230 92 12MB
English Pages [257] Year 2021
i
A Unified Theory of Polarity Sensitivity
ii
OXFORD STUDIES IN COMPARATIVE SYNTAX Richard S. Kayne, General Editor Smuggling in Syntax Edited by Adriana Belletti and Chris Collins Nominal Arguments and Language Variation Li Julie Jiang Anti-Contiguity: A Theory of Wh-Prosody Jason Kandybowicz Variation in P: Comparative Approaches to Adpositional Phrases Edited by Jacopo Garzonio and Silva Rossi Locality and Logophoricity: A Theory of Exempt Anaphora Isabelle Charnavel The Linker in the Khoisan Languages Chris Collins The Syntactic Variation of Spanish Dialects Edited by Angel J. Gallego Questions of Syntax Richard S. Kayne Exploring Nanosyntax Edited by Lena Baunaz, Liliane Haegeman, Karen De Clercq, and Eric Lander
Micro-Syntactic Variation in North American English Edited by Raffaella Zanuttini and Laurence R. Horn Japanese Syntax in Comparative Perspective Edited by Mamoru Saito Cross-Linguistic Studies of Imposters and Pronominal Agreement Edited by Chris Collins Locality Edited by Ian Roberts and Enoch Aboh Aspects of Split Ergativity Jessica Coon Variation in Datives Edited by Beatriz Fernández and Ricardo Etxepare Adverbial Clauses, Main Clause Phenomena and Composition of the Left Periphery: The Cartography of Syntactic Structures, Volume 8 Liliane Haegeman
Dravidian Syntax and Universal Grammar K.A. Jayaseelan and R. Amritavalli
Functional Heads: The Cartography of Syntactic Structures, Volume 7 Edited by Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, Cecilia Poletto
The Morphosyntax of Portuguese and Spanish in Latin America Edited by Mary A. Kato and Francisco Ordóñez
Discourse-Related Features and Functional Projections Silvio Cruschina
Deconstructing Ergativity: Two Types of Ergative Languages and Their Features Maria Polinsky
Mapping the Left Periphery: The Cartography of Syntactic Structures, Volume 5 Edited by Paola Benincà and Nicola Munaro
Rethinking Parameters Edited by Luis Eguren, Olga Fernández- Soriano, and Amaya Mendikoetxea
The Grammar of Q: Q-Particles, Wh-Movement, and Pied-Piping Seth Cable
Argument Licensing and Agreement Claire Halpert
Comparisons and Contrasts Richard S. Kayne
The Cartography of Chinese Syntax: The Cartography of Syntactic Structures, Volume 11 Edited by Wei-Tien Dylan Tsai
Mapping Spatial PPs: The Cartography of Syntactic Structures, Volume 6 Edited by Guglielmo Cinque and Luigi Rizzi
Beyond Functional Sequence: The Cartography of Syntactic Structures, Volume 10 Edited by Ur Shlonsky
The Syntax of Ellipsis: Evidence from Dutch Dialects Jeroen van Craenenbroeck
The Architecture of Determiners Thomas Leu
Scrambling, Remnant Movement and Restructuring in West Germanic Roland Hinterhölzl
A Comparative Grammar of Borgomanerese Christina Tortora Chinese Syntax in a Cross-Linguistic Perspective Edited by Y.-H. Audrey Li, Andrew Simpson, and W.-T. Dylan Tsai Functional Structure from Top to Toe: The Cartography of Syntactic Structures, Volume 9 Edited by Peter Svenonius
Functional Heads: The Cartography of Syntactic Structures, Volume 7 Edited by Laura Brugé, Anna Cardinaletti, Giuliana Giusti, Nicola Munaro, Cecilia Poletto
iii
A Unified Theory of Polarity Sensitivity Comparative Syntax of Arabic Ahmad Alqassas
1
iv
1 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 certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2021 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 license, or under terms agreed with the appropriate reproduction rights organization. Inquiries 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. CIP data is on file at the Library of Congress ISBN 978–0–19–755489–0 (pbk) ISBN 978–0–19–755488–3 (hbk) DOI: 10.1093/oso/9780197554883.001.0001 9 8 7 6 5 4 3 2 1 Paperback printed by Marquis, Canada Hardback printed by Bridgeport National Bindery, Inc., United States of America
v
CONTENTS
Acknowledgments vii Abbreviations ix 1. Issues in the Syntax of Polarity-Sensitive Items 1 1.1 Introduction: Polarity-sensitive items 1 1.2 Critical issues in PSI syntax 6 1.2.1 NPI dependencies 6 1.2.2 NCI dependencies 9 1.2.3 Enclitic negation and locus of negation 11 1.2.4 PSIs with head-like properties 17 1.3 Sentential negation and PSI syntax 20 1.3.1 Definition of sentential negation 20 1.3.2 The locus of sentential negation 24 1.4 Outline of the book 34 2. Classification of PSIs and Their Lexical Categories 37 2.1 NPIs, PPIs, and indefinites 40 2.1.1 NPIs versus PPIs 40 2.1.2 NPIs and their lexical categories 45 2.2 NCIs and their lexical categories 57 2.3 Disjunction particles in coordinate complexes 65 3. Licensing Negative Polarity Items 71 3.1 NPI licensing in the syntactic structure 71 3.2 NPI analyses in Arabic 77 3.3 Adverbial and determiner NPIs in Arabic 86 3.4 Standard Arabic negation and NPI licensing 97 3.5 Licensing Arabic PSIs in nonveridical contexts 100 4. Licensing Negative Concord Items 107 4.1 NCI licensing 107 4.1.1 The NPI approach 112 4.1.2 The NQ approach 113 4.1.3 The lexical ambiguity approach 113
vi
4.1.4 NCI analyses in Arabic: Syntactic agreement approach 115 4.1.5 NCI licensing and phase theory 135 4.2 Licensing disjunction phrases and clauses 137 4.2.1 Introduction and empirical generalizations 138 4.2.2 Theoretical issues 143 4.2.3 Compositional meaning of coordinate complexes 146 4.2.4 Analysis of negative coordination by wala 151 4.2.4.1 Status and locus of laa 152 4.2.4.2 Status and locus of wala as an NCI and disjunction operator 154 4.2.4.3 Ellipsis 159 4.2.4.4 Bare argument ellipsis and coordinate complexes 160 4.2.5 Conclusion 168 5. PSIs with Head-Like Properties 171 5.1 Head-like PSIs as adverbs with phrasal projection 171 5.1.1 Empirical generalizations 173 5.1.2 The syntactic status of head-like PSIs 177 5.1.2.1 Head-like PSIs as syntactic heads selecting a NegP 178 5.1.2.2 Head-like PSIs as syntactic adverbs in Spec-XP 180 5.1.3 Analysis of head-like PSIs as phrasal heads 181 5.1.3.1 PSIs as XPs 183 5.1.3.2 The NP subject preceding head-like PSIs 186 5.1.4 Head-like PSIs as XPs: Licensing by negative constituents 188 5.1.5 Conclusion 195 5.2 Head-like PSIs and case licensing 196 5.2.1 Distribution of PSIs with head-like properties 197 5.2.2 Temporal adverbs and their complements 198 5.2.3 Licensing non-nominative case in Arabic 200 5.2.4 Proposal for accusative case licensing on complements of PSIs 202 6. Summary and Conclusions 211 6.1 The syntax of NPIs 213 6.2 The syntax of negative indefinites 215 6.3 The syntax of head-like PSIs 221 Bibliography 225 Index 239
[ vi ] Contents
vi
ACKNOWLEDGMENTS
I would like to thank Georgetown University for supporting research in this book with the 2019 and 2020 Summer Academic Grant (SAG) Awards from the Provosts office. Parts of the research reported in this book have been presented in numerous venues: the Annual Meeting of the Linguistic Society of America, Annual Symposium on Arabic Linguistics, the Linguistics in the Gulf conference, the Georgetown University Linguistics Speaker Series, and the Georgetown University Roundtable on Linguistics. I thank the audiences at all of those talks. For their many helpful comments, I especially thank Elabbas Benmamoun, Youssef Haddad, Peter Hallman, Hamid Ouali, Usama Soltan, Ali Idrissi, and Fassi Fehri. I am very grateful to Elabbas Benmamoun for encouraging me to develop my work on polarity sensitivity and write a book to fill a lacuna in research. I also wish to thank the anonymous reviewers of this book for Oxford University Press and related talks and publications for their invaluable comments that helped me sharpen the proposals and analyses in this book. I also thank my Georgetown University colleagues, Ruth Kramer and Carole Sargent in particular, for their support and encouragement. For her editorial help, I thank Lindley Winchester for meticulous edits and helpful suggestions. I am also very fortunate and grateful to have native speaker judgments on the dialects represented in this book. For their help with data on Moroccan, Egyptian, and Qatari Arabic, I am particularly grateful for Atiqa Hachimi, Ali Idrisi, Hamid Ouali, Usama Soltan, Hanaa Kilany, Hany Fazza, Hessa Al- Noaimi, Nasser Al-Attiyah, and Sara Al-Ahmad. Needless, to say, all errors and responsibility of the outcome of the work are mine alone. I also thank the chief editor of the series, Richard Kayne, and the acquisition editor, Meredith Keffer, for their enthusiasm, time, and effort in processing this project, getting the invaluable reviews from experts in the field and supporting the publication of the book. I would like to express my appreciation to the assistant editor, Macey Fairchild, and project manager, Koperundevi, for their help with moving the manuscript to production. As ever, the love and support of my family were instrumental to the completion of this project.
vi
ix
A B B R E V I AT I O N S
1 first person 2 second person 3 third person acc accusative adj adjective adv adverb agr agreement AP active participle asp aspect comp complementizer conj conjunction cop copula CP complementizer phrase def definite dem demonstrative det determiner DP determiner phrase f feminine FocP focus phrase FP functional projection fut future gen genitive imp imperative ind indicative indef indefinite IP inflection phrase ipfv imperfective juss jussive m masculine NCI negative concord item neg negative
x
NegP negation phrase nom nominative NP noun phrase NPI negative polarity item p plural par particle part participle pfv perfective pp prepositional phrase prep preposition pron pronoun sbjv subjunctive sg singular Spec specifier TP tense phrase VP verb phrase
[ x ] Abbreviations
1
CHAPTER 1
Issues in the Syntax of Polarity-Sensitive Items
1.1. INTRODUCTION: POLARITY-S ENSITIVE ITEMS
This book is a comparative study mainly concerned with the syntax of polarity- sensitive items (PSIs) in Arabic. Scholars have not given much attention to the syntax of PSIs, especially in the generative studies on Arabic syntax. The contribution in this book fills the gap in generative research on this vital category of Arabic syntax. PSIs include the Arabic counterparts to the English “negative indefinite” (NI) expressions (Haspelmath, 2005), such as nobody, nothing, and never, which can also be negative concord items (NCIs) in Arabic. PSIs also include negative polarity items (NPIs) equivalent to the English NPIs, such as anybody and anything. Tables 1.1 and 1.2 show the inventory of Arabic PSIs in Jordanian (JA), Qatari (QA), Egyptian (EA), Moroccan (MA), and Standard Arabic (SA). PSIs belong to various types of lexical categories and interact with a wide range of syntactic categories and grammatical environments, such as negation, disjunction, questions, and conditionals. These environments share the semantic notion of nonveridicality, whereby the proposition of the sentence does not ensure the truth (Giannakidou, 1994, 1995, 1997, 1998, 1999, 2009). PSIs, however, display different degrees of sensitivity to these grammatical environments. This book focuses on the syntactic licensing approach to explain the complex patterns of sensitivity to polarity. This project aims at characterizing the micro-syntactic variation that different types of PSIs in different varieties of Arabic exhibit with respect to the various grammatical environments. The Arabic varieties discussed in this comparative study are the Jordanian variety of Levantine Arabic (JA), the Qatari variety of Gulf A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0001.
2
Table 1.1. INVENTORY OF NIS
person
English
JA
EA
MA
QA
SA
nobody
walaħada
walawaaħid
ħəttawaħəd
maħħad
laaʔaħada
maħadaaš
maħaddiš
thing
nothing
wala(ʔi)ši
walaħaaga
ħəttaħaʒa
mašay
laašayʔa
time
never
walaʕumr
walaʕumr
maʕəmmər
ʔabad
ʔabadan
maʕumriš
maʕumriš
ʔabadan
ʔabadan
wala
wala
ħətta
ma
laa
determiner
no
Note: Table 2.2 presents a more comprehensive inventory of NIs, which I discuss in c hapter 2, is summarized in table 1.1.
Table 1.2. INVENTORY OF NPIS
person
English
JA
EA
MA
QA
SA
anybody
(ʔay)ħada
(ʔay)ħadd
(ši) waħəd
(ʔay) ʔaħad
(ʔay)ʔaħad
thing
anything
(ʔay)ʔiši
(ʔay)ħaaga
(ši) ħaʒa
(ʔay) šay
(ʔay)šayʔ
time
ever
ʕumr
ʕumr
ʕəmmər
gad
qatˁ
determiner
any
ʔay
ʔay
ši
ʔay
ʔay
Note: Table 1.2 summarizes a more comprehensive inventory of NPIs in table 2.1, which I discuss in chapter 2.
Arabic (QA), EA, MA, and SA. The Southern Levantine data primarily derives from the dialect spoken in the Jordanian Houran, the author’s native variety. The GA (Gulf Arabic), EA, MA, and SA data are drawn from grammar books and scholarly studies and is complemented with novel data from the author’s native dialect, native speakers of GA, EA, MA, and a corpus study of negation in Classical and Modern Standard Arabic. The corpus study is a syntactic analysis of negation in the Qur’an (around 86,000 words) and Levantine literature (around 86,000 words). The importance of this topic for Arabic syntax in particular and syntactic theory in general has at least two aspects. First, from the literature on PSI syntax, we can gain insight into how a given language establishes dependency relations (or a lack thereof) between PSIs and negation and what syntactic configurations are necessary for those dependency relations (e.g., Alqassas, 2015, 2016; Benmamoun, 1997, 2006; Ouali & Soltan, 2014). I argue that the specifier-head (Spec-Head) and c-command configurations are the only licensing configurations for NPIs, ruling out the head–complement relation, which counters Benmamoun’s (2006) idea that the head–complement relation is licensing some of these temporal NPIs (e.g., the NPI ʕəmmər in MA in table [ 2 ] A Unified Theory of Polarity Sensitivity
3
1.2). Restricting the licensing relations to the Spec-Head and c-command configurations allows for direct mapping from syntax to semantics, where NPIs are interpreted as existential quantifiers under the scope of negation. Crucially, the head–complement configuration whereby the NPI is the syntactic head taking negation as a complement feeds the semantic component (LF; Logical Form), a structure with negation not dominating the NPI. The variation in NPIs’ behavior is thus reduced to their syntactic properties, such as their locus in the structure relative to negation, which can occupy a position higher or lower than TP (tense phrase), and whether or not they undergo movement and reconstruction at LF. The book also distinguishes the indefinite nouns ħada ‘one’ and iši ‘thing’ from other indefinite nouns by showing how they cannot escape the scope of negation while other indefinite nouns can. This sets the indefinite nouns ħada ‘one’ and iši ‘thing’ apart from other indefinite nouns as NPIs that must receive a negative polarity interpretation when they are under the scope of negation, while they receive an existential interpretation when they are not under the scope of negation. Second, analyses of the different PSIs in various languages give us insights into the complex distributional contrasts and restrictions between PSIs and their licensors within a well-controlled environment made possible by the nature of microvariation studies (see, for instance, Alqassas, 2015, for LA; Ouhalla, 2002, for MA; and Soltan, 2012, for EA). The contribution of this analysis to linguistic theory is that NIs can, indeed, be semantically negative, as will be shown from split scope facts that contrast the behavior of the NI wala-NP (noun phrase) with the NI maħadaaš ‘no one’ and maʕumrhuuš ‘never,’ thus directly contradicting Penka’s (2011, p. 106) claim that NIs cannot be semantically negative. Using the split-scope phenomenon, I will show that wala- NP ‘nobody’ NIs exhibit this phenomenon, whereas maħadaaš ‘nobody’ and maʕumrhuuš ‘never’ do not. These facts suggest that Arabic has NIs that are more negative than the English NIs ‘nobody’ and ‘never.’ In this chapter, I give an overview of PSIs in Arabic and briefly outline the critical issues in the syntax of PSIs in Arabic and their contributions to Arabic syntax and linguistic theory.1 I will also outline the theoretical underpinnings of research on Arabic negation, relying on my most recent publications on Arabic negation. I will synthesize the major and crucial findings from cross- linguistic studies on this topic and my own published studies of PSIs in Arabic (Alqassas, 2015, 2018, 2019). I will also articulate the critical issues that remain unanswered and how I approach them in the subsequent chapters in terms of methodology and scope of content. The end of the chapter contains a road map to the content and organization of the book. 1. Part of this overview appeared in an earlier version as a paper titled Negative Sensitive Items published (2018) in The Routledge Handbook of Arabic Linguistics, Oxford: Routledge.
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 3 ]
4
PSIs are a category of words that interact with negation in a heterogeneous manner. This interaction takes place in at least three different ways. First, negation is usually required in simple declarative sentences containing a PSI. Second, certain PSIs can occur as fragment answers without the presence of negation, and they receive a negative interpretation. Third, certain negative markers (NMs) are in complementary distribution with certain PSIs. At least two different categories of words fall under the term PSI, and they differ with respect to their sensitivity to negation. One important aspect that distinguishes the two categories has to do with whether the PSI must co-occur with an NM. NPIs2 always require negation in simple declarative sentences and cannot function as fragment answers. The adverbial NPIs ʕumr in JA and EA and ʕəmmər in MA all require the NM maa and cannot function as fragment answers (see chapter 2 for all other NPIs in all of the Arabic varieties discussed in this book).3 (1)
(2)
(3)
a.
ʕumr-ha maa katbat NPI-ever-her neg write.3fsg.pfv ‘She has never written a novel.’
b.
Question: Answer: katbat riwaaye? *ʕumr-ha write.3fsg.pfv novel? NPI-ever-her ‘Has she written a novel?’ ‘Never.’
(JA)
a.
ʕumr-ii maa-saafirt Masr NPI-ever-me neg-travel.pfv.1sg Egypt ‘I have never traveled to Egypt.’ (Soltan, 2012, p. 241)
(EA)
b.
Question: ʔinta saafir-t Masr ʔabl kidah? you travel.pfv-2msg Egypt before this ‘Have you traveled to Egypt before?’
(EA)
a.
nadya ʕəmmər-ha ma-ʒat Nadia NPI-ever-her neg-come.pfv.3FSG ‘Nadia never came.’ (Benmamoun, 2006, p. 144)
riwaaye novel
Answer: *ʕumr-ii ever-my ‘Never.’
(JA)
(MA)
2. Throughout the chapters, I have occasionally modified glosses and translations of examples from the literature to simplify the presentation of the data. 3. In all examples, an asterisk outside the element in brackets indicates that the word is required, while an asterisk inside the brackets means the element is not allowed in the sentence. [ 4 ] A Unified Theory of Polarity Sensitivity
5
b.
Question: Answer: ʕəmmər-ha mšat l-təmma ? *ʕəmmər-ha never -her went.3fsg to-there ever-her ‘Nadia never went there.’ ‘Never.’
(MA)
The second category of PSIs, NCIs, need not always co-occur with negation. NCIs usually require an NM in simple declarative sentences, and they can function as fragment answers. The adverbial NCI ʔabadan is an example from JA (see c hapter 2 for all other NCIs in all of the Arabic varieties I discuss in this book). (4)
(5)
(6)
a.
ʔabadan maa katbat NCI.at all neg write.3fsg. pfv ‘She has never written a novel.’
riwaaye novel
b.
Question: katbat riwaaye? write.3fsg.pfv novel ‘Has she written a novel?’
Answer: ʔabadan NCI.at all ‘not at all’
(JA)
(JA)
a.
ħətta waħəd *(ma)-ʤa (MA) even one neg-come.3msg.pfv *‘Anyone didn’t come.’ (Benmamoun, 1997, p. 273)
b.
Question: škuun šəf-ti? who see.pfv-2sg ‘Who did you see?’
Answer: ħətta not-even ‘Nobody.’
waħəd (MA) one (Ouali & Soltan, 2014, p. 162)
a.
walaa waaħid gih no one came.3msg ‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
b.
Question: ʔinta šuf-t You saw-2msg ‘Who did you see?’
(EA)
Answer: miin? walaa waaħid (EA) who no one ‘nobody.’ (Ouali & Soltan, 2014, p. 162)
Another important difference between NPIs and NCIs is that most NPIs can occur in nonnegative contexts such as yes/no questions and conditionals, whereas NCIs cannot. The adverbial NPI ʕumr can occur in those nonnegative I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 5 ]
6
contexts, whereas the adverbial NCI ʔabadan cannot do so in JA (see other examples from the other dialects in chapter 2). (7)
(8)
a.
ʕumr-ha katbat NPI-ever-her write.3fsg. pfv ‘Has she ever written a novel?’
b.
ʔiða ʕumr-ha katbat riwaaye, if NPI-ever-her write.3fsg.pfv novel, ‘If she has ever written a novel, tell me.’
a.
*ʔabadan katbat riwaaye? NCI.at all visited.3fsg novel ‘Has she written a novel at all?’
b.
*ʔiða ʔabadan katbat if NCI.at all write.3fsg. pfv ‘If she wrote a novel at all, tell me.’
riwaaye? novel
(JA)
xabbir-ni tell-me
(JA)
(JA)
riwaaye?, novel,
xabbir-ni tell-me
(JA)
In chapter 2, I discuss the fact that NPIs can occur in nonnegative contexts for all NPIs in all Arabic varieties I address in this book.
1.2. CRITICAL ISSUES IN PSI SYNTAX
The importance of studying the syntax of PSIs for syntactic theory in general and Arabic syntax in particular can be seen in at least two domains. First, the literature on PSI syntax gives us insights into the dependency relations between PSIs, their licensing categories in a given language, and which syntactic configurations are necessary to establish them (e.g., Alqassas, 2015, 2016; Benmamoun, 1997, 2006; Ouali & Soltan, 2014). Second, the different analyses of the different PSIs in various languages have implications for the syntax of negation itself, such as the status of the enclitic NM and the locus of negation (Alqassas, 2015; Ouhalla, 2002; Soltan, 2012; and others). A brief explanation of these two critical issues in PSI syntax follows.
1.2.1. NPI dependencies
In the existing literature on NPI licensing, researchers have argued for a semantic approach to characterizing the unifying properties of the wide range of licensors, including words such as NMs, disjunctive coordinators, volitional [ 6 ] A Unified Theory of Polarity Sensitivity
7
verbs (want, suggest, insist), modal verbs and adverbials, and grammatical conditions such as questions, imperatives, habituals, and the subjunctive (Giannakidou, 1994, 1995, 1997, 1998, 1999, 2009). In these works, Giannakidou develops the semantic notion of “nonveridicality” that unifies these licensing categories and conditions. I will extend this approach to Arabic NPIs in this chapter, but I will also argue that syntax is involved in licensing these items. In particular, I will discuss the role of the categorical status, syntactic configurations, and syntactic processes such as movement in licensing NPIs. Antiveridical environments (a subset of nonveridicality) include negation (including the antiadditives nobody, nothing, never, before and the comparatives and antimorphics not X), which is antiveridical in that NOT p entails that p is false (Giannakidou, 1998, 2009). Antiveridicality conditions strict NPIs that negation can license. Strict NPIs require a licensor in their local syntactic domain (i.e., syntax mediates their licensing via an agreement relation with negation). Although the NPI ʕumr ‘ever’ is a non-strict NPI, the comparative adjective cannot license it, as this example shows: (9)
ʔawwal marra (b-ħayaat-i/ *ʕumr-i) first once (in-life-my/ ever) ‘the first time I ever saw them.’
šuft-hum saw-them
The behavior of Arabic NPIs suggests that they do not conform to the same classification that English NPIs do. I discuss the distribution of Arabic NPIs in nonnegative contexts in chapter 2 and analyze them in chapter 3. The syntactic licensing question for Arabic NPIs centers on whether negation must c-command them and whether this is a requirement in surface structure (overt syntax) or in the interpretation site (covert syntax). Different types of NPIs occupy different positions with respect to negation. Determiner and nominal NPIs usually follow negation, whereas adverbial and idiomatic NPIs can follow or precede negation. (10) a.
b.
ma-ʔaʤaa-š neg-came.3msg-neg ‘No one came.’
ħada/ NPI-one/
ʔayy ħada any one
ma-šaf-š ħaaga/ ħadd neg-saw.3msg-neg NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’
(JA)
(EA)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 7 ]
8
c.
lam yara šayʔan/ ʔaħadan neg.past see.3msg NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’
(SA)
b.
ʕumr-ha maa katbat NPI-ever-her neg write.3fsg.pfv ‘She has never written a novel.’
(JA)
c.
ma-maʕ-huu-š girš/ fils ʔiħmar neg-with-him-neg NPI-penny/ NPI-cent red ‘He doesn’t have a penny/red cent’
riwaaye novel
(JA)
I argue for and develop the analysis that captures how these various NPIs establish dependency with negation in c hapter 3. The analysis in this chapter shows that the specifier-head and c-command configurations are adequate for explaining the distribution of these NPIs, thus excluding the need for the head–complement relation proposed in Benmamoun (2006). This analysis unifies all types of NPIs under the same condition for mapping their syntactic structure to the semantic component. Under my analysis, all NPIs can be interpreted as existential quantifiers under the scope of negation in both the syntactic and semantic components. Due to this, the variation in the distribution of the various NPIs in the Arabic dialects lies in the syntactic properties of the particular NPI in the particular variety. The analysis probes the following syntactic properties of the NPIs: their locus in the syntactic structure relative to negation, which can occupy a position higher or lower than TP, and the availability of head or phrasal movement in the syntax proper prior to reconstruction at the semantic component (LF). I will discuss how these various NPIs establish dependency with negation in chapter 3. Moreover, the adverbial NPIs such as ʕəmmər can precede the subject and negation, and they exhibit head-like properties. (11) a.
b.
ʕəmmər nadya ma-mšat l-təmma never Nadia neg-went.3fsg to-there ‘Nadia never went there.’ (Benmamoun, 2006, p. 145)
(MA)
ʕəmmər-ni ma-mšit l-təmma never-me neg-went.1sg to-there ‘I never went there.’ (Benmamoun, 2006, p. 145)
(MA)
Such facts motivated a distinct analysis for how adverbial NPIs establish their dependency relations with negation in Moroccan (Benmamoun, 2006).
[ 8 ] A Unified Theory of Polarity Sensitivity
9
I will discuss this and juxtapose it with facts from the equivalent NPI in the other dialects in chapter 3.
1.2.2. NCI dependencies
I discuss the syntactic configurations and processes involved in licensing the various types of NCIs in Arabic in chapter 4. I argue for a syntactic agreement approach to explain the behavior of Arabic NCIs with extra mechanisms (positing an abstract negative operator). In this way, I explain non-strict NCIs by positing an inventory of abstract negation features assigned to the various types of Arabic NCIs à la Penka (2011) for Germanic languages on different grounds. (See c hapter 4 for a discussion of Penka’s analysis and a critique of the claim that NIs cannot be semantically negative.) For NCIs, the critical issues are of two types. First, the licensing questions pertaining to NPIs carry over to NCIs. The issue is whether negation licenses preverbal adverbial NCIs overtly in a local domain where negation precedes these NCIs or covertly in their interpretation site following negation (i.e., where negation scopes over them). (12) ʔabadan/ bilmarrah Faris ever/ at all Faris ‘Faris doesn’t like fish at all.’
ma-biħibb-iš neg-like.3msg-neg
l-samak the-fish
(JA)
The second issue is how determiner NCIs in preverbal position and in fragment answers have a negative interpretation despite the absence of any NM. (13) a.
*(ma)- ʔaʤaa-š neg-came.3msg-neg ‘No one came.’
wala-ħada NCI-one
b.
wala-ħada ma-ʔaʤaa-š NCI-one neg-came.3msg-neg ‘Nobody didn’t come.’ #‘Nobody came.’
c.
Question: Answer: miin ʔaʤa ? wala-ħada who came.3msg NCI-one ‘Who came?’ ‘no one’
(JA)
(JA)
(JA)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 9 ]
01
The answer lies in whether there are two homophonous lexical items, one being a negative quantifier (NQ) and the other an NPI (i.e., the lexical ambiguity approach) or whether there is a covert negative operator that licenses preverbal NCIs. In this chapter, I argue for a covert licensing approach, putting forward an analysis that appeals to the economy condition for introducing a covert negative operator. The economy condition on inserting a negative operator is triggered by the availability of the uninterpretable negation (uneg) feature that is not c- commanded by an NM specified for the interpretable negation (ineg) feature: (14) a.
ma-ʕumr-huu-š wala-ħada zaar l-batra neg-ever-him-neg NCI-person visited.3msg the-Petra ‘No one ever visited Petra.’ (negative concord reading)
(JA)
b.
wala-ħada ma-ʕumr-huu-š zaar l-batra NCI-person neg-ever-him-neg visited.3msg the-Petra ‘No one never visited Petra.’ (double negation reading)
(JA)
The postverbal NCI does not trigger a covert negative operator; in other words, an overt NM is obligatory because of a requirement for negation (negative morphology) to be preverbal, presumably to scope over the aspectual projection (AspP) in the clause (see Penka, 2011, and references therein adopting and supporting this hypothesis). The results of this approach produce a fine-grained analysis with an inventory of the formal negative features on NCIs and NMs, as table 1.3 shows. The contribution of this analysis to linguistic theory is that NIs can, indeed, be semantically negative, which split-scope facts that contrast the behavior of the NI wala-NP with the NI maħadaaš ‘no one’ and maʕumrhuuš ‘never’ will show.
Table 1.3. VARIATION OF FEATURES OF NMS AND NIS Language
Feature of NM/Op¬
Feature of NIs
JA, EA
Op¬ [ineg∅]
walaħada/ walawaaħid: [uneg]
maa: [ineg]
maħadaaš/ maħaddiš: [ineg]
ma- . . . -š: [ineg]
maʕumriš [ineg]
Op¬ [ineg∅]
all NCIs: [uneg∅]
MA
maa: [uneg] ma- . . . -š: [ineg] QA
maa: [ineg]
all NCIs: [uneg]
SA
maa: [ineg]
all NCIs: [uneg]
Note: Table 4.1 provides a more comprehensive inventory of features on NMs, NIs, and Multiple Agree typology in Arabic, which I discuss in chapter 4.
[ 10 ] A Unified Theory of Polarity Sensitivity
1
That certain NIs are inherently negative is in direct contradiction to Penka’s (2011) claim that NIs cannot be semantically negative. Particularly, I show that wala-NP NIs display split-scope readings, whereas maħadaaš and maʕumrhuuš do not. This, in turn, means that these NIs in Arabic are more negative than the English no-NP NIs. The following contrasts from JA show that the split-scope reading is possible with the NI wala-NP ‘no one’ but not with the NI maħadaaš ‘no one’: (15) wala ʔustaað laazim ykuun mawʤuud bi-li-mtiħaan NCI-one teacher should be present in-the-exam a. ‘It is not required that there be a professor present.’ ¬ > must > ∃ b. ‘There is no professor who is required to be present. ¬ ∃ > must (16) maħadaaš laazim ykuun mawʤuud bi-li-mtiħaan no one should be present in-the-exam a. *‘It is not required that there be anyone present.’ *¬ > must > ∃ b. ‘There is no one who is required to be present. ¬ ∃ > must
This chapter also discusses the phenomenon of Multiple Agree in strict and non-strict negative concord (NC) languages. In non-strict negative concord (non- strict NC) languages, Multiple Agree is obligatory; thus, we always get a concordant reading. In strict NC languages, Multiple Agree is optional; thus, we can get both NC and a double negation (DN) readings. Arabic, on the other hand, does not have Multiple Agree with non-strict NCIs such as wala-NP. As such, it behaves like English, a DN language, despite the fact that a concordant reading is obligatory with strict NCIs, such as the never-type NCIs ʔabadan and bilmarra ‘never.’
1.2.3. Enclitic negation and locus of negation
The syntactic analyses of PSIs affect at least two main issues. The first has to do with the status of the enclitic NM. The second concerns the locus of negation in the syntactic structure. One of the intriguing facts about the enclitic NMs in Arabic dialects that have bipartite negation is that they are usually mutually exclusive with PSIs. In MA, the enclitic NM and PSIs are mutually exclusive across the board; see the following examples from Benmamoun (2006) and other examples from MA, EA, and JA in chapter 5: (17) a.
nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
ma-ʒat(*-š) neg-came.3fsg(*-neg)
(MA)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 11 ]
21
b.
ma-ʒa(*-š) neg-came.3msg(*-neg) ‘No one came.’
ħətta waħəd even one
(MA)
Ouhalla (2002) put forward an analysis in which negation in MA has a negative operator that must bind a variable. The enclitic NM, for Ouhalla, is a dummy variable that the negative operator needs so it can bind a variable. In sentences that have a PSI, the PSI introduces a variable that negation can bind, thus rendering the enclitic NM unnecessary and preventing it from co-occurring in such contexts. The relevant point here is that discussion of such facts about PSIs prompted a proposal about the syntax of negation—namely, that negation has a negative operator that needs to bind a variable. In a similar vein, Soltan (2012) presented different facts from EA, showing that the enclitic NM cannot co-occur with the NPI ʕumr when preverbal but can do so when postverbal; see the following examples from Soltan (2012, p. 241) and similar examples from JA in chapter 3. (18) a.
ʕumr-ii maa-saafirt(*-iš) ever-me neg-travel.pfv.1sg-neg ‘I have never traveled to Egypt.’
Masr Egypt
(EA)
b.
maa-saafirt-iš Masr neg-travel.pfv.1sg-neg Egypt ‘I have never traveled to Egypt.’
ʕumr-ii ever-me
(EA)
Soltan (2012) advanced an analysis in which a post-syntactic condition deletes the enclitic NM. The deletion of the enclitic marker is predicated on the hypothesized condition against the mismatch between the formal features of negation on constituents that are in a licensing relation. The enclitic NM carries a [neg] feature whereas the NPI does not. The relevant point here is that the partial incompatibility between the NPI and enclitic negation prompted a post-syntactic analysis for negation. The scope of this chapter does not allow for detailed discussion of these analyses. I discuss these analyses and my analysis of mutual exclusivity between PSIs and negation in chapter 3. The location of negation in the syntactic structure is a critical issue in the syntax of negation. The issue is whether negation scopes over the tense of the sentence or over the event of the predicate. Without going into the details of this, suffice it to say that the syntax of adverbial PSIs such as ʕumr and baʕd and their interaction with negation provide one of various arguments for the location of negation in JA and other Arabic varieties (Alqassas, 2015, 2019). My proposal is that the locus of negation can be below the tense of the sentence (i.e., not scoping over it) or be above it (i.e., scoping over it). Section 1.3 has more details on my proposal for the locus of negation. [ 12 ] A Unified Theory of Polarity Sensitivity
31
The complexity and variation in licensing mechanisms opens up more questions for future research in this area. Among the important topics for future consideration are the status of temporal NPIs; the complementary distribution between certain PSIs and enclitic negation; and the broad question of why different categories of NPIs and NCIs seem to be licensed differently within the same dialect, whereas NPIs and NCIs of the same category are licensed differently across different dialects. Benmamoun (2006) presented arguments that the pre- negative NPI ʕəmmər ‘ever’ and the pre-negative NCI baqi ‘yet’ in MA can establish dependencies with negation without being in a Spec-Head or c-command relation with negation. Benmamoun presented two empirical arguments motivating this analysis. First, the subject can intervene between the NPI and negation. Consider the example with NPI ʕəmmər: (19) a.
b.
ʕəmmər nadya ma-mšat l-təmma never Nadia neg-went.3fsg to-there ‘Nadia never went there.’ (Benmamoun, 2006, p. 145)
(MA)
ʕəmmər-ni ma-mšit l-təmma never-me neg-went.1sg to-there ‘I never went there.’ (Benmamoun, 2006, p. 145)
(MA)
Second, this NPI exhibits head-like properties (hence Benmamoun’s term head NPIs): the NPI can host clitics and can assign accusative case to the clitic pronoun, as in (19)b. Based on these facts, Benmamoun proposed that the NPI heads its own phrase (XP; X phrase) and that negation is within the local domain of that NPI phrase (i.e., the NPI phrase establishes dependency with negation by taking the negation phrase [NegP] as its complement): (20)
XP X
NegP
This analysis nicely captures the fact that the verb agrees with the subject. The subject here can occupy the Spec/TP position. Recall that like the MA ʕəmmər, the NPI ʕumr in JA can also host a clitics, and the subject of the sentence can intervene between the NPI and negation (Alqassas, 2015, 2016): (21) ʕumr-ha *(ma) katbat ever-her neg write.3fsg. pfv ‘She has never written a novel.’
riwaaye novel
(JA)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 13 ]
41
(22) ʕumr Maha *(ma) katbat Ever Maha neg write.3fsg. pfv ‘Maha has never written a novel.’
riwaaye novel
(JA)
It is possible, however, that the subject in such cases is embedded in the NPI phrase consisting of the NPI ʕumr and the subject Maha. The NPI ʕumr requires a (pro)nominal complement, which either the pronominal clitic -ha or the proper noun Maha can satisfy (Alqassas, 2016). This analysis makes it reasonable to suggest that the NPI is in a local relationship with negation (i.e., the Spec-Head configuration). This analysis rests on the argument that the NM maa is above the tense phrase (for details of this argument, see Alqassas, 2015, 2016). (23)
NegP Spec NPI
Neg Neg maa
TP T
VP Spec
V
The challenge for this analysis is to explain how the verb can agree with a subject that is buried inside the NPI phrase. However, it is highly plausible that the verb agrees with a null subject (pro in generative terms) rather than the NP Maha. Consider the following examples from JA, where the NPI ʕumr precedes not the subject but a dislocated object in (24)a and a relative clause in (24)b: (24) a.
b.
ʕumr l-walad maa ever-him the-boy neg ‘A girl never loved the boy.’
ħabbat-o love.3fsg-him
ʕumr illi budrus Never who study.3msg ‘He who studies never fails.’
maa neg
bint girl
(JA)
b-ursub asp-fail.3msg
(JA)
Crucially, the object l-walad and the relative clause illi budrus already serve as complements for the NPI. Thus, the NPI projection, represented as XP in (20), cannot select NegP as a complement. See the following illustration: (25) a. [XP X ʕumr [DP l-walad [NegP ma . . .]]] b. [XP X ʕumr [CP illi budrus [NegP ma . . .]]]
[ 14 ] A Unified Theory of Polarity Sensitivity
51
The fact that a constituent like an object or relative clause can intervene between the NPI and negation suggests that the NPI is a phrasal head that negation licenses in its Spec-NegP. The NPI selects either the object or the relative clause as a complement. See the following illustration: (26) a. b.
[NegP [XP [X ʕumr] [DP l-walad ] [Neg ma . . .]]] [NegP [XP [X ʕumr] [CP illi budrus] [Neg ma . . .]]]
I developed and argued for this analysis in Alqassas (2016). Chapter 3 extends this analysis to the counterpart NPIs in the other Arabic dialects. With regard to the complementary distribution of certain PSIs with enclitic negation (see Ouhalla, 2002, for MA; Soltan, 2012, for EA), at least two questions remain. First, why are all MA PSIs incompatible with enclitic negation, while in JA and EA only the NPI ʕumr in the preverbal position shows this incompatibility, as in (27)? The following EA data are from Soltan (2012, p. 241); the MA data are from Benmamoun (2006, pp. 143–144). (27) a.
ʕumr-o ma-zaar(*-iš) NPI-ever-him neg-visit. pfv.3MSG-neg ‘He never visited Petra.’ ʕumr-o NPI-ever-him
el-batra def-Petra
b.
ma-zaar-iš neg-visit. pfv.3MSG-neg ‘He never visited Petra.’
c.
ʕumr-ii maa-saafirt(*-iš) NPI-ever-me neg-travel.pfv.1sg-neg ‘I have never traveled to Egypt.’
d.
maa-saafirt-iš Masr ʕumr-ii neg-travel.pfv.1sg-neg Egypt NPI-ever-me ‘I have never traveled to Egypt.’
(28) a.
b.
nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
el-batra def-Petra
(JA)
Masr Egypt
(EA)
ma-ʒat(*-š) neg-came.3fsg(*-neg)
ʕumar baqi ma-ʒa(*-š) Omar yet neg-came.3msg(*-neg) ‘Omar hasn’t come yet.’
(JA)
(EA)
(MA)
(MA)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 15 ]
61
c.
ma-ʒa(*-š) neg-came.3msg(*-neg) ‘No one came.’
d.
ħətta waħəd even one ‘No one came.’
ħətta waħəd even one
(MA)
ma-ʒa(*-š) neg-came.3msg(*-neg)
(MA)
Another question that arises is whether these PSIs are, indeed, in complementary distribution with enclitic negation (i.e., whether the enclitic NM disappears in this context). A possible approach to this phenomenon is to propose that enclitic negation involves the use of a different NM, ma. This is the proposal in Alqassas (2012, 2015). The basic idea in Alqassas (2012, 2015) is that there are two different negatives: a single negation marker maa and a bipartite negation marker ma . . . š. These NMs differ in their syntactic position: single negation occurs above the TP and bipartite negation occurs below TP (see Alqassas, 2015, for empirical support and detailed arguments of this proposal). To see how this works, let us take the NPI ʕumr in JA as an example. This NPI in a preverbal position cannot co-occur with enclitic negation; it also tends to occur at the left periphery of the sentence, preceding the verb and the preverbal subject: (29) ʕumr Sara ma saafarat ever Sara neg traveled.3fsg ‘Sara never traveled by herself.’
la-ħaalha by-herself
(JA)
If bipartite negation is below TP and this NPI is above TP, it follows that this NPI is not within the local domain of negation (i.e., not in Spec,NegP). This is because the tense phrase separates the NPI from bipartite negation. Single negation, however, is above TP, and the NPI can be within the local domain of the negation phrase. Evidence in favor of locating single negation above TP appears in the following example, where an adverb intervenes between single negation and the verb. If single negation precedes the adverb and the tensed verb saafarat follows the adverb, then clearly single negation is above tense in the syntactic structure. (30) ʕumr Sara ma b-yo:m saafarat ever Sara neg in-a day traveled.3fsg ‘Sara never traveled by herself on any day.’
[ 16 ] A Unified Theory of Polarity Sensitivity
la-ħaalha by- herself
(JA)
71
In chapter 3, I discuss how this proposal can explain putative cases of complementary distribution between certain PSIs and enclitic negation in other Arabic dialects. A similar discrepancy arises in the context of NCIs: The MA NCI ħətta waħəd is in complementary distribution with enclitic negation, but the EA and JA NCIs wala-waħid/wala-ħada are not. Comparative studies between the dialects should yield explanations for these NCI contrasts. Cross- dialectal studies of PSIs can provide answers to the questions this section raises. Moreover, such studies can shed light on the extent of variation witnessed cross-linguistically: why does licensing take place at different levels in different languages and in different types of PSIs in the same language? Answers to such questions can lead licensing analyses to more adequate explanations as we will see in chapter 4.
1.2.4. PSIs with head-l ike properties
Two main issues arise from PSIs with head-like properties. The first is how these PSIs are licensed in syntax and whether positing a special syntactic configuration or reducing these contrasts to the sectional properties and syntactic operations such as movement can explain their distributional contrasts. Another issue that arises from these NPIs with head-like properties is their ability to host clitics with accusative and genitive case marking. This issue raises interesting questions pertaining to case theory and dependent case licensing. For the licensing question, I argue in this chapter that these PSIs in Arabic are adverb phrases that select a (pro)nominal or CP (complementizer phrase) complement but arguably project a clausal XP on top of the negation projection. This chapter expands on my previous work in Alqassas (2016, 2018). My analysis of these PSIs allows for direct mapping from syntax to semantics, and the analysis reduces the variation in the behavior of PSIs to their syntactic properties, such as their locus in the structure relative to negation, which can occupy a position higher or lower than TP, and whether they undergo movement and reconstruction at LF. The second issue is how these NPIs license an accusative or genitive complement. In Arabic, complements of verbs and complementizers take the accusative case, whereas complements of nouns and prepositions take the genitive case. (31) Complement of a verb (accusative) [VP zaar-ni] visited.3MSG-I.acc ‘He visited me.’
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 17 ]
81
(32) Complement of a complementizer (accusative) baʕrif [CP in-ni [TP [VP zurt-o]] know.1sg comp- I.ACC visited-him ‘I know that I visited him.’ (33) Complement of a noun (genitive) ktaab-i bi-l-maktabeh book-my.GEN in-the-library ‘My book is in the library.’ (34) Complement of a preposition (genitive) Zeed ħaka maʕ-i Zeed spoke.3MSG with-me.GEN ‘Zeed called me yesterday.’
nbaariħ yesterday
Surprisingly, the complement of a head PSI can carry an accusative case: (35) Complement of a head PSI (accusative) a. ʕumr-ni maa saafart NPI.ever-me.ACC neg traveled.1sg ‘I have never traveled.’ b.
baʕd-ni maa NCI.yet-me.ACC neg ‘I haven’t traveled yet.’
saafart traveled.1sg
Given the cross-linguistic pattern of case assignment rules that rely on the syntactic category of the assigner, temporal adverbs typically pattern with prepositions or nouns in assigning case. Such a case assignment rule incorrectly predicts that PSI complements will be assigned genitive case. The NCI adverb baʕd ‘yet’ lexically derives from the preposition baʕd ‘after.’ The NPI adverb ʕumr ‘ever’ lexically derives from the noun ʕumr ‘lifetime, age’: (36) a.
b.
door-ak baʕd-i turn-you.GEN after-me.GEN ‘Your turn is after me.’ ʕumr-i ʕišreen saneh age-my.GEN twenty year ‘My age is twenty years.’
[ 18 ] A Unified Theory of Polarity Sensitivity
91
The lexical homophones of these PSI adverbs, crucially, cannot have an accusative complement: (37) a.* door-ak turn-you.GEN b.* ʕumr-ni age-my.ACC
baʕd-ni after-me.ACC ʕišreen saneh twenty year
Verbs typically assign accusative case to nominals that are semantic arguments with a thematic role (hence the term inherent case). Recall that there is no thematic role for nominal complements of these head PSIs, which poses a contradiction given that the accusative case in such sentences is structural. One might ask whether these PSIs might be impersonal verbs that take the (pro)nominal as an impersonal subject (see Lucas, 2009, pp. 207, 221). This, in turn, would suggest that accusative case appears when nominative case is already assigned, with the T head assigning the nominative case to the null subject pro in Spec-TP and the PSI licensing the impersonal subject via accusative case. In Arabic, however, these PSIs do not share key characteristics of impersonal verbs. Impersonal verbs in Arabic select a main verb in the subjunctive mood whose subject is coindexed with the complement of the impersonal verb. The following examples from SA and JA illustrate these characteristics. SA marks the subjunctive case with the suffix -a, but JA marks it by a lack of the prefix b-, the marker of imperfective aspect: (38) a.
b.
yumkinu-ni ʔan ʔusaafir-a can.3MSG-me.ACC comp travel.1sg-sbjv ‘It is possible for me to travel.’ b-yimkin-ni ʔasaaafir asp-can.3MSG-me.ACC travel.1sg.sbjv ‘It is possible for me to travel.’
(SA)
(JA)
These PSIs, on the other hand, do not carry phi-feature morphology; they can have a complement that is not coindexed with the subject of the main verb, and the main verb is indicative. (39) a.
ʕumr-o maa šakat minn-oh mʕalmeh NPI.ever-him neg complained.3FSG about-him teacher.fsg ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 19 ]
02
b.
baʕd-o maa šakat minn-oh NCI.ever-him neg complained.3FSG about-him ‘No teacher has complained about him yet.’
mʕalmeh teacher
I propose that the accusative case of PSI complements is assigned by the NM maa. There is evidence that NMs can assign accusative case on par with complementizers in SA. This is clear from the NM laa, which can assign accusative case to the subject of a verbless sentence (see Al-Naadirii [2009], cited in Benmamoun et al., 2013): (40) laa mudarris-iin neg teachers-mpl.ACC ‘No teachers are absent.’
ɣaaʔib-uun absent-mpl.NOM
Through my analysis in c hapter 5, I am proposing a revision of the agreement- based account of case licensing. I contend that accusative case is the morphology found on nominal complements that enter into an agreement with the syntactic heads v or neg, nominative case is for nominal subjects that agree with the T head, and genitive case is for nominal complements that do not enter into a phi-agreement relation with their head. The location of negation in the syntactic structure is a critical issue in the syntax of negation. The issue is whether negation scopes over the tense of the sentence or the event of the predicate. This chapter lays out the proposal in Alqassas (2015, 2019) that the locus of negation can be below the tense of the sentence (i.e., not scoping over it) or above it (i.e., scoping over it).
1.3. SENTENTIAL NEGATION AND PSI SYNTAX 1.3.1. Def inition of sentential negation
Scholars often distinguish sentential negation from constituent negation using a variety of tests, most of which probe the scope of negation over particular variables in the sentence. Such variables include the event of the sentence and quantifiers such as every.4 What matters to the syntactic analysis of PSIs is whether negation introduces a functional projection NegP that dominates (c-commands) the PIs (Polarity Items) or not. Following Pollock (1989), I assume that sentential negation introduces a NegP functional projection. 4. This section on Arabic negation is based on my book, A Multi-Locus Analysis of Arabic Negation (Alqassas, 2019, Edinburgh University Press). Some of the research on NSI syntax in this book is based on preliminary research in that book. The present book supersedes it. [ 20 ] A Unified Theory of Polarity Sensitivity
12
Nonetheless, I depart from recent proposals that treat any NM that scopes above the event of the predicate as a case of sentential negation. Penka (2011) defined sentential negation as “negation taking scope at least above (the existential quantifier binding the event argument of) the main predicate” (p. 8). This definition includes cases of sentential negation whereby negation scopes over the event directly, as in (41)a, and cases like (41)b, where the quantifier intervenes between negation and the verb; thus, the scope of negation over the event of the predicate is indirect or not immediate. (41) a.
kull l-ʔawlaad ma-ʔaʤuu-š all the-boys neg-came.3mp.-neg ‘All the boys didn’t come.’ all > ¬
(JA)
b.
miš kull l-ʔawlaad ʔaʤu not all the-boys came.3mp. ‘Not all the boys came.’ ¬ > all
(JA)
When negation precedes the universal quantifier kull, we get the reading not > all, where negation scopes over kull. But when negation follows kull, we get the reading all > not, where negation does not scope over kull. I treat cases where negation does not immediately scope over the event of the predicate as cases in which negation does not project a NegP at the clausal level. Negation in such cases is buried inside the quantifier constituent. Evidence for this distinction appears in contrasts over whether these two different types of negation structures can license an NPI. Crucially, negation in (42)b cannot license an NPI, and the only possible reading for the indefinite noun is as a positive polarity item (PPI) equivalent to the English PPI something: (42) a.
b.
kull l-banaat ma- štarinni-š all the-boys neg-bought.3fp.-neg ‘All the girls didn’t buy anything.’
ʔiši thing
miš kull l-banaat štarin not all the-girls bought.3fp * ‘Not all the girls bought anything.’ ‘Not all the girls bought something.’
ʔiši thing
(JA)
(JA)
I will lay out more on the distinction between PPIs and NPIs in c hapter 2, and I will discuss NPI licensing in chapter 3. The basic negation facts about the Arabic varieties that I present in this book show that all have NMs preceding the verb. They all share the NM maa, which can precede imperfective verbs (43) and perfective verbs (44). I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 21 ]
2
maa y-uħibbu Zayd-un al-qira’ata neg like.imperf.3msg Zayd-nom the-reading-acc ‘Zayd does not like reading.’ (Ouhalla, 1993, p. 276)
(SA)
b.
maa t-ħibb le-ktaaba neg like.3fsg the-writing ‘She does not like writing.’
(QA)
c.
ma-ba-saafr-iš neg-asp-travel.1sg-neg ‘I don’t travel much.’
(JA)
d.
maa-b-asaafir-š kətiir neg-asp-travel.1sg-neg much ‘I don’t travel much.’ (Soltan, 2011, p. 259)
(EA)
e.
ʕumar ma-ta-y-šrəb-š Omar Neg-asp-3msg-drink-Neg ‘Omar does not drink.’ (Benmamoun, 2000, p. 83)
(MA)
(43) a.
(44) a.
kθiir much
maa kataba neg wrote.3msg ‘He has not written.’
(SA)
b.
maa katab neg wrote.3msg ‘He has not written.’
(QA)
c.
ma-saafart-iš neg-asp-traveled.1sg-neg ‘I did not travel.’
(JA)
d.
maa-saafir-t-i-š (EA) neg-traveled-1sg-ev-neg ‘I did not travel.’ (Soltan, 2011, p. 257)
e.
ma-qra-š l-wəld (MA) Neg-read.past.3MSG-Neg the-boy ‘The boy didn’t read.’ (Aoun et al., 2010, p. 96)
[ 22 ] A Unified Theory of Polarity Sensitivity
32
In JA, EA, and MA, an NM can follow the verb as well: all three varieties have the marker -š, which is usually suffixed to a verb preceded by the NM ma, as (43) and (44) show. This type of negation (ma . . . š) is known as bipartite or discontinuous negation. SA and QA do not possess this postverbal NM. In JA, EA, and MA, a nondiscontinuous form of bipartite negation usually negates nonverbal predicates: (45) a.
Zeed miš muhandis Zeed neg engineer ‘Zeed is not an engineer.’
(JA/EA)
b.
Zeed mub muhandis you neg engineer ‘You are not an engineer.’
(QA)
c.
ʕumar ma-ši muʕallim (MA) Omar neg-neg teacher ‘Omar is not a teacher.’ (Benmamoun, 2000, p. 46)
While SA lacks -š, it also possesses its own unique NMs lam and lan, which are not found in other Arabic varieties. The word lam is an NM with a past- tense interpretation that precedes imperfective verbs. The word lan is an NM with a future-tense interpretation that precedes imperfective verbs. Table 1.4 shows the inventory of NMs in the Arabic varieties discussed in this book. (46) a.
* (lam) yaštari neg.past buy.3MSG ‘He didn’t buy anything.’
šayʔan NPI.thing
(SA)
Table 1.4. INVENTORY OF NMS JA
EA
MA
QA
SA
single
maa
maa
maa
maa
maa
discontinuous
ma- . . . -š
ma- . . . -š
ma- . . . -š
—
—
nondiscontinuous
miš
miš
maši
mub
—
tensed
—
—
—
—
lam (past)
laa
lan (future)
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 23 ]
42
b.
* (lan) yaštari neg.future buy.3MSG ‘He will not buy anything.’
šayʔan NPI.thing
(SA)
1.3.2. The locus of sentential negation
There are two views in the literature on the syntax of negation regarding the position of negation in the clause structure. The first view is the parametric view, in which within a given language, NegP selects either a vP or a TP but not both (see, for instance, Benmamoun, 1992, 2000; Hoyt, 2007; Mohammad, 2000; Ouhalla, 1992; Soltan, 2007). This view restricts the locus of negation to one position in the syntactic structure; thus, I will refer to it as the uni-locus view. Under the second view, a certain language may opt for any position of negation, with the low position (selecting a vP) being the default and the higher position having special semantic/pragmatic and syntactic effects (see Alqassas, 2015, for JA; Ramchand, 2001, for Bengali; Zanuttini, 1998, for Romance languages). This second view does not restrict the locus of negation to two positions; thus, I termed it the multi-locus view or analysis in Alqassas (2019). In Alqassas (2015, 2019), I proposed that the morphologically weak ma/la is the head of a NegP projection below TP, whereas the morphologically strong maa/laa is the head of a NegP projection on top of TP. The morpheme -š is an agreement clitic that reinforces the weak proclitics ma and la. The different structural positions of negation have different syntactic and semantic and pragmatic effects. I argue that the variation in the position of NegP results not from a syntactic parameter (whether NegP selects TP or VP) but instead from certain syntactic or semantic and pragmatic properties of negation (see Alqassas, 2015, for Jordanian; Ramchand, 2001, for Bengali; Zeijlstra, 2004). According to this view, the semantic properties of negation differ in terms of whether the NM binds the event variable of the predicate (vP) or the time variable of tense inside the TP (Zeijlstra, 2004). Furthermore, the unmarked way to express sentential negation is to bind an event variable so that NegP is on top of vP (see Acquaviva, 1997; Alqassas, 2015; Giannakidou, 1999, 2000, 2006; Zeijlstra, 2004). In showing the two different negation structures in Bengali, Ramchand (2001) proposed two structural positions to explain why only the NM ni can license the NPI adverb kokhono ‘ever,’ whereas the NM na cannot do so, as the following contrasting examples show. (47) a.
Ami kokhono an I ever mangoes ‘I never ate mangoes.’
[ 24 ] A Unified Theory of Polarity Sensitivity
khai ate
ni neg
(Bengali)
52
b.
* Ami kokhono an I ever mangoes ‘I never ate mangoes.’
khai ate
na neg
(Bengali)
The NM ni happens to be the marked one. The NPI adverb is presumably in a position above TP because it is a time adverb. The NM ni can thus license this NPI because it is in a NegP on top of TP. That is, the NM binds a time variable in addition to binding the event variable. At the same time, failure of the NM na to license the NPI adverb suggests that na is in a NegP on top of vP. In other words, na binds an event variable rather than a time variable. I discussed the details of the binding relationship and its consequences in Arabic in Alqassas (2019), where I used Arabic data to analyze the relationship between negation and adverbs of time, thus accounting for many word order restrictions and semantic ambiguities in Arabic. The purpose of the present section, however, is to show the merits of the multi-locus analysis of negation in Alqassas (2019) by investigating the interaction between adverbs and subjects on the one hand and negation on the other. In Alqassas (2015, 2019), I argued that lower negation (NegP below TP) binds the event variable of the verb, whereas higher negation (NegP above TP) binds the time variable of tense. The contrasts between lower and higher negation yield different scope readings (narrow and wide) for negation in the context of time adverbials. With higher negation (maa by itself), only a wide scope reading is possible, whereas both narrow and wide scope readings are possible with lower negation, bipartite negation (see the following examples from Alqassas, 2015, p. 116–117).5
5. The locus of an element in the syntactic structure affects its scope. Consider the following examples on the scope of the adverb frequently in English (Engels, 2014): (1)
(2)
John could frequently lift 200 pounds. i. ‘John was frequently able to lift 200 pounds.’ ii. ‘John was able to lift 200 pounds several times (in a row).’ John frequently could lift 200 pounds. i. ‘John was frequently able to lift 200 pounds.’ ii. #‘John was able to lift 200 pounds several times (in a row).’
The position of the adverb lower than the copula yields an ambiguous reading, with the adverb frequently having both a wide scope reading (i.e., scoping over the modal verb could) and a narrow scope reading (i.e., scoping over the verb lift). By contrast, the position of the adverb higher than the modal verb could allows only for a wide scope reading. The different scope readings are the result of the two different positions of the adverb in relation to tense.
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 25 ]
62
maa štaɣalt la-muddit ʔusbuuʕ (# bas štaɣalt (JA) neg work.1sg.pfv for-period week (# but work.1SG.pfv θalaθt-iyyam) three-days) ‘I did not work for a week.’ i. ‘I did not work at any point during the week.’ ii. #‘I did not work for seven days, but I worked for three days.’ No t:[t< t* and t in I] ∃e:[tf ∈ T (e) = t] [working(e) & Θ1 (e, ‘I’) & ‘for a week’(I)]
(48) a.
ma-štaɣalt-iš la-muddit ʔusbuuʕ (bas štaɣalt (JA) neg-work.1sg. pfv-neg for-period week (but work.1SG.pfv θalaθt-iyyam) three-days) ‘I did not work for a week.’ (but I worked for three days) i. ‘I did not work at any point during the week.’ ii. ‘I did not work for seven days, but I worked for three days.’ ∃t:[t< t* and t in I]No e:[tf ∈ T (e) = t] [working(e) & Θ1 (e, ‘I’) & ‘for a week’(I)] b.
Building on proposals in Ramchand (2001, 2002) and Alqassas (2015), I propose that negation introduces an existential quantifier that binds the variables that the predicate and its arguments introduce. The following is a simple illustration: (49) [Op¬ ∃e,x [ . . . (e) . . . (x) . . . ]]
When negation is below TP, this quantifier binds the event variable of the predicate. Assuming a neo-Davidsonian framework of semantic representation (Davidson, 1967; Parsons, 1990), we are able to analyze the readings in examples (48)a–b as follows. In (48)a there is no time during which the event happened. In (48)b, the structure specifies that there is a time during which the event did not take place, but it leaves open the possibility that the event could have taken place at some other time, hence the possibility of either a wide or a narrow scope reading. Negation above TP thus scopes over the whole proposition (including the time variable of tense). In Alqassas (2019), I proposed that this special pragmatic effect results from having the NM maa above TP. This marker can then bind the time variable of tense (and the event variable of the verb), resulting in the semantic interpretation of absolute or emphatic negation, as in (51)a. At the same time, the NM lam is below TP and binds the event variable of the verb but not the tense variable, as in (51)b.
[ 26 ] A Unified Theory of Polarity Sensitivity
72
(50) a.
b.
maa ʔittaxaða Allaahu min waladin neg take.3msg.pfv Allaahu any child ‘Allah has not taken any son.’(Qur’an, 23, p. 91)
(SA)
wa lam yattaxið-ø waladan and neg.past take.3msg.ipfv-juss child ‘And who did not take a son’ (Qur’an, 25, p. 2)
(SA)
(51) a.
No t:[t< t* and t in Allah] ∃e:[tf ∈ T (e) = t] [taking(e) & Θ1 (e, ‘Allah’)] [NegP maa [TP [vP ʔittaxaða . . .
b.
∃t:[t< t* and t in Allah]No e:[tf ∈ T (e) = t] [taking(e) & Θ1 (e, ‘Allah’)] [TP [NegP lam [vP yattaxið . . .
Based on these scope facts and distributional contrasts for adverbs and subjects (reported later in this section),6 in Alqassas (2015, 2019), I proposed that the different scope readings result from the availability of two positions for NegP. The first position is on top of TP (the high NegP), and the second one is
6. One interesting contrast between bipartite and single negation has to do with sentential stress. Sentential (nuclear) stress falls on the preverbal negative marker maa when -š is absent but on the negated predicate when -š is used. Notice that single negation involves a preverbal negative marker that is a separate word morphologically. It is a strong morpheme that receives primary stress. This differs from bipartite negation, in which primary stress falls on the appropriate syllable in the morphological complex neg-verb-neg following the JA rules of word stress. If maa and the verb below it were one word, we would expect the stress to fall on the second syllable (the verb), as in the two-syllable word baa`zaar ‘market’ in the following example (i)a. Yet, the stress falls on maa when single negation is used to negate a verb that has a heavy monosyllabic structure such as zaar ‘visited’ (see Alqassas, 2015): (i)
a.
baa`zaar bazar ‘Petra’s bazar.’
l-batra def-Petra
b.
`maa zaar neg visit.3msg.pfv ‘He did not visit Petra.’
(JA)
l-batra def-Petra
(JA)
In the case of the negative marker in bipartite negation, there is a straightforward answer if the verb merges with the negative marker via the syntactic process of incorporation.
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 27 ]
82
below TP and on top of vP (the low NegP). In this case, the strong NMs maa and laa occupy the head position of the high NegP, whereas the weak ones, ma and la, occupy the head position of the low NegP. The following is a schematic representation of the two structural positions for negation in Arabic: (52)
a. TP T
NegP
Spec
Neg’
Neg0 ma-
AgrPolP AgrPol0 -š
VP…
b. NegP Spec
Neg’ Neg0 maa
TP T
VP…
A crucial point for bipartite negation is that the enclitic -š is not dependent on the proclitic NM ma- in certain contexts. The preverbal marker is optional. In these very contexts, ma-, when used, must be adjacent to the verb. (53) a.
(ma)-basaamiħ-k-iš neg-forgive.1sg.ipfv-you-neg ‘I will not forgive you.’
(JA)
b.* ma b-yoom basaamiħ-k-iš (JA) neg in-day forgive.1sg.ipfv-you-neg ‘I will not forgive you in any day.’ (Alqassas, 2015, p. 114)
Because the preverbal marker is optional in these contexts, adjacency is not a requirement that the enclitic imposes. It is a syntactic requirement that follows from the bipartite markers (ma and enclitic -š) being below tense and adjacent to the verb or predicate. At the same time, single negation (maa by itself) allows for syntactic constituents to intervene between negation and the verb. In Alqassas (2019), I showed that these constituents can be adverbs, subjects, or the verbal copula kunt, suggesting that maa is higher than TP. In generative syntactic terms, bipartite negation is lower than TP, whereas single negation is higher than TP. [ 28 ] A Unified Theory of Polarity Sensitivity
92
(54) a.
b.
maa (*Zeed) (*fiʕlan) bi-saafir neg (*Zeed) (*really) asp-travel.3msg.ipfv kul yoom every day ‘Zeed doesn’t really travel every day.’
(Zeed) (fiʕlan) (Zeed) (really)
maa ħada fiʕlan bi-saafir neg anyone really asp-travel.3msg.ipfv ‘No one really travels every day.’
kul every
yoom day
(JA)
(JA)
Moreover, there are contexts in which certain adverbs and definite subjects can intervene between maa and the verb, such as b-yoom: (55) wallah maa b-yoom Zeed bisaamħ-ak (JA) by-God neg in-day Zeed forgive.3msg.ipfv-you ‘By God, Zeed will not forgive you in any day.’ = ‘Zeed will never forgive you.’ (adapted from Alqassas, 2015, p. 114) (56) a.
b.
kunt miš be.1sg.pfv neg ‘I was not upset.’ maa kunt neg be.1sg.pfv ‘I was not upset.’
zaʕlaan upset
zaʕlaan upset
(JA)
(JA)
Thus, the dependency analyses I develop in this book to explain the distributional contrasts between negation and PSIs build on the proposal that NegP can occupy a position below or above TP. (57) a. b.
[TP [NegP [AspP [VP . . . [NegP [TP [AspP [VP . . .
To show the empirical breadth to which this proposal applies, I will provide more examples of negation showing contrasts between different negation structure in GA and SA from Alqassas (2019). Recall that unmarked contexts arguably involve use of the NM that projects below TP in the canonical position. Therefore, we can predict that these contexts do not allow the subject to intervene between the NM and the verb. This prediction is borne out as follows:
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 29 ]
03
(58) a.
maa xallaw šay maa xað-uu (Kuwaiti) neg leave.3mpl.pfv thing neg take.3mpl.pfv-it ‘They didn’t leave anything they didn’t take.’ (Brustad, 2000, p. 285)
b.
ʔal-awlaad maa xallaw šay maa xað-uu (QA) def-boys neg leave.3mpl.pfv thing neg take.3mpl.pfv-it ‘The boys didn’t leave anything they didn’t take.’
c.
* maa ʔal-awlaad xallaw šay maa xað-uu (QA) neg def-boys leave.3mpl.pfv thing neg take.3mpl.pfv-it
In pragmatically marked sentences, the NM maa precedes the subject. Negation in (59) presupposes that “his mother and sister did not call.” (59) a.
leeš maa ʔumm-a w-leeš maa ʔuxt-a daggat-li (Kuwaiti) why neg mother-his and-why neg sister-his call.3fsg.pfv-to.me ‘Why didn’t his mother and why didn’t his sister call me?’ (Brustad, 2000, p. 293)
b.
leeš maa ʔumm-a tasˁlat-li! why neg mother-his call.3fsg.pfv-me ‘Why didn’t his mother call me!’
c.
leeš maa tasˁlat-li why neg call.3fsg.pfv-me ‘Why didn’t his mother call me?’
(60) a.
b.
ʔumm-a? mother-his
(QA)
(QA)
ʔintu ma-antu hali (Kuwaiti) you.pl neg-you.pl family.my ‘You are not my family.’ (Brustad, 2000, p. 297) ma-anta (b-)rifiʤ-i min alyoom w-raayeħ neg-you.sg asp-friend.my from today and-going ‘You are not my friend from now on!’ (expressing anger)
The NM maa can also precede the adverb gad ‘never’ in Qatari: (61) maa gad kallamt-a neg ever talk.1sg.pfv-him ‘I’ve never talked to him.’
[ 30 ] A Unified Theory of Polarity Sensitivity
(QA)
(QA)
13
Similarly, in pragmatically marked contexts, the NM in QA and JA can precede the subject of the nonverbal predicate in verbless sentences: (62) a.
ma-ani (b-)maakil l-yoom kaamil neg-I asp-having eaten def-day entire ‘I will not be eating the whole day!’ (expressing anger)
(QA)
b.
ma-ani maakil tˁuul l-yoom neg-I having eaten throughout def-day ‘I will not be eating the whole day!’ (expressing anger)
(JA)
But the NM is adjacent to the nonverbal predicate in unmarked contexts: (63) a.
b.
ʔana mub maakil I neg having eaten ‘I haven’t eaten the whole day.’
l-yoom def-day
ani miš maakil tˁuul I neg having eaten throughout ‘I will not be eating the whole day.’
kaamil entire
(QA)
l-yoom def-day
(JA)
The same contrasts can be found in QA and JA when we compare interrogative and rhetorical questions: (64) a.
b.
ʔinta mub ʕaarif you neg knowing ‘You don’t know the answer?’
al-ʤawaab? def-answer
(QA)
mub int ʕaarif al-ʤawaab! (QA) neg you knowing def-answer ‘Don’t you know the answer!’ (The speaker presupposes that the addressee knows the answer.)
SA, by contrast, resembles JA in that it has two different NMs that are associated with the two different positions in the syntactic structure. The NM maa is adjacent to the verb, whereas the NM laa is not, suggesting that maa is above and laa below TP. The NM maa can precede subjects of verbless sentences, but laa cannot:
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 31 ]
23
(65) a.
b.
maa Muħammad-un kaatib-un (SA) neg Mohammad-nom writer-nom ‘Mohammad is not a writer’ (Aoun et al., 2010, pp. 116–117) *laa neg
Muħammad-un Mohammad-nom
kaatib-un writer-nom
(SA)
The NM maa can also precede the subjects of verbal sentences, whereas laa cannot: (66) a.
maa ʔana neg I ‘I did not say this.’
b.
maa/*laa ʔana neg/*neg I ‘I do not say this.’
qul-tu haaða say.1sg.pfv this (Fassi Fehri, 1993, p. 165) ʔaquulu say.1sg.ipfv
haaða this
(SA)
(SA)
The indefinite subject ʔaħad-un can intervene between maa and the verb, suggesting that this subject is in Spec-TP, but this is not possible with laa. The negative laa and ʔaħad can form the lexical compound laa-ʔaħada, which can occur preverbally. Because ʔaħad does not carry the nominative case, laa- ʔaħada must be a lexical compound rather than the combination of a negative head laa and a subject pronoun ʔaħad. (67) a.
b.
(68) a.
maa ʔaħad-un yašukku fii qawli-ka (SA) neg one-nom doubt.3msg.ipfv in saying-you.MSG ‘No one is questioning what you said.’ (Fassi Fehri, 1993, p. 165) [NegP maa [TP ʔaħad-un [VP yašukku fii qawli-ka maa yašukku ʔaħad-un fii neg doubt.3msg.ipfv one-nom in ‘No one is questioning what you said.’ [NegP maa [TP [VP yašukku ʔaħad-un fii laa-ʔaħada yašukku fii no one doubt.3msg.ipfv in ‘No one is questioning what you said.’ [TP laa-ʔaħada [VP yašukku fii
[ 32 ] A Unified Theory of Polarity Sensitivity
qawli-ka saying-you.ms
(SA)
qawli-ka qawli-ka saying-you.MSG qawli-ka
(SA)
3
b.
laa yašukku ʔaħad-un fii neg doubt.3msg.ipfv one-nom in ‘No one is questioning what you said.’ [TP [NegP laa [VP yašukku ʔaħad-un fii
qawli-ka saying-you.MSG
(SA)
qawli-ka
Negation in SA exhibits microvariation that characterizes how the distribution of the NMs changed over time. The change is particularly salient when comparing the frequency of maa and lam in texts from classical Arabic and modern literary Arabic. The table 1.5 shows the distribution of the NMs in the Lebanese Arabic literary corpus (LBA fiction) and the Qur’anic corpus. The data is extracted from the Qur’an and MARC-2000 corpus by Mark Van Mol at KU-Leuven.7 The NM maa occurs only 6.6 percent of the time in the LBA corpus compared with the 23.4 percent in the Qur’anic corpus. Given the multi-locus analysis adopted in this book, the wider distribution of maa in the LBA corpus is a diachronic change that favors the lower NM lam over the higher marker maa. When we compare the distribution of maa and lam on the one hand, with the distribution of laa and its variants on the other hand, we find that the former has a limited distribution. This limited distribution contrasts sharply with the distribution of laa and its variants lam and lan, which, when combined, occur
Table 1.5. DISTRIBUTION OF MAA , L AA , L AM, L AN, AND L AYSA IN CL ASSICAL AND MODERN CORPUS
NM laa
LBA Fiction
1,853
59.20%
maa
132
6.60%
732
23.40%
lam
509
25.60%
346
11%
lan
67
3.30%
106
3.30%
163
8.30%
88
2.80%
laysa Total NMs Total corpus words
1,115
Qur’an 56%
1,986
3,125
86,032
85,933
7. Negative sentences were extracted from two corpora: (1) MARC-2000 (Modern Arabic Representative Corpus—2000) and (2) the Qur’an. The negative sentences cover the SA negative markers in a range of well-defined grammatical contexts. These contexts include the distribution of the negative markers in the context of verbal and nonverbal predicates, question words, and conditional particles, among the other contexts. The corpora included two subcorpora: (1) Literary Fiction from Lebanon (86,000) words, in MarK Van Mol corpus MARK-2000 compiled at KU-Leuven, and (2) the Qur’an, which is also 86,000 words.
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 33 ]
43
84.9 percent of the time in the LBA corpus, and 73.5 percent of the time in the Qur’anic corpus.
1.4. OUTLINE OF THE BOOK
This book is organized as follows. In c hapter 2, I lay out a classification of PSIs and their lexical categories. PSIs include NPIs, free-choice items (FCIs), PPIs, and NCIs. General indefinites display different distributions than do NPIs. This is interesting in Arabic because indefinite nouns like ħada and iši function as NPIs, and they are distinct from indefinite nouns (general indefinites) that occur in the context of negation. I explain this issue in c hapter 2, where I will discuss the distinctive features of NPIs and PPIs, such as scope widening. Two different types of PSIs interact with negation in interesting ways: NPIs and NCIs. The key difference between the two is that NPIs cannot function as fragment answers without negation and can occur in nonnegative contexts, such as interrogative and conditional contexts. NCIs display the opposite behavior. In chapter 2, I will also engage recent research I conducted on disjunction in JA and expand it to include the other Arabic varieties. This chapter describes the distribution of the disjunctive particles walla/willa/ʔam ‘or’ and the negative disjunctive particle wala ‘nor’ in polarity contexts and their status as structures for coordinate complexes in Arabic. In chapter 3, “Licensing Negative Polarity Items,” I investigate the role that syntax plays in licensing NPIs. Scholars have argued for a semantic approach to characterizing the unifying properties of the wide range of licensors, such as negation, disjunction, interrogation, and subjunctives, licensors that all share the semantic notion of nonveridicality (Giannakidou, 1994, 1995, 1997, 1998, 1999, 2009, 2017). I will extend this approach to Arabic NPIs in this chapter, but I will also argue that syntax is heavily involved in licensing these items. In particular, in this chapter I will discuss the role of the categorial status, syntactic configurations, and syntactic processes such as movement in licensing NPIs.). I extend this approach to Arabic NPIs in this chapter, but I also argue that syntax is heavily involved in licensing these items. In particular, I discuss the role of the categorial status, syntactic configurations, and syntactic processes such as movement in licensing NPIs. I argue for a unified analysis with a minimal set of basic syntactic operations such as Merge and Move, and licensing configurations of c-command and Spec-Head. The analysis captures the true nature of perceived mutual exclusivity between NPIs and the enclitic NM as an epiphenomenon of the availability of multiple loci for negation. In this chapter, I also discuss the syntax of the disjunctive particles walla/ willa/ʔam ‘or.’ In chapter 4, “Licensing Negative Concord Items,” I discuss the syntactic configurations and processes that are involved in licensing the various types [ 34 ] A Unified Theory of Polarity Sensitivity
53
of NCIs in Arabic, including the coordinate complexes laa-wala ‘neither-nor’ and their interaction with other PSIs. I argue for a syntactic agreement approach. In this way, I explain the behavior of Arabic NCIs with extra mechanisms (positing an abstract negative operator) to explain non-strict NCIs by conceptualizing a revised inventory of abstract negation features that the various types of Arabic NCIs carry, thus partially building on and departing from Zeijlstra (2004, 2008) and Penka’s (2011) analysis for Germanic languages. The coordinators laa-wala ‘neither-nor’ are conspicuous analogues of the NCI wala, which can combine with NPs to create person-NCIs such as wala-ħada ‘nobody,’ in their ability to co-occur with the negative adverb maʕumriš ‘never’ in the CP layers without triggering a double negation reading. A covert negative operator licenses the coordinators laa-wala as a last resort in the absence of overt licensors. In chapter 5 “PSIs with Head-Like Properties,” I discuss two main issues that arise from PSIs with head-like properties. The first issue is how these PSIs are licensed in syntax and how a unified analysis can handle their distribution, given that they seem to be outside the (immediate) domain of their licensor. Another issue that arises from these NPIs with head-like properties is their ability to host clitics with accusative and genitive case marking. This issue raises interesting questions pertaining to case theory and dependent case licensing. Finally, in c hapter 6, I summarize the central points from each chapter and show the analysis of PSIs and its contributions to the critical issues in syntax and linguistic theory. I also discuss issues that need future research.
I s s u e s i n t h e S y n ta x of P ol a r i t y - S e n s i t i v e I t e m s
[ 35 ]
63
73
CHAPTER 2
Classification of PSIs and Their Lexical Categories
T
he term polarity-sensitive items (PSIs) describes a category of words and expressions that interact with negation in various ways. This interaction takes place in at least three different ways. First, negation is (usually) required in simple declarative sentences containing a PSI. Second, certain PSIs have a negative interpretation when used as fragment answers (the fragment answer test), despite the absence of overt negation. Third, certain negative markers (NMs) are in complementary distribution with certain PSIs. In this chapter, I begin by defining the major PSI types along with their lexical categories. I then introduce the key contrasts between two major PSI categories and show their distribution relative to their licensors by pointing out distributional restrictions such as the mutual exclusivity between particular PSIs on the one hand and bipartite negation in Jordanian Arabic (JA) and the negative marker maa in Standard Arabic (SA) on the other hand. Finally, I present novel data from the understudied coordinate complexes in Arabic. I show that coordinate complexes display key characteristics of polarity items, and I show the challenges they present to PSI syntax. In c hapter 4, I put forward a novel analysis of disjunction construction in JA and its application to the phenomenon in the other Arabic varieties. To begin, I define PSIs as a category that encompasses negative polarity items (NPIs), free choice items (FCIs), positive polarity items (PPIs), and negative concord items (NCIs). I distinguish general indefinites from NPIs given their different distributions. This is interesting in Arabic because indefinite nouns such as ħada ‘one’ and ʔiši ‘thing’ can be construed as NPIs and are distinct from other indefinite nouns (general indefinites) in the context of negation. The book also distinguishes the indefinite nouns ħada and iši from A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0002.
83
other indefinite nouns based on the former’s inability to escape the scope of negation. In contrast, other indefinite nouns can. This sets the indefinite nouns ħada and iši apart from other indefinite nouns for being NPIs that must receive a negative polarity interpretation when they are under the scope of negation and a positive polarity interpretation when they are not. In addition, I discuss the distinctive features of NPIs and PPIs, such as scope widening. There are certain types of PSIs that interact with negation in interesting ways. The two main categories of PSIs discussed in this book are NPIs and NCIs. These two categories are distinctive in that (i) NPIs cannot be used as fragment answers without negation and (ii) they can occur in nonnegative contexts, such as in interrogative and conditional contexts. NCIs, on the other hand, display the opposite behavior. NPIs can belong to a range of different lexical categories. Under the category of NPIs, for example, we find determiner NPIs, nominal NPIs, adverbial NPIs, and idiomatic NPIs. Examples of determiner NPIs include the SA/JA ʔayy, which always combines with an indefinite, like šayʔ/ʔiši: (1)
a.
* (lam) yaštari neg.past buy.3msg.pfv ‘He didn’t buy anything.’
b.
* (ma)-štaraa-š neg-buy.3msg.pfv-neg ‘He didn’t buy anything.’
šayʔ/ thing/
ʔayya any
ʔiši / thing/
šayʔ thing
ʔayy ʔiši any thing
(SA)
(JA)
The indefinite nouns ʔaħad/ħada and šayʔ/ʔiši are nominal NPIs that occupy argument positions. The temporal adverbial NPI ʕumr in JA and EA (Egyptian Arabic) and the MA (Moroccan Arabic) equivalent ʕəmmər tend to occur sentence-initially. The NPI baʕd ‘yet’ in SA, however, is a sentence-final adverb: (2)
*(lam) yaʔti neg.past come.3msg.pfv ‘He hasn’t come yet.’
baʕd yet
(SA)
Idiomatic NPIs include expressions like girš ‘piaster [small unit of currency]’ and fils ʔiħmar ‘red cent.’ (3)
ma-maʕ-huu-š girš/ fils ʔiħmar neg-with-he-neg penny/ cent red ‘He doesn’t have a penny/red cent’
[ 38 ] A Unified Theory of Polarity Sensitivity
(JA)
93
NCIs consist of determiner NCIs and adverbial NCIs. Determiner NCIs include the JA wala, the EA wala, the MA hətta, and the SA wala cited in Lucas (2009, p. 198). (4)
a.
ma-ʔaʤa-š neg-came.3MSG.-neg ‘No one came.’
b.
maa-šuf-t-i-š walaa-waaħid neg-saw-1sg-ev-neg NCI-one ‘I didn’t see anyone.’ (Ouali & Soltan, 2014, p. 161)
c.
ma -šəf-t ħətta waħəd neg-saw-1sg NCI-one ‘I didn’t see anyone.’ (Ouali & Soltan, 2014, p. 161)
d.
Ɂinnaha baqaratun laa faaridˁun wala bikrun (SA) comp-it cow.NOM neg old.NOM and-neg virgin.NOM ‘It is a cow that is neither old nor immature.’ (Qur’an 2: 68)
wala-ħada NCI-one
(JA)
(EA)
(MA)
Adverbial NCIs include the SA NCIs ʔabadan and qatˁtˁ (Lucas, 2009), the JA NCIs ʔabadan/ bilmarrah (also in LA, Hoyt, 2010), and the EA NCI xaalisˁ (Soltan, 2014). (5)
a.
lan yadrusa neg.Fut study.3MSG ‘He will not study at all.’
ʔabadan NCI.at all
(SA)
b.
lam yadrus neg.past study.3MSG ‘He didn’t study at all.’
qatˁtˁ NCI.at all
(SA)
c.
ma-daras-iš neg-studied.3MSG-neg ‘He didn’t study at all.’
d.
ma-daras-ši neg-studied.3MSG-neg ‘He didn’t study at all.’
ʔabadan/ bilmarrah NCI.at all/NCI.at all
xaalisˁ. NCI.at all
(JA)
(EA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 39 ]
04
In chapters 3 and 4, I turn to a discussion of disjunction constructions. There I elaborate on how the affirmative (nonnegative) disjunctive particle walla/ willa/ ʔam ‘or’ and the negative disjunctive particle wala ‘nor’ interact with polarity contexts despite the fact that they are structures for coordinate complexes in Arabic. I then show that the role these coordinate complexes play as polarity items has significant ramifications for licensing NPIs and NCIs, discussed in chapters 3 and 4, respectively. 2.1. NPIS, PPIS, AND INDEFINITES
At least two different categories of words fall under the term PSI and differ from each other with respect to their sensitivity to negation. NPIs form the first category. These words always require negation in simple declarative sentences and may be in complementary distribution with certain NMs. Crucially, the indefinite nouns ħada ‘one’ and ʔiši ‘thing’ always receive a negative polarity interpretation when they are under the scope of negation or other antiveridical environments. This behavior contrasts with the behavior of general indefinite nouns like kitaab ‘book,’ which do not receive a negative polarity interpretation when they are in an antiveridical environment but, instead, receive a positive polarity interpretation. In the next section, I begin discussion of NPIs by distinguishing the indefinite nouns ħada ‘one’ and ʔiši ‘thing’ as NPIs and then general indefinite nouns such as kitaab ‘book’ as PPIs. 2.1.1. NPIs versus PPIs
The indefinite nouns ʔaħadun/ ħada and šayʔan/ ʔiši are nominal NPIs that occupy the object and subject positions. One question that arises is whether they are indefinites in the scope of negation, rather than NPIs. As such, it is necessary to distinguish these nominals from indefinite nouns. First, ħada and ʔiši can occur in affirmative declarative sentences, where they are interpreted as ‘someone’ and ‘something,’ respectively. In other words, these indefinites are PPIs when they occur in such sentences (see Baker, 1970; Szabolcsi, 2004; Giannakidou, 1998, 2009, for other languages). (6)
a.
šaaf ʔiši saw.3MSG thing ‘He saw something.’
[ 40 ] A Unified Theory of Polarity Sensitivity
(JA)
14
(7)
(8)
(9)
b.
šaaf ħada saw.3MSG one ‘He saw someone.’
a.
šəft ši saw.1SG some ‘I saw something.’
ħaʒa thing
(MA)
b.
šəft ši saw.1SG some ‘I saw someone.’
ħədd one
(MA)
a.
šaaf ħaaga saw.3MSG thing ‘He saw something.’
(EA)
b.
šaaf ħadd saw.3MSG one ‘He saw someone.’
(EA)
a.
šaaf šay saw.3MSG thing ‘He saw something.’
(QA)
b.
šaaf saw.3MSG ‘He saw someone.’
(10) a.
b.
(JA)
ʔaħad one
raʔa šayʔan saw.3MSG thing ‘He saw something.’
(SA)
raʔa ʔaħadan saw.3MSG one ‘He saw someone.’
(SA)
However, unlike the indefinite noun kitaab, these nouns receive an NPI interpretation when they occur in negative sentences. In other words, they are not referential, similar to the English NPIs ‘anyone, anybody’ and cannot have a
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 41 ]
24
PPI interpretation when they are in the context of negation.1 Therefore, they cannot be analyzed as indefinites in the scope of negation. (11) a.
b.
(12) a.
b.
(13) a.
b.
ma-šaaf-iš ʔiši/ ħada neg-saw.3MSG-neg NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’ #‘*He didn’t see something/someone.’
(JA)
ma-šaaf-iš ktaab neg-saw.3MSG-neg book ‘He didn’t see a book.’ ‘#He didn’t see any book.’
(JA)
ma-šaf-š ħaaga/ ħadd neg-saw.3MSG-neg NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’ #‘*He didn’t see something/someone.’
(EA)
ma-šaf-š kitaab neg-saw.3MSG-neg book ‘He didn’t see a book there.’ ‘#He didn’t see any book there.’
(EA)
hinaak there
maa šaaf šay/ʔaħad neg saw.3MSG NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’ #‘*He didn’t see something/someone.’ maa šaaf neg saw.3MSG ‘He didn’t see a book.’ ‘#He didn’t see any book.’
ktaab book
(QA)
(QA)
1. An anonymous reviewer asks about the status of the indefinite nouns ħada ‘one’ and ʔiši ‘thing’ and whether they are bare nouns or minimizers. These nouns are always NPIs when they are in the context of negation, but they can be interpreted as minimizers in other nonveridical contexts such as questions. This is, at least, the case in my JA dialect: (i)
šuf-t ʔiši/ ħada? saw-2msg NPI.thing/ NPI.one ‘Did you even see a thing/a person?
[ 42 ] A Unified Theory of Polarity Sensitivity
34
(14) a.
b.
(15) a.
b.
ma-šəf-š ħaʒa/ ħədd neg-saw.3msg-neg NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’ #‘*He didn’t see something/someone.’ ma-šaaf-š ktaab neg-saw.3msg-neg book ‘He didn’t see a book.’ ‘#He didn’t see any book.’
(MA)
lam yara šayʔan/ ʔaħadan neg.past see.3MSG NPI.thing/ NPI.one ‘He didn’t see anything/anyone.’ #‘*He didn’t see something/someone.’ lam yara neg.past see.3MSG ‘He didn’t see a book.’ ‘#He didn’t see any book.’
(MA)
kitaaban book
(SA)
(SA)
Furthermore, it is well known in the literature that PPIs tend to escape the scope of negation (Giannakidou, 2011). Consider the quantifier some, a classic PPI example in English. The JA equivalent baʕðˁ can escape the scope of negation. This differs from ħada and ʔiši, which cannot escape the scope of negation, earlier: (16) a.
b.
(17) a.
b.
ma-štaraa-š baʕðˁ neg-bought.3MSG-neg some ‘He did not buy some books.’ ma-štaraa-š neg-bought.3MSG-neg ‘He didn’t buy anything.’
ʔiši NPI.thing
ma-štaraa-š baʕdˁ neg-bought.3MSG-neg some ‘He did not buy some books.’ ma-štaraa-š neg-bought.3MSG-neg ‘He didn’t buy anything.’
ħaaga NPI.thing
l-kutub the-books
(JA)
(JA)
l-kutb the-books
(EA)
(EA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 43 ]
4
(18) a.
b.
(19) a.
b.
(20) a.
b.
maa štaraa baʕðˁ neg bought.3MSG some ‘He did not buy some books.’ maa štaraa neg bought.3MSG ‘He didn’t buy anything.’
l-kutub the-books
šay NPI.thing
ma-šraa-š baʕðˁ neg-bought.3msg-neg some ‘He did not buy some books.’
lam yaštari baʕðˁa neg.past buy.3MSG some ‘He did not buy some books.’ lam yaštari neg.past buy.3MSG ‘He didn’t buy anything.’
(QA)
l-ktub the-books
ħaʒa thing
ma-šra ši neg-bought.3msg some ‘He didn’t buy anything.’
(QA)
(MA)
(MA)
l-kutub (SA) the-books
šayʔan NPI.thing
(SA)
In examples (16)–(20), the PPI phrase baʕðˁ l-kutub ‘some books’ escapes the scope of negation. The result is an interpretation where a subset of books were bought and another subset of books were not. This reading is impossible with ʔiši and its counterparts in Arabic dialects as shown in (11)–(15), thus giving evidence that it is an NPI rather than a PPI. See Giannakidou (2011) for the similar case in English. Finally, it is well known in the literature that PPIs such as ‘someone’ can occur in the context of some NPI licensors like English’s ‘rarely’ (Ladusaw, 1979; Szabolcsi, 2004). This is true of the JA equivalent, naadir ‘rarely.’ If ħada and ʔiši are PPIs in affirmative sentences and NPIs in negative sentences, we expect them to have both the PPI and the NPI interpretation in those contexts that allow both. Indeed, this is borne out in (21b), where ʔiši can be interpreted as a PPI and as an NPI: (21) a.
naadir maa bi-štari rarely that asp-help.3MSG ‘It is rare for him to buy anything.’
[ 44 ] A Unified Theory of Polarity Sensitivity
ʔayy ʔiši any thing
(JA)
54
b.
naadir maa bi-štari rarely that asp-help.3MSG ‘It is rare for him to buy anything.’ ‘It is rare for him to buy something.’
ʔiši (NPI).thing
(JA)
Accordingly, the JA indefinite nouns ʔiši ‘thing’ and ħada ‘person’ and their variants in Arabic are NPIs distinct from general indefinite nouns. I will gloss these indefinites as NPI whenever they appear without the NPI determiner ʔayy and its variants in Arabic. 2.1.2. NPIs and their lexical categories
In this section, I discuss the distribution of determiner/nominal, adverbial, and idiomatic NPIs in Arabic. I discuss their distribution in negative and nonnegative contexts, including NMs, disjunctive coordinators, volitional verbs (want, suggest, insist), modal verbs and adverbials, and grammatical conditions such as questions, imperatives, habituals, and the subjunctive. I also distinguish the particular polarity item as an NPI through its ability to occur in nonnegative contexts and its inability to occur as a fragment answer, which is the opposite behavior of NCIs. Under the category of NPIs, we find determiner NPIs, nominal NPIs, adverbial NPIs, and idiomatic NPIs. Examples of determiner NPIs include the SA/ JA ʔayy ‘any,’ which always combines with an indefinite, such as šayʔin/ ʔiši in examples (22)a, b, or ʔaħadin/ ħada in examples (23)a, b (see also the EA ħaaga and ħadd and the Palestinian [PA] iši and ħada in example (24), cited in Lucas, 2009, p. 207). (22) a.
* (lam) yaštari šayʔan/ neg buy.3MSG NPI.thing/ ‘He didn’t buy anything.’
b.
* (ma)-štaraa-š ʔiši / neg-bought.3MSG-neg NPI.thing/ ‘He didn’t buy anything.’
(23) a.
* (lam) yaʔti neg.past come.3MSG ‘No one came’
ʔayya šayʔ any thing
ʔayy ʔiši any thing
ʔaħadun/ ʔayyu NPI.one/ any
(SA)
(JA)
ʔaħadin one
(SA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 45 ]
64
b.
* (ma)-ʔaʤa-š neg-came.3MSG-neg ‘No one came’
ʔayy ħada any one
ħada/ NPI.one/
(JA)
(24) ma šuft-iš ħaaga/iši (ħadd/ħada)2 neg saw.1sg-neg NPI.thing NPI.one ‘I didn’t see anything (anyone).’ (Woidich, 2006, p. 337)
(EA/PA)
All of these determiner NPIs can be used as FCIs that express the freedom of choice (Vendler, 1967) and occur in nonveridical environments that are not negative, such as imperatives: ʔišrab ʔay iši drink.imp any thing ‘Drink anything.’
(JA)
b.
ʔišrab ʔayya ħaaga drink.imp any thing ‘Drink anything.’
(EA)
c.
ʔišrab ʔay šay drink.imp any thing ‘Drink anything.’
(QA)
d.
šrəb ʔayy ħaʒa drink.imp any thing ‘Drink anything.’
(MA)
e.
ʔišrab ʔayya šayʔin drink.imp any thing ‘Drink anything.’
(SA)
(25) a.
FCIs may also occur with volitional verbs (want, suggest, insist), modal verbs, and adverbials. (26) a.
baddi ʔašrab want.I drink. ipfv.sbjv.1sg ‘I want to drink anything.’
ʔay iši any thing
(JA)
2. The brackets are added by the author to indicate that either ‘thing’ or ‘anyone’ can occur in this context. [ 46 ] A Unified Theory of Polarity Sensitivity
74
b.
ʕaayiz ʔašrab want.I drink. ipfv.sbjv.1sg ‘I want to drink anything.’
ʔayya ħaaga any thing
c.
ʔabi ʔašrab want.I drink. ipfv.sbjv.1sg ‘I want to drink anything.’
ʔay šay any thing
(QA)
d.
bɣit n-šrəb want.I drink.ipfv.sbjv.1SG ‘I want to drink anything.’
ʔayy ħaʒa any thing
(MA)
e.
ʔuriidu ʔan ʔašraba want.I comp drink. ipfv.sbjv.1sg ‘I want to drink anything.’
(EA)
ʔayya šayʔin any thing
(27) a.
yimkin ʔašrab might drink. ipfv.sbjv.1sg ‘I might drink anything.’
ʔay ʔiši any thing
(JA)
b.
yimkin ʔašrab might drink. ipfv.sbjv.1sg ‘I might drink anything.’
ʔayya ħaaga any thing
(EA)
c.
yimkin ašrab might drink. ipfv.sbjv.1sg ‘I might drink anything.’
d.
yəmkən/nqdər n-šrəb might drink. ipfv.sbjv.1SG ‘I might drink anything.’
e.
qad ʔašrabu might drink. ipfv.1sg ‘I might drink anything.’
ʔay šay any thing
(QA)
ʔayy ħaʒa any thing
ʔayya šayʔin any thing
(SA)
(MA)
(SA)
Turning to examples of adverbial NPIs, we find the NPI baʕd ‘yet’ in SA, a sentence-final adverb. We also find the NPI ʕumr in JA and EA, and the MA equivalent ʕəmmər, all of which tend to occur in the preverbal position:
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 47 ]
84
(28) a.
lam yadrus neg study.3MSG ‘He hasn’t studied yet.’
baʕd NPI-yet
(SA)
b.
ʕumr-ha maa katbat riwaaye NPI-ever-her neg write.3fsg.pfv novel ‘She has never written a novel.’
c.
ʕumr-ii maa-saafirt Masr (EA) NPI-ever-me neg-travel.pfv.1sg Egypt ‘I have never traveled to Egypt.’ (Soltan, 2012, p. 241)
d.
nadya ʕəmmər-ha ma-ʒat Nadia NPI-ever-her neg-come.pfv.3FSG ‘Nadia never came.’ (Benmamoun, 206, p. 144)
(JA)
(MA)
Idiomatic NPIs include minimizers such as girš ‘piaster’ and fils ʔiħmar ‘red cent’ in JA. Comparable forms of ‘red cent’ are found in EA, shown in (27)b, and other minimizers, like qiida ʔunmula ‘an inch’ from SA, are exemplified in (27)d, and walu from MA. (29) a.
fils ʔiħmar (JA) NPI-cent red
ma-maʕ-hu-š girš/ neg-with-him-neg NPI-penny/ ‘He doesn’t have a penny/red cent’
b.
ma-maʕ-hu-š malliim ʔaħmar neg-with-him-neg NPI-cent red ‘He doesn’t have a red cent’
c.
maa maʕ-o fils neg with-him NPI-cent ‘He doesn’t have a cent.’
d.
lam tataħarrak qiida neg.past move.3FSG distance ‘She did not move an inch.’
ʔunmula tip of finger
e.
ma-ʕandu walu neg-with-him NPI-thing ‘He doesn’t have anything’
(MA)
[ 48 ] A Unified Theory of Polarity Sensitivity
(EA)
(QA)
(SA)
94
In chapter 1, I noted that the fundamental distinction can be made between NPIs and NCIs, given their behavior in the fragment answer test and their behavior with respect to negation. Specifically, NPIs require an overt NM (or nonnegative licensor such as questions and conditionals) and cannot pass the fragment answer test. This is exemplified in the NPI from SA and the Arabic dialects in the following examples. As expected, the determiner NPIs, ʔaħadun ‘one’ and ʔayyu ʔaħadin ‘anyone’ in SA, and ħada ‘one’ and ʔayy ħada in JA, all require a negative maker—see examples (30)a, (31)a, (34)a, and (35)b—and fail the fragment answer test—see examples (30)b, (31)b, (34)b, and (35)b. (30) a.
b.
(31) a.
b.
(32) a.
b.
(33) a.
ʔaħadun/ ʔayyu one/ any
*(lam) yaʔti neg.past come.3msg.pfv ‘No one came’
Question: Answer: man ʤaaʔa? *(laa) who come.3msg.pfv neg ‘Who came?’ ‘No one’ *(ma)-ʔaʤaa-š neg-come.3msg.pfv-neg ‘No one came’
ħada/ one/
ʔaħad one
ʔayy ħada any one
Question: Answer: miin ʔaʤa? *(ma) ħada who come.3msg.pfv neg one ‘Who came?’ ‘No one’ ma -šuft-iš neg-saw.1sg-neg ‘I didn’t see anyone.’ Question: ʔinta šuf-t miin? you saw-2msg who Who did you see?’
ʔayy NPI.any
ʔaħadin one
waaħid one
(SA)
(SA)
(JA)
(JA)
(EA)
Answer: (Ouali & Soltan, 2014, p. 162) *ʔayy waaħid (EA) any one ‘*Anybody.’
ma-šəft ʔayy waħəd neg-saw.1sg any one ‘I did not see anyone.’
(MA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 49 ]
05
b.
Question: škun šəf-ti? who saw-2sg ‘Who did you see?’
Answer: (Ouali & Soltan, 2014, p. 162) *ʔayy waħəd (MA) any one ‘*Anybody.’
Similar to these determiner NPIs, the adverbial NPIs ʕumr ‘ever’ in JA/EA and ʕəmmər ‘ever’ in MA all require a negative maker—see examples (30)a, (31) a, (34)a, and (35)b—and cannot pass the fragment answer test—see (30)b, (31)b, (34)b, and (35)b. Another interesting property that distinguishes these NPIs in JA is that the NPI ʕumr is mutually exclusive with the enclitic NM -š (Alqassas, 2015, 108). This is also the case for EA in (35)a (Soltan, 2012, p. 241), and for MA in (35)c (Benmamoun, 2006, p. 144). (34) a.
b.
(35) a.
b.
(36) a.
b.
ʕumr-o *(ma) zaar(*-iš) NPI-ever-his neg visit.3msg.pfv-neg ‘He has never visited Petra.’ Question: ʕumr-ha katbat NPI-ever-her write.3msg.pfv ‘Has she ever written a novel?’
riwaaye? novel
ʕumr-ii maa-saafirt(*-iš) NPI-ever-me neg-travel. pfv.1sg-neg ‘I have never traveled to Egypt.’
el-batra (JA) def-Petra
Answer: (JA) *ʕumr-ha NPI-ever-her
Masr Egypt
(EA)
Question: Answer: ʔinta saafir-t Masr ʔabl kidah? *ʕumr-ii you travel.pfv-2msg Egypt before this ever-my ‘Have you traveled to Egypt before?’ ‘Never.’ nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
ma-ʒat(*-š) neg-came.3FSG(*-neg)
(MA)
Question: Answer: (MA) nadya ʕəmmər-ha ʒat ? *ʕəmmər-ha Nadia NPI.ever-her came.3FSG NPI.never-her ‘Has Nadia ever come.’ ‘Never.’
[ 50 ] A Unified Theory of Polarity Sensitivity
(EA)
15
Likewise, idiomatic NPIs in these Arabic varieties fail the fragment answer test. Though these NPIs can occur in interrogative contexts, as shown by the use of various idioms for ‘piaster, small unit of currency,’ ‘red cent,’ and ‘an inch’ in questions like (35), they are ungrammatical when used as fragment answers (36). maʕ-o girš/ fils ʔiħmar ? with-him NPI-penny/ NPI-cent red ‘Does he have a penny/red cent?’
(JA)
b.
huwwa maʕaa-h malliim ʔaħmar? he with-him NPI-cent red ‘Does he have a penny/red cent?’
(EA)
c.
maʕ-o fils ? with-him NPI-cent ‘Does he have a cent?’
d.
hal taħarrakat qiida Q moved.3FSG distance ‘Did she move an inch?’
(37) a.
(38) a.
Question: šuu maʕ-o? what with-him ‘What does he have?’
(QA)
ʔunmula? tip of finger
(SA)
Answer: *girš/ fils ʔiħmar (JA) NPI-penny/ NPI-cent red ‘a penny/red cent’ (= nothing)
b.
Question: maʕ-ah ʔeeh? with-him What ‘What does he have?’
Answer: *malliim (EA) NPI-cent ‘cent’ (= nothing)
c.
Question: Answer: šinu maʕ-o? *fils (QA) What with-him NPI-cent ‘What does he have?’ ‘cent’ (= nothing)
d.
Question: Answer: kam taħarrakat ? *qiida ʔunmula (SA) how much moved.3FSG distance tip of finger ‘How much did she move.’ ‘an inch’ (=not an inch)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 51 ]
25
Table 2.1 summarizes the variation of NPIs in Arabic. Table 2.1. INVENTORY OF NPIS
person thing
English
JA
EA
MA
QA
SA
anybody
(ʔay)ħada
(ʔay)ħadd
(ši) waħəd
(ʔay)
(ʔay)ʔaħad
ʔay waħəd
ʔaħad
anything
(ʔay)ʔiši
(ʔay)ħaaga (ši) ħaʒa
(ʔay) šay
(ʔay)šayʔ
ʔay ħaʒa time
ever
ʕumr
ʕumr
ʕəmmər
gad
qatˁ
determiner
any
ʔay
ʔay
ši ʔay*
ʔay
ʔay
girš/ fils
malliim
walu
fils
qiida
idiom
ʔiħmar
ʔunmula
*Oulai and Soltan (2014) point out that ʔay in MA is mostly limited to positive polarity, but they use it in MA data to illustrate judgments on its distribution as an NPI.
Other licensing environments for NPIs include contexts that contain without, before, few, and/or comparative adjectives, as the following data from Greek and Arabic shows: (39) a.
b.
(40) a.
O papus pethane xoris na idhi the grandpa died.3sg without subj see.3sg apo ta egonia tu from the grandch. his ‘Grandpa died without seeing any of his grandchildren.’ (Giannakidou, 1998, p. 57)
kanena any
O papus pethane prin na idhi kanena the grandpa died.3sg before subj see.3sg any apo ta egonia tu from the grandch. his ‘Grandpa died before seeing any of his grandchildren.’ (Giannakidou, 1998, p. 57) saafar biduun maa left.3msg.pfv-neg without comp ‘He left without telling anybody.’
[ 52 ] A Unified Theory of Polarity Sensitivity
yxabbir tell.3msg.ipfv
ʔayy ħada any one
(JA)
35
b.
ʕali mišii Ali left.3msg.pfv yitkallim maʕ talk.3msg.ipfv with ‘Ali left without talking to anyone.’ (Soltan, 2014b)
c.
saafara duuna ʔan yuxbira left.3msg.pfv-neg without comp tell.3msg.ipfv ‘He left without telling anybody.’
d.
saafər blaa left.3msg.pfv-neg without ygulhaa l-ħadd tell.3msg.ipfv to-anyone ‘He left without telling anybody.’
(41) a.
min ɣeer without ʔayy waaħid any one
maa comp
(EA)
ʔayya aħadin anyone
(SA)
maa comp
maat gabl maa died. 3msg.pfv before comp ʔay ħada min ʔaħfaad-o any one of grandchildren-his ‘He died before seeing any of his grandchildren.’
(MA)
yšuuf see.3msg.ipfv
(JA)
b.
ʔabuu-haa maat ʔabl maa yi-šuuf (EA) father-her died.3msg before comp see.3msg ʔayy waaħid min ʔaħfaad-uh any/no one from grandchildren-his ‘Her father died before seeing any of his grandchildren.’ (Soltan, 2014b)
c.
maata qabla ʔan yaraa died. 3msg.pfv before comp see.3msg.ipfv ʔayya ʔaħadin min ʔaħfaadi-h any one of grandchildren-his ‘He died before seeing any of his grandchildren.’
(SA)
d.
maat qbəl maa yšuuf died.3msg.pfv before comp see.3msg.ipfv ši waħəd/ ħadd mən ħfaayd-o any one /one of grandchildren-his ‘He died before seeing any of his grandchildren.’
(MA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 53 ]
45
Arabic NPIs are also licensed in the context of the comparative adjectives and the monotone decreasing quantifiers such as few. Consider the following data from Greek, JA, EA, and MA:3 (42) a.
Apodhixtike pjo esksipni apoti perimene {kanenas/*KANINAS} proved.3sg more intelligent than expected.3sg anybody ‘She turned out more intelligent than anyone had expected.’ (Giannakidou, 1998, p. 59)
b.
Ine to kalitero vivlio pu be.3sg the better book that {pote/*POTE} os fititria Ever as student ‘This is the best book I’ve ever read as a student.’ (Giannakidou, 1998, p. 59)
c.
LIJI fitites idhan few students saw. 3pl ‘Few students saw anything.’
exo have
dhiavasi read
{tipota/*TIPOTA}. anything
galiiliin illi bi-lɣarb biʕrifuu few.pl who in-the-west know ʔayy ʔiši ʕan l-ʔislaam any thing about the-Islam ‘Few people know anything about Islam.’
(JA)
b.
naas ʔulayyil-iin fii il- ɣarb bi-yi-ʕraf-uu people few-pl in the-West asp-ipfv-know-3pl ʔayy ħaagah ʕan il-ʔislaam any/no thing about the-Islam ‘Few people in the West know anything about Islam.’ (Soltan, 2014b)
(EA)
c.
qaliilun min-hum yaʕrifu few of-them know ʔayya šayʔin ʕan l-ʔislaam any thing about the-Islam ‘Few of them know anything about Islam.’
(SA)
(43) a.
3. The nonemphatic polarity items such as kanenas are nonnegative NPIs, while the emphatic is the NCI in Greek. [ 54 ] A Unified Theory of Polarity Sensitivity
5
qlaal lli kay-ʕərfuu ši ħaʒa few who asp-know some thing ʕla l-islaam f-lɣarb about the-islam in-the-west ‘Few people know anything about Islam.’
(MA)
ʔaħmad ʔaðˁʕaf min inn-o yguul Ahmad weaker than comp-him say ʔay ʔiši l-al-mudiir any thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’
(JA)
b.
ʔaħmad ʔadˁʕaf min inn-o yʔuul Ahmad weaker than comp-him say ʔayy ħaaga li-l-mudiir any thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’ (Soltan, 2014)
(EA)
c.
ʔaħmad dˁʕiif bezzaf baš (annah-o yqdar) Ahmad weaker a lot than comp-him can yguul ši ħaʒa l-l-mudiir say some thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’
(MA)
d.
(44) a.
The distribution of these NPI categories is made more complex when we consider licensing environments other than negation and interrogation. Disjunctive constructions are one such environment and will be discussed in section 2.3. Antiveridical environments (a subset of nonveridicality) include negation, like the antiadditives nobody, nothing, never, before and the comparatives and antimorphics not X, which is antiveridical in that NOT p entails that p is false (Giannakidou, 2009). Under this definition, antiveridicality conditions strict polarity items that negation can license. The strict polarity items relevant to the discussion here are NCIs. They require a licensor in their local syntactic domain (i.e., syntax mediates their licensing via an agreement relation with negation) and will be discussed in the next section. But the relevant point here is the distinction often made in the literature for NPIs that are weak and those that are strong. NPIs that are licensed by negation are traditionally classified as strong whereas NPIs that can be licensed in nonnegative contexts are weak NPIs (Giannakidou, 1998, 1999, 2009; Giannakidou & Zeijlstra, 2017). Examples of weak NPIs include the adverbial NPI ʕumr ‘ever’ and the nominal NPI ʔay ħada. Although the NPI ʕumr ‘ever’ C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 55 ]
65
is a weak NPI, it cannot be licensed in certain antiveridical environments, created by the superlative ʔawwal marra ‘first time,’ while maintaining a past tense interpretation. (45) a.
ʔawwal marra (b-ħayaat-i/ *ʕumr-i) first once (in-life-my/ ever) ‘the first time I ever saw them.’
b.
ʔawwil marra (f-ħayaat-i/ *ʕumr-i) first once (in-life-my/ ever) ‘the first time I ever saw them.’
c.
ʔawwal marra (f-ħyaat -i/ *ʕammar-ni) first once (in-life-my/ ever) ‘the first time I ever saw them.’
šuft-hum (JA) saw-them
šuftu-hum (EA) saw-them
šəft-hum saw-them
(MA)
However, we find that comparative adjectives are able to license this same NPI in the present tense: (46) a.
ʔawwal marra ʕumr-ha first once ever-her ‘the first time ever she sees them.’
bi-tšuuf-hum asp-see-them
b.
ʔawwal marra ħada bi-šuuf-hum first once NPI.one asp-see-them ‘the first time anyone sees them.’
c.
ʔawwal marra ʕammar-ha ka-t-šuuf-hum first once ever-her asp-see-them ‘the first time ever she sees them.’
d.
ʔawwal marra ši waħəd/ ši ħadd first once NPI.one/ NPI.one ‘the first time anyone sees them.’
(JA)
(JA)
(MA)
ka-yšuuf-hum asp-see-them
(MA)
In the next chapter, I elaborate on the distribution of Arabic NPIs in nonnegative contexts.
[ 56 ] A Unified Theory of Polarity Sensitivity
75
2.2. NCIS AND THEIR LEXICAL CATEGORIES
NCIs consist of determiner NCIs and adverbial NCIs. Determiner NCIs include the JA wala, the EA wala, and the MA hətta in (47). SA also has the negative conjunctive wala in example (47)d (see Lucas, 2009, p. 198): (47) a.
ma-ʔaʤa-š neg-came.3MSG.-neg ‘No one came.’
wala-ħada NCI-one
(JA)
b.
maa-šuf-t-i-š walaa-waaħid (EA) neg-saw-1sg-ev-neg NCI-one ‘I didn’t see anyone.’ (Ouali & Soltan, 2014, p. 161)
c.
ma -šəf-t ħətta waħəd (MA) neg-saw-1sg NCI-one ‘I didn’t see anyone.’ (Ouali & Soltan, 2014, p. 161)
d.
Ɂinnaha baqaratun laa faaridˁun wala bikrun (SA) comp-it cow.NOM neg old.NOM and-neg virgin.NOM ‘It is a cow that is neither old nor immature.’ (Qur’an 2: 68)
Adverbial NCIs include “never-type words” and “still-type words,” as defined by Hoyt’s (2010) analysis of NCIs in Levantine Arabic (LA). The never-type words include the SA NCIs ʔabadan and qatˁtˁ (Lucas, 2010), the JA NCIs ʔabadan/ bilmarrah (also in LA, Hoyt, 2010), and the EA NCI xaalisˁ (Soltan, 2014). (48) a.
ʔabadan NCI.at all
lan yadrusa neg.Fut study.3MSG ‘He will not study at all.’
qatˁtˁ NCI.at all
(SA)
b.
lam yadrus neg.past study.3MSG ‘He didn’t study at all.’
(SA)
c.
ma-daras-iš neg-studied.3MSG-neg ‘He didn’t study at all.’
ʔabadan/ bilmarrah NCI.at all/NCI.at all
(JA)
d.
ma-daras-ši neg-studied.3MSG-neg ‘He didn’t study at all.’
xaalisˁ NCI.at all
(EA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 57 ]
85
Adverbial NCIs also include still-type words such as baʕd ‘yet’ in SA, baʕd/ lissa ‘yet’ in LA (Hoyt, 2010; Alqassas, 2012, 2015) and EA (Soltan, 2012), and the NCI baqi ‘yet’ in MA (Benmamoun, 2006). (49) a.
lam yadrus neg.past study.3MSG ‘He hasn’t studied yet.’
baʕd yet
baʕd-o yet-him
(SA)
b.
ma-saafar-iš neg-traveled.3MSG-neg ‘He hasn’t traveled yet.’
(JA)
c.
Mona maa-saafir-it-š lissah (EA) Mona neg-traveled.3FSG-neg yet ‘Mona has not traveled yet.’ (Soltan, 2012, p. 243)
d.
nadya baq-at ma-ʒat (MA) Nadia yet-FS neg-came.3FSG ‘Nadia hasn’t come yet.’ (Benmamoun, 2006, p. 144)
Unlike NPIs, NCIs can pass the fragment answer test. The examples following this paragraph show that for the NCIs wala-ħada ‘no one’ in JA, walaa waaħid in EA, and ħətta wəld in MA. In addition, these NCIs receive a negative interpretation when used as fragment answers, even in the absence of negation. The JA NCI wala-ħada ‘no one,’ the EA walaa waaħid, and the MA ħətta wəld all require an NM when they occur postverbally. This is not the case when these NCIs appear preverbally. While the NCI in MA must co-occur with the NM in preverbal position, the NCIs in JA and EA cannot co-occur with the NM without yielding a double negation reading (i.e., a concordant negation reading is impossible). (50) a.
b.
*(ma)-ʔaʤa-š neg-came.3MSG-neg ‘No one came.’
wala-ħada NCI-one
wala-ħada ma-ʔaʤa-š NCI-one neg-came.3MSG-neg ‘Nobody didn’t come.’ #‘Nobody came.’
[ 58 ] A Unified Theory of Polarity Sensitivity
(JA)
(JA)
95
c.
(51) a.
Question: min who ‘Who came?’ walaa waaħid no one ‘Nobody came.’
ʔaʒa? came.3MSG
Answer: wala-ħada NCI-one ‘no one’
(JA)
gih (EA) came.3msg (Ouali & Soltan, 2014, p. 164)
b.
walaa waaħid maa-gaa-š no one neg-came.3msg-neg ‘Nobody didn’t come.’ #‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
c.
Question: ʔinta šuf-t miin? You saw-2msg who ‘Who did you see?’ (Ouali & Soltan, 2014, p. 162)
Answer: walaa waaħid no one ‘nobody.’
(EA)
(EA)
Another interesting property that distinguishes the MA NCIs has to do with complementary distribution with NMs. We observe that the enclitic NM -š cannot co-occur with the MA NCI. (52) a.
*(ma)-ʒa neg-came.3MSG *‘Any boy didn’t come’
(53) a.
ħətta waħəd *(ma)-ʒa (MA) even one neg-came.3MSG ‘Anyone didn’t come.’ (Benmamoun, 1997, p. 273)
b.
c.
Question: škun šəf-ti? who saw-2sg ‘Who did you see?’ *ma-ʒa-š neg-came.3MSG-neg
ħətta wəld (MA) NCI boy (Benmamoun, 1996, p. 49)
Answer: ħətta not-even ‘Nobody.’ ħətta wəld even boy
waħəd (MA) one (Ouali & Soltan, 2014, 162) (MA) (Benmamoun, 1996, p. 49)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 59 ]
06
Closely related to PSIs is the category of negative indefinites (NIs), which are inherently negative words that express negated existential quantification. These NIs are roughly the equivalent of the Standard English no-indefinite noun compounds, such as no one, nothing, etc. In JA, they carry a negative interpretation in the absence of an NM; when an NM does co-occur with a negative quantifier, the result is double negation. In MA, there is no equivalent item, and only the NCI and NM maa are used instead. (54) a.
b.
maħadaaš rasab nobody failed.3MSG ‘No one failed the exam.’
bi-li-mtiħaan in-the-exam
maħadaaš ma-rasab-iš bi-li-mtiħaan (JA) nobody neg-failed.3MSG-neg in-the-exam ‘No one didn’t fail the exam.’ (= Everybody failed the exam) ħtta-waħəd ma-squt NCI-person neg-failed.3msg ‘No one failed the exam.’
c.
d.
(JA)
f-li-mtiħaan in-the-exam
(MA)
ma-kayn-š ši waħəd ma-squt-š f-li-mtiħaan neg-cop-neg NPI-person neg-failed.3msg-neg in-the-exam ‘No one didn’t fail the exam.’ (= Everybody failed the exam)
(MA)
Negative quantifiers will be related to the analysis of the syntactic dependency (or lack thereof) between NCIs and negation. Table 2.2 shows the variation in the NIs in Arabic.
Table 2.2. INVENTORY OF NIS English person nobody
JA
EA
MA
QA
SA
walaħada
walawaaħid
ħəttawaħəd
maħħad
laaʔaħada
maħadaaš
maħaddiš
thing
nothing
wala(ʔi)ši
walaħaaga
ħəttaħaʒa
mašay
laašayʔa
time
never
walaʕumr
walaʕumr
maʕəmmər
yet
ʔabad
ʔabadan
maʕumrhiš
maʕumriš
baqi
ʔabadan
ʔabadan
baʕd
lissah
wala
wala
det
no
[ 60 ] A Unified Theory of Polarity Sensitivity
ħətta
baʕd
ma
laa
16
The licensing environments for NCIs include clauses that contain without and before. Consider the following examples from Greek,4 JA, EA, and MA: (55) a.
b.
(56) a.
O papus pethane xoris na idhi {KANENA/kanena} the grandpa died.3sg without subj see.3sg any apo ta egonia tu from the grandch. his ‘Grandpa died without seeing any of his grandchildren.’ (Giannakidou, 1998, p. 57) O papus pethane prin na idhi {KANENA/kanena} the grandpa died.3sg before subj see.3sg any apo ta egonia tu from the grandch. his ‘Grandpa died before seeing any of his grandchildren.’ (Giannakidou, 1998, p. 57) saafar biduun maa left.3msg.pfv-neg without comp ‘He left without telling anybody.’
yxabbir tell.3msg.ipfv
b.
ʕali mišii min ɣeer maa Ali left.3msg.pfv without comp yitkallim maʕ wala waaħid talk.3msg.ipfv with any one ‘Ali left without talking to anyone.’ (Soltan, 2014b)
c.
saafər blaa left.3msg.pfv-neg without ygulhaa ħtta l-ši tell.3msg.ipfv NCI to-any ‘He left without telling anybody.’
(57) a.
maa comp waħəd person
wala ħada (JA) any one
(EA)
(MA)
maat gabl maa yšuuf (JA) died. 3msg.pfv before comp see.3msg.ipfv wala ħada min ʔaħfaad-o any one of grandchildren-his ‘He died before seeing any of his grandchildren.’
4. The emphatic KANENA is an NCI in Greek.
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 61 ]
26
b.
ʔabuu-haa maat ʔabl maa yi-šuuf (EA) father-her died.3msg before comp see.3msg wala waaħid min ʔaħfaad-uh any /no one from grandchildren-his ‘Her father died before seeing any of his grandchildren.’ (Soltan, 2014b)
c.
maat qbəl maa yšuuf (MA) died.3msg.pfv before comp see.3msg.ipfv ħtta ši waħəd mən ħfaayd-o NCI any person of grandchildren-his ‘He died before seeing any of his grandchildren.’
Unlike NPIs, however, NCIs cannot occur in the context of few and/or comparative adjectives, as the Greek and Arabic data shows: (58) a.
Apodhixtike pjo esksipni apoti perimene {kanenas/*KANINAS} proved.3sg more intelligent than expected.3sg anybody ‘She turned out more intelligent than anyone had expected.’ (Giannakidou, 1998, p. 59)
b.
Ine to kalitero vivlio pu exo dhiavasi {pote/*POTE} os fititria be.3sg the better book that have read Ever as student ‘This is the best book I’ve ever read as a student.’ (Giannakidou, 1998, p. 59)
c.
LIJI fitites idhan few students saw.3pl ‘Few students saw anything.’
(59) a.
b.
galiiliin illi bi-lɣarb few.pl who in-the-west ʔayy/*wala iši ʕan any /no thing about ‘Few people know anything about Islam.’
{tipota/*TIPOTA}. anything
biʕrifuu know l-ʔislaam the-Islam
(JA)
naas ʔulayyil-iin fii il- ɣarb bi-yi-ʕraf-uu (EA) people few-pl in the-West asp-ipfv-know-3pl ʔayy/*wala ħaagah ʕan il-ʔislaam any/no thing about the-Islam ‘Few people in the West know anything about Islam.’ (Soltan, 2014b)
[ 62 ] A Unified Theory of Polarity Sensitivity
36
c.
(60) a.
qlaal f-lɣarb lli kay-ʕərfuu few in-the-west who asp-know ši ħaʒa /*ħtta ħaʒa ʕla l-islaam some thing /NCI thing about the-Islam ‘Few people know anything about Islam.’
(MA)
ʔaħmad ʔaðˁʕaf min inn-o yguul Ahmad weaker than comp-him say ʔay/*wala ʔiši l-a l-m udiir any/No thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’
(JA)
b.
ʔaħmad ʔadˁʕaf min inn-o yʔuul (EA) Ahmad weaker than comp-him say ʔayy/*wala ħaaga li-l-mudiir any/no thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’ (Soltan, 2014)
c.
ʔaħmad dˁʕiif bezzaf baš (annah-o yqdar) yguul Ahmad weaker a lot than comp-him can say ši ħaʒa/*ħtta ħaʒa l-l-mudiir some thing/NCI thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’
(MA)
The contrasts between the NPIs and NCIs in these licensing contexts reveal that NPIs have a wider distribution than NCIs. In chapter 3, I will build on Soltan’s (2014b) analysis of EA and extend it to JA and SA. There, I will show that Giannakidou’s theory of nonveridicality (1997, 1998, 2009) fares better than Ladusaw’s (1979) monotonicity approach (= Downward Entailment) in explaining these contrasts. The focus in c hapter 3, however, is on analyzing the syntactic configurations that feed the semantic component a structure where the NPI is under the scope of their licensors. As for NCI licensing, it is clear that the nonveridicality approach cannot explain cases of ungrammaticality even when the NCI is under the scope of negation. The NCI walaħada fails to occur in embedded clauses despite having negation scoping over it in the main clause: (61) a.
ma-ðˁunni-š ʔinn-o zaar-o neg-think.1sg-neg comp visited.3MSG-him ‘I don’t think anyone visited him.’
ʔay ħada any one
(JA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 63 ]
46
b.
*ma-ðˁunni-š neg-think.1sg-neg
ʔinn-o comp
c.
ʔaħmad maa-ʔaal-š ʔin Mona (EA) Ahmad neg-said. 3msg -neg comp Mona fihm-it ʔayy/*walaa aagah understood-3fsg any/no thing ‘Ahmad didn’t say that Mona understood anything.’ (Soltan, 2014)
d.
ma-ka-ndˁun-š ʔannaho zaar-o ši waħəd (MA) neg-think.1sg-neg comp visited.3msg-him some person ‘I don’t think anyone visited him.’
e.
*ma-ka-ndˁun-š ʔanno zaar-o ħtta waħəd neg-think.1sg-neg comp visited.3msg -him NCI person ‘I don’t think anyone visited him.’
zaar-o wala ħada visited.3MSG -him NCI- one
(JA)
(MA)
The NCI wala also fails to occur in embedded clauses where a perfective preventing predicate appears in the main clause. In contrast, the same NCI can occur in the embedded clause of a participle-preventing predicate: (62) a.
b.
(63) a.
b.
mamnuuʕ tištari ʔay/ wala part.prevented buy.2sg NPI/nci ‘It is disallowed to buy anything.’ manaʕ-ha tištari ʔay/ *wala prevented.3MSG-her buy.2sg npi/ nci ‘He prevented her from buying anything.’
ʔiši thing
ʔiši thing
(JA)
(JA)
mamnuuʕ tašri ši ħaʒa /ħtta ši ħaʒa PART.prevented buy.2sg some thing/NCI some thing ‘It is disallowed to buy anything.’
(MA)
mənaʕ-ha təšri ʔay ħaʒa /*ħtta ši ħaʒa prevented.3msg-her buy.2sg NPI.thing/NCI some thing ‘He prevented her from buying anything.’
(MA)
The failure of the NCI to occur in these contexts suggests that the semantic notion of antiveridicality cannot be responsible for regulating the distribution
[ 64 ] A Unified Theory of Polarity Sensitivity
56
of NCIs.5 I take this up in chapter 4 and, instead, argue for a syntactic agreement account to license NCIs. There, I argue for a syntactic agreement account that produces a modified and detailed picture for the feature structure of the functional category of negation, NMs, and polarity items. To summarize, this chapter has demonstrated that NPIs and NCIs can be subdivided into a range of lexical categories: determiner, (pro)nominal, adverbial, and idiomatic expressions. In addition, we saw that NPIs and NCIs are distinctive given their distributions. NPIs can occur in nonnegative contexts such as interrogative and conditional sentences but cannot occur as fragment answers. NCIs display the opposite behavior. Despite the division of their distribution, I have also shown that certain PSIs, including specific types of NPIs and NCIs, cannot co-occur with the enclitic NM -š.
2.3. DISJUNCTION PARTICLES IN COORDINATE COMPLEXES
Under the category of coordinate complexes, we find PSI coordinators that make up coordinate complexes, which can disjoin phrases and clauses like DPs or CPs. These disjunction coordinators are ʔaw ‘free choice or’ and walla ‘exclusive choice or’ and compare to the determiner NPI ʔay ‘any’ in having a free choice or polarity interpretation. The similarity between the two groups is clear in cases where walla ‘exclusive choice or’ has a distribution limited to a subset of polarity environments: questions and conditionals. (64) a.
ʔišrab gahwa drink.imp coffee ‘Drink coffee or tea.’
ʔaw šaay or tea
(JA)
5. In EA, the NCI fails to appear not only with the perfective preventing predicate, but also with the participle-preventing predicate. This failure clearly adds to the mounting evidence that antiveridicality cannot explain all co-occurrence/licensing restrictions: (i)
a.
mamnuuʕ tištiri ʔayy/ *wala part.prevented buy.2sg NPI/ NCI ‘It is disallowed to buy anything.’
b.
manaʕ-ha tištiri ʔay/ *wala prevented.3MSG-her buy.2sg NPI/ NCI ‘He prevented her from buying anything.’
ħaagah thing
ħaagah thing
(EA)
(EA)
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 65 ]
6
b.
tišrab gahwa walla drink.sbjv coffee or ‘Would you like to drink coffee or tea?’
šaay? tea
(JA)
c.
ʔšrab qəhwa drink.imp coffee ‘Drink coffee or tea.’
ʔataay tea
(MA)
d.
t-šrab qəhwa wəlla ʔataay? drink.sbjv coffee or tea ‘Would you like to drink coffee or tea?’
(MA)
(65) a.
b.
wəlla or
miš mitʔakkid ʔiða bitħib gahwa neg sure whether like.3FSG. coffee ‘I am not sure whether she likes coffee or tea?’
walla or
šaay? tea
ma-ʕraft-š waaš ka-təbɣi neg-know.I-neg whether asp-like.3fsg. lqahwa wəlla ʔataay? coffee or tea ‘I am not sure whether she likes coffee or tea?’
(JA)
(MA)
However, the distribution of the disjunction coordinator walla ‘exclusive choice or’ is more limited than the determiner NPI ʔay ‘any.’ Negation can license walla but not the determiner NPI. The MA wəlla, however, is not limited and can occur in the context of negation. This behavior might be because this coordinator is synonymous with ʔaw, which is not limited in JA or MA: (66) a.
ma-štarati-š ʔay neg -bought.3fsg-neg NPI-any ‘She did not buy any coffee.’
b.
ma-štarati-š gahwa neg -bought.3fsg-neg coffee ‘She did not buy coffee or tea.’
c.
ma-šrat ħtta neg -bought.3fsg any ‘She did not buy any coffee.’
[ 66 ] A Unified Theory of Polarity Sensitivity
gahwa coffee
(JA)
ʔaw/*walla šaay or tea
(JA)
qahwa coffee
(MA)
76
d.
ma-šrat gahwa neg -bought.3fsg coffee ‘She did not buy coffee or tea.’
ʔaw/ wəlla ʔataay or tea
(MA)
These distributional facts have consequences for the correlational principles in Giannakidou-Zwarts’s (1999) theory of polarity items (see the Venn diagram in figure 2.1). This theory shows that NPIs that are licensed in nonnegative contexts are also licensed in negative contexts but not the other way around. By extension, we expect the coordinator walla ‘exclusive choice or’ to pattern with polarity items. However, we find that it patterns with FCIs, such as any, which can occur in nonveridical environments. Arabic also has the negative disjunction coordinators laa-wala ‘neither-nor.’ These coordinators can coordinate two phrases or two clauses. The coordinator laa precedes the first element of a coordination construction; see, for example, the higher phrase in (67) and the higher clause in (68). The coordinator wala precedes the second element of coordination; for example, the lower phrase in (67) and the lower clause in (68). (67) a.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
šaafuu-na saw.3mp-us
(JA)
b.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
šaafuu-na saw.3mp-us
(MA)
Nonveridical Downward entailing (minimally negative) Antiadditive (classically negative)
Antimorphic
Figure 2.1. Giannakidou-Zwarts’s (1999) theory of polarity items.
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 67 ]
86
(68) a.
laa šufna Zeed wala šufna neither saw.1p Zeed nor saw.1p ‘We neither saw Zeed, nor did we see Omar.’
ʕumar Omar
b.
maa šufna Zeed wala šufna neither saw.1p Zeed nor saw.1p ‘We neither saw Zeed, nor did we see Omar.’
ʕumar Omar
(JA)
(MA)
The negative coordination marker wala also displays the characteristics of NCIs. (69) a.
*(maa) šuft neg saw.1sg ‘I didn’t see anyone.’
wala NCI
b.
*(maa) šuft laa Zeed neg saw.1sg neither Zeed ‘I neither saw Zeed nor Omar.’
c.
*(maa) šuft neg saw.1sg ‘I didn’t see anyone.’
d.
*(maa) šuft laa Zeed neg saw.1sg neither Zeed ‘I neither saw Zeed nor Omar.’
(70) a.
b.
wala waaħad NCI one ‘No one saw us.’
waaħad one
wala nor
ħtta-waħəd NCI-person
ʕumar Omar
(JA)
ʕumar Omar
(MA)
(MA)
wala nor
šaaf-na saw.3msg-us
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
(JA)
(JA)
šaaf-na saw.3msg-us
(JA)
In interrogative contexts, walla/ willa and its variant in MA wəlla have the affirmative coordination interpretation ‘or’ and behave more like disjunction operators.
[ 68 ] A Unified Theory of Polarity Sensitivity
96
(71) Affirmative disjunction a. šuft Zeed walla/willa (šuft) saw.2MSG Zeed or saw.2MSG ‘Did you see Zeed or (did you see) Omar?’ b.
ʕumar? Omar
(waš) šufti Zeed wəlla (šufti) ʕumar? Q saw.2MSG Zeed or saw.2MSG Omar ‘Did you see Zeed or (did you see) Omar?’
(JA)
(MA)
The presupposition in this example is that the addressee saw either Zeed or Omar but not both or neither of them. In c hapter 4, I extend my NCI analysis to these coordinate complexes and show that coordinate complexes give further support to it. To conclude, this chapter showed the complex distributional patterns PSIs exhibit. The distributional patterns get more complex when we consider the type of licensors each category of PSIs has, and the word order or co-occurrence restrictions each PSIs of each category has in relation to NMs, predicates, and PSIs of other categories. Microvariation in the behavior of PSIs in Arabic is paradoxical in nature. This microvariation adds to the distributional complexity and paves the way to the analyses I put forward in the subsequent chapters of this book. In the next chapter, I argue for a unified analysis with a small set of basic syntactic processes to explain the complex distribution of Arabic NPIs. I show that the only syntactic processes involved in licensing NPIs are Move and Merge, and the only licensing configurations needed are c-command and the specifier-head configurations. The variation lies in the loci of the NPIs, the multi-locus nature of negation, and the PSI’s ability to undergo XP movement or focus-fronting. The analysis can capture the true nature of perceived mutual exclusivity between NPIs and the enclitic marker of bipartite negation. I show that these cases are an epiphenomenon of the availability of multiple loci for negation.
C l a s s i f i c at i o n of P SI s a n d T h e i r L e x i c a l C at e g or i e s
[ 69 ]
07
17
CHAPTER 3
Licensing Negative Polarity Items
3.1. NPI LICENSING IN THE SYNTACTIC STRUCTURE
This chapter investigates the role of syntax in licensing NPIs (negative polarity items). Previous literature on NPI licensing argues for a semantic approach in order to characterize the unifying properties of the wide range of licensors. These licensors include negation, disjunction, volitional verbs (want, suggest, insist), modal verbs, adverbials, questions, imperatives, habituals, and the subjective (Giannakidou, 2009) and are unified under the semantic notion of nonveridicality (Giannakidou, 1998, 1999, 2009; Giannakidou & Zeijlstra, 2017). In this chapter, I argue that, in addition to semantics, syntax is involved in the licensing of NPIs in Arabic (see Collins & Postal, 2014, 2017, for the role of syntax in licensing NPIs). Particularly, the chapter will discuss the role of the categorical status, syntactic configurations, and syntactic operations, such as Merge and Move, in the licensing of NPIs. Antiveridical environments are a subset of nonveridicality. One such environment is negation, including elements such as antiadditives nobody, nothing, never, before, comparatives, and antimorphics (not X), which are antiveridical in that NOT p entails that p is false (Giannakidou, 2009). In English, strict NPIs, which are licensed by negative licensors, are conditioned by antiveridicality and require a licensor in their local syntactic domain, i.e., their licensing is mediated by syntax via an agreement relation with negation. Among the challenges to a semantic approach to NPI licensing is capturing the variation in the category of licensors that can license a particular class of NPI. For example, Arabic NPIs do not conform to the same classification that English NPIs have. The NPI ʕumr ‘ever’ is a weak NPI that occurs in nonnegative contexts like interrogatives and conditionals, and we would expect it to occur in other nonnegative contexts such as the comparative/superlative adjective. A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0003.
27
However, the NPI ʕumr ‘ever’ is ungrammatical in the following context despite the fact that a comparative adjective precedes it: (1)
ʔawwal marra (b-ħahyaat-i/ *ʕumr-i) first once (in-life-my/ ever) ‘the first time I ever saw them’
šuft-hum saw-them
(JA)
Nonetheless, the primary question in regard to Arabic NPIs is whether they must be c-commanded by negation and whether this is a requirement in surface structure (overt syntax) or in the interpretation site (covert syntax). In chapter 2, we saw that different types of NPIs occupy different positions with respect to negation. Specifically, determiner and nominal NPIs usually must follow negation, while adverbial and idiomatic NPIs can follow or precede negation. (2)
a.
*(ma)-ʔaʤaa-š neg-came.3MSG-neg ‘No one came’
b.
ʕumr-ha maa katbat NPI-ever-her neg wrote.3FSG ‘She has never written a novel.’
riwaaye novel
(JA)
c.
ma-maʕ-huu-š girš/ fils ʔiħmar neg-with-him-neg NPI-penny/ NPI-cent red ‘He doesn’t have a penny/red cent’
(JA)
ħada/ NPI-one/
ʔayy ħada any one
(JA)
This chapter discusses how these various NPIs establish dependency with negation. I argue that NPIs are licensed only through the specifier-head (Spec- Head) and c-command configurations, not the head–complement relation, contra Benmamoun (2006). This allows for direct mapping from syntax to semantics whereby NPIs are interpreted as existential quantifiers under the scope of negation. The variation in the behavior of NPIs is thus reduced to their syntactic properties, including their syntactic location in relation to negation, which can appear higher or lower than TP (tense phrase), and whether they undergo movement and reconstruction at LF (Logical Form). In previous literature, analyses distinguished between NPIs and negative quantifiers (NQs) (e.g., Progovac, 1993; Haegeman, 1995). Within this dichotomy, NCIs (negative concord items) are generally categorized as a subset of either NPIs or NQs (Hornstein, 1995; Brown, 1999; Zeijlstra, 2004). In this
[ 72 ] A Unified Theory of Polarity Sensitivity
37
chapter, I briefly elaborate on the difference between NPIs and NQs and review previous analyses of NPIs and NCIs, respectively. The NPI anyone and the NQ no one are considered classic Standard English PSIs (polarity-sensitive items) (Brown, 1999; Zeijlstra, 2004; Haspelmath, 2005). In negative polarity environments, the use of anyone requires the English negative marker not/n’t or an NQ like no one, as in (15): (3)
a.
I haven’t seen anyone.
b.
No one has seen anyone.
The NQ no one, by contrast, does not require negation. In fact, when negation co-occurs with an NQ in Standard English, the result is double negation: (4)
a.
I haven’t seen no one (= I have seen someone)
b.
No one has seen nothing (= everyone has seen something)
NPIs can occur in a number of environments, including clause-mate negation, superordinate negation, yes/no questions, conditionals, and adversative (semantically negative) predicates like forget and doubt. As such, the English NPs anyone, anything, and so on are typical examples of NPIs.1 Discussion of NPIs in the literature has centered on identifying ways to capture the dependency between NPIs and negation, taking into account that some NPIs follow negation and others precede it. Syntactic accounts capture this dependency through specific structural configurations, such as c- command and the Spec–Head relation, although accounts take differing views regarding the level of representation at which licensing takes place—i.e., in an overt syntactic structure or covertly in LF. For instance, when an NPI follows 1. The English NPIs anything and anyone can be licensed in negative and nonnegative polarity environments, including questions: (i)
I can’t see anything/anybody. Can you see anything/anybody?
However, it is worth pointing out that anything and anyone can also appear in nonpolarity environments, as in: (ii)
Children believe anything adults say. Children like anyone who is nice to them.
These instances of any are not NPIs but free choice any (FC any) (cf. Dayal, 2004).
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 73 ]
47
negation, the configuration is interpreted as a dependency between the NPI and the negation marker, mediated through a c-command relationship (Lasnik, 1975; Jackendoff, 1969, 1972; Linebarger, 1981, 1987; Giannakidou, 1998, 2006, 2011; Benmamoun, 1997, 2006). C-command is defined here as in Reinhart (1976, p. 32): (5)
Node A c-commands node B if neither A nor B dominates the other and the first branching node dominating A dominates B.
(6)
I did not see anybody.
(7)
Overt licensing under C-command TP Spec
T T
NegP Neg
VP V
not
C-command
NP NPI (anybody)
In (7), the representation for (6), the negative marker not is the head of NegP and thus c-commands the NPI object anybody, which is the internal complement of the VP (verb phrase). Neither the Neg head node nor the NP node dominates the other, and the first node dominating Neg, namely ‘Neg,’ also dominates the NP. Previous analyses of NPIs claim that they are licensed overtly, i.e. at surface structure (Jackendoff, 1969, 1972; Lasnik, 1975). Specifically, Lasnik proposes that negation scope over NPIs at surface structure in order for the NPI to be specified [+neg] and [-referential]. This is referred to as the ‘Not-Scope Rule.’ The scope of negation here is understood to encompass any element to the right of, and c-commanded by, negation. This assumption is based on the following contrasts: (8)
a.
* Anybody didn’t see me.
b.
I didn’t see anybody.
In (8)a, the NPI composed of the quantifier any and the indefinite noun body fails to receive a non-referential reading because it is not under the scope of negation at surface structure. The ability of negation to give a [-referential] [ 74 ] A Unified Theory of Polarity Sensitivity
57
marking to quantifiers can be seen in examples where the quantified NP (noun phrase) is under the scope of negation and cannot be referential: (9)
a.
Many students (namely, James, Robert, and Sarah) attended the class.
b.
* Not many students (namely, James, Robert, and Sarah) attended the class.
Evidence from passivization and topicalization suggest that the requirement for negation to precede and c-command the NPI is a surface-level constraint, rather than a deep-structure constraint (Linebarger, 1981, 1987): (10) a.
I never saw anyone.
b.
* Anyone was never seen by me.
c.
* Anyone I never saw.
Recent analyses of NPIs in Hindi (Benmamoun & Kumar, 2006) have maintained the surface structure approach by arguing that Hindi NPIs are licensed overtly before scrambling. Evidence for this claim comes from reconstruction effects that argue against covert licensing. (11) a.
b.
Raajiiv-ne kahaa ki [vo ek bhii pikcar jo (Hindi) Rajive-ERG said that that one EMPH pictures which Sariitaai-ne lii]j uskoi pasand nahiiN hai tj Sarit-ERG took she likes neg PRS ‘Rajiv said that she does not like any picture that Sarita took.’ *Raajiiv-ne Rajive-ERG [vo that
kahaa said ek one
ki that bhii EMPH
uskoi she pikcar pictures
pasand likes jo which
nahiiN neg Sariitaai-ne Sarit-ERG
hai (Hindi) PRS lii] took
Assuming the copy theory of movement, Benmamoun and Kumar (2006) show that LF reconstruction of the constituent vo ek bhii pikcar jo Sariitaai-ne lii to a position c-commanded by negation results in a violation of Binding Principle C. Accordingly, licensing of the NPI via LF reconstruction is rejected. Instead, the authors argue that the NPI is licensed prior to the scrambling of the NPI object from its VP-internal position to pre-subject position. Nonetheless, it is insufficient to solely posit that the NPI is preceded and c- commanded by negation when explaining the behavior of NPIs in other languages. NPIs in Greek require a syntactic licensor such as the negator Dhen ‘not,’ which c- commands the NPI kanenan ‘NPI-person’ and the NCI KANENAN ‘n-person’: L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 75 ]
67
(12) Greek (Giannakidou & Zeijlstra, 2017, p. 17) a. Dhen idhe kanenan o Janis. not saw NPI-person the John ‘John didn’t see anybody.’ = John DIDN’T see anybody (#at all). b.
Dhen idhe KANENAN not saw n-person ‘John didn’t see ANYBODY at all.’
o Janis. the John
c.
∗Idhe saw
o Janis. the John
kanenan/KANENAN NPI-person/n-person
However, consider, for instance, other cases in Greek (Giannakidou, 1998), where NPIs precede negation at surface structure. Giannakidou shows that these Greek NPIs must be c-commanded by negation at LF and are therefore covertly licensed. Consider the following contrast (Giannakidou, 1998, pp. 235–236): (13) a.
b.
*Dhen lipame pu pligosa not regret.1sg that hurt ‘I do not regret that I hurt anyone.’
kanenan anyone
(Greek)
Fimes oti sinelavan kanenan dhen kikloforisan (Greek) rumors that arrested.3pl anybody not were.circulated.3pl ‘Rumors that they arrested anybody were not circulated.’
For Giannakidou, the contrast in grammaticality between (13)a, where the NPI kanenan ‘anyone’ is (unsuccessfully) c-commanded by negation, and (13)b, where the same NPI is not c-commanded by negation, is evidence that licensing takes place at LF. Under her analysis, the NPI in (13)a undergoes LF movement to a position above negation adjoined to IP. This movement, schematized in (14)a, is motivated by the presuppositional force of the complementizer pu. After this LF movement, the NPI is no longer c-commanded by negation, which leads to ungrammaticality. In (13)b, by contrast, the NPI is part of the topicalized complex NP Fimes oti sinelavan kanenan. At LF, this NP reconstructs to a position c-commanded by negation, as in (14)b, yielding the following grammatical structure: (14) a. b.
[IP [CP[ pu pligossa kanenan]i [IP dhen lipame [VP [CP ti]]]] [IP [NP fime oti sinelavan kanenan] [IP dhen kikloforisan [VP [NP]]]]
In Arabic, adverbial NPIs can precede the negative marker like this Greek data, as can be seen in example (34), repeated as follows.
[ 76 ] A Unified Theory of Polarity Sensitivity
7
(15) ʕumr-o * (ma) zaar(*-iš) NPI-ever-his neg visit.3msg.pfv-neg ‘He has never visited Petra.’
el-batra def-Petra
(JA)
However, the question of whether these NPIs are also licensed under c- command by negation is complicated by the head-like properties that they exhibit. Moreover, recent analyses have proposed other licensing configurations for these NPIs, a topic to which I will return in chapter 5. 3.2. NPI ANALYSES IN ARABIC
Postverbal determiner and nominal NPIs are dependent upon negation in that the NPI is licensed when c-commanded by negation (Benmamoun, 1996). (16) a.
b.
*(lam) yaʔti neg.past come.3msg.pfv ‘No one came’ *(ma)-ʔaʤaa-š neg-come.3msg.pfv-neg ‘No one came’
(17)
ʔaħadun/ ʔayyu one/ any
ʔaħadin one
ʔayy ħada any
ħada/ one/
(SA)
(JA) one
FP Spec
F F
[T+Neg+V]
TP Spec NPI
T
T [Neg+V]
NegP Spec
Neg
Neg [Neg+V]
VP Spec
V
NPI
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 77 ]
87
The NPIs in these examples function as subjects. Assuming a VP-internal subject position (Mohammad, 1989; Fassi Fehri, 1993), the subject NPIs are merged in Spec-VP, placing them in the c-command domain of the negative markers in the head of NegP. The analysis represented in the preceding syntactic tree also assumes that the verbal complex, including negation, moves to a functional projection above TP, following Ouhalla’s (1994) analysis of verb movement. Here, the verbal complex, including negation, continues to c-command the NPI even after the NPI moves to Spec-TP to receive overt case assignment, according to Benmamoun’s analysis (1996, 1997). When an NPI follows negation, the configuration is interpreted as a dependency relationship between the NPI and the negation marker that is mediated through c-command (Lasnik, 1975; Jackendoff, 1969, 1972; Linebarger, 1981, 1987; Giannakidou, 1998, 2006, 2011; Benmamoun, 1997, 2006; Benmamoun & Kumar, 2006). (18)
The C-command Configuration: TP Spec
T T
NegP Neg lam
C-command
VP Spec
V
NPI ayy- a ad
In (18), representation for (16)a, lam c-commands the NPI. Neither the Neg head nor the NPI dominates the other, and the first node dominating Neg also dominates the NPI. However, various types of NPIs can precede negation. Pre-negative NPIs exhibit a complex distribution that depends on the lexical category of the NPI and its linguistic variety. Nominal NPIs can be pre-negative in subordinate clauses in SA (Standard Arabic) (Alqassas, 2015, p. 112) and Levantine Arabic (LA) (Hoyt, 2010, pp. 250–251):2
2. Assuming that NPIs can be licensed by being in a specifier–head relation with negation, Benmamoun raises the question of why determiner and nominal NPIs in Arabic cannot precede negation in root clauses. Benmamoun reduces the ungrammaticality to a well-known general principle that bans nonspecific indefinite NPs from occurring preverbally (Ayyoub, 1981; Mohammad, 1989; Fassi Fehri, 1993). [ 78 ] A Unified Theory of Polarity Sensitivity
97
(19) a.
b.
ʔaðˁunnu ʔanna ʔaħadan think.I comp one ‘I think that no one came.’
lam neg.past
yaʔti come.3MSG
wa-batmanna innu ħada ma-yizʕal and-desire.1fsg that one not-anger.3MSG ‘. . . and I hope that anyone doesn’t get angry at me’
(SA)
min-ni (LA) from-me
Hoyt (2010) suggests that the NPI in LA shown in (19)b is interpreted as topical in root clauses but existential in subordinate clauses. This leads to the suggestion that these NPIs can only be pre-negative if they are in a locality relation with negation. A topic is a constituent that occupies a position to the left periphery of the clause. This position is not in a local relation with negation since other constituents may intervene between the topic and negation. Consider the following contrast, where the focused constituent Hindan can intervene between the topic Zaydan and negation in (20)c but cannot between the subject ʔaħadan and negation in (20)b: (20) a.
ʔaðʕunnu ʔanna ʔaħadan lam think.I comp NPI.one neg.past ‘I think that no one visited Hind.’
b.
*ʔaðˁunnu ʔanna think.I comp
ʔaħadan NPI.one
c.
Zayd-un Hind-an lam Zayd-nom Hind-acc neg.past ‘As for Zayd, he did not visit Hind.’
d.
(*ʔayy ħada) ma-ʔaʤaa-š any one neg-come.3msg.pfv-neg ‘No one came’
Hind-an Hind-acc
yazur visit.3MSG
Hind-an Hind-nom
(SA)
lam neg.past
yazur visit.3MSG
(SA)
yazur visit.3MSG
(SA)
(ʔayy ħada) any one
(JA)
In Benmamoun’s (1997) analysis, a pre-negative NPI in MA (Moroccan Arabic) can establish dependency with negation within a local domain. This local domain is the Spec-Head configuration where the NPI is in the specifier of the negation phrase (NegP) and the negative marker is the head of NegP. Crucially, the NPI that is in the specifier position of NegP (Spec,NegP) is within the domain of negation. The following representation illustrates:
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 79 ]
08
(21)
The Specifier-Head Configuration TP Spec
T T
NegP Spec
Neg Neg
NPI
VP Spec
lam
V
Spec-Head
Crucially, a pre- negative noun that is a topic occupies a position in the CP (complementizer phrase) layer in the left periphery of the clause. Consequently, the topic DP (determiner phrase) is not in a licensing relationship with negation (neither c-commanded by negation nor in Spec-NegP). As such, the ungrammaticality of topical preverbal NPIs is borne out. If preverbal NPIs in root clauses are topical, then they do not enter into a Spec–Head relation with negation and hence are ungrammatical. (22)
Representation for the ungrammatical NPIs in root clauses (examples (20)b, d) CP Spec *NPI
TP Spec
T T
NegP Neg
lam/ma… š
VP Spec
V
Idiomatic NPIs can precede negation despite the fact that they are not in a local relationship with negation. They, too, can be separated from negation by other constituents such as the subject, as in (23)b. (23) a.
b.
ʔaħmad ma-waffar girš/ fils ʔiħmar Ahmad neg-saved.3MSG penny/red cent ‘Ahmad did not save a penny/red cent’ girš/ fils ʔiħmar ʔaħmad ma-waffar penny/red cent Ahmad neg-saved.3MSG ‘Ahmad did not save a penny/red cent’
[ 80 ] A Unified Theory of Polarity Sensitivity
(JA)
(JA)
18
Syntactic dependency between these NPIs and negation does not follow straightforwardly from their structural configuration since the intervening subject in (23)b is not a sub-constituent of the NPI. Nevertheless, theoretical analyses in the generative tradition posit a covert level of representation (LF)3 where these NPIs are post-negative (also known as “reconstruction at LF”; see Giannakidou, 1998, for Greek; Alqassas, 2015, and references therein for JA [Jordanian Arabic]). This line of analysis contends that constituents like these NPIs reconstruct to a postverbal position where they are under the scope of negation (i.e., c-commanded by negation): (24) [CP girš/fils ʔiħmar
[TP ʔaħmad
T ma-waffar
]]4
(JA)
Evidence for this reconstruction process comes from the grammaticality of NPIs embedded in focus-fronted CPs. Consider the that-clause in (25)a. (25) a.
b.
ʔənnu ʔaħmad waffar girš, ma-ðˁunn-iš that Ahmad saved.3MSG penny neg-think.1sg-neg ‘I don’t think that Ahmad saved any penny.’
(JA)
ma-ðʕunn-iš ʔənnu ʔaħmad waffar girš neg-think.1sg-neg that Ahmad saved.3MSG penny ‘I don’t think that Ahmad saved any penny.’
(JA)
From its embedded position in the focus-fronted clause in this example, the NPI is not c-commanded by negation. The grammaticality of this construction can only be captured if the fronted CP, which contains the NPI, reconstructs to a postverbal position. At LF, the focus-fronted clause reconstructs to its base- generated position as the complement of the verb: (26) [[CP C ʔənnu ʔaħmad waffar
girš]i
[NegP Neg ma-ðˁunn-iš [CPi]]]
Accordingly, the postverbal NPI in Spec-TP is c-commanded by the negative marker in FP. As such, the NPI is licensed under c-command by negation.
3. LF, or Logical Form, is a syntactic level of representation that aims to capture semantic relations between constituents such as scope. 4. CP, the complementizer phrase, is a phrasal category in the left periphery of the sentence. This category can be occupied by fronted constituents such as the NCIs in (24) and question words (e.g., what, why, when, . . . etc.) in interrogative sentences.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 81 ]
28
Assuming that NPIs are licensed through a Spec–Head relationship with negation,5 Benmamoun raises the question of why determiner and nominal NPIs in Arabic cannot precede negation. He suggests that this ungrammaticality can be reduced to a general principle that bans nonspecific indefinite NPs from preverbal positions (Ayyoub, 1981; Mohammad, 1989; Fassi Fehri, 1993). (27) a.
b.
ʤa come.3msg.pfv ‘Some boy came.’
ši some
wəld boy
ši wəld ʤa some boy come.3msg.pfv ‘Some boy came.’ (Benmamoun, 1996, p. 62)
(MA)
(MA)
Determiner and nominal NPIs are examples of inherently indefinite, nonspecific NPs. The derivation in (17) assumes that verb-initial sentences have a topicalized verb, defined as a focus projection here. As such, preverbal NPs must be analyzed as left-dislocated elements. That is, given that the verb is in the CP layer when sentence-initial, whatever precedes it must be in a higher CP. This left-dislocated position cannot be occupied by nonspecific indefinite NPs (including determiner and nominal NPIs), which drives the ungrammaticality of these elements in preverbal position. Recall, however, that nominal NPIs in Standard Arabic can precede the verb (and negation) in subordinate clauses; see (28). It is possible to extend Benmamoun’s analysis to such examples. With the past tense negative marker in T (tense) and the NPI subject into Spec-TP (moved for the purposes of case licensing), the NPI is licensed via a Spec–Head relationship with negation. (28) ʔaðˁunnu [CP C ʔanna [TP ʔaħadan T lam [VP V yazur Hindan]]] think.I comp NPI.one neg.past visit.3MSG Hind.ACC ‘I think that no one visited Hind.’
Like SA, nominal NPIs in Levantine Arabic (LA) can precede the verb/negation in subordinate clauses (Hoyt, 2010, pp. 250–251):
5. This assumption is motivated by two observations. First, there is cross-linguistic evidence that NPIs can be licensed in Spec-NegP (Haegeman, 1995; Progavac, 1994). Second, the availability of the Spec–Head relationship as a licensing configuration for the Moroccan NCI hetta+NP motivates the assumption that this configuration is available for NPIs. [ 82 ] A Unified Theory of Polarity Sensitivity
38
(29) a.
leeš ħada ma-ʤaawab why one not-answer.3msg.pfv ‘Why didn’t anyone answer other than me?’
ɣeer-i? other-my
(LA)
b.
zaʕlaana liyinnu ħada ma-radd (LA) angry.1fs.ipfv because one not-answer.3msg.pfv ʕala mawduuʕ-ik. upon subject-your.2fsg ‘I am angry because anyone didn’t respond to your thread’
c.
wa-batmanna innu ħada ma-yizʕal min-ni (LA) and-desire.1fsg.ipfv that one not-anger.3msg.ipfv from-me ‘. . . and I hope that anyone doesn’t get angry at me’
Hoyt’s (2010) analysis distinguishes preverbal nominal NPIs in root clauses from those in subordinate clauses. In root clauses, these NPIs are interpreted as topical elements. Since topical elements must occupy a position in the CP layer, preverbal nominal NPIs in root clauses do not enter into a Spec–Head relation with negation and are therefore ungrammatical. By contrast, preverbal nominal NPIs in either subordinate clauses or in a position following question words have an existential rather than a topical interpretation. This places the NPI in an appropriate licensing configuration with negation. So far the discussion has centered on the licensing of determiner and nominal NPIs. The research suggests that both c-command and Spec-Head configuration are available for licensing these NPIs. However, we do not have a clear answer as to whether NPIs may also be licensed covertly at LF. To answer this question, we need to look at the licensing of idiomatic NPIs, which can precede or follow the verb/negation: (30) a.
b.
ʔaħmad ma-waffar girš/fils ʔiħmar Ahmad neg-save.3msg.pfv penny/red cent ‘Ahmad did not save a penny/red cent’ girš/fils ʔiħmar ʔaħmad ma-waffar penny/red cent Ahmad neg-save.3msg.pfv ‘Ahmad did not save a penny/red cent’
(JA)
(JA)
In (30), the NPI is not in a Spec–Head relation with negation since the subject can intervene between the NPI and the negative marker. Instead, the NPI must be interpreted as a left-dislocated element occupying an XP position in the CP layer. Moreover, the preverbal NPI is not c-commanded by negation.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 83 ]
48
How, then, is it licensed? Since overt licensing via c-command or the Spec- Head configuration is not possible, c-command at LF is the only possibility: (31) [CP girš/fils ʔiħmar
[TP ʔaħmad
T ma-waffar]]
(JA)
Another argument for covert licensing comes from the grammaticality of nominal NPIs, which are embedded in focus-fronted CPs (Alsarayreh, 2012, p. 133). (32) ʔənnu Maryam ðarbat ħada that Mary hit.3fsg.pfv one ‘I do not believe that Mary hit anyone.’
*(ma)-basˁaddig. neg-believe.1sg.ipfv
(JA)
The NPI in this example is not overtly c-commanded by negation since it is embedded in the focus-fronted clause. As such, the grammaticality of this construction can only be explained if c-command as a licensing condition obtains at LF. At LF, the focus-fronted clause reconstructs to its base-generated position as the complement of the verb: (33) [CP [CP C ʔənnu Maryam ðarbat
ħada]i
[NegP neg ma-basˁaddig [CPi]]
The last issue surrounding NPI licensing has to do with the NPI ʕumr ‘ever.’ This NPI tends to occur preverbally: (34) ʕumr-ha maa katbat NPI-ever-her neg wrote.3FSG ‘She has never written a novel.’
riwaaye novel
(JA)
The simplest analysis for this NPI is to assume that it is licensed in the Spec– Head relation with negation. However, this analysis is challenged by at least two empirical facts from MA (Benmamoun, 2006) and JA, shown as follows. First, the subject can intervene between the NPI and negation. (35) ʕumr Layla maa NPI-ever Layla neg ‘Layla has never written a novel.’
katbat wrote.3FSG
riwaaye novel
(JA)
Second, the NPI exhibits head-like properties in that it can host clitics and assign genitive or accusative case to these clitics; see examples (36)a–b.
[ 84 ] A Unified Theory of Polarity Sensitivity
58
(36) a.
b.
ʕumr-i *(ma) zurt NPI-ever-my neg visit.1sg.pfv ‘I have never visited Petra.’
el-batra def-Petra
ʕumr-ni *(ma) zurt NPI-ever-me neg visit.1sg.pfv ‘I have never visited Petra.’
el-batra def-Petra
(JA)
(JA)
Benmamoun (2006) convincingly argues against the use of covert movement of the negative marker to license this NPI. If the negative marker can covertly raise to a position c-commanding the NPI, this incorrectly predicts that the negative head can license the NPI ʕumr, even when the auxiliary kaan ‘be’ separates the NPI from negation. Consider the MA contrasts illustrated in (37): (37) a.
b.
ʕəmmər-u ma-kan never-him neg-be.pfv ‘He never loved Nadia.’ *ʕəmmər-u never-him
kan be.pfv
taybɣi love.3msg.pfv
ma-taybɣi neg-love.3msg.pfv
nadya Nadia
nadya Nadia
(MA)
(MA)
Instead, Benmamoun (2006) argues that this NPI acts as the head of a clausal projection, XP, in MA. (38)
XP X
NegP
Building on the notion that head–complement relations between T and VP are checking configurations in English (Bobaljik & Thrainsson, 1998) and the use of this configuration to satisfy categorical selections (e.g., V may select DP or PP [prepositional phrase]), Benmamoun proposes that the head–complement relation between the NPI and negation in (38) is a licensing configuration in MA. However, in chapter 5, I will argue against this analysis of head NPIs. There, I show that head-like PSIs cannot be analyzed as heads that project a clausal projection that selects NegP. Instead, I argue for analyzing these PSIs as adverbs with a phrasal projection that takes a pronominal complement. This pronominal complement can be coindexed with the grammatical subject or thematic object of the verb.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 85 ]
68
3.3. ADVERBIAL AND DETERMINER NPIS IN ARABIC
One intriguing property of bipartite negation is the absence of the enclitic negative marker -š in the context of PSIs. This property is reported in the literature as a phenomenon of complementary distribution between the relevant PSIs and the enclitic marker. In this section, I argue that this is an epiphenomenon of the availability of two different negation structures: bipartite and single negation; and that the NPI in the left periphery is always licensed by single negation.6 Consider the following examples of MA PSIs from Benmamoun (2006, pp. 143–144) and EA (Egyptian Arabic) PSIs from Soltan (2012, pp. 241–242) where the enclitic negative marker -š is not permitted: (39) a.
b.
(40) a.
nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
ma-ʒat(*-š) neg-come.3fsg.pfv(*-neg)
ma-ʤa(*-š) neg-come.3msg.pfv(*-neg) ‘No one came’
ħətta waħəd even one
ʕumr-i maa-saafirt(*-iš) NPI-ever-me neg-travel.1sg.pfv-neg ‘I have never traveled to Egypt.’ maa-saafirt-*(iš) Masr neg-travel.1sg.pfv-neg Egypt ‘I have never traveled to Egypt.’
c.
ma-zaar-iš ʕumr-o neg-visit.3msg.pfv-neg NPI-ever-his ‘He has never visited Petra.’
(MA)
Masr Egypt
ʕumr-ii ever-me
b.
(MA)
(EA)
(EA)
el-batra def-Petra
(JA)
Previous literature on Arabic negation has noted that the enclitic negative marker does not co-occur with PSIs (Ouhalla, 2002; Benmamoun, 2006; Soltan, 2012; among others). This is usually described as a case of complementary distribution. I will refer to this as a case of mutual exclusivity since my analysis shows that they do not compete for the same position.7 The mutual 6. The analysis of the JA phenomenon of complementary distribution as an epiphenomenon first appeared in my 2015 paper in Lingua. 7. The phenomenon of mutual exclusivity is inaccurately described as complementary distribution in the literature. Note that complementary distribution incorrectly suggests that every sentence has either the NPI or the clitic. [ 86 ] A Unified Theory of Polarity Sensitivity
78
exclusivity between the enclitic negative marker and PSIs in MA is reduced to the economy principle ‘last resort’ in Ouhalla (2002), which focuses on the LF of bipartite negation in MA. He proposes that sentential negation involves a negative operator that must always bind (i.e., quantify over) a variable to avoid vacuous quantification. This variable, Ouhalla argues, is the enclitic negative marker -ši (Ouhalla, 2002, p. 307).8 (41) [TP ʕumar [T [[NegOp ma] [V qra]]j [NegP ši [[NegOp][Vi]]j [VP Vi Omar neg read.PFV.3MSG VAR ‘Omar did not read the book.’
l-ktab]] the-book
Under this analysis, MA’s negative marker, ma, is the spell-out of the negative operator under Neg0. The enclitic –ši is a variable (glossed as VAR) in Spec-NegP. Assuming that the verb moves to T, the verb incorporates with ma and creates the complex head [[Neg Op][V]]. This head undergoes subsequent movement to the T head, leaving the enclitic –ši to the right of the verb, where it cliticizes leftward onto the complex [[Neg Op][V]]. This results in bipartite negation. Ouhalla shows that the negative marker -ši cannot co-occur with NCIs in MA, as in (42). He proposes that -ši is a variable that is inserted in negative sentences as a last resort mechanism to avoid the situation where the negative operator does not have a variable to bind. He calls this situation “vacuous quantification.” Under Ouhalla’s analysis, in sentences with an NCI, the NCI is the variable that is bound by negation. As such, he argues that -ši is a dummy variable that is only needed as a last resort mechanism to avoid vacuous quantification, thus explaining why the NCI and -ši cannot co-occur with each other in MA.9 8. For arguments of this claim, see Ouhalla (2002). 9. This suggests that the negative marker ši and the NPI ħətta ħaja ‘anything’ are competing for the same position, i.e., the Spec-NegP position. However, Ouhalla rules out this possibility by observing that more than one NPI can appear in the same sentence/clause in MA. In other words, if the negative marker ši competes with the NPI ħətta ħaja ‘anything’ for the Spec-NegP position in (42), the NPI ħətta waħəd should be competing with the NPI ħətta ħaja ‘anything.’ This is not borne out, as shown by the grammaticality of the following examples. (i)
a.
ma-ʕtat Nadia ħətta ħaja NEG-give.PFV.3FSG Nadia NCI-thing ‘Nadia did not give anything to anybody.’
l-ħətta waħəd. to-NCI-person
b.
ħətta waħəd ma-šaf NCI-person NEG-see.PFV.3msg ‘Nobody saw anything (anywhere).’
(f-ħətta blasa) (in-NCI- place)
ħətta ħaja NCI-thing
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 87 ]
8
(42) ma šaft (*ši) neg see.PFV.3fsg VAR ‘Nadia did not see anything.’
Nadia Nadia
ħətta ħaja. NCI-thing
(MA)
To support the idea that the NCI acts as a variable bound by negation, Ouhalla invokes a syntactic licensing requirement for NCI in that NCIs in MA must be licensed by negation. Thus, the NCI must either be c-commanded by negation or in Spec–Head relation with negation (Benmamoun, 1997). He supports this argument through an analysis of the following data where the subject NPI ħətta waħəd ‘anyone’ cannot be licensed if the negative marker ma is not in T (examples from Ouhalla, 2002, p. 303). (43) a.
b.
(44) a.
b.
ʕumar ma-huwwa -ši Omar neg-PRON VAR ‘Omar is not sick/in the house.’ ʕumar huwwa ma-ši Omar PRON neg-VAR ‘Omar is not sick/in the house.’ ħətta waħəd ma huwwa NCI-person neg PRON ‘No-one is sick/in the house.’
mrid/f-l-dar. sick/in-the- house
(MA)
mrid/f-l-dar. sick/in-the-house
mrid/f-l-dar. sick/in-the- house
*ħətta waħəd huwwa ma mrid/f-l-dar. NCI-person PRON neg sick/in-the-house ‘No-one is sick/in the-house.’
(MA)
(MA)
(MA)
The examples in (43) show that the negative marker ma can either appear cliticized to the copular pronoun ‘PRON,’ which is assumed to be in T, or as part of -ši. In (44), however, the negative marker can only appear as a proclitic on PRON. This contrast shows that there is syntactic dependency between the negative marker ma and the NCI. Since the NCI ħətta waħəd in both examples is not within the c-command domain of negation, it must be licensed via a Spec–Head relation with negation. In (45)a, the derivation for (44)a, the negative marker moves from the NegP projection and incorporates with the Copula under the T head. The NCI, being the subject of the sentence, is in Spec-TP. Therefore, negation can license the NCI and bind its argumental variable under the Spec–Head relation.
[ 88 ] A Unified Theory of Polarity Sensitivity
98
(45)
a.
[TP ħətta waħəd [T [[Neg Op ma]i [Copula huwwa]] [NegP [Neg OP]i [PredP..]]]
b.
*[TP ħətta waħəd [T [Copula huwwa] [NegP [Neg OP [PredP..]]]
In (45)b, the derivation for (44)a, the negative marker stays in the NegP projection while the NCI is in Spec-TP. Therefore, the ungrammaticality of (44)b follows from the inability to license the NCI since the subject NPI ħətta waħəd ‘anyone,’ which is in Spec-TP, is not in a Spec–Head relation with the negative marker ma, which is in Neg0. Without the need to go into further details of Ouhalla’s analysis, I will show that it cannot be extended to explain the distribution of bipartite negation with respect to NPIs in JA. This is first and foremost because, unlike in MA, the enclitic negative marker in JA can co-occur with the postverbal NPI, as in (46); cf. (40)b, c as counterexamples from EA and JA). (46) ma-zaar-iš neg-visit.PFV.3msg-neg ‘He never visited Petra.’
ʕumr-o NPI-ever-him
el-batra def-Petra
(JA)
Second, previous literature has proposed that the binding of variables via negation is the cause for negative interpretations of predicates at LF (Acquaviva, 1997; Giannakidou, 1999, 2000, 2006; Zeijlstra, 2004, 2008). The idea is that negation introduces an existential quantifier that binds the event variable of the predicate (and the variables of the internal and external arguments as will be explained in section 3.4). (47) [Op¬ ∃e,x [ . . . (e) . . . (x) . . . ]]
Under this view of sentential negation, the negative marker maa will always have a variable to bind. This, in turn, means that we cannot claim that -š is a variable bound by negation via a last-resort mechanism. In section 3.4, I follow this view of sentential negation. Another analysis (Soltan, 2012) reduces the lack of the enclitic negative marker in the context of preverbal NPIs in EA to morphological licensing. Soltan (2012) analyzes this phenomenon in EA. Soltan’s analysis contrasts the behavior of the NPI ʕumr with the NCI lissa, which corresponds to JA’s NCI baˁd. He capitalizes on the contrasts between these PSIs in formal features. The NPI does not have the formal feature neg while the NCI has. Therefore, Soltan posits a principle that prohibits elements with formal feature mismatch from participating in a syntactic licensing relation. The idea is that ma licenses the enclitic and the NPI and, at PF, the enclitic negative marker is deleted because
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 89 ]
09
it is formally negative, unlike the NPI. For lack of formal negativity, consider the following examples10 where ʕumr can be licensed by a question (48) and a conditional (52) but cannot be a fragment answer (48)b. (48) a.
ʕumr-o zaar el-batra NPI-ever-him visit.PFV.3MSG def-Petra ‘Has he ever visited Petra.’
b. Answer: *ʕumr-o ever-him ‘Never.’
(49) a.
ʔaħmad wisˁil? Ahmad arrive.PFV.3MSG ‘Has Ahmad arrived?’
b. Answer:
(50) a.
*šuft-iš ʔaħmad el-yoom? b. Answer: *šuft-iš see.PFV.2MSG-neg Ahmad def-day saw.2MSG-neg Intended reading: ‘Did you see Ahmad today?’
(51) *baˁd-o zaar NCI-yet-him visit.PFV.3MSG ‘Has he visited Petra yet?’ (52) ðakkir-ni ʔiða ʕumr-o remind-me if NPI-ever-him ‘Remind me if he ever visited Petra.’
baˁd-o yet-him ‘Not yet.’
el-batra? def-Petra
zaar el-batra? visit.PFV.3MSG def-Petra
(53) *ðakkir-ni ʔiða baˁd-o zaar remind-me if NCI-yet-him visit.PFV.3MSG ‘Remind me if he has visited Petra yet.’
el-batra? def-Petra
The NCI baˁd displays the opposite behavior in that it can act as a fragment answer as in (49)b, but the enclitic negative marker cannot be licensed by a question (50) and (51) or a conditional (53). Under Soltan’s analysis of EA, we are led to the conclusion that both the NCI baˁd and the enclitic negative marker in JA are formally negative, i.e., they are specified with the negation formal feature [uneg].11 Under his analysis, the NPI ʕumr is not specified for 10. Examples are from JA but they are the equivalents of the CEA (Cairene Egyptian Arabic) data reported in Soltan (2012). 11. Here, the enclitic negative marker is expected to pattern with baˁd in occurring as a fragment answer, but this is not borne out. Soltan explains that this is due to the nature of this marker as a clitic that needs a host. [ 90 ] A Unified Theory of Polarity Sensitivity
19
any negation formal feature, but the negative marker ma carries an interpretable negation feature [ineg]. This negative marker licenses the enclitic negative marker -š, which carries a [uneg] feature.12 With the preverbal NPI ʕumr in (54), ma licenses the NPI under the Spec–Head relation and licenses -š under c-command. This creates a situation where the same licensor, the negative marker ma, has one licensee (-š) that carries the formal feature [uneg] while the other licensee (the NPI) lacks it. Soltan describes this as a formal feature mismatch and proposes a constraint against it in (55). (54) a.
b.
*ʕumr-o ma-zaar-iš el-batra NPI-ever-him neg-visit.PFV.3MSG-neg def-Petra ‘He has never visited Petra.’ [PolP
ʕumr
Pol[ineg] [NegP Neg[uneg] [TP T [vP . . .]]]]
(55) Minimize formal feature mismatch (MFFM): At Spell-out, minimize formal feature mismatch on licensees of the same licenser within a local domain.
Under this analysis, MFFM blocks the pronunciation of the item that carries the [uneg] feature. In the context where the preverbal NCI baˁd is compatible with -š, as in (56), since both have the formal feature [uneg], the MFFM does not apply and -š is pronounced. (56) a.
b.
baˁd-o ma-zaar-iš NCI-yet-him neg-visit.PFV.3MSG-neg ‘He hasn’t visited Petra yet.’ [PolP
baˁd[uneg]
el-batra def-Petra
Pol[ineg] [NegP Neg[uneg] [TP T [vP . . .]]]]
However, this analysis cannot be extended to the JA data. First, the MFFM process over-applies in cases where the NCI baˁd co-occurs with single negation. Consider the following example from JA:
12. Notice here that -š is the head of a NegP projection that is not the source of negative interpretation in Soltan’s analysis. The negative marker ma heads a PolP projection, the source of negative interpretation. This analysis departs from Benmamoun’s (2000) analysis of Arabic negation where both ma and -š occupy the same discontinuous head in NegP. I agree with Soltan’s analysis of the two negative markers occupying different heads based on arguments in Alqassas (2012). Soltan’s designation of these negative markers is used here to explain his analysis.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 91 ]
29
(57) a.
baˁd-o ma zaar NCI-yet-him neg visit.PFV.3MSG ‘He hasn’t visited Petra yet.’
b. [PolP
baˁd[uneg]
el-batra def-Petra
Pol[ineg] [NegP Neg[uneg] [TP T [vP . . .]]]]
The MFFM over-applies in this case because there is no formal feature mismatch. Both licensees of the negative marker ma are formally negative. As such, we expect -š to be pronounced but find that it is not. Second, the MFFM process by definition predicts that the negative marker ma cannot license an NPI and an NCI at the same time since it is a filtering mechanism that prohibits the pronunciation of the vocabulary item that carries the [uneg] feature. When an NPI, which does not carry a [uneg] feature, and an NCI, which carries a [uneg] feature, are licensed by ma, we expect the MFFM to block the pronunciation of the NCI as it would with -š. However, this prediction is not borne out: (58) a.
b.
baˁd-o ʕumr-o maa NCI-yet-him NPI-ever-him neg ‘He has never visited Petra yet.’ ʔabadan ʕumr-o maa NCI-never NPI-ever-him neg ‘He has never spent a penny at all.’
zaar el-batra visit.PFV.3MSG def-Petra
sˁaraf spend.PFV.3MSG
As seen in these example, the NCIs baˁd and ʔabadan, which are formally negative, can co-occur with the NPI ʕumr, which is not formally negative. Counter to the expectation under Soltan’s analysis, both cases involve a formal feature mismatch between the licensee and the licensor. (59) a. b.
[FP baˁd[uneg] [PolP [FP ʔabadan[uneg] [PolP
ʕumr ʕumr
Pol[ineg] [NegP Neg[uneg] [TP T [vP . . .]]]] Pol[ineg] [NegP Neg[uneg] [TP T [vP . . .]]]]
As the derivations show, the NCI in this case is in a projection above negation (call it FP) and the negative marker has three licensees, the NCI baˁd/ ʔabadan, the NPI ʕumr, and -š. The NCIs and -š are formally negative while the NPI is not. Third, the MFFM analysis is based on the assumption that negation is above TP. This assumption is crucial for explaining why the postverbal ʕumr is compatible with -š; cf. (54)a for ungrammaticality when the NPI is preverbal.
[ 92 ] A Unified Theory of Polarity Sensitivity
39
Soltan reports this observation for CEA, but it is also true of JA as shown in the following example: (60) ma-zaar-iš neg-visit.PFV.3MSG-neg ‘He never visited Petra.’
ʕumr-o NPI-ever-him
el-batra def-Petra
To account for this, Soltan proposes that the MFFM applies in a local domain, defined as a phase (vP, CP) in the sense of Chomsky (2001). Assuming that negation is above TP and the postverbal NPI is within the vP, the MFFM does not apply in (60) since the NPI and negation are in separate phases. However, if the NPI is on the edge of the vP phase and negation is below TP, the MFFM applies. Finding an analysis that captures the alleged complementary distribution between the enclitic and a particular PSI requires us to question the premise behind its conceptualization. That is to say, we should question whether these PSIs are, in fact, in complementary distribution with enclitic negation, causing the enclitic negative marker to disappear in this context. First, if the syntactic structure of such clauses results in the disappearance of the enclitic negative marker at some level of representation (Syntax-LF in Ouhalla, 2002; or PF [phonetic form] in Soltan, 2012), the phenomenon should be referred to as ‘mutual exclusivity,’ and not complementary distribution. Complementary distribution incorrectly implies that every sentence either has the NPI or the enclitic marker. Second, under a multi-locus analysis of negation (Alqassas, 2015, 2019), we have a more adequate explanation for this phenomenon. I turn to this analysis now. The lack of the enclitic negative marker when the NPI is preverbal is better explained when recast as an epiphenomenon of the availability of two negatives (see Alqassas, 2015). Arabic exhibits two different negatives: a single negation marker and a bipartite negation marker. These negative markers differ in their syntactic position. Single negation occurs above TP and bipartite negation occurs below TP. To see how this works, let us take the preverbal JA NPI ʕumr as an example. This NPI cannot co-occur with enclitic negation. It also tends to occur at the left periphery of the sentence, preceding the verb and the preverbal subject: (61) ʕumr ʔaħmad ma saafar NPI-ever Ahmad neg travel.3msg.pfv ‘Ahmad never traveled by train.’
bi-l-qitˁaar by-def-train
(JA)
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 93 ]
49
If this NPI is only be licensed in Spec,NegP or under c-command, and if bipartite negation is below TP, it follows that bipartite negation cannot license this NPI. Being above TP, single negation, however, can license the NPI. Evidence in favor of locating single negation in this context above TP can be seen in the following example, where an adverb intervenes between single negation and the verb: (62) ʕumr ʔaħmad maa b-yoom saafar NPI-ever Ahmad neg in-day travel.3msg.pfv ‘Ahmad never traveled by train on any day.’
bi-l-qitˁaar (JA) by-def-train
In this analysis, the higher negative licenses the adverbial NPIs that are merged above TP while the lower negative does not since it is too low in the structure (below TP) for the NPI to be licensed in the specifier of NegP (see Alqassas, 2015, for JA). (63) a.
ʕumr-ha maa katbat NPI-ever-her neg write.3Fsg.pfv ‘She has never written a novel.’ *ʕumr-ha NPI-ever-her
c.
maa gad fi-sˁsˁaf ʔaħmad kallam-a neg ever in class Ahmad talk.3msg.pfv-him ‘Ahmad has never talked to him in class’
d.
*gad ever
ʔaħmad Ahmad
maa neg
riwaaye novel
(JA)
b.
fi-sˁsˁaf in class
ma-katbat-iš neg-write.3Fsg.pfv-neg
riwaaye novel
(JA)
kallam-a talk.3msg.pfv-him
(QA)
(QA)
This type of analysis is further supported by the fact that the NPI ʕumr in postverbal position can, in fact, co-occur with bipartite negation in both EA in (40)b and JA in (40)c. Given that this NPI can occur in postverbal position and that bipartite negation c-commands this postverbal NPI, it follows that bipartite negation can co-occur with this NPI. The ability of the higher negative to license NPIs is also attested with adverbs that express categorical negation, such as gad ‘never.’ Here, we also find that the adverb can intervene between negation and the verb. (64) a.
maa gad fi-sˁsˁaf kallamt-a neg ever in class talk.1sg.pfv-him ‘I’ve never talked to him in class’
[ 94 ] A Unified Theory of Polarity Sensitivity
(QA)
59
fi-sˁsˁaf in class
b.
*maa neg
kallamt-a talk.1sg.pfv-him
(QA)
c.
mub fi-sˁsˁaf kallamt-a neg in class talk.1sg.pfv-him ‘I did not talk to him in class’
(QA)
Recall from chapter 1 that there are arguments for analyzing the status of temporal NPIs, such as ʕumr/ʕəmmər ‘ever’ and baʕd/bagi ‘yet,’ as heads. However, it remains unclear whether these items act as clausal heads or phrasal heads. Consider the following example from JA, where the NPI ʕumr can be followed by a dislocated object in (65)a or by a relative clause in (65)b: (65) a.
b.
ʕumr l-walad maa NPI-ever def-boy neg ‘A girl never loved the boy.’
ħabbat-o love.3fsg.pfv-him
ʕumr ʔilli budrus NPI-ever who study.3msg.ipfv ‘He who studies never fails.’
maa neg
bint girl
bursub fail.3msg.ipfv
(JA)
(JA)
If the NPI is a clausal head in a projection above NegP and this NPI must be followed by NegP for licensing under the head-complement configuration (Benmamoun, 2006), it follows that the NPI cannot be licensed since the object l-walad and the relative clause ʔilli budrus already serve as complements for the NPI. Thus, the NPI projection cannot select NegP as a complement. See the following illustration: (66) a. b.
[XP X ʕumr [DP l-walad [NegP ma . . .]]] [XP X ʕumr [CP ʔilli budrus [NegP ma . . .]]]
A solution to this problem is to assume instead that the NPI is a phrasal head that is licensed in Spec,NegP and selects either the object or the relative clause as a complement. Refer to the following illustration: (67) a. b.
[NegP [XP [X ʕumr] [DP l-walad ] [Neg ma . . .]]] [NegP [XP [X ʕumr] [CP ʔilli budrus] [Neg ma . . .]]]
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 95 ]
69
I develop this analysis in chapter 5 where I contrast JA with MA. Recall that there is another type of negation in which the negative marker can precede the subject, referred to in this book as ‘cleft negation.’ This type of negation is exemplified by the phrase ‘Isn’t it the case that . . .’ as in the following example. (68) a.
b.
ʔinta miš/mub ʕaarif you neg knowing ‘You don’t know the answer?’
al-ʤawaab? def-answer
(JA/QA)
miš/mub int ʕaarif al-ʤawaab! (JA/QA) neg you knowing def-answer ‘Don’t you know the answer!’/‘Isn’t it the case that you know the answer!’ (Speaker presupposes that addressee knows the answer)
There is evidence that the negative marker here is above TP but in a different position than NegP. Specifically, this negative cannot license NPIs: (69) a.
ʔinta miš/mub ʕaarif ʔayya you neg knowing NPI-any ‘You don’t not know any solution’ int you
ʕaarif knowing
ʔayya NPI-any
ħal solution
ħal! solution
(JA/QA)
b.
*miš/mub neg
(JA/QA)
c.
miš/mub int mub ʕaarif al-ʤawaab! (JA/QA) neg you neg knowing def-answer ‘Isn’t it the case that you don’t know the answer!’ (Speaker presupposes that addressee doesn’t know the answer)
As such, this indicates that mub is not in a NegP projection co-commanding the NPI. The question then is where the negative is located in the syntactic structure. Presumably, it is buried in the CP domain (in a Focus projection) where it does not c-command the NPI. In this position, the negative marker would form a constituent with the pronoun inta (constituent negation). (70) *[CP miš/mub inta C [TP T [AP ʕaarif
[ 96 ] A Unified Theory of Polarity Sensitivity
ʔayya
ħal . . .
79
(70)
*[CP m iš/mub inta C [TP T [AP arif
ayya
al…
FocP DP mub
Foc’
inta
Foc 0
TP
Spec
T’ T0
AP NPI
3.4. STANDARD ARABIC NEGATION AND NPI LICENSING
The negative marker maa in SA can license the NPI ʔaħad-an in (71)a, where negation scopes over the whole proposition, but it cannot do so in (71)b, where negation precedes the fronted adverb and scopes over the focused adverb, l-baariħata: (71) a.
b.
maa qaabaltu neg meet.1sg.pfv ‘I have not met anyone’
ʔaħad-an one-acc
(SA)
*maa l-baariħata qaabaltu ʔaħadan bal al-yawma (SA) neg def-yesterday meet.1sg.pfv one-acc but def-today ‘*It was not yesterday that I met anyone, but today’
Moreover, sentential negation maa cannot be followed by a clitic left-dislocated NP (examples include an NPI to make sure that negation is sentential): (72) a.
b.
maa qaabala neg meet.3msg.pfv ‘No one has met Zayd.’ [NegP maa [TP [vP qaabala *maa neg
Zayd-un Zayd-nom
ʔaħad-un one-nom
Zaydan Zayd-acc
ʔaħad-un
Zaydan
qaabala-hu meet.3msg.pfv-him
ʔaħad-un one-nom
(SA)
(SA)
This ungrammaticality is expected if maa is the head of a NegP on top of TP. The clitic left-dislocated NP Zaydan cannot occupy a position between NegP and TP and therefore cannot intervene between maa and the verb. However,
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 97 ]
89
if maa is located in the CP layer, we expect that Zaydan can intervene between maa and the verb. This is borne out, as shown in the following example where a clitic left-dislocated NP can precede or follow contrastive focus maa (Moutaouakil, 1991, pp. 276–277). (73) a.
b.
al-kitaab-u maa allaftu-hu bal sˁaħħaħtu-hu (SA) def-book-nom neg write.1sg.pfv-it but correct.1sg.pfv-it ‘I haven’t written this book; I have only corrected it.’ [FP al-kitaab-u [FocusP maa [VP allaftu-hu bal sˁaħħaħtu-hu maa al-kitaab-u allaftu-hu bal sˁaħħaħtu-hu (SA) neg def-book-nom write.1sg.pfv-it but correct.1sg.pfv-it ‘I haven’t written this book; I have only corrected it.’ [FocusP maa [FP al-kitaab-u [VP allaftu-hu bal sˁaħħaħtu-hu
When maa does not have a contrastive focus interpretation, it acts as a sentential negative marker that can license NPIs. However, when associated with a contrastive focus interpretation, maa cannot license NPIs. We can find similar contrasts in Moutaouakil (1991, pp. 274–275): (74) a.
maa qaabaltu neg meet.1sg.pfv ‘I have not met anyone’
ʔaħad-an one-acc
(SA)
qaabaltu ʔaħadan meet.1sg.pfv one-acc
b.
*maa l-baariħata neg def-yesterday
bal but
al-yawma (SA) def-today
c.
*maa qaabaltu ʔaħadan l-baariħata bal al-yawma (SA) neg meet.1sg.pfv one-acc def-yesterday but def-today
This ungrammaticality should not be surprising given the syntactic licensing conditions described thus far. From its position inside the focused constituent l-baariħata, the contrastively focused maa does not c-command the NPI and therefore cannot license it. As such, the ungrammaticality is reduced to failure to properly license the NPI via either of the two configurations: c- command and the Spec-Head configurations (see Benmamoun, 1997, 2006 for MA; Alqassas, 2015 and 2016 for JA). The following is an illustration: (75)
FocP DP maa
XP
Foc’ Foc 0
TP…
[ 98 ] A Unified Theory of Polarity Sensitivity
9
On the other hand, the non-contrastively focused maa is located either under the head of the NegP or the head of a Focus projection (FocP) where it can c-command the NPI. As such, the analysis predicts that non-constrastively focused maa may co-occur with NPIs in SA. The following is an illustration: (76)
FocP Foc
NegP NEG maa
TP Spec
T’ T0
VP
Overall, this section has illustrated that NPIs in SA are licensed according to the previously described syntactic licensing conditions: through c-command or Spec-Head configurations. Let us now briefly discuss the issue of the nature of syntactic licensing for NPIs. Licensing NPIs does not involve formal features in the same way that NCI licensing does. NPIs do not carry uninterpretable negative features that need to be licensed by operators that contain interpretable negative features. The syntactic licensing process is based on the structural c-command and Spec-Head configurations. An anonymous reviewer raises the question of whether these licensing configurations could be reduced to the syntactic operation Agree as proposed in Chomsky (2000), or in the more flexible proposal found in Baker (2008). In principle, the licensing configurations for NPIs can be viewed as Agree relations. The operation Agree is essentially proposed as a process that establishes a functional relation between two syntactic objects (Chomsky, 2000, 2001, 2005, 2008; Miyagawa, 2010). Crucially, this functional relation does not rely on movement in these studies. Other studies (Pesetsky & Torrego, 2006; Sigurðsson, 2004, 2006) expand the domain of Agree even to both functional and lexical heads for as long as they undergo Merge, the operation that combines syntactic objects and builds phrases. Nevertheless, we do not need an expansive definition of Agree to be able to reduce c-command and Spec–Head relations to Agree. Miyagawa’s conception of Agree includes these relations under Agree. The NPI licensing process involves a lexical category of the NPI, and the functional category of negation or other licensors. The operation Agree establishes the functional relation between the class of functional licensors and the NPI. The class of functional licensors can include nonveridical operators other than negation such as the antimorphic without and before clauses, the monotone-decreasing contexts such as few and seldom, and the context of comparatives. These nonveridical operators can project a syntactic phrase and establish a functional relation with NPIs (see Chatzopoulou, 2018, for a proposal arguing for a L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 99 ]
01
Nonveridicality Phrase in Greek). I discuss the nonveridical licensors in Arabic in the next section. Later in c hapter 4, I develop an Agree-based account for NCI licensing in Arabic. 3.5. LICENSING ARABIC PSIS IN NONVERIDICAL CONTEXTS
Thus far in this chapter, I have argued that PSIs are licensed via c-command or Spec–Head relations. In this section, I will expand this list of licensing conditions to include nonveridical licensors such as the antimorphic without and before clauses, the monotone-decreasing contexts such as few and seldom, and the context of comparatives. Consider, for instance, the Greek example with the NPI kanenan ‘n-person’ in the context of the antimorphic without and the monotone decreasing few clauses: (77) a.
b.
(Greek; Giannakidou, & Zeijlstra, 2017, p. 17) xoris na dhi {kanenan/KANENAN}. without subj see. 3sg n-person ‘without having seen anybody.’ (Greek; Giannakidou, & Yoon, 2016, p. 533) Elaxisti fitites ipan very.few students said.3pl ‘Very few students said anything.’
tipota. NPI.thing
Other licensing environments for PSIs in Arabic include contexts that contain without and before: (78) a.
saafar biduun maa yxabbir left.3msg.pfv-neg without comp tell.3msg.ipfv ‘He left without telling anybody.’
b.
ʕali mišii min ɣeer Ali left.3msg.pfv without yitkallim maʕ ʔay/wala waaħid talk.3msg.ipfv with any/no one ‘Ali left without talking to anyone.’
c.
saafər blaa left.3msg.pfv-neg without ygulhaa ħtta l-ši tell.3msg.ipfv NCI to-any ‘He left without telling anybody.’
[ 100 ] A Unified Theory of Polarity Sensitivity
maa comp waħəd person
(MA)
ʔay/ wala ħada (JA) any/no one
maa comp
(EA)
(Soltan, 2014b)
1 0
(79) a.
maat gabl maa yšuuf died. 3msg.pfv before comp see.3msg.ipfv ʔay /wala ħada min ʔaħfaad-o Any/no one of grandchildren-his ‘He died before seeing any of his grandchildren.’
(JA)
b.
ʔabuu-haa maat ʔabl maa yi-šuuf (EA) father-her died.3msg before comp see.3msg ʔayy/wala waaħid min ʔaħfaad-u-h any/no one from grandchildren-EV-his ‘Her father died before seeing any of his grandchildren.’ (Soltan, 2014b)
c.
maat qbəl maa yšuuf died.3msg.pfv before comp see.3msg.ipfv ħtta ši waħəd mən ħfaayd-o NCI any person of grandchildren-his ‘He died before seeing any of his grandchildren.’
(MA)
The semantic approaches to licensing PSIs seek a unified account for the licensing of NPIs and NCIs. The primary account put forward to explain such distribution is Ladusaw’s (1980) monotonicity approach, i.e., Downward Entailment, which is defined as follows. (80) Ladusaw’s (1980) licensing condition: α is a trigger for negative polarity items in its scope iff α is downward entailing A function f is downward-entailing iff for arbitrary elements X, Y it holds that: X ⊆ Y → f(Y) ⊆ f(X).
Downward-Entailing (DE) operators include the PSI licensors we have discussed thus far. The semantic function DE unifies the licensing operators, negation, and the without and before clauses. Within the category of PSIs, licensors for NPIs and NCIs in Arabic differ. The following environment only licenses NPIs and not NCIs. Specifically, NCIs cannot occur in the context of few: (81) a.
galiiliin illi bi-lɣarb biʕrifuu few.pl who in-the-west know ʔay/*wala ʔiši ʕan l-islaam any /no thing about the-islam ‘Few people know anything about Islam.’
(JA)
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 101 ]
2 0 1
b.
naas ʔulayyil-iin fii il- ɣarb bi-yi-ʕraf-uu (EA) people few-pl in the-West asp-ipfv-know-3pl ʔayy/*wala ħaagah ʕan il-islaam any/no thing about the-Islam ‘Few people in the West know anything about Islam.’ (Soltan, 2014b)
c.
qlaal f-lɣarb lli few in-the-west who ši ħaʒa /*ħtta ħaʒa ʕla l-islaam some thing /NCI thing about the-islam ‘Few people know anything about Islam.’
kay-ʕərfuu asp-know
(MA)
These contrasts between NPIs and NCIs in regard to licensing contexts indicate that NPIs have a wider distribution than NCIs. Similar facts in other languages motivated Zwarts (1993) and van der Wouden (1997) to propose a distinction between different types (or degrees) of DE. The proposal distinguishes between three types of DE functions: (82) a.
monotone decreasing (few, seldom, hardly)
f(X)∪f(Y)⊆ f(X∩Y)
b.
antiadditives (nobody, never, nothing)
f(X)∪f(Y)⊆ f(X∩Y)
c.
antimorphics (not, sentential negation, without)
f(X∩Y)⊆ f(X)∪f(Y)
However, the problem with this analysis is that it requires licensing environments to have a DE operator. This is not the case. Such environments instead include comparative adjectives: (83) a.
b.
ʔaħmad ʔaðˁaf min inn-o y-guul (JA) Ahmad weaker than comp-him say ʔay/*wala ʔiši l-al-mudiir any/no thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’ ʔaħmad ʔadˁʕaf min inn-o y-ʔuul (EA) Ahmad weaker than comp-him say ʔayy/*wala ħaaga li-l-mudiir any/no thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’ (Soltan, 2014b)
[ 102 ] A Unified Theory of Polarity Sensitivity
3 0 1
c.
ʔaħmad dˁʕiif bezzaf baš (annah-o yqdar) Ahmad weaker a lot than comp-him ši ħaʒa/*ħtta ħaʒa ll-mudiir some thing/NCI thing to-the-manager ‘Ahmad is too weak to say anything to the manager.’
y-guul (MA) say
Due to these non-DE contexts (including comparatives, imperatives, interrogatives, and conditionals), Giannakidou-Zwarts (1999) built their theory of polarity items based partly on DE operators and formalized it using a more general notion called “nonveridicality,” as in the Venn diagram, figure 3.1. This theory shows that NPIs that are licensed in nonnegative contexts are also licensed in negative contexts, but not the other way around. Concurring with Soltan’s (2014b) analysis of EA, I contend that for JA and SA, Giannakidou’s theory of nonveridicality (1997, 1998, 2009) fares better than Ladusaw’s (1979) monotonicity approach in explaining these contrasts. The semantic notion of nonveridicality refers to the idea that the proposition of the sentence does not ensure the truth in PSI licensing contexts (Giannakidou, 1998, 2009; Giannakidou & Zeijlstra, 2017). (84) a.
A propositional operator F is veridical iff Fp entails or presupposes that p is true in some individual’s epistemic model ME(x); otherwise F is nonveridical.
b.
A nonveridical operator F is antiveridical iff Fp entails that not p in some individual’s epistemic model: Fp → ¬ p in some ME(x).
Nonveridical Downward entailing (minimally negative) Antiadditive (classically negative)
Antimorphic
Figure 3.1. Giannakidou-Zwarts’s (1999) theory of polarity items.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 103 ]
4 0 1
The notion of nonveridicality unifies the DE contexts and the non-DE contexts under the notion that the truth of the proposition cannot be certain. In other words, none of these contexts has a truth value. Nonetheless, the explanatory adequacy of this semantic approach to polarity sensitivity should not lead us to think that the conceptual intentional interface can handle all issues with licensing polarity items. Recall that syntactic configurations, c-command and Spec–Head relations, establish the functional relations between the NPI and its licensor, relations that can be viewed as Agree operation à la (Miyagawa, 2010). Moreover, we have seen the magnitude of the syntactic role in licensing when we discussed the epiphenomenon of mutual exclusivity between the NPI ʕumr ‘ever’ and bipartite negation. This NPI is merged higher than TP, and only higher negation (above TP) can license it through Spec-Head or c-command. We also saw how NPIs that move from a postverbal/post-negative position into a preverbal/pre-negative position such as the idiomatic NPI girš/ fils ʔiħmar ‘penny/red cent’ reconstruct at LF. Crucially, the licensing of such NPIs fails, and they crash at LF when they are extracted from syntactic islands. In short, syntax has a role in licensing NPIs through the basic syntactic operations of Merge, Move, and Agree, as well as c-command and Spec-Head configurations. Moreover, this semantic approach to polarity sensitivity cannot handle key contrasts in the distribution of NPIs and NCIs. Recall that unlike NPI ʔay, the NCI wala fails to occur in embedded clauses where a perfective-preventing predicate appears in the main clause. (85) a.
b.
manaʕ-ha tištari ʔay/ *wala prevented.3msg-her buy.2sg npi/ nci ‘He prevented her from buying anything.’
ʔiši thing
(JA)
mənaʕ-ha təšri ʔay ħaʒa /*ħtta ši ħaʒa prevented.3msg-her buy. 2msg NPI.thing/NCI some thing ‘He prevented her from buying anything.’
(MA)
In contrast, the same NCI can occur in the embedded clause of a participle form of preventing predicate: (86) a.
b.
mamnuuʕ tištari ʔay/ wala part.prevented buy.2sg NPI/ NCI ‘It is disallowed to buy anything.’
ʔiši thing
(JA)
mamnuuʕ tašri ši ħaʒa / ħtta ši ħaʒa part.prevented buy.2sg some thing/NCI some thing ‘It is disallowed to buy anything.’
[ 104 ] A Unified Theory of Polarity Sensitivity
(MA)
5 0 1
Clearly, the semantic notion of antiveridicality cannot explain this conspicuous behavior of these polarity items. In chapter 4, I argue for a syntactic agreement account for licensing Arabic NCIs. There, I argue for a modified and fine-grained analysis that unearths the simple invisible formal features and syntactic operations behind the complex visible landscape of Arabic polarity. One final issue to discuss in this chapter is in regard to the disjunction coordinators ʔaw ‘free choice or’ and walla ‘exclusive choice or.’ These two disjunction coordinators are somewhat parallel to the determiner NPI ʔay ‘any,’ which can also have a free choice or polarity interpretation. The similarity is clear in cases where walla ‘exclusive choice or’ has a distribution limited to a subset of polarity environments: questions and conditionals. (87) a.
b.
ʔišrab gahwa drink.imp coffee ‘Drink coffee or tea.’
ʔaw or
tišrab gahwa walla drink.sbjv coffee or ‘Would you like to drink coffee or tea?’
šaay tea
šaay? tea
(88) miš mitʔakkid ʔiða bitħib gahwa neg sure whether like.3fsg coffee ‘I am not sure whether she likes coffee or tea?’
walla or
šaay? tea
The distribution of the disjunction coordinator walla ‘exclusive choice or’ is more limited than the determiner NPI ʔay ‘any’ since negation can license the latter but not the former. (89) a.
b.
ma-štarati-š ʔay neg -bought.3fsg- neg NPI-any ‘She did not buy any coffee.’ ma-štarati-š gahwa neg -bought.3fsg-neg coffee ‘She did not buy coffee or tea.’
gahwa coffee
ʔaw/*walla or
šaay tea
By extension, we expect the coordinator walla ‘exclusive choice or’ to pattern with polarity items. Instead, we find that it patterns with free choice items such as any which cannot co-occur with negation. It is not clear why this polarity coordinator behaves in this unpredictable manner, and the question of how to correctly classify and analyze it remains open.
L i c e n s i n g N e g at i v e P ol a r i t y I t e m s
[ 105 ]
6 0 1
7 0 1
CHAPTER 4
Licensing Negative Concord Items
4.1. NCI LICENSING
This chapter discusses the syntactic configurations and processes that are involved in licensing various types of NCIs (negative concord items) in Arabic. Here, I argue for a syntactic agreement approach to explain the behavior of Arabic NCIs, including the coordinate complexes laa-wala ‘neither-nor’ and their interaction with other PSIs (polarity-sensitive items). In addition, I propose that an abstract negative operator explains the distribution of non- strict NCIs. The analysis relies on an inventory of abstract negation features assigned to the various types of Arabic NCIs, thus partially building on but departing from Zeijlstra’s (2004, 2008) analysis for Romance languages and Penka’s (2011) analysis for Germanic languages. The analysis also challenges the claim that negative indefinites (NIs) are never inherently negative (cf. Penka, 2011, for Germanic and other European languages). The rich landscape of Arabic polarity sensitivity informs the novel analysis in this chapter, which also applies to polarity coordinate complexes. I also argue that the coordinators laa-wala ‘neither-nor’ are conspicuous analogs of the NCI wala, in that it can co-occur with the negative adverb maʕumrhiš ‘never’ in the CP (complementizer phrase) layers, resulting in a concordant reading. I argue for a revised syntactic agreement account that involves the insertion of a covert negative operator to license non-strict NCIs, including the coordinators laa- wala ‘neither-nor,’ as a last resort in the absence of overt licensors. For NCIs, there are two main questions to be discussed in this chapter. First, the licensing questions pertaining to NPIs carry over to NCIs. What structural configuration and level of representation mediate the licensing of NCIs in postverbal and preverbal positions? From this, a further sub-question emerges as to whether preverbal adverbial NCIs are licensed overtly in a local A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0004.
8 0 1
domain to the left of negation or covertly in their interpretation site to the right of negation; i.e., where negation scopes over them. (1)
ʔabadan/ bilmarrah Faris ever/ at all Faris ‘Faris doesn’t like fish at all.’
ma-biħibb-iš neg-like.3MSG-neg
l-samak the-fish
The second question at issue here concerns how determiner NCIs in preverbal position and in fragment answers have a negative interpretation despite the absence of any negative marker (NM). (2)
a.
*(ma)- ʔaʤaa-š neg-came.3MSG-neg ‘No one came.’
b.
wala-ħada ma-ʔaʤaa-š NCI-one neg-came.3MSG-neg ‘Nobody didn’t come.’ #‘Nobody came.’
c.
Question: Answer: miin ʔaʤa ? wala-ħada who came.3MSG NCI-one ‘Who came?’ ‘no one’
wala-ħada NCI-one
(JA)
(JA)
(JA)
The answer to this second question can be relegated to either the use of two homophonous lexical items: an NQ (negative quantifier) and an NPI (negative polarity item), i.e., the lexical ambiguity approach, or a covert negative operator that licenses preverbal NCIs. This chapter argues for a covert licensing approach, putting forward an analysis that introduces a covert negative operator and appeals to the economy condition. As with NPIs, the syntactic dependency between NCIs and negation can be established via c-command and Spec-Head configurations (see Benmamoun, 1997, 2006, for MA [Moroccan Arabic], and Alqassas, 2015, 2016, for JA [Jordanian Arabic]). All lexical categories of NCIs can occupy a post-negative position: (3)
ma-ʔaʤa-š neg-came.3MSG.-neg ‘No one came.’
wala-ħada NCI-one
[ 108 ] A Unified Theory of Polarity Sensitivity
(JA)
9 0 1
The NCI wala-ħada follows negation and occupies the postverbal subject position. This NCI is under the scope of negation. The bipartite NM ma . . . š c- commands the NCI as in the following simplified illustration: (4)
The C-command Configuration: TP Spec
T T
NegP Neg
VP
ma …š
Spec
V
NCI
C-COMMAND
Since all lexical categories of NCIs can be post-negative, and consequently postverbal, their dependency with negation is arguably due to c-command. However, preverbal NCIs display a complex behavior when interacting with negation. As such, I will argue that there are other mechanisms used establish the dependency between NCIs and negation. As with NPIs, these mechanisms vary according to the lexical category and the Arabic variety of the NCI. In MA, the determiner NCI hetta must co-occur with negation when postverbal. Benmamoun (1996, 1997) argues that the NCI, similar to the NPI ħada, is licensed by negation under c-command. (5)
*(ma)-ʒa ħətta wəld (MA) neg-came.3MSG NCI boy *‘Any boy didn’t come’ (Benmamoun, 1996, p. 49)
This differs from his analysis of the preverbal NCI ħətta waħəd in MA. This NCI, Benmamoun argues, is licensed under a Spec–Head relationship with negation (Benmamoun, 1997). (6)
ħətta waħəd ma-ʒa (MA) even one neg-came.3MSG *‘Anyone didn’t come’ (Benmamoun, 1997, p. 273)
Recall that constituents in a Spec-Head configuration have a locality relation. This means that a constituent like the subject Salwa, in the following
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 109 ]
0 1
example, cannot intervene between the NCI and negation. Benmamoun’s examples corroborate this claim (Benmamoun, 1997, p. 281): (7)
a.
ħətta ktab ma-qrat-u even book neg-read.3FSG-it ‘Salwa didn’t read any book.’
b.
*ħətta ktab səlwa ma-qrat-u even book Salwa neg-read.3FSG-it *‘Any book, Salwa did not read it.’
səlwa Salwa
(MA)
The so-called ‘never-words’ in JA—adverbial NCIs—can be focus fronted to a pre-negative position. When in pre-negative positon, these NCIs can also be separated from negation by constituents such as the subject in (8)b. Arguably, such cases involve reconstruction at LF (see Alqassas, 2015, and references therein for Arabic; Giannakidou, 1998, of Greek). (8)
ʔabadan/ ever/
a.
Faris ma-biħibb-iš Faris neg-like.3MSG-neg ‘Faris doesn’t like fish at all.’
l-samak the-fish
b.
ʔabadan/ bilmarrah Faris ever/ at all Faris ‘Faris doesn’t like fish at all.’
ma-biħibb-iš neg-like.3MSG-neg
bilmarrah at all
l-samak the-fish
(JA)
(JA)
These examples show that the NCI can be focus fronted to a position preceding the subject. Clearly, the NCI precedes negation in this position; hence, it is not within its c-command domain. In this position, the NCI and negation are also not within a local domain, such as the Spec-Head configuration. This is clear from the fact that the subject intervenes between the NCI and negation. Arguably, instead of a c-command and Spec–Head relation, these NCIs establish dependency by reconstruction to their postverbal position where they are c-commanded by negation. At LF (Logical Form), then, the focus-fronted NCI reconstructs to its base-generated position following the verb: (9)
[CP [ abadan/ bilmarrah]i [NegP ever/ at all
Faris Faris
NEG-like.3MSG-NEG
l-samak [NCI]i]] the-fish
Finally, the adverbial NCI ‘baʕd’ in JA can also be pre-negative. Therefore, like the ‘never- words’ discussed previously, this NCI’s dependency with
[ 110 ] A Unified Theory of Polarity Sensitivity
1
negation is derived via reconstruction to a postverbal position that is c- commanded by negation (Alqassas, 2016). (10) a.
b.
ma-saafar-iš neg-traveled.3MSG-neg ‘He hasn’t traveled yet.’
baʕdo yet-him
(JA)
baʕd-o ma-saafar-iš yet-him neg-traveled.3MSG-neg ‘He hasn’t traveled yet.’
In summary, NCIs in post-negative position are licensed under c-command (i.e., they are under the scope of negation). In pre-negative position, nominal NPIs and MA determiner NCIs are licensed under the specifier–head relation (i.e., the local domain of negation), while idiomatic NPIs and JA adverbial NCIs reconstruct to a post-negative position (i.e., they are interpreted in a post-negative position), where they are c-commanded by negation. Lastly, the JA and EA (Egyptian Arabic) determiner NCIs in preverbal position do not co-occur with negation, suggesting that there is no dependency relation with negation. Turning to preverbal determiner NCIs in JA and EA, the following data shows that these NCIs occur without negation. When they co-occur with negation, the sentence has a double negation (DN) reading, as in (11)b, d. (11) a.
wala-ħada ʔaʤa NCI-one came.3MSG ‘No one came.’
(JA)
b.
wala-ħada ma-ʔaʤaa-š NCI-one neg-came.3MSG ‘No one didn’t come.’ #‘No one came.’
c.
walaa waaħid gih (EA) no one came.3msg ‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
d.
walaa waaħid maa-gaa-š (EA) no one neg-came.3msg-neg ‘Nobody didn’t come.’ #‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
(JA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 111 ]
2 1
The challenge in explaining the behavior of these preverbal NCIs is in finding a unified analysis that can explain their behavior in postverbal (and post-negative) position as well as preverbal (pre-negative position). Such cases of non-strict NCIs are partly behind the many approaches advanced to explain the syntax of PSIs. Earlier syntactic analyses of PSIs, which treated NPIs and NCIs as a single class, introduced a range of theoretical licensing mechanisms in an attempt to consolidate the behavior of these two types of elements. The major syntactic accounts are the NPI approach, the NQ approach, and the lexical ambiguity approach. The following subsections will introduce each of these analyses for historical perspective while a more recent syntactic account of NCIs, which focuses on syntactic agreement, will be introduced and discussed in section 4.1.4.
4.1.1. The NPI approach
The NPI approach treats NCIs as NPIs that do not contribute a negative reading to the sentence (Laka, 1990; Ladusaw, 1992). As such, this approach analyzes NCIs as ‘non-strict’ NPIs that do not always require the presence of a sentential NM. Non-strict NPIs, such as preverbal nominal NPIs, are licensed by an ‘empty negative morpheme’ heading a functional projection ƩP and move to a preverbal position (the specifier of ƩP) to identify this null negative morpheme preceding the verb (Spanish: Laka, 1990, p. 107). (12) a.
b.
*(No) vino not came ‘Nobody came.’ Nadie (*no) nobody no ‘Nobody came.’
nadie. (Spanish) anybody
vino. came
(Spanish)
The main challenge for this approach to NCIs becomes apparent in a cross- language evaluation of NCIs. In some languages, preverbal (and prenegative) NCIs cannot be licensed by covert negation (Benmamoun, 1997, p. 273): (13) ħətta waħəd *(ma)-ʤa NCI one neg-come.3msg.pfv ‘No one came’
[ 112 ] A Unified Theory of Polarity Sensitivity
(MA)
3 1
In this MA example, we see that the NCI ħətta waħəd requires the NM when the NCI is preverbal. NCIs can be licensed by a covert negative operator. Otherwise, the account incorrectly predicts that the NCI can be licensed without the NM.
4.1.2. The NQ approach
An alternative approach to the NPI approach treats NCIs as inherently NQs that must be licensed in the specifier of a NegP (a requirement known as the ‘neg-Criterion,’ see Zanuttini, 1991; Haegeman & Zanuttini, 1996). This analysis essentially treats NCIs as negative operators and is useful in accounting for the ability of NCIs to occur without negation preverbally and to stand alone as negative fragment answers (since NCIs are inherently negative, there is no need for a, NM ). The only requirement for NCIs in this account is that they occur in a Spec–Head relation with Neg0. While this analysis straightforwardly accounts for (some of) the distribution of NCIs, it is unable to explain why the NCI, as an inherently negative element, requires an NM when postverbal. Moreover, this approach predicts that the combination of the postverbal NCI and NM should result in a DN reading. However, this is not borne out as shown in the previous section. To explain the co-occurrence of postverbal NCIs and the NM, Zanuttini (1991) proposes that postverbal NCIs cannot move to Spec/NegP at LF—a licensing requirement—because NegP is above TP, which acts as a barrier. When an NM heads the NegP, however, TP’s barrierhood is circumvented. To account for the failure of the NM + NCI to yield a DN reading, Haegeman and Zanuttini (1996, p. 139) posit an LF process called neg-factorization: (14) [∀x¬][¬] = [∀x]¬
This process guarantees that the two instances of negation are interpreted as one single negation and yield a concord reading. However, the explanations posited here do not elaborate on why the NM cannot co-occur with preverbal NCIs (Zeijlstra, 2004; Penka, 2011).
4.1.3. The lexical ambiguity approach
The third and final approach type of syntactic analysis of PSIs that has been previously posited is the Lexical Ambiguity Approach. At its core, this approach assumes that the asymmetric behavior of non-strict NCIs is due to the lexical ambiguity of these words (Herbuger, 2001). Specifically, postverbal
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 113 ]
4 1
NCIs require the NM because they are NPIs, whereas preverbal NCIs do not require the NM because they are NQs. In other words, this approach argues that NPIs and NQs are distinct but homophonous lexical items in non-strict NC (negative concord) languages. Herburger supports this analysis with data showing that a preverbal NCI can license a postverbal NCI without inducing a DN reading (a phenomenon known as “negative spread”). The following is an illustrative example from Spanish: (15) Nadie comió NCI-person ate ‘Nobody ate anything’
nada NCI-thing
The preverbal NCI in (15) functions as a negative quantifier. As such, it produces a sentential negation reading despite the absence of an NM in the sentence. The postverbal NCI, on the other hand, functions as an NPI since it does not contribute a negative reading. If it did contribute a negative reading, the sentence would exhibit DN, which is not the case. Ouali and Soltan (2014) posit two homophonous lexical items for determiner NCIs in EA. One item is an inherently negative NCI-wala that is similar to the negative quantifier no one. This item occurs in pre-negative/preverbal position: (16) a.
walaa waaħid gih (EA) no one came.3msg ‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
b.
maa-gaa-š walaa waaħid (EA) neg-came.3msg-neg no one ‘Nobody came.’ (Ouali & Soltan, 2014, p. 164)
The other item is an NCI-wala and is not inherently negative but inherits its negative interpretation from the NM. This item occurs in a postverbal position. Crucially, under the lexical ambiguity analysis, there is no dependency relation between the preverbal NCI-wala and negation in the examples in (11). This explains why the NCI can occur without negation in (11)a, c, and it predicts a DN reading in (11)b, d. Instead, a dependency relation is defined between the postverbal NCI-wala and negation in the JA example in (3) and its equivalent in EA. This relation is most likely established via c-command, as explained previously. In order to advance this approach to PSIs, proponents of this analysis must explain why the NPI-type NCI cannot be preverbal, while the NQ-type NCI
[ 114 ] A Unified Theory of Polarity Sensitivity
5 1
cannot be postverbal. Herburger reduces these distributional constraints to independent principles for Spanish NCIs. In her analysis, she argues that NPI-type NCIs cannot be preverbal because regular Spanish NPIs cannot be preverbal. Conversely, Herburger derives the postverbal position of NQ-type NCIs pragmatically by arguing that a preverbal NQ structure would assert the existence of an event without providing a theme. In fact, Herburger contends that such sentences are possible under the right conditions, as in the following example: (17) temen que el bebe sen aitista se pasa (Spanish) fear that the baby is.sbjv autistic clitic spends el timpo mirando at nada the time looking at nothing ‘They fear the baby is autistic. He spends his time looking at nothing’
Herburger gives the following representation for this sentence: (18) e
[look(e) and Agent(baby,e) and [thing(x) and Theme(x,e)]]
This analysis faces several challenges. First, regular NPIs can be licensed in interrogative and conditional environments. If postverbal NCIs were, indeed, NPIs, we would incorrectly predict that they, too, should be licensable in nonnegative contexts like interrogative and conditional sentences. Second, if preverbal NCIs were NQs, we predict that preverbal MA NCI hətta-phrases should not require the NM. See section 4.1 for evidence to the contrary. In the previous sections, I have provided a brief overview of the previous syntactic approaches to NCIs, discussing both their advantages and disadvantages in accounting for the array of data at hand. In the next section, I will argue for one further approach to NCI licensing that relies on syntactic agreement. It will be shown that even preverbal NCIs in JA and EA are not inherently negative and they fail to contribute negative interpretation when they co-occur with the NIs maħadaaš ‘nobody’ and maʕumrhuuš ‘never.’ The analysis captures such distributional facts and the split scope contrasts between the NCI wala-ħada ‘nobody’ and these NIs.
4.1.4. NCI analyses in Arabic: Syntactic agreement approach
As with NPIs, the syntactic dependency between NCIs and negation can be established by licensing relations via structural configurations. In the previous chapter, these configurations were shown to be defined through c-command
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 115 ]
6 1
and specifier–head relations as also argued in Benmamoun (1996, 1997, 2006) and Ouali and Soltan (2014). The syntactic dependency between determiner NCIs and negation in MA is instantiated as a requirement that is dependent upon the NCI’s position. Postverbal determiner NCIs must be c-commanded by negation and, when preverbal, must be in a Spec–Head relationship with negation (Benmamoun, 1997). In addition, the NCI hetta in MA must co-occur with negation when postverbal and, being similar to the NPI ħada, is licensed by negation under c-command (Benmamoun, 1996, 1997). (19) *(ma)-ʤa neg-come.3msg.pfv *‘Any boy didn’t come’ (Benmamoun, 1996, p. 49)
ħətta wəld NCI boy
(MA)
As with the NPI subject in (17), the NCI in (19) is the subject of the sentence. It is merged in Spec-VP and moves to Spec-TP. Below this, the NM is the head of a NegP projection on top of the VP (verb phrase). The verb moves to the negative head and forms a complex V + neg, and they both move through T and land in a functional projection above TP, placing negation in a position c-commanding the NCI subject in Spec-TP. (20) FP Spec
F F
TP Spec
T T
must > ∃ b. ‘There is no professor who is required to be present. ¬ ∃ > must (52) maħadaaš laazim ykuun mawʤuud bi-li-mtiħaan no one should be present in-the-exam a. *‘It is not required that there be anyone present.’ *¬ > must > ∃ b. ‘There is no one who is required to be present. ¬ ∃ > must
(JA)
(JA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 129 ]
0 3 1
(53) wala ʔustaaz laazim yikuun mawguud fi-li-mtiħaan NCI teacher.msg should be present in-the-exam a. ‘It is not required that there be a professor present.’ ¬ > must > ∃ b. ‘There is no professor who is required to be present. ¬ ∃ > must
(EA)
(54) maħaddiš laazim yikuun mawguud fi-li-mtiħaan (EA) no one should be present in-the-exam a. *‘It is not required that there be anyone present.’ *¬ > must > ∃ b. ‘There is no one who is required to be present. ¬ ∃ > must (55) laa ʔaħada yaʤibu ʔan yakuuna mawʤuudan fi-l-ʔimtiħaan (SA) no one should be present in-the-exam a. *‘It is not required that there be anyone present.’ *¬ > must > ∃ b. ‘There is no one who is required to be present. ¬ ∃ > must
SA grammars (e.g., Raajihi, 2000) report interesting contrasts with the compound laa-NPs that are dependent on the case of the NP (noun phrase). The indefinite noun can receive either nominative or accusative case in a coordination structure including laa-NP indefinites. Either DP (determiner phrase) of coordination complexes can be nominative (default case) or accusative (case from the coordinators), with contrasts in the semantic interpretation associated with each case (Raajihi, 2000). These semantic contrasts suggest but do not clearly show a split scope reading with the nominative indefinite noun laa raʤul-un, which is not a compound, and an NI reading with the accusative indefinite compound laa-raʤul-a. (56) laa raʤul-un wala ʔimraʔat-un not man-nom.indef. nor woman-nom.indef ‘There isn’t a man or a woman at home.’ ¬ > Be > ∃
fii al-bayt at the-house
(57) laa raʤul-a wala ʔimraʔat-a fii al-bayt neither man-acc nor woman-acc at the-house ‘No man or woman is at home.’ ¬ ∃ > Be
(SA)
(SA)
To show the split scope contrasts clearly, let’s look at the following contrasts. Both of the sentences below state that there isn’t a man at home. Crucially, however, the constituent laa raʤul-un can be followed by a rectification specifying that there are men at home. The incompatibility of rectification with the accusative indefinite laa-raʤul-a suggests that it is inherently negative and cannot be used to reject part of the proposition of the sentence. [ 130 ] A Unified Theory of Polarity Sensitivity
1 3
(58) a.
b.
laa raʤul-un fii al-bayt, bal riʤaal not man-nom.indef. at the-house, but men ‘There isn’t a man at home, but a group of men.’ laa raʤul-a fii al-bayt, (*bal not man-acc at the-house, (but ‘No man is at home, but a group of men.’
riʤaal) men)
(SA)
(SA)
Given these contrasts and building on the pattern of split scope readings, we expect the nominative indefinite noun laa raʤul-un rather than the accusative indefinite laa-raʤul-a to allow the split scope reading in contexts that contain the modal yaʤib ‘must,’ as follows: (59) laa raʤul-un yaʤibu ʔan yakuuna fii al-bayt no man-nom.indef. must comp cop at the-house a. ‘It is not required that there be a man at home.’ ¬ > must > ∃ b. ‘There is no man who is required to be at home. ¬ ∃ > must (60) laa raʤul-a yaʤibu ʔan yakuuna fii al-bayt no man-acc must comp cop at the-house a. *‘It is not required that there be a man at home.’ * ¬ > must > ∃ b. ‘There is no man who is required to be at home. ¬ ∃ > must
(SA)
(SA)
These split scope contrasts are supported by the fact that the nominative indefinite noun can occur postverbally following the modal and its CP complement and thus be split from the negator laa, while the accusative indefinite cannot: (61) laa yaʤibu ʔan yakuuna fii al-bayt raʤul-un (SA) no must comp cop at the-house man-nom.indef. ‘It is not required that there be a man at home.’ ¬ > must > ∃ (62) *laa no
yaʤibu must
ʔan comp
yakuuna cop
fii al-bayt at the-house
raʤul-a (SA) man-acc
Under the analysis developed in this monograph, NIs that do not display split scope readings are inherently negative and not licensed by covert or overt negation. On the other hand, NIs with split scope readings are not inherently negative and are licensed by a covert/overt negative constituent. These NIs may also be licensed by a covert negative operator as last resort when no NM or inherently negative NI is available in their syntactic domain (c-command domain). L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 131 ]
2 3 1
The agreement analysis advocated in this book has implications for Multiple Agree phenomena in strict and non-strict NC languages. In non-strict NC languages, Multiple Agree is obligatory, resulting in an obligatory concordant reading. On the other hand, Multiple Agree is optional in strict NC languages, yielding either an NC or a DN reading. Arabic partially diverges from this taxonomy. First, Arabic dialects such as JA can have both strict and non-strict NCIs. The NCI wala-NP is non-strict while adverbial NCIs are strict: (63) Non-Strict NCIs a. *(ma)-ʔaʤaa-š neg-came.3msg-neg ‘No one came.’ b.
wala-ħada NCI-one
(JA)
wala-ħada ma-ʔaʤaa-š NCI-one neg-came.3msg-neg ‘Nobody didn’t come.’ #‘Nobody came.’
(JA)
(64) Strict NCIs a. Maha ma-bitħibb-iš l-safar ʔabadan/ bilmarrah Maha neg-like.3fsg-neg the-traveling ever/at all ‘Maha doesn’t like traveling at all.’ b.
(JA)
ʔabadan/ bilmarrah Maha ma-bitħibb-iš l-safar (JA) ever/ at all Maha neg-like.3fsg-neg the-traveling ‘Maha doesn’t like traveling at all.’
Arabic dialects with non-strict-NCIs like JA do not have Multiple Agree with non-strict NCIs such as wala-NP, thus behaving like the DN languages: German, Dutch, and Scandinavian.7
7. Nonetheless, strict NCIs like the never-type NCIs ʔabadan/bilmarra ‘never’ in JA can co-occur with non-strict NCIs: (i)
a.
wala-ħada biħib l-safar NCI-one like.3Msg the-traveling ‘Nobody likes traveling at all.’
ʔabadan/ bilmarrah NCI-ever/ at all
It is, however, still the case that JA patterns with DN languages in that a non-strict NCI fails to license another non-strict NCI it c-commands. This is unexpected given that JA diverges from other non-stirct NC languages like Italian. [ 132 ] A Unified Theory of Polarity Sensitivity
31
(65) a.
b.
wala mʕalme rassabat wala tˁaalib (JA) NCI teacher. fsg fail.3fsg.pfv NCI student *‘No teacher failed any student.’ [*NC] ?‘No teacher failed no student.’ (= ‘Every teacher failed a student’) [DN] wala mʕalme rassabat ʔayy tˁaalib NCI teacher.fsg fail.3fsg.pfv NPI-any student ‘No teacher failed any student.’ *‘No teacher failed no student.’ (= ‘Every teacher failed a student’)
(JA) [NC] [*DN]
MA is distinctive from JA since the former is a strict-NC language. As such, we expect the language to pattern more so like Romanian, with Multiple Agree being optional. This is borne out (original examples from Ouhalla, 2002, p. 307, do not show the ungrammatical DN readings): (66) a.
b.
ma-ʕtat Nadia ħətta-ħaja neg-give. 3fsg.pfv Nadia NCI-thing ‘Nadia did not give anything to anybody.’ *‘Nadia did not give nothing to nobody.’
l-ħətta-waħəd. to-NCI-person
(MA) [NC] [*DN]
ħətta waħəd ma-šaf ħətta ħaja (f-ħətta-blasa) NCI-person neg-see. 3fsg.pfv NCI-thing (in-NCI-place) ‘Nobody saw anything (anywhere).’ *‘Nobody saw nothing (nowhere).’ (= Everybody saw something’)
(MA) [NC] [*DN]
The explanatory adequacy of this agreement-based analysis for NC in Arabic can be tested by looking at the predictability of contrasts between JA and MA. The analysis predicts that the combination of an NI and the NM ma should result in a DN reading in JA and an NC reading in MA. This is, indeed, borne out (MA examples from Benmamoun, 2006, p. 145): (67) maʕumrhuuš[iNEG] maa[iNEG] ħabb never neg like.3msg.pfv *‘He has never liked coffee.’ ‘He has never not liked coffee.’ (68) a.
l-gahwa def-coffee
ma ʕəmmər nadya ma-ʒat never Nadia neg-come.3fsg.pfv ‘Nadia never came.’ *‘Nadia has never not come.’
(JA) [*NC] [DN]
(MA) [NC] [*DN]
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 133 ]
4 3 1
b.
ma ʕəmmər nadya ʒat never Nadia come.3fsg.pfv ‘Nadia never came.’ *‘Nadia has never not come.’
(MA) [NC] [*DN]
The results of this study, then, show that the typology of negative features on Arabic NIs is more diverse that what is reported for European languages in Penka (2011). The following table 4.2 shows the rich variation in the distribution of the negative features in Arabic NMs and NIs, as well as the variation in multiple agreement. As such, I propose the following (table 4.3) as a revision of Penka’s (2011) table. In it, Arabic has been included with new possible variants in the typology of negative features: The results of the micro-comparative approach unearthed new types of languages that behave differently with respect to the feature structure of NMs, NIs, and Multiple Agree. There is more variation than previously found, thanks to microvariation in Arabic. There are NIs that express negated existential quantification, pace Penka (2011). The NIs maħadaaš ‘nobody’ and maʕumrhuuš ‘never’ are inherently negative while walaħada/walawaaħid ‘nobody’ are not. The evaluation of data in this chapter unearthed a correlation between NC languages with true NIs and DN languages. Multiple Agree is not available in DN languages, or NC languages with true NIs. The correlation between the availability of Multiple Agree and DN languages is no longer exclusive, thanks to languages like JA and EA, which are largely NC languages but have NIs that are similar to those in DN languages like English. As such, languages that are partially or fully DN lack Multiple Agree, or, put differently, languages that have NIs that can occur without NMs carrying [ineg] do not have Multiple Agree.
Table 4.2. VARIATION IN THE FEATURES ON NMS, NIS, AND MULTIPLE AGREE Language
Feature of NM/Op¬
Feature of NIs
Multiple Agree
JA, EA
Op¬ [ineg∅]
walaħada: [uneg]
not available
maa: [ineg]
walawaaħid: [uneg]
ma- . . . -š: [ineg]
maħadaaš: [ineg] maħaddiš: [ineg] maʕumriš: [ineg]
MA
Op¬ [ineg∅] maa: [uneg] ma- . . . -š: [ineg]
[ 134 ] A Unified Theory of Polarity Sensitivity
all NCIs: [uneg∅]
optional
5 3 1
Table 4.3. VARIATION IN THE FEATURES ON NMS, NIS, AND MULTIPLE AGREE Type
Language
Feature of NM/Op¬
Feature of NIs
Multiple Agree
NC
JA,
maa: [ineg]
walaħada: [uneg]
not available
EA
ma- . . . -š: [ineg]
walawaaħid: [uneg] maħadaaš: [ineg] maħaddiš: [ineg] maʕumriš: [ineg]
NC
Italian, Spanish
[ineg]
[uneg]
obligatory
NC
Greek, Polish,
[uneg]
[uneg∅]
obligatory
Russian, Ukrainian NC
Romanian
[uneg]
[uneg]
optional
NC
French,
ne: [uneg]
[uneg∅]
optional
[uneg∅]
not available
pas: [ineg] MA
maa: [uneg] ma- . . . -š: [ineg]
DN
German, Dutch,
[ineg]
Scandinavian
4.1.5. NCI licensing and phase theory
As for NCI licensing, it is clear that the semantic approaches to licensing, including the nonveridicality approach, cannot explain cases of ungrammaticality even when the NCI is under the scope of negation. The reasons are as follows. First, the NCI walaħada fails to occur in embedded clauses despite negation’s scope over it in the main clause: (69) a.
ma-ðˁunni-š ʔinn-o zaar-o ʔay ħada neg-think.1sg-neg that visited.3MSG -him any one ‘I don’t think anyone visited him.’ ʔinn-o zaar-o wala ħada that visited.3MSG -him NCI- one
(JA)
b.
*ma-ðˁunni-š neg-think.1sg-neg
c.
ʔaħmad maa-ʔaal-š ʔin Mona (EA) Ahmad neg-said. 3msg -neg comp Mona fihm-it ʔayy/*walaa aagah understood-3fsg any/no thing ‘Ahmad didn’t say that Mona understood anything.’ (Soltan, 2014)
(JA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 135 ]
6 3 1
In addition, the NCI wala fails to occur in certain embedded clauses despite having a preventing predicate in the main clause. The NCI wala can occur in the embedded clause of the participle form of the preventing predicate: (70) a.
b.
mamnuuʕ tištari ʔay/ wala ʔiši part.prevented buy.2sg NPI/NCI thing ‘It is disallowed to buy anything.’ mamnuuʕ tašri ši part.prevented buy.2sg some ‘It is disallowed to buy anything.’
(JA)
ħaʒa /ħtta ši thing/NCI some
ħaʒa thing
(MA)
But not the embedded clause of a perfective one: (71) a.
b.
manaʕ-ha tištari ʔay/ *wala prevented.3MSG-her buy.2sg NPI /NCI ‘He prevented her from buying anything.’
ʔiši thing
(JA)
mənaʕ-ha təšri ʔay ħaʒa /*ħtta ši ħaʒa prevented.3MSG-her buy.2sg NPI.thing/NCI some thing ‘He prevented her from buying anything.’
(MA)
The inability to include the NCIs in these contexts suggests that the semantic notion of antiveridicality cannot be responsible for regulating the distribution of NCIs. The syntactic agreement approach, however, can explain these facts by appealing to phase theory (Chomsky, 2001). If syntactic agreement between the NCI and its licensor is sensitive to phase, it follows that NCIs in embedded clauses must be licensed by a licensor within the same phase. (72) a.
b.
[CP ma-ðˁunni-š [CP ʔinn-o zaar-o neg-think.1sg-neg that visited.3MSG-him ʔay-ħada /*wala ħada NCI-one ‘I don’t think anyone visited him.’
(JA)
[CP ʔaħmad maa-ʔaal- š [CP ʔin Mona (EA) Ahmad neg-said.3msg-neg comp Mona fihm-it ʔayy/*walaa aagah understood-3fsg any/no thing ‘Ahmad didn’t say that Mona understood anything.’ (Soltan, 2014)
[ 136 ] A Unified Theory of Polarity Sensitivity
7 3 1
To test this analysis, we can contrast the behavior of NCIs in embedded clauses that are preceded by a preventing predicate in the passive participle form with those that are preceded by a verbal preventing predicate. The prediction is that the passive participle predicate will license wala NCIs since it does not project a CP; i.e., it does not have its own phase. The verbal predicate, however, has its own CP phase and thus should be unable to license the wala NCIs. Both predictions are borne out: (73) a.
b.
[ModP mamnuuʕ [TP [VP tištari part.prevented buy.2sg ‘It is disallowed to buy anything.’
ʔay/ wala NPI/ NCI
ʔiši thing
[CP [TP [VP manaʕ-ha [CP [TP [VP tištari ʔay/ *wala prevented3MSG-her buy.2sg NPI/ NCI ‘He prevented her from buying anything.’
(JA)
ʔiši (JA) thing
Since the participle is a form of deontic modality, it is merged in a ModP that selects the subjunctive verb. As such, both the participle and the NCI are located inside the same CP phase.
4.2. LICENSING DISJUNCTION PHRASES AND CLAUSES
Disjunction phrases (DisjPs )and clauses composed using the particles laa- wala ‘neither-nor’ give further support for the syntactic agreement approach proposed and discussed in the previous sections.8 We shall see that these coordinators are NCIs that are conspicuous analogs of the non-strict NCIs such as walaħada ‘nobody’ in both their distribution with respect to NMs and NIs and their licensing configurations and operations. Coordination constructions can either conjoin or disjoin phrases or clauses. Languages vary in the strategies employed to establish these logical relationships, creating distributional contrasts and co-occurrence restrictions on the coordination particles. In this section, I describe negative coordination and propose an analysis for this unexplored topic in Arabic. I will show that laa-wala coordination is a disjunction 8. This section is based on a presentation I gave at the 33rd Annual Symposium on Arabic Linguistics held at the University of Toronto on April 5–7, 2019. I thank the organizers Atiqa Hachimi and Abdel-Khaliq Ali, and the audience for their helpful comments. Parts of this section are a revised version of appearing in the perspective on Arabic Linguistics 30. I benefited from the insightful comments and feedback from the reviewers of the manuscript, the audience of ASAL 31, Abbas Benmamoun, and Enam Al-Wer. I very much thank them, Youssef Haddad, Amel Khalfaoui, and Matthew Tucker for their helpful comments. Needless to say, any errors are mine alone.
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 137 ]
8 3 1
construction distinct from the construction involving two negative phrases/ clauses coordinated by wa-. As such, the analysis projects a DisjP headed by wala and two DPs/CPs, CP1 for the first disjunct occupying the Spec-DisjP and CP2 for the second disjunct as a complement of DisjP. The analysis shows that wala is a disjunction operator with an uninterpretable negation feature (i.e., an NCI) licensed by a negative operator. With a postverbal DisjP, the negative operator is maa; elsewhere it is a covert negative operator that dominates the DisjP. With coordinated DPs, the analysis shows that singular agreement on the verb involves clausal coordination and VP ellipsis in one disjunct.
4.2.1. Introduction and empirical generalizations
Coordination constructions are an important component of syntactic structure used to string together phrases or clauses with a compositional meaning of conjunction or disjunction. Previous studies on the syntax of Arabic have shown only a limited focus on this topic, with some research on conjunction constructions focusing specifically on the phenomenon known as first conjunct agreement. Outside of this, there is hardly any research on disjunction constructions in Arabic. To fill this gap, I take up the topic here through a description of this construction and the manner in which it contrasts with conjunction constructions in Arabic. As such, the discussion here advances an analysis for this construction that differs from previous analyses of conjunction constructions in Arabic and of disjunction constructions in English. To begin, Arabic coordinate complexes exhibit a sophisticated system interacting with the syntax of negation. The coordinate complexes that relate to the discussions of the syntax of PSIs are the following: (74) Arabic Coordinate Complexes a. b. c. d. e.
Coordinator wa ‘and’ (= conjunctive reading) Coordinator ʔaw ‘or’ (= free-choice reading) Coordinators ʔimma-ʔaw/ yaa-yaa ‘either-or’ (= disjunctive reading) Coordinators walla/willa/ʔam ‘or’ (= disjunctive reading). Coordinators laa-wala ‘neither-nor’ (= negative disjunctive reading)
The following examples illustrate the functions of these coordinate complexes: (75) Coordinator wa ‘and’ (= conjunctive reading) a. ʔišrab gahwa wa šaay drink.imp coffee and tea ‘Drink coffee and tea.’ [ 138 ] A Unified Theory of Polarity Sensitivity
(JA)
9 3 1
b.
ʔšrab lqahwa drink.imp coffee ‘Drink coffee and tea.’
ʔataay tea
w and
(MA)
(76) Coordinator ʔaw in JA wəlla in MA ‘or’ (= free-choice reading) a. ʔišrab gahwa ʔaw šaay (JA) drink.imp coffee or tea ‘Drink coffee or tea.’ b.
ʔšrab lqahwa wəlla ʔataay drink.imp coffee or tea ‘Drink coffee or tea.’
(MA)
(77) Coordinators ʔimma-ʔaw/ wəlla ‘either-or’ (= disjunctive reading) a. ʔišrab ʔimma gahwa ʔaw šaay drink.imp either coffee or tea ‘Drink either coffee or tea.’ b.
ʔšrab ʔimma lqahwa drink.imp either coffee ‘Drink either coffee or tea.’
wəlla/ ʔaw ʔataay or tea
(JA)
(MA)
(78) Coordinator walla/willa/ wəlla ‘or,’ in interrogative contexts (= disjunctive reading). a. tišrab gahwa walla šaay? (JA) drink.SBJV coffee or tea ‘Would you like to drink coffee or tea?’ b.
t-šrab l-qahwa wəlla ʔataay ? drink.SBJV coffee or tea ‘Would you like to drink coffee or tea?’
(MA)
(79) Coordinators laa-wala ‘neither-nor’ (= negative disjunctive reading) a. laa tišrab laa gahwa wala šaay (JA) neg drink.imp neither coffee nor tea ‘Don’t drink coffee or tea’ b.
maa tšrab laa neg drink.imp neither ‘Don’t drink coffee or tea’
qahwa coffee
wa laa nor
ʔataay tea
(MA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 139 ]
0 4 1
Coordination constructions for conjoining phrases or clauses in JA involve the particle wa-‘and.’9 (80) a.
b.
Zeed wa ʕumar Zeed and Omar ‘Zeed and Omar saw us.’
šaafuu-na saw.3mp-us
šufna Zeed wa šufna saw.1p Zeed and saw.1p ‘We saw Zeed and we saw Omar.’
ʕumar Omar
Negative coordination, on the other hand, involves the use of two particles laa-wala ‘neither-nor’ in a more sophisticated manner. The syntax of this construction raises questions about the status of coordination particles (e.g., neither . . . nor . . . in English; laa . . . wala . . . in JA) and the nature of the functional head that combines the two parts of a coordination construction. In JA, the coordination particles are homophonous with NMs and negative sensitive items. The coordinator laa ‘neither’ is homophonous with the NM used in negative imperatives: (81) laa neg ‘Don’t travel.’
tsaafir travel.2msg.imp
The coordination particle wala ‘nor’ is homophonous with the NCI wala in the following examples:10 (82) a.
b.
maa šuft wala neg saw.1s NCI ‘I didn’t see anyone.’
waaħad one
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
šaaf-na saw.3msg-us
9. The conjunction wa in JA is phonetically [w]. 10. The disjunction particle walla ‘or’ is pronounced as willa by other speakers of Hourani. I personally use both walla and a non-geminate variant wala. All of these variants, however, are orthogonal to the analysis in this chapter because I treat the disjunction walla ‘or’ as the affirmative counterpart of the negative coordination particle wala ‘nor.’ [ 140 ] A Unified Theory of Polarity Sensitivity
1 4
These negative coordination markers can coordinate two phrases or two clauses. The coordination particle laa precedes the first element of a coordination construction; see, for example, the higher phrase in (83) and the higher clause in (84). The coordination particle wala precedes the second element of coordination: for example, the lower phrase in (83) and the lower clause in (84). (83) a.
b.
(84) a.
b.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’ laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
šaafuu-na saw.3mp-us
šaafuu-na saw.3mp-us
laa šufna Zeed wala šufna Neither saw.1p Zeed nor saw.1p ‘We neither saw Zeed, nor did we see Omar.’ maa šufna Zeed wala šufna neither saw.1p Zeed nor saw.1p ‘We neither saw Zeed, nor did we see Omar.’
ʕumar Omar
ʕumar Omar
(JA)
(MA)
(JA)
(MA)
The negative coordination marker wala also displays the characteristics of NCIs. (85) a.
*(maa) šuft neg saw.1s ‘I didn’t see anyone.’
wala NCI
b.
*(maa) šuft laa Zeed neg saw.1s neither Zeed ‘I neither saw Zeed nor Omar.’
c.
*(maa) šuft neg saw.1s ‘I didn’t see anyone.’
d.
*(maa) šuft laa Zeed neg saw.1s neither Zeed ‘I neither saw Zeed nor Omar.’
ħtta-waħəd NCI-person
waaħad one
wala nor
(JA)
ʕumar Omar
(JA)
(MA)
wala nor
ʕumar Omar
(MA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 141 ]
2 4 1
(86) a.
b.
wala waaħad NCI one ‘No one saw us.’
šaaf-na saw.3msg-us
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
(JA)
šaaf-na saw.3msg-us
(JA)
In interrogative contexts, walla and its variant willa have the affirmative coordination interpretation ‘or’ and behave more like a disjunction operator. (87) Affirmative disjunction a. šuft Zeed walla/willa (šuft) saw.2MSG Zeed or saw.2MSG ‘Did you see Zeed or (did you see) Omar?’ b.
ʕumar? Omar
(JA)
(waš) šufti Zeed wəlla (šufti) ʕumar? Q saw.2MSG Zeed or saw.2MSG Omar ‘Did you see Zeed or (did you see) Omar?’
(MA)
The presupposition in this example is that the addressee saw either Zeed or Omar but not both or neither of them. Previous discussion on the syntax of negative coordination has focused primarily on the particle wala in the context of negative polarity or as an NCI. The analysis here treats wala as the head of a DisjP that takes the first disjunct as a specifier and the second one as a complement. As such, the coordination particle laa occupies the Spec,XP position of the first disjunct. Consider the following representations that I propose for disjoined DPs and disjoined CPs: (88) a. b.
[DisjP [CP laa [C . . . ]] [Disj wala [CP [C . . . [DisjP [DP laa [D . . . ]] [Disj wala [DP [D . . .
This analysis captures the meaning of the disjunction construction by allowing wala to head the DisjP and predicts its dependency relation with negation, as an NCI, through syntactic licensing under the scope of a negative operator. This negative operator is the NM maa when the disjoined DP is postverbal and a null negative operator dominating the DisjP elsewhere. (89) a.
[NegP Neg maa [TP [VP [DisjP [laa CP] Disj wala [CP . . . ] . . .
b.
[NegP Neg Op¬ [DisjP [laa DP] Disj wala [DP . . . ] TP . . .
c.
[NegP Neg Op¬ [DisjP [laa CP] Disj wala [CP . . . ] . . .
[ 142 ] A Unified Theory of Polarity Sensitivity
3 4 1
The subsequent sections develop this syntactic analysis for negative coordination with laa-wala. There, the discussion will contrast laa-wala with coordinated structures that use the conjunction wa-and investigate the consequences of this analysis for a variety of syntactic phenomena, such as ellipsis, first conjunct agreement, and the coordinate structure constraint.
4.2.2. Theoretical issues
A number of theoretical issues arise in the context of analyzing the syntax of negative coordination. One such issue regards whether coordination structures are tripartite (Jackendoff, 1977; Chomsky, 1981) or asymmetric (De Groot, 1959; Munn, 1993; Kayne, 1994). I argue for the asymmetric structure in Arabic. The asymmetric structure is in keeping with minimalist assumptions regarding binary branching as seen in example (90) (XP = DP, CP, etc.). (90)
a.
XP
Tripartite
XP1
and
XP2
b.
&P XP1
Asymmetric &
&
XP2
Several tests may be employed to evidence one structure over the other. In Arabic, much evidence supports the asymmetric structure for coordinated complexes. One such test reveals an asymmetry in possessive pronominalization in English (see the following examples from Zhang, 2010, p. 12): (91) a. Sally’s mother and yours have turned vegetarian. b. *Yours and Sally’s mother have turned vegetarian.
These same facts are true of Arabic. (92) a.
b.
sayyaarit Zeed w car Zeed and ‘Zeid’s car and yours are new.’ *tabʕat-ak belong-your
w and
tabʕat-ak belong-your
sayyaarit car
Zeed Zeed
ʤdaad new.P
ʤdaad new.P
(JA)
(JA)
Another test involves extraction from conjunction constructions. Asymmetrical coordination allows extraction from either conjunct (Ross, L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 143 ]
41
1967; Lackoff, 1986; Levin & Prince, 1986), as in the English data cited in Zhang (2010, p. 12): (93) a.
what kind of cancer can you eat herbs and not get __?
b.
what kind of herbs can you eat__and not get cancer?
Similar facts hold for Arabic. The following examples show that it is possible to extract the object from either the internal or the external member of the coordinate complex: (94) a.
b.
ʔayy saratˁaan btigdar tookil ʔaʕšaab which cancer can.2MSG eat.2MSG herbs wa maa ysˁiib-ak__? and neg befall-you.2MSG ‘what kind of cancer can you eat herbs and not get __?’ ʔayy ʔaʕšaab btigdar tookil__ which herbs can.2MSG eat.2MSG wa maa ysˁiib-ak saratˁaan? and neg befall-you.2MSG cancer ‘what kind of herbs can you eat__and not get cancer?’
(JA)
(JA)
If either conjunct were an adjunct, extraction from an island should not be possible because Arabic does not allow extraction from adjunct islands. (95) a.
b.
Zeed rawwaħ [PP qabl-maa Zeed went.3msg before-comp yištari ktaab l-taariix buy.3msg book the-history ‘Zeed went home before buying the history book.’ *[CP [PP qabl-maa before-comp
ʔayy ktaab]i which book yištari [PP . . . ]i buy.3msg
[C [TP Zeed Zeed
(JA)
rawwaħ went.3msg
(JA)
Overall, the results of these tests indicate that the Arabic conjuncts under discussion here are in a Specifier-Complement configuration. The second issue that arises when analyzing the syntax of negative coordination centers around the status of the coordination particle (conjunction
[ 144 ] A Unified Theory of Polarity Sensitivity
5 4 1
‘and’; disjunction ‘or’) as either the head of the coordination construction or a phrasal category at the left edge of the second XP (Zwart, 2005).11 (96) XP1
XP2 &
XP 2
For English coordination constructions, Den Dikken (2006) argues that the two elements of the disjunction construction ‘neither . . . nor . . .’ are phrasal categories in two clauses combined by an abstract functional head labeled ‘J’ (for ‘junction’). See example (97) for an illustration. (97) [JP [CP neither [C did [TP john eat]]]
[J [CP nor [C did [TP he swim . . .
As for Arabic coordination structures, they appear to differ, at least on the surface, from English in regard to coordination; unlike in English, the Arabic coordination construction uses the conjunction particle wa- and the NM la together to make up wala. As such, the null hypothesis is that wala is the negative counterpart of wa-. In other words, the question at hand is whether coordination in Arabic involves two negative clauses/phrases coordinated by the conjunction particle wa-la ‘and-not’ or two negative clauses/phrases coordinated by a disjunction particle wala ‘nor.’ (98)
&P XP1
Conjunction &
& (99)
XP2
DisjP XP1
Disjunction Disj
Disj
XP2
11. A third issue, not directly involved with this chapter, focuses on whether the two coordinated XPs are ordered in the syntactic structure. Goodall (1987) proposes that the two coordinated XPs are parallel; that is, neither XP dominates or precedes the other in syntax. The two XPs are in parallel syntactic planes. Kramer (2010) adopts this view for coordination in Amharic DPs and derives word order via morphological rules subject to various linearization sub-operations (adjacency, concatenation, and chaining) within a Distributed Morphology framework.
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 145 ]
6 4 1
In placing XP1 in the specifier of DisjP, the analysis predicts that negation does not scope over an NPI inside the internal disjunct. Despite the fact that the first disjunct is a negative clause, negation of the external disjunct cannot license the NPI in the internal disjunct. This is expected since negation of the first clause is buried inside the CP1: (100)
ma-šuf-tu-š Zeed walla (JA) neg-saw-2mpl-neg Zeed or *(ma-šuf-tu-š ) ʔayy ħada? neg-saw-2mpl-neg NPI.any one ‘Did you not see Zeed or *(did you not see) anybody?’
4.2.3. Compositional meaning of coordinate complexes
While the compositional meaning of other conjunctions and disjunctions is straightforward (the conjunction wa ‘and’ is [p ∧ q]; the disjunction coordinator ʔimma-ʔaw ‘either-or’ has the compositional meaning [p ∨ q]), it is unclear whether the coordination complex laa-wala is a negative conjunction or negative disjunction. This then leaves open the possibility for the two following compositional meanings. (101)
a. [¬p ∧ ¬q] ‘not XP and not XP’? b. [¬[p ∨ q]] ‘not XP or XP’?
To determine this, we must first ask whether wala is the negative counterpart of wa ‘and’ or ʔaw ‘or.’ The lexical composition of negative disjunction suggests that we have two negative constituents coordinated by wa ‘and’: (102)
laa XP wa laa XP → laa XP wala XP
If coordination by la . . . wala involves two negative XPs conjoined by wa-, we can expect such constructions to behave like two clauses that have been negated by bipartite negation and then conjoined by wa-: (103)
a.
ma-šuft- iš Zeed wa ma-šuft- iš neg-saw.1s-neg Zeed Conj neg-saw.1s-neg ‘I didn’t see Zeed and I didn’t see Omar.’
[ 146 ] A Unified Theory of Polarity Sensitivity
ʕumar Omar
(JA)
7 4 1
b.
laa šuft Zeed wala šuft neither saw.1SG Zeed nor saw.1SG ‘I neither saw Zeed, nor did I see Omar.’
ʕumar Omar
(JA)
The incompatibility of the disjunction marker ʔaw ‘or’ with laa ‘neither’ and with bipartite negation also suggests that the two members are coordinated by the conjunction wa ‘and.’ Empirical facts, however, show that the two constructions exhibit key contrasts. First, the disjunction operator ʔaw ‘or’ in MA is, in fact, compatible with laa ‘neither’: (104)
a.
maa ʔkkad maa neg confirmed.3msg nor ‘He neither confirmed nor denied.’
b.
*maa-ʔakkad-š ʔaw nfaa neg-confirmed.3msg-neg or denied.3msg Intended reading: ‘He did not confirm or deny.’
c.
laa al-raʔiis wala/ ʔaw ši ħadd neither the-president nor /or some one min l-wuzaraaʔ dyal-o of the-cabinet prep-his ‘Neither the president nor/or any of his cabinet members.’
nfaa denied.3msg
(MA)
(MA)
(MA)
Second, when we use the conjunction wa-, the two coordinated XPs must have identical negation (bipartite negation). On the other hand, the use of wala permits the first XP to be negated by either maa or laa: (105)
(106)
a.
ma-šuft-iš Zeed wa ma-šuft-iš ʕumar neg-saw.1SG-neg Zeed Conj neg-saw.1SG-neg Omar ‘I didn’t see Zeed and I didn’t see Omar’
b.
*laa šuft Zeed wa neither saw.1SG Zeed Conj
ma-šuft-iš ʕumar neg-saw.1SG-neg Omar
(JA)
c.
*maa neg
ma-šuft-iš neg-saw.1s-neg
ʕumar Omar
(JA)
a.
laa šuft Zeed wala šuft ʕumar neither saw.1sG Zeed nor saw.1sG Omar ‘I neither saw Zeed, nor did I see Omar’
šuft saw.1sG
Zeed Zeed
wa Conj
(JA)
(JA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 147 ]
8 4 1
b.
maa šuft Zeed wala neg saw.1sG Zeed nor ‘I didn’t see Zeed, nor did I see Omar’
šuft saw.1sG
ʕumar Omar
(JA)
Another key contrast between coordination by wa- and coordination by wala is that only the latter exhibits contrastive focus, where wala c-commands the focused constituent: (107)
(108)
ʕumar Omar
a.
laa šufna Zeed wala šufna neither saw.1p Zeed nor saw.1p ‘We neither saw Zeed, nor did we see Omar.’
b.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
a.
ma-šuft-iš Zeed wa ma-šuft-iš neg-saw.1s-neg Zeed Conj neg-saw.1s-neg ‘I didn’t see Zeed and I didn’t see Omar’
b.
miš Zeed wa miš ʕumar neg Zeed Conj neg Omar ‘It is not Zeed and not Omar that saw us.’
(JA)
šaaf-na saw.3msg-us
(JA)
ʕumar Omar
šaaf-na saw.3msg-us
(JA)
(JA)
These contrasts between wala and wa cast doubt over the claim that wala is the negative counterpart of the conjunction wa ‘and,’ which leaves us with the possibility that wala is the counterpart of ʔaw ‘or.’ The latter possibility, however, is not straightforwardly evident. If wala is a disjunctive marker, why is the disjunction marker ʔaw ‘or’ incompatible with negation? This incompatibility is unexpected given the grammaticality of such structures in English, as indicated in the following glosses and translation: (109)
a.
maa-ʔakkadi-š wala neg-confirmed.3msg-neg nor ‘He neither confirmed nor denied.’
b.
*maa-ʔakkadi-š ʔaw nafa neg-confirmed.3msg-neg or denied.3msg Intended reading: ‘He did not confirm or deny.’
[ 148 ] A Unified Theory of Polarity Sensitivity
nafa denied.3msg
(JA)
(JA)
9 4 1
c.
laa al-raʔiis wala/*ʔaw ʔay ħada min wuzaraaʔ-o neither the-president nor/or any one of cabinet-his ‘Neither the president nor/or any of his cabinet members.’
(JA)
I will, however, show that wala is the counterpart of ʔaw ‘or.’ Consider the following, where ʔaw ‘or,’ similar to English or, is compatible with negation when negation scopes over both members of coordination: (110)
a.
[TP lam [VP yuʔakkid ʔaw neg.past confirm.3msg.juss. or ‘He did not confirm or deny.’
yanfi]] deny.3msg.juss.
b.
[CP maʕumr-o [TP ʔakkad ʔaw never-him confirmed.3msg or ‘He never confirmed or denied.’
nafa]] denied.3msg
c.
[CP [AdvP Adv
d.
[CP maʕammar-o maa [TP ʔakkad ʔaw nfaa]] (MA) never-him neg confirmed.3msg or denied.3msg ‘He never confirmed or denied.’
e.
[CP [AdvP Adv
maʕumr [DisjP al-raʔiis ʔaw never the-president or ʔay ħada min wuzaraaʔ-o] any one of cabinet-his ‘Neither the president nor/or any of his cabinet members’
(SA)
(JA)
(JA)
ma-ʕammar [DisjP (laa) al-raʔiis wa laa never neg the-president or ši ħad min l-wuzaraaʔ dyal-o] maa . . . any one of the-cabinet prep-his neg . . . ‘Neither the president nor/or any of his cabinet members’
(MA)
These facts suggest that the compositional meaning is not conjoining but disjoining: (111)
a. *[¬p ∧ ¬q] ‘not XP and not XP’ b. [¬[p ∨ q]] ‘not XP or XP’
If laa-wala coordinate complexes are, indeed, disjunction constructions, the implications of this analysis may be tested through interceptions tests. Such tests are possible for ʔaw ‘or,’ where negation is inside the external member L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 149 ]
0 5 1
of coordination. Due to this, negation does not scope over the whole coordination structure and the internal member is affirmative. Such structures are, indeed, illicit and crash at LF: (112)
*ma-ʔakkad-iš ʔaw nafa neg-confirmed.3msg-neg or denied.3msg Intended reading: ‘He did not confirm or deny.’ Illicit at LF: [¬p ∨ q]
(113)
laa al-raʔiis wala/*ʔaw neither the-president nor/or ʔay ħada min wuzaraaʔ-o any one of cabinet-his ‘Neither the president nor/or any of his cabinet members.’ Illicit at LF when ʔaw is used: [¬p ∨ q] ‘*It is not A or it is B’ Licit at LF when wala is used: [¬[p ∨ q]] ‘It is not A or B’
(JA)
(JA)
Given Dikken’s (2006) analysis of neither as a Spec-CP adverb in English, it is unexpected that negation may scope over the whole disjunction construction. This suggests that ‘neither’ in English is not always a Spec.CP adverb in the coordination complexes, contra Dikken (2006). As such, it is reasonable to suggest that the disjunction adverb ‘neither’ in such structures takes a DP coordination complex as its complement: (114)
[CP Neither [DisjP [DP the president] [Disj or [DP any of his cabinet members]]]].
Due to the deictic nature of ‘neither’ in English, the structure [Neither [of them]], is permitted. ‘Neither’ is inside the external member of DP coordination complex: (115)
[DisjP Neither [DP the president] [Disj nor [DP any of his cabinet members]]]].
These empirical contrasts do not support the hypothesis that wala is the negative counterpart of wa-. In fact, there is empirical evidence that wala is a disjunction operator associated with the [+neg] feature in indicative sentences. As such, it is the negative counterpart of walla/willa. These do not carry [+neg] and occur in interrogative sentences (i.e., in affirmative disjunction):
[ 150 ] A Unified Theory of Polarity Sensitivity
1 5
(116)
Affirmative disjunction šuft Zeed walla/willa saw.2msg Zeed or ‘Did you see Zeed or (did you see) Omar?’
(šuft) saw.2msg
ʕumar? Omar
(JA)
In this discussion, I have shown that that coordination by wala is distinct from coordination by wa-. The particle wala behaves roughly like ‘nor’ in English as opposed to ‘and not.’ The distinction between conjunction and disjunction largely has to do with the fact that in disjunction structures the two members of the coordination structure can be in an exclusive relation, while they can only be in an inclusive relationship in conjunction structures. Before proceeding, I will note one further piece of evidence for the distinction between wa- as a conjunction particle and wala as a disjunction particle. In the following data, the coordination phrase laa Zeed wala ʕumar is the subject of the sentence. In sentences with this subject, the verb may appear with singular agreement. As such, this supports the claim that the coordinate structure has a disjunctive interpretation where the two members of the coordinate construction are in an exclusive relationship. Later, this will be used as support to argue that each disjunct is in a separate clause. (117)
a.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
b.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
c.
Zeed wa-ʕumar Zeed and-Omar ‘Zeed and Omar saw us.’
šaaf-na saw.3msg-us
šaaf-uu-na saw-3mpl-us
šaaf-*(uu)-na saw-*(3mpl)-us
(JA)
(JA)
(JA)
4.2.4. Analysis of negative coordination by wala
The following derivations represent the coordination constructions that I argue for in the rest of this section, with coordinated DPs in (24)a and coordinated CPs in (24)b:
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 151 ]
2 5 1
a.
(118)
DisjP DP1
b. Disj
Disj
DisjP CP1
DP2
Disj Disj
CP2
4.2.4.1. Status and locus of laa
As for the status of laa and wala in JA, I argue that laa is a phrasal category (not a head) in Spec,DP with DP coordination, or Spec,CP with CP coordination, of the first disjunct, while wala is the head of the disjunction projection. The two disjuncts are within the disjunction phrase (DisjP). The first disjunct occupies the Spec-DisjP position, while the second one is the complement of DisjP: (119)
[DisjP [CP laa [C . . .]]
[Disj
wala [CP . . .
The status of laa as a phrasal category is clear from the following contrasts with maa. First, the particles laa and maa exhibit contrasts in their distribution within coordinated CPs and DPs, as in the following JA examples: (120)
a.
laa Zeed šaaf-na wala neither Zeed saw.3msg-us nor ‘Neither Zeed saw us nor Omar saw us.’
b.
*maa neg
Zeed Zeed
šaaf-na wala saw.3msg-us nor
ʕumar Omar
šaafna saw.3msg.us
ʕumar Omar
šaafna saw.3msg.us
As the data shows, maa cannot precede the preverbal subject unlike laa, which appears before the subject in this data. This suggests that maa heads a NegP projection while laa is a syntactic adverb, XP, in the CP layer. Second, the facts indicate that laa is not an NM in a NegP projection dominating the subject. If laa were an NM, we would expect maa to appear in disjoined DPs. Since this is not possible (cf. the ungrammaticality of *maa Zeed wala ʕumar in the following example), it is reasonable to treat laa as a marker of negative coordination on par with neither in English. (121)
a.
laa Zeed wala ʕumar neither Zeed nor Omar ‘Neither Zeed nor Omar saw us.’
[ 152 ] A Unified Theory of Polarity Sensitivity
šaafna saw.3msg.us
(JA)
3 5 1
b.
*maa neg
Zeed Zeed
ʕumar Omar
wala nor
šaafna12 saw.3msg.us
In treating laa as a coordination adverb (disjunction adverb), the analysis predicts the fact that laa cannot be used for negating perfective and imperfective verbs in JA: (122)
a.
maa/*laa neg/*neg ‘He did not travel.’
saafar travel.3msg
b.
maa/*laa neg/*neg ‘He did not travel.’
bi-saafir asp-travel.3msg.
Outside the context of coordination laa is largely restricted to negative imperatives in JA: (123)
laa tsaafir neg travel.2msg .subj ‘Don’t travel.’
So far, these empirical facts suggest that laa in the coordination construction laa-wala is semantically vacuous in negative force. This is further supported with evidence that suggests laa is a disjunction particle without semantic negation. Consider the following example where a disjoined DP in postverbal position requires an NM. The inability of laa to satisfy this requirement suggests that it is semantically vacuous: (124)
*(maa) šuft laa neg saw.1sg neither ‘I saw neither Zeed nor Omar’
Zeed Zeed
wala nor
ʕumar Omar
(JA)
12. maa also cannot occur postverbally, suggesting that it is the head of a negative projection (NegP) in coordination constructions: (i)
(ii)
laa šaaf-na NEG saw.3MSG-us ‘Neither Zeed nor Omar saw us.’ maa šaafna NEG saw.3MSG-us ‘Neither Zeed nor Omar saw us.’
(*maa) NEG
Zeed wala ʕumar Zeed nor Omar
(*maa) NEG
Zeed wala ʕumar Zeed nor Omar
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 153 ]
4 5 1
4.2.4.2. Status and locus of wala as an NCI and disjunction operator
As for the status of wala, it is necessary to consider first that it is a disjunction operator and an NCI. The particle wala requires negation in the postverbal position, but not in the preverbal position (see Alqassas, 2015, 2016, 2018): (125)
(126)
a.
*(maa) šuft neg saw.1sg ‘I didn’t see anyone.’
wala NCI
b.
*(maa) šuft laa neg saw.1sg neither ‘I neither saw Zeed nor Omar.’
a.
wala waaħad NCI one ‘No one saw us.’
b.
laa Zeed wala neither Zeed nor ‘I saw neither Zeed nor Omar.’
waaħad one
Zeed Zeed
wala nor
šaaf-na saw.3msg-us
ʕumar Omar
(JA)
ʕumar Omar
(JA)
(JA)
šuft saw.1sg
(JA)
Crucially, the NCI wala is in complementary distribution with the definite determiner; see (127)a and the representation in (128). By contrast, the coordinator wala in (127)b can co-occur with the definite noun/determiner and, as such, cannot be the head of the DP. Thus, the coordinator wala in (125)b and (126)b is either a phrasal category in Spec,DP or the head of the DisjP; see the structural representations in (129)a and (129)b. (127)
a.
maa šuft neg saw.1sg ‘I didn’t see any boy.’
b.
maa šuft laa al-walad wala neg saw.1sg neither det-boy nor ‘I saw neither the boy nor the girl.’
(128)
wala NCI
DP D wala
NP
[ 154 ] A Unified Theory of Polarity Sensitivity
(*al)-walad det-boy
(JA)
al-bint (JA) det-girl
51
(129)
a.
DisjP DP1
Disj Disj
DP2 Spec
laa Zeed b.
wala
D2 D2
umar
DisjP DP1
laa Zeed
Disj Disj
DP2
wala
umar
In considering whether wala is located in the head of the DisjP or in the second disjunct, I follow den Dikken’s (2006) line of reasoning in analyzing the particle nor, and I place the particle wala in the head of the disjunction projection, the projection that den Dikken calls ‘JP.’ This is the lexicalization of the disjunction head in JA, unlike in English, where nor is the specifier of the internal disjunct (den Dikken, 2006). (130)
[DisjP [CP laa [C . . .]]
[Disj wala [CP [C . . .
Proceeding with the analysis of wala, let us consider its use in coordinated DPs. The fact that wala in DisjPs displays NCI behavior—cf. (125) and (126) —suggests that the whole DisjP interacts with a negative projection (i.e., it is licensed by an NM). Based on this, I assume that the NCI wala establishes dependency with negation via a negative agreement process whereby wala carries a [uneg] feature that gets valued by an NM carrying an [ineg] feature (Ouali & Soltan, 2014; Alqassas, 2015, 2016). This dependency relation is evidence that the entire disjoined DP is licensed by maa (the realization of the negative operator). The concordant reading of the sentence (and the absence of a DN reading) shows that wala fails to contribute a negative interpretation despite co-occurring with maa. To derive its preverbal position, I propose that DisjP undergoes phrasal movement from its postverbal position in (125)b into the preverbal position in (126)b, specifically in Spec,NegP as in (131).
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 155 ]
6 5 1
(131)
a.
[NegP maa[iNEG] [TP šuft [DisjP laa Zeed wala[uNEG] ʕumar] ]]
b.
[NegP [DisjP laa Zeed wala[uNEG] ʕumar]i Op¬[iNEG] [TP šuft [DisjP . . . ]i ]]
Ouali and Soltan (2014) propose that the preverbal wala in Egyptian is inherently negative and does not need an NM to license it. While this issue seems orthogonal to the analysis here, there is evidence that in JA even the preverbal wala fails to contribute a negative interpretation when preceded by a negative constituent (Alqassas, 2015, p. 125): (132)
a.
maʕumuriš[iNEG] wala-ħada[uNEG] zaar never NCI-person visited.3msg ‘Never has anyone visited Petra.’
b.
maa-ʕammar ħtta waħəd maa never NCI-person neg ‘Never has anyone visited Petra.’
el-batra the-Petra
zaar el-batra visited.3msg the-Petra
(JA)
(MA)
The absence of a negative interpretation in this data suggests that the disjunction wala in coordinated CPs is licensed by a negative operator. Crucially, wala in such contexts can license an NPI in the postverbal position. As such, wala c-commands the NPI in this position (see Alqassas, 2015, 2016): (133)
a.
laa štareet ʔiši, wala biʕit neither bought.1sg NPI-thing, nor sold.1sg ‘I neither bought anything nor did I sell anything,’
b.
maa šreet waalu u-maa baʕt neither bought.1sg NPI-thing, nor sold.1sg ‘I neither bought anything nor did I sell anything,’
ʔiši (JA) NPI-thing
waalu NPI-thing
(MA)
If we follow this analysis to analyze disjoined DPs in preverbal position, it is reasonable to suggest that there is a negative operator dominating DisjP and licensing the disjunctive particle wala under c-command: (134)
[NegP Neg Op¬ [DisjP [laa CP] Disj wala [CP . . . ] . . .
In addition, further evidence supports the claim that the CP coordinate complex laa-wala also fails to contribute negative interpretation:
[ 156 ] A Unified Theory of Polarity Sensitivity
7 5 1
(135)
a.
maʕumr-o[iNEG] laa Zeed zaar el-batra (JA) never-him neither Zeed visited.3msg the-Petra wala ʕumar zaar l-ʔahraam-aat. nor Omar visited-3msg the-pyramid-FP ‘Never did Zeed visit Petra or Omar visit the pyramids.’ [NC] ‘It’s never been the case that neither Zeed visited Petra nor Omar visited the pyramids.’ [*DN]
b.
maʕammar Zeed maa zaar el-batra (MA) never Zeed neg visited.3msg the-Petra wala ʕumar zaar l-ʔahraam-aat. nor Omar visited-3msg the-pyramid-FP ‘Never did Zeed visit Petra or Omar visit the pyramids.’ [NC] ‘It’s never been the case that neither Zeed visited Petra nor Omar visited the pyramids.’[*DN]
The NI licenses the coordinated CP, which is an NCI. Crucially, the co- occurrence of the NI adverb maʕumr-o[iNEG] with an NCI in as large of a constituent as a coordinated CP does not give rise to a DN reading. What we have, instead, is a concordant reading. The analysis posited thus far is supported by the fact that, as an NCI, wala in preverbal position can be licensed by a covert negative operator (see Alqassas, 2015, 2016). Here, too, the disjunction marker wala is licensed by a covert negative operator by virtue of the fact that the entire DisjP is under c-command by the negative operator. (136)
CP coordinate complex laa-wala: a. [NegP Neg Op¬ [DisjP [CP [EXTERNAL CONJUNCT] Disj wala[uNEG] [CP INTERNAL CONJUNCT b. [maʕumr-o[iNEG][DisjP [CP [EXTERNAL CONJUNCT] Disj wala[uNEG] [CP INTERNAL CONJUNCT
To test the explanatory adequacy of this analysis, we can look for cases where the coordinate complex, which is an NCI, is preceded by an NPI. Given that the analysis posits a covert negative operator scoping over the coordinate complex, and given that the adverbial NPI ʕumr can be merged in the left periphery of the clause, it follows that the NPI ʕumr should be able to occur in a position preceding the coordinate complex laa-wala. This is, indeed, borne out in JA: (137)
ʕumr-ha laa waħade zaʕʕalat-ni wala zaʕʕalt-ha NPI-ever-her neither one.F upset.3FSG-me nor upset.1s-her ‘Never has anyone either upset me or I upset her.’ [NegP Neg Op¬ [DisjP [[CP ʕumr-ha] laa CP] Disj wala [CP . . . ] . . .
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 157 ]
8 5 1
The NPI phrase is adjoined to the external CP and the negative operator is in NegP above the entire coordinated CPs. Recall from section 4.2.3 that evidence suggests that the locus of disjunction walla/willa is under the head of DisjP. This evidence stems from the particle’s use as a disjunction operator ‘or’ without negative interpretation in interrogative contexts. Consider the following example from JA: (138)
šuft Zeed walla/willa (šuft) saw.2MSG Zeed or saw.2MSG ‘Did you see Zeed or (did you see) Omar?’
ʕumar?13 Omar
Further, preverbal subjects and focused constituents cannot precede wala. Instead, they must follow it in JA. (139)
a.
laa Zeed šaaf-na wala šaaf-na neither Zeed saw.3msg-us Omar nor Omar saw.3msg-us ‘Neither Zeed saw us nor Omar saw us.’
b.
laa Zeed šufna wala neither Zeed saw.1p Omar nor ‘We neither saw Zeed nor did we see Omar.’
Omar
šufna saw.1p
To recap, this section has argued that the Arabic coordinators wa, wala, ʔaw are the lexicalization of a syntactic head that takes the internal conjunct as a complement and the external one as a specifier. Negation must scope over the two members of the coordination phrase for the derivation to be licit at LF [neg [p or q]]. With negatively disjoined DPs, I argued that the coordinator is an NCI that carries an uninterpretable negation feature. This feature is licensed by a negative operator (overt or covert), building on arguments from section 4.1. The coordinators laa-wala ‘neither-nor’ are conspicuous analogs of the NCI wala, which can combine with NPs to create person-NCIs such as wala-ħada ‘no body,’ in their ability to co-occur with the negative adverb maʕumriš ‘never’ in the CP layers without triggering a DN reading. The coordinators laa-wala ‘neither-nor,’ thus, expand the inventory of uninterpretable negation features and call for the following revision to table 4.1, as in table 4.4. These
13. Coordination by wala here behaves like an NPI because it can be licensed without negation in affirmative interrogative contexts. It is interesting to note that this behavior is not consistent with its behavior as a non-strict NCI in the declarative contexts shown previously.
[ 158 ] A Unified Theory of Polarity Sensitivity
9 5 1
Table 4.4. INVENTORY OF [neg]- FEATURES Language
Interpretable neg
Uninterpretable neg
JA, EA
Op¬ [ineg∅]
NCIs: [uneg]
maa: [ineg]
laa-wala: [uneg]
ma- . . . -š: [ineg] maħadaaš/ maħaddiš: [ineg] maʕumriš: [ineg] MA
Op¬ [ineg∅]
NCIs: [uneg∅]
ma- . . . -š: [ineg]
maa: [uneg] laa-wala: [uneg∅]
QA
maa: [ineg]
NCIs: [uneg]
SA
maa: [ineg]
NCIs: [uneg]
laa/lam/lan: [ineg]
laa-wala: [uneg]
laa-wala: [uneg]
coordinators carry the uninterpretable negation feature [uneg∅] in MA. Recall that these coordinators in MA, just like all other polarity items, always co-occur with the marker maa, which is not semantically negative. The coordinators, then, are always licensed by the covert negative operator on a par with all other polarity items. However, the coordinators laa-wala in the other varieties of Arabic carry the feature [uneg], which can be licensed by NM maa or NIs like maħadaaš ‘nobody,’ maʕumriš ‘never’. Alternatively, a covert negative operator licenses the coordinators laa-wala as a last resort in the absence of NMs or NIs that can license these NC coordinators.
4.2.4.3. Ellipsis
Let us now consider ellipsis facts relating to the second disjunct and constraints on the interaction between JA’s bipartite negation and disjunction. First, it is possible to elide the VP of the second disjunct: (140)
laa šaaf-na Zeed neither saw.3msg-us Zeed ‘Neither Zeed nor Omar saw us.’
wala nor
šaaf-na saw.3msg-us
ʕumar Omar
Crucially, it is the disjunction particle wala that licenses this VP ellipsis. Compare this data with wala to that in example (141) with the conjunction wa-. In the following data, the conjunction wa-fails to license VP ellipsis in JA:
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 159 ]
0 6 1
(141)
*Zeed ma-šaaf-naa-š wa-ʕumar Zeed neg-saw.3msg-us-neg and -Omar ‘Zeed did not see us, and Omar did not see us.’
ma-šaafnaa-š neg-saw.3msg-us-neg
Much previous research has shown that VP ellipsis can be licensed by an NM (Potsdam, 1996; Lobeck, 1995). So, can wala license VP ellipsis because it is negative or because it is a disjunction operator? There is evidence that even walla/willa as particles of affirmative disjunction can license VP ellipsis in JA. Ellipsis of the VP of the second disjunct is possible in such contexts: ʕumar? Omar
(142)
šuf-tu Zeed walla/ willa šuf-tu saw-2mpl Zeed or saw-2mpl ‘Did you see Zeed or (did you see) Omar?’
(143)
šuf-tu Zeed mbaariħ walla/ willa šuf-tu saw-2mpl Zeed yesterday or saw-2mpl ‘Did you see Zeed yesterday or (did you see) Omar?’
ʕumar? Omar
In this data, an adverb separates the two direct objects. This suggests that the two objects are not a conjoined NP of one verb but, rather, two objects of two verbs, the overt verb and the elided verb. The following section builds on the data shown here by positing that disjoined DP coordinate complexes that agree with a singular verb involve Bare Argument Ellipsis (BAE). I then contrast such structures with closest conjunct agreement structures, which, I argue, do not involve BAE.
4.2.4.4. Bare argument ellipsis and coordinate complexes
BAE is a type of ellipsis that involves the deletion of all elements in the clause except for one constituent, the thematic argument in the following example (Hudson, 1976; Hankamer & Sag, 1976; Neijt, 1979; Moltmann, 1992; Johnson, 1996; Schwarz, 1999; Depiante, 2000; Kolokonte, 2006; Culicover & Jackendoff, 2005): (144)
John bought a book yesterday, and he bought a newspaper yesterday. (Zhang, 2010, p. 24)
Syntactic tests can identify whether a sentence has BAE. One such test involves collective predicates. Sentences with BAE with a singular subject argument
[ 160 ] A Unified Theory of Polarity Sensitivity
1 6
are incompatible with collective predicates because such predicates require a plural subject. The second test discussed in this section focuses on the use of DPs with mixed gender to intercept cases of BAE. (145)
Bare Argument Ellipsis is incompatible with: a. Collective predicates b. DP coordinate complex with mixed gender as external argument
The reasoning is such that incompatibility with (a) and (b) indicates that the coordinate complex in question involves BAE whereas compatibility with (a) and (b) indicates the lack of BAE. To begin with our tests, collective predicates (and reciprocal verbs with the tafaaʕal pattern) like ltaga ‘meet’ and require a plural NP or a conjoined DP complex as external arguments (see Aoun et al., 1999; Benmamoun, 2003): (146)
(147)
(148)
a.
ltaga al-awlaad/ (*al-walad) met the-boys (*the-boy) ‘The boys met.’
b.
ltaga Zeed w-ʕumar met Zeed and-Omar ‘Zeid and Omar met.’
a.
ma-ltagaa-š Zeed w-ʕomar neg-met-neg Zeed and-Omar ‘Zeid and Omar did not meet.’
b.
*ma-ltagaa-š neg-met-neg
c.
ma-ltaqaaw-šw Zeed u-ʕomar neg-met-neg Zeed and-Omar ‘Zeid and Omar did not meet.’
d.
*ma-ltaqaaw-šw neg-met-neg
a.
ltaga l-ustaað wa met.3msg the-teacher and ‘The teacher and the students met.’
laa neither
laa neither
(JA)
(JA)
Zeed Zeed
Zeed Zeed
(JA)
wala nor
wala nor
ʕumar Omar
(JA)
(MA)
ʕumar Omar
l-tˁullaab the-students
(MA)
(JA)
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 161 ]
2 6 1
(149)
(150)
b.
*ltaga l-ustaað ʔaw l-tˁullaab met.3msg the-teacher or the-students ‘*The teacher or the students met.’
c.
*ltaga l-tˁullaab ʔaw met.3msg the-students or ‘*The teacher or the students met.’
l-ustaað the-teacher
a.
tlaqa l-ustaad w met.3msg the-teacher and ‘The teacher and the students met.’
tˁ-tˁalaba the-students
b.
*tlaqa l-ustaad wəlla met.3msg the-teacher or ‘*The teacher or the students met.’
c.
*tlaqaw tˁ-tˁalaba wəlla met.3msg the-students or ‘*The teacher or the students met.’
a.
*maa ltaga laa l-ustaað wala l-tˁullaab (JA) neg met.3msg neither the-teacher nor the-students ‘*Neither the teacher nor the students met.’
b.
*maa ltaga laa l-tˁullaab neg met.3msg neither the-students ‘*Neither the teacher nor the students met.’
tˁ-tˁalaba the-students
(JA)
(JA)
(MA)
(MA)
l-ustaad (MA) the-teacher
wala l-ustaað (JA) nor the-teacher
In all of the preceding examples, conjunction constructions are compatible with the collective predicate while disjunction constructions are not. This suggests that disjunction constructions involve BAE. In cases where the coordinate complex laa-wala triggers plural agreement, it is reasonable to assume that both conjunction and disjunction constructions involve the entire coordinated DP entering into an agreement relation with one verb with no ellipsis involved. (151)
a.
[DP
laa Zeed wala neither Zeed nor ‘Neither Zeed nor Omar saw us.’
[ 162 ] A Unified Theory of Polarity Sensitivity
ʕumar [TP šaaf-u-na (JA) Omar saw-3mp-us
3 6 1
b.
[DP Zeed wa-ʕomar Zeed and-Omar ‘Zeid and Omar saw us.’
[TP šaaf-*(u)-na saw-*(3mp)-us
(JA)
To give more evidence that the coordinated DPs in such cases involve ellipsis, consider the following example where the adverb mbaariħ intervenes between the external member of coordinate complex and the coordinator itself. These facts indicate that the two DPs are not a constituent; i.e., there is no DP coordination, and there is clausal coordination with BAE. (152)
(153)
a.
šaaf Zeed mbaariħ ʔaw (šaaf) ʕumar saw.3msg Zeed yesterday or saw.3msg Omar ‘He either saw Zeed yesterday or Omar.’
b.
šuft Zeed mbaariħ walla (šuft) saw.2msg Zeed yesterday or saw.2msg ‘Did you see Zeed or (did you see) Omar?’
c.
laa šaaf Zeed mbaariħ wala (šaaf) neither saw.3msg Zeed yesterday nor saw.3msg ‘He neither saw Zeed yesterday nor Omar.’
a.
šaaf Zeed lbaarəħ wəlla (šaaf) ʕumar saw.3msg Zeed yesterday or (saw.3msg) Omar ‘He either saw Zeed yesterday or Omar.’
b.
šəft-i Zeed lbaarəħ wəlla (šəft-i) saw.2msg Zeed yesterday or saw.2msg ‘Did you see Zeed or (did you see) Omar?’
c.
maa šaaf laa Zeed lbaarəħ wala (šaaf) ʕumar (MA) neither saw.3msg neither Zeed yesterday nor saw.3msg Omar ‘He neither saw Zeed yesterday nor Omar.’
(JA)
ʕumar? Omar
(JA)
ʕumar (JA) Omar
(MA)
ʕumar? Omar
(MA)
Considering the fact that first conjunct agreement involves a singular agreement on the verb with a plural subject in VS (verb-subject) word order, the singular agreement on the verb in example (51)a is quite interesting because the word order is SV (subject-verb). Is this construction, then, similar to first conjunct agreement? To answer that, we need to look at a subject with disjoined nouns that are mixed in gender. If my analysis is on the right track, we also expect BAE to be incompatible with mixed gender DPs, i.e., DP coordination
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 163 ]
4 6 1
without ellipsis. This is, indeed, borne out. In examples of such contexts, it is not possible to have singular agreement when the negatively coordinated DPs are mixed in gender: (154)
(155)
a.
laa Zeed wala layla (*šaaf-na)/ neither Zeed nor Layla saw.3msg-us/ ‘Neither Zeed nor Layla saw us.’
šaaf-u-na saw-3mpl-us
b.
laa Zeed wala layla (*šaafat-na)/ šaaf-u-na neither Zeed nor Layla saw.3fsg-us/ saw-3mpl-us ‘Neither Zeed nor Layla saw us.’
c.
Zeed wa layla Zeed and Layla ‘Zeid and Layla saw us.’
a.
laa Zeed wala layla (maa) (*šaaf-na)/ šaaf-u-na (MA) neither Zeed nor Layla (neg) saw.3msg-us/ saw-3mpl-us ‘Neither Zeed nor Layla saw us.’
b.
laa Zeed wala layla (maa) (*šaafat-na)/ šaaf-u-na neither Zeed nor Layla (neg) saw.3fsg-us/ saw-3mp-us ‘Neither Zeed nor Layla saw us.’
c.
Zeed w layla (*šaaft-na)/ šaaf-u-na Zeed and Layla saw.3fsg-us/ saw-3mpl-us ‘Zeid and Layla saw us.’
(*šaafat-na)/ šaaf-u-na saw.3fsg-us/ saw-3mpl-us
(JA)
(JA)
(JA)
(MA)
(MA)
The ungrammaticality of these examples is expected if the reconstructed VP must match the identity of the second VP. In other words, the ungrammaticality can be reduced to the following, a reconstructed VP that does not agree with its subject in gender: (156)
a.
*laa neither
Layla Layla
šaaf-na saw.3msg-us
wala nor
b.
*laa neither
Layla Layla
šaaf-na saw.3msg-us
wala nor
Zeed Zeed
šaaf-na saw.3msg-us
(JA)
Zeed šaaf-na (MA) Zeed saw.3msg-us
Therefore, there must be a constraint that VP ellipsis is only possible if the two verbs are identical, allowing us to recover (reconstruct) the elided VP through [ 164 ] A Unified Theory of Polarity Sensitivity
5 6 1
the second VP. When the masculine feature of the verb šaaf-na ‘saw.3msg-us’ enters into an agreement process with the feminine subject, the external disjunct Layla, the syntactic structure crashes. Support for this analysis can be seen in cases where both of the disjoined nouns are feminine. Crucially, the verb can be singular, thus giving further support to the ellipsis analysis:14 (157)
(158)
a.
laa Layla šaafat-na wala neither Layla saw.3fsg-us nor ‘Neither Layla nor Lubna saw us.’
Lubna Lubna
b.
laa Layla šaaft-na wala neither Layla saw.3fsg-us nor ‘Neither Layla nor Lubna saw us.’
Lubna Lubna
c.
maa šaaft-na laa Layla neg saw.3fsg-us neither Layla ‘Neither Layla nor Lubna saw us.’
wala nor
a.
*ʔimma either
Zeed Zeed
ʔaw or
b.
*ʔimma either
Zeed Zeed
ʔaw or
Layla Layla
c.
*ʔimma either
Zeed Zeed
ʔaw or
Layla Layla
ʕumar Omar
šaafat-na saw.3fsg-us
(JA)
šaaft-na (MA) saw.3fsg-us
Lubna Lubna
ltaga met.3msg
(MA)
(JA)
saafarat traveled.3fsg
(JA)
saafar traveled.3msg
(JA)
If the coordinated DPs involve VP ellipsis, we should be able to reconstruct the elided VP. This is, indeed, possible in JA: (159)
laa Zeed šaaf-na wala neither Zeed saw.3msg-us nor ‘Neither Zeed saw us nor Omar saw us.’
(160)
a.
laa Layla šaaf-at-na neither Layla saw-3fsg-us ‘Neither Layla nor Zeed saw us.’
wala nor
ʕumar Omar
Zeed Zeed
šaaf-na saw.3msg-us
šaaf-na saw.3msg-us
14. I thank Abbas Benmamoun for pointing out the relevance of this contrast to the analysis.
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 165 ]
61
b.
*laa neither
Layla Layla
šaaf-at-na saw-3fsg-us
wala nor
Zeed Zeed
c.
laa Layla šaafat-na neither Layla saw.3fsg-us ‘Neither Layla nor Lubna saw us.’
wala nor
Lubna Lubna
šaaf-na saw.3msg-us šaafat-na saw.3fsg-us
The ungrammaticality of these examples can be explained if the reconstructed VP must match the identity of the overt VP causing gender agreement failure. In the phi- agreement paradigm, agreement morphology (particularly grammatical gender) on the verb is determined by the first conjunct. This is observed in the following examples where the feminine DP, Layla, triggers the feminine agreement on the verb. (161)
a.
ya-drus-u Zayd-un 3msg-study.ipfv-ind Zayd-NOM ‘Zayd and Layla are studying.’
wa-Layla and-Layla
b.
*ta-drus-u Zayd-un wa-Layla 3fsg-study.ipfv.-ind Zayd-NOM and -Layla ‘Zayd and Layla are studying.’
c.
ta-drus-u Layla 3fsg-study.ipfv-ind. Layla ‘Layla and Zayd are studying.’
d.
*ya-drus-u Layla 3msg-study.ipfv-ind Layla ‘Layla and Zayd are studying.’
(SA)
(SA)
wa-Zayd-un (SA) and-Zayd-NOM
wa-Zayd-un and-Zayd-NOM
(SA)
Notice that the agreement involved here is gender agreement. It is not possible to probe number agreement in such a context because SA (Standard Arabic) has only partial agreement in VS word orders. Partial agreement means that the verb does not carry plural agreement even when the subject is plural. (162)
a.
ya-drus-u 3msg-study.ipfv-ind ‘The boy is studying.’
al-walad-u the-boy
[ 166 ] A Unified Theory of Polarity Sensitivity
(SA)
7 6 1
b.
ya-drus-u 3msg-study.ipfv-ind ‘The boys are studying.’
al-ʔawlaad-u the-boys-NOM
(SA)
To recap, this section has shown that coordinate complexes with the conjunction particle wa ‘and’ do not have BAE when the verb is singular. Though Closest Conjunct Agreement (CCA) is largely due to non-syntactic factors (adjacency, linear order, processing; see Lorimor, 2007; Cowart & McDaniel, 2008; among others), tests using collective predicates and mixed gender DPs show that CCA is sometimes a syntactic effect. Specifically, it is due to BAE in certain coordinate complexes. Coordinate complexes with the disjunction particle wala ‘nor,’ ʔaw ‘free choice or,’ or walla ‘disjunctive or’ involve BAE when the verb is singular. As for the nature of agreement and syntactic category of the head of coordination, the question is whether coordination phrases have a special syntactic category (e.g., &P) or a regular one (e.g., DP, CP). The first view is that coordination projects a special functional category. This functional category is &P (CoP, ConjP) in (Munn, 1987; Zoerner, 1995; Johannessen, 1998; Fromkin et al., 2007). Other researchers (den Dikken, 2006) label the functional category for both conjoining and disjoining JP (junction phrase) (‘J’ for Junction to cover both Conjoining and Disjoining). The second view is that coordination structures do not project a special functional projection, and they take the syntactic category of the external member of coordination: DP, CP, . . . etc. (Zhang, 2010). Case is one argument for a special functional category. If coordinators license case on the member(s) of the coordination complex, it is possible to argue that there is a functional projection through which the coordinator can enter into a licensing relation. To test that in Arabic, I will focus on SA data since it has overt case markers. The laa-wala coordinate complex in SA can license case on DPs of the external and internal conjuncts (Raajihi, 2000): (163)
laa raʤul-a wala ʔimraʔat-a fii al-bayt neither man-acc nor woman-acc in the-house ‘No man or woman is at home.’
(SA)
At first glance, the accusative case assignment seen in this data is evidence for a special &P category that has an accusative case feature. However, I urge caution in the face of the evidence for two reasons. First, it is not clear how to explain the licensing of this case feature in non-nominal complexes (e.g., coordinated verbs). Likewise, nominative case may also appear in similar contexts in SA. Consider the following examples
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 167 ]
8 6 1
which reveal that either DP of coordination conjuncts can be nominative (default case) or accusative (case from the coordinators) in a complex pattern (Raajihi, 2000): (164)
a.
laa raʤul-un wala ʔimraʔat-un fii al-bayt (SA) neither man-nom.indef. nor woman-nom.indef in the-house ‘There isn’t a man or a woman at home.’
b.
laa raʤul-a wala ʔimraʔat-un fii al-bayt (SA) neither man-acc nor woman-nom.indef. in the-house ‘There is no man or a woman at home.’
The complex case patterns exhibited in this data reveal the problems of using case as evidence of a special &P for the head of coordination constructions. Accusative case here might be morphological: related to the word formation process of the constituent [laa + indefinite noun] (see Siloni, 2001, 2002 for construct state adjectives in Hebrew). One can also argue that nominative case is also a default case and not discharged through a functional category. In short, the case facts do not give conclusive evidence that the coordinator has an intrinsic case feature that can be used to argue for &P category. I leave this issue for future research.
4.2.5. Conclusion
In this chapter, I have provided evidence for a distinction between coordination by wa-, a conjunction structure, and coordination by wala, a disjunction structure. Throughout the discussion, I have shown that negative coordination involves a disjunction phrase (DisjP) that takes the first member of the coordination structure as its specifier and the second member as its complement. The NM laa is in Spec,XP (Spec,DP, Spec,CP) of the first disjunct. The disjunction particle wala, on the other hand, is the head of DisjP. In addition, I have argued that the particle wala is the lexicalization of a disjunction operator. This disjunction particle is polarity sensitive and behaves like a non-strict NCI interpreted as ‘nor’ in indicative sentences by requiring an NM when postverbal but not when preverbal. As such, I analyzed this particle as specified for an uninterpretable negation feature [uneg] that is valued via agreement with a negative projection. Finally, I argued that the negative contrast between the disjunction particles walla/willa ‘or’ and wala ‘nor’ lies in the feature structure of coordination rather than in the presence of [+neg] (or lack thereof) under the disjunction head (contra den Dikken’s 2006 analysis of English ‘or’ and ‘nor’). I posited
[ 168 ] A Unified Theory of Polarity Sensitivity
9 6 1
that the negative disjunction wala ‘nor’ is specified for [uneg] while the affirmative disjunction walla/willa is not. As such, the dependency between the DisjP and negation is reduced to a syntactic licensing requirement in that the disjunction particle wala carries a [uneg] feature licensed by an NM or negative operator specified for a [ineg] feature.
L i c e n s i n g N e g at i v e C o n c or d I t e m s
[ 169 ]
0 7 1
1 7
CHAPTER 5
PSIs with Head-Like Properties
T
his chapter discusses two main issues that arise from PSIs (polarity- sensitive items) with head-like properties. These PSIs seem to be outside the (immediate) domain of their licensor. The first issue is how these PSIs are licensed in syntax and how a unified analysis can handle their distribution. I argue that these PSIs are adverbial phrases that do not project a clausal projection and that negation licenses these PSIs either in Spec-NegP or under c-command. This unified analysis does not appeal to the problematic head– complement relation as a putative licensing configuration. Another issue that arises from these NPIs with head-like properties is their ability to host clitics with accusative and genitive case marking. This issue raises interesting questions pertaining to case theory and dependent case licensing. I argue that negation licenses the puzzling accusative case of the pronominal complement, a conclusion with far-reaching implications for dependent case licensing in natural language.
5.1. HEAD-L IKE PSIS AS ADVERBS WITH PHRASAL PROJECTION
I analyze the status of the temporal NPI (negative polarity item) ʕumr and NCI (negative concord item) baʕd in Jordanian Arabic (JA) and explain their distribution and syntactic licensing.1 Despite their head-like properties (hosting 1. Section 5.1 is an expanded and revised version of the article “Temporal NPIs and NCIs as adverb phrases. The case of Jordanian Arabic”, Alqassas, A., pp. 129-152 in: Perspectives on Arabic Linguistics XXVIII. Papers from the Annual Symposium on Arabic Linguistics, Gainesville, Florida, 2014, Haddad, Y.A. & Potsdam, E., eds. (2016). Published by John Benjamins Publishing Company. Amsterdam/Philadelphia. I thank the editors, ASAL 28 organizers and audience for their helpful comments. All errors are mine. A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0005.
2 7 1
clitics and assigning case), I argue that both elements are AdvPs in Spec-XP positions rather than heads projecting their own clausal structure. These items can be preverbal or postverbal; they differ in their ability to precede negative constituents; they require a complement (DP [determiner phrase], CP [complementizer phrase]) that can be coreferential with the subject or object of the clause, or a pronominal complement inside a syntactic island. I contend that the similarities and differences among these properties follow from the fact that ʕumr may be base generated either preverbally and postverbally, while baʕd is always base generated postverbally and optionally moves to a preverbal position. I conclude that both c-command and specifier-head configurations can license such items, while the head-complement configuration cannot. Generally speaking, NPIs are words that display syntactic dependency with sentential negation. NPIs can be argumental—such as anyone and anything in Standard English—or adverbial—such as English at all and yet. This chapter focuses on the temporal NPI ʕumr and the NCI baʕd, whose English equivalents are the adverbs ever and yet, respectively. I use the term PSIs to refer to both the NPI and the NCI. The NCI wala-ħada ‘anyone, no one’ and the negative compound maħadaaš ‘no one’ will also be discussed from the perspective of their interaction with the NPI ʕumr ‘ever’ and the NCI baʕd ‘yet.’2 PSIs like ʕumr and baʕd are interesting because they exhibit head-like properties that put them on a par with verbs. They can host pronominal clitics and negation. The Moroccan NCI baqi ‘yet’ can even carry subject agreement inflections. They also assign case to the pronominal clitics they host. Elements with these properties are collectively referred to in the literature as head NPIs (Benmamoun, 2006; Aoun et al., 2010) or as pseudo-verbs (Lucas, 2009) and analyzed as heads that project a clausal projection selecting a NegP, for Moroccan Arabic (MA) in particular. Benmamoun (2006) suggests that such “head NPIs” are licensed under the head–complement relation with negation. In this chapter, I highlight several additional properties of these items that challenge their existing analysis as clausal heads. These properties include (i) their ability to appear both preverbally and postverbally, (ii) their requirement to have a (pro)nominal complement, and (iii) certain contrasts between these two items with respect to their interaction with the NCI wala-ħada ‘anyone, no one’ and the negative compound maħadaaš ‘no one.’ I argue that an analysis of these items as adverbs in Spec-XP captures both their head and non-head properties. The analysis is further fleshed out by treating ʕumr as a TP (tense phrase) adverb and baʕd as a VP (verb phrase) adverb that can move to higher specifier positions.
2. My JA examples are from the Jordanian Houran. [ 172 ] A Unified Theory of Polarity Sensitivity
3 7 1
Section 5.1.1 lays out the basic facts. Section 5.1.2 presents the two options for the analysis of ʕumr and baʕd: as heads in a clausal projection and as XPs (X phrases) in specifier position. A brief assessment of the benefits and drawbacks of each analysis is also presented. In section 5.1.3, I begin by presenting evidence that single negation (using the negative marker maa) is located in a NegP above TP, while bipartite negation is located below TP. Building on this claim, I then analyze the two PSIs ʕumr and baʕd as XPs in specifier position. I further develop this analysis to account for the contrast in the behavior of ʕumr and baʕd with respect to licensing by the NCI wala-ħada ‘anyone, no one’ and the negative compound maħadaaš ‘no one.’
5.1.1. Empirical generalizations
Temporal words like ʕumr ‘ever’ and baʕd ‘yet’ in JA require the presence of a negative element in sentences where they are interpreted as the negative constituents ‘never’ and ‘yet.’3 (1)
a.
b.
ʕumr-o *(maa) zaar ever-him neg visited.3msg ‘He has never visited Petra.’ baʕd-o *(maa) zaar yet-him neg visited.3msg ‘He has not visited Petra yet.’
el-batra def-Petra
el-batra def-Petra
Both items can be either initial or final; note, however, that postverbal ʕumr is a marked option, while both positions of baʕd are unmarked. (2)
a.
(ʕumr-o) maa zaar ever-him neg visited.3msg ‘He has never visited Petra.’
el-batra def-Petra
(ʕumr-o)
b.
(baʕd-o) maa zaar el-batra yet-him neg visited.3msg def-Petra ‘He has not visited Petra yet.’
(baʕd-o)
3. The same is true of their Egyptian equivalents ʕumr and lissa and the MA equivalents ʕəmmər and baqi. P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 173 ]
4 7 1
Both ʕumr and baʕd can either follow or precede the subject. This observation was initially made in Benmamoun (2006, p. 145) for MA and holds true for JA as well, as shown in (3) and (4). Benmamoun takes the ability of the subject to intervene between the PSIs and the negative marker as evidence that these elements do not form a compound with negation.4 (3)
(4)
a.
ʔaħmad ʕumr-o maa zaar Ahmad ever-him neg visited.3msg ‘Ahmad has never visited Petra.’
el-batra def-Petra
b.
ʔaħmad baʕd-o maa zaar Ahmad yet-him neg visited.3msg ‘Ahmad has not visited Petra yet.’
el-batra def-Petra
a.
ʕumr ʔaħmad maa zaar ever Ahmad neg visited.3msg ‘Ahmad has never visited Petra.’
el-batra def-Petra
b.
baʕd ʔaħmad maa zaar yet Ahmad neg visited.3msg ‘Ahmad has not visited Petra yet.’
el-batra def-Petra
Whenever either ʕumr or baʕd is not followed by the subject, it must carry a pronominal clitic. If neither a clitic nor the subject NP (noun phrase) follows the PSI, the sentence will be ungrammatical:5 (5)
a.
ʕumr*(-o) maa zaar ever*(-him) neg visited.3msg ‘He has never visited Petra.’
b.
ʕumr *(ʔaħmad) maa zaar ever *(Ahmad) neg visited.3msg ‘Ahmad has never visited Petra.’
c.
baʕd*(-o) maa zaar yet*(-him) neg visited.3msg ‘He has not visited Petra yet.’
el-batra def-Petra
el-batra def-Petra
el-batra def-Petra
4. This makes their licensing a puzzle. They are not c-commanded by negation, nor are they in a specifier–head relation with negation. 5. Hoyt (2007) presents examples from a Palestinian dialect where ʕumr appears without the pronominal clitic or the NP complement. Such cases are ungrammatical in JA. [ 174 ] A Unified Theory of Polarity Sensitivity
5 7 1
d.
baʕd *(ʔaħmad) maa zaar yet *(Ahmad) neg visited.3msg ‘Ahmad has never visited Petra.’
el-batra def-Petra
The requirement that a nominal element follow ʕumr and baʕd may be satisfied, not only by the subject NP or a pronominal subject clitic, but also by a pronominal clitic coreferential with the object, as in (6)a:6,7 (6)
a.
ʕumr-o maa ħabbat-o bint ever-him neg love-him girl ‘A girl never loved him’
b.
ha-l-walad ʕumr-o maa this-the-boy ever-him neg ‘This boy, a girl never loved him’
ħabbat-o love-him
bint girl
Despite the parallels just described, the PSIs ʕumr and baʕd do behave differently in the presence of a negative indefinite (NI). In particular, the adverb baʕd can occur before an NI, but the adverb ʕumr cannot: (7)
a.
*ʕumr-o maħadaaš zaar ever-him no one visited.3msg ‘No one has ever visited Petra.’
el-batra def-Petra
b.
baʕd-o maħadaaš zaar yet-him no one visited.3msg ‘No one has visited Petra yet.’
el-batra def-Petra
It is important to reinforce that, although I have grouped ʕumr and baʕd together as PSIs, they belong to two different classes. Recall from section 1.1 that ʕumr is an NPI while baʕd is an NCI. Next, I highlight the important
6. Hoyt (2007) presents examples from a Palestinian dialect where the pronominal clitic on the NPI and the subject NP are not coreferential. We see the same thing in example (6) from JA. 7. It is worth pointing out in this context that an NP in the left periphery can precede the NPI phrase, as ha-l-walad ‘this boy’ does in (6)b. Given the fact that the pronominal clitic of the NPI phrase is coreferential with this preposed NP ha-l-walad, it is possible that the pronominal clitic of the NPI in (6)a may be coreferential with a similar NP in the discourse. I am grateful to Youssef Haddad for pointing out that it may be the presence of an element in the left periphery that allows the pronominal clitic of the NPI phrase to corefer with the object clitic.
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 175 ]
6 7 1
properties that distinguish ʕumr (and the category of NPIs) from baʕd (and the category of NCIs). First, as pointed out in Soltan (2012) for lissa ‘yet,’ JA NCIs like baʕd are formally negative, while NPIs like ʕumr are not. (8)
a.
ʕumr-o zaar ever-him visited.3msg ‘Has he ever visited Petra?’
(9)
a.
ʔaħmad naam? Ahmad slept.3msg ‘Has Ahmad slept?’
(10) *baʕd-o zaar yet-him visited.3msg ‘Has he visited Petra yet?’
el-batra? def-Petra
b.
b.
Answer:
Answer:
ʕumr-o ever-him ‘Never’
baʕd-o yet-him ‘Not yet’
el-batra? def-Petra
(11) ðakkir-ni ʔiða ʕumr-o zaar Remind-me if ever-him visited.3msg ‘Remind me if he ever visited Petra.’
el-batra def-Petra
(12) *ðakkir-ni ʔiða baʕd-o zaar el-batra Remind-me if yet-him visited.3msg def-Petra ‘Remind me if he has visited Petra yet.’
The fact that baʕd, but not ʕumr, can be used as a fragment answer with a negative interpretation suggests that baʕd is formally negative (carrying a [uNeg] feature) while ʕumr is not. Moreover, the fact that baʕd cannot appear in nonnegative polarity environments, e.g., the interrogative in (10) and the conditional in (12), while ʕumr can, e.g., (8) and (11), also shows that baʕd can only be licensed by negation while ʕumr can be licensed in nonnegative contexts. Again, this is another fact that suggests that baʕd carries a [uNeg] feature while ʕumr does not. Second, while baʕd and ʕumr show the same basic distribution, only baʕd is unmarked in both preverbal and postverbal position, as noted previously; ʕumr, by contrast, occurs predominantly in preverbal position, and the postverbal position is marked. (13) a.
baʕd-o maa saafar yet-him neg traveled.3msg ‘He hasn’t traveled yet.’
[ 176 ] A Unified Theory of Polarity Sensitivity
(unmarked)
71
b.
maa saafar neg traveled.3msg ‘He hasn’t traveled yet.’
baʕd-o yet-him
(unmarked)
c.
ʕumr-o maa ever-him neg ‘He has never traveled.’
saafar traveled.3msg
(unmarked)
d.
maa saafar neg traveled.3msg ‘He has never traveled.’
ʕumr-o ever-him
(marked)
Third, baʕd can precede bipartite negation, while ʕumr cannot—although both elements can follow bipartite negation. Soltan (2012) makes a similar observation for Egyptian Arabic (EA), showing that lissa behaves like JA baʕd: (14) a.
b.
baʕd-o ma-saafar-iš yet-him neg-traveled.3msg-neg ‘He hasn’t traveled yet.’ ma-saafar-iš neg-traveled.3msg-neg ‘He hasn’t traveled yet.’
c.* ʕumr-o ever-him ‘He has never traveled.’ d.
ma-saafar-iš neg-traveled.3msg-neg ‘He has never traveled.’
baʕd-o yet-him
ma-saafar-iš neg-traveled.3msg-neg
ʕumr-o ever-him
5.1.2. The syntactic status of head-l ike PSIs
Conflicting empirical evidence regarding the status of these items has made it difficult to classify them categorically as syntactic heads (X0) or adverbs (XPs). Next, I lay out the empirical arguments for and against each analysis. Determining the status of these elements has important bearing on two theoretical debates. The first debate concerns the necessity for these items to project their own distinct maximal projection (as Benmamoun, 2006, contends
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 177 ]
8 7 1
for temporal NPIs in MA). The second debate has to do with the configurations under which these NPIs are syntactically licensed by negation. Standardly, Arabic NPIs and NCIs are assumed to be licensed under c-command by negation or positionally in Spec-NegP (Benmamoun, 1997, 2006). However, if these items are actually clausal heads in a maximal projection XP (X phrase) above NegP, neither of these licensing configurations can explain the syntactic dependency between negation and these items in the absence of covert neg raising.
5.1.2.1. Head-like PSIs as syntactic heads selecting a NegP
The MA NPI ʕəmmər ‘ever’ and NCI baqi ‘yet’ have been analyzed as heads of a clausal projection XP (Benmamoun, 2006), as in (17). There are three main empirical arguments for this analysis in Benmamoun (2006). First, these items can host clitics (as in all the previous examples) and can carry subject agreement inflection; this latter property is typically associated with heads in Arabic, as in the MA examples in (15–16) (Benmamoun, 2006, p.144): (15) a.
b.
nadya baq-a ma-žat Nadia yet- FS neg-came.3fsg ‘Nadia hasn’t come yet.’ lə-wlad baq-yin the-children yet-pl ‘The children haven’t come yet.’
ma-žaw neg-came.3mp
(16) nadya ʕəmmər-ha ma-žat Nadia never-her neg-came.3fsg ‘Nadia never came.’ (17)
XP X
NegP
bagi/
Second, the fact that these PSIs can be preceded by the sentential subject (as shown from previous MA examples) suggests that these items are heads of an XP with an EPP-feature that attracts subjects to its specifier. Third, the NPI ʕəmmər/ ʕumr assigns accusative case to the subject NP or pronominal clitic following it in Spec-NegP, as shown for MA in (18) and JA in (5). Hosting clitics is an exclusive property of heads in Arabic, and case
[ 178 ] A Unified Theory of Polarity Sensitivity
9 7 1
assignment here is consistent with the configuration in (17). The MA examples in (18) are from Benmamoun (2006, p. 145). (18) a.
b.
ʕəmmər nadya ma-mšat l-təmma never Nadia neg-went.3fsg to-there ‘Nadia never went there.’ ʕəmmər-ni ma-mšit never-me neg-went.1s ‘I never went there.’
l-təmma to-there
Benmamoun proposes that these items are licensed in a head-complement configuration in MA. He convincingly argues against covert movement of the negative marker to a position c-commanding the NPI and against reconstruction of the NPI to a position c-commanded by negation. Covert head movement of negation to a position c-commanding the NPI incorrectly predicts that the negative head should be able to license the NPI ʕumr even when the auxiliary kaan ‘be’ separates the NPI from negation. Consider the following examples from JA in (19) and MA in (37): (19) a.
b.
(20) a.
b.
ʕumr-o maa kaan ever-him neg was ‘He never loved her.’
yħib-ha love-her
(*ʕumr-o) ever-him
yħib-ha love-her
kaan was
maa neg
ʕəmmər-u ma-kan never-him neg-was ‘He never loved Nadia.’ *ʕəmmər-u never-him
kan was
taybɣi love
nadya Nadia (Benmamoun, 2006, p. 148)
ma-taybɣi nadya neg-love Nadia
(Benmamoun, 2006, p. 148)
Moreover, in general, covert neg movement incorrectly introduces ambiguities not supported by empirical evidence. As Merchant (2000) points out, in English, the sentence Bob doesn’t often finish on time only allows the reading not > often. If negation could move higher at LF (Logical Form), we would incorrectly predict the availability of both readings, often > not and not > often.8 8. For arguments against NEG movement at LF, see Ladusaw (1977, 1988) and Höhle (1991).
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 179 ]
0 8 1
However, analyzing these NPIs as clausal heads cannot explain the contrast in examples (7)a, b, repeated in (21): (21) a.
*ʕumr-o maħadaaš zaar ever-him no one visited.3msg ‘No one has ever visited Petra.’
el-batra def-Petra
b.
baʕd-o maħadaaš zaar yet-him no one visited.3msg ‘No one has visited Petra yet.’
el-batra def-Petra
If these items are heads that can be licensed by a negative complement, and if the negative compound maħadaaš is in a NegP projection, we incorrectly predict (21)a to be grammatical. Conversely, if the negative compound maħadaaš is not in a NegP, we incorrectly predict (21)b to be ungrammatical. Moreover, analyzing these items as clausal heads becomes less plausible once we take into account their positional flexibility; recall that these items can be either preverbal and postverbal, as shown in examples (2)a, b and repeated in (22).9 This flexibility is normally a property of XPs, such as adverbs and noun phrases, rather than clausal heads. (22) a.
(ʕumr-o) maa zaar ever-him neg visited.3msg ‘He has never visited Petra.’
el-batra (ʕumr-o) def-Petra
b.
(baʕd-o) maa zaar yet-him neg visited.3msg ‘He has not visited Petra yet.’
el-batra (baʕd-o) def-Petra
5.1.2.2. Head-like PSIs as syntactic adverbs in Spec-XP
In this section, I will illustrate how, by analyzing ʕumr and baʕd as adverbs, we can avoid the challenges faced by the clausal-head account.10 First, as adverbs, ʕumr and baʕd can freely occupy both the preverbal and postverbal positions under either the XP-adjunction analysis or the Spec-XP analysis, outlined immediately next. 9. Postverbal ʕumr is marked, but postverbal baʕd is not. 10. Hoyt (2007) considers ʕumr ‘ever’ in Palestinian Arabic to be an adverb in the CP layer. Soltan’s (2012) analysis of ʕumr ‘ever’ and lissa ‘yet’ in EA is based on the assumption that these adverbs are located in Spec-NegP when they precede negation. [ 180 ] A Unified Theory of Polarity Sensitivity
1 8
The literature on adverb placement suggests at least two strategies for allowing adverbs to occupy multiple positions. The first approach—the XP- adjunction analysis—allows phrasal adjunction to more than one maximal projection (Pollock, 1989; Iatridou, 1990; Johnson, 1991; Bowers, 1993; Potsdam, 1999; among others); the second approach—the Spec-XP analysis— base-generates adverbs in the specifier of the lowest maximal projection they modify and permits subsequent movement up to higher specifiers (Jackendoff, 1981; Alexiadou, 1994; Kayne, 1994; Cinque, 1999; among others). In principle, an analysis of ʕumr and baʕd as adverbs could adopt either of these strategies, each of which could straightforwardly accommodate their appearance in multiple structural positions. As will become clear later in this chapter, the NCI baʕd is best explained as an adverb that is merged in Spec-VP and moves to Spec-CP. This approach, as I show in section 5.1.4, will enable us to explain the licensing of baʕd when it precedes the NI. By generating baʕd postverbally and allowing it to move up to the preverbal position, we open up the option for licensing under c-command by negation. The preverbal position of the NPI ʕumr, on the other hand, is best accounted for by the XP-adjunction analysis; I will argue that this adverb is base generated flexibly, either in Spec-VP (in the case of postverbal ʕumr) or in the specifier position of a functional projection above TP, Spec-FP or Spec-NegP (in the case of preverbal ʕumr). However, the adverbial analysis alone cannot account for all the outstanding properties of these PSI elements. It remains to be determined how the subject can intervene between an PSI element and negation, how the subject can precede both the PSI element and negation, and how the PSI can assign case to a subject that intervenes between the PSI and negation. I will tackle each of these issues in turn within the context of my broader adverbial analysis of ʕumr and baʕd. 5.1.3. Analysis of head-l ike PSIs as phrasal heads
First, I assume that single negation, manifested by the marker maa, resides in the head of a NegP above TP (Alqassas, 2015). Note that, while both JA and MA have bipartite negation in most contexts, only single negation is possible with MA ʕumr and baqi and JA ʕumr. (23)
NegP Spec
Neg’ Neg0 maa
TP Spec
T’ T0
VP
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 181 ]
2 8 1
Single negation in JA (and a similar variety in Palestinian Arabic reported in Hoyt, 2007) is also attested in the context of oath words like walla ‘by God.’ What is relevant here is not the distribution of single and bipartite negation in these dialects but, rather, the contrast between the two types of negation with respect to the positioning of adverbial phrases: specifically, the fact that an adverbial phrase can intervene between the negative marker maa and the verb in the case of single negation (24), but not in the case of bipartite negation (25): (24) a.
wallah by-God
maa neg
b-yo:m in-day
basaamħ-ak (JA) forgive.1SG-you
‘By God, I will not forgive you in any day.’=‘I will never forgive you’ b.
wallaah ma-[PP fi-hal-lēle ] b-anaam ʕind-ak by-God not in-this-night sleep.1sG at-you ‘I won’t sleep with you this night.’11
c.
ʕumr-i maa b-yo:m zurt el-batra (JA) Never-I neg in-day visited.1sg def-Petra ‘I never visited Petra in any day.’
d.
wllah maa/laa n-saamħ-(l)ak -ši nhaar (MA) by-God neg forgive-you-NEG in-day ‘I will not forgive you in any day.’=‘I will never forgive you’
e.
ʕəmmər-ni maa f-ħyaat-i zərt el-batra (MA) Never-I neg in-life-my visited.1sG def-Petra ‘I never visited Petra in my life.’
(25) a.
b.
(PA)
ma-basaamiħ-k-i:š (JA) neg-forgive.1SG-you-neg ‘I will not forgive you.’ *maa b-yo:m neg in-day
basaamiħ-k-i:š (JA) forgive.1SG-you-NEG
Since the negative marker and the adverbial phrase can both precede the past-tense verb—see (24)c—and assuming that the past-tense verb moves to T in Arabic (Benmamoun, 2000), it follows that the negative marker maa in (24) 11. This example is from Schmidt and Kahle (1930, §90.6), cited in Hoyt (2007, p. 113).
[ 182 ] A Unified Theory of Polarity Sensitivity
3 8 1
must be located higher than TP. For bipartite negation in Arabic, I maintain the analysis of this structure as a negative projection below TP (Benmamoun, 2000; Aoun et al., 2010; Alqassas, 2012, 2015; among others),12 thus satisfactorily accounting for the contrast between (24) and (25).13 Below is an illustration showing the two proposed positions for negation: (26) a. b.
[NegP Neg maa
[TP
T
[TP T
[NegP Neg ma- iš
[vP . . .
[vP . . .
5.1.3.1. PSIs as XPs
Let us return now to the essential issue at hand: whether ʕumr and baʕd are best identified as clausal heads. I should point out that I do not dispute the insight from Benmamoun’s analysis that these items are heads; however, I do depart from Benmamoun’s assertion that they are clausal heads, as in (27), arguing instead that they are heads within their own (adverbial) phrase, as in (28). (27) XP X’ NegP
X
Neg’ Neg (28)
TP
NegP XP
Neg’ X’
Neg
TP
X
To begin with, note that the ability of an element to host clitics or display agreement may not be a deciding factor in determining definitively whether that
12. In Alqassas (2012, 2015) I present arguments for locating bipartite negation below TP and single negation above TP. 13. Later, I will also draw on this analysis to explain why preverbal ʕumr cannot be followed by bipartite negation.
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 183 ]
4 8 1
element is a clausal head or a phrasal head. Arabic possesses a category of elements known as “circumstantial adverbs” that display gender and number agreement: (29) ɣaadar-u left-3mp ‘They left with haste.’
musriʕ-iin fast-3mp
The element musriʕ-iin ‘fast-3mp’ displays gender and number agreement that matches the masculine plural subject agreement on the verb ɣaadar-u ‘left-3mp.’ Interestingly, musriʕ-iin can also have its own subject—cf. the pronoun hum ‘they’ in ɣaadar-u wa hum musriʕ-iin. This fact suggests that musriʕ- iin is actually an adverbial predicate inside an adverbial clause whose subject (in this case, the pronoun hum ‘they’) can be overt or covert.14 The point here is that musriʕ-iin must be a clausal head (presumably the head of a predicate phrase). By contrast, neither the NPI ʕumr nor the NCI baʕd can have its own subject. This suggests that these elements are not clausal heads, despite the fact that they can carry clitics. Second, ʕumr and baʕd require a nominal complement—either a pronoun or a noun. If we assume that ʕumr and baʕd project a causal XP above NegP, the presence of this complement should necessarily prevent these elements from entering into a head–complement relation with negation. Consider the following examples: (30) a.
ʕumr*(-o) maa zaar ever*(-him) neg visited.3msg ‘He has never visited Petra.’
b.
ʕumr *(ʔaħmad) maa ever *(Ahmad) neg ‘Ahmad has never visited Petra.’
c.
baʕd*(-o) maa zaar yet*(-him) neg visited.3msg ‘He has not visited Petra yet.’
d.
baʕd *(ʔaħmad) maa yet *(Ahmad) neg ‘Ahmad has never visited Petra.’
el-batra (JA) def-Petra
zaar visited.3msg
el-batra (JA) def-Petra
zaar visited.3msg
14. Thanks are due to Youssef Haddad for pointing this out. [ 184 ] A Unified Theory of Polarity Sensitivity
el-batra (JA) def-Petra
el-batra (JA) def-Petra
5 8 1
(31) a.
b.
ʕəmmər(-o) maa zaar ever(-him) neg visited.3msg ‘He has never visited Petra.’
el-batra (MA) def-Petra
ʕəmmər* (ʔaħmad) maa zaar ever *(Ahmad) neg visited.3msg ‘Ahmad has never visited Petra.’
el-batra def-Petra
(MA)
The obligatoriness of these nominal complements suggests that they do not originate as sentential subjects in Spec-TP but instead are directly selected by the NPI/NCI. Arabic’s status as a pro-drop language makes this particularly likely. Therefore, let us revise the representation in (28) to the structure in (32); here, the (pro)noun is under NP, and the NPI/NCI is the head of an XP (labeled “AdvP”), located in Spec-NegP. (32)
NegP Neg
AdvP Adv
NP
Neg
TP
NPI
Further motivation for this analysis comes from the observation, mentioned briefly previously, that the NPI can host an object clitic: (33) ʕumr-o maa ħabbat-o ever-him neg love-him ‘A girl never loved him’
bint girl
Thus, while the NPI requires a nominal argument, this argument can be coreferential with the sentential subject or object.15 Crucially, if ʕumr were to project a clausal XP, its complement would have to be this (pro)nominal element. Consequently, we would have to abandon the head-complement licensing configuration ʕumr is assumed to have with NegP.
15. Lucas (2009, pp. 207, 221) points out that baʕd-o and ʕumr-o are changing into impersonal verbs (pseudo-verbs) that take the logical subject of the sentence as an object pronoun. Here, though, we see that the NPI can also take the logical object of the sentence as an object pronoun.
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 185 ]
6 8 1
(34)
XP X X
NP N
NegP Neg
TP
On the other hand, if the NPI is analyzed as the head of an XP in Spec-NegP projecting a complement position, the presence of the (pro)nominal complement causes no difficulty. The head-complement configuration between the PSI and NegP is also compromised when a relative clause separates the PSI from negation: (35) a.
b.
ʕumr illi budrus never who study.3msg ‘He who studies never fails.’
maa neg
ʕəmmər lli ka-yqra maa never who asp-study.3msg neg ‘He who studies never fails.’
bursub fail.3MSG
(JA)
(ka-)ysqut (asp)fail.3MSG
(MA)
In (34), the relative clause is the complement of the NPI—i.e., it is part of the NPI phrase. If the NPI were a clausal head, the relative clause would have to be understood as a CP complement intervening between the NPI head and the NegP. Again, the NPI could not be in a head–complement relationship with NegP.
5.1.3.2. The NP subject preceding head-like PSIs
In this section, I will show how analyzing the ʕumr and baʕd phrasal heads can handle the challenge presented by the ability of subjects to precede PSIs. The relevant examples are repeated: (36) a.
b.
ʔaħmad ʕumr-o maa Ahmad ever-him neg ‘Ahmad has never visited Petra.’
zaar visited.3msg
el-batra def-Petra
ʔaħmad baʕd-o maa Ahmad yet-him neg ‘Ahmad has not visited Petra yet.’
zaar visited.3msg
el-batra def-Petra
[ 186 ] A Unified Theory of Polarity Sensitivity
7 8 1
First of all, note that there is a debate in the Arabic linguistics literature over the status of the preverbal subject. Some analyses treat this subject as a noun in Spec-TP, while others treat it as a noun in Spec-CP (left-dislocated). I will adopt the proposal that the noun preceding the NPI is left-dislocated into the CP layer and does not occupy the specifier position of the NPI. In support of this analysis, consider the following examples, where the adverb mumkin ‘probably’ can intervene between the noun ʔaħmad and the NPI: (37) a.
ʔaħmad mumkin ʕumr-o maa zaar Ahmad probably ever-him neg visited.3msg ‘Probably, Ahmad has never visited Petra.’
b.
ʔaħmad yemkin ʕəmmr-o maa zaar Ahmad probably ever-him neg visited.3msg ‘Probably, Ahmad has never visited Petra.’
el-batra (JA) def-P etra
el-batra def-Petra
(MA)
There is cross-linguistic evidence that this type of adverb can never be under the scope of negation (Cinque, 1999, p. 124). Consider the following example from JA: (38)
ʔaħmad Ahmad
ʕumr-o (*mumkin) maa (*mumkin) zaar el-batra ever-him probably neg probably visited.3msg def-Petra
Cinque (1999) analyzes the Romance equivalent of mumkin as occupying a projection (ModPepistemic) above NegP; he argues that this account provides the best explanation for the failure of negative markers preceding this adverb to scope over it. Applying this proposal to Arabic, consider that if mumkin occupies a position above NegP and if the NPI is in Spec-NegP, it follows that the subject DP ʔaħmad preceding the NPI must be in a projection higher than the one hosting the NPI (perhaps TopP, à la Rizzi, 1997). Another argument in support of the left-dislocation analysis of the pre-NPI noun phrase comes from examples like (39): (39) a.
l-walad ʕumr-o maa ħabbat-o def-boy ever-him neg love-him ‘The boy, never has any girl loved him.’
bint girl
(JA)
b.
l-waləd ʕəmmər maa bɣat-o def-boy ever-him neg love-him ‘The boy, never has any girl loved him.’
ši any
bənt girl
(MA)
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 187 ]
81
Here, we see that the noun coindexed with the complement of the NPI is, in fact, the left-dislocated object of the sentence. Therefore, it makes sense to treat the position of the noun before the NPI as a Spec-CP position rather than a Spec-NPI or a Spec-TP position.
5.1.4. Head-l ike PSIs as XPs: Licensing by negative constituents
In this section, I will demonstrate how the analysis developed here can explain the disparate behavior of the NCI baʕd and the NPI ʕumr with respect to licensing by the negative indefinites maħadaaš and wala-ħada. Recall that ʕumr cannot precede the negative compound maħadaaš or the NCI wala-ħada, while baʕd can: (40) a.
b.
(41) a.
b.
(*ʕumr-o) maħadaaš (ʕumr-o) zaar (*ever-him) no one (ever-him) visited.3msg ‘No one has ever visited Petra.’ (*ʕumr-o) wala-ħada (ʕumr-o) zaar (*ever-him) neg-one ever-him visited.3msg ‘No one has ever visited Petra.’ (baʕd-o) maħadaaš (baʕd-o) (yet-him) no one (yet-him) ‘No one has visited Petra yet.’ (baʕd-o) wala-ħada (baʕd-o) (yet-him) neg-one (yet-him) ‘No one has visited Petra yet.’
el-batra (JA) def-Petra
el-batra def-Petra
zaar visited.3msg
(JA)
el-batra (JA) def-Petra
zaar el-batra visited.3msg def-Petra
(JA)
The fact that ʕumr and baʕd behave differently with respect to the maħadaaš and wala-ħada suggests that the contrast has to do with the accessibility of NPIs versus NCIs to licensing by negation. Two questions arise. First, how is ʕumr licensed when following a negative constituent? Second, why is it that baʕd but not ʕumr can be licensed when preceding the NI? The answer to the first question lies in the adverbial properties of the NPI ʕumr. Recall that I have proposed that when preverbal, this NPI is always merged above TP. In section 5.1.3, I argued that the preverbal NPI is base generated in the specifier position of the NegP that dominates TP. In examples (40)a–b, the negative constituents maħadaaš and wala-ħada occupy a specifier
[ 188 ] A Unified Theory of Polarity Sensitivity
9 8 1
position in the CP layer (a position I will motivate later in this section). Thus, when the NPI is merged in the specifier position of a functional projection (FP) above TP, it can receive licensing under c-command by the negative constituents: (42) [CP maħadaaš/ wala-ħada [FP ʕumr-o [TP . . . [VP . . .
To answer the second question, we must consider the properties of ʕumr and baʕd as adverbs. We have seen that these two adverbial PSIs differ in their position in the structure and their ability to move. The NPI ʕumr is typically merged in Spec-FP or Spec-NegP above TP, and occasionally in Spec-VP. On the other hand, the NCI baʕd is always merged in Spec-VP and optionally moves to the CP layer.16 With this backstory in mind, let us see how the contrast in (40)/(41) follows from the properties of ʕumr and baʕd as adverbs that differ in their locus, movement, and licensing (overt or covert). To begin, we need to ascertain where the negative compound maħadaaš and the NCI wala-ħada are located in the structure. First of all, note that the negative constituent wala-NP is a non-strict NCI. When postverbal, it must co-occur with negation and has an NC (negative concord) reading. When preverbal, on the other hand, it cannot co-occur with negation under the NC reading. (43) a.
b.
ʕali *(ma-)ɣa:b-iš Ali (neg-)missed.3msg-neg ‘Ali did not miss any class.’
wala-sʕaff NCI-class
wala-sʕaff ʕali (*ma-)ɣa:b-(*iš) NCI-class Ali (neg-)missed.3msg(-neg) ‘Ali did not miss any class.’
(JA)
(JA)
16. The adverb baʕd is related to the Standard Arabic (SA) baʕd ‘yet,’ which can only appear sentence-finally, as in (i) in this note. Therefore, it makes sense to propose that baʕd is a VP adverb. The difference is that, in JA, it has acquired the ability to host clitics and to occupy a preverbal position: (i)
(*baʕd) lam ʔunhi l-kitabata (yet) neg.past finish.1sg DET-writing ‘I have not finished writing yet.’
(baʕd) (yet)
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 189 ]
0 9 1
The NCI wala-ħada can be focus fronted in JA. Consider the following examples where the NCI, which is the thematic object, precedes the preverbal subject and displays sensitivity to island constraints:17,18 (44) *wala-sʕaff ʕali nižiħ laʔinn-oh maa NCI-class Ali passed.3msg because-him neg ‘Ali passed because he did not miss any class.’
ɣa:b19 (JA) missed
Even when the NCI wala-ħada is a subject, it precedes the NPI ʔumr, suggesting that it is a topic in the CP layer rather than a subject in Spec-TP:20 (45) wala-ħada ʕumr-o zaar NCI-person ever-him visited.3msg ‘No one has ever visited Petra.’
el-batra def-Petra
(JA)
Therefore, I conclude that the preverbal NCI wala-NP in (43)b and (45) is located in the CP layer. Assuming a layered CP structure à la Rizzi (1997), I suggest that the wala-NP in (43)b is located in a focus projection, FocP, as in (46)a, while the wala-NP in (45) is located in a TopP projection, as in (46)b.21
17. Focus-fronted constituents in JA display sensitivity to adjunct islands: (i)
(ii)
xams ʔasʔɪleh ʕali ħall five questions Ali answered ‘Ali answered five questions.’ *xams ʔasʔɪleh ʕali nižħ laʔinn-oh five questions Ali passed.3MSG because-him
ħall answered
18. See Alsarayreh (2012) for similar facts. 19. The sentence is grammatical if we construe wala-sʕaff with the verb nižiħ, but ungrammatical if we construe it with ɣa:b. The latter is the intended reading as per the translation. I thank Enam Al-Wer for pointing out the grammaticality of the unintended reading. 20. There are many arguments that preverbal subjects are in Spec-CP rather than in Spec-TP (Aoun et al., 2010). Since I have argued that the NPI ʔumr is a Spec-TP adverb, it is reasonable to consider the NP preceding it to be in Spec-CP. 21. Below is a simple schematic representation for Rizzi’s (1997) layered CP: (i)
... ForceP ... TopicP ... FocusP ... TopicP. . . Fin IP
[ 190 ] A Unified Theory of Polarity Sensitivity
1 9
(46) a.
[FocP wala-NP
[TP . . . [VP . . .
b.
[TopP wala-NP
[TP . . . [VP . . .22
Now we are in a position to explain why ʕumr cannot precede the NCI wala- NP in (40)b, while the NCI baʕd in (41)b can.23 When it merges in a position above the NCI wala-NP, the NPI ʕumr effectively precedes the sentential negation; as such, ʕumr clearly cannot be c-commanded by that negation, nor can the two enter into a specifier–head relationship. The NCI baʕd, by contrast, always enters the structure in a VP-internal position. While baʕd is in this position, the NCI wala-NP can license it under c-command, a relationship that is maintained even after baʕd moves to a position preceding the NCI wala-NP in the CP layer. (47)
a. b.
The inability of ʕumr to precede the negative compound maħadaaš in (40) a receives a similar explanation. The negative constituent maħadaaš ‘no one’ and the negative constituent maʕumriš ‘never’ are lexical compounds that do not branch in syntax.24 They are inherently negative and can license NPIs that
22. In analyzing similar cases of wala-phrases in Levantine Arabic (LA), Hoyt (2010) assumes that wala-phrases are topics in the sense that they introduce a topical presupposition. Specifically, the wala- topic presupposes a set of referents that lack the property expressed by the proposition of the sentence. This explanation is consistent with the proposed structure in (46b) in this chapter. 23. Whether the preverbal NCI wala-NP is inherently negative (on a par with EA wala-NP, discussed in Ouali & Soltan, 2014) or licensed by a covert negative operator (following Zeijlstra’s, 2004, analysis of similar NCIs in Romance) is irrelevant here. In Alqassas (2015), I argue that wala-NP is not inherently negative. One piece of empirical evidence is the fact that we get a concordant reading when the preverbal NCI wala- NP co-occurs with a negative compound like maʕumriš ’never’: (i)
maʕumriš wala-ħada zaar el-batra never NCI-person visited.3msg DEF-Petra ‘No one has ever visited Petra.’
24. It is not possible to derive these compounds in syntax. To do so, we would have to move the NPIs (ʕumur and ħada) from a specifier position to the head position of the negative markers ma and -š and then move the entire complex [Neg-NPI-Neg] to a Spec-XP position. As Aoun et al. (2010) point out, this pattern of movement is theoretically problematic because it violates structure preservation (or rules of Merge in recent minimalist terms). Accordingly, I treat these elements as non-branching lexical compounds (see Hoyt, 2007, for a similar view).
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 191 ]
2 9 1
follow them. The negative compound maħadaaš can license the NPI ʕumr, and the negative compound maʕumriš can license the NPI ħada. (48) a.
b.
maħadaaš ʕumr-o zaar No one ever-him visited.3msg ‘No one has ever visited Petra.’ maʕumriš ħada zaar never one visited.3msg ‘No one has ever visited Petra.’
el-batra def-Petra
el-batra def-Petra
(JA)
(JA)
As with the NCI wala-ħada, the fact that the negative compound maħadaaš precedes the NPI ʕumr suggests that it is in the CP layer in a TopP projection. Thus, the observation that the NPI ʕumr in (40)a cannot precede the negative compound maħadaaš while the NCI baʕd in (41)a can is once again expected. (49)
Again, the NPI ʕumr in (49)a—the representation for (40)a—is not c- commanded by the negative compound, nor is it in a specifier–head relation with it. Since the NPI cannot be licensed under either of the two acceptable licensing configurations, ungrammaticality results. However, in (49)b, the representation for (41)a, the NCI baʕd originates in a VP-internal position where it is licensed via c-command by the negative compound before undergoing subsequent topicalization to the CP layer. The NPI ʕəmmər ‘ever’ in MA can either precede or follow the NCI ħtta waħəd ‘nobody.’ When it precedes the NCI, it takes the NCI as a pronominal complement. Both the NPI and its complement NCI are in Spec-NegP. In section 5.2 I discuss the details of how licensing takes place in MA and contrast that with JA. Crucially, when the NCI precedes the NPI, the NCI is not within the NPI phrase and the negative marker maa must follow the NCI. The NCI is licensed in Spec-NegP of the Neg head occupied by maa (for details see chapter 4). (50) a.
(maa)ʕəmmər ši waħəd maa (*ever-him) any one neg ‘No one has ever visited Petra.’
[ 192 ] A Unified Theory of Polarity Sensitivity
zaar el-batra visited.3msg def-Petra
(MA)
3 9 1
b.
(ħtta waħəd) maa (NCI-person) neg maa zaar neg visited.3msg ‘No one has ever visited Petra.’
ʕəmmər NPI-ever el-batra def-Petra
(ħtta waħəd) (NCI-person)
(MA)
So far, I have taken the position that the NCI baʕd is licensed overtly in its VP-internal position before it moves to the preverbal position. However, it is also possible to argue that baʕd is licensed covertly, at LF, under reconstruction to its base (VP-internal) position. Note that baʕd patterns with the other adverbial NCIs in JA in that it can be focus fronted, a property demonstrated by Alsarayreh (2012) for adverbial NCIs such as bilmarrah.25 The examples in (51) show that bilmarrah originates in a postverbal position, where is must be licensed by negation. In (51)a, bilmarrah occupies a postverbal position overtly, and thus is straightforwardly licensed under c-command by negation. In (51)b, however, the subject intervenes between negation and the NCI, making licensing by negation via either c-command or a specifier–head relationship impossible. Alsarayreh proposes that such NCIs are licensed at LF through reconstruction to their base position inside the VP:
All reconstruction implies movement (assuming the copy theory of movement, where a moved constituent leaves a covert copy in the position from which it moves: see Lebeaux, 1990; Bianchi, 1995; Sauerland, 2004; Nunes, 2004; among others). Following this movement, the lower copy gets interpreted at LF. This being the case, the inability of preverbal ʕumr in (40) to be licensed provides strong evidence that this NPI never moved from a position c-commanded by negation in the first place. Therefore, reconstruction is not possible and licensing fails, leading to ungrammaticality. In other words, the preverbal version of this NPI must be directly merged in Spec-FP or Spec-NegP
25. Alsarayreh, however, follows Benmamoun’s analysis of the MA equivalent to baʕd and baqi and analyzes baʕd as a head that projects a clausal projection above NegP and that is licensed via the head–complement relation with negation.
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 193 ]
4 9 1
above TP. In the ungrammatical examples in (40), it is simply merged too high to enter into even a specifier–head relation with negation, as explained previously. (52)
Further evidence is also available to support the claim that preverbal ʕumr is not left-dislocated. The clitic inside the NPI phrase ʕumr-o can bind pronouns inside syntactic islands, such as the complex DP ktaabt-oh ‘writing-his’ in (51)a, or the prepositional phrase minn-oh in (53)b: (53) a.? ʕumr-o maa ħabbat li-mʕalmeh ever-him neg liked.3fsg def-teacher.F ‘The teacher has never liked his writing’ b.
c.
ktaabt-oh writing-his
ʕumr-o maa šakat minn-oh mʕalmeh ever-him neg complained.3fsg about-him teacher ‘No teacher ever complained about him’
(JA)
(JA)
(maa) ʕəmmər maa tšəkkaa mənn-oh ši moʕallem (MA) (neg) ever neg complained.3Msg about-him any teacher ‘No teacher ever complained about him’
This finding suggests that the clitic inside the NPI cannot have been left- dislocated from a postverbal position bound by the pronoun inside the island. Finally, the contrasting loci of single negation (above TP) and bipartite negation (below TP) explain why bipartite negation cannot follow ʕumr in (54). See Alqassas (2015) for arguments supporting this proposal. (54) a.
baʕd-o ma-saafar-iš yet-him neg-traveled.3msg-neg ‘He hasn’t traveled yet.’
(JA)
b.
ma-saafar-iš neg-traveled.3msg-neg ‘He hasn’t traveled yet.’
(JA)
c.
*ʕumr-o ma-saafar-iš ever-him neg-traveled.3msg-neg ‘He has never traveled.’
baʕd-o yet-him
[ 194 ] A Unified Theory of Polarity Sensitivity
(JA)
5 9 1
ʕumr-o ever-him
d.
ma-saafar-iš neg-traveled.3msg-neg ‘He has never traveled.’
(JA)
e.
ʕumr-i maa-saafirt(*-iš) Masr NPI-ever-me neg-travel.1sg.pfv-neg Egypt ‘I have never traveled to Egypt.’ Soltan (2012, pp. 241)
f.
maa-saafirt-*(iš) Masr ʕumr-ii (EA) neg-travel.1sg.pfv-neg Egypt ever-me ‘I have never traveled to Egypt.’ Soltan (2012, pp. 242)
(EA)
The explanation for this distribution follows from the proposal that ʕumr is a TP adverb and baʕd a VP adverb that moves to the CP layer. When these PSIs are in postverbal position (i.e., in Spec-VP), they are automatically able to be licensed by negation, regardless of whether the locus of that negation is above or below TP. In preverbal position, however, licensing of an PSI is prevented just in the case that the PSI is merged above TP (in Spec-FP) and negation is below TP; this is precisely what occurs when ʕumr precedes the bipartite negation in (54)c. By contrast, an NCI like baʕd, which is always merged in Spec-VP, can be licensed by the low bipartite negation—either overtly, before it moves to a preverbal position, or covertly, via reconstruction at LF. (55) *[FP ʕumr [TP T [NegP Neg ma. . . š [vP . . . ]]]] (56) [FP baʕd [TP T [NegP Neg ma . . . š [vP. . . ]]]]
In both cases, the analyses developed here are possible because we are treating these PSIs as adverbs that occupy a specifier position rather than as heads that project their own clausal structure.
5.1.5. Conclusion
In this chapter, I have analyzed the status of Arabic head-like PSIs. I have argued that although these items famously exhibit head-like properties, such as hosting clitics and carrying agreement inflections (Benmamoun, 2006), they are better treated as AdvPs that occupy the specifier position of a functional phrase, rather than heads that project their own clausal maximal projections. These items take a (pro)nominal complement that can corefer with either the subject or the object of the sentence and can be as big as a relative clause. The
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 195 ]
6 9 1
analysis supports the proposal that both the specifier–head relationship and c-command are valid licensing configurations for PSIs in Arabic (Benmamoun, 1997), but it argues against the viability of the head-complement configuration as a licensing strategy for these PSIs. The NCI wala-ħada ‘anyone, no one’ and the NI maħadaaš ‘no one’ were also discussed from the perspective of their interaction with ʕumr and baʕd. The NPI ʕumr cannot precede the negative constituents wala-ħada and maħadaaš, while the NCI baʕd can. The analysis developed here derives this contrast straightforwardly from the location of these adverbs in the structure and their respective ability to move. The NCI baʕd is always merged VP-internally and receives licensing whenever negation c-commands it in that base-generated position; baʕd itself can move into the CP layer over the course of the derivation to appear preverbally/prenegatively, in which case licensing takes place either pre-movement or during reconstruction at LF. On the other hand, the preverbal NPI ʕumr is merged above TP in Spec-FP (or in Spec-NegP above TP when maa is the licensor), where it cannot be licensed by the negative constituents wala-ħada and maħadaaš.
5.2. HEAD-L IKE PSIS AND CASE LICENSING
This chapter discusses two main issues that arise from PSIs with head-like properties. The first issue is how these PSIs are licensed in syntax and whether their distributional contrasts can be explained via positing a special syntactic configuration or by reducing these contrasts to sectional properties and syntactic operations such as movement. Another issue that arises from these NPIs with head-like properties is their ability to host clitics with accusative and genitive case marking. This second issue raises interesting questions pertaining to case theory and dependent case licensing. For the licensing question, this chapter argues that these NPIs are adverb phrases that select a (pro)nominal or CP complement but project a clausal XP on top of the negation projection. This section expands on my own previous work, published in Alqassas (2016, 2018). Direct mapping from syntax to semantics reduces the variation in the behavior of PSIs to their syntactic properties, such as their locus in the structure relative to negation, and whether they undergo movement and reconstruction at LF. The third issue under investigation focuses on how these NPIs license an accusative or genitive complement. It is generally assumed that non- nominative case (e.g., accusative and genitive case) is assigned by verbs and prepositions under the head-complement configuration, whereas nominative case is assigned by a functional head T via the specifier-head configuration:26
26. All examples in this section are from JA unless otherwise indicated in the text.
[ 196 ] A Unified Theory of Polarity Sensitivity
7 9 1
(57) a.
b.
sˁadiiq-i zaar-ni friend-my visited.3MSG-me.ACC ‘My friend visited me.’ ʕumar katab l-i Omar wrote.3MSG to-me.GEN ‘Omar wrote a letter to me.’
(58) ʔana katabt I.NOM wrote.1sg ‘I wrote a letter.’
risaaleh letter
risaaleh letter
Overall, this chapter addresses how accusative case is licensed to nominal complements in two PSIs of JA, namely, the adverbial NPI ʕumr ‘ever’ and adverbial NCI, baʕd ‘yet.’ The proposed analysis challenges long-standing claims regarding how accusative case is licensed under agreement-based and dependent case accounts.
5.2.1. Distribution of PSIs with head-l ike properties
As shown throughout this book, PSIs interact with sentential negation in JA. Most relevant to this section is the adverbial NPI ʕumr ‘ever’ and adverbial NCI baʕd ‘yet.’ These require sentential negation in indicative sentences: (59) a.
ʕumr-ni *(maa) NPI.ever-me neg ‘I have never traveled.’
b.
baʕd-ni *(maa) NCI.yet-me neg ‘I haven’t traveled yet.’
saafart traveled.1sg
saafart traveled.1sg
Also, recall that the NCI baʕd can be used as a fragment answer, while the NPI ʕumr cannot: (60) a.
Question:
zaar visited.3MSG ‘Has he visited Petra?’
l-batra def-Petra
Answer:
*ʕumr-o NPI.ever-him ‘never’
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 197 ]
8 9 1
b.
Question:
zaar l-batra visited.3MSG def-Petra ‘Has he visited Petra?’
Answer:
baʕd-o NCI.yet-him ‘Not yet’
5.2.2. Temporal adverbs and their complements
Where these PSIs differ from (some of the) previous NPI and NCIs discussed throughout the book is their head-like properties. Head PSIs are unique in their ability to host clitics. PSIs in JA require this pronominal clitic, as seen in (61)a–b. It may seem that the clitic is the subject of the sentence and, as such, is required regardless of the PSI. However, Arabic is a pro-drop language. Minimal pair sentences without the PSI and clitic are grammatical in JA: (61) a.
ʕumr-*(ni) maa NPI.ever-me neg ‘I have never traveled.’
saafart traveled.1sg
b.
baʕd-*(ni) maa NCI.yet-me neg ‘I haven’t traveled yet.’
saafart traveled.1sg
c.
maa saafart neg traveled.1sg ‘I have not traveled.’
A nominal can also satisfy this requirement in the absence of a clitic: (62) a.
ʕumr *(Zeed) NPI.ever Zeed ‘Zeed has never traveled.’
maa neg
saafar traveled.3MSG
b.
baʕd *(Zeed) NCI.yet Zeed ‘Zeed has never traveled.’
maa neg
saafar traveled.3MSG
MA has the NPI ʕəmmər ‘ever’ and NCI baqi ‘yet,’ which can also host clitics and can carry subject agreement inflection. This latter property is typically associated with heads in Arabic, as in the following examples from Benmamoun (2006, pp. 144–145):
[ 198 ] A Unified Theory of Polarity Sensitivity
91
(63) a.
nadya baq-a Nadia yet-FS ‘Nadia hasn’t come yet.’
ma-žat neg-came.3fsg
(MA)
b.
lə-wlad baq-yin ma-žaw the-children yet-pl neg-came.3mp ‘The children haven’t come yet.’
(MA)
c.
nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
(MA)
d.
ʕəmmər nadya ma-mšat l-təmma never Nadia neg-went.3fsg to-there ‘Nadia never went there.’
(MA)
e.
ʕəmmər-ni ma-mšit l-təmma never-me neg-went.1sg to-there ‘I never went there.’
(MA)
ma-žat neg-came.3fsg
The (pro)nominal complement can be coreferential with the subject of the sentence, as in (61)–(62), or with a nominal complement as in the following sentence: (64) a.
ʕumr-o
maa šakat
minn-oh
mʕalmeh
NPI.ever-him neg complained.3FSG about-him ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149) b.
(JA)
teacher
baʕd-o maa šakat minn-oh mʕalmeh NCI.ever-him neg complained.3fsG about-him teacher ‘No teacher has complained about him yet.’
(JA)
The complement of these head PSIs can be CP as in the following: (65) a.
b.
ʕumr illi budrus maa NPI.ever who study.3MSG neg ‘He who studies never fails.’ baʕd illi zaar-na maa NCI.ever who visited.3msG-us neg ‘He who visited us has not yet left.’
bursub (JA) fail.3MSG (Alqassas, 2016, p. 142) rawwaħ left.3msG
(JA)
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 199 ]
02
5.2.3. Licensing non-n ominative case in Arabic
In Arabic, complements of verbs and complementizers are assigned accusative case, whereas complements of nouns and prepositions are assigned genitive case. (66) Complement of a verb (accusative) [VP zaar-ni] visited.3msG-I.ACC ‘He visited me.’ (67) Complement of a complementizer (accusative) baʕrif [CP in-ni [TP [VP zurt-o]] know.1sg comp- I.ACC visited-him ‘I know that I visited him.’ (68) Complement of a noun (genitive) ktaab-i bi-l-maktabeh book-my.GEN in-the-library ‘My book is in the library.’ (69) Complement of a preposition (genitive) Zeed ħaka maʕ-i Zeed spoke.3msG with-me.GEN ‘Zeed called me yesterday.’
nbaariħ yesterday
Surprisingly, the complement of a head PSI carries an accusative case: (70) Complement of a head PSI (accusative) a. ʕumr-ni maa NPI.ever-me.ACC neg ‘I have never traveled.’ b.
baʕd-ni maa NCI.yet-me.ACC neg ‘I haven’t traveled yet.’
saafart traveled.1sg
saafart traveled.1sg
Given the cross-linguistic pattern of case assignment rules that rely on the syntactic category of the assigner, temporal adverbs are expected to pattern
[ 200 ] A Unified Theory of Polarity Sensitivity
1 0 2
with prepositions or nouns in assigning case.27 However, such a case assignment rule incorrectly predicts that PSI complements will be assigned genitive case since the NCI adverb baʕd ‘yet’ is lexically derived from the preposition baʕd ‘after’ and the NPI adverb ʕumr ‘ever’ is lexically derived from the noun ʕumr ‘lifetime, age’: (71) a.
b.
door-ak baʕd-i turn-you.GEN after-me.GEN ‘Your turn is after me.’ ʕumr-i ʕišreen saneh age-my.GEN twenty year ‘My age is twenty years.’
The lexical homophones of these PSI adverbs, crucially, cannot have an accusative complement: (72) a.* door-ak turn-you.GEN b.* ʕumr-ni age-my.ACC
baʕd-ni after-me.ACC ʕišreen twenty
saneh year
Another issue with the case assignment facts for PSI complements is that verbs typically assign accusative case to nominals that act as their semantic arguments with a thematic role (hence the term inherent case). No such thematic role exists for nominal complements of head PSIs, which poses a contradiction given that the accusative case here is structural. It is worth mentioning that these PSIs might be impersonal verbs that take the (pro)nominal as an impersonal subject. This, in turn, would suggest that accusative case appears when nominative case is already assigned. In this context, nominative case would be assigned by the T head to the null subject pro
27. Under the distinctive features of lexical categories (Chomsky, 1975), adverbs that take complements and assign case have two options: they can either pattern with nouns or with prepositions, and they must assign genitive case, but head PSIs seem to pattern with verbs and assign accusative case. i.
N V A P
[+N, -V] [-N, +V] Head PSIs? [+N, +V] (e.g., cleverly, independently, in Emonds, 1976; Radford, 1988) [-N, -V] (e.g., after, before, inside, afterwards, in Jackendoff, 1973)
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 201 ]
2 0
in Spec-TP and the PSI licensing the impersonal subject via accusative case.28 In Arabic, however, these PSIs do not share key characteristics of impersonal verbs. Impersonal verbs in Arabic select a main verb in the subjunctive mood whose subject must be coindexed with the complement of the impersonal verb. The following examples from SA and JA illustrate these characteristics. Subjunctive case is marked with the suffix -a in SA, but JA marks it by a lack of the prefix b-, the marker of imperfective aspect: (73) a.
b.
yumkinu-ni ʔan can.3msG-me.ACC comp ‘It is possible for me to travel.’
ʔusaafir-a (SA) travel.1sg-sbjv
b-yimkin-ni ʔasaaafir asp-can.3msG-me.ACC travel.1sg.sbjv ‘It is possible for me to travel.’
(JA)
These PSIs, on the other hand, do not carry phi-feature morphology; they can have a complement that is not coindexed with the subject of the main verb, and the main verb is indicative. (74) a.
b.
ʕumr-o maa šakat minn-oh NPI.ever-him neg complained.3fsG about-him ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149) baʕd-o maa šakat minn-oh NCI.ever-him neg complained.3fsG about-him ‘No teacher has complained about him yet.’
mʕalmeh teacher
mʕalmeh teacher
As such, Arabic PSIs do not pattern with impersonal verbs.
5.2.4. Proposal for accusative case licensing on complements of PSIs
Given the complex case patterns described, I propose that the accusative case of PSI complements is assigned by the negative marker maa. Evidence from SA 28. Lucas (2009, pp. 207, 221) points out that baʕd-o and ʕumr-o are changing into impersonal verbs (pseudo-verbs) that take the logical subject of the sentence as an object pronoun. Here, though, we see that the NPI can also take a pronominal complement in the sentence as an object pronoun.
[ 202 ] A Unified Theory of Polarity Sensitivity
3 0 2
suggests that negative markers can assign accusative case on par with complementizers. This is clear from the negative marker laa, which can assign accusative case to the subject of a verbless sentence (see Al-Naadirii, 2009) as seen in example (75):29 (75) laa mudarris-iin neg teachers-mpl-ACC ‘No teachers are absent.’
ɣaaʔib-uun absent-mpl.NOM
I posit that the accusative case assigned by maa following a PSI phrase is similar to nominative case assignment. Nominative case is assigned by the head of a functional category T to the nominal occupying its specifier position. Likewise, the head of a functional category Neg assigns accusative case to a (pro)nominal in the specifier of NegP. Assuming that these PSIs are licensed under either c-command or the specifier-head configuration (Benmamoun, 1997, 2006; Alqassas, 2015, 2016), this produces the representation in (76) as the licensing configuration for these PSIs. These PSIs are licensed in the specifier of the negation phrase:30 (76)
NegP AdvP Adv NPI
Neg NP [ACC]
Neg
TP
maa
The question remains: how can a negative marker assign accusative case to the (pro)nominal complement of the head PSI? The answer may be found in the original proposals for how case is generally assigned. Under agreement-based accounts for case assignment, nominals that enter into an agreement relation with uninterpretable phi-features on T get nominative case as a marker of that phi-agreement dependency. Nominals that enter into an agreement with v get accusative case (Chomsky, 1995). If we extend this account to include nominals that enter into an agreement relation with the negative head, 29. Cited in Benmamoun et al. (2013). 30. Benmamoun analyzes the equivalent PSIs in MA as heads that project a clausal projection dominating NegP. Under his analysis, the NPI is licensed via the head– complement relation it has with negation. Such an analysis cannot be extended to the facts from JA where the PSI can take a relative clause—cf. (65) above for CP complement of the PSIs—since the NPI and negation are in no head–complement relationship or any possible licensing configuration; see Alqassas, 2016, for argument that support the representation in (76).
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 203 ]
4 0 2
the accusative case of the pronominal complement would be the marker of that agreement. To predict this, I suggest that the negative marker has person agreement with the pronominal complement of the PSI.31 This claim is supported by the fact that negative markers in Arabic display a behavior that requires the presence of the person feature. The imperative verb, for example, does not carry the inflectional morphology of person, but the second-person prefix is obligatory in negative imperatives: (77) a.
(*t-)saafir 2-travel. imp ‘travel!’
b.
laa neg.imp ‘Don’t travel.’
*(t-)saafir 2-travel
This suggests that negation enters some sort of phi-agreement with a nominal or the inflectional features of nominal morphology. If the nominal subject in (81)a or the null subject pro in (81)b gets its nominative case via agreement with the verb, then the pronominal complement of the head PSI gets its accusative case via agreement with the negative head.32 The licensing process is straightforward in the case of the PSI baʕd ‘yet.’ Recall from c hapter 4 that this PSI is merged in postverbal position where negation licenses it under c- command. As such, the analysis explains contrasts in the distribution of baʕd and ʕumr, where only the former can be left-dislocated. (78) a.
*ʕumr-o maħadaaš zaar ever-him no one visited.3msg ‘No one has ever visited Petra.’
el-batra def-Petra
b.
baʕd-o maħadaaš zaar yet-him no one visited.3msg ‘No one has visited Petra yet.’
el-batra def-Petra
31. Benmamoun (2000) argues that negation is specified for the D feature. This feature gets checked by phi-features of the verb or by the nominal subject of the sentence. 32. It is worth mentioning that adverbs in Arabic are related to accusative case in a manner rather unrelated to the issue at hand. Arabic has the accusative indefinite suffix -an, which can be suffixed to indefinite nouns to create adverbs such as lajl ‘night’→ lajl-an ‘at night.’ This form of adverb is restricted to formulaic expressions in the Arabic dialects, as in the expressions šukr-an ‘thanks,’ ahla-an wa sahl-an ‘welcome.’
[ 204 ] A Unified Theory of Polarity Sensitivity
5 0 2
We accounted for these contrasts by analyzing baʕd as an adverb that is merged postverbally and moved preverbally. It is then licensed via reconstruction at LF where it is c-commanded by negation. (79)
(80)
NegP Neg [uφ]
TP T
VP AdvP
maa
Adv NCI
VP
NP [iφ, ACC]
However, the case licensing process between negation and the pronominal complement of the NPI ʕumr is more complicated: (81) a.
b.
ʕumr-o maa [TP [T šakat minn-oh] mʕalmeh NPI.ever-him.ACC neg complained.3fsG about-him teacher.NOM ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149) ʕumr-ni maa NPI.ever-me.ACC neg ‘I have never traveled.’
[TP pro saafart traveled.1sg
(82)
The obvious problem with the agreement structure in (82) is that syntactic agreement between the negation head and the nominal complement of the NPI is not evident. The nominal complement is buried inside the NPI phrase. It is possible to see how negation licenses the NPI complement in cases where P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 205 ]
6 0 2
negation is higher than the NPI phrase. Such structures are, indeed, possible in both JA and MA: (83) a.
b.
maa ʕumr-ha katbat neg NPI-ever-her write.3fsg.pfv ‘She has never written a novel.’
riwaaye novel
ma ʕəmmər nadya ʒat never Nadia come.3fsg.pfv ‘Nadia never came.’ (Benmamoun, 2006, p. 145)
(JA)
(MA)
The question remains, though, as to how we explain the licensing of the nominal complement in cases where the negative marker appears lower than the NPI phrase. Here, I propose that there is a negative operator above the NPI phrase that licenses the pronominal complement in MA. This negative marker, which can precede the NPI in MA, is potentially the spell-out of the negative operator. An alternative analysis assumes that the whole compound ma-ʕəmmər ‘never’ in MA is an NI, is semantically negative, and can license accusative case on its nominal complement. To test the implications of this, we can look at cases where the combination of the compound ma-ʕəmmər ‘never’ and another polarity item does not give rise to a DN reading. There is, indeed, evidence that the compound ma-ʕəmmər ‘never’ can co-occur with the marker ma without yielding a DN reading. The reading in such contexts is an NC reading: (84)
NegP Neg [uφ]
TP TP
AdvP maa
Adv NPI
NP [iφ, ACC]
(85) ma neg ‘Nadia never came.’ *‘Nadia has never not come.’ (Benmamoun, 2006, p. 145)
ʕəmmər never
[ 206 ] A Unified Theory of Polarity Sensitivity
nadya ma-ʒat (MA) Nadia neg-come.3fsg.pfv [NC] [*DN]
7 0 2
(86)
NegP Neg [uφ]
AgrPolP AdvP
ma[iNEG]
Adv NPI
AgrPol NP [iφ, ACC]
AgrPol ma [uNEG]
TP
AGREE
The lower negative marker in this syntactic structure is an agreement marker, i.e., a negative concord marker. This lower negative marker, which is semantically vacuous, is then the realization of a syntactic agreement process. The lower ma is in a polarity agreement projection; call it AgrPolP. The higher ma, which carries [ineg], licenses this lower ma, which carries [uneg].33 The final question posed here regards how negation licenses the pronominal complement in cases where there is no c-command relation. I suggest that the lack of c-command is derived via movement of the NPI phrase from a position lower than negation to a position preceding it. Licensing takes place before this movement when negation still c-commands the NPI: (87) [NegP ʕumr-ha [Neg maa [TP [TP NPI-ever-her neg NPI-ever-her ‘She has never written a novel.’
katbat write.3fsg.pfv
riwaaye (JA) novel
The analysis proceeds as follows: The NPI ʕumr is merged into a position lower than NegP, at the edge of TP, where negation c-commands it and licenses the
33. An anonymous reviewer points out the relevance of semantic weakening to the Jespersen Cycle of negation, citing Chatzopoulou’s (2018) Greek study on polarity. It is widely believed that morpho-phonological weakening of a negative marker leads to the rise of a new negative marker, often from an indefinite noun like ʔiši ‘thing’ in JA (see Alqassas, 2019, for the Jespersen Cycle of negation in Arabic). However, there are studies that argue for semantic weakening as the source of the cycle. The semantic weakening of the higher negative marker maa in Moroccan would align with this view. This requires further research, but it is reasonable to posit that as a stage for the development of bipartite negation in MA. This preverbal negative marker has developed into an expletive marker that is only formally negative, as I showed in c hapters 4. This suggests that the preverbal negative marker in Moroccan is more advanced in the cycle. This weakening contrasts with JA, which also has bipartite negation, but no weakening for the negative marker maa. It is the postverbal negative marker -š that is more advanced in JA since it can occur by itself in some contexts like negative imperatives and negated labial-initial predicates (see Alqassas, 2019, chapter 6, for more on this).
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 207 ]
8 0 2
case of the pronominal complement. Then, the NPI and its pronominal complement undergo phrasal movement to Spec-NegP. This proposal weighs in on a number of broader theoretical questions. First, it challenges the agreement-based account proposed in Pesetsky and Torrego (2001) in which the nominal subject receives nominative case (structural case) through agreement with an unvalued tense feature. Internal arguments of a verb have inherent case, which is based on the thematic relationship between the nominal and the verb. This results in no unvalued case features. In other words, accusative case is analyzed as a reflection of this thematic relationship. The data of this chapter suggests that accusative case is assigned to a (pro)nominal complement that has no thematic relationship with its head. Furthermore, if this analysis is on the right track, the (pro)nominal complement carries accusative case despite having a structural, rather than thematic, relationship with the licensing head, i.e., the negative head. In challenging this conception of agreement, the analysis also calls into question the currently accepted agreement-based account to accusative case licensing. Generally, accusative case is assumed to be the morphology found on a nominal that does not enter into an agreement relation with T, as in dependent case theory, proposed by Marantz (1991). Under Marantz’s account, accusative case is the morphology found on the nominal when the non-thematic subject position is assigned nominative case. However, the agreement-based account has more merit once we consider genitive case assignment in Arabic. Recall that nouns and prepositions assign genitive case in Arabic. Crucially, the nominal complement does not enter into a phi- agreement relation with the preposition or the noun. (88)
NegP AdvP Adv
Neg NP [iφ, ACC]
Neg [uφ]
NPI
TP AdvP
Adv NPI
MOVE
[ 208 ] A Unified Theory of Polarity Sensitivity
NP [iφ, ACC]
TP pro[iφ, NOM] SUBJECT[iφ, NOM]
T… [uφ]
9 0 2
(89) Complement of a noun (genitive) ktaab-i bi-l-maktabeh book-my.GEN in-the-library ‘My book is in the library.’ (90) Complement of a preposition (genitive) Zeed ħaka maʕ-i Zeed spoke.3MSG with-me.GEN ‘Zeed called me yesterday.’
mbaariħ yesterday
As such, a revised version of the agreement-based account would define accusative case as the morphology found on nominal complements that enter into an agreement with v or neg, while nominative case is for nominal subjects that agree with the T head and genitive case is for nominal complements that do not enter into a phi-agreement relation with its head. One final note to be made here relates to a comment from an anonymous reviewer regarding whether the pronominal complement of head-like PSIs is associated with a broad subject. An example from SA reveals that the phi- features on the matrix verb coincide with an experiencer, and not the grammatical subject, in the embedded clause: (91) kid-tu ʔan yaɣliba-nii approximate-1sg comp overcome-me ‘I was almost overcome by sleep.’
l-nuʕaas-u the-sleep-NOM
(SA)
It should be noted though that the phi-features of the matrix verb can also be coindexed with the subject of the embedded verb in SA: (92) kaada ʔan yaɣliba-nii l-nuʕaas-u (SA) approximate.3msg comp overcome-me the-sleep-NOM ‘Sleep was about to overcome me.’
The suggestion seems to be that the pronominal complement of the head-like PSI might be associated with the phi-features of a broad subject. The relevant question is whether the pronominal complement of the PSI is associated with a broad subject, thus explaining the binding puzzle in cases where the pronominal complement coincides with the verb’s object and not its grammatical subject: (93) ʕumr-o maa [TP [T šakat minn-oh] mʕalmeh (JA) NPI.ever-him.ACC neg complained.3FSG about-him teacher.NOM ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149)
Doron and Heycock (1999) propose that preverbal NPs in Arabic, and other languages, are broad subjects that are directly merged in Spec-TP, i.e., it does
P S I s w i t h H e a d - L i k e P r op e r t i e s
[ 209 ]
0 1 2
not move from a VP internal position. The broad subject also binds a resumptive pronoun. The test for subjecthood in Doron and Heycock focuses on its ability to appear in ECM (exceptional case marking) contexts and following the verbal copula kaan (examples from Aoun et al., 2010, p. 65): (94) a.
b.
ðˁanan-tu l-bayt-a ʔalwaan-u-hu zaahiyat-un (SA) thought.1sG the-house-acc colors-nom-its bright-nom ‘I believe the house to be of bright colors.’ kaana l-bayt-u ʔalwaan-u-hu was.3msg the-house-nom colours-nom-its ‘The house was of bright colors.’
zaahiyat-un bright-nom
The question then becomes whether the pronominal complement of the head- like PSI is a resumptive pronoun that a broad subject binds. To answer this, we must look at cases where the NP preceding the head-like PSI is in an ECM position. If the pronominal complement of the PSI obligatorily matches the NP preceding it in an ECM position, then the pronominal complement of the PSI is a resumptive pronoun of that NP. The following examples, however, show that the pronominal complement can be coindexed with either the preceding NP or the grammatical subject of the embedded verb: (95) a.
b.
fakkart-o ʕumr-o maa (JA) thought.1sg-him NPI.ever-him.ACC neg šakat minn-oh mʕalmeh complained.3FSG about-him teacher.NOM ‘I thought no teacher has ever complained about him.’ fakkart-o ʕumr-ha maa (JA) thought.1sg-him NPI.ever-her.ACC neg šakat minn-oh mʕalmeh complained.3fsG about-him teacher.NOM ‘I thought no teacher has ever complained about him.’
Therefore, the pronominal complement cannot be a resumptive pronoun of a broad subject.
[ 210 ] A Unified Theory of Polarity Sensitivity
12
CHAPTER 6
Summary and Conclusions
T
hroughout this monograph, I have provided an in-depth discussion of PSIs in Arabic. A quick overview of these PSIs can be found in table 6.1 and reveals much of the microvariation we discussed for Arabic. Geographically, this variation comparatively grows at the phonetic, morphological, and lexical levels as we move from left to right, from the west with the Moroccan dialect, into Egyptian and Jordanian as we move along the Mediterranean, into the Arabian Peninsula with Qatari Arabic (QA), and into the formal written Arabic. Table 6.1 shows the complex and rich landscape of polarity in Arabic. This rich landscape represents the complex visibles that the unified approach and theory advocated in this book sought to explain. The simple invisibles that paint this colorful landscape of polarity lie in the following inventory of neg- features (table 6.2). The unified approach advanced in the book relies on the structural configurations that establish syntactic decencies and directly map into the semantic dependencies between the operators (licensors) and their variables (PSIs). The unified approach employed here also relies on the interaction between NPIs (negative polarity items) and PSIs in the same clause. Treating this interaction is fundamental for scrutinizing the licensing conditions for the various types of NPIs and negative indefinites (NIs). In this book, I have provided syntactic analyses of the distributional contrasts, licensing conditions, and their formal feature composition for many of these PSIs. These analyses and the microvariation that they represent have theoretically significant implications. In this chapter, I summarize the central points from each previous chapter and elaborate on the individual analyses
A Unified Theory of Polarity Sensitivity. Ahmad Alqassas, Oxford University Press. © Oxford University Press (2021). DOI: 10.1093/oso/9780197554883.003.0006.
2 1
Table 6.1. VARIATION OF PSIS IN ARABIC L ANGUAGES Negative Indefinites
person
English
MA
EA
JA
QA
SA
nobody
ħəttawaħəd
walawaaħid
walaħada
maħħad
laaʔaħada
maħaddiš
maħadaaš
thing
nothing
ħəttaħaʒa
walaħaaga
wala(ʔi)ši
mašay
laašayʔa
time
never
maʕəmmər
walaʕumr
walaʕumr
ʔabad
ʔabadan
maʕumriš
maʕumriš
ʔabadan
ʔabadan
wala
Wala
ma
laa
det
ħətta
no
Negative Polarity SA
English
MA
EA
JA
QA
person
anybody
(ši) waħəd
(ʔay)ħadd
(ʔay)ħada
(ʔay) ʔaħad
(ʔay)ʔaħad
thing
anything
(ʔay)ħaaga
(ʔay)ʔiši
(ʔay) šay
(ʔay)šayʔ
ʕəmmər
ʕumr
ʕumr
gad
qatˁ
ši
ʔay
ʔay
ʔay
ʔay
ʔay waħəd (ši) ħaʒa ʔay ħaʒa time
ever
det
any
ʔay*
* This item is mostly used as a positive polarity item in MA. Oulai and Soltan (2014) use this item for MA to illustrate microcomparisons between EA and MA and note that this item is usually used as a PPI in MA.
Table 6.2. INVENTORY OF [neg]- FEATURES Language
Interpretable NEG
Uninterpretable NEG
JA, EA
Op¬ [ineg∅]
NCIs: [uneg]
maa: [ineg]
laa-wala: [uneg]
ma- . . . -š: [ineg] maħadaaš/ maħaddiš: [ineg] maʕumriš: [ineg] MA
Op¬ [ineg∅]
NCIs: [uneg∅]
ma- . . . -š: [ineg]
maa: [uneg] laa-wala: [uneg∅]
QA
maa: [ineg]
NCIs: [uneg]
SA
maa: [ineg]
NCIs: [uneg]
laa/lam/lan: [ineg]
laa-wala: [uneg]
laa-wala: [uneg]
3 1 2
of these PSIs and their contributions to the critical issues in syntax and linguistic theory. Overall, the syntactic licensing of PSIs in Arabic bears on key theoretical issues in the cross-linguistic studies of polarity sensitivity. Such issues include the syntactic licensing configurations for these items, the feature of structure/specifications of these items, and the availability of syntactic agreement in the context of negation. Lastly, the chapter will present residual issues on the topic that can be taken up in future research.
6.1. THE SYNTAX OF NPIS
By comparing the distribution of NPIs in the various Arabic varieties, we were able to explain their distributional contrasts through a unified analysis with a small set of basic syntactic operations such as Merge and Move. The book espouses a unified approach to analyzing polarity sensitivity. The variation lies in the loci of the NPIs, the multi-locus nature of negation, and the PSI’s ability to undergo XP (X phrase) movement or focus-fronting. The explanatory power of this unified analysis is clear in its ability to capture the true nature of perceived mutual exclusivity between NPIs and the enclitic marker of bipartite negation. We showed that these cases are an epiphenomenon of the availability of multiple loci for negation in that the mutual exclusivity is between the NPI and bipartite negation, not the enclitic marker of negation. Particularly, bipartite negation is too low in syntactic structure, below TP (tense phrase), and cannot license NPIs merged above TP. The unified analysis thus reduces all the distributional restrictions of NPIs to whether they are in one of the two licensing configurations: c-command by negation and Spec-NegP. We argued that the higher negative could license adverbial NPIs that are merged above TP while the lower negative cannot do so because it is too low in the structure (below TP) for the NPI to be licensed in the specifier of NegP (see Alqassas, 2015, for JA). (1)
a.
ʕumr-ha maa katbat NPI-ever-her neg write.3fsg.pfv ‘She has never written a novel.’
b.
* ʕumr-ha NPI-ever-her
c.
maa gad fi-sˁsˁaf ʔaħmad kallam-a (QA) neg ever in class Ahmad talk.3msg.pfv-him ‘Ahmad has never talked to him in class’
ma-katbat -iš neg-write.3fsg.pfv-neg
riwaaye novel
riwaaye novel
(JA)
(JA)
S u m m a r y a n d C o n c l u s i o n s
[ 213 ]
4 1 2
(2)
(3)
fi-sˁsˁaf in class
ʔaħmad Ahmad
d.
* gad ever
maa neg
kallam-a talk.3msg.pfv-him
a.
nadya ʕəmmər-ha Nadia never-her ‘Nadia never came.’
b.
ma-ʤa(*-š) neg-come.3msg.pfv(*-neg) ‘No one came’
a.
ʕumr-i maa-saafirt(*-iš) Masr NPI-ever-me neg-travel.1sg.pfv-neg Egypt ‘I have never traveled to Egypt.’ Soltan (2012, p. 241)
(EA)
b.
maa-saafirt-iš Masr ʕumr-ii neg-travel.1sg.pfv-neg Egypt ever-me ‘I have never traveled to Egypt.’ Soltan (2012, p. 242)
(EA)
c.
ma-zaar-iš neg-visit.3msg.pfv-neg ‘He has never visited Petra.’
ma-ʒat(*-š) neg-come.3fsg.pfv(*-neg)
ħətta waħəd even one
ʕumr-o NPI-ever-his
el-batra def-Petra
(QA)
(MA)
(MA)
(JA)
The validity of this type of analysis is supported by the fact that the temporal NPI ʕumr in postverbal position can in fact co-occur with bipartite negation in both EA (Egyptian Arabic) and JA (Jordanian Arabic). Given that this NPI can occur in postverbal position and bipartite negation c-commands this postverbal NPI, it follows that bipartite negation can co-occur with the temporal NPI ʕumr. The ability of the higher negative to license NPIs is also attested with adverbs that express categorical negation, including gad ‘never’ from QA. Here, we also find that the adverb can intervene between negation and the verb. (4)
a.
maa gad fi-sˁsˁaf neg ever in class ‘I’ve never talked to him in class’
b.
* maa neg
fi-sˁsˁaf in class
kallamt-a talk.1sg.pfv-him
kallamt-a talk.1sg.pfv-him
[ 214 ] A Unified Theory of Polarity Sensitivity
(QA)
(QA)
5 1 2
c.
mub fi-sˁsˁaf kallamt-a neg in class talk.1sg.pfv-him ‘I did not talk to him in class’
(QA)
The advantage of this analysis is its predictive power regarding the mutual exclusivity between the enclitic negative marker (NM) and particular PSIs. The analysis can capture this mutual exclusivity without appealing to special grammatical operations that block the enclitic NM either via theoretically challenged post- syntactic (morphological) deletion (cf. Soltan, 2012) or through empirically challenged syntactic competition with NPIs (Ouhalla, 2002). The variable behavior of PSIs relies on the basic syntactic operations, Merge and Move, and two syntactic configurations: Spec-Head and c-command. By defining Spec-NegP and c-command as the only licensing configurations, it follows that there is direct mapping between the syntactic structure of NPIs and their semantic structure. As such, the analysis advanced in this book is superior to earlier analyses in that syntactic licensing configurations always feed the semantic component. Under this analysis, the NPI is always under the scope of its licensor and can get its negated existential quantifier interpretation, whether we consider NPIs as existential quantifiers or non-quantificational elements that undergo existential closure, as in Heim (1982).
6.2. THE SYNTAX OF NEGATIVE INDEFINITES
The micro-comparisons of NCIs (negative concord items) shows interesting and significant contrasts in the varieties of Arabic, particularly between NCIs in JA and EA, on the one hand, and MA, on the other. This comparative study shows that the NCI wala in JA and EA is not inherently negative and is always licensed by an overt or covert negative constituent (NM, negative operator, or NI) (see table 6.3). The negative operator is a last resort mechanism that licenses the NCI wala in the absence of an overt negative constituent. The MA NCI ħətta, however, is only licensed by a covert negative operator, thus giving further support to Penka (2011)’s distinction between two different types of formal negative features: overt uninterpretable negative feature [uneg] and covert uninterpretable negative feature [uneg∅]. Penka posited the formal feature [uneg∅] for NCIs that can only be licensed covertly. In turn, only a covert negative operator can license NCIs with this null uninterpretable feature (see table 6.3). The results of the microcomparative approach unearthed new types of languages that behave differently with respect to the feature structure of NMs, NIs, and multiple agree. This more complex picture shows that there is more
S u m m a r y a n d C o n c l u s i o n s
[ 215 ]
6 1 2
Table 6.3. VARIATION IN THE FEATURES ON NMS, NIS, AND MULTIPLE AGREE Type
Language
Feature of NM
Feature of NIs
Multiple Agree
NC
JA,
maa: [ineg]
walaħada: [uneg]
not available
EA
ma- . . . -š: [ineg]
walawaaħid: [uneg] maħadaaš: [ineg] maħaddiš: [ineg] maʕumriš: [ineg]
NC
Italian, Spanish
[ineg]
[uneg]
obligatory
NC
Greek, Polish,
[uneg]
[uneg∅]
obligatory
Russian, Ukrainian NC
Romanian
[uneg]
[uneg]
optional
NC
French,
ne: [uneg]
[uneg∅]
optional
MA
pas: [ineg]
[uneg∅]
not available
maa: [uneg] ma- . . . -š: [ineg] DN
German, Dutch,
[ineg]
Scandinavian Note: This table is adapted from Penka’s (2011) study of NIs, which does not include Arabic dialects. I added the variation of negative features on Arabic NMs, NIs, and availability of Multiple Agree.
variation than previously found, thanks to microvariation in Arabic. This study unearthed properties for certain NIs that are in direct contradiction to Penka’s (2011) claim that there are no true NIs in natural language. The NIs maħadaaš ‘nobody’ and maʕumriš ‘never’ are inherently negative while walaħada/ walawaaħid ‘nobody’ are not. The NI maħadaaš ‘nobody’ does not display split scope readings, licenses the determiner NCI wala, and always causes a double negation reading when co-occurring with another negative constituent. By contrast, the NI walaħada/ walawaaħid ‘nobody’ displays split scope readings, cannot license another determiner NCI, and always fails to contribute a negative interpretation when under the scope of a negative constituent. The evaluation of data in c hapter 4 unearthed a correlation between NC (negative concord) languages with true NIs and DN (double negation) languages. Multiple Agree is not available in DN languages, or NC languages with true NIs. The correlation between the availability of Multiple Agree and DN languages is no longer exclusive, thanks to dialects like JA (Jordanian Arabic) and EA (Egyptian Arabic), which are largely NC languages but have NIs that are similar to those in DN languages like English. Their similarity is in the availability of DN readings when certain NIs co-occur with negation. This then leads to the conclusion that languages with a mixed NC and DN system also lack Multiple Agree. As such, languages that are partially or fully DN lack Multiple Agree, or, put differently, languages that have NIs that can occur without NMs carrying [ineg] do not have Multiple Agree (see table 6.3). [ 216 ] A Unified Theory of Polarity Sensitivity
7 1 2
In addition, this book showed that the inventory of negative features exhibits more variation than previously assumed. It was shown that NIs (NIs) can, indeed, be inherently negative and their lexical meaning corresponds to negated existential quantification. NIs in Arabic specifically display complex NC and DN readings even in preverbal position. Some NIs display split scope readings, whereby the interpretation of negation and the indefinite noun is split by a modal while other NIs do not display split scope. With this in mind, I argued for a revised version of NC as syntactic agreement by positing a covert negative operator as last resort for NIs with split scope effects and the availability of negated existential quantification in natural language. NIs in JA include the NCI wala-ħada ‘nobody,’ which can be postverbal co- occurring with negation, or preverbal without negation. Crucially, this NCI can occur with another NI maʕumriš ‘never’ and displays similar behavior. Specifically, wala-ħada has an NC reading when following maʕumriš but a DN reading with the inverse word order: (5)
wala-ħada NCI-person
a.
* (ma) zaarat * (neg) visit.3fsg.pfv ‘She did not visit anyone.’ [NC]
b.
wala-ħada (*ma) zaar-ha NCI-person (*neg) visited.3Msg ‘Nobody visited her.’ [NC]
c.
ma-ʕumri-š wala-ħada zaar-ha neg-ever-neg NCI-one visited.3msg-her ‘No one ever visited her.’ [NC] *‘No one has never visited her.’ (= everyone visited her.) [*DN]
d.
wala-ħada[uneg] maʕumrhuuš[ineg] zaar-ha NCI-person never visited.3msg -her ‘No one has never visited her.’ (= everyone visited her.) [DN] *‘No one has ever visited her.’ [*NC]
The problem with previous analyses is that in NC languages like JA, NIs in preverbal position cannot be always treated as the realization of a negative operator (contra Zeijlstra, 2004, 2008). Under the analysis developed in this monograph, NIs that do not display split scope readings are inherently negative and not licensed by covert or overt negation. On the other hand, NIs with split scope readings are not inherently negative and are licensed by a covert/ overt negative constituent. These NIs may also be licensed by covert negative
S u m m a r y a n d C o n c l u s i o n s
[ 217 ]
8 1 2
operator as a last resort when no NM or inherently negative NI is available in their syntactic domain (c-command domain). As such, the analysis departs from Zeijlstra (2004, 2008) in positing a negative operator for all NIs that co-occur without NMs. Positing a negative operator to license the preverbal NCI wala-ħada incorrectly predicts a double negation reading when wala-ħada is preceded by the NI maʕumriš ‘never.’ This NI is clearly inherently negative, and this is in direct contradiction to Penka’s (2011) claim that NIs are always semantically nonnegative and licensed by a covert/overt negation, even in DN languages. The presence of a covert negative operator is further supported in the analysis of the preverbal NCIs wala-ħada and wala-waaħid. These preverbal NCIs occur without a NM in sentences with an NC reading, and we get a DN reading if a NM follows. The NCI wala-NP in preverbal position is a constituent that is base-generated in the left periphery of the clause (the CP [complementizer phrase] layer). In other words, it is higher than the NegP that dominates TP. Consequently, the NM neither c-commands it nor enters into a Specifier–Head relation with it. Therefore, the NCI must be licensed by a covert negative operator. The following are derivations for examples (5)a–d: (6)
*(ma) [iNEG]
Op¬ [iNEG]
(*NEG) visited.3MSG-her
visited.3MSG-her
visited.3MSG-her
This analysis predicts the readings in (5)a–d and the DN reading when the preverbal NCI wala-NP co-occurs with a NM:
[ 218 ] A Unified Theory of Polarity Sensitivity
9 1 2
(7)
The negative operator argued for in c hapter 4 was shown to be a last resort mechanism that is available to license NCIs that are located higher than TP and carry an uninterpretable negation [uneg] feature that is not c-commanded by a constituent carrying an [ineg] feature. Specifically, a preverbal NCI that follows the negative compound maʕumrhuš should result in an NC reading. This is expected because the negative compound carries an [ineg] feature that can license the [uneg] feature of the NCI. In c hapter 4, we developed a fine-grained analysis with an inventory of the formal negative features on NCIs and NMs as shown in table 6.2. This analysis built on a multi-locus analysis of negation in Arabic (Alqassas, 2019) à la Zanuttini (1997) for Romance and Ramchand (2001) for Bengali. It also contrasted the behavior of NIs in non-strict NC languages like JA and EA with strict-NC languages like MA (Moroccan Arabic). The contribution of this analysis to linguistic theory is that NIs can, indeed, be semantically negative, pace Penka (2011), and can even license complex CPs coordinated by NCIs laa-wala ‘neither nor’: (8)
To test the explanatory adequacy of this analysis, we can look for cases where the coordinate complex, which is an NCI, is preceded by an NPI. Given that the analysis posits a covert negative operator scoping over the coordinate complex, and given that the adverbial NPI ʕumr can be merged in the left periphery of the clause, it follows that the NPI ʕumr should be able to occur in a position preceding the coordinate complex laa-wala. This is, indeed, borne out:
S u m m a r y a n d C o n c l u s i o n s
[ 219 ]
0 2
(9)
a.
b.
The NPI phrase is adjoined to the external CP and the negative operator is in NegP above the entire coordinated CPs. Through the analysis of coordinate complexes, we showed that even structures as large as coordinated CPs, i.e., CPs disjoined by the coordinators laa- wala, are licensed by a covert negative operator. We thus put forward a unified analysis for the determiner NCI wala ‘no’ and the disjunction coordinator wala ‘nor.’ Both of these structures are licensed by overt negative constituents (NM or NIs) or a covert negative operator (as a last resort mechanism). We discussed the syntax of coordinate complexes formed using laa-wala ‘neither-nor’ within the larger context of coordinate complexes. The sophisticated system of coordinators includes coordinator wa ‘and’ (= conjunctive reading); coordinator ʔaw ‘or’ (= free choice reading); coordinators ʔimma-ʔaw/ yaa-yaa ‘either-or’ (= disjunctive reading); and coordinator walla/willa/ʔam ‘or,’ in interrogative contexts (= disjunctive reading). Discussing the licensing of the coordinate complex laa-wala ‘neither-nor’ within this larger context is necessary for a deeper understanding and analysis of the understudied laa-wala structures. By comparing the various coordinate complexes, we were able to show that disjunction constructions license bare argument ellipsis (BAE), while conjunction constructions do not. Crucially, by positing a covert negative operator higher than the entire coordinate complex laa-wala ‘neither nor,’ we were able to explain how such structures behave like the non-strict NCI wala-NP. Further, this analysis explains the pattern of negative agreement exhibited by laa-wala, revealing that the phrases coordinated by laa-wala are, in fact, coordinated CPs that involve clausal coordination with a BAE in one of the disjoined CPs. (10) CP coordinate complex laa-wala: a. [NegP Neg Op¬ [DisjP [CP [EXTERNAL CONJUNCT] Disj wala[uNEG] [CP INTERNAL CONJUNCT b.
[NegP Neg Op¬ [DisjP [DP [EXTERNAL CONJUNCT] Disj wala[uNEG] [DP INTERNAL CONJUNCT
The conclusions we reached in this book have significant implications for the syntax–semantics mapping. In his recent book, Chomsky defines the basic [ 220 ] A Unified Theory of Polarity Sensitivity
1 2
property of language as mapping from syntax to semantics and making language the instrument of thought. Syntax generates “an unbounded array of hierarchically structured expressions mapping to the conceptual-intentional interface . . .” (Chomsky, 2016, p.13). The nature of this mapping then is the ultimate test for the strong minimalist thesis. The results of this monograph suggest that syntactic computation of Arabic polarity (externally merged in the left periphery) is subservient to the conceptual–intentional interface. The last resort insertion of covert negation operators in the CP layer to interpret non-strict NCIs is an extra mechanism that serves the semantic interface. This mechanism, however, adds to the complexity of syntactic computation. Similarly, the exclusivity of higher NMs as licensors of head NPIs in the left periphery adds more constraints on the syntactic licensing, making it more selective and perhaps leading to (feeding) the Jespersen cycle of change.
6.3. THE SYNTAX OF HEAD-L IKE PSIS
The syntax of PSIs treats the analyst with another puzzling phenomenon: PSIs with head-like properties. On the surface, these PSIs exhibit head-like properties and do not seem to be in the (immediate) domain of negation. These PSIs precede the grammatical subject and NM of the clause. Nonetheless, these PSIs require the presence of negation in such sentences. Even more puzzling is the genitive and accusative case assignment on the pronominal following these PSIs. Throughout the discussion of these facts in c hapter 5, these observations created controversy over the status of these PSIs, their locus and phrase structure, and their licensing configurations. Building on the analysis of NPI and NCI licensing in chapters 3 and 4, respectively, I developed a unified analysis for head-like PSIs (e.g., JA adverbial NPI ʕumr, JA adverbial NCI baʕd) in chapter 5. I argued that these head-like PSIs are adverbial phrases that do not project a clausal projection and that negation licenses these PSIs either in Spec-NegP or under c-command. I showed that these PSIs select a pronominal complement that can be coindexed either with the grammatical subject or with the thematic object of the predicate. (11) a.
ʕumr * (Zeed) maa NPI.ever Zeed neg ‘Zeed has never traveled.’
saafar traveled.3msg
b.
baʕd * (Zeed) maa NCI.yet Zeed neg ‘Zeed has never traveled.’
saafar traveled.3msg
(JA)
(JA)
S u m m a r y a n d C o n c l u s i o n s
[ 221 ]
2
(12) a.
b.
ʕumr-o maa šakat minn-oh mʕalmeh (JA) NPI.ever-him neg complained.3fsg about-him teacher ‘No teacher has ever complained about him.’ (Alqassas, 2016, p. 149) baʕd-o maa šakat minn-oh mʕalmeh NCI.ever-him neg complained.3fsg about-him teacher ‘No teacher has complained about him yet.’
(JA)
In addition, I argued that negation licenses the puzzling accusative case of the pronominal complement. The advantage of this unified analysis is that it does not appeal to the problematic head–complement relation as a licensing configuration, whereby the PSI is the head and negation is the complement (cf. Benmamoun, 2006). Instead, I showed that the pronominal constituent following the PSI is its complement rather than the grammatical subject of the predicate. This complement can even be as large as CP, a relative clause. (13) a.
b.
ʕumr illi budrus maa bursub NPI.ever who study.3msg neg fail.3msg ‘He who studies never fails.’ (Alqassas, 2016, p. 142) baʕd illi zaar-na NCI.ever who visited.3msg-us ‘He who visited us has not yet left.’
maa neg
rawwaħ left.3msg
(JA)
(JA)
Under a head–complement analysis, the head-like PSI that dominates NegP would have to select the pronominal complement, and we necessarily lose the hypothesized head–complement configuration containing the PSI and negation. Analyzing head-like PSIs as phrasal, rather than clausal, constituents that take a pronominal complement paved the way for an analysis of case assignment to the pronominal complement. In the analysis, I argued that negation licenses the accusative case under an Agree process whereby negation c- commands the PSI phrase. The analysis of accusative case licensing developed in chapter 5 is empirically supported in Arabic using facts from SA (Standard Arabic) and contrasts between JA and MA. SA has cases where indefinite nouns carry accusative case markings when preceded by the NM laa. (14) laa mudarris-iin neg teachers-mpl.ACC ‘No teachers are absent.’
ɣaaʔib-uun absent-mpl.NOM
[ 222 ] A Unified Theory of Polarity Sensitivity
(SA)
3 2
I also showed that negation can appear higher than the NPI in a position that licenses the NPI and accusative case of the pronominal complement under c-command: (15) a.
b.
maa ʕumr-ha katbat neg NPI-ever-her write.3fsg.pfv ‘She has never written a novel.’
riwaaye novel
ma ʕəmmər nadya ʒat never Nadia come.3fsg.pfv ‘Nadia never came.’ (Benmamoun, 2006, p. 145)
(JA)
(MA)
Finally, I argued that negation could, indeed, raise to a position higher than PSIs where it can c-command the PSI and its accusative pronominal complement. Microvariation comparisons between JA and MA support this movement analysis through cases where the two copies of negation (two copies of maa, one above and another under the PSI) can be pronounced in MA under a concordant reading. (16) ma ʕəmmər nadya ma-ʒat neg never Nadia neg-come.3fsg.pfv ‘Nadia never came.’ (Benmamoun, 2006, p. 145) [NC] * ‘Nadia has never not come.’ [*DN]
(MA)
This property in MA nicely ties with the analysis of negative features on NMs in Arabic presented in chapter 4. The NM maa in MA is semantically vacuous. In JA, however, we argued for a movement analysis whereby the NPI is merged lower than negation and moves into Spec-NegP. (17) [NegP ʕumr-ha [Neg maa [TP NPI-ever-her neg NPI-ever-her ‘She has never written a novel.’
[TP katbat write.3fsg.pfv
riwaaye (JA) novel
The analysis has far-reaching implications to dependent case licensing in natural language. Admittedly, the implications of this analysis need more research and support from other cross-linguistic cases where negation might be involved in licensing case, which I leave to future research. To conclude, this book studied the microvariation in the syntax of PSIs (negative and positive polarity, negative concord, and NIs) in Arabic—Standard Arabic and the major regional dialects represented by Jordanian, Egyptian,
S u m m a r y a n d C o n c l u s i o n s
[ 223 ]
4 2
Moroccan, and Qatari. Through the (micro)comparative approach to studying the syntax of PSIs in Arabic, we were able to map the properties of PSIs with descriptive adequacy. We also unearthed key properties involved in syntactic licensing of PSIs and the fine-grained picture of cross-linguistic variation in their formal features.
[ 224 ] A Unified Theory of Polarity Sensitivity
5 2
BIBLIOGRAPHY
Abusch, D. (1998), “Generalizing tense semantics for future contexts,” in S. Rothstein (Ed.), Events in Grammar, Dordrecht: Kluwer, pp. 13–35. Abusch, D. (2004), “On the temporal composition of infinitives,” in J. Guéron & J. Lecarme (Eds.), The Syntax of Time, Cambridge: MIT Press, pp. 1–34. Acquaviva, P. (1997), The Logical Form of Negation: A Study of Operator-Variable Structures in Syntax, New York: Garland Publishing. Albirini, A., & E. Benmamoun (2015), “Factors affecting the retention of sentential negation in heritage Egyptian Arabic,” Bilingualism: Language and Cognition, 18(3), pp. 470–489. Alexiadou, A. (1994), Issues in the syntax of adverbs, Ph.D. dissertation, University of Potsdam. Alexopoulou, T., E. Doron, & C. Heycock (2004), “Broad subjects and clitic left dislocation,” in D. Adger, C. de Cat, and G. Tsoulas (Eds.), Peripheries: Syntactic Edges and Their Effects, Oxford: Oxford University Press, pp. 329–358. Al-Naadirii, M.A. (2009), Nahw Al-Lugha Al-ʿarabiyya, Beirut: Al-maktaba Al-assrya. Alqassas, A. (2012), The morpho-syntax and pragmatics of Levantine Arabic negation: A synchronic and diachronic analysis, Ph.D. dissertation, Indiana University, Bloomington. Alqassas, A. (2015), “Negation, tense and NPIs in Jordanian Arabic,” Lingua, 156, pp. 101–128. Alqassas, A. (2016), “Temporal NPIs and NCIs as adverb phrases: The case of Jordanian Arabic,” in Y.A. Haddad and E. Potsdam (Eds.), Perspectives on Arabic Linguistics XXVIII, Amsterdam/Philadelphia: John Benjamins Publishing Company, pp. 129–152. Alqassas, A. (2018), “Negative sensitive items,” in E. Benmamoun and R. Bassiouney (Eds.), The Routledge Handbook of Arabic Linguistics, Abington, UK; New York: Routledge. Alqassas, A. (2019), A Multi-Locus Analysis of Arabic Negation: Micro-Variation in Southern Levantine, Gulf and Standard Arabic. Edinburgh University Press. Alsarayreh, A. (2012), The licensing of negative sensitive items in Jordanian Arabic, Ph.D. dissertation, University of Kansas. Al-Tonsi, A., L. Al-Sawi, & S. Massoud (1987), An Intensive Course in Colloquial Egyptian Arabic, Center for Arabic Studies: American University of Cairo. Alwer, E. (2003), “New dialect formation: The focusing of -kum in Amman,” in D. Britain and J. Cheshire (Eds.), Social Dialectology in Honour of Peter Trudgill, Amsterdam: John Benjamins, pp. 59–67. Andersen, H. (1973), “Abductive and deductive change,” Language, 49, pp. 765–793. Anderson, S.R. (1992), A-Morphous Morphology, Cambridge: Cambridge University Press.
6 2
Aoun, J., E. Benmamoun, & L. Choueiri (2010), The Syntax of Arabic, Cambridge: Cambridge University Press. Aoun, J., E. Benmamoun, & D. Sportiche (1999), “Further remarks on first conjunct agreement,” Linguistic Inquiry, 30(4), 669–681. Ayyoub, G. (1981), Structure de la phrase en Arabe Standard, Third Cycle thesis, Paris VIII-Vincenne, France. Bahloul, R. (1996), “Negation in French and Tunisian Arabic,” in M. Eid (Ed.), Perspectives on Arabic Linguistics, Vol. 8, Amsterdam and Philadelphia: John Benjamins, pp. 67–83. Baker, C.L. (1970), “Double negatives,” Linguistic Inquiry, 1, pp. 169–186. Baker, M. (1985), “The mirror principle and morphosyntactic explanation,” Linguistic Inquiry, 16, pp. 373–415. Baker, M. (2008), The Syntax of Agreement and Concord, Cambridge: Cambridge University Press. Bakir, M. (1980), Aspects of Clause Structure in Arabic, Bloomington: Indiana University Linguistics Club. Benmamoun, E. (1992), Inflectional and functional morphology: Problems of projection, representation and derivation, Ph.D. dissertation, University of Southern California, Los Angeles. Benmamoun, E. (1996), “Negative polarity and presupposition in Arabic,” in M. Eid (Ed.), Perspectives on Arabic Linguistics, Vol. 8, Amsterdam: John Benjamins, pp. 47–66. Benmamoun, E. (1997), “Licensing of negative polarity items in Moroccan Arabic,” Natural Language and Linguistic Theory, 15, pp. 263–287. Benmamoun, E. (2000), The Feature Structure of Functional Categories: A Comparative Study of Arabic Dialects, Oxford: Oxford University Press. Benmamoun, E. (2003), “Reciprocals as plurals,” in J. Lecarme (Ed.), Research in Afroasiatic Grammar II: Selected Papers from the Fifth Conference on Afroasiatic Languages, Paris, 2000, Amsterdam: John Benjamins, pp. 53–62. Benmamoun, E. (2006), “Licensing configurations: The puzzle of head negative polarity items,” Linguistic Inquiry, 37(1), pp. 141–149. Benmamoun, E., & R. Kumar (2006), “The overt licensing of NPIs in Hindi,” in R. Singh (Ed.), Yearbook of South Asian Languages and Linguistics 2006, Berlin: Mouton de Gruyter, pp. 31–48. Benmamoun et al. (2013), “The location of sentential negation in Arabic varieties,” Brill’s Annual of Afroasiatic Languages and Linguistics, 5, 83–116. Bianchi, V. (1995), Consequences of antisymmetry for the syntax of headed relative clauses, Ph.D. dissertation, Scuola Normale Superiore, Pisa. Blau, J. (1960), Syntax des palästinensichen Bauerndialektes von Bīr-Zēt: Auf Grund der “Volkserzählungen aus Palästina” von Hans Schmidt und Paul Kahle, Walldorf, Hessen: Verlag für Orientkunde Dr. H. Vorndran. Blau, J. (1965), The Emergence and Linguistic Background of Judaeo-Arabic, Oxford: Oxford University Press. Blau J. (1981), The Emergence and Linguistic Background of Judaeo-Arabic, Jerusalem: Ben Zvi Institute. Blau, J. (2002), A Handbook of Early Middle Arabic, Jerusalem: Max Schloessinger Memorial Foundation. Bobaljik, J., & H. Thrainsson (1998). “Two heads aren’t always better than one,” Syntax, 1.1, 37–71.
[ 226 ] Bibliography
7 2
Boeckx, C., & S. Stjepanovic (2001), “Head-ing toward PF,” Linguistic Inquiry, 32(2), pp. 345–355. Borer, H. (1995), “The ups and downs of Hebrew verb movement,” Natural Language and Linguistic Theory, 13, pp. 527–606. Bošković, Ž. (1996), “Selection and the categorial status of infinitival complements,” Natural Language and Linguistic Theory, 14, pp. 269–304. Bošković, Ž. (1997), The Syntax of Nonfinite Complementation: An Economy Approach, Cambridge, MA: MIT Press. Bowers, J. (1993), “The syntax of predication,” Linguistic Inquiry, 24, pp. 591–656. Brown, S. (1999), The Syntax of Negation in Russian: A Minimalist Approach, Stanford, CA: CSLI Publications. Brown, S., & S. Franks (1995), “Asymmetries in the scope of Russian negation,” Journal of Slavic Linguistics, 3, pp. 239–287. Brustad, K. (2000), The Syntax of Spoken Arabic: A Comparative Study of Moroccan, Egyptian, Syrian and Kuwaiti Dialects, Washington, DC: Georgetown University Press. Burrdige, K. (1993), Syntactic Change in Germanic: Aspects of Language Change in Germanic, with Particular Reference to Middle Dutch, Amsterdam: Benjamins. Burzio, L. (1986), Italian Syntax, Dordrecht: Reidel. Bybee, J. (2003), “Mechanisms of change in grammaticization: The role of frequency,” in R. Janda and B. Joseph (Eds.), Handbook of Historical Linguistics, Oxford: Blackwell, pp. 602–623. Bybee, J., & P. Hopper (Eds.) (2001), Frequency and the Emergence of Linguistic Structure, Amsterdam: Benjamins. Campbell, L. (2004), Historical Linguistics: An Introduction, Cambridge, MA: MIT Press. Chatzopoulou, A. (2018), Negation and Nonveridicality in the History of Greek. Oxford: Oxford University Press. Chomsky, N. (1965), Aspects of the Theory of Syntax, Cambridge, MA: MIT Press. Chomsky, N. (1970), “Remarks on nominalization,” in R. Jacobs and P. Rosenbaum (Eds.), Reading in English Transformational Grammar, Waltham: Ginn, pp. 184–221. Chomsky, N. (1975), Reflections on Language, New York: Pantheon. Chomsky, N. (1981), Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1989), “Some notes on the economy of derivation,” in I. Laka and A. Mahajan (Eds.), Functional Heads and Clause Structure, MIT Working Papers in Linguistics, Vol. 10, Cambridge, MA: MIT Press. Chomsky, N. (1993), “A minimalist program for linguistic theory,” in Kenneth L. Hale and S. Jay Keyser (Eds.), The view from Building 20: Essays in linguistics in honor of Sylvain Bromberger, Cambridge, MA: MIT Press, pp. 1–52 [Reprinted in Chomsky (1995), 167–217]. Chomsky, N. (1995), The Minimalist Program, Cambridge, MA: MIT Press. Chomsky, N. (2000), “Minimalist inquiries, the framework,” in R. Martin et al. (Eds.), Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik, Cambridge, MA: MIT Press, pp. 1–52. Chomsky, N. (2001), “Derivation by phase,” in M. Kenstowicz (Ed.), Ken Hale: A Life in Language, Cambridge, MA: MIT Press, pp. 1–52. Chomsky, N. (2004), “Beyond explanatory adequacy,” in A. Belletti (Ed.), Structures and Beyond, Oxford: Oxford University Press, pp. 104–131. Chomsky, N. (2005), “Three factors in language design,” Linguistic Inquiry, 36, pp. 1–22.
Bibliography [ 227 ]
8 2
Chomsky, N. (2008), “On phases,” in R. Freidin, C. Otero, and M.L. Zubizarreta (Eds.), Foundational Issues in Linguistic Theory, Cambridge, MA: MIT Press, pp. 133–166. Chomsky, N. (2016), What Kind of Creatures Are We? New York: Columbia University Press. Cinque, G. (1999), Adverbs and Functional Heads: A Cross Linguistic Perspective, Oxford: Oxford University Press. Collinder, B. (1969), Survey of the Uralic Languages, 2nd revised edition, Stockholm: Almqvist & Wiksell. Collins, C., & P.M. Postal (2014), Classical neg Raising, Cambridge, MA: MIT Press. Collins, C., & P.M. Postal (2017), “Interclausal neg raising and the scope of negation,” Glossa: A Journal of General Linguistics, 2(1), pp. 1–29, DOI: https://doi.org/ 10.5334/g jgl.79 Cormack, A., & N. Smith (2000), “Head movement and negation in English,” Transactions of the Philological Society, 98(1), pp. 49–85. Cowart, W., & D. McDaniel (2008), “What kind of thing is a coordination?: Qualitative comparisons of agreement relations in coordinate and subordinate sentences,” ms. University of Southern Maine. Cowell, M. (1964), A Reference Grammar of Syrian Arabic, Washington, DC: Georgetown. Culicover, P. W., & R. Jackendoff (1997), “Semantic subordination despite syntactic coordination,” Linguistic Inquiry, 28, pp. 195–217. Culicover, P. W., & R. Jackendoff (2005), Simpler Syntax. Oxford: Oxford University Press. Dahl, Ö. (1979), “Typology of sentence negation,” Linguistics, 17, pp. 79–106. Davidson, D. (1967), “The logical form of action sentences,” in N. Rescher (Ed.), The Logic of Decision and Action, Pittsburgh: University of Pittsburgh Press, pp. 115–121. Davies, H. (1981), Seventeenth-century Egyptian Arabic: A profile of the colloquial material in Yusuf al-Sirbini’s Hazz al-Quhuf fi Sarh Qasid abi Saduf, Ph.D. dissertation, University of California, Berkeley. Dayal, V. (2004), “The universal force of free choice any,” Linguistic Variation Yearbook, 4, pp. 5–40. De Groot, A.W. (1959), Structurele syntaxis, Den Haag: Servire. de Jong, K.J., & B.A. Zawaydeh (1999), “Stress, duration, and intonation in Arabic word-level prosody,” Journal of Phonetics, 27, pp. 3–22. den Dikken, M. (2006), “Either- float and the syntax of co- or- dination,” Natural Language and Linguistic Theory, 24, 689–749. Depiante, M. (2000), The syntax of deep and surface anaphora: A study of null complement anaphora and stripping/bare argument ellipsis, Ph.D. dissertation, University of Connecticut. Diesing, M. (1992), Indefinites, Cambridge, MA: MIT Press. Diesing, M., & E. Jelinek (1995), “Distributing arguments,” Natural Language Semantics, 3(2), pp. 123–176. Doron, E., & C. Heycock (1999), “Filling and licensing multiple specifiers,” in D. Adger et al. (Eds.), Specifiers: Minimalist Approaches, Oxford: Oxford University Press, pp. 69–89. Edmonds, J. (1976), A Transformational Approach to English Syntax, New York: Academic Press. Embick, D., & R. Izvorski (1996), “Participle-auxiliary word-orders in Slavic,” in Formal Approaches to Slavic Linguistics: The Cornell Meeting, Ann Arbor: Michigan Slavic Publications, University of Michigan, pp. 210–239.
[ 228 ] Bibliography
9 2
Embick, D., & R. Noyer (2001), “Movement operations after syntax,” Linguistic Inquiry, 32(4), pp. 555–595. Enc, M. (1991), On the absence of the present tense morpheme in English, MS thesis, University of Wisconsin, Madison. Engels, E. (2012), Optimizing Adverb Positions, Amsterdam: John Benjamins. Fassi- Fehri, A. (1993), Issues in the Structure of Arabic Clauses and Words, Dordrecht: Kluwer. Fassi Fehri, A. (2000/2004), “Temporal/Aspectual interaction and variation across Arabic heights,” in J. Lecame (Ed.), The Syntax of Time, Cambridge, MA: MIT Press, pp. 235–257. Ferguson, C. (1983), “God wishes in Syrian Arabic,” Mediterranean Language Review, 1, pp. 65–83. Fromkin, V., R. Rodman, & N. Hyams (2007), An Introduction to Language, 8th edition, MA: Thomson Wadsworth. Giannakidou, A., (1995), “Subjunctive, habituality and negative polarity,” Semantics and Linguistic Theory, 5, pp. 132–150. Giannakidou, A. (1997), The landscape of polarity items, Ph.D. thesis, University of Groningen. Giannakidou, A. (1998), Polarity Sensitivity as (Non)Veridical Dependency, Amsterdam and Philadelphia: John Benjamins. Giannakidou, A. (1999), “Affective dependencies,” Linguistics & Philosophy, 22, pp. 367–421. Giannakidou, A. (2000), “Negative . . . concord?” Natural Language and Linguistic Theory, 18, 457–523. Giannakidou, A. (2001), “The meaning of free choice,” Linguistics and Philosophy, 24, 659–735. Giannakidou, A. (2006), “N-words and negative concord,” in M. Everaert and H. van Riemsdijk (Ed.), The Blackwell Companion to Syntax, Volume III, Malden, MA: Blackwell Publishing, pp. 327–391. Giannakidou, A. (2009), “The dependency of the subjunctive revisited: Temporal semantics and polarity,” Lingua, 119, pp. 1883–1908. Giannakidou, A. (2011), “Negative and positive polarity items: Licensing, compositionality and variation,” in C. Maineborn et al. (Eds.), Semantics: An International Handbook of Natural Language Meaning, Berlin: Mouton de Gruyter, pp. 1660–1712. Giannakidou, A. S. Yoon (2016), “Scalar marking without scalar meaning: non-scalar, non-exhaustive EVEN-marked NPIs in Greek and Korean,” Language, 522–556. Giannakidou, A., & H. Zeijlstra (2017), “The landscape of negative dependencies: Negative concord, N-words,” in M. Evaerert and H. van Riemsdijk (Eds.), The Blackwell Companion to Syntax, 2nd edition., London: Blackwell, pp. 1–38. DOI: 10.1002/9781118358733 Giorgi, A., & F. Pianesi (1997), Tense and Aspect: From Semantics to Morphosyntax, Oxford: Oxford University Press. Goodall, G. (1987), Parallel Structures in Syntax, Cambridge: Cambridge University Press. Haegeman, L. (1995), The Syntax of Negation, Cambridge: Cambridge University Press. Haegeman, L. (2002), “Some notes on DP- internal negative doubling,” in S. Barbiers, L. Cornips, and S. van der Kleij (Eds.), Syntactic Microvariation, Amsterdam: Meertens Institute Electronic Publications in Linguistics 2, pp. 152–184.
Bibliography [ 229 ]
0 3 2
Haegeman, L., & T. Lohndal (2010), “Negative concord and (multiple) agree: A case study of West Flemish,” Linguistic Inquiry, 41(2), pp. 181–211. Haegeman, L., & R. Zanuttini (1991), “Negative heads and the neg criterion,” The Linguistic Review, 8, pp. 233–251. Haegeman, L., & R. Zanuttini (1996), “Negative concord in West Flemish,” in A. Belletti and L. Rizzi (Eds.), Parameters and Functional Heads. Essays in Comparative Syntax, Oxford: Oxford University Press, pp. 117–179. Halle, M., & A. Marantz (1993), “Distributed morphology and the pieces of inflection,” in K. Hale & S.J. Keyser (Eds.), The View from Building, vol. 20, Cambridge, MA: MIT Press, pp. 111–176. Hallman, P. (2015), “The Arabic imperfective,” Brill’s Journal of Afroasiatic Languages and Linguistics, 7, pp. 103–131. Halpern, A. (1992), Topics in the placement and morphology of clitics, Ph.D. dissertation, Stanford University, Stanford, CA. Han, C.-H., & M. Romero (2004), “The syntax of whether/Q . . . or questions: Ellipsis combined with movement,” Natural Language and Linguistic Theory, 22, 527–564. Hankamer, J. (1979), “Deletion in coordinate structures,” in J. Hankhamer (Ed.), Outstanding Dissertations in Linguistics, New York: Garland Publishing. Hankamer, J., & I. Sag (1976), “Deep and surface anaphora,” Linguistic Inquiry, 7, 391–428. Harrell, R. (1962), A Short Reference Grammar of Moroccan Arabic, Washington, DC: Georgetown University Press. Haspelmath, M. (2005), “Negative indefinite pronouns and predicate negation,” in M. Haspelmath, M. S. Dryer, D. Gil, & B. Comrie (Eds.), The World Atlas of Language Structures. Oxford: Oxford University Press, pp. 466–469. Heim, I. (1982), The semantics of definite and indefinite noun phrases, Ph.D. dissertation, University of Massachusetts, Amherst. Heim, I. (1988), The Semantics of Definite and Indefinite Noun Phrases, New York: Garland. Heim, I. (1994), Comments on Absuch’s theory of tense. Unpublished manuscript, MIT. Hendriks, P. (2001), “Initial coordination and the law of coordination of likes,” in H. Broekhuis and T. van der Wouden (Eds.), Linguistics in the Netherlands 2001, Amsterdam/Philadelphia: John Benjamins Publishing Company, pp.127–138. Hendriks, P. (2003), “Either” as a focus particle. Unpublished manuscript, University of Groningen. Hendriks, P. (2004), Either, both and neither in coordinate structures. Unpublished manuscript, University of Groningen. Herburger, E. (2001), “Negative concord revisited,” Natural Language Semantics, 9, pp. 289–333. Hirschbühler, P., & M. Lahelle (1992/1993), “Le statut de (ne) pas en français contemporain,” Recherches Linguistiques de Vincennes, 22, pp. 31–58. Hoeksema, J. (1997), “Negation and negative concord in Middle Dutch,” in D. Forget et al. (Eds.), Negation and Polarity: Syntax and Semantics, Amsterdam: Benjamins, pp. 139–156. Höhle, T. (1991), “On reconstruction and coordination,” in Representation and Derivation in the Theory of Grammar, H. Haider and K. Netter (Eds.), Kluwer: Dordrecht, pp. 139–197. Holes, C. (1990), Gulf Arabic, London: Routledge. Holes, C. (1995), Modern Arabic: Structures, Functions and Varieties, Washington, DC: Georgetown University Press. Hornstein, N. (1995), Logical Form: From GB to Minimalism, Cambridge, MA: Blackwell.
[ 230 ] Bibliography
1 3 2
Hoyt, F. (2005a), Negative concord in two dialects of Arabic, Unpublished manuscript, University of Texas, Austin. Hoyt, F. (2005b), Negative concord and restructuring in Palestinian Arabic, Unpublished manuscript. University of Texas, Austin. Hoyt, F. (2005c), Sentential negation marking in Palestinian and Moroccan Arabic: A study in comparative morphosyntax, Unpublished manuscript, University of Texas, Austin. Hoyt, F. (2007), “An Arabic wackernagel clitic? The morphosyntax of negation in Palestinian Arabic,” in M. Mughazy (Ed.), Perspectives on Arabic Linguistics XX, Amsterdam: John Benjamins, pp. 105–134. Hoyt, F. (2010), Negative concord in Levantine Arabic, Ph.D. dissertation, University of Texas, Austin. Hudson, R. A. (1976), Arguments for a Non-transformational Grammar. Chicago: Chicago University Press. Iatridou, S. (1990), “About Agr(P),” Linguistic Inquiry, 21, pp. 551–577. Jackendoff, R. (1969), “An interpretive theory of negation,” Foundations of Language, 5, pp. 218–241. Jackendoff, R. (1972), Semantic Interpretation in Generative Grammar, Cambridge, MA: MIT Press. Jackendoff, R. (1973), “The base rules for prepositional phrases,” In S. Anderson and P. Kiparsky, (Eds.), A Festschrift for Morris Halle, New York: Holt, Rienhart and Winston, pp. 345–356. Jackendoff, R. (1977), “X-bar-syntax: A study of phrase structure,” Linguistic Inquiry Monograph 2, Cambridge, MA: MIT Press. Jackendoff, R. (1981), Xbar Syntax: A Study of Phrase Structure, Cambridge, MA: MIT Press. Jespersen, O. (1917), Negation in English and Other Languages, Copenhagen: Host. Johannessen, J. (1998), Coordination, Oxford: Oxford University Press. Johnson, K. (1991), “Object positions,” Natural Language and Linguistic Theory, 9, pp. 577–636. Johnson, K. (1996), “Gapping. Ch. 2 of: In search of the English middle field,” ms., University of Massachusetts-Amherst. Johnstone, T.M. (1967), Eastern Arabian Dialect Studies, London Oriental Series 17, London: Oxford University Press. Kayne, R. (1989), “Notes on English agreement,” Central Institute of English and Foreign Languages Bulletin, 1, pp. 40–67. Kayne, R. (1992), “Italian negative infinitival imperatives and clitic-climbing,” in L. Tasmowski Ryck & A. Zribi-Hertz (Eds.), in Hommages à Nicolas Ruwet, Ghent: Communication & Cognition, pp. 300–312. Kayne, R. (1994), The Antisymmetry of Syntax, Cambridge, MA: MIT Press. Kayne, R. (1998), “Overt vs. covert movement,” Syntax, 1, pp. 128–191. Kemenade, V. (2000), “Jespersen’s cycle revisited: Formal properties of grammaticalization,” in S. Pintzuk, G. Tsoulas, and A. Warner (Eds.), Diachronic Syntax: Models and Mechanisms, Oxford: Oxford University Press, pp. 51–75. Khan, G. (1999), A Grammar of Neo- Aramaic: The Dialect of the Jews of Arbel, Leiden: Brill. Khan, G. (2004), The Jewish Neo- Aramaic Dialect of Sulemaniyya and H: Alabja, Leiden: Brill. Kitagawa, Y. (1986), Subjects in Japanese and English, Ph.D. dissertation, University of Massachusetts, Amherst.
Bibliography [ 231 ]
2 3
Kolokonte, M. (2006), Bare Argument Ellipsis and Polarity in Modern Greek. First Newcastle Postgraduate Conference in Theoretical and Applied Linguistics. Newcastle University, United Kingdom. Koopman, H., & D. Sportiche (1991), “The position of subjects,” Lingua, 85(2/3), pp. 211–258. Kouloughli, D.E. (1988), “Renouvellement énonciatif et valeur aoristique: à propos de l’opposition maa/lam en arabe,” Langues Orientales Anciennes, Philologie et Linguistique, 1, pp. 49–72. Kramer, R. (2010), “The Amharic definite marker and the syntax-morphology interface,” Syntax, 13, 196–240. Kuroda, S.-Y. (1988), “Whether we agree or not: A comparative syntax of English and Japanese,” Lingvisticæ Investigationes, 12, pp. 1–47. Laanest, A. (1966), “Izhorskii iazyk,” in V.I. Lytkin (Ed.), IAzyki Narodov SSSR, Tom 3, Finno-Ugorskiei Samodiiskie Iazyki, Moscow: Nauka, pp. 109–111. Ladusaw, W. (1977), “The scope of some sentence adverbs and surface structure,” in M.J. Stein (Ed.), Proceedings of NELS 8, Amherst: GLSA University of Massachusetts, pp. 97–111. Ladusaw, W. (1979), Polarity sensitivity as inherent scope relations, Ph.D. dissertation, University of Texas, Austin. Ladusaw, W. (1988), “Adverbs, negation, and QR,” in H. Haghoe (Ed.), Linguistics in the Morning Calm 2, Selected Papers from SICOL-1986, Seoul: The Linguistic Society of Korea, pp. 481–488. Ladusaw, W. (1992), “Expressing negation,” in C. Baker et al. (Eds.), Proceedings of SALT II, Cornell, NY: Cornell Linguistic Circle. Laka, I. (1990), Negation in syntax: On the nature of functional categories and projections, Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA. Lakoff, G. (1986), “Frame semantic control of the coordinate structure constraint,” in Anne M. Farley et al. (Eds.), Chicago Linguistic Society 22, Part 2: Papers from the Parasession on Pragmatics and Grammatical Theory, Chicago: CLS, pp. 152–167. Larson, R. (1985), “On the syntax of disjunction scope,” Natural Language and Linguistic Theory, 3, 217–264. Lasnik, H. (1975), “On the semantics of negation,” in D.J. Hockney (Ed.), Contemporary Research in Philosophical Logic and Linguistic Semantics, Dordrecht: Reidel, pp. 279–311. Lebeaux, D. (1990), “Relative clauses, licensing and the nature of the derivation,” in Juli Carter, Rose-Marie Déchaine, Bill Philip, & Tim Sherer (Eds.), Proceedings of NELS, vol. 20, GLSA, University of Massachusetts at Amherst, pp. 318–332. Levin, N., & E. Prince (1986), “Gapping and causal implicature,” Papers in Linguistics, 19.3, 351–364. Lightfoot, D. (2006), How New languages Emerge, Cambridge: Cambridge University Press. Lightfoot, D. (2017), “Discovering new variable properties without parameters,” Linguistic Analysis (special edition, Simin Karimi & Massimo (Eds.), Piattelli- Palmarini, Parameters: What Are They? Where Are They?), 41, 3–4, pp. 409–445. Linebarger, M.C. (1981), The Grammar of Negative Polarity, Bloomington: Indiana University Linguistics Club. Linebarger, M.C. (1987), “Negative polarity and grammatical representation,” Linguistics and Philosophy, 10, pp. 325–387. Lobeck, A. (1995), Ellipsis: Functional Heads, Licensing and Identification, New York: Oxford University Press.
[ 232 ] Bibliography
32
Lorimor, H. (2007), Conjunctions and grammatical agreement, Ph.D. dissertation, University of Illinois at Urbana-Champaign. Lucas, C. (2007), “Jespersen’s cycle in Arabic and Berber,” Transactions of the Philological Society, 105, pp. 398–431. Lucas, C. (2009), The development of negation in Arabic and Afroasiatic, Ph.D. dissertation, University of Cambridge. Marantz, A. (1984), On the Nature of Grammatical Relations, Cambridge, MA: MIT Press. Marantz, A. (1991), “Case and licensing,” in G. Westphal, B. Ao, and H.R. Chae (Eds.), Proceedings of the 8th Eastern States Conference on Linguistics (ESCOL 8), Ithaca, NY: CLC Publications, pp. 234–253. Marantz, A. (1988), “Clitics, morphological merger, and the mapping to phonological structure,” in M. Hammond and M. Noonan (Eds.), Theoretical morphology, San Diego, CA: Academic Press, pp.–253–270. Martin, R. (1996), A Minimalist theory of PRO and control, Ph.D. dissertation, University of Connecticut, Storrs. Martin, R. (2001), “Null case and the distribution of PRO,” Linguistic Inquiry, 32, pp. 141–166. McCarthy, J. (1979), Formal approaches in Semitic phonology and morphology, Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA. McCloskey, J. (1997), “Subjecthood and subject positions,” in L. Haegeman (Ed.), Elements of Grammar, Dordrecht: Kluwer, pp. 197–235. Merchant, J. (2000), “Antecedent- contained deletion in negative polarity items,” Syntax, 3(2), pp. 144–150. Merchant, J. (2001), The syntax of silence: Sluicing, islands, and the theory of ellipsis. 2001. Oxford University Press: Oxford. Mitchell, E. (2004), “Evidence from Finnish for Pollock’s theory of IP,” Lingua, 116, pp. 228–244. Miyagawa, S. (2010), Why Agree? Why Move? Unifying Agreement-based and Discourse- Configurational Languages, Linguistics Inquiry Monograph 54, Cambridge, MA: MIT Press. Mohammad, M. (1989), The sentence structure of Arabic, Ph.D. dissertation, University of Southern California, Los Angeles. Mohammad, M. (1997), “A minimalist approach to verb movement in Standard Arabic,” Studia Linguistica, 51(3), p. 317–338. Mohammad, M. (1998), The Syntax of Indefinite Subjects in Equative Sentences in Palestinian Arabic, Unpublished manuscript, University of Florida, Gainsville. Mohammad, M. (2000), Word Order, Agreement, and Pronominalization in Standard and Palestinian Arabic, Philadelphia: John Benjamins. Moltmann, F. (1992), “Reciprocals and same/different: Towards a semantic analysis,” Linguistics and Philosophy, 16, 411–462. Moritz, L., & D. Valois (1994), “Pied-piping and specifier-head agreement,” Linguistic Inquiry, 25(4), pp. 667–707. Moutaouakil, A. (1991), “Negative constructions in Arabic: Towards a functionalist approach,” in K. Devengi & T. Ivany (Eds.), Journal of The Arabist, Budapest, pp. 263–296. Moutaouakil, A. (1993), al-wathiifa wa l-binya, Casablanca: ‘ocaadh. Mughazy, M. (2003), “Metalinguistic negation and truth functions: The case of Egyptian Arabic,” Journal of Pragmatics, 35, pp. 1143–1160. Munn, A. (1987), “Coordinate structure and X-bar theory,” McGill Working Papers in Linguistics, 4.1, 121–140.
Bibliography [ 233 ]
4 3 2
Munn, A. (1993), Topics in the syntax and semantics of coordinate structures, Ph.D. dissertation, University of Maryland. Munn, A. (1999), “First conjunct agreement: Against a clausal analysis,” Linguistic Inquiry, 30(4), 643–668. Neijt, A. (1979), Gapping: A Contribution to Sentence Grammar, Dordrecht: Foris Publications. Nunes, J. (2004), Linearization of Chains and Sideward Movement, Cambridge, MA: MIT Press. Ouali, H., & C. Fortin (2007), “The syntax of complex tense in Moroccan Arabic,” in E. Benmamoun (Ed.), Perspectives of Arabic Linguistics, Vol. 19, Amsterdam: John Benjamins, pp. 175–187. Ouali, H., & U. Soltan (2014), “On negative concord in Egyptian and Moroccan Arabic,” in S. Farwaneh and H. Ouali (Eds.), Perspectives of Arabic Linguistics, Vol. 25, Amsterdam: John Benjamins, pp. 159–179. Ouhalla, J. (1990), “Sentential negation, relativized minimality and the aspectual status of auxiliaries,” The Linguistic Review, 7, pp. 183–231. Ouhalla, J. (1991), Functional Categories and Parametric Variation, London/ New York: Routledge. Ouhalla, Jamal. (1992), “Focus in standard Arabic,” Linguistics in Potsdam 1, 65–92. Ouhalla, J. (1993), “Negation, focus and tense: The Arabic ‘ma’ and ‘la,’” Rivista di Linguistica, 5, pp. 275–300. Ouhalla, J. (1994a), “Focus in Standard Arabic,” Linguistics in Potsdam, 1, pp. 65–92. Ouhalla, J. (1994b), “Verb movement and word order in Arabic,” in D. Lightfoot and N. Hornstein (Eds.), Verb Movement, Cambridge: Cambridge University Press. Ouhalla, J. (2002), “The structure and logical form of sentences in Arabic,” in J. Ouhalla and U. Shlonsky (Eds.), Themes in Arabic and Hebrew Syntax, Dordrecht: Kluwer Academic Publishers, pp. 299–320. Parsons, T. (1990), Events in the Semantics of English, Cambridge, MA: MIT Press. Partee, B.H. (1973), “Some structural analogies between tense and pronouns in English,” Journal of Philosophy, 70, pp. 601–609. Partee, B.H. (1984), “Nominal and temporal anaphora,” Linguistics and Philosophy, 7, pp. 243–286. Penka, D. (2011), Negative Indefinites, Oxford: Oxford University Press. Pesetsky, D. (1992), Zero syntax II: An essay on infinitives, Unpublished manuscript, Massachusetts Institute of Technology, Cambridge, MA. Pesetsky, D., & E. Torrego (2001), “T-to-C movement: Causes and consequences,” in Michael Kenstowicz (Ed.), Ken Hale: A Life in Language, Cambridge, MA: MIT Press, pp. 355–426. Pesetsky, D., and E. Torrego (2006), “Probes, goals, and syntactic categories,” in Yukio Otsu (Ed.), Proceedings of the 7th Annual Tokyo Conference on Psycholinguistics, Tokyo: Hituzi Syobo, pp. 25–60. Plunkett, B. (1993), “The positions of subjects in Modern Standard Arabic,” in M. Eid and C. Holes (Eds.), Perspectives on Arabic Linguistics V, Philadelphia: J. Benjamins, pp. 231–260. Pollock, J.-Y. (1989), “Verb movement, universal grammar, and the structure of IP,” Linguistic Inquiry, 20, pp. 365–424. Pollock, J.-Y. (1993), Notes on Clause Structure, Unpublished manuscript, Université de Picardie, Amiens. Potsdam, E. (1996), Syntactic issues in the English imperative, Ph.D. dissertation, NM: University of Santa Cruz.
[ 234 ] Bibliography
5 3 2
Potsdam, E. (1997), “NegP and subjunctive complements in English,” Linguistic Inquiry, 28, pp. 533–541. Potsdam, E. (1999), “A syntax for adverbs,” in E. van Geldren and V. Samiin (Eds.), The Proceedings of the Twenty- Seventh Western Conference on Linguistics, Fresno: Department of Linguistics, California State University, pp. 397–411. Progovac, L. (1993), “Negative polarity: Entailment and binding,” Linguistics & Philosophy, 16, pp. 149–180. Radford, A. (1988), Transformational Grammar, 2nd edition, Cambridge: Cambridge University Press. Raajihi, A. (2000), al-Taṭbīq al-naḥwī. Bayrut: Dār al-Nahḍah al-ʻArabīyah. Ramchand, G. (2001), “Tense and negation in Bengali,” in A. Abbi, R.S. Gupta, and A. Kidwai (Eds.), Linguistic Structure and Language Dynamics in South Asia: Papers from the Proceedings of the SALA XVIII Roundtable, MLBD Series in Linguistics 15, Delhi: Motilal Banarsidass, pp. 308–326. Ramchand, G. (2002), Two Types of Negation in Bengali, Unpublished manuscript, Oxford University. Reinhart, T. (1976), The syntactic domain of anaphora, Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA. Richards, M. (2006), “Weak pronouns, object shift and multiple spell out: Evidence for Phase at the PF interface,” in C. Boeckx (Ed.), Minimalist Essays, Amsterdam/ Philadelphia: John Benjamins Publishing Co./Linguistik Aktuell. Ritter, E., & M. Wiltschko (2004), The lack of tense as a syntactic category: Evidence from Blackfoot and Halkomelem, Paper presented at the 39th International Conference on Salish and Neighbouring Languages, North Vancouver, and the Canadian Linguistics Association Annual Meeting, University of Manitoba, Winnipeg. Ritter, E., & M. Wiltschko (2005), “Anchoring events to utterances without tense,” Proceedings of WCCFL 24, Somerville, MA: Cascadilla Proceedings Project, Cascadilla Press, pp. 343–351. Rizzi, L. (1982), Issues in Italian Syntax, Dordrecht: Foris. Rizzi, L. (1989), Relativised Minimality, Cambridge, MA: The MIT Press. Rizzi, L. (1997), “The fine structure of the left periphery,” in L. Haegeman (Ed.), Elements of Grammar, Dordrecht: Kluwer, pp. 281–337. Roberts, I. (2010), Agreement and Head Movement: Clitics, Incorporation, and Defective Goals, Cambridge, MA: MIT Press. Roberts, I., & A. Roussou (2003), Syntactic Change: A Minimalist Approach to Grammaticalization, Cambridge: Cambridge University Press. Ross, J.- R . (1967), Constraints on variables in syntax, Ph.D. dissertation, MIT, Cambridge, MA. Rowlett, P. (1998), Sentential Negation in French, New York/Oxford: Oxford University Press. Rowlett, P. (2007), The Syntax of French, Cambridge: Cambridge University Press. Sauerland, U. (2004), “The interpretation of traces,” Natural Language Semantics, 12, pp. 63–127. Schmidt, H., & P. Kahle (1918), Volkserzälungen aus Palästina, vol. 1, Göttingen: Vandenhoek und Ruprecht. Schmidt, H., & P. Kahle (1930), Volkserzälungen aus Palästina, vol. 1, Göttingen: Vandenhoek und Ruprecht. Schutze, C. (1994), “Serbo- Croatian second position clitic placement and the phonology-syntax interface,” in A. Carnie, H. Harley, and T. Bures (Eds.), MIT Working Papers in Linguistics 21: Papers on Phonology and Morphology, Cambridge, MA: MITWPL, pp. 373–473.
Bibliography [ 235 ]
6 3 2
Schwarz, B. (1999), “On the syntax of either . . . or,” Natural Language and Linguistic Theory, 17, 339–370. Shlonsky, U. (1997), Clause Structure and Word Order in Hebrew and Arabic: An Essay in Comparative Semitic Syntax, New York: Oxford University Press. Shlonsky, U. (2004), “The form of Semitic noun phrases,” Lingua, 114(12), pp. 1465–1526. Shunnaq, A. (2000), Dictionary of Rural Terms in the North of Jordan, (Arabic-Arabic), A project subsidized by the Deanship of Research and Graduate Studies, Yarmouk University. Sigurðsson, H.A. (2004), “Agree and agreement—evidence from Germanic,” in W. Abraham (Ed.), Focus on Germanic Typology, Berlin: Akademie Verlag, pp. 61–103. Sigurðsson, H.A. (2006), Remarks on features. Unpublished manuscript, Lund University, Sweden. Siloni, T. (2001), “Construct states at the PF interface,” in: P. Pica & J. Rooryck (Eds.), Linguistic Variation Yearbook, vol. 1, John Benjamins, Amsterdam, pp. 229–266. Siloni, T. (2002), “Adjectival constructs and inalienable constructions,” in J. Ouhalla & U. Shlonsky (Eds.), Themes in Arabic and Hebrew Syntax, Dordrecht, Holland: Kluwer. Soltan, U. (2007), “On agree and postcyclic merge in syntactic derivations: First conjunct agreement in Standard Arabic,” in E. Benmamoun (Ed.), Perspectives on Arabic Linguistics XIX, Amsterdam/Philadelphia: John Benjamins Publishing Company, pp. 191–213. Soltan, U. (2011), “On issues of Arabic syntax: An essay in syntactic argumentation,” Brill’s Annual of Afroasiatic Languages and Linguistics, 3, pp. 236–280. Soltan, U. (2012), “Morphosyntactic effects of NPI-licensing in Cairene Egyptian Arabic: The puzzle of - š disappearance resolved,” in J. Choi et al. (Eds.), Proceedings of the 29th West Coast Conference on Formal Linguistics, Somerville, MA: Cascadilla Proceedings Project, pp. 241–249. Soltan, U. (2014a), “Splitting Neg: The morphosyntax of sentential negation in Cairene Egyptian Arabic revisited,” in R. Khamis-Dakwar and K. Froud (Eds.), Perspectives on Arabic Linguistics XXVI, Amsterdam/Philadelphia: John Benjamins Publishing Company. Soltan, U. (2014b), “On the distribution and licensing of polarity-sensitive items in Egyptian Arabic: The cases of ʔayy and walaa,” in S. Farwaneh and H. Ouali (Eds.), Perspectives on Arabic Linguistics XXIV–XXV, Amsterdam/Philadelphia: John Benjamins Publishing Company. Speas, M. (1991), “Functional heads and the mirror principle,” Lingua, 84, pp. 181–214. Sportiche, D. (1988), “A theory of floating quantifiers and its corollaries for constituent structure,” Linguistic Inquiry, 19(2), pp. 425–451. Szabolcsi, A. (2004), “Positive polarity- negative polarity,” Natural Language and Linguistic Theory, 22, pp. 409–452. Tieken-Boon van Ostade, I., G. Tottie, & W. van der Wurff (Eds.) (1998), Negation in the History of English, Berlin: Mouton de Gruyter. Travis, L. (1984), Parameters and effects of word order variation, Ph.D. dissertation, MIT, Cambridge, MA. Vallduví, E. (1994), “Polarity items, n-words and minimizers in Catalan and Spanish,” Probus, 6, pp. 263–294. Van Mol, M. (2000), Modern Arabic Representative Corpus (MARC- )2000, KU-Leuven: Leuven.
[ 236 ] Bibliography
7 3 2
Van Mol, M. (2003), Variation in Modern Standard Arabic in Radio News Broadcasts: A Synchronic Descriptive Investigation into the Use of Complementary Particles, Leuven: Peeters Publishers. Vendler, Z. (1967), Linguistics in Philosophy, Ithaca: Cornell University Press. Watanabe, A. (2004), “The genesis of negative concord: Syntax and morphology of negative doubling,” Linguistic Inquiry, 4, pp. 559–612. Wehr, H. (1953), “Zur unktion arabischer Negationen,” Zeitschrift der Deutschen Morgenländischen Gesellschaft (ZDM G), 103, pp. 27–39. Williams, E. (1981), “On the notions ‘lexically related’ and ‘head of a word,’” Linguistic Inquiry, 12, 245–274. Woidich, M. (1968), Negation und negative Satze im Agyptisch Arabischen, Ph.D. dissertation, Munich: Ludwig Maximilians Universität München. Woidich, M. (2006), Das Kairenisch-Arabische: Eine Grammatik, Wiesbaden: Harrassowitz Verlag. Wurmbrand, S. (2014), “Tense and aspect in English infinitives,” Linguistic Inquiry, (45)3, pp. 403–447. Ya‘qub, I.B. (1986), Mawsuu’at Al-Nahuu wa-l-Sarf wa-l-’I‘raab, Beirut: Daar Al-Ilm lil-Malaayiin. Zhang, N. (2009), Coordination in Syntax (Cambridge Studies in Linguistics), Cambridge: Cambridge University Press. doi:10.1017/CBO9780511770746 Zanuttini, R. (1989), “Two strategies for negation: Evidence from romance,” in J. Powers and K. de Jong (Eds.), Proceedings of ESCOL V, Columbus: Ohio State University, pp. 535–546. Zanuttini, R. (1991), Syntactic properties of sentential negation, Ph.D. dissertation, University of Pennsylvania, Philadelphia. Zanuttini, R. (1996), “On the relevance of tense for sentential negation,” in A. Belletti and L. Rizzi (Eds.), Parameters and Functional Heads. Essays in Comparative Syntax, Oxford: Oxford University Press, pp. 181–207. Zanuttini, R. (1997a), Negation and Clausal Structure: A Comparative Study of Romance Languages, Oxford: Oxford University Press. Zanuttini, R. (1997b), “Negation and verb movement,” in L. Haegeman (Ed.), The New Comparative Syntax, White Plains, NY: Longman, pp. 214–245. Zanuttini, R. (1998), Negation and Clausal structure. A Comparative Study of Romance Languages, Oxford Studies in Comparative Syntax, New York; Oxford: Oxford University Press. Zanuttini, R. (2001), “Sentential negation,” in Baltin & Collins (Eds.), The Handbook of Contemporary Syntactic Theory, Malden, MA; Oxford: Blackwell, pp. 511–535. Zeijlstra, H. (2004), Sentential negation and negative concord, Ph.D. dissertation, University of Amsterdam. Zeijlstra, H. (2008), Negative Concord Is Syntactic Agreement, Unpublished manuscript, University of Amsterdam. Zeijlstra, H. (2010), “On French negation,” in I. Kwon, H. Pritchertt, and J. Specnce (Eds.), Proceedings of the 35th Annual Meeting of the Berkeley Linguistics Society, Berkeley, CA: BLS. Zoerner, C. E. III. (1995), Coordination: The Syntax of andP, Ph.D. dissertation, University of California, Irvine. Zwart, J.-W. (2005), “Some notes on coordination in head-final languages,” in J. Doetjes & J. van de Weijer (Eds.), Linguistics in the Netherlands 2005, Amsterdam/ Philadelphia: John Benjamins Publishing Company, pp. 231–242.
Bibliography [ 237 ]
8 3 2
9 3 2
INDEX
Figures and tables are indicated by f and t following the page number. accusative case, on complements of PSIs, 202–210 adverbial negative concord items, 221 NCIs, licensing, 107, 110–111, 118–119, 132 PSI classification and lexical categories, 5, 6, 9, 39, 57, 65 PSIs with head-like properties, 193, 197 adverbial negative polarity items, 72, 157, 213, 219 Arabic, 76–77, 86–97 PSIs, classification and lexical categories, 38–39, 45, 47–50, 55 PSIs with head-like properties, 171, 172, 181–184, 188–189, 193, 197, 221 syntax issues, 4, 7–9 adverbs, temporal, 18, 38, 198–201 adverbs with phrasal projection, head- like PSIs as, 171–196 empirical generalizations, 173–177 phrasal heads, analysis, 181–188 syntactic status, 177–181 as XPs, licensing by negative constituents, 188–195 AdvPs, 149, 172, 185, 203, 205–208 in Spec-XP positions, 172, 185, 203, 205–208 Alqassas, A., 119, 127, 194, 196 PSIs, syntax issues, 15, 16, 17, 24–25 PSIs with head-like properties, 17 Alsarayreh, A., 118, 193 antiadditives, 7, 55, 67f, 71, 102, 103f antimorphics, 7, 55, 71, 102
antiveridicial environments, 7, 40, 55–56, 64–65, 71, 103, 105, 136 antiadditives, 7, 55, 67f, 71, 102, 103f antimorphics, 7, 55, 71, 102 definition, 55 negation, 55 Arabic. See also specific topics adverbial negative polarity items, 76–77, 86–97 determiner negative polarity items, 77–78, 82, 83, 86–97, 105 NCI analyses and syntactic agreement approach, 115–135, 129t, 134t, 135t negation, standard, 97–100 negative indefinites, 60, 60t negative indefinites, variations, 211, 212t negative polarity items, 51–55, 52t negative polarity items, adverbial, 76–77, 94 nominal negative polarity items, 77, 82–84 non-nominative case, 200–202 NPI analyses, 77–85 polarity-sensitive items, syntax, 2–3, 211, 212t standard negation and NPI licensing, 97–100 aspectual projection (AspP), 10 Baker, M., 120 baqi, 13, 15, 58, 60t, 118, 172, 178, 181, 198
0 4 2
bare argument ellipsis (BAE), coordinate complexes and, 160–168, 220 baʕd, 12, 18–20, 95 lexical categories, 38, 47–48, 58, 60t NCIs, licensing, 119 pre-negative, 110–111 syntax, 12 syntax, PSIs with head-like properties, 18–20, 171–177, 180–205, 221– 222 (see also polarity-sensitive items (PSIs), with head-like properties) Benmamoun, Elabbas on NCIs, licensing, 109–110, 112, 116–120, 133–134, 161 on NPIs, licensing, 72, 74–75, 77–79, 81–82, 84–86, 88, 95, 98 on PSIs and lexical categories, 50, 58 on PSIs with head-like properties, 172, 174, 177–179, 182–183, 195– 196, 198, 203 on PSI syntax issues, 2–3, 6, 8, 11–13, 15, 20, 24 bipartite negation, 11, 16, 23, 25, 28, 37, 213–214 NCIs, licensing, 146–147, 159, 173, 177 NPIs, licensing, 69, 86–87, 93–94, 104 PSIs with head-like properties, 37, 173, 177, 181–183, 194–195 case licensing, head-likes PSIs and, 196–210 accusative case on complements of PSIs, 202–210 definition and examples, 196–197 distribution, 197–198 non-nominative case, Arabic, 200–202 temporal adverbs and complements, 198–199 Chomsky, N., 93, 99, 136, 220–221 Cinque, G., 187 Closest Conjunct Agreement (CCA), 167 complementizer phrase (CP), 65, 218–222 NCIs, licensing, 127, 131, 136–137, 142, 151–152, 156–157, 167 NPIs, licensing, 80–83, 96–98, 107
[ 240 ] Index
PSI classification and lexical categories, 17, 35 PSIs with head-like properties, 172, 186–192, 195–196, 199 coordinate complexes, NCI licensing bare argument ellipsis, 160–168 compositional meaning, 146–152 covert negative operator, 215–217, 219–220 NCIs, 10, 35, 107, 108, 113, 123–124, 127–129, 131, 138, 157, 159 determiner negative concord items, 9, 39, 57, 65, 216, 220 licensing, 108–109, 111, 114, 116, 118, 119, 143 determiner negative polarity items, 2t, 7, 72 Arabic, 77–78, 82, 83, 86–97, 105 PSI classification and lexical categories, 38, 39, 45–46, 49–50, 52t, 65–66 determiner phrase (DP), 80, 130, 142, 150, 152–155, 160–168, 172, 187, 194 Dikken, Den, 145, 150, 155 disjunction particles, coordinate complexes, 65–69, 67f disjunction phrases and clauses, NCI licensing, 137–168 coordinate complexes, compositional meaning, 146–152 fundamentals, 137–138 introduction and empirical generalizations, 138–143 negative coordination analysis, by wala, 151–168 bare argument ellipsis and coordinate complexes, 160–168 ellipsis, 159–160 laa, status and locus, 152–153 wala, status and locus, as NCI and disjunction operator, 154–159, 159t theoretical issues, 143–146 Doron, E., 209–210 double negation (DN), 11 NCIs, licensing, 111, 113–114, 121– 122, 125, 127–128, 132–134, 135t, 155, 157
1 4 2
PSIs with head-like properties, 206, 216–218, 216t Downward Entailment (DE operator), 44, 101–103 ellipsis, negative coordination analysis, by wala, 159–160 enclitic negation, 11–17 epiphenomenon, 34, 69, 86, 93, 104, 213 exclusive choice or, 65, 67, 105 fragment answer test, 37 free choice items (FCIs), 37, 46, 67 free choice or, 65, 105, 167 Giannakidou, A., 1, 7, 9, 34, 40, 43–44, 55, 76, 78, 81, 89 Giannakidou-Zwart, theory of polarity items, 67, 67f, 103, 103f ħada, 3, 34, 37–46, 49, 52–53, 52t, 55–58, 61, 63–64 Haegeman, L., 113 head–complement relation, NPIs, temporal, 2–3 head-like properties, polarity-sensitive items with, 17–20, 171–210. See also polarity-sensitive items (PSIs), with head-like properties head NPIs, 13 Heim, I., 215 Herburger, E., 114–115 ħə tta waħə d, 39, 57, 59, 214 NCIs, licensing, 109, 112–113, 116– 117, 120, 133 NPIs, licensing, 86, 88–89 PSIs, syntax issues, 5, 12, 16, 17 Heycock, C., 209–210 Hoyt, F., 57–58, 79, 82, 83, 120, 182 idiomatic negative polarity items, 7, 38, 45, 51, 65, 72, 83, 104, 111 indefinite nouns, 38, 222 ħada vs. iši, 3, 34, 37–38, 40–41, 45 ʔaħaħaand šaʔaʔ iši, 38, 40 inherent case, 19 interpretable negation (iNEG), 10, 10t iši, 3, 34, 37–38, 45–46
junction phrase (JP), 155, 167 kitaab, 40–42 kokhono, 24–25 kull, 20 Kumar, R., 75 laa, 2t, 20, 24, 28, 31–33, 33t negative coordination analysis by wala, 152–153 laa-wala, 35, 67, 212t, 219–220 NCIs, licensing, 107–108, 137–140, 143, 146, 149, 153, 156–159, 159t, 162, 167 Ladusaw, W., 44, 63, 101, 103, 112 lam, 33t lan, 23–24, 33, 33t, 129t, 159t, 212t Lasnik, H., 74, 78 last resort mechanism, 35, 87, 89, 107, 124–125, 128, 131, 159, 215, 217–221 laysa, 33t lexical ambiguity approach, 108, 113–115 licensors. See also specific types of licensing covert vs. overt, 35 NCIs, 55 NPIs, 6–7, 44, 49, 71 PSIs, 3, 37, 63, 69 range, wide, 34 locality relation, 79, 109–110, 120 locus of negation, 11–17 logical form (LF), 8, 205 NCIs, 110, 113, 117–118, 121–122, 126, 150, 158 NPIs, 72, 73, 75–76, 81, 83–84, 87, 104 polarity sensitive items, head-like, 179, 193, 195, 196 PSIs, 17 reconstruction at, 81, 104, 108–109, 111–112, 122 syntax-LF, 89 l-walad, 14–15, 95, 187 ma-, 23t, 28 maa, 10t, 14–16, 18–22, 23t, 24–33, 33t, 45 maħadaaš, 3, 10–11, 10t wala-NP vs., 1, 2t, 9–11, 10t
Index [ 241 ]
2 4
Marantz, A., 208 maʕ umrhuuš, 217 negative concord items, licensing, 115, 124, 127–128, 129, 133–134 PSIs, syntax issues, 3, 10–11, 10t, 35 wala-NP vs., 1, 2t, 9–11, 10t Merchant, J., 179 Merge and Move, 71, 213, 215 microvariation, 3, 33, 69, 134, 211, 216, 223 minimize formal feature mismatch (MFFM), 91–93 monotonicity approach, 44, 101–103 multi-locus view, 24 Multiple Agree, 11, 132–134, 134t, 135t, 215–216, 216t variations, 216t NEG, 12, 123, 129t, 159t, 212t negation. See also double negation (DN) interactions, PSIs, 4 (see also specific types) NPI licensing and, Arabic, 97–100 negative concord items (NCIs), 37–38. See also specific topics adverbial (see adverbial negative concord items) baqi, 13, 15, 58, 60t, 118, 172, 178, 181, 198 c-command configurations, 2–3 complementary distribution, 154 definition and examples, 1, 2t dependencies, 9–11, 10t determiner, 9, 39, 57, 65, 216, 220 determiner, licensing, 108–109, 111, 114, 116, 118, 119, 143 idiomatic, 65 lexical categories, 57–65 licensors, 55 micro-comparisons, 215 negation requirement, 5 vs. NPIs, 5–66 specifier-head, 2–3 syntactic dependency with negation, 60, 88, 108, 115–116, 121 temporal, head–complement relation, 2–3 ʔabadan, 5–6
[ 242 ] Index
negative concord items (NCIs), licensing, 107–169 covert negative operator, 10, 35, 107, 108, 113, 123–124, 127–129, 131, 138, 157, 159 disjunction phrases and clauses, 137– 168 (see also disjunction phrases and clauses, NCI licensing) double negation, 111, 113–114, 121–122, 125, 127–128, 132– 134, 135t, 155, 157 last resort mechanism, 107, 124–125, 128, 131, 159 NCI licensing, 107–137 Arabic, NCI analyses, syntactic agreement approach, 115–135, 129t, 134t, 135t fundamentals, 107–112 lexical ambiguity approach, 108, 113–115 NPI approach, 112–113 NQ approach, 113 phase theory, 135–137 negative markers, 108–109, 112–116, 120–128, 131, 133–134, 134t, 135t, 137, 140, 142, 145, 152– 156, 159–160, 168–169 negative quantifiers, 108, 113, 114–115 NPIs, 108–119, 125–126, 146, 156–157 Spec-Head relationships, 108–110, 113, 115–118 negative coordination, 137, 168 markers, 68, 141–142 (see also wala) negative coordination, syntactic analysis, 140–142 theoretical issues, 143–146 by wala, 151–168 bare argument ellipsis and coordinate complexes, 160–168 ellipsis, 159–160 laa, status and locus, 152–153 wala, status and locus, as NCI and disjunction operator, 154–159, 159t negative disjunction, 67, 146, 169 negative indefinites (NIs), 3, 215–220, 216t, 223. See also specific types
3 4 2
Arabic, 60, 60t Arabic, variations, 211, 212t definition and examples, 1, 2t maʕumrhuuš, 3, 10t, 35 NCIs, licensing, 107, 115, 126–131, 133–134, 134t, 135t, 137, 157, 159 PSIs with head-like properties, 175, 181, 188, 196, 206, 211 semantically negative, 3 syntax, 215–221, 216t variations, 10–11, 10t, 216t wala-NP vs. maħadaš and maʕ umrhuuš, 1, 2t, 9–11, 10t negative markers (NMs), 37 complementary distribution, 4 inventory, 23, 23t NCIs, 108–109, 112–116, 120–128, 131, 133–134, 134t, 135t, 137, 140, 142, 145, 152–156, 159– 160, 168–169 NPIs, simple declarative sentences, 4 variations, 10, 10t, 37, 216t negative polarity items (NPIs), 4–5, 37– 38, 40–56. See also specific topics adverbial, 4, 7–9, 38–39, 45, 47–50, 55, 72, 157, 213, 219 Arabic, 76–77, 94 PSIs with head-like properties, 171, 172, 181–184, 188–189, 193, 197, 221 antiveridicality, 7 Arabic, 51–55, 52t definition, 172 dependencies, 6–9 determiner, 2t, 7, 38, 39, 45–46, 49–50, 52t, 65–66, 72 Arabic, 77–78, 82, 83, 86–97, 105 as existential quantifiers, 3 general indefinites vs., 37 idiomatic, 7, 38, 45, 51, 65, 72, 83, 104, 111 lexical categories, 38, 45–56 licensors, 6–7, 44, 49, 71 NCI licensing, 108–119, 125–126, 146, 156–157 vs. NCIs, 5–66 negation requirement, simple declarative sentences, 4–5
negative markers, simple declarative sentences, 4 nominal, 7, 38, 40, 45, 55–56, 65, 72, 111, 112 Arabic, 77, 82–84 post-negative, 81, 104, 108–109, 111–112, 122 vs. PPIs, 40–45 Spec-Head relationships, 2–3, 14, 72, 79–84, 88–91, 98–100, 104 syntactic dependency with negation, 81, 172 syntactic licensing, 7 syntax, 213–215 variations, Arabic languages, 211, 212t ʕə mmər, 2–5, 2t, 8, 11, 13, 15 ʕumr, 5–7, 12–16, 18–19 negative polarity items (NPIs), licensing, 71–105 adverbial and determiner NPIs, Arabic, 86–97 antiveridical environments, 71 in Arabic, analyses, 77–85 categorical status, 71 logical form, 72, 73, 75–76, 81, 83–84, 87, 104 negative quantifier, 72–73 nonveridicial contexts, Arabic PSIs, 100–105 semantic approach, 71 standard Arabic negation and, 97–100 syntactic configurations, 71 syntactic operations (Merge and Move), 71 syntactic structure, 71–77 negative polarity items (NPIs), syntax, 213–215 negative quantifier (NQ), 10, 72–73, 108, 112, 113, 114–115 negative spread, 114 NEG-Criterion, 113 NEG-factorization, 113 NEG features, inventory, 211, 212t NegP, 25, 27–29 head-like PSIs as syntactic heads selecting, 178–180 neither-nor (laa-wala). See laa-wala
Index [ 243 ]
42
nominal negative polarity items, 7, 38, 45, 55–56, 65, 72, 111, 112 Arabic, 77–78, 82–84 non-nominative case in Arabic, case licensing, 200–202 nonveridicality, 1, 7, 34, 55, 63–64, 71, 103–104, 135 noun phase (NP) subject, preceding head-like PSIs, 186–188 NPI approach, 112–113 NQ approach, 113 Ouali, H., 114, 116, 120, 123–124, 155–156 Ouhalla, J., 12, 78, 87–88, 133 &P, 143, 145, 167–168 Penka, D., 3, 9, 10–11, 21, 35, 107, 128, 134, 215–216, 218, 219 Pesetsky, D., 208 phase theory, NCI licensing and, 135–137 phi-agreement paradigm, 166 phrasal heads, head-like PSIs as adverbs with, 181–188 NP subject preceding head-like PSIs, 186–188 PSIs as XPs, 183–186 polarity-sensitive items (PSIs), 1–34. See also specific types and topics adverbial, 12, 221 complementary distribution with enclitic negation, 4, 13, 15–17, 37, 40, 59, 86–87, 93 components, 1 definition and word interactions, 37 as fragment answers, 4 fundamentals, 1–6, 2t importance, Arabic syntax, 2–3 lexical categories and interactions, 1 licensors, 3, 37, 63, 69 literature, Arabic syntax, 2 NCIs, 5–6 negation interactions, 4 (see also specific types) negative indefinites, 211, 212t negative markers, complementary distribution, 4 negative polarity, 211, 212t
[ 244 ] Index
NPIs, 4–5 syntactic licensing approach, 1 variations, Arabic languages, 211, 212t polarity-sensitive items (PSIs), classification and lexical categories, 34, 37–69 classification, 37–38 disjunction particles, coordinate complexes, 65–69, 67f free choice items, 37, 46, 67 general indefinites vs. NPIs, 37 ħada, 3, 34, 37–46, 49, 52–53, 52t, 55–58, 61, 63–64 iši, 3, 34, 37–38, 45–46 lexical categories, 38–39 NCIs and lexical categories, 57–65 NPIs, 37–38 (see also negative polarity items (NPIs)) NPIs, lexical categories and, 45–56 NPIs vs. PPIs, 40–45 positive polarity items, 37–38 (see also positive polarity items (PPIs)) polarity-sensitive items (PSIs), critical syntax issues, 6–34 enclitic negation and locus of negation, 11–17 NCI dependencies, 9–11, 10t NPI dependencies, 6–9 PSIs with head-like properties, 17–20 polarity-sensitive items (PSIs), with head-like properties, 17–20, 171–210 adverbs with phrasal projection, 171–196 empirical generalizations, 173–177 as phrasal heads, analysis, 181–188 syntactic status, 177–181 as XPs, licensing by negative constituents, 188–195 case licensing, 196–210 accusative case on complements of PSIs, 202–210 definition and examples, 196–197 distribution, 197–198 non-nominative case, Arabic, 200–202
5 4 2
temporal adverbs and complements, 198–199 double negation, 206, 216–218, 216t logical form, 179, 193, 195, 196 syntactic dependency with negation, 178 syntax, 221–224 Pollock, J.-Y., 20 positive polarity items (PPIs), 37–38 NCIs, 37–38 vs. NPIs, 40–45 post-negative negative polarity items, 81, 104, 108–109, 111–112, 122 Ramchand, G., 24–26, 219 reconstruction at LF, 81, 104, 108–109, 111–112, 122 Rizzi, L., 190 sentential negation, 20–24, 87, 89, 97, 102, 114, 172, 191, 197 definition, 20–24, 23t locus, 24–34 Soltan, Usama, 215 NCIs, licensing, 114, 116, 120, 123– 124, 155–156 NPIs, licensing, 86, 89–93, 103 PSIs, classification and lexical categories, 50, 58, 63 PSIs, syntax issues, 2, 3, 6, 12, 15, 23 PSIs with head-like properties, 176–177 Southern Levantine, 2 specifier-head (Spec-Head) relationship, 2–3 NCI licensing, 108–110, 113, 115–118 NPIs, 2–3, 14, 72, 79–84, 88–91, 98–100, 104 PSIs, 215 Spec-NegP, 15, 213, 215, 221, 223 NCIs, licensing, 113, 119, 121–122 NPIs, licensing, 80, 87 PSIs with head-like properties, 171, 178, 181, 185–187, 189, 192–194, 196, 208 Spec-TP, 19, 32 NCIs, licensing, 78, 81–82, 88–89, 116–118, 121, 126
PSIs with head-like properties, 185, 187, 188, 190, 202, 209–210 Spec-VP, 78, 116, 121, 181, 189, 195 Spec-XP AdvPs in, 172, 185, 203, 205–208 syntactic adverbs. head-like PSIs as, 180–181 split-scope phenomenon, 3 syntactic adverbs in Spec-XP, head-like PSIs as, 180–181 syntactic agreement approach, Arabic, 115–135, 129t, 134t, 135t syntactic dependency with negation head-like PSIs, 178 NCIs, 60, 88, 108, 115–116, 121 NPIs, 81, 172 syntactic heads selecting NegP, head-like PSIs as, 178–180 syntactic operations (Merge and Move), 71, 213, 215 syntax-LF, 89 temporal adverbs, 18, 38, 198–201 topic (topicalization), 75–76, 79–80, 82– 83, 125, 192 Torrego, E., 208 uninterpretable negation (uNEG), 10, 10t, 212t, 215, 216t, 217, 219, 220 NCIs, licensing, 90–92, 120–128, 129t, 134t, 135t, 155–159, 159t, 168–169 PSIs with head-like properties, 176, 207 van der Wouden, T., 102 wa-, coordination by, 138, 140, 143, 145–148, 150–151, 159–160 wala, 67 wala, negative coordination analysis by, 151–168 bare argument ellipsis and coordinate complexes, 160–168 ellipsis, 159–160 laa, status and locus, 152–153 status and locus, 154–159, 159t
Index [ 245 ]
6 4 2
walaħada (wala-ħada), 212t, 216–218, 216t NCIs, licensing, 108, 111, 115, 120, 121–128, 132, 134–135, 134t, 135t, 156, 158 PSIs, syntax issues, 2–3, 9–10, 10t, 17, 35 PSIs and lexical categories, 57–59, 60t, 63–64 PSIs with head-like properties, 2–3, 10t, 60t, 63–64, 125–128, 134– 135, 134t, 135t, 172–173, 188– 192, 196, 212t, 216, 216t wala-NP, 3 walla/willa (walla/ willa/ʔ am ‘or’), 34, 40, 68–69, 138–139, 142, 150–151, 158, 160, 168, 169, 220 X phrases (XPs) head-like PSIs as, licensing by negative constituents, 188–195 PSIs, 183–186, 213
[ 246 ] Index
Zanuttini, R., 113, 219 Zeijlstra, H., 55, 218 NCIs, licensing, 107, 113, 120–122, 128 NPIs, licensing, 71, 72, 89, 103 on PSIs, syntax issues, 24, 35 Zhang, N., 144 Zwart, J.-W., 102 Giannakidou-Zwart theory of polarity items, 67, 67f, 103, 103f ʔabadan, 5–6 ʔaħad, 2t, 32–33, 38, 52t, 97–98, 212t ʕəmmə r, 2–5, 2t, 8, 11, 13, 15 ʕumr, 5–7, 12–16, 18–19 head-like PSIS as adverbs with phrasal projection, 171–210 (see also polarity-sensitive items (PSIs), with head-like properties) NPI licensing, 71–77, 84–86, 89–95, 104, 119, 125–126, 157