substandard Negative Concord. We see here that a precise categorization is dependent on the characterization of the negation system as a whole, and on an evaluation of the pragmatic import of these constructions, i.e., whether their emphatic value ‘bleaches’ and they become more frequent and less constrained (as we would expect in a development toward Negative Concord). In a historical perspective, this may amount to a situation where redundant marking arises as a form of emphasis and where later, owing to inflationary processes, the 58 The interpretation of this example as a single-negation reading is not the only possibility: Orlandini (: –) suggests that also a double-negation reading would be appropriate in the broader context of this utterance: ‘For no one is nothing enough’ = ‘Everyone wants something’.

OUP CORRECTED PROOF – FINAL, //, SPi

The negation system in Late Latin



construction’s pragmatic import fades, leading to a possible reanalysis in terms of a Negative Concord system. Approaches that have highlighted the importance of early redundant negation in Latin defended exactly this reconstruction scenario. In my treatment of Late Latin I will propose a different one, which is rooted in phenomena appearing long before examples of ‘redundant’ negation become frequent in the texts (and, possibly, speakers are strongly influenced by their Romance substratum, i.e., in the seventh to eighth centuries cf. Molinelli , Ingham b). Starting already in Post-Classical Latin, another subtler phenomenon is observable, concerning the position of negative indefinites in the clause. I will argue that this is connected to a change in the phrase-structural status of the NM, which I will consider as the real trigger for the rise of Negative Concord. I do not deny, though, that, in general, forms of emphatic strengthening of negation by means of ‘special’ indefinites are relevant for the further development of Negative Concord: in chapter  we will see that the new nec-marked indefinites arise in precisely this way. However, in my reconstruction of the chain of phenomena leading to Romance Negative Concord, they originate as a reaction to a previous more fundamental syntactic change targeting the NM in Post-Classical Latin. .. Interim summary I briefly summarize the main conclusions upon which I will build my analysis of Late Latin: in the previous sections I have argued that in the Classical Latin negation system: (i) the negative marker n¯on is an adverbial XP attached to a projection in the TParea, above the landing site for the inflected verb; that means, according to the framework adopted here, that no NegP projection is present; (ii) The elements in () (nemo, nihil, nullus) are negative indefinites. As such, they always result in the insertion of a negative operator, independently of their position.

. The negation system in Late Latin .. Late Latin: a ‘concealed non-strict NC language’? Non-strict Negative Concord varieties allow n-words to co-occur with the negative marker only if they follow the finite verb. N-words preceding the verb suffice to express sentential negation and may not co-occur with the negative marker. I propose that Late Latin is also a language of this type, and that the absence of co-occurrence with the NM is linked to the fact that negative objects, like negative indefinites with other grammatical functions, may precede the inflected verb. In the pre-Infl area the surface behavior of non-strict Negative Concord and Double Negation languages overlaps, despite the different featural composition of the indefinite items. I argue that the prerequisites for Negative Concord (mainly, a negative marker at Stage I of a new Jespersen’s Cycle) are already present in Late Latin: Late Latin is a ‘concealed Negative Concord language’ and transmits these prerequisites to Romance. In more detail, I propose that in Late Latin the negative marker n¯on stays in the the same pre-Infl position in the TP-area that it has in Classical Latin; however, it is reanalyzed as a head (corresponding at this point to the system of Post-Classical

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

Latin, as analyzed by Danckaert , ), following the diachronically active structure-minimizing tendency known as Spec-to-Head principle or Head Preference Principle (van Gelderen ). The negative feature on the new Neg becomes formal interpretable [iNeg], i.e., syntactically active. It is realized by inserting the negative marker. However, negative indefinites also can satisfy the requirements of the Neg projection, by entering a local relation with it through movement. In the single negation reading, the Neg head becomes incompatible with negative indefinites in its c-command domain, which bring about a negative operator of their own; the only indefinites that can appear in its c-command domain are NPIs. My corpus study shows that in Late Latin ‘old’ negative indefinites are almost exclusively pre-Infl. I interpret this as a sign that, since they convey a negative operator in virtue of their lexically encoded constraints, they may not co-occur with (the realization of) another negative operator in NegP in a single-negation reading. Negative indefinites maintain a consistent OV syntax, despite the general drift towards VO (cf. Spevak , Ledgeway ) due to the loss of systematic object movement and to the reanalysis of movement operations (Danckaert ).59 .. The development of n¯on in Late Latin: a new Stage I of Jespersen’s Cycle? According to Fruyt (: ), already in at least some varieties spoken in the first century ce n¯on shows signs of weakening, occurring adjacent to the verb, and displaying some phenomena of encliticization and fusion. Some signs of the weakening of n¯on in later Latin are observable in the documentation: they are (i) the more frequent use of emphatic strengthening, by means of the devices seen in §..; (ii) the appearance of affixal uses. For (ii), Väänänen (: ) mentions a case from the Pompeii inscriptions, shown in (), where the form nosci suggests that a new verb from a sequence n¯on + scio ‘know’ had been formed (cyclically replicating the process that had produced old n˘escio) (cf. also Fruyt a: ). () Quisquis ama, valie, peria qui nosci whoever:nom love:sg be.strong:sg? die:sg who:nom ignore:sg amare love:inf = quisquit amat, valeat; pereat qui non scit amare ‘Whoever loves, may God preserve him; who does not know how to (= cannot) love should go to hell’ (CIL IV ) .. OV order with negative indefinites It has frequently been remarked that negative indefinites are very often among the last categories to undergo word-order changes (cf. Roberts :  and references). The 59 In a late text such as the Itinerarium Egeriae, fourth century ce, the VO pattern is found in , of the main active declarative clauses (data collected according to the SSWL criteria by Giovanni Pairotti, cf. http://sswl.railsplayground.net/browse/languages/Latin(Late)). In a preliminary survey of OV / VO distribution in Latin, Danckaert (: ) remarks that in Christian Latin texts VO orders appear to be significantly more frequent. However, his subsequent, detailed study of the contexts of occurrence of the various orders casts some doubt on the traditional conclusions on the indiscriminate increase of OV in Late Latin, cf. Danckaert (: ch. ).

OUP CORRECTED PROOF – FINAL, //, SPi

The negation system in Late Latin



persistence of OV orders with negative objects during the shift from OV to VO is well known from the history of Germanic (cf. Jónsson , Svenonius , Pintzuk and Taylor ) and Romance (cf. Kayne , Poletto , ). Here I show that this is the case also for Late Latin negative indefinites: my evidence comes from their distribution in object position. My corpus investigation shows that the OV order with negative indefinites is even more consistent in Late Latin texts than in early Classical Latin texts, where their flexible placement conforms to the general pattern observed for the Classical grammatical system (cf. Ledgeway : –, Danckaert  for quantitative data and analysis). Already in Post-Classical Latin (from the first century ce), a stage when crucial innovations at the clausal level start to be observed, the OV order with negative indefinites appears to be categorical. Note that also negative indefinites with other grammatical functions, in particular those with subject function, are, in principle, relevant in assessing the properties of negative indefinites, and thus of Negative Concord, in Late Latin, since also subjects may be post-Infl. However, since the basic position of subjects is pre-Infl in both Classical and Late Latin, and also negative adverbials such as numquam ‘never’ strongly tend to appear before the inflected verb in Classical and Late Latin, distributional data may be less relevant and perspicuous. Therefore I have restricted the analysis to negative indefinites with object function. The results are shown in () and (). () position of Classical Latin object negative indefinite pronouns Text Form Tot./Relev. hits OV VO other Plautus Terence Cicero letters Varro Vitruvius Livy Celsus Celsus Petronius Curtius Rufus

neminem neminem neminem all acc. all acc. neminem null* neminem all acc. all acc.

/ / / / / / / / / /

         

         

    



() position of Late Latin object negative indefinite pronouns Text Form Tot./Relev. hits OV VO Pass. Perp. Itin. Eg. Aug. serm. Vulgata Vulgata Vulgata gospels Orosius hist. Greg.Tur. Franc.

all acc. null* neminem null* neminem nihil all acc. null*

/ / / / / / / /

       

       

other



OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

In the third column, under ‘Tot.’ = ‘Total’ I provide the number of hits that are morphologically accusative (with neuter forms, showing systematic nominativeaccusative syncretism, I adopted syntactic criteria to make sure nominative forms are excluded). Under ‘Relev.’ = ‘Relevant’ I include only those forms appearing in contexts where their relative position with respect to the selecting verb can be safely assessed and is relevant to our purposes, that is, only direct objects of transitive verbs. I excluded a quite heterogeneous class of constructions: accusative subjects of certain verbs such as paenitet, fallit; complements of prepositions; complements of possibly restructuring verbs (semi-modal, such as posse, permittere, sinere), perception verbs followed by nonfinite forms (e.g., videre); cases where no verb form is present (because elided, implicit, etc.); cases of object pronouns in AcI constructions, because other factors conditioning word order may be at work.60 Note that in the excluded cases the object pronoun still systematically precedes the selecting verb; adding them to the sample, thus, would have rendered the observed effect even stronger. So, for instance, in the case of the Vulgata, the strict criteria concerning relevance for my query may obscure the more striking fact that, of the  occurrences of accusative forms of nullus,  are preverbal. The verb is a finite form in most of the cases, but some instances of objects of participles have been retained among the relevant instances. If the negative indefinite is contained in a discontinuous DP, I considered the surface linear position of the negative indefinite as relevant with respect to the pre- / postverbal position. Many of the Late Latin pre-Infl objects appear to be emphatic (focused) (a); negative indefinites are very often found in replacive (‘not x but y / y not x’) and exceptive (‘no one but x’) (b) negation. The exceptive but-phrase is introduced in Latin by nisi (literally ‘if not’). Both replacive and exceptive negation are connected to focus, the former by substituting a correction, the latter by associating with focused constituents that express the exception to a generalization. Often the negative determiners are fronted with stranding of the remnant NP (c). () a. itaque nos ex hoc neminem novimus secundum so we:nom from here no.one:acc know:pl according carnem flesh:acc ‘Wherefore we henceforth know no man after the flesh’ (Vulg. II Cor. .) b. levantes autem oculos suos neminem viderunt nisi raise:pt then eyes:acc their:acc no.one:acc see:pl not.if solum Iesum alone:acc Jesus:acc ‘When they looked up, they saw no one except Jesus’ (Vulg. Matth. .) 60 The column ‘other’ includes figures for accusative subjects of finite verbs; the other categories listed above have been excluded from the number of the relevant instances.

OUP CORRECTED PROOF – FINAL, //, SPi

The negation system in Late Latin



c. ego nullam invenio in eo causam I:nom no:acc find:sg in he:abl charge:acc ‘I find no guilt in him’ (Vulg. Ioh. .) The steady OV order for negative indefinites does not seem to be paralleled by similar phenomena affecting NPIs or other quantificational elements. Compare this picture with the distribution of quisquam ‘any’ in object position: () position of Late Latin object quemquam Text Form Tot./Relev. hits Aug. serm. Vulgata

quemquam quemquam

OV

VO

other

 

 



/ /

Compare also the distribution of other quantificational elements, as acc. omnem ‘all’ [+animate] in object position: () position of Late Latin object omnem Text Form Tot./Relev. hits Aug. serm. – Vulgata NT

omnem omnem

/ /

OV

VO

other

 

 



I interpret this situation as evidence for the fact that in Late Latin negative indefinites systematically move to a pre-Infl position for reasons connected to the licensing of the newly grammaticalized NegP projection. Late Latin negative indefinites are not reanalyzed in their feature composition: this, as we will see below, is the main reason for the fact that they are largely replaced by new indefinites in the Romance languages. In the next section, we will have a look at some crosslinguistic evidence in order to evaluate what may cause the object movement seen with Late Latin negative indefinites. I will then formulate my hypothesis to account for the distribution of negative indefinites seen in Late Latin. .. Analysis ... Object movement with negative indefinites As I have mentioned, an important observation emerging from the literature is that, often, indefinites belonging to the negation system of a language show a different syntax to that of other pronouns; typically, when they are objects, they occupy a different position, higher in the structure; diachronically, in the shift from OV to VO it is often observed that they retain the OV order longer. The question arising is of course: why? does this have to do with negation proper, or is this a general feature of quantificational nominal phrases? Many proposals of analysis have been put forward in the literature on similar phenomena. Movement could be the manifestation of a general requirement to raise quantificational material out of Nuclear Scope, as in Diesing () (either to avoid existential closure or to take scope); this operation can be overt or covert in different languages (cf. for a similar proposal Breitbarth : , n. ). Alternatively, it could

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

be understood along the lines of Beghelli and Stowell (), by assuming a rich cartography of quantificational projections. A further hypothesis is that it could have a more direct connection with the system of negation (cf. especially work on the socalled Neg Criterion, Haegeman , Haegeman and Zanuttini ). Given the variety of possible theoretical explanations, in previous diachronic studies dealing with the phenomenon, different interpretations have been proposed for quite similar data. Pintzuk and Taylor () have shown for Old English that the change from OV to VO does not affect all objects at the same time. At some point, general object movement is lost; movement, affecting only certain classes of objects, is given a different motivation. In the Old English corpus analyzed, negative objects are consistently preverbal (), differently from other types of objects, where variation is observed more often. Ingham () has proposed that the landing site of Early English preverbal negative objects is the Specifier of NegP; this movement causes the absence of the NM ne. Other authors have connected this phenomenon rather with general syntactic processes affecting quantifiers than with negation proper. For instance, for Svenonius () Negative Movement in Norwegian is a form of overt Quantifier Raising motivated by scope demarcation. Poletto (: ch. , ) follows Svenonius () in assuming that quantifiers in Early Romance obligatorily move to a dedicated position. Thus, quantifiers show a more ‘conservative’ syntax due to the presence of a separate, dedicated landing site. In addition to this, work on old and modern Romance languages has invoked an operation of non-contrastive focus fronting targeting in particular quantificational elements in order to explain some marked word orders involving, among others, negatively marked indefinites, as in () (a phenomenon known as QP-fronting, cf. Quer , Leonetti and Escandell-Vidal , Cruschina ). () Italian, from Benincà (: –) a. Niente concludi, stando in questo buco ‘You are not getting anywhere, staying in this hole!’ b. A nessuno nuoce, col suo comportamento ‘He’s not hurting anyone, with his behavior’ My data show that, in Late Latin, with negative indefinites the OV order is more systematic than with other quantificational nominal phrases. I thus interpret the Late Latin situation seen above as evidence of a change in the negation system, and the displacement of negative objects as connected to it. ... The activation of NegP and its consequences for indefinites Specifically, I argue that in Late Latin the negative marker n¯on is reanalyzed as the [iNeg] head of a clausal NegP projection, as shown in () (to be compared with the Classical Latin structure in ()).

OUP CORRECTED PROOF – FINAL, //, SPi

The negation system in Late Latin



() Reanalysis of n¯on as head of NegP in Late Latin CP FP NegP Neg  Neg0 ¯ non [iNeg]

TP VP

DP [uNeg] General changes involving clause structure conspire to provide the conditions for this reanalysis. Danckaert (: –) presents a number of facts (loss of a form of Left Edge Fronting, decay of VOAux orders, position of adverbs) pointing to the conclusion that the EPP-motivated vP-movement of Classical Latin seen in in () is lost in Late Latin. This does not lead directly to the demise of OV, but has important frequency effects, since now the individual arguments move separately out of the vP, and may thus become subject to pragmatically motivated restrictions (cf. Polo ), or to new movement requirements. One such movement-triggering requirement, I propose, is the need for negative indefinites bringing about sentential negation to realize the negative operator in the CP–TP phase. Once n¯on is reanalyzed as an X , it acquires a formal [iNeg] feature. A clausal NegP becomes syntactically active: thus, whenever sentential negation has to be conveyed, a semantic negation operator is inserted in NegP and requires overt realization in the CP–TP phase. I propose that this can be achieved either by inserting n¯on or by moving the negative indefinite to Spec, NegP: both elements can realize the semantic negation operator. This basically corresponds to the insight in Zanuttini (: ) that ‘in a language with strong features in NegP-, a negative clause must have a negative marker in NegP- or else a negative constituent in a position c-commanding NegP-’. From Spec, NegP the indefinite can further move to a left-peripheral Focus position, deriving in this way the pragmatically and syntactically marked constructions

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

we saw in (). Alternatively, we may assume that the NegP and the FocP position, at least under certain conditions, are simultaneously realized by the same element, as tentatively proposed for Romance non-strict Negative Concord languages in §... This way, the consistent preverbal position of negative indefinites is explained by the new requirement emerging with the activation of NegP in the CP–TP phase. In the Late Latin system, the absence of co-occurrence of the negative indefinites and the negative marker in a single negation reading is derived in a way similar to nonstrict Negative Concord: indefinites belonging to the negation system and preceding the inflected verb (i.e., landing in the CP–TP phase) suffice to express sentential negation. Under this analysis Late Latin is a ‘concealed Negative Concord language’: the prerequisites for Romance Negative Concord (a negative marker at Stage I of a new Jespersen’s Cycle, but also an increasing postverbal order for objects) are already present at this stage. The ‘visible’ Concord option is not yet possible, though, for two reasons: (i) negative indefinites are not reanalyzed in their feature composition (they remain [Neg]) and (ii) no n-words are present in the system. Because of (i), negative indefinites are forced to a pre-Infl position and become obsolete in the new VO grammar. In a few cases they end up being featurally reanalyzed (Old French nul and Old Italian nullo are Negative Concord elements, as we will see in chapter ), but mostly they are ousted by new, more flexible products of grammaticalization (n-words and NPIs), as shown by the extensive lexical replacement in this domain taking place in Romance. As for (ii), the negative operator licenses NPIs in its c-command domain. NPIs can bring about a focused reading and thus function as reinforcers of negation. The conventionalization of the privileged relation with the negative operator (a form of grammaticalization) leads NPIs to gradually develop into non-emphatic [uNeg] n-words. This process affects NPIs already in Late Latin: we saw in chapters  and  how aliquis ‘some or other’ expands into the ‘direct negation’ function in Late Latin, i.e., is innovatively found as a narrow-scope existential under negation. But the restructuring of the system also prompts the creation of new narrow-scope indefinites, either NPIs or elements endowed from the start with a [uNeg] feature. Latin never becomes a full-fledged Negative Concord language in our written documentation, and new n-words develop to a large extent in the individual Romance languages. However, there are some signs of a common inheritance: a shared pattern is observable, for instance, in the various Romance versions of the combinations nec unum (e.g., Old Italian niuno, Italian nessuno, Old French negun, Spanish ningun) and nec (g)entem (e.g. Italian niente, Old French nient). In chapter  I will propose that n˘ec (here ‘not even’) functions as a focus particle, and the ensuing indefinite can be argued to have originally been an emphatic reinforcer of negation. Clearly, sociolinguistic factors connected to text transmission do not give us the full picture here, and many processes remain latent. In this respect, it is all the more significant that a strong hint to the restructuring of the system, like the invariant preInfl positioning of negative indefinites, is already observable in Late Latin texts.

OUP CORRECTED PROOF – FINAL, //, SPi

The negation system in Late Latin



... Implications for theoretical models Now that I have spelled out my proposal of analysis, let me go back to (i) and further discuss its implications for synchronic and diachronic models of the syntax of negation. If the trigger to Negative Concord in acquisition is represented solely by redundancy phenomena (as foreseen in Zeijlstra’s system), with indefinites we do not expect a cline [Neg] > [iNeg] > [uNeg], as instead happens with the NM and as is frequently the case in grammaticalization processes (cf. van Gelderen ). Rather, if indefinites are reanalyzed (which does not happen in Late Latin), we have a change from [Neg] directly to [uNeg], i.e., to elements that show redundancy with respect to other markers of negation.61 In my proposal, the first change in the negation system involves the structural status of the NM and consists in its reanalysis from a [Neg] element into a [iNeg] one. If indefinites in the scope of negation also underwent this kind of reanalysis, the system would be indistinguishable from a Double Negation one, since multiple [iNeg] items would yield structures with multiple negative operators. Given the general economy constraint regulating feature projection, such a system would be impossible to learn, because the assumption of unnecessary formal features would be banned (and formal features are unnecessary every time they do not come in interpretable-uninterpretable pairs). Still, since negative indefinites appear to be tied to a specific syntactic projection (NegP) in Late Latin, we have to account for the requirement for overt movement, despite the assumed absence of formal features. I assume that the movement is motivated at the syntax-semantic interface by the fact that an activated NegP obligatorily requires an overt realization in order to yield sentence negation. In that case, operations that before could happen at LF will have to take place overtly in the syntax. In a Double Negation system, the negation operator can be overtly realized anywhere in the clause: if it is to have sentential scope, LF operations will ensure that it lands in the right node. In Negative Concord languages, instead, the position where the sentential operator is standardly interpreted has to be identified overtly. This has an important consequence for the general model: besides redundancy phenomena, there will be rich positional evidence for the activation of the NegP projection, represented by the requirement of overt identification. In non-strict Negative Concord languages the pre- / post-Infl asymmetry seems to be crucial in this respect. In strict Negative Concord languages a similar role will be played by the obligatoriness of the insertion of the [uNeg] NM marker in all configurations expressing sentential negation. For Latin, according to my analysis, the most relevant positional evidence for the activation of NegP concerns the restrictions on NIs, which obligatorily appear in the pre-Infl field from Imperial Latin on. To this we must add the positional evidence concerning the NM itself, whose syntactic independence decreases as it starts to be more strongly tied to the verbal extended projection.

61 This connects to Breitbarth’s (: ) remark on the general lack of an [iNeg] stage in this development, mentioned in §....

OUP CORRECTED PROOF – FINAL, //, SPi



Indefinites and negation in the history of Latin

Late Latin negative indefinites will internally merge the [Neg] feature they carry into the NegP projection, putting it in the right scope position and satisfying the activation requirements of NegP. If they do not rise from the vP phase, the [Neg] feature they carry will be misplaced and will not lead to sentential negation, since NegP is not identified; in a fully derivational system, we would say that NegP is not merged at all, leading to a semantically infelicitous (or strongly constrained) structure. A similar phenomenon has been observed with [uNeg] n-words taking narrow scope, with no impact on the polarity of the clause, in Negative Concord languages (Herburger ), cf. (). In this case, Zeijlstra () and Biberauer and Zeijlstra () propose that an [iNeg] silent operator is inserted lower than NegP; therefore, while formally licensing the n-word, it does not achieve sentential scope, because it fails to take scope over the event operator (for the complex issue of scope-taking from within PPs see Penka : –, –). () a. Se pasa el tiempo mirando a nada. (Sp.) refl spend the time looking at nothing ‘S/he spends the time looking at nothing.’ (adapted from Herburger : ) b. Serena si accontenta di niente (It.) Serena refl is.satisfied with nothing ‘Serena is satisfied with nothing’ (Serena is satisfied) c. Serena non si accontenta di niente (It.) Serena not refl is.satisfied with nothing ‘Serena is not satisfied with anything’ (Serena is not satisfied) In sum, the system proposed above is based on the assumption that a negative operator is the content of NegP, and that it can be realized either by an [iNeg] element like the (reanalyzed) NM or by a [Neg] element that in the new grammar has to overtly move to take scope (the negative indefinite).

. Conclusions When we discussed Haspelmath’s () semantic map for indefinite functions in chapter , we saw that many different types of items can fulfill the direct negation function, i.e., can express existential quantification under the scope of clause-mate sentential negation. We also immediately remarked that a terminological distinction was in order, reflecting the different behavior and thus the different theoretical status of narrow-scope indefinites. Some languages have indefinites uniquely specializing in the direct-negation function, other languages have classes that occur there as well as in other contexts. Some indefinites must co-occur with an additional negative marker in a single-negation reading, others cannot, and for some indefinites the possibility of co-occurrence is dependent on their position in the clause. I adopted an analysis according to which variation in this domain is due to different specifications in terms of formal features on the negative marker, on the one hand, and on the negative indefinites, on the other hand. I described Latin as a Double

OUP CORRECTED PROOF – FINAL, //, SPi

Conclusions



Negation language during the Classical period and I argued that the diachronic process leading to the Romance Negative Concord systems is initiated by a change in the featural specification and phrase-structural status of the negative marker. Changes in indefinites follow as a direct consequence, and may happen at different paces in the various Romance languages. In this sense, a change in the featural specification of the NM—in itself, a nanoparametric process affecting an individual lexical item—necessarily triggers a macroparametric process involving the adaptation of the indefinites to the new system. In the case of the change from Latin to Romance, my proposal implies that changes in the featural specification of indefinite items happening in Romance (which we will see in chapter ) are dependent on changes in the featural specification of the negative marker happening in Late Latin. I also showed how the development seen in Latin fits into our general understanding of Jespersen’s Cycle. In Latin, Jespersen’s Cycle does not involve change from a preverbal to a postverbal negative marker, at either turning point (the pre-historic Stage III > Stage I shift involving noenum and the late reanalysis involving n¯on). There is no ‘doubling’ stage either (as happens, instead, in French): we very rarely see the negative marker co-occurring with a reinforcer; we do not have traces of doubling for the pre-historic turning point, and we see reinforcers of (continuations of) n¯on first in some of the daughter Romance languages. The Latin Jespersen’s Cycle interacts with more general changes involving the clause, especially the position of objects with respect to the inflected verb. Two tendencies, one syntactic and one pragmatic in nature, may undermine the robustness of the evidence for a Double Negation system: the syntactic one consists in the structure-minimizing tendency known as Head Preference Principle; the pragmatic one is rooted in the role of NPIs in bringing about focused readings (as we will see in the next chapter). A combination of these two factors may lead to a number of consequences: (i) the reanalysis of the negative marker as a head (and consequently the activation of a NegP), as we have seen in this chapter; (ii) the further conventionalization of the licensing relation holding between the negative operator and NPIs in its scope, i.e., the grammaticalization of n-words, as we saw with the continuations of aliquis in chapter ; (iii) finally, the grammaticalization of new indefinites functioning as n-words, as we will see in chapter . This can be the birth of a Negative Concord system.

OUP CORRECTED PROOF – FINAL, //, SPi

5 Developments in negative and polarity-sensitive contexts from Latin to Romance . Introduction .. Topics and aims The motivation for the research carried out for this chapter is the persuasion that a decompositional approach, which sees in the semantic and syntactic internal structure of quantificational determiners a source of variation and change, is needed to better understand the nature and the historical origin of the differentiation observed in Romance indefinites. I concentrate on the development of new indefinites within the negation systems of Old French and Old Italian, but in fact the class of n-words that I will single out represents a pan-Romance development. I argue that the roots of this development can be traced back to Latin, and I connect it to the general restructuring of the negation system. Indefinites that depend upon negation for their licensing are the paramount example of structure-dependent indefinites, which are diachronically sensitive to the changes occurring in the surrounding syntactic context. The morpheme that will be at the center of this chapter is the Latin negative particle n˘ec ‘and not’, ‘neither’, ‘even not’. I argue that its properties can account for the origin and the nature of Early Romance indefinites participating in Negative Concord, thus proposing a novel interpretation of some long-standing puzzles in Romance linguistics. In order to achieve my goal, I will also deal with Classical and Late Latin data, since only this long-range survey can allow us to single out some important directions of development, which then determine the Romance outcomes. In chapter  we saw how the prerequisites for a Negative Concord system arise already in Late Latin. This has tremendous consequences for the whole system of indefinites, which we started to explore in chapters  and , following the development of Latin aliquis. Now I present a case study aimed at clarifying how new Concord

Indefinites between Latin and Romance. First edition. Chiara Gianollo. © Chiara Gianollo . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi

Introduction



indefinites of Romance originate and expand in time.1 Despite the large amount of variation and lexical replacement in this domain, proceeding to a large extent independently in the various languages, there are also some steady characteristics that are common to all Early Romance languages and call for an explanation in terms of inheritance from the common Latin source. One such characteristic is the grammaticalization of new narrow-scope indefinites formed with the negative morpheme ne- / ni-. The origin of this morpheme is the Latin negative particle n˘ec, which knew quite a wide range of uses. My fundamental hypothesis is that the semantic and syntactic attributes of this particle motivate its recruitment as a building block of the new indefinites, which will be shown to fall into the typologically widespread class of indefinites formed with a scalar focus particle. Moreover, I will argue that an enhanced understanding of the Latin origin of these indefinites also explains their Concord nature, which is observable in all Romance languages since the earliest stages, and gives us also some hints toward understanding their frequent NPI uses in Early Romance. The main question I will address in this chapter can be formulated as follows: what is the status of those Romance indefinites containing a negative morpheme? Pronominal series arise gradually in time, and their elements may be of heterogeneous origin. Such is the case with the series of Concord elements in Romance, pointing to an independent and complex development. One long-standing puzzle of Romance linguistics has revolved around the question of how new Romance indefinites containing a negative morpheme have become ‘ambiguous’, in that their negativity is dependent on the syntactic context in which they occur. I will contribute to the debate by presenting data from later stages of Classical Latin and from Late Latin, showing that in fact the negative formative, n˘ec, was context-dependent as well (both in a structural and in a more pragmatic sense), and showed redundancy phenomena at an early age. In particular, differently from other Latin negative items, in certain configurations it could co-occur with the NM and with negative indefinites yielding a single-negation reading. Basing on recent analysis of correlative negations in modern languages, as well as on traditional observations dating back to Jespersen (), I try to detect whether Latin n˘ec was a genuine negative element or rather an NPI. I will show that in all its uses, Latin n˘ec can be characterized as a [Neg] element, i.e., a semantically negative item, like all other elements of the Latin negation system. Secondly, I will deal with apparent counterevidence to this claim in Classical and Late Latin, consisting mainly in the fact that n˘ec shows redundant uses, where it co-occurs with another negative element in a single-negation reading. I will connect this latter fact to the focus-sensitive nature of n˘ec, which in turn also explains why it has been recruited as a formative of many Romance n-words. I will argue that redundant uses in Latin are due to focus, and not 1 In this chapter I use the label ‘Concord indefinite / element’ for Early Romance indefinites that appear in Negative Concord structures. This label will allow me to refer to them in a theoretically neutral way, since their featural specification is a matter of discussion in this chapter.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

to negation: while excluding an analysis in terms of Negative Concord for Latin, this represents a bridge toward the Romance uses, and promises to account for the broader distribution of n-words in Early Romance. .. Roadmap I will first of all (§.) present an overview of the Early Romance situation on the basis of Old French and Old Italian data, with the aim of singling out the main issues on which an in-depth evaluation of the Latin data may contribute important insights. I then proceed to investigate the process that led to the grammaticalization of Romance narrow-scope indefinites formed with a negative morpheme. In §. we will see how this morpheme is uniformly derived from Latin n˘ec. In §. I review the distribution of this element, which had uses as discourse connective, correlative negation, and focus-sensitive particle. I show how the main uses can be derived from a single lexical entry, which however is subject to semantic and syntactic reanalysis processes from Latin to Romance (§.). Subsequently, I concentrate on the ‘redundant’ uses of n˘ec (§.). In this context, I put forward a proposal as to how Romance Concord indefinites may have arisen. In §. I come back to the Romance data seen in §. and evaluate how the conclusions reached on the Latin phenomena help explain the Early Romance situation, suggesting some research directions for those issues that will have to remain open.

. The situation in Early Romance In this section I set the prerequisites for the further discussion and deal with some selected phenomena from the history of Romance, focusing in particular on data from Old French and Old Italian. My aim is to investigate the behavior of the new Romance indefinites interacting with negation and other polarity-sensitive contexts, and the development of Negative Concord. .. Diachronic puzzles with Romance n-words It is well known (and we discussed this briefly in §.) that Romance languages display redundancy in the marking of sentential negation from the beginning of attestation. For the system of indefinites, this means that those indefinites occurring in the scope of a negative operator are endowed with particular morphosyntactic features that allow them to enter Agree relations with other negatively marked elements, i.e., Negative Concord. At the same time, Early Romance varieties show remarkable variability in the behavior of indefinites in the scope of negation and other downwardentailing operators.2

2 In this respect, my research will have to be complemented in the future by Ibero-Romance data, which are very interesting in view of the attested fluctuation between a strict and a non-strict Negative Concord system in their history (cf. the data in Martins ). A reviewer suggests that the situation described for Old Italian in §.. resembles what is observed for contemporary Catalan, on which see Déprez et al. (). However, the two-population scenario depicted in Déprez et al. () (cf. also Zeijlstra ) is not immediately compatible with my account of the Old Italian facts. In general, the situation in the

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



Specifically, two clusters of variability can be singled out. On the one hand, negative polarity uses of n-words, where they are licensed by an operator other than negation, are much more widespread in Early Romance than in the later varieties. This makes the burning issue of the semantic-syntactic status of n-words (are they negative quantifiers? are they NPIs? or neither of those?) even more puzzling. On the other hand, some languages seem to oscillate between a strict and a non-strict Negative Concord system, allowing pre-Infl n-words to co-occur with sentential negation, unlike the modern use. The survey of the Old French and Old Italian data that I provide in this chapter shows, first of all, that certain indefinites belonging to a class that I will call necwords behave like Concord elements from the start in Romance. That is, under the right syntactic conditions these indefinites are able to contribute sentential negation by themselves. In addition to this, however, they also quite extensively exhibit NPI uses, where they are not carriers or markers of sentential negation, but depend on downward-entailing operators other than negation for their licensing. Moreover, already in the oldest Romance texts no special emphasis is attached to the use of necwords, and they appear to be the ‘plain’ way to express existential quantification in the scope of negation, as their modern continuations. Summarizing, an adequate account of the historical development of nec-words must therefore account for the following aspects of their distribution in Romance: (i) their Concord properties (ii) their ability to occur as NPIs in non-negative contexts (iii) their non-emphatic nature As for (i), although n-words have behaved like Concord elements from the start, they famously show a number of idiosyncrasies with respect to the contemporary Romance languages and challenge our typology of negation systems. Sometimes there is no Negative Concord where it would be expected (e.g., with post-Infl n-words), sometimes there is Negative Concord where it would not be expected (e.g., with preInfl n-words in non-strict NC systems). With respect to (ii), although also contemporary varieties show some NPI uses for n-words (cf. Acquaviva , Zanuttini , Martins , Parry , Poletto ), these are much more widespread and productive in Early Romance. The phenomena in (i) and (ii) interact: some cases of unexpected strict NC can in fact be explained as NPI uses of an n-word, licensed by a superordinate operator (long-distance licensing). NPI uses are expected for those newly grammaticalized indefinites that derive from originally positive items, like nouns expressing a generic meaning (Fr. personne, rien, Sp. nada) or a scalar minimum (Fr. pas, mie, It. mica), quantifying determiners with an NPI origin (e.g., the continuations of Lat. aliquis). They are, on the other hand, Iberian Peninsula is very intricate and exceeds the limits of this work: it requires an in-depth investigation of its own, especially given the particular relevant role that language contact and prescriptivism and other standardization processes appear to have played there (Posner , ). Enlightening discussion is found in Martins (), Poole ().

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

surprising for those indefinites that originate as negatively marked elements, like the nec-words and the Early Romance continuations of Latin nullus. Indefinite series often comprise elements of heterogeneous origins, and the processes leading to a uniform behavior potentially develop across a long time span. Within the limitations of my work, I will not be able to deal in detail with these aspects, which are certainly relevant for the variation observed in the early texts (cf. Haspelmath : –, Willis et al. a: ). Rather, in the following sections, I will single out a class of etymologically homogeneous indefinites, formed with the Latin negative particle n˘ec: this will allow control of some variables and provide answers to some very precise historical and theoretical questions. Coming now to (iii), the fact that the new Romance indefinites do not convey emphasis, or do it only optionally, is surprising in view of two facts: they can also function as NPIs, which are typically used as reinforcers of negation, and they actually originate as focus-sensitive items, but appear to have lost early on the emphatic flavor conveyed in Latin by the combination of n˘ec with an indefinite base. This points to an early grammaticalization of these indefinites, and is interesting for our general understanding of variability in NPI behavior. Each of the points that have just been introduced is the topic of an ongoing debate in Romance and general linguistics on the semantic and syntactic status of indefinites belonging to negation systems (see Penka : ch.  for a thorough discussion of the various approaches). The diachronic phenomena have been variously interpreted depending on the underlying theoretical assumptions of the different researchers. Herburger (), for instance, has used the diachronic data to support her ambiguity approach to n-words (which would be ambiguous between NPIs and negative quantifiers, as a consequence of Jespersen’s Cycle taking place, cf. also Herburger : –). Also Martins () has attributed the optionality of NC in Early Romance to the formal characteristics of the indefinites, and has proposed that in Early Romance n-words are ambiguous as to whether they are NPIs or real negative quantifiers. The difference is captured by means of a system of features, and it is argued that diachronically the pool of possible licensors for the indefinites becomes more and more restricted, as a consequence of the loss in ‘polarity versatility’ of the n-words. The new indefinites are considered the source of optionality and change by Déprez (, ) as well (cf. also Déprez and Martineau , Labelle and Espinal ). In Déprez’s framework, the difference in syntactic properties of Concord elements corresponds to a difference in their internal structure: n-words of, e.g., Standard French are determiner-like, very high in the DP, and behave as negative quantifiers; n-words of French creole varieties, instead, rather resemble positive bare NPs, and are lower in the structure, in the lexical layer. According to Déprez, the difference in internal structure influences the distribution: while determiner-like n-words are quantificationally autonomous, ‘bare-noun’-n-words contain a null determiner, i.e., a null variable which needs to establish a relation with a binder. Consequently, in Déprez’s system there are also two different types of Negative Concord (a quantificational and a dependent type).

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



In principle, following Zeijlstra’s () system as we did in chapter , the difference between a strict and a non-strict NC system could be expected to derive from an ambiguous featural specification of the NM ([iNeg] in non-strict systems, [uNeg] in strict systems, while n-words are [uNeg] in both systems). Instead, as we saw, the most influential proposals that have been put forward to account for the ‘mixed’ system of Early Romance have rather concentrated the attention on the indefinites interacting with negation. In accordance with the studies briefly surveyed above, also the conclusions I reached for Latin in chapter  suggest looking for the source of variation and change in the semantic and syntactic properties of the new Romance indefinites. The null hypothesis would be that in Early Romance we find the [iNeg] NM that we have assumed for Late Latin, and we observe the grammaticalization of Concord items, which were still missing at the Latin stage. Given the reanalysis undergone by aliquis and its continuations (chapter ), we may imagine that Concord items in Romance uniformly develop from polarity-sensitive indefinites, whose range of licensors becomes more and more restricted in the course of time. However, this null hypothesis will have to be dismissed, at least in its more general formulation. This will immediately become clear in §.., when we shall examine the data from Old French. Since the earliest stages of attestation, French shows a very innovative Negative Concord grammar, in which indefinites are already full-fledged n-words. A further, decisive piece of evidence against the idea that all Romance n-words uniformly originate from NPIs will come from the study of their Latin origin, to which I will dedicate the major part of this chapter. As already announced, I will concentrate in particular on the meaning and the distribution of Latin n˘ec. The reason for this is that if a Romance n-word is actually morphologically negative, its negative morpheme derives from n˘ec. Already in Late Latin the negative indefinites are in competition with expressions such as nec ullus, neque ullus, nec unus, neque unus (cf. Väänänen : , ). While the combination of n˘ec / n˘eque with the NPI ullus was already current in Early and Classical Latin, the combination with unus is more typical for Late Latin. Its appearance in late texts is indicative of the Latin roots of Romance formations such as e.g., Old Italian neuno, Old French neün, Old Portuguese n˜eu˜ u, Spanish ninguno ‘nobody’. Combining the correlative negative particle with the numeral for ‘one’ is a pan-Romance strategy. In the following survey, I describe the distribution of these new narrow-scope indefinites, which will provide the basis for further discussion. .. Old French According to my interpretation of the diachronic change affecting the Late Latin negative system (chapter ), Negative Concord has already started to develop at that stage. The change is triggered by the cyclical development of the negative marker, which becomes reanalyzed as the head of a pre-Infl Neg projection. This implies, following the generalization stemming from Zeijlstra’s work (, ) seen in §.., that the NM acquires a formal syntactic feature [iNeg]: this NM,

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

thus, realizes a negative operator in NegP, and can enter (multiple) Agree relations with a corresponding [uNeg] feature. Against this background, it is very interesting to observe that, since its first attestations, Old French (OF) does not directly continue the reconstructed Late Latin system with an [iNeg] NM, but shows a more innovative grammar for the NM, even in the most archaic texts: the NM behaves like a [uNeg] element of NC, co-occurring with n-words in all positions since the beginning of attestation. As for the n-words, besides co-occurring with the NM in post- and pre-Infl position, they can also co-occur with each other (Negative Spread). This is particularly striking in view of the fact that the system of negation in the earliest texts (ninth century) is very archaic in other aspects: the form of the plain sentential NM is still non (ne appears only as correlative negation, i.e., as a continuation of n˘ec), and some indefinites directly continue Latin NIs, e.g., nul ‘no’ < Lat. nullus, nunquam ‘never’ < Lat. numquam. In this respect, OF is particularly interesting because, thanks to its early attestation and to peculiar co-occurring structural conditions, we can observe two cycles of renewal in the domain of indefinites belonging to the system of negation (see also Labelle and Espinal ). The language starts out with a mixture of indefinites clearly inherited from Latin (besides nul, also aucun < Latin aliquis, on which see §..) or possibly interited from a non-documented Latin stage, and in any case formed with Latin material (nesun < Latin nec unus). Later on, in combination with changes affecting the syntax of the negative marker as well as DP-internal syntax, French develops the new set of indefinites, personne, rien, etc., which persists in the contemporary language. A rich body of work exists on these phenomena and on their interaction with the changing syntax of sentential negation (Ingham a, Hansen and Visconti , Ingham and Kallel , Labelle and Espinal , Larrivée a,b being among the most recent contributions). In this section I review the situation found in the earliest OF documents, in order to evaluate which differences have emerged with respect to Late Latin. I will thus survey the distribution of n-words in texts from the ninth to the twelfth centuries, on the basis of my own corpus data and of the existing literature.3 Already the romana lingua in the Oaths of Strasbourg ( ce) shows the characteristics discussed above very clearly. In the following example two elements that were NIs in Latin (nul, nunquam) can co-occur and behave as Concord elements: they are in the pre-Infl area and are able to express sentential negation by themselves, without a co-occurring NM. They yield a single-negation reading, showing that Negative Spread applies: () Et ab Ludher nul plaid nunquam prindrai qui meon uol cist meon fradre Karle in damno sit. ‘and I shall never make any pact with Lothair that would damage this brother of mine Charles by my will’ (Oaths of Strasbourg )

3 When not otherwise specified, I take my data from the editions used in the MCVF and in the SRCMF corpora (see §.).

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



Further on in the text, we find the NM non (also nun) and the correlative negation ne . . . ne < Lat. n˘ec. () si io returnar non l’int pois, ne io, ne neuls cui eo returnar int pois, in nulla aiudha contra Lodhuuuig nun li iu er. ‘if I cannot dissuade him from it [from breaking the oath], neither I nor anyone that I can convince, then I shall not help him in any way against Louis’ (Oaths of Strasbourg –) In this passage, the indefinite alternates between a form neul and a form nul (cf. the feminine nulla). While the latter goes back directly to the Latin NI nullus, the form neul is possibly re-segmented (or newly grammaticalized?) as ne-ull-, with a more transparent initial negative morpheme, deriving from Latin n˘ec. As mentioned above, the combination nec ullus was quite frequent in Latin. The alternation nul-neul (neül) is traditionally treated as conditioned by the metrical form of the verse (the form neul providing an additional syllable). However, given what we will see in this chapter on the role of n˘ec, I think that it is plausible that the reanalysis of nul from a [Neg] to a [uNeg] element went through a morphological reinterpretation. Possibly, therefore, the fact that Latin nullus is more successful than the other Latin NIs in its transmission to Early Romance could be attributed to its more transparent morphological structure, which allowed for analogical reanalysis. Consider also the fact that the inanimate pronoun used at this archaic stage is niënt ‘nothing’, also formed with n˘ec (cf. §. for further etymological remarks). Also the correlative particle ne(d) (later ni) seen in () derives from Latin n˘ec: as we will extensively discuss in the rest of the chapter, this form was semantically negative ([Neg]) in Latin; here, instead, it co-occurs with the negatively marked indefinite and with the NM in a single-negation reading. This is systematic in OF. An example from the Vie de Saint Alexis (mid eleventh century), showing the co-occurrence of the correlative negation and the standard sentential NM, is provided below: () Nel reconurent ne ne l’unt anterciét ‘They did not recognize him and they did not ask him’ (Alexis ) Foulet (: §) remarks that, if the particle is used to correlate two clauses, the NM ne always co-occurs (analogously to what happens in Modern French with ni).4 He also comments on the fact that the ensuing combination ne ne is very frequent in Old French (a fact that one could take as a form of reinforcement, given the emphatic flavor typically resulting from correlation). Interestingly, Foulet (: §) reports cases, impossible in Modern French, where the first conjunct of the correlation is positive: () Qu’ele li dira ausi tost, / ne ja ne celera tel chose ‘she will tell her immediately, and will not hide such a thing’ (Vergy –)

4 In archaic texts the correlative particle still has the form ne and is thus graphically indistinguishable from the NM ne. The latter is always elided before a vowel, the correlative particle only optionally (Foulet : §).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

As we will see in §., this ‘polarity switch’ use was also possible in Latin, as well as in Old Italian. The following example shows a passage from the Vie de Saint Alexis where a correlative negation, a negatively marked indefinite, and an NM co-occur in the preInfl field: () nel reconut nuls sons apartenanz / ne n[e]üls hom ne sout les sons ahanz ‘No one recognizes his lineage, nor does any man know his suffering’ (Alexis –) The first line of () shows a case of Negative Concord between the NM and a post-Infl subject. Also the Cantilène de sainte Eulalie, one of the oldest OF texts (c. ce, Northern France), already shows a strict NC structure, despite the overall archaic nature of the text (e.g., the NM is always non, never the reduced form ne): () Niule cose non la pouret omque pleier ‘nothing could ever subdue her’ (Eulalie ) In (), the indefinite is niul, another form of nul, again possibly the result of a re-segmentation involving the continuation of Latin n˘ec. It occurs as a pre-Infl subject and co-occurs with the NM. In post-Infl position omque < Latin umquam ‘ever’ is found, on which see further in this section. The crucial piece of evidence to distinguish between a Concord and a polarity use is represented exactly by those cases where nul is a subject which precedes and cooccurs with the marker of negation. A further example from the Voyage de St. Brendan (twelfth century) is given in (): () U nuls n’entret fors sul li piu ‘where none but the pious can arrive’ (Brendan ) The fact that NPIs have to be linearly preceded by an overt c-commanding operator, and thus typically do not occur in sentence-initial subject position, makes an NPIanalysis incompatible with the evidence in () and (). The examples rather point to a Concord analysis. The pattern instantiated in () is a strict NC pattern, parallel to the Modern French one.5 Examples like (), where the indefinite occurs in object function, are in principle compatible with an NPI analysis, but the fact that the negatively marked indefinite

5 Ingham (a: –) mentions a few cases in his Anglo-Norman corpus where a non-strict NC pattern is observed, with a pre-Infl indefinite negating the clause (and licensing an NPI) without cooccurring with ne: (i) Item que nul dilivera aucun overaigne ‘Next, that no one shall deliver any piece of work’ (York Memorandum Book i , c., from Ingham : ) Ingham remarks that this sort of ne-‘dropping’ was also found in earlier Anglo-Norman texts and is plausibly influenced by the contemporary loss of NC / ne-dropping in English.

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



is also possible in a position that linearly precedes the NM is an indication for the Concord nature of the indefinite at this point. () Mais n’ i truvent nul’entrethe ‘but they found no place’ (Brendan ) However, nul can also have NPI uses, as shown by cases where nul can mean ‘any’. The following example has it in the standard of comparison (cf. Labelle and Espinal , Larrivée a for more examples): () Plus vos amai que nule creature ‘I loved you more than any creature’ (Alexis ) Correlative negation too can be used as an NPI with a ‘positive’ meaning, cf. Foulet (: §), Doetjes (). The example below shows ne in the polarity-sensitive context of a wh-question: () Dont estes vos ne que querez? (Doetjes : ) ‘Where are you from and what are you looking for?’ (Fabliaux ., thirteenth century) Although the strict NC grammar is remarkably systematic in OF, some exceptions occur. In the Cantilène, a couple of lines after the strict NC example in (), a construction occurs where a negative adverb (nonque < Lat. numquam ‘never’) appears in the pre-Infl area without being doubled by a NM: () dont lei nonque chielt ‘about which she never cared’ (Eulalie ) In principle, at least two explanations are possible for this difference: either the author alternates between a strict and a non-strict grammar, or the adverb nonque has not been reanalyzed in its featural content and still behaves like a Latin semantically negative [Neg] element. In this it would contrast with omque / unc(h)es, which derives from the Latin NPI umquam and which behaves like a [uNeg] n-word in OF, positioned in the pre-Infl (i.e., pre-NegP) area and co-occurring with the NM: () Unches en Rome nen out si grant ledice ‘Never in Rome there had been such a great joy’ (Alexis ) The NM, as we saw, is non in the most archaic texts, ne afterward.6 The weakened form occupies a stable preverbal position and may be separated from the verb only by pronominal clitics, which often phonologically merge with the NM (cf. e.g., nel = ne-le in ). The fact that ne had already been critically weakened in OF and behaved

6 Only in later texts the alternation between non and ne is governed by the pragmatic factors indicated by Foulet (: §). Foulet describes the OF use of non (which disappeared at later stages) as corrective: it appears in contexts where a prejacent proposition is denied. A contrastive strong affirmation, emphasizing the adversative positive value against a prejacent negative assertion (cf. German doch), is introduced by si. Its elliptical use is still attested nowadays. However, in OF, si was not necessarily adversative, and could occur also as a confirmation of a preceding positive assertion (Foulet : §).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

like a clitic is evidenced also by the formal reinforcement it undergoes when it is used as a negative short answer: in that context ne(n) combines with a pronoun: nen-il, ne-jo ‘not-he’, ‘not-I’ > nenni ‘nay!’ (cf. Posner : ).7 However, in the earliest texts, reinforcement by means of minimizers or generalizers is extremely rare. There is not one single case in the Vie de Saint Alexis.8 In Modern French nul is used only in literary and juridical language (Riegel et al. : ). It is always singular and animate, and can be used as a pronoun as well as an adjectival determiner (agreeing in gender): () a. Nul n’est censé ignorer la loi ‘No one is supposed to ignore the law’ (Riegel et al. : ) b. Il ne l’a trouvé nulle part ‘He hasn’t found it anywhere’ In §.. we have already discussed the replacement of nul by aucun, which loses its epistemic function and many of its NPI uses and becomes specialized as narrow-scope indefinite in the scope of negation. This process conforms to the fact that the second cycle of renewal of indefinites in OF uniformly involves elements with a positive origin and with an original NPI-distribution (cf. Labelle and Espinal  for a description of this second cycle). According to the traditional reconstruction (Haspelmath : –, Catalani : –, Willis : ) the archaic indefinites originally had negative force and subsequently, owing to analogical pressure, assimilated in their use to items such as rien and aucun, which occurred without negative force in negative-polarity environments. We saw, however, that since the earliest stages nul has ‘negative force’ only inasmuch as it constantly co-occurs with the NM ne. As already stated by Foulet (: §–), nul is not able to negate a clause by itself; in negative contexts it is always accompanied by ne. The latter can occur as the only marker of sentential negation in these early texts, and it is therefore uncontroversially ‘negative’, i.e., endowed with a formal feature for negation: () n’ourent amfant ‘They did not have children’ (Alexis ) It is therefore more precise to say that nul was a [uNeg], Concord element. The open question is whether this formal feature derives directly from the reanalysis of Lat.

7 Cf. Ingham () for a detailed evaluation of the prosodic properties of OF ne. 8 Later Old French texts express emphatic negation by means of bare nouns (mie, point, goutte, pas), with various degrees of grammaticalization on the Jespersen path. Eckardt () shows that already in OF pas has the distribution of a strong NPI, occurring only in direct-negation contexts, whereas the other elements have the distribution of weak NPIs and are found in more broadly downward-entailing contexts.

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



nullus or indirectly, as I have tentatively suggested above, from the reanalysis of Latin n˘ec which is a pan-Romance phenomenon, dating back to Late Latin. Moving to NPI uses, in the early texts I have analyzed they are quite rare for nul. The later French texts (private correspondence) investigated in Ingham and Kallel () show that the presence of nul in polarity contexts is not indifferent: it is found in  of the relevant polarity contexts containing indefinites in the first period (–) and  in the second (–), with a mere  in the third (–). The corpus of legal texts from Northern France investigated by Larrivée (a) provides similar results: during the time span from  to the sixteenth century,  of the occurrences of nul show an NPI use. Despite the competition with aucun, nul recedes quite slowly, especially from negative contexts: in Ingham and Kallel’s corpus it is still found in  of the relevant negative clauses in the first period, in  in the second period, and still in  in the third period, accounting for  of the overall attestations of nul. Foulet (: §) paraphases nul in polarity contexts as n’importe qui, quiconque, qui que ce soit, i.e., he treats it as a free-choice item, something that we can interpret as the perception of a domain widening effect. A closer look at the environments where it occurs shows that the ‘positive’ uses of nul are found in typical polarity contexts, such as comparatives, antecedents of conditionals, questions (cf. ). For this reason Ingham (a: ) classifies nul as a ‘non-assertive polarity item, rather than negative indefinite’. () a. Cuidiez vous, se me disiiez / votre conseil celeement / que jel deïsse a nule gent ‘Do you believe that if you secretly told me your recommendation I would tell to anyone?’ (La Chastelaine de Vergi –, from Foulet ) b. Et s’il y trouve nul ne son mestre ne autre ‘And if he finds anyone, either his master or another’ (Règlements des Arts et des Métiers , s, from Ingham a: ) c. Se il y a frere ne sereur qui soit encuses de nul meffait, . . . ‘If there is a brother or a sister who is accused of any misdeed’ (document from Amiens , year , from Ingham a: ) NPI uses are particularly puzzling in the case of nul because of its origin: since Latin nullus was a negative indefinite, in a Double Negation system, for a continuation of nullus we would expect a reanalysis as ‘pure’ n-word (Concord element), since the Latin evidence presented it exclusively in negative contexts. That is, the polarity uses of OF can hardly qualify as remains of a previous stage, since the origin of nul is clearly not an NPI. The analysis of n˘ec that I will propose in this chapter may offer a new perspective on this problem: differently from Latin NIs, new nec-words of Romance originate as emphatic reinforcers of negation and have, thus, a clear connection with focus. This may explain why all nec-words are found in NPI uses in Early Romance, as we will discuss in §.. Given the observations above on the possible reanalysis of

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

the negative morpheme in nul, a scenario where it analogically assimilates to the new nec-words is plausible. .. Old Italian Negative Concord in Old Italian has been the subject of renewed attention in recent years (Ledgeway , Zanuttini , Garzonio and Poletto , Parry , Poletto , Franco et al. , a, Franco and Poletto , Kellert ). Many finegrained generalizations have been discovered that make sense of a very fluid situation, by remarkably restricting the role previously ascribed to optionality. I will contribute to this line of research by pointing to some aspects of the syntax of the new indefinites in the Old Italian negation systems that can be explained through a better understanding of the licensing conditions of n-words and polarity items at this stage. I will argue that, once we understand these factors better, it is not necessary to assume a shift from (optional) strict Negative Concord (cf. Martins , Zanuttini : ) to non-strict Negative Concord for Old Italian, in order to explain uses of Concord indefinites accompanied by the negative marker in the pre-Infl field of the clause. As customary, I designate as Old Italian (OI) what in fact is the Old Tuscan variety and I restrict my attention to documents from this area. It is well known that especially northern Italian dialects appear to show remarkable differences in their diachronic development, cf. Parry (). In Old Tuscan the non-strict NC grammar seems to be established by the fourteenth century (Posner : , Franco et al. a). Before that period, cases are found where the OI grammar differs from the Modern Italian grammar in some important respects: (i) some examples are found in which there is no NC: that is, in some cases post-Infl material is sufficient to express sentential negation, with no further marking in the pre-Infl area; (ii) in the pre-Infl area, optionality between a strict and a non-strict system is observed: the texts document the possibility of redundancy in the expression of negation in the pre-Infl area; (iii) n-words occur with more productivity (in a broader number of contexts and with higher frequency) than in Modern Italian in non-negative downwardentailing contexts: in other words, they can negate a clause by themselves (cf. (i) and (ii)), but they can also behave as NPIs. These phenomena have been subsumed in the literature under the heading ‘optional NC’ (non-strict NC with both pre- and postverbal neg-words). In order to evaluate the incidence of optional NC, I conducted a small corpus study over selected texts contained in the OVI corpus (cf. §.). I searched for the various forms of the negatively marked indefinites, which I list in the table in ().

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



() Forms of negatively marked indefinites in Old Italian9 Lemma Form niuno ‘no one’ niuno, niuna, niun(’), neuno, neuna niente ‘nothing’ niente, neente nullo ‘no’ nullo, nulla, nul In the various texts I analyzed, I could observe a homogeneous behavior for all these elements, showing that they already behave like a series, despite their heterogeneous origin. While niuno and niente are ‘new’ indefinites formed with the negative particle n˘ec, nullo is a direct continuation of Latin nullus, i.e., of an originally [Neg] element, as in Old French.10 OI nullo is mainly used adjectivally (Rohlfs : §), as in the following example from the Laudes Creaturarum (Umbrian): () de la quale nullu homo vivente po skappare ‘from which no living person can escape’ (Laudes Creaturarum ) It survives in this use in the Southern dialects, whereas in Modern Italian it is productive only in the predicative use (a and b), besides appearing in some fixed expressions (c) and in its nominalized form (il) nulla ‘nothing’ (d):11 () Modern Italian a. L’interesse della società per questo problema è nullo ‘Society’s interest for this problem is nonexistent’ b. I vantaggi sono nulli ‘The advantages are nonexistent’ c. Nelle ultime votazioni ci sono state molte schede nulle ‘In the last polls there have been many spoiled votes’ d. Nulla può fermarti ‘Nothing can stop you’

9 Note that the lemmatization given in the table does not necessarily correspond to the one used in the OVI corpus. 10 A further direct continuation of Latin, which is not found in my selected texts but appears  times in the OVI corpus, is the continuation of Latin nemo ‘no one’, i.e., the archaic nemo (four instances, all of them in expressions with Latin syntax and lexicon) / nimo ( instances). With respect to its Latin predecessor, nimo is reanalyzed in its featural composition and behaves like other Old Italian n-words: it occurs in the great majority of the instances in non-strict NC configurations; typically, nimo is found preverbally and does not co-occur with the NM; if postverbal (nine cases) it co-occurs with preverbal negation. It can also co-occur with another n-word (nimo puote sapere nulla di suo essere, ‘Nobody can know anything about his being (in a place)’, Palamedés pis., c.). A few NPI uses are also found, in questions and conditionals. 11 Pronominal nulla is usually thought to originate from Latin nulla res ‘no thing’. According to Rohlfs (: ), pronominal nulla is more widespread in the northern Tuscan varieties, whereas niente prevails in the eastern and southern ones.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The Modern Italian examples show that in the predicative use there is no NC. The same can be said for OI: lack of Concord is systematic where post-Infl niente and nulla are used to denote a minimal quantity or to mean ‘nothingness’ (Zanuttini : ): () e fede sanza opera, ovvero opera sanza fede, è neente a potere avere paradiso ‘and faith without works, or works without faith, is no use to reach heaven’ (Bono Giamboni, Libro ., from Zanuttini : ) Poletto (: ch. ) has shown that at no stage of OI is a difference between the featural content of ‘bare’ n-words and n-words combined with an overt nominal restriction observed. Therefore, I did not differentiate between pronominal and determiner uses. Note that, in view of this equivalence, a solution such as the one proposed by Déprez () for Old French, according to which ‘bare’ n-words are structurally lower than determiner-like n-words, does not seem to be viable for Italian. This is arguably connected to the fact that Italian argumental n-words never have a positive origin, differently from French personne, rien; they always feature an overt negative marking, in the form of the prefix ne- / ni- < Latin n˘ec.12 In my categorization, I distinguished among four types of contexts, taking as my main criterion the comparison between OI and Modern Italian: I tried to single out cases where OI behaves differently from Modern Italian. The criteria are listed in (). () a. non-strict Negative Concord (i.e., cases where OI behaves like Modern Italian), with three subcases: (i) indefinite is (the only negatively marked element) pre-Infl; (ii) indefinite is post-Infl;13 (iii) indefinite occurs after sanza / senza ‘without’;14 b. lack of Concord when indefinite is post-Infl (pre-Infl cases are naturally subsumed under a.(i)); c. pre-Infl Concord (i.e. cases of strict Negative Concord); d. cases where there is no co-occurrence with the negative marker and the negatively marked element does not have a negative meaning, i.e., cases where the negatively marked indefinite behaves as a weak NPI. The table in () shows the results for Brunetto Latini’s Rettorica, a re-elaboration (translation and commentary) of Cicero’s De Inventione dating to c..

12 Things are different with adverbials, the most significant case being that of mai ‘(n)ever’ < magis ‘more’, which indeed still nowadays has a special distribution: see Panizza and Romoli (). 13 Of course these structures are also compatible with a strict NC system; my categorization here is dependent on the way I formulate my research question: how different is the Old Italian system from the Modern Italian one? Basically, post-Infl indefinites co-occurring with a pre-Infl NM are no deciding evidence for the subtype of NC system. 14 In fact, this subcase could as well be classified under (c), i.e. it could also appear in a strict Negative Concord system: I consider it here just because it is a case where OI behaves like Modern Italian, and I count it in order to offer an exhaustive classification of all occurrences in a given text, and thus facilitate replication of the query.

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



() Negatively marked indefinites in Brunetto Latini, Rettorica15 a

b

c

d

Non-strict NC

no post-Infl Concord

pre-Infl Concord

NPI







preInfl

postInfl

‘senza’







TOT



In the first column we have those patterns that are equivalent to the Modern Italian grammar: as we see, all but one structure in the Rettorica conform to what we would expect in Modern Italian. The one case I analyze as strict NC () has a complex syntax.16 The passage of Cicero’s De inventione it translates does not contain negative elements, but helps us retrieve the sense, pointing to a negative (= single-negation) reading (some background from the context: Hermagoras offers some rhetorical precepts that according to Cicero do not really belong to rhetorics but rather to philosophy; but one cannot really say that Hermagoras was a philosopher; so the precepts are just really faulty and one has to conclude that Hermagoras was a bad rhetor): () a. Nunc vero ea vis est in homine, ut ei multo rhetoricam citius quis ademerit quam philosophiam concesserit ‘But as a matter of fact the man’s ability is such that one will more readily deny him the power of rhetoric than grant him acquaintance with philosophy’ (Cic. inv. .) b. «Ma ora è quella forza nell’uomo», cioè tal fue questo Ermagoras, che neuno che dicesse ch’e’ non sappia rettorica no·lli concederae che ssia filosofo. (Brunetto Latini, Rett. p.  l. –) ‘But the man’s ability is such—that is, such was this Hermagoras—that no one who would say that he cannot do rhetorics would accept that he is a philosopher’ In this example, thus, we have a single-negation reading despite the double pre-Infl marking. The longer text Novellino provides us with more contexts: the distribution is summarized in the table in ():

15 The preposition ‘without’ is quite consistently realized as sanza in the manuscript tradition. 16 And is altered in the  print edition: che non fu veruno che dicesse ch’elli non sappia retorica non dirà già che egli sia philosopho.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

() Negatively marked indefinites in Il Novellino a

b

c

d

Non-strict NC

no post-Infl Concord

pre-Infl Concord

NPI







preInfl

postInfl

‘senza’







TOT



Also in the Novellino the majority of cases conform to the expectations of the modern, non-strict grammar; at the same time, more variation is observed. The two cases of lack of Concord in post-Infl position involve negatively marked indefinites inside PPs, where the negation does not take sentential scope, i.e., does not negate the predicate: () a. sì che rimase a neente ‘so that he was left with nothing’ (Novellino ..) b. vedete quanti sono dugento marchi, che li avete così per neente! ‘see how much two hundred marks is, you who treat them as nothing’ (Novellino ..) This conforms to an observation that has been made before for many of the ‘no Concord’ cases in OI by Poletto () and Franco et al. (a), which have been shown to represent no exception to the Negative Concord grammar, since they do not convey sentential negation. The cases of strict Negative Concord (c), with co-occurrence of the indefinite and the negative marker, are given below: () a. e comandò ai baroni che neuno non li insegnasse spendere questo oro ‘and he ordered the barons that nobody teach him to spend this gold’ (Novellino ..) b. non donai a chi non mi insegnoe; né a neuno donai, ma ciò ch’io feci fu guiderdone, e non dono ‘I did not gift those who did not teach me; and I did not gift anyone, rather what I gave was a reward and not a gift’ (Novellino ..) c. . . . e che neuno uomo non sapea che ne fosse adivenuto ‘and that no man knew what had happened of him’ (Novellino ..) d. né nul fu Mei-di-donna chi fu ’n dietro du’an ‘And nobody [excuses me for it], Miels de domna (‘Better than a lady’) that I have fled for two years’ (Novellino ..; translation D. A. Murray) The strict Concord cases are a challenge for our typology of negation because the difference between strict and non-strict systems is captured by assuming a different

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



feature specification for the NM: [iNeg] in non-strict systems, [uNeg] in strict ones (cf. §.). They would then force us to assume an ambiguous status for the Old Italian NM, a most unwelcome move, since it would disrupt the continuity we see from Late Latin [iNeg] n¯on to Modern Italian [iNeg] non. When, after the discussion of nec-words, I come back to these issues in §., I will propose that a modification of our typology, along the lines that were sketched in §.., could provide a key to understanding these phenomena while safeguarding the continuity between Latin and Italian. For the moment, let us have a closer look at the data in (). These structures seem quite heterogeneous. The example in (d) is not helpful, since it comes from a poetic insert in a short story, which derives from a poem of the troubadour Rigaut de Berbezilh. The example in (a) occurs in a subordinate clause of a verb of command, which could be argued to license an expletive negation. The example in (c), instead, shows the typical pattern in which strict NC occurs: a constituent containing an n-word precedes the NM. The constituent is here a subject, but in other examples it can be an adverbial PP or another adverbial element. This is the most systematic pattern for the redundant expression of negation in the pre-Infl field: despite its low frequency, it seems to be telling with respect to what is going on in OI. I will come back to it in §.. For now let me point out that adverbials seem to favor the strict pattern in OI. Garzonio and Poletto (), Poletto (: ch. ) have shown that there is a consistent asymmetry between adverbial and argumental elements: only argumental n-words show optional Concord, whereas adverbs like mai ‘never’, mica ‘at all’ or adverbially used niente ‘nothing’ always co-occur with a further marker of negation when they are pre-Infl. Garzonio and Poletto () and Poletto () conclude that NC in OI is phasedependent and is activated only where elements are in the same phase; since bare adverbs do not build their own phase, when pre-Infl they always belong to the same phase as the NM and thus undergo Agree with it. While retaining Garzonio and Poletto’s original intuition that being in the CP–TP phase is the decisive factor for explaining strict NC patterns in OI, in §. I will tentatively put forward a different explanation for this distribution, connecting it to the interplay with other operators in the CP, in particular Focus operators. Example (b) is different: besides having a clear emphatic flavor (it is a polemic reaction refuting a previous statement), it features an n-word following the negative correlative né. This element is well known for causing ‘pleonastic’ expression of negation in OI. Zanuttini (: –) notes the ambiguous status of né / nè in OI: when preverbal, né can either be the only marker of negation, or co-occur with a (lower) non. This behavior parallels the ‘optional Negative Concord’ observed with negatively marked indefinites in the pre-Infl field. Zanuttini, however, remarks that the frequency of the two phenomena is different, the co-occurrence of né with non being much more frequent (the normal case, in fact) than the combination with the indefinite. According to Zanuttini (: ), in Old Italian the conjunct introduced by né / nè is always negative, unlike other Medieval varieties, like Old French, where— as we saw in §..—the correlative particle can also be used as an NPI. But in fact some rare cases are found where né occurs in an NPI context with no negative

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

import: in the following example no negation is involved and né occurs in a beforeclause (an NPI environment): () ma consideri che’l savio mette alla bilancia le sue parole tutto avanti che lle metta in dire né inn iscritta ‘he should remember that the wise man weighs his words before he commits them to speech or writing’ (Brunetto Latini, Rett. p.  l. –) Interestingly, Molinelli () notes how Old Italian né may have the role of a discourse-structuring particle (‘segnale discorsivo’), rather than a simple coordinating particle, in that it may anticipate a negative context or establish a cohesion relation with the previous context, demarcating a transition. We will see that this was an important function also for Latin n˘ec. In these (less frequent) cases, né can perform a ‘polarity switch’, i.e., correlate a negative conjunct with a preceding positive one: () e perció in mezzo della via l’uccise; né Catone non avea podere di difenderlo ‘and thus he killed him in the middle of the road; and Catone did not have the capacity to defend him’ (Brunetto Latini, Rett. p.  l. –) Given the assumption that NPIs have to be c-commanded at PF by their licensor, which is not the case in (), the example suggests that né behaves here not like an NPI, but like a [uNeg] Concord word in being able to introduce a negative operator by itself. On the other hand, the example also shows that né can co-occur with non in a single-negation reading when it is pre-Infl. A further example of this is given below: () perciò che’l consolato non si difende né non allega ragioni contra il consolatore ‘since the person being consoled does not defend himself and does not raise arguments against the person consoling’ (Brunetto Latini, Rett. p.  l. –) This is another case where we seem to have strict NC in OI, with the ‘additional’ negative element agreeing with the lower NM, something not expected under a theory that interprets the NM as an [iNeg] element. Further support for the hypothesis that né behaves like an n-word comes from cases where it is the only negative element of a clause. This shows that né does not necessarily co-occur with non when pre-Infl and introducing a clause; it also manifests a difference with respect to Old French, where the co-occurrence with a lower NM is obligatory (probably connected to the different structural status of the NM in Old French). () nessuno avea connosciuti certi figliuoli, né aveano pensato che utilitade fosse mantenere ragione et agguallianza ‘No one knew for sure their biological children, nor did they think that it would have been useful to maintain reason and equality’ (Brunetto Latini, Rett. p.  l. –) Let me conclude the discussion of the OI data by going back to the table in (). The polarity contexts considered in the fourth column are conditional clauses, superordinate negation (negation in the matrix clause and indefinite in the subordinate),

OUP CORRECTED PROOF – FINAL, //, SPi

The situation in Early Romance



yes–no questions, standard of comparison (cf. the rich exemplification in Zanuttini : –). An example of NPI use in the protasis of a conditional clause is provided below: () e desto con la mano subitamente corsi a cercarmi il lato se niente v’avessi ‘And immediately I went with my hand to check my side if I had anything there’ (Boccaccio, Decameron IV...) Frequent instances are found in yes–no questions, an environment where also Modern Italian knows NPI-uses of n-words: () hai tu mai testimonianza niuna falsa detta contra alcuno . . . ? ‘have you ever given any false witness against someone?’ (Boccaccio, Decameron I...) Modern Italian nessuno, despite being (indisputably) a Concord element under negation, still retains NPI uses in some residual environments such as yes–no questions: () Hai visto nessuno all’entrata? ‘Have you seen anyone at the entrance?’

Modern Italian

This behavior is admittedly poorly understood: often the polarity uses of nessuno have been interpreted as historical remains of a previous stage in which nessuno was an NPI, thus plausibly leading to a scenario where two different lexical entries have to be assumed. However, we see that, since the beginning of attestation, nessuno and the other n-words are able to negate by themselves, in contexts where NPIs are not licensed. The further discussion of the Latin data will lay the basis for a reintepretation of the observed diachronic path. An important test to distinguish between an NPI and an n-word has to do with the behavior of indefinites in negative short answers: given their ‘self-licensing’ ability, nwords can occur by themselves as negative short answers, whereas NPIs cannot: the elliptical structure that contains them also must include a further marker of negation. In OI negatively marked indefinites can occur in isolation as negative short answers, confirming the n-word analysis: () ella: ‘È dunque alcuno il quale riputi gli uomini poter tutte le cose?’ E io: ‘Niuno, se non chi impazzisca’ ‘she asked: ‘Is there anyone who believes that men can do everything?’ And I: ‘No one, if not someone who becomes insane’ (Alberto della Piagentina ., ) Interestingly, they also occur quite frequently in elliptical structures like the following, where they are accompanied by the NM non: () Che è ciò, Marco, ch’io ho avute sette robe e tu non niuna? ‘What is happening, Marco, that I had seven garments, and you no one?’ (Novellino, ..) Again, we have two negatively marked elements and a single logical negation. Since words like niuna can also occur in isolation as negative answers, the hypothesis that

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

here niuna behaves like an NPI and needs a co-occurring NM for licensing does not seem to be on the right track. Under an ellipsis analysis of (negative) short answers we could assume that both these elements are in a Focus Phrase, and the rest of the structure is subject to ellipsis. This structural analysis would force us to assume that the n-word and the NM agree in the pre-Infl field, i.e., that also this case is an instance of strict NC. The Modern Italian structure corresponding to () supports this analysis, since it shows a sharp difference, plausibly connected to the impossibility of strict NC (cf. further §..): in () the co-occurrence between NM and n-word would be unacceptable, as usual, in pre-Infl position: () Perché io ho avuto sette vesti e tu nessuna? ‘Why did I receive seven garments and you none?’

Modern Italian

.. Conclusions The survey of the Old French and Old Italian data conducted in this section was aimed at formulating some important points that will motivate my investigation of the Latin data. Romance varieties are Negative Concord languages from the start. The newly grammaticalized and the innovatively reanalyzed indefinites interacting with negation behave as Concord elements: they can both co-occur with negation and express negation by themselves. They are the ‘plain’ way of realizing existential quantification under the scope of negation: we never had to observe particular emphasis connected to their use in the analyzed material. These indefinites also show NPI uses, which are puzzling since their etymological origin is clearly negative (for Old French, we saw that this is true for the most archaic system, comprising the n-words nul and nient). The two Romance varieties described here have many characteristics in common (development of Negative Concord, grammaticalization of new indefinites belonging to negation systems according to similar strategies, NPI uses of such indefinites), but also show remarkable differences: Old French is a strict Negative Concord system since the earliest attestations; Old Italian is a predominantly non-strict Negative Concord system, which also displays some distributional patterns that differ from the modern grammar. We saw that the most intriguing one in a theoretical perspective is represented by the strict Negative Concord patterns that appear in a minority of cases. With these distributional generalizations in mind, let us move to investigate in depth the origin of the new Romance Concord elements. As previously announced, I will deal with the subclass of n-words that, as we saw in this section, are the most widespread in the earliest stages: the nec-words, whose negative morpheme derives from Latin n˘ec.

. The grammaticalization of nec-words .. The issue The Romance languages display great lexical variation in the realization of narrowscope indefinites under negation. This fact attests to a situation of discontinuous

OUP CORRECTED PROOF – FINAL, //, SPi

The grammaticalization of nec-words



transmission in this grammatical domain, as we have already discussed in §... There are, however, also important commonalities among Romance languages. The most fundamental one is the fact that the new narrow-scope indefinites of Early Romance invariably end up being elements of Negative Concord (that is, [uNeg] indefinites). A further shared innovation consists in the grammaticalization of indefinites formed by means of the Latin negative particle n˘ec and the cardinal numeral unus ‘one’. This pattern shows up everywhere in the Romance area (cf. REW : § and Posner  for an overview). Examples for what I will call nec-words (in order to remain neutral as to their analysis, for the moment) are given in ().17, 18 () Romance indefinite pronoun ‘nobody’ Romance nec-words for ‘nobody’ Portuguese Spanish Old French Old Occitan Provençal Old Catalan Italian Sardinian Romanian

nenhum (Old Portuguese n˜eu˜ u, n˜eg˜uu) ninguno (Old Spanish also niguno) neuns, necun, negun, nesun, nessuns negu (cf. Modern Occitan degu) neisun ningú nessuno (Old Italian also neuno, niuno, negun) nesciunu, niunu nic˘ı un

In the rest of this chapter I will formulate and support the hypothesis that there is a deep connection between the two commonalities just mentioned, the rise of Concord indefinites and the presence of n˘ec as a morphological element in many of them from the earliest stages. Namely, I will propose that the Latin negative particle n˘ec is one of the first elements to be reanalyzed as a [uNeg] element of Negative Concord, after the changes in the system of Latin negation discussed in §.. In chapter  I have argued that the prerequisites for the change from a Double Negation to a Negative Concord system obtain already at the Late Latin stage, when the negative marker n¯on is reanalyzed as an [iNeg] element. Negative Concord itself, however, was not visible yet owing to the absence of [uNeg] indefinites. In the subsequent development from Latin to Romance we see the rise of new indefinites taking narrow scope with respect to the negative operator. In nec-words the narrowscope requirement is overtly signaled by the negative morpheme, conforming to

17 Not only indefinite determiners and pronouns are formed by means of n˘ec: the morpheme is found also in adverbials, typically with focusing functions, such as It. nemmeno, neppure, neanche, meaning ‘neither’, ‘not even’. On the evolution of neanche cf. Franco et al. (, b). Garzonio () discusses the case of Old Italian né mica lit. ‘not a crumb’, where the morpheme derived from n˘ec gets lost in the course of the diachronic development leading to Modern Italian mica. 18 Portuguese also has the pronoun ninguém ‘no one’, formed by prefixing n˘ec to the wh-form quem ‘who(m)’.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

the crosslinguistically robust tendency ‘to ‘close off ’ or ‘encapsulate’ weak indefinites when occurring in the scope of clausal negation’ (Weiß b: ) mentioned in §... From the point of view of their structure, these new Romance indefinites reproduce the pattern of the Latin NIs, where the initial negative element was *ne. However, while the latter can be reconstructed also as the marker of plain sentential negation at a Proto-Latin stage (cf. §..), the particle n˘ec had a different status: it was not the plain negative marker, but a correlative negation.19 The question I want to investigate is whether the fact that the new indefinites are Concord elements from the start (§.) depends on the status of the negative morpheme itself, in terms of both its semantics and its formal features. There are two main aspects in the behavior of n˘ec that lead me to propose a decisive role for this element. The first aspect concerns the intrinsic ability of correlative negation to take part in doubling constructions. In many languages correlative negations show a special behavior with respect to their possibility of co-occurring with other negative elements. Jespersen (: ) clearly states that it is necessary to distinguish between cases of ‘cumulative negation’ (= Negative Concord) and the different phenomenon of ‘supplementary negation’ (p. ) or ‘resumptive negation’ (p. ) (modeled after Delbrück’s () Ergänzungsnegation). He exemplifies resumptive negation a.o. by means of Latin n¯on followed by ne . . . quidem or by neque . . . neque, Ancient Greek ou followed by oudé . . . oudé, English not followed by neither . . . nor or by the ‘restrictive addition’ (p. ) not even. () Jespersen (: ) a. he cannot sleep, neither at night nor in the daytime b. he cannot sleep, not even after taking an opiate With resumptive negation, ‘after a negative sentence has been completed, something is added in a negative form with the obvious result that the negative effect is heightened’ (Jespersen : ). For Jespersen, thus, this kind of negation has the nature of an afterthought and forms a syntactically and semantically separate unit, which however is pragmatically coherent with the preceding context. Resumptive negation may then result in a doubled expression of negation in the ‘same’ sentence also in languages not otherwise exhibiting Negative Concord. Jespersen himself acknowledges that in some contexts the afterthought-nature of the supplementary negation may render the result unclear and give the impression of a cumulative negation. This could of course be important in a diachronic perspective, if we assume that, given some co-occurring facilitating conditions, these data may become relevant in triggering a reanalysis in language acquisition. In §. we will see that structures similar to the English ones 19 This is a point that, I think, Bernini and Ramat (: ) miss when they suggest that non-strict NC with n-words like nessuno could be due to the fact that the string nec ipse unus grammaticalizes in subject position: they argue this on the basis of the position of the negative element, which precedes not the verb but rather the nominal phrase functioning as subject (and thus typically preverbal). I think that they are referring to the employment of n˘ec as clausal connector. But, as we will see, the pattern from which Romance nec-words more directly originate is the use of n˘ec as focus particle, adjoined to the nominal projection and thus independent of the grammatical role of the nominal in the clause.

OUP CORRECTED PROOF – FINAL, //, SPi

The grammaticalization of nec-words



in () are found in Classical Latin and undergo important developments in Late Latin. The second aspect speaking for a crucial diachronic role of n˘ec has to do with the focusing nature of correlative negation particles. This is again a crosslinguistically attested feature of correlative particles: Haspelmath (: ) mentions a number of languages where the correlative negation particle is also used as a scalar focus particle, with the meaning ‘not even’ / ‘neither’: Polish ani, Russian ni, Spanish ni, Romanian nici, Albanian as, Modern Greek oute, Hungarian sem. () Polish (from Haspelmath : ) a. Ani mnie, ani jemu sie nie udalo neither I.dat neither he.dat refl not succeeded ‘Neither I, nor he succeeded’ b. Karliczek ani slówka mi nie powiedzial Karliczek not.even word me.dat not said ‘Karliczek didn’t even say a word to me’ In chapter  I have discussed how scalar focus is an important ingredient in the grammaticalization of indefinites in the scope of negation. I will now argue that it plays a role also in the development of Romance nec-words: the use of Latin n˘ec as a scalar focus particle is well attested in Classical Latin and expands in Late Latin; this makes n˘ec a perfect candidate to appear in originally emphatic negative constructions and to combine with end-of-scale items such as unus, eventually yielding the source for many new narrow-scope indefinites in Romance. Before I start the discussion, in the rest of this section I briefly discuss some important issues concerning the etymology of the Romance nec-words. .. Etymological remarks Besides the basic formatives n˘ec and unus, some of the nec-indefinites contain other elements, whose nature and function in the grammaticalization process are not always fully understood, thus calling for further research. The debate is still ongoing on the source of the (pro)nominal element in the inanimate form. Moreover, the negative particle has been reconstructed differently for different indefinites: as we will see below, this mirrors a discussion on the (explicitly or implicitly) assumed chronology of lexicalization. Let us start with this latter issue. ... The negative particle First of all, for some of the forms, in the REW the negative morpheme is represented as an undetermined ne, for which no Latin correspondent is explicitly indicated. More specifically, the REW has ne for the Italian inanimate form niente and cognates, as well as for ne ˘ıps’ u¯ nus (REW : §), the etymon for Italian nessuno (and dialectal variant nesuno, Old French nesun, Provençal neisun, a.o.).20, 21

20 I will discuss the role of ipse in §... 21 Quite confusingly, the negative element is reconstructed as n¯e with a long vowel in § *n¯em¯ıca ‘nothing’, the origin of Romanian nimic(˘a), Old Venetian nemiga, Friulian nemige, Engadinese nimia ‘gar nicht’.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The REW explicitly reconstructs the negative particle as n˘ec only for the direct combination with unus: n˘ec u¯ nus (§) is considered the origin of forms such as Old Venetian negun, Old French and Provençal negun, Catalan ningun, Spanish ninguno.22 Rohlfs (: §) lists further forms from Italian dialects: Old Milanese nessün / nissün, Venetian nissun, Neapolitan nUsciunU / nisciunU, etc. For Romanian nic˘ıun a separate entry with the etymon n˘eque unus is provided in the REW (§). Romanian is indeed the only daughter language to continue the form n˘eque instead of n˘ec as negative correlative particle. Concerning n˘ec itself, it is grouped together with n˘eque in the REW (§). Under this entry the REW lists a number of other Romance narrow-scope indefinites, such as Italian niuno, Old French nëun, niun (whereby for the former a grouping under § n˘ec unus is mentioned as possible), Old Portuguese nehum, cf. Modern Portuguese nenhum. Apparently, thus, the dictionary wants to distinguish two stages in the formation of these indefinites: (i) one in which the particle is still Latin n˘ec, cf. § n˘ec unus; (ii) one in which the combinations already involve the Romance correlative particles derived from Latin n˘ec: e.g., Italian né, Old French ne (older),23 ni (later), Spanish and Catalan ni, listed under §. The two patterns differ with respect to the presence of the velar element, which is attested in the direct continuations of Latin n˘ec, but not in those indefinites formed with the Romance particles, which have lost the velar. Summarizing, thus, in the REW’s system we see three main sources for the Romance n-words meaning ‘no (one)’, all featuring, in one form or another, a reflex of the Latin particle n˘ec: () a. n˘ec + u¯ nus (Romanian n˘eque unus) b. ne + u¯ nus c. ne + ipse + u¯ nus There are, however, many factors that speak against this threefold distinction and lead to the proposal of a unitary treatment for all the forms in () (as is done, for instance, in the FEW, cf. FEW ). First, despite the fact that the forms without the velar element would appear to have originated already in a completely Romance grammar, syntactically and semantically the two patterns behave alike, which points to the fact that in both cases the process of lexicalization has its roots in the restructuring of the Latin system and proceeds in parallel. In this respect, consider also the fact that other formations that nowadays appear to be language-specific were much more widespread in the Medieval period. This is 22 The latter three forms also occur in the variants degun, dengun, deguno, traditionally explained as dissimilation (Ferndissimilation), although also the possibility of the influence of Old High German dihein ‘any’ has been taken into consideration. 23 Archaic texts have also ned and net, cf. FEW () s.v. n˘ec.

OUP CORRECTED PROOF – FINAL, //, SPi

The grammaticalization of nec-words



the case, for instance, with the formation at the origin of Spanish nada ‘nothing’ < rem natam ‘born thing’ and nadie ‘no one’ < (homines) nati ‘born men’: as Rohlfs (: §) reports, Old Italian shows adjectival uses of nato ‘born’ with nouns expressing very general concepts (‘soul’, ‘woman’, ‘creature’), which remind us of the constructions from which the Spanish pronouns originate. And in fact, as Väänänen (: ) remarks, similar uses are already found in early colloquial Latin. The fact that Romance varieties witness parallel strategies in this domain supports the hypothesis that the process of lexical renewal had already started during a time of linguistic cohesion. Moreover, and most importantly for the argument, a variant of the correlative particle without the velar element is already amply attested in the manuscript tradition of the Middle Ages and especially in Merovingian Latin, probably conditioned by the presence of a following word starting with a consonant (cf. FEW : :). Recall, in this respect, that in §... we have seen how an element ne was already present in the discontinuous focusing particle ne . . . quidem (which, following Fruyt b, I consider as a vestige of the pre-Latin negative particle *ne). As we will discuss in §., Late Latin shows many signs of a functional collapse between ne . . . quidem and n˘ec, which may have contributed to the formal variability between n˘e and n˘ec. The alternation n˘e / n˘ec can be thus seen as belonging to a unitary linguistic system, where the two variants of the particle must have had the same functional distribution. In the following discussion I will therefore disregard this distinction. ... The inanimate pronoun Moving now to the etymology of the nec-pronouns used for inanimates, there is no general agreement on the source of the (pro)nominal basis.24 The REW classes the inanimate forms Italian niente, Old French nient, Modern French néant, Provençal neen, nien under ne ˇınde, following an original proposal by Ascoli. However, Meyer-Lübke expresses skepticism about this origin, since it would require assuming a borrowing from French to Italian, as well as a syntactic distribution for inde favoring adjacency with the negative particle (in order to yield *neinde) for which there is no clear empirical basis in Latin (cf. also Rohlfs : §). Another hypothesis traces these forms back to nec ente ‘no being’: the usual objection here concerns the overall rarity of the word ens (a present participle of esse ‘to be’), which was considered by Quintilian (..) a neologism formed on the model of Greek by Roman philosophers and which plausibly was never current in the colloquial language. Meyer-Lübke also dismisses the proposal ne gente ‘no people’ on the grounds that it is semantically, syntactically, and formally difficult. Rohlfs (: §), instead, 24 The determiner form does not show an animacy distinction. It can be a form with adjectival morphology identical to the morphology of the animate pronoun (e.g., Spanish animate pronoun ningún and determiner ninguno/-a, Italian animate pronoun nessuno and determiner nessuno/-a), or a separate form (e.g., French animate pronoun personne and determiner aucun; Catalan animate pronoun ningú and determiner cap; Portuguese animate pronoun ninguém and determiner nenhum; Romanian animate pronoun nimeni and determiner nici un).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

favors it, proposing to analyze gente ‘people’ as a sort of generalizer, as in the Modern Italian expressions aspettiamo gente ‘we are waiting for someone’ or c’è gente ‘there is someone’. But also an interpretation of gens as ‘kind, species’, according to a frequent Latin use, is possible (cf. FEW s.v.). The noun gens comes from the root of the Latin verb gigno ‘generate’, cf. also Greek γνο ‘class, sort, kind’. This interpretation is appealing, since it brings the meaning close to the one of res nata > Spanish nada and, most importantly, it finds a parallel in the use of gens in Catalan.25 In Modern Catalan gens is used as inanimate indefinite under negation (alternating with res, especially in partitive uses), as well as negative answer particle (alone or with repetition of the NM: no gens); its use is already in the early texts and the meaning is described in dictionaries as ‘minimal quantity or intensity’ (DIEC), cf. (): () No té gens de paciència ‘S/he does not have a bit of patience’

Catalan

In the FEW (), the etymon *ne gentem is chosen as the origin for Old French nient and its various dialectal variants, including neant (cf. Modern French néant). I find this proposal the most convincing, although the loss of the velar component remains problematic in view of regular phonological processes in Romance. It could be that here an analogical process is operating: the velar element might have been resegmented as the final consonant of n˘ec, whose presence, as we have seen above, was subject to variation (cf. Old Tuscan neuno versus Old Venetian negun, Old Portuguese n˜eu˜ u alternating with n˜eg˜uu). Consequently, it may have started to be analogically dropped also in ne gente(m), cf. Old Tuscan neente.

. Syntax and semantics of Latin n˘ec If a Romance n-word is negatively marked, the negative marking always derives from Latin n˘ec (or from its long form n˘eque, in the case of Romanian), a multifunctional particle meaning, depending on context and age of attestation, ‘furthermore not’, ‘neither . . . nor’, ‘not even’. In order to understand the grammaticalization of these necwords, it is therefore necessary to understand the role of n˘eque / n˘ec in the system of Latin negation, and the changes it undergoes. In this section I will describe the functions of this particle in Latin, basing my arguments on Orlandini and Poccetti (, ) and on my own corpus work, and I will summarize the relevant aspects of the analysis that I proposed in Gianollo (b, ). .. The form As seen in §.., n˘ec is the short form, derived by apocope of the last syllable (cf. atque > ac ‘and’), of n˘eque: the latter transparently shows the univerbation of the prehistoric negative morpheme *ne and the postpositive enclitic coordination -que < PIE *-kw e.26 Both n˘ec and n˘eque can be basically characterized as negative correlatives: 25 The FEW also compares this etymology to those of German nicht ‘not’, from wiht ‘creature, being’, and of Albanian fare ‘not’ but also ‘ancestry, species’. 26 Puristic Classical authors follow a phonotactically conditioned rule according to which the nec / neque realization, similarly to the ac / atque alternation, depends on the rightward phonological context,

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



they coordinate constituents of various categories, functioning also as sentence connectors. They always give rise to a negative interpretation for the clause or smaller constituent that they introduce. Since in Classical Latin they are functionally equivalent, what I will say for n˘ec below also applies to n˘eque, unless otherwise noted. Starting already in the post-Augustan era, the form n˘ec prevails over n˘eque, foreshadowing the later Romance developments (cf. Ernout and Thomas : , Väänänen : , Wanner : ).27 The Late Latin distribution corresponds to the situation found in Romance, where typically n˘ec, and not n˘eque, is continued (with the exception of Romanian). Löfstedt (: –) remarks that also some types of early texts already show a predominant use of n˘ec: this is the case, for instance, in the Pompeian inscriptions, where only n˘ec is found, and in Petronius, where n˘eque is never attested in the dialogic part and occurs only in the narration. A similar tendency is recognizable in Christian authors such as Tertullian and Minucius Felix. Especially in some registers (juridical writing) n˘eque survives longer. Correlative n˘ec is very pertinacious diachronically: it forms the base for Romance correlative negations: It. né, Fr. ni (Old French ne and ni), Cat. ni, with the exception of Romanian nici, which continues the full form neque. Romance languages continue it also as focus particle (e.g., French ni, Spanish ni), sometimes in a reinforced form (e.g., Italian neanche, neppure, nemmeno). Crucially for us, a further aspect of its pertinacity is its function as a building block of the new Romance nec-words. .. Functions from Classical to Late Latin In Classical Latin, n˘ec and n˘eque are very frequently used for expressing sentential negation. Very often the number of their occurrences surpasses that of n¯on (as recently confirmed by the corpus study conducted on Cicero’s Ad familiares by Inghilterra ). Elaborating on Orlandini and Poccetti’s (, ) classification, in Gianollo () I distinguish three values for Classical and Late Latin n˘ec: () Uses of Latin n˘ec: (i) discourse connector (discourse-structuring particle); (ii) correlative particle; (iii) stand-alone focus particle. I derive all of them from a fundamental meaning ∧ ¬, and I discuss how, given this meaning backbone, the various values are derived. ... Discourse-structuring particle The difference between (i) and (ii) is meant to mirror the two subtypes for the coordinative value of n˘ec discussed by Orlandini and Poccetti (, ), who, based on the observations of Orlandini (a), distinguish between a ‘connective coordination’ and a ‘copulative coordination’. In Latin, i.e., whether the following word starts respectively with a consonant or with a vowel. This rule appears to be an artificial norm (cf. Löfstedt : –), and many exceptions are found in actual use. 27 Löfstedt (: –) provides some data: there are Late Latin authors, as for instance Anthimus (sixth century), who have only n˘ec; in the Mulomedicina Chironis (fourth century) only one instance of n˘eque is found, in the Itinerarium Egeriae (fourth century) two.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

the typical realization of connective coordination in a positive context is et, which usually connects propositions, whereas copulative coordination occurs as the enclitic particle -que and is most often used to coordinate phrases (although it can rarely coordinate clauses). Besides the syntactic differences, other factors can be singled out concerning the discourse value of the two strategies: with copulative coordination, corresponding to the use in (-ii), the conjuncts are already logically connected by lexical relations of affinity or complementarity; with connective coordination, corresponding to (-i), instead, the relation between the conjuncts is contingent upon the context and established in the discourse through the use of the connector itself. In this latter use, n˘ec may be rendered as ‘furthermore . . . not’, ‘moreover . . . not’, and can also have an adversative flavor ‘and on the other hand . . . not’. Distinguishing between (i) and (ii) in the texts is not always straightforward when the two conjuncts are full-fledged clauses. However, the distinction is justified empirically and theoretically in view of clear-cut cases of discourse-structuring use like (), where n˘ec connects two clauses with different polarity values, which are also two distinct discourse units. The examples show that the first conjunct can be positive: this means that n˘ec is able to perform a polarity switch when connecting two clauses, a first piece of evidence for its intrinsically negative status:28 () a. Ex his omnibus longe sunt humanissimi qui of they:abl all:abl far are:sg most.civilized:nom who:nom Cantium incolunt, quae regio est maritima Kent:acc inhabit:pl which:nom region:nom is maritime:nom omnis, neque multum a Gallica differunt consuetudine all:nom and.not much from Gallic:abl differ:pl habit:abl ‘Of all the Britons the inhabitants of Kent, an entirely maritime district, are by far the most civilised, their lifestyle differing but little from the Gallic way of life.’ (Caes. BG .) b. Accessum est ad Britanniam omnibus navibus meridiano Get is to Britanna:acc all:abl ships:abl midday:abl loco hostis est fere tempore, neque in eo around time:abl, and.not in that:abl place:abl enemy:nom is visus. seen:pt ‘All ships got into Britain at around noon, and no enemy was spotted there’ (Caes. BG .) In this use n˘ec can be defined as a discourse-structuring particle, rather than as a simple coordinating one, in that it may anticipate a negative context (thanks to 28 A dedicated corpus study would be needed to assess for how many of the examples with a positive first conjunct we could indeed derive a negative inference. In (b) one could suspect that a negative presupposition plays a role, namely that the Romans arriving in Britain—and the readers with them— could have reasonably expected to find enemies there (and indeed, as Caesar tells us immediately after, enemies were hiding close by). There are nonetheless clear cases where no negative inference can arise from the first conjunct, witness (a). In those cases we have to conclude that n˘ec can have an intrinsically negative meaning in a conjunctive non-correlative use and is thus functionally equivalent to et n¯on ‘and not’ (cf. Torrego : ), a fact to which I will come back in the analysis.

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



the prominent clause-initial position) and establish a cohesion relation, demarcating a transition. ... Correlative particle In the correlative particle use (ii) a stronger structural symmetry between conjuncts is observed when they are represented by clausal constituents, cf. (). () nec satis exaudibam, nec sermonis fallebar and.not enough hear:sg and.not conversation:gen miss:sg.pass tamen, quae loquerentur though which:nom speak:pl.pass ‘I couldn’t hear perfectly what was being said, but I didn’t miss the general drift of their conversation’ (Plaut. Epid. –) Examples like (), where the complementizer si ‘if ’ takes scope over both conjuncts, suggest that in correlative structures the conjuncts are smaller than whole CPs. Note that also in this case a polarity switch can be observed: () Si vim faciat neque pareat, interfici iubet if force:acc make:sg and.not obey:sg, kill:inf.pass order:sg ‘he ordered that he should be put to death, if he offered force or refused to obey’ (Caes. BG .) Straightforward correlative uses of n˘ec, where the conjuncts are (at least in the surface structure) smaller than clauses, are shown in (). We find n˘ec as a coordinator in polysyndetic and monosyndetic constructions, that is, either introducing also the first conjunct(s) and being repeated in each conjunct, or introducing the last one only. In Latin, thus, the particles introducing each of the coordinated elements can be identical, as they are in the Romance languages and in contrast to many other languages (e.g., English neither . . . nor, German weder . . . noch), cf. Bernini and Ramat (: –) and Haspelmath () for a typological overview. But also different combinations, involving other negatively marked items, are possible: n˘ec and n˘eque can be repeated, or correlate with each other, or correlate with other expressions of syntactic negation in the first conjunct (e.g., n¯on . . . n˘ec / n˘eque). verum, () a. nec veri simile loquere nec and.not true:gen similar:acc tell:sg and.not true:acc frutex blockhead:voc ‘You aren’t telling a true or a likely story, you blockhead!’ (Plaut. Most. ) b. si neque leges neque mores cogunt if and.not law:nom and.not custom:nom constrain:pl ‘If neither law nor custom can constrain’ (Cic. Att. ..) c. non iudicio neque disceptatione, sed vi atque not process:abl and.not debate:abl but violence:abl and impressione aggression:abl ‘Not through a process nor a debate, but with violence and aggression’ (Cic. fam. ..)

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

Borderline cases between the use in (i) and the use in (ii) are represented by instances where n˘ec causes a polarity switch between two conjuncts with a largely parallel syntactic structure, and in addition has an adversative, contrastive flavor (showing that the correlation is not simply ‘copulative’), as in (a and b): () a. Omnia habeo, neque quicquam habeo everything:acc have:sg and.not anything:acc have:sg ‘I have everything, and nonetheless I have nothing’ (Ter. Eun. ) b. Hic nunc domi servit suo patri nec this:nom now house:loc serve:sg his:dat father:dat and.not scit pater know:sg father:nom ‘Now this one is slave to his father in his home, and his father doesn’t know it’ (Plaut. Capt. ) This employment of the correlative particle would be impossible in Modern Italian, witness the contrast below: () Italian a. * Ho tutto né ho niente b. Ho tutto e non ho niente ‘I have everything, and nonetheless I have nothing’ In fact, the uses with polarity switch, both as discourse-structuring particle and in correlative structures involving a positive first conjunct and a negative one, are impossible for all Modern Romance continuations of n˘eque / n˘ec, which can only correlate conjuncts of negative polarity (cf. Orlandini and Poccetti : , Torrego : ), and otherwise resort to the combination conjunction + NM (It. e non). Recall that in §.. we saw that in Old Italian, instead, these uses are still possible (cf. ).29 When its scope is sentential, n˘ec / n˘eque appears sentence-initially, with the conjuction semantically scoping over the negation (meaning ‘and / furthermore not’). From this position high in the CP, the particle conditions the form of the indefinites in its syntactic scope: they have to be NPIs such as quisquam or ullus (since a NI would give rise to a double-negation reading): () nec magis manufestum ego hominem umquam ullum and.not more blatant:acc I:nom man:acc ever any:acc teneri vidi catch:inf.pass see:sg ‘I’ve never seen anyone caught more red-handed’ (Plaut. Men. )

29 In Modern Italian, similar polarity-switch cases are generally out. They are only marginally possible in the written literary variety (Scorretti , Manzotti and Rigamonti ). In most cases where the first conjunct has a positive form, the meaning leads nonetheless to a negative implication (cf. Manzotti and Rigamonti : ).

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



The two markers n˘eque and n¯on may co-occur, in this order, in a sort of lexicalized litotes (‘and it is not true that not . . . ’, Ernout and Thomas : , Fruyt a: –, Fruyt : ) functionalized as anaphoric discourse marker ‘and moreover’, ‘and in addition to that’ (cf. Orlandini : , n. ; frequent in Late Latin meaning ‘as well as’). Thus, there are two underly of semantic negation operators that cancel each other out. But n˘ec can also be found in particular contexts where it gives the impression of ‘concord’ with other negative items in the clause (): () a. habeo hic neminem / neque (= aut) amicum neque have:sg here nobody:acc and.not (= or) friend:acc and.not (= aut) cognatum (= or) relative:acc ‘I have no friend here or relative’ (Ter. Eun. –, from Torrego : ) b. nemo umquam neque poeta neque orator nobody:nom ever and.not poet:nom and.not orator:nom fuit qui quemquam meliorem quam se be:sg who:nom anyone:acc better:acc than himself:acc arbitraretur think:sg ‘There has never been anyone, either poet or orator, who thought anyone to be better than himself ’ (Cic. Att. .., from Orlandini and Poccetti : ) Crucially for our further argument, in examples like (a) and (b) the correlative negation co-occurs with another negative element (the negative indefinite nemo in both examples) and yet does not yield a double-negation reading (as remarked in traditional grammatical descriptions, cf. Ernout and Thomas : ). In Early and Classical Latin examples of this sort a clear break between syntactic units is often present, and is particularly appreciable in example (a), since the negative indefinite and the negative correlation belong to two different metrical units. In (b) the correlation adds a parenthetical comment further specifying the NI (a so-called epexegetic use).30 Similar uses of negative correlation are attested in many Double Negation languages, cf. the previously mentioned examples from English in () and the German data in (): () a. Und diejenigen, die gegen den Terrorismus kämpfen wollen, können das nicht tun, weder in Syrien noch im Irak, ohne koordinierte Aktionen ‘And those who want fight against terrorism cannot do it, neither in Syria nor in Iraq, without coordinated actions’ (http://www.deutschlandfunk.de/antiis-koalition-assads-rueckkehr-in-die-internationale..de.html?dram: sarticle_id=) 30 This was marginally possible even with n¯on used as constituent negator, cf. Liv. ..: ut nemo non lingua non manu promptior in civitate haberetur ‘so that no citizen was held to be readier, whether with tongue or with hand’ (Ernout and Thomas : ).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance b. Das ist ja auch nicht gewollt. Weder von Politik noch von Wirtschaft. ‘This is also not desired. Neither by politics nor by economy’ (http://www. welt.de/wirtschaft/energie/article/Das-Ende-des-deutschenSpritpreis-Wunders.html)

In the examples from Standard English and German, we note again a sharp intonational break between the unit containing the sentential NM and the subsequent unit containing the correlative negations, so that they can be analyzed without resorting to Negative Concord. This sort of pleonastic use of negation has been called ‘Ergänzungsnegation’ in Delbrück (), and ‘resumptive negation’ in Jespersen (). In Classical and Late Classical Latin there are, however, also examples where no syntactic break can be assumed, cf. the cases where they occur in syntactically embedded clauses in () (further examples in Kühner and Stegmann : I. –, Hofmann and Szantyr : –, Orlandini : –, Orlandini and Poccetti , ): () a. non possum nec cogitare nec scribere not can:sg and.not think:inf and.not write:inf ‘It is impossible for me to think or to write’ (Cic. Att. ..) b. nihil tam tutum ad custodiam nec fieri nothing:nom so safe:nom for custody:acc and.not make:inf.pass nec cogitari potest and.not conceive:inf.pass can:sg ‘Nothing so secure for imprisonment could be either built or conceived’ (Cic. Verr. ..) c. audio enim non licere cuiquam mortalium in hear:sg indeed not be.allowed:inf anyone:dat mortal:gen in nave neque ungues neque capillos deponere nisi . . . ship:abl and.not nails:acc and.not hair:acc cut:inf if.not ‘I hear indeed that on a ship it is not permitted for any mortal to cut his nails or hair, unless . . . ’ (Petr. .) In examples like (b) and (c) we see that, besides the co-occurrence of n˘ec with a semantically negative element (a negative indefinite or the negative marker), also NPIs like umquam and quisquam are present in the structure: this shows us unambiguously that the elements belonging to the negation system work otherwise as expected in a Double Negation language, and that n˘ec represents the exception. Often examples like those in () and () are considered in the literature together with the early examples of doubling that I discussed in §... I think that it is more explanatory to treat the more frequent and systematic cases where n˘ec is involved separately from the much rarer ones where other negatively marked elements (NM and NIs) co-occur. I will present my analysis in §.. In Late Latin the use of correlative n˘ec largely mirrors what we saw for Classical Latin. There is a potentially relevant difference, though: in Late Latin, unlike Classical Latin, sporadic cases are found in which the first conjunct of a n˘ec-correlation is

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



formally positive and yet its meaning is clearly negative (cf. Löfstedt : – for the apparent productivity of this construction in Late Latin, especially in copulative coordination). In these cases the negative operator expressed by n˘ec scopes over both conjuncts, and this despite the fact that only the second is explicitly marked as negative. As a result, the whole clause is negative: n˘ec is here the only marker of sentential negation. () from Löfstedt (: –) a. nodum in apice neque in cinctu neque in alia knot:acc in hat:abl and.not in belt:abl and.not in other:abl parte ullum habet part:abl any:acc have:sg ‘(The flamen dialis) does not have any knot on his hat or on his belt or in other parts’ (Gell. ..) b. id autem iumentorum . . . si languet et that:nom then beast.of.burden:gen . . . if be.inactive:sg and pabulum nec potum adpetit, desperato food:acc and.not drink:acc desire:sg despair:/.sg.imp ‘as for donkeys, if a donkey is weak and does not want to eat or drink, don’t expect anything good’ (Mulomed. ) c. imber, nix, pruina, glacies nec fulgura rain:nom snow:nom frost:nom ice:nom and.not lightning:nom nocent harm:pl ‘No rain nor snow nor frost nor ice nor even lightning can harm’ (Aenigm. Cod. Bern. .., seventh century (?)) Similar cases of negative conjunctions ‘looking before and after’ are commented upon for Classical Greek in Löfstedt (: –) and for Germanic languages (Old Norse, Old English, German) by Jespersen () (cf. Old English suðne norð ‘neither south nor north’, Beowulf ); Dutch noch is also a case in point (Doetjes ). Torrego (: ) remarks that such instances are still found in early Romance texts, e.g., seventeenth-century Spanish in (): () en toda mi vida me han sacado diente ni muela ‘I’ve never lost a tooth or a molar in my life’ (Don Quixote ., from Torrego : , n. ). See also the Old Italian examples reported by Zanuttini (: ). Remarkably, in the Latin cases in () the conjunct containig n˘ec always precedes the verb: judging from the further examples in Löfstedt (: –), Löfstedt (: –) this seems to be the case in general.31

31 Interestingly, the same cannot be said for several Early Romance examples, witness the post-Infl position in ().

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

In §. I will factor this in when providing an analysis for Classical and Late Latin n˘ec. ... Stand-alone focus particle The third function of n˘ec in (-iii) is the use as stand-alone (i.e., non-correlative) focus particle. Under this heading I consider cases where the primary function of the particle does not consist in establishing a discourse connection with the previous context or a correlation between constituents, but rather in evoking, as relevant for the interpretation, an alternative proposition to the one in focus. This happens, for instance, when a different discourse particle takes care of establishing the syntactic and semantic-pragmatic link to the previous discourse, and n˘ec adds a semantic component that can be described as an additive presupposition: ‘also not; not either’. Informally, n˘ec introduces the requirement that the proposition with which it associates must have an antecedent that qualifies as one of its alternatives and that holds true as well. Orlandini and Poccetti (, ) provide the following example: () ita primis repulsis Maharbal cum maiore thus first:abl repulsed:abl Maharbal:nom with greater:abl robore virorum missus nec ipse eruptionem strength:abl man:gen sent:pt.nom and.not himself:nom sally:acc cohortium sustinuit cohort:gen sustain:sg ‘Thus after the first forces had been repulsed (even) Maharbal himself who had been sent with greater manpower did not sustain the cohorts’ sortie’ (Liv. ..) Here the structural and discursive link to the previous context is established through ita ‘thus’, and the main assertion (Maharbal . . . nec ipse eruptionem cohortium sustinuit) is related to a previously mentioned fact (primis repulsis) that represents its antecedent: the first soldiers who had been sent had been repulsed, and Maharbal got repulsed as well. Note that structurally the antecedent proposition is positive, unlike the proposition associated with n˘ec. This use of n˘ec shares with the correlative use the fact that alternatives are contextually overtly realized. As we will see in §.., correlative particles can also in fact be analyzed as focus-inducing particles. However, while in the correlative use the alternatives are explicitly mentioned by means of the correlative construction itself, in the non-correlative focus-particle use the alternatives are anaphorically retrieved from the broader context. The standard value of n˘ec as a focus particle is that of contributing an additive presupposition. However, depending on the semantic nature of the focus associate, the interpretation can be enriched with a scalar component, so that n˘ec does not mean ‘not . . . either’ but rather ‘not . . . even’. This happens when the focus associate expresses a particularly remarkable, unexpected value on a natural or contextually established scale. In () it is indeed plausible that a commander like Maharbal, who was responsible for many victories of the Carthaginians over Rome, was considered by writer and public the likeliest candidate to resist the attack (cf. also the co-occurring intensifier ipse), and yet he didn’t, contrary to expectations.

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



Correlative and focus-particle uses of n˘ec can coexist in the same structure, and cause redundancy in the marking of negation. The Early Latin example in () shows a particularly complex situation, where the verb nolo ‘not want’ contributes a semantic negation and is followed by a complement subordinate clause containing a correlation of adverbial PPs introduced by n˘eque, as well as the object of the subordinate verb, where the NPI ullum is reinforced by the focus particle n˘ec; this latter n˘ec clearly does not correlate with the previous instances of n˘eque, given the different syntactic function of the constituents in the scope of the respective particles: () Nolo ego cum improbis te viris, gnate not.want:sg I:nom with immoral:abl you:acc man:abl son:voc mi / neque in via neque in foro nec my:voc and.not in street:abl and.not in forum:abl and.not ullum sermonem exsequi any:acc conversation:acc pursue:inf ‘My dear son, I don’t want you to pursue any conversation with immoral men, neither in the street or in the forum’ (Plaut. Trin. , from Orlandini and Poccetti : ) Also in the focus-particle use, thus, and from an early age, we see n˘ec appearing in contexts where multiple negatively marked elements contribute just one semantic negation, as in Negative Concord cases.32 The focus-inducing use becomes more frequent from the imperial prose on. In this function n˘ec often replaces ne . . . quidem ‘not even, not either’, on which see §.... The use of ne . . . quidem is particularly frequent in Classical and Post-Classical prose, and becomes much less common afterwards (Pinkster : –). Hofmann and Szantyr (: –) remark that when n˘ec substitutes for ne . . . quidem it is ‘hervorhebend-negierend’, i.e. a focused negation, used when the conveyed proposition is surprising or unexpected (‘gegen die Erwartung zutrifft’), and can be translated with German nicht einmal ‘not even’. The following example, showing a blend of both constructions, n˘ec . . . quidem, is particularly instructive as to the functional overlap: () prooemium fatigare non debet, nec epilogus introduction:nom tire.out:inf not should:sg and.not conclusion:nom quidem too ‘The introduction should not be demanding, and neither should the conclusion’ (Quint. decl. ) In Late Latin we still find n˘ec in all the functions that have been described for Classical Latin. The focus-sensitive use, however, is encountered much more often and in a broader variety of contexts, which prelude the grammaticalization of Romance nec-words. The very frequent employment as focus particle is exemplified 32 This use is attested, as we have seen, from an early age, however it is never attested in Cicero (Orlandini a: , n. ). For more information on the relative chronology of attestation for the various readings see Hofmann and Szantyr (: –). A broad corpus study would be needed to assess whether this is to be ascribed to a particular sociolinguistic value of the scalar use.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

in () with passages from the New Testament, the Itinerarium Egeriae, and Gregory of Tours. () a. dico autem vobis quoniam nec Salomon in omni say:sg yet you:dat that and.not Salomon:nom in all:abl gloria sua coopertus est sicut unum ex splendor:abl his:abl dressed:pt be:sg as one:nom from istis this:abl ‘Yet I tell you that not even Solomon in all his splendor was dressed like one of these’ (Vulg. Matth. .) b. cum ipse mons sanctus Syna totus although itself:nom mountain:nom holy:nom Sinai:nom all:nom petrinus sit, ita ut nec fruticem habeat rocky:nom be:sg so that and.not shrub:acc have:sg ‘although the holy mountain Sinai is completely rocky, so that there isn’t even a shrub on it’ (Itin. Eg. .) c. Et iratus contra regem nec valedicens and angry:nom against king:acc and.not greeting:pt.nom abscessit. leave:sg ‘And angry with the king he left without even saying goodbye’ (Greg. Tur. Franc. ..) In all the instances in (), n˘ec has the meaning ‘not even’, i.e., it has a clear scalar component. The particle may combine with a scalar minimum and contributes the emphatic meaning that not even that minimum holds (e.g., if you leave the least you are expected to do is to say goodbye, cf. (c), if there is vegetation at all you expect at least a shrub, cf. (b), etc.). Alternatively, the focus associate expresses the likeliest option (the minimum on a scale of surprise) and negates it, resulting in an emphatic assertion (e.g. the likeliest person to be splendidly dressed is Solomon, cf. (a)). There are, however, other cases where the value is rather that of an additive particle, and no scalar component is present; this is particularly clear in enumerations and other situations where at least one alternative is explicitly mentioned in the previous context: () omnis qui negat Filium nec Patrem everyone:nom who:nom deny:sg son:acc and.not father:acc habet qui confitetur Filium et Patrem have:sg who:nom acknowledge:sg son:acc and father:acc habet have:sg ‘No one who denies the Son has the Father; whoever acknowledges the Son has the Father also’ (Vulg. I Ioh. .)

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



For Late Latin, as mentioned, it has been often observed in the literature (cf. Kühner and Stegmann : II., Hofmann and Szantyr : –) that very frequently n˘ec substitutes for ne . . . quidem. The blend of both constructions (n˘ec . . . quidem) found in Imperial Latin () can be readily compared with the Late Latin example in () (cf. Hofmann and Szantyr :  for the frequency of this blend in Late Latin): () ut nec aquam quidem ei aurire liceret so and.not water:acc quidem he:dat drink:inf be.allowed:sg ‘So that it was prohibited to him to even drink water’ (Greg. Tur. Franc. ..) This testifies to the fact that the use of n˘ec as a scalar focus particle considerably expands in Late Latin, also encroaching upon the functional domain of ne . . . quidem. .. Semantic analysis ... Basic ingredients and scope relations Horn (: –) highlights the parallelism between lexicalization tendencies with quantifiers and connectives: parallel to the situation found with quantifiers, where negation extremely rarely combines lexically with the universal quantifier (the negation of universal quantifiers being built syntactically, ‘not always’, etc.), we find quite frequently lexicalizations of ‘negation + disjunction’ (‘nor’), logically equivalent to ‘conjunction + negation’, but much more rarely lexicalizations of ‘negation + conjunction’ (* ‘nand’) (cf. Jacobs : ). According to Horn, the lack of lexicalization in both cases (*‘nall’, *‘nand’, that is *¬∀, *¬∧) is due to the fact that scalar implicatures make these items superfluous. The meaning they would convey lexically is already available by implicature from the meaning of the (non-negated) existential: ‘some’ pragmatically implicates ‘not all’, ‘or’ pragmatically implicates ‘not and’ (p ∨ q implicates ¬[p ∧ q]). The meaning of * ‘nand’ is thus conveyed (though not literally expressed) by or (Horn : ).33 Therefore languages economically avoid lexicalizing this meaning. No such considerations prevent, instead, the lexicalization of ‘nor’ = ‘and not’. As emerges from the previous discussion, Latin n˘ec, n˘eque is no exception to this generalization: although the negative morpheme linearly precedes the conjunction, the scope relation is ∧ ¬. For the discourse-structuring particle the basic meaning is that of a conjunction taking wide scope on the negative operator; that is, n˘eque / n˘ec = et n¯on. For the correlative and the stand-alone focus particle I propose that the -que / -c part has the meaning of an additive focus particle: in other words, both the correlative and the non-correlative uses are focus-sensitive, and n˘ec behaves, semantically and syntactically, as a focus particle. In what follows I will deal exclusively with these focus-sensitive uses, because they motivate—as I argue—the use of n˘ec as morphological component of the new Concord indefinites of Romance.34 33 For the alternative proposal that there are lexicalized Negated Universal Quantifiers in the domain of focus particles cf. Gast (). 34 The use as discourse-structuring particle is interesting on its own, and its demise in Romance is certainly connected to the semantic and syntactic changes affecting the other uses (cf. Garzonio and

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The main point of the analysis discussed below will be to derive the contextdependent, variable readings of the -que / -c element, especially in view of the emergence of the scalar reading, a very relevant Late Latin development. As for the ne- part, I will come back to it in the syntactic analysis: the most crucial aspect of its interpretation is that n˘ec qualifies as a semantically negative element, i.e., as endowed with a [Neg] feature. This emerges particularly clearly from the polarityswitch cases, where n˘ec is uncontroversially the element contributing a negative operator. In this respect, Latin obeys the generalization, formulated by Bernini and Ramat (), that correlative negations conform to the general negation system of the language, according to the parameters discussed in chapter . Less clear is the explanation for cases where, instead, n˘ec is found in structures where negation is redundantly marked. We will come back to this issue in §.. ... Correlative and non-correlative focus-sensitive uses My main goal in this section is to motivate the choice of n˘ec as formative of the new Romance indefinites on the grounds of a crosslinguistic tendency for NPIs (and, as we will see, a special type of n-word) to interact with focus. Particles behaving similarly to n˘ec are wellattested crosslinguistically as formatives of polarity-sensitive indefinites (Haspelmath :  ff.) and have been connected to the kind of scalar focus found in emphatic negation (Lahiri , Chierchia ). In turn, as we saw in chapters  and , emphatic negation has been argued to be very significant diachronically, e.g., as driving force behind the restructuring of negation systems linked to Jespersen’s Cycle (Kiparsky and Condoravdi ). Latin n˘ec is focus-sensitive both in its correlative and stand-alone uses as focus particle. Focus particles quantify over a set of relevant alternatives introduced by a focus operator (König ), and impose varying constraints on the form of these alternatives and on the way they are retrieved. In the literature it has often been remarked that correlative particles used for coordination or disjunction (either . . . or; neither . . . nor, but also both . . . and) are more emphatic than the simple coordination or disjunction (see Haspelmath  for the frequency of what he calls ‘contrastive coordination’ in typological perspective). An explanation for this has to do with their argumentative strength (in the sense of Anscombre and Ducrot ): since the conjuncts are typically co-oriented towards making the same point in the conversation, they can in principle (although not necessarily) be informationally stronger than each single conjunct taken alone. The correlative particles explicitly mark their ‘shared topicality’, which is part of their conventional meaning, whereas it is just inferred with plain coordinators (cf. discussion in Schwenter and Waltereit ).35 Both their syntactic and their semantic behavior suggest a treatment in terms of focus particles. Hendriks () proposes for English that the element of the

Gianollo ). Further research should ascertain whether it is possible to propose a unique lexical entry for all the three uses, by better understanding the role of the conjunction part, along the lines of Zeevat and Jasinskaja (). 35 See Orlandini (a) for a thorough analysis of the argumentative value of Latin n˘ec.

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



coordination introducing the first conjunct (either, neither, both) is a focus particle. The main insights of her analysis are shared by subsequent works such as den Dikken () and Ahn ().36 In Classical Latin, n˘ec receives an additive interpretation in the correlative structures and in the stand-alone use. Additive particles, such as English too or German auch, have been treated as focus-sensitive particles (cf. König : ch.  for an overview).37 Additives convey the presupposition that at least one alternative for the focus associate holds. A negative additive is felicitous only in evoking a negated alternative to a negated alternative in focus, and results in conveying the information that neither alternative holds. Crosslinguistically, additive particles are routinely multifunctional: besides their use as focus particle, König (: ) discusses their use as ‘conjunctional adverbs’ (meaning ‘moreover’, ‘in addition’), as coordinating conjunctions, and as components of indefinites, usually with a free-choice flavor (cf. Germ. auch ‘also’ in wer auch immer ‘whoever’). Latin n˘ec is no exception in this landscape, and is in fact found, at different historical stages, in all the functions indicated by König. Haspelmath (: §) comments in particular on the crosslinguistic frequency of the fact that what he calls negative contrastive coordinators (conjoining two negative constituents or clauses) are also used as scalar particles (in §.. I mentioned some languages where this distribution is observed). The use as focus particle, although frequent, is not possible across the board for correlative elements; there are crosslinguistic differences even among closely related varieties, as we mentioned for Romance (§..). Additive particles can be nonscalar (‘also’, ‘too’, ‘as well’) or scalar (‘even’). Scalarity is an obligatory meaning component with particles such as even, but it can also be contextually triggered with plain additive particles. König (: –) remarks that it is often difficult to draw a line between additive and scalar additive uses. Not all languages have a lexicalization for the scalar function and many use an underspecified additive particle instead. However, scalar particles can also have a nonadditive source, which points to the fact that in principle they form a separate class. As König (: ) observes, additive particles can have scalar interpretations when their foci happen to denote an extreme value. We see a similar phenomenon with Latin n˘ec in the Late Latin example (), where the cardinal ‘one’ in nec unus is a natural end-of-scale element (cf. Löfstedt : – for other similar examples): () Ramessen civitas nunc campus est, ita ut nec Ramessen:nom? city:nom now field:nom be:sg so that and.not unam habitationem habeat one:acc dwelling:acc have:sg ‘the city Rameses is now a level site without a single dwelling’ (Itin. Eg. .)

36 den Dikken () acknowledges the strong association between the use of (n)either . . . (n)or and the presence of contrastive focus; however, he treats the correlative elements as phrasal adjuncts with a distribution substantially different from that found with focus particles. 37 However, they can also associate with topics, as shown by Krifka ().

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

Similar observations have been made in the literature for ne . . . quidem, and for the positive etiam ‘also’ / ‘even’ (cf. Orlandini a). But even the most unmarked conjunction, et, has scalar readings: () quis putas hic est quia et ventis imperat who think:sg this:nom be:sg that and wind:dat command:sg et mari et oboediunt ei? and sea:dat and obey:pl he:dat ‘Who then is this, that he commands even winds and water, and they obey him?’ (Vulg. Luc. .) The case of Hindi focus particle bhii (cf. Lahiri ), to which we will come back in §., is also very instructive in this respect, since it has a basic additive reading ‘also’ and it is interpreted as scalar ‘even’ when realized with focus accent. This type of ambiguity is, according to König, a contextual effect, and is not dependent on the meaning of the particle itself. Rather, it arises as a byproduct of co-occurring factors. Particularly relevant is the nature of the focused element: the scalar interpretation arises naturally when the focus associate suggests a scale, of which it represents the end point. The context dependence of the scalar reading for additive particles suggests that, in principle, we do not need an ambiguity account for them. On the other hand, the fact that scalarity is obligatory with some other particles is an indication that they form a special class and that scalarity is part of their conventional meaning.38 The split we observe in Romance between languages that use the direct continuation of n˘ec both in correlatives and as scalar focus particle and languages that developed a reinforced form for the scalar use may point to a specialization of the particle’s lexical entry in the latter type of language. In Latin, n˘ec is multifunctional at all stages, but scalar readings are much more current in Late Latin, and they emerge in the stand-alone use, and not in correlatives. In what follows I discuss the relation between the additive and the scalar interpretation, in order to understand why the scalar interpretation of n˘ec is triggered in the non-correlative use and what consequences this new reading has. ... The scalar reading Two aspects that characterize the class of scalar particles, distinguishing them from additive ones, are the type of alternatives they select and the way these alternatives are retrieved in context. My aim is to single out the factors that may have favored the development of a scalar reading for n˘ec. I propose that the type of presupposition accommodation associated with scalar readings, but not with additive ones, is responsible for this development. The main component distinguishing additive from scalar-additive particles is the ordered versus unordered nature of their alternatives. Additive focus particles (also, too, either, etc.) ‘generally do not induce an ordering, but operate over an unordered set of contextually relevant values’ (König : ). Scalar particles, instead, 38 Gast and van der Auwera () provide an overview of scalar additive particles in modern European languages, as well as useful discussion of the extensive literature on even and its crosslinguistic parallels.

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



obligatorily impose an ordering on their alternatives. On the basis of crosslinguistic comparison, two main ordering sources have been proposed: discourse-based orderings and likelihood-based orderings (König : ch. , Beaver and Clark : –, Umbach ). As an example of discourse-based ordering, consider German noch ‘in addition, still’: Umbach (, ) has shown that noch differs from auch ‘also’ in that its alternatives are ordered in a discourse-based list constructed according to the order of mention in discourse (e.g. in enumerations), or, in the temporal interpretation, to the order of times. As for the likelihood-based orderings, according to the most widespread analyses, focus-sensitive particles such as English even, German sogar, French même introduce a scale whose elements are ranked according to a likelihood measure. The proposition they take as their focus associate is then marked as the least likely alternative (first proposals in this sense are found in Horn , Fauconnier , Karttunen and Peters ). Scalarity can be modeled as a presupposition, cf. (), which is encoded in the lexical entry of the particle together with the additive presupposition. Being a presupposition, this meaning component survives in negative and other downward-entailing contexts.39 () scalar presupposition based on likelihood p: focus associate q: alternative to p in context C l: likelihood scale ∀q ∈ C [q = p → p NOT > infants are pure. There is an ongoing discussion on the scoping possibilities of even and similar particles with respect to negation and other downward-entailing operators (cf. e.g. Karttunen and Peters , Rooth , König : –, Lahiri : –). Here I will assume an analysis in which the scalar component always scopes higher than the negation operator: ‘not even x’ is always interpreted as ‘even not x’ (Karttunen and Peters , Lahiri ).40 The kind of focus involved here corresponds to the operator E in Chierchia (), in turn corresponding to Krifka’s () Emph.Assert, seen in chapter  (§..). We can already see the consequences that this analysis has for our understanding of n˘ec. In the correlative use, we have an additive particle (-c / -que) scoping above a negation operator realized by ne-. In the focus-particle use, the scope relations between the components remain unaltered: the only difference concerns the interpretation of the additive component, which becomes scalar: n˘ec is thus reinterpreted as meaning EVEN > NOT. () Additive and scalar reading with n˘ec: Additive: ALSO > NOT Scalar: EVEN > NOT

40 Many languages have different lexicalizations for the version of the scalar particle appearing in negative contexts, cf. German nicht einmal, Spanish ni siquiera, Greek oute; a survey of these particles can be found in Gast and van der Auwera (: –). According to my analysis, Latin n˘ec also belongs to this group.

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



This kind of inverse scope with respect to the surface structure is parallel to what we have seen for English not even; the surface structure in Latin is motivated by the prosodic requirement of the enclitic conjunction particle, and is reproduced by Italian neanche ‘neither / not even’, on which see Franco et al. (, b).41 In §.. I will discuss the corresponding syntactic structure. Contextually, the choice between an additive and a scalar reading depends on the way alternatives are retrieved, a further important way in which additive and scalar focus particles differ. Additive particles are anaphoric: they need their alternative(s) to be explicitly realized in the discourse context. Thus, a sentence such as () is infelicitous when uttered out of the blue or in a context where no explicit alternatives to the focus (e.g. ‘John’) have been mentioned: ()  John had dinner in New York too. Cases like () have been analyzed as presupposition failure, and it has been remarked in the literature that presupposition accommodation with additive operators seems to be impossible, or restricted to very particular cases (cf. Zeevat , Beaver and Zeevat , Schwenter and Waltereit ).42 Things are different with scalar particles: they can be used in out-of-the blue contexts, since the alternatives they evoke can always be retrieved on the basis of a contextually derived (likelihood) scale. Presupposition accommodation is, thus, routinely observed with scalar additives. I do not have enough elements to decide whether the interpretational ambiguity seen in () diachronically leads to a lexical split with n˘ec. I will assume that this is the case when n˘ec becomes a bound morpheme in the new nec-words: the scalar meaning is invariant there. I also remark that the scalar reading only emerges for the stand-alone uses, which—as we saw—are expressed by a different lexical item in some Romance languages: this could point to a lexical split. In correlations, alternatives are explicitly listed by the construction itself, thus the interpretation is always additive.43 The increased frequency of scalar readings in Late Latin could in principle be due either to a lexical split, i.e. to the incorporation of the scalar presupposition in the particle’s lexical entry, or to a paradigmatic effect, i.e. to the removal of lexical blocking due to the decrease in frequency of ne . . . quidem (which expressed a.o. the ‘not even’ meaning in Classical Latin). In §... we have observed that the scalar use as a

41 Note that Franco et al. (, b) assume a different internal structure for neanche, where né is hierarchically superior and is the head of a Conjunction Phrase, whereas anche sits in the specifier of an embedded Focus Phrase. Interestingly, in Old Italian the order anche neuno, lit. ‘also no one’ is attested, discussed in Franco et al. (b). According to my analysis, this latter order would correspond to the original scope relation between the negative particle and the additive focus particle. 42 Ahn () proposes an alternative, non-presuppositional model for too and either in which strict anaphoricity is encoded as a silent propositional anaphor that the particle, analyzed as a two-place predicate, takes as one of its arguments (the other one being the overt focus associate). 43 Since n˘ec can introduce the first conjunct, we have to conclude that it has ‘forward-looking’ properties and it is able to cataphorically satisfy its additive requirement by relating to a further alternative which is mentioned explicitly only later in discourse.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

focus particle develops for n˘ec later than its additive function as discourse-structuring particle and correlative element, and becomes particularly frequent from the PostClassical age on. Crucially, the scalar interpretation kicks in when n˘ec is used as a stand-alone particle, and the discourse cohesion is taken care of by means of other connectors or discourse particles. This kind of stand-alone use consistently requires accommodation on the part of the hearer: no alternative is explicitly given in the context. As we have seen, accommodation with additive particles is severely constrained. Thus, I propose, the need for presupposition accommodation strongly favors the scalar reading: since alternatives cannot be retrieved anaphorically, the hearer looks for a different suitable method of retrieval; especially when the focus associate is represented by a scalar minimum in a negative environment, it is only natural that the chosen method is to supply a (likelihood) scale. The combination of n˘ec with unus, from which nec-words originate, is a prime example for the context just described. In this case, since presupposition accommodation is constantly required in order to achieve the intended scalar reading, the scalar presupposition ends up being incorporated into the lexical entry. The role of presupposition accommodation in my account fits well with general models that have been proposed for semantic change, such as the ‘Invited Inferencing Theory’ by Traugott and Dasher () or the ‘Principle of Pragmatic Overload’ by Eckardt (). The role of presuppositions has been particularly highlighted by Schwenter and Waltereit (). In all these models, the necessity for a pragmatic operation on the part of the hearer may lead, if ‘non-canonical’ uses reach a certain frequency, to the conventionalization of a certain meaning component, which becomes part of the reanalyzed lexical entry. In the case of n˘ec in those combinations from which nec-words originate, the frequent accommodation of an additive presupposition by means of scale retrieval would have led the hearer to assume the scalar component to be part of the lexical entry, in the form of a scalar presupposition. This hearer-based reanalysis, in turn, would have led to a further increase of the ‘new’ uses (as we in fact observe in Late Latin) and to the establishment of the new function. The diachronic mechanism I propose is similar to the disambiguating strategy investigated by Tovena () for Italian neanche, which can be either simply additive (‘neither’) or scalar (‘not even’): Tovena shows that both readings are in principle possible only if the anaphoric constraint can be satisfied, i.e., if the alternatives are explicitly given (‘verified by the context’). If, instead, alternatives have to be accommodated, the scalar reading becomes obligatory. The reanalysis is very parsimonious, since it does not involve a change in the scope relations between the elements involved, but just the conventionalization of the scalar interpretation, which is a crosslinguistically observed possibility for additive particles. .. Syntactic analysis In this section I briefly summarize the main aspects of the syntax of n˘ec, which represent a prerequisite for the syntactic analysis of nec-words. The analysis is based on Gianollo (b: ch. ), Gianollo (), to which I refer for further details. We

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



will also have the occasion to compare n˘ec with ne . . . quidem ‘not even, not either’, highlighting their structural similarity, which underlies their functional competition as stand-alone focus particles, as well as their role in causing redundant marking on negation (as we will see in §.). ... The internal structure of n˘ec At all documented stages of Latin, the use of n˘ec is connected to the presence of a semantic negation operator in the interpretation. In all its functions, n˘ec is able to express negation by itself, and to license NPIs in its scope. This is not surprising in view of the fact that Classical Latin is a Double Negation language, where it is expected that each morphosyntactic manifestation of negation corresponds to a semantic negation operator. Therefore, it seems appropriate to treat Classical Latin n˘ec as a fundamentally [Neg] element, and to deal separately with the ‘redundant’ cases like those in (), looking for a mechanism that might explain the seemingly systematic nature of these exceptions. As emerged from the semantic analysis, n˘ec is a complex semantic element, which shows in all its uses the combination of negation with another semantic component that is responsible for additivity or scalarity, depending on the context. I proposed that the additive uses (as discourse connective or correlative particle) have in common with the focus-particle use the scopal relation between the components: namely, negation has narrow scope in both cases (AND / ALSO > NOT; EVEN > NOT). Since n˘ec is still morphologically transparent (recall that it alternates with the even more transparent form n˘eque), it seems appropriate to treat the two semantic components as morphosyntactically distinct elements; I thus propose the structure in () for correlative and scalar n˘ec:44 () Structure for correlative and scalar n˘ec FocP Foc Foc0

Op¬P Op¬

-c/-que Op¬0 [Neg]

XP

neIn the analysis proposed here, the structural relations mirror the scope relations, since the negative operator is merged at a lower level than the additive or scalar component.

44 See Gianollo (b: ch. , ), Garzonio and Gianollo () for the proposal of a parallel syntactic structure for the discourse-structuring use as well.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The reverse surface structure is due to the enclitic status of -que / -c, which forces prosodic inversion; note that ne- is itself proclitic (cf. the discussion in §..) but can be a host for the enclitic, since together they form a prosodically viable structure. In his study of pronominal syntax from Latin to Romance, Wanner (: ) finds that in Classical Latin n˘ec counts as prosodic host for the placement of prosodically weak elements. The negative operator directly attaches to the constituent it targets as a functional shell, invisible for subcategorization, in the same way as with negative indefinites of Double Negation languages (cf. () in chapter ). Also FocP is a kind of structural ‘shell’ that is projected on top of the constituent with which the particle narrowly associates, and of each focus-sensitive element in the correlation.45 Note that in correlative structures n˘ec is not the head of a Conjunction Phrase &P, but rather occupies a position in the CP-periphery of the conjunct. In correlations, the correlative element (-c/-que) of each particle heads a Focus Phrase on top of the negative affix. In the absence of evidence to the contrary, n˘ec can be given a uniform analysis, independently of whether it introduces the first or a further conjunct.46 I assume that a covert conjunction operator is the head of &P, responsible for the correlation of the two (or more) Focus Phrases (cf. Bianchi and Zamparelli , den Dikken , Szabolcsi , Mitrovi´c and Sauerland ). Similar syntatic accounts have been proposed by various authors for English correlatives (cf. in particular Hendriks , den Dikken , and Ahn ). The structure I proposed for correlation by means of n˘ec imposes no syntactic constraints on the polarity of the accompanying conjunct(s), since the projection of negation is in principle independent of the focus shell (although the two must of course co-occur if n˘ec is used, since the particle is intrinsically negative). This derives the polarity switch uses seen in Latin, where the first conjunct is positive. Correlative n˘ec, thus, is not an NPI, but a [Neg] element: it contains a negative operator and it is able by itself to negate the constituent it takes as complement (be it a CP or a smaller constituent). The associate constituent can in principle be of any syntactic category. This assumption will suffice for our present purposes, but is certainly too simplified, especially for correlative structures. Ellipsis-based accounts of similar correlative structures, such as Bianchi and Zamparelli () and Repp (), take coordination to always involve clausal constituents, or even speech acts, in the case of Repp, which undergo different types of ellipsis processes.

45 See Biberauer () for a similar proposal concerning Polarity Phrases, hosting reinforcers of negation. 46 The same conclusion is not uncontroversial for the Romance correlative particles that continue n˘ec (for instance, de Swart  proposed an ambiguity account for French ni and distinguishes two homophonous but formally different particles).

OUP CORRECTED PROOF – FINAL, //, SPi

Syntax and semantics of Latin n˘ec



In the analysis proposed in §., the connection with a CP-level Focus operator is achieved through a form of Agree: the particle is not the direct expression of the operator, but carries a formal uninterpretable feature [uFoc] inducing a relation with a higher operator at the clausal level. ... Comparison with ne … quidem Indirect support for the decompositional analysis proposed for n˘ec comes from the discontinuous particle ne . . . quidem ‘not even, not either’, which can be analyzed in an analogous way:47 () Structure of ne . . . quidem FocP Foc

ne + XPi Foc0

Op¬P Op¬

quidem Op¬0

XP ne-

i

The proposed structure is parallel to the one seen for correlative and scalar n˘ec: the two elements form a complex shell, where the focus particle scopes above the negative component (differently from the assumption of Devine and Stephens  and in conformity with our assumptions on ‘even’), and the order is reversed at PF. The negative morpheme and the particle quidem are always separated by at least one word.48 More intervening words are possible in certain rarer cases. As for the negative morpheme, despite the long vowel (n¯e), I follow Fruyt (b) in considering it a continuation of the prehistoric particle *ne (cf. §...). The negative particle is proclitic on its sister constituent. Owing to the prosodic properties of quidem (on which cf. Danckaert ), a prosodically motivated movement takes place, resulting in the observed linear order (Devine and Stephens : , Gianollo ). The complex particle has two main functions: it can be a discourse connector with an additive import (‘also . . . not’, ‘neither’), or an emphatic expression of negation in virtue of a scalar component (‘not even’), cf. ().49 47 Further data and discussion can be found in Orlandini (a: , –), Devine and Stephens (: –), Pinkster (: –). 48 The particle quidem can also occur by itself, independently of negation, and is always connected to focus: Danckaert (, ) has analyzed it as a marker of emphatic polarity. 49 On the focusing nature of ne . . . quidem see more extensively Orlandini (a: ch. ).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

() a. sese unis Suebis concedere, quibus ne di they:acc only:dat Suebi:dat yield:inf who:dat ne god:nom quidem immortales pares esse possint quidem immortal:nom equal:nom be:inf can:pl ‘(the Germani said) they yielded to the Suebi alone, to whom even the immortal gods could not be equal’ (Caes. BG ..) b. de tam magna nave ne tabulam quidem naufragus from so big:abl ship:abl ne plank:acc quidem castaway:nom habes have:sg ‘of such a great ship, as a castaway you have not one plank of it left’ (Petr. .) We see, thus, that ne . . . quidem parallels n˘ec in two of the three functions examined in the previous sections, with the exception of the copulative-correlative value. In Late Latin its two components are sometimes found adjacent (a), a use that is explicitly condemned by grammarians, indirectly testifying to its frequency (cf. TLL ...–.). Cases are also found where quidem is dropped, the first example being in one of Trimalchio’s speeches in the Satyricon (cf. b; further examples in Pinkster : ). () a. ne quidem civis Romanus est ne quidem citizen:nom Roman:nom be:sg ‘he isn’t even a Roman citizen’ (Gaius inst. .) b. hoc solum vetare ne Iovis potest this:acc only:acc forbid:inf ne Jupiter:nom can:sg ‘even Jupiter himself cannot forbid this one thing’ (Petr. .) In this latter situation, as previously remarked, formal confusion occurs between n˘ec and ne (quidem), a confusion which is connected to the contemporary functional overlap.50 .. Conclusions In this section I have dealt with the various functions of Latin n˘ec. It turned out that the correlative and the stand-alone focus particle uses are the most relevant for our main research question. Both uses are focus-sensitive, and allow for a decompositional syntactic analysis of the particle as a semantically negative focus particle. Both formally and functionally, n˘ec finds a parallel in ne . . . quidem, a discontinuous focus particle whose functions are absorbed by n˘ec in Late Latin. I discussed two ingredients that distinguish additive focus particles from scalar additive focus particles, arguing that this captures the difference (and at the same time also the diachronic link) between the 50 Unfortunately the variability in the textual transmission due to the confusion in the manuscripts and to the frequent emendations by modern editors does not allow an evaluation of the extent of this phenomenon. This is particularly regrettable in view of what we observed in §.. on the confusion between ne and n˘ec in Merovingian Latin and its significance for the Romance facts.

OUP CORRECTED PROOF – FINAL, //, SPi

From nec-words to n-words



connective and the stand-alone uses: unlike additive particles, scalar particles evoke ordered alternatives and do not need to satisfy their presupposition anaphorically in the context. The first ingredient is responsible for the strengthening uses that scalar particles have in negative environments, since combining with a negated scalar minimum amounts to yielding the informationally strongest proposition. The lack of anaphoricity, in turn, has been argued to be diachronically responsible for the conventionalization of the scalar reading with n˘ec used in combination with unus, i.e., in the structure from which the nec-words originate. Starting from these premises, in the next section we will see which semantic and syntactic processes are involved in the grammaticalization of nec-words.

. From nec-words to n-words Given the analysis of Latin n˘ec presented in §., its use as negative scalar focus particle clearly emerges as the functional motivation for its presence in the new Romance nec-words. The combination of n˘ec with the cardinal numeral functions as a negation strengthener. As we saw in §., the connection between focus and indefinites has been explored in depth in the semantic literature in order to explain the behavior of NPIs (Heim , Kadmon and Landman , Lee and Horn , Krifka , Lahiri , Giannakidou , Chierchia  being among the most significant milestones). A central role in the discussion has been played by crosslinguistic evidence that focussensitive particles are often found as morphological formatives of NPIs (Haspelmath : –, –, Lahiri ). I build on this body of work in accounting for the grammaticalization of nec-words in Romance. In this section I first propose a syntactic analysis for the internal structure of the new indefinites (§..). Afterward, I review the crosslinguistic evidence for the presence of focus particles in the morphological build-up of indefinites (§..). I also address the nature of the co-occurring morphological building blocks of nec-words (the role of unus; the function of ipse) (§..–§..). I conclude this section by addressing the loss of emphasis the nec-words experience (§..), thereby highlighting a diachronic parallelism between the development of n-words and the reanalysis of the continuations of aliquis that have been the topic of chapter . In the subsequent section (§.) I will deal more specifically with the negative component of nec-words: unlike other NPIs formed with an ‘even’-particle, nec-words are negative (‘self-licensing’, in Ladusaw’s  terminology) from the beginning, and this determines their syntactic distribution. However, the ‘even’-component we explore in this section will be shown to be relevant for their syntax as well. .. The internal syntax of nec-words In §. I analyzed n˘ec as a complex focus particle, composed of a Focus layer on top of a layer hosting the negation operator. The source construction for the grammaticalization of nec-words is the combination of this particle with the cardinal numeral unus ‘one’, either in a pronominal use or as a modifier of an indefinite description.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The negative operator attaches to the element in focus as a form of constituent negation. Notice that this is to some extent similar to what happens with negative indefinites, according to the analysis I gave in §..; however, negative indefinites are different in that they involve a form of selection and lexically require immediate closure by a negative operator. Therefore I assumed for negative indefinites that the negative operator attaches as an adjunct within the left periphery of the nominal projection. Here, instead, negation and focus form shells that are external to the extended projection of the constituent in focus. Approaches to the semantics of scalar focus particles like the one I adopted in §... require that the particle take sentential scope at LF, independently of its surface position. For reasons that will become clearer in §., I model this as a syntactically active relation of Agree between a Focus marker (the particle itself), carrying a [uFoc] feature, and a higher abstract Focus operator merged in the locus of interpretation and contributing the interpretable [iFoc] counterpart. The combination with unus in the construction from which nec-words originate yields the following structure: () nec-word: originating construction (nec unus) FocP Foc Foc0

Op¬P Op¬

-c [uFoc] Op¬0 ne[Neg]

DP NumP Num Num0 unus

NP x

In the proposed structure, unus is located in NumP, the locus of cardinal numerals and other counting devices, corresponding to Borer’s () P. At a further stage of grammaticalization, when the relation between the negative particle and the (pro)nominal base becomes lexicalized, it is plausible to assume that the negative particle becomes part of the DP-structure and is hosted in the left periphery of the DP. This possibly happens during a Proto-Romance stage, since the first consistent evidence for a word status appears in Romance and is absent from genuinely Latin documents. In the structure for the nec-word after reanalysis given below, I assume that the particle carries a [uFoc] feature, as before reanalysis, and sits in a DP-internal Focus

OUP CORRECTED PROOF – FINAL, //, SPi

From nec-words to n-words



projection. This is the projection that I posited in chapter  as landing site for indefinite algún, alcuno, etc. in inversion structures. For nec-words, I also tentatively assume that, at the time of lexicalization, the particle has lost morphosyntactic transparency and realizes just one syntactic head.51 The proposed structure finds a correspondence in the analysis given by Garzonio () for Old Italian né . . . mica, where né is considered the realization of a Focus head. At the point of reanalysis, the nec-word has plausibly already become an element of Negative Concord. In conformity with my proposal for the internal structure of n-words in §.., the indefinite does not contain the projection of a negative operator, but carries a formal uninterpretable negative feature [uNeg], which induces a syntactic relation with a higher operator. I therefore posit two uninterpretable features for the lexical entry of n˘ec at this stage: a (conservative) [uFoc] feature and an (innovative) [uNeg] feature (which I will motivate later in §.). () nec-word after reanalysis: DP TopP FocP Foc Foc0 nec [uFoc], [uNeg]

dP NumP Num Num0 NP unus

x

When the particle is grammaticalized as a component of an indefinite, the change obviously involves a restriction in the particle’s ability to select a focus: namely, the alternatives on which the particle operates must now be lexically selected by the predicate determining the restriction of the quantification (cf. discussion in Lahiri , Chierchia ). Chierchia (: –) explains this way the fact that a sentence like () is ungrammatical in Hindi (where the focus particle is lexically part of the indefinite ek bhii ‘even one’), but its English counterpart is acceptable:

51 Possibly, as a reviewer suggests, this reduction stage correlates with the phonological weakening undergone by the particle when it loses its velar component in some Romance varieties, cf. §...

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

() from Chierchia (: –) a. *ek bhii aadmii aayaa one even man came ‘Any (= Even one) man came’ b. Even one man came In Hindi alternatives are lexically restricted to be men; in an upward-entailing context, a contradiction arises between the focus particle’s requirement that the associate be the informationally strongest proposition and the fact that the alternatives (‘two men came’, ‘three men came’) are actually informationally stronger. This causes unacceptability. In English, instead, alternatives are not lexically restricted: a scale can be built contextually comprising contextually less striking alternatives (‘a woman came’, ‘a child came’) and the presupposition of even can be satisfied. .. Focus particles and polarity sensitivity We have already discussed in §. how focus and polarity sensitivity interact when the communicative goal is to yield a strengthened assertion in a downward-entailing context. In the case of nec-words, the status of the negative morpheme at the outset is clearly that of a semantically negative element ([Neg], according to our typology), therefore nec-words are not NPIs. However, much of what has been observed in the literature on the connection between NPIs and focus marking applies to nec-words as well, and will ultimately lead me to propose a revised typology for the NPI / n-word divide in §.. The fundamental insight from pragmatically based accounts for the licensing of NPIs is that the felicity of an NPI in a given linguistic environment depends on the interplay between the lexical meaning of the NPI and the pragmatic implicatures it triggers. NPIs are evaluated against the background of their alternatives (in Chierchia , NPIs are such because they obligatorily activate alternatives). Alternatives must be interpretively exploited in the context: this is the case, for instance, when the ensuing entailments lead to the strengthening (i.e., to a higher degree of informativeness) of an assertion. One way of achieving strengthening is to insert a scalar focus operator, which operates on ordered alternatives and conveys that the (proposition containing the) focus associate is the informationally strongest. In §. we saw how the meaning of this operator is taken to correspond to that of a covert even (cf., e.g., Krifka , Chierchia ). In such a framework, it is natural to see the occurrence of scalar additive particles with indefinites as a confirmation of this hypothesis. Haspelmath (: –) offers a crosslinguistic overview of indefinites formed by means of scalar focus particles meaning ‘even’. Further on (pp. –) he defines a subclass of ‘negative indefinites’ formed by negative scalar focus particles: e.g., Russian ni-kto ‘nobody’, Lithuanian nie-kas ‘nobody’, Romanian nici-odat˘a ‘never’, Classical Greek oud-eís ‘nobody’, Hungarian sen-ki ‘nobody’.52 52 Haspelmath (: –), in discussing the diachronic sources of negative indefinites, remarks that most often, and especially in European languages, the negative element of the indefinite is not the plain sentential negator, but rather a negative scalar focus particle. Remember, however, that Haspelmath’s

OUP CORRECTED PROOF – FINAL, //, SPi

From nec-words to n-words



In addition to properly scalar particles (‘even’), also nonscalar additives (‘also’) are used, which, as we saw, are known to be able to take a scalar value and cannot be strictly separated. In Haspelmath’s sample, most of the indefinites formed with focus particles have interrogative pronouns as their base, which however also function normally as indefinites in their bare form: e.g., Serbian/Croatian i-ko ‘anyone’ (i: ‘and, also, even’), Hittite kuiš-ki ‘someone’ (-ki: ‘and, also’), Sanskrit ka´s cana ‘anyone’ (cana: ‘even’), Japanese wh- + particle mo ‘also / even’ (cf. Watanabe ). In some cases, the basis is ‘one’, or a generic noun, or another indefinite item: e.g., Hindi/Urdu koii bhii ‘anybody’ (koii: ‘someone’, bhii: ‘also, even’), Dutch ook maar iemand ‘anybody’ (ook maar ‘even, at least’).53 In all these indefinites, be they positive (NPI, free-choice indefinites) or negative (n-words, NIs), the functional motivation for the appearance of the focus particle is seen in the expression of an extreme point on a pragmatic scale (Haspelmath : ). If the particle is morphologically negative, it may either convey a semantic negation operator or be a mark of an uninterpretable formal feature. That is, each of the resulting indefinites may be either a negative indefinite or an n-word (element of Concord); according to Haspelmath (: –), crosslinguistically these indefinites tend to co-occur with a further marker of sentential negation. An interesting case, resembling in some respect the Latin situation, is represented by Hungarian.54 Hungarian negative pronouns (e.g., senki ‘nobody’, semmi ‘nothing’) originate from the combination of a particle sem and an indefinite / interrogative root; in turn, the particle sem derives from the combination of an additive focus particle es with the negative particle nem (Kiss : , Gugán : ). These indefinites were negative indefinites in Archaic Old Hungarian, but have become Concord elements early in history and are still such in in Modern Hungarian (a strict NC language, Bende-Farkas ). The most famous example of a non-negative indefinite formed with a focus particle comes from Hindi and has been analyzed in a very influential study by Lahiri (). Lahiri shows that Hindi has a series of compositionally transparent NPIs formed of a weak indefinite and a (postponed) focus particle bhii ‘even’. The bases are various elements expressing a predicate ‘that is true of everything that exists’ (Lahiri : ), e.g. ek ‘one’ in ek bhii; koii ‘someone’ versus koii bhii ‘anyone’, etc. The focus particle bhii is ambiguous between an additive and a scalar reading. In plain contexts it means ‘also’: () raam bhii aayaa (Lahiri : ) Raam emph came ‘Also Raam came’ notion of ‘negative indefinite’ does not correspond to ours, since he also includes Negative Concord elements and some NPIs. 53 For ‘at least’ as focus particle see also Russian xot’. Interestingly, the latter can mean both ‘at least’ and ‘even’; other languages, instead, distinguish lexically between a particle that ‘characterize[s] the value of the focus as high on a pragmatic scale’ (‘even’: German sogar, French même, Russian daže) (Haspelmath : –) and another that characterizes low pragmatic values (‘at least’). Greek kanénas is grouped here, with kan meaning ‘at least’. The issue concerning the choice of the lower versus upper bound is thoroughly discussed by Gast and van der Auwera (). 54 I am grateful to Ágnes Bende-Farkas for discussing these data with me.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

But if the associate (raam in ()) is focused a scalar component appears, such that the assertion ‘Raam came’ has to be interpreted as the least likely option in the context. Hindi NPIs are obligatorily focused: that is, in NPI-licensing contexts only the ‘even’ reading is available. Interestingly, in Hindi NPIs can be subjects if the licensor is in the same clause, even if the latter does not c-command the indefinite at PF (Lahiri : ): () koii bhii nahiiN aayaa (Lahiri : ) anyone not came ‘No one came’ Lahiri (: –, –) attributes this to special scope-taking properties of negation in Hindi. But since focus seems more important than negation in licensing NPIs in a broad variety of contexts in Hindi, one could also argue that the crucial relation here concerns the focus operator. Lahiri () indeed has to assume that the (operator connected with the) focus particle takes sentential scope at LF, since, as we discussed in §..., for the pragmatic inferences to be felicitous focus has to operate over negated alternatives, i.e., the focus operator must outscope negation. .. N˘ec unus Now that we have seen the role of the negative formative of Romance nec-words, we can address the role of the (pro)nominal element: we will restrict our attention to those nec-words listed in §.., where n˘ec combines with unus ‘one’. I analyze unus in nec-words as a cardinality predicate expressing a scalar minimum, parallel to what we have seen for Hindi ek bhii. An important aspect (and a fundamental motivation) of the approach to NPIs as strengtheners consists in accounting for the fact that NPIs typically have existential quantificational force.55 One of the most widespread strategies to achieve existential quantification in polarity-sensitive indefinites is the use of the cardinal numeral ‘one’. In previous chapters, we had the occasion to comment on some such indefinites: Latin NPI ullus, for instance, originates as the diminutive of unus and is the base for the NI nullus. The list provided by Haspelmath (: ) comprises a.o. English any (OE ænig < an ‘one’ + -ig), Classical Greek oudeís ‘no one’, Irish aon ‘any’; Persian hiˇc < Old Persian aiva ‘one’ + emphatic particle ˇciy. The class of nec-words seen in §. testifies to the success of this strategy in Romance. Also recall that, as we saw in chapter , unus was employed in Romance to formally renew the continuations of aliquis. In §. I concluded that the addition of unus in Late Latin / Proto-Romance contributes to the salience of scalarity in the meaning of aliquis, which in turn is a motivation for its innovative use in polaritysensitive contexts.56Also in the case of nec-words unus provides the scale to be used (the scale of natural numbers) and expresses its end point. 55 But Giannakidou (, ) for Greek proposes that they can acquire universal quantificational force. 56 In Chierchia’s () framework the substitution of the wh-indefinite with the cardinality predicate ‘one’ could be interpreted as the lexicalization of totally ordered alternatives, as we will see later on.

OUP CORRECTED PROOF – FINAL, //, SPi

From nec-words to n-words



Given its general meaning (‘true of everything that exists’, Lahiri ), unus is the perfect candidate for grammaticalization. Since the alternatives are readily provided by the other numerals, it is also a very suitable element for the accommodation of the additive presupposition (conveyed by the focus particle) in contexts where no anaphoric reference to a previously explicitly mentioned alternative (set) is possible: according to my hypothesis in §..., those are the contexts that triggered the conventionalization of the scalar component in nec-words. Being a natural scalar end point, unus represents an optimal strengthening strategy, since it is logically entailed by all its alternatives. When, as a deteminer, the indefinite is combined with a lexical restriction, the alternatives are further specified along the dimension introduced by the lexical predicate.57 The cardinal numeral for ‘one’ can be found in negative contexts with the emphatic meaning ‘even a single’ at all stages in Latin (in addition to the other strengthened meanings, e.g. ‘only / just a single’). It can be used as an emphasizer of negative indefinites (‘not even a single one’) and other negative expressions already in Classical Latin, in examples like the following: () a. ut nemo unus inde praecipuum so.that nobody:nom one:nom thence conspicuous:acc quicquam gloriae domum invidiaeve ferret anything:acc glory:gen home:acc envy:gen-or bring:sg ‘in such wise that no single patrician came off with any conspicuous share of glory or unpopularity’ (Liv. ..) b. nam equidem, postquam natus sum, numquam aegrotavi indeed certainly after born:pt be:sg never be.sick:sg unum diem one:acc day:acc ‘I have never been ill even for a single day since birth’ (Plaut. Men. ) Examples of constructions where unus combines with n˘ec and ne . . . quidem like those in () may well be regarded as the source of lexicalization for nec-words.58 () a. Dic quotus et quanti cupias cenare nec say:sg how.many and how.much:gen wish:sg dine:inf and.not tibi unum addideris verbum: cena parata one:acc add:sg word:acc dinner:nom prepared:pt you:dat est. be:sg ‘Say how many you want to dine and for how much, and don’t add a word; your dinner is ready.’ (Mart. ., first century ce) 57 See Chierchia (: –) on how this restricts the focus that the particle can associate with. 58 The constructions are to be sure not particularly frequent in Classical Latin: to give an example, a query for nominative and accusative forms of unus with ne . . . quidem in the LLT-A for the period ‘Antiquitas’ (until the end of the second century ce) retrieves only  instances (some of them in Cicero). The same query for the LLT-A texts from the third until the end of the fifth century ce delivers  instances. However these results have to remain impressionistic in view of the fact that I could not control for word count, owing to the nature of the electronic corpus.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance b. ne una quidem navis amissa est ne one:nom quidem ship:nom lost:pt be:sg ‘not even one ship was lost’ (Flor. epit. .., first to second centuries ce) c. Ramessen civitas nunc campus est, ita ut nec Ramessen:nom? city:nom now field:nom be:sg so that and.not unam habitationem habeat one:acc dwelling:acc have:sg ‘the city Rameses is now a level site without a single dwelling’ (Itin. Eg. ., fourth century ce, cf. () and Bertocchi et al. : ) d. ut si quis velit dicere nec unum esse as if someone:nom want:sg say:inf and.not one:acc be:inf hominem in theatro, ita dicat: non modo illic unus man:acc in theater:abl so say:sg not only there one:nom homo non est, verum nec ullus man:nom not be:sg but and.not any:nom ‘if someone wants to say ‘there is not even a man in the theater, he will say: not just illic unus homo non est, but nec ullus’ (Boeth. in herm. comm. ., fifth to sixth centuries ce)59

Frequently in Early and Classical Latin, n˘ec combines with ullus, the NPI originating as diminutive of unus (cf. , ). This use continues in Late Latin, as shown in () (cf. also d): () Denique nocte media omnes Pectava somno finally noght:abl mid:abl all:nom from.Poitiers:nom sleep:abl falanga conpraemitur, nec ullus superfuit, band:nom overwhelm:sg.pass and.not anyone:nom remain:sg qui ex hac multitudine vigilaret. who:nom from this:abl multitude:abl keep.watch:sg ‘Finally in the middle of the night the whole band from Poitiers were overcome by sleep, and nobody from that multitude was left to keep watch’ (Greg. Tur. Franc. ..) On the basis of Lahiri’s () analysis of Hindi ek bhii ‘one-even’, the meaning of emphatic ‘even-one’ NPIs is formalized by Chierchia (: ) as follows: () ‘even-one’ NPIs, adapted from Chierchia (: ): λPλQ ∃x [one(x) ∧ P(x) ∧ Q(x)]

one σ −ALT = {λPλQ ∃x [n(x) ∧ P(x) ∧ Q(x)] : n ≥ one} 59 Boethius’ entire passage discusses the expression of negation and explicitly acknowledges an emphatic flavor for n˘ec, as we see in the example; later on in the passage Boethius also comments on the strengthening of negation ensuing from negating a diminutive: ita igitur maxima fit negatio rei parvissimae quod est unus, si ipsius diminutivum quoque subtrahat, quod est ullus ‘And in fact the maximal negation of the smallest thing which is ‘one’ takes place if the negation affects its diminutive, which is ullus’.

OUP CORRECTED PROOF – FINAL, //, SPi

From nec-words to n-words



The scalar alternatives (σ ) of one are strictly ordered; given the shape of the alternatives, in Chierchia’s framework their exhaustification has to take place by means of the E operator. The E operator corresponds to the meaning of ‘even’ (cf. Krifka’s  Emph.Assert. operator), i.e., to scalar focus, as we saw in section ..: () EALT (p) = p ∧ ∀q ∈ ALT[ p [uNeg] reanalysis. In §., concluding this chapter, I will come back to the NPI-uses of n-words seen for Old French and Old Italian in §. and discuss how they could be reconciled with the Negative Concord nature of the Early Romance grammars in view of what I concluded on Latin. .. The redundant uses Examples of redundancy that involve n˘ec and another exponent of negation in a single-negation reading are definitely the most frequent type of redundancy in negation marking in Latin. I will argue that they are very important for understanding the birth of Negative Concord. Yet, they are not themselves examples of Negative Concord. I will propose that redundancy in Classical Latin is due to the interaction with the syntax of focus, and is still compatible with a Double Negation grammar. When the general conditions for Negative Concord develop in Late Latin, these structures become optimal triggering evidence for reanalysis. In the previous sections we reached a homogeneous characterization of Classical Latin n˘ec as a [Neg] element. Correlative and scalar n˘ec have been uniformly analyzed as focus particles with identical syntactic features (their lexical entries are however distinct, in view of the additional scalar presupposition that I assumed for some standalone uses). In §.. I preliminarily proposed that the lexical entry of n˘ec involves a [uFoc] feature: I will motivate this move in what follows by showing that the focus-sensitivity of n˘ec is mirrored in the syntactic relation it entertains with a clausal Focus projection (hosting the [iFoc] counterpart). Given this characterization for the lexical entry, from Latin to Romance the process of reanalysis in () happens: () Reanalysis with correlative and scalar n˘ec a. n˘ec before reanalysis: [Neg], [uFoc] b. n˘ec after reanalysis: [uNeg], [uFoc]

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



In what follows, I will mostly concentrate on cases of redundancy involving the scalar stand-alone particle, since the latter is the element directly involved in nec-words. However, redundancy is observed also with the correlative particles (recall () and ()). In fact, redundancy with scalar n˘ec is a Late Latin phenomenon, whereas, judging from a first cursory corpus search, redundancy with n˘ec in Classical Latin is basically restricted to the correlative use.62 However, since these structures are not directly relevant to the syntax of nec-words, I will leave them aside for the moment, to come back to them briefly in conclusion of this section (§...). In §.. I provided an example in () for the co-occurrence of scalar n˘ec with another negative element in a single-negation reading in Early Latin. Cases where scalar n˘ec co-occurs with a further negative element in a single negation reading abound in Late Latin (cf. Stotz : –) and are also found in contexts of syntactic integration (i.e., not as afterthoughts). Yet these uses are not unconstrained: they are typically found in structures where a sentential negation co-occurs with a further constituent negation, as in (). () non est relictus ex eis neque unus not be:sg left:pt from they:abl and.not one:nom ‘not even one of them was left’ (Agnell. lib. pont. , ninth century) The construction in () has an emphatic flavor, and well represents the context of use where Romance nec-words originated, since the focused element is the cardinal numeral unus. The redundant behavior with respect to the marking of negation has traditionally been attributed to the particle’s copulative nature (cf. Löfstedt : ), but the link is not direct: in the examples at hand n˘ec does not have the function of a copulative correlative, and shows rather the behavior of a scalar focus particle. In Augustine’s sermons we find frequent examples of redundancy with scalar n˘ec, cf. e.g. (): () a. forte, inquit, tu avarum et cupidum diceres, si maybe say:sg you:nom greedy:acc and eager:acc say:sg if quaereret aliena: ego autem dico, cupide et desire:sg of.another:acc I:nom but say:sg eagerly and avare non appetas nec tua. greedily not strive.for:sg even.not your:acc ‘Maybe—he says—you would call him greedy and avaricious, if he desires the things of others: but I say, you should not greedily and avariciously desire even your own things’ (Aug. serm. ) b. quare non laudant adulterium nec adulteri? why not praise:pl adultery:acc even.not adulterer:nom ‘Why is it the case that not even adulterers praise adultery?’ (Aug. serm. ) 62 I performed a query in the LLT-A on Plautus’ comedies, Cicero’s Letters ad Familiares and ad Atticum, and Livy, restricting myself to structures where n¯on and n˘ec co-occur, in this order, within three words from each other. Comprehensive corpus work would be needed to assess the frequency of the different wordorder possibilities for these constructions, as well as the existence of specific preferences for ‘doubling’ with correlative and scalar n˘ec.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

Interestingly, in the example from the Vulgate in (), the associated constituent is a minimizer, passum pedis ‘a step of the foot’, which reminds us of the origin of French NM pas. () et non dedit illi hereditatem in ea nec and not give:sg that:dat inheritance:acc in it:abl and.not passum pedis step:acc foot:gen ‘He gave him no inheritance here, not even enough ground to set his foot on’ (Vulg. act. .) An important generalization can be formulated on the basis of pleonastic uses of n˘ec (be they correlative or scalar): the word order is most typically one where the particle follows the co-occurring negative element (negative marker, negative indefinite, negative adverb). If n˘ec precedes, a double-negation reading ensues in Early and Classical Latin; these combinations are often employed as a device to express emphatic affirmation (as mentioned, the combination n˘ec n¯on is in fact probably lexicalized in this function): () a. nec nil hodie nec multo plus tu hic and.not nothing:acc today and.not much more you:nom here edes, ne frustra sis eat:sg lest uselessly be:sg ‘Today you will not eat nothing (= you will eat something), but here you will not eat much more (than nothing), don’t be fooled’ (Plaut. Capt. ) b. nec hoc ille non vidit nec this:acc he:nom not see:sg ‘he was not unaware of this’ (Cic. fin. .) In later or stylistically less controlled texts, this does not seem to necessarily be the case, and we find examples where the constituent marked by n˘ec precedes the main sentential negation (in the case below, we have a subject preceding the negated predicate): advocati eorum eos defendere non possint () sic nec so and.not advocate:nom they:gen they:acc defend:inf not can:pl ‘(as she could not defend it), so may also their advocates not be able to defend them’ (Defix. Tab. , second century (?)) Despite occasional exceptions, word order seems to be an important factor influencing redundancy. In order to show this point more clearly, I need to add to the picture some data concerning the additive and scalar particle ne . . . quidem, which was introduced in §.... As we saw, n˘ec gradually encroaches on the functional domain of ne . . . quidem, resulting in enhanced frequency for the use as standalone focus particle. Interestingly, ne . . . quidem shows a redundant behavior that, in Classical Latin, is even more systematic than what is observed with n˘ec.

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



The connective particle ne . . . quidem is sufficient to yield sentential negation on its own: in most cases, as seen in the examples in §..., it occurs as the only negatively marked element in its clause. As Orlandini (a: ) remarks, even if ne . . . quidem focuses narrowly on one constituent, the negative operator takes sentential scope. () ne eo quidem tempore quisquam loco cessit ne that:abl quidem time:abl anyone:nom place:abl retreat:sg ‘Not even then did any man yield his ground’ (Caes. BG ..) Uses as short answer and in similarly elliptical structures confirm this point and exclude an analysis in terms of an NPI: () NIC: pater, da mihi ducentos nummos father:voc give:sg I:dat two.hundred:acc coin:acc Philippos, te opsecro. CHRY: ne unum quidem Philippic:acc you:acc beg:sg ne one:acc quidem hercle, si sapis. by.Hercules if be.wise:sg Nicobulus: ‘father, give me two hundred Philippics, I beg you’. Chrisalus: ‘No, not even a single one if you’re in your right mind’ (Plaut. Bacch. –) () quem? Cognatum aliquem aut propinquum? non. who:acc relative:acc some:acc or friend:acc no Thermitanum aliquem, honestum hominem ac nobilem? ne Thermitanus:acc some:acc honest:acc man:acc and noble:acc ne id quidem. it:acc quidem ‘And who is it? Some relative or friend of his? No. Some respectable and important citizen of Thermae? Not that either.’ (Cic. Verr. ..) Despite its negative value, also ne . . . quidem, like n˘ec, may cause redundant expression of negation: cases are found quite frequently in Classical Latin, much more frequently with ne . . . quidem than with nec, and currently in the literary style. The discontinuous particle ne . . . quidem ‘doubles’ the NM as well as NIs (cf. ). Orlandini (a: ) states that the co-occurrence of ne . . . quidem with a further marker of negation in a single-negation reading is crucially tied to the syntactic positioning of the elements: if ne . . . quidem is postverbal, the NM n¯on is additionally needed in clause-initial position. This generalization fits cases such as () below, showing that no redundancy occurs if ne . . . quidem precedes the verb: () a. non enim praetereundum est ne id quidem not indeed overlook:ger be:sg ne this:nom quidem ‘and the following occurrence should not be overlooked’ (Cic. Verr. .) b. ne id quidem neglegendum est ne this:nom quidem disregard:ger be:sg ‘This should not be disregarded either’ (Cic. de orat. .).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

In other words, the generalization stated by Orlandini could be understood as a pre- / post-Infl asymmetry, and the distribution of ne . . . quidem would be the same as that of an n-word in a non-strict NC language. This conclusion is of course extremely intriguing in view of the diachronic process witnessed from Latin to Romance. However, the patterns we see in the texts are more complex than this: for instance, if it is true that in redundant structures the complex particle always follows the additional negative element, it is not always the case that the complex particle is found post-Infl, witness (): () me vero nihil istorum ne iuvenem quidem movit I:acc in.fact nothing:nom this:gen ne young:acc quidem move:sg umquam ever ‘even when I was young I was never attracted by anything of that sort’ (Cic. fam. ..) Despite the complexity of the observed patterns (on which see further Gianollo b: ch. ), a generalization emerges: what we do not find is ne . . . quidem in pre-Infl position co-occurring with another lower negative element (NI or NM) in post-Infl position. That is, the order is always NI / NM > ne . . . quidem, independently of prevs. post-Infl position. Recall that we made a similar remark on redundancy with n˘ec: also in this case the additional element precedes the focus particle. Orlandini makes a further observation that is in essence confirmed in my corpus, namely that, in case of pre-Infl co-occurrence, the further negative element accompanying ne . . . quidem is found at the beginning of the clause. More precisely, either it is the first constituent or it follows Topic-like constituents and occasionally secondposition particles. I interpret the distributional evidence as indicating that the additional negative element performs a function that, in principle, could be performed by the complex particle itself, but under some structural conditions is ‘handed over’ to another eligible candidate. I propose that this function is the expression of focus. In the next section I will spell out the consequences of this hypothesis for the syntax of redundancy involving n˘ec and ne . . . quidem. .. The analysis ... Focus Concord I concluded the last section by proposing a generalization: the negative element co-occurring with n˘ec and ne . . . quidem performs a function in which it acts in place of the particle itself, so that the particle can stay lower, either post-Infl or in the pre-Infl field, following the additional element. Now, the function at issue here cannot be negation, since both n˘ec and ne . . . quidem are [Neg] elements that routinely negate by themselves, independently of their position. But, I argue, another function that the two elements have in common is the expression of focus. Association with focus can be claimed to hold also for the additional negative elements in redundant structures on the basis of the distributional evidence: the additional element is found in the high left periphery of the clause; frequently it is the first

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



constituent (e.g., a), sometimes it follows a Topic and / or second-position particles (whose positioning is partly dependent on prosodic factors) (e.g., ). This suggests that the redundant elements stand in a sort of focus chain, connected to a single Focus operator in the left periphery. I will assume that the additional element, which is linearly first, sits in the Focus Phrase, and stands in an Agree relation with the further focused constituent, introduced by ne . . . quidem. Similarly, in the cases of redundancy with n˘ec, the focus particle enables (Multiple) Agree with other occurrences of n˘ec (in the correlative use) and with the additional negative element. That is, the function that the focus particles ‘hand over’ to the additional negative element is the local relation with the abstract focus operator sitting in the Focus Phrase, in a sort of Focus Concord. Syntactic Agree is made possible by the formal unintepretable feature [uFoc] that I assumed for n˘ec and that can be extended to ne . . . quidem. The assumption that the focus-sensitivity of the Latin particles has syntactic reflexes, in that it is encoded by means of formal features and enters Multiple Agree relations, is motivated by the fact that it is well known that focus systematically causes displacement in Latin (as Devine and Stephens :  put it, Latin separates the focus from the presupposed part of the clause overtly in the syntax). Many different displacement operations in Latin can be attributed to focus requirements, as extensively shown a.o. by Polo (), Devine and Stephens (), Spevak (), Danckaert (), Ledgeway ().63 Moreover, the existence of a Focus Criterion, parallel to a Neg Criterion and a Wh-Criterion in triggering movement to the higher portion of the clause has been proposed to explain a wide array of phenomena in a relevant number of languages (cf. Haegeman , , Rizzi ,  for discussion). I therefore assume that a focus-sensitive particle such as n˘ec carries a [uFoc] focus feature and necessarily has to overtly associate with an [iFoc] focus operator. This quite systematically corresponds with a clearly detectable syntactic operation in Latin: the focus-sensitive negative item is displaced from its base position in order to establish a visible syntactic relation with a clausal Focus projection. For concreteness, following much literature, I assume the following structure for the CP-left periphery of a Latin main clause: () Structure of the CP-Left Periphery in Latin (Danckaert : ) [ForceP [TopP* [PolP [FocP [TopP* [FinP . . . ] ] ] ] ] ] The relation between the particle and the clausal Focus projection may involve movement of the focused constituent (or part of it) to the Focus projection in the split CP, which is the usual option. In this case, the focused constituent (concretely, for us, the conjuncts of a correlative construction or the simplex constituent associated with the focus particle) end up before the finite verb: this explains the fact that constituents

63 According to Danckaert (: –) both presentational (= information) and identificational (= contrastive, exhaustive) foci are found in the CP-left periphery in Latin (which fits well with the fact that the same situation obtains in Early Romance).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

introduced by n˘ec and ne . . . quidem are overwhelmingly found in the clausal left periphery.64 But the fact that, in examples such as (a), the various elements contributing focused negation are not adjacent shows that at least in some more complex cases the focused elements are contained in different functional projections. Since in Latin also the low left periphery of the vP was active and could host focused elements (cf. Devine and Stephens :  and passim, Danckaert :  and passim), it is possible that also this lower Focus position participates in the Agree relation (which could help explain the post-Infl cases found in redundant structures). The crucial point is that, in these complex cases, the overt identification of the Focus operator in the CP-left periphery obtains through an element which, while establishing a Focus chain with the lower focus-sensitive particle (ne . . . quidem or n˘ec), ‘doubles’ the negative marker contained in it, resulting in redundancy in the marking of negation. ... From Focus Concord to Negative Concord Let us move now to an obvious question that arises at this point and is of central importance for our aims: how is this kind of Focus Concord connected to the syntax of negation, and to the observed redundancy? A further, related question concerns the reason why the redundancy is systematically found only in the case of n˘ec and ne . . . quidem. As concerns the first question, my proposal goes as follows. In a Double Negation language we do not expect Concord readings because (i) we do not have formal features associated with the expressions of negation and (ii) consequently, no NegP is syntactically active. In chapter  (§.) I have analyzed Latin nemo, nihil, nullus, etc. as negative indefinites ([Neg]) and I have treated them as items conveying existential quantification and associating with a negative operator that takes sentential scope. The negative operator and the indefinite associate in the syntax, as proposed by Zeijlstra (), and can in principle scope independently of each other. The negative element does not carry formal negation features, but is semantically negative ([Neg]). Although I have not discussed negatively marked adverbials like, e.g., numquam ‘never’ in detail, their homogeneous behavior with respect to negation (fundamentally, their being able to negate a clause by themselves with no positional restrictions) also lends itself to treatment in terms of [Neg]. Being devoid of formal negation 64 Since the positive counterpart quidem, following Danckaert’s (, ) analysis, is a marker of polarity focus, one could ask whether also negative ne . . . quidem, as well as n˘ec, is connected, more specifically, to the high polarity-focus position PolP in (). Polarity or verum focus is roughly understood as focus on the polarity value of the clause. Following Romero and Han (), it has also been treated as an epistemic operator expressing the speaker’s certainty that the proposition should be added to the Common Ground. Polarity focus has other important overt manifestations in Latin, most notably the verbfirst orders in broad focus yes–no questions (Devine and Stephens : –). Examples where the NM n¯on is preposed, such as (a), seem to suggest an interpretation in terms of polarity focus. Possibly also the correlative cases, to which we will shortly turn in conclusion of this section, may represent an instance of polarity focus. But scalar focus, the operator involved in negation strengthening, cannot be excluded, given the emphatic function of many of the redundant examples.

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



features, [Neg] elements do not establish Agree relations with each other or with a specific syntactic projection for negation. Thus, following again Zeijlstra, no active NegP is assumed for Classical Latin. Negative indefinites originate through a very local relation between the projection hosting the indefinite and a negative operator. Recall from §. that I proposed that the locality of Merge between the existential quantifier and the negative operator is due to constraints on the timing of interface-related operations: the indefinite has to be combined as soon as possible with the appropriate negative operator. Since negative indefinites host no formal features and thus there is no need to wait for the insertion of the appropriate [iNeg] counterpart in a dedicated syntactic projection, the syntactic licensing operation happens as soon as the existential quantification is introduced. Now, focus-sensitive negative items necessarily have to associate with a Focus operator. I have assumed that this semantic requirement corresponds to a syntactic operation in Latin: namely, because of their [uFoc] feature, the focus-sensitive negative item has to establish a visible syntactic relation with a clausal Focus projection. Thus, in the case of n˘ec and ne . . . quidem I do assume that an Agree operation takes place. However, this operation is triggered not by formal negation features, which remain absent from the system of Classical Latin in my approach, but by focus. This, however, has consequences also for the syntax of negation. The presence of [uFoc] features on the focus-sensitive negative item forces the computation to wait for the appropriate licensor, i.e., the [iFoc] Focus operator in the CP-domain. Now, my proposal is that this is precisely what enables the ‘redundant’ option. The fact that the satisfaction of a formal requirement has to be delayed to a later stage of the derivation allows for postponing the insertion of the negative operator as well. The general hypothesis is that earliness ‘ASAP’ economy principles in syntax are subordinated to an activity condition: an operation x can be delayed if the element would anyway still remain active for further feature-induced operations after x, and if x is possible at a later stage of derivation. In the case of the insertion of the negative operator, since the latter has to take sentential scope, and thus would anyway trigger a further (covert) movement operation, it is possible to wait and to satisfy all formal requirements of the focus-sensitive element at once later on. The mechanism at issue here must have the character of an optional strategy. Recall that in the case of n˘ec in Classical Latin, doubling by means of n¯on or an NI is rare and confined to specific registers; it is actually more the exception than the rule. For ne . . . quidem it seems to be much more systematic in the case of post-Infl material, although a broader corpus study would be needed here to assess this with precision; in any case it is not obligatory, since there are examples of postverbal ne . . . quidemconstituents that are not doubled by a preverbal negative element. In Late Latin, scalar n˘ec is more frequently found in redundant constructions, but also in this case we do not have a ‘doubling’ rule. We thus say that the condition seen above allows, but in no way forces, the insertion of a further negative element (probably because other contextual conditions, e.g. prosody, allow the identification of a non-dislocated constituent as focus).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

The result of the Agree process is that the focus particle undergoing Focus Concord is morphologically, but not semantically negative. The special conditions connected to focus allow for the lack of a negative operator. An example like (a), according to this proposal, has the following (simplified) structure: () [FocP Op[iFoc] non[Neg, uFoc] [FP praetereundumi [TP est [FocvP [OpP ne[ø] [DP id ] ] quidem[uFoc] [vP ti ] ] ] ] ] Now, given the already existing, focus-driven syntactic dependency, it is quite natural to assume that, once the preconditions for Negative Concord emerge in Latin with the reanalysis of the NM (cf. chapter ), these structures become perfect candidates to yield the first [uNeg] elements for the new grammar. The syntactic dependence between the feature pair [iFoc]—[uFoc] is easily overgeneralized to capture the relation between the negative elements, by assuming a further feature pair [iNeg] (the NM)—[uNeg] (the focus particles n˘ec and ne . . . quidem). This way, the reanalysis seen in () takes place. In increasingly taking over the functions of ne . . . quidem, and especially its scalar uses, n˘ec adopts its syntax as well. This process may have been favored not only by the functional similarity, but also by concomitant phenomena affecting correlative n˘ec (on which see §...). According to my reconstruction of the diachronic path, it is possible that in the language of Augustine (fourth to fifth centuries ce) the reanalysis has already happened. We saw in §.. that negative indefinites in Augustine’s Sermons are consistently pre-Infl, a fact that I took as an indication that the NM n¯on has already been reanalyzed as an [iNeg] element and the NegP projection is already activated. The examples in (), where n¯on co-occurs with scalar n˘ec, are compatible with a Negative Concord grammar and could be analyzed as instantiating an Agree chain involving both Focus and Negation formal features, as in (): () FocusOp[iFoc] .....non[uFoc,iNeg] .....nec[uFoc,uNeg] Importantly, I assume that the reanalysis leaves the [uFoc] feature on n˘ec intact. This conforms to the analysis of nec-words I gave in §..: the featural composition that I assumed there for scalar n˘ec after reanalysis () can be assumed to obtain in Late Latin, providing the prerequisite for the pan-Romance grammaticalization strategy that involves n˘ec and an indefinite base. As discussed in the previous section, a further step in the grammaticalization process must have involved the loss of the scalar component connected with the focus particle, and the switch from E-licensing to O-licensing (cf. §..). Still, also in their non-emphatic version, the new n-word retains a formal uninterpretable feature triggering the syntactic relation with a focus operator. As we will see in §., this can be argued to have important diachronic consequences in motivating the NPI-behavior of n-words in Early Romance. Let us come now to the second question we asked above, concerning the reason for the frequency of redundant constructions with n˘ec and ne . . . quidem: if we assume [uFoc]-Agree between the focus particle and the additional negative element, this means that the latter as well carries [uFoc] features (assuming the [iFoc] counterpart to always be represented by an abstract operator). However, this would incorrectly

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



allow for across-the-board Concord among [Neg] elements, which is for sure not the case in Latin. The mechanism explored here may well account for the famous examples of redundancy with negative indefinites seen in §.., but it would be obviously wrong to let it operate unconstrained. Consider also the data seen in chapter , showing that the pre-Infl placement of negative indefinites was not obligatory in Classical Latin: negative indefinites do not obligatorily have to associate with a Focus Phrase in Latin, as focus particles instead do. Therefore I propose that the presence of a [uFoc] feature on an NI or the NM is only optional, depending on their being contextually focused or not, whereas with focus particles it is obligatory, given their function. An optional [uFoc] feature on the additional negative element is projected only if, for some reason, it precedes the focus particle. Otherwise, it is blocked: the focus particle can establish an unmediated Agree relation with the operator and no further operation is needed. This also accounts for the generalization that we saw with n˘ec: a single-negation reading occurs only if n˘ec follows the other negative element, whereas if it precedes a double-negation reading obtains. A last issue I need to address concerns the status of the morpheme ne- in both n˘ec and ne . . . quidem in the redundant cases. After the reanalysis it can be quite straightforwardly analyzed as a mark of [uNeg] features, as in modern Romance n-words; possibly, as discussed in §.., at the time of the reanalysis, the particle n˘ec has been reanalyzed as a monomorphemic item inserted as a single head, with no need for positing a ‘negative’ projection. But which role does it have before reanalysis? I argued that the negative operator is inserted higher up in the structure, when the focus-related Multiple Agree operation is performed. Why is then the negative morpheme itself spelled out ‘low’, where it is not interpretable and does not carry formal uninterpretable features either? I think that the answer has to be sought in the semantic–pragmatic function that ne- has, despite its not being [Neg] in redundant structures. Namely, ne- carries the presupposition that the alternative(s) under discussion are negative. This is a necessary component of the scalar use, since the ‘even’ component has to scope over negated alternatives to yield the desired meaning effect. I will therefore assume that ne- is merged in the structure in order for this presupposition to be legible at the interfaces. ... A note on redundancy with correlative n˘ec Now that I have delineated my hypothesis, let me briefly come back to the correlative uses of n˘ec in Classical Latin. I left them out of the previous discussion because they are not directly relevant to the grammaticalization of n-words, which is tied to scalar n˘ec. Nonetheless, the cases of redundancy found with correlative n˘ec may also have played a role in the reanalysis. In order to keep the discussion manageable, I will only briefly summarize the longer treatment in Gianollo (b: ch. ). The main observation is that, also in this case, redundancy in the marking of negation appears to be strongly linked to the syntax of focus. For instance, in example (), the NM n¯on is detached from the main predicate and located in the first

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

position in the clause, a configuration that I interpreted as focus-induced movement in §.... In example (), the constituent hominem neminem can also be argued to be in a left-peripheral CP-Focus position (with inversion between the head noun and the NI possibly due to DP-internal Focus); an indication in favor of this analysis might be a.o. the position of the adverb quidem ‘certainly’, which, according to Danckaert (), sits in a Polarity Phrase above the Focus Phrase.65 () Non istanc aetatem oportet pigmentum ullum attingere, not this:acc age:acc be.proper:sg color:acc any:acc touch:inf neque cerussam neque Melinum neque aliam and.not white.lead:acc and.not Melian.clay:acc and.not other:acc ullam offuciam any:acc makeup:acc ‘Girls of your age shouldn’t touch any color, neither white lead nor Melian clay, or any other makeup.’ (Plaut. Most. –) () quo quidem hominem neminem potuisti nec mihi who:abl certainly man:acc nobody:acc can:sg and.not I:dat amiciorem nec, ut arbitror, ad ea cognoscenda friendly:comp.acc and.not as believe:sg to this:acc know:ger.acc quae scire volebam aptiorem which:acc know:inf want:sg qualified:comp.acc prudentioremve mittere sensible:comp.acc-or send:inf ‘You could not have sent me any friendlier or, as I suppose, better qualified or more sensible informant to tell me what I want to know.’ (Cic. fam. ..) The pattern in which the correlative particles follow the co-occurring negative element (negative marker, negative indefinite, negative adverb) is characteristic for this type of structure in Early and Classical Latin. Orlandini and Poccetti (: ) remark that the opposite order, with the correlative negations coming first, is exceptional: deorsum non cresco, nisi dominum () nec sursum nec and.not up and.not down not grow:sg if.not master:acc tuum in rutae folium non conieci your:acc in rue:gen leaf:acc not put:sg ‘I will not grow an inch up or down until I have put your master into a rue bush’ (Petr. .) Redundancy with correlative uses may have different reasons. In many cases the correlate is found in sentence-final positions, and the editors often indicate with a comma a possible sharp intonational break; this could be an indication that the structure is 65 Note that quidem is positive, and yet occurs in a negated assertion; this is expected if high-polarity particles of this kind are analyzed as markers of a verum operator, as in Romero and Han (): by scoping above a (covert) epistemic modal operator, they indicate the speaker’s certainty that the assertion has to be added to the Common Ground, independently of the polarity of the assertion itself.

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



not completely integrated in the syntax of the preceding clause, representing a case of afterthought resumptive negation, which may be argued to project an independent syntactic structure. In other cases, it is clear that two CPs are being coordinated, with one of them being elliptical: in this situation, there is no genuine redundancy, since each CP has its own negative operator. See, for instance, the example below, where one can argue that the second conjunct is a full CP with ellipsis of the verb (the copula ‘be’), and with n˘ec behaving as a [Neg] element: () non est inprobum nec inutile not is dishonest:nom and.not useless:nom ‘It is neither dishonest nor useless’ (Aug. epist. .) Also Cicero’s example in () could be argued to originate this way (with gapping of the verb deero ‘fail’ in the second conjunct): () non deero officio nec dignitati meae not fail:sg duty:dat and.not dignity:dat my:dat ‘I shall not fall short of my duty and position’ (Cic. Att. ..) [&P [CP non deero officio] & [CP nec deero dignitati meae] ] Although negation is not genuinely redundant in these cases, it may certainly give the impression of so being to speakers who have reanalyzed the NM n¯on as a [iNeg] element. Therefore, these examples are significant in a diachronic perspective, since they may have represented bridging contexts for the reanalysis of n˘ec as a [uNeg] element. In this respect, consider also that correlative negation, in virtue of its intrinsic nature, always brings about multiple marking of negation, which, although not strictly redundant, may yield structural ambiguity. That is, in correlative constructions, a further motivation for the [Neg] > [uNeg] reanalysis may play a role: a trigger for the reanalysis as a Concord element could be identified in cases where a disjunctive reading of the correlation, with the negative operator taking widest scope, obtains. The negation of a disjunction and the conjunction of the negated propositions are logically equivalent (one of De Morgan’s Laws, cf. ()). () ¬x ∧ ¬y ⇔ ¬(x ∨ y) Nevertheless, they require a different syntactic structure, since in the case of a conjunction analysis of negative correlatives two negative operators have to be present, whereas under a disjunction analysis only one, wide-scope negative operator is expected (cf. Wurmbrand  for the syntactic consequences of this fact). Orlandini (a,b) notes that, especially in contexts where the particle is embedded under a non-veridical operator, n˘ec can be interpreted as an inclusive disjunction. That is, n˘eque / n˘ec is equivalent to ‘non . . . aut . . . aut’ (cf. a) in certain contexts, such as, for example, under a possibility modal (b). And indeed, as Orlandini remarks, ancient grammarians listed n˘ec among disjunctive particles, together with,

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

e.g., aut. If the negative operator takes wide scope, n˘ec is one of the possible ways to realize narrow-scope disjunction (¬(x ∨ y)).66 () a. non famem aut sitim neque frigus neque not hunger:acc or thirst:acc and.not cold:acc and.not lassitudinem opperiri weariness:acc await:inf ‘they did not await the onset of hunger or thirst, of cold or of weariness’ (Sall. Catil. ., from Orlandini a: ) b. Sine quo nec beatus nec clarus without which:abl and.not happy:nom and.not famous:nom nec tutus quisque esse omnino potest and.not safe:nom anyone:nom be:inf at.all can:sg ‘Without that, it is completely impossible for anyone to be happy or famous or safe’ (Cic. Phil. ., from Orlandini a: ) Now, in the case of a disjunction interpretation, we have only one negation operator in the semantics, but two negative expressions in the morphosyntax. According to the framework I adopted in chapter , this morphosyntactic redundancy would necessarily lead to the assumption of formal features for negation and of Agree operations connected to them. The wide-scope derivation for negation might thus have led to the analysis of n˘ec as [uNeg] element, licensed by an abstract negative operator scoping on both / all correlates. If this proposal proves to be right, we would have a straightforward diachronic link with those Romance languages where the correlative particles deriving from n˘ec behave as Negative Concord elements (cf. Herburger , Doetjes ).67 This reanalysis could make sense of the Late Latin behavior of n˘ec seen in (): there, not all correlates are introduced by the particle; more precisely, absence of marking may be observed with the first conjunct (.a,b) or with all conjuncts but the last one (.c). This could be interpreted as a sign that here n˘ec has been reanalyzed as a [uNeg] element, with an abstract negative operator evoked by the particle’s uninterpretable features and taking wide scope on all the conjuncts, even those that are not explicitly marked as negative (something that would be impossible if n˘ec were still a [Neg] element, very locally introducing a negation operator). ... Summary and outlook Let me briefly summarize what we have concluded about the redundancy seen with n˘ec. I have connected the redundancy observed in the marking of negation with the focus-sensitivity of this particle. I have shown 66 The grammarian Priscian (fifth to sixth centuries ce), whose work represented the basis for the teaching of Latin during the Middle Ages, describes this value of n˘ec as follows: sub una enim abnegatione copulat res ‘it joins items under one negation’ (Prisc. gramm. GL III..–, from Orlandini : , n. ). 67 There are tricky differences concerning the behavior of correlative negation in the Romance languages, which cannot be discussed here. These differences have sometimes suggested their treatment as NPIs, and not as elements of Negative Concord, or ambiguity analyses, where two different lexical entries are assumed, one for the particle introducing the first conjunct and one for the successive particles. The interested reader is referred to de Swart (), Scorretti (), Herburger (), Doetjes (), Aranovich ().

OUP CORRECTED PROOF – FINAL, //, SPi

Redundant uses of n˘ec and the rise of Negative Concord



that redundancy is common to another scalar focus particle, ne . . . quidem, where redundant patterns are clear at an earlier age. I have also argued that the functional overlap between n˘ec and ne . . . quidem goes hand in hand with the spread of ‘doubling’ for Late Latin n˘ec, and I have highlighted the importance of this diachronic process for the rise of nec-words and, more generally, of Romance Negative Concord. The main intuition behind my proposal is that the various elements that are part of the focus of the clause must establish a visible syntactic relation with a covert focus operator in the left periphery of the CP. I modeled this as a form of agreement between formal uninterpretable features and the interpretable counterpart on the operator. I proposed a way in which redundant negation can be argued to be parasitic on this operation of ‘Focus Concord’: given that the location of Merge of the negative operator with negative indefinites is due to constraints on the timing of interfacerelated operations, this timing may be influenced by other activity requirements of the element in question. The formal feature [uFoc] allows for delaying the insertion of the negative operator until the further operation can also be performed, in the CPleft periphery. In the course of time, this leads to a process of reanalysis, probably already in Late Latin, whereby an uninterpretable formal feature starts to be assumed also for negation. My proposal is similar in spirit to Simpson and Wu’s () analysis. Simpson and Wu () argue for a general mechanism, according to which a functional projection may associate with a corresponding focus projection. Simpson and Wu (: ) assume the existence of a form of Focus Concord, which they define as ‘[t]he optional repetition of a property of a linguistic item X in a second locus Y for the express purpose of emphasizing this property of X’. A further idea they defend is that this form of Focus Concord may decay into simple formal agreement, with concomitant loss of emphasis. Watanabe () too has proposed a connection between negative and focus features in Negative Concord. However, in Watanabe’s system, the Focus feature is a purely formal device to allow Concord as syntactic agreement in his theory of feature checking, with no semantic counterpart.68 Nonetheless, many of his interesting generalizations should be reconsidered in view of the actual semantic role that focus may play in the licensing of Concord elements and, more generally, of dependent indefinites. In this respect, a framework which attributes a crucial role to focus-sensitivity in the interpretation and in the syntax of dependent indefinites is represented by Chierchia’s () theory of polarity sensitivity (cf. Labelle and Espinal  for a further application to Romance data). My own syntactic proposal represents a way of ‘syntacticizing’ the semantic characteristic of obligatory activation of alternatives (= my [uFoc]), triggering semantic exhaustification by a covert focus operator (= my [iFoc]), which Chierchia posits as the core prerogative of NPIs. With this in mind, in the next, concluding section I will go back to the Romance data seen at the very

68 Watanabe () analyzes the n-word as an inherently negative element that is rendered an active Goal for agreement processes by an uninterpretable focus (i.e., operator) feature, creating the pair [iNeg, uFoc], where the uninterpretable feature drives agreement.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

beginning of this chapter and, on the basis of what we have seen for Latin, I will sketch a proposal of analysis, whose validity will have to be investigated by further extensive research on Early Romance.

. Concord and polarity-sensitive uses: a unification? In §. I singled out a number of issues emerging from the description of the behavior of Early Romance indefinites in negative and other downward-entailing contexts. I now come back to the two main aspects that turned out to be particularly challenging for the theoretical analysis: (i) the NPI-distribution of negatively marked n-words and (ii) the unexpected optionality seen in pre-Infl Concord patterns.69 These phenomena will be the subject of the next two sections, where I will discuss how they can be interpreted in a new light given the Latin evidence we reviewed and the interpretation I gave. I will suggest some directions for future research that I think are promising in view of the conclusions reached on Latin. Concerning NPI uses, I will discuss the possibility that the focus-sensitivity of the nec-words inherited from Latin, in coexisting with the [uNeg] formal feature, is responsible for the broader distribution of nec-words in Early Romance. As for optional strict NC in Old Italian, I will propose that a reconsideration of the featural status of the Old Italian NM may be instrumental in accounting for these patterns. This will lead me to propose a revision of the typology of negative systems that I have adopted in this work, according to the suggestions already put forward in §... As a consequence of this move, I will suggest that the complex interaction between the behavior of indefinites and general constraints on clausal syntax is the key to understanding the unexpected strict Concord patterns of Early Romance. .. NPIs or n-words? The dependence between focus and ‘doubling’ that I have proposed in my account of n˘ec and, by extension, of nec-words represents a way of encoding in the syntax the semantic operation of exhaustification by focus-sensitive operators proposed by Chierchia () as the general mechanism licensing NPIs (cf. also Chierchia , ). In fact, Chierchia () explicitly provides a syntactic implementation of his proposal (where he is mainly interested in intervention effects) and models the relation between the polarity item and the licensing operator as a form of Agree.70 The licensing requirements of NPIs are captured, in Chierchia’s () approach, by positing that NPIs always obligatorily activate alternatives, unlike other quantificational elements (e.g., some, a), which do so based on relevance (that is, depending

69 Recall that NPI distribution is much less surprising (and actually is expected) for n-words with a positive origin, which I have consequently left out of the picture. The exclusion from my concluding discussion of other unexpected patterns with n-words, like, e.g., the occasional lack of Concord with postInfl n-words, is motivated by the fact that, as discussed §., recent research (especially Garzonio and Poletto , Poletto , Franco et al. a, Franco and Poletto ) could largely make sense of them. 70 For Agree as the underlying mechanism in the licensing of both NPIs and n-words see also Biberauer and Roberts (), Longobardi ().

OUP CORRECTED PROOF – FINAL, //, SPi

Concord and polarity-sensitive uses: a unification?



on contextual conditions). The obligatorily activated alternatives of NPIs are then factored in by exhaustifying them, i.e., letting them agree with a (covert) focus-sensitive operator. Varieties of NPIs come about as the result of the different types of alternatives that they may activate (subdomain alternatives, scalar alternatives) and of the different operators that may exhaustify them. The result of (cyclic) exhaustification must obey semantic and pragmatic requirements relating to the resulting informational content: NPIs will be ungrammatical in those environments where they yield a contradiction or a statement that is informatively too weak. The different exhaustification processes determine the constraints on the distribution of NPIs on the basis of the interaction with contextual conditions. Here I discuss a further aspect of Chierchia’s framework, where a unification is attempted between the licensing conditions for NPIs and those for Concord indefinites (n-words). As I did before, I will skip the more technical details and just present the minimum of information necessary to make my point. The aim will be to suggest that NPIs and n-words probably have more in common than I have assumed until now. In my treatment, based especially on work by Zeijlstra () and Penka and Zeijlstra (), I posited a clear-cut difference between the two types of indefinites. On the one hand, NPIs have been treated as indefinites whose felicity conditions are mainly dependent on pragmatic reasons (although we also saw that they obey precise syntactic constraints, cf. the ban on subject NPIs discussed at various points). On the other hand, n-words have been analyzed as highly grammaticalized Concord elements with a [uNeg] formal feature, whose licensing depends exclusively on the syntax. In the last years, this criterion has proved to have a high heuristic power in the study of indefinites belonging to the negation system, since it has allowed the setting of precise diagnostics and a better understanding of the phenomenon of Negative Concord. However, it seems to miss some generalizations in the case of ‘mixed’ systems like Early Romance. For the modern Romance languages, where n-words are only marginally attested in NPI uses, this fact has been interpreted as a quirkiness due to language change. Although an in-depth semantic analysis of the various NPI contexts could make sense of this distribution, NPI uses have nonetheless been regarded as nonsystematic remnants of a previous system. However, my work on the Latin origin of negatively marked n-words shows that this interpretation amounts to shifting the problem back in time: given the clear negative origin of nec-words, their appearance in non-negative NPI contexts is puzzling for Old French and Old Italian as well. It seems, thus, that when we are resolutely distinguishing between NPIs and n-words we are looking at two extremes of a system, sometimes also representing two extremes of a possible diachronic cline. What is needed is an improved understanding of intermediate manifestations of the determinants of polarity sensitivity and Negative Concord. Concretely, my analysis of the Latin data suggests that different combinations of the two features [uFoc] and [uNeg] are possible: () (i) [uFoc]

(ii) [uFoc, uNeg]

(iii) [uNeg]

Type (i) would correspond to elements that function only as NPIs; their [uFoc] feature would be the syntactic implementation of their exhaustification requirement. Type

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

(ii) would be instantiated by those indefinites that, under the appropriate conditions, convey negation by themselves (n-words) and at the same time can be non-negative in NPI contexts. Type (iii) would represent n-words that can only occur in the scope of a negation operator. An account of the ‘ambiguity’ of n-words based on a dual system of features has been previously proposed by a number of authors (a.o. Martins , Jäger , Longobardi ): in those systems, a feature [aff] (= affective) or [ANY] (= polaritysensitive) occupied the place of the [uFoc] feature above. However, the [uFoc] feature has a clearer etymological foundation, as I hope to have shown in my discussion of Latin n˘ec, and allows for a diachronic explanation concerning the makeup of the lexical entry in terms of common strategies of reinforcement and emphasis, as we have discussed in this chapter and in chapter . In addition to this, Chierchia’s () assumptions that different NPIs are subject to different procedures of exhaustification, by means of different operators, do not force us to assume that indefinites endowed with a [uFoc] feature are necessarily emphatic: depending on the kind of alternatives evoked by the indefinite (something which is encoded in their lexical entry), different operators will come into play. In §. we have discussed the operators E (connected to emphasis) and O (with no emphasis involved); Chierchia’s typology is in fact much more complex, encompassing a number of varieties of the two operators, and accounting this way for finegrained distinctions in the polarity landscape (cf. Chierchia :  for a schematic overview of the inventory). The [uFoc] feature is in principle compatible with either class of operators. According to my reconstruction, Early Romance varieties inherit nec-words from Latin with the featural specification [uNeg, uFoc]. An open question is whether in the various Romance varieties they still have a preference for the scalar focus operator E, as they had in Late Latin. The reduction of licensing contexts that Romance n-words undergo in their history plausibly corresponds to a shift in the licensing operator, as suggested in §., and in particular to a progressive shift of the indefinites from Type (ii) to Type (iii). Type (iii) in (), finally, would be instantiated by n-words that do not allow NPI-uses, such as the n-words of Slavic Negative Concord languages (but also the n-words of some Romance languages: cf. §..). Their distribution is usually explained in terms of paradigmatic effects (cf. §...), but a viable alternative hypothesis is that their feature composition is just not enough to allow them in non-negative NPI contexts. Namely, in lacking the [uFoc] feature, they can only be licensed by [iNeg]. In Chierchia’s system, this would amount to saying that they do not trigger alternatives and, thus, cannot trigger exhaustification by O.71, 72

71 Chierchia (: –) has a different proposal. 72 Maybe this could offer a key to explain the fact, noted in §.., that DP-internal inversion is unusually frequent with the n-word nenhum in European Portuguese: this language does not allow NPI-uses of n-words, pointing to an analysis as Type (iii) indefinite; the focus-driven syntactic operation of inversion could be interpreted as a strategy to make up for the lack of [uFoc] features on this indefinite, adding in the syntax what is missing in the lexical entry.

OUP CORRECTED PROOF – FINAL, //, SPi

Concord and polarity-sensitive uses: a unification?



A further question concerns the role of the negative feature in Type (ii). Chierchia (: –), based on previous work by Zeijlstra (, ) and Gajewski (), attempts a unification of Negative Concord and NPI licensing. The resulting analysis should apply to the indefinites of Type (ii) proposed in (). The fundamental idea is that Negative Concord is not purely syntactic, but is also semantically motivated by the same processes that motivate polarity sensitivity in general (this explains why Negative Concord is found with existential, end-of-scale quantifiers, and not with intermediate or strong quantifiers, cf. Chierchia : ). This means that n-words undergo Agree both with a negative operator and with an exhaustifier, namely the variant of O also found with strong NPIs (OALT ).73 As in Laka () and Ladusaw (), n-words are a subkind of NPIs. With n-words, an additional feature is present, [[n-D]], which is the translation in Chierchia’s framework of [uNeg]. This feature can enter Agree with an abstract negative operator, and represents the peculiarity of n-words with respect to the rest of NPIs. The entry proposed by Chierchia for the Italian n-word nessuno ‘no one’ is given below; the second line represents the type of activated subdomain and scalar alternatives (the latter are considered to be the numerals = Num). () entry for nessuno in Chierchia (: ):

nessuno[[n−D]] = λP ∃x ∈ D [person(x) ∧ P(x)]

nessuno[[n−D]] ALT = {λP ∃x ∈ D [person(x) ∧ n(x) ∧ P(x)] : D ⊆ D ∧ n ∈ Num} The licensing of this kind of n-word is carried out by the exhaustifier OALT , which takes scope over the entire clause. Negation plays no direct role in the licensing, but is necessary to avoid contradiction in the exhaustification process (Chierchia : ). The crucial difference between n-words and (other) NPIs is that n-words have a way to rescue themselves in pre-NegP position: thanks to their [[n-D]] feature, and as in Zeijlstra’s analysis, they can evoke an abstract negation operator c-commanding them. This creates the downward-entailing environment that is necessary for exhaustification to yield the correct result. Since NPIs do not have this possibility, if pre-NegP (pre-Infl) they end up in an upward-entailing context and the result of exhaustification is a contradiction. This machinery accounts for the use of indefinites of Type (ii), the [uFoc, uNeg] indefinites, in negative and strong NPI contexts. For many of these contexts research could show that an abstract negative operator is indeed present. In the case of licensing by superordinate negation, for instance, it has been assumed that certain types of subordinate clauses are more ‘transparent’ domains for licensing than others (Progovac , Penka ). In the case of predicates such as ‘doubt’ or ‘deny’, ‘before’-clauses, and also questions and standards of comparison, it has been assumed that they contain a negative operator, which can display the behavior of an [iNeg] element (cf. discussion in Zeijlstra  and Penka : –). 73 Based on Gajewski (), Chierchia assumes that exhaustification in this case looks not only at the assertive component (as with weak NPIs), but also at the presuppositional content: cf. Chierchia (: ).

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

As for weak NPI contexts, Chierchia (: –) proposes that n-words which can occur there (e.g. Italian mai ‘(n)ever’) also have an [[n-D]] feature, but undergo a different type of exhaustification (OW ). In conclusion, Chierchia’s system provides a promising way to account for the finegrained variation observed in the distribution of NPIs and for a unification with the phenomenon of Negative Concord. Since NPI-uses of n-words receive various motivations according to the environment in which they are found, the contexts where they appear in Early Romance should be examined on a case-by-case basis, something that I will not attempt here. I hope to have shown, thanks to my investigation of the Latin origin of Romance negatively marked n-words, that the focus-sensitive nature of their negative morpheme is a fundamental ingredient for the explanation of their historical path, since it motivates their ability to occur in non-negative NPI contexts without denying their fundamental Concord nature. Early Romance n-words obligatorily evoke alternatives, which are subject to an exhaustification procedure by means of focus-sensitive operators. This [uFoc] feature is an inheritance of their Latin origin, since the negative morpheme they contain was a focus-sensitive particle in Latin. In addition to this, and, again, ultimately in virtue of their etymological origin, Romance nec-words are also characterized by an uninterpretable negative formal feature [uNeg], which has rendered them able to convey negation by themselves since the earliest stages. .. Optional strict Negative Concord in Old Italian In this section I discuss the exceptions to non-strict Negative Concord that we saw for Old Italian in §... The relevant cases are those structures where more than one negatively marked item is present in the pre-Infl field of the clause, as in (), repeated from (c): ()

. . . e che neuno uomo non sapea che ne fosse adivenuto ‘and that no man knew what had happened to him’ (Novellino ..)

These structures are not possible, in a single-negation reading, in Modern Italian; in Old Italian, they alternate with the more familiar pattern from Modern Italian (and other non-strict varieties), where either an n-word or the NM surface in the pre-Infl position. In non-strict NC systems the co-occurrence of elements endowed with a formal negative feature in the pre-Infl area is not possible in a single-negation reading. In those pragmatic contexts where they are felicitous, structures with two such elements yield a double-negation reading, as in Italian () (capitals indicate focus accent): () NESSUNO non ha pagato la quota ‘No one has not paid the fee’ = ‘Everyone has paid the fee’ According to Zeijlstra’s (, ) analysis, which we have followed until now, in cases like (), two [iNeg] operators will be present, one in NegP, realized by the [iNeg] NM, and a higher, abstract one above the additional n-word. That is, the insertion of each element endowed with a formal negative feature is connected to an [iNeg] counterpart.

OUP CORRECTED PROOF – FINAL, //, SPi

Concord and polarity-sensitive uses: a unification?



Such a language as Old Italian challenges this typology: on the one hand, we see the NM and the n-words co-occurring pre-Infl in a single-negation reading, which should be interpreted as evidence for a [uNeg] analysis of the NM, as in strict systems; on the other hand, n-words can be found pre-Infl with a negative interpretation and with no co-occurring NM, which would speak for an [iNeg] analysis of the NM. In this section I sketch an alternative analysis with the aim of contributing to the solution of this paradox. The hypothesis I want to defend is, simply put, that more landing sites were available for negatively marked elements in Old Italian than in Modern Italian: this allowed for redundancy in the expression of negation in the CP–TP area. My hypothesis shifts the burden of explanation from the featural content of the involved negatively marked elements to the clausal structure of Early Romance. At the same time, it requires a modification of previous assumptions on the featural content of the items involved, according to the observations I made in §... In §.. we saw that, according to some scholars, in Modern Italian the Focus Phrase and the Negative Phrase can be syncretic. Under a sentential negation reading, an element with formal negative features can at the same time identify the scope of negation and the informational focus of the clause in the same syntactic projection. The syncretism hypothesis cannot plausibly be extended to Old Italian, whose CPfield was richer than in the Modern variety (Poletto , especially ch. ). This is, in fact, a characteristic common to all Early Romance varieties (Benincà , , Cruschina : ch. , Wolfe ). The richness of the clausal left periphery was connected with a broader range of movement possibilities, in particular for focused constituents: they could move (as in Modern Italian) to a preverbal Contrastive Focus position, but also (and unlike Modern Italian) to a preverbal Information Focus position. In Modern Italian, instead, Information Focus is restricted to the vP-periphery. The additional movement possibilities in Old Italian explain, a.o., the remarkable amount of preverbal object placement. Moreover, we know that, in matrix clauses and in most subordinate clauses, the verb in Old Italian rises higher than in Modern Italian (where it stays in Infl). Scholars locate its default landing site in the lowest head of the CP-field. Since negation always occurs before the verb, we have to assume that non at this point is a phonological clitic that attaches to the verbal constituent and moves with it.74 The important consequence of these facts for the analysis of Old Italian examples of strict NC such as () is that we have to assume that both the NM and the n-word are higher than in Modern Italian. The NM rises together with the verb to the lowest projection of the CP-field (FinP), and the constituent containing the n-word must be even higher. A possibility is that it is located in the Contrastive Focus

74 See Poletto (: –) for the status of non in Old Italian and Frascarelli (: ) on Modern Italian. Poletto remarks that sometimes in Old Italian non can function as first-position element for the placement of clitics, which is an indication that non, besides its clitic behavior on the verb, could also still be a prosodically autonomous element. Frascarelli analyzes non as a prosodically deficient element (owing to its being monosyllabic), which phonologically cliticizes on the verb (or on the cluster formed by the syntactic pronominal clitics and the verb) and moves with it. On variation in Italo-Romance varieties cf. Poletto and Garzonio ().

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

position: Cruschina () interprets fronting of n-words in Romance as a manifestation of contrastive focus, owing to ‘the contrastive nature of these elements on a scalar dimension as a factor affecting the relevance of the fronted focus’ (Cruschina : , n. ). There is also another possibility, namely that the n-word lands in the position used for QP-fronting (cf. §..). The latter represents a further movement option connected to operator properties, which is still visible in Ibero-Romance and has been investigated in particular for Spanish, Catalan, and European Portuguese (Quer , Leonetti and Escandell-Vidal , and, for the historical development in Spanish, Poole ). An example from Modern Spanish, as well as the example, repeated from chapter , (a), from Modern Italian (where this movement is much more restricted) are given below: () a. Algo debe saber Spanish ‘S/he has to know something’ (Leonetti and Escandell-Vidal : ) b. Niente concludi, stando in questo buco Italian ‘You won’t come to anything, staying in this hole’ (Benincà : ) The nature of QP-fronting, and in particular its connection to focus, is still under discussion, and it is not clear how to connect it to similar facts in French and Italian (see discussion in Cruschina : ch. , Poletto : ch. ). Leonetti and EscandellVidal () have connected it with verum focus (on which cf. §..). Thorough corpus work would be needed to ascertain in which positions and with which informational import Old Italian n-words can be found. Franco and Poletto () have recently shown that pre-Infl n-words can make different contributions to information structure and can occupy different positions in the left periphery of the clause. But even without reaching a precise conclusion on this point, we see that there are arguments for assuming that no syncretism of the NegP and the Focus Phrase takes place in Old Italian. This automatically frees up additional positions for n-words, and suggests the hypothesis that the NM and the n-word(s) are not in complementary distribution in the pre-Infl field of Old Italian, under the single-negation reading, because they do not occupy the same position. This is different from Modern Italian, where, in order to have an n-word in addition to the NM in pre-Infl position, a ‘special’ focus shell has to be projected, with the result that an additional abstract negative operator has to be inserted, deriving the obligatory double-negation reading. In Old Italian, instead, the abstract negative operator responsible for the single-negation reading would be inserted only once on top of the last element endowed with formal negative features. () a. Modern Italian: nessuno non sapeva ‘no one did not now’ = everyone knew [Op ¬ [FocP nessuno [NegP Op ¬ non [TP sapeva . . . ]]]] b. Old Italian: neuno uomo non sapea ‘no one knew’ [Op ¬ [FocP nessuno [FinP [NegP non [TP sapeva ]i [TP ti . . . ]]]]]

OUP CORRECTED PROOF – FINAL, //, SPi

Concord and polarity-sensitive uses: a unification?



Also the cases of pre-Infl adverbial n-words, which I mentioned in §.., could be interpreted as elements that land in a focus position, which is not available anymore in Modern Italian in the single-negation reading. The behavior of argumental n-words with respect to optional Concord would be due to the fact that they are not necessarily focused, and may be found in the Specifier of NegP, with no co-occurring NM. Recall also the elliptical structure I commented upon in §.., (), repeated here as (): () Che è ciò, Marco, ch’io ho avute sette robe e tu non niuna? ‘What is happening, Marco, that I had seven garments, and you no one?’ (Novellino ..) In Modern Italian, as seen in (), we would have ‘e tu nessuna’ (a structure possible in Old Italian as well), which I analyze as an elliptical structure where everything below the Focus Phrase is deleted. For Old Italian, assuming that a similar elliptical structure holds, we could analyze niuna as sitting in a Focus position, as in Modern Italian, and locate the NM non in the high Polarity Phrase that we saw for Latin in (). Also in this case, thus, it is possible to argue that the two elements occupy two different positions. In order to make this analysis viable, it is necessary to revise the typology of negative systems that I have adopted until now, along the lines I sketched in §... The present analysis assumes that all the pre-Infl elements that carry a formal negative feature can be part of an Agree chain, allowing the insertion of a single negative operator. This is not compatible with the analysis of the NM of non-strict NC systems as an [iNeg] element, assumed until now. The question is, therefore, whether one should assume a different analysis only for Old Italian, or whether, more generally, a different theoretical interpretation of the facts is called for. Assuming a different analysis for Old Italian would force the positing of an unplausible diachronic process of reanalysis: Late Latin [iNeg] > Old Italian [uNeg] > Modern Italian [iNeg]. Moreover, we would end up analyzing Old Italian as a strict NC variety and we would therefore be faced with the problem of accounting for the (very frequent) non-strict uses. I therefore propose to revise our typology according to the observations in §.., and to assume that NMs and, in fact, all elements of Concord, are [uNeg] elements, and that the [iNeg] counterpart is always provided by the abstract negative operator. The lack of co-occurrence between NM and n-words in pre-Infl position must then receive an alternative explanation, and various sources of differentiation are possible: in Modern Italian, for instance, there are not enough positions; in Late Latin there are no [uNeg] indefinites. In principle, in the system I have adopted until now, nothing prevents pre-Infl Agree among [uNeg] elements in non-strict NC systems. In a language such as Modern Italian, this configuration obtains under very constrained conditions: in the Standard variety it is current only when the co-occurring element is mai ‘never’ (which enjoys a special status among Italian n-words, cf. Panizza and Romoli ), cf. (); much more rarely and marginally it is found in sub-Standard varieties with adverbial PPs cooccurring with a subject n-word, as in (a–c). These structures would probably not be considered grammatical by some Italian speakers, but come up in the Internet; they find parallels in combinations such as in nessun modo niente, attested in pre-Infl position in Old Italian.

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

() Mai nessuno ha raggiunto questa vetta ‘No one has ever reached this peak’ () a. in nessun modo nessun’altra entità può pensare di poter competere ‘In no way can any other entity imagine being able to compete’ (http://dotrash.blogspot.co.uk///il-sondaggio-del-martedi-secondaparte.html) b. In nessun modo nessuno verrà a conoscenza dell’indirizzo ‘In no way will anyone get to know the address’ (www.cdigorla.it/Newsletter. htm) c. Niente, in nessun modo, deve prestarsi ad equivoci ‘Nothing, in any way, should be susceptible of being equivocated’ (http://www.avvenire.it/opinioni/pagine/patriciello-ancora-inchini.aspx) These examples can be accounted for by assuming that Multiple Agree occurs between the [uNeg] features of the pre-Infl (that is, pre-NegP) n-words and the [iNeg] abstract operator inserted as Last Resort on top of the structure. No double-negation reading arises because the negative operator is inserted only once. If enough positions for the n-words can be generated, the Concord reading is possible, as expected. If also the NM is a [uNeg] element, and if there are no positional constraints for its co-occurrence with further n-words, Concord will be possible with it too, as happens in Old Italian strict NC cases. In () and () more than one (constituent containing an) n-word surfaces in pre-Infl position: this, while confirming the marginal possibility of pre-Infl Negative Concord also in a language like Modern Italian, represents at first sight counterevidence for the claim that in Modern Italian the FocP and the NegP positions are syncretic. However, one could argue that in quite marginal structures such as (a– c) the multiple n-words form in fact a single, complex constituent and occupy just one position (Spec, Foc/NegP). Their focus contribution seems to be the same, namely the emphatic negation of the assertion: the Agree process involves both [uNeg] and [uFoc] features, and just one focused constituent is built. Alternatively, one may assume that, under certain pragmatic conditions, FocP and NegP may be projected separately also in Modern Italian, cf the discussion in §... A further issue has to do with how we obtain the double-negation readings we have in Italian: for an example like (), under a [uNeg]-analysis of the NM, one would expect Concord, as in strict NC languages and as in the Old Italian pleonastic constructions, and not a double-negation reading. But also in strict NC languages double-negation readings are possible, cf. the Romanian data in §... Therefore there must be a mechanism to block Agree in those languages as well. An obvious interpretational characteristic of examples like () is the contrast in polarity: Double-negation structures are used to reject the validity of the negated proposition. Therefore here we have a situation where the n-word in focus is clearly separated from the NM in the informational partition of the clause. The two syntactic projections are separate, and a separate focus shell is projected (cf. §..). Moreover, the additional abstract negative operator licensing the n-word is plausibly added in a

OUP CORRECTED PROOF – FINAL, //, SPi

Conclusions



higher, ‘external’ position typical of denials (cf. §...). As a consequence, neither Focus Concord nor Negative Concord obtains. Under my analysis, the theoretical difference between strict and non-strict NC systems dissolves: in both cases we have a grammar where overt [uNeg] elements are paired with covert [iNeg] operators. The fundamental characteristic of NC systems, unlike DN ones, is that a designated area of the clause requires the overt expression of sentential negation (where the scope of negation must be overtly signaled). In Romance this area is the CP–TP phase. Elements endowed with a negative formal feature, i.e., the NM or the n-words, can mark the scope of the negative operator, which corresponds to the informational focus of the clause. A subject n-word, as well as any other n-word ending up in the pre-NegP area, will extend the scope of negation to cover the pre-NegP position. Co-occurrence of multiple elements endowed with formal negative features in the CP–TP phase is constrained by independent structural factors related to the syntax of the clause.75 Penka () formulates a conceptually similar hypothesis concerning DN languages: she assumes a [uNeg] feature for all elements of the negation system, and (lexically encoded) licensing conditions by an obligatorily abstract operator (hence, the specification she proposes for the feature on elements of a DN system is [uNeg ∅]). Putting both hypotheses together, the difference between Double Negation and Negative Concord systems would reduce to whether negation has to be syntactically expressed in a special area of the clause (Negative Concord) or whether the abstract negative operator can be inserted in the locus of Merge of the negative element and then reach its scope position at LF (Double Negation), with no overt syntactic dependence being created with a syntactic NegP projection.

. Conclusions In this chapter I have investigated how the structural prerequisites for Negative Concord inherited from Late Latin (according to the hypothesis I put forward in chapter ) interact with the reanalysis of the old negative indefinites of Latin and with the creation of new indefinite pronouns and determiners licensed by negation. The trigger for this process is redundancy in the marking of negation, which, I argued, arises in Latin for reasons connected to emphatic negation. While the role of emphatic strengthening is routinely acknowledged in the literature on the change from Latin to Romance, I provided a novel implementation, which rests on a formalization of the relation between emphatic strengthening and focus. I argued that a key insight into the development comes from redundancy phenomena that are systematically attested already in Classical Latin and involve the negative focus particles n˘ec and ne . . . quidem. We saw how the former prevails over the latter and 75 I leave open the problem of how to account for the obligatoriness of co-occurrence of pre-Infl n-words with the NM in ‘well-behaved’ strict NC varieties. I believe that it could be connected to the different structural status of the negative marker in languages with obligatory co-occurrence: Zeijlstra () proposes that there are two different types of X NMs: the higher ones inserted in Neg and the lower ones inserted in the verb’s extended projection, the latter invariably [uNeg].

OUP CORRECTED PROOF – FINAL, //, SPi



Developments in negative and polarity-sensitive contexts from Latin to Romance

subsumes its uses and distribution in Late Latin; the relevance of this phenomenon stems from the fact that n˘ec is a morphological formative of many of the new Romance narrow-scope indefinites. I discussed how this class of nec-words consistently behaves as an element of the Negative Concord system of Early Romance, and I tried to trace the Concord behavior to the Late Latin stage. I analyzed the various functions of Latin n˘ec and showed that in the new indefinites it has the role of a focus-sensitive particle. On the basis of crosslinguistic parallels, we saw how nec-words fit into the class of ‘even’-indefinites and we discussed their semantic–pragmatic motivation as negation-strengtheners. Based on these premises, I developed a syntactic account of redundancy phenomena in Latin and suggested a path of reanalysis to derive the birth of the formal uninterpretable feature [uNeg] responsible for Romance Negative Concord. I concluded by discussing some distributional facts concerning n-words in Early Romance. In particular, I tried to show how the conclusions reached on the basis of the Latin data help us to see some phenomena in a new light. Although nec-words are Concord elements since the start in Romance, they also have frequent NPI-uses. I argued that, given their ability to be licensed by focus-sensitive operators in certain downward-entailing environments, nec-words in Romance do qualify as a special type of NPIs. Specifically, they are NPIs in Chierchia’s () sense: they obligatorily evoke alternatives that have to be factored into the global meaning of the structure they belong to; this is done by letting them agree with different types of focus operators; the ensuing inferences are responsible for their felicity in certain environments but not in others. Differently from ‘pure’ NPIs, however, Romance nec-words have also been endowed since the start with an uninterpretable negative formal feature [uNeg], i.e., they are n-words, and can be self-licensed by evoking an abstract negative operator in the appropriate contexts. Finally, the assessment of the origin of nec-words and of their behavior in Early Romance led to a novel proposal concerning optional strict Negative Concord in Old Italian. I proposed that the optional strict patterns are not dependent on the featural specification of the elements of Concord, but arise as the result of additional structural possibilities provided by Old Italian sentence structure, more precisely by its extended left periphery. As a consequence, I revised the typology of negative systems by putting forward the hypothesis that the theoretical difference between strict and non-strict systems in terms of formal features of the negative marker dissolves once other structural factors are taken into consideration.

OUP CORRECTED PROOF – FINAL, //, SPi

6 Conclusions In this work I have investigated variation and change in the domain of indefinite pronouns and determiners, focusing in particular on those indefinites that exhibit structural dependence, i.e., that need to be licensed by a hierarchically superior operator in the clause. I have shown how approaches that see in the semantic and syntactic internal structure of quantificational determiners a source of variation and change can account in a principled way for the constraints governing the synchronic distribution of indefinites, as well as for the historical processes targeting them. In particular, I have capitalized on the hypothesis that a fundamental dimension of variation in the system of indefinites is represented by conditions imposed by indefinites on their quantificational domains and their alternatives. I adopted analyses according to which these conditions are encoded, in the form of presuppositions, in the lexical entry of the indefinite item. I have shown that these conditions may undergo diachronic change, obeying pressures arising in language use, and related to expressivity and informativeness. This way, original contextual conditions, or preferences due to the pragmatic effects of competition with other indefinite items, may end up being incorporated into the lexical entry of the indefinite. A final step in this process may be the reanalysis of a semantic–pragmatic licensing condition as a syntactic one, i.e., as the postulation of a formal feature in the lexical entry. Indefinites are organized in series and undergo recurrent diachronic shifts in their meaning and distribution. The history of Latin and Romance represents a privileged domain of inquiry in order to establish how semantic change proceeds, and how it interacts with co-occurring contextual structural conditions. In order to interpret the observed clines of development, I first provided an analysis of Classical Latin specific and epistemic indefinites, concentrating my attention on the distribution of quidam ‘a certain’ and aliquis ‘some or other’ up to Late Latin (chapter ). I then followed the development of the Romance continuations of aliquis in a number of Standard Romance languages (Portuguese, Spanish, Catalan, French, Italian): the history of these items is particularly complex and—I argued—very relevant for our understanding of cyclical change and crosslinguistic differentiation (chapter ). In the second part of the book I moved to indefinites taking narrow scope with respect to negation: I described Latin negative indefinites in the context of the Classical Latin negation system, and I analyzed some Late Latin developments, connecting them to ongoing broader changes in sentence structure (chapter ). Finally, I investigated Indefinites between Latin and Romance. First edition. Chiara Gianollo. © Chiara Gianollo . First published  by Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi



Conclusions

some Early Romance developments in the domain of indefinites in direct negation and other downward-entailing contexts (chapter ). These indefinites are innovative in a number of ways. They typically replace the ‘old’ Latin negative indefinites with new products of grammaticalization, whose distribution is dependent on the new system of Negative Concord. I showed how their nature as elements of Negative Concord can be motivated diachronically, through an analysis of their building blocks, and can explain some facts related to their synchronic distribution. Also in this case a more precise analysis of the Latin starting point enables an improved understanding of their historical development in Romance. In all the case studies presented here, the motor of the change lies in phenomena already taking place in Late Latin. It is frequently the case that Romance languages differ profoundly from the Classical Latin stage, but, for the cases discussed here, they ‘differentiate in parallel’ because they inherit from Late Latin the fundamental prerequisites for later changes. I proposed that one such parallel development is Negative Concord: I interpreted data concerning the distribution of negative indefinites in Late Latin as a sign of the fact that already around the third to fourth centuries ce Latin changes from a Double Negation to a Negative Concord language. However, the change remains ‘latent’ for a while, its evidence consisting not in phenomena of morphosyntactic redundancy, but rather in word-order facts (the consistent positioning of negative indefinites in the CP–TP area). I treated the Late Latin system as a form of ‘concealed Negative Concord’. The gradual grammaticalization of new narrow-scope indefinites leads to the situation observed in Early Romance. The change affecting the syntax of sentential negation interacts with many of the phenomena discussed. Latin negative indefinites are directly affected by the change. But the very fact that they are not suitable anymore in the new developing system has a profound effect on other indefinites encroaching on the polarity-sensitive contexts. In particular, we have seen that the continuations of Latin epistemic indefinite aliquis expand into negative-polarity functions in Romance, and in some languages become part of the negation system. The class of epistemic indefinites turns out to be an important potential bridging context in cyclical change affecting indefinites. Their lexical entry forces the assumption of a minimally widened domain. Consequently, they trigger inferences that are subject to a pragmatic felicity condition. This causes, in turn, constraints on their distribution and their scopal possibilities: in this perspective, epistemic indefinites are ‘dependent’ indefinites. I argued that this is what qualifies them as potential candidates for the Quantifier Cycle (chapter ): epistemic indefinites are a special type of ‘positive’ indefinites. They are bound to appear in the scope of a licensing operator. The nature of the licensing relation may change over time, and I proposed that one prominent factor is emphatic (scalar) focus: focus interacts with the lexically activated domain alternatives, leading to a further widening of the domain of quantification. The association with focus, once conventionalized, may lead to the grammaticalization of originally optional implicatures. Following this line of analysis, I also drew a parallel with analogous processes seen in Jespersen’s Cycle.

OUP CORRECTED PROOF – FINAL, //, SPi

Conclusions



An important aspect that emerged in my work is the fact that the conditions governing the distribution of indefinite nominal phrases can be appreciated only by looking at the surrounding context, not only in semantic–pragmatic terms, but also in structural terms. In my work, the interplay with negation has played the most distinctive role in this respect. We also observed how the behavior of indefinites interacts with the way information structure is represented in the syntax, and especially with semantic and syntactic aspects connected to focus.

OUP CORRECTED PROOF – FINAL, //, SPi

OUP CORRECTED PROOF – FINAL, //, SPi

References Abbott, Barbara. . Reference. Oxford: Oxford University Press. Abels, Klaus and Luisa Martí. . A unified approach to Split Scope. Natural Language Semantics (). –. Aboh, Enoch. . Topic and Focus within D. In Leonie Cornips and Jenny Doetjes (eds.), Linguistics in the Netherlands, vol. , –. Amsterdam: Benjamins. Aboh, Enoch, Norbert Corver, Marina Dyakonova, and Marjo van Koppen. . DP-internal information structure: some introductory remarks. Lingua . –. Acquaviva, Paolo. . The Logical Form of Negation: A Study of Operator-Variable Structures in Syntax. New York: Garland. Adams, James Noel. . Wackernagel’s Law and the Placement of the Copula esse in Classical Latin. Cambridge: The Cambridge Philological Society. Adams, James Noel. . Bilingualism and the Latin Language. Cambridge: Cambridge University Press. Adams, James Noel. . The Regional Diversification of Latin,  BC–AD . Cambridge: Cambridge University Press. Aguilar-Guevara, Ana, Maria Aloni, Angelika Port, Radek Šimík, Machteld de Vos, and Hedde Zeijlstra. . Semantics and pragmatics of indefinites: Methodology for a synchronic and diachronic corpus study. In Proceedings of the Workshop ‘Beyond semantics: corpusbased investigation of pragmatic and discourse phenomena’, –. Bochumer Linguistische Arbeitsberichte. Ahn, Dorothy. . The semantics of additive either. In Eva Csipak and Hedde Zeijlstra (eds), Proceedings of Sinn und Bedeutung , –. Georg-August-Universität Göttingen. Alexiadou, Artemis, Liliane Haegeman, and Melita Stavrou. . Noun Phrase in the Generative Perspective. Berlin: de Gruyter. Aloni, Maria and Angelika Port. . Epistemic indefinites crosslinguistically. In Yelena Fainleib, Nicholas LaCara, and Yangsook Park (eds), Proceedings of NELS , –. Amherst, MA: GLSA Publications. Aloni, Maria and Angelika Port. . Epistemic indefinites and methods of indentification. In Luis Alonso-Ovalle and Paula Menéndez-Benito (eds), Epistemic Indefinites. Exploring Modality beyond the Verbal Domain, –. Oxford: Oxford University Press. Alonso-Ovalle, Luis and Elena Guerzoni. . Double negatives, Negative Concord and metalinguistic negation. In Papers from the Chicago Linguistic Society . The Main Session, –. Chicago, IL: University of Chicago Press. Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Some epistemic indefinites. In Makoto Kadowaki and Shigeto Kawahara (eds), Proceedings of NELS , –. Amherst, MA: GLSA Publications. Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Minimal domain widening. In Natasha Abner and Jason Bishop (eds), Proceedings of WCCFL , –. Somerville, MA: Cascadilla. Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Modal indefinites. Natural Language Semantics . –.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Domain restrictions, modal implicatures and plurality: Spanish algunos. Journal of Semantics . –. Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Two views on epistemic indefinites. Language and Linguistics Compass (). –. Alonso-Ovalle, Luis and Paula Menéndez-Benito. . Epistemic indefinites: an overview. In Luis Alonso-Ovalle and Paula Menéndez-Benito (eds), Epistemic Indefinites. Exploring Modality beyond the Verbal Domain, –. Oxford: Oxford University Press. Amaral, Patrícia. . When something becomes a bit. Diachronica (). –. Anscombre, Jean-Claude and Oswald Ducrot. . L’argumentation dans la langue. Brussels: Mardaga. Aranovich, Raúl. . A polarity-sensitive disjunction: Spanish ni . . . ni. In Jean-Pierre Montreuil and Chiyo Nishida (eds), New Perspectives on Romance Linguistics. Selected Papers from the th LSRL, vol. , –. Amsterdam: Benjamins. Batllori, Montserrat. . The significance of formal features in language change theory and the evolution of minimizers. In Pierre Larrivée and Chungmin Lee (eds), Negation and Polarity: Experimental Perspectives, –. Berlin: Springer. Bayer, Josef. . What Bavarian Negative Concord reveals about the syntactic structure of German. In Joan Mascaró (ed.), Grammar in Progress: GLOW Essays for Henk van Riemsdijk, –. Dordrecht: Foris. Bayer, Josef. . Nominal negative quantifiers as adjuncts. Journal of Comparative Germanic Linguistics . –. Beaver, David and Brady Clark. . Sense and Sensitivity. How Focus Determines Meaning. London: Wiley Blackwell. Beaver, David and Henk Zeevat. . Accommodation. In Gillian Ramchand and Charles Reiss (eds), The Oxford Handbook of Linguistic Interfaces, –. Oxford: Oxford University Press. Beghelli, Filippo and Tim Stowell. . Distributivity and negation: The syntax of each and every. In Anna Szabolcsi (ed.), Ways of Scope Taking, –. Dordrecht: Kluwer. Belletti, Adriana. . Italian/Romance clitics: Structure and derivation. In Henk van Riemsdijk (ed.), Clitics in the Languages of Europe, –. Berlin/New York: de Gruyter. Belletti, Adriana. . ‘Inversion’ as focalization. In Aafke Hulk and Jean-Yves Pollock (eds), Subject Inversion in Romance and the Theory of Universal Grammar, –. Oxford: Oxford University Press. Belletti, Adriana. . Aspects of the low IP area. In Luigi Rizzi (ed.), The Structure of CP and IP. The Cartography of Syntactic Structures, vol. , –. Oxford: Oxford University Press. Bende-Farkas, Ágnes. . Focus and Negative Concord in Hungarian. In Paul Dekker and Michael Franke (eds), Proceedings of the th Amsterdam Colloquium, –. ILLC: University of Amsterdam. Benincà, Paola. . Un’ipotesi sulla sintassi delle lingue romanze medievali. Quaderni patavini di linguistica . –. Benincà, Paola. . L’ordine degli elementi della frase e le costruzioni marcate. In Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds), Grande grammatica italiana di consultazione, vol. , –. Bologna: Il Mulino. Benincà, Paola. . A detailed map of the left periphery of Medieval Romance. In Raffaella Zanuttini, Héctor Campos, Elena Herburger, and Paul Portner (eds), Negation, Tense and

OUP CORRECTED PROOF – FINAL, //, SPi

References



Clausal Architecture: Cross-linguistic Investigations, –. Washington, DC: Georgetown University Press. Bergareche, Bruno Camus and Manuel Pérez Saldanya. . From adjectives to quantifiers. Evidence from Old and Modern Catalan. Catalan Journal of Linguistics . –. Bernini, Giuliano and Paolo Ramat. . Negative Sentences in the Languages of Europe: A Typological Approach. Berlin: de Gruyter. Bernstein, Judy. . Focusing the ‘right’ way in Romance determiner phrases. Probus . –. Bertocchi, Alessandra. . Some properties of ‘ipse’. In Hannah Rosén (ed.), Aspects of Latin: Papers from the th International Colloquium on Latin Linguistics, –. Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft. Bertocchi, Alessandra. . ‘Ipse’ as an intensifier. Journal of Latin Linguistics (). –. Bertocchi, Alessandra and Mirka Maraldi. . Indefinite pronouns in conditional clauses. Journal of Latin Linguistics (). –. Bertocchi, Alessandra and Mirka Maraldi. . Menaechmus quidam: indefinites and proper nouns in Classical and Late Latin. In Carmen Arias Abellán (ed.), Latin vulgaire-latin tardif VII. Actes du VIIe Colloque international sur le latin vulgaire et tardif, –. Seville: Universidad de Sevilla. Bertocchi, Alessandra and Mirka Maraldi. . Latin free-choice pronouns in relation to one another (and to other indefinites). Journal of Latin Linguistics (). –. Bertocchi, Alessandra, Mirka Maraldi, and Anna Orlandini. . Quantification. In Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol. , –. Berlin: de Gruyter. Berwick, Robert. . The Acquisition of Syntactic Knowledge. Cambridge, MA: MIT Press. Bescherelle, Louis-Nicolas and Henri Bescherelle. . Grammaire nationale. Paris: Simon. Bianchi, Valentina and Roberto Zamparelli. . Edge coordinations: focus and conjunction reduction. In David Adger, Cécile de Cat, and George Tsoulas (eds), Peripheries: Syntactic Edges and their Effects, –. Berlin: Springer. Biberauer, Theresa. . Competing reinforcements. When languages opt out of Jespersen’s Cycle. In Ans van Kemenade and Nynke de Haas (eds), Historical Linguistics : Selected Papers from the th International Conference on Historical Linguistics, Nijmegen, – August , –. Amsterdam: Benjamins. Biberauer, Theresa and Ian Roberts. . Changing EPP parameters in the history of English: accounting for variation and change. English Language and Linguistics (). –. Biberauer, Theresa and Ian Roberts. . Negative words and related expressions: a new perspective on some familiar puzzles. In Richard Ingham and Pierre Larrivée (eds), The Evolution of Negation. Beyond the Jespersen Cycle, –. Berlin: de Gruyter. Biberauer, Theresa and Hedde Zeijlstra. . Negative Concord in Afrikaans: filling the typological gap. Journal of Semantics (). –. Biberauer, Theresa, Ian Roberts, and Michelle Sheehan. a. No-choice parameters and the limits of syntactic variation. In Robert Santana-LaBarge (ed.), Proceedings of WCCFL , –. Somerville, MA: Cascadilla Press. Biberauer, Theresa, Anders Holmberg and Ian Roberts. b. A syntactic universal and its consequences. Linguistic Inquiry (). –.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Borer, Hagit. . In Name Only Structuring Sense vol. . Oxford: Oxford University Press. Bortolussi, Bernard. . Sur la distinction entre place et position syntaxique en latin. L’exemple de quis (quelque, quelqu’un). In J. M. Marandin (ed.), Cahier Jean-Claude Milner, –. Paris: Verdier. Bortolussi, Bernard. . Quelques quelque(s) en latin. In Lucia Tovena (ed.), Déterminants en diachronie et synchronie, –. Paris: Project ELICO publications. http://elico.linguist.univparis-diderot.fr/Bortolussi.pdf. Bortolussi, Bernard. . Syntaxe des indéfinis latins. Quis, quisque, alius. Paris: Presses de l’Université Paris-Sorbonne. Bortolussi, Bernard and Lyliane Sznajder. . Quelques emplois de quis dans la Vulgate. In Camille Denizot and Emmanuel Dupraz (eds), Latin quis-qui, grec tís/tis: parcours et fonctionnements. Études sur deux interrogatifs-indéfinis-relatifs, –. Mont-Saint-Aignan: Presses universitaires de Rouen et du Havre. Breitbarth, Anne. . The History of Low German Negation. Oxford: Oxford University Press. Brennan, Jonathan. . Only finally. NYU Working Papers in Linguistics –. Brunetti, Lisa. . Information focus movement in Italian and contextual constraints on ellipsis. In Gina Garding and Mimu Tsujimura (eds), Proceedings of WCCFL , –. Somerville, MA: Cascadilla Press. Buridant, Claude. . Grammaire nouvelle de l’ancien français. Paris: SEDES. Cardinaletti, Anna. . Syntactic effects of cliticization. In Tibor Kiss and Artemis Alexiadou (eds), Syntax - Theory and Analysis, vol. . HSK, –. Berlin: de Gruyter. Carlier, Anne. . La genèse de l’article un. Langue française . –. Carlier, Anne. . Grammaticalization in progress in Old French: indefinite articles. In Deborah Arteaga (ed.), Research on Old French: The State of the Art, –. Amsterdam: Benjamins. Catalani, Luigi. . Die Negation im Mittelfranzösischen. Frankfurt: Lang. Cecchetto, Carlo. . Doubling structures and reconstruction. Probus (). –. Chatzopoulou, Katerina. . Negation and Nonveridicality in the History of Greek: University of Chicago dissertation. Chatzopoulou, Katerina. . The Greek Jespersen’s cycle: Renewal, stability and structural microelevation. In Chiara Gianollo, Agnes Jäger, and Doris Penka (eds), Language Change at the Syntax-Semantics Interface, –. Berlin: de Gruyter. Chierchia, Gennaro. . Scalar implicatures, polarity phenomena and the syntax/pragmatics interface. In Adriana Belletti (ed.), Structures and Beyond, –. 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. Cinque, Guglielmo. . Adverbs and Functional Heads. Oxford: Oxford University Press. Clackson, James and Geoffrey Horrocks. . The Blackwell History of the Latin Language. Malden, MA: Blackwell. Collins, Chris. . Economy conditions in syntax. In Mark Baltin and Chris Collins (eds), The Handbook of Contemporary Syntactic Theory, –. Oxford: Blackwell.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Combettes, Bernard. . La grammaticalisation d’un déterminant indéfini: Quelque en moyen français. Scolia . –. Condoravdi, Cleo. . Not knowing or caring who. Handout of presentation at the Workshop Context and Content, LSA Summer Institute MIT / Harvard. Condoravdi, Cleo and Ashwini Deo. . Aspect shifts in Indo-Aryan and trajectories of semantic change. In Chiara Gianollo, Agnes Jaeger, and Doris Penka (eds), Language Change at the Syntax-Semantics Interface, –. Berlin: de Gruyter. Crisma, Paola. . The ‘indefinite article’ from cardinal to operator to expletive. In Chiara Gianollo, Agnes Jäger, and Doris Penka (eds), Language Change at the Syntax-Semantics Interface, –. Berlin: de Gruyter. Croft, William. . Typology and Universals. Cambridge: Cambridge University Press, nd edn. Cruschina, Silvio. . Discourse-Related Features and Functional Projections. Oxford: Oxford University Press. Dahl, Östen. . Book review of Martin Haspelmath, Indefinite Pronouns. Linguistics and Philosophy . –. Dahl, Östen. . Inflationary effects in language and elsewhere. In Joan Bybee and Paul Hopper (eds), Frequency and the Emergence of Linguistic Structure, –. Amsterdam: Benjamins. Danckaert, Lieven. . Latin Embedded Clauses. Amsterdam: Benjamins. Danckaert, Lieven. . Quidem as a marker of emphatic polarity. Transactions of the Philological Society (). –. Danckaert, Lieven. . The Development of Latin Clause Structure. A Study of the Extended Verb Phrase. Oxford: Oxford University Press. Davison, Alice. . Negative scope and rules of conversation: evidence from an OV language. In Peter Cole (ed.), Pragmatics. Syntax and Semantics , –. New York: Academic Press. DeGraff, Michel. . A riddle on negation in Haitian. Probus . –. de la Fuente, Olegario García. . Los indefinidos en la Biblia latina. Emerita (). –. de la Villa, Jesús. . Numerals. In Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol. , –. Berlin: de Gruyter. Delbrück, Berthold. . Vergleichende Syntax der indogermanischen Sprachen, vol. II. Strasbourg: Trübner. den Dikken, Marcel. . Either-float and the syntax of co-or-dination. Natural Language and Linguistic Theory . –. Deni´c, Milica. . Bagel Problem and NPI-PPI distribution. Handout of presentation at the Workshop of the Göttingen Spirit Summer School on Negation, Universität Göttingen. Deo, Ashwini. . Diachronic semantics. Annual Review of Linguistics . –. Déprez, Viviane. . Two types of Negative Concord. Probus . –. Déprez, Viviane. . Parallel (a)symmetries and the structure of negative expressions. Natural Language and Linguistic Theory . –. Déprez, Viviane. . Atoms of negation: An outside-in micro-parametric approach to negative concord. In Richard Ingham and Pierre Larrivée (eds), The Evolution of Negation. Beyond the Jespersen Cycle, –. Berlin: de Gruyter. Déprez, Viviane and France Martineau. . Micro-parametric variation and Negative Concord. In Julie Auger, J. Clancy Clements, and Barbara Vance (eds), Contemporary

OUP CORRECTED PROOF – FINAL, //, SPi



References

Approaches to Romance Linguistics: Selected Papers from the rd LSRL, –. Amsterdam: Benjamins. Déprez, Viviane, Susagna Tubau, Anne Cheylus, and M. Teresa Espinal. . Double Negation in a Negative Concord language: An experimental investigation. Lingua . –. de Swart, Henriëtte. . Négation et coordination: la conjonction ‘ni’. In Reineke BokBennema, Bob de Jonge, Brigitte Kampers-Mahne, and Arie Molendijk (eds), Adverbial Modification, –. Amsterdam: Rodopi. de Vaan, Michiel. . Etymological Dictionary of Latin and the other Italic Languages. Leiden: Brill. Devine, Andrew M. and Laurence D. Stephens. . Latin Word Order. Structured Meaning and Information. Oxford: Oxford University Press. Devine, Andrew M. and Laurence D. Stephens. . Semantics for Latin. An Introduction. Oxford: Oxford University Press. Dhoukar, Asma. . Aucune chose, une construction pas comme les autres. In Adam Ledgeway, Michela Cennamo, and Guido Mensching (eds), Actes du XXVIIe Congrès international de linguistique et de philologie romanes. Section : Syntaxe, –. Nancy: ATILF. Dickey, Eleanor and Anna Chahoud (eds). . Colloquial and Literary Latin. Cambridge: Cambridge University Press. DIEC. Institut d’estudis catalans [Author]. . Diccionari de la llengua catalana. Segona edició. (Online version https://mdlc.iec.cat). Barcelona: Enciclopèdia catalana. Diesing, Molly. . Indefinites. Cambridge, MA: MIT Press. Doetjes, Jenny. . The chameleonic nature of French ni: negative coordination in a negative concord language. In Emar Maier, Corien Bary, and Janneke Huitink (eds), Proceedings of Sinn und Bedeutung , –. Nijmegen: Nijmegen Center of Semantics. Dupraz, Emmanuel. . Sur le sémantisme d’aliquis en latin classique. In Camille Denizot and Emmanuel Dupraz (eds), Latin quis-qui, grec tís/tis: parcours et fonctionnements. Études sur deux interrogatifs-indéfinis-relatifs, –. Mont-Saint-Aignan: Presses universitaires de Rouen et du Havre. Ebert, Christian, Cornelia Ebert, and Stefan Hinterwimmer. . The semantics and pragmatics of bestimmt and gewiss. In Seda Kan, Claire Moore-Cantwell, and Robert Staubs (eds), Proceedings of NELS , –. Amherst, MA: GLSA Publications. Eckardt, Regine. . Reanalyzing ‘selbst’. Natural Language Semantics . –. Eckardt, Regine. . Eine Runde im Jespersen-Zyklus. Negation, emphatische Negation, negativ-polare Elemente im Altfranzösischen. KOPS, University of Konstanz, http://nbnresolving.de/urn:nbn:de:bsz:-opus-. Eckardt, Regine. . Meaning Change in Grammaticalization. An Enquiry into Semantic Reanalysis. Oxford: Oxford University Press. Eckardt, Regine. . The many careers of negative polarity items. In Kristin Davidse, Tine Breban, Lieselotte Brems, and Tanja Mortelmans (eds), Grammaticalization and Language Change: New Reflections, –. Amsterdam: Benjamins. Edmondson, Jerold A. and Frans Plank. . Great Expectations. An intensive self analysis. Linguistics and Philosophy (). –. Eguren, Luis and Cristina Sánchez. . Grammar and history of Spanish cierto. In Georg Kaiser and Manuel Leonetti (eds), Proceedings of the Workshop Definiteness, Specificity and Animacy in Ibero-Romance Languages, –. Universität Konstanz: Arbeitspapier .

OUP CORRECTED PROOF – FINAL, //, SPi

References



Enç, Mürvet. . The semantics of specificity. Linguistic Inquiry . –. Endriss, Cornelia. . Quantificational Topics. A Scopal Treatment of Exceptional Wide Scope Phenomena. Berlin: Springer. Ernout, Alfred and François Thomas. . Syntaxe Latine. Paris: Klincksieck. Etxeberria, Urtzi and Anastasia Giannakidou. . Illusory specificity with Spanish algunos. Presentation at University of Trento, February , . Etxeberria, Urtzi and Anastasia Giannakidou. . A unified analysis of Spanish algún/algunos: anti-specificity, plurality, and NP-anaphora. Ms. Fˇalˇau¸s, Anamaria. . Alternatives as sources of semantic dependency. In Proceedings of SALT , –. LSA. Fˇalˇau¸s, Anamaria. . (Partially) free choice of alternatives. Linguistics and Philosophy . –. Farkas, Donka. . Specificity and scope. In Lea Nash and George Tsoulas (eds), Actes du Premier Colloque Langues & Grammaire, –. Paris: Université Paris VIII. Farkas, Donka. . Varieties of indefinites. In Proceedings of SALT , –. LSA. Farkas, Donka. . The semantics of determiners. In Carmen Dobrovie Sorin and Ion Giurgea (eds), A Reference Grammar of Romanian, ch. , –. Amsterdam: Benjamins. Fauconnier, Gilles. . Pragmatic scales and logical structure. Linguistic Inquiry (). –. FEW, Walther von Wartburg et al. . Französisches Etymologisches Wörterbuch. Atilf online version https://apps.atilf.fr/lecteurFEW/. Fleck, Frédérique. . Interrogation, coordination et subordination: le latin quin. Paris: Presses de l’Université Paris-Sorbonne. Fodor, Janet and Ivan Sag. . Referential and quantificational indefinites. Linguistics and Philosophy . –. Fortson, Benjamin W. IV. . Indo-European Language and Culture. An Introduction. London: Wiley Blackwell, nd edn. Foulet, Lucien. . ‘Quelque’. Romania . –. Foulet, Lucien. . Petite syntaxe de l’ancien français. Paris: Champion rd revised edition. Franco, Irene and Cecilia Poletto. . Optional Negative Concord with preverbal n-words. Presentation at Mini-workshop on Negative Quantification, Goethe-Universität Frankfurt, June , . Franco, Irene, Olga Kellert, Guido Mensching, and Cecilia Poletto. . A diachronic study of negative additives. Presentation at the Göttingen Spirit Summer School on Negation. Franco, Irene, Olga Kellert, Guido Mensching, and Cecilia Poletto. a. On (negative) indefinites in Old Italian. Presentation at DGfS , Konstanz. Franco, Irene, Olga Kellert, Guido Mensching, and Cecilia Poletto. b. A diachronic study of the (negative) additive ‘anche’ in Italian. Caplletra. Revista Internacional de Filologia . –. Frascarelli, Mara. . The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. Berlin: Springer. Fruyt, Michèle. a. Négation et grammaticalisation en latin. De lingua Latina. Revue de Linguistique Latine du Centre Alfred Ernout . –. Fruyt, Michèle. b. Origine de la négation n¯e dans lat. n¯e . . . quidem, n¯equam, n¯equ¯ıquam. De lingua Latina. Revue de Linguistique Latine du Centre Alfred Ernout . –.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Fruyt, Michèle. . Grammaticalization in Latin. In Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol. , –. Berlin: de Gruyter. Fruyt, Michèle and Olga Spevak (eds). . La quantification en latin. Paris: Centre Alfred Ernout. Gajewski, Jon. . Licensing strong NPIs. Natural Language Semantics . –. Garassino, Davide. . The development of diversi from Latin to Old Italian. In Kateˇrina Loudová (ed.), Proceedings of Early European Languages in the Eyes of Modern Linguistics, –. Brno: Masaryk University. Garassino, Davide. . Dalla certezza epistemica all’indeterminatezza: il percorso di certus dal latino all’italiano antico. Vox Romanica . –. Garzonio, Jacopo. . Not even a crumb of negation: on mica in Old Italian. Unpublished ms, Università di Padova. Garzonio, Jacopo and Chiara Gianollo. . From nec to né: the interaction of focus and negation. Presentation at IGG , ... Garzonio, Jacopo and Cecilia Poletto. . On niente: optional negative concord in Old Italian. Linguistische Berichte . –. Garzonio, Jacopo and Cecilia Poletto. . The negative marker that escaped the cycle: some notes on manco. In Carla Contemori and Lena Dal Pozzo (eds), Inquiries into Linguistic Theory and Language Acquisition. Papers offered to Adriana Belletti, –. Siena: CISCL Press. Gast, Volker. . The Grammar of Identity: Intensifiers and Reflexives in Germanic Languages. London: Routledge. Gast, Volker. . At least, wenigstens and company: Negated universal quantification and the typology of focus quantifiers. In Kook-Hee Gil, Stephen Harlow, and George Tsoulas (eds.), Strategies of Quantification, –. Oxford: Oxford University Press. Gast, Volker and Johan van der Auwera. . Scalar additive operators in the language of Europe. Language (). –. Gergel, Remus. . ‘Rather’: On a modal cycle. In Elly van Gelderen (ed.), Cyclical Change, –. Amsterdam: Benjamins. Gergel, Remus. . A route towards specialized epistemic indefinites: the case of ‘vreun’ (and the relationship to its cognates). Presentation at DGfS , Konstanz. Geurts, Bart. . The mechanisms of denial. Language (). –. Giannakidou, Anastasia. . Polarity Sensitivity as (Non)veridical Dependency. Amsterdam: Benjamins. Giannakidou, Anastasia. . Negative . . . concord? Natural Language and Linguistic Theory . –. Giannakidou, Anastasia. . Negative and positive polarity items. In Klaus von Heusinger, Claudia Maienborn, and Paul Portner (eds), Semantics: An International Handbook of Natural Language Meaning, vol . HSK, –. Berlin/New York: de Gruyter. Giannakidou, Anastasia and Josep Quer. . What we gain, and what we lose, with a Hamblin semantics for indefinites: Free choice and referential vagueness. Handout of presentation at the Workshop on Alternative Semantics, Nantes http://www.alternative-semantics. univ-nantes.fr/. Giannakidou, Anastasia and Josep Quer. . Exhaustive and non-exhaustive variation with free choice and referential vagueness: Evidence from Greek, Catalan, and Spanish. Lingua . –.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Gianollo, Chiara. . Constituent Structure and Parametric Resetting in the Latin DP: A Diachronic Study: Università di Pisa dissertation. Gianollo, Chiara. . Native syntax and translation effects. Adnominal arguments in the Greek and Latin New Testament. Oslo Studies in Language (). –. Gianollo, Chiara. . Prepositional genitives in Romance and the issue of parallel development: from Latin to Old French. In Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds), Parameter Theory and, Language Change, –. Oxford: Oxford University Press. Gianollo, Chiara. . Latin aliquis as an epistemic indefinite. In Sofiana Chiriacescu (ed.), Proceedings of the VI Nereus International Workshop ‘Theoretical Implications at the Syntax/Semantics Interface in Romance’, –. Universität Konstanz: Arbeitspapier . Gianollo, Chiara. . Competing constructions for inalienable possession in the Vulgate Gospels: translator’s choices and grammatical constraints. Journal of Latin Linguistics (). –. Gianollo, Chiara. a. The Latin system of negation at the syntax-semantics interface. Rivista di Grammatica Generativa . –. Gianollo, Chiara. b. Indefinites between Latin and Romance. Habilitationsschrift, Universität zu Köln. Gianollo, Chiara. . Focus-sensitive negation in Latin. Catalan Journal of Linguistics . –. Gianollo, Chiara, Agnes Jäger, and Doris Penka. . Language change at the syntax-semantics interface. Perspectives and challenges. In Chiara Gianollo, Agnes Jäger, and Doris Penka (eds), Language Change at the Syntax-Semantics Interface, –. Berlin/New York: de Gruyter. Gildersleeve, Basil L. and Gonzalez Lodge. . Gildersleeve’s Latin Grammar. London: Bristol Classical Press (reprint ). Giorgi, Alessandra and Fabio Pianesi. . Tense and Aspect: From Semantics to Morphosyntax. Oxford: Oxford University Press. Giusti, Giuliana. . Parallels in clausal and nominal periphery. In Mara Frascarelli (ed.), Phases of Interpretation, –. Berlin: de Gruyter. Giusti, Giuliana. . Le espressioni di quantità. In Giampaolo Salvi and Lorenzo Renzi (eds), Grammatica dell’italiano antico, vol. , –. Bologna: Il Mulino. Giusti, Giuliana and Rossella Iovino. . Evidence for a Split DP in Latin. University of Venice Working Papers in Linguistics . –. Giusti, Giuliana, Rossella Iovino, and Renato Oniga. . On the syntax of the Latin quantifier omnis. In Gerd Haverling (ed.), Latin Linguistics in the Early st Century. Acts of the th International Colloquium on Latin Linguistics, vol. , –. Uppsala: Studia Latina Upsaliensia. Gugán, Katalin. . Zigzagging in language history: Negation and Negative Concord in Hungarian. Finno-Ugric Languages and Linguistics . –. Haase, Albert. . Syntaxe française du XVIIe siècle. Nouv. éd. trad. et remaniée par M. Obert: Paris: Delagrave. Haegeman, Liliane. . The Syntax of Negation. Cambridge: Cambridge University Press. Haegeman, Liliane. . Negative preposing, Negative Inversion and the split CP. In Laurence R. Horn and Yasuhiko Kato (eds), Negation and Polarity: Syntactic and Semantic Perspectives, –. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi



References

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 and Raffaella Zanuttini. . Negative concord in West Flemish. In Adriana Belletti and Luigi Rizzi (eds.), Parameters and Functional Heads: Essays in Comparative Syntax, –. Oxford: Oxford University Press. Han, Chung-Hye. . Force, negation and imperatives. The Linguistic Review . –. Hansen, Maj-Britt Mosegaard and Jacqueline Visconti. . The evolution of negation in French and Italian: Similarities and differences. Folia Linguistica (). –. Hansen, Maj-Britt Mosegaard and Jacqueline Visconti (eds). . The Diachrony of Negation. Amsterdam: Benjamins. Harris, Martin. . The Evolution of French Syntax. A Comparative Approach. London: Longman. Haspelmath, Martin. . Indefinite Pronouns. Oxford: Oxford University Press. Haspelmath, Martin. . Coordination. In Timothy Shopen (ed.), Language Typology and Linguistic Description, –. Cambridge: Cambridge University Press, nd edn. Haspelmath, Martin. . Negative indefinite pronouns and predicate negation. In Matthew S. Dryer and Martin Haspelmath (eds), The World Atlas of Language Structures Online. Available online at http://wals.info/chapter/, Accessed January , . Leipzig: Max Planck Institute for Evolutionary Anthropology. Heim, Irene. . A note on negative polarity and downward entailingness. In Charles Jones and Peter Sells (eds), Proceedings of NELS , –. Amherst, MA: University of Massachusetts. Heim, Irene. . Artikel und Definitheit. In Arnim von Stechow and Dieter Wunderlich (eds), Semantik: Ein Internationales Handbuch der Zeitgenössischen Forschung, –. Berlin: de Gruyter. Heim, Irene. . Definiteness and indefiniteness. In Klaus von Heusinger, Claudia Maienborn, and Paul Portner (eds), Semantics. An International Handbook of Natural Language Meaning, vol. . HSK, –. Berlin/New York: de Gruyter. Hendriks, Petra. . Either, both and neither in coordinate structures. In Alice ter Meulen and Werner Abraham (eds), The Composition of Meaning: From Lexeme to Discourse, –. Amsterdam: Benjamins. Herburger, Elena. . The Negative Concord puzzle revisited. Natural Language Semantics . –. Herburger, Elena. . A note on Spanish ni siquiera, ‘even’, and the analysis of NPIs. Probus . –. Herburger, Elena. . Negation. In Klaus von Heusinger, Claudia Maienborn, and Paul Portner (eds), Semantics. An International Handbook of Natural Language Meaning, vol. . HSK, –. Berlin/New York: de Gruyter. Hinterhölzl, Roland and Ans van Kemenade. . Syntactic and prosodic triggers in the rise and fall of V in English. Presentation at DiGS , University of Iceland. Hinterwimmer, Stefan. . Information structure and truth-conditional semantics. In Klaus von Heusinger, Claudia Maienborn, and Paul Portner (eds), Semantics: An International Handbook of Natural Language Meaning, vol. . HSK, –. Berlin/New York: de Gruyter.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Hinterwimmer, Stefan and Carla Umbach. . Grading and hedging by ‘gewiss’. In Luis Alonso-Ovalle and Paula Menéndez-Benito (eds), Epistemic Indefinites, –. Oxford: Oxford University Press. Hoeksema, Jack. . On the grammaticalization of negative polarity items. In Susanne Gahl, Andy Dolbey, and Christopher Johnson (eds), Proceedings of the th Annual Meeting of the Berkeley Linguistics Society, –. Berkeley, CA: Berkeley Linguistics Society. Hoeksema, Jack. . Jespersen recycled. In Elly van Gelderen (ed.), Cyclical Change, –. Amsterdam: Benjamins. Hoeksema, Jack. . Dutch enig: from nonveridicality to downward entailment. Natural Language and Linguistic Theory (). –. Hoeksema, Jack. . The Middle Dutch negative clitic: Status, position and disappearance. Lingua . –. Hofmann, Johann Baptist and Anton Szantyr. . Lateinische Grammatik. Lateinische Syntax und Stilistik, vol. .. Munich: Beck. Holmberg, Anders. . The syntax of yes and no in Finnish. Studia Linguistica . –. Holmberg, Anders. . The syntax of answers to polar questions in English and Swedish. Lingua . –. Horn, Laurence R. . A presuppositional analysis of ‘only’ and ‘even’. In Papers from the Fifth Regional Meeting of the Chicago Linguistics Society, –. Chicago, IL: University of Chicago Press. Horn, Laurence R. . Metalinguistic negation and pragmatic ambiguity. Language (). –. Horn, Laurence R. . A Natural History of Negation. Chicago, IL: University of Chicago Press (Reprint  Stanford, CA: CSLI). Horrocks, Geoffrey. . Clitics in Greek. A diachronic review. In Maria Roussou and Stavros Panteli (eds), Greek Outside Greece II, –. Athens: Diaspora Books. Iatridou, Sabine and Ivy Sichel. . Negative DPs, A-movement, and scope diminishment. Linguistic Inquiry (). –. Ihsane, Tabea. . The Layered DP. Form and Meaning of French Indefinites. Amsterdam: Benjamins. Ingham, Richard. . On two Negative Concord dialects in Early English. Language Variation and Change . –. Ingham, Richard. . NegP and negated constituent movement in the history of English. Transactions of the Philological Society (). –. Ingham, Richard. a. Grammar change in Anglo-Norman and Continental French. The replacement of non-affirmative indefinite nul by aucun. Diachronica (). –. Ingham, Richard. b. Le latin tardif et le cycle des indéfinis dans la diachronie du français. De Lingua Latina. Revue de Linguistique Latine du Centre Alfred Ernout . –. Ingham, Richard. . Old French negation, the Tobler/Mussafia law, and V. Lingua . –. Ingham, Richard and Amel Kallel. . Evidence from a correspondence corpus for diachronic change in French indefinites –. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds), The Diachrony of Negation, –. Amsterdam: Benjamins. Inghilterra, Rosa. . Strutture negative nelle lettere di Cicerone: BA thesis, Università di Padova.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Iovino, Rossella. . Word order in Latin nominal expressions. The syntax of demonstratives. In Renato Oniga, Rossella Iovino, and Giuliana Giusti (eds), Formal Linguistics and the Teaching of Latin. Theoretical and Applied Perspectives in Comparative Grammars, –. Newcastle upon Tyne: Cambridge Scholars Publishing. Isac, Daniela. . Focus on Negative Concord. In Reineke Bok-Bennema, Bart Hollebrandse, Brigitte Kampers-Manhe, and Petra Sleeman (eds), Romance Languages and Linguistic Theory : Selected Papers from ‘Going Romance’ , –. Amsterdam: Benjamins. Israel, Michael. . The Grammar of Polarity. Pragmatics, Sensitivity, and the Logic of Scales. Cambridge: Cambridge University Press. Jacobs, Joachim. . Lexical decomposition in Montague Grammar. Theoretical Linguistics . –. Jacobs, Joachim. . Syntax und Semantik der Negation im Deutschen. Munich: Fink. Jacobs, Joachim. . Fokus und Skalen. Zur Syntax und Semantik der Gradpartikeln im Deutschen. Tübingen: Niemeyer. Jacobs, Joachim. . Negation. In Arnim von Stechow and Dieter Wunderlich (eds), Semantik. Ein internationales Handbuch der zeitgenössischen Forschung, –. Berlin/New York: de Gruyter. Jacobs, Joachim. . The dimensions of topic-comment. Linguistics . –. Jäger, Agnes. . History of German Negation. Amsterdam: Benjamins. Jäger, Agnes. . Anything is nothing is something. On the diachrony of polarity types of indefinites. Natural Language and Linguistic Theory (). –. Jayez, Jacques and Lucia Tovena. . Epistemic determiners. Journal of Semantics (). –. Jayez, Jacques and Lucia Tovena. . The meaning and (a bit of) the history of quelque. In Lucia Tovena (ed.), French Determiners in and across Time., –. London: College Publications. Jespersen, Otto. . Negation in English and Other Languages. Copenhagen: Høst. Jónsson, Jóhannes Gísli. . Clausal Architecture and Case in Icelandic: University of Massachusetts, Amherst dissertation. Kadmon, Nirit and Fred Landman. . Any. Linguistics and Philosophy (). –. Kahn, Charles H. . The verb ‘be’ in Ancient Greek. Dordrecht: Reidel. Karttunen, Lauri and Stanley Peters. . Conventional implicature in Montague grammar. In Choon-Kyu Oh and David A. Dinneen (eds), Syntax and Semantics : Presuppositions, –. New York: Academic Press. Kay, Paul. . ‘Even’. Linguistics and Philosophy (). –. Kayne, Richard S. . French Syntax: The Transformational Cycle. Cambridge, MA: MIT Press. Kayne, Richard S. . Overt vs. covert movement. Syntax . –. Kayne, Richard S. . English ‘one’ and ‘ones’ as complex determiners. Ms, NYU. Kellert, Olga. . Changes in the negative polarity system. Presentation at Mini-workshop on Negative Quantification, Goethe-Universität Frankfurt, June , . Keniston, Hayward. . The Syntax of Castilian Prose: The Sixteenth Century. Chicago, IL: University of Chicago Press. Kiparsky, Paul. . ‘Elsewhere’ in phonology. In Stephen R. Anderson and Paul Kiparsky (eds), A Festschrift for Morris Halle, New York: Holt, Rinehart and Winston.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Kiparsky, Paul and Cleo Condoravdi. . Tracking Jespersen’s Cycle. In Brian Joseph and Angela Ralli (eds), Proceedings of the Second International Conference of Modern Greek Dialects and Linguistic Theory, –. Patras: University of Patras. Kiss, Katalin É. . The evolution of functional left peripheries in Hungarian syntax. In Katalin É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax, ch. , –. Oxford: Oxford University Press. König, Ekkehard. . The Meaning of Focus Particles: A Comparative Perspective. London: Routledge. König, Ekkehard. . Intensifiers and reflexive pronouns. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds), Language Typology and Language Universals - An International Handbook, –. Berlin: de Gruyter. Kratzer, Angelika. . Stage-level and individual-level predicates. In Greg Carlson and Francis J. Pelletier (eds), The Generic Book, –. Chicago, IL: University of Chicago Press. Kratzer, Angelika. . Covert quantifier restrictions in natural languages. http:// semanticsarchive.net/Archive/mIzMGUyZ/CovertQuantifierDomain Restrictions.pdf. Kratzer, Angelika. . Indefinites and the operators they depend on: from Japanese to Salish. In Greg Carlson and Francis Pelletier (eds), Reference and Quantification. The Partee Effect, –. Stanford, CA: CSLI. Kratzer, Angelika and Junko Shimoyama. . Indeterminate pronouns: The view from Japanese. In Yukio Otsu (ed.), Proceedings of the rd Tokyo Conference on Psycholinguistics, –. Tokyo: Hituzi Syobo. Krifka, Manfred. . The semantics and pragmatics of polarity items. Linguistic Analysis . –. Krifka, Manfred. . Additive particles under stress. In Proceedings of SALT , –. LSA. Krifka, Manfred. . Basic notions of information structure. Working Papers of the SFB, Interdisciplinary Studies on Information Structure (ISIS) . –. Kühner, Raphael and Karl Stegmann. . Ausführliche Grammatik der lateinischen Sprache. Hannover: Hahnsche Buchhandlung, nd edn. Labelle, Marie and M. Teresa Espinal. . Diachronic changes in negative expressions: the case of French. Lingua . –. Ladusaw, William. . Polarity Sensitivity as Inherent Scope Relations: University of Texas, Austin dissertation. Ladusaw, William. . Expressing negation. In Chris Barker and David Dowty (eds), Proceedings of SALT , –. Ohio State University: Ohio State Working Papers in Linguistics vol. . Ladusaw, William. . Negation, indefinites, and the Jespersen cycle. In Joshua S. Guenter, Barbara A. Kaiser, and Cheryl C. Zoll (eds), Proceedings of the th Annual Meeting of the Berkeley Linguistics Society, –. Berkeley, CA: Berkeley Linguistics Society. Lahiri, Utpal. . Focus and negative polarity in Hindi. Natural Language Semantics . –. Laka, Itziar. . Negation in Syntax: On the Nature of Functional Categories and Projections: MIT dissertation (published  Garland). Larrivée, Pierre. a. L’unidirectionalité irréversible du changement linguistique comme conséquence de l’acquisition? Le cas d’expressions négatives déliquescentes en français ancien. SHSWeb of Conferences . –.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Larrivée, Pierre. b. The continuity of the vernacular. The evolution of negative doubling in French. In Maj-Britt Mosegaard Hansen and Jacqueline Visconti (eds), The Diachrony of Negation, –. Amsterdam: Benjamins. Larrivée, Pierre and Richard Ingham (eds). . The Evolution of Negation. Beyond the Jespersen Cycle. Berlin: de Gruyter. Lauer, Sven. . Some news about irgend- and algún. Ms., Stanford University. Ledgeway, Adam. . Grammatica diacronica del napoletano. Tübingen: Niemeyer. Ledgeway, Adam. . From Latin to Romance: Morphosyntactic Typology and Change. Oxford: Oxford University Press. Lee, Young-Suk and Laurence R. Horn. . Any as indefinite + even. Ms., Yale University. Lehmann, Winfred P. . Proto-Indo-European Syntax. Austin, TX: University of Texas Press. Leonetti, Manuel and Victoria Escandell-Vidal. . Fronting and verum focus in Spanish. In Andreas Dufter and Daniel Jacob (eds), Focus and Background in Romance Languages, –. Amsterdam: Benjamins. Levinson, Stephen C. . Presumptive Meanings. Cambridge: Cambridge University Press. Lewis, Charlton T. and Charles Short. . A Latin Dictionary. Oxford: Oxford University Press. Lightfoot, David. . Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Link, Godehard. . The logical analysis of plural and mass terms: a lattice-theoretical approach. In Christoph Schwarze, Rainer Bäuerle, and Arnim von Stechow (eds), Meaning, Use and Interpretation of Language, –. Berlin: de Gruyter. Löfstedt, Einar. . Syntactica. Studien und Beiträge zur historischen Syntax des Lateins, vol. II. Lund: Gleerup. Löfstedt, Einar. . Vermischte Studien zur lateinischen Sprachkunde und Syntax. Lund: Gleerup. Löfstedt, Einar. . Syntactica. Studien und Beiträge zur historischen Syntax des Lateins, vol. I. Zweite erweiterte Auflage. Lund: Gleerup. Löfstedt, Einar. . Philologischer Kommentar zur Peregrinatio Aetheriae. Untersuchungen zur Geschichte der lateinischen Sprache. Darmstadt: Wissenschaftliche Buchgesellschaft (reprint of  original). Lohnstein, Horst. . Satzmodus - kompositionell. Zur Parametrisierung der Modusphrase im Deutschen. Berlin: Akademie Verlag. Longobardi, Giuseppe. . Convergence in parametric phylogenies: Homoplasy or principled explanation? In Charlotte Galves, Sonia Cyrino, Ruth Lopes, Filomena Sandalo, and Juanito Avelar (eds), Parameter Theory and Language Change, –. Oxford: Oxford University Press. Longobardi, Giuseppe. . Theory and experiment in parametric minimalism. The case of Romance negation. In Rob Pensalfini, Myfanwy Turpin, and Diana Guillemin (eds), Language Description Informed by Theory, –. Amsterdam: Benjamins. Lyons, Christopher. . Definiteness. Cambridge: Cambridge University Press. Magni, Elisabetta. . Mood and modality. In Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol. , –. Berlin/New York: de Gruyter. Magni, Elisabetta, Anna Orlandini, and Paolo Poccetti. . Haud: usages et fonctions d’une négation perdue. De Lingua Latina. Revue de linguistique . –.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Manoliu-Manea, Maria. . How can somebody become nobody? Perceptive strategy and Romance negation. Revue roumaine de linguistique . –. Manzini, Maria Rita and Leonardo Savoia. . Grammatical Categories. Variation in Romance Languages. Cambridge: Cambridge University Press. Manzotti, Emilio and Alessandra Rigamonti. . La negazione. In Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds), Grande grammatica italiana di consultazione, vol. , –. Bologna: Il Mulino, nuova edn. Maraldi, Mirka. . Some remarks on Latin free-choice indefinites. In Gualtiero Calboli (ed.), Papers on Grammar, –. Bologna: CLUEB. Maraldi, Mirka. . Free choice quantification and concession in Latin. In Donna Shalev and Lea Sawicki (eds), Donum grammaticum. Studies in Latin and Celtic Linguistics in Honour of Hannah Rosén, –. Louvain: Peeters. Marouzeau, Jules. . L’ordre des mots dans la phrase latine, vol.  Les groupes nominaux. Paris: Champion. Martí, Luisa. . The semantics of plural indefinite noun phrases in Spanish and Portuguese. Natural Language Semantics . –. Martí, Luisa. . Contextual restriction on indefinites. In Anastasia Giannakidou and Monika Rathert (eds), Quantification, Definiteness and Nominalization, –. Oxford: Oxford University Press. Martins, Ana Maria. . Polarity Items in Romance: Underspecification and Lexical Change. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds), Diachronic Syntax. Models and Mechanisms, –. Oxford: Oxford University Press. Martins, Ana Maria. . Nominal negative inversion with algum / alguno. Handout of presentation at Going Romance , – December, KU Leuven, http://alfclul.clul.ul.pt/ wochwel/documents/MartinsNominalnegativeinversion.pdf. Martins, Ana Maria. . How much syntax is there in metalinguistic negation? Natural Language and Linguistic Theory . –. Martins, Ana Maria. a. Negation and NPI composition inside DP. In Theresa Biberauer and George Walkden (eds), Syntax over Time, –. Oxford: Oxford University Press. Martins, Ana Maria. b. Ordem de palavras e polaridade. Inversão nominal negativa com algum / alguno e nenhum. Diacrítica (). –. Matthewson, Lisa. . Quantification and the nature of crosslinguistic variation. Natural Language Semantics (). –. McCloskey, James. . The scope of verb-movement in Irish. Natural Language and Linguistic Theory . –. McCloskey, James. . On the morphosyntax of wh-extraction in Irish. Journal of Linguistics . –. Meillet, Antoine. . L’évolution des formes grammaticales. Rivista di scienza . Repr.  Linguistique historique et linguistique générale, –. Paris: Champion. Meisner, Charlotte and Natascha Pomino. . Synchronic variation in the expression of French negation: A Distributed Morphology approach. Journal of French Language Studies (). –. Meisner, Charlotte, Elisabeth Stark, and Harald Völker (eds). . Jespersen Revisited: Negation in Romance and Beyond. Special issue of Lingua, .

OUP CORRECTED PROOF – FINAL, //, SPi



References

Mellet, Sylvie. . Opérations de détermination. Remarques sur deux indéfinis latins: quis et aliquis. Bulletin de la Société de Linguistique de Paris (). –. Mellet, Sylvie. . A propos de quelques indéfinis latins. Classes d’occurrences et construction référentielle. Faits de Langues . –. Merchant, Jason. . The Syntax of Silence. Sluicing, Islands, and the Theory of Ellipsis. Oxford: Oxford University Press. Merchant, Jason. . Why No(t)? Style (–). –. Milner, Jean-Claude. . Le système de la négation en Français et l’opacité du sujet. Langue française . –. Mitrovi´c, Moreno and Uli Sauerland. . Decomposing coordination. In Jyoti Iyer and Leland Kusmer (eds), Proceedings of NELS , vol. , –. Amherst, MA: GLSA Publications. Molinelli, Piera. . Fenomeni della negazione dal latino all’italiano. Firenze: La Nuova Italia. Molinelli, Piera. . Double negation from Latin to Italian. In Gualtiero Calboli (ed.), Subordination and Other Topics in Latin. Proceedings of the Third Colloquium on Latin Linguistics, –. Amsterdam: Benjamins. Molinelli, Piera. . Le strutture coordinate. In Giampaolo Salvi and Lorenzo Renzi (eds.), Grammatica dell’italiano antico, vol. , –. Bologna: Il Mulino. Moscati, Vincenzo. . The Scope of Negation: Università degli Studi di Siena dissertation. Muller, Claude. . La négation en français. Geneva: Droz. Muller, Claude. . Les déterminants indéfinis négatifs aucun et pas un: portée, référence, interactions. Verbum (). –. Muller, Claude. . Valeurs communes et valeurs particulières des formes qu- en français. Langue française (). –. Muller, Claude. . Négation directe vs. négation indirecte: quelle est la place des négatifs parmi les indéfinis en français? Linguarum Varietas –. Narrog, Heiko and Johan van der Auwera. . Grammaticalization and semantic maps. In Heiko Narrog and Bernd Heine (eds), The Oxford Handbook of Grammaticalization, –. Oxford: Oxford University Press. Onea, Edgar. . Potential Questions at the Semantics-Pragmatics Interface. Leiden: Brill. Onea, Edgar and Ljudmila Geist. . Indefinite determiners and referential anchoring. International Review of Pragmatics (). –. Ordóñez, Francisco. . Word Order and Clause Structure in Spanish and other Romance Languages: CUNY PhD dissertation. Orlandini, Anna. . Semantica e pragmatica dei pronomi indefiniti latini. Lingua e Stile (). –. Orlandini, Anna. . Une analyse sémantique et pragmatique des pronoms indéfinis en latin. In Harm Pinkster (ed.), Latin Linguistics and Linguistic Theory. Proceedings of the st International Colloquium on Latin Linguistics, –. Amsterdam: Benjamins. Orlandini, Anna. . Les pronoms indéfinis et la négation. In Gualtiero Calboli (ed.), Papers on Grammar, –. Bologna: CLUEB. Orlandini, Anna. a. Négation et argumentation en latin. Grammaire fondamentale du latin, Tome VIII. Louvain: Peeters. Orlandini, Anna. b. Nec, neque ou de la disjonction. In Claude Moussy, Michèle Fruyt, Jacqueline Dangel, Lyliane Sznajder, and Léon Nadjo (eds), De lingua latina novae quaestiones: actes du Xe Colloque International de linguistique Latine : Paris-Sèvres, – avril , –. Louvain: Peeters.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Orlandini, Anna and Paolo Poccetti. . Il y a nec et nec: Trois valeurs de la négation en latin et dans les langues de l’Italie ancienne. In Franck Floricic (ed.), La négation dans les langues romanes, –. Amsterdam: Benjamins. Orlandini, Anna and Paolo Poccetti. . Three types of negation. Different values of nec in Latin and in the languages of ancient Italy. In Roger Wright (ed.), Latin vulgaire—latin tardif VIII. Actes du VIIIe colloque international sur le latin vulgaire et tardif. Oxford – septembre , –. Hildesheim: Olms-Weidmann. Özge, Umut. . ‘Strong’ Indefiniteness and Topicality. In Emmanuel Chemla, Vincent Homer, and Grégoire Winterstein (eds), Proceedings of Sinn und Bedeutung , –. ENS Paris. Panagiotidis, Phoevos. . Pronouns, Clitics and Empty Nouns. Amsterdam: Benjamins. Panagiotidis, Phoevos. . Empty nouns. Natural Language and Linguistic Theory . –. Panizza, Daniele and Jacopo Romoli. . On the processing of alternatives, exhaustification and covert negation: the case of ‘mai’. In Anamaria Fˇalˇau¸s (ed.), Alternatives in Semantics, –. London: Palgrave Macmillan. Par, Anfos. . Sintaxi catalana segons los escrits en prosa de Bernat Metge. Halle: Niemeyer. Parry, Mair. . Negation in the history of Italo-Romance. In David Willis, Christopher Lucas, and Anne Breitbarth (eds), The History of Negation in the Languages of Europe and the Mediterranean, vol. : Case studies, –. Oxford: Oxford University Press. Partee, Barbara. . On the ‘scope of negation’ and polarity sensitivity. In Eva Hajiˇcová (ed.), Functional Approaches to Language Description, –. http://www.semanticsarchive.net/ Archive/TJhNTVmN/BHPScopeofNegationPolarity.pdf. Penka, Doris. . Negative Indefinites. Oxford: Oxford University Press. Penka, Doris and Arnim von Stechow. . Negative Indefinita unter Modalverben. Linguistische Berichte Sonderheft , –. Penka, Doris and Hedde Zeijlstra. . Negative indefinites in Dutch and German. LingBuzz http://ling.auf.net/lingbuzz/. Pereltsvaig, Asya. . Negative polarity items in Russian and the ‘bagel problem’. In A. Przepiorkowski and S. Brown (eds), Negation in Slavic, Bloomington, IN: Slavica Publishers. Pereltsvaig, Asya. . Russian nibud’-series as markers of co-variation. In Natasha Abner and Jason Bishop (eds), Proceedings of WCCFL , –. Somerville, MA: Cascadilla Press. Pierluigi, Sabrina. . Definitezza e specificità nelle lingue romanze antiche: Università per Stranieri di Siena dissertation. Pieroni, Silvia. . ‘Ipse’: relationships with grammatical functions and person. In Gérald Purnelle and Joseph Denooz (eds), Ordre et cohérence en latin, –. Geneva: Droz. Pinkster, Harm. . The Oxford Latin Syntax. Volume : The simple clause. Oxford: Oxford University Press. Pintzuk, Susan and Ann Taylor. . The loss of OV order in the history of English. In Ans van Kemenade and Bettelou Los (eds), The Handbook of the History of English, –. Oxford: Blackwell. Poletto, Cecilia. . Parallel phases: a study on the high and low left periphery of Old Italian. In Mara Frascarelli (ed.), Phases of Interpretation, –. Berlin: de Gruyter. Poletto, Cecilia. . On negative doubling. Quaderni di Lavoro ASIt . –. Poletto, Cecilia. . The syntax of Focus negation. University of Venice Working Papers in Lingustics . –.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Poletto, Cecilia. . Word Order in Old Italian. Oxford: Oxford University Press. Poletto, Cecilia. . OV and VO in Romance: the special status of Quantifiers. Presentation at the CCLS Lectures, Cologne, ... Poletto, Cecilia. . Negation. In Adam Ledgeway and Martin Maiden (eds), The Oxford Guide to the Romance Languages, –. Oxford: Oxford University Press. Poletto, Cecilia and Jacopo Garzonio. . The dynamics of the PF interface: negation and clitic clusters. Lingua . –. Poletto, Cecilia and Jean-Yves Pollock. . On the Left Periphery of some Romance Whquestions. In Luigi Rizzi (ed.), The Structure of CP and IP. The Cartography of Syntactic Structures, –. Oxford: Oxford University Press. Polo, Chiara. . Word Order between Morphology and Syntax. Padova: Unipress. Poole, Geoffrey. . Focus and the development of n-words in Spanish. In Janine Berns, Haike Jacobs, and Tobias Scheer (eds), Romance Languages and Linguistic Theory : Selected Papers from ‘Going Romance’ Nice , –. Amsterdam: Benjamins. Posner, Rebecca. . Double negatives, negative polarity and negative incorporation in Romance: a historical and comparative view. Transactions of the Philological Society (). –. Posner, Rebecca. . The Romance Languages. Cambridge: Cambridge University Press. Prévost, Sophie and Catherine Schnedecker. . Aucun(e)(s)/d’aucun(e)(s)/les aucun(e)(s): evolution du français médiéval au français moderne. Scolia –. Progovac, Ljiljana. . Positive and Negative Polarity: A Binding Approach. Cambridge: Cambridge University Press. Quer, Josep. . Edging quantifiers: On QP-fronting in western Romance. In Claire Beyssade, Reineke Bok-Bennema, Frank Drijkoningen, and Paola Monachesi (eds), Romance Languages and Linguistic Theory : Selected Papers from Going Romance, –. Amsterdam: Benjamins. Ramat, Paolo. . Perché veruno significa nessuno? In Paolo Ramat and Elisa Roma (eds), Sintassi storica. Atti del XXX Congresso Internazionale della Società di Linguistica Italiana, –. Repp, Sophie. . Negation in Gapping. Oxford: Oxford University Press. REW, Wilhelm Meyer-Lübke. . Romanisches etymologisches Wörterbuch . vollständig neubearbeitete Auflage. Available online urn:nbn:de:bvb:-ubr-. Heidelberg: Winter. Riegel, Martin, Jean-Christophe Pellat, and René Rioul. . Grammaire méthodique du français. Paris: Presses Universitaires de France. Rijksbaron, Albert. . Does Ancient Greek have a word for ‘No’? The evidence from οκον . . . ο questions. Journal of Greek Linguistics . –. Rizzi, Luigi. . Issues in Italian Syntax. Dordrecht: Foris. Rizzi, Luigi. . The fine structure of the Left Periphery. In Liliane Haegeman (ed.), Elements of Grammar, –. Dordrecht: Kluwer. Rizzi, Luigi. . On some properties of Criterial Freezing. CISCL Working Papers on Language and Cognition . –. Roberts, Ian. . Diachronic Syntax. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Roberts, Ian and Anna Roussou. . Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rohlfs, Gerhard. . Grammatica storica della lingua italiana e dei suoi dialetti. Turin: Einaudi. Romero, Maribel. . Indefinites: scope, modality, discourse. Ms., University of Konstanz. Romero, Maribel and Chung-Hye Han. . On negative yes/no questions. Linguistics and Philosophy . –. Rooth, Mats. . Association with Focus. University of Massachusetts, Amherst dissertation. Rosén, Hannah. . Latine loqui: Trends and Directions in the Crystallization of Classical Latin. Munich: Fink. Rowlett, Paul. . Sentential Negation in French. Oxford: Oxford University Press. Schaden, Gerhard. . From cardinal to indefinite: discourse properties of Latin unus. Handout of presentation at the Workshop on Formal Diachronic Semantics, University of Konstanz. Schnedecker, Catherine. . L’expression ‘indéfinie’ de la personne par le SN dét+quidam: un cas de transcatégorialisation original? Talk Conférence Diachro-VII http://.. ./diachro/Diachro-resumes.pdf. Schwarzschild, Roger. . Singleton Indefinites. Journal of Semantics . –. Schwegler, Armin. . Analyticity and Syntheticity: A Diachronic Perspective with Special Reference to Romance Languages. Berlin/New York: de Gruyter. Schwenter, Scott A. and Richard Waltereit. . Presupposition accommodation and language change. In Kristin Davidse, Lieven Vandelanotte, and Hubert Cuyckens (eds), Subjectification, Intersubjectification and Grammaticalization, –. Berlin: de Gruyter. Scorretti, Mauro. . Le strutture coordinate. In Lorenzo Renzi, Giampaolo Salvi, and Anna Cardinaletti (eds), Grande grammatica italiana di consultazione, vol. I, –. Bologna: Il Mulino, nuova edn. Selig, Maria. . Die Entwicklung der Nominaldeterminanten im Spätlatein: romanischer Sprachwandel und lateinische Schriftlichkeit. Tübingen: Narr. Selkirk, Elisabeth. . The prosodic structure of function words. In James L. Morgan and Katherine Demuth (eds), Signal to Syntax: Bootstrapping from Speech to Grammar in Early Acquisition, –. Mahwah, NJ: Lawrence Erlbaum. Sihler, Andrew L. . New Comparative Grammar of Greek and Latin. Oxford: Oxford University Press. Simpson, Andrew and Zoe Wu. . Agreement, shells and Focus. Language (). –. Spevak, Olga. . Constituent Order in Classical Latin Prose. Amsterdam: Benjamins. Stanley, Jason and Zoltán Gendler Szabó. . On quantifier domain restriction. Mind and Language (–). –. Stark, Elisabeth. . Indefiniteness and specificity in Old Italian texts. Journal of Semantics (). –. Stark, Elisabeth. . Indefinitheit und Textkohärenz: Entstehung und Semantische Strukturierung Indefiniter Nominaldetermination im Altitalienischen. Tübingen: Niemeyer. Stark, Elisabeth, Elisabeth Leiss, and Werner Abraham (eds.). . Nominal Determination. Typology, Context Constraints and Historical Emergence. Amsterdam: Benjamins.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Stotz, Peter. . Handbuch zur lateinischen Sprache des Mittelalters, vol. : Formenlehre, Syntax und Stilistik. Handbuch der Altertumswissenschafts. Munich: Beck. Svenonius, Peter. . Quantifier Movement in Icelandic. In Peter Svenonius (ed.), The Derivation of VO and OV, –. Amsterdam: Benjamins. Szabolcsi, Anna. . Quantification. Cambridge: Cambridge University Press. Szabolcsi, Anna. . Quantifier particles and compositionality. In Maria Aloni, Michael Franke, and Floris Roelofsen (eds), Proceedings of the th Amsterdam Colloquium, –. University of Amsterdam. TLL. . Thesaurus Linguae Latinae online. Munich: Bayerische Akademie (Berlin: de Gruyter). Torrego, Maria Esperanza. . Coordination. In Philip Baldi and Pierluigi Cuzzolin (eds), New Perspectives on Historical Latin Syntax, vol. , –. Berlin: de Gruyter. Tovena, Lucia. . The Fine Structure of Polarity Sensitivity. New York: Garland. Tovena, Lucia. . Dealing with alternatives. In Christian Ebert and Cornelia Endriss (eds), Proceedings of Sinn und Bedeutung , vol. . ZAS Papers in Linguistics, –. Berlin: ZAS. Traina, Alfonso and Tullio Bertotti. . Sintassi normativa della lingua latina. Third edition. Bologna: Cappelli. Traugott, Elizabeth Closs and Richard B. Dasher. . Regularities in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Ekkehard König. . The semantics-pragmatics of grammaticalization revisited. In Elizabeth Closs Traugott and Bernd Heine (eds), Approaches to Grammaticalization, vol. , –. Amsterdam: Benjamins. Umbach, Carla. . Another additive particle under stress: German additive noch. In Proceedings of the Tenth Symposium on Logic and Language - LoLa, –. Hungarian Academy of Sciences. Umbach, Carla. . Strategies of additivity: German additive noch compared to auch. Lingua (). –. Väänänen, Veikko. . Introduction au latin vulgair. Troisième édition revue et augmentée. Paris: Klincksieck. Vallduví, Enric. . Polarity items, n-words, and minimizers in Catalan and Spanish. Probus . –. van der Auwera, Johan. . In defense of classical semantic maps. Theoretical Linguistics . –. van der Auwera, Johan. . The Jespersen Cycles. In Elly van Gelderen (ed.), Cyclical Change, –. Amsterdam: Benjamins. van der Sandt, Rob. . Denial. Papers from the Chicago Linguistic Society. Parasession on Negation (). –. van der Wouden, Ton. . Negative Contexts: Collocation, Polarity and Multiple Negation. London: Routledge. van Gelderen, Elly. . Grammaticalization as Economy. Amsterdam: Benjamins. van Gelderen, Elly. . Cycles of negation. Linguistic Typology (). –. van Gelderen, Elly. . The Linguistic Cycle: Language Change and the Language Faculty. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, //, SPi

References



van Gelderen, Elly. . Semantic and formal features: Feature economy in language change. In Chiara Gianollo, Agnes Jäger, and Doris Penka (eds), Language Change at the SyntaxSemantics Interface, –. Berlin: de Gruyter. van Kemenade, Ans. . Jespersen’s cycle revisited. Formal properties of grammaticalization. In Susan Pintzuk, George Tsoulas, and Anthony Warner (eds), Diachronic Syntax. Models and Mechanisms, –. Oxford: Oxford University Press. Vennemann, Theo. . Topics, subjects, and word order: From SXV to SVX via TVX. In John Anderson and Charles Jones (eds), Historical Linguistics: Proceedings of the First International Congress of Historical Linguistics, vol. , –. Amsterdam : North-Holland. von Fintel, Kai. . Restrictions on quantifier domains: University of Massachusetts, Amherst dissertation. von Fintel, Kai. . The formal semantics of grammaticalization. In Jill N. Beckmann (ed.), Proceedings of NELS , vol. , –. Amherst, MA: GLSA, University of Massachusetts. von Fintel, Kai. . Evidence for presuppositional indefinites. Ms., MIT. http://web.mit.edu/ fintel/www/presupp_indef.pdf. von Fintel, Kai. . NPI-licensing, Strawson-entailment, and context-dependency. Journal of Semantics . –. von Heusinger, Klaus. . Referentially anchored indefinites. In Ileana Comorovski and Klaus von Heusinger (eds), Existence: Semantics and Syntax, –. Berlin: Springer. von Heusinger, Klaus. . Specificity. In Klaus von Heusinger, Claudia Maienborn, and Paul Portner (eds), Semantics. An International Handbook of Natural Language Meaning, vol. . HSK, –. Berlin: de Gruyter. von Heusinger, Klaus, Claudia Maienborn, and Paul Portner (eds). . Semantics. An International Handbook of Natural Language Meaning, vol. . HSK. Berlin/New York: de Gruyter. Wallage, Phillip. . Jespersen’s Cycle in Middle English: Parametric variation and grammatical competition. Lingua (). –. Wanner, Dieter. . The Development of Romance Clitic Pronouns. From Latin to Old Romance. Berlin/New York: de Gruyter. Wanner, Dieter. . From Latin to the Romance languages. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher, and Wolfgang Raible (eds), Language Typology and Language Universals. An International Handbook, vol. ., HSK, –. Berlin: de Gruyter. Watanabe, Akira. . The genesis of Negative Concord. Linguistic Inquiry (). –. Weir, Andrew. . Some, speaker knowledge, and subkinds. In Rasmus K. Rendsvig and Sophia Katrenko (eds), ESSLLI  Student Session Proceedings, –. Weiß, Helmut. a. Indefinite pronouns. Morphology and syntax in cross-linguistic perspective. In Horst J. Simon and Heike Wiese (eds), Pronouns. Grammar and Representation, –. Amsterdam: Benjamins. Weiß, Helmut. b. A quantifier approach to negation in natural languages. Or why Negative Concord is necessary. Nordic Journal of Linguistics . –. Weiß, Helmut. c. Three types of negation. A case study in Bavarian. In Sjef Barbiers, Leonie Cornips, and Susanne van der Kleij (eds), Syntactic Microvariation, –. Amsterdam: Meertens Institute Electronic Publications in Linguistics http://www.meertens.knaw. nl/books/synmic/pdf/weiss.pdf.

OUP CORRECTED PROOF – FINAL, //, SPi



References

Wheeler, May, Alan Yates, and Nicolau Dols. . Catalan: A Comprehensive Grammar. London: Routledge. Willis, David. . Negative polarity and the Quantifier Cycle: comparative diachronic perspectives from European languages. In Pierre Larrivée and Richard Ingham (eds), The Evolution of Negation. Beyond the Jespersen Cycle, –. Berlin: de Gruyter. Willis, David. . The syntax of ‘nothing’ in Old English. Presentation at the Göttingen Spirit Summer School on Negation. Willis, David, Christopher Lucas, and Anne Breitbarth. a. Comparing diachronies of negation. In David Willis, Christopher Lucas, and Anne Breitbarth (eds), The History of Negation in the Languages of Europe and the Mediterranean. Vol. : case studies, –. Oxford: Oxford University Press. Willis, David, Christopher Lucas, and Anne Breitbarth (eds). b. The History of Negation in the Languages of Europe and the Mediterranean. Vol. : case studies. Oxford: Oxford University Press. Wolfe, Sam. . Verb-Initial Orders in Old Romance: A comparative account. Revue roumaine de linguistique LX(–). –. Wright, Roger. . A Sociophilological Study of Late Latin. Turnhout: Brepols Publishers. Wright, Roger. . Romance languages as a source for spoken Latin. In James Clackson (ed.), A Companion to the Latin Language, –. London: Blackwell. Wurmbrand, Susi. . Nor: neither disjunction nor paradox. Linguistic Inquiry (). –. Wurmbrand, Susi. . The Merge Condition: A syntactic approach to selection. In Peter Kosta, Lilia Schürcks, Steven Franks, and Teodora Radev-Bork (eds), Minimalism and Beyond: Radicalizing the Interfaces, –. Amsterdam: Benjamins. Zamparelli, Roberto. . On Certain/Specific phenomena. Handout of presentation at the ESSLLI  Workshop on Specificity, Vienna. Zamparelli, Roberto. . On singular existential quantifiers in Italian. In Ileana Comorovski and Klaus von Heusinger (eds), Existence: Semantics and Syntax, –. Berlin: Springer. Zanuttini, Raffaella. . Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: Oxford University Press. Zanuttini, Raffaella. . La negazione. In Giampaolo Salvi and Lorenzo Renzi (eds), Grammatica dell’italiano antico, vol. , –. Bologna: Il Mulino. Zeevat, Henk. . Presupposition and accommodation in update semantics. Journal of Semantics . –. Zeevat, Henk and Katja Jasinskaja. . And as an additive particle. In Mixel Aurnague, Kepa Korta, and Jésus M. Larrazabal (eds), Language, Representation and Reasoning. Memorial Volume to Isabel Gómez Txurruka, –. Bilbao: University of the Basque Country Press. Zeijlstra, Hedde. . Sentential Negation and Negative Concord. Universiteit van Amsterdam dissertation. Zeijlstra, Hedde. . Negative Concord is syntactic agreement. Ms., University of Amsterdam, available at http://ling.auf.net/lingbuzz/. Zeijlstra, Hedde. . On the syntactically complex status of negative indefinites. Journal of Comparative Germanic Linguistics . –. Zeijlstra, Hedde. . Not in the first place. Natural Language and Linguistic Theory . –.

OUP CORRECTED PROOF – FINAL, //, SPi

References



Zeijlstra, Hedde. . On the uninterpretability of interpretable features. In Peter Kosta, Steven L. Franks, Teodora Radeva-Bork, and Lilia Schürcks (eds), Minimalism and Beyond: Radicalizing the Interfaces, –. Amsterdam: Benjamins. Zimmermann, Malte. . Contrastive focus and emphasis. Acta Linguistica Hungarica . –. Zimmermann, Malte. . On the functional architecture of DP and the feature content of pronominal quantifiers in Low German. Journal of Comparative Germanic Linguistics . –. Zubizarreta, Maria Luisa. . Prosody, Focus, and Word Order. Cambridge, MA: MIT Press. Zwarts, Frans. . A hierarchy of negative expressions. In Heinrich Wansing (ed.), Negation. A Notion in Focus, –. Berlin/New York: de Gruyter.

OUP CORRECTED PROOF – FINAL, //, SPi

OUP CORRECTED PROOF – FINAL, //, SPi

Index a certain , , ,  acquisition , , , , , , ,  activity condition  Adams, James – Afrikaans  Agree , , –, , , , , , , , ,  Multiple Agree , ,  Reverse Agree ,  alcuno , –, – algo ,  alguém  alguien  algum , –, ,  algun , – algún , , , , , – aliquantus  aliquis , , , , , –, , –, –, ,  ali-series  alius – Aloni, Maria – Alonso-Ovalle, Luis –,  alternatives , , , , , , , , , , , , ,  exhaustification , , –,  ordering –,  anaphoricity , , , , ,  anti-exhaustivity ,  anti-singleton constraint , , ,  any ,  aucun , –, –,  Bavarian  Belletti, Adriana – Bertocchi, Alessandra ,  bhii , , ,  Biberauer, Theresa ,  bleaching , , , , , , , , , , ,  blocking , , , , , , , ,  Bortolussi, Bernard , , ,  bridging context , , , , , , , 

certus ,  Chierchia, Gennaro –, , , –, – clitics ‘big DP’ hypothesis  clitic doubling ,  Concord indefinite ,  NPI use –, , –, – optional Negative Concord –, , – conditional clauses , –, ,  Condoravdi, Cleo  context-sensitivity , , , , ,  conventionalization of pragmatic inferences , , , ,  of syntactic inversion ,  criterial freezing  cyclical change , , –, ,  Jespersen’s Cycle , , , –, , , , –, ,  Negative Cycle ,  Quantifier Cycle , –, , –,  Danckaert, Lieven –, ,  definite nominal phrases  degree  De Morgan’s Laws  denial , , –, , , ,  dependence , , , , , , , , , , ,  Déprez, Viviane  Devine, Andrew ,  directionality, see also semantic change, unidirectionality ,  discourse referent  discourse-structuring particle  disjunction – Double Negation , , , , , , , , , , , ,  downward-entailing contexts , , –, , , , , ,  DP  dP  DP-internal Focus , –, – DP-internal Topic 

OUP CORRECTED PROOF – FINAL, //, SPi



Index

DP-internal inversion , , , –, , –, – Dutch , 

Fruyt, Michèle , – function –, , , ,  future 

Eckardt, Regine , ,  ellipsis , , –, , , , , ,  emphasis (emphatic strengthening) , , , , , , , , , –, , , , –,  English , ,  enig , ,  epistemic indefinites –, , , , –, , , – crosslinguistic variation – plural – Etxeberria, Urtzi  even , , , 

Garzonio, Jacopo ,  generalizers , , , , ,  gens  German , , , , , , , ,  Giannakidou, Anastasia ,  Greek , , , , , , , 

feature ,  formal uninterpretable ,  [iFoc]-[uFoc] , , , ,  [iNeg]-[uNeg] , , , , ,  [Neg] , ,  reanalysis  semantic  valued-unvalued  feature economy  FinP  focus , , –, , , , –, , , , , , , , , , , , ,  ‘even’ E operator , ,  ‘only’ O operator  contrastive , , ,  Emph.Assert operator , ,  information ,  scalar , –, , ,  verum (polarity) , ,  Focus (syntactic projection) , –, , –, , , , , ,  Focus Concord –, – Focus Criterion  focus particle , , , , ,  additive – as formative of NPIs – scalar , , –,  Force  free-choice indefinites , , , , 

Haspelmath, Martin , –, , , , –, ,  haud ,  Head Preference Principle , ,  Herburger, Elena  Hindi , , – Horn, Larry  Hungarian ,  identifiability – ignorance implicature , –, ,  illocutionary operators  implicitly modalized contexts ,  indefinite article , , , , , ,  indefinites form , , –, , , , ,  meaning , , –, – quantificational analysis  Infl , , , ,  Infl-final language  pre- / post-Infl asymmetry , , –, , , ,  V-to-Infl  inflationary effects , , ,  informational strength , , , , –, ,  inheritance (inherited property) , , , , , –, , , , , , , ,  intensifier  interrogative pronouns , ,  ipse – irgendein , , –,  irrealis  Isac, Daniela  Kiparsky, Paul  König, Ekkehard –,  Kratzer, Angelika ,  Krifka, Manfred –, 

OUP CORRECTED PROOF – FINAL, //, SPi

Index Lahiri, Utpal ,  Latin corpora ,  periodization  Ledgeway, Adam ,  left periphery ,  Lexical Integrity  LF-movement  likelihood measure, see probability measure litotes  locality , , ,  mai , ,  man  Martí, Luisa  Martins, Ana Maria , –,  Matthewson-Kratzer conjecture  Menéndez-Benito, Paula –,  Merchant, Jason – minimizers , , , ,  modal indefinites, see epistemic indefinites modals –, , – multifunctionality , , , ,  nada ,  nadie  ne , ,  né ,  n¯e –, , ,  n˘e , – neanche ,  néant ,  n˘ec , , , , , , , , , , , , , , – additive meaning ,  as a [Neg] element , , – correlative particle –, , – discourse-structuring particle –,  focus particle – form – internal structure – n˘ec ullus , ,  n˘ec unus , , , , , – reanalysis ,  redundancy – scalar meaning , ,  sentence-initial  syntax –, – nec-words , , , –, –, , , , , ,  [uNeg] feature , 



etymology – internal structure – loss of emphasis – reanalysis – ne ne  n˘eque, see n˘ec ne . . . quidem , , , , , –, – negation , , , , , –, –, –, , , ,  constituent , – correlative , , , ,  covert operator , ,  direct  discontinuous ,  emphatic , –,  exceptive  illocutionary  indirect  metalinguistic, see denial position – replacive , ,  resumptive , ,  sentential – strengtheners , , , , , –, , , ,  negation system , , –, , , , , –, ,  Negative Concord , , , –, , , , –, –, , , , , , , , , – concealed – double-negation readings –, , ,  non-strict , , , , , , , –, , , ,  prerequisites in Late Latin – strict , , , , –, , ,  Negative Doubling  negative indefinites , , , , , , , –, , , –,  Latin n- series  negative marker – ‘big NegP’ hypothesis ,  cliticization , ,  featural specification , –,  internal structure  phrase-structural status , , –, , ,  position –, , ,  scope marker , 

OUP CORRECTED PROOF – FINAL, //, SPi



Index

negative-polarity items (NPIs) –, , , , –, , , , , –, , – emphatic , –,  non-emphatic  Negative Spread , , ,  Neg Criterion ,  NegP , , –, –, , , , , , –,  NegP-  nein  nemo –, ,  nenhum , ,  nessuno , , , , , ,  nesun ,  neul  neuter  ni ,  nic˘ıun  nient ,  niente ,  nihil –, ,  nimo  ningún – niuno ,  no  noenum – non , , ,  n¯on  etymology ,  focused – function  phrasal status , –, ,  position , , –,  reanalysis in Late Latin – Romance continuations  structural weakening  non nihil  non nullus  Norwegian  nul –, , , – nullo , ,  nullus , –, , , ,  NumP ,  numquam , ,  n-words , , , , , , , –, –, , , , , , , – ‘big DP’ hypothesis – interaction with focus , – internal structure 

Oaths of Strasbourg  Old English  Old French –, , –,  corpora  Old Italian –, , –, – corpora  Old Low German  omnis  omque ,  one  optionality , , ,  Orlandini, Anna , , , ,  oudeís  OV-to-VO shift , – OV orders with negative objects , – paradigmatic effects, see blocking partitivity , , ,  pas  Penka, Doris ,  personne , ,  pertinacity  PF-adjacency  phase , ,  CP–TP phase –, , ,  Phase Impenetrability Condition  vP phase –,  plural , , , , –, – Poccetti, Paolo ,  polarity switch (correlative negation) , , , , ,  Poletto, Cecilia –, , ,  Port, Angelika – Pragmatic Overload  prescriptivism  presupposition additive ,  of existence – of unique identifiability – scalar  presupposition accommodation , , ,  probability measure , , , ,  qualche ,  qualque  quantificational domain , , ,  constraints , , –, ,  covert restrictors – minimal widening , , , –, ,  narrowing  widening , , , , , , 

OUP CORRECTED PROOF – FINAL, //, SPi

Index Quantifier Raising ,  -que , , , , ,  quelque , , , – quelqu’un ,  quidam , , –, – quidem , ,  quis –, –,  quisquam , , , , , ,  QP  QP-fronting ,  redundancy (morphosyntactic) , –, , , , , , , –, –,  res  rien ,  Roberts, Ian ,  Romance , , , , –, , –, –, , , , , , –, –, , , , , , , , , ,  Catalan , , –, – French , , –, –, , , ,  Italian , , –, –, , , ,  Portuguese , , –, –,  Romanian , , , , ,  Spanish , –, ,  Venetian dialects  scalar implicature , ,  scalarity , –, , , , , , , –, , –, – selection – sem ,  semantic change –, ,  semantic reanalysis  unidirectionality , , ,  semantic map , –, , , , , , ,  of Latin indefinites –,  series , –, , , ,  shared topicality 



Shimoyama, Junko  short answers , , , , , , –, , , ,  some ,  specific indefinites , , , , , ,  specific unknown , , , , , , –, , , , ,  specificity , , –,  epistemic , ,  scopal – split scope  Stephens, Laurence ,  syncategorematic expressions  syncretism (of functional projections) , , ,  syntactic change ,  system of indefinites , , ,  paradigmatic relations , , , , ,  timing of syntactic operations , ,  earliness principles ,  Tovena, Lucia ,  ullus , , , , , ,  umquam  uno  unus –, –, , , , , –, – van Gelderen, Elly – variation , , , –, , , , , ,  veruno ,  vP ,  see also phase movement in Latin , ,  vreun  weakening phonological , ,  semantic ,  why no(t)? test – Willis, David  Zanuttini, Raffaella –,  Zeijlstra, Hedde –, –, 

OUP CORRECTED PROOF – FINAL, //, SPi

OXFORD STUDIES IN DIACHRONIC AND HISTORICAL LINGUISTICS general editors Adam Ledgeway and Ian Roberts, University of Cambridge advisory editors Cynthia Allen, Australian National University; Ricardo Bermúdez-Otero, University of Manchester; Theresa Biberauer, University of Cambridge; Charlotte Galves, University of Campinas; Geoff Horrocks, University of Cambridge; Paul Kiparsky, Stanford University; Anthony Kroch, University of Pennsylvania; David Lightfoot, Georgetown University; Giuseppe Longobardi, University of York; George Walkden, University of Konstanz; David Willis, University of Cambridge recently published in the series  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

OUP CORRECTED PROOF – FINAL, //, SPi

 Vowel Length from Latin to Romance Michele Loporcaro  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 Pan˘a 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

OUP CORRECTED PROOF – FINAL, //, SPi

 Micro-change and Macro-change in Diachronic Syntax Edited by Eric Mathieu and Robert Truswell  The Development of Latin Clause Structure A Study of the Extended Verb Phrase Lieven Danckaert  Transitive Nouns and Adjectives Evidence from Early Indo-Aryan John J. Lowe  Quantitative Historical Linguistics A Corpus Framework Gard B. Jenset and Barbara McGillivray  Gender from Latin to Romance History, Geography, Typology Michele Loporcaro  Clause Structure and Word Order in the History of German Edited by Agnes Jäger, Gisella Ferraresi, and Helmut Weiß  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 In preparation Cycles in Language Change Edited by Anne Breitbarth, Elisabeth Witzenhausen, Miriam Bouzouita, and Lieven Danckaert Morphological Borrowing Francesco Gardani

OUP CORRECTED PROOF – FINAL, //, SPi

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 τ oς and its Implications for Language Change and Structure Brian D. Joseph Reconstructing Pre-Islamic Arabic Dialects Alexander Magidow Word Order and Parameter Change in Romanian Alexandru Nicolae Referential Null Subjects in Early English Kristian A. Rusten 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 Verb Second in Medieval Romance Sam Wolfe Palatal Sound Change in the Romance Languages Diachronic and Synchronic Perspectives André Zampaulo